Genotypes, Phenotypes, and Whole Genome Sequence: Approaches from the MyLifeOurFuture Haemophilia Project

INTRODUCTION

Historical perspective

Cloning of the genes encoding coagulation factors VIII (FVIII) and IX (FIX) led to advances in therapy for haemophilia A and B and improved our diagnosis of these disorders and their genetic carrier states [1–5]. These advances allowed the development of recombinant factor products and their subsequent modifications, as well as serve as the basis for gene therapy approaches for treatment. Genetic diagnosis started with linkage analysis which utilizes a genetic marker that co-segregates with hemophilia in the family. Linkage is now rarely needed because direct DNA variant analysis is successful in determining a likely causative variant in greater than 90% of families [6,7]. Over 3000 different likely causative DNA variants have been reported to date in the F8 and F9 genes [8–12]. The discovery of F8 inversion variants in introns 1 and 22 that disrupt the F8 structural gene explained the cause of hemophilia in almost one-half of patients with severe hemophilia A [13,14], and increased knowledge of structural variation in genetic disease, in general.

The impact of F8 and F9 genotypes on haemophilia outcomes

F8 and F9 genetic information is used in reproductive planning, including in pregnancy and neonatal management [15]. The knowledge of genetic carrier status of the mother and specifics of the genotype related to hemophilia disease severity can guide evaluation of the mother’s personal bleeding risks and plans for treatment, as needed, as well as inform potential risks in the fetus and neonate. While the optimal mode of delivery for a fetus with severe haemophilia is debated, a cesarean delivery may be the safest approach for many haemophilic infants [16]. Regardless of mode of delivery, knowledge of fetal status can guide management decisions during labor and delivery, including avoidance of procedures and instrumentation that can cause neonatal trauma, such as forceps or vacuum-assisted extraction in vaginal and cesarean deliveries [15, 16]. The potential for noninvasive testing of maternal blood for diagnosis of haemophilia in the fetus should make prenatal diagnosis more common in the future [17,18]. In addition, immediate medical evaluation of the neonate and treatment, if needed, is possible when the haemophilia diagnosis is known at birth.

Additionally, specific kinds of F8 and F9 genotypes are associated with increasing the patient’s risk of developing a neutralizing antibody (inhibitor) against infused exogenous factor replacement therapy [19–22]. In haemophilia B, inhibitor formation occurs in 3–5% of patients with severe disease (<1% FIX activity), and the risk is increased further in those with partial or whole gene deletions [21]. Patients who develop inhibitors are at risk of a severe allergic reaction at the time of onset of the inhibitor [22,23]. For this reason, genetic analysis soon after diagnosis is important in clinical management of severe haemophilia B.

The presence of deleterious variants in F8 significantly impacts the risk for inhibitor formation in haemophilia A [19,20]. About 25–30% of patients with severe haemophilia A develop inhibitors within approximately 14 exposure days [24,25], which usually occurs in young children, given treatment patterns. Inhibitor risk is reported to be higher in population groups, including in African Americans, for reasons that have yet to be determined [21]. Inhibitor risk is lower in patients with mild and moderate disease (FVIII activity 1–40%) than in patients with severe disease, but inhibitors still develop with an incidence reported at 6.7% at 50 exposure days [20]. Inhibitor risk in mild and moderate haemophilia A is increased by prolonged factor exposure and is influenced by genetic variation in the F8 gene [20,26].

In patients with severe haemophilia A, the incidence of inhibitors is greatest in patients with disruptive structural variation such as large multi-exon deletions (67–88%) and comparatively less in those with haemophilia due to missense variants (<12%) [19,21]. Data from the recent Survey of Inhibitors in Plasma-Product Exposed Toddlers (SIPPET) study showed that risk of inhibitor formation was high in those with variants predicted to be null variants, while inhibitor risk was not influenced by type of product used (recombinant FVIII versus von Willebrand factor-FVIII containing plasma derived product) [27]. In contrast, in those with non-null variants, inhibitor risk was influenced by product type. This supports haemophilia A genotype being a strong driver of inhibitor formation and is consistent with the hypothesis that variants that result in little or no protein synthesis would put patients at most risk [28].

