ROS1-rearranged non–small cell lung cancer (NSCLC) makes up approximately 1% to 2% of all NSCLC, is oncogenically driven by a constitutively activated ROS1 kinase paired with certain fusion partners, and can be detected by several different assays. These patients are initially treated with tyrosine kinase inhibitors (TKIs), which target the activated ROS1 kinase. Eventually these tumors develop resistance to initial TKI treatment through secondary kinase mutations that block TKI binding or activation of bypass signaling pathways, which subvert ROS1 as the driver of the malignancy. Investigation of several TKIs that have shown efficacy in secondary resistant patients is underway.
Key points
- •
ROS1-rearranged non–small cell lung cancer (NSCLC) makes up 1% to 2% of all NSCLC.
- •
The ROS1 kinase is fused with a partner leading to constitutive activation of the ROS1 kinase domain with subsequent phosphorylation and activation of downstream signaling.
- •
Identification of ROS1 is done through immunohistochemistry, fluorescence in situ hybridization, reverse transcription polymerase chain reaction, and next-generation sequencing.
- •
Treatment includes the use of TKIs targeting the activated ROS1 kinase, including first-line use of crizotinib and entrectinib.
- •
The main secondary resistance mutations include L2026M, G2032R, G2033N, and S1986Y/F.
Introduction
According to the World Health Organization in 2018, lung cancer is tied with breast cancer for highest yearly incidence of cancer and is by far the largest cause of cancer-related mortality. Non–small cell lung cancer (NSCLC) comprises approximately 85% of all lung cancers, with genetic alterations in epidermal growth factor receptor (EGFR) entailing 10% to 25% of NSCLC and anaplastic lymphoma kinase (ALK) fusions making up another 5%, with higher relative percentages among both in nonsmokers. In patients with metastatic disease and an oncogenic driver (the sole genetic alteration leading to malignant proliferation), the use of oral medications to block this driver has been shown to be effective in managing this malignancy. ROS1 rearrangements are estimated to occur in 1% to 2% of NSCLCs and are amenable to tyrosine kinase inhibitor (TKI) therapy.
ROS1 molecular basis
In the 1980s, ROS1 was discovered as the oncogenic product of the chicken UR2 sarcoma virus and subsequently identified as proto-oncogenic in human glioblastoma cells. ROS1 is related to ALK, LTK, and the insulin receptor families, and the unaltered gene encodes a receptor tyrosine kinase, although a physiologic ligand has yet to be identified. ROS1 rearrangements have been reported in glioblastoma, NSCLC, cholangiocarcinoma, ovarian carcinoma, angiosarcoma, inflammatory myofibroblastic tumors, and spitzoid melanocytic tumors, among others.
The initial ROS1 fusion partner identified was FIG (found in glioblastoma) and is generated by an intrachromosomal homozygous deletion on 6q21 leading to the amino terminal of FIG attaching to and constitutively activating the carboxy terminal kinase of ROS1. Numerous other ROS1 fusion partners exist in NSCLC, including but not limited to CD74, EZR, SDC4, TPM3, GOPC, ZCCHC8, SCL34A2, CCDC6, MYO5C, GPRC6A, LRIG3, CTNND2, OPRM1, and SRSF6. , In vitro studies have shown that constitutive autophosphorylation of ROS1 leads to phosphorylation of SHP-2 and in turn phosphorylation of MAP/ERK kinase, ERK, STAT3, and AKT. In contrast, phosphorylation of these signals has been shown to be decreased in the presence of ROS1 inhibitors.
ROS1 and ALK share approximately 49% homology within the kinase domain, and possibly even higher homology within the ATP-binding pocket, indicating likely cross-activity between originally designed ALK inhibitors and ROS1 targets. In vitro crizotinib binds more tightly and seems to be 5 times more potent against ROS1 than ALK fusion cell lines, likely indicating higher inhibition of ROS1 fusions at studied crizotinib doses.
