Somatic Mutations in UBA1 Define a Distinct Subset of Relapsing Polychondritis Patients With VEXAS

Abstract

Objective.

Somatic mutations in UBA1 cause a newly defined syndrome known as VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome). More than 50% of patients currently identified as having VEXAS met diagnostic criteria for relapsing polychondritis (RP), but clinical features that characterize VEXAS within a cohort of patients with RP have not been defined. We undertook this study to define the prevalence of somatic mutations in UBA1 in patients with RP and to create an algorithm to identify patients with genetically confirmed VEXAS among those with RP.

Methods.

Exome and targeted sequencing of UBA1 was performed in a prospective observational cohort of patients with RP. Clinical and immunologic characteristics of patients with RP were compared based on the presence or absence of UBA1 mutations. The random forest method was used to derive a clinical algorithm to identify patients with UBA1 mutations.

Results.

Seven of 92 patients with RP (7.6%) had UBA1 mutations (referred to here as VEXAS-RP). Patients with VEXAS-RP were all male, were on average ≥45 years of age at disease onset, and commonly had fever, ear chondritis, skin involvement, deep vein thrombosis, and pulmonary infiltrates. No patient with VEXAS-RP had chondritis of the airways or costochondritis. Mortality was greater in VEXAS-RP than in RP (23% versus 4%; P = 0.029). Elevated acute-phase reactants and hematologic abnormalities (e.g., macrocytic anemia, thrombocytopenia, lymphopenia, multiple myeloma, myelodysplastic syndrome) were prevalent in VEXAS-RP. A decision tree algorithm based on male sex, a mean corpuscular volume >100 fl, and a platelet count <200 ×10³/μl differentiated VEXAS-RP from RP with 100% sensitivity and 96% specificity.

Conclusion.

Mutations in UBA1 were causal for disease in a subset of patients with RP. This subset of patients was defined by disease onset in the fifth decade of life or later, male sex, ear/nose chondritis, and hematologic abnormalities. Early identification is important in VEXAS given the associated high mortality rate.

INTRODUCTION

Somatic mutations in UBA1 affecting methionine at codon 41 have recently been reported within clonal populations of hematopoietic stem cells in association with adult-onset inflammatory syndromes (1). Patients who harbor these mutations develop a newly defined disease in late adulthood known as VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome), which is characterized by myeloid-driven systemic inflammation and progressive bone marrow failure. Prior to the identification of UBA1 mosaicism in blood, patients with VEXAS were typically diagnosed clinically with a number of inflammatory diseases, most notably relapsing polychondritis (RP), but also giant cell arteritis, polyarteritis nodosa, and Sweet syndrome. In the initial description of VEXAS, 15 of 25 patients (60%) reported inflammation of cartilaginous structures and met established diagnostic criteria for RP (1). To what extent genetic variants in UBA1 define a specific subset of RP is currently unknown.

Clinical heterogeneity in RP is well described. Subsets of patients can be delineated based upon disease severity or pattern of cartilaginous involvement (2–4). A unique subset of older patients with profound hematologic abnormalities has been reported in RP (2,5–21). Similar to VEXAS, these patients often develop hematologic abnormalities within the spectrum of a myelodysplastic syndrome. Whether UBA1 mutations exclusively define the clinical subset of patients with RP and associated hematologic abnormalities or can be detected across a broader range of clinical phenotypes within RP is unknown.

The study objectives were to define the prevalence of somatic mutations in UBA1 within a large prospective cohort of patients with RP, to compare clinical features between patients with RP with somatic variants in UBA1 and those without the variants, and to create a clinical algorithm to identify patients with genetically confirmed VEXAS among those with RP.

PATIENTS AND METHODS

Study population.

To determine the prevalence of UBA1 mutations in RP, all patients in a prospective observational cohort of RP patients at the National Institutes of Health (NIH) underwent genetic testing. Patients included in this analysis were ≥18 years of age at the time of study enrollment and met McAdam’s or Damiani’s diagnostic criteria for RP (22,23). Per protocol, each patient underwent standardized clinical assessment, including audiology assessment, computed tomography (CT) scan of the chest, otolaryngologist evaluation, and pulmonary function tests. Clinical laboratory testing included erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) level, complete blood cell count, comprehensive metabolic panel, lipid panel, antineutrophil cytoplasmic antibodies, lupus anticoagulant, anticardiolipin antibodies, rheumatoid factor, antinuclear antibody, anti–double-stranded DNA, extractable nuclear antigen, complement levels, and urinalysis.

