| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
RI and Bacterial Decoy Proteins1Department of Structural Biology and Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
RI, a receptor for IgA-Fc, recruits myeloid cells to attack IgA-coated pathogens. By competing with FcRI for IgA, bacterial decoys, like SSL7 of Staphylococcus aureus, subvert this defense. We examined how pathogen selection has driven the diversification and coevolution of IgA and FcRI. In higher primates, the IgA binding site of FcRI diversified under positive selection, a strong episode occurring in hominoid ancestors about the time of the IgA gene duplication. The differential binding of SSL7 to IgA-Fc of different species correlates with substitution at seven positions in IgA-Fc, two of which were positively selected in higher primates. Two others, which reduce SSL7 binding, emerged during episodes of positive selection in the rabbit and rodent lineages. The FcRI-IgA interaction evolves episodically under two types of positive selection: pressure from pathogen decoys selects for IgA escape variants which, in turn, selects for FcRI variants to keep up with the novel IgA. When FcRI cannot keep up, its function is lost and the gene becomes susceptible to elimination, as occurred in the mouse genome, either by chance or selection on one of the many linked, variable immune system genes. A cluster of positively selected residues presents a putative binding site for unknown IgA-binding factors. |
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
RI (CD89), a transmembrane protein expressed constitutively by myeloid cells: monocytes, macrophages, dendritic cells, neutrophils, eosinophils, and Kupffer cells (2). In combination with the FcR
chain, FcRI forms a bifunctional signaling receptor that activates cells when bound to Ag-IgA complexes, but inhibits cells when bound to IgA in the absence of cross-linking by Ag (3).
Unlike its distant homologs FcRs and Fc
RI, whose genes reside on chromosome 1 (4), FcRI is encoded by a gene (FCAR) located in the leukocyte receptor complex (LRC)3 on chromosome 19q13.4 (5). There FCAR flanks the killer cell Ig-like receptor (KIR) and NKp46 genes (6). As both structural and phylogenetic analyses indicate, FcRI is more related to the receptors encoded by LRC gene families than to other FcRs (7, 8); this chromosomal localization points to the FCAR, KIR, and other LRC genes having evolved from a common ancestor.
The extracellular part of FcRI consists of two Ig-like domains (EC1 and EC2) that are approximately orthogonal (9). Only the membrane distal EC1 domain contacts IgA directly (10, 11) and two FcRI molecules can simultaneously contact one IgA-Fc molecule. This 2:1 stoichiometry is also observed between the MHC class I-related receptor FcR and IgG (12) but differs from the one-to-one stoichiometries of the FcRIII and FcRI receptors with their ligands (13). The crystal structure of the IgA-FcRI complex (8) and mutagenesis analysis (14) identified 11 aa residues, in three clusters, in the EC1 domain that contact 19 positions of IgA-Fc: 16 on the C3 domain and 3 on the C2 domain.
Interaction with FcRI is necessary for IgA to stimulate cellular responses that lead to the attenuation or elimination of a pathogen (15), e.g., an IgA-coated bacterium. Inevitably, the FcRI-IgA interaction has become a target for interference by pathogens: Streptococcus pyogenes (group A streptococcus), group B streptococcus, and Staphylococcus aureus make decoy proteins that by binding to IgA prevent the interaction with FcRI (16, 17). This strategy can confer the bacteria with an ability to evade IgA-mediated clearance, as illustrated by the contribution of the IgA-binding part of S. pyogenes M protein to the phagocytosis resistance of the bacteria (18). Although these interactions have not been defined by three-dimensional structures, the biochemical evidence shows that the pathogen proteins target the FcRI binding site of IgA-Fc (16).
The KIR gene family, which flanks the FCAR gene, is characterized by variability and rapid evolution as seen from its diverse gene content, high allelic polymorphism, and striking divergence between species (19). These differences have been associated with disease susceptibilities and resistance in a variety of clinical settings (20). First hints to the variability of KIR were the observations that mice and humans do not use KIR for similar purposes (21) and that the mouse LRC contains no KIR genes (22). Analogously, although an FCAR gene has been found in several species of mammals, including rats, it is absent from the mouse genome (23). This shows that FCAR can be dispensable, as are the KIR genes, which also implies that it is a potential target for modification and adaptation. That FcRI has dual and conflicting functions in the prevention and generation of inflammation (3) also raises the possibility of variant FcRI for which the balance between these functions is differentially set. We, therefore, investigated variation in the FCAR of humans and chimpanzees and the role natural selection has played in the evolution of the FcRI-IgA interaction.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
Full-length FCAR sequences were characterized from three sources of cells: polymorphonuclear neutrophils (human donors 1, 2, 3, and 5), PBMCs (human donors 4, 6, 7, 8, and chimpanzees Donald, Sonia, Kipper, Brandy, and Elwood) and EBV-transformed B cell lines (human donor 9 and chimpanzees Termite, Phineas, Eve, and Harry). Cells were placed in culture with PMA (100 ng/µl) for 3–5 days in RPMI 1640 medium supplemented with 10% FBS, 100 U/ml penicillin-streptomycin, and 2 mM L-glutamine. RNA was extracted from cells using TRIzol reagent (Invitrogen Life Technologies) and cDNA was prepared using a SuperScript First Strand Synthesis kit (Invitrogen Life Technologies).
