B7-H3 Negatively Modulates CTL-Mediated Cancer Immunity

Malignant cells are under immune surveillance, wherein the immune system identifies these cells and eliminates them (1). In particular, CD8⁺–T-cell–mediated antitumor immune responses are critical for effective destruction of malignant cells. However, malignant cells evade immune surveillance via various mechanisms (2). As observed in chronic viral infectious diseases (3), effector function is lost in tumor-infiltrating T cells with an exhausted phenotype (4). T-cell dysfunction is generally mediated by a cellular-extrinsic mechanism such as regulatory T-cell infiltration (5) or a cellular-intrinsic mechanism such as the expression of numerous coinhibitory molecules on T cells (6, 7).

Programmed death-1 (PD)-1 is the most potent immune-checkpoint molecule; it is expressed on the cell membrane of T cells during persistent antigen exposure such as chronic viral infection and binds to programmed death ligand-1 (PD-L1), which is highly expressed in some virus-infected cells, cancer cells, or dendritic cells (4, 8–10). Ligated PD-1 inhibits the T-cell receptor (TCR) signaling pathway through Src homology-2 domain containing protein (SHP-2) (11), the CD28 costimulatory pathway (12, 13), or the glycolytic system mediated by the mTOR pathway (14, 15) that subsequently regulates the effector function of T cells. Therefore, anti-PD-1/PD-L1 antibodies block the binding of PD-1 with PD-L1 and reinvigorate tumor immunity (6, 16). PD-1 blockade clinically provides antitumor efficacy, and anti-PD-1 antibodies are a standard therapeutic option for some types of malignancies, including non–small cell lung cancer (NSCLC) and melanoma (17–20).

In addition to PD-1, many other cell surface coinhibitory molecules regulate T-cell exhaustion (21). During chronic infection, virus-specific CD8⁺ T cells can also coexpress Lag-3, Tim-3, and CTL-associated protein 4 (CTLA-4), and double blockade of these molecules and the PD-1 pathway increases the recovery of effector function (8, 22–24). In the case of cancer immunotherapy, the combination of anti-PD-1 antibody and anti-CTLA4 antibody provides more potent antitumor efficacy and significantly longer survival than those obtained with any of these single antibodies alone in patients with melanoma (25, 26). Coexpression of Tim-3 or Lag-3 with PD-1 in dysfunctional TILs and double blockade of these molecules enhances anticancer immunity to a greater extent than single blockade (27–29). Therefore, such combinations involving immune-checkpoint molecules provide promising strategies for some malignant diseases (30).

B7-H3 belongs to the B7 superfamily of type I transmembrane proteins, which modulate T-cell function in a costimulatory or coinhibitory manner (31–33). In normal tissue, B7-H3 protein is rarely expressed and is only found at low levels, whereas it is aberrantly expressed in numerous types of malignant diseases, including NSCLC, and is a factor for poor prognosis in some of those malignancies (34–36). However, the molecular mechanisms underlying the correlation between B7-H3 expression and poor prognosis have not been completely elucidated, and the effect of B7-H3 on the cancer–immune system interactions is controversial. Although B7-H3 was initially identified as an activator of T-cell function (32, 33), subsequent reports have shown that B7-H3 can cause T-cell dysfunction (37, 38). The receptor for B7-H3 has not been identified yet, but the crystal structure of mouse B7-H3 has revealed that the FG loop of the IgV domain of B7-H3 engages the receptor of T cells and plays a critical role in T-cell dysfunction (39). Furthermore, despite the high incidence of B7-H3 expression in malignancies, the immunologic interactions between B7-H3 and PD-L1, which is also preferably expressed in malignancies, are not clear.

In this study, we investigated the incidence of B7-H3 expression in samples obtained from patients with NSCLC and evaluated its correlation with clinicopathologic features and the efficacy of anti-PD-1 blockade treatment. In addition, anti-B7-H3 blockade was assessed for antitumor immunity and for immunologic interactions in the case of anti-PD-L1 blockade by using a syngeneic mouse tumor model.

Tumor specimens from patients with NSCLC were obtained from the Kindai University Faculty of Medicine, Kindai Sakai Hospital, National Hospital Organization Osaka Minami Medical Center, Izumi Municipal Hospital, and Kishiwada City Hospital, with approval from the Institutional Review Board and written informed consent from each patient. The inclusion criteria were as follows: (i) histopathologically proven unresectable distant metastatic or locally advanced NSCLC; (ii) availability of a paraffin-embedded tumor tissue sample; (iii) age of 20 years or more. The patients had not necessarily been treated with anti-PD-1/PD-L1 antibodies. The medical records were reviewed, and data regarding clinicopathologic features and treatment outcome were extracted (Table 1). The data have been updated as of August 23, 2017.