In addition to patients recruited within the RP cohort at the NIH, patients with UBA1 mutations identified using the original description of VEXAS (1) who met diagnostic criteria for RP were also included in this study. Data from these patients were included for clinical comparisons but were not included to estimate the prevalence of UBA1 mutations in RP. These patients were identified from other existing cohorts at the NIH and from the Leeds Teaching Hospitals NHS Trust in the UK. Every patient had a detailed clinical evaluation by their primary investigative study team, and outside clinical records were centrally reviewed by the study investigators.

Patients provided written informed consent and were enrolled in study protocols approved by local ethics review boards.

Genetic testing.

All patients in the NIH RP cohort underwent whole-exome sequencing per study protocol. Whole-exome sequencing (Otogenetics Corporation) was performed on patient peripheral leukocyte DNA using Agilent 51Mb Human Exome V5 capture and PE100–125 Illumina HiSeq2500 sequencing with an average read coverage of 100×. Quality control, kinship analysis, variant discovery, annotation, and filtering were performed as previously described (24). Briefly, sequence reads were aligned to human reference genome (GRC Build 37) with Burrows-Wheeler Aligner. These files were then processed to remove duplicate reads, refine alignment indels, and recalibrate base quality scores, according to the Genome Analysis Toolkit from the Broad Institute. Joint variant calls across multiple samples were determined using UnifiedGenotyper followed by a variant quality score recalibration using the VQSR tool (both from the Genome Analysis Toolkit).

Sanger sequencing from peripheral blood samples was used to confirm the presence of mutations in UBA1 in all patients, as previously described (1). Briefly, coding exons of UBA1 were sequenced using a BigDye Terminator version 1.1 Cycle Sequencing kit according to the instructions of the manufacturer (Applied Biosystems). Sequencing was performed on a Seq Stu-dio Genetic Analyzer (Applied Biosystems). Sequencing data were analyzed using Sequencher (Gene Codes).

Digital droplet polymerase chain reaction (PCR) was performed to quantify the variant allele fraction (VAF) for each detected variant in UBA1. Specific probes were generated for UBA1 c.121 A>G, c.121 A>C, and c.122 T>C. Reactions were performed using 11 μl 2× digital droplet PCR Supermix for probes, 900 nM target-specific PCR primers, and 250 nM mutant-specific (FAM) and wildtype–specific (HEX) probes. Twenty microliters of PCR mixture and 70 μl of droplet generation oil were mixed, and droplet generation was performed using a Bio-Rad QX100 Droplet Generator. The droplet emulsion was thermally cycled under the following conditions: denaturing at 95°C for 10 minutes, 40 cycles of PCR at 94°C for 30 seconds and at 55°C for 1 minute, and a final extension at 98°C for 10 minutes. PCR amplification in the droplets was confirmed using a Bio-Rad QX200 Droplet Reader. The threshold was determined by comparing the nontemplate digital droplet PCR results. All data were evaluated above the threshold. Quanta-Soft (Bio-Rad) was used to analyze the VAF data.

Flow cytometry.

Multipanel flow cytometry was performed for patients with RP and age and sex–matched healthy controls recruited in the NIH RP cohort and the NIH Healthy Volunteer Program. Whole blood samples in sodium heparin were collected. After red blood cell lysis using BD PharmLyse lysing buffer, the cells were washed with phosphate buffered saline (PBS) and resuspended in fluorescence-activated cell sorting buffer (PBS with 0.5% bovine serum albumin, 0.1% sodium azide, and 2 mM EDTA). The cell solution was divided between 5 roundbottom tubes (VWR). Each tube was incubated with a cocktail of antibodies specific to the cell subsets of interest. The detailed flow cytometry protocol and cell surface markers to identify each cell subset are described in the Supplementary Methods.

Clinical definitions.

For this report, all patients with RP with UBA1 mutations are described as having VEXAS-RP. All patients with RP who do not have detectable UBA1 mutations are referred to simply as having RP.

