UC Riverside Undergraduate Research Journal

University of California, Riverside 

Undergraduate Research Journal 

Table of Contents 

Zero Waste Biodiesel: Using Glycerin And Biomass 

To Create Renewable Energy 

Sean Brady ......................................................5 

Computational Prediction of Association Free Energies 

for the C3d-CR2 Complex and Comparison to Experimental Data 

Alexander S. Cheung ............................................13 

Phosphorylation of Crk Adaptor Protein by Cdc42-Activated Pak2 

and Identification of Phosphorylation Sites 

Jisun Lee ......................................................23 

Augustan Era Policy on the Rhine Frontier from 34 B.C.E.-16 C.E. 

Kyle McStay ...................................................29 

Fractal Strings and Number Theory: The Harmonic String and the Prime String 

Jason C. Payne ..................................................35 

Motion Based Bird Sensing Using Frame Differencing and Gaussian Mixture 

Deep J. Shah ...................................................47 

Love a Son, Raise a Daughter: A Cross-Sectional Examination 

of African American Mothers’ Parenting Styles 

James M. Telesford ..............................................53 

Mating-Type Distribution Of The Rice Blast Pathogen 

Pyricularia grisea In California 

R. Z. Urak .....................................................61 

Secondary Organic Aerosol (Soa) And Ozone Formation 

From Agricultural Pesticides 

Lindsay D. Yee .................................................67 

Bacterium-Induced Fluorescence-Enhancement Kinetics: 

Breaking 100-Year Old Traditions of Staining Bioanalyses 

Elizabeth Zielins ................................................75 

Copyright © 2008 by the Regents of the University of California. All rights reserved. No part of the UCR Undergraduate 

Research Journal, Volume II (June 2008) may be reprinted, reproduced, or transmitted in any form or by any means 

without consent of the publisher. 

UCR Un d e r g r a d u a t e Re s e a r c h Jo u r n a l 1

From the Administration 

Whether the area of study is art history, neuroscience, or environmental engineering, the opportunity 

to participate in undergraduate research opens new worlds for students at the University of California, 

Riverside. Students have the opportunity to unravel mysteries, explore new concepts, or advance and test 

their own hypotheses, while learning the rigor of the scientific method, the discipline of writing, and the 

creativity of experimental design. 

In an increasingly interdisciplinary academy, research often occurs at the boundaries of disciplines. 

Through guided discovery, UCR students can push not only these boundaries, but also their own imaginations. In 

the process, they gain both knowledge and invaluable experience that will help them in their future endeavors. 

The Undergraduate Research Journal chronicles the discoveries and observations of UCR students 

who have been inspired by the natural world, a mentor, or their own curiosity. I invite you to share their 

journeys as described in these pages and marvel, as I did, at the quality of the work they have achieved. 

Sincerely, 

Robert D. Grey 

Acting Chancellor 

You hold in your hands the second annual UC Riverside Undergraduate Research Journal. It provides 

a selective, peer-reviewed venue to feature the very best faculty-mentored undergraduate research and 

scholarship on our campus. We think the articles in this issue certainly meet that high standard, and provide 

a wonderful foundation for the future growth of the Journal in scope, visibility, and stature. 

The peer-review process has been very ably led by our Student Editorial Board, with advice as needed 

from the Faculty Advisory Board. So, much like the research published here, we have established a culture in 

which the editorial activities of the Journal are managed by students in a faculty-mentored process. 

The process of discovery can be filled with excitement but also riddled with frustration, as we 

search and stumble in the dark, trying to shed new light that enriches our understanding of social or natural 

phenomena, nourishes our emotions, or enlivens our souls. During this process, we travel a path on which 

no one has been before. The journal article is the culmination of that process – a formal presentation to our 

community of peers and mentors of what we found on that journey. Place this volume on your bookshelf. 

Pull it down occasionally from the shelf to re-read and to remind yourself of the journey you traveled. I 

applaud you on your efforts and wish you many more such journeys in the future. You are truly the “best and 

brightest” at UCR. 

With Best Regards, 

David H. Fairris 

Vice Provost for Undergraduate Education 

Professor of Economics 

2 UCR Un d e r g r a d u a t e Re s e a r c h Jo u r n a l

UCR Undergraduate Research Journal II 

Faculty Advisory Board 

Chuck Whitney, Creative Writing, Board Chair 

Thomas Eulgem, Botany and Plant Science 

Dimitrios Morikis, Bioengineering 

Nosang Myung, Chemical and Environmental Engineering 

William Okamura, Chemistry 

Rebekah Richert, Psychology 

Dana Simmons, History 

Howard Wettstein, Philosophy and 

Director of University Honors Program 

Student Editorial Board 

Gregory Goalwin, Political Science/International Affairs, History, 

Editor-in-Chief 

Lauren Cummings, History, Assistant Editor-in-Chief 

Michael Bogseth, Biochemistry 

Sophia Fox, Political Science/International Affairs 

Rachael Meeker, Sociology, Religious Studies 

Jeffrey Suhalim, Bioengineering 

Elizabeth Zielins, Bioengineering 

Executive Committee for Undergraduate Research 

David Fairris, Vice Provost for Undergraduate Education, 

Committee Chair 

Gary Scott, Associate Dean, College of Natural 

and Agricultural Sciences 

Steven Brint, Associate Dean, College of Humanities, 

Arts, and Social Sciences 

Chinya Ravishankar, Associate Dean, 

Bourns College of Engineering 

Leah Haimo, Associate Dean, Graduate Division 

Richard Luben, Office of Research 

Staff Coordinators 

Patsy Oppenheim, Assistant Vice Provost, 

Undergraduate Education 

Breanna Baeza, Student Assistant, Undergraduate Education 

Debbie Pence, Management Services Officer, 

Undergraduate Education 

From the Student Editorial Board 

UCR Undergraduate Research Journal Student Editorial Board, left to right: Michael 

Bogseth, Editor-in-Chief Gregory Goalwin, Assistant Editor-in-Chief Lauren Cummings, 

Elizabeth Zielins, Sophia Fox, Jeffrey Suhalim. Not pictured: Rachel Meeker. 

The members of the Student Editorial Board are honored to have 

played a part in the production of this, the second volume of the 

UCR Undergraduate Research Journal. As editors, we had the 

opportunity to work with both faculty members and our peers in 

crafting a journal that highlights the incredible depth and diversity 

of undergraduate research that is being conducted here at UCR. 

We were deeply impressed by the quality of the research and the 

professionalism of the student researchers who submitted articles 

to this year’s edition of the journal. The number of articles submitted 

more than doubled the number submitted to last year’s inaugural 

edition and we are proud to reflect this increasing awareness of the 

Undergraduate Research Journal through a commensurate increase 

in the size of the journal itself. It is our hope that undergraduate 

research will continue to be encouraged at UCR and we are 

pleased to provide an opportunity for the most outstanding student 

researchers to be recognized and celebrated for their efforts. It is 

with great pride that we present the second volume of the UCR 

Undergraduate Research Journal. 


 

 


Zero Waste Biodiesel: Using Glycerin 

And Biomass To Create Renewable Energy 

Sean Brady, Kawai Tam 

Coauthors: Gregory Leung, Christopher Salam 

Department of Chemical and Environmental Engineering 


ABSTRACT 

Biodiesel production creates glycerin as a byproduct. Although glycerin does have its commercial 

uses, even the current modest biodiesel production has outstripped US glycerin demand. We have 

combined this excess waste glycerin with waste biomass to produce combustible pellets as an 

alternative to coal for energy production. Our pellets produced an energy yield of ≈16 kJ/g, placing 

their energy content within the expected range for existing fuel pellet infrastructure. The pellets 

can be viably manufactured using simple manufacturing equipment, and can be combusted as fuel 

in existing fuel pellet and Coal burning facilities. This will greatly facilitate pellet production and 

adoption as an alternative fuel source in our increasingly resource-conscious world. 

FACULTY mentor 

Kawai Tam 


Sean Brady and recent graduates, Gregory Leung and Christopher Salam, are truly 

an inspiring team of committed, intelligent and professional individuals; it has 

been a pleasure being their faculty mentor. What started as an inspiration to fuel 

campus transportation vehicles with biodiesel manufactured from waste oil from 

food services, led the team to examine other waste streams that exist from the biodiesel process 

and other biomass waste materials available on campus. This waste stream analysis segued into a 

project that would investigate a potentially new fuel source of energy derived solely from waste 

materials. As the capstone senior design instructor and instructor in Green Engineering, I found this 

idea to be innovative because it featured the accounting of multiple waste streams in various life 

cycle analyses with a potentially beneficial result with immediate impact. As their faculty mentor, 

I provided guidance in their experimental design and framework of the project and mentorship as 

they developed proposals for student design competitions. The idea and work presented here earned 

recognition at the 2007 WERC environmental design competition with a U.S. Department of 

Agriculture award for innovative use of agricultural materials and a Phase I award at the 2007-2008 

EPA P3 student design competition. 

A U T H O R 

Sean Brady 

Environmental Engineering 

Sean Brady is a graduating senior in 

Environmental Engineering, with a focus 

on water conservation and resource 

management. His industry experience 

includes asphalt quality control, waste 

water plant operations, and perchlorate 

remediation. Although Zero Waste 

Biodiesel itself is more “material and 

energy management,” it dovetails nicely 

with the central themes of smart, lowwaste 

processes and the transformation 

of waste streams into useful products. 

Sean is currently pursuing his Professional 

Engineering License, and after 

graduation he will continue his engineering 

career in New Mexico, with his 

wife and two children. 

(Special commendation goes to two 

recent graduates who were a part of 

Sean’s research team. Christopher 

Salam, ’07, is currently a graduate 

student at the University of California, 

Davis, studying renewable energy 

sources. Gregory K. Leung, ’07, is a 

practicing engineer.) 


Zero Waste Biodiesel: Using Glycerin And Biomass To Create Renewable Energy 

Sean Brady 

Introduction 

Biodiesel is a popular alternative fuel. It is carbon 

neutral, has emissions equivalent to or below diesel, is 

biodegradable, non-toxic, and is significantly cheaper 

to manufacture than its petroleum equivalent. However 

there is one significant drawback: for every 10 gallons of 

biodiesel produced, roughly 1 gallon of glycerin is created 

as a byproduct. 

Although glycerin does have its industrial uses, 

current biodiesel production has already exceeded market 

demand, leaving large amounts of practically worthless 

glycerin in the manufacturers’ hands, leading to increased 

disposal costs. We show that by combining waste glycerin 

with waste biomass (corn husks, wheat chafe, etc.), we are 

able to produce pellets which can be easily and inexpensively 

manufactured, are suitable for existing combustion energy 

plants, and are a superior alternative to coal. 

Glycerin Pellets 

The idea of combining the waste glycerin from the 

biodiesel process with biomass is relatively new and no 

other projects using this idea have been published. The 

concept originated on a Biodiesel Internet forum where 

home brewers were brainstorming ways to utilize their 

excess glycerin. Many users discussed creating soap, 

lotions, and using the glycerin in food products, but many of 

these processes require purification, a chemically unstable 

process, and are inherently low-volume and low-demand. 

Other uses for glycerin include selling it to a company which 

Figure 1. Sample pellets containing biomass (sawdust) and 

glycerin in a ratio of 1 to 1.3. 

refines glycerin for use in food or pharmaceuticals. While 

once profitable, the current abundance of glycerin is such 

that there is little money to be made, and often money to be 

paid, in having a company pick up your glycerin. However 

one user caught our attention with the words “glycerin logs,” 

to be burned in traditional fireplaces. Although glycerin 

logs were ultimately unfeasible, it started us thinking about 

ways to create solid form, easily portable fuel sources for 

combustion energy. We ultimately found a way to absorb 

two waste streams, thereby enabling biodiesel production, by 

creating a product which could reduce the coal dependence 

in the world. 

Biodiesel 

Biodiesel has been well explored by industry and 

in academia for at least 30 years. As biodiesel is broadlydefined 

as any chemical that is a methyl ester, an acceptably 

vague definition because of the wide range of chemicals 

which can be combusted in a diesel engine. Due to this 

flexibility of definition and use, many different methods can 

be employed, even some that do not create waste glycerin. 

These processes, however, have their drawbacks. 

These often do not create a significant quantity of 

byproduct, and are usually terribly energy intensive. Often 

times the thinning procedure also leads to a weaker or 

undesirably volatile fuel. The major methods for creating 

a less viscous fuel from vegetable oil are dilution, microemulsification, 

pyrolysis, and trans-esterification 1 . Pyrolysis 

is fairly energy intensive, and leads to a loss of feed. 

Dilution and micro-emulsification processes will lead to 

lower quality fuel, and 

have large initial material 

costs. Trans-esterification, 

or the thinning process 

that chemically lowers 

the viscosity of the 

mixture, sticks out as an 

economically reasonable 

Figure 2. Biodiesel glycerin 

mixture 

process, especially if a 

market can be established 

for the post process glycerin. 

Lately, biologically-based 

reactions, such as lipasecatalyzed 

processes for 

creating biodiesel, have 



Sean Brady 

also been explored. 2 A variety of research on the qualities 

of the feed stock on oils has been conducted. 3 Due to the 

drawbacks of these processes, it would be preferable to 

remain with traditional biodiesel production, and simply 

find an adequate use of the waste glycerin. 

The biodiesel synthesis method that we used to great 

our biodiesel and waste glycerin is a trans-esterification 

process, which combines an alcohol as a thinning agent, 

and a strong hydroxide as a catalyst, with vegetable oil to 

create a viscous combustible liquid, shown in the top layer 

of Figure 2. The bottom layer is the waste glycerin which is 

our primary concern. Note that the 10:1 production ratio of 

biodiesel to glycerin is not accurately shown in the image. 

Refuse Derived Fuels (Rdf) 

The creation of waste pellets is already a significant 

industry in the developed world, converting waste 

from material and food industries to create compressed 

pellets suitable for combustion as an energy source. The 

manufacture of a pellet is easy to automate. In addition, 

the process often does not require heat input or a chemical 

change, and as such can be manufactured quickly. Many of 

these pellets are used in combustion plants, and therefore 

do not need extremely costly food-grade processing. 

These manufactured pellets are called refuse 

derived fuels, or RDFs shown in Figure 3, and are 

primarily combusted within power plants for energy 

purposes. Considering the vast array of materials already 

being formed into RDF and the ever-increasing demands 

for energy, there is plenty of room in the market for an 

additional source. 

Results 

Project/Design Approach 

Pellet formation is fairly straight forward. The raw 

materials (waste glycerin and waste biomass) are mixed 

by weight ratio and blended by hand in a large mixing 

bowl. Various ratios of glycerin(38.5g and 31.3g) to 

waste biomass(50g) were then mixed to produce a crude 

unfinished material. The pellet mixture (approximately 12g) 

is placed inside a rolled piece of newspaper wrapping, and 

the ends are folded down so that both ends of the cylinder 

are covered. No adhesive is used, and until the pellet is 

compressed this unit will remain prone to unwrapping. This 

Literature Values: Fuel Source 

Energy (kJ/g) 

Coal 4 15 – 27 

Coke 5 28 – 31 

Dry Wood 6 14.4 – 17.4 

Gasoline (octane) 3 47 

Diesel 3, 7, 8 44.8 – 47 

Bio-Diesel 3, 4 41.2 

Natural Gas (CH 4 

) 3 56 

Ethanol 3 29.7 

H 2 3 142 

Tires 28.5 – 35 

Waste Plastic 29 – 40 

Household Waste (RDF) 12 – 16 

Household (RDF) 13 – 16 

Demolition Waste (RDF) 14 – 15 

Paper Sludge (RDF) 12.5 – 22 

Waste Wood 15 – 17 

Dried Sewage 16 – 17 

Animal Waste 16 – 17 

Commercial Waste 16 – 20 

Industrial Waste 18 – 21 

Theoretical Values: Fuel Source 

1:1.3 Biomass/glycerin (manure 60%, trimmings 35%, 

leafy material 5%) (theoretical) 

Energy (kJ/g) 

11-24 

1:1.3 Glycerin/Sawdust Pellet (theoretical) 16.94 


Experimental Values: Fuel Source 

Energy (kJ/g) 


Table 1. Energy Content of Fuels 

Figure 3. A sample picture of a refuse derived fuel (RDF) 



Sean Brady 

not possible due to limited resources, mainly having used 

all available oxygen and squibs. (Update: As of the time of 

publication, access to a functional lab calorimeter has been 

obtained, and ongoing research is underway. Initial results 

proved our efforts worthwhile, coming in at 15.4, 15.6, and 

15.2 kJ/g and displaying a respectable error rate of only 

10% for our ad-hoc apparatus. These results will be part of a 

larger report to be published at a later date.) 

Figure 4. Presented are the components used to make the biomass 

glycerin waste pellets. 

raw pellet is transferred into the mold, a short length of 

PVC with one end sealed. The diameter of the PVC pipe is 

12.5 mm, and its length is four inches. The mold helps the 

pellet retain its cylindrical shape while a short metal rod, 

slightly smaller than the PVC internal diameter, is inserted 

into the open end of the mold to compress the pellet. 

Pressure was applied by hand, approximately 250 psi for 

15 seconds. This pressure not only reduced the pellet size, 

but also encouraged the glycerin to permeate the materials 

and form a single firm unit. It should be noted that this form 

of production is only used in initial laboratory experiments 

and actual commercial production will of course be large 

scale and automated. Figure 4 shows the components that 

were used in the bench scale process. 

A key element of this project is energy estimation of 

solid matter via calorific testing. Without a laboratory-quality 

bomb calorimeter, this test was conducted by combusting 

the material in an aluminum ‘boat,’ floating in water, in a 

well-insulated container. 9 While conducting this process, it 

was found that additional oxygen was needed for sufficient 

combustion to occur. The reaction eventually consumes the 

solid matter, and the temperature change experienced in the 

water is recorded and placed into the specific heat equation 

to find the energy content for the mass of the pellet. After 

multiple runs of failed calorimeter results due to accidental 

combustion of the container, wetting of the mixture, and 

failure of the squib to ignite, a single somewhat accurate 

quantitative estimate of the energy content of the solid was 

obtained, and is included in Table 1. Additional tests were 

Error Analysis 

In considering possible sources of error in this 

design, the only area of concern lies with our ad-hoc 

calorimeter and its ability to estimate energy content. 

If a proper calorimeter was available, the data, and our 

estimates derived thereof, would be much more reliable. 

Additional deviation would arise from variations inherent 

in the waste biomass, but given the medium (RDF pellets) 

the energy output is expected to vary slightly unit to unit. 

Additional testing to determine the exact variation to be 

expected will be performed once an adequate calorimeter 

is obtained. 

Results 

The energy values for glycerin waste pellets and 

competing energy sources are shown in Table 1, showing 

that theoretically and experimentally, the glycerin waste 

pellet is a very suitable source in replacing or supplementing 

low end coal and RDFs. 

Discussion 

Economical Impacts 

The economic aspects of this project are perhaps 

the most readily understood and appreciated. The fact that 

biodiesel is less expensive to produce than petroleumbased 

diesel is a significant contributor to biodiesel 

emerging popularity. However, those costs do not include 

costs associated with storage or disposal of waste glycerin, 

a consideration which is going to demand significant 

attention as biodiesel production increases. Laboratorybased 

biodiesel production on a small scale can cost less 

than a $1 a gallon; however, in a large scale biodiesel plant, 

with associated licensing and fees, production can cost 

$1.13 - $2.60 depending on the quality of the materials used. 

With waste oil (used cooking oil) the cost is expected to not 

exceed $1.58. 10 As discussed, the solid fuel pellets created 



Sean Brady 

with the waste glycerin provides a market for the glycerin, 

and may even offset the cost of biodiesel production itself. 

Potential consumers include sealed combustion residential 

heating units and all current RDF pellet customers. 

Economic concerns include both short and long 

term costs. The short term cost from transitioning from 

petroleum to biodiesel and glycerin-based electricity 

production may require a modest investment in capital and 

man-hours, however the long term benefits to a developing 

nation and global prosperity far outweigh the short term 

costs. Global agricultural and energy economies can 

prosper from the increased reliance of biodiesel and its 

glycerin waste. An increased demand in the raw products 

that are required for biodiesel production will positively 

affect existing industries. By utilizing glycerin in energy 

production, dependence on petroleum can be reduced, thus 

reducing greenhouse gases even more. 

Social Impacts 

The main goal of this project is to harness the 

discarded materials of a growing urban biodiesel industry. 

This will lead to an increase in bio-fuel desirability, and 

people will only stand to gain from a new, cheaper fuel that 

can be harnessed within the same infrastructure with which 

they are comfortable. This will help both developed and 

developing nations. 

In developing nations, where modern conveniences 

are not always accessible, many more unorthodox 

methods are already being used to obtain fuels. 11 

Developing countries see bio-fuels as a means to 

stimulate rural development, create jobs, and avoid 

foreign exchange tariffs. 12 

In developed countries, which have significantly 

entrenched fossil fuel infrastructures, biodiesel has already 

been proven to be a significant success. In the United 

States, domestic production is over 30 million gallons a 

year. 13 This fuel has been proven to be very useful, since 

diesel consumption is greater than 60 billion gallons 

per year. 14 

Regardless of the market, biodiesel synthesis 

generates a glycerin waste product, of up to 20% of the 

feed. 15, 16 This waste product is becoming the primary, 

perhaps the only, major problem with biodiesel large-scale 

manufacturing. This begs the question, “What is being done 

with the glycerin waste product?” Fortunately glycerin is 

also biological and biodegradable, as often times the product 

is thrown out or composted. There are some uses for raw 

glycerin in industry; however there are many more uses of 

glycerin after a number of purification steps. Much of the 

one million gallons of waste glycerin produced each year 

is incorporated into functional products, such as soaps and 

cleaning products, but it is essential that these purifications 

be made. Without purification, biodiesel glycerin waste 

contains methanol as well as a significant amount of salt 

content that make it difficult to be placed into isolated 

glycerin reactions. The challenge is to keep purification 

costs low and the process sustainable while still being able 

to produce products that would be less expensive than the 

current methods of glycerin alteration. 

Since both of our input streams are waste streams, 

waste glycerin and waste biomass materials (corn husks 

from the Midwest agriculture), the only costs applicable 

will be for transportation and storage of the material and 

product. This product has the potential to be an industrywide 

coal substitute or another type of RDF, as it is 

already being used in Midwest power plants. 17 Burning 

our glycerin-cellulose product lasts three times longer than 

burning the same amount of regular fire wood. 2 All in all, 

our process proposes to absorb two waste streams into a 

marketable end product that reduces use of coal and leads 

to a environmentally friendly planet. 

Environmental Impacts 

Biodiesel has been shown equivalent to diesel 

not from a performance and efficiency standpoint, but 

has also been shown to cut overall emissions by 45% or 

more, and leading to significant decreases in all non-NOx 

gases. 18 Biodiesel is non-toxic, quickly biodegradable, and 

made from a carbon renewable source. A large number of 

epidemiological studies from different parts of the world 

on air pollution from gasoline have consistently identified 

an association between ambient levels of air pollution and 

various health outcomes, including mortality, exacerbation 

of asthma, chronic bronchitis, respiratory tract infections, 

ischemic heart disease, and stroke. 19, 20 Using this product 

would not only reduce the coal burned, but also lead to an 

increased use of biodiesel, thereby improving air quality and 

reducing the immediate health effects on the population. 



Sean Brady 

At present, the United States requires nearly 

61 billion gallons of diesel fuel; if in the future 

biodiesel makes up 3% of the diesel fuel pool, over 

1.8 billion pounds of glycerol will be produced 

as a byproduct which greatly exceeds the current 

demand of 0.49 billion pounds glycerol. 21 

We hope that our work with glycerin will lead to 

a significant rise in the use of biodiesel not only from 

the surrounding community but in a global sense. Since 

this project is structured around waste products, there is 

no ultimate sacrifice or risk that the society has to make 

concerning biodiesel adoption. 

The biomass glycerin fuel pellets will burn cleaner 

than the coal currently used as fuel in many industries. 

Additionally, two pre-existing waste streams will be 

used to make the pellets, requiring no new raw materials. 

Also the energy required to make the fuel pellets will be 

substantially less than the energy required to produce coal. 

Most importantly, the fuel pellets will be a nearly carbonneutral 

fuel, meaning that the carbon that is released into 

the atmosphere when burned is carbon which has already 

been in the environment. It would not be carbon released 

from fossil fuels, which is carbon that was no longer in 

the environmental carbon cycle, causing the harmful net 

gain of carbon dioxide which has been shown to be overall 

associated with global warming. 22 

Health, Safety, & Hazards Assessments 

Biomass (sawdust, manure, grass clippings and corn 

husks.) is a very benign and chemically-inactive material. 

There are no hazards that are associated with it, aside from 

accidental ingestion and possible splinters during handling. 

Glycerin is also not a health concern. Although glycerin is a 

component in many products and pharmaceuticals, the glycerin 

which will be used in this process in not food grade. Therefore 

basic precautions regarding handling and storing glycerin 

should be followed to avoid unintentional consumption. 

Pellet production is of little health concern, although 

glycerin-biomass pellet combustion may present a hazard. 

Glycerin combustion at low temperatures encouraged the 

formation of acrolein, a toxic gas. Acrolein (2-propenal) 

formation does not occur when under high temperature 

combustion (700°C) characteristic of a plant that burns 

biomass as a source of energy. Acrolein may be cause for 

concern when the biomass glycerin pellet is left to smolder 

or burn at low temperatures (280°C), as might be expected 

in an ornamental residential hearth. 

Acrolein can be deadly in concentrations of 10 parts 

per million. The chemical is toxic if swallowed, inhaled, or 

absorbed through skin, and is a potential carcinogen. The 

vapors of this chemical can be irritating to the eye, nose, 

and throat, and contact to the skin or other body parts can 

cause burns. 

For the residential market, the pellets would only 

be available to the high temperature, sealed, properlyventilated 

furnaces available on the European market. In 

an industrial setting, secondary measures should also be in 

place such as a contained combustion chamber and postprocess 

scrubbers and incinerators, which also mitigate 

environmental risks. Of course traditional emissions such 

as carbon dioxide, carbon monoxide, nitrogen dioxide, and 

similar combustion process emissions are also produced; 

however these are not immediately toxic and should be 

removed by the scrubbers as well. 

