Targeting high-risk multiple myeloma genotypes with optimized anti-CD70 CAR-T cells

Abstract

Despite the success of BCMA-targeting CAR-Ts in multiple myeloma, patients with high-risk cytogenetic features still relapse most quickly and are in urgent need of additional therapeutic options. Here, we identify CD70, widely recognized as a favorable immunotherapy target in other cancers, as a specifically upregulated cell surface antigen in high risk myeloma tumors. We use a structure-guided design to define a CD27-based anti-CD70 CAR-T design that outperforms all tested scFv-based CARs, leading to >80-fold improved CAR-T expansion in vivo. Epigenetic analysis via machine learning predicts key transcription factors and transcriptional networks driving CD70 upregulation in high risk myeloma. Dual-targeting CAR-Ts against either CD70 or BCMA demonstrate a potential strategy to avoid antigen escape-mediated resistance. Together, these findings support the promise of targeting CD70 with optimized CAR-Ts in myeloma as well as future clinical translation of this approach.

Figure 1.. CD70 is an upregulated surface antigen in high-risk myeloma subtypes.

A. Volcano plot of transcripts in the CoMMpass tumor database (n = 50 patients, paired relapse vs. diagnosis) filtered on those encoding plasma membrane proteins. Significance cutoff p < 0.01, log₂FC > |1|. B. CD70 expression in newly-diagnosed MM tumors in CoMMpass (n = 776; release IA19) as a function of genotype. p-value by two-tailed t-test. C. Overall survival in CoMMpass patients selected for the top and bottom quintile of CD70 expression. p-value by log-rank test. D. Flow cytometry on a panel of MM cell lines comparing CD70 and BCMA expression. Representative of n = 2 per line. E. Flow cytometry on fresh MM patient tumor samples selected as CD138+/CD19− in bone marrow aspirate (top example; isotype used as negative control) or selected as CD38+/CD45 dim/− /CD19− (bottom examples; fluorescence minus one (FMO) used as negative gate). Additional patient sample data shown in Fig. S2. 13 of 27 samples had both BCMA and CD70 measured. Each sample measured once due to sample input limitations. F. Fraction of patient tumor samples defined as positive for CD70 by flow cytometry (MFI >2x that of negative control) as a function of tumor genotype determined on diagnostic fluorescence in situ hybridization (FISH) per the clinical record. p-value by Fisher’s exact test. *p<0.05.

Figure 2.. CRD1 of CD27 is required to generate functional anti-CD70 CAR-Ts.

A. Illustration of various anti-CD70 CAR-T designs. Several prior CD70 CARs have included a full-length CD27 sequence fused to CD3z (left). Here we probe 10 different scFv-based CARs (sequences in Table S3) and several CD27 extracellular domain (ECD) variants fused to a CD28-based CAR backbone. B. Crystal structure of CD27:CD70 complex (PDB: 7XK0) illustrating domains CRD1–3 (“cysteine rich domains”) of CD27 interacting with the CD70 ECD. C. CD27 ECD truncation designs (T1-T6) used in tested CAR constructs. “FL” = full-length. D. In vitro cytotoxicity assays comparing tumor lysis of different CD27 CAR designs compared to positive control of anti-BCMA CAR, “scFv 2” anti-CD70 CAR, and negative controls of “empty” CAR (no antigen binding domain) and “UTD” (untransduced T-cells). 18 hrs, lysis measured by luciferase at the indicated effector:tumor (E:T) ratios. n = 4 technical replicates, representative of 2 biological replicates. Note the elevated lysis of MM.1S cells by truncated constructs is an artifact of increased MM.1S growth in the normalized “tumor only” wells, as evidenced by no dose-response at higher E:T. E. Flow cytometry on lentivirally transduced donor T-cells reflecting binding to FITC-conjugated recombinant CD70. F. Flow cytometry of myc tag present at the N-terminus (extracellular) of all CAR constructs. G. Flow cytometry of intracellular GFP present on CAR construct. E-G: representative of 2 biological replicates. “CD27” = FL ECD CD27 construct.

Figure 3.. Both CD27-based and scFv-based CAR-Ts are effective versus myeloma models in vitro.

A. In vitro cytotoxicity of FL CD27 ECD-based CARs and 3 scFv-based CARs versus several myeloma cell line models. B. Multiplexed cytokine analysis (Eve Biosciences) of CAR T supernatant, after overnight co-culture at 1:1 E:T with LP-1 cells. n = 3 technical replicates. C-D. In vitro cytotoxicity of noted CAR designs versus ANBL-6 myeloma cells (CD70 negative, low BCMA expression) (C) and AMO-1 cells with CD70 knocked out by CRISPR/Cas9 (D). A, C, D: 18 hrs assay, lysis measured by luciferase at noted E:T ratios. n = 3 technical replicates, representative of 2 biological replicates from different T-cell donors. Mean +/− S.D. shown.

Figure 4.. CD27-based CAR-Ts lead to dramatically increased expansion in vivo in an MM model.

