A Constitutional Activating MET Mutation Makes the Genetic Link between Malignancies and Chronic Inflammatory Diseases

New high-speed technologies for genome sequencing have completely changed the study of cancer genetics, including constitutional genetics. However, they are usually restricted to identifying mutations in genes of high penetrance responsible for well-known hereditary syndromes (1). On the other hand, most genetic changes detected in whole-genome analyses are of unknown functional value. This is particularly true for genes of low to moderate penetrance.

Five percent to 10% of human cancers are considered to be linked to a constitutional genetic abnormality. Mutations in genes of high penetrance are usually identified using linkage analysis in families. For genes of lower penetrance, identifying mutations is more difficult. It is based on their cellular functions in families with cancer predisposition (2).

In the case of breast cancer, mutations in genes of high penetrance, including BRCA1 or BRCA2, are now identified in 30% of familial cancer cases (3). Genes of moderate penetrance implicated in DNA repair (CHEK2, BRIP1, PALB2, and ATM) have been characterized more recently, and genes of low penetrance are rarely identified because they require systematic studies of large populations with a low risk of cancer and possible polygenic involvement (4). Genome-wide association studies remain the predominant investigation to identify breast cancer susceptibility alleles, and potential new candidate genes (5). A recent meta-analysis of 1,059 publications found a moderate to strong association with breast cancer risk for 14 variants in 9 genes (6). However, the functional value of these 14 variants has not been demonstrated.

Next-generation sequencing applied to cancer has opened the field for individual approaches, through identification of new sporadic genomic abnormalities. Some of them, such as sporadic BAP1 mutations, have secondarily been considered as potential germline mutations in familial forms of renal cancer (7, 8), although their functional value has not so far been demonstrated.

Here, we propose an innovative individual approach by studying tissue samples from a young woman with an unusual association of cancer and other severe diseases, to identify constitutional mutations of low to moderate penetrance and demonstrate their functional value.

Whole-genome analysis was performed on breast cancer tumor cells, and on CD34⁺ cells sorted from bone marrow aspiration. In compliance with French bioethics law, the patient had been informed of the research use of the part of her samples remaining after diagnosis had been established, and did not oppose it. Informed consent was obtained according to French bioethics law (2004-800, 06/08/2004). Laser-microdissection (Palm) was used to select breast cancer tumor cells from paraffin-embedded tissue sections. CD34⁺ cells were obtained through magnetic sorting using anti-CD34 beads antibody (Miltenyi Biotec). DNA was purified using QIAmp DNA Mini Kit (Qiagen), quantified on Nanodrop, and qualified by electrophoresis.

Whole-genome analysis was performed using OncoScan-Express (Affymetrix), a molecular inversion probe (MIP)–based genotyping system that determines the genotype of 330,000 SNPs, copy number alterations, loss of heterozygosity (LOH) and somatic mutations on formol-fixed paraffin-embedded samples. MIP probes are circularizable oligonucleotides, the two ends of which carry two sequences complementary to two sequences on the genome separated by one nucleotide (where the variant to be genotyped is located). After hybridization of the genomic DNA, the reaction product is divided between two tubes, and two nucleotides are added to each tube (A/T and C/G). In the tube including the nucleotide complementary to the allele on the genome, the MIP probe is ligated and becomes circular. This structure is selected using exonucleases, and linearized. The products are amplified and hybridized onto an Affymetrix microarray for product identification (Supplementary Fig. S1). Oncoscan assay searches for 541 somatic mutations specific to cancer, with a coverage of over 200 tumor suppressors and oncogenes, and a median spacing of 1 probe per 0.5 kb for the top 10 “actionable” tumor suppressor genes, a median spacing of 1 probe per 2 kb for the top 190-plus actionable oncogenes, and a median backbone spacing of 1probe per 9 kb.

Here, data analyses for mutation detection were performed on a Nexus-7 software (BioDiscovery).

MET mutation was detected using three different methods: PCR high-resolution melting (PCR-HRM), Sanger sequencing, and restriction enzyme digestion.

