Introduction

Leukocyte interaction with the vascular endothelium has a central role in the development of inflammatory and immune responses. Thus, adhesion molecules involved in leukocyte trafficking have received considerable interest, especially in relation to the development of human diseases. The most well-studied molecules are the selectins, ICAM1 (intercellular adhesion molecule 1) and VCAM1 (vascular cell adhesion molecule 1).

Selectins are a family of C-type lectins, and humans express three family members, namely P-, E- and L-selectin, encoded by genes (SELP, SELE and SELL) located in cluster on chromosome 1. By binding to their ligands, selectins have a major role in leukocyte tethering and rolling on activated epithelia, thus favouring their extravasation and migration to inflamed tissues. The prefix for the three selectins carries a hint at their main expression site: L- and E-selectins are expressed on most leukocytes and on activated endothelial cells, respectively, whereas P-selectin is stored in platelet alpha granules and endothelial Weibel-Palade bodies (Ley, 2003). The three selectins have a common ligand, encoded by SELPLG. Exposure to different pro-inflammatory and pro-thrombotic stimuli determines the rapid translocation of SELP to the cell surface where, through interaction with SELPLG, it promotes adhesion and rolling of both leukocytes and platelets, as well as their heterotypic aggregation (Ley, 2003). Inflammatory cytokines also induce E-selectin transcription, whereas L-selectin is constitutively expressed (Ley, 2003). Experiments in knockout mouse models, as well as data from rare human conditions where leukocyte binding to selectins is impaired, indicate that SELP, SELL and SELE have a central role in the trafficking of innate immunity cells, T lymphocytes, and platelets (Ley, 2003).

ICAM1 and VCAM1 belong to the immunoglobulin superfamily; inflammatory cytokines also trigger the expression of both ICAM1 and VCAM1 on endothelial cells to increase leukocyte adherence and migration (Golias et al., 2007).

Selectins, as well as ICAM1 and VCAM1, also exist in soluble forms. The circulating amount of P-selectin and ICAM1 is influenced by polymorphisms located at their respective loci, and by variants in ABO (Miller et al., 2004; Reiner et al., 2008; Barbalic et al., 2010).

Leukocyte extravasation and tissue infiltration are necessary for efficient immune response against invading pathogens; yet, these processes may also result in tissue damage and promote chronic inflammatory states. These observations suggest on one hand that genes encoding adhesion molecules may represent natural selection targets, as is the case of several immune response loci (Sironi and Clerici, 2010); on the other hand, genetic variation at these loci might protect from or predispose to autoimmune or chronic inflammatory conditions. Indeed, polymorphisms in SELP and SELE have been associated with susceptibility to systemic lupus erythematosus (Jacob et al., 2007), whereas variants in SELL and SELE predispose to IgA nephropathy (Watanabe et al., 2006). SELP also represents a risk locus for atopy (Bourgain et al., 2003).

The increasing availability of large-scale genetic diversity data, which are being produced by next-generation sequencing platforms, opens unprecedented possibilities to the study of natural selection signatures along extended genomic regions. Still, the nature of these data also requires the development of new strategies to infer selective patterns. Here, we aimed to provide a description of the evolutionary history of the selectin gene cluster, by integrating low-coverage genome-wide resequencing data with classic population genetics analyses. Similarly, the genes encoding ICAM1, VCAM1 and SELPLG were analyzed. Results indicate that a complex selective scenario in distinct human populations has shaped the genetic diversity of the SELP gene regions where autoimmune disease risk alleles are located.

Materials and methods

Samples and sequencing

Human genomic DNA from HapMap subjects (20 individuals of Asian descent) was obtained from the Coriell Institute for Medical Research. All analyzed regions were PCR-amplified and directly sequenced; primer sequences are available upon request. PCR products were treated with ExoSAP-IT (USB Corporation, Cleveland, OH, USA), directly sequenced on both strands with a Big Dye Terminator sequencing Kit (v3.1 Applied Biosystems, Foster City, CA, USA) and run on an Applied Biosystems ABI 3130 XL Genetic Analyzer (Applied Biosystems). Sequences were assembled using AutoAssembler version 1.4.0 (Applied Biosystems), and inspected manually by two distinct operators. The genomic DNA of nine Pan troglodytes was obtained from the Gene Bank of Primates, Primate Genetics, Göttingen, Germany (http://dpz.eu/index.php). These samples have been shown to belong to the Pan troglodytes verus subspecies (Cagliani et al., 2012). Primer pairs are available upon request.

