The engineered CD80 variant fusion therapeutic davoceticept combines checkpoint antagonism with conditional CD28 costimulation for anti-tumor immunity

Abstract

Despite the recent clinical success of T cell checkpoint inhibition targeting the CTLA-4 and PD-1 pathways, many patients either fail to achieve objective responses or they develop resistance to therapy. In some cases, poor responses to checkpoint blockade have been linked to suboptimal CD28 costimulation and the inability to generate and maintain a productive adaptive anti-tumor immune response. To address this, here we utilize directed evolution to engineer a CD80 IgV domain with increased PD-L1 affinity and fuse this to an immunoglobulin Fc domain, creating a therapeutic (ALPN-202, davoceticept) capable of providing CD28 costimulation in a PD-L1-dependent fashion while also antagonizing PD-1 - PD-L1 and CTLA-4-CD80/CD86 interactions. We demonstrate that by combining CD28 costimulation and dual checkpoint inhibition, ALPN-202 enhances T cell activation and anti-tumor efficacy in cell-based assays and mouse tumor models more potently than checkpoint blockade alone and thus has the potential to generate potent, clinically meaningful anti-tumor immunity in humans.

Fig. 1. ALPN-202, a variant CD80 IgV domain Fc fusion, binds human PD-L1, CD28, and CTLA-4 with high affinity.

a Schematic illustrating the three mechanisms of action of ALPN-202: blockade of PD-1–PD-L1 interaction, PD-L1-dependent CD28 costimulation, and blockade of CTLA-4–CD80/CD86 interactions. b, c Affinity measurements of ALPN-202 and WT CD80-Fc to monomeric wild type PD-L1, CD28, and CTLA-4 were determined by surface plasmon resonance (SPR). Sensorgrams are shown in black lines and results from non-linear least squares regression analysis of the data in orange lines. Sensorgrams were global fit to a 1:1 binding model for triplicate injections of human CD28, CTLA-4, and PD-L1 against captured ALPN-202 and WT CD80-Fc surfaces. For the weak CD28–WT CD80-Fc and PD- L1–WT CD80-Fc interactions, the theoretical Rmax was used as a fixed parameter in the global fit to estimate the KD. WB, weak binding; WT, wild type; ECD, extracellular domain. d, e ALPN-202 binding to CHO cells stably expressing human or mouse PD-L1, CD28, or CTLA-4. ALPN-202 displayed higher affinity for human PD-L1 and CD28, and comparable affinity for CTLA-4, relative to WT CD80-Fc. ALPN-202 did not bind to mouse PD-L1 but did bind mouse CD28 and CTLA-4 with comparable affinity as WT mouse CD80-Fc. TA test article; MFI median fluorescent intensity. f ALPN-202 blocked binding of PD-1-AF647 to cell surface PD-L1 and CD80-AF647 to cell surface CD28 and CTLA-4 to CD80 measured by flow cytometry. g ALPN-202 blocked human PD-L1-mediated recruitment of SHP-2 to PD-1 in a cell-based assay. RLU, relative luminescence units. Experiments in (d–g) were conducted two times and data shown are representative. Source data are provided as a Source Data file.

Fig. 2. CD28 costimulation by ALPN-202 requires TCR activation and co-binding to PD-L1.

Primary human T cells, were co-cultured for 24 h with test articles (TA) and K562 cell artificial antigen-presenting cells (aAPC) expressing membrane anchored anti-CD3 clone OKT3 (maOKT3), PD-L1, or both. a In the absence of PD-L1 to anchor ALPN-202 on the aAPC, no increase in IL-2 was detected above background. b In the absence of TCR stimulation (via maOKT3), no increase in IL-2 was observed. c When both maOKT3 and PD-L1 were present on the aAPC, ALPN-202 (green line) induced a strong dose-dependent costimulatory signal above that of WT CD80 ECD-Fc or PD-(L)1 blockade alone. WT wild type; ECD extracellular domain. d The costimulatory activity of ALPN-202 was inhibited when combined with blocking antibodies to either PD-L1 or CD28, confirming the requirement for dual binding to induce costimulation. e Schematic illustrating how a monomeric (no Fc) CD80 variant Ig domain (vIgD) binds PD-L1 and engages CD28 in trans. f Overlaid SPR sensorgrams for two paired serial injections of monomeric CD28 (black line) followed by PD-L1 (blue line) or CTLA-4 (orange line) against ALPN-202 surfaces. Data were collected using two serial 180 s, 1500 nM injections. Single analyte injections of PD-L1 (pink line) or CTLA-4 (red line) were made by pairing with blank buffer injections. g Primary T cells were co-cultured with K562/maOKT3/PD-L1 cells and a titration of the monomeric ALPN-202 CD80 vIgD with saturating anti-PD-L1 antibody, anti-CD28 antibody, or Fc control for 24 h. For (a–d), and (f), each data point is mean concentration of IL-2 for each sample run in duplicate wells. Data shown are representative of three separate donors run independently. Source data are provided as a Source Data file.

