Somatic Mosaicism in Inborn Errors of Immunity: Current knowledge, challenges, and future perspectives

Abstract

Inborn errors of immunity (IEI) are a diverse group of monogenic disorders of the immune system due to germline variants in genes important for the immune response. Over the past decade there has been increasing recognition that acquired somatic variants present in a subset of cells can also lead to immune disorders or ‘phenocopies’ of IEI. Discovery of somatic mosaicism causing IEI has largely arisen from investigation of seemingly sporadic cases of IEI with predominant symptoms of autoinflammation and/or autoimmunity in which germline disease-causing variants are not detected. Disease-causing somatic mosaicism has been identified in genes that also cause germline IEI, such as FAS, and in genes without significant corresponding germline disease, such as UBA1 and TLR8. There are challenges in detecting low-level somatic variants, and it is likely that the extent of the somatic mosaicism causing IEI is largely uncharted. Here we review the field of somatic mosaicism leading to IEI and discuss challenges and methods for somatic variant detection, including diagnostic approaches for molecular diagnoses of patients.

1. Introduction

Inborn errors of immunity (IEI) are a broad spectrum of monogenic disorders of the immune system that lead to a wide spectrum of disease including susceptibility to infection, immune dysregulation, autoinflammation, severe atopy, bone marrow failure and/or susceptibility to malignancies (1, 2). To date, there are more than 485 inborn errors of immunity with a broad spectrum of molecular mechanisms (1). Disease-causing genetic variants in IEI are characterized as germline, either inherited or arising de novo. However, there has been increased appreciation of the importance of genetic variants present only in a subset of cells leading to IEI. Such genetic changes arise post-zygotically, with resulting genetic heterogeneity generally referred to as ‘somatic mosaicism’ (3, 4, 5, 6). Currently classified as ‘phenocopies’ of IEI, somatic disorders are nonetheless intrinsic to the individual, can completely phenocopy disease seen with germline variants in the same gene, and require similar therapeutic approaches including targeted therapies and hematopoietic stem cell transplantation. Disease-causing somatic mutations can arise from the earliest stages of embryonic development to any age after birth, and in some cases have the potential to be passed on to offspring, thus blurring the lines between what is an IEI versus a ‘phenocopy’. Somatic variants have been described to cause autoinflammatory diseases (7, 8, 9, 10) and now a handful of other IEIs (11, 12). Some diseases have thus far only been discovered in mosaic forms, while others have a germline equivalent of the gene.

The timing of post-zygotic mutational events leading to somatic mosaicism dictates the tissue distribution of the genetic mutation or variant. Early-onset variants (e.g., at the two cell stage) have the potential to affect all nearly cell types/lineages, whereas later–onset mutational events will be restricted to specific cell types/tissues. For example, early somatic mutations in NLRP3 present in multiple cell lineages can lead to the most severe form of cyropyrin-associated periodic fever (CAPS), known as neonatal-onset multisystem inflammatory disease (NOMID) (9, 13). Whereas adult patients with late-onset of CAPS have myeloid-restricted somatic NLRP3 variants, consistent with a mutational event occurring at later stages of life (8, 14). Variants present in both somatic and germ cells are classified as ‘gonosomal mosaicism’, although in most cases this is challenging to prove unless the genetic variant (and disease) is passed on to offspring. Indeed, many patients with ‘somatic’ mosaicism may actually have gonosomal distribution, and counseling as to such is important for reproductive health. Alternatively, genetic changes present only in germ cells (gonadal mosaicism), not causing immunologic disease in that individual who then passes on the disease-causing variants to affected children (4). For example, as reported in a case of an unaffected male who had multiple children with different partners, all having activated PI3K delta syndrome 1 (APDS1) due to the same gain-of-function variant in PIK3CD (15).

As the field evolves, there has been a growing interest in somatic variants as an underlying cause of IEI, particularly in the large cohorts of patients who remain undiagnosed despite extensive genetic testing (16). Here we discuss the current knowledge of somatic mosaicism in IEI, methods and challenges associated with detection and functional characterization of somatic variants, and implications for diagnosis and treatment of affected individuals.

2. Somatic variants in IEI

2.1. Early discoveries of somatic variants in IEI

The concept of somatic variation in autoimmunity and chronic disease was suggested as early as the 1960s (17), and the first IEI associated with a post-zygotic mutational event was described in 2004. The initial report of somatic mosaicism in IEI was described in patients with genetically undefined autoimmune lymphoproliferative disorder (ALPS), a disorder of apoptosis leading to an accumulation of autoreactive lymphocytes (18). ALPS is characterized by lymphocyte dysregulation as evidenced by elevated double-negative T cells (19). While majority of the ALPS patients carry germline variants in FAS, the Rieux-Laucat laboratory (19) reported somatic variants in FAS in ‘sporadic’ cases of ALPS, identified by Sanger-based sequencing of FAS in DNA from purified double negative T cells (DNTs). Interestingly, some patients with ALPS harbored both germline and somatic variants in the same gene (19, 20) or different genes (e.g., somatic FAS and germline CASP10), in part explaining incomplete penetrance observed in families (21). Heterozygous somatic variants in the FAS gene account for disease in ~20% of patients with a clinical diagnosis of ALPS (22). RAS-associated autoimmune leukoproliferative disease (RALD) is another example of a lymphoproliferative disease associated with somatic variants in KRAS or NRAS (23, 24, 25) (26).

