ASXL1 deficiency causes epigenetic dysfunction, combined immunodeficiency, and EBV-associated lymphoma

Monogenic inborn errors of immunity (IEIs) are a group of >550 rare diseases caused by deleterious germline variants in immune-related genes (Poli et al., 2025). Some IEI disease-causing variants occur in genes encoding epigenetic regulators that modify steady-state gene expression and cellular phenotype, such as DNMT3B and TET2 (Fu et al., 2023; Romano et al., 2022). These alterations in epigenetic regulation disrupt hematopoiesis and cellular differentiation pathways that are important for immune function (Campos-Sanchez et al., 2019; Izzo et al., 2020), giving rise to inflammation, lymphoproliferation, and increased susceptibility to infections and cancer (Haas, 2018; Stremenova Spegarova et al., 2020).

Hematologic malignancies account for the majority of cancers in IEIs and are mainly of lymphoid origin; however, myeloid leukemias and myelodysplastic syndromes (MDS) can also occur (Haas, 2018). Impaired cell-mediated immunosurveillance, chronic viral infections, and inflammation contribute to the increased risk and early onset of cancers observed in IEIs (Hauck et al., 2018; Tiri et al., 2021). Furthermore, the same mechanisms underlying immunodeficiency in certain IEIs may also promote oncogenesis in a cell-intrinsic manner, such as DNA repair or epigenetic defects, causing genomic instability (Hauck et al., 2018). Somatic variants in IEI-associated genes are also observed in hematologic malignancies, further supporting their oncogenic role (Leeksma et al., 2017). Indeed, acquired variants in the epigenetic regulatory genes DNMT3A, TET2, and ASXL1 (known as DTA mutations) are common in age-related clonal hematopoiesis and myeloid neoplasms (Genovese et al., 2014; Jaiswal et al., 2014; Shlush, 2018; Zink et al., 2017). ASXL1 (OMIM #612990) is the least characterized of the three DTA genes and has not been previously linked with immunodeficiency in humans.

ASXL1 comprises part of the Polycomb repressive deubiquitinase (PR-DUB) complex (Kassis et al., 2017), which along with Polycomb repressor complexes 1 and 2 (PRC1/2) represses target genes by modifying chromatin. In embryonic development, the establishment of histone H2A lysine 119 monoubiquitination (H2AK119ub1) by PRC1 is essential for recruiting PRC2, which leads to transcriptional repression. The process is crucial for establishing chromatin bivalency, whereby histone modifications associated with either transcriptional activation or repression occur simultaneously (Li et al., 2023; Macrae et al., 2023). The PR-DUB complex removes H2AK119ub1 at later developmental stages to maintain transcriptional regulation (Bonnet et al., 2022; Kasinath et al., 2021; Macrae et al., 2023).

In keeping with the critical role of PR-DUB in development, de novo heterozygous nonsense or frameshift ASXL1 variants cause Bohring–Opitz syndrome (BOS), a neurodevelopmental disorder characterized by severe developmental delay, microcephaly, failure to thrive, and distinct facial features (Bedoukian et al., 2018). While Asxl1-null mice recapitulate BOS phenotypes, heterozygous germline deletion or hematopoietic-specific Asxl1 deletion leads to MDS-like features (Abdel-Wahab et al., 2013; Wang et al., 2014). Mechanistic studies of BOS and ASXL1-associated hematologic malignancies implicate ASXL1 in differential DNA methylation (DNAm) patterns and altered epigenetic age (Awamleh et al., 2022; Lin et al., 2023; Nischal et al., 2013; Rahmani et al., 2021). Further, conditional knock-in of an Asxl1 variant associated with clonal hematopoiesis revealed a phenotype of impaired T cell development, function, and exhaustion that drove tumor progression in mice (Liu et al., 2022).

Here, we report the first patient with an IEI due to biallelic germline variants in ASXL1 causing ASXL1 deficiency, who presented with combined immunodeficiency complicated by vaccine-strain rubella virus granulomas and EBV-associated Hodgkin lymphoma. We characterize the patient’s clinical phenotype and treatment course, and provide a mechanistic study of the epigenetic and immunological impact of autosomal recessive ASXL1 deficiency. We anticipate that this discovery will empower the diagnosis and management of additional patients with ASXL1 deficiency.

The index patient is now a 17-year-old assigned-female-sex-at-birth born to nonconsanguineous parents. The patient’s mother has a history of rheumatoid arthritis, while the father and older sister have no notable clinical history. The index patient was born full-term and received all routine immunizations, including live-attenuated measles, mumps, and rubella (MMR) vaccine (Fig. 1 A). At 3 years of age, the patient developed chronic ulcerated skin granulomas on the right arm and left leg. The patient was hospitalized twice for pneumonia at ages 3 mo and 6 years, with the latter in the context of primary EBV infection. Laboratory tests over time showed chronic macrocytosis, T cell lymphopenia, progressive loss of B cells, and hypogammaglobulinemia (Fig. 1 B and Fig. S1, A and B).

Despite prolonged systemic courses of antimicrobials, topical corticosteroids, and wound care, the skin granulomas persisted for more than a decade, with new lesions developing in the left inguinal region. Repeated biopsies showed necrotizing granulomatous inflammation with no organisms identified on special stains (Fig. 1 C and Fig. S1 C). Serial bone marrow biopsies every 1–2 years showed mild hypocellularity, megaloblastoid changes, and eventually the presence of non-necrotizing granulomas (Fig. 1 D). Infectious studies revealed intermittent EBV viremia (ranging from 0 to 1,000s copies/ml), absent antibody titers to measles, mumps, varicella, and hepatitis B, but high titers to rubella. After vaccine-strain rubella granulomas were reported in other IEI patients (Bodemer et al., 2014), a biopsy from one of the cutaneous granulomas was tested for vaccine-strain rubella, RA 27/3, and was found to be positive by PCR.

At 14.5 years of age, the patient was diagnosed with EBV-associated stage IVA classical Hodgkin lymphoma, mixed cellularity subtype with prominent granulomatous inflammation (Fig. S1 D and Supplemental Clinical Data). The lymphoma was treated with four cycles of chemotherapy, at which time a repeat positron emission tomography–CT showed progressive disease. Treatment was then transitioned to nivolumab, an anti-programmed cell death protein 1 (PD1) antibody, immunotherapy. This was complicated by the development of autoimmune thyroiditis treated with thyroxine. 6 wk following discontinuation of nivolumab, conditioning began for a 9/10 bone marrow matched unrelated donor hematopoietic stem cell transplant (HSCT) with cyclophosphamide, fludarabine, and 200 cGy total body irradiation. Post-transplant cyclophosphamide, mycophenolate mofetil and tacrolimus were used for graft-versus-host disease (GVHD) prophylaxis. The posttransplant course was complicated by grade 1 GVHD of the upper gut and skin treated with corticosteroids. The patient was discharged on day +34 with full donor chimerism, and remains well now 18 mo after transplant with healing of the chronic skin lesions.

