Novel pathogenic alterations in pediatric and adult desmoid-type fibromatosis – A systematic analysis of 204 cases

Desmoid-type fibromatosis (DTF, aggressive fibromatosis) is a non-metastasizing mesenchymal neoplasm of deep soft tissue with a tendency towards local recurrence. Genetic alterations affecting canonical Wnt/β-catenin signaling are reported in the majority of DTF. While most sporadic DTF harbor somatic mutations in CTNNB1, germline mutations in adenomatous polyposis coli (APC) are known to occur in hereditary DTF types (FAP, Gardner-Syndrome). Additional single nucleotide variants (SNVs) in AKT1 (E17K) and BRAF (V600E) were reported in pediatric DTF with potential clinical implications. We performed targeted next-generation sequencing (NGS) in a large cohort of 204 formalin-fixed DTF samples, comprising 22 pediatric cases (patients age ≤18 years). The mutational status was correlated with clinicopathological characteristics. Overall, deleterious CTNNB1 mutations were detected in 89% of DTF, most frequently affecting the serine/threonine phosphorylation sites T41 and S45 of β-catenin. While the T41A CTNNB1 mutation was significantly more often identified in the mesenterial localization, DTF originating from extra-intestinal sites more frequently harbored the S45P CTNNB1 alteration. Beyond common mutations in CTNNB1, additional SNVs were demonstrated in 7% of the DTF cohort and in 18% of the pediatric DTF subgroup. The mutational spectrum included deleterious mutations in AKT1 (G311S/D and T312I), ALK (R806H and G924S), AR (A159T), EGFR (P848L), ERBB2 (H174Y), IDH2 (H354Y), KIT (V559D), RET (T1038A), SDHA (R325M), and SDHD (R115W), as characterized by in silico prediction tools. In conclusion, our study indicates that DTF may harbor a broader mutational spectrum beyond CTNNB1 mutations, comprising targetable alterations including the herewith first reported imatinib-sensitive KIT V559D mutation in DTF.


Results
Spectrum of CTNNB1 alterations in Dtf samples. As pathogenic mutations in the CTNNB1 gene can be responsible for the constitutive activation of the canonical Wnt/β-catenin signaling cascade, we analyzed the entire CTNNB1 coding region by targeted NGS followed by validation via Sanger sequencing. Overall, the complete cohort of 204 DTF cases was successfully analyzed (Table 1 and Supplementary Table S1). In total, deleterious CTNNB1 mutations were detected in n = 181/204 (88.7%) DTF samples, with a minor fraction detected in the pediatric (77.3%) compared to the adult (89.9%) subgroup ( Fig. 1A and Table 2). All four tumors in patients with familial adenomatous polyposis (FAP) analyzed in this study, were assigned to the CTNNB1 wild type group (n = 23/204; 11.3%). The majority of deleterious alterations in the CTNNB1 gene were restricted to the serine/  1B and Table 2). Detected allelic frequencies were in the range from minimal 6% (S45P) to 13% (T41I) and maximal 34% (T41I) to 58% (T41A) (Fig. 1C). Less     www.nature.com/scientificreports www.nature.com/scientificreports/ of the mutational subtype, recurrences were reported in 18 cases, occurring in extra-intestinal localization in 17 cases.
In three cases, more than one missense mutation was detected in CTNNB1 (Supplemental Table S1). A 26-year old female with a DTF localized at the trunk (extra-intestinal) harbored both a S45T and a S45Y mutation in CTNNB1 (case #26). The combination of S45F and S45T CTNNB1 mutations was detected in a 32-year old female patient with abdominal DTF of unknown size (case #149). A combination of uncommon S33T and G34R CTNNB1 alterations was detected in a 43-year old female patient with a 4.8 cm abdominal DTF (case #119). No additional non-CTNNB1 mutations were detected by NGS in these cases.
