NTRK fusions in osteosarcoma are rare and non‐functional events

Abstract Neurotrophic tyrosine receptor kinase (NTRK) fusions are promising molecular targets that have been described in a broad range of malignant tumours. Fusions commonly lead to the expression of chimeric proteins with constitutive tyrosine kinase activation that drives tumorigenesis. Despite a low prevalence among most solid tumours (<1%), the first encouraging results with pan‐NTRK tyrosine kinase inhibitors (TKIs) such as larotrectinib or entrectinib stimulated the search for eligible patients. Here, we report the first three cases of osteosarcoma harbouring NTRK fusions, among 113 patients sequenced. It is also the first report on NTRK fusions within a tumour type characterised by highly rearranged genomes and abundant passenger mutations. Whereas the presence of NTRK gene fusions in many tumours is considered to be one of the main driver events for tumour progression, the three chimeric transcripts described here appear non‐functional and likely represent randomly occurring passenger alterations. Particularly in tumours with complex karyotypes, it may therefore be advisable to specifically investigate the fusion transcripts for functional impact before considering targeted treatment approaches using pan‐NTRK TKIs.


Introduction
Physiologically expressed in neuronal tissue, the NTRK1, NTRK2 and NTRK3 genes promote proliferation and survival of neuronal cells through the activation of the MAP-kinase, PLC-γ and PI3K-AKT signalling pathways [1]. Gene fusions between the tyrosine kinase domain of NTRK genes and different upstream partners lead to ectopic expression of constitutively active chimeric proteins. Among the numerous fusion partners already described, most are activating translocations harbouring dimerisation domains responsible for the tyrosine kinase overactivation [1]. The list of cancer types in which NTRK fusions have been identified has kept growing since their discovery in 1982 [2]. These tumours can be divided into a group of rare malignancies displaying a high prevalence (>80%) and a group of various other cancer types, in which NTRK fusions are generally infrequent (<5%) [3].
NTRK fusions are observed in a variety of configurations differing in the combination of N-terminal partners, the NTRK gene involved, the downstream pathways activated and the tumour types affected. Nevertheless, the pan-NTRK TKI larotrectinib has shown remarkable efficacy independent of tumour type with an overall response rate > 75% [4]. Similarly, entrectinib, a Pan-NTRK/ROS1/ALK inhibitor displayed an objective response rate of 79% over different solid tumour types [5,6].
Our aim was to search for NTRK fusions in a comprehensive set of 113 osteosarcomas. Since 30-40% of patients with osteosarcoma still die of their disease despite intense and multimodal treatment regimens, innovative and treatable targets are urgently needed.

Sample collection
All tumour samples were re-evaluated by an experienced bone pathologist and confirmed the diagnosis of conventional high-grade osteosarcoma and a tumour content of >50% per sample. Ethical approval was given by the Ethikkommission beider Basel (reference 274/12) and by the Regional Ethics Committee of Lund University.

DNA sequencing for the detection of structural aberrations
The DNA sequencing strategy differed slightly for the samples from Basel and Lund. In Basel, paired-end libraries from tumour and paired-blood DNA were prepared using the Agilent SureSelectXT HumanV5 kit for whole-genome sequencing (WGS). These were sequenced together with a tumour complementary DNA on an Illumina HiSeq2500 (Cambridge, UK) (paired-end 100 bp). Sequencing reads were mapped to the GRCh37 human reference genome using BWA as described before [7]. In Lund, DNA was extracted form fresh-frozen tumour biopsies and mate pair libraries were prepared for sequencing using the Nextera mate pair sample preparation kit (Illumina, Cambridge, UK) as previously described [8]. To identify structural rearrangements, the sequence data were analysed using the structural variant callers TIDDIT and Delly2.

Circos plots
Copy number aberrations were detected by segmenting log2 values extracted from SNP array analyses using the R package 'copynumber'. For WGS, copy number segments were generated with 'cnvkit' using matched normal tissue as a baseline for copy number = 2. Copy number and structural variant data were then combined to construct circos plots using the R package 'RCircos'.

RNA sequencing
RNA sequencing in Lund was performed as described previously [8]. In Basel, sequencing libraries were prepared using the TruSeq RNA Sample Preparation Kit v2 (Illumina). Total RNA was extracted from freshfrozen tumour tissue and mRNA was then purified from 1 μg of total RNA using oligo(dT) beads. Pairedend sequencing was performed on the Illumina HiSeq 2500 in rapid run mode according to the manufacturer's protocol using the TruSeq SBS Kit v3. Sequencing reads were mapped to the GRCh37 human reference genome using STAR or Hisat2.

Fusion transcript detection
ChimeraScan, deFuse, and FusionCatcher algorithms were used to detect chimeric transcripts from RNAseq fastq files. Predicted fusions were filtered out based on the presence of chimeric spanning or encompassing reads. The sequences of reads spanning a NTRK gene were then blasted against the human transcriptome in order to exclude any ambiguity concerning the involved partners.

RT-PCR and Sanger sequencing validation
RT-PCR and Sanger sequencing were carried out as described previously [8]. In brief, the remaining mRNAs from two patients (VPS18-NTRK3; RALGPS2-NTRK3) were retrotranscribed into cDNAs. RT-PCR was performed with paired primers designed within 200 bp around the breakpoint. The amplification products were then Sanger sequenced.