For the EC1 exon analysis, we investigated 93 human donors representing four population groups (Africans, n = 55; Caucasians, n = 22; East Asians, n = 9; and South Asians, n = 7); these donor panels will be described elsewhere (P. Norman, manuscript in preparation). Genomic DNA from 91 unrelated chimpanzees was analyzed. Bonobo, gorilla, orangutan, gibbon, and squirrel monkey EC1 sequences were obtained from two individuals in each species.
This study was approved by the Stanford University administrative panels on human subjects in medical research and laboratory animal care.
Amplification, cloning, and sequencing
FCAR cDNA PCR amplifications utilized primers FCAR-F/R. Full-length variants were isolated by gel purification (QIAEX II; Qiagen), cloned into the pCR4-TOPO vector (Invitrogen Life Technologies), and sequenced.
Human and chimpanzee EC1 exons were amplified using the primers FCAR-EC1-HSA-F/R and Pt-FCAR-EC1-F1/R1, respectively. PCR products were purified (QIAquick; Qiagen) and directly sequenced. PCR products containing potentially new polymorphisms were further investigated by cloning and sequencing. Bonobo, gorilla, orangutan, and gibbon EC1 exons were obtained using the chimpanzee EC1 primers. To obtain the squirrel monkey EC1 sequences, amplifications with the chimpanzee EC1 primers were first performed on DNA from a BAC clone known to contain the squirrel monkey FCAR gene; the resulting PCR products were cloned, sequenced, and specific primers were designed (SM-FCAR-Spe-F/R).
For each analysis, two independent amplifications were performed and multiple clones from each amplification were analyzed. Sequencing was performed on a CEQ 2000XL sequencer (Beckman Coulter) and on an ABI377 DNA sequencer (Applied Biosystems). PCR amplification conditions are available upon request.
The following primers were used (all of the sequences are 5'-3'): FCAR-F, GTGCATTGAAAGGAGAGCAAC; FCAR-R, TGTCCTTCAAGTAGCTTTGTCG; Pt-FCAR-EC1-F1, TTTTACTTCCCCCACAGAGA; Pt-FCAR-EC1-R1, CTGAGAACTCTCTGAAGCAATC; SM-FCAR-Spe-F, CCACCTGAGTCTGGGCTTTC; SM-FCAR-Spe-R, CCGTGAACCTGGGTGTTTC; FCAR-EC1-HSA-F, ATTTTTACTTCTCCCACAGAGA; and FCAR-EC1-HSA-R, TGAGAACTCTCTGAAGCAATC.
Datasets
The FCAR dataset was constructed following BLAST (24) searches against the National Center for Biotechnology Information’s nonredundant database. Coding nucleotide sequences were aligned using MAFFT (25) and corrected manually; the rat sequence was trimmed in 3' because this part of the sequence could not be aligned.
The IgA C2-C3 and the SSL7 (staphylococcal superantigen-like protein 7) datasets were constructed similarly to the FCAR dataset. For the SSL7-binding analysis, IgA C2-C3 sequences for which data on SSL7 binding to IgA was available (17) were kept and translated into amino acids (the cattle and sheep sequences were excluded because their binding pattern was ambiguous).
To characterize the rabbit FCAR gene, the rabbit draft genome assembly of May 2005 available at Ensembl (www.ensembl.org) was searched. Sequences with similarity to FCAR exon 3 (encoding the EC1 domain) were obtained and aligned with the sequences of representatives of the mammalian LRC gene families. The last 
50 bp of the rabbit sequences were discarded as they could not be reliably aligned.
Accession numbers for the sequences characterized in this study are: human FCAR alleles, DQ075334–39 (001–006); chimpanzee FCAR alleles, DQ075340–44 (001–005); FCAR exon 3 sequences: EF077231 (FCAR (101C)), EF077232 (FCAR (132G)), DQ075345 (Pt-FCAR (289T)), EF077235 (Pp-FCAR (1)), DQ075346 (Gg-FCAR (1)), DQ075347 (Gg-FCAR (2)), EF077230 (Gg-FCAR (3)), EF077234 (Popy-FCAR), EF077233 (Hyla-FCAR), EF077236 (Sabo-FCAR), and EF077229 (Sasc-FCAR). Datasets used in this analysis are available upon request.
Diversity analysis
The average sequence diversity was estimated from pairwise comparisons using MEGA 3.1 (26); the SE was obtained with the bootstrap method (10,000 replicates). DNASP version 4.10.9 (27) was used to estimate 
(nucleotide diversity) and 
w (Watterson’s estimator) and their SD. KIR gene diversity in the Northern Ireland population was estimated from allele frequency data (28), whereas KIR3DL2 and KIR3DL1 gene diversity in the FCAR panel was estimated from a previous allele-level characterization (29).
