Advanced lung adenocarcinomas with ROS1-rearrangement frequently show hepatoid cell

Defining distinctive histologic characteristics of ROS1-rearranged non-small-cell lung carcinomas (NSCLCs) may help identify cases that merit molecular testing. However, the majority of previous reports have focused on surgical specimens but only limited studies assessed histomorphology of advanced NSCLCs. In order to identify the clinical and histological characteristics of ROS1-rearranged advanced NSCLCs, we examined five hundred sixteen Chinese patients with advanced NSCLCs using ROS1 fluorescence in situ hybridization and real-time polymerase chain reaction and then analyzed for clinical and pathological features. We performed univariate and multivariate analyses to identify predictive factors associated with ROS1 rearrangement. 19 tumors were identified with ROS1 rearrangement (3.7% of adenocarcinomas). 16 ROS1+ and 122 ROS1- samples with available medical records and enough tumor cells were included for histological analysis. Compared with ROS1-negative advanced NSCLCs, ROS1-rearranged advanced NSCLCs were associated with a younger age at presentation. ROS1 rearrangements were not significantly associated with sex, smoking history, drinking history and metastatic sites. The most common histological pattern was solid growth (12/16), followed by acinar (4/16) growth. 66.7% cases with solid growth pattern showed hepatoid cytology (8/12) and 75% cases with acinar growth pattern showed a cribriform structure (3/4). 18.8% cases were found to have abundant extracellular mucus or signet-ring cells (3/16). Only one case with solid growth pattern showed psammomatous calcifications. In conclusion, age, hepatoid cytology and cribriform structure are the independent predictors for ROS1-rearranged advanced NSCLCs, recognizing these may be helpful in finding candidates for genomic alterations, especially when available tissue samples are limited.


INTRODUCTION
Lung carcinomas are the leading cause of cancerrelated mortality worldwide [1]. Adenocarcinoma is the most common histologic type of lung cancer [2]. With the development of targeted therapy, adenocarcinoma are classified into subsets, such as epidermal growth factor receptor (EGFR)-mutated type [3,4], anaplastic lymphoma kinase (ALK)-rearranged type [5] and ROS1rearranged type [6]. The ROS1-rearranged non-small-cell lung carcinomas (NSCLCs) have been reported to be sensitive to the kinase inhibitor crizotinib [7].
Correlation between histology and genotype has been described in lung adenocarcinomas. For example, several studies have suggested that ALK-rearranged lung adenocarcinomas were most likely to be observed in the cases with signet-ring cells [8][9][10][11]. Boland et al found that 26% pulmonary adenocarcinoma with signet ring cell features were ALK rearrangement [8]. However, comprehensive clinical and histological characteristics of ROS1 rearrangement were not well known given the rarity of this subtype. Data describing the morphologic features of ROS1-rearranged lung adenocarcinomas are primarily based on analysis of resected primary Research Paper www.impactjournals.com/oncotarget lung tumors [12][13][14]. However, approximately 70% of patients are at advanced stage when diagnosed. For these patients, the small biopsies or cytological specimens are the only limited pathologic material available for diagnosis and molecular testing. It is still unclear whether the findings from prior studies are applicable to specimens from metastatic lung tumors, small biopsies or cytology specimens.
In this study, we confirmed 19 advanced patients with ROS1 rearrangement by real-time PCR and ROS1 FISH and compared their clinical features and histological characteristics with ROS1-subtype. Furthermore, we identified independent predictors for ROS1 rearrangement by multiple logistic regression analysis.

Clinical Characteristics of patients with ROS1rearranged NSCLCs
Of the 516 consecutive patients from the advanced NSCLCs, ROS1 rearrangements were detected in 19 cases (3.7%) using real-time PCR, 488 cases were negative and 9 cases failed. The tissues were available for FISH in 513 of 516 cases. One lymph node biopsy and two biopsies can't be detected by FISH because of few tumor cells and twelve cases were not interpretable. 18 cases were confirmed positive in both real-time PCR and ROS1 FISH. One lymph node biopsy was eliminated because of few tumor cells. In ROS1 FISH positive patients (n=18), a split pattern was observed in 17 tumors (94.4%), and an isolated red pattern was observed in one tumors (5.6%). One lymph node biopsy with few tumor cells and two cell blocks were not included in histological assessment. The clinical features of 16 patients with ROS1-rearranged were summarized in Table 1. ROS1-rearranged advanced NSCLCs were associated with a younger age at presentation. Patients with ROS1-rearranged tumors had an age range of 38-67 years and mean age of 50 years (compared with mean age of 59 years for patients with ROS1-wildtype tumors, P=0.001). ROS1 rearrangements were not significantly associated with sex, smoking history, drinking history and metastatic sites ( Table 2).

