Identifying genetic errors of immunity due to mosaicism

Inborn errors of immunity (IEI) can lead to a range of symptoms due to dysregulation of the immune response, including severe infection, autoimmunity, inflammation, atopy, malignancy, and bone marrow failure. There are ∼550 distinct IEI, and making a molecular diagnosis is critically important for the precision therapy of patients (Poli et al., 2025). IEI have also taught us a great deal about the human immune response, often with surprising findings that demonstrate unique features of our immune response compared with prior findings in model organisms (Notarangelo et al., 2020).

Broad genetic testing using next-generation sequencing (NGS), including exome and genome sequencing platforms, has become the standard of care for the diagnosis of patients with suspected IEI. Molecular diagnoses of IEI inform treatment decisions for patients, including biologic therapies and/or hematopoietic cell transplant (HCT), and allow screening for known complications in affected patients and genetic counseling for family members. In adults and children with IEI, a genetic diagnosis leads to a change in management in 25–75% of patients (Elsink et al., 2021; Quinn et al., 2022; Stray-Pedersen et al., 2017). While a genetic diagnosis of IEI has been a tremendous success, similar to other rare monogenic disorders, up to 70% of patients with a suspected IEI still lack a diagnosis despite exome or genome sequencing (Similuk et al., 2022; Wojcik et al., 2024). Thus, there is a significant gap in our ability to provide a molecular diagnosis for patients.

Somatic mosaicism, post-zygotic mutations in a subset of cells, represents one potential molecular mechanism of immune-based disease not currently detected by most exome and genome analysis strategies and are referred to here as genetic errors of immunity (GEI). Since the initial identification of somatic FAS mutations in autoimmune lymphoproliferative syndrome (ALPS) (Holzelova et al., 2004), ∼19 genes have been described to cause somatic GEI. Here, we will review the role of somatic mosaicism in GEI, including challenges in the identification of mosaic variants, the value of meticulous bioinformatic analysis, the importance of tissue selection for sequencing, and the need for rigorous functional studies to determine pathogenicity.

Primary immunodeficiencies were first described clinically in the 1950s in patients presenting recurrent infections and no serum antibodies (Bruton, 1952). Genetic causes of these diseases started to emerge in the 1980s, with an exponential growth in discovery due to advances in sequencing after 2010 (Poli et al., 2025). With increasing recognition that monogenic immune-based disease also leads to immune dysregulation, including autoimmunity and lymphoproliferation, terminology evolved to “inborn errors of immunity,” reflecting that not all of these diseases lead to infectious susceptibility in patients.

Identification of genetic causes of IEI has expanded our knowledge of the human immune system, and key immunology signaling pathways and developmental mechanisms have been identified by studying patients with IEI. For example, the discovery of the gene encoding Bruton’s tyrosine kinase (BTK) as the cause of X-linked agammaglobulinemia and absent B cells paved the way for the use of BTK-inhibitors to treat chronic lymphocytic leukemia (Herman et al., 2014; Vetrie et al., 1993). Identification of AIRE as the gene responsible for autoimmune polyendocrinopathy candidiasis and ectodermal dystrophy led to our understanding of its role in thymic tolerance and T cell selection (Peterson et al., 2004).

While IEI generally follows Mendelian inheritance patterns, including autosomal dominant (AD), autosomal recessive (AR), and X-linked (XL), there are many nuances and complexities with regard to molecular mechanisms, expressivity, penetrance, and disease phenotype even among the same gene (Poli et al., 2025). For example, variants in CARD11, a scaffolding protein important for lymphocyte activation, result in at least three distinct clinical phenotypes, including a severe combined immunodeficiency phenotype with AR loss-of-function (LOF), immune dysregulation and atopy with heterozygous dominant-negative (DN) variants, and AD gain-of-function (GOF) variants causing lymphoproliferative disease (Brohl et al., 2015; Ma et al., 2017a; Snow et al., 2012; Stepensky et al., 2013). Alternatively, disease-causing variants in different genes can result in near-identical clinical phenotypes. For example, variants in >20 genes regulating IFN-γ production and signaling lead to Mendelian susceptibility to mycobacterial disease (Ogishi et al., 2023). Thus, when assigning disease causality to genetic variants, it is important to test broadly and integrate clinical phenotype and functional data for a particular variant.

Multiple large cohort studies of non-consanguineous patients with suspected IEI have reached an ∼30–40% diagnosis rate using exome or genome sequencing (Chinn and Orange, 2020; Similuk et al., 2022; Vorsteveld et al., 2024). This diagnosis rate is consistent with large cohort studies sequencing patients with other rare diseases (Wojcik et al., 2024). This suggests that there are other mechanisms of disease not adequately ascertained from exome or genome sequencing. For example, non-coding variants leading to altered gene regulation or splicing; incomplete penetrance, including environmental triggers or epigenetic modifications altering a clinical phenotype; polygenic disease; epigenetic mechanisms including monoallelic expression; autoantibodies leading to phenocopies of IEI, such as anti-IFN-γ antibodies causing susceptibility to mycobacterial infection; and, the topic of this review, somatic mosaicism (Bastard et al., 2024; Gruber and Bogunovic, 2020; Hong et al., 2020; Staels et al., 2021; Stewart et al., 2025).

