Additive loss-of-function proteasome subunit mutations in CANDLE/PRAAS patients promote type I IFN production

Abstract

Autosomal recessive mutations in proteasome subunit β 8 (PSMB8), which encodes the inducible proteasome subunit β5i, cause the immune-dysregulatory disease chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE), which is classified as a proteasome-associated autoinflammatory syndrome (PRAAS). Here, we identified 8 mutations in 4 proteasome genes, PSMA3 (encodes α7), PSMB4 (encodes β7), PSMB9 (encodes β1i), and proteasome maturation protein (POMP), that have not been previously associated with disease and 1 mutation in PSMB8 that has not been previously reported. One patient was compound heterozygous for PSMB4 mutations, 6 patients from 4 families were heterozygous for a missense mutation in 1 inducible proteasome subunit and a mutation in a constitutive proteasome subunit, and 1 patient was heterozygous for a POMP mutation, thus establishing a digenic and autosomal dominant inheritance pattern of PRAAS. Function evaluation revealed that these mutations variably affect transcription, protein expression, protein folding, proteasome assembly, and, ultimately, proteasome activity. Moreover, defects in proteasome formation and function were recapitulated by siRNA-mediated knockdown of the respective subunits in primary fibroblasts from healthy individuals. Patient-isolated hematopoietic and nonhematopoietic cells exhibited a strong IFN gene-expression signature, irrespective of genotype. Additionally, chemical proteasome inhibition or progressive depletion of proteasome subunit gene transcription with siRNA induced transcription of type I IFN genes in healthy control cells. Our results provide further insight into CANDLE genetics and link global proteasome dysfunction to increased type I IFN production.

Figure 7. Type I IFN induction in patients’ PBMCs and fibroblasts and in healthy control PBMCs treated in vitro with proteasome inhibitors.

(A) RNAseq was performed on whole-blood RNA. Differentially expressed genes in CANDLE/PRAAS patients (fold change > 2, P < 0.05) were analyzed by Ingenuity, and IFN-regulated genes were plotted on the heat map. (B) PBMCs from healthy controls were treated with indicated concentrations of the proteasome inhibitor epoxomicin for 24 hours. Expression of IFN genes was analyzed by qRT-PCR. Fold change was calculated for each condition relative to the mean of no-treatment controls. Data represent mean ± SEM. n = 8 for IFN genes and 200 nM concentration; n = 5 for all other concentrations; n = 4 for OAS3, MX1, and IL1B; n = 3 for IL6 and TNFA. Paired t tests were done using ΔCt values. *P < 0.05; **P < 0.01. (C) Expression levels of IFNB1, IFNA7, IFNA17, IFNA5, and IFNA21/1 from 3 CANDLE/PRAAS patients and 4 healthy donors were analyzed by qRT-PCR. Fold changes were calculated over the average of 4 healthy controls. Data represent mean ± SEM. n = 3. Two-sample t tests were performed. P values are stated. (D) Expression of IFNs from whole blood of 2 active CANDLE patients (patient 1 and patient 2) were analyzed by flow cytometry. (E) Expression of IFNs in PBMCs from a healthy donor treated with the proteasome inhibitor epoxomicin at indicated concentrations was analyzed by flow cytometry. (D and E) Representative results from n = 3 and n = 2, respectively.

Figure 6. Additive depletion of proteasome subunits by siRNA simulated the digenic inheritance of proteasome mutations in patients.

Primary human fibroblasts were depleted for PSMB8, PSMB4, PSMA3, or PSMB9 as well as combinations thereof: PSMB8 plus PSMB4; PSMB8 plus PSMA3; PSMB9 plus PSMB4 by siRNA. The following siRNA concentrations were used: off-target 1, off-target 2, PSMA3, PSMB8, and PSMB9 (10 nM); PSMB4 (15 nM). (A and D) Representative results from n = 3. (A) Native PAGE substrate overlays with lysates of siRNA-treated cells show reduction of proteasome activity. (B) Quantification of native PAGE substrate overlays. (C) Knockdown control for the respective genes with qRT-PCR shows approximately 60% to 40% efficiency, respectively. (D) Immunoblot stained for α6 shows decreased total amount of proteasomes. In cells treated with siRNAs targeting the expression of 2 different subunits. a stronger decrease of proteasomes and their activity was observed. (E) mRNA expression of type I IFN and IFN-regulated genes MX1 and IP10 was significantly upregulated in cells treated with 2 different proteasomal siRNAs as assayed by qRT-PCR. All data in bar graphs represent mean ± SEM. n = 3. Samples were normalized against off-target siRNA 1+2. Paired t tests were performed. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 5. Ubiquitin aggregation and proteasome profile changes in CANDLE/PRAAS patient cells.

