Clinical outcomes in ALK-rearranged lung adenocarcinomas according to ALK fusion variants

Background Clinical outcomes of anaplastic lymphoma kinase (ALK)-rearranged non-small cell lung cancer according to ALK fusion variants are not clear. We aimed to investigate the prevalence of ALK fusion variants and to compare clinical outcomes according to ALK fusion variants. Methods A retrospective analysis was conducted on patients with advanced ALK-rearranged adenocarcinoma treated with chemotherapy and ALK inhibitors. ALK rearrangement was identified by fluorescence in situ hybridization and confirmed by immunohistochemistry. Peptide nucleic acid-mediated quantitative polymerase chain reaction assays, designed to detect 28 types of echinoderm microtubule-associated protein-like 4 (EML)-ALK rearrangements, were performed. Clinicopathological analysis and treatment outcomes with platinum-based chemotherapy, pemetrexed therapy, and ALK inhibitors—including crizotinib and ceritinib—were evaluated. Results A total of 52 patients with ALK-rearranged lung adenocarcinoma were enrolled. EML4-ALK variant 1 (v1) was the most common variant (38.5 %) followed by the non-EML4 variant (36.5 %), EML4-ALK variant 3a/b (19.2 %), and EML4-ALK variant 2 (5.8 %). No clinicopathological distinction was found between the different ALK fusion variants. Treatment response rates for each therapeutic agent did not differ according to ALK fusion variant. However, EML4 variants, especially v1, showed significantly longer progression-free survival (PFS) on pemetrexed treatment than did non-EML4 variants (median 31.1 months versus 5.7 months, P = 0.003). PFS with platinum-based chemotherapy and ALK inhibitors did not differ according to ALK fusion variant. Multivariate survival analysis using Cox’s regression model revealed v1 as the only predictive factor for prolonged PFS on pemetrexed. Conclusions Among ALK fusion variants, v1 is the most common subtype. It showed superior progression-free survival on pemetrexed than did non-EML4 variants. No survival difference was demonstrated between variants treated with crizotinib or ceritinib. Electronic supplementary material The online version of this article (doi:10.1186/s12967-016-1061-z) contains supplementary material, which is available to authorized users.


Background
Anaplastic lymphoma kinase (ALK) rearrangements, found in approximately 5 % of non-small cell lung cancers (NSCLCs), are relatively rare genetic alterations compared with epidermal growth factor receptor (EGFR) or KRAS mutations [1]. Soda et al. identified the echinoderm microtubule-associated protein-like 4 (EML4)-ALK fusion gene, and reported its transforming activity and potential as a therapeutic target in NSCLCs [2]. Subsequently, following reports of dramatic therapeutic effects of crizotinib on ALK-rearranged NSCLCs [3,4], a number of studies on the clinicopathologic characteristics of ALK-rearranged NSCLC have been conducted [5][6][7][8]. Currently, the fluorescence in situ hybridization (FISH) method is considered the gold standard for establishment of ALK-rearrangement positivity. In addition, immunohistochemistry (IHC) for ALK protein is known to have high sensitivity and specificity for recognition of ALK rearrangements and is strongly correlated with FISH results [9,10]. However, FISH and IHC cannot specify the different variants or fusion gene partners of the ALK gene, which can be identified by real time-polymerase chain reaction (RT-PCR) or next-generation sequencing technology. Crizotinib is effective for NSCLC patients harboring ALK rearrangements (~60 % of patients achieve an objective response) but almost all experience disease progression after 8-11 months [3,11,12]. We hypothesized that different ALK fusion variants would lead to different treatment responses. In the present study, we investigated the prevalence of ALK fusion partners in NSCLCs, and explored whether the efficacy of therapeutic agents differs according to ALK fusion variant.

Methods
A retrospective analysis was conducted on patients with advanced ALK-rearranged adenocarcinoma treated with chemotherapy and ALK inhibitors. This retrospective study was approved by the Institutional Review Board of Severance Hospital (No. 4-2015-0926).