Somatic mosaicism refers to genetic variants found only in a subset of somatic cells, i.e., non-germ cells, and as such they cannot be passed on to offspring. Mosaic variants are the result of postzygotic mutations that can occur at any point in embryogenesis or postnatal development. The tissue distribution of a variant depends on when and where the mutation occurred during development (Fig. 1). The frequency of a mosaic variant in each tissue, especially in early embryogenesis, depends on the ability of the cell expressing the variant to survive, proliferate, and differentiate relative to wild-type cells (Waldvogel et al., 2024). By contrast, gonadal mosaicism refers to mosaic variants found only in germline cells, which can then be passed down to offspring from unaffected parents. Finally, gonosomal mosaicism refers to variants present in both germline and somatic cells. These variants sometimes impact the health of the carrier and are capable of being passed on to offspring (Freed et al., 2014). In practice, it can be very challenging to differentiate somatic from gonosomal mosaicism, unless germ cells, i.e., sperm and eggs, are sequenced, and in most cases, gonosomal mosaicism is confirmed by the presence of a seemingly somatic variant in the germline of an affected individual’s child.

Similar to germline variants, somatic variants take many forms including copy number variants, single nucleotide variants (SNVs), insertions/deletions, and large chromosome alterations (Ogawa et al., 2022). Cells acquire somatic mutations through errors in DNA replication or repair, and it is estimated that DNA polymerases make an error every 10,000–100,000 nucleotides (Preston et al., 2010). While most of these errors are repaired, every tissue has the potential to gain mutations with cellular division. For example, one estimate is that healthy hematopoietic cells acquire 1.14 mutations per cell division (Werner et al., 2020). DNA can also be damaged through endogenous exposures such as reactive oxygen species and various exogenous mutagens (e.g., UV radiation, inhaled carcinogens). If not correctly repaired, these variants can impact cellular function (Chatterjee and Walker, 2017). Additionally, certain genome features such as methylation and transcriptional activity can alter the susceptibility of a particular locus to substitutions (Mustjoki and Young, 2021). Together, all these factors play a role in shaping somatic variants found in various tissues.

The frequency of an individual somatic variant in a tissue in part depends on its impact on the cell. Identification of somatic variants by NGS can identify a variant allele fraction (VAF), the percentage of sequencing reads at a given position on the genome that are called variants compared with the total number of reads at that site. While this does not provide information about precise numbers or types of cells harboring a genetic variant, it can provide an estimate of the percentage of cells that are affected and can be compared across different cell populations.

In some cases, somatic variants in noncoding regions can be used to trace a cell’s lineage or used as a marker of cellular aging (Jaiswal and Ebert, 2019). Coding variants that do not alter the function of a protein or cellular phenotype may persist at a low level without consequences. Some somatic variants are deleterious to the fitness of a cell and may quickly die out and never have a significant impact on the host. In other cases, somatic variants may alter cellular function without impacting cell fitness; thus, they can persist and possibly lead to disease. Finally, somatic variants often become apparent when they provide a proliferative or survival advantage to a cell type and lead to clinical disease. Interestingly, due to their limited tissue distribution, some deleterious variants can be tolerated as somatic when they are embryonically lethal in the germline state (Ogawa et al., 2022).

The immune system originates from multiple embryonic origins and developmental locations, and the timing and tissue location of somatic variants during embryogenesis determines where variants are found in the immune system. For example, a variant acquired in the yolk sac might only impact tissue-resident macrophages, whereas, a variant arising in the fetal liver might have a greater impact on the progenitors that end up in the bone marrow (Stremmel et al., 2018). During postnatal development, hematopoietic stem cells (HSCs) are responsible for generating most immune cell populations, and healthy adults have ∼100,000 active HSCs responsible for maintaining cell types that can last hours to years (Meaker and Wilkinson, 2024). Single-cell lineage tracing using mitochondrial DNA has demonstrated that in the context of normal hematopoiesis, each HSC is slightly different in terms of its output and lineage bias, and aging decreases the clonal diversity of the HSC pool (Weng et al., 2024). Additionally, HSCs accumulate mutations with age, and it is estimated that by the age 70, the average person without a hematological malignancy would have between 350,000 and 1.4 million protein-coding mutations in their HSC pool (Jaiswal and Ebert, 2019). Mutations that increase HSC fitness can result in an expansion of those cells, termed clonal hematopoiesis (CH). Such expansion is due to not only inherent proliferative capacity but also to the attrition of other clones due to aging, external stressors that eliminate other clones, or a reversion of an inherited defect (Gondek, 2021). Particularly relevant to patients with suspected GEI, clonal expansion rates can also be dictated by germline variants and inflammation (Brown et al., 2023; Hormaechea-Agulla et al., 2021). The consequences of CH are still uncertain but can be a precursor to leukemias, and there is growing interest in the role of CH in other diseases such as atherosclerotic cardiovascular disease, osteoporosis, and chronic kidney disease (Belizaire et al., 2023). The genetic and cellular mechanisms that influence the activity and longevity of HSCs, mutation rates in HSCs, and CH are important to consider as we work to understand risk factors for somatic mutations in the hematopoietic system and immune cells leading to GEI.

