Design of a Novel Integration-deficient Lentivector Technology That Incorporates Genetic and Posttranslational Elements to Target Human Dendritic Cells

Abstract

As sentinels of the immune system, dendritic cells (DCs) play an essential role in regulating cellular immune responses. One of the main challenges of developing DC-targeted therapies includes the delivery of antigen to DCs in order to promote the activation of antigen-specific effector CD8 T cells. With the goal of creating antigen-directed immunotherapeutics that can be safely administered directly to patients, Immune Design has developed a platform of novel integration-deficient lentiviral vectors that target and deliver antigen-encoding nucleic acids to human DCs. This platform, termed ID-VP02, utilizes a novel genetic variant of a Sindbis virus envelope glycoprotein with posttranslational carbohydrate modifications in combination with Vpx, a SIVmac viral accessory protein, to achieve efficient targeting and transduction of human DCs. In addition, ID-VP02 incorporates safety features in its design that include two redundant mechanisms to render ID-VP02 integration-deficient. Here, we describe the characteristics that allow ID-VP02 to specifically transduce human DCs, and the advances that ID-VP02 brings to conventional third-generation lentiviral vector design as well as demonstrate upstream production yields that will enable manufacturing feasibility studies to be conducted.

Introduction

Many pharmaceutical methods are being investigated to harness the potential of the patient's immune system to prevent or treat human diseases. Nowhere is this more apparent than in the field of cancer immunotherapy where the greatest challenge has been to induce anticancer immunity through activation of antigen-specific effector T cells (reviewed in ref. ¹). As sentinels of the immune system, dendritic cells (DCs) play an essential role in regulating cellular immune responses through antigen capture and presentation.² It would therefore be desirable to capitalize on the specific capability of DCs to drive cellular responses toward an antigen of choice when developing a platform for antigen-directed active immunotherapeutics. Delivering a cargo antigen to DCs in order to elicit an immune response has been demonstrated through both ex vivo³ and in vivo routes.⁴ However, current ex vivo methods have multiple limitations which can include the complexities of manufacturing and treatment, the risk of treatment-associated infections, and the overall cost of goods, while current in vivo delivery systems may have consequences associated with inefficient DC tropism and subsequent antigen presentation by DCs as well as undesirable off-target effects.

In order to minimize these limitations, Immune Design is developing a novel class of integration-deficient lentiviral vectors that have been engineered to deliver antigen-encoding nucleic acids to human DCs in vivo. This specificity in targeting is achieved by using a modified version of a Sindbis virus envelope that naturally exhibits tropism toward human DCs via the DC-specific ICAM-3 grabbing nonintegrin (DC-SIGN) receptor.⁵ The ability to target DCs and elicit antigen-specific immune responses has been demonstrated with a prototype of this platform.⁶ This prototype vector, like other third-generation lentiviral vectors, was generated using a split genome encoding the HIV-1 genes gag/pol and rev,⁷ and contains a self-inactivating deletion in the U3 region of the 3′LTR (ΔU3).⁸ Unlike traditional LVs that use the pan-tropic VSV-G envelope glycoprotein, this prototype utilizes a mutated Sindbis virus envelope (SVGmu) to target DCs via DC-SIGN with reduced binding to heparan sulfate.⁶

However, this prototype vector has several limitations, which include the inability of integration-deficient versions to be produced at high enough yields that would render upstream production feasible for commercial applications. Immune Design has significantly advanced this prototype vector system by the addition of several design elements to generate our novel ID-VP02 platform. These elements include two independent features that are engineered to eliminate integrase-dependent integration events, a novel envelope glycoprotein, termed E1001, that increases vector production yields and via posttranslational modification increases the ability of the vector to utilize human DC-SIGN as a receptor, and the inclusion of the accessory protein, Vpx from SIVmac, to increase the transduction efficiency of human DCs. In this report, we describe the specific elements that have been introduced into the design of ID-VP02 and we demonstrate that this novel integration-deficient lentiviral vector platform, which has been designed to deliver a cargo antigen to human DCs, can target and transduce human DCs in preclinical settings.

