Full text of DOI:10.1080/15216540050212114

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
–
IUBMB Life, 50: 105 113, 2000
c
°
Copyright 2000 IUBMB
1521-6543/00 $12.00 + .00
Original Research Article
A CD2-Based Model of Yeast a -Agglutinin Elucidates
Solution Properties and Binding Characteristics
Arabela Grigorescu,1,3 Min-Hao Chen,1,4 Hui Zhao,1,5 Peter C. Kahn,2
and Peter N. Lipke1
1Department of Biological Sciences and The Institute for Biomolecular Structure and Function,
Hunter College of the City University of New York, New York, NY 10021, U.S.A.
2Department of Biochemistry and Microbiology, Cook College, Rutgers University,
New Brunswick, NJ 08901, U.S.A.
Keywords Celladhesionprotein; homology model; immunoglobulin
superfamily; peptide mapping; residue volume; Saccha-
romyces cerevisiae; surface accessibility.
Summary
We have previously shown that the Saccharomyces cerevisiae
cell adhesion protein -agglutinin has sequence characteristics
a
of immunoglobulin-like proteins and have successfully modeled
INTRODUCTION
residues 200–325, based on the structure of immunoglobulin vari-
able-type domains. Alignments matching residues 20–200 of
-
a
a-Agglutinin is a highly glycosylated cell wall protein that
agglutinin with domains I and II of members of the CD2/CD4
subfamily of the immunoglobulin superfamily showed > 80%
conservation of key residues despite low sequence similarity over-
–
functions as a cell cell adhesion protein during mating in the
( )
yeast Saccharomyces cerevisiae 1 . a-Agglutinin is expressed
on cells of mating type a and binds to the glycoprotein ligand
a-agglutinin expressed by cells of the opposite a mating type.
Expression of both a-agglutinin and a-agglutinin is increased
after stimulation by sex pheromones, and the subsequent in-
a
all. Three-dimensional models of two -agglutinin domains con-
structed on the basis of these alignments were shown to con-
form to peptide mapping data and biophysical properties of
a
-agglutinin. In addition, the residue volume and surface acces-
sibility characteristics of these models resembled those of the well-
packed structures of related proteins. Residue-by-residue analysis
showed that packing and accessibility anomalies were largely con-
Ž ned to glycosylated and protease-susceptible loop regions of the
domains. Surface accessibility of hydrophobic residues was typi-
cal of proteins with extensive domain interactions, a Ž nding com-
–
teraction of the adhesion proteins mediates cell cell binding
and facilitates cell fusion. The binding has an association con-
stant of 109 L/mol, and the association is slow and sensitive to
(
)
cold, implying the involvement of hydrophobic groups 1, 2 .
a-Agglutinin is related to several Candida albicans cell adhe-
sion proteins involved in host pathogen interactions 3, 4 .
a-Agglutinin consists of several functional regions 1, 5 . A
19-residue N-terminal secretion signal precedes a sequence of
a
patible with the hydrodynamic properties of -agglutinin and the
–
(
)
hydrophobic nature of binding to its peptide ligand a-agglutinin.
(
)
a
The procedures used to align the -agglutinin sequence and test
the quality of the model may be applicable to other proteins, espe-
cially those that resist crystallization because of extensive glycosyl-
ation.
(
)
300 residues that makes up the ligand-binding region Fig. 1 .
C-terminal to the binding region are 300 residues constituting
«
«
–
IUBMB
Life, 50: 105 113, 2000
a heavily glycosylated stalk, which elevates the binding region
away from the surface of the cell wall. A 23-residue C-terminal
hydrophobic sequence is a signal for addition of a glycosyl phos-
phatidylinositol anchor. The glycosyl phosphatidylinositol gly-
Received 16 May 2000; accepted 4 August 2000.
Address correspondence to Dr. Peter N. Lipke, Department of
–
caniscleavedat the cellsurface, and the glycan protein complex
( )
Biological Sciences, Hunter College, 695 Park Avenue, New York, is covalently attached to cell wall polysaccharides 6 .
NY 10021. E-mail: lipke@genectr.hunter.cuny.edu
The binding region of a-agglutinin comprises residues
20 325, a region containing recurrent sequence motifs char-
acteristic of proteins of the immunoglobulin Ig superfamily
5, 7 . Comparing the a-agglutinin amino acid sequence be-
3Current address: Department of Biophysics, University of Pitts-
–
burgh, Pittsburgh, PA 15260.
