Abstract
Background
Woolly hair (WH) is a hair abnormality that is primarily characterised by tightly curled hair with abnormal growth.
Methods
In two unrelated consanguineous Pakistani families with non-syndromic autosomal recessive (AR) WH, homozygosity mapping and linkage analysis identified a locus within 17q21.1–q22, which contains the type I keratin gene cluster. A DNA sample from an affected individual from each family underwent exome sequencing.
Results
A homozygous missense variant c.950T>C (p.(Leu317Pro)) within KRT25 segregated with ARWH in both families, and has a combined maximum two-point LOD score of 7.9 at ϴ=0. The KRT25 variant is predicted to result in disruption of the second α-helical rod domain and the entire protein structure, thus possibly interfering with heterodimerisation of K25 with type II keratins within the inner root sheath (IRS) of the hair follicle and the medulla of the hair shaft.
Conclusions
Our findings implicate a novel gene involved in human hair abnormality, and are consistent with the curled, fragile hair found in mice with Krt25 mutations, and further support the role of IRS-specific type I keratins in hair follicle development and maintenance of hair texture.
The structural integrity of several ectodermal appendages, including hair, is in part maintained by keratin proteins. Mutations in several keratin-encoding genes are responsible for a wide spectrum of phenotypes involving hair, skin and nail.1 Humans have 54 α-keratin genes, which are grouped into type I (n=28) and type II (n=26), and are arranged in two clusters on chromosomes 17q21.2 and 12q13.13, respectively. Of these, 17 keratins (ie, 11 type I and 6 type II) are expressed in the hair follicle, and undergo cyclopolymerisations to form keratin intermediate filaments (KIFs), which in turn interact with keratin-associated proteins to assemble the hair shaft into a tough structure.2 Four keratin genes from each type, namely KRT25-KRT28 for type I and KRT71-KRT74 for type II, are expressed in the medulla, which is the innermost layer of the hair shaft and specifically in the inner root sheath (IRS) of the hair follicle.3 The IRS is composed of Henle’s, Huxley’s and cuticle layers, and surrounds the highly keratinised hair shaft. Both the IRS and the hair shaft are important for normal hair growth, especially in the determination and maintenance of hair texture.4
Woolly hair (WH) is a hair structural anomaly that is characterised by tightly curled scalp hair.5–8 It can be observed either as an isolated trait or as a syndromic feature such as: WH and keratoderma with or without heart disease due to mutations in DSP (MIM 125647), DSC2 (MIM 125645), KANK2 (MIM 616099) or JUP (MIM 173325); and trichohepatoenteric syndrome due to TTC37 (MIM 614589) or SKIV2L (MIM 600478) variants.9 Non-syndromic WH can be inherited either in an autosomal recessive (ARWH, MIM 278150, MIM 604379) or autosomal dominant (ADWH, MIM 194300) manner. To date, mutations in lysophosphatidic acid receptor 6 (LPAR6/P2RY5, MIM 609239) and lipase H (LIPH, MIM 607365) are known to underlie ARWH with or without hypotrichosis.5–7 On the other hand, ADWH and hypotrichosis (MIM 613981) are caused by mutations in two type II IRS-specific genes KRT71 (MIM 608245) and KRT74 (MIM 608248).8,10 Additionally, a KRT74 variant has been identified to cause pure hair and nail ectodermal dysplasia.11 All four genes are abundantly expressed in the hair follicle. In this study, a novel ARWH locus was mapped to a region containing the type I keratin cluster using data from two families, which segregate ARWH. Using exome sequencing, a rare missense variant KRT25 c.950T>C (p.(Leu317Pro)) was identified to segregate with ARWH in both families.