Results

Integration deficiency is rendered by two independent design elements

For clinical applications that require the direct administration of viral vectors but do not require sustained expression of the vector-delivered gene, such as for vaccines and antigen-directed immunotherapies, integration-defecient lentiviral vectors represent an appropriate and viable alternative to fully integration-competent lentiviral vectors for delivery of their genetic payload. The D64V integrase mutation within the gag/pol gene and the deletion of the 3′-poly purine tract within the vector genome have each been demonstrated to reduce the integration rate of lentiviral vectors.⁹ Both of these elements are included in our vector design and their use in combination is intended to provide redundant mechanisms for reducing the integration potential of ID-VP02. Alu-PCR was performed to measure the impact on integration rate of each of these two design elements alone or in combination. In this assay, 293T huDC-SIGN cells (Supplementary Figure S1) were transduced with WT- or D64V-integrase VSV-G pseudotyped vectors packaging WT (indicated with 703) or 3′-poly purine tract-deleted genomes (indicated with 704) (see also Materials and Methods). At 48 hours posttransduction, cells were analyzed for the presence of integrated provirus by nested Alu-PCR analysis.¹⁰ As shown in Figure 1a, the WT/704 and D64V/703 vectors each had integration rates that were decreased by ~2 logs as compared to WT/703 vector. In comparison, the integration rate of the D64V/704 vector was decreased by greater than 2 logs. These results demonstrate that the ID-VP02 genome has significantly reduced integration potential, and that the D64V and 704 elements independently contribute to this phenotype.

Figure 1.

Quantification of ID-VP02 integration rate. (a) 293T huDC-SIGN cells were transduced with vectors packaged with the wild type (WT) or defective integrase (D64V) and a vector genome containing the 3′-poly purine tract (3′PPT) (703) or 3′PPT deletion (704). At 48 hours posttransduction, nested Alu-PCR analysis was performed on genomic DNA extracted from the transduced cells. Error bars indicate standard error of the mean from transductions performed in triplicate. (b,c) The integration rate of WT/703 and D64V/704 vectors encoding GFP-T2A-NeoR was evaluated using two independent methods. (b) HT1080 huDC-SIGN cells were transduced with serial dilutions of the indicated vectors. Transduced cells were grown under G418 selection and neomycin-resistant colonies, representing individual integration events, were counted. Integration events were calculated as described in Materials and Methods. (c) HT1080 huDC-SIGN cells were transduced with the indicated vectors. Flow cytometry was performed at multiple time points posttransduction to determine the percentage of GFP-expressing cells. Error bars indicate standard error of the mean from flow cytometry triplicates.

To complement the nested Alu-PCR analysis, we employed two additional methods to investigate the integration rate of the ID-VP02 genome. In both methods, HT1080 huDC-SIGN cells (Supplementary Figure S1) were transduced with WT/703 or D64V/704 vector encoding green fluorescent protein (GFP) and neomycin resistance (NeoR) separated by a self-cleaving T2A linker (GFP-T2A-NeoR). Transduction with either of these vectors results in both GFP and NeoR expression. Integration rate was measured as a function of antigen expression, either by outgrowth of neomycin-resistant colonies following G418 selection or by GFP expression over time in bulk culture.

In the first method of measuring integration rate, HT1080 huDC-SIGN cells were transduced with serial dilutions of vector and grown without passaging in the presence of G418 selection. Input vector was normalized by genome copy number. Cells that expressed NeoR and survived prolonged exposure to G418, forming colonies, were presumed to harbor integrated provirus.¹¹ These colonies were counted and total integration events calculated. Using this experimental approach, the integration rate of D64V/704 vector was decreased by 3 logs relative to that of WT/703, in two independent experiments (Figure 1b).

In the second method, transduced cells were serially passaged in the absence of selection and analyzed by flow cytometry at varying times posttransduction. At day 2 posttransduction, ~40 percent of the cells transduced with WT/703 vector were GFP-positive (Figure 1c). This population remained consistent for the duration of the experiment, suggesting that GFP expression was primarily from integrated provirus. In contrast, the percent of GFP-positive cells transduced with D64V/704 vector dropped ~100-fold by day 6 posttransduction and remained low, albeit higher than the mock-transduced control, for the remainder of the experiment. These results suggest that the majority of D64V/704 transduction events yielded nonintegrated vector DNA, which expressed GFP at early times posttransduction, but was lost during subsequent cell divisions. The small percentage of GFP-expressing cells remaining by day 9 posttransduction likely represents the minority of transduction events that yielded integrated provirus. At the completion of the experiment (day 30), it was calculated that the D64V/704 vector was 386-fold decreased in its ability to undergo integration, compared to the WT/703 vector. These findings are comparable to the results from nest Alu-PCR analysis.

Taken together, the results from all three methods of measuring integration rate (nested Alu-PCR, NeoR colony outgrowth, and %GFP expression) demonstrate that the integration rate of the ID-VP02 genome is ~2 logs reduced relative to that of the standard, integration-competent third-generation lentiviral vector (WT/703).

