NLRP1 haplotypes associated with vitiligo and autoimmunity increase interleukin-1β processing via the NLRP1 inflammasome

Abstract

Nuclear localization leucine-rich-repeat protein 1 (NLRP1) is a key regulator of the innate immune system, particularly in the skin where, in response to molecular triggers such as pathogen-associated or damage-associated molecular patterns, the NLRP1 inflammasome promotes caspase-1–dependent processing of bioactive interleukin-1β (IL-1β), resulting in IL-1β secretion and downstream inflammatory responses. NLRP1 is genetically associated with risk of several autoimmune diseases including generalized vitiligo, Addison disease, type 1 diabetes, rheumatoid arthritis, and others. Here we identify a repertoire of variation in NLRP1 by deep DNA resequencing. Predicted functional variations in NLRP1 reside in several common high-risk haplotypes that differ from the reference by multiple nonsynonymous substitutions. The haplotypes that are high risk for disease share two substitutions, L155H and M1184V, and are inherited largely intact due to extensive linkage disequilibrium across the region. Functionally, we found that peripheral blood monocytes from healthy subjects homozygous for the predominant high-risk haplotype 2A processed significantly greater (P < 0.0001) amounts of the IL-1β precursor to mature bioactive IL-1β under basal (resting) conditions and in response to Toll-like receptor (TLR) agonists (TLR2 and TLR4) compared with monocytes from subjects homozygous for the reference haplotype 1. The increase in basal release was 1.8-fold greater in haplotype 2A monocytes, and these differences between the two haplotypes were consistently observed three times over a 3-mo period; no differences were observed for IL-1α or TNFα. NLRP1 RNA and protein levels were not altered by the predominant high-risk haplotype, indicating that altered function of the corresponding multivariant NLRP1 polypeptide predisposes to autoimmune diseases by activation of the NLRP1 inflammasome.

Autoimmune diseases are a diverse group of more than 80 chronic disorders in which the immune system attacks “self” tissues and cells (1), altogether affecting 3–5% of the U.S. population (2). Concomitant occurrence of multiple autoimmune diseases in some patients is well known (3–5) and results from shared genetic and perhaps environmental risk factors among these different diseases (6–8). Polymorphisms of the nuclear localization leucine-rich-repeat protein 1 (NLRP1) gene encoding the NACHT (NAIP, CIITA, HET-E, TP1) domain of NLRP1 have been associated with a number of autoimmune diseases, including vitiligo (9, 10), Addison disease (11, 12), type 1 diabetes (11), celiac disease (13), systemic lupus erythematosus (14), rheumatoid arthritis (15), systemic sclerosis (16), Kawasaki disease (17), as well as steroid responsiveness in inflammatory bowel disease (18), although the specific biological mechanism underlying these genetic associations remains unknown.

NLRP1 is a regulator of the innate immune response (19–21) and is expressed in many immunocompetent cell types, particularly the Langerhans cells of the skin (22). Pathogen- or damage-associated molecular patterns stimulate surface Toll-like receptors (TLRs) with subsequent assembly of the NLRP1 inflammasome and activation of caspase-1. Active caspase-1 then cleaves the inactive IL-1β precursor (pro–IL-1β) to the mature bioactive IL-1β (23), thereby stimulating downstream inflammatory responses (23, 24). Because of its highly inflammatory and immunostimulating properties, the processing and release of bioactive IL-1β are tightly controlled. However, in cryopyrin-associated periodic syndromes (CAPSs), which include familial cold autoinflammatory disease, Muckle–Wells syndrome, and neonatal onset multi-inflammatory disease, amino acid substitutions in the NACHT domain of NLRP3 result in constitutive activation of caspase-1 and processing and secretion of bioactive IL-1β, accompanied by marked systemic and local inflammation (21, 25–28). Blood monocytes from affected patients release significantly more mature bioactive IL-1β than cells from healthy individuals (28). Similarly, a dominant activating mutation in mouse Nlrp1a has recently been shown to result in a severe systemic inflammatory phenotype associated with greatly elevated release of active IL-1β (29).