Prior to study onset, approval was obtained from the institutional review boards of Quaid-i-Azam University, Islamabad, Pakistan, and Baylor College of Medicine and Affiliated Hospitals, Houston, Texas, USA. Written informed consent was obtained from all study participants from the two families. Family AP188 (figure 1A) was recruited from a remote village in Sindh province, Pakistan, and had four individuals (VI-1, VI-2, VI-4, VI-5) presenting with tightly curled scalp hairs (figure 1C–E). Initially, all four affected individuals possessed soft, short and sparse hairs on the scalp (figure 1D–E), but at around 4 years of age, developed tightly curled hair (figure 1C). Additionally, they had normal eyebrows, but eyelashes were sparse compared with unaffected relatives. Family AP216 was recruited from the Upper Dir district in the Khyber Pakhtunkhwa province of Pakistan, and has three individuals (II-1, II-2, II-4) with WH (figure 1B). As in AP188, the affected individuals from AP216 also presented with tightly curled scalp hair (figure 1F–H). Sparse, WH was more evident in the long-haired female individual II-2 (figure 1G). All individuals from both families showed normal craniofacial features, teeth, nails and sweating, and no evidence of palmoplantar keratoderma, heart disease, chronic diarrhoea, immune deficiencies or intellectual disability.
Figure 1.

Pedigree drawings, haplotypes and sparse, woolly hair in families AP188 and AP216. (A and B) Drawings of pedigrees AP188 and AP216 displaying segregation of the identical SNP haplotype (within rectangles) and the KRT25 c.950T>C variant with autosomal recessive woolly hair. Arrows indicate affected individuals whose DNA samples were submitted for exome sequencing. (C) Individual VI-4 from AP188 had woolly hair at age 8 years. (D and E) Individual VI-5 from AP188 had sparse, soft hairs on scalp at age 1 year. (F and H) Individual II-4 from AP216 with woolly hair at age 8 years. (G) Sparse, woolly hair is more obvious in 10-year-old female II-2 from AP216 who has long hair.
DNA samples were extracted from venous blood of eight and six individuals from families AP188 and AP216, respectively. DNA samples from selected family members (figure 1A, B) underwent a genome scan using the Infinium HumanCoreExome BeadChip (Illumina, USA), which includes ~550 000 markers. The two families’ genotype data were analysed by performing homozygosity mapping and multipoint parametric linkage analysis using the software programmes HomozygosityMapper and Allegro,12 respectively. In family AP188, a 16.7-Mb homozygous region was identified on chromosome 17q21.1–q22 (figure 1A). For this same region, a maximum multipoint LOD score of 4.14 was observed. For family AP216, a maximum multipoint LOD score of 1.33 was observed at multiple regions of homozygosity, including two 3-Mb regions on chromosomes 2q36.1–q36.3 and 14q24.2–q24.3 and a 20.4-Mb homozygous region on 17q12–q22 (figure 1B). Although the parents (I-1, I-2) from family AP216 did not self-report being related, haplotype analysis showed the presence of the same carrier haplotype between rs9972960 and rs4794665 in both parents (figure 1B). Moreover, the kinship coefficient using KING13 was estimated to be 0.05, indicating that the parents were third-degree relatives. Comparing the mapped regions from the two families, the homozygous region for family AP188 is completely contained within the homozygous region on chromosome 17 for family AP216. Although the two families are unrelated, and are from different ethnic groups within Pakistan, they segregate the same 15.8-Mb SNP haplotype (figure 1A, B), which encompasses the type I keratin cluster on 17q21.2. Hair expression of type I keratins is known,3 but no human hair-related traits have been previously associated with IRS-specific type I keratins.
