BCL2 Inhibition Reveals a Dendritic Cell-Specific Immune Checkpoint That Controls Tumor Immunosurveillance

Abstract

We developed a phenotypic screening platform for the functional exploration of dendritic cells (DC). Here, we report a genome-wide CRISPR screen that revealed BCL2 as an endogenous inhibitor of DC function. Knockout of BCL2 enhanced DC antigen presentation and activation as well as the capacity of DCs to control tumors and to synergize with PD-1 blockade. The pharmacologic BCL2 inhibitors venetoclax and navitoclax phenocopied these effects and caused a cDC1-dependent regression of orthotopic lung cancers and fibrosarcomas. Thus, solid tumors failed to respond to BCL2 inhibition in mice constitutively devoid of cDC1, and this was reversed by the infusion of DCs. Moreover, cDC1 depletion reduced the therapeutic efficacy of BCL2 inhibitors alone or in combination with PD-1 blockade and treatment with venetoclax caused cDC1 activation, both in mice and in patients. In conclusion, genetic and pharmacologic BCL2 inhibition unveils a DC-specific immune checkpoint that restrains tumor immunosurveillance.

Significance: BCL2 inhibition improves the capacity of DCs to stimulate anticancer immunity and restrain cancer growth in an immunocompetent context but not in mice lacking cDC1 or mature T cells. This study indicates that BCL2 blockade can be used to sensitize solid cancers to PD-1/PD-L1-targeting immunotherapy. This article is featured in Selected Articles from This Issue, p. 2293.

Figure 1. CRISPR KO screens for the identification of genes that act as DC immune checkpoint.

(A, B) Inducible immortalized dendritic cell (iniDC) precursors and de-induced/differentiated immature DCs (de-iniDCs) were incubated with different concentrations of cytochrome c (Cyt c, 1, 2.5, 5 mg/mL) for 24 or 48 hours before being subjected to annexin V/DAPI staining. The percentages of apoptotic (annexin V⁺ DAPI^-) and dead (DAPI⁺) cells are reported as stacked bar charts (mean ± SD with all data points). (C,D) Wild type (wt) or beta 2 microglobulin knockout (B2m^-/-) de-iniDCs were treated with navitoclax (Nav), venetoclax (Ven), or DMSO control (Ctr) before incubated with ovalbumin (OVA) or OVA peptide (S257-L264, SL8) to stimulate the activation of specific TCR-expressing B3Z T cell hybridoma. The secretion of IL2 during this process was quantified to indicate antigen presentation capability and the results are reported as bar charts (mean ± SD with all data points). Statistical analysis in A-D was performed using Tukey’s multiple comparisons test. (E) The schematic pipeline of a pooled genome-wide CRISPR KO screen and the thresholds for the flow cytometric enrichment of cells with desired gain-of-function phenotype. Cas9-expressing iniDCs (iniDC_Cas9) were infected with a lentiviral CRISPR knockout library and following antibiotic selection, dexamethasone/doxycycline (Dex/Dox) were withdrawn to differentiate transduced cells into de-iniDCs, that were then exposed to OVA and subjected to flow cytometric enrichment of mature antigen-presenting DCs. The abundance of gRNA sequences in selected cells was quantified with next generation sequencing. (F) A volcano plot displays the abundance of gRNAs/gene comparing sorted with unsorted cells as fold change (FC) versus significance (RRA score calculated with MAGeCK software). As cutoffs log2FC > 1.6 and RRA score < 0.001 have been chosen. Genes were significantly different from controls for both parameters (red dots, 715 genes) were used to perform a KEGG pathway mapping (G) and the gene enrichment factor is shown for the indicated pathways (dot size indicates number of genes, and dot color represents estimated false discovery rate (FDR)).

Figure 2. Arrayed CRISPR screening and validation of BCL2 as a target for DC stimulation

