Recurrent candidiasis and early-onset gastric cancer in a patient with a genetically defined partial MYD88 defect

Gastric cancer is caused by both genetic and environmental factors. A woman who suffered from recurrent candidiasis throughout her life developed diffuse-type gastric cancer at the age of 23 years. Using whole-exome sequencing we identified a germline homozygous missense variant in MYD88. Immunological assays on peripheral blood mononuclear cells revealed an impaired immune response upon stimulation with Candida albicans, characterized by a defective production of the cytokine interleukin-17. Our data suggest that a genetic defect in MYD88 results in an impaired immune response and may increase gastric cancer risk.


Introduction
GC is a multifactorial disease with a low survival rate in which both genetic and environmental factors are involved. Infection with Helicobacter pylori is an important risk factor for GC [1]. Approximately 1-3 % of GC cases fulfill the criteria for Hereditary Diffuse Gastric Cancer (HDGC; MIM 137215), a cancer predisposition syndrome characterized by earlyonset diffuse GC. In approximately 20 % of the families fulfilling the HDGC criteria a germline mutation can be found in the CDH1 gene encoding the E-cadherin protein [2][3][4]. Recently also CTNNA1 and MAP3K6 are described as putative gastric cancer predisposition genes [5,6], but in the majority of cases the underlying genetic cause remains unknown. In this study we report the case of a 23-year-old woman with diffuse-type GC and recurrent candidiasis, in whom we identified a novel germline homozygous MYD88 variant.

Materials and methods
Genetic analysis of the 23-year-old gastric cancer patient Genomic DNA was extracted from peripheral blood cells and hybridized to Affymetrix SNP6.0 arrays (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's protocol. The genotype was generated using the Birdseed analysis software incorporated in the Affymetrix Genotyping Console v2.1 (Affymetrix). Detected copy number variants (CNVs) were compared to a set of healthy controls, as described previously [7], to exclude regions of normal variation.
Massive parallel whole-exome sequencing was performed after DNA enrichment using the human SureSelect 50 Mb kit (version 2, Agilent Technologies, Santa Clara, CA, United States) on a 5500XL SOLiD platform (Life Technologies, Bleiswijk, The Netherlands). Reads were mapped to the hg19 reference genome using SOLiD BioScope software (Life Technologies) [8]. Variants were annotated using an in-house annotation pipeline, as described previously [8]. Briefly, from our total set of variants we selected high-confidence (C5 variant reads and/or C20 % variant reads) non-synonymous variants with a SNP frequency in dbSNP (v132) below 1 % that were present at most once in our in-house variant database containing 2096 exomes (the majority of European ancestry) [8].

Cohort screening for MYD88 variants
Our cohort of gastric cancer patients consisted of 126 additional CDH1 mutation-negative index patients that meet one of the 2010 HDGC criteria, with the exception that all types of GC histology were included [12]. Germline DNA of these patients was analyzed for mutations in MYD88 (NM_001172567.1) using either Sanger sequencing (n = 41), ion semiconductor sequencing (n = 63) (PGM, Life Technologies) or whole-exome sequencing (n = 22, see above). A cohort of 183 healthy individuals from Turkish or Pakistani descent were analyzed for the MYD88 variant using Sanger sequencing.
For Sanger sequencing the full coding sequence of MYD88, including splice junctions, was amplified using polymerase chain reaction (primer sequences and PCR conditions are available on request) and screened for mutations using Big-Dye terminator sequencing (BigDye Terminators (v 1.1) Applied Biosystems, USA). Analysis was performed on an ABI 3730 DNA Analyzer (Applied Biosystems). Subsequently, data was analyzed for variants using Chromas Lite (Technelysium, South Brisbane, Australia).
Ion semiconductor sequencing was performed using a custom designed multiplex Ion AmpliSeq TM PCR primer panel (Life Technologies) according to the manufacturer's protocol. Briefly, for the library preparation, four DNA samples from our gastric cancer patient cohort were pooled equimolarly and 10 ng of DNA was used for AmpliSeq amplification. Each pool was barcoded with Ion Xpress Barcode adapters (Life Technologies). The barcoded libraries were purified using Agencourt AMPure XP beads (Beckman Coulter Genomics, High Wycombe, UK), pooled and diluted for use in a 200-bp amplification run on an OneTouch emulsion PCR system (Life Technologies). Before loading onto the 316 chip, sequencing primer and polymerase were added to the final enriched spheres. SeqNext software (JSI Medical Systems) was used for mapping and analysis of the data. All amplicons were analyzed at a coverage of at least 1509. The identified variant was validated in the individual DNA samples of the pool using Sanger sequencing as described above.