(A) Sections from lesional skin biopsies from 2 PRAAS patients, a patient with psoriasis, and a nonlesional skin biopsy from a healthy control were stained with anti-ubiquitin antibody (Dako) for accumulation of polyubiquitinated proteins (brown staining). Scale bar: 50 μm. Representative results from n = 3. (B) Insoluble fractions of RIPA cell lysates of keratinocytes were solubilized in urea lysis buffer and separated on an SDS gel followed by immunoblotting for ubiquitin. Patients show a much stronger accumulation of ubiquitinated proteins in this fraction. Representative results are from n = 2. (C and D) Keratinocytes from healthy controls and CANDLE/PRAAS patients were lysed under native conditions and separated on a native gel. PSMB8 C1 denotes a CANDLE patient homozygous for PSMB8 mutation (T75M/T75M). The lanes for patient 2, PSMB8 C1, and HC1 were run on the same gel, but were noncontiguous. n = 1. (C) Immunoblotting was performed for α6 and β5i. Patient samples harboring β7 and/or β1i mutations were also immunoblotted for β7 and β1i. Asterisk indicates unspecific crossreaction of β1i antibody (Abcam). (D) An in-gel overlay experiment was performed for chymotryptic-like activity to visualize the activity of the proteasome. (E) Plate reader activity assay was measured from whole keratinocyte cell lysates. Patients show a strong reduction in 2 or 3 protease activities. Each of the 5 patients’ and 3 parents’ samples were normalized against HC1+2. Means were estimated from the triplicate values on the normal controls (n = 2). Data were analyzed by restricted maximum likelihood mixed models methods, using the xtmixed procedure in Release 12 of the software package Stata (StataCorp). Data represent the mean ± SD from technical triplicates. n = 1. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 4. Proteasomal activity of PBMCs.

(A) Proteasome activity of PBMCs from CANDLE/PRAAS patients (turquoise, Pt.1; orange, Pt.2; purple, Pt.4 and Pt.5; red, CANDLE patients with PSMB8 mutations), patient family members or heterozygous carriers (gray), healthy controls (HC, black), patients with an undifferentiated interferonopathy disease who are negative for proteasome gene mutations (UID, green), and active pretreatment NOMID patients (blue) was analyzed and expressed as percentage relative to the average of healthy controls. The dashed lines indicate 100% activity. Whole blood RNA samples from the same date for patient and patient family members were analyzed by qRT-PCR for IFN-regulated gene expression and an IFN score was calculated. For patient 2, we used control samples from nonrelated heterozygous carriers for the PSMB8 T75M (HZMa, HZMb, and HZMc) for comparison. Measurements for patient 4, patient 5, and their parents were performed at only 1 time point before treatment. Patients with PSMB8 mutations, C1 (T75M/T75M), C2 (C135X/C135X), and C3 (T75M/A92T), were assessed as CANDLE controls. Two-sample, 2-tailed t tests were performed, and P values were stated. Error bars indicate technical triplicates. (B) Patient proteasome activity was normalized to the UID mean for each of the proteasome activities. Paired t test was performed. (C) PBMCs from 3 healthy donors were stimulated with the indicated cytokines or none for 6, 12, or 24 hours. The cell numbers were counted and a proteasome activity assay was done. Fold changes of activity against no-treatment control was calculated. The dashed lines on the graph indicate the activity of no-treatment controls. The numbers on top of the data points are IFN scores. Data represent mean ± SEM from n = 3 samples. Two-sample, 2-tailed t tests were performed. *P < 0.05.

Figure 3. Maturation and incorporation of mutant proteasome subunits into proteasome complex in vitro in HeLa cells.

