Epigenetic control of multiple genes with a lentiviral vector encoding transcriptional repressors fused to compact zinc finger arrays

Abstract

Gene silencing without gene editing holds great potential for the development of safe therapeutic applications. Here, we describe a novel strategy to concomitantly repress multiple genes using zinc finger proteins fused to Krüppel-Associated Box repression domains (ZF-Rs). This was achieved via the optimization of a lentiviral system tailored for the delivery of ZF-Rs in hematopoietic cells. We showed that an optimal design of the lentiviral backbone is crucial to multiplex up to three ZF-Rs or two ZF-Rs and a chimeric antigen receptor. ZF-R expression had no impact on the integrity and functionality of transduced cells. Furthermore, gene repression in ZF-R-expressing T cells was highly efficient in vitro and in vivo during the entire monitoring period (up to 10 weeks), and it was accompanied by epigenetic remodeling events. Finally, we described an approach to improve ZF-R specificity to illustrate the path toward the generation of ZF-Rs with a safe clinical profile. In conclusion, we successfully developed an epigenetic-based cell engineering approach for concomitant modulation of multiple gene expressions that bypass the risks associated with DNA editing.

Graphical abstract

David Fenard and colleagues developed a lentiviral backbone for the multiplexing of up to three ZF-R sequences, allowing an efficient, stable, and specific epigenetic control of multiple genes in T cells or Tregs after a single lentiviral transduction event.

Introduction

Epigenetic modifications play a critical role in the regulation of gene expression. They are heritable and reversible changes that affect gene expression without altering the DNA sequence. In mammalian cells, Krüppel-Associated Box (KRAB)-containing zinc finger (ZF) proteins (KRAB-ZFPs) represent a family of more than 400 DNA-binding transcriptional repressors.¹ These domains are crucial for the epigenetic regulation of transposable elements and mammalian evolution.^2,3 The interaction of KRAB-ZFPs with their genomic target locus allows the recruitment of critical co-factors like KAP1/TRIM28⁴, consequently inducing the epigenetic transcriptional silencing of the target gene through various mechanisms, including DNA methylation and histone modifications. The latter involves the addition or removal of chemical groups to the histone tails (e.g., methylation, deacetylation), which can influence the accessibility of DNA to the transcriptional machinery.⁵

To specifically target a KRAB domain to a locus of interest in the genome, different DNA-binding modalities have been developed.⁶ They can be derived from the CRISPR-Cas9 system, in which the dead Cas9 protein is fused to a KRAB domain (dCas9-KRAB) and requires co-expression of a guide RNA capable of interacting specifically with the target locus.^7,8 The DNA-binding protein can also consist of a transcription activator-like effector protein fused to a KRAB domain (TALE-KRAB).^9,10 Finally, the DNA-binding domain can be derived from an engineered ZFP, containing multiple zinc fingers in a single protein, a technology developed nearly 30 years ago.^11,12,13 The last generation of engineered ZFP has the advantage (1) to target virtually any DNA sequence in the genome,¹⁴ (2) to be very compact, (3) to be an all-in-one technology, harboring DNA-binding capacity and KRAB domain on the same protein, and (4) to be of human and not bacterial origin, decreasing the chance of a pre-existing or induced immune response for therapeutic applications.^15,16,17 Engineered ZFP-KRAB repressors (ZF-repressors [ZF-Rs]), allow the precise repression of any target gene.^18,19,20,21 ZF-Rs have been developed in adeno-associated virus (AAV) gene therapy approaches targeting the central nervous system for the treatment of tauopathies²⁰ or Huntington’s disease.²¹ While the AAV delivery system is adapted for non-dividing cells like neurons, this episomal delivery of the ZF-Rs is not suitable for constitutive expression in highly dividing hematopoietic cells. For the latter, the use of third-generation self-inactivating HIV-1-derived lentiviral vectors (LVs) pseudotyped with the vesicular stomatitis virus envelope glycoprotein (VSV-G) is more adapted and extensively employed in ex vivo hematopoietic cell therapies, allowing durable responses with a favorable safety profile.^22,23,24 Furthermore, the lentiviral packaging capacity extends the multiplexing capacity of ZF-Rs compared with AAV. In this study, we optimized a lentiviral backbone design for an efficient multiplexing of up to three ZF-R sequences, allowing stable epigenetic control of multiple genes in T cells or regulatory T cells (Tregs) after a single lentiviral transduction event.

