Testing for ROS1 in non-small cell lung cancer: a review with recommendations

Rearrangements of the ROS1 gene occur in 1–2 % of non-small cell lung cancers (NSCLCs). Crizotinib, a highly effective inhibitor of ROS1 kinase activity, is now FDA-approved for the treatment of patients with advanced ROS1-positive NSCLC. Consequently, focus on ROS1 testing is growing. Most laboratories currently rely on fluorescence in situ hybridisation (FISH) assays using a dual-colour break-apart probe to detect ROS1 rearrangements. Given the rarity of these rearrangements in NSCLC, detection of elevated ROS1 protein levels by immunohistochemistry may provide cost-effective screening prior to confirmatory FISH testing. Non-in situ testing approaches also hold potential as stand-alone methods or complementary tests, including multiplex real-time PCR assays and next-generation sequencing (NGS) platforms which include commercial test kits covering a range of fusion genes. In order to ensure high-quality biomarker testing, appropriate tissue handling, adequate control materials and participation in external quality assessment programmes are essential, irrespective of the testing technique employed. ROS1 testing is often only considered after negative tests for EGFR mutation and ALK gene rearrangement, based on the assumption that these oncogenic driver events tend to be exclusive. However, as the use of ROS1 inhibitors becomes routine, accurate and timely detection of ROS1 gene rearrangements will be critical for the optimal treatment of patients with NSCLC. As NGS techniques are introduced into routine diagnostic practice, ROS1 fusion gene testing will be provided as part of the initial testing package.


Introduction
Lung cancer is the most frequent cause of cancer-related death worldwide and is usually diagnosed in advanced stages [1]. The most common histological lung cancer subgroup is nonsmall cell lung cancer (NSCLC), which accounts for 80 % of lung cancers [1]. Currently, there are two identified molecular subtypes of NSCLC that have targeted therapies approved for their treatment: mutations in the epidermal growth factor receptor (EGFR) gene and rearrangements in the anaplastic lymphoma kinase (ALK) gene; tumours harbouring these genetic alterations respond well to specific tyrosine kinase inhibitors [2,3]. In addition to EGFR and ALK, other known oncogenic drivers of NSCLC include hepatocyte growth factor receptor (MET), the GTPase KRAS, human epidermal growth factor receptor 2 (HER2), RET and ROS1 [4,5].
ROS1 is now recognised as a distinct molecular target in NSCLC [6,7]. Pre-clinical and clinical studies demonstrate that ROS1 can be efficiently inhibited by the tyrosine kinase inhibitor crizotinib [8,9], which is approved by the FDA and EMA as a treatment for patients with advanced ALK-positive NSCLC [10,11]. Crizotinib was recently approved by the FDA for patients with advanced ROS1-positive NSCLC [12]; therefore, detection of ROS1 gene rearrangements is critical for the optimal treatment of ROS1-positive NSCLC patients. In this article, we review the current state of molecular diagnostics for ROS1-positive NSCLC, discuss our experience with the relevant technologies and provide guidance on the detection of ROS1-positive tumours.