E1001 envelope confers ID-VP02 specificity to the human DC-SIGN receptor

Lentiviral vectors have been pseudotyped with envelopes from various viruses to bestow upon them either broad or specific tropisms (reviewed in ref. ¹²). The prototype Immune Design vector envelope glycoprotein, SVGmu,⁶ was designed based on the sequence of a wild-type Sindbis virus (SINV) envelope because of the natural tropism of this virus to human DCs.⁵ Like the native SINV envelope, SVGmu is able to pseudotype lentiviral vector particles and enable binding and entry to DC-SIGN expressing target cells.⁶ Unlike standard laboratory-adapted SINV envelope sequences that also have the ability to utilize ubiquitously expressed heparan sulfate as a receptor,^13,14 the ability of SVGmu to mediate binding and entry via heparan sulfate is ablated by a disruption in a heparan sulfate binding site.⁶ Cell lines stably expressing the human DC-SIGN receptor were generated (Supplementary Figure S1). These cells were then used to confirm that, as previously reported,⁶ integration-competent SVGmu pseudotyped lentiviral vector particles are able to mediate the specific transduction of DC-SIGN expressing target cells (Figure 2a). However, we found that if integration-deficient vectors (described above and in the Materials and Methods) were pseudotyped with SVGmu, the upstream production yields of biologically active SVGmu pseudotyped lentiviral vector particles were well below those that could be achieved when utilizing a native SINV envelope, SINVHR, as the pseudotyping glycoprotein (Figure 2b).

Figure 2.

E1001 envelope allows ID-VP02 to utilize DC-SIGN as a receptor for binding and entry. (a) 293T cells and 293T cells stably expressing human DC-SIGN (293T huDC-SIGN) were incubated with concentrated (×1,000) integration competent SVGmu pseudotyped vector encoding GFP. Vector titers in GFU/ml on each target cell line were calculated as described in the Materials and Methods section. (b) 293T and 293T huDC-SIGN cells were incubated with concentrated (×100) integration deficient SINVHR and SVGmu pseudotyped vectors encoding GFP. Vector titers in GFU/ml were calculated. (c) SINVHR and SVGmu pseudotyped vector envelope glycoproteins were analyzed via SDS-PAGE, followed by immunoblotting with anti-SINV antibody. (d) Schematic representation of the furin cleavage sites of SINVHR, SVGmu, and E1001. (e) SVGmu and E1001-pseudotyped vector envelope glycoproteins were analyzed via SDS-PAGE gel, followed by immunoblotting with anti-SINV antibody. (f) 293T and 293T huDC-SIGN cells were incubated with crude supernatants containing integration deficient SINVHR, SVGmu, and E1001 pseudotyped vector particles encoding GFP. Vector titers in GFU/ml were calculated.

Many viruses, including SINV, rely on host proteases to process their envelopes into the mature forms that can successfully facilitate fusion into a target cell.¹⁵ In the case of SINV, the full-length, native envelope glycoprotein is posttranslationally processed by the host cell at a furin cleavage site located between the E3 and E2 regions, which results in trimers of E2/E1 protein pairs that facilitate endosomal fusion of mature viral particles into host cells, following receptor engagement.¹⁶ The prototype SVGmu envelope was missing four amino acids in this furin cleavage site (Figure 2d). We therefore hypothesized that the difference in yields between the two pseudotyped vectors (SVGmu versus SINVHR) was due to a difference in the proteolytic processing of the two envelope glycoproteins. Using a western blot probed with antibody specific to SINV, we compared the processing of the envelopes of integration-competent vectors pseudotyped with either SVGmu or with SINVHR (Figure 2c). The observed molecular weight of SINVHR envelope was consistent with the predicted size of cleaved, mature E2 domain, whereas the majority of the SVGmu envelope glycoprotein was at the predicted molecular weight for a protein consisting of the uncleaved precursor (labeled pE2). These results demonstrate that the SVGmu envelope was inefficiently processed, likely due to the disruption of the furin cleavage site.