The objective response to treatment with anti-PD-1 antibodies, that is, including nivolumab and pembrolizumab, was evaluated according to the Response Evaluation Criteria in Solid Tumors, ver. 1.1 (40). The tumor response was evaluated by a physician every 2 to 3 months by using CT. Progression-free survival (PFS) was defined as the interval from the initiation of anti-PD-1 antibody therapy to tumor progression or death without evidence of progression. Patients without documented clinical or radiographic disease progression or who were still alive were censored on the date of last follow-up.

The murine pancreatic adenocarcinoma Pan02 cell line was purchased from the DCTD Tumor Repository, National Cancer Institute (Frederick, MD). The murine Lewis lung carcinoma cell line (3LL) was purchased from the JCRB cell bank, National Institutes of Biomedical Innovation, Health and Nutrition (Osaka, Japan). The cells were grown in a humidified atmosphere of 5% CO₂ at 37°C in RPMI1640 medium (Sigma-Aldrich) supplemented with 10% FBS and 1% penicillin–streptomycin. The anti-B7-H3 antibody (clone MJ18), anti-PD-L1 antibody (clone 10F.9G2), anti-CD4 antibody (clone GK1.5), and anti-CD8 antibody (clone 2.43) were purchased from BioXCell.

Total RNA was isolated from tumors using the RNeasy Mini Kit (Qiagen), according to the manufacturer's specifications. cDNA was synthesized using a High Capacity RNA-to-cDNA Kit (Applied Biosystems). Subsequently, RT-PCR was performed for CD8, CD274, and CD276 expression with a TaqMan Gene Expression Assay (assay ID: Mm01182107_g1, Mm00452054, Mm00506020_m1, Applied Biosystems) used per the manufacturer's instructions. GAPDH expression was used as an internal standard. Expression levels relative to the untreated tumor were determined using a relative standard curve method with ABI 7900 HT SDS software (Applied Biosystems).

All animal experiments were performed in accordance with the Recommendations for Handling of Laboratory Animals for Biomedical Research compiled by the Committee on Safety and Ethical Handling Regulations for Laboratory Animal Experiments, Kindai University (Osaka, Japan). The study was also reviewed and approved by the Animal Ethics Committee of Kindai University (Osaka, Japan). Pan02 cells or 3LL cells (5 × 10⁶ per mouse) were subcutaneously injected into the right flanks of female C57BL/6 mice (age, 8–12 weeks) obtained from CLEA Japan (Tokyo, Japan). Treatment was started on the day after tumor implantation. The mice received intraperitoneal injection of an isotype IgG control antibody, anti-B7-H3 antibody (15 mg per kg body weight in 100 μL PBS), and anti-PD-L1 antibody (10 mg per kg body weight in 100 μL of PBS) twice per week. For in vivo depletion of CD8⁺ or CD4⁺ T cells, the mice received an anti-CD8 or CD4-depleting mAb (15 mg per kg body weight in 100 μL of PBS), which was intraperitoneally injected on the day of tumor implantation and then twice per week. Tumor volumes and mouse body weights were measured twice per week. The mice were sacrificed if the tumors became necrotic or grew to a volume of 1.0 cm³. Tumor volume was defined as 1/2 × length × width². The T/C ratio was calculated according to the following equation: 100 × (average tumor volume of the treated group)/(average tumor volume of the control group).

Tumors and spleens from tumor-bearing mice were mechanically disaggregated with 200 U/mL collagenase type I (Wako) and 100 μg/mL DNase I (Wako) solution. Single-cell suspensions were treated with Fc block (2.4G2; BD Pharmingen) and then stained with a labeled primary antibody against mouse CD4 (BD Horizon), CD8a (BD Pharmingen), or PD-1 (29F.1A12; BioLegend) diluted in FACS buffer (PBS, 0.1% NaN₃, 1% BSA). The cells were examined with a BD Fortessa system and then analyzed using FlowJo software.

