A Comprehensive Comparative Analysis of the Histomorphological Features of ALK-Rearranged Lung Adenocarcinoma Based on Driver Oncogene Mutations: Frequent Expression of Epithelial-Mesenchymal Transition Markers than Other Genotype

Molecular classification of lung cancer correlates well with histomorphological features. However, specific histomorphological features that differentiate anaplastic lymphoma kinase (ALK)-rearranged tumors from ALK-negative tumors have not been fully evaluated. Eighty ALK-rearranged and 213 ALK-negative (91 epidermal growth factor receptor-mutated; 29 K-ras-mutated; 93 triple-negative) resected lung adenocarcinomas were analyzed for several histomorphological parameters and histological subtype. ALK-rearranged tumors were associated with younger age at presentation, frequent nodal metastasis, and higher stage of disease at diagnosis. ALK-rearranged tumors were more likely to show a solid predominant pattern than ALK-negative tumors (43.8%; 35/80; p<0.001). Unlike ALK-negative tumors, a lepidic predominant pattern was not observed in ALK-rearranged tumors (p<0.001). In multivariate analysis, the most significant morphological features that distinguished ALK-rearranged tumors from ALK-negative tumors were cribriform formation (odds ratio [OR], 3.253; p = 0.028), presence of mucin-containing cells (OR, 4.899; p = 0.008), close relationship to adjacent bronchioles (OR, 5.361; p = 0.001), presence of psammoma bodies (OR, 4.026; p = 0.002), and a solid predominant pattern (OR, 13.685; p = 0.023). ALK-rearranged tumors exhibited invasive histomorphological features, aggressive behavior and frequent expression of epithelial-mesenchymal transition markers (loss of E-cadherin and expression of vimentin) compared with other genotype (p = 0.015). Spatial proximity between bronchus and ALK-rearranged tumors and frequent solid histologic subtype with p63 expression may cause diagnostic difficulties to differentiate squamous cell carcinoma in the small biopsy, whereas p40 was rarely expressed in ALK-rearranged adenocarcinoma. Knowledge of these features may improve the diagnostic accuracy and lead to a better understanding of the characteristic behavior of ALK-rearranged tumors.


Introduction
Adenocarcinoma of the lung is the most common histological type of primary lung cancer [1] and is a heterogeneous tumor with diverse molecular, clinical, and pathological characteristics. Identification of molecular driver mutations and their therapeutic implications in lung adenocarcinoma have become an important area of research as evidenced by the abundance of genomic, mutational, and proteomic profiling studies. [2,3] Many studies have shown correlations between morphological features and molecular alterations in lung adenocarcinoma. Previous reports have investigated the association between epidermal growth factor receptor (EGFR) mutations and specific histological subtypes of adenocarcinoma such as lepidic (formerly known as nonmucinous bronchioloalveolar carcinoma), papillary, and micropapillary patterns. [4][5][6][7] In contrast, KRAS mutation status has been shown to be significantly associated with solid and invasive mucinous adenocarcinoma subtypes. [8,9] Therefore, these findings raise the fundamental question of whether morphological features reflect the presence of molecular alterations.
The presence of anaplastic lymphoma kinase (ALK) gene rearrangement in lung adenocarcinomas is the best predictor of response to crizotinib, an ALK tyrosine kinase inhibitor. [10,11] Fluorescence in situ hybridization (FISH) has been established as the gold standard method for the detection of ALK rearrangement in lung adenocarcinoma. The Food and Drug Administration approved crizotinib with a companion diagnostic FISH test for ALK-rearranged non-small cell lung cancer (NSCLC). Several studies have investigated the predictive value of pathological and morphological features in detecting ALK-rearranged tumors; however, the results of these studies have been inconsistent because of the limited number of ALK-rearranged tumors. [12][13][14][15][16] Solid signet-ring cell subtypes and cribriform pattern have been associated with ALK rearrangement in lung adenocarcinoma. [12,15] A few studies have reported a positive histological correlation with ALK rearrangement in lung adenocarcinoma using the new International Association for the Study of Lung Cancer, American Thoracic Society and European Respiratory Society (IASLC/ATS/ERS) classification that was published in 2011. [16,17] However, the comparative analysis of these histomorphological features and subtypes of ALK-rearranged lung adenocarcinoma based on driver oncogene mutations has not been clearly established in lung adenocarcinoma.
The aim of this study was 1) to evaluate the clinicopathological and histological features of 80 cases of ALK-rearranged resected lung adenocarcinomas, 2) to compare these features with those of ALK-negative tumors expressing well-known driver mutations associated with lung adenocarcinoma, and 3) to investigate the correlation between molecular subtype and histological features of lung adenocarcinoma based on the new IASLC/ATS/ERS classification.

