Anti-VEGF Antibody Protects against Alveolar Exudate Leakage Caused by Vascular Hyperpermeability, Resulting in Mitigation of Pneumonitis Induced by Immunotherapy

Programmed cell death 1 receptor (PD-1) on T cells plays a major role in suppressing the host's antitumor immune response by interaction with PD-L1 or PD-L2 on tumor cells, macrophages or dendritic cells, and treatment with anti–PD-1 or anti–PD-L1 antibodies generally has high clinical benefit in patients with cancer (1). However, the use of anti–PD-1 or anti–PD-L1 antibodies is often associated with potentially fatal immune-related pneumonitis, an important autoimmune toxicity with significant morbidity and mortality, often necessitating discontinuation of therapy (2, 3). Meta-analyses of published clinical trials of PD-1 and PD-L1 inhibitor therapy for patients with non–small cell lung cancer (NSCLC) show that the incidence of pneumonitis of any grade is significantly higher in patients treated with PD-1 inhibitors than in patients treated with PD-L1 inhibitors (4, 5), indicating that the appropriate clinical selection of these agents may be one of the options for reducing the risk of pneumonitis in NSCLC; however, even when using a PD-L1 inhibitor, the risk of pneumonitis remains. Therefore, the development of a method for controlling pneumonitis is required.

VEGF is an angiogenic factor (6) and also a potent inducer of vascular permeability (7, 8). It has been reported that VEGF plays a significant role in the development of several lung disorders, including chronic obstructive pulmonary disease, pulmonary hypertension, and acute lung injury (9). However, the role of VEGF in checkpoint inhibitor–related pneumonitis has not been investigated.

In this study, we used an IL2- plus IL18-induced mouse pneumonitis model (10) to investigate the roles of PD-L1, PD-L2, and VEGF in pneumonitis, and we investigated whether an anti-VEGF antibody has a mitigation effect on checkpoint inhibitor–related pneumonitis in the phase III IMpower150 trial.

Female 6-week-old C57BL/6 mice were obtained from Japan SLC. All animals were allowed to acclimatize and recover from shipping-related stress for 1 week prior to the study. The animals were allowed free access to chlorinated water and irradiated food, and were kept under a controlled light–dark cycle (12–12 hours). All animal procedures were approved by the Institutional Animal Care and Use Committee at Chugai Pharmaceutical Co., Ltd., and conformed to the Guide for the Care and Use of Laboratory Animals published by the Institute of Laboratory Animal Resources (ILAR).

Mice were treated once a day with an intraperitoneal injection of 10,000 IU/head rhIL2 (PeproTech) and 1 μg/head rmIL18 (R&D Systems). These cytokines were suspended in 200 μL of sterile PBS, and mice treated with 200 μL PBS were used as controls. Anti-mouse PD-L1 antibody (10F.9G2; BioLegend), anti-mouse PD-L2 antibody (TY25; BioLegend), anti-mouse VEGF antibody (B20–4.1.1; gifted from Genentech) or Rat IgG (MP Biomedicals) was administered intraperitoneally to the mice at a dose of 10 mg/kg three times from the day before cytokine treatment. Wet lung weight and moist rale were measured after 5 days of the treatment.

For analysis of lung-infiltrating inflammatory cells, lung tissue was excised from treated mice after 5 days of the treatment, and single-cell suspensions were obtained by mincing lung tissues and homogenizing them by disruption and digestion with a gentleMACS Dissociator and a Lung Dissociation Kit for mice (Miltenyi Biotec). Single-cell suspensions were incubated with anti-Fcγ receptor antibodies (Tonbo Biosciences) and the fixable viability dye FVD506 or FVD780 (eBioscience) at room temperature for 5 minutes, and stained with the following mAbs: mouse CD45 (30-F11), CD4 (RM4–5), CD8α (53–6.7), NK1.1 (PK136), CD11c (HL3), CD103 (M290), CD11b (M1/70), Gr-1 (RB6–8C5), F4/80 (T45–2342), PD-L1 (MIH5), and PD-L2 (TY25) from BioLegend or BD Biosciences (Franklin Lakes). The appropriate conjugated isotype-matched immunoglobulin G (IgG) was used as the control for each. Cells were analyzed using an LSRFortessa X-20 cell analyzer (BD Biosciences) and FlowJo 10 software (Tree Star).

Lung tissue was excised from treated mice after 4 days of treatment. The trachea was surgically exposed and intubated with a syringe catheter. The lungs were subjected to lavage with 0.5 mL PBS three times. Bronchoalveolar lavage fluid (BALF) was obtained from each mouse and cells in BALF were pelleted by centrifugation (500 × g, 5 minutes). The supernatants were stored at −80°C for later measurements.

