Impact of immune checkpoint gene CD155 Ala67Thr and CD226 Gly307Ser polymorphisms on small cell lung cancer clinical outcome

This study was conducted to investigate the impact of genetic variants of immune checkpoint genes on the treatment outcome in small cell lung cancer (SCLC). In the present study, 261 platinum doublet-treated SCLC patients were enrolled. A total of 96 polymorphisms in 33 immune checkpoint-related genes were selected, and their association with chemotherapy response and survival outcomes were analyzed. Among the polymorphisms studied, CD155 rs1058402G > A (Ala67Thr, A67T) and CD226 rs763361C > T (Gly307Ser, G307S) were significantly associated with SCLC treatment outcome. The rs1058402G > A had a worse chemotherapy response and overall survival (under a dominant model, adjusted odds ratio [aOR] = 0.52, 95% confidence interval [CI] = 0.27–0.99, P = 0.05; adjusted hazard ratio [aHR] = 1.55, 95% CI = 1.12–2.14, P = 0.01, respectively). The rs763361C > T had better chemotherapy response and overall survival (under a dominant model, aOR = 2.03, 95% CI = 1.10–3.75, P = 0.02; aHR = 0.69, 95% CI = 0.51–0.94, P = 0.02, respectively). When the rs1058402GA/AA and rs763361CC genotypes were combined, the chemotherapy response and overall survival were significantly decreased as the number of bad genotypes increased (aOR = 0.52, 95% CI = 0.33–0.81, Ptrend = 0.004; aHR = 1.48, 95% CI = 1.19–1.84, Ptrend = 4 × 10−4, respectively). The 3-D structural model showed that CD155 A67T created a new hydrogen bond and structural change on CD155. These changes resulted in extending the distance and losing the hydrogen bonds between CD155 and CD226, thus weakening CD155/CD226 binding activity. In conclusion, CD155 rs1058402G > A and CD226 rs763361C > T may be useful for predicting the clinical outcomes of SCLC patients after chemotherapy.

chemotherapy, but most SCLC patients relapse and eventually die. Therefore, efforts are needed to predict chemotherapy responses and find prognostic markers.
The immune system defends our bodies from infectious organisms and other invaders. The immune system is also important in preventing and eradicating cancers 4 . In general, immune cells can recognize tumor cells and destroy them. However, cancer cells have developed ways of escaping the immune system by suppressing the activation of immune cells 5 . Therefore, cancer cells can survive and spread beyond the patient's immune system. Many researchers have tried to control immune-escaping tumors [5][6][7] . Immune checkpoint inhibitors are drugs that can block inhibitor signals from cancer cells, restore the immune system, and reactivate immune cells to kill cancer cells. Recently, immune checkpoint inhibitors have been actively used in the treatment of cancers, including lung cancer [8][9][10][11] .
The genetic variants in immune checkpoints may affect the body's response to chemotherapy or lung cancer prognosis, considering the immune system's ability to prevent and kill cancer cells. We previously reported that genetic variants in immune checkpoint genes were associated with the prognosis of surgically resected NSCLC 12 . Programmed cell death-ligand 1 (PD-L1) polymorphisms were associated with chemotherapy response and clinical outcomes of advanced NSCLC 13 . Nomizo et al. reported that the polymorphisms in immune checkpoint genes had an effect on the response rate and progression survival in NSCLC patients with nivolumab treatment 14 . However, only a few studies had been conducted regarding SCLC. Therefore, we investigated the effects of genetic variants in immune checkpoint genes on the chemotherapy response and prognosis in SCLC.

