Tackling the diagnosis of HA20 in children: challenges of a highly variable clinical and genetic spectrum

Abstract

Objective

To describe the clinical, biological and genetic characteristics of paediatric-onset A20 haploinsufficiency (HA20) and to identify key clinical features that may guide early diagnosis and management.

Methods

Multicentre, retrospective observational cohort study through the French national paediatric rheumatology network and national reference laboratories for genetic autoinflammatory diseases. 17 patients from 11 unrelated families, with disease onset before 18 years of age and carrying a tumour necrosis factor alpha-induced protein 3 (TNFAIP3) mutation, were included.

Results

Median age at symptom onset was 3 years (range: 1 month–17 years) with a median diagnostic delay of 9.7 years (range: 1–43 years). Boys had a significantly earlier onset and diagnosis than girls. The most common initial manifestations were oral ulcers (64.7%), fever (64.7%), abdominal pain (47.1%), arthralgia (35.3%) and skin eruption (17.6%). Genital ulcers occurred in 17.6% at onset but in 52.9% during follow-up. Gastrointestinal involvement was frequent (abdominal pain in 76.5% of patients and colitis in 35.3%). Other features included arthralgia (52.9%), arthritis (29.4%), skin eruption (41.2%), lymphadenopathy (35.3%), hepatomegaly (11.8%), headache (17.6%) and uveitis (5.9%). C reactive protein (CRP) levels were significantly higher during flares. Three novel TNFAIP3 variants were identified. Colchicine was effective as monotherapy in 40%. TNF-inhibitors showed the highest efficacy (adalimumab 60%, infliximab 100%).

Conclusions

HA20 should be suspected in children with fever and gastrointestinal inflammation, even in the absence of oral/genital ulcers. Early onset colitis, the distinctive morphology of mucosal ulcers and a positive family history for HA20 are particularly suggestive. Early genetic testing may enable prompt diagnosis and targeted therapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• A20 haploinsufficiency (HA20) is a rare autosomal dominant monogenic autoinflammatory disorder with highly variable clinical presentation that may mimic Behçet’s disease, and paediatric data remain limited.

WHAT THIS STUDY ADDS

• A comprehensive analysis of a multicentre French paediatric HA20 cohort, identifying predominant features of recurrent fever, gastrointestinal inflammation, characteristic mucosal ulcers, novel tumour necrosis factor alpha-induced protein 3 (TNFAIP3) variants and marked phenotypic heterogeneity.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Supports early genetic testing for children presenting with unexplained recurrent fever and intestinal symptoms, promoting timelier diagnosis and more personalised treatment strategies.

Introduction

A20 haploinsufficiency (HA20) is a monogenic, early-onset autoinflammatory disease caused by loss-of-function variants in the tumour necrosis factor alpha-induced protein 3 (TNFAIP3) gene, which encodes the A20 protein.^{1 2}

A20 acts as a ubiquitin-editing enzyme that downregulates inflammation by targeting key signalling molecules such as inhibitor of nuclear factor kappa-B kinase subunit gamma (IKKγ) (also known as NF-κB essential modulator (NEMO)) and receptor-interacting serine/threonine-protein kinase 1 (RIPK1), critical mediators in the canonical nuclear factor kappa-B (NF-κB) and other signalling pathways.^{2 3} Because NF-κB is active in both innate and adaptive immunity, its dysregulation leads to a complex phenotype involving both autoinflammatory and autoimmune features.^{4 5}

Initially described in patients misdiagnosed with Behçet’s-like disease due to overlapping manifestations (oral and genital lesions, ocular, articular and cutaneous involvement),^{1 6} the clinical spectrum of HA20 has since broadened significantly. Reported phenotypes now include features mimicking systemic lupus erythematosus (SLE) and autoimmune diseases.7,10

Given its clinical heterogeneity, HA20 diagnosis requires genetic confirmation.¹¹ To date, 90 TNFAIP3 sequence variants are listed in the Infevers database,¹² including 25 pathogenic, 38 likely pathogenic, 19 variants of uncertain significance (VUSs), five likely benign, one benign and two unclassified variants.¹² Most are single-nucleotide variants or small insertions/deletions resulting in premature stop codons and disrupted protein function. Although rare, large or contiguous gene deletions in chromosome 6q have also been reported.¹³ The increasing numbers of VUS identified in patients with HA20 reflect the growing complexity of genotype-phenotype correlations.¹⁴ Variability in the functional impact of TNFAIP3 mutations may underlie the wide clinical heterogeneity observed, NEMO (NF-κB essential modulator) even within the same family, though additional genetic, epigenetic and environmental factors are also likely to contribute.15,17 Consequently, the variable expressivity and incomplete penetrance of this autosomal dominant disorder may obscure the recognition of a single underlying disease mechanism,^{1 15} often resulting in diagnostic delays that can lead to substantial morbidity and, in some cases, mortality.

Although HA20 typically presents in childhood, most reports have focused on adult cases, potentially influenced by recall bias.1518,22 Systematic data on paediatric manifestations of HA20 remain scarce, limiting early recognition and evidence-based management. Furthermore, treatment strategies are largely empirical and lack standardisation.