As the processing of SVGmu was inefficient and since SINV furin cleavage mutants grow poorly and have attenuated pathogenicity,¹⁷ we hypothesized that restoring the furin cleavage site would result in an efficiently processed envelope that would increase the yield of infectious particles during vector production. We therefore generated several envelope variants that included the furin cleavage site. At the same time, we incorporated additional amino acid changes designed to further increase the tropism of this envelope toward human DC-SIGN, while maintaining low preference toward the heparan sulfate receptor. One of these envelope variants, E1001, fulfilled these functional requirements (discussed further below) and contained the following design changes: (i) incorporation of the Sindbis virus RSKR motif of the furin cleavage site (Figure 2d), (ii) glycine substitution at amino acid 160 of the E2 protein for enhanced DC-SIGN tropism,¹⁸ and (iii) glutamate substitutions at amino acids 70 and 159 of E2 for reduced preference to heparan sulfate (Supplementary Figure S2).^17,18,19

We first examined whether restoring the furin cleavage site resulted in appropriate proteolytic processing of the E1001 envelope. To test this, we produced SVGmu and E1001 pseudotyped lentivectors encoding GFP as a marker and analyzed the envelope glycoproteins via western blot. The observed molecular weight and migration pattern of SVGmu was identical to that observed previously (Figure 2c,e). In contrast, the observed molecular weight of E1001 was consistent with that predicted for a cleaved, mature envelope containing the E2 domain (Figure 2e). This indicates that the inclusion of the native SINV RSKR motif within the E1001 amino acid sequence at the furin cleavage site between E3 and E2 was sufficient to mediate efficient processing of the E1001 envelope. We next wanted to determine if this increased efficiency of E1001 posttranslational processing would translate into increased upstream production yields of biologically active vector particles. To test this, we produced integration-deficient GFP encoding lentivectors that were pseudotyped with SINVHR, SVGmu, or E1001 and titered the crude vector supernatants on 293T cells. Pseudotyping integration-defective lentivectors with either SINVHR or E1001 resulted in comparable titers, whereas pseudotyping with SVGmu resulted in vector titers below the assay limit of quantification (Figure 2f). To evaluate the ability of E1001 to enable vector binding and entry to DC-SIGN, we titered the same preps on 293T cells stably expressing human DC-SIGN (293T huDC-SIGN). While SINVHR pseudotyped vector particles had similar transduction efficiencies on both 293T and 293T huDC-SIGN cells (difference less than twofold), a 1-log increase was observed in the efficiency of the E1001 pseudotyped vector particles to transduce 293T huDC-SIGN cells relative to parental controls (Figure 2f). These results demonstrate that the design changes introduced to E1001 significantly increased its ability to specifically bind and enter target cells via the human DC-SIGN receptor, as compared to the native SINVHR envelope.

High-mannose glycosylation of E1001 increases the ability of ID-VP02 to utilize human DC-SIGN as a receptor

Insect cells, which are the native host for wild-type SINV, have a different glycosidic pathway than do vertebrate cells, such that the envelope of viral particles produced in the insect host contains high-mannose moieties that are not present when the same virus is propagated in vertebrates.²⁰ These differences in high-mannose moieties are due to mannosidase activity that is present in vertebrates but absent in insects.²¹ It has been demonstrated that high-mannose glycosylation can be achieved by using mannosidase inhibitors in order to mimic the glycosylation state when envelope is expressed in insect cells, and that this can alter the tropism of SINV envelope pseudotyped vector particles^22,23 by increasing envelope binding to C-type lectins, such as DC-SIGN.⁵ We screened compounds that are inhibitors of mannosidase enzymes for their tolerability and their ability to produce E1001 pseudotyped vectors in mammalian cells with increased specificity toward human DC-SIGN. We produced integration-deficient E1001 pseudotyped lentivectors encoding GFP as a marker in 293T cells at 10 ml scale in the presence of either mannosidase I inhibitors 1-deoxymannojirimycin (DMNJ) (400 µg/ml) or kifunensine (1 µg/ml), or the mannosidase II inhibitor swainsonine (10 µg/ml). The resulting crude supernatants containing vector particles were titered on HT1080 cells expressing human DC-SIGN (HT1080 huDC-SIGN). Vectors produced in the presence of kifunensine had an increased efficiency to transduce human DC-SIGN expressing target cells compared to those produced in the presence of either DMNJ or swainsonine or in the absence of additive (Figure 3a). To find an optimal dose of kifunensine that would produce E1001 pseudotyped vectors with the highest tropism toward human DC-SIGN expressing cells, we produced E1001 pseudotyped lentivectors encoding GFP over a range of kifunensine doses at 1 L scale. Vector particles were concentrated by ultracentrifugation and titered on both HT1080 and HT1080 huDC-SIGN cells. Kifunensine reached maximum performance for producing vector that targeted human DC-SIGN expressing cells at 0.5–1 µg/ml, while out-performing DMNJ at its optimal dose of 400 µg/ml (Figure 3b,c). Use of increasing amounts of kifunensine during production (up to 50 µg/ml) did not further increase the ability of the vector to transduce cells expressing human DC-SIGN above that of vector produced in the presence of 1 µg/ml kifunensine (data not shown). The ability of kifunensine to increase the efficiency of E1001 pseudotyped vector particles to transduce human DC-SIGN expressing target cells was not restricted to HT1080 huDC-SIGN target cells, as similar results were observed when utilizing 293T huDC-SIGN cells as targets (Supplementary Figure S3), demonstrating a direct correlation between the use of kifunensine during vector production and targeting of human DC-SIGN. When produced in the presence of kifunensine, the upstream yields of biologically active E1001 pseudotyped vector particles (as measured by the transduction of 293T huDC-SIGN target cells) were equivalent to those of vector particles pseudotyped with the pan-tropic VSV-G envelope glycoprotein, the industry standard to which lentiviral vector production is measured (Supplementary Figure S3).