A. Murine study design and bioluminescent imaging. Disseminated MM.1S myeloma model implanted intravenously in NSG mice, n = 5/arm. Three anti-CD70 CAR designs are evaluated (scFv -2, -3, -4) compared to FL ECD CD27 CAR. Note that at later time points several mice in BCMA CAR, scFv 2 CAR, and scFv 4 CAR arms were required to be sacrificed without any evidence of tumor burden, likely due to development of graft-vs-host-disease (GvHD) based on murine symptoms. B. Quantified bioluminescence data illustrating tumor control for all anti-CD70 CARs. C. Representative flow cytometry plots of murine peripheral blood at Day 9. CAR Ts were quantified based on human CD3 (hCD3) and GFP positivity. D. Quantification of longitudinal CAR T expansion in murine peripheral blood. Split axis used to illustrate different scales of expansion for CD27 CAR versus others. Figure legend same as in B. p-value by 2-way ANOVA. ****p<0.0001. Mean +/− S.D. shown.

Figure 5.. CD27 knockout from T-cells does not abrogate expansion phenotype.

A. The LP-1 myeloma model harboring t(4;14) translocation (1e6 tumor cells/mouse) was implanted intravenously in NSG mice and 5e6 CAR T cells implanted 5 days later. n = 5 mice/arm. Note increased scale for bioluminescence measurement at final time point. T-cells from a separate donor as used in Fig. 4. Study ended after Day 21 due to development of GvHD symptoms in several mice across arms. B. Bioluminescence from the study illustrating superior tumor control by CD27-based CAR compared to others. C. CD27-based CAR exhibits superior CAR-T expansion over scFv-based anti-CD70 CARs. Gating strategy same as in Fig. 5C. p-value by 2-way ANOVA. **p<0.01. D. CD27 was knocked out donor T-cells (see Fig. S5B for flow cytometry validation) prior to transduction of each CAR, and then used in same intravenously-implanted LP-1 model as (C). Note increased radiance scale for UTD (untransduced) T-cell negative control at later time points. n = 5 mice/arm for scFv 2 and CD27-based CAR arms, n = 3 mice/arm for UTD. E. Quantification of bioluminescence shows continued tumor control by CD27-based CAR in this model. F. CD27-based CAR expansion in blood persists in this model despite CD27 KO from T cells. p-value by 2-way ANOVA. *p < 0.05, **p < 0.01. Difference in expansion between CARs from CD27 WT vs. CD27 KO cells was non-significant (Fig. S5C). Mean +/− S.D. shown.

Figure 6.. Identifying regulators of CD70 expression in multiple myeloma.

A. Knockout of the driver oncogene NSD2 in t(4;14) myeloma cell lines KMS-11 and KMS-34 led to decreased surface CD70 expression by flow cytometry. n = 3 biological replicates. p-value by t-test. *p<0.05, **p<0.01. B. MM.1S or AMO.1 myeloma cells were treated with the indicated dose of azacytidine or 0.1% DMSO for 4 days, and surface CD70 assessed by flow cytometry 3 days later. p-value by 2-way ANOVA. *p<0.05, **p<0.01. n = 3 technical replicates. C. An eXtreme Gradient Boosting (XGB) model (see Methods) extracts features of transcription factor gene expression from CoMMpass (n = 776) that best-model CD70 expression in patient tumors. 80% of data was used as a test set with 20% left out as a training set, with subsequent 5-fold cross validation (see Methods). n = 776 total samples included in the analysis. R² value corresponds to Pearson correlation between model-predicted CD70 expression in the test set and measured tumor CD70 expression in CoMMpass. D. Shapley Additive Explanations (SHAP) analysis suggest expression of transcription factors most strongly impacting CD70 expression levels in CoMMpass tumors. E. Correlation of CD70 with TFAP2A and PAX5 in the CoMMpass dataset. F. Surface CD70 by flow cytometry on LP-1 cells measured after Cas9 RNP nucleofection with sgRNA targeting TFAP2A, CD70, or scramble control. Flow cytometry mean fluorescence intensity shown as fold-change versus isotype control. p-value by t-test. ***p<0.005, ****p<0.001. n = 3 technical replicates, representative of 2 biological replicates. Mean +/− S.D. shown.

Figure 7.. Proof-of-principle dual-targeting CAR-T against both CD70 and BCMA can potentially overcome antigen escape.

A. Illustration of dual-targeting anti-CD70 + anti-BCMA CAR-T to eliminate myeloma cells that do not display CD70 or BCMA on the cell surface. B. Design of bi-cistronic dual-targeting and mono-cistronic single-targeting CAR constructs. C. Cytotoxicity of dual-targeting CAR-T versus LP-1 cell (WT, CD70 KO, or BCMA KO; see Fig. S8); n = 3 technical replicates, 24 hr assay, 3:1 E:T. p-value by two-tailed t-test. Comparisons shown for dual CAR vs. each single-targeting CAR under corresponding single gene knockout conditions. *p<0.05, ***p<0.005. Mean +/− S.D. shown.