Immunological assays
The cytokine production capacity was assessed as previously described [13]. Briefly, venous blood was collected into EDTA tubes and primary blood mononuclear cells (PBMCs) were isolated by density centrifugation of blood diluted 1:1 in PBS over Ficoll-Paque (Pharmacia Biotech AB, Uppsala, Sweden). Cells were washed three times with PBS and resuspended in RPMI 1640 (Dutch modified) supplemented with 50 mg/L gentamicin, 2 mM L-glutamine, and 1 mM pyruvate. Cells were counted on a Coulter Counter Z (Beckman Coulter, Mijdrecht, the Netherlands) and adjusted to 5 9 10 6 cells/mL. PBMCs (5 9 10 5 ) in a 100 lL volume were added to round-bottom 96-well plates (Greiner, Alphen a/d Rijn, the Netherlands) and incubated with either 100 lL of culture medium (negative control) or one of the stimuli described for 24 h (TNFa, IL-6), 48 h (IFN-c) and/or 7 days (IL-17 and IL-22). Cytokine production was tested after stimulation of PBMCs with: lipopolysaccharide (LPS) derived from E. coli (10 ng/mL), heat-killed C. albicans (10 5 microorganisms/mL), heat-killed H. pylori (10 7 microorganisms/ mL), heat-killed S. aureus (10 7 microorganisms/mL), recombinant IL-1b or IL-23 (100 ng/mL), and recombinant IL-12 or IL-18 (10 ng/mL) after 24 h and 7 days. 10 % human pooled serum was added when PBMCs were incubated during 7 days. After incubation the supernatants were stored at -20°C until assay. Experiments were performed in duplicate.

Results
Early-onset gastric cancer patient with extensive infectious phenotype The patient first presented with pain in the left upper abdomen at the age of 23 years. She was diagnosed with an H. pylori infection, for which eradication treatment was given. As the abdominal pain continued, a gastroduodenoscopy was performed, which revealed a gastric ulcer. Histopathological analyses of biopsies showed a signet ring cell carcinoma and moderate chronic active, non-specific gastritis (Fig. 1a-c). The patient was treated with neoadjuvant chemotherapy resulting in almost complete regression of the gastric tumor at the time of surgery (Fig. 1d).
Because of the early-onset diffuse GC the patient was referred for genetic counseling. Family history revealed distantly related parents of Kurdish descent, six unaffected siblings and no other relatives with GC. Her medical history revealed that she lost her fingernails (onycholysis) at the age of two years. Throughout her life, she suffered from recurrent vaginal Candida albicans infections and recurrent fungal infections, in particular dermatophytic onychomycosis, of which at clinical examination pigmented rash could be observed in her neck, between her breasts and at the rim of her fingernails. Her medical history was negative for allergies and pyogenic infections.

Identification of a novel homozygous missense variant in MYD88
The patient developed diffuse-type GC at a very young age and therefore meets the CDH1 testing criteria [12]. CDH1 germline mutation analysis was performed, but no germline CDH1 alterations were found by sequencing analysis of the entire open reading frame, including intron-exon boundaries, or Multiplex Ligation-dependent Probe Amplification (MLPA) of all exons. Therefore, further research on possible underlying genetic aberrations was performed. Copy number profiling using SNP6.0 arrays showed no significant copy number variations, but revealed large areas with loss of heterozygous calls, in agreement with parental consanguinity (data not shown).
Subsequently, whole-exome sequencing on germline DNA from the patient was performed. No putative pathogenic mutations in other cancer predisposition genes among which CTNNA1 and MAP3K6 were detected. The heterozygous missense variant in PMS1 that was validated using Sanger sequencing (Table 1) probably does not contribute to the phenotype of the patient. Because the parents were distantly related, we focused our data-analysis on homozygous variants with a variant percentage C80 % (n = 17). Five variants that were putatively pathogenic (one protein-truncating variant and four missense variants) were validated using Sanger sequencing, of which the missense variant (c.712C[T, p.(Arg238Cys)) in the myeloid differentiation primary response 88 (MYD88) gene was considered most relevant. This variant affects a moderately conserved nucleotide (phyloP 3.76) of a highly conserved amino acid in Recurrent candidiasis and early-onset gastric cancer in a patient with a genetically defined… 291 the Toll/interleukin-1 receptor (TIR) homology domain of the gene and is predicted to be damaging to protein function using in silico prediction programs. For the list of all validated variants and occurrence of variation in the corresponding genes in control databases, see Table 1.
Variation in the MYD88 gene in healthy controls and an additional cohort of gastric cancer patients is rare To assess the variation in MYD88 in control databases, we extracted all the germline variation in the coding region from MYD88 from a second in-house dataset and the Exome Variant Server (EVS) [14]. Furthermore, we used The Cancer Genome Atlas (containing 220 gastric adenocarcinomas) to determine the amount of somatic variation in the MYD88 gene [15]. Germline and somatic variation in the coding region of MYD88 is rare with less than 0.