Multiple Logistic Regression Analysis
We performed multiple logistic regression analysis to identify the independent predictors for ROS1 rearrangement status. The parameters (age, hepatoid cytology, cribriform structure and cytological atypia, P < 0.1) were included in multivariate analysis. Age, hepatoid cytology and cribriform structure were identified as independent predictors for ROS1 rearrangement. The significant prediction factors for ROS1 rearrangement were shown in Table 3.

DISCUSSION
Our analysis of 138 advanced NSCLC cases with available clinical data and enough tissue identified age, hepatoid cytology and cribriform structure as the independent predictors for ROS1 rearrangement, which may help to find candidates for genomic alterations, especially when available tissue samples are limited. Our study revealed that 3.7% (19 out of 516) of advanced lung adenocarcinomas harbored ROS1 rearrangement. This prevalence was a little higher than that of the entire adenocarcinoma cohort. This may due to all our patients were in IV stage. Most literature reported ROS1 rearrangement in 1-2% adenocarcinoma lung, but more than 3% in IV-stage lung adenocarcinoma. Bergethon et al screened 1,073 patients with NSCLC, 18 (1.7%) were ROS1-rearranged by FISH [15]. In IV-stage lung adenocarcinoma, 3.4% (11/327) were ROS1rearranged. Kim et al found 3 (3.5%) of 86 IV-stage lung adenocarcinomas harbored ROS1 rearrangements [16]. What's more, 3 cases in 98 IV-stage lung adenocarcinomas (3/98, 3.1%) were ROS1 rearrangement in the cohort of Sholl et al [17]. Many reports also found that ROS1 rearrangement was enriched in advanced group. Scheffler et al found 19 patients of 1035 (1.8%) had ROS1 rearrangement [18]. The majority of patients presented   [19].
A correlation between histomorphology and ROS1rearranged lung adenocarcinomas has been demonstrated in prior reports. The majority of the reports have focused on surgical specimens, but only limited studies assessed histomorphology of biopsies, which are procured frequently, especially from patients presenting at an advanced stage of disease and for whom crizotinib is recommended. To the best of our knowledge this is the largest series of ROS1-rearranged advanced cases studied to clarify the clinicopathological features. We found that ROS1-rearranged advanced lung adenocarcinomas were characterized by two growth patterns: solid growth (12/16) and acinar (4/16) growth. A lepidic pattern was absent in our series, which was similar with other studies [13,20]. Noguchi suggested that lung adenocarcinoma develop  metastasis (B-D), hepatoid tumor cell shows abundant eosinophilic cytoplasm, round nuclei and prominent nucleoli; the acinar-predominant growth pattern with cribriform features in a lymph node metastasis (E) and a primary lung tumor (F) www.impactjournals.com/oncotarget from atypical adenomatous hyperplasia (AAH) through lepidic pattern, and finally invasive adenocarcinoma, being the solid pattern the less differentiated adenocarcinoma component [21]. Infrequent existence with lepidic pattern in ROS1-rearranged NSCLCs may lead to the hypothesis that ROS1 may be a strong oncogenic driver in NSCLCs, and the oncogenic drive of the ROS1 fusion gene might be powerful enough to bypass the lepidic phase.
The majority of ROS1-rearranged advanced lung adenocarcinomas in our series showed a solid growth pattern (75%, 12/16). Interestingly, solid pattern often possessed hepatoid cell (66.7%, 8/12), which showed abundant eosinophilic cytoplasm, round nuclei, and prominent nucleoli. This finding was in agreement with previous reports [22]. Khozin et al reported a patient with hepatoid carcinoma of the lung harbored ALK gene rearrangement [23]. Hepatoid adenocarcinomas have been defined as AFP-producing adenocarcinomas with morphologic features similar to hepatocellular carcinoma [24][25][26]. But 8 cases in our study were negative for antibodies Hep, AFP and Gpc3, and serum alphafetoprotein (AFP) were also low (0.6-2.8ng/ml). These finding suggested that these 8 cases with hepatoid cell were not hepatoid adenocarcinomas. We defined them as "poorly differentiated adenocarcinomas with solid growth pattern and eosinophilic cytoplasm". Careful attention to this feature may lead to find the candidate of advanced NSCLC cases with ROS1 rearrangement.
Immunohistochemically, 5 of our cases coexpressed TTF-1 and p63, and 4 of them were positive for p40, and 2 cases were also positive for CK5/6. P40, an isoform of p63, has been proposed to be a specific marker of squamous cell carcinoma. In contrast with p63, the expression of p40 was more specific for squamous cell carcinoma [27]. Recent studies reported that 4 cases exhibited unequivocal squamous differentiation, coexpression for TTF1 and p63 [12,20,28]. Yonshida et al suggested TTF-1/p63 co-expression occurred frequently in ALK-positive NSCLCs [29]. But actually Rekhtman et al found that only 5.4% lung adenocarcinomas coexpress TTF-1and p63 in general [30], even less for TTF1 and p40 co-expression. Zhao et al found that ALK, ROS1 or RET gene rearrangements appeared 0 times in 214 cases of lung squamous cancer [31], suggesting that the prevalence of ROS1 rearrangement is very low in SCC histology. We suppose that histology of ROS1+ lung cancer has low differentiation with peculiar immunoprofile, which may indicate that the cell of origin of ROS1+ lung adenocarcinoma may different from most lung adenocarcinomas.
In our series, signet-ring cells presented only in two cases (2/12), which seems lower in previous reports with surgical cases (33-55%) [12,20], and much lower than ALK-rearranged lung adenocarcinoma (71%) [11]. Furthermore, only one case with solid growth pattern showed psammomatous calcifications, while Sholl et al found psammomatous calcifications were highly associated with ROS1 rearrangement (6 of 9 or 66%vs. 1 of 185 or 0.5%; P<0.001) as compared with ROS1 wild type [17]. These differences may be due to the differences in specimen source (lung resection vs. biopsy) and tumor stage (early vs. advanced). Some features frequently presented in lung resections may be missed because of the limited tissue samples. Thus, it is essential to find out the features of advanced lung adenocarcinoma with ROS1 rearrangement.
The results of our study highlight that hepatoid cytology and cribriform structure may be the key morphologic features associated with advanced lung adenocarcinomas with ROS1 rearrangement. The recognition of hepatoid cytology and cribriform structure could help to triage a specimen for appropriate molecular testing, especially when tissue is limited. However, the diagnosis of ROS1-rearrangement needs to be validated by FISH or real-time PCR.