(
)
4Current address: Abbott Laboratories, Abbott Park, IL 60054.
5Current address: Center for Experimental Therapeutics and Perfu-
sion Injury, Harvard University Medical School, Boston, MA 02115.
(
)
–
tween residues 200 325 and a consensus sequence of Ig
105

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
106
GRIGORESCU ET AL.
Figure 1. Features of a-agglutinin. DisulŽ des are shown as clear boxes below the bar, and glycosylation sites are also shown
above, with polymannosyl N-linked chains shown as brushlike, and O-linked mannosyl oligosaccharides as vertical lines. The
–
( )
disulŽ de assignments and glycosylation patterns are based on peptide mapping of residues 20 350 5 . The glycosylation pattern
( )
is inferred from sequence and carbohydrate content between residues 350 and 600 5 . GPI, glycosyl phosphatidylinositol.
variable-like domains gave a highly signiŽ cant alignment score family, creation of an optimal alignment to a sequence of a ref-
(
¡7) ( )
p < 10
8 . A model based on this alignment not only had erence protein with known structure, construction of the model,
good biophysical properties but also was consistent with muta- and assessment. Such models are now highly reliable for se-
genesis studies and with maps of proteolysis sites, glycosylation quences having 30% identity or greater, and their creation is
«
(
)
sites, and disulŽ de bonds. Circular dichroism studies conŽ rm automated and freely available on the World Wide Web 10 .
that the entire binding region of a-agglutinin has a high content However, many known sequences are well below the 30% thre-
of antiparallel b-sheets and little or no a-helix, both character- shold, and the fraction of open reading frames with low similar-
(
)
istics of IgSF proteins 2, 5, 7 .
ity to proteins of known structure will increase as we determine
Ig-like domains are usually present in tandem, especially in sequences from genome studies.
(
)
cell adhesion proteins 7, 9 . We have proposed that residues
With some care, useful models have been created for proteins
–
20 200 of the a-agglutinin binding region contain two more with considerably less identity to a known structure, especially
(
)
(
)
Ig-like domains 5, 8 . As a test of this hypothesis, we aligned in the Ig superfamily 11, 12 . For such structures, assessment
–
a-agglutinin residues 20 200 with consensus sequences for var- is particularly critical. The extensive biochemical analysis of
ious Ig domain subfamilies and found high-quality matches with a-agglutinin provides a source of important data for evaluat-
domains I and II of the CD2/CD4 subfamily. Models based on ing its status as a member of the Ig superfamily, because the
these alignments had biophysical characteristics of well-packed data can be compared with speciŽ c details of modeled Ig
proteins and were consistent with the known solution properties domains.
of a-agglutinin. In addition, the models were consistent with
biochemical mapping data that had not been used inthe construc-
tion of the models. The models are explanatory and predictive
for a-agglutinin, and may also allow prediction of differences
IdentiŽcation of Reference Structures
For sequences of low similarityto known structures, the Ž nd-
ing of suitable reference structures is the most uncertain part
of modeling. Automated global sequence comparisons are not
yet able to identify all proteins with similar folding patterns
(
)
between a-agglutinin and its homologs in C. albicans 3, 4 .
SIMILARITY-BASED MODELING
(
)
10, 13 . Indeed, some proteins have highly similar domain
The four steps for similarity-based modeling of proteins are structures but no recognizable sequence similarity, including the
(
)
identiŽ cation of the sequence of interest with a speciŽ c gene cadherin family compared withthe Ig superfamily 14 , myosins

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
CD2-LIKE YEAST CELL ADHESION PROTEIN
107
with kinesins 15 , and yeast Mata2p with homeodomain conserved in multiple alignments of a-agglutinin and the
(
)
(
)
proteins such as engrailed 16 . In fact, CD2 was identiŽ ed as C. albicans Als epithelial adhesion proteins with the CD2/CD4
(
)
a member of the Ig superfamily only after determination of its alignment of Fig. 2 4; data not shown . The “Bioinbgu” thread-
(
)
structure by crystallography and nuclear magnetic resonance ing program also identiŽ ed CD2 Protein Database Ž le 1HNF as
(
)
17, 18 .