Given the unique and severe clinical presentation, trio whole-exome sequencing (trio-WES) and subsequent whole-genome sequencing were performed. Sequencing identified compound heterozygous missense variants in ASXL1: a maternally inherited variant in exon 5 (Var1_pC111Y: NM_015338.6:c.332G>A;NP_056153.2:p.Cys111Tyr) and a paternally inherited variant in exon 13 (Var2_pS1237F: NM_015338.6:c.3710C>T;NP_056153.2:p.Ser1237Phe), as candidates. The ASXL1 variants segregated with disease in the family (Fig. 2, A and B). These are missense variants predicted to be deleterious, and were considered to be strong candidates for further functional evaluation based on their changes in amino acid polarities, high combined annotation-dependent depletion (CADD) scores (p.C111Y: 26.2; p.S1237F: 23.0), and their low allelic frequency in the Genome Aggregation Database (gnomAD) v4 (p.C111Y: 3.10e−06; p.S1237F: 4.90e−05, with no homozygotes for either variant; Fig. S2 A). Both of these variants fell outside the mutational hotspot region associated with typical truncating BOS variants in ASXL1 exon 13 (Fig. 2 C).

Molecular consequences of the patient’s ASXL1 variants were next defined. Expanded patient primary T cells (both CD4⁺ and CD8⁺) expressed approximately half the amount of total ASXL1 protein when compared to both healthy controls (HCs) and family members (P < 0.05, Fig. 2 D). ASXL1 deficiency was also demonstrated in patient cultured fibroblasts obtained from a skin biopsy (Fig. 2 E). Given the well-described role of ASXL1 in H2AK119 deubiquitination (Scheuermann et al., 2010), we quantified H2AK119ub1 in expanded T cells but did not find notable differences in bulk enrichment levels between the patient and controls (Fig. S2 B). While surprising, recent work has shown that BOS-causing germline ASXL1 variants do not alter specific histone modifications, including H2AK119ub1, but cause DNAm dysregulation (Lin et al., 2023). To define the molecular impact of each of the patient’s ASXL1 alleles, we transfected HEK293 cells with GFP-tagged plasmids containing either myc-tagged WT or the patient’s variant ASXL1 (Fig. S2 C). Both ASXL1 variants show significantly reduced myc mean fluorescence intensity (MFI) compared with WT ASXL1 in the transfected (GFP⁺) cells (P < 0.05, Fig. 2 F). Together, these studies confirm that both of the patient’s variants result in a significant reduction in ASXL1 protein abundance, and in the compound heterozygous state result in human ASXL1 deficiency.

We conducted epigenome-wide association studies (EWAS) to compare the DNAm patterns of primary patient samples with age-matched HCs compiled from publicly available data sets in three tissue types: buccal swabs (n = 725), whole blood (n = 452), and isolated peripheral blood mononuclear cells (PBMCs) (n = 285). We observed widespread differential DNAm between the patient and HCs in whole blood and PBMCs (Fig. 3 A) but not in buccal swabs (Fig. S3 A). There were 50,885 differentially methylated sites (DMs) in whole blood, 18,581 DMs in PBMCs, but only 38 DMs in buccal swabs that passed the threshold of adjusted P (P_adj) ≤ 0.05, 15.7% of which (9,438 DMs) overlapped between whole blood and PBMCs (Fig. 3 B). Emphasizing that this patient does not have BOS, we found that BOS-associated DMs were not enriched in our analysis (Fig. S3 B). Pathway enrichment analysis with 18-state ChromHMM chromatin state annotation showed that DMs in both whole blood and PBMCs were significantly enriched in regulatory elements, such as enhancers, bivalent transcription start sites, and repressed Polycomb regions, as well as some quiescent gene and heterochromatin regions in PBMCs (Fig. 3 C).

To gain further information about the patient’s DNA methylome, we assessed the patient’s tissue-specific epigenetic age. We focused on the first-generation epigenetic clocks as they specifically estimate chronological age based on DNAm patterns (Kabacik et al., 2022). In analysis of buccal swabs, the patient was estimated to have a pediatric buccal epigenetic (PedBE) clock age of 10.8 years, matching the actual age of 11 years at sample collection (Fig. S3 C). In contrast, the Horvath pan-tissue age of blood samples was higher compared with the patient’s chronological age (Fig. 3 D). Similarly, the predicted telomere length based on DNAm data was lower in blood (Fig. 3 D), and the presence of short telomeres in lymphocytes was confirmed by direct measurement (Fig. S3 D). Overall, we identified advanced epigenetic aging (as estimated by the Horvath pan-tissue clock) and short telomeres in ASXL1-deficient patient blood samples compared with HCs, which was especially prominent in lymphocyte subsets.

To examine causation, we transduced expanded patient-derived T cells with WT ASXL1 to delineate the relationship between the patient’s variants and the alterations in epigenetic regulation in lymphocytes. Whereas empty vector (EV)–transduced patient cells (patient condition) maintained a lower ASXL1 protein level compared with the EV-transduced HC cells (HC condition), WT ASXL1-transduced patient cells (rescued condition) showed increased ASXL1 protein comparable to the HC condition (Fig. 4 A).

Principal component analysis (PCA) of the T cells’ DNAm showed that the patient clustered away from the HC (including family member) cells along PC2, with the rescued cells in the middle (Fig. 4 B). Analysis of the patient versus HC comparison yielded 141,113 DMs (P_adj < 0.2 and delta-beta >0.1), of which 129,104 (91.5%) showed decreased DNAm. The pattern observed in the PCA plot was replicated in the DMs—patient samples were notable for their decreased DNAm, while the rescued condition showed an intermediate DNAm level (Fig. 4 C). Furthermore, loss of DNAm in the patient condition occurred mainly in loci that were highly methylated in HC samples; of the DMs with decreased DNAm, 80.1% were >50% methylated in the HCs (skewness test, P < 2.2e−16, Fig. 4 C). In the rescued versus HC condition, 35,457 DMs were identified, and 90,345 of the patient versus HC DMs were not significantly different in this comparison, indicating that the majority of the DMs in the patient’s cells were rescued by WT ASXL1 transduction (Fig. 4 D). Similar to our findings in blood, the genomic locations of DMs were highly enriched for repressed Polycomb and bivalent chromatin states, with the rescued cells showing higher P values than that of patient cells (Fig. 4, D and E). When performing differential gene expression on RNA-seq data of the same expanded T cell samples, we observed enriched downregulation of DNAm and other epigenetic-related pathways in the patient compared with the rescued samples (Fig. S4, A and B).

We also tested whether WT ASXL1 could normalize the epigenetic age acceleration we had observed in the patient. While the Horvath pan-tissue clock slightly underpredicted chronological age in controls, the clock assigned the patient an epigenetic age on average 27.4 years older than the chronological age. Notably, rescue with WT ASXL1 significantly lowered the epigenetic age (Fig. 4 F). Specifically, the difference between epigenetic and chronological age was significantly higher in the patient than family and unrelated HCs, and this difference was significantly reduced by WT ASXL1 (P < 0.0005) (Fig. 4 F). Two other first-generation epigenetic clocks, the Hannum and Horvath skin-and-blood clocks, corroborated this trend (Fig. S4, C and D). Based on this series of experiments, WT ASXL1 transduction rescued the DNAm differences and attenuated the advanced epigenetic aging driven by the patient’s ASXL1 variants.

