A novel anti-PD-L1/IL-15 immunocytokine overcomes resistance to PD-L1 blockade and elicits potent antitumor immunity

Abstract

Despite the demonstrated immense potential of immune checkpoint inhibitors in various types of cancers, only a minority of patients respond to these therapies. Immunocytokines designed to deliver an immune-activating cytokine directly to the immunosuppressive tumor microenvironment (TME) and block the immune checkpoint simultaneously may provide a strategic advantage over the combination of two single agents. To increase the response rate to checkpoint blockade, in this study, we developed a novel immunocytokine (LH01) composed of the antibody against programmed death-ligand 1 (PD-L1) fused to interleukin (IL)-15 receptor alpha-sushi domain/IL-15 complex. We demonstrate that LH01 efficiently binds mouse or human PD-L1 and maintains IL-15 stimulatory activity. In syngeneic mouse models, LH01 showed improved antitumor efficacy and safety versus anti-PD-L1 plus LH02 (Fc-sushi-IL15) combination and overcame resistance to anti-PD-L1 treatment. Mechanistically, the dual anti-immunosuppressive function of LH01 activated both the innate and adaptive immune responses and induced a favorable and immunostimulatory TME. Furthermore, combination therapy with LH01 and bevacizumab exerts synergistic antitumor effects in an HT29 colorectal xenograft model. Collectively, our results provide supporting evidence that fusion of anti-PD-L1 and IL-15 might be a potent strategy to treat patients with cold tumors or resistance to checkpoint blockade.

Graphical abstract

Shi et al. developed a novel anti-PD-L1/IL-15 immunocytokine (LH01) that elicits superior antitumor efficacy and better safety profile than the combination of anti-PD-L1 plus non-targeting IL-15 and, combined with bevacizumab, demonstrates synergistic antitumor efficacy in preclinical models. LH01 can activate innate and adaptive immune responses and overcome resistance to PD-L1 blockade.

Introduction

Therapeutic antibodies that block the programmed death-1 (PD-1)/programmed death-ligand 1 (PD-L1) pathway demonstrate cure-like benefits in patients with various types of cancers, but a large proportion of patients experienced a low response rate or rapidly developed resistance to these therapies with relapsed disease.¹ One of the main reasons may be the existence of an immunosuppressive tumor microenvironment (TME), which is caused by altering the immune checkpoint molecule expression, immunosuppressive cytokine secretion, oxygen nutrition status, and so on.² Cytokines play an indispensable role in regulating immune response, including innate and adaptive immunity, and are the cornerstone of cancer immunotherapy. A variety of immune-activating cytokines such as interleukin (IL)-15 have potent antitumor efficacy and can markedly prolong the survival periods of patients, which can be combined practically with immune checkpoint inhibitors (ICIs) to address the issue of resistance and increase response rate.^3,4

Recombinant human IL-15 topped the National Cancer Institute’s list of potential biopharmaceuticals for tumor immunotherapy in 2008.⁵ IL-15 has a unique mechanism of action in which it binds to IL-15 receptor alpha (IL-15Rα) expressed by antigen-presenting cells, then the IL-15/IL-15Rα complex is trans-presented to neighboring natural killer (NK) or CD8⁺ T cells expressing only the IL-15Rβ/γ receptor.⁶ In addition to inhibiting IL-2-induced activation-induced cell death, a process that leads to the elimination of stimulated T cells and induction of T cell tolerance, IL-15 can support long-lasting CD8⁺ T cell memory and effector responses against diseased cells.^7,8 Recombinant IL-15 has demonstrated clinical activity in the treatment of certain cancers, including advanced renal cell carcinoma and metastatic melanoma, and significant increases in the number of memory CD8⁺ T and NK cells were observed in patients’ peripheral blood.^9,10 Increased PD-L1 expression in tumors and decreased IL-15 levels in the TME are correlated with poor clinical outcomes.^11,12 A clinical trial showed that an IL-15 superagonist, ALT-803, can re-induce immunotherapy response in PD-1-relapsed and refractory non-small cell lung cancer (NSCLC).¹³ Unfortunately, the short half-life and the systemic toxicities of high-dose administration, which can cause fever, fall of blood pressure, and flu-like symptoms due to lack of target activity, restrict the further clinical applications of IL-15.¹⁴

Prolonging half-life and increasing targeting ability at the tumor site of this pro-inflammatory cytokine are feasible solutions to the above problems. It has been reported that complexation with the IL-15Rα-sushi domain can improve IL-15 half-life and bioavailability in vivo and be effective in mimicking IL-15 trans-presentation.^15,16 Additionally, the IL-15Rα-sushi domain is a selective and potent agonist of IL-15 action through IL-2/15Rβγ.¹⁷ Immunocytokines are also known as antibody-cytokine fusion proteins, which can utilize the targeting ability of antibody to enrich cytokines at the tumor site. On the one hand, it can enhance tumor-targeting capability and reduce the side effects of cytokines caused by systemic administration. On the other hand, this allows antibodies and cytokines to generate synergistic antitumor effects.^18,19 Hence, it is a practical strategy to generate an immunocytokine composed of anti-PD-L1 and the IL-15Rα-sushi domain/IL-15 complex to enhance antitumor activity.

