Genetic predisposition to neural crest-derived tumors: revisiting the role of KIF1B

Objective We previously described a family in which predisposition to pheochromocytoma (PCC) segregates with a germline heterozygous KIF1B nucleotide variant (c.4442G>A, p.Ser1481Asn) in three generations. During the clinical follow-up, one proband’s brother, negative for the KIF1B nucleotide variant, developed a bilateral PCC at 31 years. This prompted us to reconsider the genetic analysis. Design and methods Germline DNA was analyzed by next-generation sequencing (NGS) using a multi-gene panel plus MLPA or by whole exome sequencing (WES). Tumor-derived DNA was analyzed by SnapShot, Sanger sequencing or NGS to identify loss-of-heterozygosity (LOH) or additional somatic mutations. Results A germline heterozygous variant of unknown significance in MAX (c.145T>C, p.Ser49Pro) was identified in the proband’s brother. Loss of the wild-type MAX allele occurred in his PCCs thus demonstrating that this variant was responsible for the bilateral PCC in this patient. The proband and her affected grandfather also carried the MAX variant but no second hit could be found at the somatic level. No other pathogenic mutations were detected in 36 genes predisposing to familial PCC/PGL or familial cancers by WES of the proband germline. Germline variants detected in other genes, TFAP2E and TMEM214, may contribute to the multiple tumors of the proband. Conclusion In this family, the heritability of PCC is linked to the MAX germline variant and not to the KIF1B germline variant which, however, may have contributed to the occurrence of neuroblastoma (NB) in the proband.


INTRODUCTION
We previously described a family in which predisposition to pheochromocytoma (PCC), segregates on 3 generations with a germline heterozygous nucleotide variant of KIF1B (c.4442G>A, p.Ser1481Asn) which encodes the kinesin-like protein KIF1B (1). KIF1B isoform  KIF1B is a molecular motor protein which participates in the transport of synaptic vesicle precursors and is essential for neuronal survival and differentiation (2). In vitro, the p.Ser1481Asn variant decreases the ability of KIF1B to promote the apoptosis of primary rat sympathetic neurons (3) and thus may facilitate tumorigenesis later on. Conversely enforced expression of KIF1B resulted in an induction of apoptosis of NB cells (4). Thus, the KIF1B neuronal pro-apoptotic effect combined with the mapping of KIF1B on chromosome 1p36, a region frequently deleted in PCC and NB (5), suggested that KIF1B might function as a tumor suppressor gene (TSG) in these diseases (3). However, in our kindred, we did not identify a loss of the wild type (WT) allele of KIF1B at the somatic level in the PCC or NB of the proband, deviating from the Knudson two hits theory (1). At that time, we thus hypothesized that the p.S1481N variant of KIF1B functions in haploinsufficiency in these tumors (1) but its exact mechanisms of action remain unclear (6).
Since our initial report, the large size KIF1B gene has been rarely incorporated in the PCC/PGL gene panels which are analyzed in patients with PCC or PGL by Next Generation Sequencing to identify familial tumours. Welander et al. (7) described one PCC patient with a germline variant of KIF1B classified as disease causing in their cohort thus representing a prevalence of 1.1%. The PCC had a sporadic presentation in this women who later presented with an endometrial carcinoma. Curras-Freixes et al (8)

Materials and Methods
Clinical Data. The pedigree of the family has been updated in figure 1. In brief, the proband developed a NB of the broad ligament at 17 months which was treated by surgery, radiation therapy and chemotherapy. At age 22 y, she developed a right PCC plus a ganglioneuroma at the site of the original NB and also an ileal schwannoma. At that time, she had hypertension and high normetanephrines levels (exact data not available). Six years later, the patient underwent adrenal surgery for a left PCC associated with a mature ganglioneuroma. At the same time, a welldifferentiated leiomyosarcoma arising from the mesosigmoid was detected and surgically removed.
At 39 y, several cutaneous metastases of the leiomyosarcoma were surgically removed. At 42 y, a 9cm moderately differentiated hepatic carcinoma was diagnosed and surgically removed. Finally, at 43 y an uterine leiomyoma, and two metastases (parietal and peritoneal) from the leiomyosarcoma were removed. Her paternal grand-father (I-1) had bilateral PCC at 70 y and her father (II-2) had a lung adenocarcinoma at 47 y and prostatic cancer at 54 y. The proband's youngest brother (III-3) presented at 31 y with a cardiomyopathy complicated by a Takotsubo's syndrome which led to the diagnosis of bilateral PCC. The proband's paternal uncle (II-3) was diagnosed at 56 y with an oligosymptomatic adrenal nodule exhibiting a high [ 18 F]-DOPA uptake at PET (positron emission tomography).