Rationale for targeting ROS1 fusions in NSCLC
Although v-ROS1 had already been identified as a unique oncogenic sequence in the avian sarcoma virus (VR2) [13], a chicken retrovirus, it was only in 2003 that the genomic structure of ROS1 was fully characterised [14]. ROS1 belongs to the human receptor tyrosine kinase (RTK) family and is evolutionarily close to the ALK family, forming part of the scientific basis for using inhibitors of ALK as inhibitors of ROS1. The ROS1 gene is located on chromosome 6 (6q22) and encodes a transmembrane receptor protein with unique features. The extracellular N-terminal domain spans more than 1800 amino acids, which makes it one of the largest extracellular domains amongst all human RTKs. Despite this, no human ROS1 ligand has been found to date and the physiological function of this orphan receptor is still unclear. The C-terminal portion of ROS1 contains a kinase domain and a single transmembrane domain [9,[15][16][17].
Genomic rearrangements involving ROS1 occur in 1-2 % of NSCLCs [9,[18][19][20][21][22][23]. ROS1 gene rearrangement was initially discovered in the glioblastoma cell line V118MG [24]. In this cell line, an intrachromosomal deletion on chromosome 6 fused the 5′ region of a gene named FIG to the 3′ region of ROS1. Since then, many more novel ROS1 fusion partners have been found. Importantly, the ROS1 kinase domain is retained in all of these fusion events and the expressed fusion genes have been reported to be oncogenic. Known ROS1 fusion partners in lung cancer include FIG, CD74, SLC34A2 and SDC4, and the list is growing. CD74-ROS1 is the most frequently detected ROS1 fusion in this group of patients. With all of the known fusion genes, the ROS1 kinase domain is fully retained and the ROS1 junction point at the messenger RNA (mRNA) level invariably occurs at the 5′ end of exons 32, 34, 35 or 36 ( Fig. 1 and Table 1).
Unlike in ALK, where the fusion partner provides a dimerisation domain that induces constitutive oligomerisation and thus activation of the kinase, the mechanism by which ROS1 fusion proteins become constitutively active is not exactly known. In fact, many of the known ROS1 fusion partners do not contain dimerisation domains [22]. What is known is that several signalling pathways are activated by ROS1 fusion proteins. Expression of FIG-ROS1, CD74-ROS1 or SCD4-ROS1 in fibroblasts or Ba/F3 cells has been shown to result in auto-phosphorylation of ROS1 and phosphorylation of SHP-2, MAP-ERK kinase, ERK, STAT3 and AKT, and these effects have been blocked by pharmacological inhibition of ROS1. Subcellular localisation and downstream signalling may differ depending on the fusion partner of ROS1 [19,20,25,26], but in general, the activated pathways seem to involve common growth and survival pathways that are also activated by other RTKs.
Efficacy and safety of ROS1 inhibitor therapy ROS1 inhibition by crizotinib has been studied in a number of early-phase clinical trials in patients with advanced ROS1-positive NSCLC (Table 2). In the ROS1 expansion cohort of a phase 1 trial of crizotinib, the objective response rate (ORR) was 72 %. Median duration of response was 17.6 months and median progression-free survival (PFS) was 19.2 months. No relationship was observed between ROS1 fusion partner and duration of crizotinib treatment [8]. Furthermore, ORR with crizotinib was 80 % and median PFS was 9.1 months in heavily pre-treated patients in a retrospective study [27]. Consistent with this, in patients with advanced ROS1-positive NSCLC receiving crizotinib in a French phase 2 trial, ORR was 69 % and median PFS was 9.1 months [28]. Finally, ORR was 69 % and median PFS was 12.9 months with crizotinib in a phase 2 trial in East Asian patients with advanced ROS1-positive NSCLC [29]. Across the clinical studies in ROS1-positive NSCLC, crizotinib treatment was well tolerated, with an adverse event profile similar to that seen in ALK-positive NSCLC [3,30]. Phase 2 trials in ROS1positive NSCLC are currently ongoing.