Case Selection
This study was approved by the Institutional Review Board at Seoul National University Bundang Hospital. Written informed consent was specifically waived by the approving IRB in this study. A total of 80 surgically resected lung adenocarcinoma specimens harboring ALK-rearrangement were retrieved from the files of Seoul National University Affiliated Hospitals and the Asian Medical Center between January 2004 and June 2011. In addition, 213 ALK-negative resected adenocarcinoma specimens obtained from patients diagnosed between March 2009 and March 2010 were included in the study. Of the 213 ALK-negative tumors, 91 were EGFR-mutated, 29 were K-ras-mutated, and 93 were triple-negative (TN; wild-type EGFR, K-ras, and ALK). Patients who had a previous history of cancer, presurgical chemotherapy or radiotherapy were excluded. All cases were classified according to the seventh edition of the Union for International Cancer Control/American Joint Committee on Cancer TNM classification. [18] Clinicopathological information was obtained from the medical records and pathology reports.

Histological Analysis
All resected specimens were fixed with formalin and stained with hematoxylin and eosin (H&E). All slides, including those of normal lung tissue, were carefully reviewed by 2 of the authors (H.K. and J.H.C.). An average of 8.9 slides (range: 1-14 slides) from each case was reviewed. Recent reports have demonstrated a strong association of extracellular mucin and cribriform pattern with ALK-rearranged tumors. [19] Therefore, the presence of extracellular mucin and cribriform architecture and the proportion of mucin-containing cells were evaluated in ALK-rearranged and ALK-negative tumors. ALK-rearranged tumors tended to be centrally located near the bronchus; therefore, the anatomic relationship between the tumor and the bronchi was investigated. The following histological parameters were evaluated: tumor location in relation to the bronchus; tumor invasion to the bronchus; alterations in bronchial epithelial cells located adjacent to the tumor; psammomatous calcifications; cholesterol cleft; tumor size; pathological stage; and visceral pleural, vascular, and lymphatic invasion. Adenocarcinoma in situ and minimally invasive adenocarcinoma cases were excluded from the study. All invasive adenocarcinomas were categorized as lepidic, papillary, acinar, micropapillary, solid predominant, and invasive mucinous adenocarcinoma (IMA) according to the IASLC/ATS/ ERS classification [17].

Detection of ALK Gene Rearrangement
ALK rearrangement in formalin-fixed, paraffin-embedded tumor tissues was detected by FISH analysis using a break-apart probe specific for the ALK locus (Vysis LSI ALK dual-color, breakapart rearrangement probe; Abbott Molecular, Abbott Park, IL, USA). FISH-positive cases were defined as those with .15% split signals or an isolated red signal in tumor cells as described previously [20][21][22].

Detection of EGFR and K-ras Mutations
Genomic DNA was extracted from paraffin-embedded tissues. After deparaffinization with xylene, tissue sections were stained with H&E, and target lesions were selectively dissected to minimize contamination with normal tissue. Genomic DNA was isolated using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. EGFR mutations at exons 18-21 and K-ras mutations at codons 12, 13, and 61 were analyzed by nested polymerase chain reaction (PCR) and direct DNA sequencing as described previously. [23] PCR products were processed using the Big Dye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster, CA, USA), and sequence data were generated using the ABI PRISM 3100 DNA Analyzer (Applied Biosystems).

Immunohistochemistry
Immunohistochemistry was performed on tissue microarray sections. Four-micrometer-thick sections were transferred to poly-L-lysine-coated glass slides and incubated in a dry oven at 60uC for 1 h. The sections were then dewaxed in xylene (3 changes), rehydrated in a descending series of graded ethanol concentrations, and rinsed in Tris-buffered saline (TBS; pH 7.4). The endogenous peroxidase activity was blocked using 5% hydrogen peroxide in methanol for 15 min at 37uC. For antigen retrieval, the slides were placed in citrate buffer (10% citrate buffer stock in distilled water, pH 6.0) and microwaved for 10 min. Nonspecific staining was blocked using 1% horse serum in TBS (pH 7.4) for 3 min. The following primary antibodies were used: mucin-1 Dako). Immunostaining was developed using an avidin-biotinperoxidase complex (Universal Elite ABC Kit; PK-6200; Vectastain, Burlingame, CA, USA) and diaminobenzidine tetrahydrochloride solution (HK153-5K; Biogenex, San Ramon, CA, USA). Positive controls (samples with known reactivity for the antibody) and negative controls (omission of the primary antibody) were included in each assay.