Exome sequencing of DNA samples from affected individuals —VI-2 from family AP188 and II-4 from AP216 (figure 1A, B)—was performed at the University of Washington Center for Mendelian Genomics to average read depth of 36× and 63×, respectively. Sequence alignment, variant detection and calling were performed as previously described.14 Variant annotation was performed using Variant Mendelian Tools, an in-house pipeline that allows for exome quality control and selection of variants according to mode of inheritance, minor allele frequency (MAF) in the Exome Aggregation Consortium (ExAC) database, variant type, nucleotide conservation and prediction from multiple bioinformatics tools. Aside from being ruled out by linkage analysis and homozygosity mapping, in the two exomes, we did not identify potentially causal homozygous or compound heterozygous variants in genes that are previously known to be involved in syndromic or non-syndromic WH. Within the mapped interval for AP188, a total of 176 variants that passed quality control were homozygous in both exome sequences. Of 176 homozygous variants, only two variants had MAF <0.01 in South Asian alleles in ExAC: a synonymous FAM134C NM_178126.3:c.906T>C variant that is heterozygous in 89 out of 16 512 South Asian ExAC alleles (MAF=0.005) and in two ExAC alleles of other ethnicity; and a missense variant c.950T>C (p.(Leu317Pro)) within KRT25 (NM_18154.3) that is heterozygous in three ExAC South Asian alleles (MAF=0.0002), but was not identified in 104 876 ExAC alleles of other ethnicity. Furthermore, the KRT25 c.950T>C variant was not found in 154 in-house exome sequences from unrelated Pakistani individuals with non-hair phenotypes, while the FAM134C synonymous variant was heterozygous in three in-house Pakistani exomes. Sanger sequencing of the KRT25 missense variant using DNA samples from the remaining members of AP188 and AP216 confirmed segregation with the ARWH phenotype in both families (figure 1A, B, see online supplementary figure S1A). Additionally, Superlink15 was used to perform two-point linkage analysis of the KRT25 c.950T>C variant, and a maximum LOD score of 6.57 and 1.33 (ϴ=0) were obtained for AP188 and AP216, respectively. The KRT25 variant is absent in 462 Pakistani control chromosomes. Comparison of the exome variants within the mapped region narrowed down the haplotype shared by the two exome sequences from the two families to five exome variants that occur within a 48.7-kb region (table 1). Segregation testing of the four variants within KRT27 and KRT28 revealed that these four variants did not segregate with fragile WH. Therefore, only the KRT25 missense variant segregates with ARWH in both families.
Table 1.
Homozygous chromosome 17 variants in AP188 and AP216 individuals with autosomal recessive woolly hair and exome sequence data
| hg19 Position | 38 907 213 | 38 935 876 | 38 935 880 | 38 950 272 | 38 955 961 |
|---|---|---|---|---|---|
| Reference allele | A | C | C | T | G |
| Alternate allele | G | T | T | C | A |
| Gene | KRT25 | KRT27 | KRT27 | KRT28 | KRT28 |
| cDNA change | c.950T>C | c.850G>A | c.847-1G>A | c.1005A>G | c.185C>T |
| Amino acid change | p.(Leu317Pro) | p.(Ala284Thr) | NA | p.(=) | p.(Ala62Val) |
| Variant type | missense | missense | splice-3′ | synonymous | missense |
| ExAC South Asian MAF | 0.0002 | 0.24 | 0.06 | 0.62 | 0.31 |
| ExAC all MAF | 0.00002 | 0.17 | 0.008 | 0.58 | 0.22 |
| MAF in-house exomes | 0 | 0.22 | 0.05* | 0.69 | 0.40 |
mRNA accession numbers: KRT25, NM_18154.3; KRT27, NM_181537.3; KRT28, NM_181535.3.
ExAC, Exome Aggregation Consortium; MAF, minor allele frequency; NA, not available.
This KRT27 splice variant is heterozygous in 12 out of 154 in-house exomes from unrelated Pakistani individuals with normal hair. Additionally, an exome from an individual with microcephaly and intellectual disability, but no hypotrichosis or woolly hair is homozygous for the KRT27 splice variant. Thus, we do not believe the KRT27 splice variant is causal of ARWH.