Phylogenetic analysis
For the nucleotide datasets, three methods were used: neighbor-joining (NJ), parsimony, and Bayesian phylogenetics. NJ analyses were performed with MEGA 3.1 (26) using the Tamura-Nei method with 1,000 replicates. PAUP*4.0b10 (30) and the tree bisection-reconnection branch swapping algorithm were used for parsimony analyses with 1,000 replicates and a heuristic search. For the Bayesian analysis, we selected the model of DNA substitution using Modeltest3.7 (31) and the Akaike information criterion. Bayesian phylogenetic analyses used MrBayes3.1.2 (32); sampling was performed with one cold chain and three heated chains, which were run for 106 generations. Trees were sampled every 200 generations and the first 2,500 trees were discarded before a consensus tree was generated. Three simultaneous runs were conducted and the resulting tree topologies were compared using the Shimodaira-Hasegawa test of alternative phylogenetic hypotheses with resampling estimated log-likelihood optimization and 10,000 bootstrap replicates (as implemented in PAUP*4.0b10). This comparison was made with the maximum likelihood model defined by Modeltest. The same topology comparison was then performed between the topologies obtained with the NJ, parsimony, and Bayesian methods. Unless otherwise mentioned, the test failed to reject any of the alternative tree topologies ( = 0.05).
For the amino acid datasets, the analyses were performed similarly to the nucleotide sequences with the following differences: NJ analyses were performed using a Poisson correction; the Bayesian analyses were conducted using a BLOSUM matrix, distances, and the resulting tree topologies were statistically compared using the Templeton test with a parsimony model ( = 0.05).
Rabbit FCAR-like genomic sequences
The orthology of the rabbit FCAR-like genomic sequences and the FCAR sequences of other mammals was established by phylogenetic analyses. However, all of the rabbit sequences possess frameshifts as well as a stop codon (in the new frame) so that they all represent pseudogenes.
Selection analysis
The average rate of synonymous substitutions (dS) and rate of nonsynonymous substitutions (dN) were estimated using the Kumar method, as implemented in MEGA 3.1 (26); SE were estimated by the bootstrap method (10,000 replicates).
dN:dS (
) ratios were estimated by maximum likelihood using PAML version 3.15 (33). Site and branch-site analyses were performed using the F3 x 4 model of codon frequencies. In the site analysis, the likelihood of a tree topology was estimated for several site-specific models in which the selective pressure varied among different sites but the site-specific pattern was identical across all lineages. Three sets of LRT were conducted to compare null models that do not allow >1 (M1a, M7, and M8a), with models that do (M2a and M8). Significance was assessed by comparing twice the difference in likelihood between the models (2
L) to a 
2 distribution with one (M8a/M8) or two (M1a/M2 and M7/M8) df. For the branch-site analysis, the likelihood of a tree topology given a null model constraining = 1 for the branch of interest was compared with the likelihood of the same tree topology given an alternative model allowing >1 (LRT with 1 df). The Bayes Empirical Bayes approach (34) was used to identify codons with >1 in the site and branch-site analyses. For the branch-site analyses, the CODEML program was modified: by default the distribution of the 2 parameter is approximated using 10 categories and the maximum value is fixed at 10.5; we raised this maximum to allow a better approximation of the distribution when 2>>10.5 for the branch of interest.
For the analysis of the higher primate FcRI EC1 sequences, the best tree topologies showed minor deviations from the species tree. We investigated the likelihood of a tree topology modified to eliminate the divergence from the species tree: since the new tree topology had virtually the same likelihood as the best trees (likelihood difference <0.1), we used it for selection analysis. The same approach was used for the analysis of the mammalian IgA C2 and C3 datasets. For the primate IgA C2 dataset, the likelihood of the modified tree was markedly reduced; while the reject of the modified tree was marginal (0.08), the original tree topology was preferred for the selection analysis. These topology differences were found to have little effect on the selection analysis (data not shown).
Ancestral sequence reconstruction
Analyses were performed with CODEML (33) using the marginal reconstruction approach and the M8 model.
Distribution of the selected sites in the FcRI EC1 domain
This distribution was studied using a binomial distribution: considering 
= (0,1,2,... ,n), 
k
, p = (X = k) = nCk * pk * qn-k. The clustering of 7 of the 9 selected EC1 residues in two regions that represent 27 of the 96 EC1 residues is thus seen to be unlikely if a random distribution is assumed ( = 0.01 with k = 7, n = 9, p = 0.281 and q = 0.719). The same bias is observed when the whole region between residues 48 and 86 is considered (k = 8, n = 9, p = 0.406, and q = 0.594) or when only the variable residues are considered (two regions: k = 7, n = 9, p = 0.308, and q = 0.692; one region: k = 8, n = 9, p = 0.404, and q = 0.596).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
Analysis of cDNA from nine human donors and nine chimpanzees identified six and five FCAR alleles, respectively (Fig. 1, A and B); the commonest allele in each species was named *001 (FCAR*001 and Pt-FCAR*001). In total, there are 21 positions of nucleotide substitution (Fig. 1C): on average human FCAR differs from chimpanzee FCAR by 14.5 substitutions (Fig. 1D), whereas chimpanzee FCAR alleles differ by 1–6 (mean, 3.8) substitutions and human FCAR alleles differ by 1–3 substitutions (mean, 1.7). Such lower intraspecies diversity compared with the interspecies diversity points to the allelic differences having evolved after separation of human and chimpanzee ancestors, a possibility confirmed by full-length and domain-by-domain phylogenetic analyses (Fig. 2).
