Next-generation anti-PD-L1/IL-15 immunocytokine elicits superior antitumor immunity in cold tumors with minimal toxicity

Summary

The clinical applications of immunocytokines are severely restricted by dose-limiting toxicities. To address this challenge, here we propose a next-generation immunocytokine concept involving the design of LH05, a tumor-conditional anti-PD-L1/interleukin-15 (IL-15) prodrug. LH05 innovatively masks IL-15 with steric hindrance, mitigating the “cytokine sink” effect of IL-15 and reducing systemic toxicities associated with wild-type anti-PD-L1/IL-15. Moreover, upon specific proteolytic cleavage within the tumor microenvironment, LH05 releases an active IL-15 superagonist, exerting potent antitumor effects. Mechanistically, the antitumor efficacy of LH05 depends on the increased infiltration of CD8⁺ T and natural killer cells by stimulating the chemokines CXCL9 and CXCL10, thereby converting cold tumors into hot tumors. Additionally, the tumor-conditional anti-PD-L1/IL-15 can synergize with an oncolytic virus or checkpoint blockade in advanced and metastatic tumor models. Our findings provide a compelling proof of concept for the development of next-generation immunocytokines, contributing significantly to current knowledge and strategies of immunotherapy.

Graphical abstract

Highlights

• •The flanking moieties of IL-15 sterically mask its activities
• •Attenuated “cytokine sink” effect reduces systemic toxicities of LH05
• •The release of IL-15 superagonist from antibody restores potent antitumor effects
• •LH05 induces a Th1-type TME to treat cold tumors

Shi and Liu et al. develop a conditionally activated anti-PD-L1/IL-15 immunocytokine that employs steric hindrance to conceal IL-15 activity. Characterized by reduced toxicity and enhanced tumor targeting, this immunocytokine elicits superior antitumor immunity by converting “cold” tumors into “hot” ones, offering a compelling proof of concept for the advancement of next-generation immunocytokines.

Introduction

Antibody-cytokine fusion proteins (also known as immunocytokines) delivering these immunostimulatory payloads to tumor lesions can substantially broaden the therapeutic window of cytokine therapy. Additionally, the combination of antibody and cytokine components in immunocytokines would generate synergistic antitumor effects.¹ Some immunocytokines based on interleukin-2 (IL-2), IL-12, tumor necrosis factor alpha (TNF-α), etc. have been investigated in clinical trials. Notably, L19-IL-2 and L19-TNF, which target fibronectin’s extra domain B, have advanced to phase III trials (ClinicalTrials.gov: NCT02938299 and NCT03567889).² Earlier developed immunocytokines predominantly employed antibodies targeting highly expressed targets in the tumor microenvironment (TME), such as fibronectin and fibroblast activation protein.³ With the considerable advancements of immune checkpoint inhibitors (ICIs) in cancer immunotherapy, antibodies targeting immune checkpoints have recently emerged as the main protagonists of immunocytokines.^4,5

However, immunocytokines can be trapped by cognate receptors in circulation before reaching their target cells (the so-called “cytokine sink” effect).⁶ Only a small fraction of the immunocytokine can be taken up by the neoplastic lesion (in the best cases, 0.01%–0.1% injected dose/g of tumor), often resulting in toxicity profiles similar to that of the parental cytokine.⁷ For instance, patients treated with KD033 (an anti-PD-L1/IL-15 immunocytokine) at a dosage of 50 μg/kg experienced severe lymphocytopenia, although this dose is much lower than the clinical dose of anti-PD-L1 (10–20 mg/kg).^8,9 Therefore, it is crucial to develop practical strategies to overcome the safety challenges associated with immunocytokines and accelerate their clinical applications.

To reduce the systemic toxicity of immunocytokines, one strategy is to engineer cytokines to attenuate their activities. For example, AcTaferon, comprising human interferon α2 (IFNα2; Q124R), which is 100-fold less active on mouse cells compared to murine IFNα fused to anti-CD20, showed a significantly improved safety profile in contrast with wild-type IFNα2.⁶ However, these mutants pose a challenge to the balance between insufficient antitumor activity at low doses and the risk of systemic toxicity at high doses. Alternatively, the prodrug strategy that conditionally activates cytokines in the TME holds great promise for the development of next-generation immunocytokines.^10,11 Fu et al. engineered masked IL-2, IL-12, IL-15, and type I IFN with their natural receptors.^12,13,14,15 These pro-cytokines reactivate after being cleaved by tumor-associated enzymes within the TME. Although the receptor masking strategy can reduce the peripheral activity of the cytokine, the introduction of the masking receptor complicates the structure.

IL-15 is a highly attractive immunostimulatory cytokine renowned for its remarkable activity in treating various cancer types.^16,17 Immunocytokines with IL-15 as a payload have shown great potential in clinical applications, including KD033, targeting PD-L1, and BJ-001, targeting integrin.^18,19 We have previously developed an anti-PD-L1/IL-15 immunocytokine (LH01) that can overcome anti-PD-L1 resistance and elicit both innate and adaptive immune responses. However, as observed with abovementioned immunocytokines, LH01 also induces systemic toxicity similar to that of IL-15.²⁰

To address safety challenges and accelerate the clinical applications of immunocytokines, in the present study, we propose next-generation anti-PD-L1/IL-15 (LH05), a prodrug masking IL-15 with an innovative steric hindrance strategy. It can be selectively cleaved within the TME by a tumor-associated protease to release the IL-15/IL-15Rα-sushi domain (ILR; an IL-15 superagonist),²¹ which would have broader and more potent anticancer effects as compared to being bound to the antibody. Through this design, LH05 successfully addressed the safety concerns associated with IL-15-based immunocytokines and demonstrated enhanced efficacy, especially in “cold” tumors. Furthermore, we explored the comprehensive mechanisms by which LH05 modifies the TME. Overall, our work provides a preclinical proof of concept for the development of next-generation immunocytokines.

Results

Systemic toxicity restricts the efficacy of the anti-PD-L1/IL-15 immunocytokine in cold tumors

We have demonstrated the potent antitumor efficacy of the anti-PD-L1/IL-15 immunocytokine (LH01) in syngeneic and xenograft tumor models in previous studies. In this study, we further studied the therapeutic effect of LH01 in treating cold tumors. We observed that LH01 was well tolerated at 2.5 mg/kg but only exerted a slight antitumor activity in the RM-1 syngeneic prostate model with a “cold” immune landscape (Figures S1A–S1C). When the dosage was increased to 5 mg/kg, LH01 demonstrated significant antitumor activity (Figure S1A). However, it induced significant body weight loss and even death (half of the mice died) after two treatments (Figures S1B and S1C). In short, the dose-limiting toxicities hinder the therapeutic efficacy of LH01 in treating cold tumors.

To improve the efficacy and reduce toxicity, we attempted to mitigate the IL-15 activity. We then engineered LH03 by fusing IL-15 to the C terminus of anti-PD-L1 and the N terminus of the sushi domain via a flexible linker (Figure 1A). LH03 demonstrated decreased affinity toward IL-15Rβ compared with LH01 (Figure S1D). Besides, LH03 (half-maximal effective concentration [EC₅₀] = 39.24 μg/mL or 194.2 nM) induced 220-fold less proliferative activity than LH01 (EC₅₀ = 0.177 μg/mL or 0.88 nM) in human Mo7e cells, suggesting that it effectively masked IL-15’s immunostimulatory activity (Figure S1E). As expected, the safety of LH03 was largely improved, and no body weight loss was observed even at a dose of 10 mg/kg (Figure S1F). However, compared to the control, no significant antitumor efficacy was observed after LH03 treatment (Figure S1G). Additionally, LH03 exerted no significant antitumor effects but good tolerability at a dosage of 10 mg/kg in the MC38 and Renca models (Figures S1H–S1K). Altogether, our findings demonstrated that it is difficult to balance the toxicity and efficacy of immunocytokines by reducing cytokine activity. A practical strategy or design is imperative to address the challenges related to immunocytokine drug development.

Figure 1.

Structure-based tumor-conditional anti-PD-L1/IL-15 design

(A) Schematics of LH01, LH03, and LH05.

