In a 12.5 cM genome-wide scan for psoriasis susceptibility loci, recombination-based tests revealed linkage to the HLA region (Zmmax = 3.52), as well as suggestive linkage to two novel regions: chromosome 16q (60–83.1 cM from pter, Zmax = 2.50), and chromosome 20p (7.5–25 cM from pter, Zmax = 2.62). All three regions yielded P values ≤ 0.01 by non-parametric analysis. Recombination-based and allele sharing methods also confirmed a previous report of a dominant susceptibility locus on distal chromosome 17q (108.2 cM from pter, Zmax = 2.09, GENEHUNTER P= 0.0056). We could not confirm a previously reported locus on distal chromosome 4q; however, a broad region of unclear significance was identified proximal to this proposed locus (153.6–178.4 cM from pter, Zmax = 1.01). Taken together with our recent results demonstrating linkage to HLA-B and -C, this genome-wide scan identifies a psoriasis susceptibility locus at HLA, confirms linkage to 17q, and recommends two novel genomic regions for further scrutiny. One of these regions (16q) overlaps with a recently-identified susceptibility locus for Crohn's disease. Psoriasis is much more common in patients with Crohn's disease than in controls, suggesting that an immunomodulatory locus capable of influencing both diseases may reside in this region.
Psoriasis is a common, HLA-associated, chronic inflammatory and hyperproliferative skin disorder (1). Although epidemiologic studies clearly show that psoriasis has a strong genetic component, with heritability (h2) of 80–90%, and relative risk to siblings (λs) of 5–10 (2,3), the mode of inheritance has been difficult to ascertain. Most authors have concluded that inheritance is multifactorial (reviewed in refs 2 and 3), although it has been argued that psoriasis is a recessive trait with high disease allele frequency (4). As expected for a complex trait (5), familial aggregation of psoriasis displays an inverse relationship to age at onset (6). Prior HLA association studies have identified disease-associated alleles at several Class I and Class II loci, suggesting that one susceptibility locus resides within or near HLA (reviewed in ref. 2).
After a partial genome scan of eight families, Tomfohrde et al. reported a psoriasis susceptibility locus on distal chromosome 17q (7). More recently, Matthews et al. reported the identification of another susceptibility locus on chromosome 4q, near D4S1535 (8). Thus far, these results have not been reproduced by other investigators (9,10). Here we report the results of a 12.5 cM genome scan for psoriasis susceptibility loci in 115 families, comprised of 86 nuclear families and 29 extended families. The results identify a susceptibility locus in the HLA region, confirm previous findings of a dominant susceptibility locus on distal chromosome 17q (7), and provide suggestive evidence for the existence of novel susceptibility loci on chromosomes 16q and 20p. As a locus conferring susceptibility to Crohn's disease has recently been mapped to an overlapping interval on 16q (11), a common susceptibility locus on 16q could explain the strong concomitance of psoriasis and Crohn's disease (12).
Selected aspects of the study cohort are shown in Table 1. The total number of sib pairs was 224. Application of down-weighting (13) yields 182 independent sib-pair equivalents, with the caveat that some of the larger families contributed multiple sibships with affected sib pairs.
Results of parametric two-point linkage analysis under dominant and recessive models are shown in Figure 1. Under the recessive model, two markers on each of the chromosomes 6p, 16q and 20p gave two-point maximum lod scores >1.9, a value expected to be reached once per genome scan in the absence of linkage (14). The highest lod scores in each region were for markers TNFB (Zmax = 2.58, θ = 0.14, P = 0.00028), D16S3110 (Zmax = 2.50, θ = 0.15, P = 0.00034) and D20S851 (Zmax = 2.62, θ = 0.16, P = 0.00026). Under the dominant model, only one marker yielded a comparable lod score (D16S3032: Zmax = 2.31, θ = 0.12, P = 0.00056). This marker is 1.2 cM centromeric to D16S3110, the marker most suggestive for linkage to 16q under the recessive model.
Allele sharing in affected sib pairs and relatives
The data were analyzed for allele sharing in affected sibling pairs using the programs SIBPAIR and SPLINK. We also carried out a multipoint analysis of allele sharing in affected relatives using the GENEHUNTER program. The complete results of all three analyses can be accessed at our web site (http://www.psoriasis.umich.edu/hmg97.html. D16S3110, which yielded a recombination-based lod score of 2.50 under a recessive model, also produced the most significant ASP lod scores: 2.62 (P = 0.0003) by SIBPAIR and 1.95 (P = 0.0024) by SPLINK. With the exception of D16S3050, which resides 10.5 cM telomeric to D16S3110, no other marker reached the lod score criterion of 2.2 considered suggestive of linkage by affected sib-pair methods (14). However, D10S569 approached this value (SIBPAIR lod = 2.19, P = 0.00075).