Materials and methods
Study population. This study is observational retrospective study. This study included 261 patients diagnosed with SCLC at Kyungpook National University Hospital (KNUH) in Daegu, Korea between March 2001 and November 2017. The flow of selection of the patients is shown in Fig. 1. All the patients were treated with at least two cycles of platinum doublet chemotherapy as first-line treatment. Patients with limited-stage SCLC who underwent chemotherapy with concurrent radiotherapy were excluded to avoid the confounding effect of radiation on the chemotherapy response. If radiotherapy was conducted sequentially after 2 cycles of chemotherapy, these patients were included. In this case, the response after 2 cycles of chemotherapy was considered as the best response to chemotherapy. There were two chemotherapy regimens. One consists of cisplatin 60 mg/m 2 on day 1 and etoposide 100 mg/m 2 on day 1, 2, and 3 every 3 weeks. The other consists of cisplatin 60 mg/m 2 on day 1 and irinotecan 60 mg/m 2 on day 1, 8, and 15 every 4 weeks. The treatment was discontinued in case of disease progression, major toxicities, or according to patient's or physician's decision. Response assessment was carried out every two cycles of chemotherapy using the Response Evaluation Criteria in Solid Tumors 15 . Chemotherapy response fell into two categories: responder and non-responder. Responder means the best response was a complete response or a partial response. A stable disease or progressive disease was defined as non-responder.
This study was approved by the Institutional Review Boards of KNUH. The blood samples for genotyping were provided by the National Biobank of KNUH, which is supported by the Ministry of Health, Welfare, and Family Affairs. All blood samples were obtained before the 1st chemotherapy session. All subjects are 18 years of age or older and informed consent was obtained prior to chemotherapy.  Structure modeling of CD155 A67T variant.. In order to evaluate the effect of the rs1058402G > A of CD155, which substitutes alanine with threonine at codon 67, The mutant structure of CD155(A67T) was modeled using MODELLER v9.12 with the crystal structures of wildtype CD155 (PDB : 6ISC). The least violated 10 structures were further refined by model/refine loops (loop modeling protocol DOPE) using UCSF Chimera v1.14 and a structure showed no violation and the lowest energy was selected as a model. All images of CD226/ CD155 (wild type vs. A67T mutant) were made in PyMOL (https ://pymol .org/2/).

Patient characteristics and clinical predictors. Chemotherapy response and OS according to patients'
clinical characteristics are shown in Table 1. The overall response rate was 72.8% (95% CI = 67.4-78.2%). The median survival time (MST) was 10.4 months (95% CI = 9.1-11.1 months). The response rate was higher in the irinotecan/cisplatin regimen than the etoposide/cisplatin regimen (78.7% vs. 67.2%, P = 0.04), but OS was not different based on regimens (MST, 10.0 months vs. 10.7 months, P = 0.88, Table 1). OS was associated with age, stage, ECOG performance status, weight loss, NSE level, second-line chemotherapy, and radiation to tumor ( Table 1).
The effects of the two variants on chemotherapy response and OS did not differ according to clinical variables, such as age, gender, etc., when data were categorized by these factors, except for chemotherapy regimen on chemotherapy response (P value for homogeneity test > 0.05, Supplementary Table S3).
Combined effects of rs1058402 and rs763361 on treatment outcomes. The rs1058402GA/AA and rs763361CC genotypes were associated with worse chemotherapy response and OS. When these genotypes were considered bad genotypes, chemotherapy response was decreased as the number of bad genotypes increased (responder 80.4% with 0 bad genotype, 71.0% with 1 bad genotype, and 61.3% with 2 bad genotypes; aOR = 0.52, 95% CI = 0.33-0.81, Ptrend = 0.004; Table 3). The MST was also significantly decreased as the num-  Table 3 and Fig. 2C). PFS also significantly decreased as the number of bad genotypes increased (Ptrend = 0.009). In an analysis of only patients with extensive-stage, the combined effect of rs1058402 and rs763361 was still significant in chemotherapy response, OS, and PFS, respectively (aOR = 0.55, 95% CI = 0. 33 Functional prediction of CD155 rs1058402G > A (A67T). The rs1058402G > A changes the amino acid of alanine to threonine at codon 67 of CD155. CD155 bound to CD226 of the immune cell and acts as co-stimulatory signal 19 . We evaluated whether this amino acid change affects the function of CD155 by creating a 3-D structural model using PyMOL (https ://pymol .org/2/). As shown in Fig. 3, the alanine to threonine change at codon 67 provides distance shortening to S74 and G73 of CD155 (6.3→2.9 and 4.4→3.3 Å, respectively; Fig. 3B,C). The shortened distance between T67 and S74 created a new hydrogen bond (Fig. 3C). This change resulted in distance lengthening between S74 of CD155 and N116 of CD226 (2.7→6.0 Å) and losing the hydrogen bond (Fig. 3B,C). In addition, CD155 A67T also influenced other important interactions between G70 of CD155 and E185 of CD226 by distance extension (3.3→6.0 Å), resulting in the loss of the hydrogen bond (Fig. 3D). The increased distances and loss of hydrogen bonds will weaken the binding activity between CD155 and CD226.