Custom endocrine tumors NGS panel.
In compliance with the french regulation, each patient gave his/her written informed consent before performing genetic testing which is an integral part of the patient's care. Moreover this protocol was reviewed and validated by the Ethical Committee (Comité ). Genomic DNA extracted from blood cells was fragmented with restriction enzymes; then digested DNA was hybridized to Haloplex probes which resulted in circularization of DNA fragments and sample indexing. Target DNA was captured with streptavidin coated magnetic beads, ligated and eluted before bridge PCR amplification of the libraries. After quantification of enriched target DNA, samples were pooled for multiplexed sequencing. NGS sequencing data were aligned to hg19 human reference and annotated using two independent bioinformatic pipelines [alignment using bwa v0.7.15-r1140 followed by best practices for germline variant detection using GATK v3.7 (11) and SeqNext V4.4 (JSI) (12)]. Data were filtered using an in-house database (DVD) to remove recurrent sequencing errors. Regions with coverage <30X were re-analyzed using Sanger sequencing.  Table S1. LOH of selected nucleotide variant was searched by SNaPshot analysis as previously described (16). PCR and extension primers details are available upon request. Extension products were analyzed on Applied Biosystems 3730 along with GeneScan 120LIZ molecular marker using the Genemapper software. In addition, visualization of the variant peaks on Sanger sequence traces was done using the Mutation Quantifier tool (Surveyor program, Softgenetics).
Immunohistochemical analysis. Fumarate Hydratase expression was assessed on paraffin-embedded tumors using an anti-FH antibody 1:1000 dilution as previously described (17).
In silico analysis. Prediction of the missense variant of MAX protein was carried out with the Phyre2 server (Protein Homology/analogY Recognition Engine V2.0). We compared the deduced human amino-acid 3D structure with the 3D resolved structure of the Homo sapiens MAX protein (99% sequence identity) using The PyMOL Molecular Graphics System (v2.0, Schrödinger, LLC). The SuperPose server v.1.0 (18) was used to estimate the structural homology, measuring the average distance between the backbones of superimposed proteins.
suggestive of PCC or adrenomedullary nodule. Plasma free normetanephrines were at the upper limit of the normal values. Pre-operatively, patient I-1 had high urinary normetanephrines (5x the upper limit range) and chromogranin-A levels (x3 the upper limit range). By contrast, patient II-2, who had a regular follow-up by PET imaging due to his lung and prostatic cancers, never demonstrated an adrenal uptake of [ 18 F]-Deoxy-Glucose. Moreover, his levels of metanepherines/normetanephrines were in the normal range at each follow-up, both elements being in disfavor of a PCC.
Since MAX behaves as a TSG (20), we analyzed the tumor DNAs of patient III-3 and identified loss of the MAX WT allele in both tumors (Fig. 3) suggesting that this variant is indeed responsible for the bilateral PCC in this patient. In contrast, no LOH of the WT allele of MAX was found in the 2 PCCs of the proband (III-1), using DNA obtained from three independent samples, FFPE samples from left and right PCCs (Fig. 3), and one fresh frozen sample from the left PCC (Fig. S2). The oldest of the samples (the right PCC) in fact showed complete absence of the variant allele (Fig. 3), which could be due to allelic dropout in the PCR, a well-recognized cause of errors in DNA from suboptimal samples (21).
Although the estimated proportion of tumor cells of the left PCC FFPE sample was high (Fig. 3), this estimate was not available from the bulk frozen specimen from this tumor nor the right PCC, so it is unclear to what extent normal cell admixture may have contributed to the allelic count. We also excluded, by Sanger sequencing, the presence of an additional somatic MAX mutation, which might have functioned as the second hit in the absence of LOH in the DNA extracted from the left PCC of the proband. We further examined the DNA from the three separate fragments from the left PCC to search for additional somatic mutations in 48 cancer genes using NGS. The three fragments displayed similar variant allele frequencies across these genes, suggesting that the left PCC was homogeneous with respect to both genetic and cell composition. Moreover, these fragments lacked areas of allelic imbalance, in favor of high level of normal cells in these fragments. Only a few VUS were detected (Table S2); however, no variant frequency suggestive of LOH was observed in these three samples.
Finally, FH expression evaluated by immunohistochemisty was conserved on the uterine leiomyoma and pelvic leiomyosarcoma of the proband (Fig S3), which strongly suggests the absence of somatic pathogenic variants of FH responsible for these 2 tumors in the proband.
Since the proband (III-1) and her father (II-2) shared the same genotype for MAX and KIF1B despite very different phenotypes (Fig.1), we considered that the proband may carry additional variants in known/new susceptibility genes that modified cancer predisposition, which might have been SMAD4, STK11 and TP53. Thus, we decided to focus on the nucleotide variants which were present in the germline of III-1 but absent in II-2 in agreement with our working hypothesis. One hundred twenty-five nucleotide variants were identified as unique to III-1 (Fig S1). Class 1 and 2 variants (benign or probaby benign) were filtered out using Varsome, leading to a list of 24 variants all classified as VUS (Class 3) (Table 1). After interrogation of several resources such as PUBMED (for a link between the gene of interest and cancer), UniProt (for information on encoded protein function), HGMD (for information on germline mutations currently identified), TCGA (for somatic mutations catalogue), Protein Atlas and CTEX databases (for detailed information on tissue expression), the list was narrowed down to 5 variants occurring in 5 genes (Table1 in red). KLH7 and PKM were good candidate genes for hereditary PCC since they encode proteins expressed in adrenals (23) and are mutated (though rarely) at the somatic level in PCC (TCGA PCPG). RIPK3, TFAP2E and TMEM214 were good candidates for the non-neural crest tumors since they are mutated in sarcomas (TCGA SARC), myomatous neoplasms and in hepatocarcinomas (TCGA LIHC) at the somatic level. No protein expression data was available for TMEM214, TFAP2E and RIPK3 in the Protein Atlas (23).
Sanger sequencing confirmed the presence of the five nucleotide variants in the germline of the proband and her mother but absence in her father as expected (Table 2), none were de novo. Patient III-3 was heterozygous only for the KLHL7 variant.
Tumor DNA from 8 different FFPE tumors from the proband (Table S1) were analyzed by Snap Shot to screen for LOH of the candidate genes. No LOH of KLH7 WT allele was found in the PCC of the proband nor in those of her brother III.3 (Fig. 3). Unfortunately, the data were not informative for PKM (not shown). Regarding TFAP2E, LOH of the WT allele was found in the DNA from the uterine leiomyoma whereas no LOH occurred in the parietal and peritoneal metastasis of the leiomyosarcoma (Fig. 4). By contrast, a LOH of the WT allele of TMEM214 was found in the parietal and peritoneal metastasis of the leiomyosarcoma (Fig. 4) whereas the data were not informative in the hepatocarcinoma (not shown).