Given the methylation impairments in patient T cells, we conducted detailed immunophenotyping of peripheral blood collected before the patient’s lymphoma diagnosis. Naïve CD4⁺ and CD8⁺ T cell subsets and recent thymic emigrants were markedly reduced, suggesting poor thymic output (Fig. 5, A and B; and Table 1). The patient also showed expansion of PD1⁺ CD4⁺ and CD8⁺ memory cells, indicating T cell exhaustion (Fig. 5 C and Fig. S5 A). To further investigate whether circulating Tfh type cells contributed to the increased patient PD1 expression, we performed flow cytometry on patient PBMC collected over 5 years (three patient samples: pre-lymphoma, lymphoma pre-chemotherapy, and lymphoma during chemotherapy; one post-transplant patient sample collected 1 year after HSCT). During the time these samples were collected, the patient also had near-absent B cells with decreased memory proportions (Fig. 1 B and Fig. S5, B and C). The CD4⁺ memory T cell compartment contained severely reduced circulating Tfh (CXCR5⁺) cells (Fig. 5, D and E). Accompanied by findings of reduced IL-4 expression (Fig. 5, F and G), the data suggest that the patient had significantly restricted Tfh functions, and that PD1 was independent of CXCR5 expression suggestive of T cell exhaustion (Fig. S5 D). Notably, both the circulating Tfh and the IL-4–producing cell proportions restored to levels similar to the HCs 1 year after HSCT (Fig. 5, D–G).

We also quantified the expression of the senescence marker CD57 in patient cells. We found that patient CD8⁺ T cells exhibited elevated CD57 expression in the effector memory compartment and had a restricted T cell receptor repertoire (Table 1, Fig. 6 A, and Fig. S5 E). This was not prominent in CD4⁺ T cells (Fig. 6 A and Fig. S5 E). When further profiling the expanded T cells from patient PBMCs, we observed that in contrast to HC, the patient’s CD8⁺ T cell population failed to expand in the presence of soluble activators (Fig. 6, B and C). This skewing led to a significantly higher CD4⁺/CD8⁺ ratio that was rescued with the transduction of WT ASXL1 (Fig. 6, B and C). This high CD4⁺/CD8⁺ ratio, however, was not observed in primary patient blood (Table 1).

We then assessed pSTAT5 signaling and ROS production, as alterations in these pathways were reported in T cells containing an ASXL1 variant associated with clonal hematopoiesis of indeterminate potential (Liu et al., 2022). We observed significantly reduced pSTAT5 abundance after low-dose IL-2 (1 U/ml) treatment in expanded patient CD8⁺ T cells, which was rescued by stable WT ASXL1 expression, indicating a causative role of ASXL1 in regulating STAT5-mediated CD8⁺ T cell homeostasis (Fig. 6 D). In contrast, pSTAT5 levels in CD4⁺ T cells were not significantly different between the patient and the HCs (Fig. 6 D). The finding in CD8⁺ T cells was mirrored by the transcriptomic analysis showing downregulated IL-2-STAT5 signaling in patient samples compared with the HC, and upregulation of the pathway in the rescued condition (Fig. S4 A, top). Finally, significantly higher ROS production was observed in expanded patient CD4⁺ T cells, which was rescued upon transduction of WT ASXL1 (Fig. S5 F). In contrast, while ROS expression was lower in patient CD8⁺ T cell samples, WT ASXL1 failed to rescue the phenotype. These findings suggest that ASXL1 deficiency mediates impairment of CD4⁺ and CD8⁺ T cells by alternative modes.

Finally, we profiled the patient’s CD3⁺CD56⁺ and NK cells to assess their potential contribution to the clinical phenotype of persistent viral infection. We noted a significant contraction in the proportion of CD3⁺CD56⁺ cells in patient PBMCs compared with HCs, which was partially ameliorated one year after HSCT (Fig. 7, A–C). Interestingly, we observed a reduction of DNAm in the gene body of NCAM1—encoding for CD56—in expanded patient T cells compared with HCs (Fig. 7 D). The mRNA expression of NCAM1 was also decreased in patient samples accordingly (Fig. 7 E). Upon transduction of WT ASXL1, we observed a partial rescue of the NCAM1 DNAm level and NCAM1 transcript levels (Fig. 7, D and E), likely reflecting a parallel to the CD56⁺ expression in patient T cells after HSCT (Fig. 7 C). The patient PBMC also had a lower proportion of mature and cytotoxic NK cells (CD16⁺CD57⁺ NK cells, Fig. 7, F–H), likely further explaining the patient’s lack of cytotoxic response to viral infections.

We report the clinical, epigenetic, and immune profiling of the first case of autosomal recessive ASXL1 deficiency in a person with combined immunodeficiency and EBV-associated Hodgkin lymphoma. Single-patient studies account for approximately one third of newly identified monogenic immune disorders (Poli et al., 2025), requiring methodical validation to establish causality. Following the guidelines proposed by Casanova et al. for studies in single patients (Casanova et al., 2014), we showed that the ASXL1 variants identified in the patient are rare and monogenic, and have complete penetrance. Studies in primary patient cells demonstrated decreased ASXL1 protein abundance, allele-specific functional impairment, dysregulated DNAm, and diminished T cell help and cytotoxic capacity. Finally, we showed that epigenetic and immune defects in the patient’s cells were rescued by transduction of WT ASXL1 in vitro, in parallel with partial recovery of multiple cellular compartments 1 year after HSCT. The patient’s rare and severe clinical presentation provided further support for an underlying genetic cause (Casanova et al., 2014). T cell lymphopenia in keeping with a combined immunodeficiency was present from early childhood, followed by progressive loss of B cells. After receiving live-attenuated MMR vaccine, the patient developed vaccine-strain rubella–positive granulomas. The patient also had persistent macrocytosis and mild bone marrow hypocellularity, and developed chemorefractory advanced Hodgkin lymphoma at age 14 years. Based on our findings, patients presenting with this unique clinical profile of severe viral-associated complications and immunodeficiency should be evaluated for autosomal recessive ASXL1 deficiency.

Heterozygous germline truncating variants of ASXL1 are known to cause the neurodevelopmental disorder BOS (Hoischen et al., 2011). The hematologic and immune disorder described here is distinct from BOS in its molecular and clinical effects. We identified biallelic missense ASXL1 variants resulting in decreased ASXL1 protein abundance and abnormal DNAm. Both variants are located outside of the mutational hotspot associated with BOS, and the DNAm signature of our patient does not overlap with that seen in BOS patients (Awamleh et al., 2022). Previous studies have demonstrated no impairment of ASXL1 protein abundance in BOS and suggest that the C terminus truncating variants may be gain-of-function (Balasubramani et al., 2015; Lin et al., 2023). Mouse models indicate that Asxl1^−/− mice have developmental abnormalities and failure to thrive, recapitulating elements of the BOS patient phenotype (Abdel-Wahab et al., 2013; Wang et al., 2014), while Asxl1^−/+ mice develop an MDS-like phenotype possibly attributable to Asxl1 haploinsufficiency (Wang et al., 2014). Future studies directly comparing samples from patients with autosomal recessive ASXL1 deficiency to those with BOS are needed to further elucidate the mechanistic differences between these two diseases.