Discussion
The immune system is evidently important not only in the development and progression of cancer but also in its treatment 4,20 . The interaction between tumors and their microenvironment (including immune cells such as tumor infiltrating lymphocytes) can affect chemotherapy response [20][21][22] . Genetic variants in immune genes can also influence the host's immune activity, which affects the clinical cancer treatment outcome 12,13,23,24 . In this   www.nature.com/scientificreports/ study, we investigated the association between variants in immune checkpoint genes and the clinical outcome of SCLC. This is the first study to investigate the effects of genetic variants in immune checkpoint genes on chemotherapy response and prognosis in SCLC. We found that two variants, namely CD155 rs1058402G > A (A67T) and CD226 rs763361C > T (G307S), were significantly associated with chemotherapy response and survival outcomes. The development of drugs to block immune checkpoints has opened a new era in cancer treatment. Programmed death 1 (PD-1) and PD-L1 inhibitors are used actively in NSCLC treatment 10,11 . The PD-1/PD-L1 interaction provides inhibitory signals to suppress immune cell responses, so blocking the PD-1/PD-L1 pathway can reactivate immune cells to attack cancer. In NSCLC, the polymorphisms in PD-L1 were reported to affect chemotherapy/immunotherapy response and prognosis [12][13][14] . However, data on the association between polymorphisms in PD-1/PD-L1 and clinical outcomes in SCLC are limited. In this study, the polymorphisms in PD-1/ PD-L1 were not associated with clinical outcomes in SCLC patients (Supplementary Table S1).
CD155, also called PVR or Necl-5, encodes a transmembrane glycoprotein belonging to the immunoglobulin superfamily. CD155 was initially known to play roles in cell adhesion and migration, but recently, its roles in immunology and oncology have been noted 25,26 . CD226 encodes a glycoprotein expressed on the surface of natural killer (NK) cells, T cells, a subset of B cells, and monocytes, and plays an important role in their activation and inhibition 27,28 . CD155 on antigen-presenting cells or tumor cells binds to CD226 on T cells and NK cells, which is similar to the interaction between PD-1 and PD-L1 19,29 . The difference is that PD-1/PD-L1 interaction provides an inhibitory signal to suppress T cell response, but CD155/CD226 binding is a co-stimulatory interaction for T cell or NK cell activation 19,30 . In cancer immunology, CD155 overexpression was reported in several types of human malignancies, including lung adenocarcinoma, and was correlated with unfavorable prognosis [31][32][33] . CD226 has been reported to be involved in anti-tumor response by regulating NK cells 34,35 .
