Recessive NLRC4-Autoinflammatory Disease Reveals an Ulcerative Colitis Locus

NLRC4-associated autoinflammatory disease (NLRC4-AID) is a recently described autosomal dominant condition presenting with a range of clinical manifestations that can include macrophage activation syndrome (MAS) and severe enterocolitis [1, 2]. Here, we report the phenotype of a patient with a homozygous mutation in NLRC4 (A160T).

A 60-year-old Brazilian female patient was evaluated for recurrent episodes of systemic inflammation since six months of age. Episodes were characterized by recurrent low-grade fever, chills, oral ulceration, uveitis, arthralgia, and abdominal pain followed by diarrhea with mucus and variable erythematous macular rash. Several inflammatory episodes were triggered by gastrointestinal and urinary tract infections, mainly caused by E. coli or both E. coli and Klebsiella pathogens, respectively. The flares lasted from 7 to 10 days and occurred on average 3–4 times a year, but more than 10 flares a year could be observed. Episodes of fever correlated with increased serum levels of C-reactive protein (CRP), elevated erythrocyte sedimentation rate (ESR), and increased serum amyloid A (SAA) (Fig. 1a). A whole-body PET-CT scan realized during one of the flares while the patient suffered from abdominal pain and diarrhea (in 2016) depicted heterogeneous and diffuse bowel glycolytic activity (Fig. 1b). Colon biopsies were performed and showed unspecific lymphocyte infiltration. However ulcers, abscesses, or clear signs of inflammation could not be observed during regular macroscopical examinations. Fecal calprotectin levels, a commonly used marker to monitor neutrophil migration into gastrointestinal tissue [7], measured 1294 mg/kg before anti-IL-1β therapy was started, which is above normal range (< 50 mg/kg) and indicates gastrointestinal inflammation. Serum IgG was decreased, while IgA and IgM levels were low but within reference ranges (Fig. 1c). Flow cytometry analysis of total lymphocytes collected when the patient was not in flare showed decreased levels of B lymphocytes (CD19 + : patient 1.6% vs. reference 11%), as well as low B memory cells (CD27 + : patient 21.6% vs. reference > 27%) and elevated double negative T cells (CD4-/CD8-/TCR αβ: patient 2.1% vs. reference < 1.7%) (data not shown). Antinuclear antibody (ANA) testing fluctuated between negative (1:80) and positive (1:320) values (data not shown). The patient had a long history of corticosteroids use on demand and did not achieve control of clinical symptoms using any disease-modifying antirheumatic drugs (DMARDS), anti-TNF therapies or intravenous immunoglobulin (IVIG). Since 2017, due to the frequency and severity of flares, high doses of corticosteroids were prescribed and somewhat effective in controlling disease (Fig. 1d). Perhaps associated with prolonged use of corticosteroids the patient developed diabetes, hypertension and dyslipidaemia. High levels of SAA were observed between the flares. As the patient had a decrease in renal function, a kidney biopsy was performed and negative congo red staining ruled out renal amyloidosis (data not shown). Because Anakinra is not available in Brazil, anti-IL-1β therapy with Canakinumab was started and able to partially control the symptoms, which resulted in reduced number of flares, diminished hospitalization, allowance for corticosteroid dose reduction, and an overall improved quality of life.

Analysis of a target gene panel [3] and subsequent next generation sequencing (NGS) identified a homozygous mutation in exon 4 of the NLRC4 gene (NM_021209), c.478G > A encoding the p.A160T variant. Familial genetic segregation analysis using standard Sanger sequencing identified the healthy father, mother, and brother to be heterozygous carriers of the mutant allele (Fig. 1e). Parents were non-consanguineous, with the father of German ancestry and the mother of Syrian ancestry. Heterozygous occurrence of this allele in the general population has been reported with an estimated allele frequency 0.00081, but there are no reported healthy homozygotes in the Genome Aggregation Database (GnomAD).

To investigate potential structural consequence of the p.A160T substitution, we analyzed this residue in the crystal structure of inactive murine NLRC4 (PDB:4KXF), since no structure of human NLRC4 is available to date and proteins from both species share 75% identity on amino acid sequence level (NCBI blastp alignment). A160 is a conserved residue between both species and in the inactive structure of mNLRC4 accessible on the surface and located proximal to the nucleotide binding site and residues p.V341, p.H443, p.S171, where pathogenic dominant mutations have been identified [2, 4, 8, 9] (Fig. 1f). Whereas NLRC4-AID-associated dominantly inherited mutations are buried in the structure with direct links to the nucleotide binding site, the novel recessive mutation is solvent exposed. This could reflect its milder phenotype and the necessity to have two alleles causing a Mendelian disease. Although it is not currently known how the dominant mutations at this site promote an open, active conformation of NLRC4, the fact that A160T is adjacently located is a useful observation.