Fig. 3. Crystal Structure of ALPN-202 CD80 vIgD complexed with the wild type PD-L1 ECD.

a Representative ribbon structure from the ALPN-202 CD80 variant Ig domain (vIgD) (green) and wild type (WT) PD-L1 extracellular domain (ECD) (red) crystal structure with the interface contact residues indicated in blue (PDB: 7TPS) b Model alignments of one subunit (Chain A and Chain B) of the ALPN-202 CD80 vIgD (green)/PD-L1 ECD (red) asymmetric unit aligned with wild type CD80 ECD (gray)/CTLA-4 ECD (blue) (PDB: 1I8L) and CD28 (pink) (PDB: 1YJD). The CD28 and CTLA-4 MYPPPY motifs are indicated in yellow. c Comparison of ALPN-202 CD80 vIgD binding contacts in green, PD-1 binding contacts in blue (PDB: 4ZQK) and ALPN-202 and PD-1 common contacts indicated in magenta mapped onto the surface of PD-L1 IgV domain (gray). d Atezolizumab (orange) contacts mapped to the same surface as PD-1 and ALPN-202 CD80 vIgD contact residues on the surface of PD-L1 IgV (gray).

Fig. 4. ALPN-202 increases T cell activation more potently than CPI alone in the presence of PD-L1⁺ suppressive M2c macrophages.

Primary human T cells were co-cultured with in vitro-derived M2c macrophages and a titration of ALPN-202 or anti-PD-1, PD-L1, or CTLA-4 blocking antibodies. ALPN-202 induced (a) CD4⁺ T cell proliferation, (b) CD8⁺ T cell proliferation, and (c) IL-2 secretion in a dose-dependent manner more potently than CPI alone. Proliferation data shown are mean percent increase relative to Fc control and is from a representative donor (n = 5 donors). Error bars are SD from triplicate wells. TA test article. d ALPN-202 activity increased expression of IL-2, IL-21, IFNγ, and GM-CSF in cell culture supernatants and is dependent on binding to PD-L1 and CD28. Data are shown as fold-increase (dark shade) or decrease (white) with respect to untreated control (M2c + anti-CD3). Data shown are from five donors and were collected in two independent experiments. M1 proinflammatory M1 macrophages. e To compare ALPN-202 activity to PD-1, CTLA-4, or dual checkpoint inhibition (CPI), E6 TCR⁺ T cells were co-cultured with M2c macrophage and SCC152/hPD-L1⁺ tumor cells for 24 h. Mean ± SD for IFNγ, IL-2, and TNFα was quantitated from triplicate wells. Data are representative of four donors tested in two independent experiments. E6 TCR—T cell receptor specific for E6 peptide. Source data are provided as a Source Data file.

