Eighty-one patients who underwent surgical treatment of lung SCC between 1993 and 2016 were randomly selected from the Severance Hospital (Seoul, Korea) lung cancer database. To obtain clinical data, we retrospectively reviewed the patients' electronic medical records. Tumor stage was re-evaluated according to the seventh edition of the American Joint Committee on Cancer TNM Staging Manual14. This study was approved by the Institutional Review Board (IRB) of Severance Hospital (No. 3-2016-0019).

All tissue slides were subjected to hematoxylin and eosin staining and evaluated for the presence of the following recently identified poor prognostic factors: tumor budding, single cell invasion, and large nuclei5. Initially, the entire tumor set was scanned at ×100 magnification and subjected to a detailed review. First, tumor budding, or the presence of small tumor nests comprising fewer than five tumor cells, were counted in 10 high-power fields (HPFs) at ×200 magnification. We defined a high grade of tumor budding as more than eight tumor budding events per 10 HPFs (Figure 1A, B). Single cell invasion was also evaluated at ×200 magnification (Figure 1C), and nuclear features were assessed at ×400 magnification. We calculated the average nuclear diameter of at least 100 tumor cells in at least three HPFs per sample. A large nucleus was defined as a diameter greater than that of four small lymphocytes (Figure 1D).

The keratinization grade was determined, and tumors were classified accordingly as the keratinizing subtype, defined as a keratinizing pattern comprising ≥5% of the entire tumor, or the nonkeratinizing subtype, defined as a keratinizing pattern comprising <5% of the tumor (Figure 1E). Subsequently, the keratinization grades were refined to nonkeratinization (<5%) and low (5%–20%), moderate (20%–50%), or severe (>50% of the entire tumor) keratinization, similar to the classification used for head and neck cancers15. The keratinization patterns included cytoplasmic keratinization (Figure 1F), keratin pearl (Figure 1G), and layered keratinization (Figure 1H). All histologic parameters were evaluated independently by a pathologist (Y.J.C.).

To evaluate mTOR complex 1 (mTORC1) and Bcl-xL expression, immunohistochemistry (IHC) staining for pS6 and Bcl-xL proteins was performed using an EnVision+ system (Dako Corp., Carpinteria, CA, USA) according to the manufacturer's instructions. Briefly, sections were deparaffinized, rehydrated, and subjected to antigen retrieval via microwave heating for 10 minutes. Sections were then immersed in a H₂O₂–phosphate-buffered saline solution prior to overnight incubation with primary anti-pS6 (1:400, Cell Signaling Technology, Danvers, MA, USA) or Bcl-xL antibodies (1:600, Cell Signaling Technology). Subsequently, the sections were incubated with a peroxidase-labeled polymer for 1 hour at 4℃. IHC staining was scored independently at ×200 magnification by Y.J.C. and S.H.K., who were blinded to the clinicopathological data. A semiquantitative evaluation of pS6 and Bcl-xL was performed according to the method described in a previous study16. The staining intensity was classified as 0 (negative), 1 (trace), 2 (moderate), or 3 (strong), and the frequency of positive cells was classified as 0 (<10%), 1 (10%–50%), 2 (50%–80%), or 3 (>80%). The expression score was determined as the product of the staining intensity and frequency of positive cells.

Associations between categorical variables were analyzed using the chi-square test; this test was also used to assess the linear correlation trend of the keratinization grade with IHC scoring. Disease-free survival (DFS) and OS were estimated using the Kaplan-Meier method, and associations between factors and survival outcomes (OS) were analyzed using the log-rank test. The Cox proportional hazards model was used for multivariate analyses. SPSS version 18.0 (SPSS Inc., Chicago, IL, USA) was used for all statistical analyses, and significance was defined as a p-value of <0.05.

Among the 81 patients enrolled in this study, 56 (69.1%) and 25 (30.9%) were classified as having the keratinizing and nonkeratinizing subtypes of lung SCC, respectively, as confirmed by a pathologic review of tumor slides. The average age of the subjects was 62.8±8.7 years (mean±standard deviation); 77 (95.1%) were male, and the majority had stage T2 or T3 disease (84.0%). Among patients with the keratinizing subtype, 21 (37.5%), 19 (33.9%), and 16 patients (28.6%) were classified as having low, moderate, and severe keratinization, respectively. The keratinizing and nonkeratinizing groups did not differ significantly with respect to the distributions of age, sex, T-, N-, pStage, type of surgery, and adjuvant chemotherapy. However, most patients with the keratinizing subtype (98.0%) had a history of smoking, whereas never smokers (24.0%) were significantly more prevalent in the nonkeratinizing subtype group (Pearson's R=0.262, p=0.008, chi-square test). In a subgroup analysis of smokers, keratinization was more frequently observed among heavy smokers (i.e., ≥30 pack-year history of smoking) (p=0.012, chi-square test). This finding suggests that keratinization of lung SCC might be related to smoking (Table 1). Other morphologic parameters, such as nuclear diameter, tumor budding, and single cell invasion at the tumor edge, were not found to correlate with the keratinization of lung SCC.

Because mTOR signaling is known to regulate keratinocyte differentiation, we reviewed the relationship between keratinization and mTORC1 in 51 available lung SCC tissues, using pS6 expression as a surrogate marker. Among these 51 cases, five (9.8%), 12 (23.5%), and 23 (45.1%) exhibited strong, moderate, and trace expression, respectively. An additional 11 cases (21.6%) were negative for pS6. Viable tumors expressed pS6 beneath the shedding area, where keratinization initiates, regardless of keratinization subtype (Figure 2). However, the correlation between keratinization grade and pS6 expression failed to reach statistical significance (p=0.153, chi-square test).

In tumors, keratinization is generated by the epithelial layer and is subsequently shed from the tumor margin by apoptosis. Under a hypothesis that Bcl-xL, a key anti-apoptotic molecule, might inhibit keratinization, we investigated the relationship between keratinization and apoptosis in the keratinized area. Although we observed a negative correlation between Bcl-xL and tumor cell keratinization, this relationship did not reach statistical significance (Pearson's R=−0.264, p=0.062, chi-square test) (Figure 2).

We next analyzed the clinical outcomes of lung SCC according to keratinization. Although the keratinizing subtype was associated with a reduced DFS, this difference did not reach statistical significance (119.6 months vs. 122.7 months, p=0.459, log-rank test). However, the keratinizing subtype was associated with significantly shorter OS relative to the non-keratinizing subtype (85.2 months vs. 135.7 months, p=0.010, log-rank test) (Figure 3).

In a univariate analysis, old age (p=0.039) and a history of smoking (p=0.036) were associated with a significantly shorter OS. Among histologic parameters, a high tumor budding grade (p=0.037), presence of single cell invasion (p=0.006), and keratinizing subtype (p=0.013) were associated with a significantly poorer OS. To confirm keratinization as an independent prognostic factor for OS in lung SCC, we performed a multivariate analysis that included age, sex, smoking status, tumor stage, and keratinization. This analysis identified older age (>65 years; hazard ratio [HR], 2.154; 95% confidence interval [CI], 1.150–4.035; p=0.017), advanced stage (stage III; HR, 2.558; 95% CI, 1.207–5.422; p=0.014), and keratinization (HR, 2.389; 95% CI, 1.090–5.233; p=0.030) as independent poor prognostic factors (Table 2).