Low-level expression of necroptosis factors indicative of a poor prognosis of squamous cell carcinoma subtype of non-small-cell lung cancer

Background A programmed cell death pathway, necroptosis may synergize with DNA damage response (DDR) in opposing tumor progression. While our basic mechanistic understanding of necroptotic cell death advances rapidly, its prognostic implications have not been thoroughly examined in cancer. Methods We measured expression level of nine proteins involved in necroptosis and DDR in primary stage I non-small-cell lung carcinoma (NSCLC) samples from 394 patients using a tissue microarray. We nd that low-level expression of the necroptosis markers RIPK3 and PELI1 is associated with increased risk of patient death and that high-level expression of the key DDR factor p53 enhances this risk. These effects appear to be specic to squamous cell carcinoma (SCC) subtype of NSCLC patients but not observable in non-SCC patients. each clinical variable and protein. Proteins that were signicant at 5% level in the simple Cox model in the multiple Cox

Importantly, necroptotic factors appear to in uence tumor progression differentially in a manner speci c to individual tumor types [4,6,7]. We focus our investigation on lung cancer, which is relatively less represented in preclinical necroptosis studies yet accounts for a substantial proportion (~ 20%) of cancer-related deaths globally [14]. A major type of lung cancer, non-small-cell lung cancer (NSCLC) has a particularly grave prognosis albeit a slowly improving survival trend in recent years. Thus, development of NSCLC biomarkers would enable us to detect the disease at a preclinical stage, to predict treatment outcome more accurately, and to employ better targeted drugs [15,16]. Development of prognostic biomarkers for early-stage NSCLC is another interest of research in this eld because the approximate value for the 5-year survival rate is a disappointing 70% [17]. Importantly, current adjuvant treatment does not confer a survival advantage and treatment decisions are not guided by biomarker [9]. Squamous cell carcinoma (SCC) constitutes an aggressive subtype of NSCLC, distinguishing itself from non-SCCs not only by distinct histology and genomic landscapes but also by unique clinical characteristics of age, smoking habit, and comorbid diseases [18,19]. Given the signi cant differences between these subtypes, biomarkers speci c to SCC are greatly needed.
To this end, we here investigated whether the expression of proteins involved in necroptosis and their combined expression with DDR proteins have any effects on the survival of patients with stage I NSCLC. We present data about the relationship between expression level of those proteins, NSCLC subtypes, and patient survival as follows.

Patients and tissue specimens
Consecutive patients with pathological stage I NSCLC who had been treated at the Inha University Hospital (Incheon, South Korea) were retrospectively included in the study ( Supplementary Fig. 1). Patients who underwent surgical tumor resection between 1 January 1997 and 31 December 2011 were initially considered for inclusion. The cancers were staged using the 7 th edition Tumor-Node-Metastasis staging system [20]. Data from each patient who had been followed-up for at least 3 years after surgical resection and whose tumor had been completely resected (R0 resection) were included in the analysis. Data from patients who had undergone limited resection (segmentectomy or wedge resection), received any anticancer treatment before surgical resection, or had a history of other cancer were excluded.
Information about smoking habits, Eastern Cooperative Oncology Group performance status, CCIS, tumor size, T stage, lymphatic or vascular invasion (from the pathological specimens), type of surgery, and use of adjuvant chemotherapy was collected. To increase the quality of the information, radiological examination (e.g., chest computed tomography scan, positron-emission tomography) and surgical pathology reports, and clinical information, were reviewed independently for each patient. All formalinxed, para n-embedded (FFPE) blocks were sent to a laboratory (SuperBioChips, Seoul, South Korea) for tissue microarray (TMA) construction and immunohistochemical staining. Study protocol was approved by the institutional review board of the inha university hospital and informed consent requirement was waived.