Conclusion 

In conclusion, this design project was pursued with the 

hopes of promoting biodiesel and sustainability. In order to 

create an effective energy source, this group has explored the 

combining of two waste streams, biomass and biodiesel waste 

glycerin, to form an energy source that is equivalent to refuse 

derived fuels (RDFs) in energy content. Additionally, this 

process will lessen the impact of refuse on landfills and reduce 

our dependence on fossil fuels. Our project has shown that 

the energy content of the bench scale test is similar with the 

theoretical energy content that was calculated. The verification 

of energy content shown in our bench scale process can easily 

be reproduced in an industrial scale, where a pellet making 

module can be attached to a biodiesel plant. The module has a 

payback period of 10 years where after 10 years a profit will be 

made. The fixed capital investment needed for this attachment 

facility is $4.85 million. This cost is relatively low, making 

this project a feasible idea that can be commercialized, easily. 

With further research and design, this concept for creating 

an effective source of energy from these waste streams can 

be applied and improved further to industry and eventually 

become a viable commercialized product 



Sean Brady 

References 

Endnotes 

1 

Schwab A.W. et al, Preparation and Properties of Diesel 

Fuels from Vegetable Oils, 1987. 

2 

Lloyd A. Nelson, et. Al. Lipase catalyzed production of 

biodiesel, 1996. 

3 

Fukuda et. Al, Biodiesel Fuel Production by 

Transesterification of Oils, 2001. 

4 

Herington, E. F. G. Calorific values of solid, liquid, 

and gaseous fuels. http://www.kayelaby.npl.co.uk/ 

chemistry/3_11/3_11_4.html 2006. National Physical 

Laboratory. 01 Sept. 2006. 

5 

Crumpler, Paul “Industrial Wood Waste,” From the 

Source - Fall 1996. 

6 

T. Cherry, J. Shorgun. The Social Costs of The Social 

Cost of Coal: A Tale of Market Failure and Market 

Solution, Working Papers, Department of Economics, 

Appalachian State University Sept. 2002. 

7 

Akers, S., Conkle, J., Thomas, S., Rider, K. Determination 

of the Heat of combustion of Biodiesel Using Bomb 

Calorimetry. Journal of Chemical Education. 2006 83, 2. 

8 

Lambert, M. Excess Properties of H2 – D2 Liquid 

Mixtures. Physical Review Letters. 4, 11 (1960):55-56 

9 

The Bomb Calorimeter consisted of a 2’x1’x1.5’styrofoam 

ice chest filled with 200 mL of water. Weighed quantities 

of the glycerin/biomass mixture was floated in an 

aluminum foil boat, and the ice chest lid was placed 

on top and taped. The atmosphere was flushed and 

replaced entirely with pure oxygen, and then the pellets 

were ignited with a squib. The water temperature was 

measured prior to ignition, and after combustion. Many 

iterations were needed to come up with this one design 

that worked somewhat properly. 

10 

Bender, M. Economic feasibility review for communityscale 

farmer cooperatives for biodiesel. Bioresource 

Technology 1999, Vol. 70, 1:81-87 13 Oct. 2006 

11 

Heath, Jim. How about bio gas? Isle of Man Weekly 

Times Jan. 2 2004. 

12 

M. Kojima and T. Johnson, Potential for Biofuels for 

transport in Developing Countries. ESMAP, October 

2005, Washington, D.C. U.S.A. 

13 

Suppes, Galen. Chemical Engineering Professor 

Develops New Biodiesel Process. 2006 Research at MU: 

News & Press Release. 20 Sept 2006. 

14 

Eckhardt, Angela. Freedom Fuel: How and why biodiesel 

policy should reflect freedom. Rural Oregon Freedom 

Project, 26 Oct. 2006. 

15 

Briggs, Michael. Wide Scale Biodiesel Production from 

Algae. 2004. http://www.unh.edu/p2/biodiesel/article_ 

alge.html UNH Biodiesel Project. 14 Sept. 2006 

16 

Thompson, J.C., He, B.B. Characterization of Crude 

Glycerol from Biodiesel Production from Multiple 

Feedstocks. Applied Engineering in Agriculture Vol. 22, 

2:261-265 

17 

McElroy, A. K., Jesson, H., Zeman, N., Kotrba, R., Nillas, 

D. Proposed Biodiesel Plant List. Biodiesel Magazine. 

June. 2006: id=943 

18 

Norton, Patricia. OSWER Innovations Pilot: Reducing 

Production Costs and Nitrogen Oxide (NOx) from 

Biodiesel. EPA, 2004 18 Sept. 2006. 

19 

A. Sydbom, A. Blomberg, S. Parnia, N. Stenfors, T. 

Sandström and S-E. Dahlén. Health effects of diesel 

exhaust emissions. Eur Respir J 2001; 17:733-746 

20 

Salvi S., Holgate S. Mechanisms of particulate matter 

toxicity. Clin. Exp. Allergy 1999; 29: 1187-1194 

21 

Virent Energy Systems. Conversion of Glycerol Stream 

in a Biodiesel 2004 http://www.virent.com/whitepapers/ 

Biodiesel%20Whitepaper.pdf Virent Energy Systems 

Inc. 24 August 2006. 

22 

Global Warming Basics Pew Center on Global Climate 

Change. www.pewclimate.org/global-warming-basics/ 

February 18, 2007 



Computational Prediction of Association Free Energies for the 

C3d-CR2 Complex and Comparison to Experimental Data 

Alexander S. Cheung, Dimitrios Morikis 

Graduate Student Assistants: Jianfeng Yang, Chris A. Kieslich 

Department of Bioengineering 


Abstract 

The complement system functions to clear pathogenic threats from the body and is part of the innate 

immune system. The association between complement protein C3d and B or T cell-receptor CR2 

(complement receptor 2) represents a crucial link between innate and adaptive immunities. The 

goal of this study is to computationally predict association abilities of C3d and CR2 mutants by 

theoretically calculating electrostatic free energies of association with and without solvation effects. 

We demonstrate that incorporation of solvation effects is necessary to accurately predict previously 

published experimental data for the association abilities (relative to the parent proteins) of specific C3d 

and CR2 mutants. We show that a proportional relationship exists between the predicted solvation 

free energy differences and the experimental data. Additionally, an inversely proportional relationship 

is demonstrated between the calculated solvation free energy differences and previously calculated 

ionization free energy differences. Our results yield new insights into the physicochemical properties 

underlying C3d-CR2 association. Moreover, our results can also be extended to any complex with 

excessively charged components. This is a basic study, aimed toward understanding the theoretical 

basis of immune system regulation at the molecular level, which can be the groundwork for the 

design of regulators with tailored properties, vaccines, and biotechnology products in general. 

Faculty Mentor 

Dimitrios Morikis 


Alexander joined my lab during his freshman year, being very enthusiastic and 

ambitious about doing research. During his first two years, he matured as a 

researcher by learning about research methodology, biomolecular modeling and 

simulation, and immune system function and regulation. As a Junior, with all the 

core requirements in mathematics, physics, chemistry, and biology, he was ready to undertake a 

project of higher challenge. He did so with great success as indicated by the current publication. 

Alexander used a computational protocol implemented by graduate student Jianfeng Yang, ’07, 

and modified by graduate student Chris Kieslich, both coauthors in this paper, to construct a 

number of theoretical mutants and to predict the association abilities for an immune system protein 

and its receptor. He performed a meticulous analysis of his data in view of previous experimental 

work and theoretical calculations. This type of work is important for the design of immune system 

therapeutics. Alexander has the ability to process large amounts of complex information. Although 

he is a quiet participant in a typical research discussion, he comes back to our next meeting with 

data from well-planned and well-executed calculations and with impeccable presentations. 

A U T H O R 

Alexander S. Cheung 

Bioengineering 

Alexander Cheung is a third year Bioengineering 

student. He is a Chancellor’s 

Scholar and is a member of the Medical 

Scholars Program and the Tau Beta Pi 

National Engineering Honors Society. He 

has worked in the Morikis lab since his 

first year, primarily focused on studying 

the physicochemical properties of 

complement system proteins through 

the use of computational methods in 

order to better understand the molecular 

basis of immune system function and 

regulation and the molecular origins of 

autoimmune diseases. This summer he 

will perform research at the University 

of California, San Francisco, as part of 

the UCSF SRTP program. Alexander’s 

goal for next year is to obtain hands-on 

wet lab experience by bridging his theoretical 

predictions on immune system 

inhibition with experimental validation. 

After completing his undergraduate 

degree, Alexander is planning to attend 

graduate school. 


Computational Prediction of Association Free Energies for the C3d-CR2 Complex and Comparison to Experimental Data 


Introduction 

The immune system of higher vertebrates is made 

up of both innate and adaptive immunity [1]. The innate 

immune system functions primarily through the activity of 

leukocytes that indiscriminately work to rid the body of 

any foreign pathogenic substances. Activated by the innate 

immune system, the adaptive immune system is principally 

made up of memory B and T cells. These cells not only 

combat pathogenic threats, but also have the ability to 

remember specific pathogens and thus mobilize more 

rapidly and launch stronger attacks against the pathogen 

each time it is again encountered. By efficiently working 

together in concert, the innate and adaptive immune 

systems protect the body against disease. 

The complement system is a key component of 

innate immunity but also serves as a link between innate 

and adaptive immunities [2]. The complement system 

is activated through a complex cascade of catalytic 

reactions which involve cleavage of complement proteins 

into fragments and formation of protein complexes. The 

association of the d-fragment of complement component 

C3 (C3d), a globular serum protein, with complement 

receptor 2 (CR2), a modular B or T cell surface receptor, 

is a crucial link between the innate and adaptive immune 

systems. CR2 consists of 15 modules, called complement 

control protein (CCP) modules, of which only the first two 

modules, CCP1 and CCP2, are known to interact with C3d 

([3] and references therein). In this paper we investigate 

the interaction between C3d and the first two modules of 

CR2, herein referred to as CR2. 

There are numerous studies in literature that implicate 

charge and electrostatics as having a role in the interaction 

between C3d and CR2 [3-10]. It has been established that 

electrostatics are essential for many biological functions, 

including protein-ligand interactions [3-5], protein stability 

[4,11], catalysis [12], conformational transitions [13], and 

protein ionization [3-5,11-14]. In previous studies, the 

Morikis group has proposed that electrostatics drive C3d- 

CR2 recognition [3-5] and used the ionization properties 

of C3d and CR2 to demonstrate a correlation between 

ionization free energy differences, and association data 

from experimental mutagenesis studies [5]. Based on these 

studies, a two-step model for association was proposed, 

with the first step being recognition, and the second, 

binding [3-5,14]. According to this model, recognition 

is driven solely by long-range electrostatic interactions 

whereas binding involves long- and short/medium-range 

electrostatic interactions (including hydrogen bonds, salt 

bridges, and van der Waals forces) as well as short-range 

hydrophobic interactions and entropic effects. Recognition 

is responsible for the formation of a weak, nonspecific 

encounter complex, whereas binding is responsible for 

the formation of a strong and specific final complex. 

Conventional thinking dictates that only mutations at the 

binding interface should affect binding abilities. However, 

the study by Zhang et al [5] demonstrated that mutations 

remote from the association interface can also affect binding 

abilities (evaluated as ionization free energy differences), 

explaining controversial experimental data that previously 

seemed contradictory. This effect of distant mutations on 

association is possible only if recognition, as an initial step 

in association, involves electrostatic attraction between 

protein macrodipoles to form the encounter complex. 

Thus, the results of the study performed by Zhang et al [5] 

validated the two-step model for C3d-CR2 association. 

The goals of our study are to examine whether 

similar correlations exist between electrostatic free energies 

of association and relative binding ability as well as to 

evaluate the effect of solvation on the electrostatic free 

energies of association and examine whether correlations 

exist between solvation free energy differences and relative 

binding ability. We analyzed the effect of the same C3d 

and CR2 mutations as in the study by Zhang et al [5] on 

C3d-CR2 association. These are mutations for which there 

are published experimental data for C3d [9] and CR2 [6]. 

Figure 1. Molecular representation of C3d-CR2 demonstrating the 

topology of the mutated amino acids. Basic and acidic amino acids 

are shown in blue and red, respectively. Glu116 was used as the 

contact amino acid in association interface and is colored in green. 




Figure 1 shows a molecular model of the backbone trace for 

the C3d-CR2 complex with the side chains to be mutated in 

stick representation. As shown in Fig. 1, the range of these 

mutations spans the entire volume of the C3d-CR2 complex. 

Side chains in blue denote basic amino acids whereas those 

in red denote acidic amino acids. Since C3d is excessively 

negatively charged (total charge –4e), most of the amino 

acids selected by the experimentalists for mutation were 

acidic so that the effect of a loss in this excess of negative 

charge [9] on association could be evaluated. Similarly, in 

CR2 which is excessively positively charged (total charge 

+8e), only basic amino acids were selected for mutation 

by the experimentalists in order to evaluate the effect of 

a loss in this excess of positive charge on association [6]. 

Nonpolar solvation energy is not considered in this article 

but will be calculated for this ongoing study in the future. 

Methods 

Electrostatic calculations in this study are based on 

the solution of the linearized Poisson-Boltzmann equation 

[15], given by 

where ϕ is the dimensionless electrostatic potential in units of 

k B 

T/e, ε is the dielectric coefficient, ε 0 

is the vacuum permittivity, 

κ is the ion accessibility function, q is the charge within the 

protein, e is the electron charge, k B 

is the Boltzmann constant, 

and T is the absolute temperature. The ion accessibility function 

relates to ionic strength, I, according to 

(2) 

(1) 

The ionic strength relates to ion concentration, c, according to 

where c i 

is concentration for ion type i, and z i 

is the ion 

charge (or valence). Parameters ϕ, ε, κ, and q are position 

dependent. The calculation methodology is based on 

embedding the protein in a grid and assigning values 

for q, ε, and κ at each grid point. The charge, q, refers 

to discrete charges within the protein. These are unit 

charges of acidic or basic amino acids (for physiological 

pH: Asp, Glu, Lys, Arg, His, and N- and C-termini) and 

(3) 

Figure 2. Flowchart of our computational protocol. 

partial charges of chemical groups having electric dipole 

moments (e.g., amide, groups, carbonyl groups, etc). The 

dielectric coefficient, ε, is assigned two discrete values: 

one for the protein interior (within the molecular surface; 

typically in the range of 2-20) and another for the protein 

exterior (solvent; about 80). The ion accessibility, κ, is 0 

within the ion accessibility surface and assumes values 

according to Eqs. (2) and (3) outside the ion accessibility 

surface. The protein molecular surface is defined using a 

probe sphere with a radius of 1.4 Ǻ (representing a water 

molecule) whereas the ion accessibility surface is defined 

using a probe sphere with a radius of 2.0 Ǻ (representing 

salt ions such as Na + or Cl − ). The Poisson-Boltzmann 

equation is used to calculate values for ϕ at each grid point. 

Electrostatic potentials are converted to electrostatic free 

energies according to 

where ρ refers to the charge density of both solute and 

solvent charges. 

Our computational protocol is summarized in Figure 

2. The first step was to obtain the three-dimensional atomic 

coordinate file (PDB file) with Code 1ghq from the Protein 

Data Bank (PDB) [16]. This structure was derived from 

X-ray diffraction data [7]. 

The second step (Fig. 2) was the manual removal 

of unnecessary information in the PDB files, including the 

PDB file header and footer, water and heteroatom entries 

(zinc ions and D-acetyl-N-glucosamine), and the N-terminal 

(4) 




methionine of C3d which is an artifact added by the protein 

expression system. One of the chains in the PDB file, chain 

C, a CR2 molecule that is not in contact with C3d, was also 

removed. This is because chain C is irrelevant in this study, 

since it is known that CR2 behaves as a monomer in solution 

[7]. What remains is a C3d molecule, consisting of 306 amino 

acids, in contact with a CR2 molecule, consisting of 129 

amino acids. We used the program SPDBV (Swiss Protein 

Data Bank Viewer) [17], version 3.7, to renumber the amino 

acids and atoms in the PDB file (displacing every atom by 

-1, resulting in a total of 435 amino acids consisting of 3399 

atoms) and subsequently separate the two components of the 

complex to create one PDB file with C3d alone and one PDB 

file with CR2 alone. After separation, amino acids and atoms 

were renumbered for CR2 to begin at residue 307 and atom 

2413. Thus three PDB files constitute the final output of 

this step: one PDB file consisting of the C3d-CR2 complex, 

one consisting of C3d, and one consisting of CR2. These 

three PDB files are considered the “parent” PDB files. By 

generating each of the component parent PDB files from the 

complex parent PDB file in this way, it is ensured that the 

atomic coordinates of each component of the complex in 

each of their respective component files are identical to their 

atomic coordinates in the complex file. This is crucial for 

the accurate calculation of free energy differences. Finally, 

we used the program WHATIF [18] to add the missing 

C-terminal oxygen atom of C3d in the C3d-CR2 complex. 

The third step (Fig. 2) was the construction of the 23 

specific mutants. This is done using WHATIF, a home-made 

python script [19] that calls WHATIF, the three parent PDB 

files, and three input text files, each of which list the amino 

acid substitutions to be made in one of the three parent PDB 

files. Each of the 23 mutations had to be performed twice: 

once on the complex PDB file and again on the individual 

component file containing the mutation(s). Thus, the script 

was run three times, each time using as inputs one of the 

three parent PDB files and its corresponding input text file. 

For the purpose of consistency, each of the parent PDB files 

was also run through WHATIF manually without making any 

mutations. The outputs of this step are 23 mutant complex 

PDB files, 9 mutant C3d PDB files, 14 mutant CR2 PDB 

files, and the three parent PDB files (49 PDB files total). 

These 49 PDB files comprise 24 sets (parent and 23 mutants) 

of 3 PDB files each (C3d, CR2, complex). 

The fourth step (Fig. 2) was the removal of a 

WHATIF-specific header added to each PDB file in the 

last step, and the change of the nomenclature of C-terminal 

oxygens from O’’, which is recognizable by WHATIF, to 

OXT, which is recognizable by PDB2PQR (to be used in 

the next step). These two tasks are accomplished using 

home-made python scripts [19]. 

The fifth step (Fig. 2) involved the use of the 

program, PDB2PQR [20] 1.2.1, to prepare the coordinate 

PDB files for use with APBS (see below). Through the use 

of a home-made python script, each of the 49 PDB files was 

run through PDB2PQR. The outputs were 49 files in PQR 

format, each containing three-dimensional atomic coordinate 

data as well as charge and van der Waals radii assigned 

according to the PARSE parameter file [20]. The default 

options for debumping and hydrogen bond optimization 

were left on. Debumping refers to local optimization to 

eliminate unfavorable van der Waals clashes (overlap or 

partial overlap of atomic radii). The hydrogen bond network 

optimization algorithm assures that optimal hydrogen bonds 

are present by 180 o -flipping the rings of histidine or of planar 

amine groups of glutamines or asparagines. This option is 

necessary because electron densities from X-ray diffraction 

data do not discriminate between the 0 o - and 180 o -flip states 

of these amino acid side chains. 

The purpose of steps 1-5 described above is to 

create the proper input files for use with the program 

APBS (Adaptive Poisson-Boltzmann Solver) [22]. The 

sixth step was calculation of electrostatic potentials using 

Figure 3. Hypothetical thermodynamic cycle. Horizontal processes 

represent association in vacuum (top) and in solution (bottom). 

Vertical processes represent solvation of the components (left) 

and of the complex (right). Electrostatic potential surfaces are 

visualized at ±30 kT/e for association in vacuum (top) and at ±1 

kT/e in solution (bottom). 




APBS. The inputs for an APBS calculation are a PQR file 

and an input text file with the calculation parameters. The 

electrostatic potential of the complex and of each of the 

individual components was calculated at two different 

conditions as shown in Fig. 3. For each APBS calculation, 

the protein or protein complex was embedded in a box with 

129 × 129 × 129 grid points having coarse grid dimensions 

of 140 Ǻ × 110 Ǻ × 120 Ǻ and fine grid dimensions of 105 

Ǻ × 85 Ǻ × 90 Ǻ. The grid dimensions were chosen by 

performing preliminary test calculations using structures 

of the complex to ensure that there was no truncation of 

the largest electrostatic potential when plotted at ±1 k B 

T/e. 

For each of the 24 structures, two sets of three calculations 

were performed, one set being in vacuum and the other 

in a realistic protein-solvent environment. Each of the 

two sets included calculations for the complex and the 

two individual components. The vacuum calculations 

were performed using the same dielectric constant for 

the protein interior and solvent (ε p 

= ε s 

= 2) and at ionic 

strength corresponding to 0 mM ionic strength. The proteinsolvent 

calculations were performed using low dielectric 

constant for the protein interior (ε p 

= 2) and high dielectric 

constant for the solvent (ε s 

= 78.5), and at ionic strength 

corresponding to 150 mM ionic strength. To eliminate grid 

artifacts when comparing the results of the calculations, 

the individual components, C3d 

and CR2, were each positioned 

in the grid exactly as they 

were in the C3d-CR2 complex, 

respectively. A home-made 

python script [19] was used to 

automatically perform a set of 24 

calculations (for parent proteins 

and mutants). Each of the 24 

calculations includes a subset 

of 6 calculations, as described 

above. Each of the 24 APBS 

calculations generates a file 

with the electrostatic potential 

matrix and a log (OUT; Fig. 2) 

file describing the calculation 

progress and providing the 

electrostatic free energy of 

association in solution and the 

solvation free energy difference. 

The seventh step (Fig. 2) was visualization of the 

electrostatic potentials using the program VMD (Visual 

Molecular Dynamics) [23] version 1.8.5 and data analysis 

using MATLAB (The Mathworks, Inc., Natick, MA) version 

r2007b. Distances between mutations and the association 

site contact residues were measured using SPDBV. 

In earlier stages of this study, we performed the 

calculations manually, making mutations with SPDBV and 

performing the conversion of the PDB files to PQR files using 

the online version of PDB2PQR version 1.3.0 (http://agave. 

wustl.edu/pdb2pqr/index.html). Variations in the calculated 

electrostatic free energies of association in solution (without 

solvation effects) between the script-based high-throughput 

protocol and the manual online-based protocol of up to 21% 

were observed. Variations of up to 3% were observed in the 

solvation free energy differences. 

Results 

Figure 3 describes the hypothetical thermodynamic 

cycle we used in our calculations of electrostatic free 

energies of association. The horizontal steps describe 

association in vacuum (top) and in solution (bottom) and 

the vertical steps describe solvation of the free components, 

C3d and CR2 (left), and of the C3d-CR2 complex (right). 

Table 1. List of mutations, calculated solvation free energy differences, calculated association 

free energies in solution, experimental binding ability data, previously calculated ionization free 

energy differences, and distances of mutated residues from the association site. 




Association in vacuum is described simply by Coulomb’s 

law, using the same dielectric coefficient for protein and 

solvent in the absence of ions (ε p 

= ε s 

= 2; κ = 0). Association 

in solution is described by the Poisson-Boltzmann equation 

with different dielectric coefficients for the protein and 

solvent in the presence of ions (ε p 

= 2; ε s 

= 78.5; κ ≠ 0). 

Solvation describes the transfer of the protein or protein 

complex from vacuum into solution. 

The electrostatic free energy of association is given by 

(5) 

where 

, with tip and base referring to 

the direction of the arrows in the thermodynamic cycle. We 

solve Eq. (5) for 

Table 1 lists the mutants (also shown graphically 

in Fig. 1), solvation free energy differences (Eq. 5), 

calculated electrostatic free energies of association in 

solution without solvation effects (Eq. 6), experimental 

binding ability data for C3d mutants [9] and CR2 

mutants [6], previously published calculated ionization 

free energy differences [5], and calculated distances 

of each mutated amino acid from the association site 

contact residues. The distances were calculated using 

the C3d-CR2 structure (Fig. 1) and Glu116 (in C3d) and 

Arg390 (in CR2) as the association site contact residues. 

Distances were measured between the central atoms of 

the ionization sites: C γ for Asp, C δ for Glu, N ζ for Lys, C ζ 

and for Arg. Experimental binding abilities are relative 

to the parent proteins [6,9]. The key for the experimental 

binding abilities is as follows: +++++, 2-fold increase; 

(6) 

which represents the solvation free energy difference 

upon C3d-CR2 association between solution and vacuum 

environments. 

In order to perform this calculation, we must 

know , , and , which are 

determined by calculating the 6 electrostatic free energies 

for C3d, CR2, and C3d-CR2 in vacuum and in solution and 

by taking their differences according to the thermodynamic 

cycle of Fig. 3 and Eqs. (5) and (6). As can be seen in Eq. 

(6) the solvation free energy difference, ΔΔG solvation , 

is equal to the electrostatic free energy of association in 

solution, 

minus the Coulombic free energy 

of association in vacuum, (also seen in the 

thermodynamic cycle of Fig. 3). 

To assess the effect of solvation in association, we 

also calculated the electrostatic free energy of association 

in solution alone (bottom horizontal step only in Fig. 3) 

according to 

(7) 

The values of and as 

described by Eqs. (6) and (7) were calculated 24 times: 

once for the parent proteins and once for each of the 23 

sets of mutant protein. 

Figure 4. (A) ΔΔG solvation (calculated using the complete 

thermodynamic cycle of Fig. 3) versus relative association ability 

with a linear fit drawn in red. (B) 

(calculated using 

the bottom horizontal reaction only of the thermodynamic cycle 

of Fig. 3) versus relative association ability. 




++++, 90–120%; +++, 70–90%; ++, 40–70%; +, 20–40%; 

−, 0–20%. Because CR2 consists of two CCP modules, 

CCP1 and CCP2, connected by a flexible loop, we list in 

Table 1 the location of each mutation in CR2. It should be 

noted that only CCP2 is in contact with C3d in the crystal 

structure. However, we demonstrate in our study that 

mutations in CCP1 and the loop, though remote from the 

binding interface, also affect the association process. 

Figure 4A illustrates a proportional relationship 

between the calculated ΔΔG solvation (Eq. 6) and 

experimental relative association ability. In general, as 

increases, the association ability is also 

observed to increase. The correlation coefficient between 

these two sets of values was found to be 0.7, representing 

a solid relationship between ΔΔG solvation and the relative 

association ability. Figure 4B shows (Eq. 7) 

also plotted against the experimental values of relative 

association ability. The correlation coefficient between 

these two sets of values was found to be 0.0, indicating that 

there is no proportionality relationship between them. This 

is shown visually in Fig. 4B. This result is significant as it 

illustrates that simply calculating by itself is not 

sufficient to predict association abilities since no correlation 

was observed between the values and association 

abilities. This result demonstrates the necessity to calculate 

the complete thermodynamic cycle, including effects such 

as solvation, in order to obtain reasonable predictions for 

association ability from free energy calculations. 

