ALK, ROS1, and NTRK Rearrangements in Metastatic Colorectal Cancer

Background
ALK, ROS1, and NTRK fusions occur in 0.2% to 2.4% of colorectal cancers. Pioneer cases of metastatic colorectal cancer (mCRC) patients bearing rearrangements who benefited from anti-ALK, ROS, and TrkA-B-C therapies have been reported previously. Here we aimed at characterizing the clinical and molecular landscape of ALK, ROS1, and NTRK rearranged mCRC.


Methods
Clinical features and molecular characteristics of 27 mCRC patients bearing ALK, ROS1, and NTRK rearranged tumors were compared with those of a cohort of 319 patients not bearing rearrangements by means of Fisher's exact, χ2 test, or Mann-Whitney test as appropriate. Overall survival curves were estimated with the Kaplan-Meier method and compared using the log-rank test. A Cox proportional hazard model was adopted in the multivariable analysis. Deep molecular and immunophenotypic characterizations of rearranged cases, including those described in The Cancer Genome Atlas database, were performed. All statistical tests were two-sided.


Results
Closely recalling the "BRAF history," ALK, ROS1, and NTRK rearrangements more frequently occurred in elderly patients (P = .02) with right-sided tumors (P < .001) and node-spreading (P = .03), RAS wild-type (P < .001), and MSI-high (P < .001) cancers. All patients bearing ALK, ROS1, and NTRK fusions had shorter overall survival (15.6 months, 95% confidence interval [CI] = 0.0 to 20.4 months) than negative patients (33.7 months, 95% CI = 28.3 to 42.1 months), both in the univariate (hazard ratio [HR] = 2.17, 95% CI = 1.03 to 4.57, P < .001) and multivariable models (HR = 2.33, 95% CI = 1.10 to 4.95, P = .02). All four evaluable patients with rearrangements showed primary resistance to anti-epidermal growth factor receptor agents. Frequent association with potentially targetable RNF43 mutations was observed in MSI-high rearranged tumors.


Conclusions
ALK, ROS1, and NTRK rearrangements define a new rare subtype of mCRC with extremely poor prognosis. Primary tumor site, MSI-high, and RAS and BRAF wild-type status may help to identify patients bearing these alterations. While sensitivity to available treatments is limited, targeted strategies inhibiting ALK, ROS, and TrkA-B-C provided encouraging results.

Despite these pioneer case reports, it has not been clearly established whether ALK, ROS1, or NTRK rearranged tumors represent a distinct, although rare, disease subtype that should be detected early in order to adopt a tailored management strategy that may include targeted treatments.
Although a few reports have described the occurrence of ALK, ROS1 and NTRK fusions in CRC (Supplementary Table 1), there is still limited knowledge about clinical and pathological characteristics, prognosis and sensitivity of these tumors to available treatments including anti-EGFR monoclonal antibodies (MoAbs) such as cetuximab and panitumumab. Similarly, except for some preclinical reports 11,19 , comprehensive molecular and functional data to clarify whether these alterations confer oncogene addiction and to suggest perspectives on optimal treatment strategies are not available yet.
We therefore carried out a global effort aimed at characterizing the molecular and clinical landscape of ALK, ROS1 and NTRK rearranged mCRCs. Even though a broader list of gene fusions has been described in CRC, including those affecting RET, HER2 and BRAF 2,8,22,23 , we specifically focused on mCRC with ALK, ROS1 and NTRK rearrangements since their phylogeny is closely related and they are frequently grouped as targets of newly developed agents such as entrectinib 24 .