Definitions of organ involvement were applied to disease-relevant features. Ear chondritis was defined as physician-observed tender swelling of the pinna with associated redness, cauliflower ear, or other evidence of cartilage damage including floppy ears and thickened cartilage. Nose chondritis was defined as tenderness over the bridge of the nose with or without redness or swelling, tip of the nose tenderness with associated swelling or redness, saddle nose deformity, nasal crusting, nasal ulcers, or nasal septal perforation. Airway chondritis was defined as tracheomalacia, bronchomalacia, tracheal thickening, or sub-glottic stenosis. Tracheomalacia was defined as anterior or lateral flattening of the tracheal wall of ≥50% visualized during bronchoscopy or dynamic CT scan. Bronchomalacia was defined as bronchial collapse visualized during bronchoscopy or dynamic CT scan. Tracheal thickening was defined as ≥3-mm wall thickness as measured by chest CT. Pulmonary infiltrates was defined by opacification of air spaces visualized in chest CT or radiography. Subglottic stenosis was defined as pathologic narrowing of the subglottis visualized by direct laryngoscopy. Arthritis was defined as physician-observed synovitis/tenosynovitis or arthralgias with associated morning stiffness lasting longer than 1 hour. Vestibular/cochlear damage was defined as documented sensorineural hearing loss by audiometry and/or documented vestibular dysfunction by vestibular testing. Ocular inflammation was defined as physician-observed scleritis, episcleritis, iritis, or uveitis. Skin involvement was restricted to neutrophilic dermatosis or vasculitis confirmed by biopsy-proven skin pathology.

Clinical algorithm to identify patients with UBA1 mutations.

Decision tree analysis was used to create an algorithm to identify patients with VEXAS-RP among a cohort of patients with RP. Univariate receiver operating characteristic (ROC) curves were examined for each continuous variable to determine the optimal threshold based on maximized Youden’s index to differentiate patients with VEXAS-RP from other patients with RP, rounded to the nearest ten (25). The random forest method was used to rank variable importance, and a decision tree was fitted (R version 3.6.2; “RandomForest” Package). Missing values were imputed using proximity from the random forest using the “rfImpute” function. Random forest was trained with diagnosis (VEXAS-RP versus RP) as the outcome measure. Clinical and lab-oratory features as covariates were studied with 100,000 trees generated. Variable importance was ranked based on the mean decrease in Gini impurity.

Because the objective was to develop a clinical screening strategy to identify cases of VEXAS-RP among a population of patients clinically diagnosed as having RP who would then undergo subsequent diagnostic confirmation by genetic testing, priority was assigned to maximize sensitivity (i.e., not missing any VEXAS-RP cases, few false negatives) rather than specificity (i.e., misclassification of RP as VEXAS-RP, few false positives). Thus, higher cost penalty was imposed on VEXAS-RP misclassification by assigning a weight of 10 per VEXAS-RP case versus 1 per RP case. A decision tree was constructed using package “rpart,” and covariates were selected from the top-ranked variables of importance generated by random forest. Similarly, a weight of 10 was assigned per VEXAS-RP case versus 1 per RP case. Sensitivity and specificity were calculated after applying the algorithm to the whole cohort. The top-performing models were evaluated, and variables were prioritized that were objective, easy to measure, with excellent performance characteristics that optimized sensitivity while retaining excellent specificity.

Statistical analysis.

Additional statistical analyses were performed using JMP version 14.0.0 or GraphPad Prism 8. Wil-coxon’s rank sum test was used to compare distribution between groups. Fisher’s exact test was used to compare frequencies between groups. Spearman’s rank correlation was used to assess associations between continuous variables. Analysis of covariance was used to study the associations between cell counts (outcome measure), diagnosis (VEXAS-RP versus RP), and daily prednisone dose, modeling for interaction effects.

RESULTS

Genetic findings in study participants.

Ninety-two patients were included from the ongoing prospective, observational NIH RP cohort (ClinicalTrials.gov identifier: NCT02257866). All enrolled patients underwent exome sequencing as part of the study design. Of these, 7 patients were found to have UBA1 somatic mutations at p.Met41, for an overall prevalence of genetically confirmed VEXAS-RP within a cohort of patients with RP of 7.6%.

Six additional patients with VEXAS-RP were included from other cohorts: 4 patients from other cohorts at the NIH and 2 patients from the Leeds Teaching Hospitals NHS Trust in the UK. Genetic information on each patient with VEXAS-RP is summarized in Table 1. All 3 variants in UBA1 that were previously associated with VEXAS (p.Met41Val, p.Met41Leu, p.Met41Thr) were identified in the cohort. The VAF of UBA1 mutations in peripheral blood ranged from 32.6% to 89.4%. No significant correlations were observed between VAF, age at symptom onset, and disease duration.

