Establishment of a robust CRISPR activation platform
CRISPRa has the potential to be used for the development of accurate disease models because the induction of oncogenic driver genes from their endogenous promoters can be more modest and physiologically relevant compared to that achieved with cDNA constructs. We therefore sought to establish a robust and widely applicable CRISPRa system for the generation of faithful pre-clinical cancer models and the identification of targets that could translate into improved therapies for cancer patients. We adapted the SAM system originally described in a two-vector configuration to be expressed from a single construct to achieve similar expression of all the components required for CRISPR mediated gene induction.9 To this end, we linked the dCas9-VP64 via a T2A sequence to the MS2-p65-HSF1 expression cassette (Fig. 1a). In addition, we incorporated a GFP sequence as a marker via a second T2A sequence downstream of the dCas9-VP64-T2A-MS2-p65-HSF1 coding sequence. We incorporated the cassette into a lentiviral vector for ease of manipulating cells22. We initially sought to validate the efficiency of the CRISPRa cassette in vitro. We therefore introduced lentiviral vectors encoding the CRISPRa cassette and one of three unique sgRNAs targeting the BCL-2 promoter into cell lines derived from the Eµ-Myc transgenic mouse model of lymphoma23 (Supplementary Fig. 1). Western blot analysis of two independent lymphoma cell lines confirmed that all sgRNAs caused a substantial increase in BCL-2 expression, detectable even before puromycin selection of sgRNA-transduced cells (Fig. 1b).
Eµ-Myc lymphomas are highly reliant on the pro-survival protein MCL-1 for their sustained survival24, 25, 26 and hence the elevated BCL-2 expression increased resistance to the MCL-1 selective inhibitor S63845 (Fig. 1c). Interestingly, the enforced expression of BCL-2 in these lymphoma cells did not sensitise them to the BCL-2 inhibitor venetoclax (Fig. 1d), likely because these tumour cells did not develop under conditions of high BCL-2 expression and therefore do not depend on BCL-2 expression for continued survival. These experiments confirmed that our CRISPRa cassette could induce strong upregulation of genes in cell lines in vitro but also highlighted the limitations of engineering established cell lines for studies relevant to more sophisticated cancer models. We therefore sought to progress this system to an in vivo setting, by generating transgenic mice with a similar configuration for CRISPRa mediated transcriptional upregulation.
Generation of CRISPRa transgenic mice
For flexible expression of the CRISPRa system in vivo, we targeted the dCas9-VP64-T2A-MS2-p65-HSF1 cassette into the ubiquitously expressed Rosa26 locus. We used the CTV vector,27 which has a loxP flanked stop cassette between the CAGS promoter and the cDNA, allowing for temporal or cell type specific induction of gene expression upon CRE mediated deletion (Fig. 2a). Crossing the CRISPRa transgenic mice to a CRE deleter strain removed the stop cassette, allowing expression of the CRISPRa components in all tissues. We refer to the resulting strain as dCas9a-SAM. Transgene insertion and expression was confirmed by PCR on DNA isolated from mouse tails and flow cytometric analysis for GFP on cells isolated from the thymus, spleen, bone marrow and lymph nodes (Fig. 2b, c). Moreover, intracellular staining and flow cytometric analysis of thymus and bone marrow cells with Cas9-specific antibodies confirmed that the CRISPRa components were expressed in vivo (Fig. 2d). Since the dCas9 protein is fused to the VP64 activation domain, we next sought to determine whether the constitutive expression of a transcriptional activator protein had any impact on the cellular composition of the mice. We analysed the haematopoietic compartment because it is particularly sensitive to cytotoxic stress and even subtle changes in gene expression. We found that homozygous dCas9a-SAM mice showed no differences in the frequencies and numbers of the different immune cell subsets in the thymus, bone marrow, spleen, and lymph node compared to wildtype mice (Supplementary Fig. 2). Importantly, aged (older than 12 months) dCas9a-SAMKI/KI mice showed no signs of disease, further substantiating that constitutive expression of the CRISPRa system in all tissues of the animals does not cause marked toxicity or substantive changes to the proportions of all tested cell types.