In the present study we have investigated the hypothesis that common inherited variation in human NLRP1, which is associated with genetic risk of autoimmune disease, similarly results in constitutively increased processing and release of bioactive IL-1β. By next-generation DNA sequencing, we identified common NLRP1 haplotypes containing multiple predicted functional variants that are associated with disease. We next studied the in vitro production of cytokines from monocytes of healthy individuals homozygous for the most prevalent high-risk NLRP1 haplotype and compared the responses to monocytes from subjects with the low-risk reference haplotype. Our findings indicate that these multivariant NLRP1 haplotypes predispose to common autoimmune diseases by activation of the NLRP1 inflammasome.

Results

NLRP1 Sequencing Identifies Conserved Disease-Associated Haplotypes That Differ by Multiple Nonsynonymous Substitutions.

To identify genetic variation in NLRP1 that might contribute to autoimmune susceptibility, we resequenced the complete NLRP1 gene region in probands of the 114 families with multiple cases of vitiligo and other autoimmune diseases in which we originally established genetic linkage and association with NLRP1 (9, 10). The 161.5 kb region sequenced (chr17:5386564–5548062) included the entire 70.4 kb NLRP1 structural gene, as well as 60.2 kb upstream and 30.9 kb downstream. We observed a total of 656 sequence variants (Dataset S1), including 261 not annotated in the Short Genetic Variations database (dbSNP), Build 135. Within the exons, these included 16 nonsynonymous variants, with no observed nonsense variants, splice junction variants, or coding region indels. We verified all previously unannotated nonsynonymous variants by Sanger sequencing in the corresponding patients, and we integrated the observed sequence variants with high-density NLRP1 region SNP data in these same families (9).

The NLRP1 region consists of three main haplotype blocks, comprising the NLRP1 structural gene, the 5′-untranslated sequence and extended promoter region, and the 3′-untranslated and 3′- flanking region. Of the 16 nonsynonymous variants, 12 were genotyped directly or were sufficiently frequent to permit accurate genotype imputation (r² > 0.3) in the corresponding families. Previous family-based association analyses showed that rs12150220 (L155H) was the single variant most significantly associated with vitiligo (9). However, the L155H variant occurs on three main haplotypes, each differing by multiple additional nonsynonymous substitutions (Fig. 1 and Dataset S1), with no apparent ancestral recombinants observed through the region. Thus, the corresponding NLRP1 polypeptides would likewise differ by multiple amino acid substitutions, although all include L155H. The reference haplotype 1 was the most prevalent haplotype (frequency, 0.53). Haplotype 2 (overall frequency, 0.42) comprised three subtypes: 2A, 2B, and 2C. Haplotype 2A was the most prevalent high-risk haplotype (OR, 1.6; frequency, 0.39), containing three nonsynonymous substitutions (L155H–V1059M–M1184V) compared with reference haplotype 1. Haplotypes 2B and 2C were less frequent (frequency, 0.01); both appear to be derived from haplotype 2A and respectively contained three (L155H–V1059M–M1184T) and four (L155H–V939M–V1059M–M1184V) nonsynonymous substitutions. Haplotype 3 was even higher risk (OR, 3.7; frequency, 0.05) and contained a remarkable nine nonsynonymous substitutions (L155H–T246S–T782S–T878M–T995I–M1119V–M1184V–V1241L–R1366C), with only a single observed apparently intermediate form that contained L155H, M1184V, R1366C, and many synonymous and intronic variants otherwise specific to haplotype 3. In addition to these major multivariant haplotypes, we also observed four singleton nonsynonymous substitutions, G354D, E869K, T1113M, and R1289H, of which only T1113M cosegregated with autoimmunity in the corresponding multiplex family. Of the 16 observed nonsynonymous substitutions, bioinformatic analyses predict that six are probably or possibly deleterious (Fig. 1), although individual in silico functional predictions are of uncertain validity in the context of such multivariant haplotypes. Of these, L155H is the only predicted deleterious substitution shared by both high-risk haplotypes 2A and 3, which additionally share the predicted benign variant M1184V.

