Mutations in proteasome subunit β type 8 cause chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature with evidence of genetic and phenotypic heterogeneity




Chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE syndrome) is an autoinflammatory syndrome recently described in children. We undertook this study to investigate the clinical phenotype, genetic cause, and immune dysregulation in 9 CANDLE syndrome patients.


Genomic DNA from all patients was screened for mutations in PSMB8 (proteasome subunit β type 8). Cytokine levels were measured in sera from 3 patients. Skin biopsy samples were evaluated by immunohistochemistry, and blood microarray profile and STAT-1 phosphorylation were assessed in 4 patients and 3 patients, respectively.


One patient was homozygous for a novel nonsense mutation in PSMB8 (c.405C>A), suggesting a protein truncation; 4 patients were homozygous and 2 were heterozygous for a previously reported missense mutation (c.224C>T); and 1 patient showed no mutation. None of these sequence changes was observed in chromosomes from 750 healthy controls. Of the 4 patients with the same mutation, only 2 shared the same haplotype, indicating a mutational hot spot. PSMB8 mutation–positive and –negative patients expressed high levels of interferon-γ (IFNγ)–inducible protein 10. Levels of monocyte chemotactic protein 1, interleukin-6 (IL-6), and IL-1 receptor antagonist were moderately elevated. Microarray profiles and monocyte STAT-1 activation suggested a unique IFN signaling signature, unlike in other autoinflammatory disorders.


CANDLE syndrome is caused by mutations in PSMB8, a gene recently reported to cause “JMP” syndrome (joint contractures, muscle atrophy, microcytic anemia, and panniculitis-induced childhood-onset lipodystrophy) in adults. We extend the clinical and pathogenic description of this novel autoinflammatory syndrome, thereby expanding the clinical and genetic disease spectrum of PSMB8-associated disorders. IFN may be a key mediator of the inflammatory response and may present a therapeutic target.

Autoinflammatory diseases were first characterized more than 10 years ago by episodic, systemic, and organ-specific inflammation (1, 2). These disorders differ from autoimmune diseases in that they primarily result from perturbations in the innate immune system, rather than in adaptive immunity, although overlapping features may occur (2, 3). During the past decade, the genetic basis for many autoinflammatory diseases has been revealed (1). Elucidating the underlying molecular basis for these monogenic disorders has increased the understanding of inflammation and has led to improved therapy, particularly interleukin-1 (IL-1) inhibition for patients with cryopyrin-associated periodic syndromes (4).

Chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE syndrome) is a newly described autoinflammatory condition, which had recently been reported in 5 patients (5, 6). It is characterized by onset during the first year of life, recurrent fevers, purpuric skin lesions, violaceous swollen eyelids, arthralgias, progressive lipodystrophy, hypochromic or normocytic anemia, delayed physical development, and increased levels of acute-phase reactants. Variable clinical features include hypertrichosis, acanthosis nigricans, and alopecia areata (5, 6). The skin biopsy findings of a characteristic atypical, mixed mononuclear and neutrophilic infiltrate further confirm the diagnosis (5). A genetic etiology was suggested, possibly inherited in an autosomal-recessive pattern. In order to elucidate the molecular basis of CANDLE syndrome, we performed genome-wide analysis and sequencing in 8 families with 9 affected patients.



The present study included 9 patients (3 Spanish, 3 Hispanic, 1 Ashkenazi Jew, and 2 Caucasian Americans) from 8 families who were seen at 5 international centers (2 in Spain, 1 in Israel, and 2 in the US). The study was approved by the institutional review boards at the respective sites, and written informed consent was obtained from the parents. Blood samples were obtained from 8 patients and, where available, from their unaffected family members. All of the patients included in this study were discussed among the lead investigators of the 5 centers to ensure the diagnosis of CANDLE syndrome before they were included. All patients had to have episodic fevers, typical erythematous eruptions, arthralgia/arthritis, and evidence of systemic inflammation (elevation of acute-phase reactant levels). In addition, a skin biopsy sample showing an atypical mixed myeloid, neutrophilic, and histiocytic infiltrate positive for myeloperoxidase and CD68 had to be present. Patients 1–4 were previously described by Torrelo et al (5), and patient 5 was reported by Ramot et al (6); patient 3 died at age 14 years, and blood samples were not available for analysis (5). Patients 6–9 have not been described before.

Genetic analysis.

A genome-wide analysis of single-nucleotide polymorphisms (SNPs) was performed using the GeneChip Human Mapping 250K SNP Array of Affymetrix. For this analysis we used blood samples from patients 1, 2, 4, 5, and 9. Genome-wide homozygosity analysis was performed with HomozygosityMapper (7).

In parallel with the conventional Sanger sequencing, we also performed whole-exome sequencing in patient 5. DNA captured with the Agilent SureSelect Human All Exon 50Mb kit was sequenced on an Illumina HiSeq 2000 platform. Sequence data were analyzed by Beijing Berry Genomic Co., Ltd. using a custom bioinformatics pipeline.