PCR-HRM analyses of MET mutation in human samples (with forward primer 5′-GCCTATCCAAATGAGGAGTGTGT-3′ and reverse primer 5′-TCTGTTTTAAGATCTGGGCAGTGA-3′) were carried out on the the CFX96 Real Time System (Bio-Rad) on a total volume of 20 μL containing 5 μL of genomic DNA (20 ng), 15 μL of SsoFast EvaGreen Supermix 1× (Bio Rad), and 0.4 μmol/L of both forward and reverse primers. The PCR was performed with an initial denaturing step at 94°C for 2 minutes, followed by 45 cycles of denaturation (95°C for 5 seconds) and annealing (60°C for 10 seconds). After PCR, a postamplification melting curve program was initiated by heating to 95°C for 1 minute, cooling to 50°C for 1 minute, and continuously increasing the temperature by 0.2°C to finally reach 95°C. Each PCR run included a no-template control, and each sample was run in triplicate. Postamplification fluorescent melting curves were analyzed with the Bio-Rad Precision Melt Analysis software (Bio-Rad).

Sequencing of the shift fragment determined by PCR-HRM was performed using the Sanger method. Twenty microliters of PCR products were purified using affinity columns (Qiagen). Labeling was performed using BigDye-Terminator-v1.1-Sequencing-Kit (Applied Biosystems) in both forward and reverse directions. The reaction was run with an initial denaturing step at 94°C for 3 minutes, 25 cycles at 94°C for 10 seconds, and annealing temperature at 60°C for 20 seconds. BDX-terminator purified products were run on a 16-capillary automated sequencer (ABI-PRISM-3130xl-GeneticAnalyzer, Applied Biosystems). SeqScape-Software v 2.5 (Applied Biosystems) enabled nucleotide change determination.

For restriction enzyme digestion, about 10 to 15 μL of the amplified PCR product was digested with five units of Xmn1 restriction enzyme (Promega), and electrophoresis was performed on a 2% agarose gel (Invitrogen).

For microsatellite analyses, allelotyping of a physical distance of approximately 6.5 Mb between markers D3S1597 and D3S3611 was performed using 2 polymorphic microsatellite markers flanking VHL gene and listed in the UCSC database. The primer sequences used for PCR analyses are:

One primer of each microsatellite marker was end-labeled with one of the following fluorophores: HEX or CyS5 (Applied Biosystems). PCR was conducted as follows: for each microsatellite marker, equal amounts of DNA (20 ng each of constitutional and tumor-extracted) were subjected simultaneously in 2 different PCR tubes to one cycle of soaking at 95°C in a CFX96 Real Time System (Bio-Rad), then to 40 cycles of amplification at 95°C for 30 seconds and 60°C for 1 minute. The amplification was terminated by a final extension step at 72°C for 10 minutes. The resulting PCR products were analyzed on an Applied Biosystems 3130xl genetic analyzer using the GeneMapper software. This technique allowed the estimation of allele size and the quantitative evaluation of allele ratio. Homozygous markers were quoted “not informative”. Informative cases were scored as LOH when the intensity of the signal for one allele in tumor tissue specimens was decreased by >50% in comparison with allelic signal observed in normal tissue specimens. Allelic imbalance was scored when the signal for one allele was decreased >20% and <50%.

C57BL/6 mice were purchased from Charles River Laboratories at 3 to 4 weeks of age (Charles River Laboratories). All experiments were performed in accordance with NIH guidelines, and the European Union recommendations (2010/63/UE). The University Institute Board Ethics Committee for experimental animal studies approved this study (N°2012-15/728-0115).

Because the pR988C,c2962C>T MET mutation is located on a codon identical in human and mouse genes, we decided to establish knock-in transgenic mice to decipher the function of this point mutation.

MET human sequence at the locus of interest:

3′TCTTTAACAAGCTCTCTCTTTCTCTCTGTTTTAAGATCTGGGCAGTGAATTAGTTCGCTATGACGCAAGAGTACACACTCCTCATTTGGATAGGCTTGTAAGTGCCCGAAGTGTAAGTCCAACTACAGAGATGGTTTCAAATGAGT5′

MET mouse sequence at the same locus of interest:

3′TCTTTAACAAGCTCTTTCTTTCTCTCTGTTTTAAGATCTGGGCAGTGAATTAGTTCGCTACGATGCAAGAGTAGCAGAAATTGTTCGAGAAAGAAAGAGAGACAAAATTCTAGACCCGTCACTTAATCAAACGATGCTAGCTTCTC5′

c2962C_T MET mutation is highlighted in yellow, and differences between human and mouse are in bold.