Testing
Fluorescence in situ hybridization
Fluorescence in situ hybridization (FISH) is a technique that can identify the ROS1 fusion partner and is often considered the definitive standard. FISH involves labeling the 3’ (centromeric) portion of the fusion breakpoint with a fluorochrome and the 5’ (telomeric) portion with a different-colored fluorochrome in order to identify when they have been split apart. There are 2 possible positive rearrangement patterns:
- •
Classic (break-apart) pattern: 1 normal fusion signal (native ROS1) with 2 separated 3′ and 5′ signals
- •
Atypical (isolated 3′ signal) pattern: 1 normal fusion signal (native ROS1) with 1 3′ signal without the corresponding 5′ signal
Thresholds for FISH positivity usually involve the presence of greater than 15% split signals (including both classic and atypical patterns). ,
Immunohistochemistry
Immunohistochemistry (IHC) uses the D4D6 monoclonal rabbit antibody to stain slides for ROS1 protein positivity. The scoring for IHC positivity ranges from 0 to 3+ with slight variations in the definitions between different categories. Broadly, they are:
- •
0 = No detectable staining in tumor cells
- •
1+ = Faint cytoplasmic staining in tumor cells that does not exceed background cell staining
- •
2+ = Cytoplasmic staining exceeding background cell staining in 0% to 50% of tumor cells
- •
3+ = Cytoplasmic staining exceeding background cell staining in greater than 50% of tumor cells
IHC staining is inexpensive and fast; however, the International Association for the Study of Lung Cancer (IASLC) recommends that although it may be used as a screening tool, there should be confirmation of ROS1 positivity via a complementary method. ,
Multiple studies have shown the potential use of IHC as a screening tool given its high relative sensitivity to specificity, depending on the IHC staining cutoff chosen compared with FISH. For instance, 1 study showed a cutoff of 2+ or higher provided a sensitivity of 100% and specificity of 92%, whereas a cutoff of 3+ only provided a sensitivity of 87.5% and specificity of 98%. Similarly, a separate study showed a sensitivity of 100% and specificity of 97.8% with an IHC cutoff of 2+ or greater. Although most studies indicate sensitivity close to 100% with an IHC cutoff of 2+ or higher, 1 study did note a sensitivity of 76.9% and specificity of 95.7% with an IHC of 2+ or higher compared with a sensitivity of 100% and specificity of 93.6% with an IHC of 1+ or higher.
False-positivity of IHC can be related to numerous factors. For instance, in 1 study, most invasive mucinous adenocarcinomas showed ROS1 reactivity on IHC with no ROS1 fusion found, possibly related to artifact from mucin. In addition, nonmalignant areas of biopsy samples can stain positive on IHC, including in osteoclast-type giant cells and in areas of type II pneumocyte hyperplasia and bronchiolar metaplasia. , ,
Reverse transcriptase polymerase chain reaction
Reverse transcriptase (RT) polymerase chain reaction (PCR) testing uses RNA reverse transcribed into complementary DNA, and then amplified using specific ROS1 gene fusion primers via PCR, and subsequently sequenced for confirmation. The main drawbacks of RT-PCR include the requirement for a larger amount of tissue, inability to identify fusions that are not prespecified, and a higher false-positive rate. One study noted that although 100% of FISH-positive samples were identified by RT-PCR, only 65% of RT-PCR–positive samples are FISH positive with a sensitivity of 100% and a specificity of 85.1%. False-positives are posited to be from small subsets of tumor cells with ROS1 positivity, detected by the amplification of PCR, or RNA that is present on the transcriptional level that does not undergo further translation and is likely not clinically significant. , , ,
Next-generation sequencing
Next-generation sequencing uses sequencing in parallel on a massive scale to identify a large number of separate genetic alterations. Individual DNA or RNA is separated spatially and undergoes amplification to eventually identify predetermined genes. Amplification can be achieved either through the use of amplicon sequencing via multiple PCR reactions or hybrid capture sequencing. Sequencing in parallel of the predetermined genes in combination with computational data processing allows the organization and identification of genetic alterations, including gene rearrangements. An RNA-based anchored multiplex PCR assay of 319 fixed paraffin-embedded samples showed a 100% sensitivity and 100% specificity of gene rearrangement identification compared with FISH.
ROS1 clinical data
Crizotinib
Given ALK and ROS kinase domain homology, success with crizotinib in ALK fusion–positive NSCLC, and evidence of more effective binding of crizotinib at tolerated doses, crizotinib became the first TKI to be investigated to target ROS1 fusion–positive NSCLC at the dose of 250 mg by mouth twice a day. An early retrospective analysis evaluated the efficacy of different therapies in a ROS1 fusion–positive NSCLC population of 51 patients. The 15 patients who received crizotinib in the second line or later were compared with 49 patients who received pemetrexed-based chemotherapy in any line of therapy versus 44 patients who received non–pemetrexed-based chemotherapy in any line of therapy. Progression-free survival (PFS) was 294 days in the crizotinib group versus 179 days in the pemetrexed-based chemotherapy group and 110 days in the non–pemetrexed-based chemotherapy group. When evaluating the pemetrexed-based chemotherapy group, when patients are split into those with tumors that express low levels of thymidylate synthetase (TS) messenger RNA (mRNA) versus high levels, the PFS further separates to 184 days versus 110 days.