Objectives

This observational study aimed to improve the understanding of HA20 in paediatric population by characterising its phenotype, genotype, disease course and treatment strategies, with particular focus on childhood presentation.

Methods

Patient recruitment and data collection

Clinical data were collected through a call for cases of patients diagnosed with HA20 through the Francophone Society of Paediatric Rheumatology and Inflammatory Diseases network. In parallel, cases identified through the national reference genetics laboratories for autoinflammatory disorders (Montpellier and Paris, France) were included.

Inclusion and exclusion criteria

Patients with disease onset before the age of 18 years and carrying a TNFAIP3 mutation classified as pathogenic or likely pathogenic were included.

Exclusion criteria comprised confirmed alternative diagnoses with overlapping clinical features, such as paediatric Behçet’s disease, as defined by the Paediatric Behçet’s Disease criteria (≥three of six main clinical features: recurrent oral ulcers, genital ulcers, skin, ocular, neurological or vascular signs)²³; typical inflammatory bowel disease (IBD), as defined by the Paediatric Gastroenterology Hepatology and Nutrition Revised Porto Criteria, based on the combination of endoscopic, histological and imaging findings consistent with either ulcerative colitis (continuous mucosal inflammation starting from the rectum) or Crohn’s disease (discontinuous, transmural inflammation with granulomas and small bowel involvement)²⁴; or primary immunodeficiencies.

Clinical assessment and investigations

Data collected included demographic characteristics, clinical features, laboratory and imaging findings and TNFAIP3 genotyping (NM_006290.4) results. Disease course and response to treatment were also analysed.

Flares were defined as episodes of fever and/or recurrences of mucosal ulcers beyond isolated oral ulcers, with or without additional HA20-associated features. Treatment was considered effective when symptom control was achieved, defined as fewer than three flares per year, resolution of mucosal or joint symptoms and normalisation of inflammatory markers.

Statistical analyses

Descriptive statistics were used to summarise patient characteristics. Categorical variables were compared using Fisher’s exact test. For continuous variables, normality was assessed using the Shapiro-Wilk test and homogeneity of variance using Levene’s test. Normally distributed data were analysed with the unpaired Student’s t-test or Welch’s t-test when variances were unequal, while non-normally distributed variables were compared using the Mann-Whitney U test. Results are expressed as mean±SD for normally distributed variables and as median (IQR) for non-normally distributed variables. All statistical analyses were two-tailed, and a p value<0.05 was considered statistically significant (α=0.05).

Results

Patient characteristics

Data from 17 patients (41.2% female) from 11 unrelated, non-consanguineous families and one consanguineous family were analysed. Most patients were of Caucasian origin (15/17, 88%), with one of Indian (5.9%) and one of Maghrebian (5.9%) descent. A family history of HA20, defined as at least one first-degree or second-degree relative diagnosed with HA20 or Behçet’s Disease-like and/or carrying the same TNFAIP3 variant, was present in 13 patients (76.5%).

The median age at symptom onset was 3 years (range: 1 month–17 years), and 70.5% of patients developed symptoms before the age of 5 years. Boys had a significantly earlier onset than girls (median 1.5 years vs 8 years; p=0.017, figure 1a). The median age at diagnosis was 13 years (range: 2–51 years), with a median diagnostic delay of 9.7 years (range: 1–43 years). Boys were diagnosed significantly earlier than girls (median 9.5 years vs 23 years; p=0.023, figure 1b). Eight patients (47%) experienced disease flares, lasting a median of 7 days and occurring approximately every 30 days.

Figure 1. Age at onset and age at diagnosis by sex. Age at disease onset (a) and age at diagnosis (b) were significantly lower in boys compared with girls (p<0.05). Boxes show median and IQR; diamonds indicate mean values, * p<0.05.

Demographical characteristics of our cohort are summarised in table 1.

Table 1. Demographic characteristics of patients with HA20.

Feature	N (%)	Median (range; IQR)
Patients	17 (100)
Sex
Male	10 (58.8)
Female	7 (41.2)
Origin
Caucasian	15 (88)
Indian	1 (5.9)
Maghrebian	1 (5.9)
Positive family history	13 (76.5)
Age at inclusion (years)		17.6 (5.6–54.7; 11.6)
Age at onset (years)		3.0 (0.1–17; 5.0)
Age at diagnosis (years)		13.5 (2–51; 14.0)

Values are expressed as median (IQR) or N (%).

HA20, A20 haploinsufficiency.

Clinical presentation at disease onset

Oral ulcers (11/17 patients, 64.7%) and fever (11/17, 64.7%) were the most frequently reported symptoms, followed by abdominal pain (8/17, 47.1%), arthralgia (6/17, 35.3%), diarrhoea (4/17, 23.5%), skin rash (3/17, 17.6%) and arthritis (3/17, 17.6%). Less frequent features included recurrent pharyngotonsillitis (1/17, 5.9%) and hepatomegaly (1/17, 5.9%).