Immunological assays reveal functional relevance of the MYD88 variant
Since MYD88 plays a central role in the immune response, we analyzed the cytokine production of peripheral blood mononuclear cells (PBMCs) of the patient, stimulated with heat-killed C. albicans and H. pylori. As expected based on the infections seen in patients with MYD88 deficiency [16], a specific defect in the production of Th17 cytokine IL-17 (Fig. 2, left panel) was observed upon re-stimulation of the patients' PBMCs with C. albicans. In healthy individuals that have been exposed to H.pylori, an enhanced IL-17 production is observed after re-stimulation of the PBMCs with this pathogen [17]. When we performed this assay for the patient, no difference was observed compared to controls (Fig. 2, right panel). The immune response to Toll-like receptor stimuli and S. aureus was normal (data not shown).

Discussion
In this study we report the case of a young woman with diffuse-type GC in which we did not find a germline mutation in the known cancer predisposition genes like  [18]. In mice it was shown that MYD88 deficiency leads to an early and rapid development of H. pylori-induced gastric dysplasia [19].
GC was not reported in the twenty-four patients described with MYD88 deficiency, of which the oldest patient was aged 20 at time of publication [16]. However, 'classic' MYD88 deficiency has a high mortality at an early age, and therefore patients likely die too young to develop GC [20]. In contrast to the functional analyses described for these MYD88 deficiency patients, our assays indicate that our patient suffers from a milder form of MYD88 deficiency, which matches her clinical phenotype and is probably explained by the type of MYD88 mutation, which to the best of our knowledge was not described before.
Therefore, we hypothesize that, in our patient, the partial MYD88 defect causes an impaired immune response, resulting in recurrent C. albicans infections and nonspecific gastritis increasing the risk of signet ring cell carcinoma of the stomach wall. Whether GC is more frequently diagnosed in patients with recurrent fungal infections is unknown, but it is known that patients with simultaneous C. albicans and H. pylori infection have an increased risk of developing gastric ulcers [21]. C. albicans is able to degrade E-cadherin in vitro [22], which may be a tumor initiating event in the development of gastric signet ring cell carcinoma similar to E-cadherin inactivation by mutations in its gene CDH1. In line with our hypothesis that impaired immune response might lead to an increased risk of gastric cancer, the co-occurrence of C. albicans and signet ring cell carcinoma of the stomach was previously described in a patient with acquired immune deficiency syndrome [23], who as a group have a 1.5-fold increased risk to develop GC [24]. Moreover, a tenfold increased GC risk is seen in patients with common variable immunodeficiency disorders (CVIDs) [25].
Due to neoadjuvant chemotherapy the patient received prior to surgery, no molecular analyses were possible on resection material. Therefore, we cannot exclude that, although this has not been described for other DGC patients thus far, two somatic mutations in either CDH1 or one of the recently described candidate genes CTNNA1 or MAP3K6 [5,6] are causal for the DGC development. The fact that the cancer has arisen by chance can also not be excluded, but based on the extremely young age of the patient we consider this very unlikely.
No additional germline mutations in MYD88 were found in our cohort of 126 patients suspected of hereditary gastric cancer. This implies that germline mutations in MYD88 in GC patients are rare. Rather than screening complete cohorts of patients for this type of mutations, we suggest more thorough anamnestic analysis of GC patients to unravel whether a (partial) immunodeficiency could be causal for GC development.
In conclusion, we identified a functionally relevant homozygous missense variant in MYD88 and a defective Th17 response in a patient with recurrent infections and early-onset gastric cancer. The missense variant in MYD88 is the likely cause of persistent fungal infections that ultimately may have led to the development of early-onset gastric cancer in this patient. Future research is needed to unravel whether such immune deficiencies are a more common genetic risk factor for gastric cancer development.