A "Function-First" Approach to Identify Regulatory T cell-Targeting Antibodies for Immunotherapy

Abstract

Despite advances in cancer immunotherapy, treatment response is still highly variable. One contributing factor is the tumor microenvironment and specifically the presence of suppressive immune cells such as regulatory T (Treg) cells. Being able to target these specifically, while leaving effector T-cell populations untouched, is an attractive strategy that may overcome some of these issues, improving responses. To generate antibodies specific for tumor-associated Tregs, lymphocytes were isolated from tumor-bearing mice and panned against the n-CoDeR phage antibody library. Using the target-agnostic F.I.R.S.T. discovery platform, they were evaluated in ex vivo and in vivo models to determine tissue and cell selectivity and specificity and ability to deplete Tregs and elicit tumor control in subcutaneous tumor models. A total of 24 antibodies were identified and explored, representing a range of specificities from pan-T cell to Treg and tumor Treg specific. Relative expression/binding of these mAbs on tumor Tregs was not a predictor of subsequent Treg deletion efficacy or tumor control, whereas tumor Treg selectivity was. One mAb in particular demonstrated tumor-specific depletion of Tregs, leaving those in the spleen and blood untouched. This Fc:FcγR-mediated tumor-specific Treg depletion was important for antitumor effects. Target deconvolution showed that this mAb binds a distinct epitope within ICAM-1, which is hypothesized to mediate its selectivity toward tumor Tregs. These data validate the target-agnostic discovery approach as a viable means to identify new therapeutic antibodies.

Figure 1.

Generation of tumor Treg–specific mAbs. A, Schematic of phage panning strategy employed to isolate Treg-associated scFv. B, A total of 320 scFv phages from the Treg phage pool were screened for binding to primary tissue and cell lines by flow cytometry. The geometric MFI (gMFI) of binding was then plotted and clustered to reveal a grouping of scFv, which demonstrated similar Treg-enriched specificity for further evaluation. C, scFv from the Treg cluster (outlined in the box labeled Treg binders) in (B) were produced as bivalent full-length IgG and specificity toward T cells confirmed by flow cytometry on immune cells isolated from CT26 tumors and spleen. Shown are representative examples demonstrating tumor T cell–enriched, tumor CD4 binders, and tumor Treg–specific binding. D, The mean mean fluorescence intensity values from (C) were plotted for all mAbs screened in this manner and are presented as a heatmap to show relative binding levels to different T-cells populations from different locations. Iso, isotype; MFI, mean fluorescence intensity; Spl, spleen; Tum, tumor.

Figure 2.

Tumor-specific depletion of Tregs confers therapeutic responses in the CT26 model. A, CT26 tumor–bearing mice were treated with three doses of mAb over 7 days, before blood, spleen, td-LN, and tumor were collected 3 days after the last dose. The tissue was processed into a single-cell suspension and the proportion of immune cell populations was assessed by flow cytometry. The fold change in the proportion of Tregs (based on percentage of total CD45⁺) when compared with isotype-treated mice was plotted, for which a decrease below 1 indicated depletion and an increase indicated expansion of the Treg population. The number of mice for each group is provided on the x-axis. B, The CD8:Treg ratio for the same individuals in (A) was also plotted. Statistically significant changes across the tissues for each individual mAb were determined by one-way ANOVA. C, The mean fold change in Treg population was plotted for each mAb across each tissue and arranged based on the Euclidean clustering in R to determine similar activities between mAbs. This revealed four distinct groups (emphasized by bold border). Where a statistically significant change from the isotype control was found, the corresponding cell within the heatmap was marked based on the P value obtained. D and E, The mean fluorescence intensity of mAb binding (from Fig. 1D) was plotted against the average Treg fold change (from A) for the spleen (D) and tumor (E). Linear regression was assessed in Prism and the R² value for each tissue provided in the plot. F, CT26 tumor–bearing mice were treated with three doses of mAb and tumor growth tracked for survival. The plots are grouped based on the clusters of Treg-depleting activity identified in C. Survival data are compiled across multiple independent experiments—each one controlled for with an isotype control–treated group. *, P = 0.05; **, P = 0.005; ***, P = 0.0005. P values > 0.05 were left blank. FC, fold change; Iso, isotype; MFI, mean fluorescence intensity.

Figure 3.

Benchmarking newly identified mAbs against anti-CD25 mAb in CT26 tumors. Mice with CT26 tumors subcutaneously were treated with three doses of 200 μg of mAb on days 0, 4, and 7 (in which day 0 is once tumor becomes palpable). A, Tumor growth was measured three times a week and curves are presented for each mouse. The mean of the tumor growth for the isotype control group is plotted in the black dashed line for all plots. B, Overall survival between the groups is presented. Number of mice for each group is present on the plots. Responses are representative of at least two independent experiments. C, Schematic of in vivo tolerability study; 28-day-old mice were treated with 10 mg/kg of either anti–CTLA-4 or 12-D10 (alone or combined with anti–PD-1) twice weekly for 3 weeks. Serum was collected 24 hours after last dose, and spleen was collected on day 65. D, Serum ELISA measuring cardiac troponin I (TNNI3), spleen mass, and the MCV are presented. One-way ANOVA was performed. *, P = 0.05; **, P = 0.005; ***, P = 0.0005; ****, P < 0.0001. P values > 0.05 were left blank. Exp, experiment; Iso, isotype.