To note, genital ulcers were present in only three out of 17 (17.6%) patients at initial presentation, but increased in prevalence with age, reaching 52.9% during follow-up.

Two patients (patients 7 and 9) developed Kawasaki disease (KD) before the onset of HA20-related symptoms. Patient 7 presented with KD at 13 months of age with coronary artery involvement (maximum Z score +7.3, according to McCrindle criteria) and was treated with a single course of intravenous immunoglobulin (2 g/kg) and aspirin for 6 months, with complete resolution of coronary abnormalities. Patient 9 presented with KD at 2 years of age and was treated with intravenous immunoglobulin and corticosteroids, without coronary involvement during follow-up.

No significant difference in symptoms was observed according to gender or age of onset.

Clinical manifestations at disease onset are detailed in table 2 and online supplemental table 1a.

Table 2. Clinical characteristics of patients with HA20 at disease onset and during the disease course.

Feature	At onset (N, %)	During disease course (N, %)
Oral ulcers	11 (64.7)	13 (76.5)
Fever	11 (64.7)	13 (76.5)
Abdominal pain	8 (47.1)	13 (76.5)
Arthralgia	6 (35.3)	9 (52.9)
Diarrhoea	4 (23.5)	7 (41.2)
Genital ulcers	3 (17.6)	9 (52.9)
Skin rash	3 (17.6)	7 (41.2)
Maculopapular	2 (11.8)	3 (17.6)
Pseudo folliculitis	1 (5.9)	2 (11.8)
Purpuric/necrotic	0 (0)	1 (5.9)
Palmoplantar pustulosis/psoriasis	0 (0)	1 (5.9)
Arthritis	3 (17.6)	5 (29.4)
Fatigue	2 (11.8)	7 (41.2)
Lymphadenopathy	2 (11.8)	6 (35.3)
Autoimmune thyroiditis	2 (11.8)	2 (11.8)
Autoimmune hepatitis	2 (11.8)	2 (11.8)
Autoimmune gastritis	1 (5.9)	1 (5.9)
Autoimmune diabetes	1 (5.9)	1 (5.9)
Recurrent pharyngotonsillitis	1 (5.9)	2 (11.8)
Hepatomegaly	1 (5.9)	2 (11.8)
Colitis	0 (0)	6 (35.3)
Headache	0 (0)	3 (17.6)
Pericarditis	0 (0)	1 (5.9)
Uveitis/red eye	0 (0)	1 (5.9)

Values are expressed as N (%).

HA20, A20 haploinsufficiency.

Disease course

During follow-up, the frequency of oral and genital ulcers increased to 76.5% (13/17) and 52.9% (9/17), respectively. Ulcers were painful, single or multiple, long-lasting, isolated or associated with other symptoms. They were typically deeper and clinically distinct from common aphthous ulcers and recurrent aphthous stomatitis (RAS)-type Behçet’s ulcers. Common aphthous ulcers refer to typical minor aphthous stomatitis lesions: small (<10 mm), shallow, round or oval ulcers with a yellowish base and a surrounding erythematous halo, affecting non-keratinised oral mucosa, healing within 7–14 days, and not associated with systemic disease²⁵; RAS-type Behçet’s ulcers refer to minor (<10 mm) or major (>10 mm) lesions with well-defined erythematous margins, homogeneous yellow base, round or oval shape and shallow rather than deep depth²⁶ (figure 2).

Figure 2. Genital ulcer in a patient with HA20. Genital ulcer in a 14-year-old patient with HA20 (patient 6, family 4). The ulcer is deep and irregular, without a well-defined erythematous halo and displays a heterogeneous yellowish base. HA20, A20 haploinsufficiency.

Gastrointestinal involvement was the second most frequent manifestation, ranging from isolated, often recurrent, abdominal pain (13/17 patients, 76.5%) to watery or bloody diarrhoea (7/17, 41.2%). Colitis, often indeterminate, was reported in 6/17 patients (35.3%); endoscopic findings included ulcerated colitis with atypical features, not consistent with typical IBD (patient 5), superficial colonic ulcers (patient 7), right-sided colitis extending to the sigmoid (patient 10), aphthoid colonic ulcers (patient 12 and 15) and non-specific, ‘Crohn-like’ colitis (patient 17). Patient 1 underwent endoscopy due to digestive symptoms and mildly elevated faecal calprotectin, revealing small aphthoid colonic ulcerations with moderate lymphoid nodular hyperplasia and normal histology, not consistent with IBD or colitis. One case of gastritis was recorded (patient 8).

Fever (13/17 patients, 76.5%) and fatigue (7/17, 41.2%) were common during flares. Ocular involvement occurred in one patient (5.9%, patient 12), presenting as anterior non-granulomatous uveitis.

Musculoskeletal involvement was frequent, with arthralgia (9/17, 52.9%) being more common than non-erosive arthritis (5/17, 29.4%). Commonly affected joints included knees, ankles, wrists and proximal interphalangeal joints or metacarpophalangeal joints of hands. Patient 7 developed chronic osteoarthritis of the first toe, ultimately attributed to an inflammatory origin, given the negative infectious work-up (no bacterial growth on blood cultures and bone biopsy) and the absence of response to broad-spectrum antibiotic therapy.