Figure 5(A,B) shows the value of the change in 

ΔΔG solvation for each mutant relative to that of the parent 

plotted against the respective distance between association 

site contact residue Glu116, and the amino acid mutated 

in each of the mutants (distances listed Table 1). Figures 

5A and B show that although a somewhat larger effect 

is observed when the amino acid mutated is closer to the 

association site, mutation of amino acids remote from the 

association interface can also result in a sizeable change in 

the solvation free energy difference. From Fig. 5, it can be 

seen that regardless of distance from the association site, the 

magnitude of the change in ΔΔG solvation of the majority of 

the mutants lies between about 50-150 kJ/mol, particularly 

in CR2. The magnitude of the change in ΔΔG solvation for 

mutations made in C3d is a somewhat larger, but this more 

pronounced effect is most likely the result of CR2 having a 

higher magnitude of excess charge (8e) than C3d (4e). Thus, 

Figure 5. Plot of (ΔΔG solvation ) x 

– (ΔΔG solvation ) parent 

versus the distance between the mutated amino acid and 

association site contact residue Glu116, where X denotes the 

mutant. (A) C3d mutants. (B) CR2 mutants. (B) Acidic and basic 

amino acids are labeled in red and blue. 

because all the amino acids that were mutated were charged 

residues, the mutations had a larger effect in C3d which has 

less excess charge to compensate for the mutation. Mutations 

had a less dramatic effect in CR2 than in C3d since CR2 has 

a charge excess magnitude that is twice as large enabling it to 

better compensate for and effectively mask, to some extent, 

the effects of the mutation. It can also be seen from Fig. 

5A,B that in C3d, which is excessively negative, an increase 

in ΔΔG solvation is observed when a basic residue is mutated 

to an uncharged residue (Ala), increasing the magnitude of 

its excess charge. Likewise, a decrease in ΔΔG solvation 

is observed when an acidic residue is mutated, effectively 

decreasing the magnitude of its excess charge. A similar 

trend is observed in CR2. When a basic residue is mutated 

to an uncharged residue (Ala), effectively decreasing the 




magnitude of the excess charge in CR2 by 1e, a decrease 

in ΔΔG solvation is observed. Additionally, when a basic 

residue is mutated to an acidic residue, an even greater 

decrease in ΔΔG solvation is observed since such a mutation 

decreases the magnitude of the excess charge in CR2 by 2e. 

Intuitively, these results correlate well to those 

illustrated in Fig. 4A. In Fig. 4A, a proportional relationship 

is observed between ΔΔG solvation and association ability. 

In Fig. 5A,B, an increase in ΔΔG solvation was observed 

when the mutation increased the magnitude of the excess 

charge on the component (e.g. mutation of a basic amino 

acid to a neutral amino acid in C3d), while a decrease in 

ΔΔG solvation was observed when the mutation diminished 

the magnitude of the excess charge on the component (e.g. 

mutation of and acidic amino acid to a neutral amino acid in 

C3d). Because C3d and CR2 have opposite excess charge, 

an increase in the magnitude of the excess charge on 

either component can increase the long-range electrostatic 

attraction between the two components, speeding up 

formation of the encounter complex and thus ultimately 

resulting in an increase in binding ability. 

Similar plots have been generated in which the change 

in ΔΔG solvation is plotted against the distance between 

Arg390, the CR2 residue at the association site (not shown; 

data in Table 1). Trends similar to Fig. 5 were observed. 

Discussion 

The interaction of complement protein fragment C3d 

with the first two modules of CR2 has been the subject of 

many intensive studies among immunologists, because it is an 

essential link between innate and adaptive immunities. Earlier 

efforts by the Morikis Group have focused on shedding light 

on the underlying structural and physicochemical properties 

underlying C3d-CR2 association [3,5]. In particular, a twostep 

model was proposed according to which electrostatics 

drive the nonspecific recognition between C3d and CR2 and 

contribute to their specific binding. An earlier study by Zhang 

et al [5] presented calculations of ionization free energy 

differences as a function of pH, ionic strength, and mutations, 

which compared favorably with experimental data. 

In the present study, we examine the effect of 

solvation on the electrostatic free energies of association. 

We have calculated electrostatic free energies of association 

using two different methods. The first calculation 

produced a difference between electrostatic free energy of 

association and Coulombic free energy of association in 

vacuum, using the complete thermodynamic cycle of Fig. 

3, which incorporates solvation effects (vertical processes) 

in the calculation. The second calculation produced the 

electrostatic free energy of association in solution by using 

the bottom horizontal process only, which does not account 

for solvation. We have performed our calculations for the 

parent complex and the same mutants as in [5] for which 

there are published experimental association data [6,9]. We 

have demonstrated that the inclusion of solvation effects 

is necessary to accurately predict association abilities, by 

comparing our theoretical data to the experimental data. 

This is shown in Fig. 4 where a positive correlation is 

observed between the predicted solvation free energy 

differences and the experimental binding ability data (Fig. 

4A). In contrast, no correlation is observed when solvation 

effects are not taken into account (Fig. 4B). 

Moreover, we compare our calculated solvation free 

energy differences to the ionization free energy differences 

of Zhang et al [5] (Fig. 6). A solid correlation is observed 

between the two sets of data, with a correlation coefficient 

of -0.8. The negative sign indicates that solvation effects are 

unfavorable while ionization free energies are favorable. 

Inclusion of favorable Coulombic interactions (top horizontal 

process in the thermodynamic cycle) are expected to produce 

overall electrostatic free energies of association, which, when 

compared to the ionization free energy differences, will yield 

Figure 6. ΔΔG ionization versus ΔΔG solvation . ΔΔG ionization is from [5]. A 

linear fit is drawn in red. 




a positive correlation coefficient. This additional calculation 

is work in progress. Previous studies have demonstrated that 

the interaction between C3d and CR2 is driven primarily by 

electrostatics [3,5-10] and this is the focus of our present study. 

However, we are planning to incorporate nonpolar effect in 

the free energies of association. Nonpolar interactions are 

also present as in any binding process, but they are operative 

only at the binding step and involve the limited binding 

interface, without contributions from the rest of the protein 

volumes. Nonpolar contributions to the association free 

energies are modeled as differences in the solvent accessible 

surface area (SASA) of the two proteins upon association 

multiplied by a surface tension constant (γ), ∆G nonpolar = γ 

∆SASA. Upon complex formation there is loss in SASA 

of each of the components, corresponding to the surface 

areas that form the binding interface. As such, only mutated 

residues located at the binding interface would produce 

unique changes in SASAs. This is because upon association, 

compared to their free states, residues at the interface will 

be deprived from solvated environments. On the other hand, 

residues that are not located at the binding interface will be 

subject to the same solvation environments in both their free 

and bound states, and thus differences in SASAs produced 

by mutations are cancelled out when calculating ∆SASAs. 

For the purposes of this article, as only one of the mutated 

residues is located at the binding interface (Arg390) and 

thus is the only mutant that will result in a unique change 

in SASA, for all other mutants, the nonpolar free energy 

represents only a change in the calculated electrostatic free 

energies by a uniform constant. 

Our calculations are consistent with the previously 

proposed two-step model for C3d-CR2 association [3,5]. 

According to this model, the amino acids that affect C3d- 

CR2 association are not only located at the association 

interface, but also remote from the association interface. 

This is because all amino acids throughout the volumes 

of components of the complex contribute to their overall 

electrostatic potential, which is excessively negative for 

C3d and positive for CR2. Elimination of even a single 

charge through mutation to a neutral amino acid, regardless 

of the residue’s position, could affect the spatial distribution 

of the electrostatic potential and thus the recognition ability 

of the protein. This is shown in Fig. 5 which demonstrates 

that mutations distant from the association interface can 

still have a sizeable effect on the relative magnitude of the 

solvation free energy differences with respect to the parent 

protein complex. 

Collectively, the results of this study provide 

considerably better insight into the physicochemical 

properties underlying C3d-CR2 association. Further 

analysis of C3d-CR2 association will be performed, in 

which the vacuum interactions will be explicitly calculated 

using Coulomb’s law, and nonpolar contributions will be 

incorporated into the free energy by calculating the loss of 

solvent-accessible surface area upon binding. 

Acknowledgments 

This work was supported by NSF grant 0427103 

and an REU (Research Experience for Undergraduates) 

Supplement 0611503 (DM). 

References 

1. Alberts, B, Johnson, A, Lewis, J, Raff, M, Roberts, 

K, & Walter, P (2002) Molecular Biology of the Cell, 

Garland Science, New York, NY. 

2. Morikis, D & Lambris, JD (2005) Structural Biology 

of the Complement System, Taylor & Francis/CRC 

Press, Boca Raton, FL. 

3. Morikis, D, & Lambris, JD (2004) The electrostatic 

nature of C3d-complement receptor 2 association, 

Journal of immunology 172:7537-7547. 

4. Morikis, D & Zhang, L (2006) An immunophysical 

study of the complement system: examples for the pH 

dependence of protein binding and stability, Journal of 

Non-Crystalline Solids 352:4445-4450. 

5. Zhang, L, Mallik, B, & Morikis, D (2007) 

Immunophysical exploration of C3d-CR2 interaction 

using molecular dynamics and electrostatics, Journal 

of Molecular Biology 369:567-583. 

6. Hannan, JP, Young, KA, Guthridge, JM, Asokan, 

R, Szakonyi, G, Chen, XJS, & Holers, VM (2005) 

Mutational analysis of the complement receptor type 

2 (CR2/CD21)-C3d interaction reveals a putative 

charged SCR1 binding site for C3d, Journal of 

Molecular Biology 346:845-858. 

7. Szakonyi, G, Guthridge, JM, Li, DW, Young, K, Holers, 

VM, & Chen, XJS (2001) Structure of complement 

receptor 2 in complex with its C3d ligand, Science 




292:1725-1728. 

8. Guthridge, JM, Rakstang, JK, Young, KA, 

Hinshelwood, J, Aslam, M, Robertson, A, Gipson, 

MG, Sarrias, MR, Moore, WT, Meagher, M, Karp, 

D, Lambris, JD, Perkins, SJ, & Holers, VM (2001) 

Structural studies in solution of the recombinant 

N-terminal pair of short consensus/complement repeat 

domains of complement receptor type 2 (CR2/CD21) 

and interactions with its ligand C3dg, Biochemistry 

40:5931-5941. 

9. Clemenza, L, & Isenman, DE (2000) Structureguided 

identification of C3d residues essential for its 

binding to complement receptor 2 (CD21), Journal of 

Immunology 165:3839-3848. 

10. Nagar, B, Jones, RG, Diefenbach, RJ, Isenman, DE, 

& Rini, JM (1998) X-ray crystal structure of C3d: A 

C3 fragment and ligand for complement receptor 2, 

Science 280:1277-1281. 

11. Mallik, B, Zhang, L, Koide, S, & Morikis, D (2008) pH 

dependence of stability of the 10th human fibronectin 

type III domain: a computational study, Biotechnology 

Progress 24:48-55. 

12. Morikis, D, Elcock, AH, Jennings, PA, & McCammon, 

JA (2001) Native state conformational dynamics of 

GART: a regulatory pH-dependent coil-helix transition 

examined by electrostatic calculations, Protein Science 

10:2363-2378. 

13. Morikis, D, Elcock, AH, Jennings, PA, & McCammon, 

JA (2001) Proton transfer dynamics of GART: 

the pH-dependent catalytic mechanism examined 

by electrostatic calculations, Protein Science 

10:2379-2392. 

14. Zhang, L, & Morikis, D (2006) Immunophysical 

properties and prediction of activities for vaccinia virus 

complement control protein and smallpox inhibitor of 

complement enzymes using molecular dynamics and 

electrostatics, Biophysical Journal 90:3106-3119. 

15. Wu, J & Morikis, D (2006) Molecular thermodynamics 

for charged biomacromolecules, Fluid Phase Equlibria 

241:317-333. 

16. Berman, HM, Westbrook, J, Feng, Z, Gilliland, G, 

Bhat, TN, Weissig, H, Shindyalov, IN, & Bourne, PE 

(2000) The Protein Data Bank, Nucleic Acids Research 

28:235-242. 

17. Guex, N, Peitsch, MC (1997) Swiss-model and the 

Swiss-Pdb-Viewer: an environment for comparative 

protein modeling, Electrophoresis 18:2714-2723. 

18. Vriend, G (1990) WHAT IF: A molecular modeling and 

drug design program, Journal of Molecular Graphics 

8:52-56. 

19. Yang, J (2007) Implementation of a high throughput 

computational protocol for calculation and clustering 

of electrostatic potentials of protein families, M.S. 

Thesis, University of California, Riverside. 

20. Dolinsky, TJ, Nielsen, JE, McCammon, JA, & 

Baker, NA (2004) PDB2PQR: an automated pipeline 

for the setup of Poisson-Boltzmaann electrostatics 

calculations, Nucleic Acids Research 32:W665-W667 

21. Sitkoff, D, Sharp, KA, & Honig, B (1994) Calculation 

of alkane to water solvation free energies using 

continuum solvent models, Journal of Physical 

Chemistry 98:1978–1988. 

22. Baker, NA, Sept, D, Joseph, S, Holst, MJ, & 

McCammon, JA (2001). Electrostatics of nanosystems: 

application to microtubules and the ribosome. 

Proceedings of the National Academy of Sciences 

(USA) 98:10037-10041. 

23. Humphrey, W, Dalke, A, & Schulten, K (1996) VMD 

- Visual Molecular Dynamics Journal of Molecular 

Graphics 14:33-38. 


Phosphorylation of Crk Adaptor Protein by Cdc42-Activated Pak2 

and Identification of Phosphorylation Sites 

Jisun Lee 1 , Jin-Hun Jung 2 , Jolinda A. Traugh 2 

1 

Department of Biology, 2 Department of Biochemistry 


ABSTRACT 

The p21-activated protein kinase Pak2 is activated in response to a variety of stresses. Pak2 is activated 

by binding of Cdc42(GTP) followed by autophosphorylation or by caspase3 cleavage. Pak2 modifies 

various targeted substrates. In this study Crk, an adaptor protein, was phosphorylated by Pak2 using 

(γ- 32 P)ATP. Phosphorylation of Crk was analyzed by SDS-polyacrylamide gel electrophoresis and 

phosphorimaging. Results from scintillation counting showed 0.6 mol/mol of 32P incorporated into 

Crk. Phosphoaminoacid analysis showed that Crk was phosphorylated on serine, not threonine. 

To identify the sites of phosphorylation, phosphorylated Crk was digested by the proteases Glu-C 

or trypsin and subjected to two-dimensional phosphopeptide mapping. The maps yielded two 

phosphopeptides, which suggested two phosphopeptides contained the same phosphorylation site or 

two distinctive phosphorylation sites on Crk. Potential phosphorylation sites in two sequences of Crk 

are located near the SH2 and SH3 domains. 

Faculty Mentors 

Jolinda Traugh 

Department of Biochemistry 

Jisun Lee worked as a dishwasher in my laboratory for several years, and also prepared electrophoresis 

buffer and solutions for staining and destaining proteins. She became interested in our research on the 

protein kinase Pak2, and received a campus Nova Award last summer to enable her to work in the 

laboratory full time. She has mastered a number of techniques to analyze phosphorylated proteins and 

identify the sites phosphorylated by Pak2. Jisun is a delight to work with. Her interest and motivation 

in writing this paper attests to her rapid growth and outstanding potential as a scientist. 

A U T H O R 

Jisun Lee 

Biology 

JinSun Lee is a fourth year student 

majoring in Biology. Her research interest 

involves the study of phosphorylation 

of proteins such as Crk, and she is 

currently working on the identification of 

phosphorylation sites of Crk. She hopes 

that research in this field may contribute 

to study of progression of tumor cells 

and its therapeutic solution in the future. 

Jisun would like to thank her family and 

two faculty mentors for their support 

and generous advice throughout her 

academic years. She plans to continue 

her studies in graduate school and work 

for the public interest in the future. 

Jin-Hun Jung 

Department of Biochemistry 

We have been studying a protein kinase, Pak2, which is activated under a variety of 

stress conditions, such as DNAdamaging and hyperosmolarity. Jisun Lee has been 

working on phosphorylation of a proto-oncoprotein Crk by Pak2. She has made 

a significant contribution to identification of the site of phosphorylation on Crk, 

which can lead us to study afunctional relationship of the two proteins in cells. She has successfully 

developed the research skills and related knowledge on her project. It has been a great pleasure to 

work with her. 


Phosphorylation of Crk Adaptor Protein by Cdc42-Activated Pak2 and Identification of Phosphorylation Sites 

Jisun Lee 

Introduction 

The protein Crk, chicken tumor virus no. 10 regulator 

of kinase, is a key adapter protein that functions in several 

signal transduction pathways. Crk has an important role as 

a linkage between tyrosine kinases and small G proteins, 

and leads to the regulation of cell growth, motility, 

apoptosis, and transcription (1). Also, as an oncoprotein, 

Crk is responsible for malignant features of cancers (1). 

Crk is composed of two isoforms, CrkI and CrkII. The 

activity of CrkI has been studied more than of CrkII. CrkII 

has three domains, Src homology 2 (SH2), N-terminal 3 

(SH3n), and C-terminal 3 (SH3c) domains (2). CrkII has 

a phosphorylation site on Tyr221 between N-terminal 

and C-terminal domains of SH3 (3) as indicated in Fig. 1. 

This phosphorylated tyrosine provides an intramolecular 

binding interaction with the SH2 domain of CrkII (3, 4). 

In this study, CrkII is phosphorylated by Pak2 and possible 

phosphorylation sites are identified. 

Pak2, p21-activated kinase 2, is activated in 

response to various cell stresses, such as DNA damaging 

agents or ionizing radiation (5). Pak2 is activated either 

by binding of the small G protein Cdc42 or by cleavage 

with caspase 3, followed by autophosphorylation (5, 6). 

There are 7 serine and 1 threonine sites that are identified 

as autophosphorylation sites for Pak2 (6) as shown in Fig. 

2. The sequence on substrates that allows recognition and 

phosphorylation by Pak2 is represented as (K/R)RXS (7). 

The basic amino acids lysine or arginine at the -3 position 

and arginine at -2 position and any type of amino acid at -1 

position on the substrate, allow phosphorylation by Pak2 (7). 

Consequently, the features of this sequence can be applied to 

identify possible phosphorylation sites on Crk for Pak2. 

In this research, the phosphorylation of CrkII by 

Cdc42-activated Pak2 was studied to examine whether the 

Crk is a good substrate for Pak2 and analyzed the level 

of Crk phosphorylation by Pak2. The characteristics of the 

determinants for phosphorylation by Pak2 were applied and 

analyzed by phosphopeptide mapping and possible sites 

were identified. By studying phosphorylation of Crk with 

Pak2, basic links between Pak2 and Crk can be achieved. 

Furthermore, the regulation of Crk’s critical functions in 

regulation of cell growth and apoptosis by Pak2 can be 

studied in further research for therapeutic treatment of 

human cancers. 

Results 

GST-Crk was phosphorylated by GST-Pak2 in Vitro 

Pak2, Crk, and Cdc42 were identified based upon their 

molecular weights as shown in Coomassie Blue staining in 

Fig. 3 (top panel). To observe the phosphorylation of Crk by 

Cdc42-actived Pak2, phosphorimaging was used. During the 

time course, there was a significant increase in phosphorylation 

of Crk by Cdc42-activated Pak2 in Fig. 3 (bottom panel). 

Autophosphorylation of Pak2 that was activated by Cdc42 

was shown in the phorphorimaging as well. 

Figure 1. Schematic structure of Crk 

Figure 2. Phosphorylation sites of Pak2. Seven phosphorylation 

sites of serine and one phosphorylation site of threonine and a 

caspase cleavage site is indicates within the structure of Pak2. 

Figure 3. Phosphorylation of Crk by Pak2. Top panel: Crk (10 

ug) was incubated with active Pak2 (1 ug) over time, analyzed 

by SDS-PAGE, and stained with Comassie blue. Bottom panel: 

radiolabeled Crk was detected by phosphorimaging. 



Jisun Lee 

32 

P incorporation into Crk 

To examine the effects of the phosphorylation of 

Crk, the protein bands of Crk were subjected to scintillation 

counting. Each phosphorylated Crk band was excised from 

the polyacrylamide gel and quantified for 32 P incorporation. 

The amount of 32 P was adjusted to the actual amount of Crk 

protein due to the fact that there was a slight difference 

between the amounts of proteins in the gel. Fig. 4 shows 

Figure 4. Gamma- 32 P incorporation into Crk (pmol/pmol). The 

molar ratio of 32 P and Crk was obtained from the data in Fig. 3 

and plotted over time. 

that phosphate incorporation was nearly linear over time. 

The maximum phosphorylation obtained at 80 min reached 

a level of 0.6 pmol of 32 P incorporated per pmol of Crk. A 

similar amount of 32 P was observed at 100 min. 

2-D Mapping of Crk for analysis of the phosphorylaiton site 

To study the sites of phosphorylation of Crk, 

Crk was analyzed by phosphopeptide mapping. Two 

proteinases, Glu-C and trypsin were used in the evaluation. 

The protein bands in the gel were cleaved by trypsin or 

Glu-C to generate phosphopeptides. Glu-C cleaved Crk 

after glutamic acid, and trypsin cleaved after arginine and 

lysine. In the first dimension electrophoresis the peptides 

migrated by size, and in second dimension chromatography 

the peptides migrated by charge and size. Fig. 5 displayed 

the products of the phosphopeptide mapping. In these two 

dimensional maps, two main spots with similar sizes were 

detected at similar places in both maps, which suggested 

similar sizes of sequences with similar charges. Closely 

positioned phosphopeptides with similar charges and sizes, 

could be interpreted as one phosphopeptide that contains 

two phosphorylation sites or two different phosphopeptides 

that have their own phosphorylation site. 

Figure 5. Two-dimensional phosphopeptide mapping of Crk phosphorylated by Pak2. Phosphorimages of phosphopeptide maps of Crk. 

Right: peptide digested with trypsin. Left: peptide digested with Glu-C. The origins are indicated with an arrow. 



Jisun Lee 

Figure 6. Sequence of the Crk. SH2 and SH3 domains, and candidate phosphorylation sites are indicated. 

Discussion 

CrkII was phosphorylated by Pak2 that was activated 

by the binding of Cdc42 followed by autophosphorylation. The 

phosphorylation and scintillation counting of 32 P showed Crk as 

an efficient substrate for Cdc42-activated Pak2, with 0.6 pmol/ 

pmol of 32 P incorporation into Crk after 80 min of reaction 

time. To determine the sites of phosphorylation on Crk, the 

peptides corresponding with the spots on the 2-D maps were 

compared regarding their positions in the sequence of Crk. In 

Fig. 6, the sequences underlined with black lines indicate Glu-C 

digested peptides and the sequences underlined with green lines 

indicate trypsin digested peptides. The peptides in the boxes are 

candidate phosphorylation sites determined by the (K/R)RXS 

sequence containing the determinants for potential substrates 

phosphorylated by Pak2. Due to the fact that the spots in each 

of the phosphopeptide maps were close to each other, exact 

numbers of the sites of phosphorylation cannot be identified. 

Glu-C cleaved at glutamic acid, and trypsin cleaved at lysine 

and arginine. According to similar sizes of phosphopeptides 

cleaved in both 2-D maps, and the (K/R)RXS phosphorylation 

sequence, the underlined sequences are only sequences that 

contain similar sizes and charges for each phosphopeptide. Thus 

identification of phosphorylation sites via mass spectrometry 

would be essential for further study on Crk. 

Materials and Methods 

Expression and Purification of Proteins 

Crk and Cdc42 were transformed in Escherichia 

coli individually and Pak2 was expressed in TN5B-4 insect 

cells. The proteins were purified by glutathione affinity 

with glutathione-Sepharose 4B beads. GST was the tag 

used to purify these proteins. GST-Pak2, GST-Cdc42, and 

GST-Crk were removed from the glutathione beads in 10 

mM reduced glutathione in 50 mM Tris-HCl (pH 8.0). Each 

protein and three different concentrations of bovine serum 

albumin (BSA) were subjected to SDS-Polyacrylamide 

gel electrophoresis (PAGE). The gel was stained and the 

densities of stained protein bands were quantified via 

ImageJ for their concentrations. 

Phosphorylation assay 

Phosphoryation of Crk with Pak2 was carried out in a 

total volume of 26 ul and incubated at 30°C for 1 min to 100 

min. Crk (1 µg) was phosphorylated by Pak2 (0.1 µg) that 

was activated by Cdc42 (1 µg) and autophosphorylated in a 

mixture of 50 mM Tris-HCl (pH 7.4), 50 mM NaCl, and 0.18 

mM GTP_S. The reaction was radiolabeled with (γ- 32 P)ATP 

(500 cpm/pmol) in a volume of 26 ul containing 20 mM Tris- 

HCl (pH 7.4), 10 mM MgCl 2 

, 30 mM 2-mercaptoethanol, 

and 0.2 mM ATP. Each reaction was terminated by adding 



Jisun Lee 

SDS-sample buffer. The reactions were subjected to SDS- 

Polyacrylamide gel electrophoresis (PAGE) on a 10% 

polyacrylamide gel. The gel was stained with Coomassie 

Blue, de-strained, and dried followed phosphorimaging. 

The dried gel bands were then quantified and counted for 

incorporation of 32 P into Crk by scintillation counting. 

Phosphopeptide Mapping 

Two phosphorylated Crk protein bands that exhibited 

most high levels of phosphorylation were selected for 

phosphopeptide mappings. The gel band incubated for 80 min 

was digested with trypsin, and the other gel band incubated 

for 60 min was digested with Glu-C. The amount of the 

endoproteinases used a ratio of 1:20 of protease:protein. The 

trypsin-treated gel was incubated at 35°C and Glu-C-treated 

gel was incubated at 26°C overnight. Following lyophilization 

of phosphopepides, they were subjected to two-dimensional 

phosphopeptide mapping. The phosphopeptide mapping was 

composed of two stages. The first dimension of the mapping 

was electrophoresis in buffer containing butanol and glacial 

acetic acid (pH 3.1) for 2 hours at a voltage of 600, and the 

second dimension of the mapping was chromatography in 

buffer containing butanol, pyridine, and glacial acetic acid for 

6 hours. The 2-D maps were visualized by phosphorimaging. 

Acknowledgments 

I would like to thank Dr. Jolinda A. Traugh for 

giving me such a great opportunity throughout the year. In 

addition, I really appreciate Jin-Hun Jung, Linda Xu, and 

Poly Tuazon for their generous advice and support. 

References 

1. 

2. 

3. 

4. 

5. 

6. 

7. 

Kobashigawa, Y., Sakai, M., Naito, M., Yokochi, M., 

Kumeta, H., Makino, Y., Ogura, K., Tanaka, S., and 

Inagaki, F. (2007) Structural basis for the transforming 

activity of human cancer-related signaling adaptor 

protein CRK, Nature Structural & Molecular Biology, 

14, 503-510. 