Figure 3.

Glycosylation modification of E1001 increases the ability of ID-VP02 to effectively utilize DC-SIGN as a receptor for binding and entry. (a) HT1080 cells stably expressing human DC-SIGN (HT1080 huDC-SIGN) were transduced with E1001 pseudotyped vectors encoding GFP that were produced either with kifunensine, swainsonine, or DMNJ in the culture media or in the absence of additive. Percent GFP⁺ cells are plotted on the y-axis; volume of crude supernatants containing vector particles (μl) is plotted on the x-axis. (b) E1001-pseudotyped vectors encoding GFP were harvested from producer cells treated with varying doses of kifunensine (0.125, 0.25, 0.5, 1 µg/ml) or with DMNJ (400 µg/ml) or from untreated cells. HT1080 huDC-SIGN cells were transduced with these concentrated vector preps. Percent GFP⁺ cells are plotted on the y-axis; volume of concentrated vector particles (μl) is plotted on the x-axis. (c) E1001-pseudotyped vectors from b were titered on HT1080 and HT1080 huDC-SIGN cells. Vector titers in green fluorescence units (GFU)/ml and corresponding targeting ratios were calculated as described in the Materials and Methods section. (d) Schematic representation of the specificities of the endoglycosidases Endo H and PNGase-F. (e) E1001-pseudotyped vectors generated in the absence or presence of 1 µg/ml kifunensine were treated with endoglycosidases that cleave oligosaccharides only if they are of a high-mannose structure (Endo H) or that cleave nonspecifically (PNGase F). The effects of oligosaccharide cleavage on the electromobility of the E1001 envelope glycoprotein was analyzed with a gel-shift western blot assay on a SDS-PAGE gel by electrophoresis of treated virus samples, followed by immunoblotting with anti-SINV antibody. (f) Samples of E1001-pseudotyped vectors from b were treated with Endo H. The effects of oligosaccharide cleavage on the electromobility of the E1001 envelope glycoprotein was analyzed as in e.

The E1001 envelope consisting of E2/E1 domains contains four N-linked glycosylation sites, two of which are structurally hidden from mannosidases (therefore always high-mannose) and two of which are exposed (therefore vulnerable to mannosidase and modifiable by kifunensine during vector production).²² We tested whether the use of kifunensine as an additive during vector production would result in a homogenous rather than heterogeneous glycosylation pattern of the E1001 glycoprotein, as defined by all four oligosaccharides terminating in mannose residues, by utilizing a gel-shift assay that would reveal differences in the electromobility of E1001 glycoproteins with either two N-linked high-mannose oligosaccharides or four N-linked high-mannose oligosaccharides (schematized in Supplementary Figure S4). The presence or absence of high-mannose at these sites would be measured by the sensitivity of the envelope glycoprotein to two endoglycosidase enzymes: Endo H, which specifically cleaves N-linked high-mannose oligosaccharides; or PNGase F, which cleaves N-linked oligosaccharides irrespective of their terminal structure (schematically represented in Figure 3d). E1001 pseudotyped vectors were generated in the absence or presence of 1 µg/ml kifunensine and then were treated with either Endo H or PNGase F or left untreated. The electromobility of the E1001 glycoproteins from these vectors was then analyzed via sodium dodecyl sulfate–polyacrylamide gel (SDS-PAGE) electrophoresis and immunoblotting using anti-SINV antibody. As shown on the right half of Figure 3e, the electromobility of the E1001 glycoprotein from vector produced in the presence of kifunensine (1 µg/ml) was increased after digestion with either Endo H or PNGase F relative to that of untreated vector (no digestion). The equivalence in the sensitivity of this sample to these two enzymes, as measured by the equivalence in their migration pattern, indicates a homogenous glycosylation pattern of the E1001 envelope with respect to high-mannose structure (i.e., all four glycosylation sites are sensitive to both Endo H and PNGaseF, thus all four sites contain high-mannose). In contrast, as shown on the left half of Figure 3e, the no kifunensine sample (no additive) resulted in different electromobility patterns after digestion with EndoH or PNGase F, indicating that not all four glycosylation sites are high-mannose (i.e., only two out of four sites are sensitive to Endo H digestion and therefore lack terminal mannose structures). Furthermore, the difference in two versus four high-mannose glycosylation sites can be visualized in this assay by comparing the Endo H lanes between the no additive and kifunensine treatments. These data demonstrate that the use of 1 µg/ml of kifunensine during vector production generates a homogenous glycosylation pattern of the E1001 envelope with respect to its high-mannose structure.