Splenocytes or tumor-infiltrating lymphocytes (TILs) isolated from tumor-bearing mice were incubated in a 96-well plate. The cells were incubated in the presence of Pan02 cells and Brefeldin A (50 mg/mL, Wako) for 6 hours at 37°C. Surface staining for CD8 was performed as described above, and the cells were fixed and permeabilized with the Cytofix Cytoperm kit (BD Biosciences). For the detection of intracellular IFNγ, granzyme B, and Foxp3, the cells were incubated for 30 minutes in Perm/Wash buffer, followed by incubation in the same buffer with anti-IFNγ (BD Horizon), anti-granzyme B (BD Biosciences), or anti-Foxp3 (Invitrogen) for 30 minutes; the cells were then washed and analyzed as described above. To correct for background variations between experiments, we subtracted the percentage of IFNγ⁺ cells among CD8⁺ T cells without stimulation from the percentage of IFNγ⁺ cells following stimulation, for each type of mouse splenocyte.

Membranous tumor B7-H3 expression was evaluated by IHC in NSCLC tumor tissues based on previous report (34). Sections (thickness, 4 μm) were immunostained for B7-H3 with an anti-B7-H3 mouse mAb (Daiichi Sankyo Co. Ltd.; diluted 2.5 μg/mL) using a Bench Mark XT automated system (Ventana Medical Systems). The IHC expression of B7-H3 in the membranes of cancer cells was interpreted by two pulmonary pathologists (O. Maenishi and Y. Takashima) in a blinded manner. The B7-H3 staining pattern was scored as 0, 1+, 2+, or 3+ (Supplementary Fig. S1) in the following manner: 0, no membranous staining, ≤10% tumor cells with faint/weak membranous staining; 1+, >10% tumor cells with faint/weak membranous staining; 2+, >10% tumor cells with weak or moderate membranous staining or ≤10% tumor cells with strong membranous staining; 3+, >10% tumor cells with strong membranous staining.

On the basis of a previous report (41), the number of CD8⁺ TILs was evaluated at an absolute magnification of 200 × (0.15 mm² per field). Five fields of tumor regions were randomly chosen for each TIL count. TILs were counted by one pathologist, and the density of TILs in the tumor was calculated by dividing the number of TILs by the sum of the area (mm²) of the viewed fields. TILs were defined as cells positive for CD8 regardless of staining intensity.

Statistical analyses were performed using SPSS version 22.0 (IBM Corp.). To investigate whether a factor, including B7-H3 IHC expression, affected responsiveness and survival, logistic regression analysis, and multivariable Cox regression were performed. For cases in which the outcome variable was time to event, Kaplan–Meier curves were constructed. For statistical hypothesis testing, the unpaired t test was used for continuous variables, Fisher exact test for categorical variables, and log-rank test for time to event, where two-sided P values were obtained. Graphical depictions of data were obtained using GraphPad Prism 5.0 for Windows (GraphPad Software, Inc.).

We evaluated B7-H3 expression levels in NSCLC tumors by using IHC. A cohort consisting of 82 patients with advanced or recurrent NSCLC was used, of which 50 patients were treated with anti-PD-1 therapy. The B7-H3 staining pattern was classified as 0, 1+, 2+, and 3+ according to membranous staining intensity and frequency in cancer cells (representative cases are shown in Fig. 1A): 74% of tumor samples expressed B7-H3 with a staining pattern of 1+, 2+, or 3+, and 26% did not express B7-H3, that is, the staining pattern was 0 (Fig. 1B). Normal lung tissue was not stained with B7-H3 (Fig. 1A). For further analysis, tumors with a B7-H3 staining pattern of 1+, 2+, or 3+ were defined as B7-H3–positive and those with a staining pattern of 0 as B7-H3–negative. The data regarding clinicopathologic characteristics and responsiveness to anti-PD-1 therapy for all patients are shown in Table 1. These factors were analyzed for correlation with B7-H3 expression. Compared with the findings for nonsquamous cell lung carcinoma, more squamous cell lung carcinomas were found to be positive for B7-H3 expression (69% vs. 81%). In particular, among nonsquamous cell lung carcinomas, EGFR-mutant NSCLCs had a lower incidence of B7-H3 positivity than wild-type EGFR NSCLCs (47% vs. 81%, P = 0.022). The incidence of B7-H3 positivity in tumors from responders to anti-PD-1 therapy was lower than that in tumors from nonresponders. A subgroup of responders to anti-PD-1 therapy had a 63% incidence of B7-H3 positivity, but nonresponders had a 97% incidence of B7-H3 positivity (P = 0.0007). In addition, B7-H3 expression was not related to the PD-L1 TPS. These results suggest that the B7-H3 expression level varied among NSCLCs and may be correlated with EGFR gene expression status and efficacy of anti-PD-1 therapy.