|
|
) is 10.2 ± 2.2 x 10–4, slightly higher than the genome average (8 x 10–4) but within the expected range of genetic variation (Fig. 1D) (36). For chimpanzee FCAR, is 20.4 ± 6.9 x 10–4, also higher than the average of 13.2 x 10–4 (37). A similar difference between human and chimpanzee is observed when the diversity is assessed from the number of segregating sites (w). Particularly divergent is EC2, where chimpanzees have five polymorphic positions, humans have one and there is a 5-fold difference in and w (Fig. 1D). Since four of the five chimpanzee EC2 substitutions are synonymous, it is likely that these species differences reflect population history rather than difference in natural selection.
The FCAR gene flanks the KIR locus, which evolves rapidly and encodes polymorphic MHC class I receptors with two or three extracellular Ig domains. Assessing diversity using w showed FCAR to have diversity similar to KIR2D but less than KIR3D (Fig. 1E). In contrast, FCAR is lower than that of all KIR except KIR2DS4. This could reflect differences in the types of natural selection operating on FCAR and KIR genes. For example, KIR allele frequencies are subject to balancing selection (38).
EC1 has been subject to positive selection and EC2 to purifying selection in higher primates
We investigated natural selection on FCAR by using a pairwise comparison to estimate the synonymous (dS) and nonsynonymous (dN) substitution rates (Fig. 3A). Analysis of mammalian FCAR revealed a significant excess of synonymous substitutions ( = 0.05), an effect that became marginal when the analysis was restricted to primates. Because nonsynonymous substitutions concentrate in exon 3 (EC1) and synonymous substitutions concentrate in exon 4 (EC2) of catarrhine primates (hominoid and old world monkey) FCAR (Fig. 1C), we examined these exons individually (Fig. 3A). EC2 has a significant excess of synonymous substitutions in all taxonomic groups. Catarrhine EC1 has a significant excess of nonsynonymous substitutions, whereas when all mammals or all primates were considered the numbers of synonymous and nonsynonymous substitutions were equivalent. This showed that the IgA-binding EC1 domain of catarrhine FcRI was subject to positive selection, while the EC2 domain, which does not contact IgA, was subject to purifying selection.
|
The IgA binding site of FcRI has diversified under positive selection in higher primates
To identify sites in EC1 that were targets for selection, the dN:dS () ratios for each position of sequence variation were investigated by maximum likelihood using a codon-based substitution model (Ref. 39 and Fig. 3, B--D). In this analysis, the two models that permit >1 (M2a and M8) are significantly more likely ( = 0.01) than their equivalents that do not (M1a, M7, and M8a). These results concur with those from the pairwise comparison in emphasizing that positive selection has acted on the EC1 domain of higher primate FcRI. Eight positively selected positions (positions 21, 48, 61, 65, 71, 78, 85, and 86) were identified by M8 (p > 0.9), four of them (positions 48, 61, 65, and 85) also having good support with M2a (p > 0.9; Fig. 3C).
To see how positive selection has affected different taxonomic groups of higher primates, selection analysis was performed on four branches of the phylogenetic tree for EC1 (Fig. 3, B and E). It revealed a strong episode of positive selection in the hominoid ancestor (branch 1, = 0.01), which involved positions 48, 55, 61, and 85 (p > 0.95). When branch 1 was excluded from the analysis, positive selection was still detected ( = 0.05), showing it has also occurred on other branches of the tree. Individually, however, the evidence for positive selection on branches 2, 3, and 4 was marginal, only approaching significance for branch 4 (0.05). Natural selection has thus contributed generally to the diversification of higher primate EC1 sequences and was particularly strong in the hominoid ancestor.
Together, the site and branch analyses identified nine positively selected positions in FcRI EC1: residues 21, 48, 55, 61, 65, 71, 78, 85, and 86. Seven of these positions cluster in two regions that represent 27 of the 96 EC1 residues (boxes in Fig. 4A), a distribution significantly different from random ( = 0.01). That these two regions contain 10 of the 11 IgA binding sites and 7 of the 8 sites known to affect the IgA binding indicates natural selection on EC1 has targeted the binding site of FcRI for IgA-Fc. Mutation at two of the positively selected positions (H85 and Y87) is known to reduce the binding affinity of human FcRI for human IgA (Refs. 11 and 14 and Fig. 4A). A third residue (K55) makes three contacts with IgA-Fc in the crystallographic structure and contributes 9% of the FcRI surface buried upon interaction with IgA-Fc (Ref. 8 and Fig. 4, B and C).