Over the past two decades, at least 19 somatic GEI have been described, most of which phenocopy germline disease either fully or with a milder phenotype, while others have a distinct clinical phenotype from their germline counterpart. Over the last 5 years, there have been discoveries of disorders that are primarily somatic (Table 1). Somatic GEI are currently classified as “phenocopies” of IEI (Poli et al., 2025). With their growing recognition, particularly in diseases that are largely somatic without a germline counterpart, we use the terminology GEI to encompass the variety of ways in which genetic variation can lead to disease (inborn, somatic, and epigenetic). The discovery of somatic GEI has also changed the way we think about monogenic causes of immune dysregulation in adults because they can arise at any age, and their phenotype can change over time (Staels et al., 2021).

Due to their restricted tissue distribution, somatic GEI are highly variable in their resemblance to germline GEI (Table 1). Individuals with somatic variants in FAS display a clinical phenotype that can be indistinguishable from individuals with germline FAS variants (Dowdell et al., 2010; Holzelova et al., 2004). Patients with somatic NLRP3 GOF can have surprisingly severe and early-onset cryopyrin-associated periodic syndrome (CAPS) with neurologic disease (Lasigliè et al., 2017). Some somatic GEI exhibit a partial phenocopy of their respective germline disease, such as patients with somatic STAT3 or CYBB variants leading to hyper-IgE syndrome and chronic granulomatous disease (CGD), respectively (Hsu et al., 2013; Wolach et al., 2005). In both of those cases, the disease is due to LOF variants, and the milder disease is likely due to the percentage of cells expressing the variant. In some cases, variation in clinical phenotype is partially due to the tissue restriction of the somatic variant. For example, patients with germline GOF variants in RAP1B have complex syndromic features. However, when the same pathogenic variant had a limited tissue distribution, patients have more limited disease with thrombocytopenia and immune deficiency due to high integrin-affinity conformation that impacts platelet activation and lymphocyte migration and survival (Benavides-Nieto et al., 2024).

Alternatively, somatic GEI can have phenotypes distinct from their germline counterparts. This may in large part be due to the cell expressing the variant. For example, a patient with multiorgan Sweet syndrome, including fever, rash, and neutrophilia, had a somatic GOF in PIK3R1 in the skin but not peripheral blood neutrophils (Bhattacharya et al., 2023). By contrast, in the germline state, GOF in PIK3R1 causes immune dysregulation with immunodeficiency and lymphoproliferation (activated PI3K delta syndrome 2) (Lucas et al., 2014).

Some of the most interesting, and challenging to recognize, are GEI associated with genes not previously described to have a germline disease. Patients with somatic variants in UBA1, a ubiquitination enzyme not previously described as causing a GEI, were described as having treatment-refractory autoinflammation termed VEXAS (vacuoles, E1 enzyme, XL, autoinflammatory, somatic) syndrome (Beck et al., 2020). Following this discovery in 2020, VEXAS is now the most common genetic autoinflammatory disease seen in adults, with estimates that 1 in 4,200 men over 50 have a pathogenic somatic variant in UBA1 (Beck et al., 2023). Patients with somatic GOF variants in TLR8 were first described with a somewhat surprising phenotype of neutropenia, lymphoproliferation, and bone marrow failure termed INFLTR8 (inflammation, neutropenia, bone marrow failure, and lymphoproliferation caused by TLR8), with somatic TLR8 variants in five of six patients, and one germline patient in the original cohort (Aluri et al., 2021a; Boisson and Casanova, 2021).

The ability to test for and accurately diagnose somatic GEI has significant implications for the treatment of patients and advances our understanding of the immune system. For example, prior to the discovery of INFLTR8 associated with GOF variants in TLR8, it was unclear whether this toll-like receptor (TLR) was relevant for human disease, particularly due to differences in mouse and human TLR8 and redundancies in TLR8 and TLR7 recognition of ssRNA (Aluri et al., 2021a, 2021b; Cervantes et al., 2012). The discovery of INFLTR8 and diagnosis of patients led to successful HSC transplant (HSCT) in patients for this otherwise fatal disease. Similarly, the recognition of GOF somatic variants in STAT5B and JAK1 leading to GOF in the encoded proteins led to the successful use of JAK inhibitors in these patients (Eisenberg et al., 2021; Gruber et al., 2020; Ma et al., 2017b; Milner et al., 2015). These cases illustrate how recognition of somatic variants as confined to a specific immune cell population can lead to directed therapy including HSCT or biologic therapy.