In this study, we characterized the biochemical activity of LH01, a bifunctional fusion protein designed to overcome resistance to PD-1/PD-L1 blockade via improving the target activity of IL-15 and blocking the PD-L1 pathway concurrently. The anti-PD-L1 moiety of LH01 is based on atezolizumab, which has been approved to treat different types of cancer.^20,21,22 We compared the antitumor efficacy of LH01 versus anti-PD-L1 + LH02 in murine carcinoma models and preliminarily investigated the mechanism by which LH01 overcomes resistance to anti-PD-L1 treatment. For the first time, we evaluated the synergistic antitumor effect of combinative administration of LH01 and bevacizumab.

Results

Biochemical characterization of the bifunctional fusion protein LH01

The fusion of IL15Rα-sushi domain (Ile 31 to Val 115) and human IL15 mutant (IL-15N72D) to the C terminus of the anti-PD-L1 monoclonal antibody was expected to improve the target activity and reduce untoward effects of IL-15 (Figures 1A and S1). A new molecule called Fc-sushi-IL15 (LH02) was also designed as a non-targeting control (Figure 1A). The mature LH01 protein, whose light chain and heavy chain migrate as approximately 25-kDa and 70-kDa proteins under reducing conditions on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) respectively (Figure 1B), was purified by one-step protein A affinity chromatography (Figure 1C). The purification processes of secreted anti-PD-L1 and LH02 proteins are the same as that of LH01. The light and heavy chains of anti-PD-L1 migrate as an approximately 25-kDa and 50-kDa protein, respectively, and LH02 migrates as an approximately 50-kDa protein under reducing conditions on SDS-PAGE (Figure S2).

Figure 1.

Design, preparation, and biochemical characterization of LH01

(A) Schematic representation of fusion proteins: IL-15Rα sushi domain/IL-15 was directly fused to the C terminus of the atezolizumab (LH01) or atezolizumab’s Fc portion (LH02). (B) Purified LH01 was characterized by reduced SDS-PAGE. (C) LH01 was purified by protein A affinity chromatography. (D and E) Binding of LH01 to plate-bound mouse (D) or human (E) PD-L1. Data were analyzed using the one site-total to calculate the EC₅₀ values and are shown as mean ± SD. (F and G) The biological activity was compared with IL-15 monomer at different concentrations by determining the proliferative potential in human Mo7e cells (F) and murine CTLL-2 cells (G). Data were analyzed using the four parameter fit logistic equation to calculate the EC₅₀ values, and graphs are shown as mean ± SD.

In enzyme-linked immunosorbent assays (ELISAs), LH01 bound human or mouse PD-L1 with a profile similar to that of the anti-PD-L1 antibody (half-maximal effective concentration [EC₅₀] = 18.8 ± 2.6 and 10.9 ± 1.4 ng/mL [or 94.1 ± 13.0 and 75.3 ± 9.7 pM], 18.79 ± 3.2 and 7.8 ± 2.7 ng/mL [93.95 ± 16.0 and 54.7 ± 18.9 pM], respectively), indicating that the binding of the anti-PD-L1 moiety was not affected (Figures 1D and 1E). As shown in Figures 1F and 1G, LH01 exhibited weaker proliferative capacity than IL-15 in human Mo7e cells (EC₅₀ = 149.5 and 2.2 ng/mL [or 0.74 and 0.17 nM]), whereas markedly reduced proliferative activities of LH01 were observed compared with IL-15 in mouse CTLL-2 cells (EC₅₀ = 194.5 and 0.039 ng/mL [or 970.4 and 3.03 pM]), which may be explained by the finding that the IL-15Rα-sushi domain was able to bind IL-15 with high affinity and thus inhibited proliferation driven through the high affinity IL-15Rα/β/γ signaling complex of the CTLL-2 cells.¹⁷ In short, LH01 retained strong proliferative capacity in both human Mo7e and mouse CTLL-2 cells.

Prolonged half-life and improved tumor-targeting distribution of LH01

In order to provide medication guidance for the following animal experiments, we explored the pharmacokinetic properties of LH01. Plasma concentrations of LH01 and IL-15 climbed to peaks and then decreased over time, but LH01 decreased much more slowly than IL-15 (Figure 2A). LH01 peaked around 8 h at a concentration of 3,231 ng/mL, whereas IL-15 peaked about 1 h at a concentration of 39 ng/mL (Table 1). The half-lives were calculated to be about 12.52 h for LH01 and 1.02 h for IL-15 monomer, indicating that the fusion of IL15 and the sushi domain of the IL-15Rα to the C terminus of the anti-PD-L1 monoclonal antibody have markedly prolonged the half-life of IL-15 by more than 12-fold (Table 1). To further trace LH01 distribution, we collected various tissues 24 h after mice received treatment with LH01. The in vivo biodistribution of LH01 displayed a certain specificity, with the concentration of LH01 in tumor tissues being 2-fold higher than that in normal tissues (Figure 2B).