Responsiveness to anti-PD-1 therapy may be influenced by some factors such as the PD-L1 TPS or the EGFR gene expression status in NSCLC. Therefore, using logistic regression analysis, we investigated whether the responsiveness to anti-PD-1 therapy was affected by several factors, including not only the B7-H3 IHC staining pattern but also the age, sex, smoking habit, histologic features, EGFR genomic status, anti-PD-1 antibody, anti-PD-1 treatment line, and PD-L1 TPS in 50 patients with NSCLC who underwent anti-PD-1 therapy (Supplementary Table S1). Among those factors, the OR for B7-H3 expression was notably greater than those for the other factors specified in Table 2 [OR (negative/positive) = 17.5; 95% confidence interval (CI), 1.939–158.0; P = 0.0031]. The ORs for PD-L1 TPS and EGFR gene status were not greater than that for B7-H3 expression; the OR (95% CI) was 0.7937 (0.2247–2.803) for PD-L1 TPS (<50/≥50) and 0.5294 (0.06351–4.413) for EGFR gene status (wild-type/mutant).

Among the four subgroups with different B7-H3 IHC staining patterns, the group with a B7-H3 IHC staining pattern of 0 had the highest response rate at 88% (vs. 1+, 2+, 3+, P = 0.0031; Fig. 1C). The other groups, which had B7-H3 IHC staining pattern of 1+, 2+, and 3+, had an almost equivocal response rate. Regardless of the B7-H3 IHC staining pattern, except for 0, B7-H3 expression was associated with refractoriness to anti-B7-H3 therapy.

Second, in 50 patients with NSCLC who underwent anti-PD-1 therapy, we performed Cox regression analyses to investigate whether the PFS was affected by several factors, including B7-H3 IHC staining pattern, age, gender, smoking habits, histologic features, EGFR genomic status, anti-PD-1 treatment line, and PD-L1 TPS. The results are shown in Supplementary Table S3. The results revealed that only two factors of B7-H3 IHC staining pattern and PD-L1 TPS were strongly related to PFS (HR, 0.082; 95% CI, 0.016–0.421; P = 0.003 for B7-H3 IHC staining pattern; HR, 3.559; 95% CI, 1.445–8.764; P = 0.006 for PD-L1 TPS). Figures 1D and E show the survival curves for B7-H3 IHC staining pattern and PD-L1 TPS. The scatter diagram for the B7-H3 IHC staining pattern and PD-L1 TPS indicated that there was no correlation between these factors (Fig. 1F). Furthermore, the B7-H3–positive ratio was not significantly different between the subpopulation with PD-L1 TPS more than 50% and the subpopulation with PD-L1 TPS less than 50% (88.2% vs. 78.1% respectively, P = 0.4668; Fig. 1G).

These results suggest that greater efficacy of anti-PD-1 therapy would be achieved for patients with NSCLCs that are B7-H3–negative, which might be independent from the PD-L1 expression level, than for NSCLCs that are B7-H3–positive.

Anti-PD-1 therapy blocks the coinhibitory interaction between PD-L1/PD-1 in CD8⁺ TILs, and then provides an antitumor immune response. Therefore, we examined the relationship between PD-L1 expression and CD8⁺ TILs in NSCLC tissue samples by using IHC. The CD8⁺ TIL level was significantly greater in tumors with PD-L1 TPS ≥ 50% than in those with PD-L1 TPS < 50% (P = 0.0038; Fig. 2A). This observation suggests that a high amount of CD8⁺ TILs mediates the potent antitumor efficacy of anti-PD-1 therapy in the subpopulation of NSCLC with high expression of PD-L1. Then, we evaluated the relationship between B7-H3 expression and CD8⁺ TILs in the tumor subpopulation with lower PD-L1 expression (<50% PD-L1 TPS). Tumors with B7-H3 expression showed a low number of CD8⁺ TILs; however, tumors without B7-H3 expression had a higher count of CD8⁺ TILs than those with B7-H3 expression (P = 0.0103; representative cases also shown in Fig. 2B).

Thus, high PD-L1 expression did not involve the loss of CD8⁺ TILs; rather, this cell population was increased. Presumably, CD8⁺ TILs produce the inflammatory cytokine IFNγ, which increases PD-L1 expression in tumors. In contrast to PD-L1, tumors that did not express B7-H3 had a high level of CD8⁺ T cells despite having PD-L1 TPS of less than 50%. These observations raised the hypothesis that B7-H3 could impair CD8⁺ TILs, and also that B7-H3–targeted therapy could increase CD8⁺ TILs, which would enhance the efficacy of anti-PD-1 immunotherapy.