We attribute the immune disorder observed in our patient to ASXL1 deficiency, as supported by parallels between the patient’s phenotype and Asxl1^−/+ and Asxl1-mutant mice. This includes impaired self-renewing capacity of hematopoietic stem cells (Abdel-Wahab et al., 2013; Fujino et al., 2021; Wang et al., 2014; Zhang et al., 2018); severely diminished naïve T cells and T cell exhaustion (Liu et al., 2022); bone marrow hypocellularity (Fujino et al., 2021); and presence of immature red blood cells and macrocytosis (Abdel-Wahab et al., 2013; Wang et al., 2014). While differences in ASXL1’s function in mice versus humans limit direct comparisons between the two species, the shared immune and hematologic features observed in the ASXL1-deficient patient and the Asxl1-mutant mice support that altered ASXL1 underlies the patient’s phenotype.

EWAS showed significant alterations of DNAm in patient T cells, PBMCs, and whole blood, but not buccal swabs; DNAm decreased predominately in T cells and whole blood, but was increased in PBMCs. The contrasting significance and directionality of EWAS results across tissue types suggest cell type–specific impacts of abnormal ASXL1 on global DNAm. This was further supported by the pronounced increase in Horvath pan-tissue epigenetic age and reduced telomere length observed in patient lymphocytes. Global loss with regional gain of DNAm is key features of aging, as epigenetic marks are ineffectively maintained through the S phase of mitosis (Sen et al., 2016). We showed the patient’s ASXL1 variants drive DNAm loss and increased epigenetic age in T cells. This is supported by recent findings showing high epigenetic age differences in BOS patients with ASXL1 truncating variants (Awamleh et al., 2022), highlighting the role of ASXL1 in cellular aging. Cellular aging may also be associated with the abnormal ROS production seen in ASXL1-deficient patient CD4⁺ T cells. Oxidative stress and epigenetic regulation are both linked to cellular aging, and the interplay between the two mechanisms is well documented (Davalli et al., 2016; Gao et al., 2017; Guillaumet-Adkins et al., 2017; Kandlur et al., 2020; O’Hagan et al., 2011; Schroeder et al., 2013). Asxl1-mutant mice also display mitochondrial dysfunction and cellular senescence (Fujino et al., 2021; Liu et al., 2022); this aligns with our findings linking altered ASXL1 with epigenetic dysregulation and ROS production.

The granulomas that emerged at age 3 years represented a particularly enigmatic feature of this patient’s clinical course. These lesions persisted for more than a decade, were refractory to treatment, and were labeled as idiopathic until recognized as being positive for vaccine-strain rubella virus. Rubella virus was first identified in the granulomas of three IEI patients in 2014 (Bodemer et al., 2014), and is now linked to ∼60% of idiopathic cutaneous granulomas across a variety of IEIs (Buchbinder et al., 2019; Gross et al., 2022; Perelygina et al., 2021; Perelygina et al., 2020; Zhang et al., 2024). Impaired T and NK cell cytotoxicity specifically has been implicated in susceptibility to rubella vaccine–induced granulomas (Gross et al., 2022). Rubella strains identified in affected patients often feature mutations in CD8⁺ T cell and neutralizing B cell epitopes used to evade cellular and humoral responses (Perelygina et al., 2019), suggesting that defects in both modes of adaptive immunity can contribute to persistent rubella infection.

In this patient’s case, the rubella vaccine granulomas were found to be secondary to deficiency of the epigenetic regulator ASXL1. DNAm patterns are distinct across cell populations and are crucial for hematopoiesis and the maintenance of cellular identity (Baumann et al., 2019; Izzo et al., 2020; Smith et al., 2025). The patient demonstrated profoundly abnormal DNAm patterns and deficiencies in key immune cells for viral host defense. CD3⁺CD56⁺ cells were decreased in the ASXL1-deficient patient, with altered NCAM1 (CD56) DNAm and expression levels. Immunophenotyping demonstrated decreased CD8⁺ T cell proliferative capacity and impaired IL-2-STAT5 signaling that was rescued by introduction of WT-ASXL1, and reduced CD16⁺ CD56dim NK cells, which express granzyme and perforin, and are responsible for antibody-dependent cellular cytotoxicity (Vivier et al., 2024). Diminished cytotoxicity coincided with impaired humoral immunity, as reflected in low circulating Tfh prevalence and hypogammaglobulinemia, ultimately creating an environment susceptible to infection and ineffective in forming antiviral memory.

Our detailed molecular and clinical evaluation of human ASXL1 deficiency provides further insight into the role of immune dysregulation in cancer predisposition. Global hypomethylation has been proposed as a replicative barrier against highly proliferative and premalignant cells by inducing cellular senescence (Cruickshanks et al., 2013; Johnstone et al., 2022). In this case of ASXL1 deficiency, however, T cell senescence may have contributed to the patient’s impaired antiviral and antitumor immunity, predisposing to EBV-associated Hodgkin lymphoma (Zhao et al., 2020). Indeed, recent modeling of an ASXL1 variant associated with clonal hematopoiesis showed that it promoted the progression of solid tumors via impaired intra-thymic T cell development, naïve-memory T cell imbalance, and T cell exhaustion (Liu et al., 2022). Consistent with these findings, we also observed the near-absence of recent thymic emigrants and naïve T cells, decreased CD8⁺ proliferative capacity, and increased senescence in the patient. As NK cells play pivotal roles in antitumor immunity, and CD8⁺ T cells have been shown specifically to inhibit EBV-associated lymphoproliferation, the patient’s immune landscape likely resulted in limited immunosurveillance capacity (Cohen, 2000; Kim et al., 2007; Liu et al., 2021; Pietersma et al., 2008; Rickinson and Moss, 1997; Swann and Smyth, 2007). Some of these mechanisms may also play roles in other hematopoietic malignancies, specifically acute myeloid leukemia, chronic myelomonocytic leukemia, and MDS (Boultwood et al., 2010; Micol and Abdel-Wahab, 2016; Russell and Graham, 2013), where somatic ASXL1 variants are commonly described (Micol and Abdel-Wahab, 2016). Taken together, ASXL1 deficiency may drive the maintenance of an oncogenic milieu.

This study has many strengths, notably the detailed immunological description of the first human case of biallelic ASXL1 deficiency and well-controlled epigenetic profiling with family controls and a population reference data set across multiple tissues. However, there were also limitations. First, this was a single-patient study, and ultimately, the full phenotype of human ASXL1 deficiency will only be defined through diagnosis of more patients. Second, while we observed global differential DNAm in patient whole blood and PBMCs, we explored the effect of WT ASXL1 transduction only in patient T cells. Sample limitations restricted us from working on other cell types (e.g., neutrophils) in a controlled setting. We demonstrated that the patient’s missense variants disrupt ASXL1 at a posttranslational level; however, the precise mechanisms by which this occurs remain unknown. Future work will aim to further dissect how p.C111Y and p.S1237F destabilize ASXL1 protein. Similar to previous findings, we showed the effects of ASXL1 variants to be highly cell type–specific (Nagase et al., 2018), and therefore, examining different cellular models could elucidate the roles of ASXL1 in hematopoiesis and disease. Finally, although our patient has benefitted clinically from allogeneic HSCT, follow-up was only out to 18 mo at the time of writing; we are therefore unable to comment on long-term outcome after HSCT.

