Dear Editor,

Inhibiting PD-1/PD-L1 interaction is a highly promising therapeutic modality.1 However, due to the low overall response rate in patients, researchers have attempted to combine PD-L1 inhibitors with other antitumor agents for cancer therapy. Studies have shown that combination immunotherapy of PD-L1 antibodies with CXCL12 inhibitors exhibited synergistic and better antitumor efficacy than monotherapy, indicating the potential clinical utility of targeting both PD-L1 and CXCL12 as dual immunotherapy to treat cancer.2,3 However, there are several disadvantages for combination therapy, including unpredictable PK/PD and overlapping toxicities. A potential alternative to combination therapy would be to use a single molecule with dual or multi-targeting capability, as the PK/PD of a single molecule is easily predictable. For example, dual-targeting bispecific antibodies (bsAbs) have gained significant attention in the field of anticancer drug discovery in recent years. Many PD-1/PD-L1-based bsAbs (e.g., anti-PD-L1/TGF-β, anti-PD-1/CTLA-4, and anti-PD-1/LAG-3) have entered clinical trials as dual immunotherapy for treating cancer. However, bsAbs-based dual immunotherapies also suffer from the common drawbacks (e.g., immunogenicity, poor pharmacokinetics) as antibodies, thus it would be of high significance to develop small molecule PD-L1 inhibitor-based dual immunotherapy, as small molecules may overcome the above drawbacks of antibodies.

We have previously reported PD-L1-targeting bifunctional molecules as potential anticancer agents.4 To continue our interest in this area, we designed a set of compounds targeting both PD-L1 and CXCL12 simultaneously as potential dual immunotherapy based on the hypothesis that PD-L1 and CXCL12 are two critical biomacromolecules controlling the immunosuppressive tumor microenvironment. Firstly, we analyzed the pharmacophores of PD-L1 inhibitors and CXCL12 inhibitors (Fig. 1a). The tail group of PD-L1 inhibitors and the hydroxyl moiety of CXCL12 inhibitors were exposed to solvent, making them suitable sites for conjugating the two inhibitors via a linker. Thus, twenty-one bifunctional molecules were designed, synthesized (Supplementary Scheme S1), and bioevaluated (Supplementary Table S1).

Fig. 1
figure 1

a Design of bifunctional molecules targeting PD-L1/CXCL12 based on the pharmacophores of CXCL12 inhibitors and PD-L1 inhibitors. b The binding affinity of CP21 to hPD-L1 as measured by SPR. c CD spectra of hPD-L1 in complex with or without CP21. d Docking of CP21 to hPD-L1 (PDB code: 6R3K). e ITC study for CP21 in complex with hPD-L1. f Intracellular Ca2+ release was measured by spectrofluorometry in THP-1 cells. g Effects of CP21 and BMS-1233 (positive control) on HepG2 cell mortality in Jurkat T cell&HepG2 co-culture system (n = 5, **P < 0.05). h Therapeutic effects of CP21 (h, 125 mg/kg; l, 75 mg/kg), BMS-1233, and CXCL12i (inhibitor) against B16-F10 melanoma tumors in mice (n = 5). i Average tumor weights (**P < 0.05, n = 5). Percentages of CD3+ cells (j) and CD3+CD8+ cells (k), and the ratio of CD8+/CD4+ cells (l) in tumor tissues (n = 3, **P < 0.05)

Full size image

Among them, CP21 showed the strongest PD-L1-inhibitory effects with IC50 of 78.6 nM (HTRF assay). Furthermore, CP21 displayed similar binding affinity (SPR assay) to both h(human)PD-L1 (KD = 66.9 nM, Fig. 1b) and m(mouse)PD-L1 (KD = 70.1 nM) (supplementary Fig. S1a). In addition, CD (Circular dichroism) assay revealed that when hPD-L1 or mPD-L1 was mixed with CP21, the conformation of their secondary structures changed similarly, as compared to the vehicle which contains only hPD-L1 (Fig. 1c) or mPD-L1 (Supplementary Fig. S1b). Moreover, the microscale thermophoresis (MST) assay confirmed that CP21 could bind to mPD-L1 with a KD of 654.1 nM (Supplementary Fig. S1c, d). Next, the binding affinity of CP21 to hCXCL12 and mCXCL12 was also determined by SPR and CD. CP21 bound to hCXCL12 (KD = 160 nM, SPR) and mCXCL12 (KD = 76.5 nM, SPR) in a dose-dependent manner (supplementary Fig. S1e, f). Furthermore, CD spectroscopy revealed an altered conformation in the secondary structures of hCXCL12 and mCXCL12 upon the addition of CP21 (Supplementary Fig. S1g, h). The above results suggest that CP21 could cross-bind to hPD-L1/hCXCL12 or mPD-L1/mCXCL12 proteins with high affinity.