In contrast to other inflammasomopathies that are mainly driven by increased release of IL-1β, IL-18 was suggested to be the main cytokine responsible for the clinical manifestations of NLRC4-AID [10]. Patient serum cytokine levels during anti-IL-1β treatment with Canakinumab evidenced increased levels of IL-18 (276.80 pg/ml) compared to healthy controls (HCs) (mean 85.82 pg/ml) and a remaining slight increase in IL-1β levels (patient 46 pg/ml, healthy controls mean 8.62 pg/ml) (Fig. 2a). At the time of serum collection for cytokine measurements, the patient had already received continuous Canakinumab treatment for 8 months (1 injection of 4 mg/kg per month). Serum was collected just before another dose of Canakinumab was injected. Since no patient data on cytokine serum levels prior to Canakinumab treatment are available and Canakinumab binding to circulating IL-1β has been reported to prolong cytokine half-life [11], the interpretation of serum IL-1β levels remains indicative. Nonetheless, elevated serum IL-18 levels that are independent of Canakinumab treatment suggest increased NLRC4 inflammasome activation. In order to investigate the disease-causing potential of the mutation, patient monocyte-derived macrophages (MDMs) were stimulated with lipopolysaccharide (LPS), a Toll-like receptor (TLR) 4 agonist, and then ATP or transfected flagellin to activate NLRP3 and NLRC4 inflammasome formation, respectively (Fig. 2b). Treatment with LPS and cytoplasmic flagellin alone did reveal some differences in IL-1β and IL-18 release, suggesting increased baseline and ligand-induced activity of NLRC4 A160T in patient MDMs without LPS-priming (Fig. 2b). However, substantially elevated IL-1β and IL-18 levels were observed in patient-derived cells upon activation of both investigated inflammasomes, with the most pronounced difference in IL-18 released following specific stimulation of NLRC4 in primed cells (LPS + DOTAP/FLA) (Fig. 2b). TNF levels, measured as inflammasome-independent priming control, were comparable in all tested conditions (Suppl. Figure 1a). Interestingly, we found that longer-term 24 h stimulation with LPS did not alter IL-18 production, but 24 h FLA (extracellular) showed an increase in IL-18 release for the patient sample, which may suggest internalization of FLA and an increased response of NLRC4 A160T at later time points. Overall, these results suggest NLRC4 A160T autoactivation to some extent, and that environment, infection, or other triggers are able to exaggerate responses due to NLRC4 A160T, which is consistent with flares of disease experienced by the patient. Increased responsiveness of MDMs from another NLRC4-MAS patient to NLRP3 stimulation (LPS + ATP) has been reported in the literature [1] and studies have suggested an interaction of NLRP3 with NLRC4 and recruitment to the same inflammasome complex [12, 13], which could provide a mechanistic basis to explain why NLRP3 stimuli also appear to trigger increased responses in NLRC4 A160T MDMs; however, this explanation remains speculative.

As an in vitro readout for inflammasome assembly and activation on a single cell level, formation of apoptosis-associated speck-like protein containing a CARD (ASC) specks was quantified by flow cytometry [5]. HEK293T cells stably expressing ASC-RFP were transiently transfected with mCitrine-tagged NLRC4 wild type (WT) or A160T. Increased ASC speck formation was observed with NLRC4 A160T suggesting spontaneous inflammasome assembly. To resemble a heterozygous carrier, WT and A160T NLRC4 were co-transfected at a 50:50 ratio, which resulted in slightly elevated ASC speck formation compared to WT NLRC4. The dominant mutation NLRC4 S171F [4, 8] results in stronger inflammasome activation than A160T, which is recessively inherited in this case (Fig. 2c). Protein levels of NLRC4 WT, WT/A160T, and S171F are comparable following overexpression in this assay, but slightly reduced for NLRC4 A160T (Fig. 2d). This finding was consistent between experimental repeats, which may suggest that in some circumstances, this mutant is less stable and degraded more rapidly.