Figure 2.

Prolonged half-life and improved tumor-targeting distribution of LH01

(A) Male Balb/c mice aged 9 weeks were intraperitoneally injected with 24.0 μg of LH01 or 3.6 μg of IL-15 (equimolar of IL-15 molecules). The pharmacokinetics curves of LH01 and IL-15 monomer are plotted (n = 5). (B) MC38 tumor-bearing mice (n = 4) were intraperitoneally (i.p.) injected with LH01 at a dose of 1 mg/kg. Tissues were collected at 24 h after injection. The concentrations of LH01 were measured by ELISA. Both graphs show mean ± SEM. ∗p < 0.05.

Table 1.

Pharmacokinetic parameters of IL-15 and LH01

Parameters	IL-15	LH01
Half-life (h)	1.02	12.52
Tmax (h)	1	8
Cmax (ng/mL)	39.26	3231.91
AUC (0→∞) (ng × h/mL)	79.58	81,819.12
MRT (h)	1.62	20.27

Calculated with PK Solver 2.0 for a noncompartmental model.

Tmax, peak time; Cmax, peak concentration; AUC, area under the curve; MRT, mean resident time.

LH01 improves antitumor efficacy and safety versus anti-PD-L1 + LH02 combination

We first explored the antitumor effects of LH01 among different doses in murine MC38 and CT26 colon carcinoma models. LH01 greatly inhibited the growth of both tumors at all the three inspected dosages (Figures 3A and 3B). In the MC38 model, LH01 demonstrated the antitumor activity in a dose-dependent manner (tumor growth inhibition [TGI], 38.5% [1 mg/kg], 56.4% [3 mg/kg], 71.1% [5 mg/kg]) (Figure 3A). LH01 at 3 mg/kg induced similar reduction in CT26 tumor burden compared with 5 mg/kg (TGI, 61.6% [1 mg/kg], 75.3% [3 mg/kg], 79.4% [5 mg/kg]) (Figure 3B). Next, we compared the therapeutic efficacy of LH01 versus anti-PD-L1, LH02, or anti-PD-L1 plus LH02 in murine MC38 and CT26 tumor models. Considering that five of eight CT26 tumor-bearing mice died on day 9 after receiving two intraperitoneal treatments of LH02 (1 mg/kg), we decided against using equimolar doses of LH01 and anti-PD-L1+LH02. Instead, LH01 was compared with anti-PD-L1 (5 mg/kg) and LH02 (0.5 mg/kg). Both MC38 and CT26 colorectal tumors were resistant to anti-PD-L1, while LH02 monotherapy showed moderate antitumor effects with TGI of 32.2% and 30.9%, respectively (Figures 3C and 3D). In MC38 model, anti-PD-L1 + LH02 showed greater antitumor activity than anti-PD-L1 monotherapy, but displayed no significant antitumor efficacy versus LH02 monotherapy (Figure 3C). In CT26 model, anti-PD-L1 + LH02 exerted enhanced antitumor activity compared with anti-PD-L1 or LH02 monotherapy (Figure 3D). Importantly, LH01 at 5 mg/kg induced remarkable reduction in both MC38 and CT26 tumor burden versus anti-PD-L1 plus LH02. Moreover, in the MC38 and CT26 models, LH01 significantly increased median overall survival (mOS) from 27 to 32.5 days and from 22 to 27 days, respectively, compared with anti-PD-L1 + LH02 (Figures 3E and 3F). Notably, in CT26 tumor-bearing mice, anti-PD-L1 + LH02 significantly increased the spleen weight compared with the PBS group, while there was no obvious spleen weight gain in the LH01 group at doses of 1 or 3 mg/kg (equivalent to 0.25 and 0.75 mg/kg of LH02, respectively), indicating that LH01 exerted good tumor-targeting capability (Figure 3G). LH01 was well tolerated in both tumor models, as neither the MC38 nor the CT26 tumor-bearing mice obviously lost weight after treatment (Figures 3H and 3I). In MC38 tumor models, anti-PD-L1 + LH02 showed good tolerability, but in CT26 tumor models, two of six mice died after receiving two intraperitoneal treatments due to systemic toxicity. Collectively, these data illustrate that LH01 has greater antitumor activity than anti-PD-L1 + LH02 and a favorable tolerability.

Figure 3.

LH01 reduces CT26 and MC38 tumor burden and ameliorates safety versus LH02 + anti-PD-L1