Detection of ROS1 gene rearrangements
As mentioned above, ROS1 gene rearrangement is one of several addictive oncogenic events which may drive a proportion of pulmonary adenocarcinomas. Since ROS1-positive tumours are very sensitive to treatment with tyrosine kinase inhibitors such as crizotinib, detecting this rare genetic alteration may be an important step in the diagnostic work-up of a patient with lung adenocarcinoma.
The traditional approach to detecting ROS1 gene rearrangement is by the use of so-called dual 'break-apart' fluorescence in situ hybridisation (FISH) probes, where the rearrangement separates the two ends of the ROS1 gene and thus the two probes. The rearrangement event, when oncogenic, fuses the portion of the ROS1 gene bearing the tyrosine kinase domain with another partner to create a ROS1 fusion gene. An alternative approach to the identification of the abnormal DNA sequence created by the rearrangement event is to use massive parallel 'next-generation' sequencing (NGS). A variety of approaches using this technology may be used, and commercial platforms are now available, for use with test kits covering a range of fusion genes, including ROS1. Following transcription, fusion gene mRNA provides another possibility for detection with polymerase chain reaction (PCR) technology using a multiplex platform capable of detecting a range of known ROS1 fusion gene transcripts. For oncogenic activity, the ROS1 fusion gene transcript must be translated into protein with tyrosine kinase activity. Elevation of ROS1 protein levels in tumour cells may be detected by immunohistochemistry (IHC) using antibodies against ROS1 protein. This leaves open a possible scenario, rather like testing for ALK gene rearrangement, where elevations in the protein may be used as a surrogate marker for the presence of a ROS1 gene rearrangement. Furthermore, a positive IHC test is likely to be an indication of a functional rearrangement, since the protein must be present for oncogenic activity and the protein tyrosine kinase is the target of drug therapy. Details of these various testing approaches are discussed below.

Fluorescence in situ hybridisation
The screening strategy for ROS1 rearrangement was developed based on the experience of ALK testing in lung adenocarcinomas [17,23,32]. While ROS1 alterations may be detected with a variety of techniques, most laboratories rely on FISH assays using a dual-colour break-apart probe design. These involve labelling the 3′ (centromeric) part of the fusion breakpoint with one fluorochrome and the 5′ (telomeric) part with another fluorochrome. It is important to choose a 3′ probe colour [16,22,[35][36][37] that allows ROS1 (most commonly a green 3′ fluorochrome) and ALK (an orange 3′ fluorochrome) tests to be distinguished [38], particularly if the two tests are to be run together on one slide, in parallel. There are two positive ROS1 rearrangement patterns. One is the break-apart pattern ('classic' pattern) with one fusion signal (native ROS1) and two separated 3′ and 5′ signals. The other positive pattern is an isolated 3′ signal pattern, usually an isolated green signal, ('atypical' pattern) with one fusion signal (native ROS1) and one 3′ signal without the corresponding 5′ signal (Fig. 2). Table 3 summarises the criteria for ROS1 FISH interpretation in NSCLC [23,32,35,36].
For optimal FISH results, there are a number of relevant factors. Use of sections older than 6 months may result in poor hybridisation. In the post-analytic phase, it is important that only intact tumour cells with non-overlapping nuclei are scored. Furthermore, the use of an automated software system (e.g. BioView Duet system, Rehovot, Israel) can facilitate FISH scoring. It should be noted that FISH testing for ROS1 (and ALK) is not restricted to histological tissue sections, but is also applicable to cytological specimens [39,40].