The design of CRISPR/Cas9 plasmid was performed using two sgRNAs (http://crispr.mit.edu/, sgRNA 5 and sgRNA 8).

The sgRNA-Cas9 coexpression plasmid (CMV-T7-hspCas9-T2A-GFP-H1-gRNA linearized SmartNuclease vector) was obtained using the Precision Cas9 SmartNuclease System from SBI (System Bioscience). The plasmid was transfected into E.Coli using the One-Shot TOP10 Chemically Competent E. Coli Kit (Invitrogen). After kanamycin selection of transfected bacteria, plasmid DNA was extracted using the PureLink HiPure Plasmid Midiprep Kit (Invitrogen). Female C57BL/6J mice were injected with pregnant mare serum gonadotropin (PMSG) and human chorionic gonadotropin (hCG) with a 48-hour interval, and mated with male C57BL/6J mice. Cas9 mRNA (2 ng/μL) and sgRNAs (5 ng/μL for each sgRNA) targeting sgRNA5 or sgRNA8 were mixed and injected into the pronuclei of the one-cell embryos. After transfection, the zygotes were cultured in M16 Medium at 37°C, until blastocyst stage (around 100 cells) and transferred into pseudopregnant mice. This step was performed in the animal facilities of the Cochin Institute (MDC, CF). When founders were identified, they were transferred to the animal facility of the university institute of hematology for breeding. For each generation, a genotyping was performed on 3-week-old mice.

PCR-HRM and Sanger sequencing were performed for mouse genotyping as described above, using forward primer 5′-GTTTTGTTATTATCCGGGCTCTTCCTGT-3′ and reverse primer 5′-GATATTTCTCAGGATAGTAAACTGAATT-3′.

Restriction enzyme digestion was performed using XmnI enzyme as described above.

The number of off-targets was assessed using copy number variation.

Droplet Digital PCR (ddPCR) was performed using the QX100 ddPCR workflow system (Bio-Rad) to analyze copy number variation (CNV). The mix contained 20 ng of genomic DNA, 10 μL of ddPCR Supermix for probes (no dUTP; Bio-Rad), 1 μL of MET probes (Mm00192978_cn, Life Technologies), and 1 μL TFrc probes (TaqMan copy number Reference Assay, 4458366, Life Technologies) per well, with a final mix volume of 20 μL. The PCR mix was then poured into the middle wells of DG8 cartridges for the QX100 Droplet Generator (Bio-Rad), and 70 μL of droplet generation oil (Bio-Rad) was added to the bottom wells of the same cartridges. After sealing, the cartridges were placed in the Bio-Rad QX200 Droplet Generator System (Bio-Rad). The final volume of droplets in oil was approximately 40 μL. PCR was carried out on the CFX96 Real Time System (Bio-Rad), with an initial denaturing step at 95°C for 10 minutes, followed by 40 cycles of denaturation (95°C for 15 seconds), and annealing (60°C for 1 minute). A postamplification melting curve program was initiated by heating to 98°C for 10 minutes and then cooling down to 12°C. Each PCR run included a no-template control. The results of ddPCR were generated using QX100 Droplet Reader (Bio-Rad), and analyzed using QuantaSoft software (Bio-Rad). The ratio of pR988C MET-positive droplets to TFrc-positive droplets was calculated.

Blood and bone marrow analyses were performed at three different time points for each mouse. At euthanasia, all organs, including bone marrow, were systematically dissected for further tissue analyses. All tissue samples were similarly processed and cut into two parts: one was formaldehyde-fixed and paraffin-embedded, one was snap-frozen in liquid nitrogen and stored in Hôpital-Saint-Louis Tumorbank. All tissue sections were performed with HM-340 automatic microtome (Microm) at 5 μm and stained with hematoxylin–eosin (H&E). Two pathologists (A. Janin, G. Bousquet) performed the tissue analyses.

For auto-immunity assessment, the level of auto-antibodies in each mouse serum was measured using ELISA for anti-SSA, anti-SSB, anti-dsDNA, anti-CENT (Sygnosis), and anti-CCP (Interchim) according to the manufacturers' protocols.

Blood cell counts were obtained using MS9-5 Hematology Analyzer (Melet Schloesing).