Statistical Analysis
Pearson's chi-square test, Fisher's exact test, and one-way analysis of variance were used to evaluate the association of clinicopathological and histological variables with lung adenocarcinoma genotype. Multivariate logistic regression analysis was used to determine the most significant morphological features associated with ALK-rearranged tumors. All statistical tests were twosided, and a p value ,0.05 was considered statistically significant. All analyses were performed using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA).

Histomorphological Findings
Comparison of the morphological features between ALKrearranged and ALK-negative tumors. Focal cribriform formation was present in 40.0% (32/80) of ALK-rearranged tumors ( Figure 1A). Significant extracellular mucin and mucincontaining cells, including goblet cells ( Figure 1B) and signet-ring cells ( Figure 1C), were observed in 57.5% (46/80) and 62.5% (50/ 80) of ALK-rearranged tumors, respectively. Three ALK-rearranged tumor cases showed morphology similar to IMA, including a predominant lepidic pattern of goblet cell proliferation with abundant extracellular mucin. We noted that mucin containing tumor cells resemble the non-neoplastic goblet cells in segmental bronchus or bronchiole level. Several tumor glands resembled adjacent bronchial gland in morphology (Figure 2A). In some ALK-rearranged cases, tumor cells invaded the adjacent bronchiolar epithelium and showed the appearance of 'budding off' of small epithelial cell clusters into the lumen ( Figure 2B). Furthermore, flat atypical epithelial lesions that resembled adjacent tumor cells infiltrated the non-neoplastic bronchial epithelium. This close anatomic relationship with the bronchus was observed in 86.3% (69/80) of ALK-rearranged tumors. Psammoma bodies ( Figure 1D) and cholesterol clefts ( Figure 1E) were observed in 56.4% (44/80) and 40.5% (32/80) of ALK-rearranged tumors, respectively. All the morphological characteristics described above were less evident in ALK-negative tumors (p,0.001).  lung adenocarcinoma classification are summarized in Table 3. ALK-rearranged tumors showed various histological patterns. The frequency of a solid predominant pattern was significantly higher in ALK-rearranged tumors than in ALK-negative tumors (43.8%; 35/80; p,0.001). In contrast, the frequency of acinar predominant histology was significantly lower in ALK-rearranged tumors than in ALK-negative tumors (28.7%; 23/80; p,0.001). In contrast to ALK-negative tumors, lepidic predominant histology was not observed in ALK-rearranged tumors (p = 0.003). For EGFRmutated and TN tumors, acinar predominant pattern was the most frequently observed histology (EGFR-mutated: 58.2%, 53/ 91; TN: 40.9%), followed by papillary predominant pattern  Figure 3).

