Highly efficient transduction of human plasmacytoid dendritic cells without phenotypic and functional maturation

Dendritic cells (DC) are antigen-presenting cells (APC) with a role in controlling the balance between immunity and immunological tolerance [1, 2]. In humans, at least two subsets of DC are known in the blood, myeloid DC (also known as interstitial or dermal DC), and plasmacytoid DC (pDC) and Langerhans cells (LC) in the tissues [3]. Plasmacytoid DC also called "natural interferon producing cells" (NIPC), represent 0.2–0.8% of peripheral blood cells and have also been found in the spleen, bone marrow, tonsils, lymph nodes, foetal liver and thymus [2, 4‐6]. Plasmacytoid DC are well known for their ability to recognize and respond to a variety of viruses [6]. They recognize viral genomic nucleic acids of dsDNA viruses [7‐10] and ssRNA viruses [11‐13] via Toll-like receptor 9 (TLR9) and TLR7, respectively in the acidified endosomes without becoming infected themselves. Plasmacytoid DC are characterized by their high secretion levels of type I interferon in response to viruses [14, 15], which not only have direct inhibitory effects on viral replication, but also can promote the function of natural killer cells, B cells, T cells and myeloid DC [16]. Human pDC do not express lineage specific markers, but are characterized by the expression of HLA-DR, CD4, CD123, BDCA2 and BDCA4 [3]. These scarce cells can be generated from CD34⁺ progenitor cells [17]. Recently, a pDC cell line called GEN2.2 established from leukemic pDC was described as sharing most of the phenotypic and functional features of normal pDC [18] and so represents a good model for study of the physiology of their normal counterpart [19].

Over the classical antigen loading methods usually considered, such as peptide or protein loading, gene modified DC offer potential advantages: 1) they ensure long-lasting expression of the antigen and production of an entire array of epitopes presented by the autologous HLA molecules, 2) antigens are delivered to both endogenous MHC class I and class II antigen presentation pathways [2, 20]. Lentiviral vectors (LV) pseudotyped with the vesicular stomatitis virus envelope glycoprotein (VSVG) are efficient gene delivery vectors for dividing and non-dividing cells and were shown to be applicable to many cell types including human conventional DC and LC [21‐26]. Transduced DC and LC retained their immature phenotype, were able to respond to maturation signals, and maintain immunostimulatory potential in both autologous and allogeneic settings [22, 26, 27]. To our knowledge, the transduction capacity of LV into pDC has not yet been evaluated. LV can be pseudotyped with a variety of envelope glycoproteins [28, 29] such as the gibbon ape leukaemia virus envelope (GaLV) or the feline endogenous virus envelope (RD114) which have been reported to be efficient in the transduction of hematopoietic cells [30‐32]. The elongation factor-1α (EF1α) and phoshoglycerate kinase (PGK) promoters were shown to have an activity in a human CD34⁺ cell and in cultured cord blood cells and transgene-expressing myeloid DC were obtained from them [23, 26, 33, 34].

One of the alternate vectors used to transduce monocytes or DC was the recombinant adeno-associated virus (rAAV) with a genome conventionally packaged as single-stranded molecules (ss) [35‐37], characterized by its ability to transduce both dividing and non-dividing cells. Recombinant AAV is unique among viral vectors that are being developed for gene therapy applications in that the wild-type virus counterpart has never been shown to cause human disease. So far, transduction efficiencies of DC subsets have been shown to be low and variable [36, 38].

In this study, we compared the transduction efficiency into a human pDC cell line and in CD34-pDC, with i) LV pseudotyped with different envelopes encoding E-GFP. In this context, we also tested different promoters: two promoters with high activity in hematopoietic cells (PGK and EF1a) and two promoters described as muscle-specific [39‐41] (C5–12 and desmin) in order to evaluate the promoter leak in pDC, ii) rAAV of different serotypes. We found that efficient gene transfer into pDC can be achieved mainly with VSVG-pseudotyped LV under the control of PGK and EF1 promoters. Surprisingly, promoters described as muscle-specific were also highly active in pDC. Gene transfer into pDC could be of high importance for the design of new DC-based vaccines, or for induction of peripheral tolerance for dedicated therapeutic applications.