DISCUSSION
The occurrence of a bilateral PCC at a young age in a relative (patient III. The interpretation of the pathogenesis of the PCCs occurring in the proband and her grandfather is more intricate. They both have the MAX p.Ser49Pro plus the KIF1B p.Ser1481Asn germline variants. We were unable to identify LOH of KIF1B (1) or MAX WT alleles in the proband's PCC despite the analyses of multiple separate fragments from the most recently removed PCC which was not fixed in Bouin's reagent known to be deleterious for DNA (21). However, as discussed above, our additional data on this tumor does not allow us to rule out the contribution of high levels of nontumoral tissue to the lack of detectable LOH. Although we cannot exclude that both KIF1B and MAX may contribute to the PCC phenotype or to the clinical variability in this family, KIF1B pathogenic variants have rarely been described in patients with PCC/PGL since our initial report in 2008 (1,7,8). Only one, a p.Tyr835Cys variant, was reported by Welander et al. in a 54-yr old woman with a unilateral PCC and an endometrial carcinoma (7). This patient had no germline variants in any of the 11 other major susceptibility genes for PCC/PGL but no somatic LOH of the wild-type allele was found, so preventing any definitive conclusion on the pathogenic relevance of this novel variant.
Given the phenotypic variability of this family, with multiple non-PCC/PGL cancers in patient III.1, and to evaluate the possibility that other susceptibility events were at play, we performed WES on the germline DNAs from patient III.1 and II.2. Our working model was that patient III.1 had de novo or maternally-inherited mutations in new/known susceptibility genes for hereditary cancers or, alternatively, in "modifier genes" which could have an impact on cancer-promoting pathways (24).
These additional nucleotide variants, combined with the KIF1B and MAX germline variants, might explain the very severe phenotype of this patient. WES excluded a pathogenic mutation in other genes predisposing to familial PCC/PGL and also in 30 other hereditary cancer susceptibility genes (22). Among the 125 variants which were private to the proband most were classified as benign or probably benign and were filtered out; the 24 remaining variants were all classified as VUS. Based on bibliographic data, tissue pattern of expression, biological function of the encoded protein, and a catalogue of somatic mutations, we narrowed down the list to 5 variants in 5 different genes which may be implicated either in the pathogenesis of PCC (KLHL7 and PKM) or sarcomas (RIPK3, TMEM214 and TFAP2E). We did not detect a LOH of the WT allele of KLHL7 or PKM in the proband's PCC. Thus, we cannot assign any causality in the pathogenesis of the proband's PCCs to one of these variants.
We propose that in this family the genetic susceptibility to PCC/PGL is linked to the MAX nucleotide variant which is however associated with an incomplete penetrance since patient II-2 did not develop any symptomatic adrenal lesion. The KIF1B isoform  p.Ser1481Asn variant, which is partly defective in the apoptotic culling of neural crest progenitors, may also contribute to the occurrence of NB in the childhood of patient III-1, similar to other rare observations in NB (3). Finally, the leiomyosarcoma and hepatic carcinoma of the proband, and her father's lung adenocarcinoma may be MAX and KIF1B independent. The involvement of TFAP2E and TMEM214 variants is an attractive hypothesis but the pathways acting in disease development remain to be determined.