Somatic mosaicism in inherited bone marrow failure syndromes

Abstract

Inherited bone marrow failure syndromes (IBMFS) are a heterogenous group of diseases caused by pathogenic germline variants in key pathways associated with haematopoiesis and genomic stability. Germline variants in IBMFS-related genes are known to reduce the fitness of hematopoietic stem and progenitor cells (HSPC), which has been hypothesized to drive clonal selection in these diseases. In many IBMFS, somatic mosaicism predominantly impacts cells by two distinct mechanisms, with contrasting effects. An acquired variation can improve cell fitness towards baseline levels, providing rescue of a deleterious phenotype. Alternatively, somatic mosaicism may result in a fitness advantage that results in malignant transformation. This review will describe these phenomena in IBMFS and delineate their relevance for diagnosis and clinical management. In addition, we will discuss which samples and methods can be used for detection of mosaicism according to clinical phenotype, type of mosaicism, and sample availability.

Introduction

A mosaic classically refers to a form of art assembled with small pieces of different materials of various colours, shapes, and patterns. Similarly, from a biological perspective, every individual is a complex mosaic of genetically different cells. In genetics, mosaicism is a phenomenon intrinsic to human life, characterized by the emergence of two or more cell populations with distinct genotypes within an individual [1,2]. Throughout life, thousands of variations accumulate in normal human tissues to a different extent, making evident that aging itself is a natural driver of mosaicism [[3], [4], [5], [6], [7], [8], [9]]. Although clonal expansion has been mostly attributed to cell-intrinsic factors (fitness effects provided by a variation that affects a cellular function, such as proliferation, differentiation, and susceptibility to apoptosis), extrinsic factors such as aging and changes in the bone marrow environment also shape clonality [[10], [11], [12], [13]]. In healthy individuals, most clonal cells do not impact cellular functions and some leave no detectable genomic footprint [11]. The vast majority of detectable clones are selected by an universal pattern of clonal dominance through positive selection seen in both cancers and healthy tissues [10,14].

In the blood, somatic mosaics are termed clonal haematopoiesis (CH). CH is a known prognostic marker in hematologic disorders such as myelodysplastic syndromes (MDS) and aplastic anaemia (AA), and a driver of myeloid malignancies (such as acute myeloid leukaemia (AML), MDS, and myeloproliferative neoplasms) [12,15,16]. In bone marrow failure syndromes (BMF), the prototypic mosaic disease is paroxysmal nocturnal haemoglobinuria, a clonal disorder due to somatic variants in PIGA [[17], [18], [19]]. Although it is rare, somatic alterations in genes associated with autoinflammatory diseases and immunodeficiencies can lead to marrow failure [13]. We focus here on mosaicism in the less explored setting of inherited bone marrow failure syndromes (IBMFS), first discussing the genetic concepts of mosaicism, and subsequently reviewing the roles and clinical implications of mosaicism in particular IBMFS.

Mosaicism, or post-zygotic variations, are DNA alterations acquired from a single zygote throughout life, being mainly classified as gonadal, if confined to germ cells, or somatic if acquired during or after the first mitotic division in the zygote (non-germ cells) [1,2,20]. The different types of mosaicisms and methods available for detection are reviewed elsewhere [1,[21], [22], [23], [24]].

Mosaicism can be broadly classified depending on its size and type as structural variations (SV), genomic alterations without copy number change, small insertions and deletions (indels), and single nucleotide variants (SNV)(Table 1) [22,25]. At the karyotype level, types of SV include large copy-number variations (CNV; duplications and deletions), translocations, inversions, and aneuploidy (Table 1). Larger SVs are commonly detected by standard cytogenetics with FISH or chromosome (chr) banding techniques, when larger than 3 Mb. Since 2004, higher resolution techniques have allowed the identification of submicroscopic CNVs (>50bp and <3 Mb) in human genomes; array-based comparative genomic hybridization (CGH array) was first developed to characterize the presence of global CNV (>50 kb) [23,26], but single-nucleotide polymorphism (SNP) arrays were subsequently introduced to simultaneously detect CNVs and allelic alterations such as loss of heterozygosity (LOH; Loss of DNA segments in one of the two parental chromosomes leading to allelic imbalance) and copy-neutral LOH (CN-LOH), in which allele segments encompassing heterozygous disease-causing variants are lost and replaced with a duplicate of the wildtype homolog [27].