(A) A scheme illustrates the arrayed CRISPR KO screening procedure. Cas9-expressing iniDCs (iniDC_Cas9) were transfected with individual gRNAs in the absence of Dex/Dox to establish specific gene KO and simultaneously differentiate the precursors into immature de-iniDC_Cas9, which were then exposed to OVA, washed and cocultured with B3Z T cell hybridoma cells that would be activated to secret interleukin 2 (IL2), indicating antigen cross-presentation capacity. (B) Log2 fold change (FC) of IL2 quantity for each KO as compared with non-target gRNA transfected controls and corresponding p-values (obtained with t-test) were used to generate a volcano plot. As cutoffs log2FC > 1 and p < 0.001 were chosen. gRNAs led to significant change of IL2 production from non-targeting control gRNA for both parameters are shown in red. Bcl2-targeting gRNA-transfected de-iniDC_Cas9 cells were cloned by flow cytometry and expanded. (C, D) Clones exhibiting Bcl2 protein levels undetectable by immunoblot (C) were evaluated for their capability to cross-present OVA to B3Z cells, as indicated by the secretion of IL2 (D). The significance of increased antigen cross-presentation over wild type (wt) de-iniDC_Cas9 cells were calculated by means of t-test. P-values are labelled in the figure as comparing between the indicated groups. (E) Differentiated iniDCs (de-iniDC) were treated with the indicated chemical inhibitors of Bcl2 family proteins, at different concentrations (1, 5, 10 μM), overnight before incubation with soluble OVA in the presence of the indicated treatments for 6 h. Venetoclax (Ven), navitoclax (Nav), and ABT737 are targeting Bcl2; A-1331852 and WEHI-539 are targeting Bcl-XL; S63845 is targeting Mcl1. The cells were washed and cocultured with B3Z T cells to evaluate their capability in priming the latter to secret IL2 as quantified by ELISA. (F)Treatments with Nav and Ven were applied to bone marrow-derived DCs (BMDCs) for 2~6 h before exposure to OVA to test their capability to activate B3Z cells. After treatment with venetoclax at the highest concentration, (G) De-iniDC viability was assessed by annexin V/DAPI staining and the percentage of apoptotic (annexin V⁺ DAPI^-) and dead (DAPI⁺) cells is reported as stacked bar chart, mean ± SD with all data points. Statistical significance was calculated by means of two-way ANVOA with Dunnett’s multiple comparisons test. (H) Bulk RNASeq data obtained with wild type (wt) de-iniDC, Ven-treated wt de-iniDCs and Bcl2^-/- de-iniDCs was deconvoluted using the BayesPrism algorithm and aligned with scRNA-sequencing signatures from Brown et al (59). Barplot represents virtual proportions (Theta_Fraction) of scRNAseq DC signatures in bulk RNAseq samples. (I) Bulk RNASeq data also indicates a set of genes that are regulated in Ven-treated wt de-iniDCs and Bcl2^-/- de-iniDCs as compared to untreated wt de-iniDCs. (J) Based on the commonly upregulated genes, the GO term enrichment analysis for immune cell activation related genes is depicted as gene numbers with log transformed adjusted p values for these two comparisons.

Figure 3. Pharmacological inhibitors of BCL2 activate conventional Type 1 DCs in vivo.

(A-D) Orthotopic MCA205 fibrosarcoma-bearing mice received two intraperitoneal (i.p.) injections venetoclax or solvent controls (Ven) at day 0 (when tumors became palpable) and day 2. The blood, tumor, tumor-draining lymph node (tdLN), and non-tumor-draining LN (inguinal LN on the opposite side of the tumor) were harvested at day 4 and dissociated into single cell suspensions for multiplex immunostaining and flow cytometric analysis. The mean fluorescence intensity (MFI) of indicated markers on the Type 1 conventional DCs (cDC1, defined as F4/80^- MHC-II⁺CD11c⁺CD103⁺CD11b^- among viable leukocytes) as well as cDC2 cells (CD103^-CD11b⁺) were normalized to the average value of the Sol condition, which are depicted as scattered dot plots (n=10 animals/group). Statistical significance was calculated using one-way ANOVA test with Dunnett’s multiple comparisons, as compared to Sol. (E) Cryopreserved PBMCs from acute myeloid leukemia (AML) patients, before- and after- (1 week) treatment with Ven plus azacytidine (Aza), were subjected to multiplex immunostaining and high dimensional flow cytometric analysis to quantify the expression of distinct DC activation related markers on the surface of different subpopulations. The ratio of MFI (after : before treatment) of CCR2, CCR7, XCR1, CD5, CD86, and HLA-DQ on different DC subsets from individual patients are depicted as scatter plot (n=5 patients). (F-I) PBMCs from healthy donors were treated in vitro with Ven, Aza, or their combinations and were then subjected to the same staining and analyses as above. The ratio of MFI (mono or combined treatment: DMSO control) from individual healthy donors are depicted as scatter plots (n=5 donors). A paired t test was performed to calculate the p values.

Figure 4. Bcl2 deletion or inhibition in DCs improves their potency in tumor control.

(A) Schematic presentation of the experiment. Orthotopic lung cancers were established by intravenous (i.v.) injection of luciferase-expressing TC1 cells at day 0. Wild type (wt) or Bcl2^-/- de-iniDC_Cas9 cells were differentiated and fed with the lysate of TC1 cells to be activated before i.v. injection into the tumor-bearing mice. Once the tumors became detectable by bioluminescence (~ day 7), the animals were treated with 4 DC infusion cycles in combination with neutralizing antibodies to PD-1 (αPD-1) or corresponding isotype control antibody (αIso) as illustrated in the scheme. (B) Representative images of tumor development under different treatment conditions. Bioluminescence signals were quantified as total flux of photons to indicate tumor size in the lung and are reported as tumor growth curves as, means ± SEM (C). Final tumor size distribution at endpoint is shown in (D). (E) In some cases, de-iniDCs were pre-treated for 4 h with navitoclax (Nav) or venetoclax (Ven) before exposure to tumor lysate, and the adoptive DC transfer were used in combination with neutralizing antibodies to CD4 and CD8 (αCD4/CD8). (F,G,H) Tumor growth curves in mice receiving PBS as a control or de-iniDCs pretreated with navitoclax (Nav) or venetoclax (Ven) in vitro before injection (F) or de-iniDCs with different genotypes combined with control antibodies (G) or monoclonal antibodies depleting CD4⁺ and CD8⁺ T cells (H). Statistical significance was calculated by means of the Type II ANOVA test for tumor growth curves and one-way ANOVA test for the tumor size distributions, n = minimum of 7 mice/group. P-values are indicated to indicate statistical significance in intergroup comparisons.