Results

ZF-R design and characterization

To evaluate the ZF-R technology efficiency, multiplexed in a single lentiviral particle, four human genes were selected for their potential therapeutic application. These genes are the beta 2-Microglobulin (B2M) and the class II major histocompatibility complex transactivator (CIITA), which regulate cell surface expression of the major histocompatibility complex class I (MHC-I) and class II (MHC-II), respectively, and are often knocked out in allogeneic T cell products to prevent rejection or graft-versus-host disease (GvHD)^25,26; the CD3zeta (CD3Z) gene, a component of the T cell receptor (TCR)/CD3 complex, essential for its expression and signaling^27,28; and the CD5 gene, a target repressed in cell therapy expressing an anti-CD5 chimeric antigen receptor (CAR) to prevent fratricide.²⁹ For each target gene, approximately 200 research-grade ZFPs targeting the flanking regions of the transcriptional start site (TSS) were designed and linked to the human KOX1 KRAB repressor domain. The repression efficiency of these ZF-Rs was evaluated in human T cells 2 days after ZF-R mRNA electroporation by quantifying target gene mRNA expression using RT-qPCR. As shown in Figure 1, highly efficient ZF-Rs for all four target genes were identified in these screens. The targeted B2M locus showed that only a small set of ZF-Rs targeting the exon 1 region was highly efficient (Figure 1A), highlighting the importance of designing an extensive library of ZFPs to target efficiently a large portion of the TSS region. The ZF-Rs targeting CD5, CD3zeta and CIITA (Figures 1B–1D) exhibited a potent repression across a wider genomic region neighboring the TSS. In later screens, alternative KRAB domains⁷ were tested and several of them (e.g., human ZIM3, ZNF324) harbored an equal or improved repression activity compared with KOX1 (Figure S1). Finally, in line with the transient nature of mRNA transfection experiments, the transcript expression levels of target genes returned to basal levels within a few days (Figure S2), confirming that stable ZF-R expression in the rapidly dividing T cells is required for durable gene silencing.

Figure 1.

Screening of functional B2M, CD5, CD3z, and CIITA ZF-Rs in primary T cells

Schematic of the B2M (A) CD5 (B), CD3z (C), and CIITA (D) target genes with ZF-R candidates binding near the TSS. Triangles represent ZF-R binding locations on the target gene and the red color intensity correlates with repression efficiency evaluated by RT-qPCR two days after ZF-R mRNA transfection in T cells. The triangle orientation indicates the binding orientation of ZF-Rs. The selected lead ZF-Rs are highlighted by a yellow star.

Bidirectional LV design for efficient ZF-R multiplexing

To stably integrate ZF-R sequences in the genome, we set out to generate a lentiviral delivery system. We initially aimed for an efficient, easily interchangeable LV design allowing functional co-expression of up to two ZF-Rs, and a truncated version of the nerve growth factor receptor (dNGFR) to monitor and enrich transduced cells. Strategies to multiplex transgenes can be based on the use of picoviral self-cleaving 2A peptides.^30,31,32 Despite some functionality alone in a bicistronic setting, 2A peptides have shown reduced efficiency when combined within a multicistronic cassette, resulting in a gradual loss of transgene expression toward the end of the 2A-linked peptide chain.^31,33 To mitigate this potential limitation, we generated a bidirectional LV architecture, a typical approach in LV backbone design.^34,35 It consists of an antisense cassette expressing the dNGFR protein and a sense cassette accommodating codon-diversified ZF-Rs, targeting B2M and CD5 genes either in simplex or duplex format connected by a self-cleaving Thosea asigna virus 2A sequence (Figure 2A). Diverse KRAB domains, such as the transcriptional repressor domain of KOX1, ZNF324, and ZIM3,⁷ were used for ZF-R design to avoid repeated sequences that could be deleterious for LV production and infectivity upon multiplexing (Figure S3). Furthermore, the human phosphoglycerate kinase (PGK) and elongation factor 1 alpha (EF1a) promoters have been selected to form bidirectional transcriptional units, as described previously.³⁵ To identify the desirable lentivector design ensuring robust repression of the indicated ZF-R targets, we first compared two different LV architectures, each harboring the described transcriptional units in inverted orientation with respect to each other (pLV.PGK-EF1a and pLV.EF1a-PGK). Following T cell transduction, both LV architectures resulted in a comparable genomic integration capacity, as shown with the vector copy number per cell (VCN/cell) regardless of the cargo (Figure 2B, left). Consistently, dNGFR-positive cells were detected nearly at the same frequency, indicating that the orientation of the transcriptional units did not impact the transduction efficiency (Figure 2B, right). ZF-R-mediated gene repression was next assessed by flow cytometry on dNGFR-enriched cells at day 10 after transduction. Monocistronic sense cassettes resulted in high percentages (>90%) of CD5(−) or MHC-I(−) cells, and no significant differences were observed when comparing the two LV architectures (Figure 2C). In contrast, bicistronic sense cassettes were less efficient in co-repressing MHC-I and CD5, with the EF1-PGK bidirectional promoter resulting in a dramatically reduced percentage of CD5(−)/MHC-I(−) cells as compared with its inverted counterpart (19.4% vs. 62.6%, respectively). These findings were further corroborated by RT-qPCR, revealing a significantly more pronounced decrease of CD5 and B2M mRNA transcripts when ZF-R expression was driven by the EF1a promoter, rather than PGK (Figure S4). To further optimize this LV design, the PGK promoter was replaced with a previously reported shorter version, namely PGK200,³⁶ with the intent to decrease cargo size and improve infectivity (Figures 2A and S3). This PGK200-EF1a backbone improved LV infectivity as shown with the slight increase in VCN/cell, as well as the higher percentages of dNGFR positive cells compared with a PGK-EF1a backbone (Figure 2B). In addition, constructs harboring the bicistronic sense cassette showed a striking increase in the frequency of CD5(−)/MHC-I(−) cells (>80%), suggesting that the indicated modification did not only improve transduction (Figure 2B), but also repression efficiency (Figure 2C) and with nearly no impact on dNGFR expression levels (Figure S5). To repress three different genes simultaneously with a single LV particle, the dNGFR sequence was replaced with a ZF-R sequence targeting the CD3z promoter (Figures 2D and S3). In line with previous studies showing that CD3z is required for the cell surface expression of the TCR/CD3 complex,^27,28 this triple ZF-R construct resulted in the concomitant downregulation of MHC-I, CD5, and the TCR/CD3 complex (Figures 2D and S6). Altogether, these data defined the pLV.PGK200-EF1a architecture as an LV backbone of choice for efficient ZF-Rs multiplexing.