Somatic variants can lead to GEI through multiple mechanisms (Fig. 2). First, similar to germline disease-causing variants, somatic variants must impact the function of an encoded protein of the immune system. Next, whether this altered protein function is sufficient to cause disease depends upon the cell(s) expressing that variant. For example, TLR8 GOF leads to disease caused by monocytes, the cell type that expresses the TLR8 transcript and protein (Aluri et al., 2021a). However, a similar mutation restricted to B cells would be unlikely to cause disease as B cells do not express TLR8 transcript.

Other genetic factors can influence the pathogenicity of a somatic variant including the presence of other germline (or possibly somatic) variants in the same gene on the same or opposite allele, loss-of-heterozygosity (LOH) (loss of the healthy allele), and “driver” mutations that may allow passenger somatic variants to affect a larger number of immune cells. Somatic variants can act as a “second-hit” to cause disease, leading to either an acquired biallelic defect or loss-of-heterozygosity in a patient harboring a germline variant insufficient to cause clinical disease in the heterozygous state, both of which are seen in FAS- and FADD-associated ALPS (Magerus-Chatinet et al., 2011; Pellé et al., 2024). The acquisition of a driver mutation can increase the number of cells carrying the somatic GEI. For example, a patient with a somatic NLRC4 GOF coincidentally acquired a mutation in Tet2, a known CH gene, leading to the expansion of cells with NLRC4 GOF and worsening clinical phenotype over time (De Langhe et al., 2023).

A number of cellular mechanisms can also influence disease. Somatic variants present in a small proportion of cells can exert a dominant impact on the immune response. For example, TLR8 GOF variants with a VAF of 8–26% lead to high production of proinflammatory cytokines such as IL-1β and IL-18 by monocytes, with broad effects on the immune system. Notably, patients’ VAFs did not correspond with disease severity, and TLR8 variants did not confer a survival advantage (Aluri et al., 2021a). Similarly, patients with low-level mosaicism in NLRP3 (VAF 4–35%) presented with urticarial rash and fever consistent with CAPS (Tanaka et al., 2011). Alternatively, some somatic GEI can confer a selective survival and/or proliferative advantage, leading to the accumulation of pathogenic cells. In VEXAS, myeloid precursor cells with UBA1 pathogenic variants have a selective advantage in the bone marrow over wild-type cells. For example, in the original report, one patient with VEXAS had a 50% VAF in their HSCs that translated to a 70% VAF in their monocytes, demonstrating the selective advantage of the variant on precursor cells (Beck et al., 2020). Similarly in somatic ALPS, deleterious FAS variants confer a survival advantage to double-negative T cells (DNTs) by inhibiting apoptosis. The majority of DNTs carry the somatic FAS variants, whereas the same somatic FAS variants are detected in only a small fraction of HSCs and other immune cells (Dowdell et al., 2010; Holzelova et al., 2004). Finally, in some cases, the somatic variants can lead to intrinsic cellular defects and disease that is typically milder than the germline equivalent, as seen with CYBB and STAT3 LOF (Table 1).

Interestingly, there have been three somatic GEI associated with XL genes (Table 1), all with different mechanisms. VEXAS syndrome is due to LOF variants in UBA1, and no female patients with somatic mutations have been identified, likely because this gene escapes X-inactivation and the expressed healthy allele is compensatory (Tukiainen et al., 2017). There was one report of a female with Turner syndrome and therefore only one X chromosome with VEXAS syndrome (Stubbins et al., 2022). By contrast, INFLTR8 is due to GOF variants in TLR8, and thus while not yet identified, we anticipate that females with mosaic pathogenic variants will be affected as TLR8 also escapes X-inactivation in human immune cells, and both the variant and healthy alleles would be expressed in the same cell (Youness et al., 2023). Consistent with this, we have identified a female patient with infantile-onset germline disease due to TLR8 GOF without evidence of X chromosome skewing (unpublished data). Finally, the patient reported with somatic CGD due to an LOF variant in CYBB was a 74-year-old woman who presented with late-onset XL CGD due to X-chromosome skewing favoring the disease allele (Wolach et al., 2005).

Somatic variants cause immune dysregulation through multiple mechanisms, and when diagnosing somatic GEI, it is important to understand and investigate both the consequence of the pathogenic variant and other factors (e.g., cell-type specificity, co-occurring variants, and expression patterns) that might influence disease.

In addition to leading to GEI, somatic variants in the immune system can cause the reversion or amelioration of germline IEI. Reversion mutations can occur through multiple mechanisms including a copy neutral loss of heterozygosity, a second variant that reverts the error back to wild-type, or a back mutation where the alternate base pair changes back to the wild-type base pair (Miyazawa and Wada, 2021). The immune system may be particularly amenable to reversion mutations as HSCs need to proliferate to replenish immune cells.