To test the hypothesis raised by the clinical observation, first we examined whether anti-B7-H3 blockade could increase CD8⁺ TILs by using a syngeneic mouse model. The mouse pancreatic cancer Pan02 cell line had been previously reported to show B7-H3 expression (42). We subcutaneously implanted Pan02 cells in C57BL/6 mice, which were then peritoneally injected with anti-B7-H3 antibody at 15 mg/kg twice per week or with isotype IgG control antibody for 35 days. Anti-B7-H3 blockade inhibited growth of the xenografted tumor as compared with that in the isotype IgG control group with a T/C ratio of 56% on day 35 after inoculation (Fig. 3A).

Then, we examined whether anti-B7-H3 blockade systemically enhanced the T-cell–mediated antitumor reaction by using flow cytometer. Splenocytes from tumor-bearing mice were restimulated with Pan02 cancer cells. Cancer cell stimulation elicited 13.4-fold more IFNγ production in CD8⁺ splenic lymphocytes from mice treated with anti-B7-H3 blockade than in the IgG control on day 30 (P = 0.002; Fig. 3B). In addition, cancer cells stimulated 7.0-fold more IFNγ production in CD4⁺ splenic lymphocytes from mice treated with anti-B7-H3 blockade than in the IgG control on day 30 (P = 0.027; Fig. 3C).

Subsequently, we compared tumor-infiltrating lymphocytes (TILs) with anti-B7-H3 blockade to the IgG control by using flow cytometry. Minimal CD8⁺ TILs were present in tumors treated with the IgG control antibody on day 30 (Fig. 3D). In contrast, anti-B7-H3 blockade strikingly increased the CD8⁺ TILs; 4.9-fold more CD8⁺ TILs were noted per total tumor compared with those in the isotype control, or 11-fold more per mm³ tumor (P < 0.0001, Fig. 3D). In addition, CD4⁺ TILs were minimal in tumors treated with the IgG control on day 30. However, anti-B7-H3 blockade markedly increased CD4⁺ TILs compared with those noted for the control on day 30 (Fig. 3E). The number of CD4⁺ TILs per total tumor was 1.97-fold higher in anti-B7-H3-treated tumors than in the isotype control, or 4.5-fold higher per mm³ tumor (P < 0.0001 Fig. 3E). The number of CD4⁺ T-cells was increased in tumors treated with the anti-B7-H3 antibody, whereas the number of Foxp3⁺ CD4⁺ TILs was decreased by 4.3-fold compared with the isotype control (P < 0.0001; Fig. 3F).

These observations suggested that anti-B7-H3 blockade provided an antitumor reaction mediated by CD4⁺ or CD8⁺ TILs; therefore, we examined whether CD4⁺ or CD8⁺ T-cell depletion using an anti-CD4 or anti-CD8 antibody could prevent the antitumor reaction provided by anti-B7-H3 blockade. CD8⁺ T-cell depletion using an anti-CD8 antibody impaired the antitumor efficacy of the anti-B7-H3 blockade (Fig. 3G). However, CD4⁺ T-cell depletion using an anti-CD4 antibody minimally affected the antitumor efficacy of anti-B7-H3 blockade (Fig. 3G). These results suggest that anti-B7-H3 blockade provided antitumor efficacy mediated by CD8⁺ TILs rather than CD4⁺ TILs.

Finally, we evaluated the function of CD8⁺ TILs in tumor-bearing mice with anti-B7-H3 blockade by using flow cytometry. Anti-B7-H3 antibody treatment caused CD8⁺ TILs to produce more IFNγ than the isotype control (P = 0.0264; Fig. 3H). In addition, anti-B7-H3 antibody treatment caused CD8⁺ TILs to produce more granzyme B than the isotype control (P = 0.0153; Fig. 3I).

In addition, we examined whether anti-B7-H3 blockade could increase the number of functional CD8⁺ TILs in other cancer models. Lewis lung carcinoma (3LL) cells were previously reported to express B7-H3 (43). We subcutaneously implanted 3LL cells in C57BL/6 mice, which were then peritoneally injected with an anti-B7-H3 antibody at 15 mg/kg twice per week or with the isotype IgG control antibody for 10 days. Anti-B7-H3 blockade inhibited growth of the xenografted tumor as compared with that in the isotype IgG control group with a T/C ratio of 36.5% on day 10 after inoculation (P < 0.0001; Fig. 3J). Flow cytometry showed that the CD8⁺ TILs were significantly increased in anti-B7-H3–treated tumors compared with those in the isotype control by 3.4-fold per mm³ tumor (P = 0.046, P = 0.0031, respectively; Fig. 3K). CD8⁺ T cells produced 2.6-fold more granzyme B in anti-B7-H3–treated tumors than in the isotype control (P = 0.019; Fig. 3L). Anti-B7-H3 treatment thus increases CD8⁺ TILs and recovers their effector function. B7-H3 potentially causes immune evasion for tumors via damage to CD8⁺ TILs.