Construction of tissue microarrays and immunohistochemistry
TMA construction from individual FFPE samples was performed as previously described [8]. Each tissue array block contained up to 50 specimens, which allowed all 394 specimens to be contained in 20 blocks. Commercially available antibodies were chosen for RIPK3, MLKL (EPR17514), and PELI1 (F-7); these proteins were chosen as they are known to play essential roles in necrosome formation and modulation (Supplementary Table 1) [3]. P53 (DO-7), gH2AX (JBW301), ATM (7C10D8), Chk2pT68, BRCA1pS1423, and ERCC1 (8F1) were chosen because of their well-de ned roles in cellular responses to DNA damage [1,8,9].
Serial sections (4 mm) from the FFPE blocks were analyzed for protein expression by immunohistochemistry. All antibodies were tested using a human control TMA panel [21]. A pathologist performed the test procedure with manufacture's information and evaluated the quality of positive and negative controls in immunohistochemistry.
All immunohistochemical staining was performed with a BenchMark XT autostainer (Ventana Medical Systems, Tucson, AZ) and i-View detection kit (Ventana Medical System) by following a standard operating procedure at SuperBioChips Laboratory, Seoul, South Korea.
Evaluation of immunohistochemistry TMA samples containing more than 500 malignant cells were considered for evaluation. Scanned images of the stained slides were viewed using Aperio ImageScope program (version 11.2.0.782; Aperio Technologies, Vista, CA, USA) at 20x objective magni cation. Automated digital image analysis was performed using Genie classi er and Nuclear v9 algorithm (Aperio Technologies). The percentages of stained cells and intensities of staining were estimated in the cytoplasm for MLKL and PELI1 and in the nucleus for RIPK3, p53, gH2AX, ATM, Chk2, BRCA1, and ERCC1 by two pathologists (H.J.J. and L.K.). The evaluators were blinded to each patient's results for the clinical and pathological variables and survival status.
H-score method was applied to evaluate protein expression. The percentage of tumor cells with positive cytoplasmic or nucleus staining on each TMA core was calculated and assigned a score of either 0 (0% stained tumor cells), 0.1 (1-9%), 0.5 (10-49%), or 1 (≥ 50%). The staining intensity was assigned a score of either 0 (weakest intensity), 1, 2, or 3 (strongest intensity). Each H-score was obtained by multiplying the proportion by the staining intensity. The median values of all mean H-scores were used as the cutoff values for the classi cation of the expression of the nine proteins as either "low" or "high" (Supplementary Fig. 2).

Survival measurements
The overall survival as the endpoint of this study was calculated from the time of diagnosis to the time of the last follow-up or death due to any cause. Dates of death were obtained principally from review of the medical records. To increase the quality of the information regarding the survival status of patients lost to follow-up, missing information was either obtained by contacting the patients or their relatives by phone or mail (13 patients) or obtained from the Ministry of Public Administration and Security, South Korea (29 patients).

Statistical analysis
Distributions of the clinical variables and levels of protein expression between SCC and non-SCC groups were tested using either chi-square test or Mann-Whitney U test as appropriate as possible. Cox model was performed to estimate HR and its 95% CI for each clinical variable and protein. Proteins that were signi cant at 5% level in the simple Cox model were included in the multiple Cox model. Performance of the model was evaluated using Harrell's c-index. P values were corrected using the FDR for multiple comparisons [22]. Statistical signi cance was accepted when p < 0.05. Statistical analyses were performed using a statistical software package (SPSS version 19.0, SPSS, Chicago, IL).

Characteristics of patients
Clinical characteristics of 394 patients with stage I NSCLC (197 with SCC, 175 with adenocarcinoma, and 22 with other histology) are presented in Table 1. The median age of the patients was 66 years. 160 patients (40.6%) have since died. The median survival time was 9.8 years. Each patient was classi ed into either SCC or non-SCC groups. The proportions of never smokers and women were greater in non-SCC (82% and 81% of the entire cohort, respectively, chi-square p < 0.001) than in SCC. Non-SCC group also had greater proportions of patients who were younger, had smaller tumors, lower T stages, and/or received a lobectomy.
Associations between survival and age, size of tumor, and T stage were statistically signi cant (Supplementary Table 2). Gender and Charlson comorbidity index score (CCIS), smoking habit, histology result, lymphatic or blood vessel invasion, type of surgery, and use of adjuvant chemotherapy were not statistically signi cant prognostic factors, however.
Expression of proteins involved in necroptosis or DDR Expression levels of proteins (low versus high) were compared between the SCC and the non-SCC groups using a tissue microarray (Table 2). High-level expression of RIPK3, p53, gH2AX, ATM, Chk2, BRCA1, and ERCC1 was more common in SCC group than in non-SCC group. By contrast, low expression of MLKL and PELI1 was more common in SCC group than in non-SCC group.
Association of necroptosis and DDR proteins with patient survival Analysis of the entire cohort for screening revealed that expression of neither necroptosis nor DDR proteins was signi cant for survival of the patients at the 5% level (Supplementary Table 3). However, histology results indicated that low expression of RIPK3, PELI1, or BRCA1, and high expression of p53 were signi cantly detected in the simple Cox proportional hazards model (Cox model) for SCC group patients (p = 0.011, 0.019, 0.029, and 0.016, respectively). However, the effects of the proteins were not apparent in the patients with non-SCC. In SCC group, the prognostic effect of expression of these four proteins were statistically signi cant after adjusting clinical or pathological variables [adjusted hazard ratio (aHR) (95% con dence interval (CI)), p value, and p value corrected by false discovery rate (FDR) (q value), 2.292 (1.242-4.228), 0.008, and 0.016, respectively, for low RIPK3; 2.007 (1.058-3.806), 0.033, and 0.033, for low PELI1; 2.555 (1.382-4.721), 0.003, and 0.012, for high p53; and 2.088 (1.133-3.850), 0.018, and 0.024, for low BRCA1] ( Fig. 1 and Table 3). The estimated Harrell's c-index was 0.73 (0.69-0.77), indicating that the nal model was accurate enough.