For bone marrow analysis, cells aspirated from femurs of each mouse were suspended in PBS/SVF 2%. Red Blood Cells Lysis Buffer (Sigma Aldrich) was used before staining. Hematopoietic stem cell (HSC), multipotent hematopoietic progenitors (MPP), myeloid progenitors (MP), granulocyte monocyte precursor (PreGM), granulocyte-monocyte progenitor (GMP), megakaryocyte erythroid progenitor (MEP), Erythroid progenitor (EryP), and megakaryocyte progenitor (MkP) populations were stained with a panel of rat anti-mouse antibodies (all Sony Biotechnologies) including the PacBlue lineage cocktail (B220, Ter-119, CD3e, Gr-1, CD11b), PercpCy5.5 CD117 (Clone 2B8), PE/Cy7 Sca-1(Clone D7), BUV737 CD16/32 (Clone 2.4G2), Brilliant Violet 650 CD150 (Clone TC15-12F12.2), Alexa Fluor 647 CD105 (Clone MJ7/18), Brilliant Violet 605 CD41 (Clone MWReg30). The XMP population appears as an undefined myeloid progenitor. All experiments were performed using a FACs LSR Fortessa (BD Biosciences) according to the gating strategy shown in Extended Data (Behrens and al., 2016) and analyzed with FlowJo software (FlowJo, LLC).

The mice were treated with cabozantinib (Sellekchem), a MET inhibitor in the tyrosine kinase domain, at the dose of 30 mg/kg, by daily gavage over one month. The drug was diluted in 100 mL of NaCl with 2% of DMSO.

To assess treatment efficacy, we performed bone marrow aspirations and blood analyses at three time points: before treatment, after one month of treatment, and after a wash-out period of 2 weeks following the end of treatment.

Auto-immunity and bone marrow analyses were performed as described above.

Remaining MET-mutated transgenic mice older than 9 months were divided into two groups of three mice, one group exposed to external mutagenic factors for the development of skin cancers, and the second group exposed to the same mutagenic factors concomitantly with a daily administration of cabozantinib at 30 mg/kg by gavage. Six wild-type mice of the same age and from the same background served as control, and were exposed to the same mutagenic factors than mutated transgenic mice.

We started tumor induction by a single topical application of 20 μg of 7,12-dimethylbenz[a]anthracene (DMBA) (Sigma-Aldrich) in 0.1 mL acetone on the backs of mice. Two days later, we started UVB exposition at 380 mJ/cm² twice a week until the development of skin tumors, using Waldmann UV 109 B UV lamp. For the group of mutated mice receiving cabozantinib, the drug administration was initiated 14 days before we started tumor induction. Skin tumors were counted from the moment they appeared, and tumor growth was measured twice weekly in two perpendicular diameters with a caliper. Tumor volumes were calculated as following: V = L × l² ÷ 2, L being the larger diameter (length), l the smaller (width). At euthanasia, mice were dissected as describe above.

A 36-year-old woman was diagnosed with concomitant localized breast cancer overexpressing HER2, and polycythemia vera with JAK2 V617F mutation, together with definite rheumatoid arthritis (9).

A thorough analysis of the patient's family history showed a third-generation papillary thyroid cancer in a paternal aunt (III,9), and four patrilineal breast cancers and one colon cancer in the second generation (Supplementary Fig. S2). However, in our patient, no mutation was found in breast cancer predisposition genes of high penetrance (BRCA1, BRCA2, TP53, PTEN) or of moderate penetrance (CHEK2, PALB2 and ATM).

To look for mutations in other candidate genes of low to moderate penetrance in our patient (index-case), we chose an original approach, and performed genomic analyses for copy number variations and point mutations on laser-microdissected tumor cells from the breast cancer, and on CD34⁺ cells sorted from bone marrow aspiration. Using ONCOSCAN technology dedicated to formalin-fixed tumor samples (Supplementary Fig. S1), and a threshold of 9 for probability scores to filter the data, we identified 17 mutations (Table 1). Over 541 mutations tested, only pR988C, c2962C_T MET mutation was common to the two cell populations, with a probability score of 48.5 and 34.1 for breast cancer and CD34⁺ cells, respectively (Table 1 in bold).