Population genetic analyses

Tajima’s D (Tajima, 1989), Fu and Li’s D* and F* (Fu and Li, 1993) statistics, as well as diversity parameters θW (Watterson. 1975) and π (Nei and Li. 1979) were calculated as previously described (Fumagalli et al., 2009). Calibrated coalescent simulations were performed using the cosi package (Schaffner et al., 2005) and its best-fit parameters for African American (AA), CEU and Asian (AS) populations with 10 000 iterations. Coalescent simulations were conditioned on mutation rate and recombination rate was derived from University of California Santa Cruz (UCSC) tables (http://genome.ucsc.edu/, snpRecombRateHamap table).

The maximum-likelihood-ratio Hudson–Kreitman–Aguadé (MLHKA) test was performed using the MLHKA software (Wright and Charlesworth, 2004), as previously proposed (Fumagalli et al., 2009). For human populations, 16 reference loci were randomly selected among loci shorter than 20 kb that have been resequenced in AA, CEU and AS by The National Institute of Environmental Health Sciences (NIEHS); the only criterion was that Tajima's D did not suggest the action of natural selection (that is, Tajima’s D is higher than the 5th and lower than the 95th percentiles in the distribution of NIEHS genes). As for chimpanzee, the MLHKA was performed using the 16 resequenced regions as reference loci.

The DIND (derived intra-allelic nucleotide diversity) test was performed as previously proposed (Barreiro et al., 2009). Significance thresholds for each MAF interval were computed via coalescent simulations incorporating a demographic model (Schaffner et al., 2005).

Genotype data for 5 kb regions from 238 resequenced human genes were derived from the NIEHS SNPs Program web site (http://egp.gs.washington.edu). In particular, we selected genes that had been resequenced in populations of defined ethnicity including CEU, AA and AS (NIEHS panel 2). The iES (integrated extended haplotype homozygosity) score was calculated as previously described (Tang et al., 2007).

Details on FST statistic calculation among continental groups in the HGDP-CEPH panel are available as Supplementary Methods. Genotype information was obtained for the SELP regions and for 2000 randomly selected RefSeq genes. Sliding-window analysis was performed on overlapping 20 single-nucleotide polymorphism (SNP) windows moving with a step of 5 SNPs.

Haplotype analysis and TMRCA calculation

Haplotypes were inferred using PHASE (version 2.1) (Stephens and Scheet, 2005). Details on haplotype analysis and TMRCA calculation are available as Supplementary Methods.

Minigene construction, transfection and transcript analysis

The genomic DNA of NA10851 HapMap subject (heterozygous for haplotypes belonging to the two main clades) was used as a template to PCR amplify a genomic fragment encompassing SELP exons 11–16. PCR amplifications were performed with high-fidelity polymerase (Pfu DNA Polymerase, Promega Corporation, Madison, WI, USA). The ∼6500 bp PCR products were cloned in the pDisplay expression vector (Invitrogen, Carlsbad, CA, USA) to generate the pDisplay-SELP-HapA and pDisplay-SELP-HapB constructs. The two plasmids were checked by direct DNA sequencing using primers located on the pDisplay vector (HA-For: 5′-GATGTTCCAGATTATGCTG-3′; MYC-Rev: 5′-CAGATCCTCTTCTGAGATG-3′).

HeLa cells were maintained in DMEM medium supplemented with 10% fetal-calf serum and transfected using Lipofectamine-2000 (Invitrogen) in accordance with the manufacturer’s instructions. Cells were harvested 48 h after transfection and total RNA was extracted by TRIzol reagent (Invitrogen), following the manufacturer’s protocols. Purified RNA was reverse transcribed to cDNA using random hexamers and Ready-To-Go You-Prime First-Strand Beads (GE Healthcare, Little Chalfont, Buckinghamshire, UK). RT–PCR was performed with SELP primers located in exon 12 and 16. PCR products were separated on a 2% agarose gel.