In conclusion, our study identified a novel human IEI caused by germline biallelic deleterious ASXL1 variants resulting in ASXL1 deficiency. This discovery will empower the diagnosis of future patients with ASXL1 deficiency, and expands upon the clinical spectrum of ASXL1-associated disease. We recommend genetic testing at the first sign of combined immune deficiency as a powerful precision medicine approach to guide diagnosis and inform therapeutic decision-making.

Peripheral blood samples were collected from the index patient, as well as from the patient’s father, mother, and sister, along with three unrelated healthy individuals at the BC Children’s Hospital Research Institute. Similarly, skin biopsies were collected from the patient and six healthy individuals. All participants from whom we collected biological samples and/or their parents/guardians provided written informed consent for sample collection, sequencing, data analysis, and the publication of findings and photographs. Data from additional samples were obtained from publicly available databases, as outlined below. The research protocols were approved by the University of British Columbia Clinical Research Ethics Board (H15-00641).

Genomic DNA was isolated from peripheral blood using standard procedures, and exome/genome sequencing was performed on the Illumina HiSeq platform (Macrogen). Sequencing data were analyzed using an updated version of our in-house, open-source, semi-automated bioinformatics pipeline as described previously (Tarailo-Graovac et al., 2016; van Kuilenburg et al., 2019). trio-WES was performed on genomic DNA from the patient and both parents.

Based on family history and pedigree, once a candidate variant list was generated by the pipeline, variants identified through the de novo, autosomal recessive, and compound heterozygous inheritance models were first assessed. All high-impact, putative loss-of-function variants were reviewed. Next, the candidate variant list was screened for variants in genes associated with IEIs, especially those previously linked with vaccine-strain rubella–associated granulomas. Genes associated with DNA repair deficiency disorders and telomeropathies were also investigated.

In addition, a phenotype-driven approach was also employed to identify relevant variants in genes that may be associated with clinical features (using associated Medical Subject Headings [MeSH] and Human Phenotype Ontology [HPO] terms) presented by the proband, such as multiple skin lesions, necrotizing granulomatous inflammation (postvaccine rubella strain identification), idiopathic chronic macrocytosis, abnormal T cell/B cell count, mild megaloblastoid changes, mild dysplastic features, infections, pneumonia, and failure to thrive.

Compound heterozygous ASXL1 variants were considered the top candidates of interest. The ASXL1 variants, p.C111Y and p.S1237F, were further considered as potentially pathogenic based on in silico tools. The pathogenicity predictors included CADD (p.C111Y: 26.2; p.S1237F: 23.0), PolyPhen-2 (p.C111Y: 1.0; p.S1237F: 0.883), and mutation taster (p.C111Y: 1.0; p.S1237F: 0.778) (Adzhubei et al., 2010; Rentzsch et al., 2019; Schwarz et al., 2014). Furthermore, the p.S1237F variant is present in the COSMIC database (catalog of somatic mutations in cancer) for its association with basal cell carcinoma (Bonilla et al., 2016; Sondka et al., 2024).

Whole-genome sequencing was also performed to scan for potential somatic mosaic variants and to exclude variants in noncoding regions. Compound heterozygous ASXL1 variants still emerged as the variants of highest significance and were further verified by Sanger sequencing.

Briefly, standard sequencing of genomic DNA from all individuals of the family was performed to confirm genotype segregation with the disease in the family using the following primers: Var1 p.C111Y: Forward 5′-GAA​TGA​CTG​CCT​TGC​TTT​AAG​A-3′, Reverse 5′-AAA​GGT​TAG​TCC​CTA​GTG​CC-3′; Var2 p.S1237F: Forward 5′-TGC​CAG​ATA​GCT​GTG​AAA​CA-3′, Reverse 5′-CAC​ATT​TGG​AGA​TGA​GAA​ACG​A-3′.

For generation of T cell blasts, T cells were isolated from PBMCs (EasySep Human T Cell Isolation Kit, Cat # 17951; STEMCELL) and activated for 12 days. Briefly, T cells were treated with exogenous IL-2 (10 U/ml) and ImmunoCult Human CD3/CD28 T Cell Activator (Cat # 10971; STEMCELL) for 12 days and maintained at a concentration of 10⁶ cells/ml every 2 days. After 12 days, exogenous IL-2 concentration was increased to 1,000 U/ml.

To generate fibroblasts, a skin biopsy was obtained, manually cut into smaller pieces, and placed in separate wells of a 6-well plate in complete Dulbecco’s modified Eagle’s medium (DMEM) composed of 10% heat-inactivated FBS (Gibco-BRL Life Technologies; Life Technologies), 2 mM l-glutamine (HyClone; Thermo Fisher Scientific), 1 mM sodium pyruvate (Gibco-BRL Life Technologies), and 1× antibiotic–antimycotic (Gibco-BRL Life Technologies, Thermo Fisher Scientific) for 3 wk at 37°C or until a confluent monolayer of fibroblasts had formed. Primary fibroblasts were then lifted and frozen at −80°C for future assays.

Full-length WT ASXL1 cDNA (NM_015338; Cat# RG219354; OriGene Technologies) was cloned into a pCMV6-AC-IRES-GFP vector (Cat# PS100027; OriGene Technologies) using the restriction enzymes AsiSI (Cat# R0630S; New England Biolabs) and MluI-HF (Cat# R3198S; New England Biolabs). The p.C111Y and p.S1237F variants reported in this study were generated in the cloned vector through site-directed mutagenesis. The following primers were used for site-directed mutagenesis (lowercase = mutated site): Var1 p.C111Y: forward 5′-TGT​GGA​GAG​Cta​tGG​GTC​TAA​TG-3′, reverse 5′-CAT​TAG​ACC​Cat​aGC​TCT​CCA​CA-3′; Var2 p.S1237F: forward 5′-GCA​GTT​CTC​Ttt​tTT​TAG​TTG​TGA​AG-3′, reverse 5′-CTT​CAC​AAC​TAA​Aaa​aAG​AGA​ACT​GC-3′.

The expression of variant plasmids was induced transiently in HEK293 cells using Lipofectamine 3000 (Cat# L3000150; Thermo Fisher Scientific). The transient expression of ASXL1 variant plasmids in HEK293 cells was accomplished using a Lipofectamine 3000 kit (Thermo Fisher Scientific) according to the manufacturer’s recommendations. Briefly, HEK293 cells were seeded at 2.0 × 10⁵ cells/well in a 24-well plate in 0.5 ml of DMEM with 10% FBS (Gibco, Life Technologies) and incubated for 24 h at 37°C. Cells were transfected with 250 ng of plasmid DNA using the P3000 and Lipofectamine 3000 reagents.