These data show a similar hypothesis as in ALK NSCLC, that patients with ALK or ROS1 fusions may be more susceptible to pemetrexed-based chemotherapy because of decreased TS levels as a result of the driver mutation. Some controversy exists, because a separate retrospective analysis evaluating the use of pemetrexed-based chemotherapy in patients with different driver mutations found a PFS of 7.5 months in ROS1 patients, 5.3 months in with ALK fusion, 3.7 months in EGFR-mutated patients, and 4.1 months in nondriver patients, which was statistically significant, although the sample size of ROS1 patients was small. In this group of patients, although patients with ROS1 fusion seemed to respond to pemetrexed-based therapy for longer, there was no correlation with TS levels.
The first prospective evaluation of crizotinib in ROS1 fusion–positive NSCLC was in PROFILE 1001, an expansion cohort of the phase I study of crizotinib as a MET and ALK inhibitor, which found the maximum tolerable dose of crizotinib to be 250 mg by mouth twice a day. Fifty patients with advanced NSCLC with ROS1 rearrangements, largely previously treated, received crizotinib and showed an overall response rate (ORR) of 72% and a disease control rate (DCR) of 90%. Median PFS was 19.2 months and the overall survival (OS) rate at 12 months was 85%. PFS results of crizotinib in ROS1-rearranged NSCLC were promising and more than doubled that of ALK-rearranged NSCLC (PFS of 9.1 months). Recently updated results of PROFILE 1001 were published showing a median PFS of 19.3 months and a median OS of 51.4 months.
Soon after the publication of PROFILE 1001, multiple other studies reported the efficacy of crizotinib with ROS1-rearranged NSCLC. In the EUROS1 retrospective analysis of 31 patients with previously treated ROS1-rearranged NSCLC, the ORR was 80% and DCR was 86.6%, similar to PROFILE 1001. However, median PFS was noted as much lower, at 9.1 months. A much larger phase II study in East Asia was conducted with 127 largely previously treated patients with ROS1-rearranged NSCLC showing an ORR of 71.7% and DCR of 88.2%. Median PFS was 15.9 months and a 12-month OS was 83.1%. Interestingly, when separating patients into those with baseline brain metastases and those without, median PFS was 10.2 months and 18.8 months, respectively. These data may elucidate the higher PFS in PROFILE 1001 given originally patients with brain metastases were excluded and only later allowed by amendment, likely indicating enrollment of fewer of these patients.
Three trials (EUCROSS, AcSé, and METROS) are ongoing to evaluate the efficacy of crizotinib in ROS1-rearranged NSCLC. In the EUCROSS trial, the 18 patients treated showed an ORR of 83%. In the AcSé trial, the 37 patients treated showed an ORR of 71%, a DCR of 85%, and a median PFS of 10 months. The METROS trial has no reportable data. ,
An additional retrospective analysis of 49 patients with ROS1-rearranged NSCLC was completed in 2018, which showed an ORR of 83.3% and DCR of 97.2%. Median PFS was 12.6 months and median OS was 32.7 months. A further analysis was completed to evaluate for differences in outcomes when stratified for ROS1 fusion partners, with CD74 as the most common fusion partner (52.8% of patients analyzed). Differences in PFS between CD74 fusions and non-CD74 fusions were 12.6 months and 17.6 months respectively. Differences in OS between CD74 fusions and non-CD74 fusions were 24.3 months and 44.5 months. Both PFS and OS differences trended toward significance; however, on multivariate analysis, they were not found to be statistically significant. A confounding factor for worse prognosis may have been brain metastases, given that only patients with CD74 fusions at initiation of crizotinib had brain disease, with no patients with non-CD74 fusion found to have brain metastases; brain metastases before crizotinib treatment was an independent poor prognostic factor. However, although these data may indicate that brain metastases are more common in patients with CD74-ROS1 NSCLC, during treatment an equivalent percentage of CD74 and non-CD74 patients developed brain metastases. The small sample sizes limit the ability to generalize; however, it does raise the question of whether or not patients with CD74-ROS1 NSCLC develop brain metastases earlier than other fusions and subsequently do worse.