Next, we assessed the efficiency of the CRISPRa components for gene activation in primary cells derived from the dCas9a-SAM mice. To this end, we stimulated splenocytes isolated from dCas9a-SAMKI/KI mice with LPS or Concanavalin A plus Interleukin-2 for the generation of activated B or T cell blasts, respectively. We transduced these cells with the Bcl-2 sgRNA1 construct (that co-expresses puromycin resistance) and selected transduced cells with puromycin for 2 days. Analysis of BCL-2 expression by intracellular flow cytometry and Western blotting revealed upregulation of BCL-2 protein in both B and T cells (Fig. 3a). To assess the potential of the CRISPRa system for targeting various genes, we introduced sgRNAs for CD19 or IRF4 into B cells. As expected, CD19-specific sgRNAs elevated surface expression of CD19 above the basal level, while introducing sgRNAs for IRF4, known to induce differentiation of B cells into plasma cells,28 enhanced the frequency of cells positive for the plasma cell marker CD138 (Syndecan) (Fig. 3b, c). Similarly, transducing activated T cells with sgRNAs targeting the CD4 promoter enhanced expression of CD4 on the cell surface (Fig. 3d). Since these genes are already transcriptionally active in B cells or T cells, respectively, we next challenged our CRISPRa system by attempting to induce expression of genes that are normally transcriptionally silent in these cells. We first introduced sgRNAs targeting the B cell marker CD19 into T cells. Remarkably, this elicited CD19 expression in almost 50% of T cells (Fig. 3e). Similarly, introduction of sgRNAs targeting the T cell specific CD4 gene promoter into B cells resulted in B cells with CD4 expression (Fig. 3f). These data demonstrate that we have developed a powerful CRISPRa mouse model that displays no detectable toxicity as well as the potency required to induce targeted gene expression in primary cells, even of genes that are normally silenced within a specific fully differentiated cell type.
Exploiting CRISPRa in vivo for the development of aggressive lymphomas
Having validated the efficiency of the CRISPRa mouse for gene induction in primary cells, we set out to test its applicability for developing disease models. Initially, we sought to confirm whether induction of a gene product by CRISPRa could indeed affect disease aetiology, for example, by modulating the latency of tumour development. We know from previous reports that deleting the p53 tumour suppressor accelerates Eµ-Myc-driven tumourigenesis.29 To replicate the effect of loss of p53 activity using CRISPRa, we transduced Eµ-Myc/dCas9a-SAMKI/+ HSPCs with sgRNAs to induce expression of MDM2 (an E3 ligase that degrades p53 protein)30 (sgMdm2), or non-targeting sgRNAs as controls. The transduced HSPCs were transplanted into lethally irradiated C57BL/6-Ly5.1 recipient mice which were observed for tumour development (Fig. 4a and Supplementary Fig. 3a). As occurs for p53 knockout in the Eµ-Myc background, we observed accelerated tumour onset in the mice reconstituted with Eµ-Myc/dCas9a-SAMKI/+/sgMdm2 transplanted HSPCs compared to controls. To determine the levels of p53 in the lymphomas expressing sgMdm2, we derived cell lines and induced expression of p53 with the MDM2 inhibitor Nutlin3a.31 Western blot analysis clearly demonstrated a reduction in p53 protein levels in Eµ-Myc/dCas9a-SAMKI/+/sgMdm2 lymphomas compared to controls consistent with elevated MDM2 levels (Supplementary Fig. 3b). Having shown that our CRISPRa model is indeed powerful enough to affect lymphomagenesis by induced expression of MDM2 leading to reduced p53, we next sought to develop a long sought after model of aggressive DHL, for which all previous attempts to mimic this devastating disease have failed. Such a lymphoma model would facilitate the identification and validation of novel therapeutic strategies for patients with DHL. To this end we used the same approach as described above, this time using one of the validated sgBcl-2 constructs described in Fig. 1. Irradiated mice were injected with the Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 HSPCs or control Eµ-Myc/dCas9a-SAMKI/+/non-targeting sgRNA HSPCs, and 6 weeks post transplantation, BCL-2 expression was analysed in haematopoietic cells of the mice by flow cytometry. The analysis revealed increased BCL-2 expression in peripheral blood cells of mice transplanted with Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 HSPCs (Fig. 4b). Accordingly, these mice went on to develop aggressive lymphomas with a median latency of 68 days, compared with a median latency of 132 days for mice transplanted with the same HSPCs that had been transduced with a non-targeting sgRNA (Fig. 4c). Characterisation of Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 lymphomas revealed a B cell phenotype (CD19/B220 double positive; Fig. 4d) and high BCL-2 protein expression (Fig. 4e), which are both also observed in human DHL18. These mature B cell lymphomas could readily be derived into cell lines in vitro (Fig. 5). This model of aggressive lymphoma therefore contrasts with a previous one that utilised Eµ-Myc/Eµ-Bcl-2 double transgenic mice which developed lymphomas exhibiting an immature haematopoietic progenitor phenotype16 that could not be grown in vitro32 (Fig. 4, 5).