Study design and participants
In the clinical step (Figure 1

Statistical analysis
We investigated the association of ALK, ROS1 and NTRK rearrangements with the following variables collected at the diagnosis of mCRC: age, gender, ECOG performance status (0, 1), primary tumor location (right colon, left colon, rectum), primary tumor resection, mucinous histology, time to metastases (synchronous, metachronous), number of metastatic sites (1, >1), metastatic sites (lung, lymph nodes, liver, peritoneum), RAS and BRAF status (mutated, wild-type), 6 MMR status (proficient, deficient). Fisher's exact test, χ 2 test or Mann-Whitney tests were used when appropriate to assess the associations of the ALK, ROS1, NTRK rearrangements with investigated characteristics. Statistical significance was set at p=0.05 for a bilateral test.
We investigated the impact of ALK, ROS1 and NTRK rearrangements on overall survival (OS), defined as the time from diagnosis of metastatic disease to death or last follow up for alive patients.
OS analysis was determined according to the Kaplan-Meier method and survival curves were compared using the log-rank test. The correlation of ALK, ROS1, NTRK status and clinicopathological characteristics with OS was assessed in univariate analysis. In order to minimize the bias of multiple comparisons, according to the false discovery rate correction, statistical significance was set at p=0.009 for a bilateral test. Cox proportional hazard model was adopted in the multivariate analysis, including as covariates variables correlated with survival with p<0.1 in the univariate analyses. Hazards' proportionality was assumed.
All analyses were carried out by means of Prism 7 for Mac OS X v7.0.

Translational analyses
As shown in Figure 1

Study population
Based on a systematic literature review, we identified 24 published cases of ALK, ROS1 or NTRK rearranged CRCs (Supplementary Table 1). Nineteen were staged as metastatic, and informative medical records were retrieved for fifteen of them. Taking advantage of screening programs 7 worldwide, we were able to identify 12 additional cases. Therefore, the final population consisted of 27 ALK, ROS1, NTRK rearranged mCRCs (Figure 1; Supplementary Table 2) including a newly described SCYL3-NTRK1 fusion (Supplementary Figure 1). We compared the clinical and pathological features of ALK, ROS1, and NTRK rearranged mCRCs with a cohort of ALK, ROS1, and NTRK negative patients (n=319), screened for phase 1 studies at three Institutions ( Figure 1).

Clinical and pathological features of ALK, ROS1 and NTRK rearranged mCRC
As shown in Table 1 Of note, RAS mutations were much less frequent in rearranged than in other tumors (7.4% vs 48.3%; p<0.001). Only one (3.7%) rearranged sample showed the co-occurrence of SLC34A2-ROS1 fusion and BRAF V600E mutation. Overall, right-sided primary location, RAS wild-type and MSIhigh status, in addition to female gender, were particularly associated with NTRK rearrangements.
Notably, patients with right-sided, RAS and BRAF wild-type, MSI-high mCRCs had 54-and 453- values with specific regard to NTRK rearrangements were 75% and 99%. 8 Molecular reports from next-generation sequencing DNA analyses performed on rearranged cases were retrieved ( Figure 1). Additionally, molecularly annotated genomic variants from seven CRC samples harboring ALK or NTRK3 fusions (Supplementary Figure 2 and 3) in the TCGA database were gathered. First, we focused on the subset of genes previously reported as the most frequently mutated in CRCs (Figure 2A) 23  An explorative analysis of selected genes implicated in immune-escape mechanisms 27 was conducted by retrieving the transcriptomic profiles of the seven rearranged samples for which RNA seq data was available from the TCGA and these were compared with non rearranged MSI-high CRC samples also from TCGA ( Figure 2C). Although the analysis suggested that the presence of rearrangements did not impact the typical MSI-high phenotype represented by the upregulation of immunoinhibitory molecules 27 , the small number of samples limits the power of this observation.

Prognostic impact of ALK, ROS1 and NTRK rearrangements in mCRC
Finally, we explored the clinical impact of ALK, ROS1 and NTRK rearrangements in the metastatic setting (TCGA samples were excluded from survival analyses, since they were mostly found in earlier disease stages and had incomplete follow-up data). When looking at OS results, at a median 9 follow-up of 28. 5  p<0.001) ( Figure 3A). When applying the false discovery rate correction, the association of ALK, ROS1 and NTRK rearrangements with OS was still statistically significant (p<0.005). In the multivariable model (

Therapeutic implications of ALK, ROS1 and NTRK rearrangements in mCRC
All the patients with rearranged tumors that were treated with cetuximab or panitumumab (N=4) experienced disease progression as best response during the treatment with anti-EGFR agents One patient with EML4-ALK rearrangement and MSI-high tumor received single agent anti-PD-1 treatment with nivolumab and achieved a durable response (Supplementary Figure 5). Notably, the IHC staining of this tumor revealed intense staining for CD4, CD8, CD68 and especially PDL-1, with an abundant intra and extratumoral lymphocytic infiltration (Supplementary Figure 6).