To the best of our knowledge, two uncommon alterations were detected in CTNNB1: (I) S33L (97_98delin-sCT) in an intra-abdominal DTF of a 60-year old male without additional mutations (case #181) and (II) S45KA in a 74-year old male with an abdominal 10.5 cm DTF (case #51). Both variants were predicted as potentially deleterious by in silico analysis.  Table S3). In contrast, all five variants identified in FGFR3, MET and PDGFR were classified in silico as potentially neutral/tolerated. Notably, deleterious mutations were more often demonstrated simultaneously to CTNNB1 alterations affecting the phosphorylation site T41 compared to S45 (n = 13/21; 61.9% vs. n = 5/21; 23.8%; p = 0.0278) with three deleterious mutations exclusively identified in two CTNNB1 wild type pediatric DTF cases. In total, deleterious non-CTNNB1 mutations were demonstrated in n = 4/22 (  www.nature.com/scientificreports www.nature.com/scientificreports/ In one case (case #202), an imatinib-sensitive KIT (V559D) mutation was detected in addition to the CTNNB1 (T41A) mutation. The tumor was localized in the Musculus rectus abdominis of a 43-year old female patient and displayed a diffuse infiltrative growth.

Spectrum of non-CTNNB1 alterations in
A combined S45P and S45F mutation was detected in a 25-year old male with an 8.0 cm DTF affecting the trunk (case #37). In this tumor, additional mutations were identified in AR (A159T) and SDHA (T36I). In silico analysis of the additional mutations indicated no pathogenic potential. Of note, in one case, a similar combination of S45P and S45F CTNNB1 alterations and an A159T mutation in AR was detected with no clinical data available (case #38).
The known KIT polymorphism M541L was demonstrated in 16% of the analyzed DTF cases. Analyzing the mutational spectrum of single nucleotide substitutions from all n = 183/204 mutated DTF samples, C > T substitution was identified as major process (Fig. 3B).

DiScUSSion
In this study, we attempted to refine the mutational spectrum in DTF beyond the well-established alterations affecting the canonical Wnt/β-catenin signaling pathway 2-6 . We conducted a comprehensive molecular characterization of 204 DTF cases applying NGS.
Based on our previous results analyzing mesenteric DTF 20 , we first examined the mutational CTNNB1 status in correlation with clinicopathological features. Previous studies reported genomic CTNNB1 alterations in 67-92% of sporadic DTF cases demonstrated by conventional Sanger sequencing and limited to CTNNB1 exon 3 7-9,19-21 (summarized in Table 3). In our present study, we conducted NGS with an increased sensitivity (≥10% allelic frequency) and detected deleterious CTNNB1 mutations in an enlarged fraction of n = 181/204 (89%) DTF cases. Due to low allelic frequencies, 4.3% of CTNNB1 mutations were exclusively traceable by NGS (The mean region coverage depth was 377×) and not detected by Sanger sequencing. With regard to clinical data, 4 cases (one pediatric) were known to be associated with FAP and harbored, as expected, no mutation in CTNNB1.
As reported by previous studies, the majority of deleterious alterations in the CTNNB1 gene were restricted to the serine/threonine phosphorylation sites T41 and S45. In our study, T41A (n = 111/204; 54.4%), S45F (n = 40/204; 19.6%) and S45P (n = 18/204; 8.8%) were the most frequent amino acid exchanges affecting sites involved in the modulation of the interaction of β-catenin with several kinases: Sequential phosphorylation at T41, S33 and S37 is mediated by the glycogen synthase kinase 3 beta (GSK-3β), while S45 is phosphorylated by the casein kinase-1 alpha (CK1α) targeting β-catenin for ubiquitination and subsequently proteasomal www.nature.com/scientificreports www.nature.com/scientificreports/ degradation 5,22,23 . Mutations affecting these phosphorylation sites lead to stabilization and translocation of β-catenin into the nucleus where it acts as a transcriptional regulator. A schematic overview of β-catenin indicating relevant protein domains, phosphorylation sites, interaction partners and associated cellular functions is shown in Fig. 4.