To further investigate the possible binding modes of these dual-acting compounds, we conducted molecular docking analysis for CP21 with PD-L1 and CXCL12 proteins. As expected, CP21 fitted nicely in the inner cavity of the PD-L1 dimer (Fig. 1d). Calorimetric data (Fig. 1e and supplementary Fig. S2a) from the ITC assay suggested that the excellent binding affinity of CP21 to PD-L1 was primarily due to hydrophobic interactions, consistent with molecular modeling studies. Additional molecular docking study also suggested that CP21 could cross-bind with h/m PD-L1 through different modes (supplementary Fig. S2b–e). Furthermore, CP21 bound well to the hydrophobic cleft formed by CXCL12 (supplementary Fig. S2f), which was further confirmed by ITC analysis (supplementary Fig. S2g, h). The well-defined molecular modeling studies of CP21 in PD-L1 and CXCL12 may explain the high binding affinity of CP21 to its target proteins.

To validate CP21 as a dual inhibitor of CXCL12 and PD-L1 in vitro, we first evaluated the effects of CP21 on CXCL12-mediated Ca2+ cellular responses in THP-1 cells using a calcium flux assay.5 As shown in Fig. 1f, CP21 dose-dependently inhibited CXCL12-induced calcium responses as measured by spectrofluorometry. Flow cytometry analysis further confirmed that CP21 abolished the response of CXCL12-mediated Ca2+ flux (supplementary Fig. S3a). Moreover, CP21 dose-dependently inhibited CXCL12-mediated cell migration (supplementary Fig. S3b).

To investigate the immunomodulatory effects of CP21, we assessed the mortality rate of HepG2 cells utilizing a Jurkat T&HepG2 cell co-culture model. As illustrated in Fig. 1g, CP21 dose-dependently promoted the death of HepG2 cells with effects similar to that of BMS-1233 (supplementary Fig. S4a, b). The immunomodulatory effects of CP21 were further confirmed by the MDB-MB-231& PBMC, Hep3B&PMBC, and Jurkat T&B16-F10 co-culture models (supplementary Fig. S4c–g). Collectively, these results suggest that CP21 could target both CXCL12 and PD-L1 in vitro.

The pharmacokinetic properties of CP21 were evaluated in male SD rats with CP21 administered intravenously (i.v.) and orally (p.o.). As detailed in Table S2, oral gavage of CP21 (18 mg/kg) exhibited low but acceptable plasma exposure (AUC(0–t)= 127.5 ± 4.8 ng/mL h). While intravenous administration of CP21 (1 mg/kg) displayed more favorable pharmacokinetic properties, such as higher exposure (AUC = 2032.5 ± 11.44.8 ng/mL h), and lower clearance rate (0.5 ± 0.1 L/h/kg).

Having identified the excellent PD-L1/CXCL12 inhibitory potency in vitro for the bifunctional molecule CP21, we next evaluated the anticancer activity of CP21 in vivo using a mouse B16-F10 model. As presented in Fig. 1h, i, CP21 dose-dependently decreased tumor weights and volumes. Oral administration of CP21 at 125 mg/kg exhibited a tumor growth inhibition (TGI) of 58.5%, better than either of the monotherapies. In addition, the body weights of mice did not change markedly during the treatment (supplementary Fig. S5). In addition, CP21 exhibited similarly high in vivo antitumor efficacy in a CT-26 colon syngeneic tumor model (supplementary Fig. S6).

To examine the immunomodulatory effects of CP21, a flow cytometry assay was carried out to analyze the TILs (tumor-infiltrating lymphocytes). As shown in supplementary Fig. S7 (representative examples), the percentage of TILs (CD3+) in the CP21 treatment group was dose-dependently higher than that in either of the monotherapy groups. In addition, data from multiple repeated experiments/animals also showed that the percentages of CD3+, CD3+CD8+, and the ratio of CD8+/CD4+ T cells were dose-dependently/significantly increased in the CP21 treatment groups, higher than either of the monotherapy groups (Fig. 1j–l). The above results indicate that CP21 could activate the immune microenvironments in tumors.

In addition, serum biochemistry showed that CP21 did not induce hepatotoxicity and nephrotoxicity in mice (Supplementary Fig. S8a). Importantly, CP21 treatment did not cause cardiotoxicity based on the creatine kinase isoenzyme assay (Supplementary Fig. S8b). Furthermore, H&E staining revealed that there was no apparent morphological aberration in the CP21 treatment group (Supplementary Fig. S8c), which was in agreement with the results of serum biochemistry.

Collectively, CP21 is a bifunctional small molecule targeting PD-L1 and CXCL12 with high affinities. It also demonstrated high in vivo antitumor efficacy by activating the immune system. Hence, CP21 represents a first-in-class dual PD-L1/CXCL12 inhibitor deserving further investigation as a potential dual immunotherapy agent for cancer treatment.