(A) MC38 tumor cells (5 × 10⁵, subcutaneously) and (B) CT26 tumor cells (1 × 10⁶, subcutaneously) were implanted into the right flank of female C57BL/6 and Balb/c mice, respectively. Mice were randomized into five groups based on tumor size and treatment initiated when tumors reached 50–100 mm³. Mice were treated with PBS, anti-PD-L1 (5 mg/kg) + LH02 (0.5 mg/kg), or LH01 (1 mg/kg, 3 mg/kg, or 5 mg/kg) on days 8, 11, 14, and 17 for MC38 model and days 5, 8, 11, and 13 for CT26 model, respectively, and the progression curves of tumor volumes are depicted (n = 6). Tumors were removed and weighed after euthanasia. (C–F) MC38 or CT26 tumor-bearing mice were randomized into five groups based on tumor size, and treatment was initiated when tumors reached 50–100 mm³. Mice were treated with PBS, anti-PD-L1 (5 mg/kg), LH02 (0.5 mg/kg), anti-PD-L1 (5 mg/kg) + LH02 (0.5 mg/kg), or LH01 (5 mg/kg) on days 8, 11, 14, and 17 for MC38 model (n = 10) and days 5, 8, 11, and 13 for CT26 model (n = 8), respectively. (C) and (D) are tumor progression curves of MC38 and CT26 models, respectively; (E) and (F) are survival of MC38 and CT26 models, respectively. (G) Spleens of CT26 tumor-bearing mice were removed and weighed after euthanasia (n = 6). (H and I) Body weights of MC38 tumor-bearing mice (H) and CT26 tumor-bearing mice (I) were recorded (n = 6). All graphs show mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

LH01 induces both innate and adaptive immune cell activation in tumors

IL-15 is a pleiotropic cytokine that plays a vital role in regulating innate and adaptive immunity and can strongly expand CD8⁺ T and NK cells with much weaker regulatory T cell (Treg)-stimulating activity.²³ To explore the changes in peripheral blood, splenic and intratumoral CD8⁺ T, NK, or Treg populations, we performed flow cytometry analysis of dissociated peripheral blood, tumors, and spleens from CT26 tumor-bearing mice. LH01 markedly increased CD8⁺ tumor-infiltrating lymphocytes (TILs) compared with PBS or LH02 + anti-PD-L1, which suggested that LH01 can selectively activate CD8⁺ T cells in the tumor (Figure 4A). Besides its effects on CD8⁺ T cells, LH01 treatment also increased the tumor-infiltrating NK cells (TINKs) and dramatically decreased the tumor-associated Tregs (Figures 4B and 4C). In comparison with the PBS group, LH01 treatment only elicited slight increase in splenic CD8⁺ T cells, while anti-PD-L1 + LH02 treatment markedly increased splenic CD8⁺ T and NK cells (Figures 4D, 4E, and S3A–S3C). LH01 treatment demonstrated much weaker increases in splenic CD8⁺ T and NK cells than anti-PD-L1 + LH02 treatment, indicating preferable tumor-targeting capacity and safety of LH01. To our surprise, we found that LH01 treatment remarkably increased splenic Tregs versus the PBS group, which may be beneficial in reducing systemic toxicity (Figures 4F and S3D). Both anti-PD-L1 + LH02 and LH01 treatments resulted in comparable increases in peripheral blood CD8⁺ T and NK cells (Figures 4G, 4H, S3E, and S3F). These results illustrated that LH01 can specifically stimulate the tumor-infiltrating CD8⁺ T and NK cells, resulting in better antitumor efficacy than anti-PD-L1 + LH02.

Figure 4.

LH01 increases both adaptive and innate immune cell activation. Flow cytometry analysis of peripheral blood, spleens, and tumors from CT26 tumor-bearing mice treated as described in Figure 3

(A–F) The percentages of intratumoral CD8⁺ T cells (A), NK cells (B), and Tregs (C) and splenic CD8⁺ T cells (D), NK cells (E), and Tregs (F) are shown for populations of CD3⁺, CD45⁺, and CD4⁺ lymphocytes, respectively. The cell numbers of CD8⁺ T (G) and NK cells (H) in peripheral blood were counted. Data are reported as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

LH01 overcomes resistance to PD-1/PD-L1 blockade in an MC38 model correlated to inducing immunostimulatory TME

Our results showed that mice did not respond to anti-PD-L1 treatment at a dose of 10 mg/kg with primary resistance to therapy, while LH01 displayed obvious therapeutic improvements (Figure 5A). Both anti-PD-L1 and LH01 treatments relieved inhibition of T cells via the PD-1/PD-L1 axis, and resulted in a significant increase in CD8⁺ TILs compared with the control group (Figures 5B and 5C). The above meant that impairment of T cell function caused by immunosuppressive TME may contribute to resistance to PD-L1 blockade. As an important feature of TME, high reactive oxygen species (ROS) are detrimental to the survival and function of T lymphocytes.²⁴ We explored whether LH01 can inhibit the apoptosis induced by oxidative stress. The results of flow cytometry demonstrated that the apoptosis rate was significantly higher than that of the control group (57.4% ± 3.0% versus 23.9% ± 1.7%) after incubating T lymphocyte cell line CTLL-2 with 50 μM H₂O₂ for 18 h, whereas the addition of LH01 could dramatically reverse the apoptosis induced by H₂O₂ (57.4% ± 3.0% versus 6.2% ± 0.6%). LH02 can also markedly reverse the apoptosis caused by H₂O₂ (Figure 5D). Furthermore, we detected H₂O₂ levels within tumors and found that H₂O₂ levels were significantly lowered by LH01 treatment compared with PBS, indicating LH01 has the potential to nurture a favorable TME (Figure 5E).