Cutaneous manifestations were observed in seven out of 17 patients (41.2%) and included palmoplantar pustulosis, pseudo folliculitis rash and maculopapular eruptions. One patient (patient 5) developed necrotic lesions on the fingers.

Recurrent pericarditis was reported in one patient (patient 4).

Neurological symptoms were noted in three patients (17.6%), all complaining of headache. Two (patients 7 and 13) underwent lumbar puncture (LP) and brain MRI; both LPs were negative; brain MRI was normal in patient 13 and showed nonspecific hyperintense signals in patient 7 (figure 3). The third patient (patient 11) did not undergo further investigations, as neurological symptoms resolved after treatment.

Figure 3. Brain MRI in a patient with headache. Axial FLAIR brain MRI in a 15-year-old patient (patient 7, family 5) showing non-specific white matter hyperintensities involving the periventricular occipital region and bilateral frontal subcortical areas (white arrows). FLAIR, fluid-attenuated inversion recovery.

Hepatomegaly and lymphadenopathy were reported in 2/17 (11.8%) and 6/17 (35.3%) patients, respectively. Two patients (patient 6 and 8) had been diagnosed with autoimmune hepatitis: patient 8 based on clinical and ultrasound evidence of hepatomegaly, a threefold increase in transaminases and biopsy findings; patient 6 based on clinical and ultrasound hepatomegaly and positive antismooth muscle antibodies (1:400).

Regarding autoimmunity, two patients (patient 1 and 8, 11.8%) had autoimmune thyroiditis (elevated thyroid-stimulating hormone (TSH) and low free thyroxine (FT4), positive antithyroid peroxidase (anti-TPO) antibodies, negative antithyroglobulin and anti-TSH receptor antibodies) and were treated with levothyroxine. One patient (patient 10, 5.9%) had autoimmune diabetes, and two patients (patients 6 and 8, 11.8%) had autoimmune hepatitis, with patient 8 also showing autoimmune gastritis. No patient fulfilled the clinical or biological criteria for connective tissue disease or isolated juvenile idiopathic arthritis.

Clinical characteristics over the disease course are summarised in table 2 and online supplemental table 1a.

Laboratory findings

Acute-phase reactants were markedly elevated during disease flares, with median C reactive protein level of 80.5 mg/dL (IQR 112.08) versus 5.5 mg/dL (IQR 8.65) outside flares (p=0.001, figure 4); erythrocyte sedimentation rate (ESR) data were available for nine patients during inactive phases and for five during active phases; only four had paired values. The mean ESR was 19.7 mm/hour (SD 17.2) outside flares and 50.8 mm/hour (SD 67.6) during flares (p=0.319). White blood cell counts were significantly higher during flares, with median count of 10.4×10⁹/L (IQR 6.1) versus 6.83×10⁹/L (IQR 1.4) outside flares (p=0.039).

Figure 4. CRP levels during flares versus outside flares. CRP levels were markedly higher during flares compared with outside flares periods (p<0.005). Left: distribution of CRP values (boxplot). Right: individual patient trajectories coloured according to percentage change. CRP, C reactive protein, * p<0.05.

Immunological tests revealed positive antinuclear antibodies (ANA) in 7/13 patients (54%), including two with high-positive titres (>1:1000) and five with low-positive titres (<1:320); extractable nuclear antigen antibodies were negative in all 10 patients tested. Rheumatoid factor was positive in 1/6 patients (16.6%). Patient 7 had positive anti-TPO antibodies, without clinical, biochemical or ultrasound evidence of autoimmune thyroiditis. Anti-Saccharomyces cerevisiae antibodies (ASCA) were positive in 1/5 patient (20%). Gastric autoantibodies were detected in one patient with autoimmune gastritis (patient 8). All tested patients (n=8) were negative for antiphospholipid antibodies. HLA-B51 was negative in all five tested individuals. Interferon signature was positive in 5/6 tested patients, with a median score of 9.8 (range: 3.3–38, normal<2). Faecal calprotectin levels were mildly elevated during flares in 3/8 tested patients (37.5%) (mean 60 mg/g, range: 53–375, normal<50).

Outside flares, hypergammaglobulinaemia was observed in 4/11 patients (36.3%), with elevated immunoglobulin G (IgG) levels in 5/10 (50%), elevated IgM in 1/10 (10%) and elevated IgA in 3/11 patients (27%).

Cytokine assays (n=2) showed normal interleukin-1β (IL-1β) and IL-6 levels and slightly elevated IL-18 levels (mean 552 pg/mL, normal<350). Serum amyloid A was elevated in 3/6 tested patients (median 44 mg/L; range: 18–160, normal <6).

Biological characteristics of our cohort are summarised in online supplemental table 1b.