Haemophilia genotype also impacts the bleeding phenotype. Overall, patients with severe haemophilia B have less bleeding than those with severe haemophilia A. It has been postulated this is due to the nature of type of gene variant stemming from the observation that compared to severe haemophilia A, missense variants are more common in patients with severe haemophilia B [29]. In concept, missense variants may result in more production of a dysfunctional protein than a true null variant, and even that small amount of coagulation factor protein, undetected by the current factory activity assays, could result in modulation of bleeding in this disease. An uncommon, but extreme example of genotype affecting haemophilia phenotype is haemophilia B Leyden. In Leyden, genetic variants in the promoter region of F9 result in severe deficiency early in life, but with puberty FIX level increase significantly with commensurate resolution of bleeding symptoms [30]. In genetic carriers of haemophilia A, the presence of the causative variant is a strong predictor of bleeding in females, even in those with borderline or normal measured factor activity levels [31].

Variants in genes outside of F8 have been shown to be associated with varying factor levels or an increase or decrease in bleeding symptoms. For example, co-inheritance of haemophilia and prothrombotic genetic variants has been reported to decrease bleeding events and time to first bleed in some but not all studies [32–34]. Additional factors impact phenotype, including the pharmacokinetics of infused exogenous factor replacement and for haemophilia A, levels of von Willebrand factor (VWF) [35,36]. These and other factors, known and unknown, are in part the result of heritable variation in genes outside of the F8 and F9 genes.

Response to desmopressin therapy in patients with haemophilia A is modulated by the underlying genotype. Desmopressin increases endogenous FVIII levels in some patients with moderate and mild haemophilia A by provoking release of VWF and FVIII into the circulation. Some haemophilia A causative variants, particularly in the A3, C1, and C2 FVIII domains, impact normal VWF-FVIII binding and lead to a shortened half-life of circulating endogenous FVIII. A number of studies have correlated individual responses to desmopressin with the underlying causative DNA variant [37–39].

Gene therapy trials are underway for both haemophilia A and B [40]. Early Phase 1/ 2 results have produced responses in the subnormal to normal range in the majority of patients, warranting the development of Phase 3 studies. While the precise functional gene variant is not necessarily informative at this stage for patients’ response to gene transfer, knowledge of genotypes will permit future analyses of therapeutic outcomes based upon genotypes and intrinsic inhibitor risks.

EXPANDING KNOWLEDGE IN HEMOPHILIA AND ITS OUTCOMES THROUGH OMICS

Some haemophilia-causing DNA variants are located outside the usual targets for haemophilia DNA sequencing diagnostics, likely affecting regulatory regions beyond the coding sequences usually tested, such as variants positioned deep within introns [41–43]. In addition, our understanding of clinically relevant haemophilia phenotypes, including variations in bleeding severity and risk of inhibitor development, likely involve genetic and epigenetic variation outside the F8 and F9 genes.

Advances in technology and our understanding of the human genome now allows more comprehensive methods to study genomic variation [44]. Linking DNA variants, be they single nucleotide variants (SNVs) or structural variants such as copy number variants (CNVs), to protein structure and human phenotypes has greatly advanced our understanding of disease over the past decade [45]. In addition, studies of the transcriptome and proteome are key to advancing our understanding of disease states and their associated clinical outcomes [46]. For example, the immune transcriptome should perceptibly change in patients as they are exposed to factor concentrate during the time in which they develop an inhibitor and again in response to immune tolerance, if instituted [47]. When genomic data are paired with standardized phenotypic data, such approaches are powerful tools for scientific discovery and precision medicine.

