Calcium-Binding Protein Protocols Calcium-Binding Protein Protocols

244 Weljie and Heringa
sequences as increasing the domain size might provide more “signal” and
lower the amount of “noise” in these alignments.
2. C2-domain proteins: The C2-domain proteins present a difficult case for automatic
global alignment algorithms as two topologies exist for this motif. Structural
superpositioning of the two topologies shows that strand 8 from topology II
is equivalent to strand 1 in topology I, while both strands are sequentially similar.
Manual alignments demonstrate that sequences with the topology I fold overhang
at the N-terminal end by one strand, and the other group overhangs at the
C-terminus (3). Hence, in the current example, these overhangs were not considered,
and the alignments encompassed the middle of strand 2 to the middle of
strand 8 in topology I, and the middle of strand 1 to the middle of strand 7 from
topology II. The C2 domains are found as single, double, and triple domains,
hence the same strategy was employed as with the EF-hand alignments; individual
C2 domains were extracted and then aligned. The results from Fig. 2 demonstrate
that both programs separated the secondary structure elements from
sequences that were reasonably highly conserved. However, the PRALINE algorithm
continued to keep secondary structure elements together as the homology
decreased, with a few exception in the C-terminal region, whereas the
CLUSTALX program had serious problems with the low homology sequences.
On the other hand, the PRALINE alignment shows a greater number of gaps
within the aligned β-strand regions, which might hamper their recognition in the
absence of structural knowledge. Both alignments were submitted to the JPRED
server to ascertain which alignment provides the best prediction (see Fig. 2).The
JPRED server was able to predict strands 1, 2, 3, and 5 from both alignments
reasonably well. Strand 6 was not predicted from either alignment, and only the
PRALINE alignment saw strand 4, although it was shifted towards the C-terminus.
The PRALINE alignment was also more consistent in matching the residues
thought to be important in calcium-ion coordination. These alignment differences
are significant if one is interested in including homologs of lower identity for
subsequent analysis, especially when attempting to predict secondary or tertiary
structure. The phylogenetic trees generated from both alignments were much
more stable than those from the EF-hand alignments, presumably because of the
Fig. 2. (see opposite page) Selected sequences from an alignment of C2-domain
regions as described in the text. (A) The resulting alignment from PRALINE, and (B)
from ClustalX. Note that the order of the sequences in (B) has been modified to facilitate
comparison with (A). The first column is the sequence ID with the occurrence
number appended if there are multiple C2-domains from the same sequence used in
the complete alignment, and the second column in (A) is the accession code, followed
by the alignment. In both alignments, residues of synaptotagmin known to have β-strand
secondary structure are denoted by “S” above the alignments. Secondary structure
predictions as performed by the JPRED server for the top sequence in the alignments
(but based on the complete alignments) are given under each of the two alignments.