Genetic findings

In 16 out of 17 patients, a next-generation sequencing (NGS) panel for autoinflammatory diseases was performed. Depending on the centre, analysis used either a targeted 8-gene panel (Cochin laboratory, Paris) or an extended 300-gene panel (Montpellier laboratory). One patient underwent an NGS panel for inborn errors of immunity. All patients carried a pathogenic or likely pathogenic TNFAIP3 variant. 13 patients (76.5%) had at least one first-degree or second-degree relative carrying the same TNFAIP3 variant. One patient (patient 17) had an asymptomatic parent carrying the same variant. The remaining three patients were the only family members who underwent genetic testing, as their relatives were asymptomatic. Genetic analysis revealed a predominance of loss-of-function TNFAIP3 variants in the ovarian tumour (OTU) deubiquitinase domain of A20, with 8/17 patients harbouring OTU‐localised mutations classified as pathogenic or likely pathogenic. Recurrent stop‐gain variants included p. Arg87* (3/17 patients, three families), p. Arg183* (3/17 patients, same family) and p. Trp164* (1/17 patient). Two unrelated patients carried the novel pathogenic p. Pro457Alafs*16 variant (patients 9 and 15), present in Infevers database. Two novel mutations, c. 295+2T>G, p.? (patient 6) and c. [29T>A; 49A>T], p. [(Leu10*); p. (Lys17*)] (patient 14), were identified. The mutations were classified as likely pathogenic according to ClinGen and American College of Medical Genetics and Genomics (ACMG) recommendations (pathogenic very strong (PVS1) +pathogenic moderate (PM2)=9 points). Patient 6 experienced recurrent fever, oral and genital ulcers and autoimmune hepatitis, and patient 14 developed mucosal ulcers, arthritis, lymphadenopathy and abdominal pain. Phenotypic heterogeneity, even among relatives carrying identical TNFAIP3 mutations, was observed in our cohort. For instance, in family 2, patient 4 presented with recurrent fever, oral ulcers, arthritis and pericarditis. Her mother displayed oral and genital ulcers, arthralgia, lymphadenopathy and skin eruption (pseudo folliculitis), while her grandfather had oral and genital ulcers and skin eruption (pustulosis/pseudo folliculitis). Additional relatives in the same family (aunt and two nieces of the grandfather) were reported to have recurrent ulcers and chronic hepatitis, and another cousin had ulcers, skin eruption and abdominal pain. In family 8, patient 10 presented with oral and genital ulcers, colitis and autoimmune diabetes. His half-brother had fever, arthralgia, headache and gastrointestinal symptoms without documented colitis, whereas his half-sister exhibited recurrent fever, oral and genital ulcers, skin eruption (maculopapular) and arthralgia. Their mother had fever, oral and genital ulcers, arthritis, skin eruption (maculopapular) and uveitis, and the maternal grandfather displayed fever, ulcers and arthralgia.

Genetic findings are summarised in table 3. Family pedigrees are shown in online supplemental figure 1.

Table 3. Genetic analysis of patients with HA20 (TNFAIP3 variants, NM_006290.4).

Family ((patient(s))	Nucleotide change	Protein change	Classification	Domain	Reference(s)
Family 1
Patient 1–3	c. 547C>T	p. (Arg183*)	Pathogenic	OTU	1 12 17
Family 2
Patient 4	c. 259C>T	p. (Arg87*)	Pathogenic	OTU	12 19 30
Family 3
Patient 5	c. 492G>A	p. (Trp164*)	Pathogenic	OTU	12 21
Family 4
Patient 6	c. 295+2T>G	p.?	Likely pathogenic	OTU	Not published
Family 5
Patient 7	c. 259C>T	p. (Arg87*)	Pathogenic	OTU	12 19 30
Family 6
Patient 8	c. 259C>T	p. (Arg87*)	Pathogenic	OTU	12 19 30
Family 7
Patient 9	c. 1368dup	p. (Pro457Alafs*16)	Pathogenic	Disordered (Interaction with TNIP1)	12
Family 8
Patient 10–13	c. 971_975del	p. (Leu324Glnfs*7)	Pathogenic	OTU	1 12
Family 9
Patient 14	c. 29T>A; 49A>T	p. (Leu10*) p. (Lys17*)	Likely pathogenic	OTU	Not published
Family 10
Patient 15	c. 1368dup	p. (Pro457Alafs*16)	Pathogenic	Disordered (Interaction with TNIP1)	12
Family 11
Patient 16	c. 707T>C	p. (Leu236Pro)	Likely pathogenic	OTU	12 44
Family 12
Patient 17	c. 821delC	p. (Pro274Hisfs*13)	Likely pathogenic	OTU	12 46

HA20, A20 haploinsufficiency; OTU, ovarian tumour; TNIP, TNFAIP3-interacting protein.

Imaging findings

Brain MRI was abnormal in one patient (patient 7), presenting with headache, showing non-specific white matter hyperintensities (figure 3).

Cardiovascular imaging, including echocardiography, chest X-ray and echocardiograph of supra-aortic trunks, identified pericarditis in one patient (patient 4).

Abdominal ultrasound revealed hepatomegaly in 3/15 patients (20%), two of whom had clinically detectable hepatomegaly.