Figure 5.

LH01 overcomes resistance to anti-PD-L1 treatment related to inducing immunostimulatory TME

(A) C57BL/6 mice transplanted subcutaneously with MC38 cells (right) were treated with PBS, anti-PD-L1, or LH01 as described in Figure 3, and the progression curves of tumor volumes are depicted (n = 6). (B) The percentage of CD8⁺ TILs is shown for populations of CD3⁺ lymphocytes.

(C) Immunohistochemistry analysis of CD3⁺ T cells.

(D) Well-grown CTLL-2 cells were planked with 2 × 10⁵ cells per well on 12-well cell culture plates. Flow cytometric analysis of CTLL-2 cells untreated, treated with 50 μM H₂O₂, 50 μM H₂O₂ + LH01, or 50 μM H₂O₂ + LH02 for 18 h. The proportion of late apoptotic cells was statistically analyzed (n = 4). (E) Intratumoral levels of H₂O₂ were detected. (F and G) Quantitative real-time PCR analysis was performed to measure the expression levels of pd-l1, tgf-β1 (F), and ifng, tnfα, and tbet (G) in tumor tissues. Data are reported as the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Transforming growth factor β1 (TGF-β1) has a key role in preserving a cold TME by promoting regulatory T cell differentiation, reducing CD8⁺ T cytotoxicity, suppressing CD4⁺T cell differentiation, and blocking NK cell proliferation.^25,26 It has been revealed that TGF-β participates in the resistance to PD-1/PD-L1 blockade.^27,28 Anti-PD-L1 treatment markedly increased expression levels of TGF-β1, which may partly explained the resistance to its treatment (Figure 5F and Table S1). Intriguingly, compared with the anti-PD-L1 group, LH01 treatment did not significantly alter PD-L1 expression levels, but remarkably reduced TGF-β1 levels (Figure 5F). In addition, LH01 treatment induced a significant elevation of ifng, tnfa, and tbet gene transcription in the tumor compared with anti-PD-L1 treatment, indicating a T helper (Th) 1-skewed TME (Figure 5G; Table S1). The above results suggested that LH01 created a favorable and immunostimulatory TME for effector T cells to exert enhanced antitumor efficacy and sensitize tumors to immunotherapy.

Combination therapy with LH01 and bevacizumab exerts synergistic antitumor effect

Previous studies have shown that angiogenesis is an essential process for proliferation of solid tumors and vascular endothelial growth factor (VEGF) can elicit immunosuppressive effects in the TME, suggesting that anti-angiogenic agents and LH01 could generate synergistic antitumor efficacy.^29,30 Bevacizumab is a molecularly targeted drug that can inhibit tumor angiogenesis by binding to VEGF-A around the tumor.³¹ There are already clinical trials combining anti-PD-L1 antibody and VEGF/VEGF receptor (VEGFR) inhibitors (e.g., NCT04356729; NCT04698213) to treat cancers, so we propose that the combination of LH01 and bevacizumab would display synergistic antitumor effects. In the HT29 xenograft model, mice experienced a slight and significant reduction in tumor volume and tumor weight after receiving bevacizumab and LH01, respectively (Figures 6A and 6B). Two mice died after receiving four intraperitoneal treatments of LH01, possibly due to graft-versus-host disease caused by infused human peripheral lymphocytes (Figure 6C). Combination therapy of LH01 and bevacizumab markedly reduced tumor volume and tumor weight (Figures 6A and 6B) versus LH01 or bevacizumab alone and showed synergistic antitumor activities, as the combination index (CI) was calculated to be 1.09. Larger areas of necrosis were observed in the combination regimen than the other three groups (Figure 6D). Both LH01 monotherapy and combination therapy obviously reduced Ki67 expression levels compared with PBS or anti-VEGF treatment, which demonstrated a poor proliferative and metastatic ability of tumor cells (Figure 6D). Our results indicate that LH01 is a promising candidate to exert enhanced antitumor activities in combination with angiogenesis inhibitors.

Figure 6.

Combining LH01 with bevacizumab enhances antitumor activity

(A and B) NOD-SCID mice were inoculated subcutaneously with 3 × 10⁶ HT29 cells, and subsequently received 3 × 10⁶ fresh human PBMCs intravenously on the same day. Mice were randomized into four groups and treatment initiated when tumors reached 40–80 mm³. Mice were treated intraperitoneally with LH01 (3 mg/kg), bevacizumab (10 mg/kg), or LH01 (3.0 mg/kg) + bevacizumab (10 mg/kg) at days 5, 8, 11, 14, 17, and 20 (n = 6). (A) Tumor volumes were measured every 3 days. (B) Tumor weight (day 23). (C) Body weights of mice. (D) Tumor tissues were fixed, followed by H&E staining and immunohistochemical staining for Ki67. CI was calculated based on the formula Ea + b/(Ea + Eb − Ea × Eb). Data are reported as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Discussion