Discussion
Here we showed that ALK, ROS1 and NTRK rearrangements identify an uncommon CRC molecular subtype with specific clinical, pathological and molecular features. The investigated fusions (and particularly those affecting NTRK) were more frequent in elderly females with right-sided tumors, spreading to extra-regional lymph nodes. However, the most clinically relevant association was found with MSI-high and RAS wild-type status, which are two relevant and commonly used biomarkers for patient selection for immunotherapy and anti-EGFRs, respectively. This type of clinical and molecular associations resemble very closely what observed for codon 600 BRAF mutations and, interestingly, BRAF V600 mutations and gene fusions were almost invariably mutually exclusive. Since MSI-high status is reported in less than 5% of mCRCs 28 , the frequency of MSI-high rearranged tumors is unexpectedly high (48.1%), even considering the right-sided location 29 . The frequency of MSI-high status in ALK, ROS1 and NTRK rearranged tumors seems similar or even higher than in BRAF V600E mutated mCRCs, where it reaches 30-35% 24,28 . While the association between right-sided tumors, MSI-high and BRAF mutations is well established, we report for the first time a strong association with right-sided tumor location and MSI-high status also for gene fusions. Of note, while frame-shift mutations occurring in MSI-high cancers are heterogeneously represented in tumor sub-clones 30 , gene rearrangements appear as "founder" events, as they are present in most, if not all, tumor cells. Nevertheless, since defective mismatch repair is also an early event in CRC carcinogenesis, the adenoma-carcinoma sequence should be further elucidated for this rare subtype., Future studies exploring the role of food carcinogens and/or peculiar microbiota components in the right colon are also warranted to clarify the potential link between MSI status and kinase rearrangements.
When compared with negative samples, ALK, ROS1, and NTRK rearranged tumors show a low frequency of RAS and BRAF oncogenic mutations. A low prevalence of BRAF V600E mutation was reported in the group of negative tumors (5.8%), probably as a consequence of the poor prognosis and rapid progression of BRAF mutant tumors, preventing these patients to receive later lines of therapy and therefore to be screened for phase 1 trials. Therefore, we were unable to identify a statistically significant difference in terms of BRAF mutations between rearranged and not rearranged tumors (p=1.000) in the present series. However, the observation that ALK, ROS1 and NTRK rearrangements co-occur rarely with other common driver events in the RTK-RAS pathway, and specifically RAS and BRAF codon 600 mutations, supports the hypothesis that gene fusions drive oncogene addiction. Indeed, previous reports indicate that NTRK1 and ALK rearranged CRC preclinical models and patients respond to pharmacological blockade of the fusion kinase 6,11,15,19,20 . In spite of the relatively low prevalence of gene fusions, the identification of patients with tumors bearing these alterations may be simplified and enriched by the evaluation of four simple and easy-to-collect variables (i.e. primary tumor location, RAS, BRAF and MSI-high status), which are available for the vast majority of patients. Therefore, in an evidence-based perspective of resource sparing, the molecular screening for gene rearrangements should not be denied to patients with RAS and BRAF wild-type and/or MSI-high mCRC.
A high prevalence of RNF43 frameshift mutations was reported among ALK, ROS1  Our observations argue that the early enrolment of patients with tumors bearing ALK, ROS1 and NTRK rearrangements in clinical trials with matched targeted agents should be highly encouraged, as this subset of patients may in fact be uniquely poised to benefit from targeted strategies.
Nevertheless, benefit from targeted strategies against ALK, ROS1, and TrkA-B-C may be transient and mechanisms of acquired resistance may occur early 17,20 . This is quite reasonable particularly when considering the impressive mutational burden of MSI-high tumors that may promote in these tumors the early emergence of acquired resistance.
The combination of targeted agents and immunotherapy approaches in MSI-high rearranged tumors may be a promising strategy to be further investigated, supported by a strong molecular rationale, and by the absence of impact of rearrangements on MSI-high associated immunophenotype.
The major limitation of this study is the choice of the control group. Although a wider series of negative cases, especially those analyzed by MSK-IMPACT or FoundationOne tests, would have been more appropriate, both MSK-IMPACT and FoundationOne are DNA-based assays and do not 13 completely cover intronic regions, thus making possible to miss some gene fusions. Moreover, clinical data were not available for the vast majority of these patients. Therefore, a cohort of wellannotated patients screened at three Institutions for a phase 1 trial and quite representative of the general population of mCRC patients was adopted as control group.