Massive parallel sequencing of short DNA fragments (75-700bp) has revolutionized the characterization of mosaicism at the nucleotide level [21,22]. With improved sensitivity allowing detection of SNVs and indels as low as 1–5%, the limit of detection due to intrinsic technical error rates, DNA sequencing technologies (most commonly Illumina) can be used to either screen entire exomes and genomes or a select panel of genes (targeted sequencing). By using unique molecular identifiers, error-corrected massively parallel sequencing (ECS) has allowed the detection of clonal haematopoiesis at variant allele fraction (VAF) as low as 0.01% [28]. Long-read sequencing technology has also been introduced to screen DNA fragments with an average length of 10–100 kb, but up to 2 Mb [24]. Traditional tools such as Sanger sequencing and PCR based approaches can also detect mosaic events when present above a technical limit of detection. More recently, highly sensitive platforms have been increasingly used to study mosaicism and characterize clonal trajectories at the single-cell level, being particularly useful to determine co-occurrence of mosaic events [[29], [30], [31]].

Post-zygotic variations are generally tissue-specific. In blood, CH emerges at the level of haematopoietic stem cells (HSC) and is associated with a myeloid lineage bias, but recent single-cell DNA data also has shown CH in B and T lymphocytes [32].

Both germline and somatic variations can be identified in the blood, and their distinction is important for disease characterization. Mosaic SNVs and CNVs are often identified based on their VAF. While germline variants are expected to be at approximate VAF of either 50% or 100% (heterozygous or homozygous, respectively), if below or above these ranges (<30% or 70–80%), variants are generally identified as somatic. However, this approach is faulty and cannot definitively characterize all mosaic events: the VAF of mosaic clones and germline variants can overlap, and revertant mosaicism can alter the frequency of germline variants into unexpected ranges.

Definitive confirmation of germline or somatic origin requires parallel screening of different tissues from a single individual, usually to compare blood to non-haematopoietic tissues (germline controls). Germline variants will be present at similar frequencies in all tissues screened, and somatic variations predominate in a specific organ (Fig. 1A) [33]. An exception occurs in revertant mosaicism: since the mutant genotype is corrected to the wildtype, the frequency of a germline variation will be lower in haematopoietic tissue compared to germline specimens. In rare cases, mosaicism can occur during the first mitotic divisions of the zygote and screening of tissues derived from different germ layers is required to confirm a variant's haematopoietic origin [20,34].

In IBMFS, epithelial fibroblasts are commonly used as germline controls (Fig. 1A), requiring skin biopsies. Buccal swabs are more easily collected but can be contaminated with myeloid cells and not reliable in some cases [33]. Familial genetic testing and screening of sorted blood subpopulations, particularly purified CD3⁺ T lymphocytes, are common strategies used to confirm somatic mosaicism in blood. Although CH can be seen in T cells derived from the mutated HSC, it is usually at lower frequencies than in myeloid cells [32]. In gonadal mosaicism (also referred as germline mosaicism), sperm is commonly used to confirm whether a variation is restricted to the father's germ-cells (Fig. 1A).

IBMFS are caused by pathogenic germline variants in key pathways associated with haematopoiesis, ribosome assembly, and genomic stability. These diseases are characterized by a reduced “fitness” of HPSC, a term used in this review to characterize cells' growth rates and survival [35,36]. In contrast to acquired diseases, clonal selection in IBMFS is intrinsic to the pathogenic effects of disease-causing germline variants and appears to be shaped by restricted cell fitness rather than by extrinsic factors, thus acting as a genomic footprint of the underlying molecular dysfunction [37]. In IBMFS, somatic mosaicism is dichotomous, in that it can lead to reversion or compensatory rescue of cell fitness towards normal, but may also confer an uncontrolled proliferative advantage resulting in malignant clonal evolution (Fig. 2) [[37], [38], [39]]. After a clone is selected, clonal expansion will be determined by a cell's competitive fitness in the marrow [38].