Patients and samples
The study was approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang University. The Ethics Committee waived the need for consent for the use of the samples in this research study. Our study group consisted of 516 consecutive cases of IV-stage lung adenocarcinoma diagnosed at the First Affiliated Hospital of Zhejiang University between 2014 and 2016, 19 of which were ROS1+ and 497 of which were ROS1-. ROS1 rearrangements were detected using realtime PCR and FISH, as described below. The specimen types consisted of small biopsies (n=499) and cell blocks (n=13) or Ebus (n=4). The clinical characteristics of all

Fluorescence In Situ hybridization
Slices of FFPE tissues, 4μm thick, were used to evaluate the presence of ROS1 gene fusion by FISH, using a break-apart probe for ROS1 (6q22 ROS1 Break Apart FISH Probe; Abbott Molecular, Des Plaines, IL, USA). Nuclei were counterstained with DAPI (Zytovision, Bremerhaven, Germany) and slides were examined with a BX51 fluorescence microscope (Olympus, Tokyo, Japan) equipped with a triple-pass filter (DAPI/Green/Orange). Samples were considered to be ROS1 positive if more than 15% of tumor cells showed split red and green signals (signals separated by ≥1 signal diameter) or single 3' signals. Otherwise the samples were considered as FISH negative.

Real-time PCR
ROS1 was amplified by multiplex real-time PCR using a Stratagene Mx3000P real-time PCR system (Stratagene, CA) with an AmoyDx® ROS1 fusion gene detection kit (Amoy Diagnostics Co., Ltd, Xiamen, China). All types of ROS1 gene fusions detected with the detection kit were listed in Supplementary Table S2. A case was considered positive for ROS1 rearrangement if FAM Ct value was <30. Otherwise the samples were considered as negative. The sample may contain two or more fusion patterns simultaneously. The status of the EGFR mutation and ALK rearrangement were also analyzed by real-time PCR and IHC, according to methods previously described [34].

Statistical analysis
Statistical analysis was performed using SPSS software version 17.0 (SPSS Inc., Chicago, IL). Continuous variables were analyzed using Student T test. The categorized variables were analyzed using Chi-square test. Multivariate analysis was performed to determine the discriminating power of each histologic parameter between ROS1-positive and ROS1-negative cohorts by multiple logistic regressions. P<0.05 was considered statistically significant.