The attractiveness of dendritic cells as a target for genetic manipulation is a consequence of their ability to initiate and orchestrate primary immune responses, including tolerogenic responses [1, 45, 46]. At least two circulating subsets of DC have been described: myeloid DC and pDC with evidence of functional differences in their ability to regulate the T-cell responses, to produce antiviral type I IFN and to cross-present exogenous antigens to CD8⁺ T cells [47]. We previously showed that VSVG-pseudotyped HIV-1 vectors are good candidates for efficient transduction of monocyte- and CD34⁺-derived LC, without inducing phenotypic and functional maturation [26]. More recently, we also showed that self-complementary duplex strands but not single strands rAAV2/1 and 2 were also very efficient in transducing major DC subsets generated in vitro, including CD34⁺-derived pDC [38].

In this study, we extended LV transduction to pDC, using different pseudotyped HIV-1 vectors encoding E-GFP under the control of different promoters and showed that VSVG-pseudotyped LV encoding E-GFP under the control of EF1 or C512 promoters are the most efficient combinations, leading to transduction of 60% to 90% of the pDC cell line, GEN2.2 [18] and CD34-pDC. Of note, we showed that transduction did not alter alloreactive presentation properties of pDC. Furthermore, pDC transduced with LV expressing a MART-1 peptide was as efficient as Mo-DC for activation of a specific CD8⁺ T cell clone. Altogether, these results show that antigen-loading of pDC through ex-vivo LV transduction may represent a relevant immunotherapy approach for particular clinical applications. Indeed, compared with antigen loading protocols using whole tumor cell lysates or recombinant tumor-associated antigen peptides, LV transduction offers the advantage of direct antigen processing from cytosolic proteins and of long lasting antigen expression.

Previous publications [30‐32] reported efficient transduction levels of hematopoietic cells with LV pseudotyped with GaLV or RD114 envelopes. Here, the highest pDC transduction levels were obtained with the VSVG envelope, which was also previously shown to efficiently transduce human hematopoietic progenitor and leukaemia cells [26, 48, 49] as well as fully differentiated human monocyte-derived DC [50, 51], with a long lasting expression. The EF1α promoter was shown to have a stronger activity than the PGK promoter in a human CD34⁺ cell line [33] and in cultured cord blood cells [33, 34] and allowed to obtain transgene-expressing myeloid DC [23]. Here, we showed that after a single exposure to VSVG-pseudotyped LV, the percentage of E-GFP expressing pDC was 2.6 fold higher when the expression was driven by the EF1 compared to the PGK promoter. The average copy number of the vector in transduced pDC under both conditions was similar (3–4 copies per cell), as determined by real-time quantitative PCR (data not shown). This indicates that the integration levels are similar with both constructions but that, as previously described, the promoter activity is different. We also evaluated two other promoters described to be muscle restricted [39‐41], the desmin and synthetic C512 promoters which have been shown in gene therapy studies to specifically target muscles and to drive gene expression in a context of ss rAAV vectors [41]. As in our previous report [38], we showed here that even with an ubiquitous promoter like CMV, only a very low transduction efficiency could be reached with ss rAAV in the different DC subsets. So, in order to investigate the potential leak of these promoters in human DC subsets, we constructed and produced LV vectors carrying the two different cassettes. Surprisingly, we showed that the percentages of E-GFP expressing pDC with desmin and C512 promoters were very high and equivalent to those obtained with PGK and EF1 promoters, respectively. The average copy number in pDC for desmin and C512 promoters were 4 and 1 copies per cell, respectively, showing that the C512 promoters was at least as efficient as an ubiquitous promoter (data not shown). In contrast to the desmin promoter, the C512 promoter was also active in monocyte-derived DC and LC (around 10% of E-GFP⁺ cells) and in a human colorectal carcinoma (HCT116) (data not shown). Nevertheless, transgene expression with these cassettes in ss AAV vectors was not detectable (data not shown). Taken together, these data suggest that the use of desmin or C5–12 promoters in ss rAAV, for clinical gene therapy protocols, will not induce transgene expression in DC subsets. Nevertheless, the use of these promoters in sc rAAV, which are highly efficient for transducing major DC subsets [38] might elicit high immune responses against the transgene.