Fig. 5. ALPN-202 demonstrates potent and dose-dependent anti-tumor activity in vivo in a MC38/human PD-L1 tumor model.

a ALPN-202 induced a small but significant decrease in tumor growth relative to Fc control in a parental MC38 tumor model. (n = 10 mice/group). b, c In MC38/hPD-L1 tumor studies, ALPN-202-mediated anti-tumor activity (green) was superior to wild type WT CD80 ECD-Fc (blue), anti-PD-L1 (durvalumab, orange), or isotype control (black) (n = 10 mice/group). WT wild type; ECD extracellular domain. d A subset of tumors from study shown in (c) were harvested 72 h after the first dose and RNAseq performed. Transcripts per kilobase million reads (TPM) of select genes for individual tumors are shown. Individual points represent TPM for individual tumor samples and bars indicate mean values ± SEM. e Signature transcripts from the indicated immune cell populations were identified using the ImmGen Population Comparison application [https://www.immgen.org/ImmGenpubs.html]. The heatmap compares the fold increase (red) or decrease (blue) in mean transcript levels of either ALPN-202-treated or anti-PD-L1 treated tumors relative to Fc-control-treated tumors (n = 4 tumor samples per group). f ALPN-202-treated MC38/hPD-L1 tumors (n = 12 mice/group) showed a dose-dependent increase in anti-tumor activity. g 72 h after treatment there was a dose-dependent increase in intratumoral p15E⁺ tumor antigen-specific CD8⁺ T cells. No change was observed in inguinal lymph nodes. N = 5 samples from each group (individual symbols) and bar height indicates mean percentage of p15E⁺ CD8⁺ T cells. Error bars are SEM. Gating strategy is summarized in Supplementary Fig. 7. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple comparisons test (d, g) or by repeated-measures two-way ANOVA for treatment effects (a, b, c, f). P values shown are ALPN-202 vs Fc control. For all studies except (f, g), antibodies and ALPN-202 were treated at 100 µg/dose while Fc control was 75 µg/dose. For all graphs, the means ± SEM are shown. Source data are provided as a Source Data file.

Fig. 6. ALPN-202 increases intratumoral inflammatory cell infiltrate in part by hPD-L1-dependent mCD28 costimulation.

a, b The anti-tumor activity of ALPN-202 was significantly blocked by combining with (a) anti-hPD-L1 or (b) anti-mCD28 blocking antibodies (n = 12 mice/group). Mean tumor volumes ± SEM are shown. Tumors were collected from mice 48 h after the second dose, dissociated, and analyzed by flow cytometry. ALPN-202 treatment significantly increased the percentage of infiltrating total CD3⁺ T cells (c), CD8⁺ T cells (d), and granzyme B⁺ CD8⁺ T cells (e) relative to Fc Control treated animals. In each case, this influx in T cells was reduced when ALPN-202 was combined with anti-hPD-L1 or anti-mCD28 antibodies. Statistical significance was determined by repeated-measures two-way ANOVA for treatment effects (a, b) or one-way ANOVA with Dunnett’s multiple comparisons test (c–e). Bar height indicates mean value ± SEM and symbols represent individual tumor samples (n = 5 tumors per treatment group) (c–e). Source data are provided as a Source Data file.

Fig. 7. ALPN-202 combines with anti-PD1 and anti-CTLA-4 antibodies and is active in a humanized tumor model.

a, b Tumor growth curves for (a) hPD-L1-expressing MC38 tumors (n = 10 mice/group) or (b) B16-F10 tumors (n = 9 mice/group) treated with ALPN-202 (green line), anti-mPD-1 (orange line), a combination of ALPN-202 and anti-mPD-1 (blue), or an isotype control (black). c MC38/hPD-L1 tumors (n = 9 mice/group) treated with ALPN-202 (green), anti-mCTLA-4 formatted with either mIgG2b Fc (red) or inert Fc (orange), or combinations of ALPN-202 with each mCTLA-4 antibody (dark blue, light blue respectively). d Tumor growth curves (n = 9 mice/group) from NSG mice implanted with human E6 TCR⁺ transgenic T cells as well as human SCC152/hPD-L1 tumor cells treated with ALPN-202 either every three days for four doses (dark green; Q3Dx4) or once every seven days for three doses (light green; Q7Dx3), anti-hPD-L1 (durvalumab, orange) or Fc control (black). NSG NOD-scid IL2Rγ^null; E6 TCR, T cell receptor specific for E6 peptide. All experiments were repeated at least two times and data shown are representative. Statistical significance was determined by repeated-measures two-way ANOVA for treatment effects. For all graphs, means ± SEM are shown. Source data are provided as a Source Data file.