Lung tissue was excised from treated mice after 5 days of treatment. After opening of the thorax, the lungs were immediately fixed by intratracheal instillation of 10% neutral buffered formalin. After gross examination, the excised tissues were placed in 10% formalin. Sections (4-μm thickness) were cut from paraffin-embedded tissues. Deparaffinized sections were stained with hematoxylin and eosin (HE).

Concentrations of mouse SP-D were measured with a Quantikine ELISA kit (R&D Systems). VEGF was quantified by using a Quantikine ELISA kit (R&D Systems). Albumin was quantified by using a specific ELISA kit (Bethyl Laboratories).

The IMpower150 trial (NCT02366143) was a randomized phase III study that evaluated the safety and efficacy of atezolizumab in combination with carboplatin + paclitaxel with or without bevacizumab compared with treatment with carboplatin + paclitaxel + bevacizumab in metastatic NSCLC (11). The IMpower150 study was done in full accordance with the guidelines for Good Clinical Practice and the Declaration of Helsinki, and all patients gave their written informed consent. We performed the human investigations after approval by the research ethics committee of Chugai Pharmaceutical Co., Ltd. Participants were randomized in a 1:1:1 ratio to the ACP arm (atezolizumab + carboplatin + paclitaxel), the ABCP arm (atezolizumab + carboplatin + paclitaxel + bevacizumab), and the BCP arm (carboplatin + paclitaxel + bevacizumab).

To evaluate statistical significance in the mouse model experiments, data was analyzed with the Wilcoxon test. For two groups, P < 0.05 was considered to indicate a significant difference. For multiple groups, P values were adjusted by the Holm−Bonferroni method (12) by using JMP version 10 software (SAS Institute).

BALF of patients with checkpoint inhibitor–related pneumonitis often suggests lymphocytic inflammation of the alveoli, and BALF cytology often presents an increased proportion of lymphocytes (13). Therefore, to investigate immune-related mechanisms for pneumonitis aggravated by PD-1 signal blockade, we examined the effects of anti-PD-L1 and anti-PD-L2 antibodies in an IL2- plus IL18-induced murine pneumonitis model with symptoms of interstitial lymphocyte infiltration in the lungs. The wet lung weight was significantly higher in the group with dual PD-L1/PD-L2 blockade compared with that in the IL2 plus IL18 alone group (Fig. 1A). Serum pulmonary surfactant protein-D (SP-D) is a useful clinical biomarker for the diagnosis and management of pneumonitis and it also increases in patients with checkpoint inhibitor–related pneumonitis (14–16). In the current model, the serum SP-D level was significantly higher in the group with dual PD-L1/PD-L2 blockade compared with that in the IL2 plus IL18 alone group (Fig. 1B).

Next, to investigate the relation between VEGF signaling and checkpoint inhibitor–related pneumonitis, we examined the effect of VEGF blockade in the IL2- plus IL18-induced pneumonitis model. The wet lung weight was significantly lower in the VEGF blockade group compared with the group with dual PD-L1/PD-L2 blockade (Fig. 1A). The level of SP-D in serum was also significantly decreased by the addition of VEGF blockade (Fig. 1B).

To investigate the mechanism of action through which dual PD-L1/PD-L2 blockade aggravated pneumonitis in the IL2- plus IL18-induced pneumonitis murine model, we analyzed the expressions of PD-L1 and PD-L2 in lung tissues of mice. In normal lung tissue, PD-L1 and PD-L2 were expressed most strongly on CD103⁻ CD11c⁺ dendritic cells (DC; Fig. 2). IL2 plus IL18 treatment upregulated PD-L1 expression on CD103⁻ DCs, CD103⁺ DCs, CD11b⁺ Gr1^high neutrophils, and F4/80⁺ CD11c⁻ macrophages. In contrast, IL2 plus IL18 treatment did not upregulate PD-L2 expression on these cells (Fig. 2).

Next, we analyzed the inflammatory condition of the lung tissue. Histological evaluation of mice treated with IL2 + IL18 + anti–PD-L1 antibody + anti-PD-L2 antibody revealed that the alveolar wall and interstitium had large interstitial infiltration of mononuclear cells and granulocytes (Fig. 3).

Flow cytometry analysis revealed that IL2 plus IL18 significantly increased CD3⁻ NK1.1⁺ NK cells, neutrophils, and macrophages in the lungs and significantly decreased CD8α⁺ T cells, while addition of dual PD-L1/PD-L2 blockade significantly increased CD8α⁺ T cells and CD4⁺ T cells (Fig. 4). However, additional treatment with anti-VEGF antibody did not change the number of inflammatory cells in the lung tissues (Fig. 4).