Discussion
In this study, we performed a detailed comprehensive analysis of the histomorphological features associated with ALK-rearranged lung adenocarcinoma based on comparisons with well-known driver oncogene mutations. We found that ALK-rearranged tumors were significantly associated with younger age at presentation, frequent nodal metastasis, and higher stage of disease at diagnosis. Furthermore, ALK-rearranged lung adenocarcinoma exhibited several histological characteristics that differentiated it from other genotypes: cribriform formation, presence of mucin-containing cells, close relationship to adjacent bronchioles, presence of psammoma bodies, and solid predominant histological subtype. Correlation of histological characteristics with molecular alterations in lung adenocarcinoma may provide a new approach to refine pathological classification and its clinical relevance. To the best of our knowledge, this is the largest comprehensive analysis comparing the histomorphological features of resected ALKrearranged tumors with other genotypes.
Our results revealed that the predominant histological subtype varied according to the status of driver mutations in ALK, EGFR, and K-ras. In contrast to ALK-negative tumors, ALK-rearranged   tumors were significantly associated with solid predominant subtype and not acinar or papillary predominant subtypes. ALKrearranged tumors exhibited aggressive behavior such as nodal metastasis, advanced disease stage at diagnosis, and lymphovascular invasion. Further, loss of E-cadherin and expression of vimentin, representing EMT phenotype, was frequently observed in ALK-rearranged tumors. EMT phenotype is the characteristic finding of ALK-rearranged tumors compared with other genotypes, and this could potentially be a contributing feature to the frequent metastases and high tumor stage seen in ALK-rearranged tumors.
In contrast, EGFR-mutated tumors were significantly associated with acinar or papillary predominant subtypes. Although the lepidic predominant subtype was not significantly correlated with EGFR mutation, a significant correlation between lepidic component presence and EGFR mutation was observed. EGFR-mutated tumors have been shown to exhibit nonaggressive behavior, such as decreased nodal metastasis, lymphovascular invasion, and EMT feature. In the present study, the frequency of acinar and solid predominant patterns was higher in K-ras-mutated tumors than micropapillary and lepidic predominant patterns; however, it is difficult to assess the pathological relevance of these findings because of the small number of K-ras-mutated tumors. TN tumors were associated with acinar, papillary, and solid predominant patterns. Together our findings indicate that there is a strong association between genetic status and histological type based on the new IASLC/ATS/ERS lung adenocarcinoma classification.
Histomorphological features specific to ALK-rearranged tumors have been reported, including cribriform formation and the presence of mucin or mucin-containing cells and psammoma bodies. [16,19] We also found that these features were strongly associated with ALK-rearranged tumors. We also identified that a close relationship to the adjacent bronchial epithelium is a unique feature of ALK-rearranged tumors. This close relationship with the bronchus was observed in 86.3% of ALK-rearranged tumors. Mucin containing tumor cells resembling the non-neoplastic goblet cells in segmental bronchus or bronchiole level were observed, and several tumor glands resembled the adjacent bronchial gland in morphology. In a few ALK-rearranged cases, tumor cells invaded the adjacent bronchiolar epithelium and showed the appearance of ''budding off'' of small epithelial cell clusters into the lumen. Furthermore, flat atypical lesions that resembled adjacent tumor cells infiltrated the non-neoplastic bronchial epithelium. ALK-rearranged tumors were more likely to be centrally located and easily obtained from the bronchoscopic biopsy procedure. Our findings suggest that ALK-rearranged tumors, in contrast to EGFR-mutated tumors, may represent non-TRU-type adenocarcinoma and therefore originate from bronchial epithelial cells. However, the expression of MUC-1, SP-A, SP-B, and SP-C was not different between ALK-rearranged and ALK-negative tumors. In contrast, TTF-1 and p63 expression was significantly different between ALK-rearranged and ALK-negative tumors, especially EGFR-mutated tumors. TTF-1 positivity was lower in ALK-rearranged tumors than in EGFR-mutated tumors, whereas p63 positivity was higher in ALK-rearranged tumors than in EGFR-mutated tumors. Although the frequency of TTF-1 positivity in ALK-rearranged tumors suggested they were of TRUtype origin, type II pneumocytes and Clara cells, which are characteristic of TRU-type, are typically negative for p63. [14] Our results suggested that a different mechanism mediates the development of ALK-rearranged tumors. We also evaluated the expression of p40 (gNp63) protein by immunohistochemistry. In contrast to p63, p40 positivity was less frequently observed in ALKrearranged tumors (2.9%). p63 and p40 have been shown to be overexpressed especially in squamous cell carcinoma of lung and regarded as a marker of squamous differentiation. [27][28][29][30][31] However, several studies reported that p63 expression was seen in variable frequency (up to 30%) and extent (10 to 70% of tumor cells) in lung adenocarcinoma, whereas p40 was rarely expressed in adenocarcinoma. [32] In addition, Sakai et al. reported 7 out of 9 ALK-rearranged tumors expressed p63, but none of ALK-  rearranged tumors expressed p40. [33] Our results are similar to those of these studies. We suggested that overexpression of p63 might have functional roles related with carcinogenesis or tumor differentiation rather than squamous markers in lung adenocarcinomas, but further investigation of the significance of p63 expression in lung adenocarcinoma is warranted. In the diagnostic point of view, spatial proximity between bronchus and ALKrearranged tumors and frequent solid histologic subtype with p63 expression may cause diagnostic difficulties to differentiate squamous cell carcinoma in the small biopsy, whereas p40 was rarely expressed in ALK-rearranged adenocarcinoma. Awareness of these features may help pathologists diagnose accurately.
In conclusion, ALK-rearranged lung adenocarcinoma exhibited distinct clinicopathological and morphological features compared with other genotypes. Knowledge of these features may improve the diagnostic accuracy and lead to a better understanding of the characteristic behavior of ALK-rearranged tumors.