Reversion mutations are likely under-recognized as they typically only come to light in a patient with an identified IEI who has a milder course or spontaneously improves over time. For example, patients with XL severe combined immunodeficiency due to germline variants in the IL2 receptor γ chain have seen improved T cell repertoire and clinical conditions due to a reversion mutation in the T-cell precursor (Speckmann et al., 2008). However, studies specifically evaluating reversion mutations suggest that they may be more common in IEI. For example, an investigation of 34 patients with DOCK8 deficiency found that 50% had evidence of reversion mutations (Jing et al., 2014). A large study of 272 patients with Wiskott–Aldrich syndrome identified reversion in 11% (Stewart et al., 2007). Some reversion mutations may manifest due to positive selection, such as the outgrowth of revertant HSCs in patients with the bone marrow failure syndrome Fanconi anemia leading to at least partial correction of the clinical phenotype (Lo Ten Foe et al., 1997). Similarly, CD8 T cells with revertant mutations in SH2D1A in patients with XL lymphoproliferative disease were under selective pressure specifically in response to EBV infection (Palendira et al., 2012). The presence of somatic reversion variants also may not eliminate the need for definitive treatment such as HCT (Chan et al., 2005; Moncada-Vélez et al., 2011). Importantly, IEI can still be passed on to offspring as the pathogenic variant is present in gonadal tissue. Therefore, despite the role of somatic reversion mutations in improving clinical disease, they can also make accurate diagnosis and treatment of germline GEI challenging.

Most somatic GEI have been discovered in patients who presented with a phenotype similar to a known IEI but without a pathogenic germline variant. The initial somatic GEI variants were identified in patients with an ALPS phenotype in 2004 (Holzelova et al., 2004). Somatic variants in FAS were identified by Sanger sequencing of that gene in purified DNTs, a subset of T cells important in the pathogenesis of FAS-ALPS due to their inability to undergo FAS-mediated apoptosis. Somatic FAS variants, either alone or as second hits, are now understood to be responsible for ∼20% of all ALPS cases (López-Nevado et al., 2021). Similarly, a somatic variant with a VAF of 16.7% in whole blood was identified in NLRP3 using Sanger sequencing and subcloning methods in a patient presenting with autoinflammatory disease that appeared consistent with CAPS (Saito et al., 2005). Subsequently, 26 germline mutation-negative cases with a CAPS phenotype were re-evaluated, and somatic mosaicism for pathogenic NLRP3 variants was found in 18 patients (Tanaka et al., 2011). Sanger sequencing allowed these groups to interrogate a gene that they were highly confident could contain a variant in a relevant cell type; however, the discovery capacity of this method is limited as it requires gene-specific amplification and cannot reliably detect VAFs below 5% (Yan et al., 2021).

Using a broader approach, Mensa-Vilaró et al. tested for disease-causing somatic mosaicism using amplicon-based deep sequencing (ADS) of 100 variants in 24 genes in 36 families with a high suspicion for mosaicism. They identified mosaicism in 23 patients (∼64%). ADS achieved ∼5,000× read-depth and identified variants with VAFs as low as 0.8% (Mensa-Vilaró et al., 2019). While this method allowed for increased sequencing depth, it required the group to use multiple different primer sets per gene to mitigate the problem of allele selectivity during initial PCR amplification and was limited in the number of genes and variants tested.

While exome sequencing is of a lower depth (∼200×), it has been successfully used to identify somatic mosaicism. VEXAS syndrome was discovered using exome sequencing data independent of clinical phenotype (Beck et al., 2020). Exome sequencing was also used to identify TLR8 variants that established INFLTR8 as a somatic GEI (Aluri et al., 2021a). Both TLR8 and UBA1 are on the X chromosome and were initially described in biologic males. Having only one X chromosome increased the ratio of variant to wild-type allele reads and more confidence in identifying these somatic variants at a lower overall depth.

With advances in sequencing technologies, there are multiple ways to go about identifying somatic variants. Considerations for identifying and verifying somatic variants as pathogenic causes of GEI are discussed below, and the sequencing methodology should consider the structure and diversity of the cohort, sample availability, and the specific research question.

Tissue sample selection is critical to identifying somatic variants. For GEI, using peripheral blood DNA, peripheral blood mononuclear cells, or a purified population of immune cells may be sufficient to identify most somatic GEI. Lymph nodes, bone marrow, and spleen may also be good sources to test for somatic GEI, for example, in ALPS where DNTs are enriched, expansions of DNTs were found in the lymph node of a patient (Holzelova et al., 2004). However, in practice, samples from these and other sites of resident immune cells such as the thymus are not always readily accessible. Similarly, sampling tissues known to be impacted by inflammation can identify somatic GEI. For example, a somatic PIK3R1 GOF variant was detected in neutrophils in the skin of a patient but was not in peripheral blood neutrophils (Bhattacharya et al., 2023). Somatic mosaicism in the gut epithelia of people with inflammatory bowel disease increases their risk for cancer (Ogawa et al., 2022) but also potentially alters the crosstalk between the epithelia and the immune system.