|
RI binding site for IgA-Fc during higher primate evolution. The changes are particularly striking in the branch leading to the hominoid ancestor, a time frame corresponding to the duplication and diversification of the IgA locus (40). Higher primate IgA-Fc has been subject to positive diversifying selection
To investigate the possibility that changes in IgA imposed selection upon the EC1 domain of FcRI, we examined sequence divergence in the C2 and C3 domains of higher primate IgA (Fig. 5, A and B). Although C2 is more variable than C3, it has only three contact residues with FcRI compared to 16 for C3. Moreover, only three of the FcRI-binding positions are variable. Of these positions, 258 in C2 differs only in gibbon, and 387 and 389 in C3 are part of a cluster of nine substitutions, between positions 377–402, that distinguishes hominoid from old world monkey IgA and comprises most of C3 variation observed (Fig. 5B). Position 387 in human IgA interacts with positions 53, 55, and 57 in FcRI (8), of which position 55 is one of the sites positively selected in the hominoid ancestral branch (Fig. 3E). Ancestral sequence reconstruction indicates that substitution at position 387 (T387S) occurred during the same time interval. These observations suggested that some of the selected changes in higher primate FcRI could be a consequence of changes in IgA.
|
2 and C3 (Fig. 5, C and D). Pairwise sequence comparison of dN:dS revealed excess synonymous substitutions in higher primate C2 and C3 domains. For C3, the deviation from neutrality was significant ( = 0.05), but not for C2 (0.05; Fig. 5C). Further analysis of C3, using the same maximum likelihood approach applied to FcRI EC1, showed that models allowing >1 (M2a and M8) are as likely as models that do not (M1a, M7, and M8a; Fig. 5D). This result agrees with the pairwise analysis, suggesting that the observed substitutions at the two FcRI-binding positions in C3, positions 387 and 389, are not the result of positive selection.
In contrast, maximum likelihood analysis for the C2 domain revealed evidence for positive diversifying selection, there being a significantly increased likelihood for model M8 over M7 and M8a (Fig. 5D). The increased likelihood for the conservative model M2a over model M1a was marginal, but both M2a and M8 identified positively selected positions, of which position 319 was with a particularly strong support (M8: p > 0.95, M2a: p = 0.9) (Fig. 5A). This difference with the pairwise analysis, which indicated a marginal overall excess of synonymous substitutions, suggests that only a small set of positions in the domain are selected and that the majority accumulates synonymous substitutions. Consistent with this, M2a and M8 both predict a small fraction of selected residues: 7 and 12%, respectively (data not shown). When mapped onto the crystallographic structure of the FcRI IgA-Fc complex, the selected positions were seen to segregate to two different parts of the C2 domain: 245, 296, 326, 331, and 333 are close to the hinge region, whereas 317 and 319 are closer to the C2-C3 junction (Fig. 5E). Although residues 317 and 319 are near the site of interaction with FcRI, they are not in, or close to, sites interacting with FcRI (Fig. 5F). In conclusion, the positively selected positions in the C2 domain of IgA-Fc do not contact FcRI. These substitutions could influence FcRI binding indirectly or, alternatively, reflect selection imposed by IgA-Fc-binding proteins other than FcRI.
Evidence for pathogen-mediated selection on rodent and rabbit IgA-Fc
Another potential source of selective pressure on FcRI is bacterial decoy proteins like SSL7 of S. aureus, which compete with FcRI for binding to IgA (16, 17). Mutagenesis of IgA shows that the SSL7 and FcRI binding sites overlap and involve residues 257, 258, and 440–443 (41). SSL7 binds well to human, chimpanzee, and pig IgA, weakly to horse and rat IgA, and does not bind to either mouse or rabbit IgA (17). With a parsimony approach, we identified seven variable positions in IgA that discriminate the modes of binding to SSL7 (Fig. 6A). Substitution at positions 317, 319, 320, 325, and 442 distinguish IgAs that bind SSL7 from IgAs that do not; substitutions at positions 326, 346, and 442 distinguish strong from weak binding IgAs. Five of these positions are on the IgA surface near the C2-C3 junction, the inferred binding site for SSL7 (Refs. 17 and 41 and Fig. 6, B-D); whereas positions 325 and 326 are away from the junction and near the hinge. C3 442 is the only position where variation correlates with all three binding groups and mutation at this residue is known to affect human IgA binding to SSL7 (41). Furthermore, variation at 317 and 319 has been positively selected in higher primate IgA (Fig. 5A), raising the possibility that the pressure to change came from pathogen proteins like SSL7. Although there is no evidence for positive selection at positions 320, 346, and 442 in higher primates, reconstruction of ancestral IgA sequences is consistent with their selection in mouse and rabbit, species in which IgA has evolved resistance to SSL7 binding. Importantly, none of the residues predicted to give this resistance are ancestral: they all are recently evolved (Fig. 6E).
