et al.
et al.
et al.
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
RESEARCH ARTICLES<br />
790<br />
C-termin<strong>al</strong> fragment (residues 213 to 522), bound<br />
to ISD (Fig. 5C). A longer C-termin<strong>al</strong> fragment<br />
containing residues 161 to 522 did bind to ISD,<br />
which suggests that the sequence 161 to 212 may<br />
be important for DNA binding. However, del<strong>et</strong>ion<br />
of residues 161 to 212 from h-cGAS did not<br />
impair ISD binding, which suggests that cGAS<br />
contains another DNA binding domain at the N<br />
terminus. Indeed, the N-termin<strong>al</strong> fragment containing<br />
residues 1 to 160 <strong>al</strong>so bound ISD (Fig.<br />
5C). Thus, cGAS may contain two separate DNA<br />
binding domains at the N terminus. Our attempts<br />
to express the cGAS fragment containing residues<br />
161 to 212 in E. coli or HEK293T cells have not<br />
been successful, so at present we do not know<br />
wh<strong>et</strong>her this sequence <strong>al</strong>one is sufficient to bind<br />
DNA. Non<strong>et</strong>heless, it is clear that the N terminus<br />
of h-cGAS containing residues 1 to 212 is both<br />
necessary and sufficient to bind DNA.<br />
Different del<strong>et</strong>ion mutants of h-cGAS were<br />
overexpressed in HEK293T-STING cells to d<strong>et</strong>ermine<br />
their ability to activate IRF3 and induce<br />
IFN-b and the cytokine tumor necrosis factor–a<br />
(TNF-a) (Fig. 5C and fig. S6A). The protein fragment<br />
containing residues 1 to 382, which lacks<br />
the C-termin<strong>al</strong> 140 residues including much of<br />
the Mab21 domain, failed to induce IFN-b (Fig.<br />
5C,right)orTNF-a or to activate IRF3 (fig. S6A),<br />
which suggests that an intact Mab21 domain is<br />
important for cGAS function. As expected, del<strong>et</strong>ion<br />
of the N-termin<strong>al</strong> 212 residues (residues<br />
213 to 522), which include part of the NTase<br />
domain, abolished the cGAS activity (Fig. 5C<br />
and fig. S6A). An intern<strong>al</strong> del<strong>et</strong>ion of just four<br />
amino acids (Lys 171 ,Leu 172 ,Lys 173 ,andLeu 174 )<br />
within the first helix of the NTase fold preceding<br />
the cat<strong>al</strong>ytic residues <strong>al</strong>so destroyed the cGAS<br />
activity (fig. S6A).<br />
Del<strong>et</strong>ion of the N-termin<strong>al</strong> 160 residues did<br />
not affect IRF3 activation or cytokine induction<br />
by cGAS (Fig. 5C and fig. S6A). In vitro assay<br />
showed that this protein fragment (residues 161<br />
to 522) still activated the IRF3 pathway in a DNAdependent<br />
manner (fig. S6B). Thus, the N-termin<strong>al</strong><br />
160 amino acids of h-cGAS, whose primary sequence<br />
is not highly conserved evolutionarily, appear<br />
to be largely dispensable for DNA binding<br />
and cat<strong>al</strong>ysis by cGAS. In contrast, the NTase and<br />
Mab21 domains are important for cGAS activity.<br />
cGAS is predominantly loc<strong>al</strong>ized in the cytosol.<br />
To d<strong>et</strong>ermine wh<strong>et</strong>her cGAS is a cytosolic<br />
DNA sensor, we prepared cytosolic and nuclear<br />
extracts from THP1 cells and an<strong>al</strong>yzed the loc<strong>al</strong>ization<br />
of endogenous h-cGAS by immunoblotting.<br />
h-cGAS was d<strong>et</strong>ected in the cytosolic<br />
extracts but was barely d<strong>et</strong>ectable in the nuclear<br />
extracts (Fig. 6A). The THP1 extracts were further<br />
subjected to differenti<strong>al</strong> centrifugation to separate<br />
subcellular organelles from one another<br />
and from the cytosol (Fig. 6B). Similar amounts<br />
of h-cGAS were d<strong>et</strong>ected in S100 and in the pell<strong>et</strong><br />
after 100,000g centrifugation, which suggests that<br />
this protein is soluble in the cytoplasm but that<br />
a substanti<strong>al</strong> fraction of the protein is associated<br />
with light vesicles or organelles. The cGAS pro-<br />
A B<br />
IB: GST<br />
C<br />
Streptavidin<br />
Pull-down<br />
Lane: 1 2 3<br />
10%<br />
Input<br />
ISD<br />
Bio-ISD<br />
GST-RIGI(N)<br />
GST-m-cGAS<br />
GST-h-cGAS<br />
G<br />
Full-length<br />
1-212<br />
213-522<br />
1-160<br />
161-522<br />
∆161-212<br />
1-382<br />
213 S214 E225 D227 Mab21<br />
NTase<br />
h-cGAS<br />
1 160 212 382 522<br />
tein was not d<strong>et</strong>ected in the pell<strong>et</strong> after 5000g<br />
centrifugation, which contained mitochondria<br />
and endoplasmic r<strong>et</strong>iculum (ER) as evidenced<br />
by the presence of VDAC and STING, respectively.<br />
cGAS was <strong>al</strong>so not d<strong>et</strong>ectable in the pell<strong>et</strong><br />