Results

Sliding-window analysis of the selectin cluster

In humans, three selectin genes (SELP, SELL and SELE) are located in a ∼150 kb region on chromosome 1. In order to analyze population genetic differentiation along the selectin cluster, we exploited data from the 1000 Genomes Pilot project deriving from the low-coverage whole-genome sequencing of 179 individuals with different ancestry (1000 Genomes Project Consortium et al., 2010). In particular, individuals from three distinct ethnic groups have been sequenced: CEU, Yoruba from Nigeria (YRI) and Japanese plus Chinese (East Asian, AS). Using these data we calculated FST (Wright, 1950) in sliding-windows moving along the chromosome 1 genomic region where SELE, SELL and SELP are located. As an empirical comparison, the same procedure was applied to 2000 randomly selected human genes, so as to calculate percentiles of genetic differentiation for pairwise population comparisons. As shown in Figure 1, prominent peaks in FST occur within the SELP gene region. Specifically, AS/YRI population genetic differentiation exceeds the 95th percentile in two regions covering exons 9–10 and 11–13. In these two regions FST is also high for CEU/AS and YRI/CEU comparisons (Figure 1). The 1000 Genomes Pilot project data have an advantage over other data sets (for example, HapMap and the HGDP-CEPH panel) because they do not rely on the genotyping of selected SNPs, resulting in a less-marked bias towards intermediate frequency variants. Still, these data suffer from lack of power for rare variants (1000 Genomes Project Consortium et al., 2010), and possible biases within repetitive elements (Cagliani et al., 2012). Thus, we repeated FST analysis using both HapMap data and genotype information from the HGDP-CEPH panel: we obtained results very similar to those from the 1000 Genomes Pilot project data (Figure 1 and Supplementary Figure 1). These were also confirmed by FST analysis using Sanger resequencing data for SELP, SELL and SELE from the SeattleSNPs Variation Discovery Resource (Supplementary Figure 2). Finally, we calculated the iES (Tang et al., 2007) along the selectin cluster without identifying regions with unusual score (Supplementary Figure 3). These results warrant further analysis on the selection pattern of the two SELP regions, hereafter referred to as SELPex9–10 and SELPex11–13.

Figure 1
figure 1

Sliding-window analysis of FST along the selectin cluster. (a) Data from the 1000 Genomes Pilot Project were used to calculate FST in sliding windows of 20 SNPs moving along the selectin gene cluster with a step of 5 SNPs. Color codes refer to population comparisons: red, YRI/CEU; blue, YRI/AS; green, CEU/AS. Horizontal dashed lines represent the 95th percentile in the distribution of FST calculated for sliding windows deriving from 2000 randomly selected human genes. (b) The location of SNPs genotyped in the HGDP-CEPH panel is reported with the value of FST calculated across continental groups. The horizontal line corresponds to the 95th percentile in the distribution of all SNPs genotyped in the panel.

Full size image

Sliding-window analysis of SELPLG, ICAM1 and VCAM1

The same approach described above was applied to analyze FST along SELPLG, ICAM1 and VCAM1. No unusual FST value was observed in ICAM1 and VCAM1 using data from the 1000 Genomes Pilot project; conversely a region of high genetic differentiation between YRI and AS was observed in the 5′ portion of SELPLG (hereafter referred to as SELPLG5′) (Figure 2). High genetic differentiation was not detected using the HGDP-CEPH panel data (Figure 2), possibly because of the small number of SNPs genotyped in SELPLG; similarly, analysis of HapMap SNPs (Supplementary Figure 4) revealed no unusual FST value in the gene, nor did data derived from SeattleSNPs (Supplementary Figure 5). Yet, in this latter case the analysis was restricted to CEU and AA, as no AS individual was resequenced by the Program.

Figure 2
figure 2

Sliding-window analysis of FST for SELPLG, ICAM1 and VCAM1. FST calculation and color codes as in Figure 1. (a) Data from the 1000 Genomes Pilot Project. (b) Data from the HGDP-CEPH panel.