Figure 2.

LV backbone design for efficient ZF-R multiplexing

(A) Schematic representations of bidirectional LV backbones harboring two promoters, either EF1a-PGK, PGK-EF1a, or PGK200-EF1a, to co-express the dNGFR marker and one or two ZF-Rs. The different bidirectional LV backbones express either no ZF-R (CTRL), a single ZF-R (CD5 ZF-R or B2M ZF-R) or a combination of CD5+B2M ZF-Rs. (B) Transduction efficiency of the indicated LV vectors was assessed by measuring the VCN/cell (left) and the percentage of dNGFR⁺ cells by flow cytometry (right) at D4 post-transduction. Data are presented as mean ± SEM. White circles indicate data obtained from each T cell donor (n = 4–6). (C) CD5 and MHC-I repression in dNGFR-enriched T cells transduced with the indicated LV vector was assessed by flow cytometry at day 10 after transduction. Data are represented as the mean ± SEM (n = 4–6 donors). (D) Representative plots showing concomitant MHC-I, CD3, and CD5 repression in T cells transduced with the triple ZF-R lentivirus at day 12 after transduction. The Mann-Whitney test was performed and only statistically significant differences were indicated: ∗p ≤ 0.05; ∗∗p ≤ 0.01.

Detection of a repressive epigenetic profile following the ZF-R-mediated silencing of the B2M gene

To interrogate the chromatin-modifying activity of the B2M ZF-R, we compared the epigenetic status of the B2M locus in T cells transduced with a bidirectional LV vector expressing either B2M-ZF-R or no ZF-R as a control. Binding of the activated RNA polymerase II was first evaluated by chromatin immunoprecipitation followed by qPCR (ChIP-qPCR) using an antibody targeting the phosphorylated form of the enzyme at serine 5 (RNAP II-S5P), which is normally associated with the initiation of RNA synthesis. Indeed, control cells showed a strong enrichment of RNAP II-S5P that peaked in proximity to the TSS (Figure 3A). In contrast, B2M-ZF-R-expressing T cells exhibited a dramatic decrease in the RNAP II-S5P signal near background levels, thus correlating with a strong MHC-I protein downregulation. Furthermore, because KRAB domains cooperate with chromatin-remodeling factors such as histone lysine methyltransferases, we also interrogated the levels of histone 3 lysine 9 trimethylation (H3K9me3) on the B2M locus. Consistently, the loss of RNAP II-S5P in ZF-R-expressing T cells was accompanied with the acquisition of this repressive mark, whose enrichment also peaked around the TSS (Figure 3B). Within the TSS region, we did not detect the enrichment of another known repressive mark, namely, H3K27Me3, but we observed the concomitant loss of H3K4Me3, an epigenetic modification linked to active transcription (Figure S7). Overall, these data indicate that the silencing induced by the B2M ZF-R occurs through the epigenetic modulation of the local chromatin.

Figure 3.

Epigenetic repression profile following the ZF-R mediated silencing of the B2M gene

ChIP-qPCR analysis for the phosphorylated form of RNA Pol II at serine 5 (RNAP II-S5P) (A) and H3K9me3 epigenetic mark (B) performed on the B2M gene of control and B2M ZF-R-expressing T cells and on the unrelated GAPDH gene (right histograms). Data show enrichment events per 1,000 cells (n = 3 independent T cell donors). Data are represented as mean ± SEM.

B2M and CD5 co-repression efficiency and durability in ZF-R-expressing T cells

To monitor the durability of the ZF-R-mediated epigenetic repression, ZF-R-expressing T cells were injected in an immunodeficient mice model and monitored for 10 weeks in vivo (Figure 4A). ZFR-expressing T cells were enriched based on dNGFR expression (>93%) and expanded for 2 weeks in vitro, showing a stable MHC-I and CD5 repression (Figures S8A and S8B), and a stable CD4/CD8 T cell ratio (Figure S8C). Then, these MHC-I(−) and/or CD5(−) T cells were injected intraperitoneally into NOD-Prkdc^scid-IL2rg^Tm1/Rj (NXG) mice and human T cell engraftment (hCD45⁺) was monitored every week in the mouse peripheral blood (gating strategy Figure S9). As expected, without any human cytokine support like interleukin-2, engraftment slowly decreased over time for the control and B2M ZFR-expressing T cells (Figure 4B). In contrast, T cells negative for CD5 were able to proliferate efficiently (Figure 4B), highlighting in this setting the negative role of CD5 on T cell activation/proliferation as described previously.^37,38,39 Interestingly, MHC-I and/or CD5 downregulation was very stable during the entire experiment, independent of the proliferation status (Figure 4C). Importantly, ZF-R-expressing T cells isolated from the spleen or the murine blood (week 10) displayed the same repression profile as before injection (Figure 4D). These data suggest that the LV is an optimal delivery system for sustained expression of multiplexed ZF-Rs, allowing durable and concomitant repression of multiple genes.