The Trans-Omics in Precision Medicine (TOPMed) program of the National Heart, Lung, and Blood Institute (NHLBI) of the U.S. National Institutes of Health (NIH) has as its goal to use genomic and other –omic data to improve diagnosis and care for patients with heart, lung, blood, and sleep disorders and advance precision medicine. The MyLifeOurFuture (MLOF) project is a nationwide U.S. initiative which provides genotyping for patients and families affected by haemophilia and supports a research repository. The MLOF research cohort is part of the TOPMed program, poised to leverage -omics resources to advance our understanding of haemophilia and its treatment.

THE MY LIFE OUR FUTURE PROGRAM

Program structure

In 2012, two separate surveys, one distributed to hemophilia providers through the American Thrombosis Hemostasis Network (ATHN) and the other distributed to the patient community through the U.S. National Hemophilia Foundation (NHF), found that only approximately 20% of the hemophilia patients in the U.S. had their genotype determined. The most common reasons cited for not having the testing performed were the lack of medical insurance coverage and availability of testing. MLOF was developed to provide wide-scale access to free hemophilia genotype analysis for patients in the U.S. and to expand research in haemophilia to better understand disease mechanisms, complications, and more precisely inform treatment through the development of a research repository containing clinical (phenotypic) data linked to biological samples and genetic data.

The program is a formal, multi-sector collaboration among four entities: ATHN, NHF, Bloodworks Northwest (BWNW) formerly known as the Puget Sound Blood Center, and Bioverativ, formerly Biogen. Each of the four partners brings a distinct expertise and together govern the project through Steering Committee membership. ATHN, working with 135+ affiliated hemophilia treatment centers (HTC)s, provides HTC provider education, a secure infrastructure for data collection known as ATHN Clinical Manager, and oversees review of applications for use of the research repository and phenotypic data distribution for approved projects. Participating HTC providers contracted through ATHN enroll patients, obtain samples for genotyping, and provide clinical results to their patients. BWNW facilitates receipt of samples from sites, serves as the central genotyping laboratory, interprets and returns clinical genotype reports to HTC sites, and manages the research sample repository including distribution of samples for approved projects. NHF, a U.S. national patient advocacy organization with 52 chapters, educates the bleeding disorders community about the initiative through its publications, annual meeting, and local chapter events. These efforts help keep the community informed about project status and supports recruitment. Bioverativ, a biotechnology company with products for treatment of hemophilia, provides scientific collaboration and financial support. Decisions of the Steering Committee and its working groups are informed by HTC provider and patient input gathered through numerous surveys, social media forums, conference calls, and in person meetings. Project management and public relations services are provided by KYNE (New York, NY).

MLOF began with enrollment of patients followed at HTCs in the U.S. with a diagnosis of haemophilia A or B (See Figure 1 for project process). In 2015, the project was expanded to include known and potential genetic carriers of haemophilia A or B. Participants, and for children, their parents (participants/parents), were given the option of enrolling in the MLOF Research Repository, which received human subjects Institutional Review Board approval at each site. Participants/parents gave written informed consent, which included consent for whole genome sequencing and deposition of data into large databases such as the NIH database of Genotypes and Phenotypes (dbGaP). Permission to recontact through their HTC to request additional samples was also given. Participants are represented in the Research Repository by a study ID only; identity is known only to the enrolling HTC. For participants/parents who do not consent to the Research Repository, clinical genotyping is provided.

Figure 1. MLOF Process.

Patients are approached at their local U.S. hemophilia treatment center (HTC) and a sample for genotyping sent to Bloodworks Northwest (BWNW). Patients, or for children, their parent, can consent to have clinical data, through the ATHNdataset, and additional blood samples (plasma, serum, RNA, DNA) placed in the Research Repository. Researchers access the repository through application to the Research Review committee through the American Thrombosis and Hemostasis Network (ATHN).