Full size image

Population genetics of SELP and SELPLG in human populations

In order to perform further population genetics analyses we turned to Sanger resequencing data. As mentioned above, the SELP and SELPLG transcription units have been fully resequenced in CEU and AA by the Seattle SNPs Variation Discovery Resource. As for AS, we resequenced SELPex9–10, SELPex11–13 and SELPLG5′ in 20 HapMap AS individuals. These data were used to calculate θW, an estimate of the expected per site heterozygosity (Watterson, 1975) and π, the average number of pairwise sequence nucleotide differences between haplotypes (Nei and Li, 1979). In order to compare the values we obtained for the SELP and SELPLG regions, we calculated θW and π for 5 kb windows (hereafter referred to as reference windows) deriving from 238 genes resequenced by the NIEHS Program in the same populations; the percentile rank corresponding to the SELP and SELPLG regions in the distribution of reference windows is reported in Table 1. No exceptional θW or π value was obtained for SELPex9–10 and SELPLG5′ in the three populations. Conversely, both indexes were extremely high in SELPex11–13 in AA and CEU, although in this latter population the rank of π did not reach the 95th percentile. Peaks of nucleotide diversity in this region were also obtained for YRI using a sliding-window approach with the 1000 Genomes Pilot Project data (Supplementary Figure 6)

Table 1 Nucleotide diversity and MLHKA test for the two analyzed SELP and SELPLG regions
Full size table

High nucleotide diversity might be suggestive of balancing selection, as neutral variants tend to be maintained with the selected allele(s). In order to formally rule out the possibility that the excess of nucleotide diversity is due to higher local mutations rates, we applied a MLHKA test (Wright and Charlesworth, 2004). This is based on the concept whereby, under neutral evolution, the amount of within- and between-species diversity is expected to be similar for all loci in the genome. Therefore, the test compares polymorphism and divergence levels for a region of interest with those calculated for other neutrally evolving genomic segments. Sixteen reference loci were used for the analysis (Fumagalli et al., 2009), and the results for SELPex11–13 indicated a significant excess of polymorphism compared with divergence in all populations (Table 1). No deviation from neutrality was observed for the SELPex9–10 and SELPLG5' regions (Table 1).

As mentioned above, natural selection acting on specific gene regions can determine a distortion in the site frequency spectrum. Common neutrality tests based on the SFS include Tajima’s D (DT) (Tajima, 1989), and Fu and Li’s D* and F* (Fu and Li, 1993). DT tests the departure from neutrality by comparing θW and π, and positive values indicate an excess of intermediate frequency variants. Fu and Li’s F* and D* are also based on the site frequency spectrum (SFS) and differ from DT, in that they also take into account whether mutations occur in external or internal branches of a genealogy. As population history, in addition to selective processes, is known to affect the SFS, we evaluated the significance of neutrality tests by performing coalescent simulations that incorporate demographic scenarios (see methods) (Schaffner et al., 2005). As explained above, we also applied an empirical comparison by calculating the percentile rank of DT, F* and D* for the SELP and SELPLG gene regions relative to reference windows. No deviation from neutrality was observed in AA, whereas DT and F* were significantly high in SELPex9–10 for CEU (Table 2). In this population, D* was extremely high in SELPex11–13. As for AS, SFS-based statistics tended to display low values in both SELP regions (Table 2). These data were confirmed by running coalescent simulations with different demographic models (Supplementary Table 1). No departure from neutrality was observed for SELPLG5′.

Table 2 SFS-based statistics for the analyzed SELP and SELPLG gene regions
Full size table

SELP haplotype analysis and selective pattern in distinct populations

The power of SFS-based statistics to reject neutrality depends on several factors such as the timing and strength of the selective event, and complex selective patterns may leave signatures that are difficult to detect. Thus, to gain further insight into the evolutionary history of SELPex11–13, we performed haplotype analysis using two approaches: a neighbor-joining network (Bandelt et al., 1999) and a maximum-likelihood coalescent method implemented in GENETREE (Griffiths and Tavare, 1995). This latter indicated that the haplotype genealogy of SELPex11–13 is split into two major clades (clades A and B) with a coalescence time of 3.51 million years (MY) (SD: 0.569 MY) (Figure 3). Both clades are further split into relatively deep subclades, one of them, in clade A, showing a TMRCA of more than 2.2 MY. As shown in the haplotype network (Figure 3), no AS chromosome contributes to clade B, whereas all other major clade A haplotypes are represented across all populations. These analyses, and the deep coalescence time of SELPex11–13 haplotypes suggest that this region has been a target of multiallelic balancing selection in CEU and AA. Conversely, the low diversity levels observed in AS for the two SELP regions indicate that a selective sweep may have occurred in this population. Indeed, calculation of Fay and Wu's H revealed significantly negative values for both SELPex9–10 (H=−17.83, P<0.01, Table 2) and SELPex11–13 (H=−6.93, P=0.01, Table 2). In order to further explore this possibility, we applied the DIND test (Barreiro et al., 2009). This is based on the ratio of intra-allelic diversity associated with the ancestral and derived alleles (iπA/iπD) plotted against the frequency of the derived allele (Figure 4): a high value of iπA/iπD for variants with high frequency of the derived allele is suggestive of positive selection, as the neutral diversity associated with the derived allele is limited despite its high frequency in the population. We applied the DIND test to SELPex9–10 and SELPex11–13; again, statistical significance was calculated by coalescent simulations that incorporate a demographic model (see methods). Significantly, high values of DIND were obtained in SELPex9–10 for rs3917750, rs3917751, rs3766126, rs2223303, rs2205894 and rs3917777 (Figure 4). These SNPs define a haplotype with a frequency of 0.95, 0.45 and 0.10 in AS, CEU and AA, respectively. Visualization of haplotypes in this region indicated that the derived alleles of the 5 SNPs reported above define a set of highly homogeneous haplotypes in both CEU and AS, as also assessed by calculation of nucleotide diversity for the distinct haplotype groups (Figure 4).

