UBA1 variant signature in unselected HN patients
Among the 8,976 patients with HN, 86 (1%) were found to have at least one presumably somatic UBA1 variant, including two patients who each had two distinct UBA1 variants (Table 1). Sixty-seven variants in 66 patients were classified as PV, while the remaining 21 variants were classified as VUS, with insufficient evidence to support pathogenicity. In all PV, the canonical loss of start-codon p.M41T (n = 24, 35% of all PV, Fig. 1b) was the most frequent recurrent variant, followed by p.M41L (n = 21, 31%), p.M41V (n = 14, 11%) and c.118-1G > C (n = 2, 3%), a previously reported splice site variant upstream to p.M41.1,3 Additionally, novel splice site variants, c.118 − 10_118-1 del (n = 1) and c.118-5_118-1 del (n = 1), along with three previously reported likely VEXAS-causal missense variants p.Y55H (n = 1), p.G477A (n = 1) and p.A478S (n = 1), were also classified as PV (Fig. 1b).1,3,10 Furthermore, 17 novel missense variants including two recurrent variants (p.D506N and p.I890F, below the protein sequence in Fig. 1b), along with 2 frame-shift variants (Fig. 1b, p.S816fs and p.L954fs) and 1 nonsense variant (p.Y1055*), were classified as VUS. PVs were concentrated in the first inactive adenylation domain (IAD) and the first active adenylation domain (AAD), whereas VUSs occurred in all functional domains across the protein sequence, except for the second IAD and the second AAD.
Genetic profiles in UBA1-mutant patients with HN
Concomitant somatic variants were detected by NGS in patients with presumably somatic UBA1 variants (Fig. 2a and Table 1). Thirty-one (47%) patients with UBA1 PV exhibited at least one concomitant variant, a significantly lower frequency compared to HN patients with UBA1 VUS (Fig. 2a-b, 85%, p = 0.004,). A lower somatic mutation burden, defined as the number of somatic variants per patient, was noted in patients with pathogenic variants (PV, mean ± SEM, 2.0 ± 0.2) compared to those with UBA1 VUS (Fig. 2c, 4.0 ± 0.5, p = 0.0001). UBA1 clone sizes were notably larger in PV (51.7% ± 3.3) than those in VUS (Fig. 2d, 27.1% ± 6.0, p = 0.0005). Fifty-five PV patients (83%) exhibited UBA1 variant VAFs higher than those of the leading non-UBA1 variants, indicating that UBA1 PV were the founding clones in these patients (Fig. 2e). In contrast, it appeared that only 40% of UBA1 VUS were the leading clones (Fig. 2f, 83% in PV vs 40% in VUS, p = 0.0001). In patients with UBA1 PV, DNMT3A was the most commonly mutated gene (23%), followed by TET2 (12%) and ASXL1 (6%, Fig. 2a, g-i). Conversely, the most prevalent concomitant variant in UBA1 VUS patients was TET2 (35%, p = 0.03), followed by ASXL1 (25%, p = 0.02), whereas DNMT3A variants were only detected in 2 patients (10%, p = 0.03, Fig. 2a, g-i). Notably, mutations in genes involved in tyrosine kinase or RAS signaling pathways were significantly more prevalent in VUS patients (Fig. 2a, g-i), mirroring those in HN without somatic UBA1 variants.
Clinical and pathologic characteristics of patients with HN and UBA1 variants
Corresponding to the genetic characteristics, patients with UBA1 PV and VUS displayed distinct epidemiological and clinical features, along with pathologic findings and diagnoses (Fig. 2a). Consistent with earlier studies,1,5,11 there was a pronounced male predominance among patients with PV (Table 2, 98%), which was slightly reduced in patients with VUS (80%, P = 0.002). The median age at variant detection was 69 years for patients with PV and 68 years for those with VUS (Table 2). The hallmark hematologic feature distinguishing PV and VUS was macrocytic anemia (Table 2, MCV mean ± SEM, 105.3 fL ± 1.4 in PV and 96.2 ± 4.1 fL in VUS, P = 0.01), observed in 31 patients (84%) with PV for whom hematologic information was available. Vacuoles were observed in bone marrow hematopoietic progenitor cells in 47 patients (89%) with PV (100% of patients with documented VEXAS syndrome diagnosis) and in 2 patients (Table 2, 10%, P < 0.0001) with VUS, while a hypercellular marrow was evident in nearly all patients (98% in PV and 90% in VUS). Among the PV patients for whom clinical diagnoses were available, 51 had documented VEXAS syndrome, while 4 did not (Fig. 2a). The majority of inflammatory disorders observed in patients with UBA1 variants consisted of ANA-negative autoimmune disorders. Systemic autoinflammation was observed in 90% of PV patients, whereas 30% (Fig. 2a) of VUS patients exhibited inflammation affecting at least one anatomical site. The most commonly affected anatomical site was the skin, followed by joints and lungs, resulting in respiratory failure (Fig. 2a). Other affected sites included muscles, eyes (uveitis, and retrobulbar cellulitis, RC), ears, nose, and muscles.