In 2005, somatic variants causing autoinflammatory disease were reported in patients with cryopyrin-associated periodic syndrome (CAPS), an autosomal-dominant inherited systemic autoinflammatory disease (27). CAPS is caused by gain-of-function (GOF) variants in the NLRP3 (NLR family pyrin domain containing 3) gene leading to a spectrum of disease characterized by fevers, rashes, and arthropathy (28). While the majority of these patients harbor germline NLRP3 variants, at the time ~40% were genetically undefined. Saito et al. identified NLRP3 mosaicism in a mutation-negative NOMID patient, with variant allele frequency (VAF) of ~20% using a sub-cloning strategy to sequence multiple copies of the NLRP3 gene from patient cells (27). Following this observation, an international collaborative study of 26 mutation-negative NOMID patients, identified somatic NLRP3 mosaicism in 18 of the 26 patients (69.2%), with estimates of mosaicism from ~4–35%, again using a sub-cloning strategy (9).

While Sanger sequencing of a purified population of cells or sub-cloning strategies were able to detect disease-causing mosaicism, these methods generally lack sensitivity for detecting somatic mosaicism. Sanger sequencing requires knowledge of the cell type harboring the variant, is low-throughput with sequencing of only highly-targeted regions of a known gene, and is not quantitative. Next-generation sequencing (NGS)-based strategies including targeted and unbiased approaches (exome and genome) are the current standard for diagnosis of germline IEI, and together with other high-throughput sequencing strategies have allowed for enhanced detection of somatic mosaicism in IEI (see Box 1 for key sequencing terms used in this text).

Box 1-. Key terms used in this text.

Term	Definition
Allele	One of 2 or more alternate forms of a gene at the same location.
Variant	A change in DNA that is different from a reference genome sequence.
Read depth	Number of sequencing reads computationally aligned to a reference sequence for a given genomic position. For example, the number of times a particular fragment of DNA was sequenced. Exome sequencing usually has a “read depth” of 100–300 at a particular location, while genome sequencing generally has a lower “read depth” of ~30–50. For detection of somatic mutations, knowing the read depth of the sequencing results will help to predict whether there was sufficient sensitivity to detect a somatic variant with a low VAF.
Variant allele frequency (VAF)	In the context of somatic mutations, VAF describes the percentage of sequencing reads at a given location with the variant of interest. For example, if 5% of sequencing reads at a given location reported variant, the VAF would be .05 (or 5%). For genes with 2 alleles (i.e., an autosome or X-chromosome in a female) this suggests that 10% of the cells tested carry one variant allele and one reference allele.
Targeted gene panel (TGP)	As a clinical test, this refers to the focused evaluation and sequencing of a pre-defined set of genes. Depending upon the test used, this may include only protein-coding regions or may also include regulatory regions known to be associated with disease.
Exome sequencing (ES)	Sequencing method that enriches and sequences protein-coding regions of a genome, referred to also as whole exome sequencing (WES). This approach facilitates discovery of novel genetic variants. However, exome sequencing does not always provide adequate coverage of the entire exome, and some IEI genes are not well sequenced by exome or genome. For example IKBKG (the gene causing NEMO syndrome) or NCF1 (the most common cause of autosomal recessive chronic granulomatous disease, CGD) due to pseudogenes or repetitive regions that are difficult to sequence and align.
Whole genome sequencing (WGS)	Method for sequencing the entire genome, both protein-coding and non-coding regions. In addition to broad genomic coverage, WGS provides the ability to better predict structural variants. Similar to exome sequencing, some IEI-associated genes are not well-sequenced by this method.

2.2. Germline IEI with somatic ‘mimics’

Germline IEI ‘mimics’ are caused by somatic variants in genes previously associated with IEI, often leading to clinical manifestations similar to the germline disease, as is the case with somatic FAS variants in ALPS.

Autoinflammatory diseases appear to be particularly enriched for disease-causing somatic mosaicism, likely due to the dominant nature of these diseases in which a gain-of-function mutation even in a small percentage of cells is sufficient to initiate a cascade of systemic inflammation (Table 1). In addition to somatic NLRP3-associated disease discussed above, somatic mosaicism has been reported in five other autoinflammatory syndromes with known germline disease (Table 1). These include Blau syndrome due to somatic mosaicism in the NOD2 gene (VAF~ 5–11% depending on the tissue tested (29)), STING-associated vasculopathy with onset in infancy (SAVI) due to TMEM173 mosaicism in 2 unrelated patients (30, 31), and high frequency somatic NLRC4 mosaicism in NOMID (VAF-60–68%) (32). TNF receptor‐associated periodic syndrome (TRAPS) due to somatic variant in TNFRSF1A has been reported in two unrelated patients, with one patient having an early onset of the disease and similar VAF in multiple tissues including his sperm cells (33). The second patient presented with late onset of disease (34), and the variant was only detected in the patient neutrophils, B cells, and NK cells (VAF 25–45%), consistent with a late mutational event. Germline JAK1-GOF has been previously described in a family with multi-organ immune dysregulation in the affected mother and her two children (35). Work from Bogunovic laboratory, identified a somatic JAK1 variant with a VAF ~ 27% in the peripheral blood of a patient with clinical presentation similar to germline disease. In this case, the variant was present in multiple tissues including PBMCs, granulocytes, cheek swabs, and epithelial tissue suggesting an early mutational event. By combining genotyping with custom scRNA sequencing technology, they identified that the expression of variant mRNA transcript was highest in the CD56^hi NK cells (69%) within the immune cell compartment, suggesting a selective advantage for the variant in this cell type.