Figure 3
figure 3

Haplotype analysis for SELPex11–13. (a) GENETREE: mutations are represented as black dots. The absolute frequency of each haplotype is reported. (b) Median joining network: each node represents a different haplotype, with the size of the circle proportional to frequency. Branch lengths are proportional to the number of nucleotide differences. Circles are color-coded according to population (yellow: AA, blue: CEU, red: AS). The most recent common ancestor (MRCA) is also shown (black circle). Nonsynonymous substitutions are shown in red and indicated by the arrows.

Full size image
Figure 4
figure 4

Haplotype analysis for SELPex9–10. (a) The DIND test was performed for AS. P-values were calculated through coalescent simulations that incorporate demographic scenarios (Schaffner et al., 2005). The continuous, dashed and dotted lines indicate the median, the 95th and the 99th percentiles, respectively. (b) Visualization of SELPex9–10 haplotypes; polymorphic positions are color-coded according to their allelic state (dark gray: derived, light gray: ancestral). The position of the 5 SNPs identified through the DIND test are shown and inferred haplotypes are ordered on the basis of their allelic status.

Full size image

The soluble form of P-selectin is thought to result both from alternative splicing of the transmembrane domain and from shedding of the membrane-bound form (Vestweber and Blanks, 1999). Given that the transmembrane domain is encoded by exon 14, we wished to verify whether the two major haplogroups we observed for SELPex11–13 carry variants with an effect on exon 14 inclusion in SELP transcripts. To this aim, we constructed SELP minigenes. In particular, DNA fragments spanning exons 11–16 were cloned in an expression vector. Two distinct fragments were cloned, deriving from chromosomes carrying clade A or clade B haplotypes. RT–PCR amplification of SELP transcripts isolated from HeLa cells transfected with both minigenes generated similar amounts of two products: a 433 bp transcript deriving from the full-length SELP and a 323 bp PCR fragment corresponding to exon 14 skipping. Albeit not quantitative, these results suggest that exon 14 skipping is not associated with a specific haplotype.

Population genetics of SELP in chimpanzee

To analyze SELP evolutionary history in Pan troglodytes, we resequenced SELPex9–10 and SELPex11–13 in nine unrelated chimpanzees. No polymorphic variant shared with humans was observed. The total number of segregating sites was 39 and 33 in SELPex9–10 and SELPex11–13, respectively; their analysis indicated that most nonsynonymous substitutions (five out of six) occur in exons 12 and 13 (Supplementary Table 2). In order to assess whether the two SELP regions show unusual levels of nucleotide variability, we resequenced 16 genomic regions in these same individuals to be used as an empirical comparison (Cagliani et al., 2012). Calculation of θW that SELPex9–10 and SELPex11–13 show by far the highest values of diversity compared with the 16 reference regions (Figure 5); as for π, the value for SELPex11–13 was also the highest. This result was confirmed by application of the MLHKA test, which yielded significant results (k=3.81 and 7.72, P=0.0021 and 5.41 × 10−6 for SELPex9–10 and SELPex11–13, respectively). Calculation of SFS statistics indicated that both SELP regions display the lowest negative values for DT, D* and F* (Supplementary Table 3). Further analysis indicated that variants in SELPex9–10 and SELPex11–13 are in tight linkage disequilibrium in the chimpanzee sample (Supplementary Figure 7), and one single individual carries a haplotype, which is highly divergent from those observed among the remaining eight primates. This translates in a haplotype genealogy with extremely long basal branches and a deep coalescent time. GENETREE analysis yielded a maximum-likelihood estimate of θ (θML) of 9.2, resulting in an effective population size (Ne) of 14081. Using this method, the TMRCA of the chimpanzee SELP haplotype lineages amounted to 3.52 MY (SD: 424 KY).