To identify additional mutations in PSMB8 (proteasome subunit β type 8) in the 2 patients who were heterozygous for the PSMB8 mutation T75M and in the patient who did not have any mutation in PSMB8 on exonic sequencing, we sequenced the other 5 exons and all of the introns of the longest isoform of PSMB8 (Ensemble nomenclature PSMB8.001). Alternative and cryptic splicing events were ruled out by sequencing the complementary DNA (cDNA). We searched for genomic deletions by long-range polymerase chain reaction (PCR) with primers spanning the entire gene and ruled out the possibility of a heterozygous deletion at the primer binding site by sequencing the c.224C>T mutation using the long-range PCR amplicon as template. In patients 6 and 7 both the maternal and paternal copies were present at the PSMB8 locus, and the PSMB8 cDNA was of full length (8). Assuming digenic inheritance in patient 7, we sequenced all the other β subunits from PSMB1 to PSMB10 and 2 α subunits (PSMA6 and PSMA7). PSMB8 messenger RNA levels were similar to those in healthy controls by quantitative reverse transcription–PCR from cDNA from peripheral blood.

In silico modeling.

A structural model of both mutations identified was assembled. Structures were based on a previously reported model (9). Figures were made with Pymol (Schrödinger).

Cytokine analysis.

Serum was collected from patients 6, 7, and 8 and stored at −80°C. Cytokine concentrations were measured using the Bio-Plex system (Bio-Rad) in batches including patient control and healthy serum.

Microarray analysis.

Total RNA was extracted from blood samples collected in PAXgene tubes (from patients 4, 6, 7, and 8) and processed as recommended by the manufacturer (Qiagen). RNA integrity was analyzed with an Agilent 2100 Bioanalyzer. Complementary DNA synthesis and target amplification was done with the NuGEN Ovation Whole Blood Solution kit. Affymetrix HU-133 Plus 2.0 gene chips were used for hybridization. Data analysis was done with GeneSpring 11.5 software and Partek software after removal of nonannotated genes. Genes considered differentially expressed in comparison to the mean expression in healthy controls had at least 2-fold higher expression (fold change >2), with a P value less than 0.05 adjusted for multiple hypothesis testing and controlling for the false discovery rate (FDR) (FDR <0.05). Groups were compared using Welch's 2-tailed t-test. Differentially expressed genes were then analyzed using Ingenuity Pathway Analysis ( to investigate dysregulated canonical pathways and gene ontology. Results for interferon (IFN)–induced genes were plotted as a heatmap with up-regulated genes in shades of red and down-regulated genes in shades of blue.

Cell stimulation and STAT-1 phosphorylation assay.

Peripheral blood mononuclear cells (PBMCs) were isolated by standard Ficoll density-gradient centrifugation and frozen in liquid nitrogen. Cells were thawed, washed, and resuspended in 0.1% bovine serum albumin/phosphate buffered saline (PBS) at 2 × 106/ml and then aliquoted at 0.5 ml per tube. For studies with the small molecule JAK kinase inhibitor tofacitinib, [3-[(3R,4R)-4-methyl-3-[methyl(7H-pyrrolo[2,3-d] pyrimidin-4-yl)amino]piperidin-1-yl]-3-oxopropanenitrile] (synthesized by Dr. Craig Thomas, National Institutes of Health Chemical Genomics Center), cells were treated with the inhibitor for 15 minutes before stimulation. Cells were stimulated with various concentrations of IFNγ or control PBS buffer at 37°C for 15 minutes, fixed with 4% paraformaldehyde, and then stained with phycoerythrin-conjugated anti-CD14 and Alexa Fluor 647–conjugated pSTAT-1 (BD PharMingen) and analyzed using a BD FACSCanto instrument according to standard procedures. Data were analyzed with FlowJo software (Tree Star).

Immunohistochemistry and special stains.

Punch biopsy samples of lesional skin were fixed in 10% neutral buffered formalin and processed routinely. Serial tissue sections of 5 μm thickness were made and spread on poly-L-lysine–coated glass slides. Immunohistochemical staining was carried out using the Ventana Benchmark XT fully automated slide preparation system (Ventana Medical Systems) and the following antibodies: antimyeloperoxidase (1:1,000 dilution; Dako), anti-CD163 (1:100 dilution; Novocastra), and anti-CD68 (KP1) (1:400 dilution; Dako). Staining was developed with 3,3′-diaminobenzidine, and slides were counterstained with Mayer's hematoxylin and mounted. Leder staining (naphthol-AS-D chloroacetate esterase or specific esterase), which identifies cells of the granulocyte lineage from the early promyelocyte stage to mature neutrophils, was carried out using the NAPHL AS-D chloroacetate kit (no. 91C-1KT) according to the protocol of the manufacturer (Sigma-Aldrich).


Clinical characteristics of CANDLE syndrome patients.