Table 1. Somatic mosaicism causing disease with corresponding germline IEI.

Disease phenotype		Gene	Chr	Types of mosaicism	Cell types/ tissues affected	VAF	Mechanism	Sequencing to detect & validate	Validation/evidence for variant pathogenicity	Ref.
Autoimmune lymphoprolife rative syndrome (ALPS)		FAS	Chr10	Somatic	PBMCs, DNTs	1–50%	LOF	S, ES	Defective FAS mediated lymphocyte apoptosis	(19, 21, 22, 97)
RAS-associated autoimmune leukoproliferative disease (RALD)		KRAS	Chr12	Somatic	T,B, NK cells	NA	GOF	S	Defective lymphocyte apoptosis, BIM levels & proliferation of primary lymphocytes	(23, 24, 26)
		NRAS	Chr1	Somatic	PBMCs	50%	GOF	ES, Pyr	BIM expression in patients cells, defective lymphocyte apoptosis	(11, 25, 26)
Auto inflammatory disorders	CAPS	NLRP3	Chr1	Somatic	Multiple tissues	2–45%	GOF	S, sub-cloning & S, ES, AS	NF-κβ reporter activity in cell lines transfected with variant or WT response in the presence of ASC; necrotic cell death of THP-1 cells transfected with the variant	(8, 9, 11, 13, 14, 51, 98, 99, 100, 101)
	NLRC4 GOF	NLRC4	Chr2	Somatic	Multiple tissues	30%	GOF	ES	Patient iPSC-derived myeloid cells with WT or variant response at baseline and after LPS stimulation	(32)
	TRAPS	TNFRSF1A	Chr12	Somatic GS	B, NK cells; Multiple tissues, sperm cells (GS)	18–30%; 4–21%	GOF	Targeted panel, S	Previously reported as a germline variant; response to treatment with interleukin-1 (IL-1) blockade	(34)
	Blau syndrome	NOD2	Chr16	Somatic GS	Multiple tissues	4.9–11%; 0.9–12.9%	GOF	S, AS	Previously reported as germline variant	(11, 29, 34)
	SAVI	TMEM173	Chr5	Somatic	Multiple tissues	NA	GOF	S	IFNβ1 reporter activity in cell lines transfected with variant or WT; IFN response gene signature in patient primary samples	(30, 31)
	JAK1 GOF	JAK1	Chr1	Somatic	Multiple tissues	27%	GOF	ES ddPCR	Patient EBV-B cell lines with variant or WT JAK1/STAT response at baseline and after cytokine stimulation, phosphorylation of STATs in primary cells, scRNAseq with genotyping; rescue of JAK1 hyperactivity in patient cells after treatment with tofacitinib	(102)
Chronic Granulomatous disease		CYBB	ChrX	Somatic	Leukocytes	NA	LOF	S	NADPH oxidase activity in patient neutrophils	(103)
Hyper IgE syndrome		STAT3	Chr11	GS	Multiple tissues	NA	LOF	S	Recurrent infections and CMC, milder phenotype; transmission of pathogenic STAT3 variant to offspring with classical features of HIES	(36)

Abbreviations: S, Sanger; Pyr, pyrosequencing; ES, exome sequencing; AS, amplicon sequencing; GS, gonosomal, CAPS, cryopyrin-associated autoinflammatory syndrome; CINCA, chronic infantile neurological, cutaneous, and articular syndrome; DNT, double-negative T cells; BIM, BCL-2-interacting mediator of cell death; ASC, apoptosis-associated speck-like protein containing a caspase recruitment domain; WT, wild type; GOF, gain-of-function; LOF, loss-of-function; NA, not available; LPS, lipopolysaccharide; TRAPS, tumor-necrosis-factor-receptor-associated periodic syndrome; SAVI-STING, associated vasculopathy with onset in infancy; CMC, chronic mucocutaneous candidiasis; HIES, Hyper IgE syndrome

Somatic mosaicism has been reported in other IEIs such as described in a female patient with late-onset of chronic granulomatous disease (CGD) due to a somatic variant in CYBB. The variant was present in a heterozygous state and was detected only in the DNA derived from her white blood cells, but was completely expressed in the mRNA of these cells, suggesting skewed X-inactivation. The disease phenotype is consistent with a late mutational event, leading to clonal expansion of the mutated bone marrow stem cell due to growth advantage over healthy cells.

In some cases, the somatic disease may not completely ‘mimic’ the germline disease and present with a milder, incomplete phenotype as was described in in two unrelated males with an intermediate hyper-IgE syndrome (HIES) phenotype identified with mosaic variants in STAT3, the gene causing AD-HIES. Both patients had normal numbers of Th17 cells, but presented with chronic mucocutaneous candidiasis (CMC), staphylococcus infections, and elevated IgE and transmitted the STAT3 variant to their children (36). Investigation of such mosaic individuals with an incomplete phenotype has the potential to shed novel insights into the underlying disease mechanism, for example by studying the intrinsic effects of cells with or without the variant.

2.3. Discovery of monogenic immune disease due to primarily somatic mutations

One of the most exciting advances in the field has been the discovery of diseases that are largely somatic only through re-analysis of otherwise ‘negative’ exome sequencing and other more targeted strategies to detect mosaic variants (Table 2).

Table 2. Disorders that are predominantly somatic leading to ‘mimics’ of inborn errors of immunity.