In IBMFS, mosaic mechanisms are dependent on the specific disease and associated germline causes (Fig. 1B). Somatic pathogenic variants in myeloid cancer genes (oncogenic hits), CNV and large chromosomal abnormalities can cause malignant transformation to MDS/AML and increased risk of solid organ malignancy [40,41] [19,[42], [43], [44]]. In contrast, somatic rescue (SR) involves the gene that carries a germline pathogenic variant [37]. Mechanisms of direct SR are characterized by spontaneous correction or compensation of the disease-causing germline variant in a subset of cells (Fig. 1B). The correction of the mutant genotype to the wildtype is called revertant mosaicism, a natural form of gene therapy (specific to the mutated gene and nucleotide; Fig. 1B). Revertant variants (“back mutation”) and CN-LOH (also named uniparental disomy; UDP) are most common (Fig. 1B and Table 1) [37,45]. Reversion of genotype by CN-LOH requires a wildtype allele, which is not present in an autosomal recessive or X-linked diseases. Other revertant modifications include substitutions not correspondent to the wildtype but less deleterious than the original sequence, and second-site variants within the same gene that can neutralize or weaken the pathogenic effects of an inherited defect. Of note, direct SR can sometimes be harmful for patients, as noted with the emergence of clones with larger deletions and aneuploidies. Indirect SR is observed when compensatory SNVs arise in genes within the same molecular network as the germline-mutated gene, such as the TERT promoter and EIF6 mutations in telomere biology disorders and Shwachman-Diamond syndrome (SDS), respectively [[46], [47], [48]].

Telomere biology disorders (TBDs) are a spectrum of multiorgan disorders caused by pathogenic germline variants in a group of genes related to telomere maintenance and associated with excessive telomere shortening [49]. Dyskeratosis Congenita (DC) refers to the classical phenotypic presentation of the clinical triad of oral leukoplakia, nail dystrophy, and skin pigmentation, and very short telomeres compared to age-matched controls. DC presents with early-onset BMF, with or without pulmonary and liver findings [50]. However, many patients with telomere diseases do not have classic DC but present with heterogenous clinical manifestations and variable age of onset [51]. BMF, pulmonary fibrosis, and hepatic cirrhosis are the most serious manifestations of telomere disease.

The first example of mosaicism and SR in the telomeropathies were seen in DC individuals with germline TERC variants. Two brothers without BMF were unexpectedly found to have higher frequencies of the wildtype TERC allele over the mutant in blood by Sanger sequencing [52]. Since the proband with DC had the same heterozygous variant, the identification of unbalanced proportion of wildtype and mutant cells in some family members prompted investigation of revertant mechanisms by SNP-array, which confirmed the presence of a CN-LOH of the chr3q in myeloid and B cells from mosaic patients. A similar mechanism was indentified in four additional TERC patients.[52] Revertant mosaicism has been observed in another DC case with a DKC1 “back mutation” by whole exome sequencing (WES); the male proband was found with 14% of wildtype but expected to have 100% of mutant reads due to X-linked disease [53]. A second-site truncating TINF2 somatic variant found in cis with a TINF2 pathogenic germline variant was also reported as a compensatory mechanism in a patient with pulmonary fibrosis; mosaicism led to a nonsense-mediated decay of the mutant allele with diminished expression of the germline variant [54]. Other large mosaicisms reported in TBDs include X-inactivation and CN-LOH of chr1q [53]. Although there is limited evidence that these SR mechanisms can revert a patient's phenotype, they had been linked to mild disease presentations.

Recently, two independent groups have identified TERT promoter (TERTp) variants as clonal markers specific for TBD [46,47]. Recurrent TERTp variants at positions −124C>T, −146C>T, and −57A>C are positively selected in myeloid, B and natural killer blood cells of 10% of TBD patients, particularly with TERT and TERC germline variants, but also PARN and NHP2 [46,47,55]. In contrast to the historical link between telomerase reactivation and tumorigenesis, TERTp clones have not been associated with malignant transformation or telomere elongation in patients with telomere diseases. These clones likely compensate restricted cell fitness by upregulating TERT expression and telomerase activity.

CH is also important in the surveillance of patients with telomere diseases, as they have an increased risk of solid cancers and MDS/AML (Fig. 2C) [56]. Cancer susceptibility has been best documented in classical DC; after fifteen years of follow-up, the cumulative incidence of solid tumours and MDS/AML in DC was 20% by age of 65 and 10%–20% by age of 50 [56]. In non-DC patients, MDS/AML is the most common type of cancer, often with karyotypic aberrations involving chr7 [57]. The MDS in these patients features typical somatic variants in myeloid cancer genes, such as DNMT3A, TET2, ASXL1, SF3B1, U2AF1, RUNX1, TP53, and PPM1D [40,[57], [58], [59]]. However, data are still emerging, and the long-term prognostic impact of most of these gene variants in TBD is uncertain. CH associated with myeloid malignancies appears to be rare in TBD patients without MDS/AML [53,57,58].