Figure 4.

Efficient and durable MHC-I and CD5 repression in human ZF-R-expressing T cells engrafted in vivo

(A) Schematic of the in vivo experimental protocol. (B) Monitoring of the engraftment efficiency of hCD45⁺ cells during 10 weeks. (C) Durability of MHC-I and/or CD5 downregulation in ZF-R-expressing T cells injected in NXG mice (n = 5/group). (D) Representative fluorescence-activated cell sorting (FACS) plots showing MHC-I and CD5 downregulation in dNGFR⁺ T cells before injection and 10 weeks after injection in blood and spleen. On the right panel, the histogram represents the percentage of concomitant MHC-I(−)/CD5(−) T cells in spleen (n = 5 mice). Data are represented as mean ± SEM.

ZF-R-mediated downregulation of MHC-I and MHC-II in CAR-expressing Tregs

To extend the evaluation of the ZF-R technology to different T cell subsets, ZF-Rs were delivered in CAR-expressing Tregs (CAR-Tregs).⁴⁰ Tregs were transduced with a vector expressing a CAR directed against the human HLA.A2 protein⁴¹ and two ZF-Rs directed against B2M and CIITA to downregulate MHC-I and MHC-II expression, respectively (Figure S3), an approach previously described to produce off-the-shelf hypoimmunogenic CAR-T cells.^25,26 As shown in Figures 5A and 5B, concomitant repression of B2M and CIITA genes led to the generation of approximately 70% of CAR-Tregs with no cell surface expression of MHC-I and MHC-II molecules. Constitutive expression of these ZF-Rs had no impact on CAR-Treg activation (i.e., CD69 expression), neither after stimulation through the TCR/CD3 complex using anti-CD3/CD28 beads nor through the CAR using a fluorescent dextramer coated with HLA.A2 proteins (Figure 5C). Furthermore, the phenotype of ZF-R-expressing CAR-Tregs remained stable in vitro (Figure 5D), strongly suggesting that the integrity and functionality of primary cells is not impacted by the constitutive expression of ZF-Rs per se.

Figure 5.

ZF-R-mediated downregulation of MHC-I and MHC-II in CAR-Tregs

(A) FACS plots of CAR-Tregs transduced with the CAR alone (CAR) or multiplexed with B2M and CIITA ZF-Rs showing co-repression of MHC-I and MHC-II expression at day 7. (B) Percentage of double-negative MHC-I and MHC-II cells measured at day 12. (C) Expression of the CD69 activation marker was assessed by flow cytometry one day after incubation with either anti-CD3/CD28 beads to stimulate the TCR (positive control) or with Dextramer-A2 to stimulate the HLA-A2 CAR. (D) The expression of FoxP3 and Helios Treg markers was assessed at day 7 via intracellular staining after cell surface staining and gating on CD4⁺, CD25⁺. and CD127low-positive cells. Histograms represent the mean ± SEM. All the data were obtained from three independent T cell donors.

Assessment of the ZF-R specificity

Despite high efficiency and a favorable safety profile of the research-grade ZF-Rs tested in this study (i.e., no impact on cell viability, expansion, phenotype, and functionality), non-specific interactions of ZF-Rs to genomic DNA (gDNA), described as off-target sites, can occur. Microarray analyses were performed to identify the number of deregulated genes (DEGs) in T cells transduced with either the parental B2M ZF-R LV (Figure 6A) or the CD5 ZF-R LV (Figure S10), the two ZF-Rs highly potent in long-term in vivo monitoring. The results highlighted a strong repression of the B2M and CD5 genes with a limited number of other DEGs. The development of clinical grade ZF-Rs will require high on-target specificity and nearly undetectable off-targets. We have previously shown that an arginine residue in each zinc-finger DNA binding domain makes a non-specific contact with the DNA phosphate backbone; the mutation of this arginine to glutamine, known as the phosphate-contact variant (PCV), can modulate DNA binding affinity and reduce the off-target binding of zinc finger nucleases.⁴² To exemplify this strategy on the research-grade B2M ZF-R, PCV variants of the parental B2M ZF-R were generated. The genomic sequence of the more significant DEGs were aligned with the on-target sequence, either fully (18 consecutive base pairs) or partially. Only the PMVK gene presented a significant sequence homology to the B2M on-target sequence (Figure 6B), with 12 consecutive base pairs, covering the first (ZF-1 and -2) and the second two-finger module (ZF-3 and -4). Therefore, one PCV mutation was incorporated in either ZF-2 (PCV2) or ZF-4 (PCV4) to decrease the affinity of one of the two modules implicated in this off-target. As shown in Figure 6C, microarray analysis of T cells expressing the PCV2 variant showed a strong improvement in the specificity profile and these results were confirmed by RT-qPCR (Figure S11). The PCV4 variant was excluded because its specificity profile was still harboring two DEGs. Of note, only one DEG persisted in the presence of the PCV2 variant, namely, PATL2, a gene located only 200 bp upstream of the B2M locus (see Figure S12). RT-qPCR analysis showed a partial repression of PATL2 (approximately 65%). This deregulation of the PATL2 gene could be the consequence of an on-target effect of the B2M ZF-R since the localization of the H3K9Me3 epigenetic mark is extended by approximately ±2 kb of the target site, as shown in Figure 3B. Finally, despite a decrease in the non-specific affinity of PCV2 and PCV4 for the gDNA, the on-target interaction and B2M repression was still highly efficient, leading to more than 90% of MHC-I downregulation (Figure 6D).