The observation of sparse WH in affected individuals who are homozygous for the KRT25 c.950T>C variant recapitulates the phenotype in mutant mice. Two previously described mutant mouse strains carry Krt25 missense mutations. Re/Re mice who are homozygous for a c.1142T>C (p.(Leu381Pro)) variant within the helix termination motif (HTM) at the end of the rod 2B domain of K25 have strongly curled, extremely fragile fur.16 On the other hand, M100573 mice have ragged, fragile vibrissae and carry a dominant c.1135T>A (p.(Tyr379Asn)) variant, which is also within the HTM of mouse K25.16 At young age, both strains demonstrate bent hair follicles and break points in the Henle’s layer of the IRS, but in the M100573 strain with the dominant variant, the hair becomes less wavy with age.16 Furthermore, because of heterodimerisation of K25 with type II keratin K71, heterozygous Re/+ mice have mislocalised K71 protein and irregular Henle’s and Huxle’s layers within the IRS.16 In AP188 and AP216 however, family members who are heterozygous for the KRT25 c.950T>C variant have normal head and facial and body hair throughout life. This may be due to differences in hair structure between mouse and human, and/or the specific K25 motif that is affected by mutation. It should be noted that non-occurrence of the p.(Leu317Pro) variant within either the HTM or helix initiation motif of K25 does not preclude causality for the ARWH phenotype. For example, a p.Phe274Ser variant within the coil 1B domain of K74 is known to cause hair and nail ectodermal dysplasia in humans.11
The K25 p.(Leu317Pro) variant is predicted to be damaging or deleterious by multiple bioinformatics tools in the dbNSFP database, including Combined Annotation Dependent Depletion (CADD scaled score=15.0), fathmm, likelihood ratio test, logistic regression, MutationAssessor, MutationTaster, PolyPhen2 (HVAR), PROVEAN, SIFT and support vector machine. Analysis of human K25 with InterProScan showed that the variant occurs within a type I keratin signature at residues 308–323. Alignment of the human sequence with other type I keratins in seven mammals, one marsupial, one bird, one amphibian and two fish shows strong amino acid conservation of Leu317 and several residues, including Arg304, Gln307, Leu309, Ile311, Leu313, Gln314, Ser315, Lys320 and Leu323 (see online supplementary figure S1B). Based on molecular modelling, it is predicted that these conserved residues contribute to the formation of the α-helix well within the rod 2B domain of K25 (see online supplementary figures S1C and S2A, C). The replacement of leucine at residue 317 with a proline is expected to disrupt the α-helix, and cause changes to the local domain and the overall protein structure (see online supplementary figure S2B, D).
The KRT25 transcript and the encoded protein are expressed in all three layers of the follicle IRS and the hair shaft medulla.3 Likewise, other type I keratins namely KRT27 and KRT28 have the same expression pattern, while of the type II keratins, only KRT71 is expressed in all three IRS layers.3 The WH phenotype observed in individuals homozygous for the KRT25 c.950T>C variant implies that heterodimerisation between K25 and K71 is essential for stable hair structure, and is not compensated by interaction of K71 with K27 or K28. This is further supported by the abnormal localisation of K71 and irregularity of Henle’s and Huxley’s layers of the IRS in Krt25-mutant Re/+ mice.16 Additionally, KRT71 mutations independently cause hypotrichosis and WH in humans.7
Hair keratins are known to heterodimerise such that a type I keratin is required to partner with a type II keratin to form KIFs that are basic building blocks for hair structure.17 Interestingly, disulfide bonding between cysteine residues of partnering keratins has been shown as an important mechanism for the assembly of keratins into KIFs.18 Human K25 has cysteine residues at positions 325 and 335 that are highly conserved in mammalian K25 sequences (see online supplementary figure S1B). Due to the p. (Leu317Pro) variant, the region containing these cysteine residues is predicted to become highly disorganised (see online supplementary figure S2D), which possibly interferes with disulfide bonding with K71 and affects proper heterodimerisation and microfibril structure within the hair follicle. Because K25 is also localised to the medulla, structural defects from improper heterodimerisation with other type II keratins are also expected to affect the hair shaft. For example, type II keratin K75 is also expressed in the medulla.3 Mutant mice with a dominant Krt75 mutation have rough hair coat and hair shaft defects,19 and chickens with a KRT75 deletion also have frizzled feathers with defects within medulla.20 Thus, it is postulated that the p. (Leu317Pro) variant also impairs heterodimerisation of K25 with K75 within the medulla of the hair shaft.
In conclusion, we identified a homozygous missense mutation within KRT25 that causes ARWH in humans, which is consistent with findings in mutant mice. The variant is predicted to disrupt the α-helical rod 2B domain of K25 and to interfere with heterodimerisation with type II keratins, which are expected to lead to irregular and fragile hair that is characteristic of the WH phenotype. Taken together, our findings further support the role of type I keratins in hair follicle development and maintenance of hair texture. The identification of a KRT25 mutation as a cause of WH in humans can help direct efforts in the development of therapies for hair keratin disorders.1
Supplementary Material
Acknowledgments
We thank the family members for their invaluable cooperation and participation in this study.
