Two-sample t-test test was used to compare continuous variables such as age between two patient groups. The Chi-square test with exact p value from Monte Carlo simulation was used to compare categorical variables such as sex. In addition, mean+/-SD were reported for continuous variables; column percentages were reported for categorical variables. Fisher’s exact test was used to compare prevalence of different genotypes between Group 1 patients and the control population, Group 3. The significance level is set at p<0.05 and all analysis was performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

A total of 63 adult patients with SAIDs were included in this study, among whom 44 (69.8%) carried combined variants in two or more genes, and there were 19 (30.2%) YAOS patients with characteristic clinical features and the specific NOD2 variants. All 44 patients in Group 1 were Caucasian and 93% were female; mean age was 44 ± 13 years, and disease duration 15 ± 13 years at the time of diagnosis. The latter underscores a prolonged diagnostic delay due to lack of recognition. The demographic, clinical, and laboratory data of patients in both groups are listed (Table 1), and these parameters were compared between the two groups. There were no statistically significant differences between the two groups except for a higher rate of chest pain in Group 2. All patients presented with a constellation of inflammatory symptoms, including recurrent fever, rash, arthralgias/distal leg swelling, gastrointestinal complaints, sicca-like symptoms, and eyelid swelling. Other notable symptoms were myalgias, oral ulcers, chest pain/pleuritis/pericarditis, asthma, and hearing loss. Most patients (80%) reported no family history of periodic fever syndromes.

Of the 44 SAID patients in Group 1, all patients carried NOD2 monoallelic or biallelic variants, as well as other gene variants. Most patients carried digenic variants, while oligogenic variants were found in a minority of patients. Among the digenic variants were NOD2/MEFV, NOD2/NLRP12, NOD2/NLRP3, and NOD2/TNFRSF1A in descending order of frequency. The oligogenic variants were NOD2/NLRP3/NLRP12, NOD2/MEFV/NLRP12, NOD2/MEFV/TNFRSF1A, NOD2/NLRP12/TNFRSF1A, and NOD2/MEFV/NLRP3 (Table 2). Most gene combinations were composed of low frequency/penetrance and rare variants, with the digenic variants in NOD2 and MEFV being the most common. To compare the distribution and frequency of these combined SAID-associated genetic variants in patients (Group 1) with those in the control population (Group 3), we used and analyzed the WES data of 2,952 subjects with European-American ancestry (Table 2). The combined gene variants identified in patients (Group 1) were either absent (23/44) or significantly lower in the ARIC control population. All 19 patients in Group 2 carried NOD2 variants, compound heterozygote mostly.

Clinical phenotypes with features suggestive of autoinflammatory disease were found for all patients at presentation. Following detailed phenotypic evaluations and phenotype-genotype correlations, 16/44(36%) of patients in Group 1 were diagnosed as YAOS, 3/44(6.8%) as atypical FMF, and the remaining with mixed diagnoses of two or more SAIDs, such as YAOS, FMF, NLRP3-AID, NLRP12-AID, and TRAPS. Having NOD2 as the denominator in all patients with combined variants, we asked if there were similarities between these patients with NOD2 ± other gene variants. Our results demonstrated no statistically significant differences in demographics, clinical phenotypes and laboratory results between the two groups except for lower frequency of chest pain in Group 1 (Table 1), suggesting similar but mixed clinical phenotypes among patients with NOD2 ± other gene variants. No significant internal solid organ damage or dysfunction was found for either group. However, some patients in both groups experienced frequent disease flares over a prolonged period of time, resulting in impaired ability to function physically and mentally. Nearly half of the patients in either group received IL-1 inhibitors (Canakinumab or Anakinra), many after trials of colchicine or sulfasalazine, with good response.

SAIDs are generally associated with variants in a single gene locus, though variant combinations in two or more relevant genes can be identified in individual patients as in our study. As a result, challenges arise for diagnosis and management. In the current study, genetic variants classified as variants of uncertain significance in NOD2 and MEFV genes based on the Infevers database were found to be the most frequently inherited https://infevers.umai-montpellier.fr/web/index.php.