Fig. 1.

NLRP1 nonsynonymous substitutions and haplotypes. The 1,473 amino acid NLRP1 protein and functional domains are indicated with nonsynonymous substitutions overline: red, predicted deleterious; yellow, possibly deleterious; green, nondeleterious (predicted by SeattleSeq Annotation). The vertical arrow indicates the posttranslational autolytic cleavage activation site at S1213 (31). Phase of the unique E869K, T1113M, and R1289H variants cannot be assigned. M1184T derives from M1184V via a second change. OR, odds ratio; p, reference haplotype frequency; q, variant haplotype frequency.

Increased Processing of Mature IL-1β from Monocytes of Healthy Subjects with the Prevalent High-Risk NLRP1 Haplotype.

To determine whether the multivariant NLRP1 polypeptides encoded by disease-associated NLRP1 haplotypes differentially affect NLRP1 function, we screened healthy individuals to identify subjects alternatively homozygous for the NLRP1 reference haplotype 1 or for the prevalent high-risk NLRP1 haplotype 2A. We genotyped a total of 115 unrelated healthy non–Hispanic/Latino European-derived white (EUR) individuals without evidence of autoimmune disease for 14 SNPs spanning the NLRP1 structural gene and extended promoter region, which together distinguish haplotype 1, haplotype 2A, haplotype 2B, and haplotype 3, although they do not distinguish the rare haplotype 2C from the prevalent haplotype 2A (Table S1). Among these healthy subjects, the frequency of haplotypes 1, 2A+2C, 2B, and 3 were 0.52, 0.43, 0.01, and 0.04, respectively, which was not significantly different from the distribution of NLRP1 haplotype frequencies observed in the 114 multiplex family probands (P = 0.91).

To determine whether the predominant high-risk NLRP1 haplotype 2A affects function by increasing processing of pro–IL-1β to mature bioactive IL-1β, as in CAPS patients with NLRP3 mutations, we studied healthy subjects homozygous for NLRP1 reference haplotype 1 (n = 6) as well as subjects homozygous for the common high-risk haplotype 2A (n = 10). From each subject we cultured freshly obtained peripheral blood monocytes on three separate occasions over a minimal period of 3 mo. Adherent monocytes were cultured for 24 h either without stimulation (resting) or activated by the TLR4 ligand lipopolysaccharide (LPS), the Nucleotide-binding oligomerization domain-containing protein 2 (NOD2) ligand muramyl dipeptide (MDP), the TLR2 ligand Staphylococcus epidermidis, or a combination of LPS plus MDP. After 24 h of culture, the percent cytotoxicity was assayed by measurement of release of lactate dehydrogenase (LDH). As shown in Fig. S1, there was no significant increase in LDH release associated with each of the four stimulated culture conditions, indicating that these stimuli do not cause extensive cell lysis.

Next, for each subject’s monocyte culture, the levels of both IL-1β and pro–IL-1β were determined in the intracellular compartment (cell lysates) and in the extracellular compartment (supernatant media) by use of specific, non–cross-reacting ELISA, one measuring processed (mature) IL-1β and the other measuring the IL-1β precursor (pro–IL-1β). The values from these two ELISAs for each compartment were added to calculate total IL-1β. The total amount of IL-1β produced was then calculated by adding the values of pro– and mature IL-1β in the two compartments. The percent of mature IL-1β processing from pro–IL-1β was then calculated. As shown in Fig. 2A, there was no difference in the total amount of IL-1β produced by monocytes of the two alternative haplotypes 2A/2A versus 1/1 under each condition of stimulation. In contrast, as shown in Fig. 2B, even in unstimulated (resting) monocytes, the mean percent level of processed IL-1β was 1.8-fold higher (16.3%) in cells from individuals homozygous for high-risk haplotype 2A compared with cells from individuals homozygous for the reference haplotype 1 (8.9%). This difference was highly significant (P < 0.0001). Similarly, increased processing of IL-1β by monocytes from the haplotype 2A homozygotes compared with haplotype 1 homozygotes was also observed for each stimulation condition (P < 0.0001 for each condition), although the increase was proportionately less than in the unstimulated cells, suggesting that this mostly reflects the basal difference in IL-1β processing.

