Exploring the global landscape of genetic variation in coagulation factor XI deficiency

Coagulation factor XI (FXI) is a 160-kDa glycoprotein mainly synthesized in hepatocytes and secreted into the circulation as a zymogen.¹  In the coagulation cascade, activated FXI (FXIa) plays a fundamental role in the activation of factor IX, a key step for thrombin generation at the site of vessel injury.¹ 

Hereditary FXI deficiency (Mendelian Inheritance in Man, MIM*264900) is an autosomal hemorrhagic disorder characterized by mildly to severely reduced levels of coagulation FXI in plasma. The deficiency can be classified as quantitative or qualitative, based on concordance/discordance of FXI antigen and activity levels.²  In severe FXI-deficient patients (ie, those showing FXI levels <20 IU/dL), the disease is usually associated with mild-to-moderate bleedings, principally after trauma or surgery. The bleeding history of patients with partial FXI deficiency is often unpredictable, making them not easily distinguishable from the severe ones.³  Genotype-phenotype correlation in these patients is also jeopardized by the possible coexistence of other hemostatic defects.⁴ 

Hereditary FXI deficiency is mostly associated with genetic defects in the FXI gene (F11), which is composed of 15 exons spanning ∼24 kb on 4q35.2. According to the FXI Deficiency Mutation Database (http://www.factorxi.org/),⁵  >190 disease-causing mutations have been reported, with a great preponderance of missense mutations (66%). A high level of allelic heterogeneity is found in most populations, where prevalence of 1 in 10⁶ has been reported.²  On the other hand, important founder effects have been reported in specific populations, associated with higher prevalence. In particular, prevalent ancestral mutations were found in Basques from France (p.Cys38Arg), in French patients from Nantes (p.Gln88X), in English patients (p.Cys128X), and, above all, in Ashkenazi Jews (the so-called type II p.Glu117X and type III p.Phe283Leu mutations).^6-9  Indeed, 1 in 450 Ashkenazi Jews are expected to suffer from type II homozygous, type III homozygous, or type II-III compound heterozygous FXI deficiency.⁹ 

In this work, we attempted to define the global landscape of F11 genetic variation and to determine population-specific carrier rates in FXI deficiency by exploring exome data available through the Exome Aggregation Consortium (ExAC) resource.

For assessing prevalence of FXI deficiency in different ethnic groups, we extracted data from the ExAC database, which contains exome data from 60 706 unrelated individuals sequenced as part of several disease-specific studies (not including FXI deficiency), as well as population-based genetic studies.¹⁰ 

Among the reported variants, we considered only: (1) disruptive mutations (nonsense, frame-shift, and splice-site variants, the latter including only those affecting the first 2 or last 2 intronic nucleotides) and (2) nonsynonymous variants annotated as deleterious by 7 of 7 prediction programs. These software, all comprised in the dbNSFP database, were: SIFT,¹¹  PolyPhen2 (2 different algorithms),¹²  MutationTaster,¹³  MutationAssessor,¹⁴  the Likelihood Ratio Test (LRT),¹⁵  and the Functional Analysis Through Hidden Markov Model (FATHMM).¹⁶ 

We performed prevalence calculations using 2 different approaches. In one case, we considered all variants belonging to groups 1 and 2. In the other case, we restricted the mutation list to group 1 and to those variants of group 2 already described in FXI deficiency, by searching them in publicly available databases: the Factor XI mutation database, the Leiden Open-source Variation database (LOVD; http://www.lovd.nl/3.0/home),¹⁷  the Expert Protein Analysis system (ExPASy; https://www.expasy.org/),¹⁸  the ClinVar resource (http://www.ncbi.nlm.nih.gov/clinvar/),¹⁹  and the public release of the Human Gene Mutation Database (HGMD; http://www.hgmd.cf.ac.uk/ac/index.php).²⁰ 

The ExAC database offers information on sequence variation in 60 706 unrelated individuals collected through the contribution of 16 exome-based projects.¹⁰  High-quality variants were collected in subjects of diverse ethnicities, including Africans, non-Finnish Europeans, Finnish Europeans, admixed Americans (Latinos), South Asians, East Asians, and other ethnicities. Data on phenotypes for these individuals are not available, but it is known that those suffering from severe pediatric diseases were not included.

Among the reported variants, we retrieved all of the null mutations as well as the nonsynonymous variants annotated as damaging by 7 of 7 prediction programs. We found a total of 420 variants in 121 412 different alleles (Table 1); the number of variants decreased to 351 considering only those mutations that have already been reported as associated with FXI deficiency in publicly available repositories (Table 2; in Table 3, we listed the mutations predicted as dangerous by 7 of 7 prediction programs, but not annotated in the literature). All variants were present in heterozygous individuals, with the exception of 1 subject (European, non-Finnish) who was homozygous for the type II mutation.