WT and mutant versions of the subunits PSMA3/α7, PSMB4/β7, and PSMB8/β5i/LMP7 were ectopically expressed with V5 epitope tags in HeLa cells. (A) Schematic representation of V5-tagged subunits and their maturation behavior. Of note, most of the β subunits (but not α subunits) are expressed as proforms and have to be matured by autocatalytic propeptide cleavage during proteasome assembly. Maturation of active site subunits such as β5i is a 2-step procedure resulting in proform, intermediate proform, and mature form of the subunit. (B) Immunoblot of HeLa cells transfected with V5 epitope–tagged subunits and detected by a V5-specific antibody. MOCK, empty vector backbone. Asterisks indicate unspecific background staining. β–Tubulin served as loading control. Expression of α7-R233del-V5 or β5i-C135X-V5 revealed almost no detectable protein, whereas the -9G>A mutation in β7-V5 or the β5i-T75M and -G201V mutations caused decreased protein expression. The β7-3aadel-, β5i-G201V-, -K105Q-, and -A92T-V5 versions display altered maturation of these β subunits (see accumulation of preforms or intermediate forms). (C) Incorporation of V5-tagged subunits into the proteasome complexes was assessed by native PAGE analysis and immunoblot against the V5 epitope tag. The β7-3aadel-V5 and β5i-G201V-V5 exhibit the most prominent incorporation defects. (D) With the exception of the -9G>A mutation in PSMB4-V5, all mutant subunit mRNAs were equally expressed in HeLa cells, as evaluated by RT-PCR. Expression of endogenous GAPDH mRNA served as loading control. Transfection efficiency was determined by FACS analyses of cotransfected EGFP vector. (B–D) Representative results from n = 3.

Figure 2. Analysis of mutated gene expression.

(A) Using WT and mutant allele (5′UTR: c.-9G>A) specific primers to assess PSMB4 expression, the mutant transcript shows significantly lower expression than WT transcript. This is observed in patient 1 and his mother (Pt.1-M) when comparing WT (blue bar in Pt.1-M) and maternal PSMB4 mutant (red) transcript level. Patient 1 and his father (Pt.1-F) carry the PSMB4 (β7) 3-aa deletion detected as the G allele here; in the patient, the G allele is the amplified mutant paternal allele; in the father the expression level includes WT and the 3-aa deletion allele. (B) WT and mutant (c.696_698del, p.R233del) PSMA3 allele from patient 2 shows equal transcription for both. (C) PSMA3 c.404+2T>C mutation on splicing of exon 5 leads to unstable transcription of the mutant allele. (D) Reduced expression of PSMB4 is seen in the asymptomatic father (Pt.4+5-F) and in patient 4 and patient 5 who are heterozygous for the c.44_45insG mutation allele. The father may compensate with increased transcription of the WT allele. The mother (Pt.4+5-M) has 2 WT alleles for PSMB4 (she carries the p>G165D mutation in PSMB9). Contr., healthy adult control. (A–D) Tests were conducted as technical triplicates. (E) Sequencing of the PCR amplified product shows skipping of exon 5. (F) Mutant transcript lacking exon 5 is amplified with the junction-specific primers (patient 3 and his mother [patient 3-M], both carry the c.404+2T>C [p.H111Ffs*10] mutation). The amplification of mutant transcript from cDNA was performed using 2 different junction-specific exon 4/6 spanning primer sets that failed to amplify genomic DNA, as expected. DW, distilled water. The lanes of the upper panel and of the lower panel were run on the same gel, but were noncontiguous.

Figure 1. Clinical findings and CANDLE/PRAAS-associated mutations in 4 proteasome-encoding genes and in silico modeling.

(A) Marked facial edema during flare. (B) Lipoatrophy later in life. (C) CANDLE rash during acute flare. (D) Abdominal protrusion due to intraabdominal fat deposition. (E) Pedigrees and identified genotypes of patients and their direct relatives. Underline in red indicates maternal, in blue, paternal, and in green, de novo inheritance of mutant allele. (F) Schematic organization of PSMB4, PSMA3, PSMB9, and PSMB8 genes (exon-intron structure, black rectangles represent coding sequences, white rectangles represent UTRs) with positions of the identified mutations. (G) Species conservation of mutated aa (yellow). Hs, Homo sapiens; Pt, Pan troglodytes (chimpanzee); Mm, Mus musculus (mouse); Oc, Oryctolagus cuniculus (rabbit); Bt, Bos tauris (cattle); Clp, Canis lupus familiaris (dog); Xl, Xenopus laevis (frog), Dr, Danio rerio (zebrafish). Alignment was performed with ClustalW. (H) PSMB8 and PSMB9 mutations were modeled based on the x-ray structure of the mouse immunoproteasome (PDB entry code: 3UNH) (46), and the mutations in PSMA3 and PSMB4 were based on the bovine 20S proteasome (PDB entry code: 1IRU) (19). Mutated subunits α7 (orange), β7 (cyan), and β1i (purple) are located at the opposite side of the 20S particle compared with β5i (red). (I) Top view of α ring. Subunit α7 (orange) with mutant residue R233 (balls) highlighted. (J) Detailed perspectives of ribbon models of mutant proteins. Mutated residues are depicted in yellow with relevant interaction aa side chains shown with stick models. Novel mutations are highlighted in yellow rectangles. Catalytic active sites in β1i and β5i are marked with asterisks. H, heterozygous, NM, nonmutant.