|
2 and C3 were examined for positive selection using branch-site maximum likelihood models (Fig. 7). For the C2 domain, there was no evidence of positive selection along the two branches studied (Fig. 7A, branches 1 and 2). For C3, the analyses revealed positive selection along the three branches where substitutions occurred that reduced IgA binding to SSL7 (Fig. 7, B and C, branches 1–3). These are branches leading to the rabbit ancestor (branches 1 plus 3) and to the rodent ancestor (branches 1 plus 2). Independent analyses of branches 1 and 3 showed evidence of positive selection for position 442, which reached significance when the two segments were analyzed jointly. Six other positions were positively selected on branch 3 (p > 0.9): positions 382, 384, and 441 are contact sites for FcRI, while positions 386, 390 and 392 are close to FcRI contact sites (Fig. 7C). The alanine to histidine 442 substitution that occurred on branches 1 plus 3 likely reduced or eliminated the interaction of IgA with SSL7 and FcRI, because mutating alanine 442 toarginine in human IgA abrogated its binding to SSL7 and FcRI (41, 42). In summary, the evidence is consistent with a model in which positively selected change at position 442 was accompanied or followed by an episode of positive selection that targeted the binding of IgA to FcRI, or possibly to another IgA-binding protein or receptor that uses the same sites as FcRI to contact IgA.
|
3 on branches 1 plus 2 leading to the mouse and rat ancestor parallels that observed for branches 1 plus 3 leading to the rabbit ancestor. There is evidence for positive selection on both pathways and it occurs at the same positions: 392, 397, and 441 (Fig. 7). Evidence for positive selection on position 442 in the rodent lineage was also obtained (p = 0.75) but was weaker than for the rabbit lineage (p = 0.95). After this episode of selection, the mouse and rat lineages split to evolve independently, with the result that rat IgA binds SSL7 weakly and for mouse IgA there is no detectable binding. Ancestral sequence reconstruction favors a model in which the rodent ancestor had asparagine at position 442, as in mouse, and that serine 442 of rat emerged after the split of the mouse and rat lineages (Fig. 7B). Because the emergence of asparagine 442 in the rodent ancestor likely decreased IgA binding to FcRI, as well as to bacterial decoy proteins like SSL7, the acquisition of serine 442 by rat IgA may have improved its affinity for FcRI. Such improvements were not possible for the mouse, because at some time during its evolution the FCAR gene was lost from the genome. The rabbit may also lack a functional FCAR gene, as all FCAR sequences in the draft rabbit genome appear to be pseudogenes (data not shown; see Materials and Methods).
This analysis has identified periods of positive selection in which variants of IgA emerged with reduced affinity for bacterial decoy proteins like SSL7. This involved positive selection for substitutions at positions in the C3 domain of IgA that correlate with IgA affinity for SSL7. This correlation suggests that pressure from pathogens has driven changes in IgA along the lineages leading to mouse and rabbit. Comparable analysis of SSL7 sequences from different bacterial strains shows that they too have diversified under positive selection (data not shown), consistent with models in which there is dynamic coevolution of bacterial decoy proteins with host IgA.
A cluster of positively selected residues on IgA-Fc identifies a potentially novel binding site
In the rabbit and rodent lineages, we see that substitutions at position 442 in the C3 domain of IgA were accompanied by episodes of positive selection. Reconstruction of ancestral sequences indicates that substitution of asparagine for alanine in the rodent ancestor was a reversion of the change from alanine to asparagine that occurred in the placental mammal ancestor (N442A, Fig. 7B). We therefore investigated whether this earlier change at position 442 had also been accompanied by an episode of positive selection, as would be evidenced by additional positively selected residues in the predicted IgA-Fc of the placental mammal ancestor (Fig. 7).
Branch-site models indicate positive selection acted on both C2 ( = 0.001; branch 3, Fig. 7A) and C3 ( = 0.05; branch 4, Fig. 7B) domains in the placental mammal ancestor. Seven positions were positively selected (p > 0.9), four of them with strong support (p > 0.99). Surprisingly, none of these positions is at, or close to, the binding site for FcRI (Fig. 8). Positions 301 in C2 and 422 in C3 are separated and locate to the two junctions of the two H chains, the former being a conserved cysteine in placental mammals that possibly forms a disulfide bond (43). In contrast, the other five residues are tightly clustered at the C2 surface, in a pattern characteristic of a binding site. No protein is known to bind to this site of IgA, suggesting it has either been a target of pathogen proteins other than SSL7 and its relatives or is the ligand for a mammalian IgA-binding protein other than FcRI. In this regard, the molecular basis for several putative IgA receptor activities on a variety of cell types has yet to be defined (44, 45).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
RI. Interaction of IgA-Fc with FcRI allows pathogen-specific IgA to opsonize the pathogen, facilitating its uptake and processing by phagocytes and dendritic cells (2). Subverting this aspect of host immunity are pathogen decoys, exemplified by the SSL7 protein of the bacterium S. aureus, which by binding to IgA-Fc prevents its engagement by FcRI. Thus, host FcRI and pathogen SSL7 compete for binding to IgA-Fc. In this situation, IgA or FcRI variants that favor their mutual interaction over IgA binding to SSL7 should confer host advantage, whereas mutations in SSL7 that increase affinity for IgA should confer pathogen advantage. The results of our analyses provide evidence for episodes of such selection.