Figure 5
figure 5

Nucleotide diversity and haplotype analysis in chimpanzee. (a) Plot of θW and π values for the 16 regions we resequenced (white circles) and for the two SELP regions (black). (b) GENETREE analysis of the SELP region covering exons 9–13: mutations are represented as black dots and the absolute frequency of each haplotype is reported.

Full size image

Discussion

In this study, we analyzed the evolutionary history of genes encoding major molecules involved in leukocyte–epithelium interaction. A previous analysis of immune response genes with available sequences in public databases (Casals et al., 2011) included ITGAL and ITGA2, encoding leukocyte integrin molecules also involved in white blood cell trafficking; the two genes were found to be neutrally evolving. To our knowledge, no other study has addressed the evolutionary history of adhesion molecules involved in leukocyte trafficking in human populations. Here, we exploited the availability of the 1000 Genomes Project low-coverage data to perform a sliding-window analysis of FST along a ∼150 kb region encompassing SELL, SELE and SELP. Peaks of significantly high population genetic differentiation were identified through an empirical approach based on the comparison with a large number of human genes, and were restricted to two regions in SELP where several amino-acid changes are located. A region in SELPLG also showed high population genetic differentiation. Extremely high values of FST strongly suggest the action of local adaptation, as allele frequencies are driven to differ more than that would be expected on the basis of drift or demography alone. Thus, we focused on these regions, and, in order to overcome the possible biases in the SFS, which are introduced by next-generation sequencing approaches (1000 Genomes Project Consortium et al., 2010), we analyzed Sanger sequencing data to infer selection patterns. Results indicated that the genomic region covering SELPex11–13 displays high nucleotide diversity in Africans and CEU, and a higher level of within-species diversity compared with inter-specific divergence. Analysis of inferred haplotypes in this region revealed the presence of two deeply separated clades with a TMRCA of ∼3.5 MY. Coalescence times for neutrally evolving autosomal human loci range between 0.8 and 1.5 MY (Tishkoff and Verrelli, 2003); thus, the TMRCA of the SELP haplotype clades is much deeper than expected under neutrality. Indeed, clade A is further split into two lineages that coalesce at ∼2.2 MY (Figure 3). Overall, these data are consistent with a model of multiallelic balancing selection and may explain the values of SFS-based statistics (Tajima's D and Fu and Li’s F* and D*), which are not strikingly positive in SELPex11–13, as the skew toward intermediate frequency variants tends to be less marked in a multiallelic selection model than in the case of biallelic selection (Navarro and Barton, 2002). In this respect, it is worth mentioning that, although SELPex9–10 showed high values for DT and F*, these features are not accompanied by a local increase in nucleotide diversity (which is a hallmark of balancing selection (Charlesworth. 2006)), suggesting that the balancing selection target(s) is located in SELPex11–13, and that linkage disequilibrium might cause the distortion of the SFS in exons 9–10. As for SELPLG, despite the high population genetic differentiation identified in the initial scan, no test suggested departure from neutrality, confirming the previous indication that high FST alone is not sufficient to demonstrate natural selection (Xue et al., 2009).