When we focused on copy number variations, the most common events were LOH in noncoding sequences (Supplementary Table S1). One LOH, located in chromosome 3p25.3 with von Hippel Lindau (VHL) and FANCD2 genes, was common to the breast cancer and CD34⁺ cells. No common hotspot VHL mutation was identified in either breast cancer or CD34⁺ cells.

Using DNA-HRM and sequencing, we confirmed the presence of pR988C, c2962C>T heterozygous MET mutation in laser-microdissected breast cancer cells, CD34⁺ bone marrow cells, and constitutional DNA from hair bulb (Fig. 1A). Because we had the opportunity to test the thyroid cancer of our patient's paternal aunt (III;9 on Supplementary Fig. S2), we found the same pR988C, c2962C>T MET mutation, enabling us to confirm that this was a familial mutation from the paternal side. MET protein was also overexpressed by tumor cells in the thyroid cancer (Fig. 1B).

For the VHL locus, using microsatellite markers (Supplementary Fig. S3), breast cancer cells had VHL LOH, in accordance with ONCOSCAN results. CD34⁺ bone marrow cells and constitutional DNA from our patient had a profile of allelic imbalance, suggesting mosaicism for VHL LOH. In contrast, the microsatellite profile of the aunt's thyroid cancer (III;9) was normal, confirming that VHL abnormalities occurred de novo in our patient.

In a series of 105 sporadic breast cancers from Saint-Louis Hospital (Paris, France), we used DNA-HRM to detect pR988C, c2962C>T MET mutation. It was identified in one patient with a triple negative ductal carcinoma of the breast.

We also tested total blood DNA from 96 patients with JAK2 V617F-mutated polycythemia vera. Using DNA-HRM, the pR988C, c2962C>T MET mutation was identified in one 87-year-old patient. Surprisingly, she also had personal history of rheumatoid arthritis and cryoglobulinemia.

Because the pR988C, c2962C>T MET mutation is located on a codon identical in human and mouse genes, we decided to establish knock-in transgenic mice to decipher the function of this point mutation.

Using CRISPR-Cas9 technology, one cell stage embryos from C57Bl6 females were microinjected with ssDNA oligonucleotides as mutation donor, two types of sgRNA and two concentrations of Cas9 enzyme and sgRNA, either into the cytoplasm or into the nucleus. After incubation and cell division up to the blastocyst stage (around 100 cells), the blastocysts were tested for the integration of the pR988C, c2962C>T MET mutation. Using the enzymatic restriction method and DNA-HRM, we were able to ascertain whether our method was successful before embryo reimplantation. We could thus determine the optimal condition, with a ratio of 50 ng/20 ng of Cas9/sgRNA microinjected directly into the nucleus resulting in 19% to 25% of the embryos harboring the mutation (Supplementary Table S2).

Twenty-one days after reimplantation, a first generation of 17 mice was obtained, including 4 mutated animals (23.5%), 2 of them with both pR988C MET mutation and an indel (mice 8, 10, 12, and 19, see Supplementary Fig. S4). Second and third generations, obtained through heterozygous mating and mainly successful for Founders 10 and 12, led to the elimination of indel, and of potential off-targets for 2 mice (Supplementary Figs. S4 and S5B). A total of 23 mutated mice were thus obtained, characterized by genotyping using enzymatic restriction, DNA-HRM and sequencing. Copy number variation was performed for all mice with a ratio mutated MET/TFrc reference gene ranging from 0.5 to 12 (Supplementary Fig. S5). Unfortunately, we were not able to obtain successful mating after the third generation, either through heterozygous or homozygous coupling.

From the age of 9 months, all mutated mice had histologic signs of Sjögren syndrome, characterized by a dense periductal lymphocyte infiltrate (Chisholm score of III or IV) in the salivary and lachrymal glands. To confirm a potential autoimmune disease associated with the MET mutation, we used ELISA to monitor serum levels of anti-SSA and anti-SSB murine antibodies. Both were significantly higher in mutated mice than in wild-type mice (P < 0.05). In 11 of 12 mutated mice, we also found histologic signs of synovial hyperplasia and an inflammatory infiltrate in the distal but not proximal joints. Serum levels of anticyclic citrullinated peptide (anti-CPP also called ACPA) and of anti-dsDNA antibodies were significantly higher in the mutated mice than in the wild-type mice (P < 0.01; Fig. 2).