A separate analysis of ROS1 fusion partners was completed on the 127-patient East Asian cohort discussed earlier. Again, CD74 was the most common fusion partner for ROS1, with 49.1% of samples analyzed containing this fusion. When comparing different pairs, including CD74, SDC4, non-CD74, and non-SDC4 in different combinations, the OS and PFS were comparable. When evaluating samples in 15 patients for allelic fraction (AF) and separating patients along the median value into high and low AF, differences were observed, which may convey a prognostic value. Patients with low AF showed an OS of 552 days compared with high AF, which showed an OS of 411 days. PFS differences were not statistically significant.
Ceritinib
With positive results in the use of crizotinib in ROS1-mutated NSCLC, ceritinib, as the initial second-generation ALK inhibitor with proven benefit in crizotinib-naive and crizotinib-treated patients with ALK-mutated NSCLC, was investigated for its utility in ROS1-mutated NSCLC given ceritinib’s ability to inhibit ROS1 in vitro. A phase II study of 32 patients with ROS1-rearranged NSCLC who had progressed on standard therapy was conducted using ceritinib 750 mg by mouth daily. Thirty of the 32 patients were crizotinib naive and the 2 patients who had progressed on crizotinib were not evaluable because of progression or death. When taking the cohort altogether, the ORR was 62%, the DCR was 81%, and the median PFS was 9.3 months. When excluding the 2 crizotinib pretreated patients, the ORR improved to 67%, the DCR to 87%, and the median PFS to 19.3 months. The median OS was 24 months with a 12-month OS of 56%. In addition, intracranial ORR was 25% and DCR was 63% in patients with baseline central nervous system (CNS) metastases. This trial showed favorable outcome data in crizotinib-naive patients with ROS1-rearranged NSCLC; however, it indicated a low likelihood of response in patients previously treated with crizotinib, although the sample size was miniscule. In addition, side effects, namely gastrointestinal (GI) toxicity including diarrhea and nausea, were higher with ceritinib, raising questions of tolerability, with 68% of patients requiring dose adjustments and 72% requiring dose interruption.
Lorlatinib
Lorlatinib, a highly selective third-generation ALK and ROS1 TKI, is approved as second-line therapy for patients with ALK-positive NSCLC given retained potency against secondary ALK resistance mutations. Compared with crizotinib, ceritinib, alectinib, and foretinib in vitro, lorlatinib shows at least a 10-fold improved potency against ROS1. In addition, in vitro data support retained activity of lorlatinib against ROS1 G2032R and L2026M secondary resistance mutations. In the phase I study of lorlatinib in patients with ALK fusion–positive or ROS1 fusion–positive locally advanced or metastatic NSCLC, 54 patients were evaluable. Among those, 12 harbored ROS1 fusion with 7 patients previously receiving ROS1 targeted TKIs. For this group, the ORR was 50% with a DCR of 67% and an overall PFS of 7.0 months. The intracranial ORR was 3 of 5 patients (60%). The recommended phase 2 dose was 100 mg by mouth daily.
The phase I/II expansion of lorlatinib treatment of ROS1 fusion–positive NSCLC is currently ongoing, with 59 patients enrolled with 51 tumor samples evaluable. Thirty-eight patients previously received ROS1-targeted TKIs, with G2032R being the most frequent secondary resistance mutation with an ORR of 29.4%. Among patients previously treated with TKIs, the ORR was 23.8% if no secondary mutation was found and 33.3% if a secondary resistance mutation was found. Of patients with the G2032R secondary mutation, all had stable disease. Thirteen patients were ROS1 TKI naive and had an ORR of 76.9%.
Entrectinib
The multitargeted TKI entrectinib, for ALK, ROS1, and NTRK1/2/3, has shown promising results for the treatment of ROS1-mutated NSCLC and shows a potency for ROS1 30 times stronger than crizotinib. Data from 2 phase I studies, ALKA-372-001 and STARTRK-1, were compiled providing efficacy data. These studies enrolled 119 patients with advanced solid tumors who had previously received at least 3 lines of therapy (including 27% of patients receiving previous ALK or ROS1 TKIs) for their malignancies with any type of alteration in ALK, ROS1, or NTRK1/2/3 allowed. Responses were only observed in patients who harbored gene fusions and not in those with amplifications, copy number variants, insertions, or deletions. In addition, patients with ALK or ROS1 fusion who previously progressed on targeted TKIs did not show favorable responses. Two dose-limiting toxicities were observed at the 800-mg by mouth daily dose level and thus 600 mg by mouth daily was chosen as the recommended phase 2 dose.