The rs1058402G > A is a missense mutation. The rs1058402G > A changes alanine amino acid to threonine at codon 67 of CD155. 3-D structure model showed that changing alanine to threonine at codon 67 shortened the distance to S74/G73 of CD155. The shortened distance made a new hydrogen bond between T67 and S74 of CD155, and this change resulted in distance extension and loss of hydrogen bond between the S74 of CD155 and N116 of CD226. In addition, the structure modification due to rs1058402G > A (A67T) increased the coupling distance between the G70 of CD 155 and E185 of CD226, and resulted in hydrogen bond loss. In the interaction between CD155 and CD226, the S74/G70 of CD155 and N116/E185 of CD226 are important binding points 19 . In this study, the 1058402G > A (A67T) was significantly associated with worse chemotherapy response and OS in SCLC patients. As shown in the 3-D structural model, the 1058402G > A (A67T) increased the binding point distance between CD155 and CD226. The increased distance and loss of hydrogen bonds lead to a weakening of the binding force of CD155/CD226, thereby reducing the co-stimulatory signal to immune cells. The decreased immune response would had a negative effect on SCLC clinical outcomes. However, further functional studies are needed to determine whether CD155 rs1058402G > A (A67T) affects the CD155/CD226 interaction. This is the first review to report that a CD155 variant was associated with cancer clinical outcomes. This variant may also affect the therapeutic effect of the drug to be developed in the future, considering that CD155 is being studied as a new therapeutic target in tumor immunology 26,36 .
In the present study, CD226 rs763361C > T was associated with better chemotherapy response and OS in SCLC patients. The rs763361C > T is located in exon 7 encoding the cytoplasmic tail of CD226, which harbors two phosphorylation sites and is a non-synonymous mutation 37 . The rs763361 C-to-T change results in the replacement of glycine to serine at codon 307. A Gly307Ser substitution believed to affect CD226 expression by changing the phosphorylation or altering RNA splicing by disrupting the exon-splicing silencer sequence 37,38 . CD226 rs763361C > T is associated with multiple autoimmune diseases, including multiple sclerosis and rheumatoid arthritis [37][38][39][40] . Regarding cancer, the T allele of rs763361 has been reported to increase the risk of NSCLC in the Chinese Han population 41 . However, the exact biological function of CD226 rs763361C > T is not wellknown, and further researches are needed. The effect on clinical outcomes may be due to other causal variants in LD. The CD226 rs1790947G > T has a strong LD (r 2 = 0.91) with CD226 rs763361C > T and is located in the 3′ untranslated region. Using RegulomDB (http://www.regul omedb .org/), the rs1790947G > T was likely expected to affect binding and linked to CD226 expression. RegulomeDB is a novel approach and database, which provides interpretation of regulatory variants in the human genome 42 . The association patterns between the rs1790947 genotypes and chemotherapy response and OS were similar to that of rs763361 (Supplementary Table S5). Hence, the selected polymorphism was rs763361C > T, but rs1790947G > T could actually affect CD226 expression and clinical outcomes.
In this study, polymorphisms in the CD155 and CD226 was associated with survival outcomes in SCLC patients who received chemotherapy. CD155 and CD226 are one of targets in immuno-oncology 43,44 . Therefore, variants in CD155 and CD226 may have a greater impact on clinical outcomes with immunotherapy alone or in combination with chemotherapy than with chemotherapy alone. Recently, immunotherapy has been shown to be effective in the treatment of SCLC 45 . The addition of atezolizumab to etoposide/carboplatin in the first-line treatment of extensive-stage SCLC showed significant longer OS and PFS than chemotherapy alone 46 . Because CD155 and CD226 are co-stimulatory signal for immune cells, investigating the role of variants in CD155 and CD226 in SCLC patients who received combination immunotherapy and chemotherapy may yield interesting results.
Although a theoretical model for the interactional changes caused by mutations of CD155 and CD226 has been presented, one of limitations is that it has not been confirmed experimentally. Unlike NSCLC, surgical treatment is not usually performed in patients with SCLC. The difficulty of obtaining sufficient tumor tissues for experiments is an obstacle to overcome in SCLC studies. Observational retrospective study at a single center can be another limitation of this study.
In summary, we investigated the association between the variants in immune checkpoint genes and SCLC prognosis. Two variants, namely CD155 rs1058402G > A (A67T) and CD226 rs763361C > T (G307S), were