The three regions yielding the lowest GENEHUNTER P values were 16q, 17q and 6q. The markers yielding the most significant results in each of these regions were D16S3255 (P = 0.0018), D17S785 (P = 0.0034) and D6S270 (P = 0.0068). A number of markers yielded lod scores >1.0 (P < 0.016) and at least one non-parametric P value less than 0.05 (Table 2). These criteria define the boundaries of the candidate regions mentioned in the abstract. As suggested by Curtis (15), the allele sharing P value criterion for inclusion in Table 2 has been made less stringent than that applied to recombination-based linkage tests, due to the lesser power of allele sharing methods to detect linkage. As reflected in Table 2, inspection of the complete data reveals generally good agreement between recombination-based and allele sharing results (see web site).
Refinement of parametric models
We tested the 6p, 16q and 20p candidate regions identified above under seven additional recessive models (all possible combinations of affecteds-only versus all subjects, locus heterogeneity versus homogeneity and extended pedigrees versus nuclear family splits). Results for the markers yielding the highest recessive lod scores for each region are summarized in Table 3. The lod score at TNFB increased to 3.31 after allowing for genetic heterogeneity, and affecteds-only analysis further increased the lod score to 3.52 (P = 0.000029). We assessed the effects of performing these additional tests by comparing the mean lod score for all 86 markers on chromosomes 6, 16 and 20 under the standard recessive model (0.66) to the mean of the highest lod scores obtained under any of the eight models (0.86) (16). The difference in mean lod score of 0.20 is much less than the increase in Zmax of 0.72 obtained for TNFB under heterogeneity, and of 0.93 obtained under the optimal model. Consistent with this result, application of the admixture test to TNFB revealed significant evidence for heterogeneity under linkage (P = 0.04, see Table 3). In contrast, lod scores for D16S3110 and D20S851 failed to increase more than the mean lod increase of 0.2 described above, and the admixture test provided no evidence for heterogeneity (Table 3). Partitioning large kindreds into nuclear families had little effect on the lod scores derived for any of the three markers tested (data not shown).
Confirmation of linkage to 17q
Tomfohrde et al. reported evidence for a dominant psoriasis susceptibility locus between D17S802 and D17S928 (108.2–128.8 cM from pter) (7). We were previously unable to confirm this result; however, a trend towards allele sharing was noted at three markers within this interval (9). The study cohort has now been expanded to include five additional extended families and 86 nuclear sib pair families (Table 1). In the current study, D17S802 yielded a recombination-based lod of 1.34 (P = 0.0065) under dominance. Applying the admixture test using the HOMOG program, we found significant (P < 0.03) evidence for heterogeneity under linkage (Table 4), as did Tomfohrde et al (7). As shown in Table 4, the maximum lod score for D17S802 increased to 2.09 under heterogeneity (P = 0.00097). As some members of two of the pedigrees used by Tomfohrde et al. have been included in our cohort (see Materials and Methods), all 17q markers were analyzed with and without these pedigrees. However, this manoeuvre produced little or no change in the outcome (data not shown). In fact, the lod scores at D17S802 actually increased after exclusion of the redundant individuals (Table 4).
Although our understanding of the pathogenesis of psoriasis has greatly increased in recent years, particularly as regards the role of the immune system (17), the cause of this common skin disease remains unknown. Based on the high heritability and λs values characteristic of this disease (2,3), we have undertaken a complete genome scan of 726 individuals at a marker density of ∼12.5 cM. While previous partial and complete genome scans for psoriasis have appeared in the literature (7,8), this cohort is by far the largest reported to date. We found three regions with parametric lod scores >1.9, when only one such region would be expected by chance (14). Moreover, one of these loci lies within HLA, a region already indirectly implicated as a susceptibility locus by association studies (reviewed in ref. 2). In addition, we found confirmatory evidence for a previously-reported locus at the distal end of chromosome 17q (7).
Evidence for linkage to HLA
In the initial genome scan, TNFB yielded Zmax = 2.59 at θ = 0.14 under a recessive model. Affecteds-only analysis increased this value to 2.83 at θ = 0. The admixture test provided significant evidence (P < 0.04) for heterogeneity under linkage, and increased Zmax further, to 3.52 at θ = 0. While the latter result exceeds the recently-proposed criterion proposed for declaring significant linkage in a genomewide scan (lod = 3.3) (14), it is important to note that this criterion does not take into account the effects of testing multiple genetic models or the use of multiple analysis methods (15). We applied an empirical correction for multiple testing based on our own data, following Ginns et al. (16). The correction factor of 0.2 for a total of eight genetic models was substantially less than the value of 0.9 obtained following the formula of Terwilliger and Ott [correction factor = log10(n), where n = the number of tests performed] (18). While the corrected lod score of 3.32 still exceeded the criterion value of 3.3, multiple-testing concerns remain (see below).