To independently verify these results, Flp-In 293 T-REx cell lines stably expressing ASC-EGFP and NLRC4 WT, A160T, or S171F were generated (Suppl. Figure 1b). Site-specific genomic integration of a single copy of NLRC4 under a tetracycline-inducible promoter yielded a highly controlled and consistent experimental system. Importantly, western blot analysis confirmed similar expression levels between different cell lines (Suppl. Figure 1c), which leads us to speculate that reduced A160T stability might only occur when expressed above endogenous levels (Fig. 2d). Without stimulation, ASC specking was induced by autoactivating mutation S171F, but not by A160T or WT NLRC4 (Suppl. Figure 1d). Instead, for this system, increased ASC-speck formation by NLRC4 A160T over WT was only observed following NLRC4 activation by transient transfection of the sensor protein NAIP and S.typhimurium type three secretion system (T3SS) needle protein PrgI-expressing plasmids (Fig. 2e, f). However in this experiment, protein expression levels of NLRC4 A160T were slightly below WT levels (Fig. 2f). This was attributable to a small difference in mRNA transcript expression in this case (qRT-PCR data not shown), rather than intrinsic defects in protein stability which may result is slight underestimation of the effect caused by NLRC4 A160T.

To recapitulate the effect of NLRC4 A160T in a human monocyte-like cell line with endogenous expression of proteins relevant for inflammasome activation and cytokine release, we employed CRISPR/Cas9-mediated gene editing and homology-directed repair to genetically modify THP-1 cells and introduce the A160T mutation on genomic level (Suppl. Figure 2a). Monoclonal cell lines were generated and sequences of the NLRC4 genomic locus as well as mRNA transcripts confirmed by Sanger sequencing (Suppl. Figure 2b). This experimental approach yielded one clone with in-frame insertion of the correct mutation (c.478G > A) on one allele and 1 bp deletion on the second allele, generating an A160T/KO genotype. A matching WT/KO control clone and 3 WT/WT clones carrying one or two functional NLRC4 WT alleles, respectively, were selected as controls for subsequent stimulation experiments. Furthermore, a KO/KO clone with homozygous 143 bp deletion inducing a premature stop codon was included (Suppl. Figure 2b). qRT-PCR analysis of NLRC4 transcript levels demonstrates that expression in A160T/KO and WT/KO clones is around 50% reduced compared to THP-1 pool and WT/WT clones with 2 functional NLRC4 alleles (Suppl. Figure 2c). Therefore, the WT/KO clone was used as comparator group in subsequent stimulation assays.

NLRC4 stimulation of monoclonal cell lines with PrgI and TLR1/2 agonist Pam3CSK4 (P3C)-mediated priming suggests enhanced inflammasome signalling by NLRC4 A160T (clone A160T/KO) as observed by significantly elevated release of IL-1β and IL-18 into supernatants and increased cell death. Spontaneous signalling after P3C-priming alone was not observed, indicating that A160T does not cause NLRC4 autoactivation and requires a NLRC4-specific stimulus to elicit pathogenicity (Fig. 2g).

To exclude clonal differences between these THP-1 cell lines, we tested inflammasome priming by quantification of IL-8, a NF-κB-induced cytokine (Suppl. Figure 2d), and activation of a different inflammasome sensor, NLRP3 (Suppl. Figure 2e). Small differences were observed in IL-8 levels in response to P3C priming alone, however that was not consistent in the PrgI-treated samples (Suppl. Figure 2d). Cells stimulated with P3C and Nigericin, to activate NLRP3, also had smaller differences in IL-1β and IL-18 production but induced cell death did not differ (Suppl. Figure 2e). This data suggests that there are likely some clonal differences between the cell lines, but not to the same extent as for NLRC4 activation.

Unlike IL-1β, which is a NF-κB-induced cytokine, the IL-18 precursor protein is constitutively expressed in various cell types, including intestinal epithelial cells and was previously shown to be involved in intestinal inflammation [14]. Since the severity of inflammatory bowel disease (IBD) was correlated with increased IL-18 secretion in clinical studies [14], we aimed to investigate the genome-wide association of NLRC4 (A160T) with Crohn’s disease (CD) and ulcerative colitis (UC), the two common forms of IBD. Using data from the IBD exomes portal (Broad institute), we found significant enrichment of the NLRC4 (A160T) allele in UC patients compared to healthy controls, with an odds ratio (OR) of 2.546 ((95% 1.778–3.644), P = 0.01305) (Table 1). Other genetic variants have been reported to be associated with UC, with odds ratios comparable to NLRC4 (A160T). These include variants in inflammatory pathway related molecules, such as NLRP7 (S361L): OR = 4.79, P = 0.0039 [15] and NOD2 (R38M): OR = 2.904, P = 0.03291 (IBD exomes portal). Of note, the correlation between NLRC4 (A160T) and CD was less clear (OR = 1.62 (95% 1.119–2.346), P = 0.2615), suggesting this locus as a risk factor predominantly for the development of UC.