Figure 5. Systemic administration of Bcl2 inhibitors exerts immune-dependent anticancer efficacy.

Orthotopic lung cancers were established by intravenous (i.v.) injection of luciferase-expressing TC1 cells on day 0. Once the tumors became detectable by bioluminescence (~ day 7), the animals were intraperitoneally (i.p.) treated with either solvent (Sol), navitoclax (Nav), or venetoclax (Ven) at day 7, 9, and 11 in combination with neutralizing antibodies to CD11b (αCD11b) or CD4 and CD8 (αCD4/CD8), or the isotype control antibody (αIso) as illustrated in the scheme (A). Representative images of tumor development under different treatment conditions are shown in (B). Bioluminescence signals were quantified as total flux of photons to indicate tumor sizes on the lung which are reported as tumor growth curves (C, mean ± SEM). The percentage of overall survival is reported in (D). Statistical significance was calculated by means of the type II ANOVA test (C), or logrank test (D), n = minimum of 8 mice/group. P-values are labelled in the figure and indicate statistical significance as compared with PBS or between the indicated groups.

Figure 6. Genetical depletion of Type 1 conventional DCs interferes the antitumor effect by Bcl2 inhibitors.

(A) Lethal dose irradiated mice were transplanted with bone marrow from WT or Batf3-KO donors to establish a conditional change in the immune system. In a pilot experiment, the splenocytes and BM cells were isolated from bone marrow reconstituted mice to verify the deletion of Batf3 by qPCR (B) and the depletion of cDC1s by flow cytometric analysis (C,D). For the tumor growth experiment, orthotopic lung cancers were established in bone marrow reconstituted mice by intravenous (i.v.) injection of luciferase-expressing TC1 cells on day 0. Wild type (wt) or Bcl2^-/- de-iniDC_Cas9 cells were differentiated and activated with the lysate of TC1 cells before i.v. injection into the tumor-bearing mice. Once the tumors became detectable by bioluminescence (~ day 7), the animals were infused with 4 cycles of de-iniDCs, with or without combination of intraperitoneal (i.p.) treatment by either solvent (Sol) or venetoclax (Ven). Representative images of tumor development under different treatment conditions are shown in (E). Bioluminescence signals were quantified total flux to indicate tumor size on the lung which are reported as tumor growth curves (F,G, mean ± SEM) The percentage of overall survival is reported in (H). Statistical significance was calculated by means of the type II ANOVA test (F,G), or logrank test (H), n = minimum of 6 mice/group. P-values are labelled in the figure and indicate statistical significance as compared with PBS + Sol or between the indicated groups.

Figure 7. Systemic depletion of Type 1 conventional DCs interferes the antitumor effect by Bcl2 inhibitors.

Orthotopic lung cancers were established in wild type C57BL/6 mice by intravenous (i.v.) injection of luciferase-expressing TC1 cells on day 0. Once the tumors became detectable by bioluminescence (~ day 7), the animals were subjected to treatment as schemed in (A). Briefly, mice were pre-treated (i.v.) with either 5 mg/mouse of cytochrome C (Cytc) or the corresponding vehicle (PBS) before other treatment and maintained the injection every other day for two weeks. Then intraperitoneal (i.p.) treatment with either solvent (Sol), navitoclax (Nav), or venetoclax (Ven) was applied, and the combination with PD-1 (αPD-1) blockade or equivalent isotype antibody (αIso) followed 8 days after chemical treatment. Bioluminescence signals were quantified as total flux to indicate tumor size on the lung which are reported as tumor growth curves (B,C, mean ± SEM). The percentage of overall survival is reported in (D). Statistical significance was calculated by means of the type II ANOVA test (B,C), or logrank test (D), n = 9 mice/group. (E-H) Orthotopic TC1 lung tumor-bearing mice were treated as described above. Spleen, lung, and blood were harvested 7 days later and were dissociated into single cell suspensions for multiplex immunostaining and flow cytometric analysis. The percentage of PD-1⁺ cells in total T cells (E), CD8⁺ T cells (F), CD4⁺ T cells (G), as well as Tregs (defined as CD4⁺FOXP3⁺CD25⁺,H) among total T cells were evaluated with FlowJo and depicted as dot plots (n=8 animals/group). Statistical significance was calculated using one-way ANOVA test. (I,J) Lung tumor growth in TC-1 lung cancer-bearing mice that were treated with Cytc, Ven, antiPD-1, or their combinations, were monitored and depicted as growth curves (I); overall survival is reported as Kaplan–Meier curves (J). Statistical analysis was performed as (B-D). P-values are labelled in the figure to indicate statistical significance as compared with PBS + Sol or between the indicated groups.