Confirmation of linkage to 17q
Bowcock and co-workers previously identified a dominant psoriasis susceptibility locus on the distal end of chromosome 17q, near D17S802 (7). While we were previously unable to confirm linkage to this locus (9), we re-addressed this question after expansion of our cohort from 24 extended kindreds to 29 extended kindreds and 86 nuclear sib pair families (Table 1). The expanded cohort yielded a iecombination-based P value for D17S802 (P = 0.0065; Table 2) that exceeded the criterion of P = 0.01 required for independent confirmation of previously-documented significant linkage (14). Other points of agreement between Tomfohrde et al. (7) and our own study include the dominant character of the locus and the presence of genetic heterogeneity. Thus, the lod score for D17S802 was 1.34 under the dominant model as opposed to only 0.08 under the recessive model (Table 2). The admixture test (18) was significant for heterogeneity under linkage, and allowance for heterogeneity increased the maximum lod scores at D17S802 to 2.09 (Table 4). It is worth noting that two adjacent 17q markers, D17S785 and D17S802, yielded the second-lowest GENEHUNTER P values in the scan, ranking behind only a cluster of tightly-spaced markers on 16q (Table 2). This would be consistent with a dominant model for the 17q locus, because GENEHUNTER is more sensitive to dominant traits than are ASP-based methods when all affected relatives are included in the analysis (19).
A common locus for psoriasis and Crohn's disease on 16q?
All methods of analysis support the existence of a susceptibility locus on chromosome 16q near marker D16S3110 (Fig. 1, Table 2). Unlike HLA and 17q, there was no evidence for genetic heterogeneity at this locus (Table 3). Of interest, a locus conferring susceptibility to Crohn's disease has recently been mapped to the interval between D16S409 and D16S419 (56.2–65.6 cM from pter) (11). This interval has substantial overlap with the psoriasis candidate region (60–83.1 cM from pter, as defined using the criterion for inclusion in Table 2). Interestingly, five independent case-control studies have found that the prevalence of psoriasis is markedly and significantly increased in patients with Crohn's disease (reviewed in ref. 12). Combining these five studies with one small study in which no significant increase was detected, there were 53 cases of psoriasis in 594 patients with Crohn's disease (8.9%) versus 63 cases of psoriasis in 4639 controls (1.4%), yielding an aggregate relative risk of 7.1 (χ2= 139, P < 1 × 10−9). Much of this increased risk appears to be due to genetic factors, as 10% of 136 Crohn's disease patients had a history of psoriasis in at least one first-degree relative, as compared to 2.9% of 136 controls (P < 0.02) (12). While the histopathology of psoriasis and Crohn's disease differ substantially (20), both diseases are characterized by immune infiltration of stratified epithelia and underlying connective tissue, without evidence for a transmissible agent. Taken together, the immunopathologic, disease concomitance, and linkage results provide indirect but provocative evidence for a 16q locus involved in the immunopathogenesis of both diseases.
Other candidate regions
Recombination-based and allele sharing methods also support the existence of a susceptibility locus on 20p, in the 17.5 cM region flanked by markers D20S906 and D20S917 (7.5–25 cM from pter) (Zmax = 2.62). In contrast to 16q, however, to date no disease loci have been mapped to 20p (21). Evidence for a second dominant psoriasis susceptibility locus near D4S1535 on chromosome 4q has recently been reported in six Irish-English families (8). Of the chromosome 4q markers analyzed in the present study, D4S413 yielded the most significant P values by recombination-based (lod = 1.01, equivalent P = 0.015) and allele sharing (P = 0.004) techniques (Table 2). However, this marker is 40.6 cM centromeric to D4S1535, which failed to yield any P values <0.05 in our hands (data not shown). Several other markers displayed recombination-based lod scores >1.0 and allele sharing P values <0.05 (Table 2). A larger data set and/or confirmation by other groups will be necessary to assess the significance of 20p, 4q and these additional loci.