Fanconi anaemia (FA) is caused by pathogenic variants in a group of genes involved in the FA/BRCA DNA repair pathway; patients typically present with BMF, congenital abnormalities, and propensity to cancers [45]. Increased chromosomal breakage upon treatment with DNA cross-linking agents such as diepoxybutane (DEB) or mitomycin C (MMC) is a hallmark of FA cells and a specific diagnostic functional test for FA [60]. Mosaicism in FA was first inferred from the presence of two lymphocytic populations in a chromosomal breakage assay, one sensitive to DEB/MMC, the other normal. Resistance of FA lymphocytes to genotoxic agents was later found to be caused by revertant mechanisms, which are frequent; these mechanism are observed in 15–25% of FA patients and most commonly described in peripheral blood of patients with germline variants in FANCA, FANC, and FANCD2 [45,60]. The mechanisms of SR reported in these patients include “back mutation” of pathogenic variants, CN-LOH, intragenic recombination, and second-site variants in cis with a germline variant (Fig. 1, Fig. 2D) [45,61].

In all these cases, mosaicism was initially suspected based on the results of the DEB/MMC assay, regardless of the FA-subtype. The chromosomal breakage assay is therefore critical not only for diagnosis of FA but also for identification of FA mosaic patients. Mosaicism is generally confirmed or suspected when resistant FA lymphocytes after genotoxic exposure are seen at levels of 20% and 60% (specific ranges are standardized within each laboratory) [45,61]. The presence of FANCD2 monoubiquitination in FA cells is also a marker of somatic mosaicism as it is absent in cells with a disrupted FA/BRCA pathway but normal in revertant cells [61,62]. In mosaic cases, FA diagnosis is confirmed by chromosomal breakage in fibroblasts, expected to show divergence from minimal abnormality in blood [63]. However, in one recent study, reversion of a tandem intragenic duplication FANCB to wildtype was not only present in almost all of proband's PB cells but also in 8% of fibroblasts by PCR-based assays [62].

Revertant mosaicism, mostly characterized in PB T-lymphocytes, has been associated with improved blood counts and milder clinical phenotype, due to increased fitness of the corrected cells [61,64]. Reversion can emerge in the HSC pool, in some patients leading to blood count normalization [[64], [65], [66]]. In a recent literature review, normalization of all PB lineages was seen in 30% of FA mosaic patients; they also rarely developed cancer and BMF [45]. From a diagnostic perspective, revertant mosaicism can partially or completely restore the activity of the FA/BRCA pathway, leading to inconclusive or negative chromosomal breakage data in peripheral blood. Such a misdiagnosed FA patient is at risk of toxicity to standard conditioning for hematopoietic stem cell transplantation (HSCT). To overcome false-negative results, many centers use fibroblasts in paired analysis with blood for chromosomal breakage assays [63]. With advances in DNA sequencing approaches and FA molecular diagnosis, revertant mosaicisms have been increasingly detected in routine WES and targeted panels. In these cases, screening of bulk blood cells may identify germline variants at an average VAF that is lower than the expected range and can be misinterpreted as somatic alterations.

FA patients also have a high propensity to develop MDS/AML and epithelial malignancies, and requires close surveillance [56]. FA mosaicism in the setting of MDS/AML has been mainly characterized by RUNX1 alterations and karyotype changes involving gains in 1q and 3q, chr7 abnormalities. Except for gain of 1q, these alterations increase the risk of leukemic transformation [44,67,68]. Gain of 3q results in increased EVI1 expression and has been linked with poor overall survival of FA patients [69,70].