Combined effects of necroptosis and DDR protein expression on patient survival
Combined effects of a protein pair were analyzed to determine whether any combination of proteins involved in necroptosis and DDR pathway had a synergy on patient survival. We noted that the combination of RIPK3 with p53 potentiated the multiplicative effects in the patients with SCC (p value and q value, 0.001 and 0.005) ( Fig. 2A and Supplementary Table 4). Patients with low RIPK3 and high p53 expression had markedly worse survival than patients with high RIPK3 and low p53 expression (aHR (95% CI), 8.394 (2.856-24.677)) ( Fig. 2A and Supplementary Table 4). Similarly, the combination of PELI1 with p53 potentiated the multiplicative effects in patients with SCC (p value and q value, and aHR (95% CI), 0.007, 0.018, and 6.760 (2.173-21.024) for low PELI1 and high p53) ( Fig. 2B and Supplementary Table 4). Combined effect of RIPK3 with BRCA1 was associated with survival of patients with SCC, but its signi cance was not observed (Supplementary Table 4).

Discussion
In summary, our analysis has found that expression of proteins involved in necroptosis, such as RIPK3, is a robust prognostic factor for patients with stage I SCC, but not for patients with non-SCC. We also note that the effect of the low expression of necroptotic factors such as RIPK3 on patient survival becomes more profound when the same patient displays a higher p53 expression.
Studies using cancer cell lines and animal models have found that RIPK3 has diverse effects on tumor progression [4,6,7]. Our data show that SCC patients with low RIPK3 expression have an approximately 2.3 times increased risk for death relative to those with high RIPK3 expression. This observation indicates that as found in patients with esophageal SCC, RIPK3 inhibits tumor progression in the patients with SCC [10]. Another necroptosis factor PELI1 also has diverse effects on survival in patients with diffuse large B cell lymphoma or melanoma [23,24]. We found that SCC patients with low expression of PELI1 had an approximately 2.0 times increased risk for death when compared with those with high expression. Diversities of tumor types or tumor microenvironments likely account for these differences in the roles for RIPK3 and PELI1 in tumor progression.
Our data also reveal a synergistic effect of either RIPK3 or PELI1 when combined with p53. Patients with a low-RIPK3 and high-p53 expression had an approximately 8.4 times increased risk for death relative to those with a high-RIPK3 and low-p53 expression. Similarly, the combination of low PELI1 expression and high p53 expression increased the risk of death up to 6.7 times when compared with high PELI1 and low p53 expressions. Notably, a large proportion of patients (approximately 45%) had either a low-RIPK3 and high-p53 expression or a high-RIPK3 and low-p53 expression. This result suggests that the use of these potential biomarkers could considerably bene t the population of patients with SCC. The result that the combination of low RIPK3 and high p53 expression can promote tumor progression was expected because high expression of p53 indicates an abnormal function.
RIPK3 or PELI1 may somehow interact with p53 in patients with SCC. In human mbrial epithelium, RIPK3 function depends on the status of p53, and PELI1 can regulate p53 function by promoting cytoplasmic localization of MDMX [12,24]. Furthermore, in this study, this relationship between RIPK3 or PELI1 and p53 was found. This issue merits further investigation using cell lines and preclinical models.
The results of this study should be interpreted with some caution. The results may be limited in generalization to other cohorts because we adopted a retrospective study design and they were not validated using an independent cohort. However, the sample size was large enough to evaluate the effects of protein expression. Our present study was designed to maintain the homogeneity of the study population. We used stringent criteria for inclusion or exclusion of patients' data to minimize selection bias.
Information on clinical or pathological variables as potential confounders was comprehensively collected and extensively considered during the analysis. We did not provide a biological basis directly supporting the differential prognostic effects between the two histological groups. However, these ndings are supported by previous mechanistic studies with cancer cell lines or animal models [4-7, 10-13, 24, 25], as well as differences in genomic architectures and clinical pro les between these groups [18,19]. All in all, our ndings demonstrate a clear translational relevance of decreased levels of select necroptosis factors in predicting a poor prognosis of SCC-subtype NSCLC.

Conclusions
This is the rst study to demonstrate prognostic implications of necroptosis proteins and their association with expression of DDR proteins in patients with stage I NSCLC. Our results indicate that RIPK3 and PELI1 or their individual combination with p53 can help classify patients with SCC into groups of either low or high death risk, providing a novel insight to clinicians about which patients would require a thorough follow-up.     Survival of patients with stage I SCC according to the expression of RIPK3, PELI, p53 and BRCA1 in adjusted models. aHR: hazard ratio after adjusting age, gender, smoking habit, Eastern Cooperative Oncology Group performance status, Charlson comorbidity index score, tumor size, T stage, lymphatic or vascular invasion, type of surgery, and adjuvant chemotherapy.