Fig. 2.

IL-1β production from adherent monocytes. (A) Mean ± SEM levels of total IL-1β produced from monocytes cultured from haplotype 1 homozygotes (Hapl 1; n = 6 subjects) and haplotype 2A homozygotes (Hapl 2A; n = 10 subjects). (B) Mean ± SEM percent processed IL-1β from monocytes cultured from haplotype 1 homozygotes (Hapl 1) and haplotype 2A homozygotes (Hapl 2A). Each subject was tested on three occasions, and on each occasion, in triplicate wells. The stimulants are indicated under the horizontal axis: RPMI (no stimulant), LPS (1 ng/mL), MDP (10 μg/mL, LPS + MDP, S. epidermidis (Materials and Methods).

To demonstrate the specificity of these findings for IL-1β, we also determined IL-1α (Fig. 3A) and TNFα (Fig. 3B) production in the same samples. There was no significant difference in the levels of these cytokines in cells from subjects homozygous for the high-risk versus low-risk haplotypes. These data thus indicate that the difference in percent processed IL-1β between the haplotype 2A and haplotype 1 homozygotes is not due to differences in the numbers of monocytes, to expression levels of TLR, or to mutations in the gene encoding the intracellular receptor for MDP, NOD2.

Fig. 3.

Production of IL-1α and TNFα from adherent monocytes. (A) Mean ± SEM levels of intracellular IL-1α produced from monocytes cultured from haplotype 1 homozygotes (same samples as in Fig. 2A) and haplotype 2A homozygotes (same samples as in Fig. 2B). (B) Mean ± SEM level of extracellular TNFα in same samples shown in Fig. 2B. The stimulants are indicated under the horizontal axis as in Fig. 2.

NLRP1 RNA and Protein Expression and Stability.

High-risk NLRP1 haplotype 2A contains three nonsynonymous substitutions, L155H–V1059M–M1184V, of which only L155H is predicted to be individually deleterious. To exclude the possibility that greater IL-1β processing associated with haplotype 2A versus haplotype 1 might reflect differentially elevated NLRP1 mRNA expression from haplotype 2A versus haplotype 1, we measured steady-state NLRP1 RNA levels in lymphoblastoid cells from normal subjects homozygous for haplotype 1 or for haplotype 2A, and found no significant difference (Fig. 4A). Similarly, to exclude the possibility that the observed haplotype-specific difference might reflect greater amounts of the NLRP1 protein or protein stability, we measured steady-state NLRP1 protein in cells of individuals homozygous for haplotype 1 and haplotype 2A, again finding no significant difference (Fig. 4B and Fig. S2).

Fig. 4.

NLRP1 RNA and protein levels. Levels of NLRP1 (A) RNA and (B) protein in lymphoblastoid cells from haplotype 1 homozygotes (Hapl 1) and haplotype 2A homozygotes (Hapl 2A). NLRP1 protein was assayed by Western blot (Fig. S2) and quantitated using ImageJ. NLRP1 RNA and protein were normalized to GAPDH RNA or β-actin protein, respectively.