What could be the clinical significance of combined variants in individual SAID patients who present with adult-onset disease? We did not observe significant differences between the groups in clinical phenotypes, which likely extends to disease course and therapy. As seen in Table 1, YAOS was diagnosed more frequently among patients with combined NOD2 and other SAID gene variants, and the majority of patents with YAOS carried compound NOD2 variants or rare variants. This might indicate more influences of the NOD2 variants on phenotypic expression in the genetic background containing multiple SAID genes. We stratified patients in both groups into subgroups: YAOS patients with NOD2 variants only (n=19), patients diagnosed with YAOS and NOD2/other gene variants (n=16), and patients with mixed or other diagnoses (n=28). Further analyses showed statistically significant differences between YAOS patients (n=35) vs patients with mixed or other diagnoses (n=28). Lower extremity swelling as a characteristic finding for YAOS was significantly higher in YAOS patients (69%) than patients with mixed or other diagnoses (43%, P value=0.0404), whereas headache was significantly higher in patients with mixed or other diagnoses (Table 3, Supplemental). In addition, chest pain was significantly higher in YAOS patients with NOD2 variants only than those with NOD2+other gene variants (Table 4, Supplemental). These data indicate that patients with NOD2+other gene variants may not have more severe diseases and poorer outcomes than those with variants in a single gene locus. Another possibility is that combined variants could be synergistic, antagonistic, or perhaps both, depending on the type of gene combinations.

To understand the role of these intracellular sensors in autoinflammatory diseases, individual genes(NOD2, NLRP3, NLRP12, and MEFV) and their downstream pathways are schematically depicted in Figure 1 based on literature review (1, 2). NLR interactions are complex, and NOD2 may function in concert with other immune sensors to regulate immune and inflammatory responses in different tissues. NOD2 shares a similar molecular structure with NLRP3 and NLRP12, and has an important role with regard to innate immune responses in gut. NOD2 recognizes a bacterial wall component, Muramyl Dipeptide (MDP), and functions in defense against microbial infection, in the regulation of the inflammatory process, and apoptosis (2). NOD2 mutations are linked to Crohn’s disease (40% patients), Blau syndrome, and YAOS. NOD2 signaling pathway involves receptor interacting protein kinase 2 (RIP2) that can be activated dependently on or independently of NOD2 (22). There is a cross-talk between NOD2 and Toll-like receptors (TLRs). NOD2 activation causes interferon regulatory factor 4(IRF4) expression, which in turn binds to tumor necrosis factor receptor associated factor 6(TRAF6) and RIP2, leading to NF-kB activation (23, 24). NOD2 mutations have been recently shown to cause loss of NOD2 cross-regulatory function involving IRF4 (25). In murine models, NOD2, together with NLRP3, caspase-1, Apoptosis-associated Speck-like protein containing a Caspase activation and recruitment domain (ASC), and RIP2, is required for MDP-induced IL-1 release (26). The NLRP3 inflammasome mediates intestinal inflammation in NOD2-deficient mice (27). NLRP3 also interacts with IRF4 (28).The function of NLRP12 is yet unclear in humans. In mice, NLRP12 has been shown to play a role in the proteasomal degradation of NOD2 and to promote bacterial tolerance and colonization by enteropathogens. NLRP12 suppresses MDP-induced NF-κB activation by targeting the NOD2/RIP2 complex. MDP tolerance is lost in murine monocytes deficient for NLRP12 (29, 30). There may be indirect interactions between NLRP12 and IRF4 (31). MEFV gene is an immediate early gene for interferon-gamma (32). These data suggest cooperative and interrelated roles of some NLRs in disease, and that IRF4 could be a key transcription factor for the orchestration of NLR functions in immune cells. Further study of the role of IRF4 within the context of NLR interactions in autoinflammatory disease is needed. In addition, we previously reported a combination of NOD2 and UBA1 mutations in a patient with an autoinflammatory disease and VEXAS syndrome, noting that both genes are involved in, or regulated by, the ubiquitin pathway (33). Taken together, our study has extended the understanding of the enormous complexity of the genetic influences that underlie autoinflammatory diseases whose clinical characteristics are often superimposable. In other words, similar clinical phenotypes in SAIDs may be caused in different genetic background, suggesting the biological complexity behind the disease phenotypes. Based on our study results and the literature data, it would be reasonable to perform higher-order molecular testing, as first-level testing may not suffice to reveal the underlying genetic mechanisms.