Table 1 summarizes the demographic characteristics and clinical presentation of the 9 affected children. Our series included 5 previously reported patients (patients 1–5) (5, 6) and 4 previously unreported patients from nonconsanguineous families: 2 Caucasian males and 1 Hispanic female from the US (patients 6–8) and 1 Caucasian male from Spain (patient 9) (Figures 1A–G). Most patients presented within the first 2–4 weeks of life (all by age 6 months) with fever and repeated attacks of erythematous and violaceous, annular cutaneous plaques, lasting for a few days or weeks and leaving residual purpura. Later during infancy, patients developed persistent periorbital erythema and edema, finger or toe swelling, and hepatomegaly with variable elevation of acute-phase reactant levels. Other common clinical features that developed in the first years of life included progressive loss of peripheral fat (lipodystrophy), failure to thrive, lymphadenopathy, and hypochromic or normocytic anemia. More variable findings included perioral swelling, parotitis, conjunctivitis/episcleritis, acanthosis nigricans and hypertrichosis, chondritis, aseptic meningitis, interstitial lung disease, nephritis, epididymitis, hypertriglyceridemia, and intermittent positivity for antinuclear antibodies or antineutrophil cytoplasmic antibodies (Table 1 and Figure 1). Magnetic resonance imaging of the thigh showed irregular enhancement of fat, suggesting panniculitis, but no myositis (Figures 1H and I). Synovial enhancement, consistent with joint discomfort and arthralgia and/or arthritis, was also seen (Figure 1J).

Table 1. Patient demographic characteristics, clinical disease manifestations, and treatments*
 Patient 1Patient 2Patient 3Patient 4Patient 5Patient 6Patient 7Patient 8Patient 9
  • *

    ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; LFT = liver function test; ND = not done; BOOP = bronchiolitis obliterans with organizing pneumonia; NA = not applicable; TSH = thyroid-stimulating hormone; LDL = low-density lipoprotein; HDL = high-density lipoprotein; LAC = lupus anticoagulant; cANCA = classic antineutrophil cytoplasmic antibody; ANA = antinuclear antibody; NSAIDs = nonsteroidal antiinflammatory drugs; PUVA = psoralen ultraviolet A; IVIG = intravenous immunoglobulin.

  • Patient previously reported.

  • Information on drug doses and treatment durations could only be retrieved for patients 6–8, and dose ranges and maximal exposure times were as follows: methotrexate, >4 years; tacrolimus, 4 mg/day for >12 months; infliximab, 5–12.5 mg/kg/dose every 4 weeks up to 12 months; adalimumab, 20–40 mg every 10 days for 4 months, then discontinued; anakinra, 1–5 mg/kd/day for 3 months; tocilizumab, 5–12 mg/kg/dose every 14 days, 2 patients are still receiving treatment at >6 months, 1 developed a drug reaction warranting discontinuation. Patients 6 and 7 are receiving prednisolone and tacrolimus in addition to 1 biologic agent; patient 7 is receiving prednisone in combination with tocilizumab. Prednisone is tapered to control symptoms of fever, rash, and joint pain with doses of 0.3–3 mg/kg, and pulses have been administered in between doses to control symptoms.

Demographic characteristics         
 Age, years1310Deceased212.55.53.562.5
 Ethnicity/originSpanishSpanishHispanicHispanicAshkenazi JewishAmerican CaucasianAmerican CaucasianHispanicSpanish
 Clinical outcomeAlive and failure to thriveAlive and failure to thriveDeceased at age 14 yearsAlive and failure to thriveAlive and failure to thriveAlive and failure to thriveAliveAliveAlive
Common clinical disease manifestations         
 Age at clinical presentation1 month6 months<6 months1 week1 month2 weeks2 months2 weeks1 month
 Symptoms at initial presentationFever and skin lesionsFever and skin lesionsFever, violaceous plaques and nodules, and hepatomegalyViolaceous annular plaquesFever and skin lesionsRash and foot swelling, periorbital erythemaPeriorbital erythema and swellingFever, violaceous plaques, and periorbital erythemaFever and skin lesions
 Recurrent fevers and elevated acute-phase reactant levels (ESR, CRP)+++++++++
 Annular plaques+++++++++
 LipodystrophyYes/peripheral, especially face and upper limbsYes/upper limbs and cheeksYes/peripheral and faceYesYesYes/upper and lower bodyYes/upper body, mainly proximalYes/upper bodyYes
 Finger swelling+++++++
 Hepatosplenomegaly++ (only hepatomegaly)+ (only hepatomegaly)+ (only hepatomegaly)+ (only hepatomegaly)++Upper limit of normal
 Low weight and height+++++++
 Muscle atrophy++++
 Prominent abdomen++/−+++++
 Violaceous eyelids+++++++++
 Anemia and other hematologic manifestationsHypochromic anemia and increased platelet countsHypochromic anemiaHypochromic anemiaHypochromic anemia and increased platelet countsHypochromic anemiaHypochromic anemiaNeutropenia and thrombocytopeniaHypochromic anemia in the pastHypochromic anemia
 Elevated LFT values+++++Intermittent++
More variable findings         
 Perioral swelling+++++
 Acanthosis nigricans+++
 Ear and nose chondritis+Slight nose chondritisPossible+
 Wide-spaced nipples++++
 Aseptic meningitis++
 Basal ganglia calcifications++NDNDND
 Lung manifestationsInterstitial lung diseaseBOOP-like
 Recurrent infectionsOtitisOtitisOtitis and recurrent sinusitisOtitis and recurrent sinusitisRecurrent upper respiratory infections+
 Metabolic manifestationsHypertriglyceridemiaHypertriglyceridemiaElevated TSH, triglycerides NDMild hypertriglyceridemiaElevated TSH, high LDL and triglyceridesLow HDL
 Autoantibodies (intermittent)Coombs-positive hemolytic anemia, LAC positiveAlopecia areata, cANCA positiveANCA 1:20, now negativeANA positive in pastHypergammaglobulinemia
TreatmentsNSAIDs, colchicine, dapsone, methotrexate, etanercept, anakinraNSAIDs, PUVA, methotrexateNSAIDs, methotrexate, azathioprine, infliximab, cyclosporine, rituximab, pulsed steroidsIVIG, anakinra, dapsone, ibuprofen, pulsed steroidsHydroxychloroquine, pulsed steroidsTacrolimus, methotrexate, prednisolone, infliximab, methylprednisolone, adalimumab, anakinra, tocilizumabTacrolimus, methotrexate, prednisolone, tocilizumab, methylprednisolonePrednisone, methylprednisolone, anakinra, adalimumab, tocilizumabNone
Figure 1.