The antibodies targeting the immune checkpoint have become the protagonist of immunocytokines with the breakthrough progress of ICIs for tumor treatment in the past few years. Recently, Martomo et al. preliminarily explored the antitumor activity of anti-PD-L1/IL-15 fusion protein KD033 on various solid tumor models in mice, whereas they did not provide further rationale for KD033 to treat patients with cold tumors or resistance to ICIs.³² LH01 differs from KD033 in structure: the antibody part is atezolizumab and the sushi domain is 85 amino acids in length with higher binding affinity to IL-15 than the 65 amino acids of KD033.¹⁵ Interestingly, our results show that LH01 can overcome primary resistance to PD-1/PD-L1 blockade by downregulating TGF-β1 levels within the TME without markedly affecting PD-L1 expression. Additionally, we demonstrate that LH01 can induce the development of an inflamed TME through enhancing the populations of CD8⁺ TILs and TINKs with a decrease in Treg populations and upregulating Th1-type cytokines. Resistance is a major obstacle to cancer immunotherapy, and its mechanisms are varied and complicated. Amelioration of primary resistance to anti-PD-L1 therapy by using LH01 may be related to converting inherently immunosuppressive TME to immunosupportive TME.

The format of an immunocytokine has a significant impact on its targeting activity. The homodimeric format usually possesses a high binding avidity to the target and a long residence time at the tumor site. In our study, we found that the LH01 did not substantially increase weight of the spleen of tumor-bearing mice at doses of 1 and 3 mg/kg compared with the control group, which indicated that LH01, a homodimeric fusion protein, had good tumor-targeting activity. It should be noted that LH01 has a favorable safety at a high dose (5 mg/kg) in both CT26 and MC38 models, with no difference in mice body weights. On the other hand, fast blood clearance profiles caused by “cytokine sink effect” may be beneficial to reduce untoward effects associated with the use of potent pro-inflammatory cytokine payloads, which perhaps partially explains the good safety of LH01 in mice.

In principle, certain immunocytokines could mimic the action of bispecific antibodies (BsAbs). The cytokine moiety can engage in a binding interaction with its cognate receptor on the surface of T cells, thus creating an immunological synapse with the tumor cells.³³ Bispecific T cell engagers (BiTEs), one kind of BsAb, have been attracting a great deal of attention due to their unique mechanism of action and significant antitumor activity. BiTEs demonstrated remarkable efficacy in B cell hematologic malignancies, but the use of such new drugs to treat solid tumors is unsatisfactory.³⁴ BiTEs can direct T cells to specific tumor antigens and activate T cells directly, whereas the immunosuppressive factors (e.g., ROS, hypoxia, TGF-β) can impair the proliferation and survival of T and NK cells in the TME, which can significantly reduce its antitumor activity.^35,36 However, the pro-inflammatory cytokine moiety of immunocytokines can convert the tumor immunosuppressive microenvironment to a certain extent, and promotes the activation and proliferation of T and NK cells, which is supported by our results that LH01 can inhibit the apoptosis of CTLL-2 under high levels of ROS and downregulate the TGF-β1 levels in the TME. In terms of BiTEs, a major restriction of tumor-associated antigen selection in solid tumors is that low-level expression is often found in normal tissue, exposing the patients to a risk of “on-target, off-tumor” toxicity.³⁶ In the case of immunocytokines, it seems not so demanding for antigen specificity, and the adverse effects of immunocytokines are mainly caused by cytokine moiety.

Given that LH01 is well tolerated in preclinical models, we believe that this bifunctional fusion protein represents a promising candidate for inclusion in combination therapy regimens. We have validated this in our murine models; combining LH01 with a VEGF-A inhibitor bevacizumab elicits enhanced and superior antitumor activity compared with either agent alone. Vascular abnormalities resulting from elevated levels of proangiogenic factors (e.g., VEGF and angiopoietin 2) are a hallmark of most solid tumors.³⁷ Additionally, proangiogenic factors have been reported to play a vital role in immunosuppressive TME.³⁸ For instance, VEGF can directly elevate PD-L1 expression on dendritic cells, resulting in impaired function of T cells, and VEGF can also directly bind to VEGFR2 on Tregs and myeloid-derived suppressor cells (MDSCs), which increases these immunosuppressive cells in the TME.^39,40 Our results further indicate that the combination of the other inhibitors of the VEGF signaling pathway, including small molecule receptor tyrosine kinases inhibitors (sunitinib, sorafenib, and pazopanib), with LH01 has the potential to generate greater antitumor effects.

Our study has some limitations. First of all, the mechanisms by which LH01 overcomes resistance to anti-PD-L1 remain to be further studied because a variety of factors, including other immune checkpoints, cancer neoantigens, soluble major histocompatibility complex (MHC)-related molecules, and cytokines, in the TME also affect anti-cancer immune responses.⁴¹ Moreover, more experiments are needed to verify whether overcoming resistance to anti-PD-L1 is related to inhibition of TGF-β1 and ROS production by LH01. In addition, we noted that the CT26 tumor-bearing mice showed slightly ungroomed hair without weight loss after the third administration of LH01 at a dose of 5 mg/kg, which was associated with side effects caused by cytokine IL-15. However, no significant pathological changes were observed in organs (heart, liver, spleen, lung, and kidney) of mice treated with LH01 at a dose of 5 mg/kg compared with PBS treatment (data are not shown). Given that most of immunocytokines still produce the same adverse effects as cytokines in clinical trials,⁴² further efforts should be made to improve safety by structure-based design.