The majority of patients with combined gene variants carried both low-frequency/low- penetrance and rare variants. These variants are MEFV Glu148Gln (E148Q; rs3743930; MAF=0.07 in gnomAD, as high as 0.30 in Asian populations), NLRP3 Gln705Lys (Q705K; rs35829419; MAF=0.038), NLRP12 Phe402Leu (F402L; rs199985574; MAF=0.05), and TNFRSF1A Arg121Gln (R92Q; rs4149584; MAF=0.012), and selected NOD2 variants. These previously reported risk alleles have been associated with susceptibility to SAIDs or they play a role in modifying disease expression (10, 12, 30, 34–37). Notably, these low-frequency variants were found to be concurrent with rare ones in most patients. Furthermore, the distribution and frequency of these variant combinations in patients were either absent or significantly lower in the control population. These data suggest that these variant combinations are clinically significant. Genetic variants contributing to disease lie on a spectrum from rare alleles with large effect sizes to more common alleles with small effect sizes. There is a “gray zone” between these two extremes, as in this study, which is poorly defined with regard to terminology, classification and clinical reportability (38). These autoinflammatory diseases with low penetrance variants have been recently classified as the new category of “Genetically Transitional Disease” (GTD). GTD straddles the old distinction between monogenic and polygenic, where a large-effect mutation is necessary, but not sufficient, to cause disease (39). GTD emphasizes the key role of genetic background in modifying both penetrance and expressivity of a mutation or variant. This concept may also be important for the genetic counseling of these patients, supplementing the traditional interpretation of monogenic autosomal dominant or recessive diseases.

Genome-Wide Association Studies (GWAS) have revealed thousands of genetic variants associated with hundreds of human diseases. However, those that reach genome-wide statistical significance explain only a small fraction of heritability. Common variants may explain more than 50% heritability of many common diseases, including Crohn’s disease, type I diabetes, and multiple sclerosis (40). Overall, the composite effect of variants of small effect may have equal or greater impact than a rare pathogenic mutation with a large effect for more common human phenotypes (41). This concept and theory may in part help explain the implication of our findings, i. e., the coexistence of low penetrance variants in individual patients. Our data also suggest that a combination of genetic defects in different genes, which converge to a common pathogenic pathway, may have a synergistic effect and predispose individuals to SAIDs. For example, CAPS, NLRP12-AID and TRAPS have been classified as IL-1 mediated autoinflammatory diseases based on the patients’ response to IL-1 inhibitors (42, 43). The variant combinations in our study also support the conventional wisdom that genetic background influences both penetrance and phenotypic expressivity of gene mutations (39). Genetic background refers to all other related genes that may interact with the gene of interest to potentially influence specific phenotype in concert with environment (44). Our data may also serve as an example to understand the interactions between genes of interest, genetic background, and environmental or other factors in genetic diseases. Typical monogenic diseases are autosomal dominant disease (Huntington’s disease, HD) and recessive disease (Cystic fibrosis, CF). In these Mendelian diseases, there are slight but not predominant differences in female: male ratio, 54.5%/45.5% for HD (45), and 47.1%/52.9% for CF (46). One explanation is that primary mutations are highly penetrant and major players in these diseases. The candidate gene interactions with genetic background contribute to disease. For example, primary CFTR gene interacts with its modifier gene in the background contribute to CF (47). Unlike monogenic diseases, diseases like YAOS are mostly associated with reduced penetrance variants and are considered as GTD. Based on the GTD model of necessity and insufficiency, the candidate gene, for example, NOD2, genetic background containing certain SAID genes, and other factors such as sex hormones could play more important roles in the disease. Hormonal imbalance is known to cause inflammation or immune response (48), whereas androgen acts via its receptor on macrophages to suppress inflammation or immune response (24). Consequently, this could skew towards female predominance in the disease. Concretely, NOD2 variants are the denominator in all combinations of variants in our study; if NOD2 is considered as the candidate gene, other gene variants such as MEFV, NLRP3, NLRP12 andTNFRSF1A may be entertained as modifying alleles within genetic background. As noted in Table 3, among the overwhelming majority of 44 patients with combined variants, each carried a different combination of a limited number of the SAID genes to constitute a different but related genetic background, as analogous to the ten Arabic numerals from 0 to 9 for writing numbers or codes. Although most NOD2 variants are low-penetrant, their effect could be upwardly influenced by additional germline genetic alleles (39). As most of these gene variants are gain-of-function, they might have been under a positive evolutionary selection as they could provide better immune responses against various pathogens (49).Future genome-based studies in large cohorts of patients may help identify more SAIDs-associated modifying alleles. Acquired somatic mutations in the same genes may further contribute to disease expressivity in adult-onset autoinflammatory and autoimmune diseases.

