Autosomal dominant (AD) gain-of-function mutations in the NLRP3 gene are known to cause cryopyrin-associated periodic syndrome (CAPS). More recently, AD mutations in NLRP3 have also been linked to non-syndromic sensorineural hearing loss DFN34 as well as to Keratoendotheleiitis fugax hereditaria (1).

The first description of somatic NLRP3 mosaicism in a patient with chronic infantile neurological cutaneous articular syndrome (CINCA), also known as neonatal-onset multisystem inflammatory disease (NOMID) occurred almost 20 years ago (2). Since then, several studies (3–5) have confirmed the important role of mosaicism in NLRP3 mutation-negative CAPS patients. These were prompted by the clinical observation that up to 50% of CAPS patients with identical clinical features and response to anti-IL-1 treatment, show no mutation in NLRP3 by conventional DNA sequencing, suggesting the presence of either genetic heterogeneity, or low levels of somatic NLRP3 mutations (2, 6, 7). Reported frequencies of NLRP3 mosaicism in the literature vary from 10%–69% (3) of “mutation-negative” CAPS cases, whilst a more recent review established a general estimate of around 19% for NLRP3 mosaicism (4).

The frequency of mutant cells can be quite low, even lower than 5% in blood, yet still cause disease. Therefore, the most significant limitation for an adequate genetic diagnosis is the requirement of methods sensitive enough to pick up mutations present at very low read frequencies with statistical confidence (8). Advances in next-generation sequencing (NGS) methods such as targeted gene panels (TGP) with high depth coverage or amplicon-based deep sequencing (ADS) made these the methods of choice to detect somatic mosaicism by several groups (7–9). Advances in ADS technology could increase the success rate of genetic diagnosis for mutation-negative CAPS from 60 to 80%, which would greatly benefit the care of these patients by increasing diagnostic yield and help prevent potentially irreversible complications by allowing earlier targeted treatment with IL-1 blockade (8).

Although the phenotype of patients with somatic NLRP3 mosaicism is very typical of the disease, a few studies tried to ascertain if there were any differences in relation to CAPS patients with germline mutations (3, 9). Thus, patients with CINCA/NOMID and NLRP3 mosaicism were found to have a lower incidence of intellectual disability, and milder neurological symptoms overall following genotype matched comparison. One possibility to explain this could be the level of mosaicism in central nervous system cells or glial cells (3). Regarding Muckle-Wells Syndrome (MWS), the differences between patients with germline and mosaic mutations were slightly broader. MWS patients with mosaicism seemed to have a reduced incidence of AA-type amyloidosis when compared to patients with germline mutations, an increased incidence of recurrent arthritis, as well as older ages at the onset of the disease and that of sensorineural deafness. Moreover, absence of family history is a requisite in patients with mosaicism, whilst patients with germline mutations may have affected first-degree relatives. However, one of the most important and impactful differences was the significant delay in obtaining access to treatment with anti-IL1 drugs in patients with mosaicism, which was only achieved once the molecular diagnosis of mosaicism was secured (9).

The differences in clinical severity between mosaic and germline CAPS patients can be due to several factors, including the type of amino acid exchange, its location in the cryopyrin protein, the functional consequence of the mutation, as well as the level (percentage of cells) and tissue distribution of the mosaicism (7, 9). Previous mosaicism studies in paediatric patients with NLRP3 mosaicism showed presence of the mosaic mutation in comparable levels both in the myeloid and lymphoid lineages as well as affecting cells from ectodermic origin (3, 7, 10), unlike what has been described in adult late-onset CAPS patients (11–13).

In this study, we assessed the possibility of somatic mosaicism in a cohort of paediatric patients from a single UK centre with a clinical diagnosis of CAPS and no NLRP3 mutations identified through conventional DNA sequencing. In identified cases, additional mosaicism studies focused on tissue distribution, stability over time, genotype-phenotype correlation, and possibility of transmission to offspring. Given the difficulty of detecting low-level mutations, we have also tried to compare different sequencing methods regarding the ability for detecting somatic mosaicism.