Treatment and outcomes

All patients required therapeutic intervention. Colchicine was prescribed in 15/17 patients (88.2%), and was effective as monotherapy in six (40%), though often continued as adjunctive therapy.

Additional treatments included monotherapy or combination regimens with conventional and biologic disease-modifying anti-rheumatic drugs, Janus kinase (JAK) inhibitors or apremilast. Systemic corticosteroids and non-steroidal anti-inflammatory drugs (NSAIDs) were used during flares in 7/17 (41.2%) and 3/16 (18.8%), respectively.

Treatment was considered effective when symptom control was achieved, defined as fewer than three flares per year, resolution of mucosal or joint symptoms and normalisation of inflammatory markers.

Among anti-TNF agents, etanercept was used in 4/17 patients (23.5%) with clinical response in one (25%), adalimumab was administered to 5/17 patients (29.4%), with response in three (60%) and infliximab in 3/17 (17.6%), effective in all cases (100%). Regarding IL-1 inhibitors, 5/17 patients (29.4%) received anakinra, achieving remission in 2/5 and partial improvement in 2/5; patient 4 received canakinumab, without clinical improvement. As for JAK inhibitors, one patient (patient 7), refractory to colchicine, methotrexate and adalimumab, received baricitinib (6 mg/day), with partial clinical improvement and subsequent disease control on infliximab. Two patients (patients 11 and 15) refractory to previous therapies achieved remission with apremilast monotherapy and another (patient 12) with apremilast in combination with colchicine. Methotrexate relieved joint symptoms in 2/4 patients (50%).

One patient (patient 8) presenting with a polyautoimmune profile (autoimmune hepatitis, autoimmune gastritis and autoimmune thyroiditis) was successfully treated with azathioprine.

Therapeutic strategies and responses are summarised in table 4, online supplemental table 1c and online supplemental figure 2.

Table 4. Treatments used in patients with HA20 and their effectiveness.

Treatment	Patients treated	Effectiveness
	N (%)	N (%)
Colchicine (n=17)	15 (88.2)	6 (40)
0.5 mg/day	2 (13.3)
1 mg/day	11 (73.4)
2 mg/day	2 (13.3)
NSAID during flares (n=16)	3 (18.8)	NA
Oral steroid during flares (n=17)	7 (41.2)	7 (100)
Anti IL1 (n=17)
Anakinra (80–200 mg/d)	5 (29.4)	2 (40)
Canakinumab (150 mg/4 weeks)	1 (5.9)	0 (0)
JAK inhibitors (n=17)
Baricitinib 6 mg/day	1 (5.9)	0 (0)
TNF inhibitors (n=17)
Etanercept (0.8 mg/kg/week)	4 (23.5)	1 (25)
Infliximab 5 mg/kg/6 weeks	3 (17.6)	3 (100)
Adalimumab 20–40 mg/week	5 (29.4)	3 (60)
Methotrexate (n=17) (10–17.5 mg/week)	4 (23.5)	2 (50)
Apremilast (n=17)	3 (17.6)	3 (100)
Azathioprine (n=17)	1 (5.9)	1 (100)

Treatment was considered effective when symptom control was achieved—defined as fewer than three flares per year, resolution of mucosal or joint symptoms and normalisation of inflammatory markers.

HA20, A20 haploinsufficiency; IL, interleukin; JAK, Janus kinase; NA, not available; NSAID, non-steroidal anti-inflammatory drug; TNF, tumor necrosis factor.

Discussion

To our knowledge, this is the first study characterising a French cohort of patients with childhood-onset HA20, with a specific focus on clinical presentation at disease onset. Our findings provide new insights into the phenotypic and genetic spectrum and highlight key differences in clinical features compared with previously reported HA20 cohorts and overlapping conditions, such as Behçet’s disease, periodic fever, aphthous stomatitis, pharyngitis and adenitis (PFAPA) syndrome and IBD.

In our cohort, the median age at onset (3 years, range 1 month–17 years) was consistent with previous reports (median 4–4.5 years, range 3 months–17 years).^{7 27} Notably, none of these studies identified sex-related differences in age at onset or diagnosis. In contrast, we found that boys presented and were diagnosed at a significantly younger median age than girls, possibly reflecting a more overt and painful nature of genital ulcers in boys. In addition, boys have been reported to exhibit higher innate pro‐inflammatory cytokine production (eg, IL-1β, TNF-α) than girls in early life,²⁸ which might contribute to earlier clinical expression of HA20 in males, although this hypothesis requires further exploration.

Fever occurred in 76.5% of our patients. This aligns with East Asian cohorts (67–80%),^{27 29 30} but contrasts with lower frequencies in American and European series (~50%).^{7 15} The higher prevalence in our study may reflect the paediatric focus, in which autoinflammatory features are often more pronounced. Moreover, in adults, fever may be underreported or misattributed to common childhood infections during the early course of their disease.