Table 1.

Genetic characteristics of the patients with VEXAS-RP^*

Patient	Mutation type	VAF, %	Age at symptom onset, years	Disease duration, years†	Death
1	p.Met41Thr (c.122 T>C)	54.9	45	10	No
2	p.Met41Val (c.121 A>G)	82.6	55	3	Yes
3	p.Met41Val (c.121 A>G)	77.6	56	11	Yes
4	p.Met41Thr (c.122 T>C)	84.2	63	8	Yes
5	p.Met41Thr (c.122 T>C)	85.9	64	6	No
6	p.Met41Thr (c.122 T>C)	69.9	53	3	No
7	p.Met41Thr (c.122 T>C)	73.1	64	3	No
8	p.Met41Leu (c.121 A>C)	32.5	70	6	No
9	p.Met41Leu (c.121 A>C)	76.2	64	4	No
10	p.Met41Thr (c.122 T>C)	68.1	56	2	No
11	p.Met41Thr (c.122 T>C)	89.3	56	6	No
12	p.Met41Leu (c.121 A>C)	96.8	64	10	No
13	p.Met41Thr (c.122 T>C)	96.2	68	8	No

^*VEXAS-RP = vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome with relapsing polychondritis; VAF = variant allele fraction.

^†Defined as time from symptom onset to time of DNA testing.

Clinical comparisons between VEXAS-RP and RP.

Of the 98 patients included in the study, the majority were female (n = 72; 73%) and white (n = 90; 92%) (Table 2). The median age at symptom onset was 38 years (interquartile range [IQR] 30–47), and the median disease duration was 8 years (IQR 3.3–13). The most common clinical manifestations across the whole cohort were nose chondritis (n = 83; 85%), arthritis (n = 83; 85%), costochondritis (n = 72; 73%), vestibular symptoms (n = 65; 68%), and ear chondritis (n = 61; 62%). The most common complication was intensive care unit admission (n = 20; 21%), followed by unprovoked deep vein thrombosis (n = 12; 12%) and death (n = 6; 6%). Each of the 6 patients who died had severe, progressive disease at the time of death; however, the exact causes of death were unknown.

Table 2.

Clinical characteristics of the patients with VEXAS-RP compared to RP^*

	All patients (n = 98)	Patients with RP (n = 85)	VEXAS-RP (n = 13)	P
Demographic characteristics
White ethnicity	90 (92)	77 (91)	13 (100)	0.59
Male sex	26 (27)	13 (15)	13 (100)	<0.0001
Age at current visit, median (range) years	47 (18–82)	45 (18–82)	62 (48–71)	<0.0001
Disease duration, median (IQR) years	8 (3.25–13)	8 (4–15)	4 (2–5)	0.0068
Age at symptom onset, median (range) years	38 (12–74)	37 (12–74)	56 (45–70)	<0.0001
Clinical symptoms
Fever	33 (34)	20 (24)	13 (100)	<0.0001
Weight loss	14 (14)	10 (12)	4 (31)	0.091
Ear chondritis	61 (62)	48 (56)	13 (100)	0.0015
Nose chondritis	83 (85)	71 (84)	12 (92)	0.68
Airway chondritis	37 (38)	37 (44)	0 (0)	0.0015
Costochondritis	72 (73)	72 (85)	0 (0)	<0.0001
Arthritis	83 (85)	77 (91)	6 (46)	0.0005
Hearing loss	30 (32)	25 (29)	5 (50)	0.28
Vestibular symptoms	65 (68)	62 (73)	3 (27)	0.0044
Skin involvement	33 (34)	22 (26)	11 (85)	<0.0001
Periorbital edema	6 (6)	2 (2)	4 (32)	0.0025
Laboratory values
ESR, median (IQR) mm/hour	12 (6–22)	11 (5–19)	66.5 (42–110)	<0.0001
CRP, median (IQR) mg/liter	2.9 (0.8–9.6)	1.9 (0.6–6.3)	17.7 (9.6–99.5)	<0.0001
Lupus anticoagulant positive	25 (26)	18 (21)	7 (54)	0.019
RF positive	4 (4)	0	4 (31)	0.0002
ANCA positive	0	0	0	1.00
Platelet count, median (IQR) ×10³/μl	246 (201–299)	258 (227–312)	145 (100–169)	<0.0001
Hemoglobin, median (IQR) gm/dl	13.2 (12–14)	13.4 (12–14)	10 (8–12)	<0.0001
MCV, median (IQR) fl	93.05 (90–98)	92.2 (89–95)	105 (102–115)	<0.0001
Absolute lymphocyte count, median (IQR)	1.6 (1.1–2.3)	1.78 (1.4–2.4)	0.92 (0.5–1.2)	<0.0001
Absolute monocyte count, median (IQR)	0.49 (0.3–0.6)	0.5 (0.4–0.6)	0.26 (0.1–0.3)	<0.0001
CT scan abnormalities
Pulmonary infiltrates	16 (16.33)	6 (7.06)	10 (77)	<0.0001
Complications
Death	6 (6)	3 (4)	3 (23)	0.029
ICU admission	20 (21)	16 (19)	4 (33)	0.24
Need for transfusion	6 (6)	0	6 (46)	<0.0001
Unprovoked DVT	12 (12)	4 (5)	8 (62)	<0.0001
Myelodysplastic syndrome	3 (3)	0	3 (23)	<0.001
Multiple myeloma	1 (1)	0	1 (7)	0.04
MGUS	1 (1)	0	1 (7)	0.04
Medications
Prednisone dose at time of CBC assessment, mean (IQR) mg	7.25 (0–20)	5 (0–20)	17.5 (5–30)	0.065
Receiving prednisone at time of CBC assessment	61 (63)	51 (60)	10 (91)	0.045
No. of steroid sparing agents ever, mean (IQR)	3 (2–4)	2 (2–4)	4 (3–7)	0.0043