Network analysis of SELPex11–13 indicated that distinct haplotypes have different frequencies in the three human populations we analyzed, with AS chromosomes being absent from clade B; this explains the limited nucleotide diversity observed for this population, and suggests a different selective scenario for SELP in Asia (see below). Analysis of SNPs located on the branches of the haplotype network indicated that the Val640Leu (rs6133) polymorphism separates the two major haplotype clades and might therefore represent a selection target. The derived 640Leu allele, which is common in populations of African descent, has previously been associated with atopy, thrombo-embolic stroke and systemic lupus erythematosus, suggesting that it might affect some properties of SELP, which are relevant to immunologic/inflammatory functions (Bourgain et al., 2003; Zee et al., 2004; Morris et al., 2009). The variant has also been shown to modulate circulating levels of soluble P-selectin in some studies but not in others (Reiner et al., 2008; Barbalic et al., 2010). As for the two subclades within A group haplotypes, all variants separating the two branches are non-coding, suggesting that the selection target might be represented by one or more regulatory variants. Two nonsynonymous variants, namely Thr756Pro (rs6136) and Asp603Asn (rs6127), differentiate major haplotypes within clade A (Figure 3). The position of these variants in the haplotype genealogy indicates that they do not represent selection targets or, at least, that the selective pressure acting on them is extremely weak or recent. Nonetheless, these variants have been the subjects of intense investigation, as they were reported to affect cardiovascular functions in humans, although results have often been inconsistent among studies and in populations with different ethnic origin (Tregouet et al., 2002; Miller et al., 2004; Volcik et al., 2007). As already suggested by some authors (Tregouet et al., 2002), results herein clearly show that haplotype-based association analysis, rather than single SNP genotyping, should be applied when SELP genetic variation is to be associated with human phenotypic traits. Indeed, the SELP haplotype structure is complex and coding variants (in addition to putative regulatory variants) define distinct major haplotypes. These have extremely different frequencies in distinct human populations, also resulting in diverse statistical power to detect associations depending on sample ethnicity.

Given that a fraction of soluble P-selectin is thought to result from alternative splicing of exon 14 (Vestweber and Blanks, 1999), we also tested whether the two major haplotypes encompassing exons 11–13 might affect exon inclusion. SELP transcript levels in lymphoblastoid cell lines were low (not shown), not allowing quantification of the alternatively spliced product. For this reason we adopted a minigene approach with transfection in HeLa cells. On the one hand, this allows splicing to occur for the two haplotypes in the same conditions and genetic background, therefore removing possible confounding effects. On the other hand, a major drawback of this approach is that the minigene is expressed from an exogenous promoter, and increasing evidences suggest that polymerase elongation rate and promoter elements have an impact on alternative splicing regulation (Kornblihtt et al., 2004). Under the experimental conditions we used, the two SELP haplotypes yielded similar amounts of full-length and alternatively spliced transcripts, suggesting no major effect of SELPex11–13 variants in exon 14 inclusion. However, further analyses will be required to determine whether promoter elements, environmental cues or biochemical stimuli affect the haplotype-specific splicing of SELP.

As mentioned above, striking differences in allele/haplotype frequencies, which are reflected in high FST values, are commonly regarded as a signature of local adaptation. Our analysis strongly suggest that while genetic diversity has long been maintained in African and CEU populations owing to a balancing selection regime, a SELP haplotype has been driven to high frequency in AS populations as a result of positive selection. Indeed, AS populations show reduced diversity at the two analyzed SELP regions, especially SELPex9–10, and neutrality tests highlight an excess of low-frequency alleles. Calculation of Fay and Wu (2000), which is sensitive to the presence of high-frequency-derived alleles, yielded a strongly negative result in the region encompassing exons 9–10. Consistently, application of the DIND test confirmed the hypothesis of positive selection for SELP and identified a haplotype, which has been driven to a frequency of 0.95 in AS whereas not reaching 0.5 in either CEU or AA. In both AS and CEU populations, a set of highly homogeneous haplotypes is defined by the derived alleles of the 5 SNPs we identified through the DIND test, whereas haplotypes carrying ancestral alleles at these loci display much higher diversity (Figure 4). It is worth mentioning that the two SELP regions we analyzed are located nearby, and therefore cannot be thought of evolving independently. In principle, the loss of the balanced polymorphism at SELPex11–13 in AS might explain the reduced diversity observed at SELPex9–10 in this population; yet, this scenario is not expected to result in the presence of a set of highly homogeneous haplotypes. Thus, we suggest that the selection targets in the two SELP regions have been competing, and that the selective sweep in the region encompassing exons 9–10 has determined the loss of the balanced SNP(s) in AS and reduced its frequency in CEU. In this respect, it is also worth mentioning that no significant iES score was detected in the region encompassing SELPex9–10; yet, simulation analyses have indicated that DIND has more power than haplotype-based tests for frequencies of the selected haplotype ranging from 10 to 98% (Barreiro et al., 2009). Variants defining the SELPex9–10 high frequency haplotype are all non-coding and inspection of available resequencing databases (for example, 1000 Genome Project; NIEHS Exome Project) indicated that they are not in linkage disequilibrium with any coding variant. Thus, the selection target in AS populations might be represented by a regulatory variant that possibly affects SELP expression.