(B) The proliferative potential of LH05 and uPA-cleaved LH05 was compared with LH01 in human Mo7e cells (n = 3 technical replicates). Data were analyzed using the four-parameter-fit logistic equation to calculate the EC₅₀ values.

(C and D) Binding of anti-PD-L1, LH01, and LH05 to plate-bound human (C) or mouse (D) PD-L1 (n = 2 technical replicates). Data were analyzed using the one site-total to calculate the EC₅₀ values.

(E and F) Binding of anti-PD-L1, LH01, and LH05 to CHO-K1 cells stably expressing human (E) or mouse (F) PD-L1 (n = 3 technical replicates). Data were analyzed using the one site-total to calculate the EC₅₀ values.

All graphs are shown as mean ± SD. See also Figures S1 and S2.

Steric masking of IL-15 activity and in vitro activation of LH05 by tumor-specific protease

We sought to develop an engineered IL-15 blockade that retains its antitumor activity while limiting systemic exposure. Considering that the ILR complex has been reported as an IL-15 superagonist, we devised a next-generation IL-15-based immunocytokine (LH05) by incorporating a protease-cleavable linker between the antibody and ILR, which can employ steric hindrance caused by the Fc fragment and the sushi domain to mask IL-15 activity. The cleavable linker acts as a switch for IL-15 activity. Before its cleavage, IL-15 is shielded by the joint forces of the Fc fragment and sushi domain. After its cleavage, ILR would be released, restoring the antitumor activity. The cleavable linker was chosen for its sensitivity to protease, which is overexpressed in various human carcinomas: urokinase-type plasminogen activator (uPA) (Figure S2A). A schematic of LH05 is shown in Figure 1A. We simulated the conformational structures of LH01 and LH05 by using AlphaFold, which showed that the IL-15 portion in LH01 was free and that the receptor-binding sites were exposed. Contrarily, the IL-15 portion of LH05 was restricted due to steric hindrance caused by the Fc fragment and the sushi domain (Figures S2B and S2C).

SDS-PAGE analysis revealed that LH05, but not LH03, can be cleaved after incubation with uPA (Figure S2D). LH05 (EC₅₀ = 30.31 μg/mL or 147.5 nM) induced 168-fold less proliferative activity than LH01 (EC₅₀ = 0.177 μg/mL or 0.88 nM) in Mo7e cells. When LH05 was cleaved, it restored the activity by more than 30-fold (EC₅₀ = 4.9 nM) (Figure 1B). LH05 also showed a decreased IL-15Rβ binding affinity compared with LH01 due to IL-15 masking, explaining its weaker proliferative activity (Figure S1D). In ELISAs, both fusion proteins bound to human PD-L1 with a profile similar to that of the anti-PD-L1 antibody (EC₅₀ = 21.83, 29.95, and 10.02 ng/mL or 109.04, 145.80, and 69.28 pM for LH01, LH05, and anti-PD-L1, respectively) (Figure 1C) as well as similar affinity for mouse PD-L1 as the anti-PD-L1 antibody (EC₅₀ = 22.23, 34.90, and 10.32 ng/mL or 111.03, 169.90, and 71.36 pM for LH01, LH05, and anti-PD-L1, respectively) (Figure 1D). Furthermore, we used flow cytometry to assess the affinity of anti-PD-L1, LH01, and LH05 toward CHO-K1 cells stably expressing human or mouse PD-L1. The results showed that all three proteins exhibited comparable affinity for the cell-surface antigens (human PD-L1: EC₅₀ = 812.6, 633.3, and 764.7 ng/mL or 4.05, 3.08, and 5.28 nM; mouse PD-L1: EC₅₀ = 1115.0, 1035.0, and 1529.0 ng/mL or 5.55, 5.03 and 10.55 nM for LH01, LH05, and anti-PD-L1, respectively) (Figures 1E and 1F). Our results demonstrated that the anti-PD-L1 portion of LH05 was unaffected, and the ILR portion would be preferentially released within the TME to restore IL-15 activity.

LH05 exhibits an excellent safety profile in vivo

Given its lower immunostimulatory activity in vitro, we suppose that LH05 would attenuate the expansive capacity of peripheral lymphocytes and reduce systemic toxicity in vivo. To confirm whether LH05 has a significantly improved safety profile compared to LH01, we treated mice with PBS, LH01 (5 mg/kg), or LH05 (10 mg/kg). After two LH01 treatments, all mice experienced dramatic body weight loss and eventually died within 6 days. Contrarily, none of mice treated with LH05 lost weight or died even after six injections (Figures 2A and 2B). Compared with the PBS treatment, LH01 treatments induced a 229.3% increase in spleen weight, whereas double doses of LH05 only resulted in a 70.9% increase, indicating that LH05 can effectively shield IL-15 activity in circulation (Figure 2C).

Figure 2.

LH05 significantly reduces systemic toxicity compared with LH01

(A and B) Female BALB/c mice were intraperitoneally injected with PBS, LH01 (5 mg/kg), or LH05 (10 mg/kg) every 3 days, with body weight changes (A) and survival (B) monitored (n = 8).

(C) Spleens of mice were extracted and weighed after euthanasia on day 5 (n = 4).

(D) The percentages of splenic CD8⁺ T cells are shown for populations of CD3⁺ lymphocytes (n = 4).

(E) The CD8⁺ T cells in peripheral blood were counted (n = 4).

(F) The percentages of splenic NK cells are shown for populations of CD45⁺ lymphocytes (n = 4).

(G) The NK cells in peripheral blood were counted (n = 4).

(H and I) Blood samples (n = 4) were collected after euthanasia on day 5, and plasma cytokine levels were measured using ELISAs (H); ALT and AST plasma levels were also quantified (I).

All graphs show the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant by an unpaired t test or one-way ANOVA. The log rank (Mantel-Cox) test was used to assess survival. See also Figure S3.

Interestingly, LH05 treatment did not lead to a substantial increase in splenic CD8⁺ T proportion or peripheral blood CD8⁺ T cell counts compared to PBS treatment (Figures 2D, 2E, and S3A–S3E). It retained some stimulatory activities on splenic and peripheral blood natural killer (NK) cells, although they were much weaker than those of LH01 (Figures 2F, 2G, and S3A–S3E). Moreover, unlike LH01, LH05 did not significantly trigger cytokines, such as IFN-γ and IL-6, further indicating that the risk of systemic toxicity induced by LH05 was greatly reduced (Figure 2H). LH01 treatments also caused increased plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels, whereas LH05 treatments did not (Figure 2I). Additionally, neither LH01 nor LH05 substantially increased the plasma creatinine and urea levels compared to PBS, implying that no renal injury occurred (Figures S3F and S3G). Overall, these findings suggest that LH05 was effectively sheltered against peripheral activity and adverse effects in vivo.

LH05 extends half-life due to the attenuated “cytokine sink” effect

Immunocytokines may rapidly disappear from circulation before reaching tumor tissues due to the ubiquitous expression of their cognate receptors.⁶ Given its considerably reduced affinity for the IL-15 receptor, it’s anticipated that LH05 would confer pharmacokinetic properties superior to LH01. As expected, the plasma concentrations of LH05 exhibited a slower rate of decline than those of LH01, with calculated half-lives of 8.40 and 3.45 h, respectively, following intravenous injection. These findings suggest that the reduced “cytokine sink” effect of the masked prodrug could prolong the half-life of LH01 by approximately 2.4-fold (Figure 3A). Moreover, we also evaluated the pharmacokinetic properties of LH05 in cynomolgus monkeys. The half-life of LH05 was determined to be 87.03 h and 211.14 h at doses of 0.2 mg/kg and 0.6 mg/kg, respectively (Figure S4A).

Figure 3.

Prolonged half-life and improved tumor-targeting distribution of LH05

(A) Male BALB/c mice were injected intravenously with 1 mg/kg LH01 or LH05 (equimolar). The plasma concentration-time curves were plotted (n = 5). Pharmacokinetic parameters were calculated with PK Solver 2.0 for a non-compartmental model (Cmax, peak concentration; AUC, area under the curve; MRT, mean resident time).