Immunohistochemistry
Given the rarity of ROS1 rearrangements in NSCLC, screening of tumours by IHC may allow unnecessary FISH analysis in ROS1-negative cases to be avoided and thus dramatically reduce the cost of testing. Based on available data, IHC is an effective screening tool to detect ROS1-positive NSCLC, with a sensitivity of 100 % in most studies and a variable specificity ranging from 92 to 100 %, depending on the threshold used to define positivity [20,36,37,[41][42][43][44][45]. These results are based on the use of the ROS1 (D4D6) rabbit monoclonal antibody (Cell Signaling Technology, Danvers, MA, USA) applied at dilutions ranging from 1:50 to 1:1000 with various antigen retrieval methods and use of different amplification and detection systems, in automated instruments or manually. Tumour specimens with known ROS1 rearrangement, or a cellblock of the HCC78 cell line harbouring the SLC34A2-ROS1 fusion gene, can serve as positive controls [20] ( Figs. 3a and 4a). In contrast to ALK, where the ganglion cells of the appendix serve as an adequate external control, there is currently no good external benign tissue control for ROS1. Currently, there is no universally accepted system for how to score IHC results. The thresholds used include either any staining above faint background (if present) or moderate or strong staining (2+/3+). Another option is the use of an Hscore with optimal threshold for ROS1 positivity defined as >100 [40] or >150 [43]. Notably, weak and focal staining was found in 31 % of 253 ROS1 wild-type lung carcinomas in one study [44]. However, this had hardly any influence on specificity when appropriate thresholds of positivity were used (i.e. H-score >150). Thus, none of the reported scoring methods have been shown to be clearly superior to the others since all resulted in very good to excellent correlation with FISH results.
Positive ROS1 IHC typically reveals finely granular cytoplasmic staining (Fig. 3). However, the staining pattern may depend on the function and subcellular location of the gene fusion partner [44,46]. Globular ROS1 immunoreactivity has been described in tumour specimens with the CD74-ROS1 fusion, and membranous staining has been observed in tumours with the EZR-ROS1 fusion [41,44]. Interestingly, ROS1 expression levels in ROS1positive lung cancers and cell lines can vary from cell to cell, suggesting dynamic ROS1 protein expression despite homogeneous presence of ROS1 gene rearrangement (Figs. 3 and 4). Detection of ROS1 protein expression in ROS1-positive adenocarcinomas with signet ring cells is challenging since the cytoplasm is largely replaced by non-reactive mucin [44]. The same pitfall has already been shown for ALK IHC [47]. One should be aware that weak ROS1 expression is occasionally detectable in nonneoplastic hyperplastic type II pneumocytes (Fig. 3e, h) and in alveolar macrophages. In bone metastases, there is strong granular cytoplasmic staining of osteoclast-type giant cells (Fig. 3f).
Lung cancer is often diagnosed by cytology alone, necessitating ROS1 testing in cytological specimens. For cellblocks, the same IHC protocols can be used as for formalin-fixed paraffin-embedded (FFPE) tissue specimens. In the laboratory of one of the authors (L.B.), ROS1 IHC on the BondMax immunostainer is routinely used on conventional, ethanol-fixed and Papanicolaoustained cytological specimens, including smears or cytospin preparations (Fig. 4). Notably, in the case of limited cytological material, FISH for confirmation of a positive IHC result can be applied to the immunostained slide if 3-amino-9-ethylcarbazole (AEC) was used as a red chromogen. Although IHC on cytological specimens is common practice in many laboratories, immunocytochemistry performed on smears and/or cytospin slides may be much more influenced by various pre-analytical factors [48,50]. Thus, it should be performed only in laboratories with experience and appropriate quality assurance in place. The simultaneous use of cell blocks would allow testing sequentially for several biomarkers. ROS1 IHC has a great advantage over FISH in that it can detect rare positive cells or cell groups within a majority of non-neoplastic reactive cells that would be easily missed by FISH. This can be particularly helpful in cytological specimens where architectural tissue context is lost. Although it has been proposed to score ROS IHC only in specimens containing ≥20 tumour cells [41], a positive result in even only a few clearly neoplastic cells can be considered diagnostic.
In summary, IHC is a cost-effective method that can be used to efficiently screen patients with lung cancer for ROS1 rearrangements. Given the laboratory-dependent variability of specificity, confirmation of positive or doubtful ROS1 IHC by FISH or another method is highly recommended.