DNA extraction and analysis are performed at the University of Washington and BWNW, which also maintains the DNA, plasma, serum, and RNA samples for the MLOF Research Repository. Initial F8 and F9 DNA analysis is performed utilizing a next generation sequencing (NGS) screen employing molecular inversion probes (MIPs) for DNA capture [48,49], targeting all F8 and F9 exons, splice sites, promoter and untranslated DNA regions, and 5’ and 3’ of coding sequences. MIPs also simultaneously detect known F8 intron 1 and intron 22 inversions using a custom MIP adaptation of the inverse-shifting PCR methodology described by Rossetti et al. [50]. Candidate F8 or F9 variants identified by NGS are confirmed in the BWNW CLIA-certified laboratory using a second method. In haemophilia patients in whom no likely deleterious DNA variant is identified, and in females with very low factor levels unexplained by their genotype, multiplex ligation-dependent probe amplification (MLPA) is used to detect structural variants such as partial or whole gene duplications. A clinical report is returned to the HTC with variant results classified guided by the most recent American College of Medical Genetics and Genomics (ACMG) guidelines for interpreting variant pathogenicity [51]. HTCs share the result with patients and families. Data are entered in the ATHN Clinical Manager database and shared with the CDC and EAHAD databases [8–11]. Findings from the first 3000 patient samples analyzed and details on the methodology have been published [52].

MLOF interim results

The initial phase of MLOF completed enrollment in December 2017 with 11,356 haemophilia patients, genetic carriers, and potential carriers enrolled at 111 HTCs. The distribution of subjects enrolled is shown in Table 1. In the 9453 subjects in whom F8 and F9 genetic analysis is complete, 687 unique previously unreported variants in 1111 individual subjects were found. The detection of a large number of variants previously unknown in haemophilia was surprising given the substantial body of work of others DNA sequencing the F8 and F9 genes in patients with haemophilia. A potentially causative F8 or F9 variant was not identified by this analysis in ~1.6% (112/7137) of male haemophilia patients (F8: 13 severe, 16 moderate, 58 mild; F9: 0 severe, 7 moderate, 18 mild). As of February, 2018, 1223 females were still undergoing genetic analysis to determine carrier status.

Table 1.

Characteristics of MLOF Subjects with Haemophilia Enrolled through December 2017^*

	Haemophilia A						Haemophilia B
Sex	Male			Female			Male			Female
	5688			532			1448			152
% in Repository	84%			83%			82%			89%
Severity**	S	Mo	Mi	S	Mo	Mi	S	Mo	Mi	S	Mo	Mi
No. by Severity	3115	1022	1551	16	17	499	498	579	371	1	2	149

^*On 7821 subjects in whom sequence analysis is complete as of 01/17/2018. This does not include data on confirmed carriers (996) in whom factor activity levels are normal or unknown, or in 633 females confirmed to be non-carriers of haemophilia by our genetic analysis.

^**S, severe; Mo, moderate; Mi, mild

Defining Pathogenicity for Variants in F8 and F9

Outside of the region encoding the FVIII B domain, the F8 and F9 genes had previously purported to be highly conserved with little benign variation in the coding regions, and thus it was common clinical practice to assume that DNA variants detected in either the F8 or F9 gene are the cause of the patient’s haemophilia A or B, respectively. This assumption has been questioned for F8 [53], and a study of F8 genetic variation in the 1000 Genomes Project supports the presence of considerable benign variation in the F8 gene across ethnic groups [54]. Interrogation of the ExAC database for F8 and F9 variants captured in whole exome sequencing of ~66,000 individuals further supports that there is considerable rare variation in both genes [55].

In the MLOF project, knowledge of the subjects’ haemophilia A or B status and simultaneous sequencing of both the F8 and F9 genes in all subjects makes us uniquely positioned to assess benign variation in the purported unaffected gene. We have identified numerous likely non-deleterious variants in both the F8 and F9 genes (see ref#52 for those in first 3000), and normal factor levels in males proved that at least 10 variants previously reported in haemophilia are benign variants. Understanding that not all F8 and F9 genetic variation causes haemophilia is essential to avoid over-assigning clinical significance to variants in the interpretation of clinical hemophilia genotype data.