Figure 4.

12-D10 binds ICAM-1 and is able to maintain specificity and depletion activity in different tumor models. A and B, Flow cytometry data confirming that 12-D10 binds Treg cells from the tumor of CT26-bearing mice. N = 3 from one independent experiment. C and D, Flow cytometry data confirming that 12-D10 also binds Treg cells from the tumor of MC38-bearing mice. Representative data from one independent experiment, with N = 3. E, Compiled CT26 survival data demonstrating that treating with three doses of 12-D10 cures approximately 50% of mice. N = 17, pooled from three independent experiments. F, 12-D10-mIgG2a results in a clear depletion of Treg cells from the tumor, while leaving Tregs in the spleen and td-LN untouched. G, 12-D10⁺ and 12-D10⁻ Tregs from MC38 tumors were co-stained for additional T-cell markers. 12-D10 binds mouse ICAM-1 as demonstrated by (H) ELISA, (I) flow cytometry of mouse ICAM-1–transfected (but not control non-transfected) CHO-S cells, and (J) can be blocked by pre-incubating the antibody with recombinant ICAM-1 protein. **, P = 0.005; ***, P = 0.0005; ****, P < 0.0001. P values > 0.05 were left blank. ISO, isotype; LN, lymph node; MFI, mean fluorescence intensity; RLU, relative light units; Spl, spleen; Tum, tumor.

Figure 5.

12-D10 binds the fourth domain of mouse ICAM-1. A, MFI of 12-D10-PE and a commercial anti–mouse ICAM-1 PE mAb binding to Treg, CD8⁺ T cells, and NK cells isolated from CT26 tumor–bearing mice. B, Model of extracellular mouse ICAM-1 structure [using alphafold (35)] with each Ig-like domain colored and denoted. C, Domain truncated variants of mouse ICAM1 were cloned (represented by the schematics) and tagged with a rituximab-specific peptide, Rp3. These were transiently transfected into CHO-S cells and binding of 12-D10, and the commercial clone, was assessed. Binding pattern indicates that 12-D10 binds the fourth domain (yellow) of mouse ICAM1, whereas the commercial clone binds the first domain (red). D, Cross-blocking assay performed in which either unlabeled 12-D10 or ligand was incubated with mouse ICAM-1–expressing cells before adding a fluorescently conjugated variant of the mAb or recombinant ligand-His (p150, 95) detected with anti–His-APC. No blocking was seen indicating that 12-D10 bound a different region to the ligand. Mean and SD of three independent transfections presented. E, Amino acid sequence of domain 4 from human and mouse ICAM-1. Species-different residues are italicized, whereas those highlighted in yellow are predicted to be exposed on the surface of the molecule using the pymol sasa_relative command (62). F, Representative example of 12-D10 losing binding to mutant T289I. G, Ratio of 12-D10 and RTX binding to CHO-S cells transfected with point mutations of the fourth domain ICAM-1. H, Visualization of binding data highlighting species differences that do not affect binding (green), location of T289 (red), proposed dimerization site (orange), and the binding site of ligand (p150, 95; cyan). Image used the resolved crystal structure of human ICAM1, PDB 2OZ4. Anti-His, anti-histidine; FSC-A, forward scatter area; Iso, isotype; LN, lymph node; MFI, mean fluorescence intensity; Ritux, rituximab; Spl, spleen; Tum, tumor.

Figure 6.

12-D10-mIgG2a activity is dependent on FcγR interaction and depletion of Tregs. A, Survival graph of WT BALB/c CT26 tumor-bearing mice treated with either 12-D10-mIgG2a or 12-D10-mIgG1-NA. B, Survival data from WT or FcR-γ chain KO mice bearing CT26 tumors treated with 12-D10-mIgG2a or an isotype control. Six mice per group from a single experiment. C and D, Fold change in Tregs (C) and the CD8:Treg ratio (D) compared with isotype-treated mice in the spleen, td-LN, and tumor assessed as presented in Fig. 2A and B. N = 5 of a single experiment. E, Schematic showing the treatment regimen for PD-1 and 12-D10 combination in MC38 tumors in WT C57BL/6 mice. F, Individual tumor growth and (G) survival of mice from this experiment. Numbers of mice per group are presented on the plots in (F) and are combined from three independent experiments, except for the CTLA-4–only group, which was from a single experiment. *, P = 0.05; **, P = 0.005; ***, P = 0.0005. P values > 0.05 were left blank. D, day; Iso, isotype.

Figure 7.

Identification of similar anti-human ICAM-1 clones. A, Schematic and representative data demonstrating that clones 5-A11 and 3-E11 bind domain 4 of human ICAM-1 in transfected cells whereas enlimomab binds domain 1. B, 5-A11, 3-E11, and enlimomab are used to stain Tregs, CD4⁺, CD8⁺ and non-CD3 immune cells isolated from healthy blood (n = 3) and tumor samples (n = 5) by flow cytometry. Ritux, rituximab.