Multiple testing and data interpretation
For polygenic traits, there is no foolproof method to protect against false positive results without imposing very strict criteria for linkage (15,18). In this report, we have described the significance of our results using terms recently suggested by Lander and Kruglyak (14). However, these criteria are based on the assumption that only one method is being used to analyze the data. We have analyzed multiple markers using multiple methods and genetic models, and are aware that this increases the risk of a false positive result (18). Our evaluation of one aspect of this risk following Ginns et al. (16) suggests that the ‘penalty’ for testing multiple models that are only partially independent may be relatively small. The same considerations may apply to the use of multiple methods, because allele sharing and recombination-based methods are intrinsically correlated. Thus, all four analytical methods used identified 16q, 20p and HLA as candidate regions. This result gives us confidence that the use of multiple methods is not markedly inflating the number of regions identified. Overall, we agree with Elston and colleagues (22) that nominal P values provide a workable measure of the strength of the evidence for linkage, especially when the results of all tests performed are reported.
We are aware that applying lod score ‘corrections’ which assume independence would decrease the lod score obtained for HLA to below the criterion lod for ‘significant’ linkage. However, our findings are supported by several other lines of evidence. First, long-standing HLA association studies provide substantial a priori evidence for a susceptibility locus within HLA. The associations observed are quite strong and have been reported worldwide (reviewed in ref. 2), making it unlikely that the results are due to population substructure. Second, we have recently presented highly significant evidence for linkage of psoriasis to HLA-B and -C in our kindreds by association, recombination-based linkage, transmission disequilibrium, and segregation tests (22). These results, which take into account the known linkage disequilibrium between psoriasis and certain HLA haplotypes, verify the long-standing finding that failing to account for the presence of marker-trait disequilibrium can greatly decrease power to detect linkage (23). Third, evidence for linkage to the HLA region has recently been reported in two independent data sets (24,25). In aggregate, these data provide very strong evidence for the existence of a susceptibility locus within HLA.
These results provide a solid genetic rationale for further high-resolution genetic and physical analysis of the HLA region and distal 17q. They also identify at least two novel candidate regions for further scrutiny, including one which may also predispose to Crohn's disease. While it will be a major challenge to narrow the candidate intervals to a manageable size by genetic techniques, it seems likely that our increasing knowledge of human genomic polymorphism and psoriasis pathogenesis will suggest candidate genes that can be studied by association analysis, a highly powerful technique (26). Thus, although formidable, the challenges facing psoriasis genetics appear not to be insurmountable.
Materials and Methods
Ascertainment, clinical evaluation and recruitment of subjects
This study contains pedigrees identified from three sources: (i) University of Michigan and Ann Arbor Veterans Hospital Dermatology clinic records, including referrals from community physicians (60 pedigrees, 378 subjects); (ii) the National Psoriasis Foundation Tissue Bank (NPTB) (12 pedigrees, 71 subjects); and (iii) University of Kiel Dermatology clinic records, including referrals (43 pedigrees, 277 subjects). Most of the American pedigrees were from southeastern Michigan (nine extended and 51 nuclear families). The NPTB pedigrees (three extended and nine nuclear families) were from various parts of the United States. Most German pedigrees were from the Schlesswig-Holstein region of Northern Germany. Two of the NPTB families (PS1 and PS5) were part of a previous study that reported linkage of psoriasis to chromosome 17q (7). In these families, fewer individuals were available than published previously, due to unavailability of cell lines or DNA. Pedigree information is summarized in Table 1. Pictures of all pedigrees are available at our web site (http://www.psoriasis.umich.edu/hmg97.html).
Ascertainment was based on juvenile onset of psoriasis in the proband, defined as age ≤40 years (6). Ninety-two percent of affected subjects had an age of onset of 40 years or less. The other inclusion criteria were (i) at least one other affected sibling or (ii) at least four affected members in an extended kindred. Diagnosis was based on examination of the skin, nails and joints by a dermatologist. Individuals were considered affected if two or more skin, scalp, nail or joint lesions were characteristic of psoriasis (1) or if a single lesion covered more than 1% of total body surface area. Parents and siblings of affected sib pairs were collected whenever possible. Informed consent was obtained from all subjects, under protocols approved by the Institutional Review Boards of each participating institution.