Figure 6.

Design of PCVs of B2M ZF-R to improve specificity

(A) Volcano plot representation of pooled RNA array data from dNGFR-enriched T cells (n = 3 donors) transduced with parental B2M ZF-R (n = 12) compared with empty LV vector (n = 6). Threshold for significant deregulation (dotted lines) at false discovery rate (FDR)-adjusted p values of less than 0.05. The B2M target gene is marked with a yellow dot, other downregulated (red dots) or upregulated genes (green dots) more than 2-fold are annotated. (B) Schematic of the sequence homology between B2M target DNA sequence and the deregulated PMVK gene. Zinc finger proteins with Arg-to-Gln PCV mutation are highlighted with a yellow star. (C) Volcano plot representation as in (A) with PCV2 and PCV4 B2M-ZF-R variants (n = 12) compared with empty LV vector (n = 6). (D) Efficiency of MHC-I downregulation in dNGFR-enriched T cells in quadruplicate (n = 3 donors) expressing parental, PCV2, and PCV4 B2M ZF-R.

Discussion

In this study, we sought to design a versatile engineering strategy allowing concomitant and stable silencing of multiple genes in the same target cell. The development of engineered human-derived ZF-Rs for the epigenetic repression of target genes has the advantage of mirroring endogenous ZFP-KRABs already operating on the human genome. Furthermore, ZF-Rs lack the genotoxicity associated to nuclease-mediated gene silencing (i.e., DNA double-strand breaks, translocations, etc.). For these reasons, this gene modulation technology holds promising therapeutic applications, as described previously for either KRAB domains fused to ZFP or other DNA-binding modalities.^{18,19,20,21,43}

During the ZF-R screening procedures, we observed that the number of highly active ZF-Rs targeting the TSS region of a given gene was variable. B2M and CD5 ZF-R screenings led to a small or high number of effective ZF-Rs, respectively, while both genes are highly active in T cells. High chromatin accessibility for multiple CD5 ZF-Rs can be the consequence of low nucleosome occupancy, or high nucleosome occupancy with low stability and/or an increase in inter-nucleosomal spacing.^44,45,46 Due to this chromatin complexity and unknown nucleosome pattern for a given gene in human T cells, the strategy that we applied for the ZF-R screenings was always to design a large number of ZFPs targeting the TSS and the surrounding regions (±0.5 kb).

The intracellular co-delivery of multiple epigenetic silencers for therapeutic applications can be achieved with HIV-1-derived LVs. They are well adapted for ex vivo engineering of hematopoietic cells, allowing permanent expression of silencers. Deployment of a single lentiviral particle to control multi-gene expression gives major advantages, including (1) large cargo capacity, (2) concomitant delivery of all epigenetic repressors in each transduced cell, and (3) greater simplicity in the process development and consequently a cheaper manufacturing cost. Nonetheless, the cloning design for multiple transgenes into the proviral DNA remains a challenge. Here, we showed that the bidirectional architecture is well adapted to express up to three ZF-Rs. Importantly, the type and the orientation of the two promoters are critical. Generation of MHC-I and CD5 co-repressed T cells was improved from 20% to 80% just by swapping the dual promoter architecture from EF1a-PGK to PGK200-EF1a. The EF1a promoter, known to be stronger than PGK, allows a better expression of the ZF-Rs. However, we also observed that the EF1a promoter in sense orientation was negatively impacted by the proximity of the antisense PGK promoter, a side effect strongly attenuated by shortening the PGK promoter from 500 to only 200 bp (PGK200). Based on these data, we hypothesized that a competition for common transcription factors between PGK and EF1a may occur.

Next, we observed that the constitutive expression of ZF-Rs after lentiviral integration led to stable gene repression during the 2-week period of culture in vitro. To extend this monitoring, ZF-R-expressing T cells were injected into immunodeficient mice and showed again a strong capacity to silence target genes during the entire follow-up period of 10 weeks. This silencing is driven by the deposition of the H3K9me3 repressive mark, correlating with a concurrent loss of the H3K4me3 active mark. Notably, the H3K27me3 repressive mark was undetectable. Indeed, the H3K9me3 mark is known to be mediated by the KRAB complex, while H3K27me3 is primarily mediated by the EZH2 methyltransferase.⁴⁷ Furthermore, this epigenetic silencing following ZF-R lentiviral delivery was not modulated by the cell proliferation status, suggesting that the epigenetic marks are constantly deposited at each proliferation cycle. Indeed, during the ZF-R screening with transient mRNA expression, we showed that the gene repression was lost in proliferating T cells after few days of culture.