Shwachman-Diamond syndrome (SDS) is a classical IBMFS, for which the dichotomous mosaic mechanisms discussed in this review were recently elucidated [48]. SDS is an autosomal recessive disease with a high predisposition to MDS/AML, and caused by biallelic variants in SDBS and other genes important for ribosomal biogenesis [40]. Initial reports showed del20q, chr7 abnormalities, and somatic variants in TP53 as recurrent in SDS, but not strongly correlating with leukemic transformation [71,72]. Del20q, which encompasses the EIF6 gene, in fact provided a good prognosis [73]. Somatic mutations in EIF6 were later found to be a clonal marker of SDS with a distinct role from TP53: while EIF6 variants compensated for the germline SBDS defect and ameliorated cell fitness (indirect SR), TP53 variants were linked to increased predisposition to clonal evolution. [48] TP53 clones were small and stable over years, not promoting leukemic clonal expansion unless monoallelic TP53 clones became biallelic by acquisition of a second TP53 variant, a TP53 deletion, or 17p CN-LOH (Fig. 2E). From characterization of these clonality patterns preceding leukemic transformation, new strategies for improved surveillance of pre-leukemic clones and risk stratification for MDS/AML can now be implemented.

SAMD9/SAMD9L (SAMD9/9L) syndromes are a new class of autosomal dominant disorders with manifestations ranging from BMF-like disease to MDS with monosomy 7 (−7). They are caused predominantly by missense germline gain-of-function (GoF) variants in two paralogue genes located on chr7q: SAMD9 (Sterile Alpha Motif Domain Containing 9) and SAMD9L (Sterile Alpha Motif Domain Containing 9 like). [74] Both genes are negative regulators of cellular proliferation and patients’ variants expressed in vitro cause growth arrest with increased apoptosis and translation impair [[75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86]]. The initial reports associated germline SAMD9 variants to a fatal, early-onset condition with Myelodysplasia, Infections, Restriction of growth, Adrenal hypoplasia, Genital phenotypes, and Enteropathy (MIRAGE), while SAMD9L pathogenic variants were described in families with progressive neurological phenotype, multi-lineage cytopenias and BM hypoplasia (Ataxia-Pancytopenia syndrome) [[75], [76], [77],87]. The disease spectrum of SAMD9/9L has further broadened to include non-syndromic familial MDS and AA [[80], [81], [82],88], autoinflammatory disease [89], and steroid-resistant nephrotic syndrome [90]. To date, 64 distinct germline variants had been reported in roughly 110 patients. [[75], [76], [77], [78], [79], [80],82,84,85,[87], [88], [89], [90], [91], [92], [93], [94], [95], [96], [97], [98], [99], [100]].

SAMD9/9L syndromes have a high propensity for somatic mosaicism in haematopoiesis, leading to either loss or modification of the ‘toxic’ germline SAMD9/9L variant (Fig. 2A). One distinctive mechanism is ‘adaptation by aneuploidy’ caused by non-random loss of chr7 containing the mutated SAMD9/9L allele. This results in monosomy 7 which is considered a (pre) malignant event with predisposition to acquire additional somatic mutations in known leukaemia genes like SETBP1, ASXL1, ETV6, and RAS pathway genes [78,82,88,101]. Conversely, truncating somatic loss-of-function (LoF) SAMD9/9L variant in cis to the germline SAMD9/9L and uniparental isodisomy of chr7q (UPD7q) are considered rescuing events [102]. The SAMD9/9L LoF variants were shown to ameliorate the growth-inhibitory effect of the germline mutant in vitro [[76], [77], [78],82,93,99,101]. However, clinical improvement in such patients is rare and has been implicated in only two related individuals carrying a germline SAMD9 T778I and different post-zygotic truncating SAMD9 mutations (R221X, R285X) [103]. Alternatively, UPD7q ensuing replacement of the mutant SAMD9/9L allele by wildtype copy results in true genetic reversion, with the potential to outcompete the monosomy 7 clone (referred to as transient −7) with long-term support of functional haematopoiesis, as long as 20 years after diagnosis [80,101]. From the diagnostic perspective, both monosomy 7 and UPD7q decrease the allelic burden of SAMD9/9L variants, posing a diagnostic challenge of germline variant being misinterpreted as somatic. Therefore, routine sequencing of germline specimens is a paramount in the diagnosis process of these syndromes.

The implication of these somatic rescue events on the treatment of SAMD9/9L disease is debatable. The general consensus for paediatric MDS with monosomy 7 is to perform HSCT as early as possible [102,104]. It remains to be answered if young SAMD9/9L patients with monosomy 7 and mild hematologic phenotype would benefit from a watch-and-wait strategy, since in some patients the monosomy 7 clone can disappear over time (thus far reported in young patients) [105].