(B) RM-1 tumor-bearing mice (n = 4) were intravenously injected with 1 mg/kg LH01 or LH05, and tissues were collected at 18 h and 24 h post injection, respectively. The concentrations of LH01 or LH05 were measured using ELISA.

(C) The cleavage efficiency of LH05 in tumors and organs were evaluated at 18 h and 24 h post injection (n = 4).

(D) Cy5.5-labeled LH05 (1 mg/kg) was intravenously injected into RM-1 tumor-bearing mice. LH05 accumulation in the tumors was tracked by the IVIS Spectrum in vivo imaging system (n = 3).

All graphs show the mean ± SD. ∗p < 0.05, ∗∗∗p < 0.001 by an unpaired t test. See also Figure S4.

To investigate the tissue distribution of LH01 and LH05, mouse tissues were collected 18 h, 24 h, 48 h, and 72 h after treatment. Tumor tissue concentrations of LH05 were significantly higher than those of LH01 18 h and 24 h after treatment (Figure 3B). However, both LH01 and LH05 at a dosage of 1 mg/kg were nearly entirely eliminated across various organs 48 h and 72 h after administration. Subsequently, we adjusted the dosage to 3 mg/kg, which demonstrated a consistent pattern of LH05 concentrations exceeding those of LH01, attributed to its longer half-life (Figure S4B).

Although uPA has been reported to be highly expressed in multiple tumors, it is also found in normal tissues, such as liver, spleen, and kidney.²² This poses a risk of non-selective cleavage of LH05 in healthy tissues. Therefore, we further investigated the selectivity of LH05 cleavage between tumor and normal tissues. Our results showed that LH05 was predominantly cleaved in tumor tissues when compared to any other tissue at all measured time points. Notably, we detected a relatively higher degree of cleavage of LH05 in the spleen, although significantly lower than that observed in the tumor (Figures 3C and S4C). Therefore, we compared the concentrations of biologically active IL-15 (ILR) in both tumor and spleen tissues. The data revealed that ILR levels in tumors progressively exceeded those in spleens over time, particularly at 24 h and 48 h (Figures S4D and S4E). Notably, the ILR concentrations were generally comparable in tumors and spleens. Considering the safety data in Figure 2, it is obvious that this concentration of active ILR in the spleen is not sufficient to induce significant toxicity. To trace the fate of LH05 at the tumor site, we conjugated LH05 with a Cy5.5-maleimide tracer and injected it into tumor-bearing mice for whole-body imaging at different time points. The bioluminescence data indicated a gradual increase in tumor fluorescence intensity, which reached a peak at approximately 5 h and started to decline from 11 h (Figure 3D). These findings demonstrate the accumulation and preferential cleavage of LH05 in tumors.

LH05 exhibits potent antitumor activity with significantly reduced toxicity

In order to guide research on the in vivo anti-tumor efficacy, we first measured uPA expression across various tumor cell lines. RM-1 and Renca cells exhibited relatively high uPA expression levels, while MC38 cells demonstrated low expression (Figures S4F–S4H). We first investigated the antitumor effects of LH05 in the RM-1 prostate carcinoma model. LH01 was well tolerated at 2.5 mg/kg but exerted much weaker antitumor activity than LH05 at 10 mg/kg. Notably, LH05 was well tolerated, and no mice had obvious weight loss. The LH02 (an IL-15 superagonist) dosage used in this study was 0.25 mg/kg, as established in previous research.²⁰ We observed that LH05 also demonstrated superior antitumor efficacy compared to anti-PD-L1+LH02 (Figures 4A–4C).

Figure 4.

LH05 exerts potent antitumor efficacy with reduced toxicity

Tumor growth curves and survivals were plotted. The body weights of tumor-bearing mice were recorded.

(A–C) On days 9, 12, and 15 (n = 12), RM-1 tumor-bearing mice were intravenously injected with IgG (10 mg/kg), anti-PD-L1 (10 mg/kg) + LH02 (0.25 mg/kg), LH01 (2.5 mg/kg), or LH05 (10 mg/kg).

(D–F) On days 8, 11, 14, and 17 (n = 10), MC38 tumor-bearing mice were intravenously injected with IgG (10 mg/kg), anti-PD-L1 (10 mg/kg) + LH02 (0.25 mg/kg), LH01 (2.5 mg/kg), or LH05 (10 mg/kg).

(G–I) On days 9, 12, 16, and 21 (n = 6), Renca tumor-bearing mice were intravenously injected with IgG (10 mg/kg), anti-PD-L1 (10 mg/kg) + LH02 (0.25 mg/kg), LH01 (2 mg/kg), or LH05 (10 mg/kg).

All graphs show the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant by an unpaired t test or one-way ANOVA. The log rank (Mantel-Cox) test was used to assess survival. See also Figure S4.

We further explored the antitumor effects of LH05 in the murine MC38 colon carcinoma model. LH05 exhibited antitumor efficacy comparable with anti-PD-L1+LH02, but it was somewhat weaker than LH01 (Figure 4D). Although LH05 did not improve the median overall survival as much as LH01 (28 vs. 32.5), the difference was not statistically significant (Figure 4E). Notably, LH01 and anti-PD-L1+LH02 induced significant body weight loss after two treatments, but not LH05, suggesting that the superior antitumor effects of LH01 were at the expense of toxicity (Figure 4F).

The above results illustrate that the antitumor efficacy of LH05 relies on uPA expression in the TME. To support this conclusion, further experimentation using the Renca renal cell carcinoma model revealed that LH05 generated a greater antitumor effect than LH01 or anti-PD-L1+LH02 (Figures 4G and 4H). Besides, among the treatments, LH01 induced the most significant decreases in body weight in Renca tumor-bearing mice (18.14% from day 9 to day 33, p < 0.001), indicating its systemic toxicity (Figure 4I). The above findings from the three tumor models suggest that LH05 has superior tolerability while maintaining an uncompromised overall therapeutic effect in a proteolytic cleavage-dependent manner in comparison to LH01.

LH05 induces both innate and adaptive immune responses for tumor control

In RM-1 tumor-bearing mice, anti-PD-L1+LH02 significantly increased the spleen weight compared with immunoglobulin G (IgG), but there was no obvious spleen weight gain observed in the LH05 treatment at a dose of 10 mg/kg (equivalent to 2.5 mg/kg of LH02 for IL-15), implying that LH05 had very weak peripheral immunostimulatory activity (Figure 5A). We then performed flow cytometric analysis to explore the changes in splenic and intratumoral CD8⁺ T or NK populations. We observed that LH01, LH05, and anti-PD-L1+LH02 treatments markedly decreased the frequency of splenic CD4⁺ T cells compared to IgG treatment (Figure 5B). The percentage of splenic CD8⁺ T cells increased significantly in the LH05 and anti-PD-L1+LH02 groups but not in the LH01 group (Figure 5C). Furthermore, both LH05 and anti-PD-L1+LH02 treatments led to a significantly decreased splenic CD4/CD8 ratio than the other treatments, indicating a stronger immune response (Figure 5D). All other treatments markedly increased the splenic NK cells when compared to IgG treatment, but the percentage of NK cells was much lower in the LH01 group than in the LH05 group (Figure 5E). Interestingly, although LH05 treatment significantly increased the percentage of splenic CD8⁺ T and NK cells, it did not induce spleen weight gain compared to IgG treatment.

Figure 5.

LH05 induces both adaptive and innate immune cells activation

(A) The spleens of RM-1 tumor-bearing mice were extracted and weighed after euthanasia (n = 5).

(B and C) The frequency of splenic CD4⁺ T cells (B) and CD8⁺ T cells (C) for CD3⁺ lymphocytes, respectively.

(D) The ratio of CD4⁺ to CD8⁺ T cells was calculated.

(E) The percentage of splenic NK cells for CD45⁺ lymphocytes was determined.

(F) The expression of the memory cell markers CD62L and CD44 on splenic or intratumoral CD8⁺ T cells was assessed.