Cell lines, markers and genotyping
Methods used for establishment of lymphoblastoid cell lines (LCL) from blood samples and DNA preparation have been described in detail elsewhere (9). Microsatellite markers were selected from the CEPH-Genethon set (27,28) or from the Genome Data Base (http://gdbwww.gdb.org), on the basis of spacing, heterozygosity, and experimental tractability. The CEPH-Genethon map was used to establish inter-marker distances for the genome scan (27,28). Marker distances for some intervals on chromosome 6p were integrated into the map using information available from the Genome Data Base and the published physical distance between markers D6S258 and D6S276 (600 kb) (29), assuming 1 cM = 1 Mb. Whenever possible, the most telomeric markers from each chromosome were selected, along with additional markers spaced at ∼12.5 cM intervals (287 markers). Preliminary analysis of 459 subjects identified six regions with elevated P values or lod scores (data not shown). Based on this information, 53 additional closely spaced markers were typed for chromosomes 4q, 6p, 6q, 11p, 16p and 20p, yielding a total of 340 markers. PCR-based genotyping was performed as described (9), using 32P-labeled oligonucleotide primers purchased from Research Genetics or synthesized at the University of Michigan core facility. The PCR products were fractionated on 6% polyacrylamide-urea gels, and allele sizes were determined using the published allele sizes of CEPH individual no. 134702 as a reference. Reference allele sizes are available at http://www.genethon.fr/genethon_en.html.
Data processing and statistical analysis
Data were checked for Mendelian inheritance errors using the LINKAGE package (18) and Pedmanager (available by anonymous ftp from ftp-genome.wi.mit.edu/distribution/software). Allele frequencies were calculated from founders using Pedmanager. Two-point parametric linkage analysis was performed using LINKAGE (version 5.1) under dominant and recessive models. Maximum lod scores were calculated using ILINK. Disease allele frequency was assumed to be 0.05 under dominant and 0.25 under recessive models (30). Assuming a population prevalence of 1% (31) with 10% of psoriatics affected due to non-genetic causes, Hardy-Weinberg equilibrium yielded the penetrances fDD = fDd = 0.092, fdd = 0.0011 for the dominant model and fDD = 0.144, fDd = fdd = 0.0011 for the recessive model. X chromosome penetrances for males were fD = 0.18, fd = 0.0011 under dominance, and fD = 0.036, fd = 0.0013 under recessivity. The effects of assuming genetic heterogeneity were assessed using HOMOG (version 3.35) (32). Affecteds-only analysis was performed by dividing all penetrances by 1000, as recommended by Terwilliger and Ott (18). Assessment for the effects of testing with multiple recessive linkage models was performed as described by Ginns et al. (16). Conversion of maximum lod scores to equivalent P values was performed assuming that the likelihood ratio statistic follows a χ2 distribution with one degree of freedom and that the test is one-sided.
Non-parametric sib-pair allele sharing analysis was performed using SPLINK (version 1.07) (33) and SIBPAIR (version 2.1) (34), both of which downweight for the presence of more than one affected sibling pair per sibship. SPLINK was run using the default flag settings. Multipoint non-parametric analysis was performed using GENEHUNTER (version 1.1) in the non-parametric mode. This program uses a hidden Markov model to calculate the inheritance distribution conditional on all marker genotypes, and compares it to the distribution of the trait by parametric or non-parametric methods (19). The program was configured for large pedigrees and simultaneous scoring of all possible affected relative pairs (settings were max bits = 16, skip large = off, analysis = NPL, score = all, increment = step 5). In two cases, pedigrees were split to ensure that the program retained all affected individuals.
We thank Dr Margaret Terhune for clinical evaluation of a significant portion of our subjects, and community physicians across the United States for referring psoriasis patients and evaluating them for the study. The excellent technical assistance of Sherri Kokx, Brent Rasmussen and Emily Malvitz is sincerely appreciated. The Pedmanager program was kindly provided by Dr Eric Lander's laboratory. This research was supported by USPHS awards P30 HG00209-03 and R01 AR4274-01 (JTE, RN, SWG, JJV), by award DFG-WE 905/1-1 from the German Research Foundation (TH, SJ, EC), by the Ann Arbor Veterans Administration Hospital (JTE), and by the Babcock Memorial Trust. Portions of these studies were conducted at the General Clinical Research Center (GCRC) at the University of Michigan, funded by a grant (M01EE00042) from the National Center for Research Resources, National Institutes of Health, USPHS.
affected sib pair
Centre d'tude du Polymorphisme Humain
human leukocyte antigen
lymphoblastoid cell line
National Psoriasis Foundation Tissue Bank
polymerase chain reaction
most p-terminal marker on chromosome (28)
Note Added in Proof
After this manuscript was submitted for publication, Trembath et al. (Hum. Mol. Genet.6, 813–820) has extended the finding of ref. 25 in a larger data set. Using a novel linkage strategy that extracts full non-parametric information, they identified a major susceptibility locus in the HLA region on chromosome 6p, which is in agreement with the findings of this paper. In addition, possible linkage regions were identified on chromosomes 2, 8 and 20. The chromosome 20 marker with the lowest P value, D20S186, is 9.4 cM centromeric to D20S851, the marker for which we obtained the highest parametric lod score. We did not find significant lod scores for chromosomes 2 and 8.