Characterization of the novel SCYL3-NTRK1 fusion
For Patient #13 harboring the novel fusion, SCYL3-NTRK1, a set of PCR primers was generated to further confirm the result (SCYL3: 5'-GGAGGAGAACGAACCAAGAT; NTRK1: 5'-CATGAAATGCAGGGACATGG). Total nucleic acid was reverse-transcribed and amplified by PCR using SuperScript III One-Step RT-PCR System with Platinum Taq High Fidelity (ThermoFisher, Carlsbad, CA). The PCR products were assessed on a 2100 Bioanalyzer electrophoresis system (Agilent, Santa Clara, CA). A parallel no template control was also included to determine the presence of any background hybridization.

Criteria for evaluation of primary resistance to anti-EGFR monoclonal antibodies
To assess the association of ALK, ROS1 and NTRK status with primary resistance to anti-EGFR MoAbs, we restricted the analysis to RAS and BRAF wild-type patients receiving cetuximab or panitumumab as single agents or in combination with irinotecan, only in strictly defined irinotecanrefractory patients (i.e. those with documented disease progression during or within three months from the last irinotecan-containing therapy). We excluded patients receiving an anti-EGFR agent in combination with chemotherapy, except in the case of disease progression as best response indicating primary resistance to the whole treatment. Thus, we were able to focus on the true impact of ALK, ROS1 and NTRK translocations on treatment resistance. of pan-Trk, ROS1 and ALK antibodies and a single DAB detection system. Strong staining intensity was seen in almost 100% of tumor nuclei indicating the elevated expression of at least one of the targeted proteins. Panel B: NTRK1 FISH (Abnova SPEC NTRK1) was performed on the same specimen and resulted in break-apart positivity for the NTRK1 gene in 100% of nuclei. Panel C: An RNA-based NGS assay using AMP-technology was performed to identify the fusion/fusion partner. This patient exhibited an intrachromosomal inversion and rearrangement that leads to a novel in-frame fusion of SCYL3 exon 11 to exon 12 of NTRK1 (upstream of the NTRK1 kinase domain in exons [13][14][15][16][17]. Panel D: To confirm the novel fusion, RT-PCR was performed using primers specific to the SCYL3-NTRK1 gene rearrangement. Lane a is a nucleic acid size ladder, annotated in base pair sizes by the column titled 'Bp'; lane b is the RT-PCR product obtained from the patient specimen using rearrangement primers. The arrow indicates the specific RT-PCR product, which migrated at the expected 126 bp size; lane c is a no template control using the same primers as in lane b, which resulted in absence of a PCR product at the expected 126 bp (the strongest product generated migrates at 66 bp).  Table S4) carrying outlier ALK, NTRK1, NTRK2, NTRK3, ROS1 gene expression, defined as the 95th percentile for each gene based on z-score normalization. Gene fusion identification in the RNA sequencing reads from these 154 samples was performed by applying a custom pipeline (see Online Methods) and the FusionMap 1 algorithm. A total of 7 fusions (red circles) in the selected kinases were found.