Clinical features of chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE syndrome). AD, Facial features and eruption in patients with CANDLE syndrome (patients 6 [A], 7 [B], 8 [C], and 9 [D]). Note the violaceous periorbital discoloration and edema, perinasal erythema, and significant fat loss most prominent in B. E and F, Finger swelling and violaceous lesions on the heel and toes (patients 9 [E] and 6 [F]). G, Discrete erythematous nodules and postinflammatory hyperpigmentation (patient 6). H and I, T2-weighted magnetic resonance image of the thighs fo patient 6 (H), suggesting loss of subcutaneous fat, particularly on the posterior aspect of both thighs, and STIR image from the same patient (I), showing increased signal intensity in the subcutaneous fat, suggesting panniculitis (asterisks). J, Synovial enhancement with rice body formation extending to the suprapatellar pouch (patient 6).

Responses to treatment were variable (Table 1). Most clinical symptoms, including cutaneous eruption, joint pain, and fever, responded to high doses of steroids (1–2 mg/kg/day), but rebounded with tapering (∼0.5 mg/kg/day). Responses to steroid-sparing agents were inconsistent. Methotrexate in combination with calcineurin inhibitors permitted administration of lower doses of steroids; however, skin and joint flares with fever in between led to the further addition of biologic agents. Tumor necrosis factor α (TNFα) inhibitors provided temporary improvement in some cases but gave rise to flares in others. Anti–IL-1 therapy did not allow a decrease in steroids, and IL-6–blocking agents normalized acute-phase reactant levels and anemia but had limited success in reducing the cutaneous eruption and improving fatigue (see Table 1 for dose ranges of biologic agents). Lipodystrophy invariably progressed despite immunosuppressive therapy.

Histologic evaluation identifies a dense dermal infiltrate of immature neutrophils and activated macrophages.

Hematoxylin and eosin–stained skin biopsy samples from all patients were assessed. Characteristic features included a dense interstitial infiltrate of mononuclear cells with nuclear atypia and both mature and immature neutrophils, with areas of karyorrhexis (Figures 2A and B). Interstitial dermal collagen degeneration was seen. Immunohistochemical staining disclosed a mononuclear infiltrate composed of immature neutrophils/myeloid precursors (Leder stain positive [Figure 2C] and myeloperoxidase positive [results on myeloperoxidase staining not shown for neutrophils]), as well as atypical mononuclear cells that were most likely activated macrophages (positive for CD68 and CD163, negative for Leder stain) (Figures 2D–F).

Figure 2.

Histopathology of lesional skin. A, Hematoxylin and eosin (H&E) staining shows a dense perivascular and interstitial mononuclear dermal infiltrate. The overlying epidermis appears unremarkable. B, Areas of karyorrhexis are shown (H&E staining). C, Leder staining (naphthol-AS-D chloroacetate esterase or specific esterase) specifically identifies cells of the granulocyte lineage from the early promyelocyte stage to mature neutrophils. 1, Boxed area at higher magnification shows that some of the mononuclear cells are Leder stain positive, suggestive of cells of granulocyte lineage that are immature neutrophils (mononucleated cells; arrow). Rare mature neutrophils (segmented nucleus) are seen throughout. 2, Boxed area at higher magnification shows karyorrhectic nuclear debris surrounded by a rim of Leder stain–positive cytoplasm (arrow). DF, Shown is immunohistochemical staining for the macrophage markers CD68 (D) and CD163 (E) and for myeloperoxidase-positive cells (F).

Genetic analyses reveal that mutations in PSMB8 cause CANDLE syndrome.