Positive selection on FcRI in higher primates
During the 35 million years that higher primates have diverged from a common ancestor (46), we have obtained evidence for positive, diversifying selection on FCAR, which has led to differences in FcRI between new world monkeys, old world monkeys, and hominoids. Such selection is also evident among five chimpanzee FCAR alleles, but not among five human FCAR alleles (Fig. 3A). The targets for this positive selection were nine positions in EC1, the domain of FcRI that binds IgA-Fc. In contrast, the EC2 domain, which does not contact IgA-Fc, has been highly conserved by negative, purifying selection. Of the five EC1 positions giving the strongest evidence for positive selection, positions 55 and 85 make direct contact with IgA-Fc while positions 48, 61, and 65 are all in the vicinity of the binding site. This correlation infers that during the evolution of higher primates, new forms of FcRI were selected for IgA-binding properties that differed from those of existing forms. The strongest episode of positive selection we detected was on the evolutionary branch leading to the hominoid ancestor. During this same time period, the single IgA gene was duplicated to give daughter genes that subsequently evolved to give the modern and functionally differentiated IgA1 and IgA2 genes (40). This may have been a period during which IgA function expanded, incorporating changes into both IgA and FcRI. Emphasizing the episodic nature of the positive selections on FcRI, and the benefit of examining particular branches or time frames of mammalian evolution, was the absence of significant evidence for positive selection on FCAR when the mammals were analyzed as a group.
Positive selection on IgA-Fc in primates, rodents, and rabbit
Although both the C2 and C3 domains of IgA-Fc make contact with FcRI, C3 plays the major role and is more conserved. Although variability in higher primates occurs at two FcRI-contact residues of C3, neither position was positively selected. Conversely, good evidence for diversifying selection was obtained for the higher primate C2 domain. It is striking that all seven of the selected positions in C2 are situated away from the FcRI binding site. Thus, the positive selection detected on IgA did not come from direct pressure to improve interaction with FcRI. Substitution at two of the positions, 317 and 319, correlates with differential binding to the bacterial decoy protein SSL7, which binds IgA, prevents its association with FcRI and subverts the host immune response. This correlation suggests that pathogen-driven selection has diversified higher primate C2 domains. For the other IgA positions implicated in SSL7 binding, including the critical position 442, there was no evidence for positive selection in higher primates.
Further evidence for pathogen-driven selection on IgA-Fc was obtained from the rodent and rabbit lineages. In these species, substitution at positions 346 and 442 in C3 has reduced or eliminated the binding of IgA to SSL7. In addition to positions 346 and 442, most of the sites affected by these episodes of positive selection are in or near the FcRI binding site, suggesting that these changes also reduced the binding of IgA to FcRI as well as its interaction with SSL7. Consistent with this evolution, mouse IgA does not bind human FcRI (47) and this has been correlated with changes at C3 positions 441 and 442 (42).
Pathogen-mediated selection on IgA may have led to loss of FCAR from the mouse genome
The results of our study suggest a two-stage model for the pathogen-driven evolution of IgA-Fc and FcRI (Fig. 9). The first stage represents a cycle of coevolution involving successive adaptations that provide temporary advantage first to the pathogen and then to the host, but with completion of the cycle they are brought back to where they started (48). The second stage shows how the cycle can break in two alternative ways, one of which can lead to complete and irrevocable loss of FcRI function, the circumstance that now pertains to the mouse.
|
RI. The competition between decoy and FcRI imposes pressure on the host, selecting a variant form of IgA that has no affinity for the decoy while retaining some functional, affinity for host FcRI. After selection of the variant IgA, its function can be improved by selection of a variant FcRI that has a higher affinity for the new IgA than did the previously dominant FcRI form. Emergence in the host of a novel IgA-FcRI combination that is resistant to the decoy imposes pressure on the pathogen. This selects for a variant decoy having sufficient affinity for the new form of IgA that prevents IgA binding to FCAR and compromises host immunity. Thus begins the cycle anew.
Because FcRI and the decoy compete for binding to IgA, each round of the coevolutionary cycle will tend to increase the similarity in the sites on IgA-Fc that interact with FcRI and the decoy. Such convergence reduces the potential for host IgA variants to reduce significantly affinity for the decoy without losing functional binding to FcRI. Stage 2 of the model considers what might happen when this situation arises and considers two possible outcomes. In the first outcome the host species evades the decoy through selection of a variant IgA that has no binding to either the decoy or FcRI. After such selection the FCAR gene becomes effectively nonfunctional, a situation in which loss of the FCAR gene or its degradation to a pseudogene would be selectively neutral. Such a progression can explain how the FCAR gene was deleted from the mouse genome and inactivated in the rabbit genome. Such events could occur through chance or as a consequence of selection on a linked gene. The FCAR gene is present in a region dense with families of immune system genes that are prone to gene duplication and deletion (49), features that increase the likelihood of deletion either through chance or selection on a linked gene. Supporting such a model, analysis of the draft genome of the dog (Canis familiaris) revealed a partial FCAR gene adjacent to Nkp46 (accession number for the genomic segment: AAEX02033729; our unpublished observations) with features indicating that dog FCAR is a pseudogene. Thus, the loss of FCAR appears a common occurrence in mammals.