Given the well-defined haemophilia phenotype and our in-depth understanding of the F8 and F9 genes and their roles in the pathogenesis of haemophilia, “over-calling” variants as disease-causing is likely less common in haemophilia than in many other disorders. Instances of incorrect interpretation of DNA variant pathogenicity that impacted clinical care in other diseases have been documented [56]. To reduce this risk, guideline for the interpretation of DNA variants were developed, including by the American College of Medical Genetics and Genomics [51,57]. Traditionally in haemophilia, only one affected member of a family has been genotyped. This precludes capturing familial segregation data, detection of compound variants, and genetic confirmation of suspected de novo variation that can be useful to support interpreting the pathogenicity of a variant.

A unique example of challenges in determining pathogenicity of variants using ACMG criteria is illustrated in the story of the F9 variant previously reported in DNA extracted from the remains of Tsarevich Alexei of the Russian royal family who was the great-grandson of Queen Victoria [58]. The splice variant (c.278–3A>G) is purported to cause aberrant splicing through bioinformatics analysis [58], although this has not yet been demonstrated in functional transcription assays [59]. Furthermore, although this variant is likely to impact the splice site, it does not impact the canonical splice positions −2, −1, 1, and 2 needed to assign pathogenicity in the absence of other data [51]. Additionally, because there are no affected descendants of Queen Victoria with haemophilia and genotype information, segregation data in the family is absent. Thus, from what is known of the royal family alone, there is insufficient data to classify c.278–3A>G. There was a report of this same F9 variant in a boy with severe hemophilia B [59] in Spain. The boy’s mother, but not his maternal aunt or maternal grandmother, was found to be a genetic carrier. His maternal grandfather was reported by history to not have haemophilia, but unfortunately his DNA was not available for analysis. The authors proceeded with studies to try to demonstrate this variant arose de novo in the family by ruling out non-paternity in the boy’s mother (a carrier) and her sister (a non-carrier) using STR haplotype analysis of the distal X chromosome. Although their findings were consistent with a de novo variant, given the small number of family members studied and the limitations in the genetic techniques used by the investigators at that time, this evidence is not strong enough to prove de novo (as other scenarios could explain this finding) and inform interpretation of the pathogenicity of this variant. This variant was also identified in MLOF, but further evidence to inform ACMG interpretation could not be discovered. Thus, while this famous variant may have changed world history, we do not yet have enough evidence to reassign this variant from a Variant of Uncertain Significance to Likely Pathogenic (>90% certainty) or Pathogenic (100% certainty) by ACMG criteria [51].

This case illustrates the importance of complete clinical, family, and genetic datasets. Through the MLOF project we are working to obtain data in families to establish segregation of variants or de novo occurrence to support classification of F8 and F9 variants as functional (pathogenic) or benign. To aid in this classification we plan to share data on our laboratory website (www.bloodworksnw.org/labs/genomics), listing ACMG support criteria and classification for the variants found by our laboratory including those in the MLOF project. Through these and other efforts, we should be able to strengthen the interpretation of genetic data in haemophilia, further inform annotation of known or newly discovered variants, and provide more certainty of genetic diagnoses for our patients.

MLOF and OMICS THROUGH THE NHLBI TOPMed PROGRAM

While a major purpose of MLOF was to genotype the F8 and F9 genes, participants were offered the opportunity to contribute their samples to a research repository for future study. Overall, 81% of participants elected to enroll in the research repository, which collected DNA, RNA, plasma, and serum protein samples and provides links to clinical data in ATHN. These data and samples are available to interested investigators through an application and review process.