An analysis was performed on the phase II eligible population, which included patients with ALK, ROS1, or NTRK1/2/3 fusions who had never received ALK/ROS1 TKIs and who were exposed to doses equivalent to 600 mg by mouth daily. Of these patients, 14 had ROS1-rearranged solid tumors, 13 of which were NSCLC. Within this subgroup, the ORR was 86% with 2 complete responses. Median PFS was 19.0 months with 1 patient with continued response at 32 months at data analysis. In addition, 8 patients with ALK, ROS1, or NTRK1/2/3 mutations in the phase II eligible population had brain metastases, with a response in 63% of patients, showing penetration through the blood-brain barrier.
An abstract reporting the results for ROS1 fusion–positive NSCLC from a pooled analysis of the phase I studies discussed earlier (ALKA-372-001 and STARTRK-1) in addition to the ongoing phase II study STARTRK-2, which enrolled locally advanced or metastatic ALK, ROS1, or NTRK1/2/3 fusion solid tumors and only allowed prior treatment with crizotinib in NSCLC with brain-only recurrence, was presented at AACR (American Association for Cancer Research) 2019. Of the patients enrolled, there were 53 TKI-naive patients with ROS1 fusion–positive NSCLC. The reported ORR was 77% with 3 complete responses. PFS was 26 months for patients without CNS disease and 14 months for patients with CNS disease at baseline. Intracranial ORR was 55%.
Repotrectinib
Repotrectinib, a next-generation ALK, ROS1, and NTRK1/2/3 fusion inhibitor, is also under investigation for the treatment of ROS1 fusion-positive NSCLC. Although entrectinib seems to have limited utility in the setting of previous ROS1 TKI use, preclinical data indicate activity of repotrectinib against secondary solvent front mutations in ROS1, namely G2032R and D2033N, as well as theorized effective CNS penetration. In the phase I clinical trial TRIDENT-1, enrolling patients with locally advanced or metastatic solid tumors positive for ALK, ROS1, or NTRK1/2/3 fusion, 31 ROS1 fusion–positive tumors (29 NSCLC) were evaluable under different dose escalation cohorts. The ORR was 70% among TKI-naive patients and 11% among TKI-refractory patients, with a patient with CD74-ROS1 NSCLC with brain metastases and confirmed G2032R secondary mutation showing a partial response, including in the CNS.
DS-6051b
DS-6051b is a TKI developed with a high affinity for both ROS1 and NTRK1/2/3. Both in vitro and in vivo mice xenograft models showed affinity of DS-6051b for wild-type ROS1 as well as G2032R-mutated ROS1 with DS-6015b more effectively inhibiting G2032R tumor xenograft growth than lorlatinib. The US phase I study of DS-6051b in advanced solid tumors with either ROS1 or NTRK [Neurotrophic Tropomyosin Receptor Kinase] fusions enrolled 9 patients with ROS1 fusions, 7 of whom were previously treated with ROS1 TKIs. Of the 6 evaluable patients, all were previously treated with crizotinib and showed an ORR of 33.3% and DCR of 66.7%. The maximum tolerated dose was 800 mg by mouth daily.
In a Japanese phase I study of DS-6051b in advanced ROS1 fusion–positive NSCLC, 15 patients were enrolled. Twelve of the patients had measurable disease and 9 of the 12 were ROS1 TKI naive. The ORR was 58.3% for the overall cohort, 66.7% in TKI-naive patients, and 33.3% in TKI-pretreated patients, whereas the DCR was 100% overall. The recommended phase 2 dose was determined to be 600 mg by mouth daily. The discrepancy between the US and Japanese phase I trials can be ascribed to an increased area under the curve in Japanese patients on 800 mg by mouth daily, which decreased when corrected for body weight, indicating a likely consequence of different median weights for patients in the United States and Japan. Table 1 provides a summary of the numerous studies discussed evaluating TKI efficacy in ROS1-rearranged NSCLC.