Results
Next-generation DNA and RNA sequencing was performed across 113 osteosarcomas, including samples from primary tumours and metastases (76 and 37 cases, respectively). Assessment and comprehensive analyses of chimeric transcripts were carried out using Chi-meraScan [10], FusionCatcher [11] and defuse [12] and resulted in the detection of NTRK fusions in three 108 B Ameline, KH Saba et al patients (2.7% of cases, n = 113; Figure 1 and see supplementary material, Figure S1). All gene fusions were verified and validated by the existence of split-reads in genome sequencing data ( Figure 2) and/or RT-PCR (see supplementary material, Figure S2). Somatic copy number variations and structural variant assessment derived from genome sequencing furthermore showed that the NTRK fusions occurred in the context of heavily recombined genomes (Figure 2). In the first case, we analysed a 23-year old female with lung metastases and identified a novel fusion of NTRK2 with an upstream partner UFD1. The fusion led to a premature stop codon by introducing a reading frame shift in NTRK2 upstream of the tyrosine kinase domain with subsequent shortening of the coding sequence of the NTRK2 transcript ( Figure 3A).
The second analysis of a locally recurring osteosarcoma of a 22-year old female revealed an intrachromosomal rearrangement between NTRK3 and the VPS18 gene that was likely generated through chromothripsis of chromosome 15 ( Figure 2B). No fusion transcripts were fully transcribed ( Figure 4B). Moreover, the rearrangement led to the introduction of a premature stop codon, preventing any translation of NTRK3 functional domains ( Figure 3B). A VPS18-NTRK3 gene fusion has already previously been reported by Okamura et al in an unspecified tumour type, although no details about the translocation breakpoint were provided [6].
The third osteosarcoma was derived from a local recurrence of a 51-year old female and showed a dicentric translocation between the 5 0 untranslated region of the RALGPS2 gene and exons 4-17 of NTRK3. The predicted fusion transcript lacked an endogenous start codon ( Figure 1C) and, although it introduced an alternative in-frame start codon just after the breakpoint, this was not sufficient to induce transcription of the downstream functional domains of NTRK3 (Figures 3C and 4C).
Immuohistochemistry was performed using a pan-Trk antibody and did not yield immunoreactivity in any of the three osteosarcomas with NTRK rearrangements (data not shown), indicating a lack of detectable protein.
All tumour samples with NTRK fusions had been obtained and archived >3 years before this study was conducted so none of the patients has been considered for pan-NTRK TKI treatment.  Regarding additional genetic alterations observed, the three tumours all showed copy number losses of CDKN2A and CDKN2B. This concomitant deletion has already been reported in several NTRK-fusion positive tumours of different entities [13,14].

Discussion
To the best of our knowledge, this is the first study to describe NTRK gene fusions in osteosarcoma. A previous pan-cancer study did not reveal a single NTRK fusion in the 53 osteosarcomas included [6]. As expected, osteosarcoma does not belong to the group of tumours that show a high prevalence of NTRK fusions as observed in a small set of rare neoplasms including secretory carcinoma of the breast / salivary gland or infantile fibrosarcoma [3,6].
NTRK gene fusions are commonly considered oncogenic drivers regardless of tumour type but chromosomally unstable tumours like osteosarcomas might challenge this notion. The high amount of chromosomal instability increases the likelihood of abundant and randomly occurring passenger alterations that might also involve the NTRK genes. Whether  The studies published so far included patients with evidence of NTRK rearrangements based on immunohistochemistry (IHC), FISH or RNA/DNA sequencing methods, and some studies only required a tumour to be 'positive for a molecular alteration' of NTRK1-3 [4,5] (ClinicalTrials.gov number: NCT02097810, NCT02568267, NCT02122913, NCT02576431 and NCT02637687). Recent studies suggesting that IHC with a pan-NTRK antibody reliably identifies NTRK1-3 rearrangements pave the way for rendering individual patients suitable for pan-NTRK TKI therapy based solely on surrogate markers [9]. Accordingly, the three patients with NTRK fusions described here would have met the criteria for enrollment in these studies. However, DNA and RNA sequencing of the two first patients demonstrate the introduction of premature stop codons by these fusions and a lack of transcription (Figure 4), which is tantamount to a loss-of-function of the chimeric proteins due to the absence of tyrosine kinase domains. Frameshifts in NTRK fusion transcripts have been reported in only a single case of a primary undifferentiated neuroendocrine carcinoma so far [15]. Finally, the third case exemplifies a recombined transcript whose functionality cannot be assessed by our analysis, although the absence of both an endogenous start codon and detectable RNA makes subsequent translation highly unlikely. None of the three tumours had detectable expression of NTRK proteins, as expected in the absence of transcription.
In summary, the role of NTRK gene fusions as driving/oncogenic events in the three osteosarcoma cases described here can be virtually excluded. At the same time, the functionality of NTRK chimeric transcripts detected in other tumour types with highly rearranged genomes should be interpreted with caution. Chromoanagenesis could be a potential mechanism to explain non-functional NTRK gene fusions as observed in the second tumour. As long as FISH or any other breakpoint-independent technique alone are used as inclusion criteria for clinical trials investigating pan-NTRK TKIs, patients will be included who will most probably not respond to treatment. Hence, sequencing of the fusion transcripts in at least all highly rearranged tumours, but preferably in all tumours, should be considered before initiating targeted treatment. and the Crafoord Foundation. Michaela Nathrath was supported by the Cura Placida Stiftung.