Our findings also challenge previous assumptions about hallmark features of HA20. Contrary to reports suggesting the predominant presence of oral and genital ulcers (73–100%),^{7 15} these symptoms were not constant in our cohort, being present in only 64% and 17% of patients at onset and in 76% and 53% over the disease course, respectively. These findings suggest that mucosal ulcers should not be considered mandatory to suspect HA20 in children, especially at disease onset. Notably, there is limited literature describing the morphology of these ulcers and their differential diagnosis with common aphthous ulcers and RAS-type Behçet’s ulcers. In our cohort, ulcers were typically deeper, often irregular in shape (not necessarily round or oval), lacking a clear erythematous halo, were long lasting, sometimes >10 mm in diameter and frequently healed with residual scarring. These features contrast with the classical Behçet’s disease (BD)-related ulcers, which, according to the international Delphi consultation by Poveda-Gallego et al,²⁶ are generally superficial, with a homogeneous yellowish base, a well-defined erythematous margin and a regular round or oval shape. To note, in BD, the distinction from RAS is often challenging: about 76% of experts in that study agreed that BD and RAS ulcers could not be reliably differentiated based on clinical images alone, emphasising that diagnosis of BD should rely on the broader multisystemic context and established BD criteria. Furthermore, recent genetic studies have identified shared susceptibility loci among BD, RAS and PFAPA—including IL10, IL12A, signal transducer and activator of transcription (STAT4), C-C chemokine receptor (CCR1–CCR3) and IL23R–IL12RB2—supporting a ‘Behçet’s spectrum disorder’,³¹ which may also explain the morphological resemblance of oral lesions across these conditions. Conversely, to our knowledge, no genome-wide or large-cohort study has explicitly demonstrated shared common variant loci between HA20 and these disorders.

HA20 ulcers also differ from the lesions associated with IBD, which include specific lesions, such as indurated tag-like folds (firm, white reticular mucosal folds), cobble stoning (fissured, swollen, corrugated mucosa), mucogingivitis (oedematous, granular gingiva), lip swelling with fissures or deep linear ulcerations and non-specific oral lesions, such as aphthous stomatitis, pyostomatitis vegetans (multiple pustules on an erythematous mucosa), angular cheilitis and glossitis.³² Limited awareness of typical HA20 ulcer morphology may bias clinical description and patient selection and may delay diagnosis, thereby prolonging morbidity.¹⁶

Although gastrointestinal involvement is recognised in HA20,^{7 33} our findings suggest it may occur more frequently, affecting over 75% of patients, than previously reported (46–56%).^{5 7 15} This is consistent with recent data from Shiraky et al (75.9%),²⁷ underscoring its diagnostic relevance. Intestinal inflammation, manifesting as abdominal pain, diarrhoea and/or colitis, may represent a key diagnostic clue for HA20. As previously described,34,38 HA20-associated colitis often mimics early-onset IBD, possibly reflecting shared involvement of the NF-κB pathway, which plays a central role in some monogenic forms of IBD.³⁹ This mechanistic overlap might contribute to the high frequency of mucosal and intestinal involvement observed in HA20, although the underlying pathogenic mechanisms remain to be fully elucidated. Several features may help differentiate HA20 from typical IBD: endoscopic findings are often inconclusive for typical IBD, tending to show atypical Crohn-like or ulcerative colitis-like patterns; oral ulcers in HA20 differ from IBD-related oral lesions; genital ulcerations may coexist and a strong familial history of HA20 is frequent. While clinical overlap may lead to initial misclassification as IBD, these distinctive features should raise suspicion for HA20 and prompt genetic testing.

Intestinal manifestations may also help distinguish HA20 from Behçet’s disease,^{6 23} where gastrointestinal involvement is less common. Beyond ulcer morphology and intestinal manifestations, additional differentiators include the earlier age of onset in HA20 (median 3 years vs 11 years in BD), the lower prevalence of vascular and ocular involvement, the distinct aspect of cutaneous manifestations (maculopapular or pseudo folliculitis in HA20 vs necrotic folliculitis or erythema nodosum in BD) and the type of neurological involvement, as headache, common in HA20, is not part of the classification criteria for paediatric BD.^{23 40}

Distinguishing HA20 from PFAPA syndrome can be challenging, as both may present with recurrent fever, oral lesions and pharyngitis. Some patients with HA20 are initially diagnosed as PFAPA¹⁴; in our cohort, patient 1 and patient 16 were first labelled PFAPA, reflecting this clinical overlap. However, several clinical features may help distinguish these two conditions. Features favouring HA20 include inconstant and not strictly periodic, non-stereotyped fever, oral ulcers that differ from common aphthous stomatitis of PFAPA, genital or gastrointestinal ulcerations and frequent systemic involvement (arthritis, skin rash, uveitis). In contrast, PFAPA typically shows strictly periodic febrile episodes with a stereotyped pattern and lacks systemic autoimmune/inflammatory organ involvement.⁴¹ Systematic genetic screening for TNFAIP3 variants in all PFAPA patients is not currently recommended; genetic testing should be considered in atypical cases or where the family history is suggestive of HA20.

A comparative summary of the main differential diagnostic features distinguishing HA20, Behçet’s disease, IBD and PFAPA syndrome is provided in online supplemental table 2.