Funding This work was funded by the National Institutes of Health grant U54 HG006493 and the Higher Education Commission, Islamabad, Pakistan.
Footnotes
Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/jmedgenet-2015-103255).
WEB RESOURCES
Clustal Omega, http://www.ebi.ac.uk/Tools/services/web_clustalo
dbNSFP, sites.google.com/site/jpopgen/dbNSFP
Exome Aggregation Consortium (ExAC), exac.broadinstitute.org
HomozygosityMapper, homozygositymapper.org
InterProScan, http://www.ebi.ac.uk/Tools/pfa/iprscan
OMIM, http://www.omim.org
Protein BLAST, blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Proteins
RCSB Protein Data Bank, http://www.rcsb.org/pdb
UCSC Genome Bioinformatics, genome.ucsc.edu
Variant Mendelian Tools, varianttools.sf.net/VMT
Collaborators University of Washington Center for Mendelian Genomics.
Contributors Conceived and designed the experiments: MA, MJB, DAN, WA and SML. Performed the experiments: MA, SIR, IU, AA, HD, UWCMG and JDS. Analysed the data: MA, KL and RLPS-C. Contributed reagents/materials/analysis tools: SIR, IU, SS, JDS, UWCMG, JS, WA and SML. Contributed to writing: MA, RLPS-C, KL, WA and SML.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Institutional review boards of the Quaid-i-Azam University and Baylor College of Medicine and Affiliated Hospitals.
Provenance and peer review Not commissioned; externally peer reviewed.
References
- 1.McLean WHI, Moore CBT. Keratin disorders: from gene to therapy. Hum Mol Genet. 2011;20:R189–97. doi: 10.1093/hmg/ddr379. [DOI] [PubMed] [Google Scholar]
- 2.Stenn KS, Paus R. Controls of hair follicle cycling. Physiol Rev. 2001;81:449–94. doi: 10.1152/physrev.2001.81.1.449. [DOI] [PubMed] [Google Scholar]
- 3.Langbein L, Rogers MA, Praetzel-Wunder S, Helmke B, Schirmacher P, Schweizer J. K25 (K25irs1), K26 (K25irs2), K27 (K25irs3), and K28 (K25irs4) represent the type I inner root sheath keratins of the human hair follicle. J Invest Dermatol. 2006;26:2377–86. doi: 10.1038/sj.jid.5700494. [DOI] [PubMed] [Google Scholar]
- 4.Harel S, Christiano AM. Genetics of structural hair disorders. J Invest Dermatol. 2012;132:E22–6. doi: 10.1038/skinbio.2012.7. [DOI] [PubMed] [Google Scholar]
- 5.Shimomura Y, Garzon MC, Kristal L, Shapiro L, Christiano AM. Autosomal recessive woolly hair with hypotrichosis caused by a novel homozygous mutation in the P2RY5 gene. Exp Dermatol. 2009;18:218–21. doi: 10.1111/j.1600-0625.2008.00788.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Shimomura Y, Ito M, Christiano AM. Mutations in the LIPH gene in three Japanese families with autosomal recessive woolly hair/hypotrichosis. J Dermatol Sci. 2009;56:205–7. doi: 10.1016/j.jdermsci.2009.08.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Khan S, Habib R, Mir H, Umm-e-Kalsoom, Naz G, Ayub M, Shafique S, Yamin T, Ali N, Basit S, Wasif N, Kamran-Ul-Hassan Naqvi S, Ali G, Wali A, Ansar M, Ahmad W. Mutations in the LPAR6 and LIPH genes underlie autosomal recessive hypotrichosis/woolly hair in 17 consanguineous families from Pakistan. Clin Exp Derm. 2011;36:652–4. doi: 10.1111/j.1365-2230.2011.04014.x. [DOI] [PubMed] [Google Scholar]
- 8.Fujimoto A, Farooq M, Fujikawa H, Inoue A, Ohyama M, Ehama R, Nakanishi J, Hagihara M, Iwabushi T, Aoki J, Ito M, Shimomura Y. A missense mutation within the helix initiation motif of the keratin K71 gene underlies autosomal dominant woolly hair/hypotrichosis. J Invest Dermatol. 2012;132:2342–9. doi: 10.1038/jid.2012.154. [DOI] [PubMed] [Google Scholar]