Ten patients with a clinical diagnosis of CAPS who were NLRP3 mutation-negative by conventional Sanger sequencing were recruited to this study. Six patients had a CINCA/NOMID phenotype, whilst four had a MWS phenotype. Onset of symptoms occurred during the first year of life in all patients. Male sex was over-represented in this cohort with only two female patients. Patient 5 was found to have the germline p.G779V NLRP3 mutation, which was previously missed due to a restrictive Sanger approach that did not include exon 5 (exon 6 in the new updated transcript NM_001243133.2) (14). Patient 6 had the germline p.D512Y mutation in the NOD2 gene and in retrospect the phenotype was compatible with Blau syndrome.

Of this paediatric UK cohort of 10 patients who were initially considered as “mutation negative CAPS”, one patient was found to have a NOD2 mutation and an eventual diagnosis of Blau syndrome; one patient had a germline mutation in NLRP3 that was missed by an historic Sanger sequencing approach that only focused on exons considered to harbour mutation hotspots; and NLRP3 mosaicism was present in 4/10 (40%), confirming the important role of somatic NLRP3 mosaicism some CAPS patients. Thus, a confirmatory genotype was identified in 60% in this “mutation negative” series.

Two studies have been pivotal to the understanding of somatic mosaicism in CAPS by establishing evidence of unification of the intracellular and extracellular inflammatory cascades that results in amplification of the inflammatory response, thus explaining why even only a small proportion of mutated leucocytes can result in severe systemic inflammatory disease (23–25). It has been demonstrated that extracellular specks released during pyroptosis can act as an initial danger signal and bypass the sensor-activating portion of the inflammasome cascade, as well as recruit NLRP3 and caspase-1, thus acting as a functional extracellular inflammasome (23–25). Extracellular ASC specks have also been identified in the serum of patients with somatic NLRP3 mutations (23). Together, these reports demonstrate how mutations in only 4% of leukocytes could lead to a full-blown severe CAPS phenotype (25), since it is now apparent that the biology of the inflammasome is not restricted to the intracellular compartment.

The p.F566L mutation identified in Patient 1 has been previously reported in 2 unrelated patients with CINCA/NOMID syndrome in mosaic state with estimated allelic mosaicism of 11.5% and 14.6%, respectively (3). The p.E567K NLRP3 mutation found in Patient 2 has been previously described in a patient with MWS with an estimated mosaicism of 6.5% (26). Functional studies performed with these two variants confirmed their suspected pathogenicity through induction of THP-1 necrosis-like programmed cell death and ASC-dependent NF-kB activation (3, 26). The p.G564D mutation identified in Patient 3 has also been described in somatic form in a patient with late-onset CAPS and its pathogenic role was confirmed by functional studies showing increased ASC aggregates in patient's serum (13). The p.G307V mutation in Patient 4 had also been previously described in a CINCA/NOMID patients as either germline (22) or mosaic (3). Therefore, all these mutations fulfil several categories of evidence of pathogenicity according to the guidelines of the American College of Genetics and Genomics/ American College of Pathology for gene variant classification (27), as seen in Supplementary Table S1.

In this cohort we found mutations thus far only described in a somatic form and others described as germline in a more severe (CINCA/NOMID) form. This seems to support the notion that some mutations are so damaging that may be incompatible with life in a germline form, and that somatic mutations tend to be milder than if occurring as germline. Therefore, the clinical severity of a somatic NLRP3 mutation would be a function not only of its allelic frequency, but also of its impact on the function of the cryopyrin protein (9).

To assess the distribution of mosaicism, we selected samples representative of the three embryonic layers: blood and saliva (mesoderm), buccal swab, hair root and nails (ectoderm) and urine (endoderm/mesoderm) (28). As expected, a slightly different pattern was noted between this paediatric cohort and the adult late-onset patients reported in the literature. In late-onset adult patients, increased allelic frequencies were found in the myeloid compartment in comparison to the lymphoid compartment, inclusively with significant enrichment in the former seen in some patients, with relatively small allelic frequencies in other samples likely corresponding to leukocyte contamination. This distribution places the mutational event at the level of either a common myeloid progenitor or a pluripotent hematopoietic stem cell. In late-onset CAPS, a possible explanation to at least partially account for the difference between myeloid and lymphoid cell lineages could be an impairment of differentiation of the lymphoid lineage, with selective increase of the myeloid lineage, which has been known to occur in the elderly (12). In this paediatric, early-onset cohort, however, this was not the case: in our patients, similar allelic frequencies were found between myeloid and lymphoid cells.