the most likely homolog of a-agglutinin, with a Z 7.1 com-
=
(
Potential reference structures were identiŽ ed by BLAST and pared with Z 3.8 for the next best match. This match was
=
JPRED and a search of the Protein Data Bank for b-rich struc- found by a referee for this manuscript; the program is avail-
(
)
)
tures 13, 19 . Of the identiŽ ed protein families, only Ig super- able at http://www.cs.bgu.ac.il/ bioinbgu. The alignment gen-
«
(
( )
family structures aligned within conserved sequence motifs see erated by Bioinbgu similar to one previously discussed 5 and
–
)
below .
is identical to that in Fig. 2 in strands A C of domain I, although
For a-agglutinin, matches to the Ig family were found by a it diverges thereafter. These matches conŽ rm that a-agglutinin
visual sequence motif comparison, which had originally iden- has CD2-like sequence characteristics.
tiŽ ed sequences around the Cys residues as being similar to
The match scores for the alignment shown in Fig. 2 are much
(
)
those in Ig-like proteins 1, 5, 8 . Of the six Cys residues in better than those for matches to a previous IgV consensus, which
–
the binding region of a-agglutinin, Ž ve are in sequences of 6 8 showed only 50% and 33% identity for domains I and II, respec-
–
( )
residues characteristic of the Cys residue regions in Ig proteins. tively 5 . Therefore, the alignment of residues 20 200 of a-
Because several such similarities were available, we could then agglutinin to the published structure-based sequence alignment
(
)
search for other critical structural motifs, such as the conserved of CD2 and CD4 17 was the basis for modeling a-agglutinin
–
Trp residues and the Asp Arg ion pairs that stabilize many Ig domains I and II.
(
)
domain cores 7, 20, 21 .
Construction of the Model
The model was constructed by using the Homology features
Alignment
(
)
of the program QUANTA Insight/Quanta, San Diego, CA . The
We generated alternative alignments of the a-agglutinin do-
mains I and II regions to Ig consensus sequences, both by hand
and by using several pairwise and multiple alignment algo-
rithms. The most effectivemethod was hand-generated; italigned
conserved consensus residues in regions of maximum b-sheet
potential, allowing gaps only in loop and turn regions 8 .
With this approach, a-agglutinin residues 20 120 domain I
were matched to an Ig Variable Domain IgV consensus in
22 of 24 positions with a single one-residue gap data not
main chain atoms were assigned the coordinates of the corre-
(
)
sponding atoms of CD2 Protein Data Bank Ž le 1HNF in those
regions with no gaps or only small gaps. These residues included
all of the b strands, as well as domain I loops AB, BC, CC0,
C0C00, DE, and EF, and domain II loops BC, CC0, EF, and FG.
Then side chains were built using the QUANTA AMINO.RTF
library. Each b-strand region was then “regularized,” to repair
single residue gaps or insertions and to correct conŽ guration of
the backbone.
( )
–
(
)
(
)
(
)
shown; 20 .
An independent analysis of Ig variable-like domains of low
sequence similarity deŽ ned a V-frame proŽ le that identiŽ es a
Separatesearches were performed tobuild the coordinates for
residues of the loops C00D and FG from domain I and AB from
(
)
small number of key sites 21, 22 ; residues that occupy these
positions have crucial structuralroles andarehighly conserved in
Table 1
a
–
all IgV structures. Again, we were able to align residues 20 120
(
)
Sequence identities and similarities %
of a-agglutinin to such a proŽ le, based on the consensus and b-
strand probability maxima in 15 of 18 positions for a-agglutinin
Core
CD2 human CD4 human consensusb
(
)
(
)
(
)
domain I Fig. 2 and Table 1 . Randomized sequences of a-
–
agglutinin residues 20 122 could not be Ž t to the IgV consensus
First domains
(
)
(
)
20 , nor to the V-frame proŽ le 22 , nor to a secondary structure
(
—
)
(
)
)
(
)
a-Agglutinin
16 40
16 34
83 88
(
)
prediction proŽ le 5, 8 .