A region of homozygosity shared by 4 patients (patients 1, 2, 4, and 5), but not by patient 9, was identified in chromosome 6p21 (haplotype cluster rs6924453–rs3763341) (Figure 3A). This region spans ∼2.7 Mb and includes 164 genes, including 120 protein-coding genes, encompassing the major histocompatibility complex (MHC). Given that CANDLE syndrome is an immune-mediated disease with prominent involvement of the skin and adipose tissue, we performed direct sequencing of the following candidate genes: AGPAT1 (Gene ID 10554), TRIM27 (Gene ID 5987), ITGB1 (Gene ID 3688), SLC39A7 (Gene ID 7922), NOTCH4 (Gene ID 4855), SLC44A4 (Gene ID 80736), and PSMB8 (Gene ID 5969).

Figure 3.

Genetic analyses. A, Genome-wide homozygosity peaks discovered by HomozygosityMapper analysis. Arrow indicates the chromosomal region of homozygosity shared by 4 patients (patients 1, 2, 4, and 5). B, Sequence analysis of PSMB8 (proteasome subunit β type 8), showing the wild type, the heterozygous c.405C>A mutation in the parents of patient 5, and the homozygous c.405C>A mutation in patient 5. C, Sequence analysis of PSMB8, showing the wild type, the heterozygous c.224C>T mutation in the parents of patients 1–4 and in patients 7 and 9, and the homozygous c.224C>T mutation in patients 1–4 and 8. Arrows in B and C indicate locations of the mutations. D, Overview of PSMB8 coverage by exome sequencing demonstrating the identification of a homozygous mutation in PSMB8. Blue stripes represent the sequenced forward strands and red stripes represent the reverse strands. Vertical dashed lines indicate the homozygous mutation c.405C>A with both forward and reverse strands showing the same nucleotide change.

Sequencing of the 6 exons of the PSMB8 gene in patient 5 revealed a homozygous c.405C>A mutation in exon 3, changing cysteine at amino acid 135 to a stop codon (p.C135X; in accordance with ENST00000374882 transcript) (Figure 3B) (see Supplementary Table 1, available on the Arthritis & Rheumatism Web site at The cysteine at position 135 is highly conserved across species. A homozygous missense mutation, c.224C>T, which leads to the substitution of threonine by methionine at position 75 (p.T75M), was found in patients 1, 2, 4, and 8. Of the 4 patients with the same mutation, only 2 shared the same haplotype, indicating a mutational hot spot. Interestingly, 2 patients were heterozygous for the mutation (patients 7 and 9) and, despite extensive analysis, no second mutation has been found; 1 patient (patient 6) showed no mutation in PSMB8. The genotype of the deceased patient 3 was deduced from that of her sister (patient 4), since both had the same disease. There were no mutations identified for the neighboring subunit PSMB2 and the other immunoproteasome-specific subunits PSMB9 and PSMB10 (Figure 3C), but further analyses are ongoing (see Supplementary Table 1, available on the Arthritis & Rheumatism Web site at None of these sequence changes was observed in 750 healthy controls, including 100 Ashkenazi Jews.

Whole-exome sequencing done in parallel generated 2.4 Gb of mappable sequence data and achieved 20-fold coverage of the targeted exome. On average, 57.74% of the bases originated from the targeted exome, with 81% of the targeted bases covered at least 4 times. Initial variant sites calling (>4 times) resulted in the identification of 20,012 genetic variants, including 153 within the shared homozygous region on chromosome 6p21. Eight individual reads were mapped back to the genomic position of the c.405C>A PSMB8 mutation (Figure 3D), confirming the homozygous nonsense mutation.

In silico analysis suggests improper formation of the immunoproteasome.

We modeled mutations identified in PSMB8 in a ribbon diagram (Figure 4A). The β units of the immunoproteasome are organized in 2 overlying rings containing 7 subunits each. Two of the mutations, one previously reported in adults with severe lipodystrophy (p.T75M) (10) and the novel nonsense mutation, p.C135X, identified in the Jewish patient are shown (left image). The p.C135X mutation leads to a large deletion of the terminal 141 amino acids of the inducible β5i subunit. The deleted residues would normally interact with the neighboring β4 subunit in the same ring, but also bind to the β4 unit of the adjacent ring (right image). The nonsense mutation therefore provided us with structural evidence that the immunoproteasome, which requires the incorporation of the β5i subunit, cannot form. The T75 amino acid that is mutated in the other patients is represented by an arrow indicating the mutated T75M residue.

Figure 4.