(G–I) The percentage of intratumoral CD8⁺ T cells (G) within the population of CD3⁺ lymphocytes and the frequency of IFNγ⁺ (H) or granzyme B⁺ (I) CD8⁺ T cells were assessed.

(J–L) The percentage of intratumoral NK cells (J) within the population of CD45⁺ lymphocytes and the frequency of IFNγ⁺ (K) or granzyme B⁺ (L) NK cells were determined.

All graphs show the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant by an unpaired t test. See also Figure S5.

To differentiate the immunostimulatory effects of LH05 treatment in the spleen and tumor, we investigated the CD8⁺ T cell phenotypes. The results revealed that LH05 significantly induced the generation of effector memory CD8⁺ T cells within the tumor, whereas central memory CD8⁺ T cells were more prevalent in the spleen. Moreover, the proportion of naive CD8⁺ T cells in the spleen was higher than that in the tumor (Figures 5F and S5A).

LH05 treatment resulted in comparable increases in CD8⁺ tumor-infiltrating lymphocytes (TILs) than LH01 (Figure 5G). To assess whether LH05 treatment enhanced the effector function of CD8⁺ T cells, we determined the IFN-γ and granzyme B⁺ expression of CD8⁺ TILs. We found that LH05 treatment significantly increased the frequencies of both CD8⁺ IFNγ⁺ and CD8⁺ granzyme B⁺ T cells within the tumor compared with IgG treatment, but no significant difference was observed between the LH05 and LH01 groups (Figures 5H and 5I). Additionally, LH05 induced similar proportions of effector memory CD8⁺ T cells compared to LH01 within the TME (Figure S5B). However, LH05 treatment led to significantly higher levels of tumor-associated activated NK cells than LH01 treatment, potentially accounting for the enhanced antitumor efficacy of LH05 (Figure 5J). LH05 treatment increased the frequencies of IFN-γ and granzyme-expressing CD335⁺ NK cells compared to IgG treatment, but no significant difference was observed between the LH01 and LH05 groups (Figures 5K and 5L). Altogether, these data suggest that LH05 can activate the CD8⁺ T and NK cells for tumor inhibition.

Both CD8⁺ T and NK cells recruited by LH05 contribute to its antitumor efficacy

First, we calculated the correlation coefficients of IL-15 expression and immune infiltration levels by employing the Microenvironment Cell Population-counter, xCELL, and Cell-type identification by estimating relative subsets of RNA transcripts absolute mode algorithms. Then, we depicted the landscape of IL-15 correlating with immune cell infiltrates in various The Cancer Genome Atlas cohorts. Our resulting heatmap showed a statistically significantly positive correlation between IL-15 expression and immune infiltration of NK cells, particularly activated NK cells, and the central and effector memory subset of CD8⁺ T cells in the majority of cancers (Figure S6).

To ascertain which cell type contributes to LH05-mediated tumor control, we depleted the CD8⁺ T or NK cells in RM-1 tumor-bearing mice with respective depletion antibodies. The results showed that the depletion of NK cells completely abrogated the antitumor efficacy of LH05, indicating that NK cells played an essential role (Figure 6A). Depleting CD8⁺ T cells also compromised LH05’s therapeutic effect, suggesting that CD8⁺ T cells are also required for antitumor immunity (Figure 6A). We then used FTY720 to further determine whether the pre-existing immune cells within the tumor or recruited cells are indispensable for LH05’s anticancer effect. The experiment revealed that inhibiting lymph node egress almost entirely eliminated LH05’s efficacy (Figure 6B). Additionally, the RM-1 tumor is known as a typical “cold tumor,” with few pre-existing T cells. Altogether, these findings suggest that LH05’s antitumor activity is primarily dependent on CD8⁺ T and NK cells that infiltrate the TME from the circulation, making it a promising candidate in the treatment of “cold tumors.”

Figure 6.

The recruited CD8⁺ T cells and NK cells contribute to LH05-mediated antitumor efficacy

(A and B) Growth curves of RM-1 tumors of mice treated with IgG, LH05, and CD8⁺ T or NK cell depletion (A) or FTY720 (B) in the presence or absence of LH05 (n = 6).

(C and D) Tumor tissues were extracted for RNA-seq analysis (n = 3). The heatmaps depict gene expression alterations of chemokines and receptors (C) and leukocyte-mediated cytotoxic effectors (D) in response to three LH05 treatments, as indicated by the log2 (fold change) values.

(E and F) The expression levels of Pdl1, Cxcl9, and Cxcl10 (E) and Ifng, Tbet, and Tnf (F) in the TME were measured using quantitative real-time PCR (n = 4).

All graphs show the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant by an unpaired t test or one-way ANOVA. See also Figure S6.

To further evaluate the impact of LH05 treatment on immune responses, we conducted RNA sequencing (RNA-seq) of RM-1 tumors treated with or without LH05. A gene set enrichment analysis revealed that LH05 treatment positively modulated the expression profile of chemokines and receptors. Specifically, there was a decrease in the expression of pro-tumor chemokines and receptors, such as Cxcl1 and Ccl28, whereas the expression of chemokines and receptors with antitumor properties, such as Cxcl9 and Cxcl10, was increased (Figure 6C). These findings suggest that LH05 may potentially modulate the TME by altering the chemokine and receptor signaling balance toward an antitumor immune response. Moreover, LH05 treatment led to increased expression of genes related to leukocyte-mediated cytotoxic effector activity (Figure 6D). These results are consistent with the findings from those obtained by flow cytometry (Figure 5), indicating that LH05 has immunostimulatory effects on the TME.

As reported with LH01,²⁰ the Pdl1 level was also significantly up-regulated by LH05, implying an enhancement of antitumor immune responses. CXCL9 and CXCL10 are critical factors that facilitate immune cell migration to the TME and bring “heat” to tumors.²³ LH05 treatment resulted in a dramatic increase in Cxcl9 and Cxcl10 expression, potentially explaining the recruitment of CD8⁺ T cells in RM-1 tumors (Figure 6E). Compared to IgG treatment, LH05 treatment also significantly increased the expression of Ifng, Tnf, and Tbet in the tumor, suggesting a T helper (Th) 1-skewed TME (Figure 6F). It has been reported that IL-15 promotes intratumoral immune cell functions via a cytokine network involving XCL1, IFN-γ, CXCL9, and CXCL10.²⁴ Taken together, upon reaching the TME, LH05 activates the immune response through ILR release, leading to the recruitment of CD8⁺ T and NK cells, promoting their expansion and cytotoxicity and inducing Th-1-type cytokine secretion to exert a potent antitumor immunity.

LH05 restores response to immunotherapy in U251 cold tumors

Glioblastoma (GBM) is a highly malignant primary brain tumor with a 5-year survival rate of <5% despite various treatment strategies.²⁵ GBMs are considered “cold” tumors, characterized by poor lymphocyte infiltration and an immunosuppressive TME, which poses challenges for ICIs.²⁶

Given LH05’s ability to overcome ICIs resistance and re-induce immunotherapeutic responses, we evaluated its antitumor efficacy in the U251 GBM xenograft model. LH01 was used as a control with a dosage of 3 mg/kg because of safety concerns. The results showed that LH01 failed to elicit a robust immune response to inhibit tumor growth when compared to that of PBS. Contrarily, LH05 treatment significantly suppressed tumor growth compared with PBS or LH03 treatment, indicating that LH05 can be cleaved in the TME and trigger profound antitumor immunity (Figures 7A and 7B). Furthermore, LH05 treatment significantly reduced Ki67 expression of tumors when compared with the other three treatments, demonstrating the reduced proliferation and metastasis ability of tumors (Figure 7C). Overall, LH05’s ability to overcome immunotherapy resistance and stimulate antitumor immunity in the GBM model highlights its potential as a therapeutic strategy for other tumor types with similar immunosuppressive characteristics.

Figure 7.