Non-in situ technologies
In addition to FISH and IHC, a number of non-in situ approaches based on real-time PCR (RT-PCR) or NGS have been developed for the detection of ROS1 gene rearrangements. RT-PCR assays require multiple specific primer sets to discriminate amongst known fusion variants, which can be confirmed by subsequent sequencing [50]. RT-PCR has been successfully utilised to identify positive cases with a sensitivity of 100 % and a specificity of 85-100 %, using FISH as the reference standard method [37,42]. Multiplex RT-PCR is easy to perform, rapid and relatively inexpensive but may be challenging using RNA extracted from FFPE samples [52]. In addition, as the list of ROS1 fusion partners is quite large and still growing, RT-PCR is likely to miss rare variants. These reasons have limited the use of the technique in clinical practice. Recently, a very sensitive RT-PCR-based method was devised to detect the overexpression of 3′ regions of fusion transcripts involving tumour genes constitutionally repressed or expressed at very low levels [53]; this approach has been successfully applied to ALK gene fusions in lung cancer [53,54]. Unfortunately, this method cannot be easily applied to ROS1, since the gene is also expressed in normal and hyperplastic lung tissue [15,55]. An alternative transcript-based method for detecting ROS1 fusion genes is also available. The NanoString assay, capable of detecting known fusion gene transcripts and employing a dual capture and reporter probe system, provides a convenient and commercially available assay that has shown good concordance with FISH and IHC results for ROS1 [50,55].
A series of innovative approaches to detect gene fusions in multiple targets has been developed using NGS (Table 4). It is remarkable that some of these comprehensive assays require as little as 10 ng of RNA [56], with relatively low failure rates in paraffin-embedded tissue (5.6 % in the authors' experience [unpublished data]). A very sensitive NGS technique to assess ROS1 and other gene rearrangements in lung cancer is anchored multiplex PCR that targets only the gene of interest, allowing the detection of the specific alteration irrespective of fusion  NSCLC non-small cell lung cancer partner. Validation of a gene rearrangement panel using 319 FFPE samples showed 100 % sensitivity and 100 % specificity compared with reference assays [51].
These promising results suggest potential application of non-in situ methodologies in clinical practice, as stand-alone methods or as complementary tests within algorithms for the selection of patients to be treated with ROS1, RET or NTRK inhibitors [57]. However, published data for these assays are still limited.

Concordance between FISH, IHC and PCR
There is good correlation between FISH and IHC using clone D4D6 with a highly sensitive amplification kit. Although some discrepant cases have been reported, ROS1 testing by IHC seems to be highly sensitive, but less specific, also when compared with ALK IHC for detection of the corresponding gene rearrangement. As suggested by others [41], IHC testing of specimens containing at least 20 tumour cells and application of an H-score cut-off of >100 are highly concordant with ROS1 rearrangement by FISH or RT-PCR.
Currently, there is very limited published information on the concordance of IHC, in situ hybridisation (ISH) and nonin situ tests for the detection of ROS1 gene rearrangements in lung adenocarcinoma [36,37,42,58]; less than 30 cases with gene rearrangements have been subjected to comparative study of the three methods. Four of the ROS1 tests currently hold in vitro diagnostic (IVD) and CE-marked status ( Table 5). The general consensus seems to be that IHC, ISH and non-in situ methods all are promising for the detection of ROS1-positive cases, with concordance rates well above 90 %. Preliminary conclusions from limited studies suggest a role for IHC as a screening tool, but so far the lack of an IVD-classified IHC assay is problematic. Advantages of RT-PCR analysis include the highest sensitivity reported and the ability to identify translocation partners. On the other hand, there are potential issues around the quality and quantity of mRNA that may be obtained from routine, FFPE NSCLC diagnostic tissues. Furthermore, familiarity and availability of FISH as a technique in detecting other markers, such as ALK, is also important. Perhaps the use of more than one technique could be of value until further experience of testing and companion diagnostics with IVD status have emerged.
General recommendations for ROS1 testing Current guidelines either do not refer to ROS1 testing [59] or mention it briefly without making any strong recommendation [33]. With recent changes in the status of crizotinib for the treatment of ROS1-positive NSCLC, the case for recommending ROS1 testing will now increase. Certainly, assuming drug treatment is available, response rates in treated patients whose tumours bear a ROS1 rearrangement are impressive [8]. Fig. 3 a-f Examples of ROS1 IHC in histological NSCLC specimens (D4D6 antibody, Ventana BenchMark XT; DAB chromogen). a HCC78 cell line (cellblock; ×400). b NSCLC with diffuse, strongly positive staining (×200). c NSCLC with diffuse, granular cytoplasmic staining (×400). d Adenocarcinoma with heterogeneous staining (×200). e Nonneoplastic type II pneumocytes with weak ROS1 staining (×630). f Bone metastasis of a ROS1-negative NSCLC showing strong granular staining of non-neoplastic osteoclastic giant cells (×400). g-h Aberrant immunostaining of ROS1 in a transbronchial biopsy with lung adenocarcinoma. g H&E stain, asterisks show tumour cells. h ROS1 IHC in adjacent hyperplastic type II pneumocytes (arrows) but not in tumour cells (asterisks) The group of patients more likely to bear a ROS1 fusion gene is largely the same as that currently recommended for testing for EGFR gene mutation and ALK gene rearrangement. Clinical features such as gender, ethnicity and smoking status are not used to select patients for EGFR or ALK testing [33,59].
Currently, however, as mentioned above, ROS1 testing is often part of a second phase of testing in a patient whose tumour is negative for more common, routinely tested alterations such as EGFR and KRAS mutation and ALK gene rearrangement and who is a lifelong never or long-time ex-smoker. This is based upon the observations that these various addictive oncogenic driver events tend to be mutually exclusive in occurrence and are almost exclusively found in adenocarcinoma. As mentioned above, as NGS techniques are introduced into routine diagnostic practice, ROS1 fusion gene testing will be provided as part of the testing 'package', whether or not this particular test was actively sought by a treating physician.
At present, when ROS1 testing is required, it will be reasonable to test the same tumours currently being selected for EGFR mutation and ALK gene rearrangement. Although this should ideally occur in parallel, this is not possible in all cases. Therefore, in order to save tissue and time, it is wise to cut extra blank sections at the first cutting session [60]. FISH remains the core test for the time being. Laboratories may use IHC as a screening tool, but with our current state of knowledge, a positive IHC test should be confirmed by FISH testing. If laboratories chose to use a multiplex PCR approach instead of FISH, they should be aware of the possible pitfalls of this highly sensitive and specific technique, in terms of sample quality and risks of test failure.
There are a number of general principles that should be observed in order to ensure high-quality laboratory testing for predictive biomarkers in NSCLC. The use of adequate control materials, awareness that test outcomes may be confounded by pre-analytical issues in tissue handling and processing and the need for laboratories to participate in external quality assessment programmes are discussed in the following sections.