To extend our evaluation of the ZF-R multiplexing technology, we designed a next-generation CAR-Treg approach. Converting polyclonal Tregs into antigen specific Tregs by introducing a CAR gained significant attention as a potential treatment option for autoimmune diseases and solid organ rejection.^40,41 We engineered CAR-Tregs co-expressing ZF-Rs targeting B2M and CIITA genes, both required for cell surface expression of MHC-I and MHC-II, respectively.^25,26 The ZF-R-mediated repression for both was potent and stable. The constitutive expression of B2M and CIITA ZF-Rs per se and the consequent downregulation of MHC-I and MHC-II did not impact either the CAR functionality or the Treg phenotype, expanding the field of application of ZF-Rs to diverse hematopoietic cells.

Finally, we shared a strategy to optimize the specificity of research grade ZF-R to generate potential clinical-grade candidates. We showed that human T cell transduction with the parental B2M or CD5 ZF-Rs, very efficient for long-term stable repression in vivo, led to some DEGs as shown in microarray analysis. These genes could be true off-targets or biologically co-DEGs, downstream of the target gene. As an example, going through gene ontology studies, we did not find any link between B2M expression and the identified DEGs. We also performed an in silico analysis of the DEGs for any homology with the on-target sequence and found only one off-target candidate, namely, the PMVK gene, harboring 12 consecutive base pairs fully homologous to the B2M on-target site. Based on this homology, a PCV mutation was incorporated into the zinc finger 2 (module A) and/or zinc finger 4 (module B) of the parental ZF-R. While PCV2 and PCV4 maintained a strong repression of B2M expression, the double mutant PCV2-4 was way less efficient to repress B2M (approximately 30%) and was, therefore, excluded from the study (data not shown). The PCV4 B2M ZF-R was also excluded due to a remaining DEG (i.e., TRIM47), while the PCV2 B2M ZF-R mutant showed a dramatic decrease for all DEGs in the microarray analysis and in RT-qPCR validation assays. Altogether, these data confirm that the disruption of the non-specific interaction of the zinc finger with the DNA phosphate backbone is a powerful and essential engineering step for increasing ZFP specificity to a level required for clinical grade ZF-Rs.⁴² Interestingly, both the parental and the PCV2 B2M ZF-R resulted in some deregulation of the surrounding PATL2 gene, located only 200 bp upstream of the B2M promoter (Figure S9). While B2M was strongly repressed (>90%), the PATL2 gene was partially repressed (65% as compared with control). Due to its close proximity to B2M, PATL2 gene repression could potentially be an on-target effect of the B2M ZF-R. This slight deregulation of the closest neighboring gene is probably the consequence of a halo effect of the KRAB domain, a phenomenon previously observed for the PTMS gene following the repression of the LAG3 gene located approximately 1 kb downstream.⁴⁸ It is important to note that the modulation of expression of PATL2 did not have any impact on T cell phenotype, viability, expansion, or B2M repression stability. This is consistent with the absence of a functional role of the PATL2 gene in the biology of human T cells. Nonetheless, it highlights the importance of monitoring the expression of surrounding genes during the development of clinical grade repressors, either ZFP-KRAB, dCas9-KRAB, or TALE-KRAB, to evaluate any potential halo effect of the KRAB domain along the few kilobases of the target gene locus.

In conclusion, we were able to concomitantly deliver multiple functional epigenetic repressors in a single target cell. Considering that ZFP binding domains can also be fused to compact transcriptional activation domains,^11,49 this cell engineering approach opens the way to innovative therapeutic applications requiring concomitant transgene expression (e.g., CAR, TCR) and modulation of expression of multiple genes with a performant and cost-effective single lentiviral transduction event.

Material and methods

Human primary cell and cell line culture

For screening of ZF-Rs, human T cells were obtained from fresh Leukopaks purchased from StemCell Technologies (Seattle, WA, USA). T cell isolation was performed on the CliniMACS system using anti-CD4 and CD8 magnetic beads (Miltenyi Biotec, San Jose, CA, USA). Next, T cells were suspended in CryoStor 10 (StemCell Technologies, Seattle, WA, USA), aliquoted, and stored in liquid nitrogen. After thawing, T cells were cultured in complete X-vivo15 (Lonza, Hayward, CA, USA), supplemented with 100 IU/mL IL-2 (Thermo Fisher Scientific Waltham, MA, USA), 5% human AB serum (Valley Biomedical, Winchester, VA, USA), and activated with anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific) at a bead:cell ratio of 1:3. For other experiments, primary cells were isolated from buffy coats of healthy donors (EFS, Marseille, France). Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll gradient centrifugation. T cells and Tregs were isolated by immunomagnetic selection. Briefly, T cells were isolated using the EasySep Human T cell isolation kit (StemCell Tech., Saint-Egrève, France) and cultured in complete X-Vivo15 medium. Tregs were isolated using the EasySep human CD4⁺CD127lowCD25⁺ Regulatory T cell Isolation Kit (StemCell Tech) and cultured in X-Vivo15, supplemented with 1,000 IU/mL IL-2, anti-CD3/CD28 dynabeads (1:1 ratio) and 100 nM rapamycin (Sigma-Aldrich, Saint-Quentin-Fallavier, France). Jurkat E6.1 cells (Sigma-Aldrich) were cultured in X-Vivo15 medium.

ZF-R design and screening

Design and assembly of ZF-Rs

ZFP backbones, based on the human ZFP Zif268/EGR, were designed using an archive of prevalidated one- and two-finger modules as previously described.⁵⁰ The resulting ZFPs were cloned into the pVAX vector (Thermo Fisher Scientific) to produce a fusion with KRAB repressor domains described in this study.