It is important to note that in cases of mosaicism presenting in the first weeks of life, in theory, the mutational event could have occurred at any stage, from early embryonic development until a mutation occurring in a myeloid progenitor or pluripotent hematopoietic stem cell at birth or in the first weeks of life. In Patient 4 the somatic mutation was not detected in other tissues (buccal swab, hair, nails and urine) therefore likely placing the occurrence of the mutational event at the level of a pluripotent haematopoietic stem cell. In Patient 2, the presence of lower allelic frequencies of the somatic mutation in buccal cells and urine could likely derive from leukocyte contamination of the samples, although an earlier occurrence of the mutational event cannot be ruled out at this point. This could have practical clinical consequences for this patient, because if the mutation has also affected the gonadal tissue, it could potentially be transmitted to offspring in germline form.

We have confirmed that even very low levels of NLRP3 mosaicism may cause CAPS (the lowest being 1.9% in one case) as described previously (10). Patient 4 had a slight rise in mosaicism which eventually enabled the diagnosis, but without any obvious clinical impact to phenotype or therapeutic response. It is therefore possible that some patients who are NLRP3 mutation-negative even after investigation for somatic mosaicism, may actually have a mutation with extremely low frequency which could be under the threshold for detection of past (or current) methods. Thus, repeating a deep sequencing method at a later point in such patients may be warranted, either with a more sensitive method or to look for an increase in allelic frequency above the threshold for detection. However, significant increases in mosaicism frequency over time might have impact on treatment response as reported by Rowczenio et al. (13), although it remains to be seen if this will occur in this paediatric cohort. It has also been suggested that IL-1 blockade might also influence mosaicism levels. However, what has been documented thus far is a decrease in mosaicism levels following anti-IL-1 treatment (10, 12), probably due to treatment reducing the increased haematopoiesis caused by the IL-1 overproduction and consequently normalizing the total white cell and neutrophil counts.

The difficulties in detecting the somatic NLRP3 mutation in paediatric Patient 4 highlight the importance of choice of sequencing method as well as existing limitations despite the fantastic advances in NGS technologies in recent years. The comparison of results of different sequencing methods on mosaicism detection showed that TGPs appear to offer the optimal compromise between breadth and depth for the genetic diagnosis of these patients, as they can simultaneously identify mutations in several genes even at very low-levels of mosaicism. Genetic panels are now currently used as standard practice in many genetic laboratories, although many still use restrictive Sanger approaches. However, even in those using NGS gene panels, a survey found that only 17% of respondent laboratories were including routine testing for somatic mosaicism in their practice (29), even though consideration of low-level somatic mosaicism has been included in recent guidelines for best practice in the genetic diagnosis of autoinflammatory diseases (30). Furthermore, to the best of our knowledge, there is only one published study comparing sequencing methods in autoinflammation, which compared the effectiveness of a TGP to Sanger sequencing in the diagnosis of autoinflammatory diseases (31). Nevertheless, the tendency in recent years has been towards favouring methods with higher breadth.

In this series, if WES had been the only method used, at least half of this cohort would have remained without a genetic diagnosis due to comparative lack of sensitivity for the detection of low-level mosaicism. An interesting question in this regard would be to see if whole-genome sequencing (WGS) could offer advantages over WES. Two of our patients were previously recruited to WGS studies, however, in both cases there was no evidence of the presence of mosaicism in WGS data, even after careful manual review of the read alignment at the given genomic location (data not shown). Somatic mutations with an allelic frequency of 20% can be described as low-level in WGS projects (32), therefore it is not entirely convincing that even the mosaicism case with an allelic frequency of the mutation of 14.5% would be easily detected by WGS.

There are several still unanswered but potentially relevant questions to be addressed: namely the distribution of somatic mosaicism in additional non-haematopoietic tissues and different leukocyte populations; additional time points to assess mosaicism stability over time; and to assess the possibility of vertical transmission of the mutated allele to offspring.

The work described herein confirmed the importance of somatic NLRP3 mosaicism in the genetic diagnosis of paediatric cases of “mutation-negative” CAPS from a single centre UK cohort. Furthermore, mosaicism levels may vary over time in some patients and could be picked up later in life if repeat testing is undertaken. The comparison of sequencing methods suggests that TGPs may offer an advantage in this setting, whilst the diagnostic yield for mosaicism detection using WES or WGS may be lower.