In conclusion, LH01 elicits superior antitumor efficacy and a good safety profile in preclinical models. LH01 possesses the potential to help T cells resist damage from unfavorable factors and overcome primary resistance to PD-1/PD-L1 blockade related to modulating TME through stimulating positive factors such as Th1-type cytokines and inhibiting negative factors such as ROS and TGF-β1. Our work provides support for clinical use of LH01 for treatment of patients with resistance to ICIs or cold tumors. LH01 can be combined more practically with other therapies to target even more pathways to improve clinical benefit. Altogether, LH01 represents a promising candidate for further clinical investigation.

Materials and methods

Cloning, expression, and purification

The plasmids encoding LH01, LH02, and anti-PD-L1 were constructed as shown in Figure S1. The DNA sequences of IL-15 mutant (IL-15N72D) and IL-15Rα sushi domain (Ile 31 to Val 115) were amplified by PCR using the pIL-15 and psIL-15Rα/Fc we reported previously as template.⁴³ All the plasmids were constructed by inserting the DNA fragments into the vector we used before.⁴³ The light and heavy chain expression plasmids of LH01 or anti-PD-L1 were mixed at 2:1 and co-transfected using liner polyethylenimine (PEI) with a molecular weight of 25 kDa (Polysciences, Warrington, PA, United States). LH02 was produced by transfecting HEK293E cells with Fc-sushi-IL15-expression plasmid alone. LH01, anti-PD-L1, and LH02 were all purified by affinity chromatography using a protein A affinity column (GE Healthcare, Piscataway, NJ, United States) and analyzed in reducing condition on SDS-PAGE.

Cell lines

HEK293E and CTLL-2 cell lines were kept in our laboratory and cultured as previous descriptions.⁴³ Mo7e, MC38, and CT26 murine colon carcinoma cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, United States). Mo7e cells were grown in RPMI 1640 (Gibco, Waltham, MA, United States) containing 10% FBS (Gibco) and 10 ng/mL human GM-CSF (Sino Biological, Beijing, China). MC38 and CT26 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS. All cells above were maintained under aseptic conditions and incubated at 37°C with 5% CO₂.

Measurement of LH01 binding and pharmacokinetics by ELISAs

ELISAs for PD-L1 binding

ELISAs were performed using standard methods. Briefly, 96-well ELISA plates (Corning, Corning, NY, United States) were coated by incubating with 1.0 μg/mL of recombinant human or mouse PD-L1 (Novoprotein, Nanjing, China) at 4°C overnight, then washed four times with PBST (PBS, 0.05% Tween 20) and blocked with 5% BSA for 2 h at room temperature. After washing the plates, serial dilutions (1:3) of LH01 and anti-PD-L1 were added to the plates in duplicate and incubated at room temperature for 2 h. Plates were washed four times and incubated with Peroxidase AffiniPure Goat Anti-Human IgG (H + L) (Jackson ImmunoResearch, West Grove, PA, United States, 1:10,000 dilution) at room temperature for 1 h. After being washed, TMB single-component substrate solution (Solarbio, Beijing, China) was added to the plates and incubated in the dark for 3–5 min. After terminating the reaction with 2 M sulfuric acid, absorbance was read at 450 nm.

Pharmacokinetic evaluation of LH01and IL-15 by ELISAs

Plasma samples were drawn from mice 0.5, 1, 2, 4, 8, 12, 24, and 48 h after treatment with LH01, and 0.5, 1, 2, 4, and 6 h after treatment with IL-15 monomer. A 96-well ELISA plate, previously coated overnight at 4°C with 1.0 μg/mL of recombinant human PD-L1, was incubated with plasma samples for 2 h from mice treated with LH01. The following experimental procedure was the same as described above. The human IL-15 ELISA Pair Set (Sino Biological) was used for the quantitative determination of IL-15 monomer.

Cell proliferation assay

Mo7e cells were wash with human GM-CSF free medium (RPMI1640 + 10% FBS) and seeded into 96-well plates with 2 × 10⁴ cells in a volume of 50 μL per well. After 4 h of starvation, serial dilutions (1:3) of LH01 or IL-15 were added to the plate in sextuplicate at 50 μL per well to achieve a final density of 2 × 10⁴ cells/100 μL/well. After being incubated for 96 h at 37°C with 5% CO₂, the cell viability was measured using Cell Counting Kit-8 (Dojindo, Kumamoto, Japan). Absorbance was read at 450 nm, and the final optical density 450 (OD₄₅₀) value was calculated as the reading of sample well minus the reading of blank well containing medium. The method used in the CTLL-2 cell proliferation assay was identical to that of Mo7e, except that the number of cells used was 1 × 10⁴ and the incubation time was 72 h.