Overall, these data suggest that distinct human SELP variants have been targeted by natural selection possibly as a response to different environmental pressures. In an attempt to analyze the selective history of P-selectin in our closest living relatives, we sequenced the region spanning exons 9–13 in nine unrelated Pan troglodytes. Our results indicated that SELP displays extremely high diversity when compared with other loci resequenced in the same individuals. Yet, this feature is due to the presence of a single highly divergent haplotype that originates a gene phylogeny with extremely deep basal branches that coalesce at 3.5 MY. This estimated TMRCA largely pre-dates the split of the chimpanzee and bonobo lineages, as well as the dispersal of modern chimpanzee populations (Becquet et al., 2007). Thus, these data lend themselves to at least two possible explanations: either distinct SELP haplotypes have been maintained in these primates as a result of long-standing balancing selection or the ancestral primate population was structured. Indeed, archaic population structure with migration can determine the occasional sampling of loci with deep TMRCAs from extant populations. Unfortunately, the still limited knowledge of chimpanzee demographic history, and the small number of sequenced individuals complicate the interpretation of this result. Nonetheless, it is interesting to notice that in both humans and chimpanzees several amino-acid replacements tend to cluster in the region covering exons 11–13, although in either species the MK test was not significant when orangutan or macaque were used as out-groups (not shown). This gene region encodes the SELP sushi domains (also known as complement control protein modules, or short consensus repeats), which are thought to be relevant for the interaction between P-selectin and leukocytes (Ley, 2003). As mentioned above, the binding of SELP to SELPLG has a central role in mediating leukocyte tethering, rolling and extravasation to sites of acute and chronic inflammation. Thus, the selective pressure(s) acting on SELP may be related to its central role in the elicitation of immune and inflammatory responses, as infectious diseases are commonly considered a powerful selective force. SELP has been involved in the pathogenesis of severe malaria, a disease that is commonly regarded as the strongest selective pressure acting on the human genome (Pozzoli et al., 2010). P-selectin contributes to the cytoadherence of infected erythrocytes to the brain microvasculature (Rowe et al., 2009) and mice that specifically lack endothelial expression of Selp are protected against cerebral malaria (Combes et al., 2004). The parasite ligand for P-selectin is thought to be represented by PfEMP1 (Plasmodium falciparum erythrocyte membrane protein 1), as the two proteins bind in vitro (Senczuk et al., 2001). PfEMP1 is a highly variable protein which has a central role in parasite immune evasion by changing binding tropism. Thus, an interesting possibility is that the diversity of human (and possibly chimpanzee) SELP results, at least partially, from a host–pathogen genetic conflict with Plasmodium species that also accounts for local changes in allele frequency of human polymorphisms.

Immune and inflammatory responses are vital to organism survival, but need fine tuning to prevent autoimmunity and tissue damage. Interestingly, the interaction of SELP with SELPLG has been shown to contribute to the generation of tolerogenic dendritic cells (Urzainqui et al., 2007), and P-selectin binding to pentraxin 3 activates a negative feed-back loop that decreases leukocyte recruitment at inflammation sites (Deban et al., 2010). In line with these observations and with its role in immune responses, polymorphisms in SELP have been associated with autoimmune manifestations or chronic inflammatory diseases. Risk alleles for autoimmune conditions have been shown to be common targets of natural selection in humans (Sironi and Clerici, 2010), possibly because of their conferring a selective advantage against ancestral infections. This hypothesis might apply to the Val640Leu SNP in SELP, which is a likely selection target and has been associated with atopy and systemic lupus erythematosus (Bourgain et al., 2003; Morris et al., 2009). Nonetheless, data herein suggest that further genetic association studies for autoimmune diseases or other phenotypic traits would benefit from taking the complex SELP haplotype structure into account and from the analysis of possible regulatory variants in the gene, as they might have represented additional targets of natural selection.

Data archiving

Data have been deposited at Dryad: 10.5061/dryad.211vn8m4.