When evaluating somatic variants, it is helpful to have multiple tissue samples from the same individual. Within a given tissue, purified cell populations or single cells can provide insight into the origin of a somatic variant. A difficult aspect of tissue selection for somatic GEI is that affected immune cells containing the somatic variant may be phenotypically indistinguishable from cells without the variant. This raises complications not only for sample collection but also for variant identification, as many somatic variant callers are designed to compare phenotypically normal tissue to a “tumor” tissue population known to contain somatic variants (Teer et al., 2017).

Ideally, sequencing methods to identify mosaic variants should cover the region of interest with high quality and adequate depth to distinguish true somatic variants from sequencing errors. This can be challenging when somatic variants have a low VAF. For example, a genetic variant present in only 1% of cells would have a VAF of 0.5% if present on an autosome, which is below the limit of detection for many sequencing methods. Here, we discuss several NGS options for identifying somatic variants (Table 2).

Genome sequencing is comprehensive in terms of coverage across the genome, including non-coding regions, and for the identification of structural variants, and it can be done without PCR, which has the potential to introduce errors (Zhou et al., 2022). However, genome sequencing generally provides 20–30× coverage of any given genomic location. While high-depth genome sequencing is possible, it remains prohibitively expensive to be used consistently to identify somatic variants. Exome sequencing has higher depth, typically ∼200× at any given position, and while relatively gene agnostic, it primarily covers protein-coding regions and has lower sensitivity for structural variants (Raca et al., 2023). In theory, exome sequencing would provide ∼10 alternate reads for a variant with a VAF of 0.05 (present in 2.5% of cells if the variant is on an autosome or 5% of cells for XL). In practice, this can vary based on the gene and complexity of the region. Finally, targeted sequencing panels focus on specific genes with depths generally >500× using PCR to amplify regions of interest or oligonucleotide probes to capture regions of interest for sequencing (Pei et al., 2023). All three methods are susceptible to erroneous calls due to sequencing artifacts and are limited by their ability to align reads with certainty when there is complexity (e.g., pseudogenes, highly repetitive regions, large contiguous tandem repeats, and gene duplications) (Corominas et al., 2022; Ebbert et al., 2019; Ryan and Corvin, 2023).

Newer technologies aimed at reducing errors in sequencing may help to aid in the detection of somatic variants leading to GEI. Various error-corrected sequencing methods have been developed to limit the contamination of results with sequencing artifacts and lower the limit of detection for somatic variants. These methods attempt to address common errors in NGS including PCR amplification errors and DNA damage. Adding unique molecular identifiers (UMIs) to reads before amplification allows for the detection of errors due to PCR. Methods such as DuplexSeq and Concatenating Original Duplex for Error Correction (CODEC) find consensus between the top and bottom strands of DNA to reduce sequencing errors (Bae et al., 2023; Schmitt et al., 2012; Menon and Brash, 2023).

Exome, targeted, and genome sequencing are all limited in their ability to identify mutations that co-occur in the same cell. Single-cell genome sequencing has the potential to detect variants that co-exist in the same cell. Aside from the cost, which in some cases can be >10-fold higher, an issue that can arise with single-cell genome sequencing is that one copy of an allele might fail to amplify leading to an uneven distribution of sequencing depth across the genome (Kashima et al., 2020). Alternatively, some groups have developed ways to genotype cells for specific somatic variants during the process of single-cell RNA sequencing (scRNA-seq) or single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) to obtain genotype as well as expression data (Gruber et al., 2020; Izzo et al., 2024).

With the increasing interest in somatic variants on a single-cell level, there is also interest in utilizing other single-cell data to detect somatic variants. The Numbat algorithm detects copy number variation utilizing scRNA-seq data (Gao et al., 2023). The SComatic algorithm enables the use of ATAC-seq and scRNA-seq to identify somatic variants without matched DNA sequencing. Not only does this allow for a multiomic approach to somatic variant identification but it allows for retrospective analysis of single-cell data sets for somatic variants (Muyas et al., 2024).

In summary, multiple sequencing methods and approaches have the potential to detect somatic variants with varying degrees of sensitivity. Each has its own limits of detection influenced by the methodology of the error correction and sequencing method. The process of sample collection and preparation can also introduce DNA damage and potential false positives (Beeler and Bolton, 2023) and sequencing artifacts may exist regardless of the sequencing method used and require careful calling and filtering strategies to identify true variants.

With sequencing data in hand, the next challenge is identifying somatic variants. Multiple variant calling strategies have been created, utilizing different algorithms to assess sequencing reads. Generally, variant callers were built and tested to identify somatic variants in a specific context (Ha et al., 2023). Considerations when choosing a variant calling strategy include the method of sequencing performed, estimated VAFs of somatic variants in the sample, and types of variants being identified (e.g., single-nucleotide versus structural variants) (Xu, 2018) (Table 2).