(
)
(
(
94 100
)
)
CD2 human
15 44
Matches of the a-agglutinin sequence were better with CD2
or CD4 as the reference protein than with other members of the
Ig superfamily. The CD2/CD4 subfamily consists of proteins
with tandem Ig-like domains. CD2, has a V-like domain with
nine0 b strands in two sheets consisting of strands ABED and
Second domains
(
)
)
(
)
)
)
(
a-Agglutinin
16 44
16 46
92 100
(
)
(
CD2 human
—
26 45
100
c
(
(
(
)
Third domain,
14 33
14 33
74 79
a-agglutinin
00
(
)
C C CFG, respectively 22 . The second domain is of the trun-
cated 7 strand S-type, with b-sheets composed of strands ABE
and C0CFG, respectively CD4 has four Ig domains in a V-S-
V-S tandem array. The proŽ le-based match for domain I was
extended into a match for domain II of 11 of 12 key posi-
tions for an alignment of domain II with truncated S-type do-
a
(
)
The percent identity and similarity in parentheses at aligned posi-
(
)
tions isshown for the alignmentsof Fig. 2 bottom panel . The similarity
setswere: AFILMPVWY, DEKR, ND, NQ, EQ, KRH, ST, FHWY, and
AG.
b
(
)
Based on conserved consensus residues bolded and named in
(
)
Fig. 2 bottom panel .
cSee reference 8.
(
)
mains Fig. 2 and Table 1 . Consensus residues were similarly

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
108
GRIGORESCU ET AL.
(
)
Figure 2. Alignment of a-agglutinin domains I and II. Sequence alignments based on the structures of CD2 and CD4 17 are
(
)
(
)
shown above the best consensus matchalignment of a-agglutinin for domain I top and domain II bottom . The b strands are boxed
(
)
and labeled. V-frame residues 22 are marked at the bottom of the alignments, and the residues in these positions are shown in bold
face. The consensus residue characteristicsare: AB1, Gly; B1, aliphatic; B3, large hydrophobic; B5, Cys or large hydrophobic; B7,
neutral or hydrophobic; C2, hydrophobic; C4, Trp; D, basic; D2, hydrophobic; E4, hydrophobic, usually Leu; E6, hydrophobic;
EF2, hydrophobic, Ser, or Thr; EF6, Asp; F1, Gly, Ala, Asp; F3, Tyr; F5, Cys or large hydrophobic; G6, G8, and G10, hydrophobic
(
)
22 .
domain II. Loops from homologous positions in other members Ž de could not be accommodated by several other models, in-
[ ]
of the Ig SF were chosen as reference structures. The Loop Mod- cluding one based on the alignment we proposed earlier 5
eling function of QUANTA was applied to each of the result- and the alignment proposed by the threading program “bioin-
)
ing loops of the a-agglutinin structure to generate low-energy bgu” . Peptide mapping has identiŽ ed 19 residues in domains I
conformations. and II that are exposed to proteases and a glycosylation site
CHARMM minimization in vacuo was then performed for at Asn109. All proteolysis sites follow residues exposed on the
Ê
(
)
the entire structure to a root mean square force of 2 kcal/A. domain surfaces Figs. 3 and 4 , and most are at the ends of
¡
The energy of the reported structure is 6674 kcal/mol. The strands or in loops. Lys154, the only residue cleaved by pro-
¡
0
–
models are shown in Fig. 3.
tease Arg-C, is exposed in the C C loop of domain II. Cleavage
at this site leaves the protein intact, as expected for a cleav-
age within a domain. Dissociation of the Arg-C proteolytic
Evaluation
( )
The models ofdomains Iand IIwere evaluatedfor congruence fragments is possible only after unfolding of the domain 2 .
with peptide mapping data and other biochemical data and were The corresponding region of CD2 is also poorly structured
(
)
analyzedfor residue volume occupancy and surface accessibility
17, 18 .
Biophysical Characteristics. The two modeled domains
Biochemical Tests. The models of domains I and II are have all-b structures with well-packed and extensive hydropho-
(
)
5, 23, 24 .
(
)
consistent with known biochemical data. The current model bic cores Fig. 3 . Important tests for protein structures are
of domain I includes a disulŽ de bond between the F and G residue volume and surface accessibility. These tests have id-
(
strands, consistent with the mapped position note that thisdisul- entiŽ ed poorly packed regions of models based on X-ray