Feller, M., Stephan. (2001) Crk family adaptorssignalling 

complex formation and biological 

roles, Nature Structural & Molecular Biology, 20, 

6348-6371. 

Rosen, M. K., Yamazaki, T., Gish, G. D., Kay, C. M., 

Pawson, T., and Kay, L. E. (1995) Direct demonstration 

of an intramolecular SH2-phosphotyrosine interaction 

in the Crk protein, Nature 374, 477-479. 

Feller, S. M., Knudsen, B., and Hanafusa, H. (1994) 

C-Abl kinase regulates the protein binding activity of 

c-Crk, EMBO J. 13, 2341-2351. 

Roig, J., and Traugh, J. A. (1999) p21-activated protein 

kinase gamma-PAK is activated by ionizing radiation 

and other DNA-damaging agents: Similarities 

and differences to alpha-PAK, J. Biol. Chem. 274, 

31119-31122. 

Gatti, A., Huang, Z., Tuazon, P. T., and Traugh, J. A. 

(1999) Multisite autophosphorylation of p21-activated 

protein kinase gamma-PAK as a function of activation, 

J. Biol. Chem. 274, 8022-8028. 

Tuazon, T. P., Spanos, W. C., Gump, E. L., Monning, C. 

A., and Traugh, J. A. (1997) Determinants for Substrate 

Phosphorylation by p21-Activated Protein Kinase 

(gamma-PAK), Biochemistry, 36, 16059-16064. 




Kyle McStay 1, 2 , Michele R. Salzman 1 

1 

Department of History, 2 Department of Classical Studies 


ABSTRACT 

According to the view of certain historians, military affairs in the Roman Empire were 

conducted according to an intentional and centrally controlled “Grand Strategy.” Whether or 

not such an empire-wide strategy existed is highly debatable. As a result of this uncertainty, 

the question of provincial strategy arises; was there a consistent strategy on the provincial 

level? This paper addresses that question by means of a case study of the province of Germania 

during the reign of Augustus, the first emperor of Rome. I argue that there was indeed no such 

coherent “German Policy” during the reign of Augustus, but rather a series of policies, each 

with different objectives based on changing conditions. 


Michele R. Salzman 

Department of History 

Kyle’s fascination with military policy and foreign relations under Augustus 

emerged in his senior history seminar in the fall of 2007. His decision to focus 

on the defeat of Varus under the emperor Augustus meant that he was addressing 

one of the most vexed moments in this emperor’s rule. How could the Romans be 

defeated by the Germans whom they had just conquered? Kyle’s research into the ancient sources 

and archaeological evidence enabled him to construct a largely negative argument against those who 

would see a “Grand Strategy” for the Roman empire under Augustus. Kyle’s ability to argue against 

this line of analysis led him to read closely the ancient sources and made for several searching 

discussions about the nature of historical narratives and the ways of approaching evidence. It 

became increasingly evident that Kyle had not only a passion, but real abilities in pursuing problems 

in ancient history. I am pleased that he will go on to a graduate program in ancient history at UCR 

where he can pursue these and other issues. 

A U T H O R 

Kyle McStay 

History and Classical Studies 

Kyle McStay is a graduating senior with 

a double major in History and Classical 

Studies. His main areas of interest 

are the later Roman Empire and, more 

generally, the Roman military system. 

In particular, Kyle is interested in the 

Christianization of the empire and 

its effects on political, military, and 

social frameworks; Roman military history, 

especially the way in which major 

military disasters affected the political 

history of the empire; and the history of 

the Eastern Roman Empire after the fall 

of the West, an area he believes to be 

frequently underplayed or ignored outright 

in many history programs. In fall 

of 2008, Kyle will continue his studies 

in graduate school here at UCR. 



Kyle McStay 

During the last three decades it has been forcefully 

argued, most notably by Edward Luttwak 1 , that Roman 

military affairs during the imperial period were conducted 

according to a “Grand Strategy.” This system envisages 

an intentionally planned and unified set of strategic 

principles controlled by the emperor, by which all tactical 

operations were governed. Whether or not such a system 

actually existed, or indeed, whether such a system could 

have functioned at all due to the vast size of the Empire, 

the dubious state of available geographic knowledge, and 

the relatively slow speed at which information and troops 

traveled in antiquity, remains highly suspect. 2 While the 

existence of an empire-wide strategic policy continues to 

be hotly debated, even the existence of a unified strategic 

policy on the provincial level is a question that is still 

controversial and largely unexplored. 

What follows is an attempt to address that question 

by means of a case study of the province of Germania 

between 34 B.C.E. and 16 C.E. During that period, no 

less than seventeen campaigns were conducted in the 

Rhineland 3 by the legates of Augustus, the first emperor of 

Rome, or members of his family against numerous Gallic 

and Germanic tribes. To characterize these campaigns 

as being the result of a preconceived policy designed 

specifically to expand imperial territory by incorporating 

Germania into the Empire is inaccurate. Such a view fails 

to take into account the relationship between Gaul 4 and 

Germania. Also, such a view demonstrates a fundamental 

misconception of the nature of the Rhine River itself in 

Roman strategic thought. Without an understanding of 

these complex relationships, it is impossible to understand 

Augustus’ “German Policy,” which was in fact not a 

single policy at all, but was rather a set of several distinct 

yet interrelated strategies. Each of these strategies was 

reactionary in nature; all were based on particular sets of 

conditions and were designed to produce distinct results. 

Augustus’ first two strategies in the Rhineland were 

motivated by the need to ensure internal stability in Gaul. 

Germanic tribes from both sides of the Rhine frequently 

invaded Gaul; however, the Germanic tribes often came 

not to raid, but to assist Gallic tribes revolting from Roman 

control. As such, Gallic security was linked to controlling 

these Germanic tribes. This Germanic assistance in Gallic 

uprisings was the primary motivating factor for many of the 

campaigns across the Rhine during this period. Between 31 

and 28 B.C.E. three Gallic tribes, the Morini, Treviri, and 

the Aquitani revolted, each with assistance from Germanic 

tribes from across the Rhine. 5 This forced Marcus Agrippa, 

who had been commissioned to put down the revolts, to 

campaign on both sides of the river. 6 Agrippa was again 

posted to Gaul in 20-19 B.C.E., as according to Cassius 

Dio, one of our best sources for these events, “…the 

inhabitants were not only quarrelling among themselves, 

but were being harassed by the Germans.” 7 This was a 

familiar situation; Augustus placed great importance on 

the region, as is evidenced by the fact that he assigned 

Agrippa, his most able and trusted lieutenant, to subdue 

both Gallic uprisings. 8 

Up to this date, Augustus’ policy of retaliating against 

specific tribes that had committed specific offenses had not 

proved to be a deterrent against invading Gallic territory. In 

17-16 B.C.E. three Germanic tribes, the Sugambri, Usipetes, 

and the Tencteri crossed the Rhine and inflicted a defeat 

upon Marcus Lollius, which was severe enough, according 

to the Roman historian Velleius Paterculus, to cause 

Legio V Alaudae to lose its eagle standard, though it was 

subsequently returned. 9 After this defeat, Augustus himself 

hurried to the Rhine to take command, but by the time he 

arrived the Germans had retreated back across the river, and 

had decided to make peace. 10 Augustus’ personal intervention 

shows how serious he considered this event to be. 

The Emperor did not return to Rome until 13 B.C.E. 

The defeat of Lollius clearly advertised the failure of his 

policy of limited retaliation, and his reaction to this failure was 

a significant alteration of his previous strategy. His new policy 

would be aggressive, and would consist of yearly, preemptive 

invasions across the Rhine. Just as the previous strategy was 

aimed at securing Gallic stability by punishing those Germanic 

tribes that had either participated in Gallic rebellions or had 

invaded and plundered Roman controlled territory, the new 

strategy was aimed at ensuring Gallic stability by intimidating 

and subduing the Germanic tribes across the Rhine, thereby 

eliminating them as possible threats to Roman control of 

Gaul; though the goal was the same, the methods employed to 

achieve it were radically different. 

Augustus used the three years he spent in Gaul 

preparing the Rhineland for the launch of his new aggressive 

strategy. It is very likely that it was during this period that 

the legionary bases along the Rhine were established, 11 as 

the large scale offensive operations Augustus was planning 



Kyle McStay 

would have needed large supply bases, and the five or six 

legions stationed along the Rhine by 13 B.C.E. would have 

needed extensive bases to winter in and to operate from. 

Three of these bases, Vechten (Fectio), Xanten (Vetera), 

and Mainz (Mogontiacum), stood at the heads of the 

three main invasion routes into Germania, and all of the 

Rhineland bases were built on the west bank of the Rhine. 

Bases were also established along the Lippe River, 

the invasion route opposite Xanten mentioned above, but 

these were most likely established by Augustus’ stepson 

Drusus between 12 and 9 B.C.E. or later. 12 The most 

important site along the Lippe is the legionary fortress at 

Haltern (Aliso), which may have included a naval station 

and large stores complex. 13 It is important to note that all 

of these fortresses, both those on the Rhine and those on 

the Lippe, were wood and turf structures, which is a strong 

indication that they were not yet intended to be permanent 

installations. 14 The construction of these fortresses does not 

imply a shift towards a defensive strategy, and Augustus 

had no plans of establishing a border along the Rhine, 

Elbe, or anywhere else; he was still employing offensive 

strategies with the intent to secure Gaul from invasion. 

According to Erich Gruen, this is because: 

[t]he Rhine was an artificial and largely ineffectual 

barrier…It represented at best a frontier zone 

rather than a demarcated border. And harassment 

of Roman Gaul by trans-Rhenane intruders was a 

continual menace. 15 

The years between 16 and 13 B.C.E. clearly 

marked a departure from previous policy. Major military 

installations were established to support regular, largescale 

offensive operations into an area that had previously 

been invaded only sporadically. Five or six legions, along 

with comparable numbers of auxiliary infantry and cavalry 

had been transferred from the interior of Gaul, Spain, and 

other locations to the Rhineland fortresses in order to carry 

out those invasions. 16 Augustus was clearly no longer 

considering local retaliatory incursions, but this does not 

imply an intention to annex Germania; indeed, the sources 

make no mention of any such intention at that time. 

The commander he chose to carry out his new policy 

was his stepson Drusus. The result was five consecutive 

yearly campaigns across the Rhine, the first of which 

was launched in 12 B.C.E. According to Dio, the Gauls 

were again “discontented at their subjugation,” 17 and the 

Germanic Sugambri crossed the Rhine to aid them. Drusus 

quickly quelled the Gauls and then drove the Sugambri 

back across the river. His next moves illustrate clearly 

that a new policy was in place; he crossed the river and 

proceeded to attack tribes all over northern Germania, none 

of which, other than the Sugambri, are mentioned as having 

participated in the Gallic uprising. He passed through the 

territory of the Usipetes, though Dio makes no mention 

of any fighting against them. Drusus then laid waste the 

territory of the Sugambri, following which his army sailed 

up the Rhine to the ocean and gained the alliance of the 

Frisians along the coast. The army then went inland and 

attacked the Bructeri along the Ems River and the Chauci 

between the Ems and Weser Rivers. At the end of this 

campaign, Drusus and his army returned to their bases 

along the Rhine for the winter. 18 That Drusus returned to 

the Rhine following each campaign strongly indicates that 

annexation was not then under consideration. 

No mention is made of any overt hostile actions that 

could have provoked the four campaigns that followed. This 

is further evidence that Augustus was pursuing a new and 

more aggressive preemptive strategy. In 11 B.C.E., Drusus 

crossed the Rhine and subdued the Usipetes, marching all 

the way to the Weser River, after which he was attacked by 

the Cherusci and the Sugambri. Though the return to the 

Rhine was difficult and fraught with hardships, Drusus did 

manage to build and garrison the fortress at Haltern, before 

he reached his main bases. 19 

Augustus accompanied Drusus and his other 

stepson, the future emperor Tiberius, to Gaul during the 

winter of 11-10 B.C.E. The occasion was the inspection of 

the Altar of Roma and Augustus which had been set up at 

Lyons (Lugdunum), signifying the allegiance of the Gallic 

tribes to both Augustus and the Roman state. Dio states 

that the purpose of this visit was also to keep watch on the 

Germanic tribes more closely; once again, the pacification 

of Germania and the internal stability of Gaul are intimately 

linked. 20 At the beginning of spring, 10 B.C.E., Drusus set 

out across the Rhine and attacked the Sugambri and the 

Chatti tribes, which had formed an alliance. Once again, 

Drusus returned to the Rhine at the onset of winter. 21 

The most far-reaching successes came during the 

fourth campaign. Beginning in the spring of 9 B.C.E., Drusus 



Kyle McStay 

set out from Mainz and again attacked the Chatti. After 

fierce fighting along the upper Main River, he defeated the 

Marcomanni, who afterwards migrated eastward. The army 

then turned north, crossed the Weser, and reached the Elbe 

River. Drusus became the first Roman commander to achieve 

this. For unknown reasons, Drusus did not cross the Elbe, but 

turned back toward the Rhine. At some point he suffered a 

broken leg, and died before his army reached the river. 22 

Tiberius took over command after the death of 

Drusus, and launched a new campaign in 8 B.C.E. during 

which, according to Velleius Paterculus, he traversed every 

part of Germania with no loss to his own army. 23 But 

what had Drusus and Tiberius actually accomplished? It 

seems clear that these five campaigns had been invasions, 

not conquests. The same tribes were attacked year after 

year, indicating that those tribes remained unconquered 

at the end of each campaign. For the time being however, 

the Germanic tribes east of the Rhine were unwilling to 

continue to fight the Romans, so it is probable that Drusus 

and Tiberius had at least succeeded in weakening the tribes 

and intimidating them into accepting peace; Augustus’ 

strategy appeared to have worked. 

By this time, Augustus’ thoughts seem to have been 

turning away from simply securing Roman control of Gaul 

and towards incorporating Germania into the Empire. The 

willingness of the Germanic tribes to make peace and 

to remain peaceful, at least for the time being, appear to 

have convinced Augustus of the safety of Gaul and that 

Germania could be made into a province as well. Evidence 

to support the claim that Augustus was now turning towards 

the peaceful incorporation of Germania into the Empire is 

provided by the erection of an Altar of Roma and Augustus 

at Cologne (Oppidum Ubiorum) around 8 B.C.E., similar 

to the altar that had been erected at Lyons during the winter 

of 11-10 B.C.E. The altar signified allegiance to Augustus, 

and as Tacitus tells us, the priest of the cult was a member 

of the Cherusci tribe, 24 another indication of the believed 

loyalty of the Germanic tribes. 

Further offensive operations were deemed 

unnecessary, and Augustus was content with the planting 

of garrisons and the construction of additional fortifications 

along the Rhine and also along the Lippe. The cessation 

of preemptive invasions is further evidence that Augustus 

was now concerned with making Germania into a proper 

province, not with securing Roman control of Gaul. 

Germania seemed to be pacified, and indeed, Dio states 

that the area was slowly Romanizing with the presence of 

Roman garrisons, the growth of cities, the establishment of 

markets, and the introduction of peaceful assemblies. 25 The 

culmination of Augustus’ new policy of integration came in 

6 C.E., with the appointment of Publius Quinctilius Varus 

to the governorship of Germania. 

Varus’ career before being appointed governor of 

Germania was largely one of administration; 26 that he had 

not had extensive military experience is itself a strong 

indication that Augustus believed Germania to be ready 

to be fully integrated into the Empire. No other reason 

presents itself which can adequately explain why he would 

have appointed a man such as Varus to a governorship that 

had, until 6 C.E., been held exclusively by viri militares 

(military men). 27 It would appear that Varus had been 

appointed to do what he was accustomed to, which was to 

introduce peacetime administration. 28 

Correspondingly, Varus set out to speed up the 

process of Romanization. According to Dio and Velleius 

Paterculus, Varus began levying taxes and exercising judicial 

powers to which the Germanic tribes were not accustomed, 

nor were they disposed to accept this new imposition of 

authority. 29 The Germanic tribes began to lull Varus into a 

false sense of security by appearing to submit to his judicial 

authority, so that, as Velleius Paterculus states, “he came to 

look upon himself as a city praetor administering justice 

in the forum, and not a general in command of an army 

in the heart of Germany.” 30 In September of 9 C.E., word 

was brought to Varus that a revolt had broken out far from 

the Rhine. This was done to lure Varus deep into enemy 

territory, while simultaneously allowing him to believe 

that he was traveling through friendly territory, so that he 

might become lax on the march, which is apparently what 

occurred. 31 Accordingly, Varus set out with all three legions 

of his army, six cohorts of auxiliary infantry, and three alae 

of auxiliary cavalry, a force totaling about 21,000 infantry 

and 1,500 cavalry. 32 

This force was taken by surprise somewhere in the 

Teutoburg Forest, terrain that made it almost impossible 

for the Romans to deploy, thus negating all of their tactical 

strengths. According to Dio and Velleius Paterculus, the 

battle was a four day nightmare of ambushes and vicious 

close quarters fighting during which the Romans constantly 

tried to regroup and escape back to their bases along the 



Kyle McStay 

Lippe and the Rhine. 33 However, by the fourth day the 

ranks of the rebels had swelled to insurmountable numbers, 

and the Roman column was hemmed in on all sides; Varus 

and his officers committed suicide. Over the four days of 

the battle, Varus’ force was annihilated almost to a man. 34 

In the aftermath of this defeat, one of the worst 

in Roman history, all of the fortresses on the Lippe were 

captured almost immediately except for Haltern, which 

endured a horrendous siege until the garrison was able to 

escape back to the Rhine. 35 According to Dio and Velleius 

Paterculus, it was only the quick action of Lucius Asprenas 

who moved his forces into the fortresses on the lower 

Rhine that prevented the rebellion from spreading across 

the river. 36 This defeat completely destroyed Augustus’ 

hopes of making Germania into a province and effectively 

marked the end of his third policy. The defeat of Varus was 

so complete that it took Tiberius, who rushed to the Rhine 

as soon as news of the disaster reached Rome, almost three 

years to stabilize the situation. 

Augustus reacted by initiating his fourth policy, 

which was one of retribution. In 13 C.E. Drusus’ son 

Germanicus was appointed to supreme command of 

the Rhineland armies. He was to vigorously engage in 

offensive operations against the Germanic tribes, but as 

Tacitus tells us, the motivation behind these campaigns 

was to wipe the stain of Varus’ defeat from Rome’s military 

reputation, not to extend the Empire. 37 Further evidence 

that the goal of these campaigns was not re-conquest is that 

Germanicus made no effort what-so-ever to establish new 

fortresses anywhere between the Rhine and Elbe, nor did 

he attempt to reoccupy any of the previously established 

posts, except possibly at Haltern. 38 It seems clear that all 

hope of conquering Germania had been abandoned, and reestablishing 

Rome’s reputation for military dominance was 

now the main concern. 

In 16 C.E., after two years of hard fighting, Tiberius, 

by then having acceded to the Principate upon the death of 

Augustus in 14 C.E., recalled Germanicus from Germania. 

With Rome’s reputation apparently restored, Tiberius 

initiated the fifth and final policy, which was one of defense 

and consolidation. Offensive operations across the Rhine 

were halted, but all eight legions in the Rhineland remained 

stationed along the river. It seems obvious that by this point, 

Tiberius, “who knew Germany if any Roman did … saw the 

impossibility of recovering the territory lost after the Varian 

disaster.” 39 Tiberius must have been aware that the Germanic 

tribes from beyond the Elbe were migrating westward, many 

along the same invasion routes the Romans had used to go 

east. Pacifying Germania now meant, unlike during the time 

of Drusus, not only pacifying the current population but also 

denying the migrating tribes access to the lands west of the 

Elbe. According to Collin Wells, 

After a youth spent augmenting the Empire and 

a middle-age in defending it, he [Tiberius] set his 

face against further expansion. His concern was for 

consolidating what Rome already possessed. 40 

Though Dio states that Augustus’ will contained 

instructions that Tiberius not expand the Empire further, 41 

it seems likely for the reasons stated above that he would 

not have tried to do so in Germania even without such 

advice from Augustus. 

To conclude, though the policies of Augustus 

changed over time, some consistencies run throughout his 

rule. Swift and public retaliation for any defeat or perceived 

weakness; the personal intervention of Augustus or his 

stepsons in every crisis situation; and the promotion of the 

imperial cult were common factors throughout his reign. 

However, these common factors do not imply a unity of 

purpose among his policies in Germania, all of which were 

essentially reactionary. His decision to implement yearly 

invasions across the Rhine was a reaction to the defeat of 

Lollius in 17 B.C.E.; his decision to attempt incorporation 

of Germania was a reaction to the apparent success of 

his policy of preemptive aggression; the initiation of the 

punitive campaign of Germanicus was a reaction to the 

annihilation of Varus in 9 C.E.; and Tiberius’ decision to 

implement a defensive policy was a reaction to Germanicus’ 

success. These policies were never the result of a unified 

set of strategic goals, and as such, no “Grand Strategy” for 

Germania existed under Augustus. 



Kyle McStay 

References 

Primary Sources 

1. Cassius Dio. The Roman History: 

Books 48-49: Trans. Earnest Cary for Loeb Classical 

Library, 1914-1927. Text on LacusCurtius at 

http://penelope.uchicago.edu/Thayer/E/Roman/Texts/ 

Cassius_Dio/home.html 

Books 50-56: Trans. Ian Scott-Kilvert. London: 

Penguin Books, 1987. 

2. Gaius Velleius Paterculus. The Roman History. Trans. 

Fredrick W. Shipley for Loeb Classical Library, 1924. 

Text on LacusCurtius at http://penelope.uchicago.edu 

/Thayer/E/Roman/Texts/Velleius_Paterculus/home.html 

3. Publius Cornelius Tacitus. Annals. Trans. Alfred John 

Church and William Jackson Brodribb, 1942. E-text 

edition by Bruce. J. Butterfield, 1997. http://mcadams. 

posc.mu.edu/txt/ah/tacitus/ 

Secondary Sources 

1. Gruen, Erich S. “The Expansion of the Empire Under 

Augustus.” The Cambridge Ancient History, 2nd ed. 

Vol. 10. Eds. Alan K. Bowman, Edward Champlin, and 

Andrew Lintott. Cambridge, New York, Melbourne: 

Cambridge University Press, 1996. pp. 147-197. 

2. Ruger, C. “Germany.” The Cambridge Ancient 

History, 2nd ed. Vol. 10. Eds. Alan K. Bowman, 

Edward Champlin, and Andrew Lintott. Cambridge, 

New York, Melbourne: Cambridge University Press, 

1996. pp. 147-197. pp. 517-534. 

3. Wells, C[olin] M[ichael]. The German Policy of 

Augustus. Oxford: Clarendon Press, 1972. 

Endnotes 

1 For detailed discussion of this view see: Luttwak, 

Edward N. The Grand Strategy of the Roman Empire, 

From the First Century A.D. to the Third. Baltimore 

and London: Johns Hopkins University Press, 1976. 

2 For detailed criticism of the views of Luttwak see: 

Mattern, Susan P. Rome and the Enemy: Imperial 

Strategy in the Principate. Berkeley, Los Angeles, 

London: University of California Press, 1999. 

3 Gruen, pp.169-171, 178-188 

4 The area known to the Romans as “Gaul” roughly 

encompassed the territory of modern-day France and 

Belgium. 

5 Dio LI.21.5-6 

6 Dio XLVIII.49.3 

7 Dio LIV.11.2 

8 Dio XLVIII.49.3, LI.21.5-6 

9 Velleius Paterculus II.97.1 

10 Dio LIV.20.4 

11 Wells, pp. 93-148 

12 Wells, pp. 149-154 

13 Haltern is likely, but not certainly, the fortress 

referred to as “Aliso” by Cassius Dio, Velleius 

Paterculus, and Tacitus (see Wells, pp. 152-153, 

pp.192-198, for discussion of available evidence). 

14 Wells, pp. 99-100 

15 Gruen, p. 179 

16 Wells, pp. 94-95 

17 Dio LIV.32.1 

18 Dio LIV.32.1-3 

19 Dio LIV.33.1-5; Gruen p. 181 

20 Gruen, p. 181; Dio LIV.36.4 

21 Dio LIV. 36.4 

22 Dio LV.1.2-5; Velleius Paterculus II.97.2-3 


24 Tacitus, Annals, 1.57.2 

25 Dio LVI.18.2 

26 Wells, pp. 238-239 

27 Wells, p. 239 

28 Wells, p. 239 

29 Dio LVI.18.3; Velleius Paterculus II.118.1 


31 Dio LVI.18.5 

32 Ruger, p. 527 

33 Dio LVI.20.1-3; Velleius Paterculus II.119 

34 Dio LVI.19.4-20.5; Velleius Paterculus II.119 


36 Velleius Paterculus II.120.3; Dio LVI.22.2 

37 Tacitus, Annals 1.3.6 

38 Wells, pp. 198-206 for detailed discussion 

of available evidence. 

39 Wells, p. 244-245 

40 Wells, p. 243-244 

41 Dio LVI.33.5 


Fractal Strings and Number Theory: 

The Harmonic String and the Prime String 

Jason C. Payne, Michel L. Lapidus 

Department of Mathematics 


Abstract 

The purpose of this paper follows two veins which converge at one critical juncture- a bridge 

between two seemingly disparate concepts. The first goal is to provide a brief survey of the topics 

of fractal strings and their complex dimensions, which is achieved through the introduction of 

their geometric and spectral zeta functions. Following this is consideration of two important 

examples of generalized fractal strings- the harmonic string and the prime string. Through this 

two discussions, the goal is to establish a strong connection between the concrete study of the 

geometry and spectrum of fractal strings and the abstract world of number theory, which is 

achieved by way of the infamous Riemann zeta function. 


Michel L. Lapidus 

Department of Mathematics 

Michel Lapidus is Professor of Mathematics at UCR and also teaches in the 

departments of Physics, Electrical Engineering and Computer Science. He works 

at the crossroad of many research areas, including Mathematical Physics, Fractal 

Geometry, Dynamical Systems, Parital Differential Equations, Noncommutative 

Geometry, and Number Theory. His recent research books include “The Feynman Integral and 

Feynman’s Operational Calculus” (Oxford Univ. Press, 2000, paperback: 2001; joint with G. W. 