LH05 exerts enhanced antitumor efficacy in U251 cold tumors and synergizes with an oncolytic virus or checkpoint blockade to control advanced and metastatic cold tumors

(A) NOD CRISPR Prkdc Il2r Gamma mice were inoculated subcutaneously with 2 × 10⁶ U251 cells and received 4.0 × 10⁶ fresh human peripheral blood mononuclear cells intravenously on day 4. Mice were treated with PBS, LH03 (10 mg/kg), LH01 (3 mg/kg), or LH05 (10 mg/kg) intraperitoneally on days 5, 8, and 11 (n = 5). Tumor volumes were measured.

(B) On days 15, mice were euthanized, and tumors were removed and weighed.

(C) Immunohistochemical staining for Ki67 was performed on the tumor tissues.

(D and E) Treatment was initiated when tumors reached 200 mm³. On days 9, 12, and 15 (n = 8), B16-F10 tumor-bearing mice were intravenously injected with IgG (10 mg/kg) or LH06 (10 mg/kg). An oncolytic virus (5 × 10⁵ plaque-forming units) was injected intratumorally on days 9 and 12. Shown are tumor volumes (D) and immunofluorescence staining for CD8 (green) and Foxp3 (red) of tumor tissues (E) (scale bar, 50 μm; n = 2).

(F and G) B16-F10 tumor cells (5 × 10⁵) were intravenously inoculated into C57BL/6 mice. On days 5, 8, and 11 (n = 8), mice were intravenously injected with IgG control (10 mg/kg), anti-PD-1 (10 mg/kg), LH06 (10 mg/kg), and LH06 (10 mg/kg) + anti-PD-1 (10 mg/kg). The mice were sacrificed on day 15. The colonies on the lungs were counted.

(H) LH05 was incubated with human serum at 37°C for 24 or 72 h before the cleavage was measured by ELISA (n = 4).

(I) LH05 was incubated with human cancer homogenate or adjacent normal tissues homogenate at 37°C for 24 h, and the cleavage efficiency was detected by ELISA (n = 3 for glioma, 8 for lung cancer, and 5 for colon cancer).

Data are shown as the mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant by an unpaired t test or one-way ANOVA. See also Figure S7.

Combination therapy of LH06 with an oncolytic virus or checkpoint blockade exerts synergistic antitumor effects in advanced and metastatic melanoma

The B16-F10 melanoma model is well accepted as a typical “cold” tumor. Given the low uPA and high matrix metalloproteinase (MMP) expression in B16-F10 cells (Figures S4F–S4H), we substitute the urokinase substrate linker in LH05 with an MMP-2/14 substrate linker to generate LH06. In vitro, upon MMP cleavage, LH06 can restore its proliferative activity on Mo7e cells by more than 30-fold, similar to that of LH05 (Figures S7A and S7B).

In a late-stage B16-F10 cold tumor model, we assessed the combination therapy of LH06 and V200, an oncolytic poxvirus. The results demonstrated that LH06 monotherapy elicited potent antitumor effects, comparable to the efficacy observed with V200 monotherapy (Figure 7D). Notably, the combination of LH06 and V200 exhibited substantial synergistic antitumor effects, with a combination index of 1.23. Immunofluorescence analysis revealed that LH06 increases CD8⁺ T cells in the tumor without increasing regulatory T cells (Tregs) compared to IgG treatment (Figure 7E). Unexpectedly, the CD8⁺-to-Treg ratio, a critical indicator of successful immunotherapy, was comparably upregulated in V200 and LH06 + V200-treated groups. We speculate that the combination treatment showed stronger antitumor activity compared to the V200 monotherapy, perhaps due to LH06 enhancing the effector function of CD8⁺ T cells rather than increasing CD8⁺ T cell infiltration. Moreover, histopathological section analysis illustrated that LH06 treatment did not induce liver or kidney damage (Figure S7C). We further investigated the infiltration of IL-6 and IFN-γ in these organs, finding them undetectable by our assay (data not shown). In conclusion, the combination of LH06 and V200 exhibited strong antitumor activity without observable toxicity in the advanced B16-F10 melanoma model.

Anti-PD-1 has demonstrated effectiveness in treating metastatic melanoma, yet response rates remain relatively low.²⁷ Moreover, recent research demonstrated that the combination of anti-PD-1 and anti-PD-L1 can be synergistic as they mediate overlapping and non-overlapping interactions between tumors and immune cells (PD-1/PD-L1) or between immune cells in the TME (PD-L1/CD80).¹⁸ We assessed the antitumor effects of the combination of LH06 and anti-PD-1 in the B16-F10 metastatic melanoma model. LH06 and anti-PD-1 monotherapies significantly suppressed lung tumor metastasis with similar levels of effectiveness (Figures 7F and 7G). Notably, the combination therapy demonstrated a synergistic effect that outperformed the monotherapies. These findings indicate that LH06 has considerable potential for the treatment of advanced and metastatic tumors.

LH05 is stable in human serum and susceptible to patient tumor-specific proteolytic cleavage

To evaluate the translational potential of the tumor-conditional anti-PD-L1/IL-15 for clinical use, we verified LH05’s efficient and selective cleavage in various primary human tumor samples. LH05 was first incubated with serum from healthy human donors (n = 4) for 24 or 72 h. LH05 underwent slight cleavage after 24 h, with only approximately 40% cleavage occurring within 72 h (Figure 7H), demonstrating favorable stability in human serum.

We then obtained diverse tumors and their corresponding peri-tumoral tissues from patients to assess the specificity of LH05 cleavage. As expected, the cleavage efficiency varied across individuals. Colon tumors, for instance, displayed proficient LH05 cleavage (>40% within 24 h), but lung and glioma tumors exhibited only 35% or less after 24 h (Figure 7I). Notably, no LH05 cleavage was observed in any homogenates of the adjacent normal tissues, indicating that LH05 is stable in non-tumorous human tissues (Figure 7I). Overall, these data suggest that LH05 has a low risk of systemic toxicity, owing to its substantial peripheral stability and specific activation in human tumors. However, selecting the appropriate tumor types is crucial to guarantee efficient cleavage in vivo.

Discussion

Immunocytokines are designed to enhance the targeting activity of cytokines, but only a modest 10-fold increase in targeted activity is reportedly achieved, which provides a limited increase in the therapeutic index.²⁸ In clinical studies, the majority of immunocytokines still has a dose-limiting toxicity similar to the parental cytokines.²⁹ To achieve more effective modalities of immunocytokines, further reducing systemic toxicity and increasing antitumor activity are imperative.

One solution for reducing systemic toxicity of immunocytokines is to engineer cytokines with reduced affinity for their cognate receptors.^30,31,32 Decreasing affinity toward the cognate receptor can reduce the “cytokine sink” effect, thereby extending half-life. Additionally, the lower biological activity of the engineered cytokines allows for higher doses and the related immunocytokine accumulation at the tumor site, such as the IL-2 mutants with reduced affinity for IL-2Rα or IL-2Rβ/γ.^33,34,35 However, these mutants reduced the affinity of immunocytokines for both tumoral and peripheral lymphocytes, posing a challenge to the balance between insufficient antitumor activity at low doses and the risk of systemic toxicity at high doses.

Prodrug-based strategies for conditionally activating cytokines in the TME can potentially improve their safety profile while maintaining the antitumor activity. One of the most promising directions for achieving tumor-localized cytokine activation is by leveraging tumor-associated proteases. Until now, various masking domains have been used to shield cytokines, including native cytokine receptors, antibody fragments, anti-cytokine antibodies, and peptides.¹¹ Fu et al. have reported cognate receptor-masked IL-2, IL-12, IL-15, and IFN-α prodrugs.^12,14,15 WTX-124, an IL-2 prodrug, comprising native human IL-2 linked to an antibody fragment (inactivation domain) and a single-domain antibody targeting human albumin (half-life extension domain), has entered a phase I clinical trial (ClinicalTrials.gov: NCT05479812) by Werewolf.³⁶ However, the introduction of the masking domain could complicate the structure and increase the immunogenicity risk.