ZF-R mRNA production and transfection

Templates for in vitro transcription were generated by PCR from pVAX-ZFP-KRAB plasmids (forward primer: GCAGAGCTCTCTGGCTAACTAGAG; reverse primer: T(60)CTGGCAACTAGAAGGCACAG). mRNA transcripts were synthesized using the mMESSAGE mMACHINE T7 ULTRA Transcription Kit (Thermo Fisher Scientific) as per the manufacturer’s instructions and purified using the Agencourt RNAClean XP Kit (Beckman Coulter, Brea, CA, USA) on a Kingfisher instrument (Thermo Fisher Scientific). T cells were cultured in complete X-Vivo15 medium for three days prior to mRNA transfection. ZF-R mRNA transfection was performed using 1–10 μg ZFP mRNA per 2 × 10E5 cells on a BTX Multi-well Electroporation 96-well plate with a 2-mm well gap with a BTX ECM 830 Electroporator with Plate Handler HT-200 (BTX Harvard Apparatus, Holliston, MA, USA). The transfection percentage was greater than 90% as determined from co-transfection with a control GFP-encoding mRNA. Forty-eight hours after transfection, cells were lysed and analyzed for gene expression by RT-qPCR.

Gene expression analysis using RT-qPCR

mRNA samples from ZF-R-transfected T cells were reverse transcribed to cDNA using a Power SYBR Green Cells-To-CT kit (Thermo Fisher Scientific) and qPCR reactions were performed using QuantiFast Multiplex PCR Master Mix (w/o ROX) (Qiagen, Redwood City, CA, USA). Each biological replicate was assessed in technical quadruplicate on a 384-well CFX real-time qPCR instrument (Bio-Rad, Hercules, CA, USA) and analyzed using the Bio-Rad CFX Manager 3.1 software. The housekeeping genes EIF4A2 and ATP5B were used for RT-qPCR normalization. The RT-qPCR probe/primer sets used are listed in Table S1.

LV production and titration

LVs were produced using the classical 4-plasmid lentiviral system. Briefly, HEK293T cells (Lenti-X, Ozyme, France) were transfected with plasmids expressing HIV-1 Gag/pol (pMDLg/pRRE), HIV-1 Rev (pRSV.Rev), the VSV-G (pMD2.G) (Didier Trono, EPFL, Geneva, Switzerland), and a third-generation transfer plasmid expressing the transgene. At 24 h after transfection, viral supernatants were harvested, concentrated by centrifugation, aliquoted, and frozen at −80°C for long-term storage. Infectious titers expressed in transducing units per milliliter (TU/mL) were obtained after transduction of Jurkat T cells with a serial dilution of viral supernatants and transduction efficiency evaluated after 3–4 days by monitoring cell surface dNGFR expression using an anti-CD271-APC antibody (Miltenyi Biotec, Paris, France) or HLA-A∗02 Dextramer-APC (Immudex, Copenhagen, Denmark) for HLA.A2 CAR.

Lentiviral transduction and dNGFR enrichment

Lentiviral constructs used in this study are described in Figure S2. Lentiviral transduction was performed two days after T cell or Treg isolation and activation. Briefly, cell suspension (2 × 10⁶ cells/mL) were incubated with LV supernatants at a final concentration of 4 × 10⁷ TU/mL in complete X-Vivo15 medium. After 6 h at 37°C and 5% CO₂, cell suspensions were diluted eight times with fresh complete medium. After 3–5 days, transduction efficiencies were measured using flow cytometry to determine the percentage of dNGFR- or CAR-positive cells. For dNGFR enrichment, transduced cells were isolated with the EasySep Human CD271 Positive Selection Kit II with EasyEights EasySep Magnets (StemCell Technologies) according to the manufacturer’s protocol.

In vitro Treg activation assay

In the activation assay, Tregs were stimulated for 24 h through the TCR (positive control) with anti-CD3/CD28-coated dynabeads (1 bead:2 cells ratio) (Thermo Fisher Scientific) or through the HLA.A2-CAR with 2.5μL HLA-A2∗02 dextramer (Immudex). Cells were then harvested and analyzed by flow cytometry.

VCN and mRNA quantification in LV-transduced T cells

The VCN/cell was measured by qPCR. gDNA was extracted using the DNeasy Blood and tissue kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. To determine the VCN, amplification of the human albumin (ALB) gene and the HIV-1 Psi vector sequence were performed using respectively custom qPCR HEX assay and custom qPCR FAM assay (Bio-Rad, Marnes-la-Coquette, France) with previously published primer sequences.⁵¹ qPCR reactions were performed using 200 ng gDNA and the Master Mix Taqman Universal II (Applied Biosystems, Villebon-sur-Yvette, France), according to the manufacturer’s recommendations. All reactions were performed using the CFX Opus 96 Dx System (Bio-Rad). A standard curve was used to convert cycle threshold values to copy numbers by amplifying serial dilutions of a unique plasmid containing equimolar ratios of the ALB and PSI target sequences.