We found signs of thyroiditis in 8 of 12 mice. One 10-month-old mutated mouse (II:10) also developed a skin squamous cell carcinoma on the ear. Wild-type mice had no sign of Sjögren syndrome, arthritis or cancer.

Blood cell counts, obtained at 1, 6, and 9 months for 12 mutated mice, showed a significant increase in platelet numbers in the mutated mice compared with the wild-type mice from 1 month on (mean numbers at 9 months of 1,078.10³/mm³ (±137) and 775.10³/mm³ (±282), respectively, P < 0.05 (Fig. 3A). There was a trend for an increase in red blood cell numbers (mean numbers at 9 months of 10.7 × 10⁶/mm³ and 9.4 × 10⁶/mm³ respectively, P = 0.07; Fig. 3A).

Flow cytometry analyses of bone marrow were obtained for 8 mutated mice and 8 wild-type mice to assess percentages of different lineages (Fig. 3B and C). For LK populations, we found that the mutated mice had higher bone marrow cell densities than wild-type mice, with absolute cell counts of 24.1 × 10⁶ (±2.6) versus 9.4 × 10⁶ (±1.1), respectively (P < 0.01, Fig. 3D). These higher cell densities were found in all three myeloid progenitor lineages (Fig. 3E). The largest numbers were found for the granulocyte macrophage progenitors, GMP (Fig. 3F).

Tissue section analyses confirmed the higher bone marrow density in mutated mice than in wild-type mice. In addition, voluminous ovaries were found in two mutated mice. Microscope examination showed extramedullary hematopoietic proliferation with bone differentiation in these two voluminous ovaries (Fig. 3G).

Activation of the MET/HGFR transmembrane receptor induces MET dimerization, phosphorylation of tyrosines Y1234 and Y1235 in the tyrosine kinase (TK) domain, and of Y1349 and Y1356 in the carboxyl terminal region, leading to downstream signals (Supplementary Fig. S6; ref. 10). The highly conserved juxta-membrane domain (JM) acts as a negative regulator of signal transduction, enabling MET ubiquitination and degradation (11). Mutations in the JM domain could suppress this negative regulation (12), and thus lead to constitutive activation of MET.

The pR988C, c2962C>T MET mutation is located in the JM domain (Supplementary Fig. S6). Using immunohistochemistry on frozen liver samples and an anti-phosphoMET recognizing the phosphorylated tyrosine 1349 in the murine carboxyl terminal domain of MET receptor, we showed a higher pMET expression on mutated compared to wild-type liver samples (Fig. 4A). To decipher the functional value of pR988C, c2962C>T MET mutation, we administered a MET inhibitor of the TK domain, cabozantinib, to 4 mutated mice for 28 days. On successive bone marrow aspiration samples and blood analyses (day 0, day 28, and day 42), this MET pathway inhibition led to a decrease of GMP in the bone-marrow of the mutated mice and a drop in red blood cell counts and hemoglobin level, and to a normalization of serum concentration of anti-CCP and anti-SSB antibodies at day 28 in the mutated mice. These effects were followed by redundancy after the 14-day wash-out period (Fig. 4).

To sum up, CRISPR-Cas9 technology enabled us to establish pR988C MET-mutated transgenic mice that reproduced the autoimmune diseases and myeloproliferation found in our index-case. The use of an anti-MET drug validated the functional role of this mutation, and of the MET pathway activation in the maintenance of myeloproliferation, rheumatoid arthritis, and Sjögren syndrome.

We hypothesized that pR988C MET mutation is of moderate penetrance for cancerogenesis and thus requires additional mutational factors to induce cancers. Because one of the transgenic mice spontaneously developed a skin squamous cell carcinoma of the ear, we decided to expose 3 transgenic mice and 6 control wild-type mice to two additional mutagenic factors, including one chemical factor DMBA, and Ultraviolet B radiation UVB (Fig. 5A). After 2 months, the mean number of skin cancers per mouse was significantly higher in the MET-mutated mice compared with wild-type mice (7 vs. 1.8, P < 0.05). One of the MET-mutated mice also developed a thyroid carcinoma.

To confirm the role of pR988C MET mutation in activating MET pathway for cancerogenesis, we gave cabozantinib to 3 MET-mutated mice before exposing them to DMBA and UVB.