- 9.Ramot Y, Zlotogorski A. The twisting tale of woolly hair: a trait with many causes. J Med Genet. 2015;52:217–23. doi: 10.1136/jmedgenet-2014-102630. [DOI] [PubMed] [Google Scholar]
- 10.Wasif N, Naqvi SK, Basit S, Ali G, Ansar M, Ahmad W. Novel mutations in the keratin-74 (KRT74) gene underlie autosomal dominant woolly hair/hypotrichosis in Pakistani families. Hum Genet. 2011;129:419–24. doi: 10.1007/s00439-010-0938-9. [DOI] [PubMed] [Google Scholar]
- 11.Raykova D, Klar J, Azhar A, Khan TN, Malik NA, Iqbal M, Tariq M, Baig SM, Dahl N. Autosomal recessive transmission of a rare KRT74 variant causes hair and nail ectodermal dysplasia: allelism with dominant woolly hair/hypotrichosis. PLoS ONE. 2014;9:e93607. doi: 10.1371/journal.pone.0093607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gudbjartsson DF, Thorvaldsson T, Kong A, Gunnarsson G, Ingolfsdottir A. Allegro version 2. Nat Genet. 2005;37:1015–16. doi: 10.1038/ng1005-1015. [DOI] [PubMed] [Google Scholar]
- 13.Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen WM. Robust relationship inference in genome-wide association studies. Bioinformatics. 2010;26:2867–73. doi: 10.1093/bioinformatics/btq559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Santos-Cortez RLP, Lee K, Azeem Z, Antonellis PJ, Pollock LM, Khan S, Irfanullah, Andrade-Elizondo PB, Chiu I, Adams MD, Basit S, Smith JD, University of Washington Center for Mendelian Genomics. Nickerson DA, McDermott BM, Jr, Ahmad W, Leal SM. Mutations in KARS, encoding lysyl-tRNA synthetase, cause autosomal recessive nonsyndromic hearing impairment DFNB89. Am J Hum Genet. 2013;93:132–40. doi: 10.1016/j.ajhg.2013.05.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Fishelson M, Geiger D. Exact genetic linkage computations for general pedigrees. Bioinformatics. 2002;18(Suppl 1):S189–98. doi: 10.1093/bioinformatics/18.suppl_1.s189. [DOI] [PubMed] [Google Scholar]
- 16.Tanaka S, Miura I, Yoshiki A, Kato Y, Yokoyama H, Shinogi A, Masuya H, Wakana S, Tamura M, Shiroishi T. Mutations in the helix termination motif of mouse type I IRS keratin genes impair the assembly of keratin intermediate filament. Genomics. 2007;90:703–11. doi: 10.1016/j.ygeno.2007.07.013. [DOI] [PubMed] [Google Scholar]
- 17.Smith TA, Parry DA. Three-dimensional modeling of interchain sequence similarities and differences in the coiled-coil segments of keratin intermediate filament heterodimers highlight features important in assembly. J Struct Biol. 2008;162:139–51. doi: 10.1016/j.jsb.2007.11.005. [DOI] [PubMed] [Google Scholar]
- 18.Feng X, Coulombe PA. A role for disulfide bonding in keratin intermediate filament organization and dynamics in skin keratinocytes. J Cell Biol. 2015;209:59–72. doi: 10.1083/jcb.201408079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Chen J, Jaeger K, Den Z, Koch PJ, Sundberg JP, Roop DR. Mice expressing a mutant Krt75 (K6hf) allele develop hair and nail defects resembling pachyonychia congenita. J Invest Dermatol. 2008;128:270–9. doi: 10.1038/sj.jid.5701038. [DOI] [PubMed] [Google Scholar]
- 20.Ng CS, Wu P, Foley J, Foley A, McDonald ML, Juan WT, Huang CJ, Lai YT, Lo WS, Chen CF, Leal SM, Zhang H, Widelitz RB, Patel PI, Li WH, Chuong CM. The chicken frizzle feather is due to an α-keratin (KRT75) mutation that causes a defective rachis. PLoS Genet. 2012;8:e1002748. doi: 10.1371/journal.pgen.1002748. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