Although treatment with the anti-VEGF antibody prevented the increases in wet lung weight and serum SP-D level—an index useful for the evaluation of pneumonitis—it did not change inflammatory cell infiltration into lung tissues. A characteristic of this immune-related pneumonitis model is a moist rale caused by air mixing with fluid exudate in the bronchial tubes. A moist rale is also a typical clinical finding in patients with checkpoint inhibitor–related pneumonitis. The mice with symptoms of moist rale showed that alveolar exudate formed in the lung tissue (Fig. 3). Moist rale occurred in the group with PD-L1 blockade and in the group with dual PD-L1/PD-L2 blockade, but it did not occur in the group with PD-L1 + PD-L2 + VEGF blockade (Fig. 5A).

To ascertain whether VEGF blockade protects against vascular-to-alveolar leakage in this model, we measured the albumin concentration in BALF. The albumin concentration in the group with dual PD-L1/PD-L2 blockade was significantly higher than in the IL2 plus IL18 group and in the PD-L1 blockade group (Fig. 5B). Addition of VEGF blockade to the dual PD-L1/PD-L2 blockade treatment significantly decreased the albumin concentration in BALF (Fig. 5B).

Because VEGF blockade protected against vascular-to-alveolar leakage, we evaluated the expression of VEGF in lung tissues, expecting to see an increase. Unexpectedly, IL2 plus IL18 significantly decreased VEGF levels in the lung tissue and dual PD-L1/PD-L2 blockade did not change these decreased VEGF levels (Fig. 5C).

We showed that VEGF blockade prevented immune-related pneumonitis from aggravating in the murine model. Therefore, to investigate whether VEGF modulates the risk of immune-related pneumonitis in patients with NSCLC treated with immune checkpoint blockade therapy, we compared the status of pneumonitis between the atezolizumab + carboplatin + paclitaxel (ACP group) and the atezolizumab + bevacizumab + carboplatin + paclitaxel (ABCP group) in the phase 3 IMpower150 trial. Incidence rates of pneumonitis of any grade were 4.3% in the ACP group and 2.8% in the ABCP group (Fig. 6). Incidence of symptomatic grade 2–4 pneumonitis was 3.8% in the ACP group and 2.0% in the ABCP group (Table 1). Rates of “not recovered/not resolved” outcomes in patients with pneumonitis were 29.4% in the ACP group and 9.1% in the ABCP group (Fig. 6). Among patients with pneumonitis, 5.9% in the ACP group and 27.3% in the ABCP group did not use steroids (Table 2).

We demonstrated that dual treatment with anti–PD-L1 antibody plus anti-PD-L2 antibody worsened the index for evaluation of pneumonitis in an IL2- plus IL18-induced murine pneumonitis model. Because pneumonitis does not develop in PD-1^−/− C57BL/6 mice but is accelerated by PD-1 deficiency in the MRL background which has the risk of the autoimmune damage in lung (17), the aggravation of pneumonitis due to dual PD-L1/PD-L2 blockade may need a pre-existing inflammatory immune microenvironment in the lungs. In patients with NSCLC, preexisting pulmonary fibrosis is reported to be a risk factor for anti–PD-1–related pneumonitis, supporting this hypothesis (18, 19). It has been reported that PD-L2 plays a role in the homeostasis of inflammation in the lungs (20). Our data showed that PD-L2 was expressed on DCs in murine lung tissue in a steady state and that that expression continued without a change even under the inflammatory conditions induced by IL2 plus IL18. Expression of PD-L1, on the other hand, was upregulated under the IL2- plus IL18-induced inflammatory conditions. These results suggest that not only PD-L2 but PD-L1 also may be active in the lungs under inflammatory conditions. In fact, we demonstrated that not only dual PD-L1/PD-L2 blockade but also blockade of PD-L1 alone caused symptoms such as moist rale under the IL2- plus IL18-induced inflammatory conditions. Although there is a higher incidence of pneumonitis with the use of PD-1 inhibitors than with the use of PD-L1 inhibitors (4), it is necessary to be prepared for pneumonitis also when using PD-L1 inhibitors.