Considering the ability of anti-B7-H3 treatment to increase CD8⁺ TILs, we expected that anti-B7-H3 could enhance the antitumor immune reaction of anti-PD-L1 blockade. Furthermore, whereas anti-B7-H3 blockade showed significant antitumor efficacy, the tumors were not completely eradicated; they persisted and showed growth at later time points. Because Pan02 cells express not only B7-H3 but also PD-L1 (44), it is possible that CD8⁺ TILs gradually undergo functional exhaustion because of PD-1–mediated suppression even after anti-B7-H3 antibody treatment. Despite the significant increase in the number of tumor-infiltrating CD8⁺ T cells with anti-B7-H3 antibody treatment, these cells still expressed PD-1 at levels comparable with those of the IgG control on day 15 (Fig. 4A and B). Therefore, we tested whether the anti-B7-H3 and anti-PD-L1 double blockade would enhance the antitumor immune reaction. Pan02 cells were subcutaneously implanted in C57BL/6 mice, which were then treated with peritoneal injection of anti-B7-H3 antibody at 15 mg/kg twice per week, peritoneal injection of anti-PD-L1 antibody at 10 mg/kg twice per week, or isotype IgG control for a maximum of 84 days. Anti-B7-H3 monotherapy significantly prevented tumor growth until day 31 (P = 0.04 on day 31), whereas these tumors rapidly grew with a decrease in treatment efficacy (Fig. 4C). Similarly, anti-PD-L1 monotherapy significantly prevented tumor growth until day 31 (P = 0.02 on day 31). However, subsequently, the tumor size rapidly increased despite anti-PD-L1 blockade (Fig. 4C). In contrast to the findings for these monotherapies, the combination of both antibodies robustly prevented tumor growth, with a T/C ratio of 8% on day 39 (Fig. 4C). Furthermore, in individual cases, anti-B7-H3 monotherapy and anti-PD-L1 monotherapy did not achieve complete tumor eradication (Fig. 4D). However, use of the combination of both antibodies enabled tumor eradication in two of six mice on day 60. Anti-B7-H3 monotherapy and anti-PD-L1 monotherapy minimally improved survival, whereas the combination doubled the survival time, relative to that in the control group (Fig. 4E). In addition, we evaluated CD8⁺ TILs by using flow cytometry. Tumors were obtained from mice treated with the anti-B7-H3 antibody alone, anti-PD-L1 antibody alone, and a combination of both antibodies on day 32 (n = 6 for each group). The CD8⁺ TIL count was greater in tumors with double blockade of B7-H3 and PD-L1 than in tumors with monotherapy (Fig. 4F). In contrast, the CD4⁺ TIL count was equivalent between these groups (Fig. 4F).

Together, these results suggest that anti-B7-H3 blockade enhanced the antitumor efficacy of anti-PD-L1 blockade presumably due to an increase in the number of CD8⁺ TILs by anti-B7-H3 blockade. Alternatively, anti-B7-H3 blockade alone incompletely prevented tumor growth because CD8⁺ TILs still maintained the exhaustion mediated by PD-L1/PD-1 signaling despite anti-B7-H3 blockade. Although double blockade of B7-H3 and PD-1 signaling potently and durably inhibited tumor growth, toxicity such as body weight loss was minimal (Supplementary Fig. S2). Furthermore, the anti-B7-H3 antibody had no effect on PD-L1 expression, and also the anti-PD-L1 antibody had no effect on B7-H3 expression (Fig. 4G). Therefore, we expected that the combination effect of double blockade occurs through an intrinsic mechanism of CD8⁺ TILs, rather than an extrinsic mechanism such as PD-L1 or B7-H3 upregulation in cancer cells.

The current study observed that the benefit of an anti-PD1 inhibitor was sufficient in patients with PD-L1–expressing NSCLC plus abundant CD8⁺ TILs, but limited in those with B7-H3–expressing NSCLC. A low number of CD8⁺ TILs was observed in tumors with B7-H3 expression, whereas CD8⁺ TILs were abundant in tumors without B7-H3 expression despite having PD-L1 TPS less than 50%. This relationship implies that anti-PD-1 therapy provided antitumor efficacy mediated by CD8⁺ T cells in tumors without B7-H3 expression. A mouse study revealed that anti-B7-H3 blockade increased CD8⁺ TIL number with enhanced effector function. Finally, anti-B7-H3 blockade potently enhanced antitumor reaction by anti-PD-L1 blockade. Our findings suggest that anti-B7-H3 blockade especially with anti-PD-1/PD-L1 blockade is a novel and promising treatment strategy for patients with NSCLC with aberrant B7-H3 expression.