After 24 h, cells were harvested and stained with a live/dead stain (Fixable Viability Stain 780, Cat# 565388; BD Biosciences) for 5 min at 37°C. The expression of the live/dead stain and GFP was then acquired on the BD LSRFortessa (BD Biosciences) and analyzed using FlowJo software (BD Biosciences). Cells were then fixed with Cytofix (Cat# 554655; BD Biosciences) for 10 min at room temperature and permeabilized using Perm III for 20 min on ice (Cat# 558050; BD Biosciences). The cells were then stained with either primary antibody for myc tag (1:1,000 dilution, Cat# 2276; Cell Signaling Technology) or mouse IgG1 isotype control (1:500 dilution) for 1 h at room temperature in 1× Perm/Wash (Cat# 554723; BD Biosciences), followed by staining in secondary goat anti-mouse secondary antibody (1:200 dilution, Alexa Fluor 405, Cat# A-31553; Thermo Fisher Scientific) for 30 min at room temperature, also in 1× Perm/Wash.

Stable WT ASXL1 expression was achieved by infection of expanded patient-derived T cells with lentiviral particles as described previously (Kutner et al., 2009; Sharma et al., 2022), and expanded cells were used for downstream DNAm and intracellular flow cytometry analyses.

To generate lentivirus vectors, the WT ASXL1 (NM_015338) plasmid (Cat# RG219354; OriGene Technologies) was cloned into a GFP-tagged Lenti vector (Cat# PS100071; OriGene Technologies) using AscI (Cat# R0558L) and MluI-HF (Cat# R3198L) (both from New England Biolabs). The plasmid was packaged using third-generation packaging plasmids and transfected into HEK293T cells. Culture medium was collected, centrifuged, filtered, concentrated, and stored at −80°C until use. Expanded T cells were infected with lentiviral particles and expanded in complete RPMI 1640 (GE Healthcare) supplemented with 10% FBS for 9 days.

Expanded patient T cells, patient T cells rescued with WT ASXL1, and HC T cells stably transduced with a GFP EV had their transcriptome measured. Briefly, RNA was extracted as previously described (Sharma et al., 2022) using RNeasy Mini Plus Kit (Qiagen) according to the manufacturer’s recommendations. A ribosomal depletion step was then carried out on the RNA before library construction and sequencing on the Illumina platform. 150 base pair paired-end sequencing was conducted for a targeted total of 50 million reads per sample.

Demultiplexed read sequences were aligned to a reference sequence using RNA-Seq Alignment app (v1.1.1) on Illumina BaseSpace, using Spliced Transcripts Alignment to a Reference aligner and Cufflinks 2 for assembly and estimation of gene expression. Expression data were normalized to reads between samples using the edgeR package in R (Robinson et al., 2010). Normalized counts were filtered to remove low counts using the filterByExpr function in edgeR (Chen et al., 2016).

Differential expression was accomplished using limma (Ritchie et al., 2015). We performed two sets of comparisons: HC (n = 3) versus patient (n = 3) samples, and patient versus rescued (n = 3) samples. Gene set enrichment analysis was conducted on Gene Set Enrichment Analysis software, using 1,000 permutations with weighted (P = 1) enrichment score for preranked analysis based on t statistics from differential expression analysis (Subramanian et al., 2005). Two MSigDB gene sets, Hallmark and Reactome, were used in the analysis, where default gene set exclusion criteria were set, with gene sets above the size of 500 or below 15 being filtered out in the analysis (Liberzon et al., 2015; Liberzon et al., 2011). The sample-level enrichment analysis scores were then determined by subsetting for the core enrichment genes in the significantly enriched pathways, followed by a z-score calculation of these genes’ expression in 1,000 bootstraps.

Blood and buccal swabs were collected from the patient, and genomic DNA was extracted from whole blood, PBMCs, buccal cells, and cultured T cells. Whole blood was collected in Vacutainer CPT Cell Preparation Tubes (Becton Dickinson). PBMCs were then isolated by standard Ficoll-Paque (GE Healthcare) density centrifugation. Buccal swabs were collected with Isohelix Buccal Swabs (Cell Projects Ltd.) and Oragene DNA sample collection kits (DNA Genotek), respectively. Genomic DNA was extracted from whole blood, PBMCs, and expanded T cells with DNeasy kits (Qiagen). Buccal DNA was extracted with Isohelix Buccal DNA Isolation kits (Cell Projects Ltd.). Extracted DNA (750 ng) was subjected to bisulfite conversion with an EZ-96 DNAm kit (Zymo Research), and aliquots of 160 ng were assayed with DNAme arrays. DNA extracted from primary tissues and T cells was assayed using Infinium HumanMethylationEPIC BeadChip (EPIC) and HumanMethylationEPIC v2 BeadChip arrays (Illumina), and scanned with the Illumina iScan 2000 to obtain IDAT (intensity data files) that capture raw DNAme intensities. Two patient PBMC samples were prepared and assayed as technical replicates on the EPIC arrays.

To establish a pediatric population reference set, publicly available DNAme microarray data on blood and buccal samples were downloaded from the Gene Expression Omnibus (GEO) database, including both EPIC arrays and their predecessors, the Illumina Infinium HumanMethylation450 BeadChip (450k) arrays. We gathered all cohorts in whole blood, PBMCs, and dried blood spots, and excluded cord blood samples. To ensure consistent data preprocessing and minimization of batch effects, we used only cohorts for which raw IDAT files were available in the database. The methylation data and metadata were then compiled manually. Missing metadata was curated from the GEO data set publications and included variables such as age, sex, tissue, ethnicity, and disease status. For disease status, we used the designation of HC for samples collected from individuals without persistent severe health conditions, as described in the cohort metadata and the associated publication. Only cohorts that included age information and had samples in the pediatric age range were retained. After quality control (QC), outlier removal, age matching (6–18 years), and selection for HCs, 725 buccal samples, 452 whole blood samples, and 285 PBMC samples remained. The cohorts and demographics are described in Tables S2, S3, and S4.

For the analysis of primary tissues, the HC cohorts and patient samples were preprocessed together. For T cells, two batches were collected, one with n = 1 and the second with n = 3 for both patient and rescued samples. The two T cell batches were processed and analyzed separately. All subsequent analyses were performed in R version 4.0.3 (R Foundation for Statistical Computing). The IDAT files were normalized with the preprocessFunnorm function in the minfi package, which included two types of normalization: Noob (normalized out-of-band) normalization and functional normalization (Fortin et al., 2014; Fortin et al., 2017). The two normalization steps utilized the readings from control and out-of-band probes to adjust for dye bias from the two-color channels and probe-type bias from type I and type II probes on the arrays, respectively. Samples from the 450k and EPIC microarrays were normalized separately, and then combined by subsetting to the shared probes.