after 20,000g centrifugation, which contained predominantly<br />
ER and heavy vesicles (Fig. 6B).<br />
We <strong>al</strong>so examined the loc<strong>al</strong>ization of cGAS<br />
by confoc<strong>al</strong> immunofluorescence microscopy<br />
of L929 cells stably expressing Flag–m-cGAS<br />
(Fig. 6C). The cGAS protein was distributed<br />
throughout the cytoplasm but could <strong>al</strong>so be observed<br />
in the nuclear or perinuclear region. After the<br />
cells were transfected with Cyanine 3 (Cy3)–labeled<br />
ISD for 2 or 4 hours, punctate forms of cGAS were<br />
observed, and they overlapped with the DNA<br />
fluorescence. Such coloc<strong>al</strong>ization and apparent<br />
aggregation of cGAS and Cy3-ISD was observed<br />
in more than 50% of the cells under observation.<br />
These results, tog<strong>et</strong>her with the biochemic<strong>al</strong> evidence<br />
of direct binding of cGAS with DNA, suggest<br />
that cGAS binds to DNA in the cytoplasm.<br />
Discussion. We have developed a strategy<br />
that combines quantitative mass spectrom<strong>et</strong>ry<br />
with convention<strong>al</strong> protein purification to identify<br />
biologic<strong>al</strong>ly active proteins parti<strong>al</strong>ly purified from<br />
crude cell extracts. This strategy may be gener<strong>al</strong>ly<br />
applicable to proteins that are difficult to<br />
purify to homogeneity because of very low abundance,<br />
labile activity, or scarce starting materi<strong>al</strong>s.<br />
10%<br />
Input<br />
Streptavidin<br />
Pull-down<br />
IB: Flag<br />
ISD<br />
Bio-ISD<br />
Streptavidin<br />
Pull-down<br />
10%<br />
Input<br />
ISD<br />
Bio-ISD<br />
Bio-RNA<br />
IB: Flag Flag-h-cGAS<br />
Lane: 1 2 3 4<br />
IFN-β RNA (fold)<br />
0 40 80 120<br />
Fig. 5. cGAS is a DNA binding protein. (A) The indicated GST fusion proteins were expressed and<br />
purified from E. coli andthenincubatedwithstreptavidinbeadsinthepresenceofISDorbiotin-ISD.<br />
Bound proteins were eluted with SDS sample buffer and d<strong>et</strong>ected by immunoblotting with a GST antibody.<br />
(B) Flag–h-cGAS was expressed and purified from HEK293T cells and then incubated with streptavidin<br />
beads as described in (A), except that a Flag antibody was used in immunoblotting and a biotin-RNA was<br />
<strong>al</strong>so tested for binding to cGAS. (C) Flag-tagged full-length or truncated human cGAS proteins were<br />
expressed in HEK293T cells and affinity-purified. Their ability to bind biotin-ISD was assayed as described<br />
in (B). Right panel: Expression plasmids encoding full-length and del<strong>et</strong>ion mutants of h-cGAS were transfected<br />
into HEK293T-STING cells, and IFN-b RNA was then measured by qRT-PCR.<br />
15 FEBRUARY 2013 VOL 339 SCIENCE www.sciencemag.org<br />
As a proof of principle, we used this strategy to<br />
identify the mouse protein E330016A19 as the<br />
enzyme that synthesizes cGAMP. This discovery<br />
led to the identification of a large family of cGAS<br />
that is conserved from fish to human, form<strong>al</strong>ly<br />
demonstrating that vertebrate anim<strong>al</strong>s contain evolutionarily<br />
conserved enzymes that synthesize<br />
cyclic dinucleotides, which were previously found<br />
only in bacteria, archaea, and protozoa (11–13).<br />
Vibrio cholerae can synthesize cGAMP through<br />
its cyclase DncV (VC0179), which contains an<br />
NTase domain but has no obvious primary sequence<br />
homology to the mamm<strong>al</strong>ian cGAS (12).<br />
Our results not only demonstrate that cGAS<br />
is a cytosolic DNA sensor that triggers the type<br />
I interferon pathway, but <strong>al</strong>so reve<strong>al</strong> a mechanism<br />
of immune sign<strong>al</strong>ing in which cGAS generates<br />
the second messenger cGAMP, which binds<br />
to and activates STING (4), thereby triggering<br />
type I interferon production. It remains to be d<strong>et</strong>ermined<br />
wh<strong>et</strong>her STING evolved first to d<strong>et</strong>ect<br />
cyclic dinucleotides from bacteria, or to d<strong>et</strong>ect<br />
endogenous cGAMP in the host as a mechanism<br />
of responding to cytosolic DNA. Although<br />
STING can directly d<strong>et</strong>ect certain cyclic dinucleotides<br />
produced by some bacteria, the deployment<br />
of cGAS as a cytosolic DNA sensor<br />
greatly expands the repertoire of microorganisms<br />
d<strong>et</strong>ected by the host immune system. In principle,<br />
<strong>al</strong>l microorganisms that can carry DNA<br />
on February 14, 2013<br />
www.sciencemag.org<br />
Downloaded from