The regimen with cabozantinib significantly reduced the number of skin cancers when compared with MET mutated mice (P < 0.05, Fig. 5B).

Here, we demonstrated the feasibility of an individual approach in constitutional genetics: (i) by studying several samples from different lesions in a same young patient, (ii) by using high-speed technologies dedicated to whole-genome analysis, (iii) by using CRISPR-Cas9 technology to establish knock-in transgenic mice for a variant of interest, within few months. We were thus able to demonstrate the functional value of the pR988C, c2962C>T MET mutation, and its involvement in the genesis of myeloproliferative neoplasms and autoimmune diseases, an important link not previously identified.

Sjögren syndrome and rheumatoid arthritis are frequently associated (13), and familial aggregations have been reported (14). Genome-wide association studies have identified more than 10 susceptibility alleles of unknown functional value, such as HLA-DRB1, IRF5, STAT4, or BLK (15–19). Here, we demonstrated the role of constitutional genetics in the initiation and development of Sjögren syndrome and rheumatoid arthritis, with high phenotypic penetrance of pR988C, c2962C>T MET mutation in our murine model. On the basis of these experimental results, we inquired about the occurrence of these two auto-immune diseases in the family of our index-case, and found that her father (III,8) and her paternal grand-father (II,3) had rheumatoid arthritis (Supplementary Fig. S11).

To confirm the role of the MET pathway in the development of Sjögren syndrome and rheumatoid arthritis, we used the drug cabozantinib to inhibit the MET pathway in our mouse model. After 4 weeks of drug exposure, serum concentrations of specific auto-antibodies, anti-SSB and anti-CCP, were normalized. Previously, the role of HGF, the only ligand of MET (20), has been proposed in rheumatoid arthritis. In 136 patients, high HGF plasma levels were associated with severe joint destruction (21), and experimental HGF blockade using NK4, a HGF inhibitor, decreased arthritis lesions in mice (22). Here, we demonstrated the role of MET receptor activation in Sjögren syndrome and rheumatoid arthritis.

As a whole, these clinical and experimental results open the way for a new targeted therapy consisting in testing MET inhibitors, already available in clinical practice, among patients with rheumatoid arthritis, Sjögren syndrome, and MET pathway activation. A preliminary step is to assess the incidence of constitutional pR988C, c2962C>T MET mutation in cohorts of patients with rheumatoid arthritis or Sjögren syndrome, using cohorts like the French “ESPOIR” cohort, which includes patients with early arthritis followed up over a number of years (23, 24).

Patients with autoimmune diseases have a 20% increased risk of developing myeloproliferative neoplasms (25), including polycythemia vera as in our index-case. The chronic inflammation of autoimmune diseases could generate mutations in hematopoietic stem cells (26), particularly for JAK2 V617F mutation (27), what found in our index-case (see Table 1). In 96 other patients with JAK2-mutated polycythemia vera, we identified the pR988C MET mutation in blood DNA in one case, and this woman also had cryoglobulinemia and rheumatoid arthritis. The genetic origin of myeloproliferative neoplasms is established for somatic mutations in JAK2, CALR or MPL, which drive clonal proliferation, but only candidate polymorphisms have been identified through genome-wide association studies for constitutional genetics (28). Here, we demonstrated that the pR988C, c2962C>T MET mutation was sufficient to generate a myeloproliferation in all mutated mice. In addition, the involvement of the three myeloid lineages is an argument for a functional effect of this mutation at an early stage of myeloid progenitor commitment. As in Sjögren syndrome or rheumatoid arthritis (21, 29), the HGF level is increased in the serum of patients with polycythemia vera, independently from the JAK2 V617F mutation (30).

Our index-case developed rheumatoid arthritis and polycythemia vera in a context of constitutional pR988C, c2962C>T MET mutation. We were able to reproduce this phenotype in mice transgenic for the MET mutation. But our index-case had also developed breast cancer, and her paternal aunt had thyroid cancer with pR988C MET mutation.