Jonson), “Fractal Geometry and Number Theory” (Birkhauser, 2000), “Fractal Geometry, Complex 

Dimensions and Zeta Functions: Geometry and spectra of fractal strings” (Springer-Verlag, 2006) 

[both joint with M. van Frankenhuysen], and most recently, “In Search of the Riemann Zeros: Strings, 

fractal membranes and noncommutative spacetimes” (Amer. Math. Soc., Jan. 2008). Professor 

Lapidus is a Fellow of the American Association for the Advancement of Science, and has been a 

member of the American Mathematical Society Council since January 2002. Over the last nine years, 

he has worked with nine undergraduate research projects and undergraduate Honors Theses. 

A U T H O R 

Jason C. Payne 

Pure Mathematics 

Jason Payne is a graduating senior majoring 

in Pure Mathematics. His research 

interests include fractal geometry and 

complex analysis and their applications 

in both analytic and algebraic number 

theory, and Riemannian geometry. He is 

currently working on a senior thesis on 

volume formulas in arbitrary (potentially 

fractal) dimensions, and how they can 

be combined with the study of fractal 

strings in order to provide a generalization 

of Gauss’s Circle Problem. Some 

fellow math majors and he are creating 

of an official math club at UCR. After 

graduating this spring, Jason will begin 

the Mathematics Ph.D. program here at 

UCR where he will continue his research 

in fractal geometry, complex analysis, 

and number theory. 




Introduction 

Introduction to Fractal Strings, and their Geometric and Spectral Zeta Functions 










Self-Similar Strings 
















Generalized Fractal Strings 







References 


Motion Based Bird Sensing Using Frame Differencing 

and Gaussian Mixture 

Deep J. Shah 1 , Deborah Estrin 2 

Student Assistant: Afrouz Azari 

Graduate Student Assistants: Teresa Ko, Shaun Ahmadian, Mohammad Rahimi 

1 

Department of Electrical Engineering, University of California, Riverside 

2 

Department of Computer Science, UCLA 

ABSTRACT 

Background segmentation is a general technique that aims at detecting the moving objects in a 

sequence of continuous scenes or video stream by separating the non-moving background from 

the moving foreground object. The success and weakness of a foreground and/or background 

segmentation method depends on several external factors such as the scene selection, lighting, and 

weather conditions. Hence, we perform a comprehensive quantitative and qualitative analysis of 

two such background segmentation methods - frame differencing and Gaussian mixture - for a 

continuously changing environment with articulated foreground objects. We try to detect birds as our 

foreground objects in image streams with varying backgrounds. This study can be used to evaluate 

the effectiveness of a segmentation technique in a constantly changing outdoor environment. We 

find that the Gaussian mixture approach is more accurate than the frame differencing approach; 

however, as a trade-off the Gaussian mixture approach takes far more time and memory to run. The 

segmentation results produced using these techniques are the foundation for any further analysis 

that biologists need to better understand bird motion, bird feeding habits, bird flight and other bird 

behaviors. 

Key Terms: 

1.) Background segmentation – Segment the background from foreground. 

2.) Frame differencing – Finding absolute difference between frames. 

3.) Centroid – The center of a motion detected region. 

4.) Ground truth – The actual data collected on site. 


Deborah Estrin 

Department of Computer Science, UCLA 

The Center for Embedded Networked Sensing (CENS) is a National Science 

Foundation Science and Technology Center. One of the driving applications at 

this center is in the automated sensing of ecosystem health indicators. Deep joined 

our lab in 2007 as a participant in our intensive summer undergraduate internship 

program. His principle task was assisting in the development of an image recognition program 

designed to identify avian activity. Specifically, Deep was able to detect birds utilizing techniques 

in frame differencing and image segmentation. The detection of birds in the natural environment is 

a critical step in aiding ecologist in the study of avian behavior through image sensors. Deep Shah 

was a delight to have in our program; his light-hearted humor combined with his determination to 

produce quality work made for a wonderful addition to our labs. 

A U T H O R 

Deep J. Shah 

Electrical Engineering 

Deep Shah is a graduating senior majoring 

in Electrical Engineering. His interests 

are in the field of signal processing 

and communications in Electrical Engineering. 

He has been published in the 

Journal of Computational Electronics and 

has presented research work at Southern 

California Conference for Undergraduate 

Research and at the CENS undergraduate 

research symposium. He is a Bourns College 

of Engineering (COE) Ambassador, 

Success Counselor, a COE representative 

in the ASUCR Senate, and served as a 

student editor in the inaugural edition of 

UCR Undergraduate Research Journal. He 

will begin his graduate studies in Electrical 

Engineering at UCLA in fall, 2008. 



Deep J. Shah 

Introduction 

In the past, several background segmentation 

techniques have been used to identify the objects of interest 

in a scene. The object of interest can be defined as something 

that is different in a scene in comparison to previous scenes. 

This comparison is performed by comparing a scene with 

an object, to an ideal scene which just has the presence of 

the background. The regions which observe a considerable 

change between the current scene and the previous scenes 

are the areas of interest, as they may indicate the location 

of the new object. 

The term “background segmentation” refers to 

identifying the difference between an image and its 

background using any of the techniques mentioned in [4] 

and then thresholding the results to identify an object of 

interest. There are several papers that describe various 

methods of performing background subtraction. However, 

very few papers describe the algorithms in sufficient detail 

to make the re-implementation easy [4]. 

Hence, this paper thoroughly describes two 

background segmentation techniques - frame differencing 

and Gaussian mixture. We then investigate the usefulness 

and shortcomings of each of these methods in an 

environment that involves a constant moving background 

due to bird flight, leaf movement, lighting changes, and 

other movements in the sky. We try to monitor the birds as 

our objects of interest. 

Methods: Segmentation Techniques 

In the following sections, we describe two approaches 

that we use for bird detection. The first section describes how 

frame differencing is performed and the second section displays 

the framework behind the Gaussian mixture approach. 

Frame Differencing 

This section describes one of the most common 

techniques used in background segmentation. As the name 

itself suggests, frame differencing involves taking the 

difference between two frames and using this difference 

to detect the object. The approach we use is very similar to 

this and is a two part process. First, the object is detected 

using frame differencing. Then this detected object is 

compared with the ground truth to learn the reliability of 

this approach. 

In our implementation we load two consecutive 

frames from a given sequence of video frames. These color 

frames are converted to gray scale intensity. Otsu’s Method 

[1] is then used to determine the threshold value of the gray 

scale images. The threshold value is determined such that 

the pixel values on either side of this value are established 

to be either a background or a foreground pixel. 

Following Otsu’s method, the two consecutive gray 

scaled images are differentiated and their absolute difference 

is used to identify the movement between frames. The noise 

collected due to differencing is removed by applying the 

threshold value to the images. The threshold value that we 

find in our case varies between [0.43 - 0.45]. Pixels below the 

threshold are removed from the differenced frame leaving 

behind our object of interest. As described in equation 1, the 

absolute difference between two frames needs to be greater 

than the threshold for the object to be detected. 

| F2 − F1 

| > T - (1) 

Here F 

1 

is the initial frame, F2 

is the following 

frame and T is the threshold value. 

Gaussian Mixture 

In this section, we describe another technique 

that is commonly used for performing background 

segmentation. In their paper [5], Stauffer and Grimson 

suggest a probabilistic approach using a mixture of 

Gaussians for identifying the background and foreground 

objects. The probability of observing a given pixel value 

p at time t is given by [5]: 

t 

K 

( 

, 

i= 

1 

P p t 

) = ∑ω i, tη( 

pt 

, µ 

i, 

t 

, Σi 

t 

) - (2) 

where K is the number of Gaussian mixtures that are 

used. The number of K varies depending on the memory 

allocated for simulations. The normalized Gaussian η is 

a function ofω i, t 

, µ 

i, t 

, Σ 

i, t 

which represent the weight, 

mean, and the covariance matrix of the i th Gaussian at 

time t respectively. The weight indicates the influence of 

the i th Gaussian at time t. In our case we choose K = 5, to 

maximize the distinction amongst pixel values. 

Since this is an iterative process, all the parameters 

are updated with the inclusion of every new pixel. Before 

the update takes place, the new pixel is compared to see 

if it matches any of the K existing Gaussians. A match is 



Deep J. Shah 

determined if | pt 

- µ 

i,t| 

< 2.5σ , where σ corresponds 

to the standard deviation of the Gaussian. Depending on 

the match, the Gaussian mixture is updated in the following 

manner as mentioned in [4, 5]: 

If the pixel value p 

t 

matches the i th Gaussian, 

then the i th Gaussian component values are updated in the 

following manner: 

σ 

ω 

i , t 

= ( 1−α) 

ωi, 

t−1 

+ α - (3) 

µ 

i, t= ) 

t−1 

t 

( 1− 

ρ µ + p ρ - (4) 

2 

T 

= (1 − ρ) 

σ 

− 

+ ρ( 

p − µ ) ( p − µ ) - (5) 

2 

i, t 

i, 

t 1 t t t t 

where - (6) 

In this case the variable 1 / α defines the speed at 

which the distribution parameters change. 

If the pixel p 

t 

matches the i th Gaussian, then the 

remaining K-1 Gaussians are updated in the following manner: 

ω - (7) 

i, t 

= ( 1−α) 

ωi, 

t−1 

µ - (8) 

i , t= µ 

i, 

t−1 

σ - (9) 

2 2 

i, t 

= σ 

i, 

t−1 

If the pixel pt 

fails to match any of the K 

Gaussians, then the Gaussian with the least likelihood of 

being the background is removed and replaced with a new 

distribution with the following parameters: 

ω = A very low Weight - (10) 

i,t 

µ 

i,t= 

Pixel value p 

t 

- (11) 

σ = A high Variance - (12) 

2 

i,t 

The values for weight and variance can vary based 

on the significance that is given to a pixel which is least 

likely to occur in a particular setting. 

All the Gaussian weights are renormalized after the 

update is performed. The K Gaussians are then re-ordered 

based on their likelihood of existence. This likelihood is 

ωi, 

t 

determined using the ratio [5]. Both ω 

i, t 

being high 

and/or 

σ 

i, 

t 

σ being low leads to a higher ratio. This leads 

i, t 

to an open-ended list of all K Gaussians, where the most 

likely background models with a high ratio are to the left 

and the less probable transient background distributions 

are to the right [5]. 

Then b distributions are modeled to be the 

background and the remaining K - b distributions are 

modeled as the foreground for the next pixel. The value 

for b is determined as described in [5]: 

b 

∑ 

B = arg min ( w > T ) - (13) 

b 

i= 

1 

where T is some threshold value which measures the proportion 

of the data that needs to match the background and then the first 

B distributions are chosen as the background model. 

We use MATLAB to perform all our computations and 

implementation for both the segmentation approaches. The 

video recordings and images used for this study are collected 

near the bird feeding station at the James Reserve, CA. Our 

sampled videos are each a minute and six seconds long in 

length, with each video containing about 1000 frames. There 

are 15 frames taken per second at a resolution of 640×480 

pixels using the Sony NTSC SNC-RZ30N network camera. 

The camera is located about 15 feet away from the feeder 

station. The location of the feeder station with respect to the 

camera is shown in Figure 1: 

Results 

Location of the bird 

feeder s tat i o n 

Figure 1. The bird feeder station region that is used for capturing 

video samples. 

To compare the validity of all the results obtained 

by using different background segmentation techniques 

we evaluate our results with the actual ground truth. 


The data collected can be classified in two 

different ways to understand the detection of a bird. Table 

1 shows the sum of the distances from a centroid to N of 

i 



Deep J. Shah 

its other nearest centroids. A centroid is defined to be the 

center of a motion detected object. The centroids found are 

checked to see if they match the ground truth by detecting 

their presence in the bounding box. A bounding box is a 

rectangular region covering all the edges of the detected 

object inside it. The size of this box is determined by the 

size of the object that it covers. If a centroid matches, then 

it is classified as a centroid inside the bounding box. If it 

does not match then it is classified as a centroid outside the 

bounding box as shown in Table 1. 

N 

Centroids inside 

bounding box 

(Average distance 

in pixels) 

Centroids outside 

bounding box 

(Average distance 

in pixels) 

1 4.0668 8.3747 

2 10.9104 23.4571 

3 20.3761 44.9948 

4 31.8322 74.1308 

5 48.9304 110.5066 

Table 1. Distance from a centroid to N nearest centroids in the 

frame showing presence of a bird. 

For this part, we use dilation on differentiated regions 

and connect them with m (m = 6 in this case) surrounding 

pixels. This is done to connect pixels of the same object 

which might have been left undetected due to the threshold 

value being too high. Table 2 shows the number of times 

a centroid is detected when the bird is present or absent in 

a video sequence. The value for the number of centroids 

detected when bird is absent in Table 2 does not include 

the centroids that might have been detected in frames 

that actually did not have any presence of bird. The final 

column displays total number of centroids that are detected 

in a given video sequence. 

Number of occurrences 

Video 

Sequence 

Number 

Centroid 

detected 

when bird 

present 

(TP) 

Centroid 

detected 

when bird 

absent 

(FP) 

Centroid 

undetected 

when bird 

present 

(TN) 

Total 

Centroids 

detected 

in all 

1 10 33 17 484 

2 23 106 49 545 

3 5 36 13 346 

4 11 39 24 337 

5 13 78 36 436 

Table 2. Number of true and false occurrences for bird detection 

using centroids. 

Frames 51-53 from the video sequence number 

4 are displayed in Figure 2. These images show the best 

potential accuracy of the frame differencing approach in 

trying to detect a moving object. The bird gets detected in 

these frames but the frame also detects the location of the 

bird in the previous frame. This is because when absolute 

differencing is performed between two consecutive images, 

the bird location in the previous frame is different from 

that in the current frame. This leads the algorithm to spot 

two birds in the differenced image. This can be seen in 

the center and right image of Figure 2, where a bounding 

box is located at the same location as bird’s location in the 

previous frame. 

Gaussian Mixture: 

In comparison to frame differencing, we use a 

conceptual approach to present the results that we obtain 

for Gaussian mixture application. Figure 3 shows the 

change in a pixel value over a range of 1000 frames. A 

sudden dip in the pixel value around frame number 500 

may represent a moving object over that pixel region and 

Figure 2. Application of frame differencing for bird detection. Frames 51-53 from video sequence 4 are displayed from left to right in 

order. 



Deep J. Shah 

Figure 3. Pixel distribution over the sequence of 1000 frames 

before applying Gaussian Mixture. 

Figure 4. Distribution of most probable pixel mean value after 

applying Gaussian Mixture shows a step increase in its mean. 

Figure 5. Histogram displaying the distribution of the pixel value 

for a particular pixel over range of 1000 frames. 

Figure 6. Model of a pixel as a mixture of K Gaussians. 

hence it would be modeled as the foreground. Figure 4 

shows the mean value of most probable Gaussian curves 

after it gets updated and re-ordered. Figure 4 also shows 

that the mean of a particular pixel gets updated at around 

every 300 frames. 

Figure 5 is a histogram of the pixel values for the 

same 1000 frames as used above. The distribution fits 

very closely to a regular Gaussian curve. Figure 6 displays 

three Gaussian curves for the same pixel with a mean 

and standard deviation as seen in the figure. Comparing 

Figure 5 and 6 we see that the Gaussians for each pixel gets 

updated appropriately and attempts at providing the best 

background model. 

Discussion 

Distances of centroids outside the bounding box are 

almost twice as much as the distances of centroids inside the 

bounding box to N nearest centroids, as displayed in Table 

1. This demonstrates a good algorithm that could be used to 

weed out the noise centroids (outside bounding box) from 

the bird centroids (inside bounding box). However, this 

approach fails to be effective as there are several centroids 

that exist even in frames that do not have any bird. Due to 

this it becomes extremely challenging to distinguish noise 

from a bird when it comes to frames that have several nonbird 

centroids close to each other. 



Deep J. Shah 

The Gaussian mixture and frame differencing 

approaches are compared below to identify how they fair 

against each other. 

Accuracy 

Speed 

Memory 

Requirements 


Detects a lot of noise. 

Less accurate. 

Inversely proportional 

to the number of pixels in 

a frame. 

Proportional to 

image size. 

Gaussian 

Mixture 

Adapts itself to background 

changes. 

Multi-modal distribution 

reduces noise. 

Inversely proportional to the 

number of pixels per frame 

× the number of Gaussian 

per pixel (K = 5) 

Proportional to image size 

× the number of Gaussians 

used (K = 5). 

We see that each method has its own advantages and 

disadvantages related to it. Gaussian mixture will provide 

far more accurate results as it is less susceptible to noise. 

Contrastingly, frame differencing requires less memory 

and is more rapid in its simulations. These factors play 

an important role in designing a system that has its own 

specific requirements and constraints 

Summary and Future Work 

Frame differencing is a very primitive technique 

that could be implemented very easily. In comparison 

Gaussian Mixture approach requires several resources for 

it to be effective. Hence, such two contrasting approaches 

are used for the analysis performed in this study. Further 

on a similar study can be conducted to understand how 

other background segmentation techniques such as Pfinder, 

LOTS, W 4 , Halevy, and Cutler, which are not described in 

this paper, would perform in a similar situation. This will 

help us identify an approach that would be most suitable to 

a given system’s unique requirements. 

Acknowledgements 

I would like to thank the following people and 

organizations for their support through funding, resources 

and mentorship: Center for Embedded Network Sensing at 

UCLA, National Science Foundation, Dr. Deborah Estrin, 

Wesley Uehara, and Karen Kim. 

References 

1. 

2. 

Otsu, N. “A Threshold Selection Method from Gray 

– Level Histograms.” IEEE Transactions on Systems, 

Man, and Cybernetics. 9. 1(1979): 62-66. 

Piccardi, Massimo. “Background Subtraction 

techniques: A Review.” The ARC Center of 

Excellence for Autonomous Systems (CAS). Faculty 

of Engineering, UTS, Sydney, 2004. 

3. Gonzalez, Rafael R., and Eddins, Steven E. Digital 

Image Processing Using MATLAB. New Jersey: 

Pearson Education, Inc., 2004. 

4. 

5. 

Alan M. McIvor, “Background Subtraction 

Techniques”, Image and Vision Computing, New 

Zealand. Hamilton, New Zealand, November 2000. 

Stauffer C, Grimson W. E. L. “Adaptive background 

mixture models for real-time tracking.” 1999 IEEE 

Computer Society Conference on Computer Vision 

and Pattern Recognition. 06/23/1999 – 06/25/1999, 

Fort Collins, CO, USA. 06/08/2002. 


Love a Son, Raise a Daughter: A Cross-Sectional Examination 

of African American Mothers’ Parenting Styles 

James M. Telesford 1, 2 , Carolyn B. Murray 1 

1 

Department of Psychology, 2 Department of Sociology 


ABSTRACT 

The primary focus of this study is to answer the question: “Do African American mothers ‘raise’ their 

daughters but ‘love’ their sons?” This element of Black folklore has been around for more than two 

decades, but it has little empirical evidence (Randolph, 1995). Indirect support for the belief is found 

in studies reporting that parents are more permissive with children of the opposite sex (Williams, 

1988). As part of a larger four-year longitudinal project examining socialization and personality 

development in African American families, 94 mothers and their 7-year-old (n=26), 10-year-old (n=26), 

13-year-old (n=23), or 16-year-old (n=19) daughters or sons were videotaped while discussing a 

topic upon which they disagreed but were directed to come to a consensus. Four raters assessed 

these dyads on the degree of warmth and control exhibited by the mothers. In addition, the children 

were examined to discover whether there were gender differences in the way they behaved with their 

mothers. While no evidence was found for the mothers behaving differently with their sons, there 

was clear evidence that boys behaved differently than girls with their mothers. 


Carolyn B. Murray 

Department of Psychology 

James Telesford, as a research assistant and Honors student for the past two years, 

has shown himself to be a serious, bright, and highly motivated researcher. In my 

capacity as his research mentor I exposed James to my data set that investigated 

African American (AA) family socialization practices. James carved out data 

to investigate whether mothers “raise” their daughters but “love” their sons. This element of AA 

parenting folklore has been in existence for more than two decades, but it has never been empirically 

verified. The research procedures required mothers and their children to interact with each other while 

being videotaped. James’ thesis is unique in that interactions among African Americans are seldom 

studied and even more rarely videotaped. Most of the existing literature was collected via paper 

and pencil survey instruments, often retrospectively. James’ research should do much to enlighten 

psychologists and other professionals’ understanding of AA family communication. This can itself 

lead to greater appreciation for strengths within the AA family and to improved interventions when 

addressing problematic issues. 

A U T H O R 

James M. Telesford 

Psychology and Sociology 

James Telesford is a graduating senior 

with a double major in Psychology and 

Sociology. His broad research interests 

include racism, discrimination, stereotypes, 

and institutional influences on 

cognition. Currently, he is completing his 

Honors thesis for the University Honors 

Program. His thesis focuses on African 

American mother/child dynamics, specifically, 

examining the stereotype that 

the “African American mother ‘loves’ 

her son but ‘raises’ her daughter.” He 

hopes to further pursue his research 

in graduate school. He thanks his faculty 

mentor for her unwavering support 

and guidance, and his parents for their 

unconditional love and support. 


Love a Son, Raise a Daughter: A Cross-Sectional Examination of African American Mothers’ Parenting Styles 


There are clear inequities between racial groups with 

regard to enrollment in higher education. For example, 

European Americans between the ages of 18-19 are much 

more likely than African Americans to be enrolled in 

college (U.S. Bureau of the Census, 2005). While these 

inequalities are evident between races, there are also more 

subtle disparities within racial groups that should not be 

ignored. These within group differences are particularly 

striking with regard to gender. In 1967, an estimated 22% 

of African American males were enrolled in college, while 

only 15% of African American females were enrolled 

(U.S. Bureau of the Census, 2005). In contrast, the U.S. 

Bureau of the Census (2005) reported that 35% of African 

American males were enrolled in college, while 45% of 

African American females were enrolled. In a mere 40 

years, a significant shift has been observed, such that more 

African American females are now enrolled in institutions 

of higher learning than are African American males. Indeed, 

it has been demonstrated that more African American males 

are incarcerated in the prison system than are enrolled in 

colleges (Western & Petit, 2005). 

Since the publication of the Moynihan report (1965), 

the African American mother has been blamed for the 

break down of the African American family and has been 

considered the primary cause for all that ails the community. 

Indeed, a folk saying has emerged in the African American 

community that the mother “loves” her son, but “raises” 

her daughter – implying that this difference in parenting 

has produced the high incarceration rate of Black men. 

If this statement holds true, it has important implications 

for both incarceration rates and enrollment rates on college 

campuses. Perhaps it is the lax parenting style that the mother 

exhibits toward her son that fails to prepare him for society. He thus 

ends up incarcerated rather than enrolled in college. The converse 

of this is that the mother’s instructive and controlling style toward 

her daughter is the reason for the corresponding higher college 

enrollment among Black women. While this statement has often 

been expressed, it has received sparse attention in the research 

literature. Indirect support for the belief is found in studies reporting 

that parents are more permissive with children of the opposite sex 

(Williams, 1988), and that cross-sex parenting emphasizes noncontingent 

or “unconditional” love, whereas same sex parenting 

emphasizes performance-conditional love (Jones & Berglas, 

1978). Thus, the focus of this paper is four-fold. First, we will 

review the literature regarding African American mothers, sons, 

and daughters. Second, we will examine whether or not African 

American mothers treat their children differently. Third, we will 

examine if the children themselves behave differently when 

interacting with their mothers. Finally, we will determine if there 

are any discrepancies in the way that the mother and children 

interact with each other with respect to the age of the children. 

The African American Mother 

The African American mother faces truly unique 

challenges. She must contend with both racism and sexism 

(Cauce et al., 1996). Everyday, she is faced with negative 

stereotypes about her race and gender on television and in 

magazines and newspapers. In addition, according to the 

U.S. Bureau of Labor Statistics (2002), African American 

females are more likely than any other group to face 

economic challenges. 

Besides economic, racial, and gender based 

difficulties, the African American mother faces the difficulty 

of imparting an Afrocentric ideology within a Eurocentric 

culture. Whereas a Eurocentric cultural view focuses on 

individualism, materialism, reason, and differences, an 

Afrocentric cultural view stresses community, spirituality, 

affect, and similarities (Kambon, 1998). These differing 

cultural views often come into conflict. For example, when a 

successful African American executive must work Sundays 

in order to pursue individualistic and materialistic goals, 

she may give up spirituality by not attending church and 

thus lose the sense of community that the church provides 

through fostering affective growth within her family. 

African American Mothers and Sons 

Without a doubt, the African American mother is 

extremely important to the African American son. For 

instance, research has shown that a positive, supporting 

relationship with his mother negatively correlates with 

a son’s likelihood of exhibiting internalizing, anxious, 

depressed, and/or withdrawn behavior (Murata, 1994). With 

the increasingly Black female single-parent family condition 

(Smith & Smith, 1986), Black mothers are left with the 

challenge of teaching their sons how to become men. While 

the literature on this topic is sparse, Lawson Bush (2000) has 

attempted to show how Black mothers accomplish this goal. 

One way in which they do this is by telling their sons stories 




of how their ancestors dealt with their hardships (King & 

Mitchell, 1990). Another technique is by instilling guilt 

within their sons by explaining to them how behaviors that 

are not consistent with good morals and principles upset and 

disappointed them (King & Mitchell, 1990). 

Some researchers have found that parents support 

and validate their African American sons much more 

so than daughters, and this is done possibly to protect 

them from the threats of discrimination in the real world 

(Smetana, Abernethy, & Harris, 2000). Others have shown 

that when caring for their chronically ill sons, African 

American mothers perceive them as being frail and less 

healthy; the mothers are thus more likely to limit their 

sons’ involvement in certain activities, such as sports (Hill 

& Zimmerman, 1995). The concern with protecting their 

sons from the cruel world of discrimination may be one 

explanation for why the African American mother “loves” 

her son – at least to the extent that such “love” is centered 

around “protection.” 

African American Mothers and Daughters 

On the other hand, African American mothers are 

shown to interact with their daughters in very different 

ways than they do with their sons, yet have similar goals in 

mind. Mothers are trying to protect both sons and daughters 

from covert and overt societal discrimination. However, 

for sons, mothers shelter them from discrimination by 

providing validation and support (Smetana, Abernethy, 

& Harris, 2000), and explaining how ancestors have dealt 

with it (King & Mitchell, 1990). For daughters, they try to 

teach them how to handle discrimination by being strong 

and independent (Hill-Collins, 2001). 

While it is definitely difficult for a mother to teach her 

son how to become a man, teaching a daughter to become a 

strong, independent woman contains diametrically opposed 

concepts. On one side, the daughter is supposed to identify 

with the mother to learn the roles of her gender. On the 

other side is the patriarchal society of the United States, 

where men are often valued over women. As a result, the 

daughter may be motivated to resist identifying with the 

mother in order to conform to the conventional view of 

femininity (Hill-Collins, 2001). 