QC was performed at both sample and probe levels. Samples with bisulfite conversion rate <75 or average detection P > 0.01 were removed. Three other outlier detection methods—detectOutlier from lumi (Du et al., 2008), outlyx from watermelon (Pidsley et al., 2013), and Z-test on the first principal component (PC)—were implemented to identify outliers with distinct methylation patterns. Furthermore, sex prediction was performed with the getSex function in minfi. Samples with mismatches in reported and predicted sex were also removed. Finally, probes in which >5% of samples had bead count <3 or an average detection P > 0.05 were filtered out. After QC, the missing values were imputed in a probe-wise manner based on the DNAm patterns of other probes with the lowest Euclidean distance. The method was implemented with the imputeKNN function in the impute package, where k = 10.

Cell-type proportion of whole blood and PBMC DNAm data was estimated from the imputed data with the Houseman reference–based prediction method (Houseman et al., 2012), as implemented using the estimateCellCounts2 function from the FlowSorted.Blood.EPIC package with the FlowSorted.BloodExtended.EPIC reference (Salas et al., 2022). In buccal swabs, cell-type proportion was predicted with the hepidish function in the EpiDISH package (Zheng et al., 2018). We compared the estimated proportion of each cell type between the patient and HCs using the one-sample t test.

Given that the patient shows unique cell-type proportions distinct from the HC, it is essential to adjust for cell type–associated DNAm variation to prevent confounding with the disease-associated DNAm signature. In the primary tissues, this was accounted for with a regression model. To overcome the compositional nature of cell-type proportion data, we first transformed the proportion into an isometric log ratio followed by PCA to extract components that capture the variabilities corresponding to cell-type differences, yet are not bounded or correlated. We then fitted a regression model of the DNAm beta value as a function of the cell-type PCs (Jones et al., 2017) and extracted the residuals as the cell type–adjusted DNAm values. For the analysis in T cells, we accounted for the CD8⁺ proportion directly as a covariate in our EWAS.

Given that we only included samples for which raw IDAT files were available, we also had access to the chip ID and position at which each sample was run. We identified three known sources of batch effect, i.e., the array platform (450k or EPIC), chip ID, and chip row. Therefore, we adjusted the data for these three known batch effects with the ComBat function in the sva package (Leek et al., 2012), which uses an empirical Bayesian framework to estimate the batch effect on samples across probes and account for them with a regression model. Each cohort can also be considered as a batch, but that is often confounded with different tissues, disease status, and age range, so we did not include cohort as a known batch variable. We confirmed that the batch effect of the three variables was removed with PCA. Platform ID, chip ID, and chip row were originally strongly correlated with PC1 through PC5 on Pearson’s correlation test (multiple test P_adj < 0.001), and most of the effects were removed by ComBat. For the T cell DNAm data, we accounted for batch in the limma regression model, as too few batches were present, and we did not detect a strong batch effect within each time point.

The epigenetic ages of the patient, family control, and unrelated HC samples (blood, buccal, and T cell) were estimated using the Horvath pan-tissue clock (Horvath, 2013), Hannum clock (Hannum et al., 2013), and Horvath skin-and-blood clock (Horvath et al., 2018). The buccal samples’ epigenetic age was estimated with the PedBE clock (Vivier et al., 2024). These estimations were performed using the methylclock R package (Pelegi-Siso et al., 2021).

Cell-type proportion prediction, subsequent accounting for intersample cellular heterogeneity, and adjusting for other sources of batch effects were performed prior to differential methylation analysis. For differential DNAm analysis, linear regression was performed with limma (Ritchie et al., 2015). For each primary tissue, the regression model accounted for sex and age, whereas for cultured T cells, we additionally accounted for batch and CD8⁺ T cell proportion measured by flow cytometry.

The expanded T cells for the DNAme arrays were collected over 3 days to capture some biological and technical variability. In the linear regression model, we accounted for concomitant variables in our analysis, including sex, age, and technical batch, to specifically isolate the effect of patient ASXL1 genotype. We identified 276,769 DMs with a statistical cutoff with P_adj ≤ 0.05 and effect size cutoff with delta-beta >0.1. The same regression model comparing rescue versus control yielded 208,692 DMs passing the statistical and effect size thresholds. We then included flow-derived CD8⁺ proportion with other concomitant variables in the model to account for cellular heterogeneity. With the new model, we no longer observed any DMs passing the statistical threshold of P_adj < 0.05, showing that ASXL1 variants led to an imbalance in CD4⁺ and CD8⁺ T cell proportions, and that ASXL1 has a strong cell population–specific effect on DNAme pattern. We relaxed the statistical threshold to P_adj < 0.2 and detected 141,113 DMs. Downstream analyses are performed with this model and detection threshold. Multiple test correction was performed with the Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995). Skewness of the differential DNAm pattern was assessed with skewness test for normality, implemented with the skewness.norm.test function in the normtest package (Shapiro et al., 2012).

Overrepresentation analyses were conducted using the gometh function in the missMethyl package to identify enriched ChromHMM chromatin states. The ChromHMM states were exported from the Roadmap Epigenomics database and overlapped with the Illumina annotation file for the EPIC arrays (Roadmap Epigenomics Consortium et al., 2015). The method accounts for the background probes on the DNAm arrays to minimize coverage biases. The probes were annotated to the corresponding genes, and the enriched pathways with multiple test P_adj < 0.05 were identified.

Intracellular flow cytometry was performed to examine the abundance of ASXL1, H2AK119ub1, and IL-2–induced phospho-STAT5 (pSTAT5) in expanded patient-derived primary T cells, pSTAT5 in PBMCs, and ASXL1 in primary fibroblasts as follows.

Expanded T cells were first washed and stained with CD4⁺ BUV563 (Clone: SK3, Cat# 612912; BD Biosciences), CD8⁺ BUV395 (Clone: RPA-T8, Cat# 563796; BD Biosciences), and Fixable Viability Stain 780 (Cat# 565388; BD Biosciences) for 15 min. Cells were then fixed (fibroblasts were lifted, spun, and fixed) using BD Cytofix (Cat# 554655; BD Biosciences) for 20 min at room temperature and permeabilized using Perm III for 30 min on ice (Cat# 558050; BD Biosciences). The cells were then stained with either primary antibody for ASXL1 (Cat# MA5-36238; Thermo Fisher Scientific) or mouse IgG1 isotype control for 1 h at room temperature in 1× Perm/Wash (Cat# 554723; BD Biosciences). Intracellular staining was performed separately for H2AK119ub (Cat# 8240S; Cell Signaling Technology) in expanded T cells. Subsequently, the cells were washed and stained with goat anti-mouse secondary antibody (Alexa Fluor 405, Cat# A-31553; Thermo Fisher Scientific) for 30 min at room temperature. ASXL1 expression was then measured using a BD FACSymphony flow cytometer (BD Biosciences) and analyzed using FlowJo software (BD Biosciences).

Detection of pSTAT5 was performed similar to the protocol outlined above in expanded T cells. After permeabilization, cells were stained with mouse anti-human pSTAT5 (pY694) PE-Cy7 (Cat# 560117; BD Biosciences). The following antibodies were used for extracellular staining of PBMCs: CD3 FITC (Clone: UCHT1, Cat# 555332; BD Biosciences), CD4⁺ BUV563 (Clone: SK3, Cat# 612912; BD Biosciences), CD8⁺ BUV395 (Clone: RPA-T8, Cat# 563796; BD Biosciences), CD27 BV605 (Clone: L128, Cat#: 562655; BD Biosciences), CD45RA BV421 (Clone: HI100, Cat# 562885; BD Biosciences).