Currently, there is limited research on immunocytokine prodrugs. Only Askgene has reported an anti-PD-1/IL-15 prodrug, ASKG915, that utilizes IL-15Rβ to mask the IL-15 activity.³⁷ In this study, we propose a next-generation immunocytokine prodrug strategy with two features: (1) cytokine activity is masked by steric hindrance, and (2) the cytokine would not be confined to the antibody moiety but be released after a tumor-associated proteolysis. With this strategy, we constructed LH05, which has a prolonged plasma half-life and improved safety profile due to the attenuated “cytokine sink” effect in circulation. As expected, LH05 exhibits potent antitumor efficacy in a proteolytic cleavage-dependent manner, with significantly lower systemic toxicity than wild-type anti-PD-L1/IL-15. Our results highlight distinct advantages of our design: it avoids the addition of extra proteins or peptides, minimizing structural complexity and potential immunogenicity, and after cleavage, the released ILR can elicit broad-spectrum immune responses that enhance antitumor efficacy.

Mechanically, the excellent efficacy of LH05 can be attributed to both the PD-L1 trans-delivery of ILR to the TME and the release of active ILR after cleavage. Previously reported PD-1 cis-targeted IL-2/IL-15R agonists, including PD-1-laIL-2, αPD1-IL15m, and αPD1-IL15-R, can selectively deliver IL-2 or IL-15 to PD-1⁺CD8⁺ TILs and bypass NK cells.^38,39,40 All of these immunocytokines showed an antitumor efficacy that was dependent on intra-tumoral CD8⁺ T cells but not on NK or lymph node T cells. However, in our study, due to the released ILR, LH05 stimulated both the adaptive and innate immune cells, illustrating a more comprehensive antitumor role than the PD-1 cis-delivered immunocytokines.

In this research, we employed three “cold” tumor models: RM-1, U251, and B16-F10. They are characterized by an immune-hostile and immunosuppressive TME that abrogates T cell infiltration and activation. LH05 showed significant antitumor effects across all models, underscoring its therapeutic potential for cold tumors. To further explore the mechanisms behind LH05’s effectiveness, we focused on the TME. Specifically, CXCL9 and CXCL10 are pivotal for recruiting effector T cells from the circulation into the tumor and establishing a “hot” TME.²⁵ Our findings revealed that LH05 treatment significantly upregulated Cxcl9 and Cxcl10 mRNA levels without affecting Treg-attracting chemokines Ccl17 and Ccl22 (Figure S7D).⁴¹ Additionally, LH05 treatment significantly elevated the expression levels of Ifng, Tnf, and Tbet, suggesting a Th1-biased TME. Importantly, the improved safety profile of conditionally activated LH05 enables administration of higher doses, which is beneficial for enhancing antitumor effects.

In summary, LH05 represents a class of next-generation immunocytokines that distinguishes itself from previously reported molecules, including immunocytokines and conditionally activated cytokines. In preclinical models, LH05 demonstrated a favorable safety profile and superior antitumor efficacy. It holds great potential as a promising candidate for further clinical investigation in patients with ICI resistance or cold tumors.

Limitations of the study

In this study, all the molecules we developed are humanized. Upon administration of these molecules to immunocompetent mice, we observed the generation of anti-drug antibodies (ADAs), which could potentially hamper their antitumor activities. Notably, ADA levels were comparable between anti-PD-L1 and IgG groups, while LH01, LH03, and LH05 exhibited increased immunogenicity, possibly attributed to the antibody-cytokine fusion or the immunoadjuvant effects of IL-15. Nevertheless, the similar ADA levels observed for LH01, LH03, and LH05 suggest their similar immunogenicity, thereby highlighting that contrasting the antitumor effects of LH05 and LH01 could yield valuable insights (Figure S7E).

Other limitations include that it is necessary to assess the antitumor effects of LH05 in a broader range of cold tumor models. Moreover, individual differences in tumor-associated protease expression levels, including uPA, MMPs, or matriptase, could add uncertainty to the clinical application of such products, which should also be considered in all prodrug strategies.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
APC/Cy7 anti-mouse CD45.2 (clone:104)	BioLegend	Cat#: 109824; RRID: AB_830789
FITC anti-mouse CD3ε (clone:500A2)	BioLegend	Cat#: 152304; RRID: AB_2632667
PE anti-mouse CD4 (clone:557308)	BD Biosciences	Cat#: 557308; RRID: AB_396634
BV510 anti-mouse CD8a (clone: 53–6.7)	BD Biosciences	Cat#: 563068; RRID: AB_2687548
APC anti-mouse CD8a (clone: 53–6.7)	BD Biosciences	Cat#: 553035; RRID: AB_398527
FITC anti-mouse CD8a (clone: 53–6.7)	BioLegend	Cat#: 100705; RRID: AB_312744
BV421 anti-mouse Nkp46 (clone: 29A1.4)	BioLegend	Cat#: 137611; RRID: AB_10915472
Alexa Flour 647 anti-mouse Nkp46 (clone: 29A1.4)	BD Biosciences	Cat#: 560755; RRID: AB_1727464
BV786 anti-mouse IFN-γ (clone: XMG1.2)	BD Biosciences	Cat#:563773; RRID: AB_2738419
PE/Cy7 anti-human/mouse Granzyme B (clone: QA16A02)	BioLegend	Cat#: 372213; RRID: AB_2728380
PE anti-mouse/human CD44 (clone: IM7)	BioLegend	Cat#: 103007; RRID: AB_312958
APC anti-mouse CD62L (clone: MEL-14)	BD Biosciences	Cat#: 553152; RRID: AB_398533
FITC Goat anti-human IgG-Fc secondary antibody	Invitrogen	Cat#: A18818; RRID: AB_2535595
Fc block-anti-mouse CD16/32	BioLegend	Cat#: 101302; RRID: AB_312801
InVivoMAb anti-mouse CD8a (clone 2.43)	BioXCell	Cat#: BE0061; RRID: AB_1125541
InVivoMAb anti-mouse NK1.1 (clone PK136)	BioXCell	Cat#: BE0036; RRID: AB_1107737
Peroxidase AffiniPure Goat Anti-Human IgG (H + L)	Jackson ImmunoResearch	Cat#: 109035003; RRID: AB_2337577
Rabbit anti-mouse CD8 alpha	Abcam	Cat#: ab217344
Rabbit anti-mouse Foxp3	CST	Cat#: 12653T
Goat anti-mouse IgG-Fc Secondary antibody (HRP)	SinoBiological	Cat: SSA006
Bacterial and virus strains
V200	Provided by Converd Inc.	N/A
Biological samples
Tumor and adjacent normal tissues	Shanghai Chest Hospital; Hangzhou First People’s Hospital; Huashan Hospital	N/A
Chemicals, peptides, and recombinant proteins
Collagenase IV	Yeasen	Cat#: 40510ES76
Hyaluronidase	Yeasen	Cat#: 20426ES80
Polyethylenimine	Polysciences	Cat#: 23966
FTY720	Sigma-Aldrich	Cat#: 162359-56-0
Mouse PD-L1	Novoprotein	Cat#: CJ88
Human PD-L1	Novoprotein	Cat#: C315
Urokinase/uPA	SinoBiological	Cat#: 10815-H08H
Human MMP-2	SinoBiological	Cat#: 10082-HNAH
CD122 (IL-2/15Rβ)	SinoBiological	Cat#: 10696-H08B
TMB single component substrate solution	Solarbio	Cat#: PR1200
Lymphocyte separation medium	DAKEWE	Cat#: 7211011
SulfoCy5.5 SE	Fluorescence	Cat#: 1057-1mg
PrimeScript RT Master Mix	Takara	Cat#: RR036A
Hieff qPCR SYBR Green Master Mix	Yeasen	Cat#: 11199ES08
Critical commercial assays
Zombie Red Fixable Viability Kit	BioLegend	Cat#: 423109
Transcription Factor Buffer Set	BD Biosciences	Cat#:562574; RRID: AB_2869424
Mouse IL-6 ELISA Kit	Multi Sciences	Cat#: EK206/3
Mouse IFN-γ ELISA Kit	Multi Sciences	Cat#: EK280/3
Cell Counting Kit-8	Donjindo	Cat#: CK04
Three color mIHC Fluorescence kit	Recordbio	Cat#: RC0086-23
Deposited data
The differential expression between tumor and adjacent normal tissues for uPA	Dana Farber Cancer Institute- X Shirley Liu Lab42	TIMER2.0: http://timer.cistrome.org/
The correlation of IL-15 expression with immune infiltration level in diverse cancer types	Dana Farber Cancer Institute- X Shirley Liu Lab42	TIMER2.0: http://timer.cistrome.org/
RNA-sequencing of RM-1 mouse prostate cancer	This paper	PRJNA1006452
Mendeley dataset	This paper	https://doi.org/10.17632/7k6wwcy568.1
Experimental models: Cell lines
HEK293E	Provided by J.W. Zhu	N/A
Mo7e	Procell	CL-0686
RM-1	ATCC	CRL-3310
MC38	Provided by L.K. Gong	N/A
Renca	ATCC	CRL-2947
B16-F10	Provided by L.K. Gong	N/A
U251	Procell	CL-0237
Human PD-L1-CHO-K1	Provided by L.K. Gong	N/A
Mouse PD-L1-CHO-K1	Provided by L.K. Gong	N/A
Experimental models: Organisms/strains
Mouse C57BL/6J	SLAC Animal	N/A
Mouse Balb/c	SLAC Animal	N/A
NCG	GemPharmatech	N/A
Software and algorithms
GraphPad Prism v.8	GraphPad Software	https://www.graphpad.com/
FlowJo v.10	FlowJo	https://www.flowjo.com/
Adobe Illustrator CC	Adobe	https://www.adobe.com/
BioRender	BioRender	https://www.biorender.com/