The Hill and Zimmerman (1995) study discussed 

previously showed that, when caring for their female 

children with Sickle Cell Disease, mothers were more 

likely to allow their daughters freedom, see them as 

healthier, and see them as better able to handle the physical 

pain of disability than their sons. These startling contrasts 

existed even though no known gender differences exist 

within those afflicted with Sickle Cell Disease (Hill & 

Zimmerman, 1995). This emphasis by mothers on selfreliance 

toward their daughters may explain why African 

American mothers are seen as “raising” their daughters. 

Influencing the Black Mother 

As most social interactions require the engagement 

of both parties, with one party responding to the other, it 

may be the children themselves that are behaving differently 

toward the mother. Indirect evidence for this idea comes 

from Cowan and Avants (1988) who found that: 1) sons 

were more likely to use Autonomous Influence strategies 

with their mothers (e.g., telling or asking the mother if they 

could do what they wanted to do), and 2) daughters were 

more likely to use Anticipating Noncompliance strategies 

with their mothers (e.g., crying, persistence, anger, 

begging, and/or pleading). Thus, the researchers concluded 

that the son might have more influence over the mother by 

using self-sufficient strategies during persuasion, whereas 

the daughter may have less influence and must anticipate 

their mothers’ noncompliance. However, this study was 

conducted on a Caucasian sample and thus may not be 

representative of a Black behavior. 

Preliminary Exploratory Questions 

The present study investigated several research 

questions. First, we examined if males, in contrast to 

females, behave differently when interacting with their 

mother. Second, we examined whether or not mothers, when 

interacting with younger and older children, differentially 

express affection and guidance. Finally, we examined 

whether or not Black mothers “love” (show more affection 

towards) their sons and “raise” (express more guidance 

for) their daughters.” In sum, are there gender differences 

and/or age differences in the way mothers’ parent, and/or 

the way children interact with their mothers? 




Methods 

Participants consisted of 94 African American 

mothers and their children. The children were 7 (n=26), 

10 (n=26), 13 (n=23), or 16 (n=19) years of age. 44% of 

the sample consisted of male children, while 56% were 

female children. The participants were drawn from a larger 

four-year longitudinal project examining socialization and 

personality development in African American families. 

The mother-child dyads were videotaped while 

discussing a moral dilemma upon which they disagreed but 

about which they were directed to come to a consensus. Four 

trained raters then viewed the videotaped interactions and 

assessed these dyads on the degree of warmth and control 

exhibited by the mothers, and warmth and independence 

exhibited by the children. The ratings form consisted of 51 

items rated on a 5 point Likert-type scale, with 0 indicating 

the behavior was “not at all characteristic” and 4 indicating 

that the behavior was “completely descriptive” of that item. 

The form was divided into three sections: 20 descriptors 

for only the child’s behavior, 19 descriptors for only the 

mother’s behavior, and 12 questions about the interaction 

as a whole. 

Results and Discussion 

Inter-rater reliability was assessed using Chronbach’s 

Alpha to determine the extent to which the four raters 

agreed on the interactions they were viewing. Results 

indicated that the raters were reliable in assessing the dyads, 

as alpha had a range of .72 to .91. Next, a 2 (gender: male 

and female) X 4 (age group: 7, 10, 13, and 16) multivariate 

analysis of variance (MANOVA) with the child and mother 

descriptors serving as the within-subjects variables was 

conducted to examine the exploratory questions. 

Exploratory question one was “do males in contrast 

to females behave differently when interacting with 

their mothers?” Several main effects resulted for gender 

differences in the way children were perceived as behaving. 

Daughters were seen as significantly more initiating, 

assertive, happy, talkative, warm, loving, stubborn, and 

challenging than were sons when interacting with their 

mothers (See Table 1). 

A significant interaction effect resulted for child’s 

description as stubborn with gender by age, (F(1, 3) 

Child Descriptors 

Sons 

Daughters 

Initiating 1.47(1.37) 1.88(1.33) 

Assertive 1.67(1.48) 2.12(1.49) 

Happy 2.13(1.14) 2.70(1.08) 

Talkative 2.10(1.28) 2.74(1.18) 

Warm 2.23(1.12) 2.80(1.08) 

Loving 2.31(1.17) 2.84(1.08) 

Stubborn .51(.99) .77(1.22) 

Challenging .44(.92) .68(1.12) 

Mother Descriptors 

Comfortable 3.02(.82) 3.22(.75) 

Relaxed 3.05(.84) 3.26(.73) 

Happy 2.36(1.06) 2.68(1.06) 

Note. All nonsignificant descriptors were omitted; N=308. 

Table 1. Means (SD) for Descriptor Variables by Child’s Gender 

Child Descriptors 

7-Year-Olds 10-Year-Olds 13-Year-Olds 16-Year-Olds 

Happy 2.82(1.02) 2.26(1.05) 1.96(1.16) 2.44(1.21) 

Loving 2.89(1.03) 2.41(1.12) 2.30(1.27) 2.60(1.16) 

Dependent 1.73(1.14) .98(.81) .86(.99) .83(.94) 

Initiating 1.35(1.27) 1.52(1.32) 1.56(1.30) 2.42(1.32) 

Assertive 1.75(1.51) 1.54(1.40) 1.81(1.44) 2.53(1.46) 

Mother Descriptors 

7-Year-Olds 10-Year-Olds 13-Year-Olds 16-Year-Olds 

Concerned 2.68(1.18) 2.75(1.02) 2.91(.92) 2.36(1.14) 

Assertive 2.64(1.08) 2.38(1.30) 3.00(1.14) 3.21(.92) 

Comfortable 3.31(.67) 2.97(.86) 2.68(.88) 3.33(.66) 

Warm 3.42(.74) 2.95(.85) 2.70(.88) 2.83(1.17) 

Caring 3.43(.73) 3.13(.81) 2.93(1.06) 2.82(1.10) 

Happy 2.91(.95) .10(.44) .16(.63) .09(.43) 

Loving 3.44(.73) 3.20(.78) 2.91(1.12) 2.81(1.09) 

Note. All nonsignificant descriptors were omitted; N=308. 

Table 2. Means (SD) for Descriptor Variables by Child’s Age 

=3.88, p < .01; see figure 1), such that 16-year-old girls 

were perceived as more stubborn than 16-year-old boys. A 

second significant interaction effect was found on the child’s 

description as challenging with gender by age (F(1, 3) = 

3.23, p < .05; see figure 2), such that 16-year-old girls were 




Figure 1. Interaction effect for “stubborn” as a function of the 

child’s age and child’s gender. 

also perceived as more challenging than 16-year-old boys. 

Exploratory question two was “do mothers express 

affection and guidance differently across the age groups?” 

The MANOVA indicated that mothers behave differently 

with children based on their ages (See Table 2). Mothers 

were viewed as significantly more likely to be concerned 

when interacting with 13-year-old children than with 

16-year-old children, perhaps reflecting the transition from 

puberty to young adulthood. More evidence of this shift to 

young adulthood is the fact that mothers of the 16-year old 

children were more likely to be viewed as assertive with 

their children than the mothers of 7 or 10-year-old children. 

Also, mothers were seen as much more comfortable around 

their 7-year-old children (presumably before puberty) and 

16-year-old children (presumably after puberty), than with 

10 and 13-year-old children (presumably during puberty). 

Mothers were rated as more warm, caring, happy, and 

loving toward their 7-year-old children than with other age 

groups. These latter results may indicate that children at 

the age of 7 require a more loving relationship. A second 

probable interpretation is that 7 year olds do not cause their 

mothers as much stress as do older children. Further, it was 

discovered that children at different ages interact differently 

with their mothers. Several main effects resulted with 

the child descriptors by the child’s age. 7-year-olds were 

viewed as significantly happier and more loving than 10 

and 13-year-olds, but not 16-year-olds. 7-year-olds were 

also seen as more dependent than all other age groups. 

16-year-olds were rated as significantly more initiating and 

assertive than all other age groups (See Table 2). These 

Figure 2. Interaction effect for “challenging” as a function of the 

child’s age and child’s gender. 

differences most likely reflect the different developmental 

stages of the children. 

The major exploratory question was “do mothers 

show more affection towards their sons than toward their 

daughters, and express more guidance towards their daughters 

than toward their sons?” Several main effects resulted with 

the mother descriptors by the child’s gender. Mothers were 

perceived as more comfortable, relaxed, and happy when 

interacting with their daughters (See Table 1). While these 

main effects do not support the notion that “African American 

mothers ‘love’ their sons, but ‘raise’ their daughters,” it does 

show support for the idea that mothers interact differently 

with children based on the gender of the child. However, 

there were no perceived gender based differences in the way 

mothers used control with their children. 

The MANOVA resulted in a significant interaction 

effect for mothers as uninterested by child’s gender 

by child’s age, (F(1, 3) = 2.92, p < .03). Mothers were 

perceived as being more interested when interacting with 

their 16-year-old daughters than when interacting with 

their 16- year-old-sons. The previously reported finding 

that 16-year-old girls were significantly more challenging 

and stubborn than were boys may help explain mothers 

showing more interest when interacting with the girls. 

In addition, this shows support for the findings of other 

researchers that the Black mother socializes her daughter 

to be strong and independent (Collins, 2001). When the 

daughters are challenging and stubborn with their mothers, 

the mothers may positively reinforce these qualities by 

approaching the interaction with more interest. 




Limitations 

One limitation to our study is that the task the dyads 

were required to complete may be biased toward eliciting 

certain behaviors from daughters over sons. Future 

research should design and employ tasks that are both 

gender balanced and engaging (e.g., sports, clothes, etc.). 

Another limitation is that this was qualitative data taken 

from participants in front of a video camera. The video 

camera may have skewed the behaviors of the participants 

such that they did not truly act, as they would have in a 

real world situation. Future research should also include 

self-report data on parent-child interactions that actually 

occurred in their lives. Despite these limitations, this study 

is significant in that it represents the first examination of 

how the child’s gender affects the manner in which their 

mothers socialize African-American children. 

Conclusion 

With the ever-rising incarceration rates of young 

African American males, the growing gender disparities 

with regards to enrollment rates in college for African 

Americans, and these situations being blamed on the 

African American mother, it becomes extremely important 

to investigate the relationship that the mother has with her 

children. This study represents the first attempt at this. 

While no differences were found in the way that the mother 

uses control with her children, the mother did appear to 

have a more warm and guiding-producing relationship with 

her daughter. This finding runs contrary to the stereotype 

that the “African American mother ‘loves’ her son, but 

‘raises’ her daughter.” Further, it was found that the 

children themselves have different interactional patterns 

with their mothers. It was also discovered that there were 

age/developmental differences in the way that both the 

mothers and the children interacted with each other. Further 

research should investigate this important relationship, as 

well as further disentangle the variables of age and gender 

within the relationships. 




References 

1. 

2. 

3. 

4. 

5. 

6. 

7. 

Brewer, R. M. & Heitzeig, N. A. (2008). The 

racialization of crime and punishment: Criminal 

justice, color-blind racism, and the political economy 

of the prison industrial complex. American Behavioral 

Scientist, 51, 625-644. 

Cauce, A. M., Hiraga, Y., Graves, D., Gonzales, N., 

Ryan-Finn, K., & Grove, K. (1996). African american 

mothers and their adolescent daughters: Closeness, 

conflict, and control. In B. J. Ross Leadbeater, & 

N. Way (Eds.), Urban Girls: Resisting Stereotypes, 

Creating Identities (pp. 100-116). New York: New 

York University Press. 

Chappell, A. T. & Maggard, S. R. (2007). Applying 

black’s theory of law to crack and cocaine dispositions. 

International Journal of Offender Therapy and 

Comparative Criminology, 51, 264-278. 

Cowan, G., & Avants, S. K. (1998). Children’s 

influence strategies: Stucture, sex differences, and 

bilateral mother-child influence. Child Development, 

59, 1303-1313. 

Hill, S. A., & Zimmerman, M. K. (1995). Valiant girls 

and vulnerable boys: The impact of gender and race 

on mothers’ caregiving for chronically ill children. 

Journal of Marriage and the Family, 57, 43-53. 

Hill-Collins, P. (1991). Black mother-daughter 

relationships. In P. Bell-Scott (Ed.), Double Stitch (pp. 

52-57). Boston: Beacon Press. 

Jones, E. E., & Berglas, S. (1978). Control of 

attributions about the self through self-handicapping 

strategies: The appeal of alcohol and the role of 

underachievement. Personality and Social Psychology 

Bulletin, 4, 200-206. 

8. Kambon, K. K. K. (1998). African/Black Psychology in 

the American Context: An African-Centered Approach. 

Tallahassee, FL.: Nubian Nation Publications. 

9. King, J & Mitchell, C. (1990). Black Mothers to Sons: 

Juxtaposing African-American Literature with Social 

Practice. New York: Peter Lang. 

10. Lawson Bush, V. (2000). Black mothers/black sons: 

A critical examination of the social science literature. 

The Western Journal of Black Studies, 24, 145-154. 

11. Moynihan, D.P. (1965). The Negro Family: The Case 

for National Action. Washington, D.C.: Government 

Printing Office. 

12. Murata, J. (1994). Family stress, social support, 

violence, and sons’ behavior. Western Journal of 

Nursing Research, 16, 154-168. 

13. Randolph, s. (1995). African American children in 

single-mother families. In B. Dickerson (Ed.), African 

American Single-Mothers: Understanding Their Lives 

and Families (pp. 117-145). Thousand Oaks, CA: 

Sage Publications. 

14. Smetana, J. G., Abernathy, A., & Harris, A. (2000). 

Adolescent-parent interactions in middle-class 

African American families: Longitudinal change and 

contextual variations. Journal of Family Psychology, 

14, 458-474. 

15. Smith, E. J. & Smith, P. M. (1986). The Black female 

single-parent family condition. Journal of Black 

Studies, 17, 125-134. 

16. U.S. Bureau of the Census. (2005). The Population 18 

and 19 Years Old by School Enrollment Status, Sex, 

Race, and Hispanic Origin. Retrieved March 9, 2008, 

from http://www.census.gov/population/socdemo/ 

school/TableA-5b.xls 

17. U.S. Bureau of Labor Statistics. (2002). Unemployment 

Rate - Black or African American. Retrieved March 

9, 2008, from ftp://ftp.bls.gov/pub/special.requests/lf/ 

aat24.txt 

18. Western, B. & Pettit, B. (2005). Black-White wage 

inequality, employment rates, and incarceration. 

American Journal of Sociology, 111, 553-578. 

19. Williams, S. W. (1988). Parent-adolescent conflict 

management styles as functions of age, gender, 

family cohesion, and family adaptability. Dissertation 

Abstracts International, 48(7-A), p. 1899. 



Mating-type distribution of the rice blast pathogen 

Pyricularia grisea in California 

R. Z. 1, 2 , G. W. Douhan 3 , F. Wong 3 

1 

Department of Biology, 2 Department of Biochemistry, 

3 

Department of Plant Pathology and Microbiology 


ABSTRACT 

Pyricularia grisea causes the plant disease known as rice blast and is one of most devastating diseases 

of rice worldwide. The fungus was discovered for the first time in California in 1996 and attempts to 

control the disease have been made by using resistant cultivars and fungicide applications. The longterm 

objective of this research project is to study the genetic structure of this fungus to determine if 

it has changed over time due to factors such as fungicide usage, resistant rice cultivar deployment, 

new introductions, and sexual reproduction of the fungus. The objective of this specific study was to 

determine the mating-type distribution of P. grisea. Each isolate of this fungus possesses a single gene 

for mating type with one of two alleles, either MAT 1-1 or MAT 1-2. Only isolates of opposite matingtype 

are able to sexually reproduce, otherwise all reproduction is from asexually derived spores. We 

used a PCR assay to determine the mating-type of 168 isolates collected from diseased rice in North 

Central California. One hundred and sixty isolates were found to be MAT1-1 and no PCR products 

were amplified from 8 isolates. This result suggests that P. grisea populations associated with rice crops 

in California are reproducing clonally. However, further work using additional molecular markers is 

being pursued to better understand the population dynamics of this important plant pathogen. 

A U T H O R 

R. Z. Urak 

Biology and Biochemistry 

Ryan Urak is studying a double major 

in Biology and Biochemistry with an 

Anthropology minor. He enjoys all 

aspects of research ranging from biology 

to humanities. His current research 

involves molecular work with DNA; 

efforts that he hopes will aid him in a 

profession of Forensic Anthropology. 

Faculty Mentors 

G. W. Douhan 


In my lab we examine the population genetics of fungal species over spatial scales 

ranging from centimeters to large landscapes using a number of molecular methodologies 

including multilocus genotyping and multi-gene genealogies. My program also focuses 

on developing rootstocks for avocado that are resistant to Phytophthora cinnamomi, 

the most destructive and important pathogen of avocado worldwide. Ryan started in my lab in fall 

of 2007 as a third year student. With his interest in anthropological forensics, he trained in molecular 

biology methods, including DNA isolation, polymerase chain reaction, and genotyping techniques. 

Ryan’s research project utilized some of these tools and answered one of our first questions about the 

reproductive biology of Pyricularia grisea. This fungus is likely not going through a sexual cycle under 

field conditions, which is our first step at understanding the reproductive biology and population genetic 

structure of this important pathogen. 

F. Wong 


My research and extension program focus on the management of diseases of turf and 

landscape in California. Of highest importance is developing control strategies for 

emergent and invasive diseases. Pyricularia grisea is an emergent fungal pathogen that 

was recently found to be causing significant damage to rice (1997), perennial ryegrass 

and kikuyugrass (2003). Because of the economic value of rice, and turf used on golf courses and sports 

fields, the control of P. grisea is important to California agriculture. As part of a project funded by the 

University of California Exotic and Invasive Pests and Diseases Program, Ryan helped examine the 

population structure of the pathogen from rice in California. Information generated from this study is 

important in understanding how the pathogen is reproducing, surviving and spreading in California, and 

how maximize the effectiveness of management programs. 


Mating-type distribution of the rice blast pathogen Pyricularia grisea in California 

R. Z. Urak 

Introduction 

Rice is an important agricultural commodity that 

supplies approximately 23 % of the per capita energy 

for six billion people worldwide (Maclean, 1997). There 

are many serious plant diseases of rice, including the 

ascomycete fungus Pyricularia grisea (Telemorph: 

Magnaporthe grisea) which causes the disease known as 

rice blast (Correll, 2000). Pyricularia grisea can infect 

most sections of the plant, but infections of the node or 

the panicle are the most damaging phases of the disease 

(Ou, 1985). When P. grisea infects rice and produces 

neck rot or panicle blast, it will either kill the host plant 

or prevent seed development, respectively. P. grisea also 

causes disease in other graminacious species besides rice 

(Malca, 1957; Bain, 1972; Ou, 1985; Sundaram, 1972) and 

there are reports of this pathogen in more than 85 countries 

(Agarwal, 1989). 

P. grisea was first identified in California in 1996 

(Greer, 2001), which was unexpected due to P. grisea’s 

common association with high humidity conditions 

(Webster, 1992) which is unlike temperate Sacramento 

Valley where the disease occurs. Seeds, crop residue, and 

secondary hosts are all possible origins for the introduction 

of into California and could have been the primary sources 

of inoculum for the disease (Agarwal, 1989; Lee, 1994; 

Ou, 1985; Rao, 1994; Teng, 1994). Now that P. grisea is 

present in some rice fields in California, usage of fungicides 

and deployment of resistant cultivars is the best course of 

action to control the disease. 

By studying the genetic structure of this invasive 

pathogen, we will be able to make inferences on how the 

fungus spreads, reproduces, and how the population is 

responding or adapting to current management practices. 

The overall objective of this research project is to analyze 

the genetic structure of P. grisea from collections of isolates 

from the initial introduction of the disease in the mid 1990’s 

to more recently sampled isolates using molecular markers. 

This would shed light on whether or not any changes in 

the genetic structure of the fungus have occurred over 

an approximate 10 year period which may be linked to 

fungicide usage, resistant rice cultivar deployment, new 

introductions, or due to sexual reproduction of the fungus. 

P. grisea is a heterothallic fungus with a single 

mating type gene that produces two alleles, MAT 1-1 

and MAT 1-2. The pathogen requires both mating types 

in order for sexual reproduction to occur (Yoder, 1986), 

and mating type alleles have been used as a marker to 

measure population diversity in this pathogen (Viji, 

2002). The specific objective of this part of the project 

is to use mating-type to measure and assess population 

diversity in California populations of the pathogen from 

rice and determine if sexual reproduction is possible in 

these populations. Populations collected from both early 

outbreaks in 1997 and 1998, and more recent ones in 2007, 

were targeted. These results will provide initial information 

concerning potential sexual reproduction of P. grisea in 

California rice fields and give a preliminary measure of any 

changes in population diversity since the initial discovery 

of the pathogen on rice in 1996. 

Results 

All 33 isolates from the historical 1997 and 1998 

collections were all identified through PCR as mating-type 

MAT 1-1 (Table 1). Of the 135 P. grisea isolates collected 

in 2007, 127 of the isolates were the mating-type MAT 1-1 

(Table 1) (Figure 1). Isolates collected from fields A, D, E 

and F all produced the 552 bp product associated with the 

presence of MAT 1-1. For populations collected from fields 

B and C, eight isolates did not produce a PCR product 

specific for MAT 1-1. Subsequent PCR assays using the 

MAT 1-2 specific primers also failed to produce a PCR 

product specific for this allele. 

Discussion 

We detected only a single mating-type occurring in 

populations of P. grisea isolated from rice in California. 

This suggests that P. grisea is not sexually reproducing, 

which can be important in the population dynamics of this 

pathogen. Sexual reproduction can reshuffle genetic material 

via recombination, thereby bringing together new alleles, 

which may influence pathogenicity to various rice cultivars 

and could influence the efficacy of fungicides. MAT 1-1 

has been identified in past studies as the dominant matingtype 

associated with rice. In a survey of 467 P. grisea rice 

isolates from 34 countries in Europe and Africa, only mating 

type MAT 1-1 was found (Notteghem, 1992). Research 

in Japan also yielded similar results, as all surveyed rice 



R. Z. Urak 

Mating –type 

Year Field MAT 1-1 MAT 1-2 No amplification 

2007 A 26 0 0 

“ B 26 0 1 

“ C 16 0 7 

“ D 16 0 0 

“ E 17 0 0 

« F 26 0 0 

1997 97 10 0 0 

1998 98 23 0 0 

Table 1. Mating-type distribution of P. grisea based on 

mating-type specific PCR. 

Columns 

1 2 3 4 5 Ladder 

500 bp 

Figure 1. Agarose gel showing the PCR product (550 bp) for the 

MAT1-1 allele (in columns 1-5) from some of the representative 

isolates of P. grisea used in this study. 

isolates belonged to MAT 1-1 (Kato, 1982). Despite these 

consistencies, a mating-type analysis of Californian isolates 

is important because sexual reproduction can affect diversity 

and dissemination, and both mating-types have been found 

in other studies. For example, Dayaker (2000) found that 39 

of the 74 isolates from India were MAT 1-1 and the other 35 

isolates were MAT 1-2. 

California P. grisea isolates also showed that there 

was no significant change in mating-type from 1997-1998 

to 2007, strongly suggesting that the pathogen survives and 

reproduces through asexual means. As for the few isolates 

that produced no results, it is believed that this is a case of 

failed PCR, perhaps due to user error of DNA concentration 

in the reactions, or that the DNA preparations contained 

inhibitors, which are common sources of failed reactions. 

Further work and more meticulous determinations of DNA 

concentrations in sample extracts will address these issues, 

but the current data suggests that the MAT 1-2 allele is not 

present in the California populations of P. grisea. 

Most studies on the population structure of P. grisea 

associated with rice support the idea that this pathogen is 

highly clonal, as evident by the use of molecular makers 

and population genetic analysis to identify the fertility of the 

isolates (Viji, 1998). The predominance of a single matingtype 

in most turfgrass (Tredway, 2003) and perennial 

ryegrass (Viji, 2002) isolates also supports that P. grisea is 

primarily an asexual fungus, though both mating-types have 

been associated with other turf hosts such as kukuya grass 

(Wong, F. unpublished). However, P. grisea shows strong 

host preference, which suggests that gene flow does not occur 

between isolates associated with separate host species. 

The severity of P. grisea has decreased since the 

initial 1996 introduction because, as previously postulated, 

P. grisea cannot flourish in the environmental conditions 

that exist in California. California rice production takes 

place in a climate that is permissive for rice blast but is 

too arid to allow the onset of significant epidemics in most 

years. However, control is still needed when conditions are 

favorable for the disease (Greer, 2001). Growers primarily 

use resistant rice cultivars (Ou, 1985) but fungicides are 

also used when disease pressure is significant. Azoxystrobin 

is the standard and consequently the only fungicide for 

rice available in California (Greer, 2001). Azoxystrobin 

effectively inhibits spore germination and is therefore 

a protectant prior to infection (Clough, 1998). The few 

resistant rice cultivators and single fungicide have been 

the only means to control rice blast in California during 

the past decade. Our long-term research on the population 

biology of P. grisea isolates collected approximately over 

a ten year period may give insight into how these cultural 

practices have influenced the population structure and the 

reproductive biology of this important pathogen. 



R. Z. Urak 


Sampling 

Isolates from 1997 and 1998 were obtained from Dr. 

Tom Gordon, Department of Plant Pathology, University 

of California, Davis; these were originally isolated and 

described by Greer (2001). Recovery of all of the isolates 

from long-term storage was not possible and recovered 

isolates were simply grouped by year regardless of 

location, and referred to as the 1997 and 1998 collections. 

In 2007, six rice fields (A through F) from Yuba county 

were sampled by randomly collecting diseased panicle 

tissue from along a single transect approximately five 

meters from the edge of the fields. The tissues were placed 

in paper bags, brought back to the laboratory, and air-dried 

in the fume hood for several days. 

DNA extraction and analysis 

To isolate P. grisea, infected tissue was surface 

sterilized, and the lesions were cut in half. The tissues were 

placed in petroleum jelly that was positioned on the lids of 

acid potato dextrose agar (aPDA) plates so that they would 

be elevated. After 48 hours, the lids were tapped to release 

the spores and the plates were allowed 4-7 days for growth. 

Five germinated single spores were randomly selected and 

removed from the aPDA plates. These were then transferred 

to clean aPDA plates. PDA was also used to regrow the 

1997 and 1998 collection of P. grisea isolates. 