To specifically test the effect of the L155H substitution on NLRP1 protein stability, we transfected HeLa cells with expression constructs containing NLRP1 cDNA with the 155H versus 155L substitutions on the background of the reference haplotype, again observing no difference in the amount of steady-state NLRP1 protein (Fig. S3). Together, these data demonstrate that there is no differential NLRP1 transcription, translation, or protein stability associated with NLRP1 haplotype 2A versus haplotype 1. Thus, we conclude that increased IL-1β processing in cells from NRLP1 haplotype 2A homozygotes results from altered functional properties of the corresponding multivariant NLRP1 polypeptide carrying the L155H–V1059M–M1184V substitutions.

Discussion

This study identifies a series of common NLRP1 haplotypes that differ by multiple amino acid substitutions in the corresponding NLRP1 polypeptides. At least two of these haplotypes are associated with susceptibility to vitiligo and other autoimmune diseases, likely resulting in part from the L155H substitution shared by these haplotypes. Adherent monocytes cultured from the peripheral blood of healthy individuals homozygous for the common disease-associated haplotype 2A exhibited a 1.8-fold greater amount of IL-1β processing compared with cells from individuals homozygous for the reference haplotype 1. This difference was also observed in resting monocytes—that is, without exogenous stimulation. These data thus suggest that the predominant NLRP1 multivariant haplotype associated with common autoimmune diseases up-regulates IL-1β processing via the NLRP1 inflammasome, in a manner analogous to that by which dominant activating mutations of NLRP3 result in elevated IL-1β processing and severe autoinflammatory disease.

The specific mechanism by which the NLRP1 inflammasome manifests greater functional activity due to haplotype 2A remains unknown. Increase in processing to mature IL-1β was not due to increased levels of pro–IL-1β, increased numbers of monocytes, or greater expression of TLR4 and TLR2. Indeed, we observed that total IL-1β, total IL-1α, and total TNFα production were the same in cells of the two haplotypes, supporting the conclusion that increased processing of IL-1β by monocytes from haplotype 2A homozygotes is directly attributable to NLRP1 itself. However, haplotype 2A is not associated with significantly altered NLRP1 transcription, or expression level or stability of the NLRP1 protein, suggesting that the predominant effect is due to altered function of the corresponding multivariant NLRP1 polypeptide encoded by this haplotype.

Compared with the reference NLRP1 polypeptide encoded by haplotype 1, the NLRP1 polypeptide encoded by haplotype 2A contains three amino acid substitutions, L155H–V1059M–M1184V. Although only L155H is predicted to be deleterious, recently M1184V has been shown to markedly increase the rate of posttranslational autoproteolytic activation of NLRP1 polypeptide (30, 31), indicating that this substitution is also functionally significant. Sharing of both the L155H and M1184V (or the derivative M1184T) substitutions across disease-associated haplotypes suggests the possibility that these two substitutions together may result in greater disease risk than either would alone. Indeed, the haplotype 3 NLRP1 polypeptide contains nine nonsynonymous substitutions, L155H–T246S–T782S–T878M–T995I–M1119V–M1184V–V1241L–R1366C, including L155H and M1184V, and this haplotype is largely inherited intact due to virtually complete linkage disequilibrium and appears to confer even greater disease risk than does haplotype 2A.

In patients with autosomal dominant CAPS, nonsynonymous substitutions in the NACHT domain of NLRP3 result in even greater processing and secretion of IL-1β, about fivefold that in normal individuals (28). This relatively modest increase in IL-1β secretion causes patients severe local and systemic inflammation and in some cases renal amyloidosis and death (25–27). A similar severe autoinflammatory phenotype has recently been associated with a mutation of Nlrp1 in the Neut1 mouse (29). In contrast, nonsynonymous mutations in the common NLRP1 haplotype 2A are located outside the NACHT domain, result in ∼1.8-fold up-regulation of IL-1β processing in monocytes from healthy individuals, and are associated only with increased risk of autoimmune disease, rather than with frank disease. CAPS and similar diseases associated with increased IL-1β secretion are effectively treated with IL-1β blocking agents (32), supporting the concept that modestly increased IL-1β secretion is causal for disease and suggesting that similar therapeutic approaches may be beneficial in treating or even preventing autoimmune diseases in individuals who are genetically predisposed due to NLRP1.