As described by several authors,42,46 loss of A20 function drives both excessive inflammation and selective loss or dysfunction of specific B-cell or T-cell subsets, contributing to autoimmune phenotypes resembling systemic lupus.^{9 10} In our cohort, two patient developed autoimmune thyroiditis, two developed autoimmune hepatitis and one developed autoimmune diabetes; one patient presented with both autoimmune hepatitis and gastritis. No cases of SLE or isolated juvenile idiopathic arthritis were observed, although ANA were positive in 54% and RF in 16%. Overall, the autoinflammatory phenotype predominated over the autoimmune features, likely reflecting the paediatric focus of our study. Autoimmune manifestations may emerge later in the disease course.

Consistent with previous reports,^{5 7 15} a positive family history of HA20 was present in most cases (76.5%), underscoring its importance as a diagnostic clue that may guide genetic testing in suspected patients. One patient had an asymptomatic parent carrying the same TNFAIP3 variant, suggesting incomplete penetrance. In three additional cases, only the affected patient underwent genetic testing, as relatives were asymptomatic. Therefore, it was not possible to determine whether these variants occurred de novo or were inherited from asymptomatic carriers, precluding definitive conclusions about penetrance. This finding highlights the importance of systematic family screening and clinical awareness of subclinical or oligo-symptomatic presentations in HA20. Our data also show the phenotypic heterogeneity of HA20, even among relatives carrying identical TNFAIP3 mutations.

Chen et al¹⁴ proposed that mutations in the OTU domain are associated with more severe phenotypes, whereas musculoskeletal symptoms may predominate in patients with ZnF domain mutations. Most of our patients (58 %) harboured OTU nonsense/frameshift mutations, and only one missense variant was identified, which precludes drawing any firm conclusions regarding genotype-phenotype correlations.

Notably, in our cohort, three novel mutations were identified. The novel p. Pro457Alafs*16 mutation was identified in two unrelated patients, suggesting a possible mutational hotspot. This variant, located in the C-terminal region of A20, is predicted to impair its regulatory function in NF-κB signalling, consistent with its classification as pathogenic on Infevers.^{12 47} Two other novel, likely pathogenic TNFAIP3 mutations (c. 295+2T>G and complex allele c. 29T>A/49A>T) were identified. The clinical phenotype of these patients was similar to that of other individuals carrying pathogenic or likely pathogenic TNFAIP3 variants. In addition, one of the patients in our cohort (patient 16) corresponds to a case already reported in the literature, as part of the family described by El Khouri et al⁴⁴; the remaining patients represent newly characterised cases.

Currently, no standardised treatment guidelines exist for HA20, and management remains largely individualised according to the predominant clinical features. Several studies have described commonly used therapies, though efficacy data are limited. In our cohort, colchicine, effective across Behçet’s disease spectrum disorders,⁴⁸ was effective as monotherapy in 40% of patients, mainly for mucosal ulcerations, consistent with previous reports (35–55%).^{7 14 27} Methotrexate provided symptomatic relief of joint manifestations in 50% of cases, a higher rate than reported by Elhani and Shiraki (25%).^{7 27} Among second-line therapies, TNF inhibitors, particularly infliximab, appeared to offer the greatest clinical benefit, with response rates ranging from 25% to 100% in our cohort, especially in colchicine-refractory ulcers or combined colitis and arthritis, in line with prior series (59–77%).^{15 27 33} Previous reports^{15 16} have described variable efficacy of IL-1 blockade with anakinra in HA20, from partial to complete remission. In our cohort, five patients received anakinra, achieving complete remission in 40% of cases. Canakinumab was administered to one patient, without benefit, consistent with Shiraki et al.²⁷ In contrast with earlier reports describing the effectiveness of JAK inhibitors,^{49 50} baricitinib, given to one patient, produced only partial clinical improvement despite a mildly positive interferon signature. Notably, three refractory patients showed good clinical response to apremilast for mucosal ulcerations; to our best knowledge, no previous cases reporting the efficacy of apremilast in HA20 have been published. Overall, these findings highlight the therapeutic heterogeneity of HA20 and the need for individualised strategies tailored to disease manifestations and response.

This study has several limitations. Its retrospective design, the small sample size and the predominance of nonsense/frameshift mutations limit exploration of genotype-phenotype correlations, disease modifiers and the full cytokine profile. Recall bias may have affected symptom reporting, such as fever, particularly in adult patients. Furthermore, treatment regimens varied substantially across patients, reflecting both disease heterogeneity and evolving therapeutic approaches over time, limiting robust comparisons with previous studies.2733 48,50

Conclusion

HA20 should be considered in the differential diagnosis of recurrent fever and gastrointestinal symptoms in children, even in the absence of oral and genital ulcers. Early-onset colitis, the distinctive morphology of mucosal ulcers and a positive family history for HA20 are particularly suggestive of HA20. Larger, multicentre studies and functional analyses of novel TNFAIP3 missense mutations are needed to better elucidate the genotype-phenotype correlations, improve diagnostic accuracy and guide therapeutic management.

Data availability statement

Data are available upon reasonable request.