In the second outcome, the host does not acquire a new IgA variant and the cycle is broken at a point where FcRI function is retained but compromised by the decoy. Between decoy and FcRI a status quo is maintained. If other host mechanisms are able to control or clear the infection, then the host species will survive. This possibly corresponds to the situation in higher primates, where SSL7 binds to human, chimpanzee, and baboon IgA (17), and there is no evidence of selective pressure on the IgA sites that contact FcRI.
The IgA-FcRI interaction in mammals is seen to be a frequent target of natural selection, which contributes to the plasticity of the system. This plasticity is however matched by the pathogens. Several characteristics of the pathogen decoy proteins indicate that targeting the IgA-FcRI is advantageous for the bacteria. For example, the episode of positive selection observed in the evolutionary lineage leading to S. aureus SSL7 sequences, or the fact that unrelated proteins of S. pyogenes (group A streptococcus), group B streptococcus, and S. aureus evolved to contact IgA using the same area, or even the same sites as FcRI (16, 41). Similarly to the loss of FcRI in mouse, dog, or rabbit, at least one strain of S. aureus (COL) lacks SSL7 (50), suggesting that these bacteria can survive and propagate without it. Such capacity of the pathogens to match or even exceed the plasticity of the host is particularly striking in rabbits, whose IgA-FcRI system underwent major changes, including expansion of the IgA locus, loss of FcRI function, and positive selection on IgA C3. Although it is unknown whether these changes in the rabbit were caused by S. aureus SSL7 or by another pathogen protein with a similar binding pattern for IgA, it should be noted that modern rabbits are not immune to S. aureus and epidemics of high-virulence strains can lead to death (51).
Evolution of IgA receptors
Ab-mediated immunity at mucosal surfaces is provided by IgA, a function that is dependent upon transcytosis by the polymeric Ig receptor. IgA and polymeric Ig receptor are both characteristic of birds and mammals (52), whereas FcRI being restricted to mammals is of more recent origin. This phylogeny and the presence of relatively little serum IgA in birds suggest IgA principally evolved to serve mucosal immunity and only later developed as an important Ig of blood, lymph, and tissue fluids. The current consensus view is that FcRI mainly functions to bind serum IgA, which in humans is predominantly monomeric compared with the dimers of secreted IgA (2). In this scenario, the origin of FcRI can be seen as an important event in the emergence of IgA as a more effective serum Ig. However, it is also possible that serum IgA was a functioning component of the immune system before the origin FcRI. Because Ab effector function usually depends upon cellular FcRs, it is therefore plausible that another leukocyte IgA-FcR, one which has yet to be well characterized, evolved before FcRI and still functions today. In the mouse, which is a natural knockout for the FCAR gene, this other IgA-Fc receptor could compensate for the absence of FcRI and help explain the animals’ viability. Indirect evidence obtained here for a second receptor is the positively selected cluster of residues we identified on the C2 domain of IgA-Fc (Fig. 8). This cluster looks like a binding site, a candidate ligand for another soluble or cell surface factor that binds to IgA-Fc. Several putative cellular receptors for IgA have been reported (44, 45), but have yet to be defined at the molecular level and provide candidate receptors for this orphan ligand.
|
Acknowledgments |
|---|
|
Disclosures |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
|
Footnotes |
|---|
1 This study was supported by National Institutes of Health Grants AI031168 and AI024258 (to P.P.). P.J.N. was a Lymphoma Research Foundation Fellow. 
2 Address correspondence and reprint requests to Dr. Peter Parham, 299 Campus Drive West, Fairchild Building, Stanford University, Stanford, CA 94305. E-mail address: peropa{at}stanford.edu
3 Abbreviations used in this paper: LRC, leukocyte receptor complex; KIR, killer-cell Ig-like receptor; dS, rate of synonymous substitution; dN, rate of nonsynonymous substitution; LRT, likelihood ratio test; NJ, neighbor-joining; SSL7, staphylococcal superantigen-like protein 7.
Received for publication February 23, 2007. Accepted for publication April 9, 2007.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
RI as an inhibitory receptor that controls inflammation: dual role of FcR ITAM. Immunity 22: 31-42. [Medline] receptors: old friends and new family members. Immunity 24: 19-28. [Medline]RI and its complex with IgA1-Fc. Nature 423: 614-620. [Medline]RI. J. Biol. Chem. 278: 27966-27970. RI (CD89) and bovine Fc2R are located in their membrane-distal extracellular domains. J. Exp. Med. 189: 1715-1722. receptor essential for interaction with IgA. J. Immunol. 162: 2146-2153. 
prime;dS) tailed Z test. The significance level (
strands in the secondary structure (