Several genotype-phenotype correlations have been published so far with the S45F alterations found to predict a higher tendency for local recurrence in DTF [7][8][9] . In accordance with results published by Colombo et al., we detected CTNNB1 mutations in association with S45 significantly more often in extra-abdominal than in intra-abdominal or abdominal DTF cases (n = 36/106; 36.0% vs. n = 16/78; 20.51%, p = 0.0222). As previously indicated for mesenteric DTF 20 , we observed a significantly higher incidence of deleterious T41 alterations in intra-abdominal compared to abdominal and extra-abdominal DTF cases (n = 43/59; 72.9% vs. n = 63/125; 50.4%; p = 0.0042). Recently, the French Sarcoma group published data from a nationwide prospective study indicating that tumor localization constitutes a major prognostic factor. Two-year event free survival was significantly higher in favorable localizations than in unfavorable sites which was mainly extra-intestinal localization (except the lower limb) 14 .
We observed phenotypic and genetic differences comparing the pediatric and adult DTF groups.  19 . These findings suggest a more complex mutational spectrum occurring at least in a subset of DTF than expected so far, including potentially targetable alterations that could be exploited for therapeutic benefit. To define the spectrum of non-CTNNB1 alterations in our large DTF cohort, all protein coding exons of 26 additional cancer-associated genes were analyzed by NGS. In 15.2% of DTF cases, additional non-CTNNB1 alterations could be detected,   To the best of our knowledge, we herewith report the first DTF case with a synchronous CTNNB1 (T41A) and imatinib-sensitive KIT (V559D) mutation (case #202). The tumor was localized in the Musculus rectus abdominis of a 43-year old female patient, displayed a diffusely infiltrative growth pattern and showed nuclear expression of β-catenin. The imatinib-sensitive KIT (V559D) alteration is known to cause ligand-independent tyrosine kinase activity and is supposed to play a central role in the pathogenesis of GIST [27][28][29] . If this KIT (V559D) mutation represented an additional oncogenic alteration in DTF, one could hypothesize that imatinib therapy might be effective. However, the clinical impact of additional deleterious mutations in the case of CTNNB1 mutated DTF cases remains uncertain. Tyrosine kinase inhibition has been evaluated in a phase 2 study of the German Interdisciplinary Sarcoma Group (GISG-01) 16 . The authors describe a positive correlation of CTNNB1 mutational status and progression arrest rate after imatinib treatment, especially in DTF tumors harboring a CTNNB1 S45F alteration which were overrepresented in that study. In accordance with previously published results 30,31 , the known KIT polymorphism M541L was demonstrated in 16% of all analyzed DTF cases. In a prospective series by Dufresne et al., no predictive effect of the M541L variant on the efficacy of imatinib therapy for patients with DTF was seen 30 .
As the current study was retrospectively designed and focused on the mutational spectrum of different DTF subgroups, one of the major limitations of this study is the sparse clinical follow-up information available to us. Complementary studies comprising larger DTF cohorts are needed to correlate the mutational spectrum of both, the pediatric as well as the adult DTF subgroup to clinical outcome taking into consideration e.g. the primary localization as prognostic factor. Additional information of potential therapeutic benefit will further be gained by expanding the NGS panel used in this study (limited to the exonic region of only 27 genes) to include additional genes with targetable alterations. Furthermore, germline APC mutations were not validated for patients with FAP.
In conclusion, we report deleterious CTNNB1 mutations in n = 181/204 (89%) patients with DTF confirming previous studies. Further mutational analysis indicates that a subset of DTF harbor a more complex mutational spectrum comprising not only deleterious alterations in CTNNB1 but also potentially targetable alterations in non-Wnt/β-catenin signal transduction effectors that could be further exploited for therapeutic approaches. As an example, we herewith report the first imatinib-sensitive KIT (V559D) mutation in DTF which deserves further clinical investigation.