Expression levels of indicated mRNA transcripts in LV-transduced cells were analyzed by RT-qPCR. Total RNA was extracted from dNGFR-enriched cells using the RNAqueous total RNA isolation kit (Ambion, Austin, TX, USA) and subsequently treated with Turbo DNA free-kit (Ambion). Reverse transcription was performed with Verso cDNA Synthesis kit (Thermo Fisher Scientific). Gene-specific primers and probes are described in the Table S1 and S2. Experiments were carried out in duplicate for each data point. The relative quantification in gene expression was determined using the 2-ΔΔCt method.⁵² Using this method, fold changes in gene expression were normalized to the housekeeping GAPDH mRNA.

ChIP-qPCR assay

ChIP-qPCR was performed using the ChIP-IT PBMC kit (53042; Active Motif, Carlsbad, CA, USA) according to manufacturer’s instructions. Briefly, NGFR-enriched T cells from three different healthy donors (purity >90%) were harvested and fixed with 1% formaldehyde for 10 min, lysed, and prepared for sonication. gDNA was fragmented to a mean length of 300–1,500 bp using the Epishear probe sonicator (Active Motif) and was controlled on 1.5% agarose gel. Next, immunoprecipitation with 12 μg fragmented chromatin was performed using the following antibodies from Abcam (Cambridge, UK): anti-H3 (ab1791), anti-RNA polymerase II CTD repeat YSPTSPS phospho S5 (ab5131), anti-H3K4me3 (3690, Active Motif), anti-H3K27me3 (ab6002), anti-H3K9me3 (ab8898), and an unrelated IgG control (ab171870). qPCR reactions were performed in duplicate to detect specific genomic regions using SYBR Green (BioRad) and specific primers (Table S3). Results were analyzed with the ChIP-IT qPCR Analysis kit (53029; Active Motif) to calculate binding events detected per 1,000 cells. Pol II phospho S5 and H3K9me3 signals were normalized with H3 signals performed with anti-H3 antibody on the same chromatin.

NXG mouse model for in vivo monitoring of human ZF-R-expressing T cells

NXG mice were obtained from Janvier Labs (Le Genest-Saint-Isle, France). Eight-week-old NXG mice were housed in containment isolators and habituated for 1 week prior to experimental use. The animal protocol was approved by the French animal ethics committee. Mice were randomly assigned to groups. For in vivo experiments, T cells were transduced at D2, dNGFR-enriched at D6, and injected intraperitoneally at D12 (3 × 10⁶ T cells per mouse). Throughout the experiments, body weight and GvHD score were monitored twice a week by operators blinded to treatment. Once a week, blood samples were collected by retro-orbital sampling under local anesthesia. At the end of the experiment, mice were euthanized and tissue and blood samples were collected and processed for immunophenotyping.

Human primary cells immunophenotyping

For cellular immunophenotyping, T cells were stained with conjugated monoclonal antibodies (mAb) targeting CD3, CD4, CD8, CD5, and CD271/NGFR (Miltenyi Biotec) MHC-I and MHC-II (BD Biosciences, Le Pont de Claix, France). For HLA-A2 CAR-Treg labeling, cells were stained at cell surface with conjugated mAb targeting CD4, CD25, and CD127 (Miltenyi Biotec) and the CAR was detected after incubation with a conjugated HLA-A2∗02 dextramer (Immudex). Following fixation/permeabilization, Tregs were stained with conjugated mAb targeting the intracellular Helios (eBiosciences, Life Technologies Corp., Carlsbad, CA, USA) and FoxP3 proteins (BD Biosciences). For in vivo experiments, mouse spleen and bone marrow samples were passed through a 70-μm cell strainer to obtain a single cell suspension. Red blood cells were lysed with red blood cell lysis buffer (Merck, Fontenay-Sous-Bois, France). Next, cells were incubated with mouse Fc block (BD Biosciences) and stained with the indicated antibodies. Marker expressions were measured by flow cytometry and analyzed using FlowJo software (BD Biosciences).

Microarray studies

T cells were transduced on day 2 with either the B2M-ZF-R LV, the CD5-ZF-R LV or the empty LV control vector. Four days later, ZF-R-expressing T cells were enriched using the dNGFR marker and total RNA extracted on day 12. Total RNA was analyzed for gene expression changes using a microarray analysis (Clariom S Human HT, Thermo Fisher Scientific). Each replicate (100 ng total RNA) was processed according to the manufacturer’s protocols for sample preparation, hybridization, fluidics, and scanning. Robust multi-array average (RMA) was used to normalize the raw signal from each probe set. Fold change analysis was performed using Transcriptome Analysis Console 4.0 (Thermo Fisher Scientific) with analysis type-expression (gene) and summarization-SST-RMA options selected. Gene expression levels in ZF-R-expressing T cells were always compared with T cells transduced with an empty control LV. Change calls are reported for transcripts (probe sets) with a >2-fold difference in mean signal relative to control and an false discovery rate-adjusted p value of less than 0.05 (one-way ANOVA and unpaired t test for each probe set) with n = 19,669 transcript clusters assessed.

Statistical analysis

Statistical analyses were performed using GraphPad Prism v.10.1.0 (GraphPad Software Inc., San Diego, CA, USA).

Data and code availability

All requests for data will be reviewed by Sangamo Therapeutics, Inc., to verify whether the request is subject to any intellectual property or confidentiality obligations. If deemed necessary, a material transfer agreement between the requestor and Sangamo Therapeutics, Inc., may be required for the sharing of some data. Any data that can be freely shared will be released.