Disease phenotype	Gene	Chr	Types of mosaicism	Cell types/ tissues affected	VAF in blood or cell type	Mechanism	Sequencing to detect & validate	Validation/evidence for variant pathogenicity	Ref.
Hypereosinophilic syndrome	STAT5b	Chr17	Somatic	Lymphoid and myeloid cells	10–50%	GOF	Targeted panel	STAT5b responsiveness to cytokines in patient T cells	(38, 40)
Inflammation, neutropenia bone marrow failure, and lymphoproliferation caused by TLR8 (INFLTR8)	TLR8	ChrX	Somatic	Multiple tissues	8–26%	GOF	ES, ddPCR	Patient iPSC-derived myeloid cells with WT or variant TLR8 response to TLR8 ligand	(12)
VEXAS syndrome	UBA1	ChrX	Somatic	Myeloid cells	35–80% in blood 60–95% in myeloid cells	LOF	ES ddPCR	Ubiquitinylation in patient primary cells, Uba1 deficient zebrafish models to recapitulate patient phenotype	(10, 44, 45, 47, 48, 104, 105, 106, 107)
Immune dysregulation with autoinflammation and dysmorphy	IL6ST	Chr5	Somatic	Multiple tissues	15–40%	GOF	ES, AS	Patient EBV line with variant GP130/STAT3 response at baseline and after cytokine stimulation	(39)
Sweet Syndrome	PIK3R1	Chr13	Somatic	Neutrophils	100%	GOF	ES	Transduced cell line (HL60) with variant or WT differentiated into neutrophils with measurement of response to inducers of chemotaxis, phospho-S473 AKT phosphorylation and IL-1R1 protein levels	(49)

Abbreviations: GOF, gain-of-function; LOF, loss-of-function; ES, exome sequencing; ddPCR, droplet digital PCR; AS, amplicon sequencing; VAF, variant allele frequency; WT, wild type; VEXAS, vacuoles, E1 enzyme, X-linked, auto inflammatory, somatic, APDS1, Activated phosphoinositide 3-kinase delta syndrome 1

Several disorders leading to GOF in cytokine signaling have been discovered as primarily somatic in affected individuals. Somatic GOF variants in STAT5B cause a hypereosinophilic syndrome without a known germline equivalent identified by the Milner laboratory in two unrelated female patients and 1 male patient by sequencing on a targeted gene panel (37, 38). The variant was detected in immune cells, with a selective advantage in T cells and eosinophils based on evidence of higher VAF (≥50%) in these cells and lower VAF in B cells, and dendritic cells. A mosaic inframe 4 amino acid deletion in the IL6ST gene leading to constitutive GP130 cytokine receptor signaling was identified in a pediatric patient presenting with a novel syndrome of immune dysregulation with autoinflammation and dysmorphy (39). The deletion was confirmed by amplicon deep sequencing and was found to be present in multiple sample types (blood, urine sediment, hair bulbs and buccal swab) with most samples having a VAF of 30–40%, and a lower VAF of 15% in the blood, suggestive of a disadvantage to peripheral blood immune cells (39). Loss of function variants in the IL6ST gene has been previously reported in patients with elevated IgE levels (40, 41, 42), however, normal IgE levels and persistent elevation of inflammatory markers with constitutive activation of IL6 signaling distinguishes the mosaic IL6ST syndrome from its germline equivalent.

We recently described somatic variants in the TLR8 gene as a primary mechanism of a monogenic IEI termed INFLTR8 (inflammation, neutropenia, bone marrow failure, and lymphoproliferation caused by TLR8) (12). Originally identified in 6 boys, patients all presented with neutropenia causing infectious susceptibility and lymphoproliferation. Exome sequencing identified novel genetic variants in the X-chromosome gene TLR8, which encodes toll-like receptor 8 (TLR8), an endosomal TLR that recognizes single-stranded RNA. Variants were mosaic in 5/6 patients, with four patients harboring the same nucleotide change. The sixth patient had a different de novo germline variant, and died at a young age due to severe infections. In patients with mosaicism, a VAF of 8–26% was detected in the peripheral blood with similar allele frequencies in sorted immune cells, saliva, and fibroblast lines (5–30%). The variant was not detected in the fibroblasts of one patient, suggesting mosaicism was restricted to the hematopoietic compartment in that patient.

What turns out to be the most common mosaic immune disease was recently discovered in 2020 by Beck et al. (10), VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic) syndrome. This disease is caused by somatic variants in in the X-linked gene UBA1, which encodes for ubiquitin-activating enzyme 1. Present primarily in older men, but also reported in women (43, 44, 45, 46), this disorder leads to a broad clinical phenotype with generally treatment-refractory autoinflammation with symptoms including severe inflammation, rash, rheumatologic, and/or hematologic disease, including myelodysplastic syndrome (10, 46, 47, 48). Variants in this disorder are largely found only within the myeloid compartment with a VAF~35–80% in peripheral blood and VAF~60–95% in myeloid cells. Since the initial discovery of VEXAS syndrome in late 2020, this disease is now recognized as what is likely the most common cause of monogenic autoinflammation in adults, with hundreds of patients reported with a broad clinical phenotype. A recent study from Beck and colleagues screening for UBA1 variants in a large healthcare system identified a high prevalence of pathogenic UBA1 somatic variants, with detection in ~1 in 4200 men and 1 in 26,000 women older than the age of 50 years (46).