Animal experiments

Female Balb/c, C57BL/6, and NOD-SCID mice aged 6–8 weeks were purchased from Shanghai SLAC Laboratory Animal Co. and reared under specific pathogen-free conditions. All experiments were approved by the Animal Care and Use Committee of Shanghai Jiao Tong University. All mice were treated humanely throughout the experimental period. Human peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll density gradient centrifugation to serve human T lymphocytes with procedures previously we described.⁴³ For antitumor studies, tumors were measured every 2 days using digital calipers, and volumes were calculated as (length × width²)/2. TGI was calculated as TGI(%) = 100 × (1 − T/C). T and C were the mean tumor volume of the treated and control groups, respectively. The humane endpoint for tumor size was 1,500 mm³.

Flow cytometric analysis of peripheral blood, splenic, and intra-tumoral CD8⁺ T, NK, or regulatory T cells

Tumor tissues (150 mg) were finely minced and digested with 4 mL of lysis solution (2 mg/mL collagenase IV and 1.2 mg/mL hyaluronidase). The digested tumor tissues were filtered through 200-mesh nylon net to obtain the cell suspension, centrifuged, then the supernatant was discarded, and the cells washed once with 6 mL of fluorescence-activated cell sorting (FACS) buffer (PBS + 2% FBS). The cells were re-suspended in 6 mL of FACS buffer, and filtered through 200-mesh nylon net again to obtain a pre-treated single cell suspension. The spleens were gently ground and lymphocytes were isolated with lymphocyte separation medium. The lymphocytes of peripheral blood were also isolated with lymphocyte separation medium (Dakewe, Beijing, China).

Cell samples were blocked with anti-mouse CD16/CD32 monoclonal antibody 2.4G2 (BD Biosciences, San Jose, CA, United States) at 4°C for 15 min and incubated with surface marker antibodies at 4°C for 25 min. For the detection of Tregs, cell samples were further incubated with FOXP3 Fix/Perm buffer (BioLegend, San Diego, CA, United States) for 20 min and FOXP3 Perm Buffer for 15 min at room temperature before anti-FOXP3 was added.

The antibodies and reagents were used as follows: anti-mouse CD45-Percp/Cyanine 5.5 (BioLegend), anti-mouse CD45.2-PE (BioLegend), hamster anti-mouse CD3e-FITC (BD Biosciences), rat anti-mouse CD4-PE (BD Biosciences), rat anti-mouse CD8a-APC (BD Biosciences), rat anti-mouse Nkp46-Alexa Flour 647 (BD Biosciences), anti-mouse CD25-Brilliant Violet 421 (BioLegend), and anti-mouse/rat/human FOXP3-Alexa Fluor 647 (BioLegend). Flow cytometry was performed on a CytoFLEX cytometer (Beckman Coulter) and analyzed by FlowJo 10 (TreeStar, Ashland, OR, United States).

Flow cytometric analysis of cell apoptosis

Flow cytometry was performed to detect the apoptosis of CTLL-2 cells by using Annexin V-FITC/PI Apoptosis Detection Kit (Vazyme) and analyzed by FlowJo 10 (TreeStar).

RNA isolation and qRT-PCR analysis of mRNA expression

Total RNAs of pre-treated tumor tissues were extracted by using Ultrapure RNA Kit (Cwbio, Beijing, China). cDNA was synthesized using a PrimeScript RT Master Mix (Takara, Tokyo, Japan), and real-time qRT-PCR was performed on an Applied Biosystems 7500 Fast Real-Time PCR System (Thermo Fisher Scientific, Eugene, OR, United States) using Hieff qPCR SYBR Green Master Mix (Yeasen, Shanghai, China). The primer sequences are listed in Table S1. All results were normalized to GAPDH expression and calculated using the 2^−(ΔΔCt) method.

H₂O₂ detection

Tumor tissue homogenates were used. The assays were performed using the Hydrogen Peroxide Assay Kit (Beyotime, Haimen, China) according to the manufacturer’s instructions.

Histopathological and immunohistochemistry analysis

The tumor tissues were fixed in 4% paraformaldehyde, and then embedded in paraffin, sectioned (4 μm), and stained with hematoxylin and eosin (H&E). After dewaxing and hydration, the tumor sections were treated with heat-induced epitope retrieval and 3% H₂O₂ for 15 min to block the endogenous peroxidase activity. Next, the tumor sections were blocked with 5% BSA for 30 min and incubated with anti-mouse CD3 rabbit antibody (Servicebio, Wuhan, China) or anti-human Ki67 rabbit antibody (Servicebio, Wuhan, China) at 4°C overnight. Afterward, the sections were incubated with the HRP-conjugated goat anti-rabbit secondary antibody (Servicebio) for 50 min. Finally, the sections were stained with DAB detection kit (Dako, Copenhagen, Denmark) and hematoxylin. Then the slides were observed under the OLYMPUS BX53 Microscope and photographed.

Statistical analysis

Statistical analyses were performed in Prism 7.0 (GraphPad, San Diego, CA, United States). The statistical significance of differences between experimental groups was determined with two-tailed Student’s t test and analysis of variance. Survival was analyzed using log rank (Mantel-Cox) test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001).