MET is a proto-oncogene (31), and nonsense activating mutations have been found in the TK domain of patients with hereditary papillary kidney cancers (32–34). Somatic mutations have mainly been identified in the TK- and SEMA-domains of MET in childhood hepatocellular carcinomas, and in head and neck, gastric, and lung cancers (Supplementary Fig. S9; ref. 20). The pR988C, c2962C>T MET mutation is not located in the TK- or SEMA-domain but in the JM domain (35, 36). This mutation has been reported in thyroid and lung cancers, and to date is considered as sporadic (37, 38). Here, the analysis of several lesions from the same patient, combined with family investigation, led us to demonstrate that, in this case, the pR988C mutation of MET was constitutional. When we tested a series of 105 breast cancers without history of familial breast cancer, we found this MET mutation in one other case; unfortunately, we were not able to perform any familial investigation, as the patient had died. These familial investigations, mainly based on retrospective data analyses, can be difficult to perform. However, our study does raise the question of whether constitutional mutations contribute to cancers so far considered as sporadic.

We did not reproduce spontaneous breast or thyroid cancers, but the systematic pathologic study found lymphocytic thyroiditis in 14 of 23 transgenic mice when, in patients, chronic lymphocytic thyroiditis is associated with a 2-fold increased risk of papillary thyroid cancer (39). In addition, one of the transgenic mice developed a skin squamous cell carcinoma of the ear, as observed in mice transgenic for HGF with a strong activation of the MET pathway (40). We reproduced experimentally the occurrence of skin carcinomas in our transgenic mice, and reduced significantly the number of cancers by concomitant administration of a MET inhibitor demonstrating that MET activation is sufficient for increasing the susceptibility to skin cancer. The activation of the MET pathway may be too low for spontaneous carcinogenesis in our transgenic mice with pR988C MET mutation, and likewise in our patient. Additional molecular abnormalities are probably required, including mutations in other oncogenes or tumor suppressor genes. Mutations in the JM domain suppress MET receptor ubiquitination and degradation in human lung cancer cell lines (12, 20), but the level of MET pathway activation is not known. In another transgenic model for MET-activating mutation in the TK domain, considerable amplification of the MET gene was required for the development of breast cancer (20, 41). High serum levels of HGF were found with breast cancer development in mice transgenic for HGF (42). Autocrine/paracrine loops could also explain this increased activation of the MET pathway in cancer cells compared to normal cells, as described in multiple myeloma (43, 44).

In our index-case, we found a constitutional LOH including FANCD2 in addition to the pR988C MET mutation. FANCD2 is a gene implicated in DNA repair (45), as are BRCA genes, responsible for hereditary breast cancers. This reinforces the role of constitutional genetics in the development of cancers usually considered as sporadic.

Our approach combining the study of cell and tissue samples of several lesions in the same patient, the use of high-throughput genomic analyses, and of CRISPR-Cas9 technology enabled us to demonstrate for the first time a functional genetic link between Sjögren syndrome, rheumatoid arthritis, polycythemia vera, and cancers. It opens a field of application for this individual approach in the domain of constitutional genetics.

J. Kiladjian reports receiving commercial research grants from Novartis and is a consultant/advisory board member for Novartis, Celgene, and AOP Orphan. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. El Bouchtaoui, I. Loisel-Ferreira, C. Ferreira, A. Janin, G. Bousquet

Development of methodology: M. El Bouchtaoui, I. Loisel-Ferreira, C. Fedronie, G. Falgarone, G. Bousquet

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. El Bouchtaoui, M. Do Cruzeiro, C. Leboeuf, R. Ait El Far, M. Ziol, M. Espié, G. Falgarone, B. Cassinat, J.-J. Kiladjian, A. Janin, G. Bousquet

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. El Bouchtaoui, C. Leboeuf, G. Falgarone, J.-J. Kiladjian, J.-P. Feugeas, A. Janin, G. Bousquet

Writing, review, and/or revision of the manuscript: M. El Bouchtaoui, M. Do Cruzeiro, C. Leboeuf, I. Loisel-Ferreira, C. Ferreira, M. Espié, G. Falgarone, B. Cassinat, J.-J. Kiladjian, A. Janin, G. Bousquet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Leboeuf, G. Falgarone, G. Bousquet

Study supervision: A. Janin, G. Bousquet

Other (establishment of mouse model MET by CRISPR approach and maintenance of the line): M. Do Cruzeiro

We would like to thank Mrs A. Swaine for revising the English language. We also would like to thank Prof. Jessica Zucman-Rossi, Prof. Hugues de Thé, and Dr. Marie Dutreix for their helpful critical review of the manuscript. This work was supported by Inserm, University Paris 7.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.