Bhattacharya et al. recently identified a gain-of-function missense variant in the PIK3R1 gene restricted to skin infiltrating neutrophils of a 51 year-old patient with Sweet syndrome (49). This clinical syndrome is generally characterized by fever, leukocytosis, and painful skin rash with a dense dermal neutrophilic infiltrate. Using genomic and transcriptomic studies on cells from affected skin from this patient, they identified multiple possible somatic variants in skin-infiltrating neutrophils from an affected area, including in PIK3R1, which was validated by Sanger sequencing as present in neutrophils from the blood. Transcriptional data from the skin suggested that there was increased PI3K signaling. Sanger sequencing confirmed the presence of PIK3R1 variant in peripheral blood neutrophils. Functional studies in a cell line demonstrated that the variant resulted in activation of the PI3K pathway and increased expression of IL1R1 protein, increased neutrophil respiratory burst, and neutrophil migration in response to IL1β. Germline GOF variants in PIK3R1 are the cause of activated PI3Kdelta syndrome (APDS) type 2, however these patients primarily harbor exon deletions, exon skipping, or splice site mutations, without any apparent effect on neutrophil migration responses as described in the Sweet syndrome patient (50). Identification of this underlying mechanism of disease enabled treating this patient with IL1R blockade, leading to improvement of clinical disease.

2.4. Other forms of mosaicism relevant to IEI

Genetic variants limited to somatic cells are not inherited. However, vertical transmission of the variant is possible if it is also present in the germ cells (gonosomal mosaicism). For patients with IEI, gonosomal mosaicism has been best described in healthy parents who pass disease on to their children, leading to unexpected recurrence of disease in a family. For example, Jiménez-Treviño, Santiago et al. (51) re-investigated 4 families with a presumed de novo NLRP3 mutations, and detected gonosomal mosaicism in the healthy mother of one family. The variant was present in the healthy mother at a low frequency (VAF ~2–3%) in hematopoietic and epithelial cells. In some cases, a parent may have a mild phenotype, as was the case of a father who presented with late-onset mild presentation of Blau syndrome including presence of multiple noncaseating granulomata, maculopapular skin rash, and bilateral anterior uveitis, with a relatively benign course, and lack of joint involvement. In contrast, both his two daughters with the same variant in their germline had early-onset disease with a more severe phenotype and (52). Investigation of the mildly-affected father demonstrated that he harbored the same NOD2 mutation at a low VAF of ~8–13% in blood, mucosa, and urinary cells. The inheritance of the disease-causing variant demonstrated gonosomal mosaicism in this family.

The distinction between a germline variant that is de novo or due to vertical transmission from a parent with gonosomal mosaicism is challenging, however this has significant clinical implications for the families and counseling for recurrence risk. Deep sequencing of parental blood for the pathogenic mutation identified in a ‘proband’ may help to rule out the presence of gonosomal variants in the parents and better assist in recurrent risk counseling in families, while sequencing of germ cells (e.g., sperm or egg) would be required for gonadal mosaicism.

In some rare instances, somatic variants can arise in a gene that harbors a disease-causing mutation, leading to partial or complete loss of its functional consequence in the affected cells. This naturally occurring phenomenon of revertant mosaicism can occur as back mutations that results in spontaneous correction of the mutant sequence. For example, spontaneous inversion of an inherited mutation observed in the T cells of a patient with Wiskott Aldrich syndrome (WAS) (53) or due to second-site mutations leading to compensatory changes (54). Somatic reversion mutations are seen in several IEI, for example Wiskott Aldrich syndrome, multiple forms of severe combined immunodeficiency (55, 56, 57, 58, 59), and DOCK8 deficiency (60).

3. Methods and challenges associated with detection and functional characterization of somatic variants

Exome and genome sequencing by NGS has led to the rapid discovery of germline IEIs over the past decade (1). While these sequencing techniques have enabled the discovery of some somatic mosaicism in some IEI, for example VEXAS syndrome, there are significant limitations for detection of somatic variants including coverage depth, sequencing errors or artifacts, and the ability of variant detection strategies to identify variants with low allele frequency. With more than 60% of patients with apparent IEI undiagnosed following exome or genome sequencing (16), the IEI field is at the tip of an iceberg of sorts with regards to discovery of somatic variants causing disease. Discovery of disease-causing somatic mosaicism in the research lab and the clinic will require a multi-pronged approach to genetic discovery and testing of variants.

Two of the most critical points for discovery of genetic mosaicism are ensuring that 1) the proper source of DNA is tested, and 2) the sequencing strategy, including variant calling, is sufficiently sensitive to detect disease-causing variants.

3.1. Considerations for tissue tested

The first step requires selecting the correct tissue or cell to subject to genetic testing, with a goal of sequencing the tissue most likely to have a VAF detectable by the sequencing strategy chosen. For example, somatic mosaicism limited to the hematopoietic compartment or greatly enriched in a particular immune cell type, for example in DNTs of ALPS patients, is unlikely to be detected in a buccal swab DNA sample. This is particularly relevant for adult-onset disease, in which individuals are unlikely to have carried the somatic variant since birth. For most IEI or immune-related disorders, peripheral blood is probably a good starting source of genetic material, as it contains the relevant cell types, i.e. circulating immune cells. However, one could envision that this may not always be the case, for example the concept that a somatic mutation in a tissue-resident immune cell would lead to accumulation of that cell at the site of disease, e.g. enteropathy, encephalopathy or skin inflammation. Two recent studies highlighted the role of somatic evolution in ulcerative colitis (UC) epithelium, driven primarily by accumulating somatic mutations in genes related to the IL-17 pathway (61, 62). The clonal expansion of somatically mutated clones in the inflamed epithelium of these patients is suggested to increase the risk for cancer in UC patients. By contrast, when mutational events occur at an early stage of embryogenesis, the variant may be distributed across multiple tissues, but lead to primarily an immune phenotype due to expression patterns, as seen with TLR8 and NOD2. In such cases, if there is not a selective survival advantage for a particular immune cell, the VAF may be similar in different tissues (e.g., saliva, fibroblast, immune cells) as was the case with NOD2, and some patients with TLR8 GOF, making the DNA source less relevant.