In general, our analysis revealed prevalence data for severe FXI deficiency significantly higher than those reported so far. For instance, non-Finnish Europeans show a rate of 12.9 in 10⁶, East Asians display the exceptional rate of 20.3 in 10⁶, whereas for Latinos we observed a rate of 4.6 in 10⁶. Africans and South Asians show lower and similar rates (around 2 in 10⁶), whereas FXI deficiency seems to be quite rare among Finns (0.1 in 10⁶) (Table 2).

Concerning specifically the type II and type III mutations, these were confirmed to be present almost exclusively in Europeans, with the only exception of South Asians, in which the type II mutation was quite frequent (12.5% of all mutated alleles) (Table 4).

Finally, this analysis allowed us to highlight at least 4 recurrent and ethnic-specific mutations: the missense variant p.Phe223Leu in Africans (23.5% of all mutated alleles), the nonsense p.Gln263X and the frameshift p.Leu424CysfsX mutations in East Asians (28.2% and 20.5% of all mutated alleles, respectively), and the missense variant p.Ala412Thr in Latinos (25% of all mutated alleles) (Table 4).

The genetic bases of quantitative FXI deficiency are almost invariably constituted by mutations within the F11 gene. The best source of publicly available information concerning FXI deficiency–causing mutations is the FXI Deficiency Mutation Database.⁵  Since 1999, 191 mutations have been reported, highlighting the high allelic heterogeneity of the disease. The absence of important symptoms, often characterizing FXI-deficient patients, could be responsible for an underestimation of the actual prevalence of the disease. In this respect, genetic-epidemiologic research could be of help to improve these estimations. Of course, this kind of approach has remained largely unexplored in the case of rare diseases, mainly due to the absence of a significant amount of genetic data. This was particularly true for countries with limited resources, considering that most epidemiologic studies have been performed on individuals from North America and Europe.^21,22  Today, the progress in next-generation sequencing technologies is revolutionizing the field of genetic epidemiology, with huge amounts of exome data from large international consortia becoming increasingly available. Hence, we decided to explore the global landscape of genetic variation in FXI deficiency by taking advantage of the extraordinary resource represented by the ExAC database, which offers information on sequence variation in >60 000 individuals, including those coming from “developing” countries.

In ExAC, all of the 14 protein-coding exons of the F11 gene were sequenced at sufficient depth (on average 69× coverage, ∼10% of individuals with coverage >100×), with data available for a total of 121 412 alleles. From this data set, we retrieved 420 potentially deleterious alleles, corresponding to 96 different null mutations or missense variants predicted to be damaging by 7 of 7 algorithms. Their distribution was 25% nonsense, 60.4% missense, 6.3% splicing, and 8.8% frameshift mutations. This distribution was significantly different (χ² = 11.4; 2-tailed P = .0098) from that reported in the FXI Deficiency Mutation Database (13.7% nonsense, 66.3% missense, 9.5% splicing, 10.5% frameshift mutations; the “undefined” mutation and the variant located in the promoter region were excluded from this calculation), essentially for an inflation of nonsense mutations, which, however, have a clear impact on the protein. For the sake of comparison, we also compared the mutation type distribution between the ExAC variants predicted to be damaging by all algorithms and those listed in the FXI Deficiency Mutation Database that were “predicted” by the same algorithms to be deleterious. Though annotated in the disease database as associated with FXI deficiency, only 56% of missense variants were predicted as potentially deleterious by 7 of 7 algorithms, leading to a different mutation distribution: 19.1% nonsense, 52.2% missense, 14% splicing, and 14.7% frameshift mutations (the distribution again was significantly different, though to a different extent; χ² = 9.7; 2-tailed P = .022).

Based on ExAC data, we calculated exceptionally high prevalence rates, with values up to 29.3 in 10⁶ and 16.5 in 10⁶ among East Asians and Europeans (Table 1). We also tried to be more conservative, taking into account, among missense variants, only those already reported in the FXI-related literature/databases, and again the calculated prevalence was higher than expected (eg, 20.3 in 10⁶ and 12.9 in 10⁶ in East Asian and European populations) (Table 2). Our calculations could in theory even be an underestimate, considering that we applied a very conservative selection of deleterious variants (ie, we considered only missense changes predicted by 7 of 7 algorithms, as well as splicing defects affecting the invariant dinucleotides GT-AG). Indeed, on the basis of the observation that only half of the missense mutations listed in the FXI Deficiency Mutation Database would survive this selection, we can speculate that we could have “missed” a significant fraction of genetic defects responsible for reduced FXI levels. In addition, some variants may have been missed for technical problems related to exome sequencing, for example: (1) gross deletions and rearrangements may go undetected; (2) calling of indel mutations still represents a tricky step in exome analysis, so they can go unnoticed; and (3) mutations in the promoter region, and those located outside of the canonical splice sites, are not included by design.

