Pan-tumor survey of ROS1 fusions detected by next-generation RNA and whole transcriptome sequencing

Receptor tyrosine kinase (RTK) fusion has been recognized as oncogenic structural gene rearrangements in solid malignancies [1]. Among the 58 human RTKs [2], there are US Food and Drug Administration (FDA) approved treatments in anaplastic lymphoma kinase (ALK), c-ROS1 (ROS1), rearranged in transformation (RET), fibroblastic growth factor receptor (FGFR2-3), and neutrophin receptor tyrosine kinase (NTRK1-3) fusion positive tumors. With the exception of NTRK and RET fusions, all of the US FDA approvals are tumor-specific: ALK (non-small cell lung cancer [NSCLC]), ROS1 (NSCLC), and FGFR2-3 (urothelial, cholangiocarcinoma). Although these RTK fusions are found in all solid tumors albeit in a lower frequency, the main biology of the pathological process is univerval and not tumor histology-specific. Therefore, it is important to identify RTK fusions systematically beyond the specific tumor histologic types with approved treatments to expand the horizon of RTK fusion patients who may benefit from the expanded approval of treatments by raising awareness among clinicians, pharmaceutical companies and regulatory authorities to screen and enroll these patients in future clinical trials.

In this study, we performed a large-scale pan-tumor survey of ROS1 fusions detected by next generation RNA sequencing to identify and characterize the molecular characteristics of ROS1+ solid tumors.

In this first large scale survey of ROS1 fusions identified by RNA NGS where only in-frame messenger RNA (mRNA) transcripts were reported, we identified 259 ROS1+ tumor samples by RNA NGS of tumor samples spanning 16 tumor types. Many inter/intragenic rearrangements have been reported in the literature using pure DNA NGS, but whether an in-frame mRNA were eventually transcribed (and the exact ROS1 fusion variant) remained to be determined especially if clinical response was not reported [6, 7].

From the example of NTRK fusions, it is generally accepted that RTK fusions are likely an universal actionable driver among the vast majority if not all tumor types harboring that RTK fusion [1]. While ROS1+ NSCLC tumors were the dominant tumor type at 78.8% and has US FDA approved treatment of two ROS1 TKIs [8, 9], it implies that potentially more than 20% of the ROS1+ solid tumor may benefit from currently approved ROS1 TKIs.

Importantly, ROS1+ GBM constituted the second largest ROS1+ tumors at 6.9%. In fact ROS1 fusion was first identified in glioblastoma multiforme in 1987 [10]. Thus, it not surprising that ROS1+ GBM constituted the second most common ROS1+ solid tumors. Although limited by the numbers of ROS1+ glioblastoma identified and thus statistically not significant, the presence of ROS1 fusions in glioblastoma seems to indicate a poor prognosis. Entrectinib, a first-generation ROS1 TKI, has CNS activity and second generation ROS1 TKIs in clinical development such as repotrectinib, taletrectinib and NVL-520, also has either demonstrated CNS activity clinically or in pre-clinical models and thus could be considered as a potential treatment option for ROS1+ GBM patients [9, 11‐13].

ROS1+ breast adenocarcinoma was the third most common ROS1+ solid tumors but only at 3%. Other ROS1 fusion positive tumors that have been previously reported in the literature such as ROS1+ melanoma [14] and ROS1+ soft tissue tumor (inflammatory myofibroblastic tumor [IMT]; [15] were also identified in the database.

Molecularly, the exon fusion breakpoints in exons 32, 34–36 with fusion at exon 34 was the most common especially among ROS1+ NSCLC while exon 35 were found as fusion breakpoints for ROS1+ non-NSCLC tumors. Among the junctional reads of the major ROS1+ tumors, the highest was among NSCLC, followed by GBM. Of note, ROS1+ breast adenocarcinoma, although with limited number of samples, had a tenfold lower junction reads than those of ROS1+ NSCLC. Additionally, two ROS1+ breast adenocarcinoma had fusion breakpoints at exon 17 and exon 27, respectively.