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Huili Lu (roadeer@sjtu.edu.cn).

Materials availability

All unique/stable reagents generated in this study are available from the corresponding author Huili Lu (roadeer@sjtu.edu.cn) with a completed Materials Transfer Agreement (https://www.gene.com/scientists/mta).

Data and code availability

• •All data generated in this study have been included in the article and the Supplementary Materials. The data used for analyzing the uPA expression level, and IL-15 expression correlating with the immune cell infiltration in tumors were obtained from TIMER (Tumor IMmune Estimation Resource) database, listed in the key resources table. RNA-seq data have been deposited in NCBI’s Sequence Read Archive under the accession code PRJNA1006452. Raw data from Figures 7C and 7E, S2D, and S7A were deposited to Mendeley Data: https://doi.org/10.17632/7k6wwcy568.1.
• •This paper does not report the original code.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Experimental model and study participant details

Cell lines

HEK293E and Mo7e cell lines were kept in our laboratory and cultured as previous described.²⁰ RM-1, MC38, Renca, B16-F10, and U251 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS. CHO-K1 cell line was maintained in Ham’s F-12K containing 10% FBS. All of the cells mentioned above were kept in aseptic conditions and incubated at 37°C with 5% CO₂.

Animal experiments

All animal experiments were approved by the Animal Care and Use Committee of Shanghai Jiao Tong University. Sex-matched Balb/c and C57BL/6 mice aged 6–8 weeks were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. Female NCG mice aged 6–8 weeks were purchased from Jiangsu GemPharmatech LLC. All mice were raised in pathogen-free environments and received humane treatment throughout the experimental period. Human peripheral blood mononuclear cells (PBMCs) were purchased from Shanghai Milestone Biotechnologies. For antitumor studies, 5 × 10⁵ RM-1, MC38, Renca, or B16-F10 tumor cells were subcutaneously implanted into the right flank of male, female, female, and female C57BL/6J mice, respectively. Mice were randomized into four groups based on tumor size, with treatment initiating when tumors reached 50–100 mm³.Tumors were measured every two or three days using a digital caliper, and volumes were calculated as (length×width²)/2. Tumor Growth Inhibition (TGI): TGI (%) = 100 × (1-T/C). T and C were the mean tumor volumes of the treated and control groups, respectively.

Human tumor samples

Colon, lung, and brain tumors as well as their adjacent peritumoral tissues were collected with the approval of the Medical Research Ethics Committee of Hangzhou First People’s Hospital (No. 2022[262]), Shanghai Chest Hospital (No. KS(Y)22140), and Huashan Hospital (No. KY2022-913). All subjects provided broad informed consent for the research use of their biological samples.

Method details

Cloning, expression, and purification

LH01, LH02, and anti-PD-L1 were constructed and produced as previously described.²⁰ For LH03 or LH05 construction, the human IL-15 mutant (IL-15N72D)/IL-15Rα-sushi domain (Ile31 to Val 115) complex (ILR) was fused to the C-terminal of anti-PD-L1 heavy chain via a (GGGGS)₃ (non-substrate) linker or a GSSGGSGGSGGSG-LSGRSDNH-GSSGGSGGSGGSG (substrate) linker, respectively, in the pMF09 vector we reported before.²⁰ The amino acid sequence of LH05 is shown in Figure S7F. LH06 was constructed by substituting the urokinase substrate linker (LSGRSDNH) of LH05 with an MMP-2/14 substrate linker (SGQLLGFLTA). The plasmids were mixed with 25 kDa linear polyethylenimine and transiently transfected in HEK293E cells. All fusion proteins were purified by using a protein A affinity column (GE Healthcare) and analyzed on SDS-PAGE in the reducing condition.

ELISA to evaluate the affinity of anti-PD-L1/IL-15 for PD-L1

ELISAs were conducted following standard procedures. Briefly, 96-well ELISA plates (Corning) were coated with 1.0 μg/mL of recombinant human or mouse PD-L1 overnight at 4°C, followed by washing four times with PBST (PBS, 0.05% Tween 20) and blocked with 5% bovine serum albumin for 2 h at room temperature. After washing the plates, serial dilutions (1:3) of LH01, LH05, or anti-PD-L1 antibody were added in duplicate to the plates and incubated at room temperature for 2 h. The plates were washed four times and then incubated with Peroxidase AffiniPure Goat Anti-Human IgG (H + L) (1:10,000 dilution) at room temperature for 1 h. After washing, the plates were incubated with TMB single component substrate solution in the dark for 3–5 min. The reaction was stopped with 2 M sulfuric acid, and absorbance was read at 450 nm with a reference at 630 nm (Teacan, Infinite 200 PRO).

Flow cytometry analysis for affinity to the cell surface PD-L1

The CHO-K1 cells stably expressing PD-L1 were seeded into 48-well plates at a density of 1 × 10⁵ cells in a volume of 200 μL PBS per well. Serial dilutions (1:3) of LH01, LH05, or anti-PD-L1 antibody were added to the plates and incubated at 4°C for 1 h. After washing cells twice, The Goat anti-human IgG-Fc (FITC) antibody were added in to the plates and incubated at 4°C for 30 min. After washing cells twice, the mean fluorescence intensity (MFI) was detected using ACEA Novocyte (Agilent, Technologies, USA).

Pharmacokinetics studies of immunocytokines

Plasma samples were collected 5 min, 0.5 h, 1 h, 4 h, 8 h, 24 h, and 48 h after intravenous injection with 1 mg/kg LH01 or LH05. A 96-well ELISA plate, previously coated overnight at 4°C with 1.0 μg/mL of recombinant human PD-L1, was incubated for 2 h with plasma samples from mice. The following experimental procedure for ELISA was the same as described above.

Quantitative biodistribution studies of immunocytokine

Heart, liver, spleen, lung, kidney, and tumor tissues of RM-1 tumor-bearing mice were collected and minced. About 100–150 mg of tissues were weighed and homogenized in 10% PBS before being centrifuged to obtain supernatant. We employed two ELISA assays to quantify the amount of LH01 or LH05, either cleaved or un-cleaved, for each homogenate. The above ELISA assay developed to evaluate anti-PD-L1/IL-15 affinity for PD-L1 was used to detect the total amount of LH01 or LH05 in both cleaved and un-cleaved forms. Since both LH01 and LH05 can bind IL-15Rβ, an ELISA assay was developed to detect LH01 or un-cleaved LH05 with IL-15Rβ coated on the plate and Peroxidase AffiniPure Goat Anti-Human IgG (H + L) (1:10,000 dilution) as the detection antibody. In detail, 2.0 μg/mL of recombinant human IL-15Rβ was coated overnight at 4°C, and then incubated with supernatant from tissue homogenate for 2 h. The following experimental procedure for ELISA was the same as described above.