With regard to the role of VEGF in checkpoint inhibitor–related pneumonitis, we demonstrated that anti-VEGF antibody could control symptoms of checkpoint inhibitor–related pneumonitis in the murine model, especially the vascular-to-alveolar leakage and moist rale due to the leakage. Although moist rales/crackles are not very commonly manifested symptoms in the patients with immune checkpoint inhibitor-related pneumonitis, they has been listed as one of the clinical findings suspecting the onset of severe pneumonitis recommending a chest CT (21, 22). Furthermore, we demonstrated that in patients with NSCLC, the incidence of symptomatic pneumonitis induced by atezolizumab plus chemotherapy was lower in the group also treated with bevacizumab than in the group without bevacizumab. And even in the broadly defined pneumonitis, including interstitial lung disease, pulmonary fibrosis and etc., this trend does not change (23). Therefore, it is suggested that checkpoint inhibitor–related pneumonitis occurs when two conditions are both met: presence of intrinsic VEGF and aggravation of preexisting inflammation caused by PD-1 signal blockade. On the other hand, our results showed that the level of VEGF expression in lung tissues decreased in the IL2 + IL18 + anti-PD-L1/PD-L2 antibody-induced pneumonitis model. Although the lungs have the highest level of VEGF gene expression among all normal tissues (24), VEGF levels in BALF from patients with acute lung injury are reported to be lower than levels in normal BALF, and a possible explanation for these decreased levels of VEGF may be dilution resulting from alveolar flooding (25). Therefore, the same phenomenon may be happening in our pneumonitis model, and our study demonstrated that despite this lowered level of VEGF, it is still sufficient and critical for the massive alveolar exudate leakage seen in the immune-related pneumonitis model. Because increase in immune cell infiltration caused by PD-1 signal blockade was not affected by co-administration of anti-VEGF antibody, it may not be considered as a fundamental treatment measure to cure immune-related pneumonitis at least from the results using current mouse pneumonitis model. However, as there has been little information about the precise roles of infiltrated immune cells in immune checkpoint inhibitor–related pneumonitis, further studies are required to understand the contribution of each type of immune cells to the development of each symptom.

Although checkpoint inhibitor–related pneumonitis is one of the most serious immune-related adverse events, no preventive measures have been established. Although steroid therapy is an effective treatment, it carries the risk of suppressing the antitumor effect of checkpoint inhibitors (26). We demonstrated that among patients with NSCLC, the clinical outcome of pneumonitis induced by atezolizumab plus chemotherapy was better in patients treated with bevacizumab than in those without bevacizumab. Furthermore, the percentage of patients with pneumonitis who used steroids was lower in the group treated with bevacizumab than in the group without. Anti-VEGF antibodies such as bevacizumab are reported to have additive or synergistic effects when administered in combination with anti-PD-L1 antibodies (such as atezolizumab) in NSCLC, renal cell carcinoma, and hepatocellular carcinoma (11, 27, 28). Anti-VEGF antibodies plus checkpoint inhibitors may be a treatment method that can control at least a part of checkpoint inhibitor–related pneumonitis without loss of therapeutic effects. On the other hands, mechanism of checkpoint inhibitor–related pneumonitis is not clear and there is a limit what we can learn from our research in an IL2- plus IL18-induced murine pneumonitis model and the particular group of patients. Further study will be needed to clarify the clinically relevant mitigation effects of bevacizumab on immune-related pneumonitis.

In conclusion, we have shown that an anti-VEGF antibody protected against vascular-to-alveolar leakage of protein and fluid, resulting in mitigation of checkpoint inhibitor–related pneumonitis in a murine model. Our findings suggest that bevacizumab may be a meaningful therapeutic option for patients receiving checkpoint inhibitors, not only in terms of enhancing their antitumor effect but also in terms of reducing some side effects.

T. Iwai M. Sugimoto, K. Yorozu, M. Kurasawa, and O. Kondoh are an employee of Chugai Pharmaceutical Co., Ltd. No disclosures were reported by the other authors.

T. Iwai: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Sugimoto: Conceptualization, formal analysis, supervision, validation, methodology, writing–original draft, project administration, writing–review and editing. H. Patel: Formal analysis, validation, investigation, visualization, methodology. K. Yorozu: Formal analysis, validation, investigation, visualization, methodology. M. Kurasawa: Validation, investigation, visualization, methodology. O. Kondoh: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

We thank Nobuyuki Ishikura, Daiko Wakita, Kaname Yamamoto, and Kazushige Mori (Product Research Department, Chugai Pharmaceutical Co., Ltd.), and Ayumi Kato and Wataru Kudo (Discovery Pharmacology Department, Chugai Pharmaceutical Co., Ltd.) for their helpful discussion and support in this study; and we thank Hiromi Sawamura, Masako Miyazaki, and Saki Otsuki (Product Research Department, Chugai Pharmaceutical Co., Ltd.) for their technical assistance in the experiments. This study was supported by Chugai Pharmaceutical Co., Ltd., and F. Hoffmann-La Roche, 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.