Consistent with previous studies (34–36, 45), tumors from patients with NSCLC, especially squamous cell lung carcinoma and wild-type EGFR NSCLC, preferentially showed high B7-H3 expression levels. Furthermore, to our knowledge, this study is the first to show that B7-H3 expression levels are also correlated with innate refractoriness to anti-PD-1 blockade therapy in NSCLC. B7-H3 expression has been repeatedly reported as a factor for poor prognosis in patients with NSCLC (34, 36, 45). B7-H3 expression might result in not only poor prognosis, but also refractoriness to anti-PD-1/PD-L1 immunotherapy in NSCLC by impairing CD8⁺ T-cell–mediated tumor immunity. In addition, Koyama and colleagues reported the upregulation of alternative immune checkpoints, including Tim-3, in tumor progression following response to anti-PD-1 therapy, suggesting that other immune checkpoints are targetable biomarkers associated with adaptive resistance to PD-1 blockade (46). Although this study did not evaluate B7-H3 expression levels during tumor progression in relation to development of adaptive resistance to anti-PD-1/PD-L1 therapy, the findings indicate that B7-H3 is potentially involved in adaptive resistance to the anti-PD-1/PD-L1 blockade, which requires further examination.

This study showed that anti-B7-H3 blockade increased the number of CD8⁺ T cells and recovered their effector function. Given these results, the subpopulation of NSCLCs expressing B7-H3 might have functionally impaired CD8⁺ T cells. Other studies reported that B7-H3 gene silencing in NSCLC H1299 cells promoted the proliferation of T cells and stimulated these cells to secrete IFNγ (36). B7-H3 has been reported to exert dual functions, that is, costimulatory (32, 33) or coinhibitory functions (37, 38), whereas our current results indicate that B7-H3 dominantly functions in a coinhibitory manner in tumor immunity and impairs anti-PD-1 therapy in NSCLC.

This study showed that B7-H3 expression levels were not significantly related to PD-L1 expression levels. Other researchers have reported that PD-L1 expression is dominantly found in regions of active T-cell infiltration in the tumor microenvironment, which produce the inflammatory cytokine IFNγ (47). In contrast to the reported findings for PD-L1, this study observed that B7-H3 expression was inversely correlated with T-cell infiltration. The molecular mechanism regulating B7-H3 expression remains unclear. However, Xu and colleagues reported that B7-H3 protein expression is inversely correlated with miR-29 levels and that the miR-29–binding site on B7-H3 is conserved in evolution, suggesting that a miRNA-regulatory mechanism is responsible for the differing expression patterns of B7-H3 (48). Given these results, B7-H3 expression might be regulated in tumors independently from PD-L1 expression, and analyses of the expression of both proteins might help to further optimize the subpopulation for anti-PD-1/PD-L1 blockade therapy in NSCLC. Previous research showed that tumors with PD-L1 TPS less than 50% are less responsive to anti-PD-1 therapy (18), whereas in this study, subpopulations with PD-L1 TPS less than 50% had high responsiveness, with an 86% response rate in tumors lacking B7-H3 (Supplementary Table S2).

This study showed that dual blockade of B7-H3 and PD-L1 clearly enhanced antitumor potency compared with that obtained with single blockade. Similarly, others have reported that dual coinhibitory signaling inhibition, such as Tim-3/PD-1 (27, 28), Lag-3/PD-1 (29), CTLA-4/PD-1 (25), or VISTA/PD-1 inhibition (30), also rescues CD8⁺ T cells more vigorously from exhaustion than single signaling blockade. In this study, anti-PD-L1 and anti-B7-H3 double blockade provided more CD8⁺ TILs than PD-L1 or B7-H3 single blockade. Increased CD8⁺ TILs potentially provided a synergistic antitumor effect. For further understanding of the precise mechanism underlying the increase in CD8⁺ TILs by anti-B7-H3 and anti-PD-1 dual blockade, it is necessary to identify the receptor for B7-H3 and its downstream pathway. Recently, checkpoint-inhibitor combination therapy, especially dual anti-PD-1 and anti-CTLA-4 blockade therapy, was found to provide enhanced efficacy but greater toxicity than single blockade in melanoma (49). In contrast, anti-B7-H3 blockade with anti-PD-L1 blockade did not lead to severe immunologic toxicity or body weight loss. We suggest that anti-B7-H3 blockade combined with anti-PD-1 might be a promising strategy against refractoriness to anti-PD1/PD-L1 therapy in B7-H3–expressing NSCLC because of its efficacy and tolerability. Drugs targeting B7-H3 including MGA271 (50) are currently being developed. MGA271 has shown tolerability and preliminarily shown efficacy, and a trial for combination with other immune checkpoint inhibitors including anti-PD-1 antibody has been initiated (50).