Guidance on the use of controls
With ISH techniques, the case for study serves as a control when consistently presenting signals, both in tumour cells and in the accompanying normal cells (lymphocytes, fibroblasts, non-neoplastic lung epithelium). A pre-hybridisation assessment of digestion is useful in difficult samples (e.g. very small biopsies with low tumour content).
With IHC, ROS1 protein expression may also be detected in normal cells, namely histiocytes/giant cells, reactive type II pneumocyte hyperplasia and bronchiolar metaplasia at the  Multiplex amplicon RNA massive parallel sequencing Moskalev et al. [54] Personalised analysis of rearranged ends (PARE) Leary et al. [87] Anchored multiplex PCR (AMP) Zheng et al. [51] NGS next-generation sequencing tumour periphery or in subpleural areas. In most cases, the expression in these cells is weak to moderate (1+/2+ in intensity), and it is unclear whether protein stability may be affected by pre-analytical variables (e.g. time of fixation) [43]. To control for appropriate analytical conditions of ROS1 testing by IHC, it is mandatory to include a piece of tissue from a ROS1 FISH-positive tumour on the same slide of the neoplasm of interest or on a separate slide to use in the same run.

Pre-analytical variables and factors affecting quality of biopsies and surgical samples
Regarding the ROS1 epitope, the influence of pre-analytical factors has not been investigated systematically, but experiences from our group and several others [36,61] show that the protein is relatively stable and may be detected reliably by IHC. In addition, the corresponding genomic alterations can be reproducibly detected by ISH [62]. To improve nucleic acid stability, new techniques of fixation may become useful [63]. Nonetheless, attention should be paid to a number of basic requirements in order to avoid false-negative results.

Resection material
Surgical material such as lobectomies should initially be handled macroscopically (documentation and gross sectioning) following a standardised protocol. To greater standardise the work-up of resection material, vacuum preservation might be considered [64]. It is important that these procedures are conducted in a standardised way.