Immunophenotyping on T, B, and NK cells with intracellular cytokine detection was also conducted with PBMCs from the patient, family members, and HCs. The PBMCs were stimulated with a cell stimulation cocktail (PMA/ionomycin, Cat# 00-4970-93; eBioscience Cell Stimulation Cocktail; Thermo Fisher Scientific) for 4 h at 37°C in the presence of protein transport inhibitors (brefeldin A and monensin, Cat# 00-4980-93; eBioscience Protein Transport Inhibitor Cocktail; Thermo Fisher Scientific). Cells were stained with two different antibody panels (listed in Table S1) using eBioscience Foxp3 Transcription Factor Staining Buffer Set (Cat# 00-5523-00; Invitrogen; Thermo Fisher Scientific). Samples were acquired on the BD FACSymphony flow cytometer (BD Biosciences). Data were analyzed using FlowJo (BD Biosciences).

Patient clinical immunophenotyping was carried out on whole blood at an accredited clinical flow cytometry laboratory according to standard protocols. Briefly, whole blood was collected and stained using dried antibody reagents (DURAClone tubes; Beckman Coulter) according to the manufacturer’s recommendations. Panels focused on the analysis of T cell subsets (DURAClone IM T Cell Subsets, Cat# B53328; Beckman Coulter) and B cell subsets (DURAClone IM B Cell, Cat# B53318; Beckman Coulter). Samples used for the B cell subsets were washed three times prior to staining. Aliquots of 100 μl of fresh blood were added to the dried reagent tubes and incubated in the dark at room temperature for 15 min prior to lysis with ImmunoPrep Reagent System (Beckman Coulter). Samples were then fixed with 1.0 ml of 1% paraformaldehyde prior to flow cytometry. All flow data were acquired on a Navios flow cytometer and then analyzed using Kaluza software (Beckman Coulter).

Formalin-fixed, paraffin-embedded lymph node tissue was cut into sections 3 µm thick for routine H&E staining or 5 µm thick for immunohistochemistry. Bone marrow trephine biopsies were processed in B-plus solution followed by decalcification and formalin fixation. Paraffin-embedded marrow tissue was then cut into sections 3 µm thick for routine H&E staining or 4 µm thick for immunohistochemistry.

Fig. S1 shows the extended immune phenotype and biopsy image of the patient. Fig. S2 shows the further functional assessment of the ASXL1 variants. Fig. S3 shows deeper analysis of the DNAm patterns and telomere length analysis. Fig. S4 characterizes the expanded T cells of the patient. Fig. S5 describes B and T cell phenotyping of the patient. Table S1 shows the list of antibodies used in the flow cytometry panels. Tables S2, S3, and S4 show cohort and demographic information of HC whole blood, PBMCs, and buccal swab samples used in the analysis of DNAm data. Supplemental Clinical Data provides a detailed chronological account of the patient’s clinical presentation of Hodgkin lymphoma, diagnostic workup, histopathological findings, treatment course (including chemotherapy, immunotherapy, and HSCT), and posttreatment outcomes.

The data underlying this study are not publicly available due to the potential identifying nature of the genomic data in a single patient. The data are available from the corresponding author upon reasonable request via emailing the corresponding author.

We are grateful to our colleagues, Karlie Edwards, Dr. Chaini Konwar, Dr. Hilary Brewis, Hannah-Ruth Engelbrecht, and Alan Kerr, who critically contributed to the flow and style of the manuscript. Thanks are also due to Dr. Nicole Gladish who facilitated the downloading and merging of ChromHMM data from the Roadmap Epigenomics database. We are also grateful for the help in patient management and organization of the investigation effort we received from Dr. Anne K Junker.

This work was supported by the Canadian Institutes of Health Research grant EGM-141897 (to M.S. Kobor), Canadian Institutes of Health Research grant PJT-178054 (to S.E. Turvey, C.M. Biggs), Genome British Columbia grant SIP007 (to S.E. Turvey), Michael Smith Health Research BC grant HPI-2018-2041 (to C.M. Biggs), and the BC Children’s Hospital Foundation and the Mining for Miracles initiative. M.S. Kobor is the Edwin S.H. Leong University of British Columbia Chair in Healthy Aging, C.M. Biggs is a Michael Smith Health Research BC Health Professional-Investigator, and S.E. Turvey holds a Tier 1 Canada Research Chair in Pediatric Precision Health and the Aubrey J. Tingle Professor of Pediatric Immunology position.

Author contributions: M.P. Fu: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, and writing—original draft, review, and editing. M. Sharma: conceptualization, data curation, formal analysis, methodology, and writing—original draft, review, and editing. P. Yousefi: investigation, validation and writing—review and editing. S.M. Merrill: conceptualization, formal analysis, and writing—review and editing. R. Tan: investigation. S. Samra: methodology and writing—review and editing. A. Setiadi: resources and writing–review and editing. L. Golding: writing—review and editing. B.P. Modi: data curation, formal analysis, and writing—review and editing. K.L. Del Bel: conceptualization, methodology, project administration, and writing—review and editing. R.J. Deyell: conceptualization, resources, and writing–review and editing. J. Rozmus: investigation, methodology, resources, and writing—review and editing. W. Rehmus: resources, visualization, and writing—review and editing. K.J. Hildebrand: writing—review and editing. E. James: writing—review and editing. G. Blanchard-Rohner: investigation and writing—review and editing. S. Lin: investigation. K.E. Shopsowitz: investigation and writing—review and editing. J. Terry: investigation, resources, visualization, and writing—review and editing. A.F. Lee: resources and writing—review and editing. B.I. Drögemöller: validation and writing—review and editing. A. Matthews: data curation and formal analysis. M. Tarailo-Graovac: data curation, formal analysis, investigation, and writing—review and editing. L. Sauvé: conceptualization and writing—review and editing, H. Mitchell: investigation. J.S. Prendiville: writing—review and editing. J.L. MacIsaac: investigation. K. Dever: data curation, investigation, methodology, project administration, and resources. D.T.S. Lin: investigation and resources. M. Meijer: writing—original draft, review, and editing. C.J.D. Ross: conceptualization, formal analysis, funding acquisition, methodology, resources, supervision, and writing—review and editing. S.R.M. Dobson: investigation and writing—review and editing. S.M. Vercauteren: data curation and formal analysis. W.W. Wasserman: formal analysis, project administration, supervision, and writing—review and editing. C.D.M. van Karnebeek: conceptualization, funding acquisition, resources, supervision, and writing—review and editing. M.L. McKinnon: conceptualization and writing—review and editing. M.S. Kobor: resources, supervision, and writing—review and editing. S.E. Turvey: conceptualization, funding acquisition, project administration, resources, supervision, and writing—review and editing. C.M. Biggs: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.