In vitro cleavage of immunocytokines with uPA or MMP-2

In vitro cleavage was performed by incubating 10 μg LH03 or LH05 with 0.25 μg uPA in phosphate buffer saline in a total reaction volume of 10 μL at 20°C for 12 h. Human MMP-2 (1 μg) was firstly activated via incubation with 0.2 mM APMA at 37°C for 1 h in TCNB activation buffer (50 mM Tris, 10 mM CaCl₂, 150 mM NaCl, 0.05% (w/v) Brij 35, pH7.5). 25 μg LH06 was incubated with the above reaction buffer at 37°C for 12 h.

LH05 stability in human serum and cleavage of LH05 by human tumors

Human serum was purchased from Shanghai Xinfan Biotechnology Co., Ltd. 1 μL LH05 (2 μg) was added to 9 μL human serum, then incubated at 37°C for 24 or 72 h. Homogenization of human tumors was performed using FastPrep tissue homogenizer (MP Bio, USA). Supernatant was collected by centrifugation at 10000g for 15 min. For the cleavage experiments, 9 μL tissue lysate was incubated with 0.2 μg LH05 (0.2 mg/mL) at 37°C for 24 h.

The un-cleaved and total LH05 was quantified by the ELISA described above. To confirm the feasibility of the above ELISA methods, LH03 containing non-cleavable linker and uPA-activated LH05 were included as negative and positive control, respectively.

Mo7e cell proliferation assay

Mo7e cells were washed with human GM-CSF free medium (RPMI1640 + 10% FBS) before being seeded into 96-well plates at a density of 2×10⁴ cells in a volume of 50 μL per well. After 4 h’ starvation, serial dilutions (1:3) of LH01, LH03, or LH05 (treated with or without uPA) were added to the plate in sextuplicate at 50μL per well to achieve a final density of 2×10⁴ cells/100 μL/well. Cell viability was measured using the Cell Counting Kit-8 kit after 96 h of incubation at 37°C with 5% CO₂. The absorbance was read at 450 nm with an ELISA reader (Teacan, Infinite 200 PRO), and the final OD450 value of the sample wells have subtracted the blank reading.

Flow cytometry analysis

About 150 mg of tumor tissues was cut into small pieces and re-suspended in digestion buffer [RPMI1640 medium containing collagenase IV (2 mg/mL) and hyaluronidase (1.2 mg/mL)]. Tumors were digested for 60 min at 37°C and then filtered through a 200-mesh nylon net to obtain the cell suspension. The cells were washed by RPMI 1640 and filtrated through a 200-mesh nylon net again, and then resuspended in FACS buffer (PBS +2% FBS) to obtain pre-treated single cell suspension. Splenic lymphocytes were isolated from the spleens with lymphocyte separation medium after the spleens were gently ground.

Cell samples were blocked with anti-mouse CD16/CD32 mAb at 4°C for 30 min before being incubated with surface marker antibodies at 4°C for 30 min. The Zombie Red Fixable Viability Kit was used to exclude dead cells. For the detection of intracellular IFN-γ and granzyme B, cell samples were further fixed and permeabilized by Fixation/Permeabilization Kit. Flow cytometry was performed using CytoFLEX cytometer (Beckman Coulter, USA) or ACEA Novocyte (Agilent, Technologies, USA) and analyzed by FlowJo 10 (TreeStar, USA) or NovoExpress (Agilent, Technologies, USA).

Detection of plasma ALT, AST, CREA, UREA, IFN-γ, and IL-6

Plasma levels of ALT, AST, CREA, and UREA were measured with a Roche biochemical analyzer (Roche, Switzerland). Plasma levels of IFN-γ and IL-6 was determined by mouse IFN-γ and IL-6 ELISA Kit according to the manufacturer’s procedures, respectively.

Fluorescence imaging

LH05 was labeled with Cy5.5 and excess dye unbound LH05 was removed via ultrafiltration. Fluorescently labeled LH05 (1 mg/kg) was intravenously injected into tumor-bearing mice. Fluorescence was measured with PE IVIS Spectrum at different time points.

Depletion of immune cells in mice

To deplete the individual immune cell types, RM-1 tumor-bearing mice were intravenously injected with IgG control (10 mg/kg) or LH05 (10 mg/kg) on days 9, 12, and 15. For cell depletion, mice were intraperitoneally given 200 μg of anti-NK1.1 antibody or 200 μg anti-CD8α antibody on days 7, 9, and 13. Tumor growth curves were plotted. To study the effect of lymphocytes egressing from lymph nodes, RM-1 tumor-bearing mice were administered with IgG (10 mg/kg, i.v.), LH05 (10 mg/kg, i.v.), or FTY720 with or without LH05. FTY720 (25 μg) was intraperitoneally administered every other day beginning 8 days after tumor cell inoculation.

RNA isolation and quantitative RT-PCR analysis

Total RNAs were extracted from tissues using the Ultrapure RNA Kit (Cwbio, China), and cDNA was synthesized using a PrimeScript RT Master Mix. Real-time qRT-PCR was performed on an Applied Biosystems 7500 Fast Real-Time PCR System (Thermo Fisher Scientific, USA) using Hieff qPCR SYBR Green Master Mix. The primer sequences are listed in Table S1. All results were normalized to GAPDH expression and calculated using the 2^-(ΔΔCt) method.

RNA sequencing

Total RNA was extracted from RM-1 tumor tissues. cDNA library construction, sequencing, and data analysis were performed by the Shanghai Majorbio Bio-Pharm Technology Co., Ltd. using the Majorbio cloud platform. High-quality reads were aligned to the mouse reference genome (GRCm39) using Bowtie2. We normalized the expression level of each gene to the fragment of the exon model per million mapped reads (FPKM) based on the expectation maximization method. NOISeq method was used to screen out differentially expressed genes (DEG). Statistical significance for DEG was fold change >1.5 with p values <0.05. We accessed the Gene Set Enrichment Analysis website to obtain gene sets associated with immunity. Immune signature scores are defined as the mean log₂ (fold change) among all genes in each gene signature. The FPKM of these genes were logarithmically (fold-change) converted, and heat maps were generated by GraphPad Prism V.8 software.

Histopathological and immunohistochemistry analysis

The tumor tissues, livers, and kidneys were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned (4 μm). The livers and kidneys sections were prepared and stained with hematoxylin-eosin (H&E). The U251 tumor sections were incubated with anti-human Ki67 rabbit antibody (Servicebio, China) at 4°C overnight. Next, the sections were incubated with the HRP-conjugated goat anti-rabbit secondary antibody (Servicebio, China). Finally, the sections were stained with a DAB detection kit (Dako, Copenhagen, Denmark). Co-staining of CD8 and Foxp3 of the B16-F10 tumor sections was performed using a Three color mIHC Fluorescence kit based on the tyramide signal amplification (TSA) technology according to the manufacture’s instruction. Sections were imaged on an Olympus BX53 Microscope and a Pannoramic DESK slide scanner.

ADA detection

Human IgG, anti-PD-L1, LH01, LH03, and LH05 were individually diluted to 2 μg/mL, coating ELISA plates, which were subsequently incubated overnight at 4°C. Plasma samples from various treatment groups of RM-1 tumor-bearing mice (described as Figure 4A), underwent gradient dilution and were added to the wells. The plates were then incubated at 37°C for 2 h. The plates were washed and then incubated with goat anti-mouse IgG-Fc secondary antibody at 37°C for 1 h. The following experimental procedure for ELISA was the same as described above.

Quantification and statistical analysis

Prism 8.0 software (GraphPad, USA) was used for statistical analysis. The unpaired two-tailed Student’s t test and one-way ANOVA were used to determine the statistical significance of differences between experimental groups (∗: p < 0.05, ∗∗: p < 0.01, ∗∗∗: p < 0.001). The log rank (Mantel-Cox) test was used to assess survival.

Published: May 1, 2024