Functional assessment of mutations in PSMB8 (proteasome subunit β type 8). A, Structural model of the 20S proteasome. Left, A ring of β subunits in a proteasome; the upper right corner shows the position of the ring in the 20S proteasome. The intact residues of the β5i subunit in the patient with the C135X mutation are shown in red and the deleted residues are shown in gray. The T75 amino acid is represented by an arrow indicating the mutated T75M residue. Right, 2 β4 subunits interacting with 2 β5i subunits on 2 adjacent rings; the upper right corner shows their position in a 20S proteasome. B, Cytokine expression in patients 6–8 with chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE syndrome), healthy controls, and patients with neonatal-onset multisystem inflammatory disease (NOMID). Values are the mean ± SEM. Horizontal bars show significant P values. C, Results of whole-blood microarray analysis performed in 4 CANDLE syndrome patients (P4 and P6–P8) and 4 age- and sex-matched healthy controls (C1–C4). A color-coded heatmap was generated on 42 interferon (IFN)–regulated genes from a list of transcripts that were >2-fold differentially expressed compared with their expression in healthy controls (P < 0.05). Red indicates increased expression levels and blue indicates decreased expression levels compared with the mean expression in healthy controls. IP-10 = IFNγ-inducible protein 10; MCP-1 = monocyte chemotactic protein 1; IL-1Ra = interleukin-1 receptor antagonist; TNFα = tumor necrosis factor α.

Functional analyses suggest an inflammatory response involving the IFN pathway.

To assess the dysregulated inflammatory response in CANDLE syndrome patients, we performed cytokine profiling in serum from peripheral blood obtained from 3 patients—1 patient with PSMB8 mutations on both alleles, 1 patient with only 1 identified PSMB8 mutation to date, and 1 patient without a detectable mutation in PSMB8. All showed very high but variable levels of IFNγ-inducible protein 10 (IP-10). Mean IP-10 levels in CANDLE syndrome patients were 77-fold higher than those observed in healthy controls and >30-fold higher than those in untreated patients with an IL-1–mediated autoinflammatory syndrome, neonatal-onset multisystem inflammatory disease (NOMID). Other cytokines that were significantly elevated in CANDLE syndrome patients compared to healthy controls were monocyte chemotactic protein 1 (MCP-1) and RANTES (Figure 4B). Notably, levels of IL-6 were modestly elevated not only in CANDLE syndrome patients but also in patients with NOMID compared to controls, and in CANDLE syndrome patients IL-1 receptor antagonist (IL-1Ra) levels were also elevated.

The very high levels of IP-10 suggested excessive IFN responses in CANDLE syndrome patients (see Supplementary Figure 1, available on the Arthritis & Rheumatism Web site at To probe for evidence of excessive IFN signaling in CANDLE syndrome patients in vivo, we assessed the transcriptome in whole-blood microarray analysis in 4 CANDLE syndrome patients and 4 age- and sex-matched healthy controls.

CANDLE syndrome patients had 507 genes (650 transcripts) that were >2-fold differentially expressed compared with their expression in healthy controls (P < 0.05) (see Supplementary Table 2, available on the Arthritis & Rheumatism Web site at Differentially expressed genes were analyzed using Ingenuity Pathway Analysis to identify dysregulated canonical pathways, and the IFN pathway was the most differentially regulated in CANDLE syndrome patients (P = 4.73 × 10−6). Of the list of IFN-induced genes in Ingenuity Pathway Analysis, most were induced by IFNγ (n = 42), and 6 were also regulated by IFNα/β. The genes were plotted on a color-coded heatmap, and the patterns of increased and decreased differentially expressed genes were strikingly similar among CANDLE syndrome patients, regardless of the presence or absence of detectable PSMB8 mutations (Figure 4C). IP-10 (CXCL10), which was highly expressed in the patients' serum, was among the IFN-induced up-regulated genes. We also compared our list of differentially expressed genes to IFN-regulated genes reported in, and 119 of the 507 differentially expressed genes were found to be IFN regulated.

Since STAT-1 is a downstream mediator of IFNα/β and IFNγ signaling, we studied STAT-1 phosphorylation in the monocytes in response to IFNγ stimulation. Compared with monocytes from healthy controls and from a patient with NOMID, monocytes from CANDLE syndrome patients showed stronger STAT-1 phosphorylation in response to IFNγ at all concentrations used for stimulation (Figure 5A).

Figure 5.

STAT-1 phosphorylation in monocytes from CANDLE syndrome patients in response to IFNγ stimulation, and inhibition by a JAK kinase inhibitor (tofacitinib). A, Shown is STAT-1 phosphorylation in response to 15 minutes of IFNγ stimulation in monocytes from CANDLE syndrome patients, healthy controls, and a NOMID patient. The dashed line indicates an isotype control, the shaded line indicates the response in a healthy control, and the solid line indicates the response in a CANDLE syndrome patient. Bottom, The bold dotted line indicates the response in a NOMID patient. B, Peripheral blood mononuclear cells from a healthy control and from patient 7 (P7) were stimulated with 10 IU/ml IFNγ for 15 minutes in the absence of tofacitinib and in the presence of 0.1 μmoles or 0.5 μmoles of tofacitinib. A dose-dependent inhibition of STAT-1 phosphorylation was seen in both the patient and the healthy control. APC = allophycocyanin (see Figure 4 for other definitions).