^*Except where indicated otherwise, values are the number (%) of patients. VEXAS-RP = vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome with relapsing polychondritis; IQR = interquartile range; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; RF = rheumatoid factor; ANCA = antineutrophil cytoplasmic antibody; MCV = mean corpuscular volume; CT = computed tomography; ICU = intensive care unit; DVT = deep vein thrombosis; MGUS = monoclonal gammopathy of undetermined significance; CBC = complete blood cell.

A complete list of clinical comparisons is detailed in Table 2. Compared to patients with RP (n = 85), patients with VEXAS-RP (n = 13) were exclusively male (100% versus 15%; P < 0.001), with a greater prevalence of fever (100% versus 24%; P < 0.001), ear chondritis (100% versus 56%; P = 0.0015), skin involvement (85% versus 26%; P < 0.0001), pulmonary infiltrates (77% versus 7%; P < 0.0001), and periorbital edema (32% versus 2%; P = 0.0025). Neutrophilic dermatosis (46% versus 0%; P < 0.001) and cutaneous vasculitis (38% versus 1%; P < 0.001) was more common in patients with VEXAS-RP than in those with RP. The pulmonary infiltrates observed in VEXAS-RP were not infectious and were variable in severity, resulting in mild symptoms such as cough and shortness of breath or more severe symptoms leading to mechanical ventilation. Repeat chest imaging, when available, demonstrated complete resolution of the infiltrates in response to treatment with glucocorticoids. Compared to VEXAS-RP, patients with RP had a significantly higher prevalence of airway chondritis (44% versus 0%; P = 0.0015), costochondritis (85% versus 0%; P < 0.0001), and arthritis (91% versus 46%; P = 0.0005). Representative images of key clinical features that differentiate between VEXSAS-RP and RP are shown in Figure 1.

Figure 1.

Clinical manifestations in patients with VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome) with relapsing polychondritis (VEXAS-RP) and patients with RP. A and B, Pulmonary parenchymal disease (e.g., inflammatory infiltrate) is common in VEXAS-RP (A), while disease of the large airways (e.g., tracheomalacia) is seen only in RP (B). C and D, Chondritis of the ear and nose without resultant cartilage damage is common in VEXAS-RP (C), while cauliflower ear and saddlenose deformity are features seen only in RP (D). E, Skin involvement (e.g., leukocytoclastic vasculitis and neutrophilic dermatosis) is a defining feature of VEXAS-RP that is not typically seen in RP.

There were treatment differences observed between the 2 groups. In general, VEXAS is treatment-refractory to medications other than glucocorticoids (1). On average, patients with VEXAS-RP received a higher mean number of steroid-sparing medications compared to patients with RP (4 versus 2; P = 0.0043) (Table 2). The prevalence of glucocorticoid use and daily prednisone dose was also greater in patients with VEXAS-RP compared to those with RP.