Biopsies
Biopsies almost always are transferred into the fixation solution immediately after removal from the patient. The small tissue fragments should be fixed for no longer than 24 h; as shown for several antigens, a gradual decrease in antigenicity may appear over time. Prior or in parallel to ROS1, several immunohistochemical, ISH and molecular markers may need to be analysed to confirm the subtype of NSCLC and the immunological and molecular profile. This means that a minimum amount of tissue/cells is required for reliable analysis.
In the context of personalised medicine, this multi-parameter analysis plays an increasing role, which has led to the suggestion to provide at least three to four endobronchial biopsy specimens for pathology.

External quality assessment
For the successful treatment of patients, it is of great importance that molecular test results are accurate, highly reliable, clearly understandable to the clinician and reported within an acceptable turnaround time [59]. In 2012, the European Society of Pathology (ESP) proposed an external quality assessment (EQA) scheme to promote high-quality biomarker testing in NSCLC for EGFR mutation analysis and ALK rearrangement detection. From 2014 onwards, ROS1 testing was also included [65]. The EQA was performed at the beginning of the development of ROS1 testing. The rate of false negativity for IHC on a limited number of ROS1-positive cases was approximately 15 %. The rate of false positivity was <10 %. For FISH, although the number of evaluable cases was limited, no false-negative scores were present in a ROS1-positive control cell line. Overall, at an early stage of ROS1 testing, the EQA showed promising results, emphasising the need for regular EQA monitoring.

Integration of ROS1 into current testing algorithms
Since clinical trials with crizotinib in ROS1-positive patients have used FISH, this method has been considered the 'gold standard' for determining ROS1 positivity by the FDA in the USA. European guidelines currently recommend ROS1 rearrangement testing in patients with advanced NSCLC who have previously tested negative for EGFR mutation and ALK rearrangement, including all stage IIIB/IV histological subtypes in non-smokers and the non-squamous cell carcinoma subtype in current or ex-smokers [33]. As the demand for ROS1 testing increases, it is reasonable that ROS1 rearrangement be considered for testing concurrently with ALK rearrangement and EGFR mutation. Cutting extra blank sections for ROS1 testing (and also for additional tests such as PD-L1) at the first cutting session is good practice to avoid tissue waste, especially when the amount of tumour tissue is scarce.
Although validation with large series is needed, IHC could also become a good preliminary, rapid screening method. An algorithm based on IHC screening with further confirmation by a ROS1 break-apart FISH assay in positive or doubtful cases seems appropriate. Nevertheless, in the near future the possibility of using transcript-based methods in a single-tube assay to detect several oncogenic fusions involving the ALK, RET, ROS1 and NTRK1 genes could drastically limit the use of IHC and FISH tests. The algorithm presented in Fig. 5 is proposed for use in routine clinical practice.

Conclusions
With the recent FDA approval of crizotinib for patients with advanced ROS1-positive NSCLC, ROS1 fusion proteins comprise one of only three oncogenic drivers in NSCLC for which an approved targeted therapy is available. Like ALK, ROS1 gene rearrangements may be detected using a break-apart FISH assay, IHC and a number of non-in situ methods. Although FISH was the methodology used in the clinical trials of crizotinib, IHC can also be used as a screening approach if carefully validated. The real-world possibility of falsepositive FISH or IHC results strengthens the case for confirmation of positive cases by a second methodology. Targeted NGS is a valid alternative if cost and turnaround time are reasonable. As testing for ROS1 becomes increasingly important for patients with advanced NSCLC, it will be key to share experience and recommendations on how to accurately implement these diagnostic methodologies into routine practice. Regardless of which testing method(s) is used, it is key that routine testing for ROS1 in the clinical setting be carefully validated, with appropriate controls and participation in EQA schemes. To achieve efficient molecular testing in NSCLC and an optimal turnaround time for test results, we propose that EGFR, ALK and ROS1 are tested for upfront and in parallel in NSCLC specimens.