The NHLBI TOPMed program is using high throughput -omics technologies to characterize molecular abnormalities or signatures associated with heart, lung, blood, and sleep disorders, to conduct analyses of data generated by this program, and to build and share tools for advanced data analysis. Through this resource, DNA from 5141 MLOF subjects have undergone whole genome sequencing (WGS) (TOPMed phases 2 and 3), and resources for DNA sequencing more individuals or performing other -omics may be available in the future. This WGS data should lead to new knowledge, as prior genetic studies in haemophilia have focused mostly on a gene targeted sequence analysis, an approach that is insensitive to DNA variants outside the targeted region and cannot fully characterize the extent of structural variants (SVs) such as large deletions and duplications.

Characterizing SV using NGS approaches remains a challenge. Study of the haemophilia WGS data, particularly of the F8 genomic region given the frequency of SV associated with haemophilia A, will advance our knowledge and applications in this area. A focused preliminary analysis of the first 2181 genomes has allowed us to explore using a WGS approach to analyze the F8 and F9 genomic regions for SV. We find that small and large SVs can be detected, including variants that are likely causal for severe haemophilia (factor levels < 1%) as they are predicted to disrupt gene structure. Moreover, the identification of SVs which impact both a haemophilia gene and neighboring gene(s) may have additional clinical implications and raises the possibility of previously unsuspected haemophilia-associated syndromes.

In addition, we have used initial WGS data to evaluate the VWF gene that encodes von Willebrand Factor (VWF), a large plasma glycoprotein which protects FVIII in circulation from proteolysis. Deficiencies in VWF lead to von Willebrand Disease (VWD), a disease that can phenocopy haemophilia A by causing low FVIII levels. We studied the VWF gene in WGS data from the 2,181 subjects with haemophilia sequenced in Phase 2 of the TOPMed WGS effort: 1955 had haemophilia A (200 mild, 352 moderate, 1403 severe) and 226 had haemophilia B (1 mild, 4 moderate, 221 severe). WGS identified 17 VWF gene variants previously reported in VWD [60]. Overall, 7.1% of all MLOF subjects had at least one VWD-associated VWF variant identified, a finding that may have significance with regard to FVIII levels and/or bleeding phenotypes in these individuals. VWF variants were more prevalent in subjects with mild-moderate haemophilia A in whom no causative F8 variant could be identified, suggesting that some patients diagnosed with haemophilia A in this cohort may have low FVIII due to VWD instead. VWF variants in patients with pathogenic F8 variants could impact the type of bleeding symptoms and bleeding severity in hemophilia A. While VWF is known to impact FVIII levels, other genetic variation also likely impacts hemophilia phenotypes. A multi-gene or WGS approach to analysis may be used in the future to better understand phenotypic variation in haemophilia and other bleeding disorders.

Future analysis of the MLOF data, including through investigator-initiated studies using the MLOF Research Repository, will provide new insight into haemophilia disease mechanisms, treatment, and complications. Applications for use of the Research Repository are reviewed by an independent review committee through ATHN (www.athn.org/what-we-do/national-projects/my-life-our-future.html). The first cycle of investigator-initiated studies were approved in 2017. The phenotypic and treatment data available through the ATHNdataset, biologic samples, and sequence data in the MLOF Research Repository, in concert with the TOPMed data, provide valuable resources for further investigation of inhibitors and other factors. The MLOF TOPMed cohort contains data relevant to this mission, including 634 haemophilia A subjects who have developed FVIII inhibitors (241 current; 393 past documented). In addition, WGS and limited phenotypic data will be available to researchers through the NIH database of genotypes and phenotypes (dbGaP).

SUMMARY

MLOF is the largest genetic program in haemophilia to date, providing genetic information for patients and their families to help inform clinical care and reproductive planning and to the haemophilia community to inform a better understanding of haemophilia genotype interpretation. In addition, the project exceeded its goal of 6000 patients enrolled in the Research Repository. Through the resources in this repository, the ATHN clinical database, and the NHLBI TOPMed Program, we are building an invaluable resource for research in haemophilia and associated disorders that couples omics data with extensive phenotypic data. Furthermore, MLOF has demonstrated the power of diverse organizations working together to advance science, expand knowledge and improve health outcomes.