Patients with VEXAS-RP also had unique laboratory findings, including significantly greater values of maximum ESR and CRP level and a greater prevalence of detectable lupus anticoagulant, rheumatoid factor, thrombocytopenia, anemia, and macrocytosis. Additionally, total lymphocyte and absolute monocyte counts were significantly lower in patients with VEXAS-RP compared to patients with RP (Table 2).

Unique interaction effects were observed between cell counts in association with daily prednisone dose in patients with VEXAS-RP compared to RP. Adjusting for prednisone dose, absolute neutrophil counts were similar in patients with VEXAS-RP compared to RP (β estimate 0.05, P = 0.94). Neutrophil count was negatively associated with increased prednisone dose in VEXAS-RP; however, increased prednisone dose was positively associated with neutrophil count in RP (P for interaction = 0.02) (Figure 2A). Adjusting for prednisone dose, absolute monocyte count and absolute lymphocyte count were significantly lower in patients with VEX-AS-RP (β estimate −0.10, P < 0.01) compared to RP (β estimate −0.46, P < 0.01), without a statistically significant interaction effect (P = 0.06 and P = 0.68, respectively) (Figures 2B and C).

Figure 2.

Association of cell counts and prednisone dose in patients with VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome) with relapsing polychondritis (VEXAS-RP) and patients with RP. A, Absolute neutrophil counts were similar in patients with RP and those with VEXAS-RP but differed in relation to daily prednisone dose (P for interaction = 0.02). B and C, Absolute monocyte count (B) and lymphocyte count (C) were higher in patients with RP compared to patients with VEXAS-RP, and a significant interaction effect with daily prednisone dose was not observed. Symbols represent individual subjects.

Immunologic profiling of VEXAS-RP and RP.

Similarities and differences in immune cell subset abundance were identified in patients with VEXAS-RP compared to patients with RP and healthy matched controls. The most striking differences were observed in the B lymphocyte and monocyte compartments (Figure 3). Reduction in naive B lymphocytes and nonclassical/intermediate monocyte counts relative to healthy controls was observed in both VEXAS-RP and RP, and these cell populations were both significantly lower in VEXAS-RP when compared directly to patients with RP. Patients with VEXAS-RP had significantly lower CD4+/CD8+ T cell percentages when compared to controls but not when compared to other patients with RP. Patients with VEXAS-RP had a greater percentage of activated CD8+ T cells (CD3+, CD4−, CD8+, HLA–DR+) when compared to controls and patients with RP. Patients with RP had a significantly higher relative percentage of Th17 cells compared to healthy controls, but not when directly compared to VEXAS-RP. Complete results from these analyses are shown in Supplementary Table 1 and Supplementary

Figure 3.

Immune cell subset differences between patients with VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome) with relapsing polychondritis (VEXAS-RP) and patients with RP. Monocyte (A) and B cell (B) subsets were significantly different in both RP and VEXAS-RP relative to matched healthy controls. Intermediate and nonclassic monocytes and B cell subsets were significantly reduced in VEXAS-RP when directly compared to RP. Data are shown as box plots, with lines inside the boxes showing the median, boxes showing the interquartile range, and bars outside the boxes showing the minimum and maximum values. Symbols represent individual subjects. * = P < 0.05; ** = P < 0.01; *** = P < 0.001; **** = P < 0.0001.

Clinical algorithm to identify patients with VEXAS-RP.

ROC curves demonstrated that a mean corpuscular volume (MCV) >100 fl and a platelet count <200 ×10³/μl optimally differentiated VEXAS-RP from RP. The relative performance of the individual variables used to identify VEXAS-RP among cases of RP is shown in Supplementary Figure 2. The top-performing decision tree was comprised of male sex as the first node of the tree, followed by an MCV >100 fl as the second node, and a platelet count <200 ×10³/μl as the third node (Figure 4). With this algorithm, all patients with VEXAS-RP were correctly identified (100% sensitivity), and 3 patients with RP were incorrectly classified as VEXAS-RP (96% specificity). Sixteen patients with RP without UBA1 mutations had an MCV >100 fl (n = 8) or a platelet count <200 ×10³/μl (n = 10), including the 3 men who were incorrectly predicted to have VEXAS-RP using the clinical algorithm. Deeper sequencing of peripheral blood by digital droplet PCR in these 16 patients did not uncover additional UBA1 mutations at a VAF threshold above 0.01% (data not shown).