To assess the effect of various treatments received by the patients on IFN-induced genes, blood samples were obtained at multiple visits from 2 patients, including 1 patient treated at different times with anti-TNFα and anti–IL-6 therapy (see Supplementary Figure 2, available on the Arthritis & Rheumatism Web site at Although temporary clinical improvement was seen with anti-TNFα and anti–IL-6 treatment (Goldbach-Mansky R: unpublished observations), the “IFN signature” did not improve. IL-6–blocking therapy normalized IL-6–inducible genes (data not shown) and C-reactive protein levels; however, skin lesions, fatigue, or joint pain did not improve significantly and peripheral fat loss progressed, suggesting a possible association between the IFN signature and disease activity.

JAK kinases are critical signaling molecules mediating IFN signaling on the IFN receptors. To determine the effect of a JAK kinase inhibitor, tofacitinib, on the excessive IFN response in CANDLE syndrome patients, we assessed its inhibiting effect on STAT-1 phosphorylation in monocytes stimulated with IFNγ. Tofacitinib decreased STAT-1 phosphorylation in a dose-dependent manner both in monocytes from a CANDLE syndrome patient and in monocytes from a healthy control (Figure 5B). Tofacitinib also inhibited IFNγ-induced IP-10 production in PBMCs in a dose-dependent manner, and its inhibitory effect was more efficient than with the IL-1 receptor antagonist anakinra or anti–IL-6 blockade with tocilizumab (data not shown). Since tofacitinib and other agents blocking the IFN pathways are not available to use as treatments in our patients, we could not assess the effect of blocking the IFN signaling pathway on the clinical symptoms of and laboratory findings in CANDLE syndrome patients.


In the current study we identified mutations in the PSMB8 gene as the cause of CANDLE syndrome, extending the phenotypic spectrum of a novel, recently described autoinflammatory syndrome caused by PSMB8 mutations (10). We also identified dysregulation of the IFN signaling pathway and suggest that the IFN pathway may be a target for treatment in these patients.

After the original report of CANDLE syndrome in 4 children (5), a syndrome diagnosed in 3 adult patients with joint contractures, muscle atrophy, microcytic anemia, and panniculitis-induced childhood-onset lipodystrophy was reported under the acronym “JMP” (10). Patients with “JMP” were recently demonstrated to carry a mutation in the PSMB8 gene (8). The patients described were homozygous for the same mutation, p.T75M, that we found in 5 of our patients (8, 10).

Although CANDLE syndrome patients have some overlapping features with “JMP” patients, including cutaneous eruption and lipodystrophy (10), none of our patients has developed joint contractures, and muscle atrophy was not a prominent disease feature, although 2 patients (patients 1 and 7) developed an acute, self-limited attack of myositis. CANDLE syndrome patients, on the other hand, showed several key features that have not been described in the “JMP” patients, particularly recurrent febrile episodes, elevated acute-phase reactant levels, and a characteristic neutrophilic dermatosis with a mononuclear interstitial infiltrate including “immature” neutrophils in the dermis that seems pathognomonic for CANDLE syndrome. In fact, 2 CANDLE syndrome patients have been misdiagnosed as having acute cutaneous myelogenous leukemia. Nevertheless, the detection of the same and additional mutations in PSMB8 unifies these disorders as a novel immunoproteasome-associated autoinflammatory syndrome. Clinical reports of children with disease manifestations that closely resemble those of CANDLE syndrome from families in Japan and Lebanon (11–13) might allow for the discovery of further molecular causes of this disease. While our data in young children illustrate manifestations of early, severe, and potentially lethal disease and alert us to the fact that muscle involvement and joint contractures may not present until later in life, the findings in the adult patients likely illustrate the natural course of the disease in untreated or partially treated patients (10, 11).

In 1 of our patients, no mutation in PSMB8 was discovered, and 2 patients showed a mutation in only 1 allele, despite sequencing of the entire exonic and intronic sequence of the gene. In the boy without a mutation on either allele (patient 6), we similarly found no mutation in sequencing the 2 other immunoproteasome-specific subunits. These data indicate genetic heterogeneity underlying CANDLE syndrome and raise the prospect of other genes as the genetic cause of CANDLE syndrome.

The gene mutated in CANDLE syndrome, PSMB8, encodes the inducible β5i subunit of the proteasome, a protein complex that consists of 2 α rings and 2 β rings. Each ring is formed of 7 different globular α or β subunits. Proteasomes are evolutionarily conserved cylindrical structures that are critical for protein degradation (14). Upon IFN stimulation, critical subunits of a constitutive proteasome, the β1, β2, and β5 subunits, are replaced with inducible i subunits, β1i, β2i, and β5i, to form immunoproteasomes which are highly expressed in hemopoietic cells (15).

The functions of the immunoproteasomes have been studied in vitro and in animal models. The immunoproteasome can generate antigenic peptides for class I MHC presentation (16), but recent data in psmb8/lmp7-knockout mice (17) suggest an important additional role in maintaining cell homeostasis by removing accumulating proteins marked for degradation from the cells (18). Cellular stress such as infection or radiation leads to type I IFN–induced production of reactive oxygen species and newly synthesized proteins that are particularly sensitive to oxidation (19–21). Failure to process/degrade protein will result in formation of ubiquitin-rich cytoplasmic aggregates or inclusions and consequently increase cellular sensitivity to apoptosis (18). It is thought that the excessive demand for protein processing/degradation is mainly met by cytokine-mediated up-regulation of the ubiquitination machinery and increased assembly of the highly efficient immunoproteasome (22, 23).