The method of sequencing is relevant because the callers were designed and tested with data from specific sequencing platforms. For example, some variant callers (e.g., UMI-VarCal [Sater et al., 2020] and Deep SNVMiner [Andrews et al., 2016]) were designed for UMI-based sequencing. Their analysis of the consensus sequences is superior to variant callers that analyze raw-reads at lower VAFs (∼1%) in this context (Xiang et al., 2023). The range of VAFs being investigated can also affect the performance of callers and the ability to accurately call variants with VAFs of <2.5% is particularly challenging as they must be distinguished from sequencing errors. Finally, like germline callers, some somatic callers perform better in identifying SNVs than insertions/deletions and other structural variants (Krøigård et al., 2016).

Given the large interest in somatic mutations and cancer, many variant callers were designed with specific parameters including controls to distinguish tumor from germline DNA, requiring a matched “healthy” sample. Other variant callers can function without a healthy control, but may not perform as well (e.g., Mutect2) (Teer et al., 2017). Machine-learning algorithms are a promising approach to variant calling; however, for somatic mutations, many were trained using specific cancer data sets, which could impact their performance when used outside of that context (Yang et al., 2023). In the context of somatic GEI, particularly when sampling peripheral blood, there may not always be a phenotypic distinction between healthy and pathogenic cells, and disease-causing somatic variants do not always confer a selective advantage to a particular cell type, and furthermore, they may be present at a similar VAF in other tissues including buccal swab and skin biopsy (e.g., TLR8 and NRLP3 GOF). Therefore, it is important to choose a caller that can function without healthy control or compare to a panel of healthy individuals (Table 2).

Given the breadth of options for somatic variant callers, multiple studies have analyzed the impact of combining variant callers on the accuracy of somatic variant detection (Griffith et al., 2015). Consensus between variant callers has shown improved accuracy; however, there is variation in the best way to combine variant callers. For example, for somatic variant detection in cancer and CH, studies have recommended using up to seven variant callers and accepting only variants called by at least six callers (Trevarton et al., 2023) using four variant callers for SNVs and three for INDELs (Wang et al., 2020) or using two callers with different algorithms (Beeler and Bolton, 2023). With the continued evolution of sequencing techniques and the continual development of variant callers, the right strategy for a given research question will continue to evolve, particularly with growing interest in somatic variants leading to diseases other than cancer.

Due to the nuances of different algorithms for identifying somatic variants, validation of the presence of a variant prior to ascribing or investigating a somatic variant as disease-causing is essential. A simple manual review of sequencing data to assess the quality of mapped reads can identify errors in automated variant callers, and a protocol for reviewing somatic variants using Interactive Genome Viewer has been identified focused on cancer (Barnell et al., 2019). This method involves visualizing the sequence read data and ensuring that the somatic variant has high mapping quality, reads are not restricted to one sequencing direction (positive or negative strand), and that they are not in a low complexity region. Manual review can also be used to help set filtering criteria to ensure that true variants are not removed and poor-quality variants are removed.

Consideration should be given to validate the presence of potentially disease-causing variants through resequencing a new sample or library using the same platform (e.g., exome or targeted sequencing panel), an alternative NGS approach (e.g., long-read sequencing), or an orthogonal method such as droplet digital PCR (ddPCR) to ensure that identified somatic variants were not a result of sequencing artifact or introduced through the process of library preparation (Table 2). ddPCR has the advantage of being a benchtop method that utilizes primer/probe-based PCR to rapidly detect specific variants and determine the VAF of a sample. However, disadvantages of ddPCR include the expense of the probesets for each variant, which can be equivalent to the cost of exome sequencing, and difficulty detecting structural variants (Hou et al., 2023).

Identifying the presence of the variant in the transcriptome using RNA-seq can be another powerful validation tool to identify the cells producing the transcript. Ideally, scRNA-seq could be used for this, as was done with somatic JAK1 variants (Gruber et al., 2020). Some challenges with scRNA-seq are that commonly used platforms use short-read sequencing with low coverage outside the targeted 3′ or 5′ ends of the transcript, allelic imbalance leading to uneven coverage across both alleles, and cell-to-cell variability in gene expression causing uneven coverage and depth between cells (Dou et al., 2024). scRNA-seq is also limited in its ability to detect small structural variants as well as variants with decreased RNA expression (e.g., frameshift variants and nonsense variants) (Cooper et al., 2024).

Similar to the approach for germline disease, assigning pathogenicity for somatic GEI requires demonstrating that the variant leads to altered function of the encoded protein in a manner consistent with the immunologic phenotype (Casanova et al., 2014). Prediction of whether a variant will alter the protein function can be made using a combination of ever-growing algorithms that evaluate likely functional consequences, impact on splicing, and evolutionary conservation of a particular locus (Table 3). These algorithms, together with disease databases including ClinVar, CBioPortal, and the Catalogue of Somatic Mutations in Cancer (COSMIC), can help to prioritize variants for further functional validation, particularly if they are known to be pathogenic in the germline state. Similarly, population databases including gnomAD can help to identify variants that are not pathogenic since a variant observed as germline in a healthy population would be highly unlikely to be disease-causing in the somatic state (Table 3).