Isolates were allowed to grow at room temperature 

for approximately one week before the hyphae were scraped 

and placed into cetylrimethylammonium bromide (CTAB) 

extraction buffer (2% CTAB, 100 mM Tris, pH 8.4, 10mM 

EDTA, and 0.7 M NaCl) (Gardes and Bruns, 1993). The 

resultant mixture was incubated at 65˚C for one hour. An 

equal volume of chloroform was then added to the mixture 

prior to it being centrifuged for 30 minutes. The extracted 

supernatant was transferred to a separate tube and the DNA 

was precipitated using 3M NaOAc and 70% isopropanol. 

After the mixture was centrifuged for another 30 minutes, 

the supernatant was poured out and the pellet cleaned with 

an ethanol rinse. TE buffer (10mM Tris, 1mM EDTA, 8 pH) 

was used to resuspend the pellet. The suspension was then 

incubated for an hour at 65˚C. The genomic DNA mixture 

was purified by the addition of RNase and one hour of 

heating at 35˚C. Five microliters of each DNA extraction 

were mixed with 0.5 µl of the nucleic acid stain SYBR Green 

I (Molecular Probes, Eugene, OR, USA) in TE, separated in 

a 0.8 % agarose gel, and visualized under UV light. 

The gene encoding for the mating-type was 

amplified by the polymerase chain reaction (PCR) using these 

primers: L1 (5’- ATGAGAGCCTCATCAACGGCA) and 

L2 (5’- ACAGGATGTAGGCATTCGCAGGAC) for MAT 

1-1 and T1 (5’ ACAAGGCAACCATCTGGACCCTG) and 

T2 (5’-CCAAAACACCGAGTGCCATCAAGC) for MAT 

1-2 (Tredway, 2003). Two microliters of Genomic DNA was 

used as the template in a 20 µL PCR mixture (Thermo Pol 

Buffer, 10mM dNTP mix, 5 µM of each primer, and TAQ 

polymerase) using a Mycycler (BioRad) thermalcycler. 

The mixture was subjected to 30 cycles of amplification. 

Thermocycling conditions consisted of an initial hold at 

94˚C for 3 minutes, followed by 30 cycles of 94˚C (30 

sec), 60˚C (30 sec), and 72˚C (1 min), and a final hold of 

72˚C for 8 min. The PCR products were then stained and 

visualized as previously described using a 1.5% agarose gel. 

Any genomic DNA that was not amplified by the MAT 1-1 

primer was processed again with the MAT 1-2 primer. 



R. Z. Urak 

References 

1. 

2. 

3. 

4. 

5. 

6. 

7. 

8. 

9. 

Agarwal, P. C., Mortensen, C. N., and Mathur, S. B. 

1989. Seed-borne diseases and seed health testing of 

rice. Tech. Bull. No. 3, Phytopathological Papers No. 

30. CAB International Mycological Institute, Kew, U. 

K. 

Bain, D. C., Patel, B. N., and Patel, M. V. 1972. Blast 

of ryegrass in Mississippi. Plant Dis. Rep. 56: 210 

Clough, J. M., and Godfrey, C. R. A. 1998. The 

strobilurin fungicides. In: Fungicidal Activity: 

Chemical and Biological Approaches to Plant 

Protection. D. Hutson and J. Miyamoto, eds. John 

Wiley and Sons, Ltd., Chichester, U. K. 

Correll, J. C., Harp, T. L., Guerber, J. C., Zeigles, R. 

S., Liu, B., Cartwright, R. D., and Lee, F. N. 2000. 

Characterization of Pyricularia grisea in the United 

States using independent genetic and molecular 

markers. Phytopathology 90: 1396-1404 

Dayakar, B. V., Narayanan, N. N., and Gnanamanickam, 

S. S. 2000. Cross-compatibility and distribution of 

mating type alleles of the rice blast fungus Magnaporthe 

grisea in India. Plant Dis. 84: 700-704 

Gardes M, and Bruns T. D. 1993. ITS primers with 

enhanced specificity for Basidiomycetes- application to 

the identification of mycorrhizae and rusts. Molecular 

Ecology 2: 113-118 

Greer, C. A., and Webster, R. K. 2001. Occurrence, 

distribution, epidemiology, cultivar reaction, and 

management of rice blast disease in California. Plant 

Disease 85 (10): 1096-1102 

Kato, H., and Yamaguchi, T. 1982. The perfect stage of 

Pyricularia oryzae Cav. In culture. Ann. Phytopathol. 

Soc. Jpn. 48: 607-612 

Lamey, H. A. 1970. Pyricularia oryzae on rice seed in 

the United States. Plant Dis. Rep. 54: 931-935. 

10. Lee, F. N. 1994. Rice breeding programs, blast 

epidemics and blast management in the United States. 

In: Rice Blast Disease. R. S. Zeigler and S. A. Leong, 

eds. CAB. International, Wallingford, U. K. 

11. Maclean, J. 1997. Rice Almanac International Rice 

Research Institute, Los Banos, Phillippines. 

12. Malca, I., and Owen J. H. 1957. The gray leaf spot 

disease of St. Augustinegrass. Plant Dis. Rep. 41: 

871-875 

13. Manandhar, H. K., Lyngs Jorgensen, H. J., Smedegaard- 

Petersen, V., and Mathur, S. B. 1998. Seedborne 

infection of rice by Pyricularia oryzae and its 

transmission to seedlings. Plant Dis. 82: 1093-1099. 

14. Notteghem, J. L, and Silue, D. 1992. Distribution of the 

mating type alleles in Magnaporthe grisea population 

pathogenic on rice. Phytopathology 82: 421-424. 

15. Ou, S. H. 1985. Rice Diseases. 2nd ed. Commonwealth 

Mycological Institute, Kew, U. K. 

16. Rao, K. M. 1994. Rice Blast Disease. Daya Publishing 

House, Delhi, India. 

17. Sundaram, N. V., Palmer, L. T., Nagarajan K., and 

Prescott, J. M. 1972. Disease survey of sorghum and 

millet in India. Plant Dis. Rep. 56: 740-743. 

18. Teng, P.S. 1994. The epidemiological basis for blast 

management. In: Rice Blast Disease. R S. Zeigler and S. 

A. Leong, eds. CAB International, Wallingford, U. K. 

19. Tredway, L. P. and Stevenson K. L. 2003. Mating type 

distribution and fertility status in Magnapothe grisea 

populations from turfgrasses in Georgia. Plant Disease 

87 (4): 435-441. 

20. Viji, G and Gnanamanickam, S. S. 1998. Mating type 

distribution and fertility status of Magnaporthe grisea 

population from various hosts in India. Plant Disease 

82: 36-40. 

21. Viji, G. and Uddin, W. 2002. Distribution of mating 

type alleles and fertility status of Magnaporthe grisea 

causing gray leaf spot of perennial ryegrass and St 

Augustinegrass turf. Plant Disease 8 (8): 827-832 

22. Webster, R. K., and Gunell P. S., eds. 1992. Compendium 

of Rice Diseases. American Phytopathological Society, 

St. Paul, MN 

23. Yoder, O. C., Valent, B., and Chumley, F. G. 1986. 

Genetic nomenclature and practice for plant pathogenic 

fungi. Phytopathology 76: 383-385 



Secondary Organic Aerosol (Soa) And Ozone Formation 

From Agricultural Pesticides 

Lindsay D. Yee, Bethany A. Warren, David R. Cocker III 



Abstract 

A large portion of California’s economy is based on agriculture, which depends on heavy use of 

pesticides to limit the amount of crop damage from insects, fungi, and unwanted weed growth. 

Pesticides are known to have direct health effects on humans. In this work, we investigate their 

potential to react within the atmosphere to form ozone and secondary organic aerosols (SOA). A 

series of photo-oxidation experiments were conducted within dual 90m 3 reactors of our environmental 

chamber. Individual pesticides were added to a surrogate volatile organic compound (VOC)/nitrogen 

oxides (NO x 

) mixture and were irradiated with a 200 kW Argon arc-lamp to study increased ozone 

formation and SOA production. The gas surrogate consists of ethene, propene, n-butane, trans-2- 

butene, toluene, octane, and m-xylene. The pesticide compounds tested include carbon disulfide 

(CS 2 

), kerosene, 1-3-dichloropropenes, S-ethyl N, N-di-n-propyl thiocarbamate (EPTC), and methyl 

isothiocyanate (MITC). Initial results show that some pesticides (i.e. EPTC, kerosene) increased 

SOA formation up to ten times over the base case surrogate mixture, while decreasing the ozone 

formation. Other pesticides (i.e. CS 2 

, MITC) increased the SOA formation by as much as twelve 

times in the surrogate mixture while increasing ozone levels. 


David Cocker 


Lindsay Yee joined our research group as a Research Advancement Program 

(RAP) scholar upon entering her freshman year at UC Riverside. Her academic 

and research skills are exceptional and she is a true pleasure to work with in the 

lab. Her undergraduate research has focused on estimating secondary organic 

aerosol formation (SOA), which is fine particle formed within the atmosphere from gaseous 

organic precursors. This current work reflects her investigation of the atmospheric chemistry of 

agricultural pesticides and their propensity to form fine particles – important from an air quality 

perspective as well as in the transport and environmental fate of such compounds. 

A U T H O R 

Lindsay D. Yee 

Environmental Engineering 

Lindsay Yee is a graduating senior in 

Environmental Engineering. Her research 

interests are in air quality, with an 

emphasis on secondary organic aerosols 

(SOA), a major contributor to fine 

particulate matter. She has conducted 

four years of undergraduate research 

investigating SOA formation using the 

environmental chamber at the College 

of Engineering Center for Environmental 

Research and Technology (CE-CERT). 

Her academic research interests and her 

outreach efforts through the Society of 

Women Engineers led to her selection as 

a National Science Foundation Graduate 

Research Fellow. Yee will expand on her 

research pursing a Ph.D. at the California 

Institute of Technology in fall of 2008. 


Secondary Organic Aerosol (Soa) And Ozone Formation From Agricultural Pesticides 


Introduction 

California Agriculture relies heavily on the use 

of chemical pesticides as insecticides, fungicides, and 

herbicides for crop protection and production. While 

pesticides have been studied for their adverse effects on 

human health, little has been studied regarding their reactivity 

in the atmosphere. Previous pesticide atmospheric studies 

conducted by Carter et al. have looked at the reactivity of 

pesticides like methyl bromide and chloropicrin . The goal 

of this case study was to look at the possible fate of a new 

group of selected pesticides in the atmosphere through 

secondary reactions. Simulated atmospheric reactions in the 

presence of a pesticide allowed for quantitative measure of 

the additional reactivity that each pesticide contributed to 

the system in terms of the additional ozone and secondary 

organic aerosol formed. 

Five pesticides of interest to the California 

Air Resources Board were tested in this work: 

1,3-dichloropropenes (trade name Telone ® ), Carbon 

disulfide (CS 2 

), S-Ethyl N,N-di-n-propyl thiocarbamate 

(EPTC), Kerosene, and Methyl Isothiocyanate (MITC). 

1,3-dichloropropenes are often planted with crops to fight 

nematodes 2 . CS 2 

is applied to nuts, apples, and other fresh 

fruit crops while EPTC is often applied to potatoes, corn, 

peas, and alfalfa as an herbicide for weed management. 

Kerosene is often applied as an oil pesticide for insecticide 

use on the almond, avocado, cotton, grape, lemon, and 

cauliflower crops 3 . MITC is another common soil fumigant, 

on the EPA’s list of high volume chemicals for being 

produced in over 1 million pounds per year 4 . 

When a reactive organic gas (ROG) undergoes photooxidation 

in the atmosphere reacting with an oxidizing agent 

like ozone, hydroxyl radical, or nitrate radical, a myriad 

of products are formed. Some with higher vapor pressure 

remain in the gas phase while others with lower vapor 

pressure condense and become secondary organic aerosol. 

Within the surrogate mixture used in the experiment, the 

ROGs are the aerosol forming aromatic hydrocarbons, 

m-xylene or toluene. 

Fine particulate matter, defined as particles with 

diameter less than 2.5 μm (PM 2.5 

), adversely affects human 

health 5, 6 ,decreases visibility 7, 8 , damages property, and affects 

global climate change 9 . SOA can contribute significantly 

to atmospheric PM 2.5 

. In fact, it was estimated that it can 

contribute up to 70% of fine particulate matter (PM 2.5 

) in 

urban air sheds 10 . Tropospheric ozone formation results from 

photochemical reactions of volatile organic compounds 

(VOCs) in the presence of nitrous oxides (NO x 

) 11 . High 

ozone levels at ground level also remain a human health 

concern as it causes many respiratory problems 12 . 

A surrogate gas mixture with known ozone and SOA 

formation potential can be used to study the effects of each 

selected pesticide when added to the system. A surrogate 

mixture was developed by Carter et. al to emulate urban 

atmospheric reactivity 15 . In the presence of a pesticide, changes 

in the chemical mechanisms and pathways are anticipated, 

resulting in final ozone and SOA concentrations different 

from the surrogate profile. While the scope of this work does 

not extend to the actual mechanisms behind the results, initial 

determination of a pesticide as an ozone or SOA enhancing or 

depressing agent under the experimental conditions was gained. 

Further parameters of changes in particle count and particle size 

were also determined to gain a greater understanding of each 

pesticide’s effects on SOA characteristics. 

Results 

Experimental results from five representative 

pesticide experiments are presented here. A summary of 

the results can be found in Table 1. 

Final ozone concentration in parts per billion by 

volume (ppb), mass volume of particles in micrograms per 

cubic meter (µg/m 3 ), count of particles, and diameter of 

particles were recorded. The final volume was calculated 

from the raw data collected by the SMPS, including a 

correction which accounts for particle losses on the walls 

of the reactors 15 . As shown in Table 1, each run utilizes 

one side (indicated as A or B) of the dual reactors strictly 

for the surrogate mixture, while the other side is used for 

the surrogate mixture with a pesticide added. For example, 

in Run 554, Side A only contained the surrogate mixture 

and Side B contained the same concentration of surrogate 

as Side A plus 103ppb of dichloropropenes. This method 

allowed for direct comparison of the surrogate versus 

surrogate plus pesticide results. 

Effects on Ozone Formation 

Figure 1 shows the ozone formation during 

experiment from the time the light source is turned on 




Run Light Pesticide 

Pesticide 

Added 

O3 Final (ppb) 

PM Final 

Volume 

(μg/m 3 ) 

PM Final 

Count 

Final Diameter 

(nm) 

554A Arc None 0 ppb 146 7.8 31700 60 

554B Arc Dichloropropenes 103 ppb 165 7.0 29000 62 

590A Arc None 0 ppb 140 6.9 25100 118 

590B Arc EPTC 250 ppb 129 51.0 70900 76 

597A BL None 0 ppb 118 6.8 21100 72 

597B BL CS2 630 ppb 133 32.7 63500 97 

599A BL MITC 990ppm 362 62.8 81100 112 

599B BL None 0 ppb 54 5.1 16500 72 

602A BL Kerosene 1.0 ppmC 99 50.1 16000 192 

602B BL None 0 ppmC 110 5.0 15300 79 

Table 1. Results of the experiment in terms of final ozone concentration, mass concentration of SOA formed, particle count, and particle 

size (by diameter) is recorded below for each Run. Under the “Light” column, Arc refers to the Argon arc lamp and BL refers to the black 

lights being used for that particular run. 

(Time = 0) to the time the light source is turned off for 

the surrogate mixture. Ozone formation profiles for the arc 

light (dashed trend lines) and black light (solid trend lines) 

surrogate experiments are established here. The difference 

in these profiles results from the differing light intensity 

of the Argon arc light and black light sources available to 

initiate the photochemistry and ozone formation. The ozone 

results of the arc light surrogate plus pesticide experiments 

are directly compared to the surrogate profile shown in 

Figure 2, revealing dichloropropenes as an ozone forming 

enhancer by about 20 ppb in Run 554, and EPTC as an 

ozone depressant by 11 ppb in Run 590. For the pesticides 

run under black light conditions (Figure 3), ozone formation 

decreased in the presence of kerosene by 11 ppb in Run 

602, while CS2 increased ozone by 15 ppb in Run 597. 

Most dramatic was the near seven fold increase of ozone 

formation in the presence of MITC from 54 ppb to 362 ppb 

in Run 599. The difference in ozone surrogate profiles by 

light source does not allow for direct comparison of all the 

pesticides to establish a relative order of ozone enhancing 

potential; however, each pesticide was identified as either 

an ozone enhancer or ozone depressant. 

Effects on SOA Formation 

Figure 5 (see page 72) presents the ozone and SOA 

formation for each pesticide run. SOA is shown as a mass 

concentration of SOA formation during experiments. Total 

aerosol mass formed was calculated from the final particle 

size distribution, after wall loss correction, and assuming 

unit density 13 . In the case of SOA formation, all surrogate 

only experiments had the same SOA profile regardless of 

Ozone (ppb) 

200 

180 

160 

140 

120 

100 

80 

60 

40 

20 

Ozone Formation from Surrogate by Light Source 

0 

0 100 200 300 400 500 

Time (min) 

1.1ppmC Surrogate Arc 

1.1ppmC surrogate BL 

Figure 1. Ozone formation profiles for the surrogate gas mixture 

under the Argon arc light source and Black light source. Each run 

has a surrogate ozone formation profile shown. Ozone formation 

potential is higher under the arc light for the surrogate gas 

mixture; this is due to differing intensities of the light sources. 




Ozone (ppb) 

200 

180 

160 

140 

120 

100 

80 

60 

40 

20 

Ozone Formation from Arc Light Experiments 

. 

0 

0 100 200 300 400 

Time (min) 

Surrogate + Dichloropropenes 


Surrogate + EPTC 

Ozone (ppb) 

Ozone Formation from Black Light Experiments 

200 

180 

160 

140 

120 

100 

80 

60 

40 

20 

Surrogate + MITC 

Surrogate + CS2 


Surrogate + Kerosene 

0 

0 100 200 300 400 500 

Time (min) 

Figure 2. Ozone formation in the presence of Dichloropropenes 

and EPTC compared to the 1.1 ppmC Surrogate only. 

Figure 3. Kerosene was the only pesticide run under black light 

conditions to result in lower ozone than the surrogate base case. 

light source used. This allows for a direct comparison of 

all pesticides in terms of their particle formation potential 

(Figure 4). 

Clearly, all the pesticides enhanced SOA formation 

except for dichloropropenes, where no significant difference 

from the surrogate was observed. CS 2 

caused an almost five 

times increase in SOA and the particle count tripled. The 

diameter of the particles also increased by 25 nm, suggesting 

that CS 2 

has increased the number of particles formed as 

well as their size (Table 1). Yet, while EPTC depressed 

ozone formation (Figure 5, see page 72), the SOA mass 

concentration was seven times larger and particle count was 

SOA Formation 

80 

Volume (mg/m 3 ) 

70 

60 

50 

40 

30 






1.1ppmC surrogate 

20 

10 

0 

0 50 100 150 200 250 300 350 400 

Time (min) 

Figure 4. The volume of SOA formed during experiment is shown for all runs for overall comparison. This comparison can be made 

because the SOA formed by the surrogate mixture alone is the same regardless of light source, as shown by the overlapping surrogate 

SOA profiles for all five runs. The graphical analysis allows for a predicted trend of lowest to highest SOA formation potential amongst 

the tested pesticides, determined as: dichloropropenes, CS 2 

, EPTC, kerosene, and MITC. 




almost tripled. In addition, diameter size of the particles in 

the presence of EPTC was smaller (Table 1). It is possible 

that a large nucleation burst at the onset of particle formation 

led to the lower final diameter achieved in the presence 

of EPTC. Kerosene, on the other hand, while decreasing 

ozone slightly has ten times the SOA mass concentration 

(Figure 5, see page 72) while maintaining a close particle 

count but with increased diameter size (Table 1). MITC had 

very dramatic and direct impacts on the surrogate system, 

putting out ozone concentrations seven times higher and an 

aerosol mass concentration twelve times higher (Table 1, 

Figure 5). In the presence of MITC particle count was a 

staggering 81,100 cm -3 compared to 16,500 cm -3 and they 

were significantly larger too (Table 1). In addition, SOA 

formed within 50 minutes of the experiment, compared to 

around 100 minutes for the surrogate base as seen in the 

MITC SOA plot in Figure 5 (see page 72). MITC could be 

initiating SOA formation pathways earlier with nucleation 

and/or even serving as a reactive organic gas to directly react 

and contribute to the SOA yield. It could also be affecting 

kinetics of the reaction. 

Discussion 

These preliminary results suggest that many of these 

agricultural pesticides impact the atmospheric chemistry 

that would normally occur from just the surrogate base. 

Reaction mechanisms and pathways could be altered, 

bypassed, or changed completely in the presence of a 

pesticide. Reaction kinetics and timing of nucleation could 

be affected, as in the case of MITC, EPTC, and CS 2 

. The 

number and physical properties of the particles formed 

were also changed. All of these properties are related to 

their atmospheric transport. Moreover, there is no direct 

correlation or predictor that a pesticide affects ozone 

formation in the same direction it does for SOA formation. 

A pesticide independently impacts the yields of ozone and 

particulate matter formed. Yet a trend of lowest to highest 

SOA forming potential was proposed. The results of this 

study, in addition to the quantified maximum incremental 

reactivity as determined by Carter and Malkina 14 for these 

pesticides, could serve as motivation for new limitations 

on the use of certain pesticides in order to meet federal 

and state air quality standards. Future work would include 

further investigation into the reasons behind these trends. 

Further experimental detail on all the pesticide experiments 

completed can be found in Carter and Malkina 14 . 


UC Riverside/CE-CERT Environmental Chamber 

The experiments were performed using the UC 

Riverside/CE-CERT Environmental Chamber, consisting of 

dual 90m 3 reactors made from 2 mil FEP Teflon ® film. Pure air 

from an AADCO ® air purification system (NO x 



Ozone (ppb) 

180 

160 

140 

120 

100 

80 

60 

40 

20 

0 

Ozone from Dichloropropenes Run 554 



0 100 200 300 400 

Time (min) 

Ozone from CS2 Run 597 


10 

9 

8 

7 

6 

5 

4 

3 

2 

1 

0 

SOA from Dichloropropenes Run 554 


1.1 ppmC surrogate 

0 100 200 300 400 

Time (min) 

SOA from CS2 Run 597 

Ozone (ppb) 

140 

120 

100 

80 

60 

40 


20 

1.1 ppmC Surrogate BL 

0 

0 100 200 300 400 500 

Time (min) 


80 

70 


60 


50 

40 

30 

20 

10 

0 

0 100 200 300 400 

Time (min) 

Ozone from EPTC Run 590 

SOA from EPTC Run 590 

160 

140 

120 

80 

70 

60 



Ozone (ppb) 

100 

80 

60 

40 


20 

1.1 ppmC surrogate Arc 

0 

0 50 100 150 200 250 300 350 400 

Time (min) 


50 

40 

30 

20 

10 

0 

0 100 200 300 400 

Time (min) 

Ozone from Kerosene Run 602 

SOA from Kerosene Run 602 

120 

80 

Ozone (ppb) 

100 

80 

60 

40 



70 

60 

50 

40 

30 

20 



20 

1.1 ppmC surrogate BL 

10 

0 

0 50 100 150 200 250 300 350 400 

Time (min) 

0 

0 50 100 150 200 250 300 350 400 

Time (min) 

Figure 5. All pesticides were plotted over time in terms of their ozone formation and the volume concentration of SOA formed. Comparison 

of the ozone formation and SOA formation plots for each pesticide show that each pesticide independently affects ozone and SOA 

formation. For example, while kerosene depressed ozone formation, there was a dramatic increase in SOA formed in its presence. 




Ozone (ppb) 

Ozone from MITC Run 599 

200 

180 

160 

140 

120 

100 

80 

60 


40 

20 


0 

0 100 200 300 400 500 

Time (min) 


80 

70 

60 

50 

40 

30 

20 

10 

0 


1.1ppmC surrogate 

SOA from MITC Run 599 

0 50 100 150 200 250 300 350 400 

Time (min) 

from a TSI model 88 Kr neutralizer, a Dynamic Mobility 

Analyzer (DMA) TSI 3081 long column, and a TSI Model 

3760 Condensation Particle Counter (CPC). The SMPS 

was located within the chamber enclosure and pulled three 

samples per side at a 60 second scan rate. The column was 

ramped from -40 to -7000 V to monitor particle diameters 

from 28-730 nm. The DMA air flows consisted of 2.5 L 

min -1 for sheath and excess flows and 0.25 L min -1 for 

aerosol and monodisperse flows. 

Use of the Surrogate 

A case study of the pesticide effects on the 25ppb 

NO x 

(1/3 NO 2 

, 2/3 NO), 1.1ppmC surrogate runs was 

performed. The surrogate mixture contained ethene, 

propene, n-butane, trans-2-butene, toluene, octane, and 

m-xylene. Experiments were run by injecting surrogate in 

both reactors and then injecting atmospherically relevant 

concentrations of pesticide into only one reactor. This 

allowed for a direct comparison of the SOA and ozone 

formation profiles for the surrogate-only reactor to the 

surrogate plus pesticide reactor. 


We gratefully acknowledge funding from National 

Science Foundation CAREER 0449778 and California 

Air Resources Board Contract No. 04-334 for support 

of this project. We thank Kurt Bumiller, Irina Malkina, 

and William P.L. Carter for their knowledge, design, and 

expertise in conducting these experiments. 

References 

1. Carter, W. P. L., D. Luo and I. L. Malkina (1997b): 

Investigation of that Atmospheric Reactions of 

Chloropicrin,” Atmos. Environ. 31, 1425-1439. 

2. EPA. R.E.D. Facts 1,3-Dichloropropene; 

EPA/787/F/98/014; Environmental Protection 

Agency: Washington, DC, 1998. 

3. Department of Pesticide Regulation. 

Summary of Pesticide Use Report Data 

2005 Indexed by Chemical; California 

Environmental Protection Agency: Sacramento, 

CA, 2006. 

4. Scorecard The Pollution Information Site. Chemical 

Profile for METHYL ISOTHIOCYANATE (CAS 

Number: 556-61-6), http://www.scorecard.org. 

5. Schwartz, J.; Dockery, D.W.; Neas, L.M.; et al. Is daily 

mortality associated specifically with fine particles? J. 

Air Waste Manage. Assoc. 1996, 46, 927-39. 

6. EPA. Air Quality Criteria for Particulate Matter; 

EPA/600/P-95/001cF; Environmental Protection 

Agency: Washington, DC, 1996. 

7. Eldering, A.; Cass, G.R. Source-oriented model for air 

pollutant effects on visibility. J. Geophy. Res. 1996, 

101, 19343-19369. 

8. Larson, S. M; Cass, G.R. Characteristics of summer 

midday low-visibility events in the Los Angeles area. 

Environ. Sci. Technol. 1989, 23 (3), 281-289. 