Overall, given the variables in where the somatic mutation originally arose and whether or not it imparts a selective growth advantage to a particular cell type, several approaches to tissue selection could be selected including whole blood, disease-specific tissue (e.g., small bowel with enteropathy, intestinal epithelium of patients with ulcerative colitis (61, 62)), and/or a purified immune cell identified as aberrant in the particular disease. Ultimately, using a combination of tissue sources depending on the immunologic disease may have the best chances of success for discovery of disease-causing somatic mutations.

3.2. Detection of somatic variants in IEI by NGS

3.2.1. Sequencing strategies

The second major challenge of somatic variant detection is identifying the best sequencing strategy to detect a disease-causing variant. Somatic variants with VAFs as low as 0.01–0.04 (1–4%) have been detected in some autoinflammatory disorders (Table 1), suggesting that low level of somatic mutation is sufficient to cause disease pathogenesis. The lower limit of VAF that can lead to an IEI is unknown, and likely differs based on the disease phenotype. For detection of somatic mosaicism in IEI, important considerations include depth and accuracy of sequencing.

The most common approaches to unbiased sequencing and gene discovery are by exome and genome sequencing. Exome sequencing generally achieve read depths of 100–300x, and based on the variant calling strategy used, has the potential to identify VAFs as low as 0.1 (10%) in an unbiased manner (63). Manual review of exome sequencing data has the potential to identify lower frequency variants, but is limited by the need to focus on a specific gene/location and the time for such analysis by a person. Genome sequencing (WGS) generally has 30–50x coverage, further limiting detection of somatic variants to those with a relatively high VAF. For instance, WGS at 50x coverage was able to detect 95% of the variants at VAF~15% or higher, and only 10% variants with VAF<15% in primary AML tumor samples (64, 65). Newer unbiased techniques, such as high-depth (>1000x) exome sequencing or single-cell sequencing techniques used in the cancer field to identify somatic mutations driving disease hold significant potential for disease discovery of new somatic IEI and other diseases (66, 67). A current limitation of these approaches is cost, which is likely temporary given the rapid rate at which sequencing technologies have advanced and costs have dropped. In the case of single-cell exome/genome sequencing, a limitation is the ability to know the correct cell to sequence and the number of cells that must be sequenced to identify the relevant somatic mutation, since IEI lack the clear designations of healthy/tumor seen in cancer. Single-cell techniques are also more reliant upon amplification steps for library preparation, introducing more potential for sequencing errors downstream (68, 69).

An alternative approach is to perform targeted high-depth sequencing, either using amplicon-based sequencing, meaning PCR-based amplification of specific genes, or probe-based capture assays to enrich an identified set of genes. Such techniques are relatively rapid and feasible to achieve high sequencing depth in a cost effective manner. This was a strategy used by Mensa-Vilaro and colleagues in a 2019 study focused on the contribution of somatic variation to autoinflammatory and immune dysregulation syndromes (11). They performed NGS-based method of amplicon-based deep sequencing of a set of ~24 genes, known to cause IEI, with sequencing depth of at least 1000x. Testing of patients from 36 families with negative standard testing identified likely disease-causing somatic or gonosomal variants in 23 patients (~64%), with VAFs ranging from 0.8 to 40%. One patient was also identified as having a reversion mutation. Parents from another 92 families in which the affected child had an apparent de novo disease-causing germline variant identified that there was actually gonosomal mosaicism in 7% of unaffected parents, highlighting the potential importance of such investigation in determining risk of recurrence in a family. Thus, in cases where the clinical presentation matches a well-defined category of IEI, a customized panel of known genes can be constructed and sequenced at high depth (11).

3.2.2. Identifying somatic variants and assigning pathogenicity

Detection of somatic variants requires a customized approach that takes into account specific challenges of the disease group. For instance, variant calling in cancer is typically performed by comparing the DNA sequence of tumor tissue to normal tissue from the same patient. In the case of IEI, a matched normal sample that can be used as a reference is often lacking, which makes the identification of somatic variants more challenging. In general, detection of somatic variants includes alignment of sequencing data to the reference genome, followed by application of variant caller tools (primarily designed for cancer), such as Varscan (70), Strelka2 (71), Mutect (72) to identify potential somatic variants. The best tools for this are evolving, and likely to change over time, particularly with more widespread application outside of the cancer field. However, a recent study by Solis-Moruno et al. compared the detection capacity of eight variant callers using high coverage exome sequencing data from patients with pathogenic somatic variants in IEI genes and identified VarDict and VarScan2 as having the best detection rate for somatic variants (70, 73, 74).