Clinicopathologic analysis
The following clinicopathologic parameters were recorded: age, sex, smoking status [never smokers, former smokers (quit smoking >1 year before diagnosis), and current smokers], pack-year smoking history (defined as the number of cigarette packs smoked per day multiplied by the number of years of smoking), sites of metastasis, and pathological tumor stage at diagnosis. For histological analysis, intra-and/or extracellular mucin, ALK-related growth patterns including cribriform and solid signet ring cells, features of nuclei, and psammomatous calcification were examined. All samples were reviewed by experienced pulmonary pathologists (Y.J.C. and H.S.S.). Treatment methods, treatment responses, overall survival, and progression-free survival (PFS) were assessed. Tumor response was determined according to Response Evaluation Criteria in Solid Tumors, version 1.1 [13].

EGFR and KRAS mutation analysis
To determine the EGFR and KRAS mutation status, DNA was extracted from formalin-fixed, paraffin-embedded (FFPE) tissues using the DNeasy Isolation Kit (Qiagen, Valencia, CA, USA), according to the manufacturer's instructions. For the EGFR gene, direct DNA sequencing of exons 18-21 was performed using the PNA Clamp ™ EGFR Mutation Detection Kit (PANAGENE, Daejeon, Korea). For the KRAS gene, direct DNA sequencing of codons 12 and 13 was performed. Each tumor was classified as positive or negative for a mutation after comparison with the wild-type gene sequence.

ALK fluorescence in situ hybridization and immunohistochemistry
To identify ALK rearrangements, FISH was performed using a break-apart ALK probe (Vysis LSI Dual Color, Break Apart Rearrangement Probe; Abbott Molecular, Abbot Park, IL, USA). ALK rearrangement was scored as positive when >15 % of tumor cells displayed split or isolated signals containing a kinase domain. IHC was performed using an ALK antibody (rabbit monoclonal, clone D5F3, Cell Signaling Technology, Danvers, MA, USA) and Ventana automated immunostainer BenchMark XT (Ventana Medical Systems, Tucson, AZ, USA), as previously described [14].

RNA extraction and cDNA synthesis
Total RNA was extracted using the PureLink ™ FFPE Total RNA Isolation Kit (Invitrogen Carlsbad, CA, USA) with the following protocol modifications. The resulting RNA was eluted in 50 µL of elution buffer. The concentration and purity of the extracted RNA were determined by spectrophotometry. The extracted RNA was stored at −80 °C until required. We used 250 ng of total RNA to generate cDNA using the Super Script VILO cDNA synthesis kit (Invitrogen).

Statistical analysis
Clinicopathologic parameters were compared using the Chi square (for categorical parameters) and Mann-Whitney U (for continuous parameters) tests. Survival was evaluated using the Kaplan-Meier method, and statistical differences in survival times were determined using the log-rank test. A Cox proportional hazards model was used to assess risk factors for PFS of each therapeutic agent. Statistical analyses were performed using SPSS 19.0 (SPSS Inc. Chicago, IL, USA), and a P value <0.05 was considered significant.