9. Pilinis, C.; Pandis, S.; Seinfeld, J.H. Sensitivity of 

direct climate forcing by atmospheric aerosols to 

aerosol size and composition. J. Geophys. Res. 1995, 

100, 18739-18754. 




10. Turpin, B.J.; Huntzicker, J.J. Secondary formation 

of organic aerosol in the Los-Angeles basin—a 

descriptive analysis of organic and elemental carbon 

concentrations. Atmos. Environ. 1991, 25A (2), 

207-215. 

11. Seinfeld, J.H.; Pandis, S.N.; Atmospheric Chemistry 

and Physics: from Air Pollution to Climate Change; J. 

Wiley: Hoboken, N.J., 2006; 2nd ed. 

12. Environmental Protection Agency. Ground-level 

ozone: Health and the Environment, http://www.epa. 

gov/air/ozonepollution/health.html 

13. Forstner, H.J.L.; Flagan, R.C.; Seinfeld, J.H. Secondary 

organic aerosol from the photooxidation of aromatic 

hydrocarbons: molecular composition. Environ. Sci. 

Technol. 1997, 31, 1345-1358. 

14. Carter, W. P. L.; Malkina, I. L. Investigation of 

Atmospheric Ozone Impacts of Selected Pesticides; 

Final Report to the California Air Resources Board; 

Contract No. 04-334; 2007. 

15. Carter, W. P. L.; Cocker, D. R., III; Fitz, D. R.; 

Malkina, I. L.;Bumiller, K.; Sauer, C. G.; Pisano, J. T.; 

Bufalino, C.; Song, C. A new environmental chamber 

for evaluation of gas- phase chemical mechanisms and 

secondary aerosol formation. Atmos. Environ. 2005, 

39, 7768-7788. 


Bacterium-Induced Fluorescence-Enhancement Kinetics: 

Breaking 100-Year Old Traditions of Staining Bioanalyses 

Elizabeth Zielins, Valentine I. Vullev 

Graduate Student Assistant: Marlon Thomas 



Abstract 

The virulence and increasing antibiotic resistance of certain bacterial strains creates a need 

for efficient and timely detection of environmental pathogens. We evaluate the kinetics of the 

fluorescence enhancement of cationic dyes as an assay for differentiation between bacterial species. 

For several benzothiazole cationic dyes, such as 3-3’-diethylthiacyanine, we observed fluorescence 

enhancement in the presence of vegetative bacteria and bacterial spores. Different bacterial species 

manifested different rates of emission enhancement. Although staining, particularly fluorescence 

staining, has been a broadly used technique for the identification of bacterial species, the kinetics 

of the staining process has not been examined in detail. Analyses of the kinetics of emission 

enhancement for a series of fluorophores in the presence of one species of bacteria (or spore) can 

be used to create a set of kinetic parameters specific to that bacterial type. We hypothesized that 

these kinetic parameters can be utilized as “fingerprints” for detection and identification of bacterial 

species. We used three different vegetative bacteria and three different bacterial spores as model 

organisms for the collection of preliminary data that demonstrated the feasibility of our hypothesis. 

Conducting kinetic emission assays with various concentrations of bacteria and fluorophores 

allowed us to determine the first order time constants of the kinetics of emission enhancement. 

These time constants reflect the migration of the dye from the surrounding media to the fluorogenic 

microenvironment within the bacterial cell wall. Furthermore, we observed that the time constants 

were concentration independent and species specific. 

A U T H O R 

Elizabeth Zielins 

Bioengineering 

Elizabeth Zielins is a third year student 

majoring in Bioengineering. At UCR, she 

has discovered a fascination with how 

the physical and mathematical sciences 

can be applied towards the study 

and manipulation of ever-changing biological 

systems. Her ultimate goal is to 

combine a career in medicine with clinical 

research in bioengineering. Specifically, 

she plans to specialize in pediatric 

reconstructive plastic surgery and pursue 

research in tissue engineering and 

regenerative medicine. 


Valentine I. Vullev 

Department of Engineering 

The research in my laboratory is cross-disciplinary and involves students with 

different backgrounds and interests. Utilizing our expertise in spectroscopy for 

addressing fundamental issues and developing tools for microbiology resulted in 

the project described in this particular publication. Due to her experience and strong 

background in cell and microbiology, Elizabeth Zielins was a perfect fit for this project. Under the close 

supervision and mentoring from Marlon S. Thomas (a second-year graduate student and coauthor 

of this paper) and Duoduo Bao (a first year-graduate student), Elizabeth expediently expanded her 

skills and knowledge into the areas of spectroscopy and photophysics. Her talent, intelligence and 

perseverance, indeed, helped Elizabeth rise to the occasion. With Marlon Thomas, Elizabeth Zielins 

formed a “power team” that brought this project to its current stage in about six months. 


Bacterium-Induced Fluorescence-Enhancement Kinetics: Breaking 100-Year Old Traditions of Staining Bioanalyses 


Introduction 

This paper describes kinetic studies on the 

fluorescence enhancement of 3,3’-diethylthiacyanine 

(THIA) in the presence of six different bacterial species. 

Currently, one quarter of human deaths worldwide 

are caused by bacterial infections. In the United States in 

the early 2000s, 33 million illnesses were caused by foodborne 

bacteria alone. Of these illnesses, 10,000 were fatal 1,2 . 

In the modern healthcare setting, it can take up to 72 hours 

to correctly identify an unknown microbial pathogen. 3-5 In 

the meantime, patients are often treated with inappropriate 

antibiotics. 5 Besides having a neutral effect on the 

patient’s health, such treatments may lead to the buildup 

of antibiotic-resistant bacterial strains. For example, 

the percentage of healthcare-associated Staphylococcus 

aureus (staph) infections caused by Methicillin-resistant 

Staphylylococcus aureus (MRSA) has risen from 2% in 

1974, to 64% in 2008. 1,2,6-8 

In the modern healthcare setting, techniques 

used to identify unknown bacterial strains range from 

biochemically-based methods such as real-time PCR, 

DNA sequencing, and immunostaining, to methods in 

cell staining. 9,10 While cell staining is perhaps the most 

economical and likely to produce expedient results, current 

methods in cell staining (including fluorescence assays) are 

inherently limited: (1) they yield only a Boolean outcome; 

and (2) they detect only species that are sought for. 3,11 

Ever since the development of the Gram stains 

(more than 100 years ago), 12,13 the identification of bacterial 

species using staining techniques has been based solely on 

the initial and final appearance of the cells (i.e., before and 

after the staining process). Hence, the staining analyses 

produce only Boolean outcomes: i.e., the reagents either 

stain (positive) or do not stain (negative) the analyzed 

bacteria. Gram staining may identify a bacterial species 

as gram positive. Further testing, however, is required to 

determine which gram positive species the sample belongs 

to. The general (Gram) staining tests dictate the types of 

further analyses (with increased specificity) which can 

be used for identification of the bacterial species. Due to 

insufficient specificity of the initial staining tests, however, 

the choices for the set of analyses strongly depend on the 

pathologist’s intuition. As a result, only species for which 

one is looking are finally identified, leaving key determining 

factors of the diagnosis quite susceptible to human error. 

Herein we present the development of cost-efficient 

assays for expedient analysis of bacterial samples, utilizing 

the kinetics of fluorescence enhancement upon staining. 

Certain cationic dyes manifest enhanced fluorescence upon 

binding to bacterial spores or vegetative bacteria. 14 Our 

findings revealed that for six different bacterial species, 

each had different kinetic signatures. Moreover, with the 

method we describe, the presence of unknown species 

can be detected even if their kinetic signatures are not 

associated with any of the previously investigated bacteria 

(whose kinetics have been characterized). 

Such specificity can be achieved due to the fact 

that the kinetics of emission enhancement reflects the 

migration of dye molecules from the aqueous solvent to 

the fluorogenic microenvironment within the bacterial cell 

walls. The increased fluorescence of the bacterium-bound 

dye could be due to either the polarity or the viscosity of the 

microenvironment. Our photophysical studies indicated that 

the observed emission enhancement is a result of migration 

of the dye from the relatively non-viscous aqueous media 

to the viscous environment of the cell wall. 

Results and Discussion 

A symmetric cyanine dye, 3,3’-diethylthiacyanine 

(THIA), manifests orders of magnitude increase in its 

fluorescence when in the presence of bacterial spores 

or vegetative bacteria (Figure 1). At the same time, the 

presence of bacteria does not cause wavelength shift in the 

Figure 1. Emission spectra of THIA (6.43µM in 2 mM Tris 

buffer, pH = 8.5) in the presence and absence of B. subtilis 

(~10 7 cells/mL). λ ex 

= 420 nm. 




solvent Φ f 

× 10 3 ε η 0 

/ cP 

water 0.975 81 0.89 

70% methanol 0.821 40 1.3 

glycerol 141 43 930 

methanol 0.515 33 0.55 

iso-propanol 0.634 20 2.3 

Table 1. Fluorescence quantum yield, Φf (λ 

e x 

= 420 nm), of THIA 

in solvents with different polarities (relative dielectric constants, 

ε) and viscosities, η0. 

a 

absorption maxima of THIA (data not shown), indicating 

that ground-state phenomena are not responsible for the 

observed fluorescence enhancement through alterations in 

A (eq. 1). 

We investigated the dependence of the fluorescence 

quantum yield, Φ f 

, of THIA in solvents with various 

polarities and viscosities (Table 1). The quantum yield did 

not exhibit dependence on the dielectric properties of the 

solvents: the correlation coefficient for Φ vs. ε was -7.5 x 

10 -3 (Figure 2a). The correlation coefficient r reflects the 

interdependence of the two quantities. A strong correlation 

(dependence) results in a value of r close to 1 or -1. A lack 

of correlation results in r having a value close to zero. 

We did, however, observe a three orders of magnitude 

increase in Φ f 

for THIA in glycerol compared with other 

solvents that have more than 100 times smaller viscosities. 

Indeed, the correlation coefficient for Φ f 

vs. η was close to 

unity (Figure 2b), indicating a strong dependence of the 

fluorescence properties of THIA on the viscosity of the 

surrounding media. 

From these findings, we can therefore conclude 

that the reason for the observed fluorescence enhancement 

of THIA in the presence of bacterial species is due to an 

increase in the viscosity of the microenvironment of the 

dye. As the free dye in solution binds to the bacteria, it 

migrates from the relatively non-viscous aqueous media 

to the gel-like microenvironment of the cell wall—which 

has a significantly higher viscosity (Scheme 1). The 

fluorogenic effect of the viscous microenvironment could 

be due to suppression of the non-radiative decay processes 

resultant from molecular vibrational and rotational 

b 

Figure 2. Linear correlation between the fluorescence quantum 

yield, Φ f 

, of THIA and (a) dielectric constant, ε; and (b) the 

viscosity, η, of the solvent media, with the corresponding 

correlation coefficients. The distortion of the linear fit is a result 

from the logarithmic representation, introduced for convenient 

visualization of the spread among the data points. 

Scheme 1. Interaction between THIA (structure shown) and a 

bacterial cell causing the emission enhancement. 




a 

b 

Figure 3. Fluorescence kinetics curves (λ ex 

= 420 nm, λ em 

= 475 nm) for: (a) 2 mM Tris buffer (pH = 8.5) and B. sphaericus (106 

cells/mL in Tris buffer); and (b) THIA (6.43µM in Tris buffer) and THIA (6.43µM) in the presence of B. sphaericus (106 cells/mL) in 

Tris buffer. 

a 

Figure 4. Fluorescence kinetics curves (λ ex 

= 420 nm, λ em 

= 475 nm) 

for THIA (6.43µM in 2 mM Tris buffer, pH = 8.5): (a) in the presence 

of two Gram positive and one gram negative vegetative bacteria; 

and (b) in the presence of three different bacterial spores. 

motion. The structure of THIA shows flexible bonds 

within the π-conjugation between the two ring systems 

(Scheme 1). Binding to a rigid microenvironment impedes 

molecular motions normally allowed by bond flexibility. 

Consequentially, radiative (fluorescence) processes become 

the predominant pathways for decay of the lowest singlet 

excited state. 

A principal goal of our research was to investigate the 

dynamics of the fluorescence enhancement of THIA induced 

by bacterial species. The addition of bacteria to a solution of 

b 

Table 2. Time constants, t, for the evolution of the fluorescence 

enhancement of THIA (three different concentrations) in the 

presence of three vegetative bacteria and three bacterial spores 

(with different cell densities). Dashes indicate low signal-to-noise 

ratio for reliable data fits. 

THIA produces a fluorescence increase in the time domain 

of seconds and minutes (Figure 3). Monoexponential fits of 

the kinetics data allowed us to extract the time constants, τ, 

of the emission-enhancement processes: 

Here, F is the measured fluorescence intensity over 

time, t; F 0 

is the initial fluorescence intensity of the dye 

without bacteria; F ∞ 

is the increase in the fluorescence 

(2) 




intensity (due to the bacteria); and t 0 

is the time of injection 

of the bacteria into the dye solution. 

We investigated the interaction of THIA with six 

bacterial species: three vegetative bacteria (B. sphaericus, 

B. subtilis, and E. coli) and three bacterial spores (B. 

globigii, B. pimulis, and B. thuringiensis). Each vegetative 

bacterium species exhibited unique emission-enhancement 

time constants, regardless of its Gram-stain classification 

(Figure 4a). Bacterial spores manifested similar segregation 

in the values of their emission enhancement time constants 

(Figure 4b). 

Furthermore, the measured time constants did not 

show significant dependence on the dye concentration 

or cell density of the samples (Table 2), though the 

characteristics of fluorescence enhancement appear to be 

most discernable when 6.43µM THIA is used. It is also 

clear that, aside from limiting whether or not the data can 

be fit with a kinetic function, bacterial concentration does 

not significantly affect the values of the time constants. 

An inspection of Figure 4 and Table 2 shows that the 

time constants for each vegetative bacterium species fall within 

distinct ranges: i.e., for E. coli, 15 < τ < 20 s; for B. subtilis, 

5 < τ < 11 s; for B. sphaericus, 30 < τ < 40 s. We ascribe 

the observed differences in the time constants to the different 

composition of the cell walls of the vegetative bacteria. 

Although the variations in the compositions of the 

spore coats for different species are not truly substantial, 17-19 

the values of the time constants for the bacterial spores still 

fell within different regions: i.e., for B. globigii, 75 < τ < 85 

s; for B. pimulis, 15 < τ < 20 s; and for B. thuringiensis 10 

< τ < 15 s. It should be emphasized that the time constant 

values for B. pimulis and E. coli appear to be within the 

same region. This overlap, however, is most probably 

coincidental because there is no similarity between the E. 

coli cell wall and the B. pimulis spore coat. 17-24 

An ANOVA analysis (two-factor with replication) was 

performed on the data in order to test whether the measured 

time constants have a significant dependence on the cell 

densities of the bacterial species, and the concentrations of 

dye. The first “null” hypothesis states that the time constants 

for the different bacterial species are the same, while the 

second “null” hypothesis states that the time constants 

for the different dye concentrations are the same. For our 

tests, we set a confidence level of 95% or p=0.050. The 

ANOVA analysis generated three p-values: the first and the 

second p-values are for the first and second null hypotheses, 

respectively. The third p-value represents the interaction 

between the first two parameters. The p-value for the first 

null hypothesis was 5.66 x10 -46 , which is considerably 

smaller than p = 0.050, indicating that the time constants 

for different bacteria are unique. The p-value for the second 

null hypothesis was 0.12, which is greater than p=0.050, 

suggesting that the time constants do not significantly 

depend on the dye concentration. The interaction between 

the two parameters, bacterial species and dye concentration, 

however, gives a p-value of 0.0020. This is an additional 

indication of the uniqueness of the time constants for each 

bacterial species, at a given dye concentration. 

Conclusions 

We demonstrated that the kinetics of fluorescence 

staining with a cyanine dye, THIA, is species-specific 

and does not show concentration dependence. We believe 

that the kinetics of the staining processes will produce 

“fingerprint” features for detection and identification of 

bacterial species. Our findings on the media-dependence 

of THIA fluorescence indicate that viscosity-sensitive 

dyes which stain bacteria will allow for the development 

and expansion of emission-enhancement methodology for 

detection and identification of microorganisms. 


A solution of 643µM THIA was prepared by 

dissolving solid THIA in a 70% ethanol solution. The 

solutions for the bacterial analyses were prepared via 10x, 

100x, and 1000x dilutions of the stock dye solution in 2mM 

Tris buffer (pH 8.5). 

Bacterial cultures of two gram positive species, 

Bacillus subtilis and Bacillus sphaericus, and one gram 

negative species, Escherichia coli, were prepared on solid, 

Luria broth agar media. Colonies from these cultures were 

transferred to liquid media and allowed to grow for no 

longer than 24 hours. When not in use, both liquid and solid 

media stock solutions were stored in a -4°C refrigerator. For 

use in experimental measurements, fractions of the liquid 

cultures were centrifuged, washed twice, and re-suspended 

in 2mM Tris buffer solution. 

Bacterial spores Bacillus globigii, Bacillus pimulis, 




and Bacillus thuringiensis, were donated by the U.S. Army 

Research Laboratory and treated as we have previously 

described. 15 For stock solutions, bacterial spores were 

suspended in 2mM aqueous Tris buffer, stored at 4°C and 

used within 24 hours (for prevention of germination). Prior 

to making spectroscopic measurements, 2µL TWEEN 

40 was added to each stock solution in order to reduce 

the aggregation of the spores (in order to increase the 

homogeneity of the solution). 

In order to determine the cell and spore density of the 

suspensions of vegetative bacteria and bacterial endospores, 

cell counts were performed using a hemocytometer at 40x 

magnification (transmission optical microscope). 

Fluorescence cuvettes (four polished sides) were 

filled with 3mL of aqueous THIA solution (64.3µM, 6.43µM, 

and 643nM in Tris buffer) for absorption and emission 

measurements. 3µL, 30µL, or 300µL of bacterial suspensions 

in Tris buffer were added to the 3mL of dye solution in each 

cuvette. Due to differences in the bacteria concentrations 

of the liquid culture stock solutions, the final cell densities 

of the dye-bacteria solutions ranged from 10 5 to 10 8 cells/ 

mL. For each bacterial species, nine measurements were 

conducted: i.e., the three amounts of bacterial suspensions 

were added to each dye concentration. 

The absorption spectra were recorded using a UV/Vis 

spectrophotometer (Varian Cary 50 Bio), at a wavelength 

range between 350 and 750nm. All measurements were 

taken in 1cm plastic cuvettes. Absorption measurements 

were collected of all dye-bacteria solutions and blank (no 

bacteria) dye solutions. 

The excitation and emission spectra, as well as 

the emission-enhancement kinetics, were recorded with 

a fluorescence spectrophotometer (Fluorolog 3-22). For 

the emission spectra and the kinetics measurements, the 

excitation wavelength was set near the absorption maximum 

of the dye, λ ex 

= 420nm. Measurements of fluorescence 

intensity at the spectral maximum were collected over tenminute 

periods for all samples. Control experiments with 

samples containing pure buffer, only THIA in buffer, and 

only bacteria in buffer were also conducted, allowing us 

to establish the baseline fluorescence (autofluorescence) 

of the Tris buffer solution and of the bacteria solutions 

in absence of dye. About 50 seconds into the ten-minute 

measurement period, 3, 30, or 300µL of bacterial 

suspension was added. In order to eliminate fluctuations 

in fluorescence intensity due to diffusion of the bacteria 

through the dye solution, the samples were continuously 

stirred during the measurements. The samples were also 

kept at about 37°C in order to prolong the viability of 

the bacteria. Emission, excitation, and absorption spectra 

of the dye-bacteria solutions were taken before and after 

the kinetics measurements. The kinetic measurements 

were repeated twice (with one-to-two month separations 

between the measurements). 

From the fluorescence and absorption data, the 

quantum yield, Φ, of THIA for different solvents (glycerol, 

methanol, 70% methanol, isopropanol, and water) was 

estimated: 

where S is the areas under the fluorescence spectra; A is 

the absorption at the excitation wavelength; and n is the 

medium index of refraction. The subscript “0” indicates the 

quantities for the fluorescence standard, coumarin 151 in 

ethanol, Φ 0 

= 0.49. 16 

Quantification and analysis of the spectroscopic 

and quantum yield data was performed using IGOR and 

Microsoft Excel software. 


Funding for this work was provided by the Nathan 

and Violet David Foundation, the Howard Hughes Medical 

Institute, the UC Riverside Medical Scholars Program, 

and the U.S. Department of Education. We also extend our 

gratitude to Ms. Duoduo Bao for the productive discussions 

and support. 

(1) 




References 

1. J. K. Clark CD, Jones JL, et al. , “ Incidence of hand 

infections and their bacterial flora,” Presented at 

the American Academy of Orthopaedic Surgeons 

74th Annual Meeting. Feb. 14-18, 2007. San Diego. 

(2007). 

2. W. F. Duke, M. C. Robson and T. J. Krizek, “Civilian 

wounds, their bacterial flora and rate of infection,” 

Surgical forum 23, 518-520 (1972). 

3. E. A. Mothershed and A. M. Whitney, “Nucleic acidbased 

methods for the detection of bacterial pathogens: 

Present and future considerations for the clinical 

laboratory,” Clinica Chimica Acta 363, 206-220 

(2006). 

4. D. V. Lim, J. M. Simpson, E. A. Kearns and M. F. 

Kramer, “Current and developing technologies for 

monitoring agents of bioterrorism and biowarfare,” 

Clinical Microbiology Reviews 18, 583-607 (2005). 

5. S. Owens, “Killing the right bug. Phenotypic and 

genotypic approaches for identifying infectious 

agents and antibiotic resistance can help to slow the 

increasing antibiotic resistance among infectious 

bacteria,” EMBO reports 2, 757-760 (2001). 

6. R. M. Klevens, R. Edwards Jonathan, C. Tenover 

Fred, L. C. McDonald, T. Horan and R. Gaynes, 

“Changes in the epidemiology of methicillin-resistant 

Staphylococcus aureus in intensive care units in US 

hospitals, 1992-2003,” Clinical infectious diseases 

: an official publication of the Infectious Diseases 

Society of America 42, 389-391 (2006). 

7. M. Burd, H. Humphreys, G. Glynn, E. Mitchell, P. 

McDonald, H. Johnson, B. McDonnell, D. Doyle and 

A. Rossney, “Control and the prevention of methicillinresistant 

Staphylococcus aureus in hospitals in Ireland: 

North/South Study of MRSA in Ireland 1999,” The 

Journal of hospital infection 53, 297-303 (2003). 

8. A. Tambic, “[Methicillin-resistant Staphylococcus 

aureus (MRSA), a predictor of the end of the 

antibiotic era – diagnosis, epidemiology, therapy and 

dissemination prevention],” Lijecnicki vjesnik 119, 

166-171 (1997). 

9. J. C. Cheng, C. L. Huang, C. C. Lin, C. C. Chen, Y. C. 

Chang, S. S. Chang and C. P. Tseng, “Rapid detection 

and identification of clinically important bacteria by 

high-resolution melting analysis after broad-range 

ribosomal RNA real-time PCR,” Clinical chemistry 

52, 1997-2004 (2006). 

10. N. Kobayashi, T. W. Bauer, H. Sakai, D. Togawa, 

I. H. Lieberman, T. Fujishiro and G. W. Procop, 

“The use of newly developed real-time PCR for the 

rapid identification of bacteria in culture-negative 

osteomyelitis,” Joint Bone Spine 73, 745-747 (2006). 

11. M. S. Veena and J. W. L. van Vuurde, “Indirect 

immunofluorescence colony staining method for 

detecting bacterial pathogens of tomato,” Journal of 

microbiological methods 49, 11-17 (2002). 

12. M. L. Kaplan and L. Kaplan, “The Gram Stain and 

Differential Staining,” J Bacteriol 25, 309-321 

(1933). 

13. C. Stahl and E. Olsen, “Gram-negative staining of 

gram-positive intestinal bacteria with particular 

reference to examinations of faeces in infants,” Acta 

paediatrica 39, 372-380 (1950). 

14. G. Jones, II and P. Landsman, “New detection targets 

for amyloid-reactive probes: spectroscopic recognition 

of bacterial spores,” Proceedings of SPIE-The 

International Society for Optical Engineering 5795, 

1-7 (2005). 

15. V. I. Vullev, J. Wan, V. Heinrich, P. Landsman, 

P. E. Bower, B. Xia, B. Millare and G. Jones, II, 

“Nonlithographic Fabrication of Microfluidic 

Devices,” Journal of the American Chemical Society 

128, 16062-16072 (2006). 

16. S. Nad and H. Pal, “Unusual Photophysical Properties 

of Coumarin-151,” Journal of Physical Chemistry A 

105, 1097-1106 (2001). 

17. A. Driks, “Proteins of the spore core and coat,” Bacillus 

subtilis and Its Closest Relatives, 527-535 (2002). 

18. P. Setlow, “Resistance of bacterial spores,” Bacterial 

Stress Responses, 217-230 (2000). 

19. A. D. Russell, “Bacterial resistance to disinfectants: 

present knowledge and future problems,” The Journal 

of hospital infection 43 Suppl, S57-68 (1999). 




20. H. Frank and D. Dekegel, “Electron microscopical 

studies on the localization of different components 

of cell walls of gram-negative bacteria,” Folia 

Morphologica 12, 227-233 (1967). 

21. E. V. Anisimova, A. O. Badyakina, N. V. Vasil’eva and 

M. A. Nesmeyanova, “Changes in the composition of 

anionic membrane phospholipids influence protein 

secretion and cell envelope biogenesis in Escherichia 

coli,” Microbiology 74, 147-152 (2005). 

22. S.-Y. Li, J.-V. Holtje and K. D. Young, “Comparison 

of high-performance liquid chromatography and 

fluorophore-assisted carbohydrate electrophoresis 

methods for analyzing peptidoglycan composition of 

Escherichia coli,” Analytical Biochemistry 326, 1-12 

(2004). 

23. E. Painbeni, M. Caroff and J. Rouviere-Yaniv, 

“Alterations of the outer membrane composition in 

Escherichia coli lacking the histone-like protein HU,” 

Proceedings of the National Academy of Sciences of 

the United States of America 94, 6712-6717 (1997). 

24. S. Morein, A.-S. Andersson, L. Rilfors and G. 

Lindblom, “Wild-type Escherichia coli cells regulate 

the membrane lipid composition in a “Window” 

between gel and non-lamellar structures,” Journal of 

Biological Chemistry 271, 6801-6809 (1996).

UC Riverside Undergraduate Research Journal

Create successful ePaper yourself

Delete template?

Save as template?