Our data pose the question of why such surprisingly high prevalence rates of homozygotes/compound heterozygotes can be calculated for FXI deficiency. As also stated earlier, the most obvious possible reason is that its prevalence has always been underestimated because of the relatively mild symptomatology associated with the disease.²  Another possibility could be related to a selective “advantage” conferred by FXI-null mutations to heterozygous or even homozygous carriers. Of note, Tucker and coworkers^23,24  observed evidence for disseminated intravascular coagulation after sepsis induced by cecal ligation and puncture in wild-type mice, but not in FXI knockout^−/− mice; this survival advantage could be related to an alteration of the cytokine response to infection and to a reduced activation of the contact system.²⁵  On the other hand, it could be conceivable that we do not observe in the general population the predicted rates of homozygotes/compound heterozygotes because of possible defects during fetal implantation for FXI-deficient women/embryos, or because of problems associated with pregnancy. To discriminate between the “early lethality” vs “absence of bleeding” hypotheses, we compared the expected and actual frequency of all variants deposited in the ExAC server in F11 and prothrombin (F2) genes. F2 was chosen on the basis of its comparable size and exon-intron structure with F11, and because its complete deficiency is considered to be incompatible with life.²⁶  Conversely to F2, this analysis showed a good tolerance of F11 to loss-of-function mutations (Table 5), thus strongly supporting the hypothesis that the mildness of bleeding symptoms is the most likely explanation. Moreover, it has to be underlined that, unlike for women showing fibrinogen or factor XIII deficiency, the incidence of spontaneous miscarriage in women with FXI deficiency is similar to that of the general population.²⁷ 

A potential limitation of this study is that our calculations were based on the assumptions that all mutations contribute equally to the phenotype, and that the analyzed populations are truly panmictic. Considering the first assumption, some of the identified FXI mutations have been shown to have a dominant-negative effect in the heterozygous state due to the dimeric structure of circulating FXI, leading to a dominant pattern of inheritance.²  Therefore, the actual prevalence of moderate to severe FXI deficiency could be even higher than the one we calculated under the assumption of a “true” recessive disorder. Considering the assumption of random mating, it has to be kept in mind that >1 billion people live in societies where consanguineous marriages are accepted, and, worldwide, 15% of all newborns have consanguineous parents.²⁸  Consanguinity is a well-known factor able to significantly influence the prevalence rates for autosomal-recessive disorders.²⁹  Consanguinity rates vary greatly between and within countries, with highest prevalence in populations, like the ones that we analyzed, coming from North Africa, the Middle East, and South Asia.^28,30,31  This could result in “true” prevalence data even higher than the ones we calculated from allele frequencies.

A final consideration pertains to some outcomes of our analysis: the confirmation that, among ExAC populations, type II and type III mutations are virtually exclusive to Europeans, and the discovery of specific prevalent mutations in other populations (up to 28% of all mutated alleles; Table 4). This information, tentatively indicating the F11 regions to prioritize for the molecular screening, could be of great value for diagnostic purposes, especially for countries where public health resources are scarce. Moreover, our results stress the notion that the availability of population-based genomic data will greatly improve our knowledge of the mutational spectrum as well as the prevalence of rare human genetic diseases.

Finally, recent evidence suggests the possible utility of FXI-targeted anticoagulants in treating atherosclerosis³²  and hypertension,³³  as shown in mouse-model studies, and in preventing thrombosis in patients undergoing knee arthroplasty.³⁴  These therapies, however, might not be suitable to FXI-deficient patients, and, before considering such treatments, measurement of FXI plasma levels should be recommended as FXI deficiency, besides the well-known high frequency among Ashkenazi Jews, is also unexpectedly frequent in other populations.

The authors thank the Exome Aggregation Consortium and the groups that provided exome variant data for comparison. A full list of contributing groups can be found at http://exac.broadinstitute.org/about. Special thanks to Patricia Corcoran (Tucson, AZ) for her meticulous proofreading.

M.M. was a recipient of the World Federation of Hemophilia Clinical Research Grant Program 2015.

Contribution: All authors participated in the conception and design of the study, analysis and interpretation of data, and revision of the manuscript; R.A. conceived the study, wrote the manuscript, and contributed to the statistical analysis; E.M.P. was responsible for the statistical analysis of the data; V.R., F.P., M.M., and O.S. participated in the writing of the manuscript and critical review; and S.D. contributed to the writing of the manuscript and supervised the entire study.

Conflict-of-interest disclosure: F.P. has received honoraria for participating as a speaker at satellite symposia and educational meetings organized by Ablynx, Bayer, Grifols, Novo Nordisk, and Sobi; is the recipient of research grant funding from Ablynx, Alexion, Kedrion Biopharma, and Novo Nordisk paid to Fondazione Luigi Villa; has received consulting fees from Freeline, Kedrion Biopharma, LFB, and Octapharma; and is a member of the following scientific advisory boards: Ablynx and F. Hoffmann-La Roche Ltd. The remaining authors declare no competing financial interests.

Correspondence: Stefano Duga, Department of Biomedical Sciences, Humanitas University, Via A. Manzoni 113, 20089, Rozzano (Milan), Italy; e-mail: stefano.duga@hunimed.eu.