Understanding known mechanisms of GEI in specific genes is also important when thinking about potentially pathogenic somatic variants. For germline GEI with AD or XL inheritance patterns, pathogenic somatic variants would be predicted to have the potential to cause disease. However, for AR genes, particularly AR LOF variants, without a known AD mechanism of disease, somatic variants would be less likely to be pathological due to the presence of the healthy wild-type copy. Thus, grading the potential pathogenicity of somatic variants requires both an understanding of the molecular consequence of the variant as well as known consequences of perturbations to that gene to immune function (Table 3).

There are no established criteria for grading somatic variant pathogenicity in the context of GEI. While the American College of Medical Genetics criteria for germline variants cannot be directly extrapolated to somatic variants (Richards et al., 2015), they can still be useful for considering whether a somatic variant might be pathogenic. For example, accounting for a variant’s population frequency and in silico predictions of a variant’s impact on the protein. Efforts have been made to create grading criteria for somatic variants in the context of other diseases such as cancer and neurologic diseases; however, their ability to be extended to somatic GEI is limited (Horak et al., 2022; Leon-Quintero et al., 2024; Li et al., 2017). These grading scales assess a variant’s relevance to the pathogenesis of their specific disease. A grading system for somatic GEI would ideally incorporate an understanding of known mechanisms of GEI/IEI for specific genes and functional information about the variant.

After potentially damaging variants are identified, similar to germline variants, it is essential that functional validation in a cell type relevant to the patient’s immunologic disease, and ideally a primary cell from the patient, is performed in a rigorous manner (Casanova et al., 2014). This can be challenging for mosaic variants if the affected cell cannot be isolated from patients. Identifying the cell type harboring the variant can be a first step, and in some cases, may be sufficient to assign pathogenicity if previously well-described to lead to a selective advantage (e.g., UBA1 variant enriched in monocytes or FAS variant enriched in DNTs). This can be accomplished through sequencing of sorted immune cell populations, or potentially scRNA-seq as previously discussed. With the discovery of somatic TLR8 variants, we utilized patient-derived induced pluripotent stem cells to differentiate patient-derived myeloid cells with and without the variant and demonstrated enhanced responsiveness and cytokine production with TLR8 stimulation to demonstrate GOF (Aluri et al., 2021a). Similar to the investigation of germline variants, a variety of gene-editing approaches in vitro and in vivo can also be used to determine the functional consequences of a variant. Together, these strategies can help to identify whether identified somatic variants lead to the observed clinical phenotype, which is ultimately important for designing therapeutic strategies for patients.

Somatic GEI discovery will benefit from current advances being made in genetics, bioinformatics, and oncology research. Improvements in genome mapability will enhance our ability to identify somatic variants. Many current variant callers, regardless of algorithm, have specific filters to remove any variant called in regions that are prone to mapping errors (Xu, 2018). The telomere-to-telomere reference genome is the first gapless assembly of the human genome that will reduce mapping errors and improve subsequent variant identification in coding and non-coding regions of the genome (Beeler and Bolton, 2023; Nurk et al., 2022). Somatic GEI identified thus far have all been due to coding variants that directly alter proteins, and we have little understanding of how non-coding variants that might alter the regulation of genes can impact GEI. An initiative specific to somatic mosaicism is the Somatic Mosaicism Across Human Tissues (SMaHT) network, a large collaborative endeavor that aims to better understand mosaicism across tissues and lifespan in health and disease. Understanding the landscape of mosaicism in other tissues has the potential to uncover new sources of somatic GEI, for example, tissue-specific mosaicism that leads to altered interactions with immune cells.

With the expansion of somatic GEI identification and the expansion of our knowledge of mosaicism in general, it will also be important for us to detect and classify somatic variants based on pathogenicity in clinical testing.

Somatic GEI are monogenic causes of immune dysregulation caused by post-zygotic mutations. They allow for the onset of immune dysregulation at any point in a person’s life. The advent of NGS has allowed us to investigate and discover new somatic GEI. Diagnosing somatic GEI is important because it allows for the accurate diagnosis and treatment of patients. The discovery of somatic GEI remains challenging. It requires careful evaluation of sequencing methods, variant calling and validation, and functional studies to demonstrate pathogenicity.

Work in the Cooper lab was supported by National Institutes of Health /National Institute of Allergy and Infectious Diseases grant R21AI168957. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

This work was supported by the Children’s Discovery Institute Center for Pediatric Immunology at Washington University and St. Louis Children’s Hospital and the Jeffrey Modell Diagnostic and Research Center for Primary Immunodeficiencies at St. Louis Children’s Hospital.

Author contributions: E.G. Schmitz: Conceptualization, Visualization, Writing - original draft, Writing - review & editing, M. Griffith: Writing - review & editing, O.L. Griffith: Conceptualization, Supervision, Writing - review & editing, M.A. Cooper: Conceptualization, Supervision, Writing - original draft, Writing - review & editing.