The persistent IFN signature in CANDLE syndrome patients on microarray and the increased STAT-1 phosphorylation in monocytes from CANDLE syndrome patients in response to IFNγ stimulation could reflect ongoing “cellular stress” in these patients. In concordance with the current understanding of the immunoproteasome function, we have suggested a disease model (see Supplementary Figure 1, available on the Arthritis & Rheumatism Web site at which proposes that defects in immunoproteasome function may lead to accumulation of damaged proteins, resulting in more cellular stress and a vicious circle of increased IFN signaling. Interestingly, CANDLE syndrome flares are observed with infections and other stressful events. Some cells such as fat or muscle cells may be subject to cellular apoptosis due to accumulation of damaged proteins. In fact, a Japanese patient with severe fat loss, muscle atrophy, and suspected CANDLE syndrome died of cardiac failure at age 47 years. Histologic examination of skeletal muscle at autopsy revealed intramitochondrial paracrystalline inclusions and cytoplasmic and myeloid bodies in muscle cells (24). Whether the inclusions seen constitute accumulation of oxidant-damaged/aggregated proteins that cause cell death is an attractive hypothesis to account for muscle loss later in life, but studies on the cell-specific effect of the immunoproteasome deficiency are needed to explain the observed visceral effects of the mutations.

IP-10 is a type I or type II IFN–induced protein that functions as a CXC motif chemokine, also known as CXCL10 (see Supplementary Figure 1, available on the Arthritis & Rheumatism Web site at It is produced in a variety of cell types, including endothelial cells, keratinocytes, fibroblasts, mesangial cells, monocytes, dendritic cells, neutrophils, and activated T cells (25). IP-10/CXCL10 is an important chemoattractant for effector T cells; its serum level has been shown to correlate with the extent of T cell infiltration in the tissue. IP-10 may contribute to the pathology in CANDLE syndrome by acting as a chemoattractant for T cells into tissues such as the skin with the described inflammatory infiltrate. Further evaluation is needed to determine whether the inflammatory skin infiltrate of immature neutrophils and activated myeloperoxidase-positive histiocytic cells is a consequence of IFN signaling. However, the important roles of IFNγ in the recruitment of neutrophils through the induction of CCL3 (26) and in the stimulation of myeloperoxidase production in monocyte/macrophages (27) are consistent with a possible pathogenic role of IFN in skin.

Given our own data and the clinical reports of devastating disease manifestations in adults, the outcome of untreated disease is expected to be poor. The partial responses and the continued need for steroids despite treatment with targeted therapies including IL-1Ra, TNFα blockers, and IL-6R inhibitors (see Table 1) highlight the need for better understanding of the disease pathogenesis and for identification of more effective targets for therapeutic intervention. One can test the hypothesis of whether the persistence of the IFN signature after treatment with targeted agents offers a clue to a more effective intervention, since agents blocking IFN signaling (including JAK inhibitors) are in clinical trials.

CANDLE syndrome and the other PSMB8-associated syndromes illustrate the profound effect of immunoproteasome dysfunction on inflammation and organ function. In the current study we have established PSMB8 as the causative gene in CANDLE syndrome and have identified clinical and histologic features that can establish early diagnosis. The mutations in PSMB8, also found in “JMP,” illustrate the clinical and genetic spectrum of this novel immunoproteasome-associated autoinflammatory syndrome. Further studies to fully explore the role of the immunoproteasome in autoinflammatory diseases and to identify other mutations in PSMB8 mutation–negative CANDLE syndrome patients are needed and are ongoing.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Drs. Zlotogorski and Goldbach-Mansky had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Liu, Ramot, Torrelo, Paller, Babay, Lee, Chen, Vera, Zhang, Goldbach-Mansky, Zlotogorski.

Acquisition of data. Liu, Ramot, Torrelo, Paller, Si, Babay, Sheikh, Lee, Chen, Vera, Goldbach-Mansky, Zlotogorski.

Analysis and interpretation of data. Liu, Ramot, Torrelo, Paller, Si, Babay, Kim, Lee, Chen, Vera, Zhang, Goldbach-Mansky, Zlotogorski.


The authors would like to thank Nikki Plass, RN, Deborah Stone, MD, Dawn Chapelle, RN, Sapna Patel Vaghani, MD, and Rhina Castillo, MD for their help in organizing patient visits and patient examinations; Adam Reinhardt, MD, Diane Brown, MD, Paulina Navon-Elkan, MD, and Kristina Rother, MD for their collaboration on patient treatment; Ivona Aksentijevich, MD for her help with the interpretation of the genetic data; Hang Pham, MT for her help with the cytokine analysis; Max Gadina, PhD for his help with the initial cytokine analysis and the sharing of the synthetic CP-690550 (tofacitinib); and Isabel Colmenero, MD, Luis Requena, MD, and Heinz Kutzner, MD for their help in immunohistochemistry studies.