Patient selection
Between March 2000 and February 2015, ALK-rearrangement was confirmed in 76 patients at our institution; all had adenocarcinomas. Results of RT-PCR analysis for ALK fusion partners were available for 52 of the 76 patients. These 52 patients were included: 19 were diagnosed by transbronchial lung biopsy or fiber optic bronchoscopy biopsy of the primary lung tumor, 6 by endobronchial ultrasound lymph node biopsy, 16 by lobectomy or wedge resection of the primary lung tumor, 5 by pleural biopsy, and 5 by biopsy of distant metastatic lesions.
Among the ALK fusion variants, tumors with EML4-ALK variants showed more frequent lung and/or pleural involvement without distant metastasis compared with the non-EML4 variants (P = 0.015). Among the 22 EML4 variant tumors with lung and/or pleural involvement, the majority were v1 (N = 13, 65.0 % of v1) and v3a/b (N = 8, 80.0 % of v3a/b). Other clinicopathologic parameters and histologic features did not differ according to ALK fusion variant.
Most patients received platinum-based chemotherapy or pemetrexed therapy as first-line chemotherapy, before receiving crizotinib or ceritinib. Single-agent pemetrexed, EGFR tyrosine kinase inhibitors (TKIs), and single-agent platinum therapy were used as second-line or further lines of therapy. Patient treatment history before ALK inhibitor use is summarized in Additional file 1: Table S2. Overall, 37 patients received ALK inhibitors, including crizotinib (N = 32, 61.5 %), ceritinib (N = 14, 26.9 %), and alectinib (N = 2, 3.8 %). ALK inhibitors were administered as second-or further-line therapy in most patients, except for 8 patients who received crizotinib (N = 7) and ceritinib (N = 1) as first-line therapy. Five patients (two v1 and three v3a/b), who received crizotinib showed PD while most patients showed at least SD and PR to ALK TKI treatment. Overall, the ORR was 53.1 % with crizotinib and 57.1 % with ceritinib. Treatment response rates to ALK inhibitors did not differ according to ALK fusion variant ( Table 2).

Efficacy of chemotherapy regimens and treatment response, according to ALK fusion partners
A 42 year-old woman, who harbored a v2 variant, showed complete response (CR) to ceritinib. Her brief clinical history and detailed histologic features are summarized in Additional file 1: Fig. S1.

Progression-free survival with each therapeutic agent, according to ALK fusion variant
In patients who received first-line platinum-based chemotherapy, there was no significant difference in PFS according to ALK fusion variant (Fig. 2). With regard to pemetrexed, EML4-ALK fusion variants showed significantly superior PFS compared to non-EML4 variants (Fig. 3a). When further analyzing the subtypes of EML4-ALK variants, v1 showed significantly better PFS than did the others (Fig. 3b). No significant difference according to ALK fusion variant was found in PFS of patients treated with crizotinib or ceritinib (Fig. 4). In the univariate Cox proportional hazards analyses for PFS on pemetrexed, v1 was determined to be a predictive factor for prolonged PFS, while non-EML4 was identified as a risk factor for shorter PFS. However, in the multivariate analysis, v1 was the only significant predictive factor of longer PFS (Table 3).

Discussion
In the present study, patients harboring the EML4-ALK variant, especially v1, had significantly prolonged PFS on pemetrexed therapy, whereas no difference in PFS was observed for those treated with ALK inhibitors, according to the ALK fusion variants. Since FISH was used as the gold standard method for enrollment in clinical trials of ALK inhibitors, information on ALK fusion variants was limited in previous studies, and the efficacy of chemotherapy or targeted therapy according to fusion variant was not established. In the present study, we confirmed the ALK rearrangements in patient with NSCLCs using FISH and IHC and additionally specified the EML4-ALK variants using RT-PCR, which identifies the largest number of EML4-ALK variants to date.  In the present study, EML4-ALK fusion v1 was the most common variant, identified in 38.5 % of all patients and accounting for 60.6 % of all EML4-ALK variants. This finding is consistent with previous studies [15,16]. Non-EML4 variants were the second most common, identified in 36.5 % of patients, which is slightly higher than that previously reported [3,16]. Although the RT-PCR methods used in the present study were designed to detect 28 types of EML4-ALK rearrangements, only v1, v2, and v3a/b were identified in our patients. Among the EML4-ALK variants, v3a/b and v2 were the second and third most common types, as previously described [3,15].