For further characterisation and experimentation, cell lines were derived from the Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 lymphomas and non-targeting control lymphomas. All sgBcl-2 tumour-derived lines displayed high expression of BCL-2 and the pro-apoptotic BH3-only protein BIM, compared with the control lymphoma cell lines derived from tumours with non-targeting control sgRNAs (#219, #220), as detected by both Western blotting (Fig. 5a) and intracellular flow cytometric analysis (Fig. 5b). We further noted that expression of the related pro-survival protein BCL-XL was highly variable across the individual lymphomas (Fig. 5a). Significantly, we found that all cell lines derived from sgBcl-2 lymphomas were sensitized to venetoclax treatment (mean IC50=0.11 µM), which is in striking contrast to the standard lymphomas that arise in Eµ-Myc transgenic mice (mean IC50>1 µM; Fig. 5c and Supplementary Fig. 4a, b). In addition, whilst the BCL-2 expressing lymphoma lines were less sensitive overall to treatment with the MCL-1 inhibitor S63845 than control Eµ-Myc lymphoma lines, some of the BCL-2 expressing lines that displayed lower venetoclax sensitivity were still similarly sensitive to the MCL-1 inhibitor (Fig. 5d and Supplementary Fig. 4c).
These data confirm that we have been able to establish a novel model of aggressive B cell lymphoma, that is very similar to human DHL, i.e. double expression of c-MYC and BCL-2 and surface expression of CD1918. Our results suggest that both the BCL-2 inhibitor venetoclax and MCL-1 inhibitors (already in clinical trials for B cell malignancies but not DHL)33 could be used for the treatment of this disease in humans.
Identification of venetoclax resistance factors in the new model of aggressive lymphoma using genome wide CRISPR activation screens.
An important clinical issue is the emergence of resistance to venetoclax in patients on therapy. The generation of Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 lymphoma cell lines that are highly venetoclax sensitised provided a system in which whole genome CRISPR activation screens could be carried out to identify genes that confer drug resistance when upregulated. We transduced six replicates each of two venetoclax sensitive murine BCL-2 expressing lymphoma lines with a recently described mouse genome-wide sgRNA library.34 The cells were cultured for two weeks after transduction to permit induction of gene expression and were then subjected to treatment with vehicle (DMSO) or the indicated concentrations of venetoclax for a further two weeks (Fig. 6a). DNA samples were collected, and next generation sequencing was performed to identify the sgRNAs enriched in venetoclax-treated versus control cell populations. At all concentrations, venetoclax treatment led to a strong enrichment of a subset of sgRNAs compared to the DMSO treated control samples (Supplementary Fig. 5a). Notably, we found enrichment of sgRNAs upregulating two pro-survival BCL-2 family members, BCL-XL and MCL-1, that, based on current literature, would be expected to mediate resistance to venetoclax19 (Supplementary Fig. 5b). To our surprise, however, we found that sgRNAs targeting the underappreciated pro-survival BCL-2 family member A1 were the most dominant sgRNAs enriched by venetoclax treatment in both cell lines (Fig. 6b). This was particularly evident in IC80 doses of venetoclax, where multiple sgRNAs targeting A1 were highly significantly enriched (FDR < 0.05) compared to DMSO treated control groups (Supplementary Table 1 and Supplementary Data). To confirm that upregulation of A1 can confer protection from venetoclax induced killing, a sgRNA targeting the Bcl2a1a promoter was transduced into two Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 cell lines. Upregulation of A1 expression in these cells was confirmed by Western blotting (Fig. 6c). Cell competition assays in vitro confirmed that A1-activated cells possessed a striking survival advantage over parental Eµ-Myc/dCas9a-SAMKI/+/sgBcl-2 cells in the presence of venetoclax (Fig. 6d), confirming that upregulation of A1 confers resistance to venetoclax treatment. Similar results were obtained when cells were engineered to upregulate MCL-1 expression rather than A1 (Supplementary Fig. 5c, d). In addition to the genes encoding pro-survival BCL-2 pro-survival proteins, we also identified a number of other genes which may be interesting to investigate in the context of venetoclax resistance (Supplementary Fig. 5).