Resistance to anti-PD-1/anti-PD-L1: galectin-3 inhibition with GB1211 reverses galectin-3-induced blockade of pembrolizumab and atezolizumab binding to PD-1/PD-L1

Abstract

Background: Galectin-3 (Gal-3) is a β-galactoside-binding lectin that is highly expressed within the tumor microenvironment of aggressive cancers and has been suggested to predict a poor response to immune checkpoint therapy with the anti-PD-1 monoclonal antibody pembrolizumab. We aimed to assess if the effect of Gal-3 was a result of direct interaction with the immune checkpoint receptor.

Methods: The ability of Gal-3 to interact with the PD-1/PD-L1 complex in the absence and presence of blocking antibodies was assessed in in vitro biochemical and cellular assays as well as in an in vivo syngeneic mouse cancer model.

Results: Gal-3 reduced the binding of the checkpoint inhibitors pembrolizumab (anti-PD-1) and atezolizumab (anti-PD-L1), by potentiating the interaction between the PD-1/PD-L1 complex. In the presence of a highly selective Gal-3 small molecule inhibitor (GB1211) the binding of the anti-PD-1/anti-PD-L1 therapeutics was restored to control levels. This was observed in both a surface plasmon resonance assay measuring protein-protein interactions and via flow cytometry. Combination therapy with GB1211 and an anti-PD-L1 blocking antibody reduced tumor growth in an in vivo syngeneic model and increased the percentage of tumor infiltrating T lymphocytes.

Conclusion: Our study suggests that Gal-3 can potentiate the PD-1/PD-L1 immune axis and potentially contribute to the immunosuppressive signalling mechanisms within the tumor microenvironment. In addition, Gal-3 prevents atezolizumab and pembrolizumab target engagement with their respective immune checkpoint receptors. Reversal of this effect with the clinical candidate GB1211 offers a potential enhancing combination therapeutic with anti-PD-1 and -PD-L1 blocking antibodies.

Keywords: PD-1 - PDL-1 axis; atezolizumab; binding; computational chemistry; galectin-3 (Gal3); immuno-oncology; pembrolizumab.

Figure 1

(A) Connolly surfaces of PD-1 (purple) binding PD-L1 (green) carrying four N-glycans (sticks) each at the known N-glycan sites for PD-1 (N49, N58, N74 and N116. (18) and PD-L1 (N35, N192, N200, and N219 (19). Gal-3 is shown as a Connolly surface (salmon) bound to a terminal D-galactoside residues of the PD-1 N58 and PD-L1 N192 glycans or to the inhibitor GB1211 (space-filling atoms with green carbons). The glycans were built manually onto the X-ray structures of full-length PD-L1 (4Z18) and on PD-1 in the PD-1/PD-L1 complex (4ZQK), whereafter the structures were superposed to show a possible binding of full-length PDL1 to PD1. The X-ray structure of Gal-3 in complex with GB1211 (7ZQX) was used to visualise inhibited Gal-3 and to manually superpose Gal-3 to the glycoprotein N-glycans. (B) The overlay of the binding interface between pembrolizumab (yellow) to PD-1 and atezolizumab (cyan) to PD-L1 and are shown as transparent surfaces for reference. The structures and interaction positions of pembrolizumab and atezolizumab were taken from their X-ray structures with PD-1 (5GGS) and PD-L1 (5XXY), respectively. Note, glycan structures are generic branched complex N-glycans and may not represent the most prevalent structures experimentally determined. The glycans were energy minimised to represent possible minimum conformations, but not necessarily the global minimum conformations.

Figure 2

(A) Profile of GB1211 binding to Gal-3 as measured by SPR. (B) Gal-3 potentiates the binding of PD-L1 to PD-1 with a 3-fold increase in affinity. (Bii) GB1211 blocks the binding of Gal-3 (1 µM) to immobilised PD-1 and PD-L1 with an IC₅₀ of 0.57 µM and 0.50 µM, respectively. (C) Gal-3 potentiates the binding of PD-L1 to immobilised PD-1 that reduces the ability of atezolizumab to bind the PD-1/PD-L1 complex. GB1211 reverses the Gal-3 blockade restoring the binding of atezolizumab to PD-L1. (D) Gal-3 reduces the binding of pembrolizumab to immobilised PD-1 that is reversed by GB1211. (E) Atezolizumab binding to PD-L1 on the surface of Raji-hPD-L1 cells measured by flow cytometry. Gal-3 reduces the ability of atezolizumab to bind to the Raji-hPD-L1 cells, with the Gal-3 inhibitor GB1211 able to reverse this blockade. (F) Pembrolizumab binding to PD-1 on the surface of Jurkat-hPD-1 cells measured by flow cytometry. Gal-3 reduces the ability of pembrolizumab to bind to the Jurkat-hPD-1 cells, with the Gal-3 inhibitor GB1211 able to reverse this blockade. Data shown is mean ± SEM (n=3) with arrows showing the restorative effect of GB1211 on ICI. Atez, atezolizumab.

Figure 3

LLC1 subcutaneous tumor (A) volumes and (B) weights. Results represent the mean ± SEM (vehicle n=28 tumors from n=16 mice, GB1211 (10 mg/kg) n=12 tumors from n=8 mice, anti-PD-L1 (200 µg) n=28 tumors from n=16 mice, GB1211 (10 mg/kg) + anti-PD-L1 (200 µg) n=14 tumors from n=8 mice). Two-way ANOVA with Tukey’s post-hoc test was used to test for differences in tumor volume with only significant changes shown. One-way ANOVA and Dunnett’s post-test were used to compare tumor weight with only significant changes shown.

Figure 4

post-hocLLC1 subcutaneous tumor digests were analysed for tumor infiltrating CD8+ T cell populations by flow cytometry. Frequency of (A) CD3+, (C) CD4+ and (E) CD8+ cells expressed as a % of the total immune (CD45+) population and Ki-67 expression as mean fluorescence intensity (MFI) in (B) CD3+, (D) CD4+ and (F) CD8+ T cells. Results are expressed as mean ± SEM (vehicle n=12, GB1211 n=10, anti-PD-L1 n=10, GB1211+anti-PD-L1 n=12). Analysed via one-way ANOVA with Dunnett’s post-test with only significant changes shown except for drug combination p-value shown in panel 4 (E).

Figure 5

Schematic depicting the relationship between Gal-3 and the checkpoint receptor PD-1, with the proposed negative effect of high Gal-3 on pembrolizumab binding leading to blockade of the ICI’s T cell activation and tumor suppression effects. The proposed reversal of Gal-3’s inhibitory effect on pembrolizumab’s target engagement is also shown with the Gal-3 inhibitor GB1211. The same effect on atezolizumab and PD-L1 is also proposed from the data in this study.