Table 1 Clinicopathologic characteristics and histological analysis
Most of the patients in the present study received ALK inhibitors as second-or further-line treatment, and the ORR to ALK inhibitors was far better than to platinumbased or pemetrexed chemotherapy. Crizotinib was used most commonly, followed by ceritinib and alectinib. Crizotinib was initially developed as a c-Met inhibitor, but was found to be an efficient inhibitor of ALK phosphorylation and signal transduction [17], and to be effective in the treatment of ROS1-rearranged NSCLCs [18,19]. Ceritinib [20,21] and alectinib [22,23] are second-generation ALK inhibitors that can be used in patients with crizotinib resistance or intolerance. In the present study, the treatment response rate to ALK inhibitors was no different between the EML4 group and non-EML4 group, consistent with a previous study [3]. There was also no difference among EML4-ALK variants. Recently, during preparation of our manuscript, Yoshida et al. reported on the frequency of ALK fusion variants and the therapeutic efficacy of crizotinib according to the different variants in patients with ALK-rearranged NSCLCs; this approach was similar to that of our study [16]. They evaluated 35 patients with ALK-rearranged NSCLCs, and found that v1, the most common variant, was associated with superior PFS on crizotinib than non-v1 variants. This was not observed in our study. Further studies are required to investigate these discrepant findings. These two studies had similar limitations that could affect study results: both were retrospective studies with a small sample size. In addition, the treatment line for crizotinib differed across studies.
Although Yoshida et al. did not evaluate the therapeutic efficacy of second-generation ALK inhibitors such as ceritinib, we examined the therapeutic efficacy of ceritinib according to ALK fusion variants, and found no difference in the use of ceritinib as crizotinib. Notably, however, one patient with v2 variant achieved a CR on ceritinib treatment. The tumor of this patient exhibited an extensive papillary and micropapillary pattern with partly retained alveolar wall architecture, distinguishing it histologically from the usual pattern of ALK-rearranged tumors. In the present study, all v2 patients showed PR to crizotinib and CR to ceritinib. Moreover, both the ORR and DCR were 100 %, although v2 was found in only 3 patients-too small a number to draw a general conclusion. A previous in vitro study reported v2 as being the most sensitive to ALK inhibition, and explained that v2 has a shorter half-life compared with the other variants and is far more unstable since it has the longest N-terminus of EML4 [24].
Pemetrexed, a folate antimetabolite that inhibits enzymes used in purine and pyrimidine synthesis, has been approved for malignant mesothelioma and NSCLC of non-squamous histology. Previous studies demonstrated an association between ALK rearrangement in NSCLC and prolonged PFS in patients treated with pemetrexed [25,26]. The studies measured mRNA level of thymidylate synthase (TS), one of the catalytic enzymes thought to reduce sensitivity to pemetrexed, and showed a significantly lower TS mRNA level in ALK-rearranged tumor cells [26]. However, a more recent study with a larger cohort refuted these observations: it showed similar PFS of patients with ALK-rearranged and ALK-wild type NSCLCs [27]. They also measured mRNA level of TS and concluded that ALK rearrangement was associated with a lower TS mRNA level, but that PFS on pemetrexed treatment was not affected by ALK status  [27]. So far, all previous studies on pemetrexed efficacy confirmed ALK rearrangement using the FISH method. Thus, frequency of ALK fusion variants in each study was unknown [25][26][27]. One possible scenario for the observed discrepancy regarding pemetrexed efficacy is that proportion of v1 might have differed in each previous study, because in the present study pemetrexed showed a significantly better PFS when used to treat v1 variants than when used to treat other variants. Although we could not examine the TS level in tumors because the  remaining tumor tissue was not available due to previous extensive molecular examination, further validation is needed to clarify the mechanism of prolonged PFS of v1 on pemetrexed observed in the present study. Although crizotinib is the most efficient and verified target agent for patients with ALK-rearranged NSCLC, pemetrexed would be a good treatment option if patients harbor the v1 variant and cannot afford crizotinib. Subtyping ALK variants might predict the efficacy of pemetrexed.

Conclusions
In conclusion, our study showed different PFS on pemetrexed treatment according to ALK fusion variant in lung adenocarcinoma. EML4-ALK variants, especially v1, had