One of the biggest challenges associated with somatic variant calling include distinguishing a ‘true’ low frequency variant call from sequencing artifacts which may arise due to DNA contamination, sequencing errors, or read-mapping problems (66). Some groups have addressed this issue by using an internal panel of healthy individuals to identify and remove recurrent sequencing artifacts. In this approach, sequencing data from healthy individuals is compiled into a reference panel to identify and filter out variant calls associated with sequencing artifacts specific to a laboratory’s sequencing or analysis pipeline (76–78). Another approach that can significantly improve somatic variant detection accuracy is use of molecular barcoding in NGS enrichment protocols. UMIs incorporate a unique barcode onto each molecule within a given sample library (79). Quantification of template molecules is done by counting the number of unique barcodes rather than the number of total reads, which allows the analysis software to filter out duplicate reads, and PCR errors thereby improving the quantification accuracy. Including trio analysis, with affected individuals and their parents (or other family members) can be helpful for determining whether variants are inherited or potentially de novo or somatic.

Once somatic variants are identified, assessment of variant pathogenicity is an important step when considering whether a genetic variant is potentially disease-causing. The American College of Medical Genetics (ACMG) (78) has developed guidelines for the determination of pathogenicity of germline variants, with classifications of pathogenic, likely pathogenic, a variant of uncertain significance (VUS), likely benign, or benign, based on evidence including variant frequency in population databases, functional and biological data, and computational data. While such guidelines have not been described for somatic variants in non-cancerous diseases, such strategies can still be applied for somatic variant filtering and classification.

For rare germline disease, a minimum minor allele frequency (MAF, frequency of germline variant in a population) less than 0.01 (<1%) in common population databases, such as gnomAD (79), can be used as a starting point for filtering variants associated with autosomal recessive disease. However, given that somatic mutations are likely to be private to the individual and cause disease when expressed by a minority of cells, limiting analysis to only ‘novel’ or known disease-causing variants not previously reported in unaffected individuals would be an appropriate approach for diagnosing somatic IEIs.

Variants can be prioritized by prediction of their functional consequence using a combination of functional analysis tools such as a CADD score (80), and predictions of pathogenicity from algorithms used for germline disease such as SIFT (81), PolyPhen (82), and Mutation Taster (83). Searching mutation databases and published literature to determine if the gene candidate has been previously implicated in human diseases can provide additional evidence. In addition to population databases, searching for variants in cancer genomic resources that catalog somatic mutations found in malignancies such as COSMIC, Cancer hotspots, The Cancer Genome Atlas (TCGA) and The Candidate Cancer gene database (84, 85, 86) may be helpful, as there is overlap with some IEI and cancer-associated genes, for example STAT3, STAT5b, IL6ST, KRAS, NRAS (87, 88, 89).

Once a potentially pathogenic somatic variant is identified, validation using an orthogonal platform such as droplet digital PCR (dd-PCR) (90) or deep amplicon-based sequencing (91) is required to confirm the presence of the variant. Ideally, this would include sequencing DNA from multiple tissue types of distinct developmental origin, for example saliva or buccal swab, purified immune populations, nails or hair, and/or skin fibroblasts. Such studies are important to confirm the somatic nature of the variant, and also provide a clue on the timing of the mutational event (e.g., during embryogenesis if in multiple tissues versus restricted to the hematopoietic compartment).

3.2.3. Demonstrating pathogenicity of somatic variants

If the somatic variant detected has not previously been shown to lead to altered function of the protein and/or the disease phenotype present in the patient, functional studies are an important step to ascribe pathogenicity and disease association. Information from sequencing may strongly support pathogenicity, for example, the dramatic enrichment of UBA1 variants in myeloid cells or the recurrence of the same somatic variant in patients with a similar clinical phenotype as in the case of the initial reports of UBA1(10) and TLR8 (12). In cases where expression of the variant is limited to a cell type (e.g. DNTs in somatic ALPS or monocytes in VEXAS), it may be possible to test the function of the encoded protein in those primary cells. However, if the variant is distributed across multiple cell types, and enrichment of variant bearing cells is not feasible, alternative strategies need to be employed.

There are a variety of in vitro systems that can be used to test somatic variants, using the same principles as those outlined for correlating a germline variant with disease in a single patient with an IEI, such as demonstrating altered function of the encoded protein in a relevant cell type or in vivo disease model (92). Techniques such as gene-editing and overexpression of variants make it possible to test the function in cases where the role of the protein in the immune response is known. A significant difference between germline and mosaic variants is that working with primary or patient-derived cells is more challenging. One way to potentially overcome this is by combining genotyping with scRNAseq to link genetic and transcriptomic information in single cells to interrogate signaling pathways altered by the variant, as performed in the initial report of somatic JAK1-GOF (93). Alternatively, generation of patient-derived cell lines with or without the variant can provide a tool for examining function. Patient specific iPSC-derived myeloid cell lines bearing WT or NLRC4 or NLRP3 variants were used to confirm that the cytokine hallmark of autoinflammation was present in macrophages expressing the identified variants but not wild-type cells (9, 32, 94). Similarly, we demonstrated that patient-derived clonal iPSC lines expressing either variant TLR8 or the WT gene had a GOF phenotype when differentiated into macrophages or neutrophils from patients with INFLTR8 (12). Finally, similar to germline variants, model organisms can be a powerful tool to evaluate pathogenicity of variants. For example, the biological consequence of somatic loss-of-function variant found in patients with VEXAS syndrome was successfully recapitulated by CRISPR–Cas9-mediated genome editing in zebrafish models.