In our study, ROS1 fusions were also detected at similar frequency as previously reported ROS1+ NSCLC tumors of approximately 2% [16] in other major tumors such as breast and pancreatic cancers. This is consistent with prior reports of identification of ROS1 beyond NSCLC [17‐24]. While the second most common tumor type glioblastoma is considered rare, the third most common tumor type in our study was breast cancer, which is one of the most common types of tumor. A previous Chinese study of 1440 breast cancer patients described a total of 30 RTK events including 3 with ROS1 [18]. Although the hormonal status of these patients were not described in the paper, a prior case report on inflammatory breast cancer harboring CD74-ROS1 was triple negative [17] and so were 3 out of 7 cases of ROS1+ breast cancer in our study. Triple negative breast cancer patients are known to have less treatment options. Thus, it is important to profile tumors beyond NSCLC for ROS1 fusions given there are now two approved tyrosine kinase inhibitors for the treatment of ROS1+ NSCLC.

By far, the most common ROS1+ tumor type was NSCLC. We identified 8 of the 24 ROS1+ NSCLC fusion partners reported in the literature [25]. However, 4 (CD74, EZR, SLC34A2 and SDC4) of the fusion partners made up the vast majority of the ROS1+ NSCLC fusion partners. Neel et al. have demonstrated that different fusion partners affect the subcellular localization of the ROS1 fusions [26] while Li et al. has described that ROS1+ NSCLC patients with CD74-ROS1 fusion partners are more likely to present with brain metastases and showed a trend toward improved survival in the non-CD74-ROS1 group when they were treated with crizotinib [27], suggesting the possibility that fusion partners may have differential responses to therapy. As in the case with ALK-rearranged NSCLC, concurrent mutations such as TP53 may also play a role on differential responses to targeted therapy [28] and further exploration in the space of fusion partners and concurrent mutations in ROS1+ NSCLC is eagerly awaited.

Also notable in this brief report is the fact that to our knowledge, this is the first large scale survey of PD-L1 expression among ROS1+ NSCLC. PD-L1 expression was detected in 81.2% of the ROS1+ NSCLC samples where the PD-L1 expression was known. The majority of the PD-L1 positive ROS1+ NSCLC (> = 1%) were high expressors (54.5%, 104/191). In these patients, clinicians may be tempted to use single agent pembrolizumab as the first-line treatment of ROS1+ NSCLC given the overall survival benefit of Keynote-024 results for PD-L1 expression (> = 50%) and the FDA expanded approval of pembrolizumab approval of pembrolizumab for PD-L1 >  = 1% based on the Keynote-042 results as only EGFR + and ALK + NSCLC were excluded and ROS1+ NSCLC were not excluded from these studies [29, 30].

However, single agent immune checkpoint inhibitor appears to have limited activity in actionable driver mutation positive NSCLC. It is generally recognized that single agent immunotherapy is not effective in EGFR mutated NSCLC [31, 32]. Although evidence in ROS1 fusion positive NSCLC is limited, a global registry has shown limited ORR of immune checkpoint inhibitors in NSCLC harboring oncogenic alterations with reported ORR of ROS1 fusion + patients being 17% (n = 7) and 12% (n = 125) for EGFR mutated patients [33]. While better than the ORR of 12% in EGFR mutated patients, immunotherapy as a single agent may not be as effective as other options in ROS1 fusion + NSCLC. Although statistically non-significant and limited analysis due to small sample size, our study also showed that the time on treatment was less with immunotherapy versus ROS1 targeted TKIs in ROS1+ NSCLC. Further prospective data on efficacy as well as safety are warrented.

Overall the incidence of ROS1+ NSCLC detected in this database was lower than the generally accepted approximately 2% incidence in the literature [16]. Targted RNA NGS and WTS are the most vigorous platforms in detecting RTK fusions where the transcribed RNA are detected and the reading frame is checked to ensure it is "in frame". In this study, there was no difference in the detection rates of ROS1+ by ArcherDx fusion assay (0.52%, 55/9393 and WTS (0.55%, 155/28173) in the NSCLC cohort which constituted the majority of the ROS1 fusions.

Despite limitations, we were able to determine the characteristics of ROS1 fusions in a tumor agnostic manner. ROS1 fusions were identified in multiple tumor types. Only 78.8% with NSCLC have approved ROS1 targeted therapy. Further studies to investigate the role of tyrosine kinase inhibitors in the approximately one fifth of solid tumors with ROS1 fusions would be warranted. Finally, a more detailed examination of the clinical effects of other co-existing mutations along with underlying biological and molecular mechanism including high PDL1 score and having concurrent alterations such as TP53 to account for the differences in survival outcomes of various ROS1 fusions is eagerly awaited.