These relatively common autoimmune diseases are inherited as complex traits, in which no single gene variant is sufficient to cause disease. Thus, although common NLRP1 variation is an important predisposing factor to autoimmune disease, the high-risk NLRP1 haplotypes are clearly not sufficient by themselves to result in disease. Rather, increased mature IL-1β may contribute to the pathogenesis of autoimmunity, perhaps by functioning as an “adjuvant,” facilitating presentation of autoantigens that may trigger or provide specificity for the autoimmune response. It will be of considerable interest to extend these studies to systematically compare IL-1β processing in males versus females homozygous for haplotype 2A, as ∼70% of autoimmune diseases occur in females.

Materials and Methods

Study Populations.

Vitiligo family probands comprised 114 unrelated individuals of self-reported EUR ancestry from North America or the United Kingdom. Each was an affected proband of a family with multiple cases of generalized vitiligo and at least one case of any of the other autoimmune diseases epidemiologically associated with vitiligo (33), previously used to establish linkage and association of NLRP1 with vitiligo (9). All cases met strict clinical diagnostic criteria for generalized vitiligo. Normal subjects were 117 unrelated individuals of EUR ancestry from North America with no history of any autoimmune disease, no recent illness, and not consuming chronic medications, including nonsteroidal anti-inflammatory drugs. This study was approved by the University of Colorado Denver Combined Institutional Review Board, and all donors gave written, informed consent.

NLRP1 DNA Sequencing.

A total of 161.5 kb spanning the NLRP1 region of chromosome 17p13.2 (chr17:5386564–5548062) was sequenced in the 114 unrelated EUR vitiligo probands. Primers were designed using Primer3 (http://primer3.sourceforge.net/webif.php), amplifying 517 PCR products sized 200–600 bp (Table S2), spanning all 17 exons and introns, 60.2 kb of 5′ flanking DNA, and 30.9 kb of 3′ flanking DNA. For each patient, 5 μg genomic DNA was used to prepare DNA sequencing libraries and was sequenced as described previously (34). Functional consequences of all exonic sequence variants were assessed using FuncPred (http://snpinfo.niehs.nih.gov/snpinfo/snpfunc.htm), SeattleSeq Annotation 131 (http://snp.gs.washington.edu/SeattleSeqAnnotation131/), and SKIPPY (http://research.nhgri.nih.gov/skippy/index.shtml).

Isolation of Monocytes and Cell Stimulation.

Healthy individuals homozygous for either NLRP1 haplotype 1 (n = 6) or haplotype 2A (n = 10) were identified, and peripheral venous blood was collected into heparinized tubes in the morning from each subject on three separate occasions over a period of at least 3 mo. Peripheral blood mononuclear cells (PBMC) were prepared by differential centrifugation of blood over Ficoll-Paque (Sigma-Aldrich). To isolate monocytes, PBMC were suspended at 5 × 10⁶ cells/mL in RPMI 1640 containing penicillin-streptomycin (Cellgro), and 100 µl of PBMC suspension was added to 96-well flat-bottom polystyrene plates and incubated at 37 °C for 1 h. Nonadherent cells were then removed, and the adherent monocytes were incubated in 200 μL RPMI 1640 (unstimulated), or in RPMI 1640 containing 1 ng/mL peptidoglycan-free LPS (Sigma), 10 μg/mL MDP, a combination of LPS plus MDP, or 1 × 10⁶ microorganisms/mL heat-killed S. epidermidis in triplicate wells. The release of LDH was determined using the LDH-Cytotoxicity Assay Kit II (BioVision). After 24 h, cell supernatants were collected and cells were lysed by adding 100 μL of 0.5% triton X-100. Supernatants and cell lysates were stored at –80 °C.

Additional detailed materials and methods are presented in SI Materials and Methods.