In this study, the limited sample size in clinical observation potentially caused a selection bias. Second, we obtained tissue samples mostly at diagnosis, whereas patients were treated with anti-PD-1 as a second-line or later therapy. Intervention with other therapies might have affected the B7-H3 expression. Furthermore, B7-H3 has been repeatedly reported to be a factor for poor prognosis, and our single cohort study could not exclude the possibility that the poor PFS of the B7-H3–positive resulted from this characteristic. However the current clinical observations, combined with the preclinical findings, suggest that B7-H3–expressing NSCLC escape from antitumor immunity via CD8⁺ T-cell repression, and dual PD-1 and B7-H3 signaling blockade therapy is a promising treatment strategy for B7-H3–expressing NSCLC.

K. Yonesaka reports receiving commercial research grants from Daiichi Sankyo Co. Ltd. K. Haratani reports receiving speakers bureau honoraria from AstraZeneca KK, Bristol-Myers Squibb, Ono Pharmaceutical Co., and Pfizer Japan Inc., and is a consultant/advisory board member for AstraZeneca KK. K. Tanaka reports receiving speakers bureau honoraria from AstraZeneca KK, Bristol-Myers Squibb, Merck Serono Co., and Ono Pharmaceutical Co Ltd. K. Nakagawa reports receiving commercial research grants from A2 Healthcare Corp., AbbVie Inc., AC Medical Inc., Astellas Pharma Inc., Bristol Myers Squibb, Chugai Pharmaceutical Co.,Ltd., Daiichi Sankyo Co., Ltd., Eisai Co., Ltd., Eli Lilly Japan K.K., EPS Associates Co., Ltd., EPS International Co.,Ltd., Gritstone Oncology, Inc., ICON Japan K.K., inVentiv Health Japan, Japan Clinical Research Operations, Kyowa Hakko Kirin Co.,Ltd., MSD K.K., Nippon Boehringer Ingelheim Co.,Ltd., Novartis Pharma K.K., Ono Pharmaceutical Co.,Ltd., PAREXEL International Corp., Pfizer Japan Inc., PPD-SNBL K.K, Quintiles Inc., Taiho Pharmaceutical Co.,Ltd., and Takeda Pharmaceutical Co.,Ltd., and speakers bureau honoraria from Astellas Pharma Inc., AstraZeneca K.K., Eli Lilly Japan K.K., MSD K.K., Nippon Boehringer Ingelheim Co.,Ltd., Novartis Pharma K.K., Ono Pharmaceutical Co.,Ltd, and Pfizer Japan Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Yonesaka, K. Hirotani, K. Nakagawa

Development of methodology: K. Yonesaka, K. Haratani

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Yonesaka, K. Haratani, R. Kato, N. Takegawa, T. Takahama, K. Tanaka, H. Hayashi, M. Takeda, O. Maenishi, K. Sakai, T. Okabe, K. Kudo, H. Kaneda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Yonesaka, S. Takamura, O. Maenishi, K. Sakai, Y. Chiba, K. Nakagawa

Writing, review, and/or revision of the manuscript: K. Yonesaka, S. Takamura, K. Tanaka, K. Sakai, Y. Chiba, Y. Hasegawa, M. Yamato, K. Hirotani, K. Nishio, K. Nakagawa

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Yonesaka, K. Haratani, S. Takamura, H. Sakai, S. Kato, K. Sakai, Y. Hasegawa, K. Hirotani, M. Miyazawa

Study supervision: K. Yonesaka, S. Takamura, M. Miyazawa, K. Nakagawa

This study was financially supported by Daiichi Sankyo Co. Ltd. We thank Haruka Yamaguchi, Yume Shinkai, Michiko Kitano, and Mami Kitano at the Department of Medical Oncology, Kindai University Faculty of Medicine, and Shinji Kurashimo at the Central Research Facilities, Kindai University Faculty of Medicine, for technical support. This study was financially supported by Daiichi Sankyo Co. Ltd.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.