Materials and Methods
tumor specimens/clinicopathological features. In total, 204 DTF tissue specimens were selected from the archive of the GIST/Sarcoma Registry (University Münster, Germany) and the Gerhard-Domagk-Institute of Pathology (Münster University Hospital, Germany). According to the current WHO classification 1 , all DTF diagnoses were reviewed by four experienced pathologists (EW, IG, SH, WH) based on morphological criteria, clinical information, immunohistochemical and molecular analyses as described before 20 . Scientific and www.nature.com/scientificreports www.nature.com/scientificreports/ retrospective analysis of the cohort of mesenchymal tumors was approved by the Ethics Review Board (University of Münster, 2016-091-f-S). Written informed consent from the patients was not requested and was waived by the Ethics Review Board. The study did comply with the principles set out in the United States Department of Health and Human Services Belmont Report and the World Medical Association Declaration of Helsinki. Clinicopathological data were available for 192 (94%) DTF patients (female: n = 118, 61% and male: n = 74, 39%). The cohort included 22 pediatric (≤18 y/a; mean years of age 11.9) and 170 adult (>18 y/a; mean years of age 44.9) DTF patients. With regard to clinical data, 4 cases (2%; one pediatric) were known to be associated with FAP and 200 (98%) cases were sporadic. Median tumor size was 7.0 cm (range 1-28 cm). Tumor sites were categorized into (I) extra-intestinal (n = 106, 58%), (II) intra-abdominal/mesenteric (n = 59, 32%) and (III) abdominal/ abdominal wall (n = 19, 10%). All tumors within the pediatric subgroup were exclusively localized extra-intestinally. Detailed clinicopathological patient and tumor characteristics are summarized in Table 1 and individualized in Supplementary Table S1. that are commonly mutated in cancer. As described before [32][33][34] , Target enrichment was conducted by means of the GeneRead DNAseq Panel PCR V2 Kit (Qiagen). Purification and size selection steps were processed utilizing Agencourt AMPure XP magnetic beads (Beckman Coulter). The GeneRead DNA Library I Core Kit (Qiagen) was applied to perform end repair, A-addition and ligation to NEXTflex-96 DNA barcodes (Bioo Scientific). The HiFi PCR Master Mix (GeneRead DNA I Amp Kit, Qiagen) and NEXTflex oligonucleotides (Bioo Scientific) were used for the amplification of adapter-ligated DNA. Sequencing was performed on the MiSeq system (Illumina; 12.5 pM library pools, 2% PhiX V3 control and MiSeq Reagent v2 chemistry). We used the Quantitative Multiplex FFPE Reference Standard (Horizon Discovery, ref.no. HD200, 11 somatic variants verified at 0.8-24.5% allelic frequency) as isogenic quality control for routine performance evaluation and monitoring of NGS workflow integrity (pre analytical DNA extraction, NGS workflow and post-analytical bioinformatics). The CLC Biomedical Genomics Workbench software (CLC bio, Qiagen) was used for NGS data analysis. Validation by Sanger sequencing (CTNNB1 exon 3) was performed using the BigDye Terminator Cycle Sequencing Kit (v3.1, Life Technologies). Following primer set was applied: 5′-CTG ATT TGA TGG AGT TGG ACA TGG CCA TG-3′ (forward) and 5′-CCA GCT ACT TGT TCT TGA GTG AAG GAC TGA G-3′ (reverse).
In silico tools to predict the potential deleterious impact of detected variants. The potential impact of NGS-detected variants in the coding regions was predicted using following in silico tools as reported 32 40 . The PolyPhen-2 method utilizes physical and evolutionary comparative considerations to predict amino acid changes on protein structure and function. Scoring is in the range from 0 (neutral) to 1 (deleterious) and potential functional significance is categorized into benign, possibly damaging, and probably damaging. The PROVEAN web-based algorithm classified gene variants as either potentially neutral or deleterious (cutoff −2.5). SIFT was used with the default settings, classifying gene variants from 0 (damaging) to 1 (tolerated). The CADD algorithm is trained to differentiate 14.7 million high-frequency human-derived alleles for integration of diverse annotations into a single measure (C score). Scoring correlates with allelic diversity, annotations of pathogenicity, disease severity, complex trait associations, and experimentally measured regulatory effects. Calculated C scores rank a pathogenic alteration relative to all potential substitutions of the human genome. To increase the prediction accuracy and the level of confidence, a combination of in silico algorithms based on functional parameters, protein structure and evolutionary information was used. We combined the individual output from ≥3 of five in silico prediction tools and produced a single consensus outcome summarized in Supplementary Table S3.

Statistical analysis.
Clinical parameters of the cohort are based on descriptive statistics (comprising numbers and percentages, median and mean standard deviation). Chi-square-, Fisher's exact-and t-tests (two-tailed with a 95% confidence interval) were calculated (SPSS 20 software, IBM and GraphPad Prism).