Figure 4.

Decision tree algorithm. Flow chart details how 3 clinical variables (male sex, mean corpuscular volume [MCV] >100 fl, and platelet count <200 ×10³/μl) can identify patients with genetically confirmed VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic syndrome) with relapsing polychondritis (VEXAS-RP) (i.e., mutations at p.Met41 in UBA1) with 100% sensitivity and 96% specificity, when applied to a cohort of patients diagnosed as having RP.

DISCUSSION

The recent discovery of somatic mutations in UBA1 at p.Met41 causing VEXAS represents an important advancement in RP research. The present study deepens our understanding about the relative contribution of UBA1 mutations in hematopoietic stem cells as a causal disease mechanism in RP. Mutations in UBA1 were detected in 7.6% of patients with RP. The prevalence estimate from this study aligns with a prior clinical study from France, where older male patients with hemopathies represented 9% of a cohort of 142 patients with RP; however, these patients were not genetically tested for UBA1 mutations (2). All patients identified as having VEXAS-RP had a VAF of >30% and were detected and confirmed using multiple modalities, with more sensitive genetic approaches failing to identify additional cases. These patients can be readily identified based on a distinct clinical profile of ear/nose chondritis, male sex, disease onset in adulthood, and concomitant hematologic abnormalities including macrocytic anemia and thrombocytopenia.

Findings from this study provide further context to clinical heterogeneity in RP. Our group previously used latent class analysis within the NIH RP cohort to identify 3 main patterns of disease (4). Type I RP was characterized by cartilaginous destruction of the ears, nose, and upper airway. Type II RP was defined by lower airway–predominant disease. Type III RP was defined by nondestructive involvement of primarily the ears, nose, and joints. In this study, patients with VEXAS-RP would be categorized as having type III RP. Cartilage involvement in these patients was confined mainly to the ears, nose, and joints without obvious resultant cartilaginous damage. No patient with VEXAS-RP had chondri-tis of the large airways; however, infiltrative inflammatory disease of the lungs was common. While patients with VEXAS-RP may have a less severe form of chondritis than other patients with RP, treatment-refractory disease leading to further complications and death was more common in patients with VEXAS-RP. These data support the concept that RP is a clinically heterogenous disease and that defining subsets of patients with RP based on clinical phenotype is a useful framework to investigate divergent causal mechanisms of disease.

While this study explicitly focused on patients with VEXAS who met diagnostic criteria for RP, VEXAS is a pleomorphic dis-ease. The clinical hallmark of VEXAS is the presence of treatment-refractory severe systemic inflammation in association with profound hematologic abnormalities which can evolve into overt hematologic malignancy. The systemic inflammation observed in VEXAS is myeloid-driven and affects multiple tissues including cartilage, skin, lung parenchyma, and blood vessels. Consequently, patients with VEXAS often meet established clinical diagnostic criteria for a range of rheumatic diseases, including RP, giant cell arteritis, polyarteritis nodosa, and Sweet syndrome (1). Identification of a shared genetic etiology that spans multiple clinical diagnoses may provide insight into shared mechanisms that foundationally underlie systemic inflammation and may reveal novel therapeutic approaches across a spectrum of rheumatic diseases.

Findings from this study inform the clinical identification of patients with VEXAS-RP. A simple algorithm based on easily measured clinical parameters demonstrated nearly perfect accuracy in identifying which patients with RP would be genetically diagnosed as having VEXAS. For patients with chondritis of the ear and nose who are clinically diagnosed as having RP, genetic testing for UBA1 mutations should be strongly considered if the patient is male and has an MCV >100 fl or a platelet count <200 × 10³/μl. Because VEXAS is a newly identified genetic disease, the associated clinical spectrum of disease will likely further expand with future investigations. While the proposed clinical algorithm is useful to identify which patients with RP likely have VEXAS, more research is needed to inform specific genetic screening guidelines for VEXAS in parallel with an evolving understanding of the complete clinical spectrum of the disease. Additionally, because patients with VEXAS may develop hematologic malignancies including multiple myeloma or myelodysplastic disease, early identification of UBA1 mutations may end up being a powerful method to identify individuals with rheumatic diseases who are at risk of hematologic malignancy and may benefit from increased surveillance.