The term “Digenic disease” refers to combinations of variants in two genes (50), and encompasses disorders in which both genes are required for expression or situations where a modifier gene significantly influences phenotype. The term “Oligogenic disease” refers to variant combinations in multiple genes. Based on these concepts, variant combinations in our study may be classified as “Digenic” in the majority of cases and “Oligogenic” in the minority. These variant combinations in our patients may be significant for the following reason. FMF is generally classified as a monogenic recessive disease, caused by biallelic missense gain-of-function mutations in MEFV. However, approximately 25% of FMF patients only carry monoallelic MEFV mutations (51). In fact, several studies have shown the monoallelic association with FMF (52–55), suggesting it is a dominant disease in some cases, where GTD model may apply. In our study, most patients with NOD2/MEFV variants were heterozygous for MEFV E148Q. Several studies of its pathogenic role in classic FMF were conducted with mixed results (35), but most literature data have favored its contributory role in autoinflammatory phenotypes that may or may not be classified as typical FMF as is the case with two recent independent Turkish studies (36, 56). An Israeli study showed that MEFV E148Q is likely a contributory genetic factor when coinherited with M694V (35). Based on our current study, we would assume that patients with carriage of both NOD2 and heterozygous MEFV mutations could mimic biallelic compound heterozygotes of MEFV mutations leading to autoinflammatory diseases with mixed phenotypic expressivity. This could explain a proportion of FMF patients with monoallelic MEFV mutations.

In our study, approximately 20% of patients reported a family history of similar symptoms. A co-segregation analysis of three families in our cohort found that symptomatic relatives shared digenic or oligogenic variants with the probands. Functional studies of some individual low penetrance variants involved in our study were previously conducted by others with ambigous results, specifically in regard to the MEFV E148Q variant. Cells expressing NLRP3 Q705K have mildly increased caspase 1 activity and cleavage, and such patients responded to an IL-1 inhibitor therapy (34). In another study, human monocytic cell lines transduced with Q705K produced significantly higher level of IL-1β and IL-18 than wild type, indicating a gain-of-function (57). In addition, we previously demonstrated abnormal NOD2 expression, NOD2 pathway activation, and a cytokine profile in patients harboring NOD2 variants, IVS8 + 158 and R702W (58).

Genotype-phenotype correlation may be readily apparent in patients with monoallelic or biallelic variants in a single gene, but can be challenging concerning combined variants from different genes. Additionally, NOD2 was the denominator in all the combined variants, and our study indicates that these NOD2 variants together with other relevant SAID-associated gene variants contribute to disease pathogenesis. We, therefore, have suggested a new term at the American College of Rheumatology Annual Meeting in 2022, mixed NLR-associated Autoinflammatory Disease (NLR-AID) to denote NLR involvement in the mixed diagnosis, which could be assigned an ICD10 code in the future for insurance billing and research purposes. Biologic therapy with IL-1 inhibitors are generally effective for mixed NLR-AID, YAOS, as well as hereditary periodic fever syndromes (59, 60). One would ask whether mixed NLR-AID could represent an independent entity that might be caused by a hidden pathogenic mutation. To clarify about it, whole exome or genome sequencing would be needed.

In conclusion, this is the largest single site-cohort of autoinflammatory disease patients with NOD2+ other gene variants. Most patients underwent a diagnostic odyssey during a prolonged “mysterious” illness. Unlike many common diseases for which there are readily available guidelines or consensus, SAIDs are rare diseases, and best evidence may come from case reports and case series (61). We provide rational interpretations and our experience with regard to diagnosis, classification, and management.

This is a single center study with benefit of uniformity and standardization of the study population. Relative to studies of common diseases, the sample size of our current study is small because the disease entity and associated clinical scenario is rare. We hope that more studies using similar cohorts of patients with these combined gene variants should be performed to replicate our findings.