3.3. Approach to patients with concern for somatic variants causing an IEI

For clinicians caring for patients, somatic variants should be considered in patients with a clinical diagnosis of IEI in the absence of family history, where broad genetic testing has not been revealing (Figure 1). The identification of older individuals with somatic variants causing autoinflammatory disease, for example due to NLRP3 and UBA1, also supports the expansion of genetic testing to this older group of patients, a population in which germline sequencing has traditionally not been as revealing.

Figure 1. Genetic testing algorithm for consideration of somatic mosaicism in IEI.

Patients suspected to have an IEI based on clinical presentation and clinical laboratory testing should have genetic testing for monogenic disease. Clinical testing here refers to tests performed in a clinically-certified laboratory [e.g., CLIA (clinical laboratory improvements amendments) certification], where results are a part of the patient’s medical record and guide clinical management. Research-based testing refers to testing performed in a research-setting, with informed consent, and generally not a part of the patient’s medical record. Targeted gene panels for known IEI-associated genes are widely available for clinical testing, and can be a first-tier approach for molecular diagnosis. If panel testing does not result in a diagnosis, or as *first-line depending on availability, consider trio exome or genome sequencing, including any additional affected family members if possible (siblings). Some IEI genes (dotted line box) have incomplete coverage on NGS platforms [see (108) for additional information]. If genetic testing results in variants of uncertain significance (VUS) in an IEI-associated gene or novel gene not previously associated with an IEI, functional studies should be performed in the clinical or research testing to prove pathogenicity. Identification of additional patients with the same clinical phenotype and variants in that gene can provide additional supportive evidence. After ES or genome testing, if no relevant gene is reported, research based re-analysis or research-based exome/genome sequencing can be performed. If genetic testing fails to identify any relevant gene in a patient without family history of disease (‘sporadic’), or there is incomplete penetrance in the family [for example, some germline variants may have low clinical penetrance, and a “second hit” is required for disease onset as in somatic ALPS (20)], consider evaluation for disease-causing somatic mosaicism. If exome/genome data is available, it can be re-analyzed for somatic variants. If a somatic variant predicted to be associated with the disease phenotype is identified, the presence and variant allele frequency of the genetic variant should be validated on an orthogonal platform e.g. droplet digital PCR or amplicon deep sequencing. If the genetic variant was previously functionally validated and is consistent with known disease, the patient can be diagnosed. For novel somatic variants, functional testing should be pursued to determine pathogenicity. Figure created using Biorender (https://biorender.com/)

IEI: Inborn errors of immunity; CMA: Chromosomal microarray; NGS: Next-generation sequencing; VUS: Variant of uncertain significance

Clinical testing for somatic variants is limited. Several commercial labs offer sequencing for somatic ALPS and RALD, the longest-recognized somatic IEI. While not optimized for somatic variants, some clinical labs will also report out potentially somatic variants from IEI gene panels or exome sequencing, but a ‘negative’ result here does not rule out disease-causing somatic mosaicism. Until clinical testing is available for other somatic IEI, detection of somatic disease is largely research-based. However, with the ever-changing landscape of commercial testing, clinicians may consider seeking out testing marketed to other diseases (such as cancer) but applicable to IEI as well. High-depth sequencing for somatic mosaicism of some genes associated with both IEI and cancers may be included in commercially available somatic panel targeting malignancies, for example the inclusion of some exons of STAT5B and more recently UBA1 on the MyeloSeq^™-HD panel, with a detection limit of ~2% VAF (95, 96). If a patient has had clinical exome or genome sequencing, it may be possible to obtain these data files from the clinical testing and re-review the data with a focus on somatic variants, however this takes significant expertise in genomic analysis and is not readily available to many clinicians. While the current algorithm (Figure 1) suggests starting with germline variant detection, with more widespread availability of somatic sequencing in the future, or in certain clinical cases (e.g., suspected VEXAS syndrome), concurrent germline and somatic testing (or somatic first) may be appropriate.

Until there is more broadly-available clinical sequencing for somatic variants relevant to IEI, it will take the astute clinician to recognize the potential for somatic mosaicism leading to disease partnering with researchers to investigate patients. While time consuming, such studies have significant potential to benefit patients by offering a molecular diagnosis and potentially altering treatment based on that diagnosis.

4. Future perspectives and conclusions

Unbiased broad genetic testing has advanced our understanding of genetic determinants for IEI, however there are many patients who remain without a molecular diagnosis. Identification of disease-causing somatic mosaicism leading to immunologic disease has opened up a new field of genetic investigation of undiagnosed patients. As the field has established with diagnosis of germline IEI, discovery of somatic mosaicism causing immune dysfunction will lead to enhanced diagnosis and precision therapies for patients. Detection of disease-causing somatic mosaicism is currently challenging, and still largely in the realm of research rather than clinical medicine. As research in the field grows, and we expand our thinking about groups of patients (e.g., older individuals) may have a genetic etiology for their immunologic disease, it remains to be seen how common somatic mosaicism leading to IEI phenotypes will actually be. Recent discoveries in the field of diseases that are largely somatic-only also bring to light what terminology we should use for such disorders. Currently somatic variants causing IEI are classified as ‘phenocopies’. But what about disease due to somatic mosaicism in individuals who were born with that mosaicism and can potentially pass disease on to their offspring? Additional discovery of somatic IEI will only further blur the lines between what is inborn versus a phenocopy, and will be worth the field considering disease classifications in the future.