These data refine our understanding about the pathophysiology of VEXAS within the broader context of RP and may have therapeutic implications. In VEXAS, early marrow progenitor cells, including myeloid and lymphoid progenitors, display mosaicism for UBA1 variants; however, somatic mutations are lineage-restricted in peripheral blood to myeloid cells and are absent from lymphocytes. Decreased circulating T and B cell counts in VEXAS compared to other patients with RP and healthy controls suggest that mutant lymphoid cells in marrow fail to produce mature lymphocytes resulting in decreased total cell numbers. Reduction in total and intermediate B lymphocytes and skewing of monocyte differentiation toward classical monocytes in both VEXAS-RP and RP, relative to controls, highlights the potential for shared pathophysiology in both patient populations. B cell depletion therapy (e.g., rituximab) is not efficacious as a treatment for VEXAS or RP, which is consistent with the observation that total B cell counts are often reduced in association with disease in these patients (1,26).

Elevated serum levels of cytokines related to monocyte/ macrophage activation have been reported in association with VEXAS and other forms of RP (1,27). Similarly, therapies that target macrophage-related cytokines are at least partially efficacious in both VEXAS and other forms of RP (28–32). Activation of T lymphocytes was observed in patients with VEXAS-RP relative to patients with RP and controls, possibly due to cell-nonautonomous effects secondary to myeloid-driven inflammation. Finally, a Th17 association was observed in patients with RP relative to controls, but not in patients with VEXAS-RP relative to controls. This observation may have therapeutic implications for patients with RP, as many existing therapeutics that target Th17 pathways are not commonly administered to patients with RP.

Activated neutrophils contribute heavily to inflammation in VEXAS, exemplified by abundant neutrophilic infiltrate on histologic specimens and enhanced spontaneous neutrophil extracellular trap formation in ex vivo studies (1). While total lymphocyte and monocyte counts were reduced in patients with VEXAS-RP compared to other study patients, absolute neutrophil counts were comparable. Compensatory mechanisms of neutrophil production and release in VEXAS likely explain this observation. Circulating neutrophil counts typically increase in response to glucocorticoid treatment due to demargination of peripheral neutrophils, delayed migration of neutrophils into tissue, and increased release of immature neutrophils from marrow (33). Indeed, in the present study, absolute neutrophil counts were positively associated with glucocorticoid treatment in patients with RP. In contrast, a paradoxical inverse relationship between neutrophil count and glucocorticoid dose was observed in VEXAS-RP, suggesting that there is a failure of neutrophil demargination, altered migration, and aberrant bone marrow dynamics in this disease. Future studies examining the role of neutrophil production and migration in VEXAS are warranted.

There are some study limitations to consider. An accurate estimate of the prevalence of VEXAS within RP was limited by the relatively small sample size in this study and was potentially impacted by referral bias. Distinct bone marrow findings, including vacuolization of myeloid progenitor cells, have been reported in VEXAS; however, patients with other forms of RP do not routinely undergo bone marrow biopsy, precluding comparisons of mar-row findings. Patients in this study were genetically screened for VEXAS using peripheral blood. Screening bone marrow aspirate could potentially increase sensitivity to detect UBA1 variants in smaller clonal populations; however, genetic findings in peripheral blood have accurately reflected findings in bone marrow in patients with VEXAS (1), and deep sequencing of blood did not uncover additional cases.

In conclusion, this study determined that 7.6% of patients with RP have somatic mutations in UBA1 that are detectable in blood. Among patients with RP, patients with mutations in UBA1 can be readily identified based on key clinical symptoms including older age at disease onset, male sex, chondritis that spares the airway and chest wall, and hematologic abnormalities including macrocytic anemia and thrombocytopenia. These patients often develop progressive bone marrow failure and should be screened for hematologic malignancies. Discovery of effective therapies, while important for patients with RP in general, is particularly important for patients with VEXAS-RP due to the associated high mortality rate. UBA1 is required to initiate ubiquitylation, which is essential for modulating signaling pathways and targets proteins for degradation via the proteasome or autophagy–lysosome system. While it remains possible that defects in the ubiquitin proteasome system unrelated to genetic defects in UBA1 contribute to dis-ease pathophysiology in a broader group of patients with RP, the profound clinical and immunologic differences delineated in this study strongly suggest that patients with VEXAS-RP are distinctly different from other patients with RP.