Concurrent HN were frequent among patients with UBA1 variants. MDS was the most common concurrent hematologic neoplasm in PV patients (8/51, 16%), followed by 3 PCNs (3/51, 6%, 1 CML (1/51, 2%), and 1 LPL (1/51, 2%) (Fig. 2a). The diverse genetic profiles previously outlined in VUS patients (Fig. 2a-i) were also reflected by the heterogeneity of pathological findings and a wide clinical spectrum including 5 MDS, 4 AML (0% in PV vs 22% in VUS, P = 0.004), 3 CMML (0% in PV vs 22% in VUS, P = 0.02), 2 Philadelphia chromosome-negative MPN, 1 SM-AHN, and 1 B-lymphoblastic leukemia (Fig. 2a). The higher average blast counts in VUS patients (Table 2, mean ± SEM, 2.4% ± 0.4 in PV and 29.1% ± 12.3 in VUS, P = 0.0002) also reflected the diverse spectrum of WHO entities, including AML. Most patients presented with anemia accompanied by elevated serum erythropoietin (EPO), with no observed instances of copper deficiency (Fig. 2a).
Rapid dynamics of acquired UBA1 PV accelerates the progression of low-risk MDS
VEXAS is associated with an increased risk of developing a myeloid neoplasm (MN), characterized by cytopenia(s), myelodysplasia, and cytogenetic and molecular abnormalities.1,3,5 Here, we compared the clonal metrics and dynamics in UBA1 PV patients to age- and gender-matched UBA1 wildtype (WT) patients with CCUS and MDS. The average number of somatic variants in UBA1 PV patients (Fig. 3a, 2.0 ± 0.2), was similar to that in UBA1 wildtype CCUS (2.0 ± 0.2) but significantly lower compared to MDS patients (Fig. 3a, 4.0 ± 0.2, P < 0.0001). Additionally, the clone sizes in all variants and those excluding UBA1 PV (20.2 ± 1.7 and 17.6 ± 1.7, respectively) were significantly larger than in CCUS (12.6 ± 0.4, P < 0.0001) but smaller than in MDS patients (30.5 ± 0.7, P < 0.0001). Furthermore, PV patients showed a similar mutation prevalence in individual genes and within major gene categories, including TP53, compared to those in CCUS, but with a lower prevalence compared to MDS patients (Fig. 3c-d). Four UBA1 PV patients were sequenced sequentially over a time interval ranging from 1 month to 8.5 months, and no acquisition of additional somatic variants was seen (supplementary Fig. 1, pts. 1, 10, 16, and 34). In three patients with clone expansional trajectories, growth in VAF from the 25th to the 75th percentile was estimated over 4.8 months (Fig. 3e), and this rapid clonal dynamic was reflected by the peak distribution of normalized VAF at 30–40% (Fig. 3f). In contrast, in UBA1 WT CCUS and MDS patients, the most prevalent mutations in TET2 and DNMT3A showed slow clone dynamics, and the changes in VAFs from 25th to 75th percentile was noted over a median time of 24.4 months in TET2 and 13.7 months in DNMT3A, respectively (Fig. 3g). The typical slow clonal expansion in TET2 and DNMT3A was also mirrored by a similar bimodal distribution of VAFs with two stable states in VAFs of 10% or less and VAFs of 40% and above (Fig. 3h). In contrast to healthy individuals with clonal hematopoiesis, in whom somatic mutations typically displayed low and consistent fitness advantages, UBA1 PV clones in affected patients exhibited notably higher fitness levels compared to control TET2 and DNMT3A mutations, despite the already elevated fitness levels in these UBA1 WT MDS and CCUS patients (Fig. 3i).
Most patients received prednisone, while a few received a hypomethylating agent (HMA) azacitidine, and one was bridged to stem cell transplant (Fig. 2a). The OS at 5 years was 80%, and the median OS was 77 months (Fig. 3k) in UBA1 PV patients, significantly more favorable compared to the age-matched UBA1 WT male MDS patients treated at HCI (Fig. 3j-k, 38 months, p = 0.007). Data enrichment through literature search further confirmed the OS in UBA1 PV patients, often accompanied by low blast count in bone marrow (Table 2), similar to the age matched low-risk but superior to standard-risk MDS patients (Fig. 3j-k). This superior OS appeared independent of patients’ age, hematologic parameters, presence of autoinflammation, blast count in bone marrow, somatic variant burden, VAFs of UBA1 PV, and presence of other concurrent variants excluding TET2 and DNMT3A, regardless of cytogenetic abnormalities or treatment (Fig. 3l-m). The presence of concurrent DNMT3A and/or TET2 variants conferred an adverse prognosis (Fig. 3l-m, hazard ratio [HR] = 2.65; 95% confidence interval [CI], 1.13–21; p = 0.03), similar to a previous study.5