Next, we assessed whether there was a correlation in the extent of the kifunensine-mediated high-mannose modifications of the E1001 glycoprotein with vector activity. E1001 pseudotyped vectors generated in the presence of varying amounts of kifunensine (evaluated above in Figure 3b,c) were treated with Endo H followed by an analysis of the electromobility of E1001 via sodium dodecyl sulfate–polyacrylamide gel electrophoresis and immunoblotting using anti-SINV antibody (Figure 3f). In Endo H–treated samples, the electromobility of the E1001 envelope was observed to increase in correlation with increasing amounts of kifunensine used during vector production. The varying degrees of high-mannose on these vectors produced from varying kifunensine amounts (as indicated by the degree of gel-shift) correlates with the biological activity of the vector preparations observed in Figure 3b,c. Vector produced with DMNJ, conversely (Figure 3f), when treated with Endo H, resulted in a band that was much more heterogeneous than that seen with vector produced with kifunensine. This heterogeneity of the band in DMNJ vector falls within the heterogeneity of vector produced with lower doses of kifunensine (between 0.125 and 0.25 µg/ml), which correlates with the biological activity of the different vectors (i.e., biological titer of vector produced with DMNJ falls in between the titer of vector produced with 0.125 and 0.25 µg/ml kifunensine; Figure 3b,c). Collectively, these results demonstrate that the use of kifunensine during vector production alters the glycosylation of the E1001 envelope and increases the ability of E1001 pseudotyped vectors to effectively utilize DC-SIGN as a receptor for vector binding and entry. We therefore selected kifuensine for use as a glycosylation modifier at a concentration of 1 µg/ml for use during the production of ID-VP02 (see Materials and Methods).

ID-VP02 transduces human DCs via the endogenous DC-SIGN receptor in a manner that is dependent on a high-mannose envelope glycoprotein and Vpx

Having demonstrated the ability of E1001 pseudotyped integration-deficient lentiviral vectors produced in the presence of kifunensine to transduce target cell lines ectopically expressing DC-SIGN, we next assessed the ability of the modified E1001 envelope to mediate the binding and entry step required for transduction of human DCs. Although the genetic and glycobiological modifications of the E1001 envelope under evaluation are predicted to facilitate preferential binding to carbohydrate recognition domains present on DC-SIGN and mediate the entry of the vector particle into human DCs, recently, a restriction factor directed against HIV-1–based vectors has been described in human myeloid-derived cells such as DCs and macrophages that would prevent transduction to occur.^24,25 This restriction factor, SAMHD1, prevents the reverse transcription of lentiviral genomes in the cytoplasm after viral entry by depleting intracellular pools of dNTPs.³ Unlike HIV-1, other lentiviruses, such as SIVmac and HIV-2, encode and incorporate into virus particles a viral protein termed Vpx that is able to target SAMHD1 for degradation.^24,25 Thus, we investigated whether our lentiviral vector would transduce human DCs via a two-step process, and whether our vector might benefit from the incorporation of both our modified E1001 envelope and Vpx into the vector design.

Previous studies to determine the molecular elements required to package SIVmac Vpx into HIV-1–based vector particles have yielded conflicting results. In early reports, Vpx was delivered into target DCs by preinfecting these cells with SIVmac-derived virus-like particles that packaged Vpx;²⁶ these cells were subsequently successfully transduced with HIV-1–based vectors that could be either integration-competent^26,27 or integration-defecient.²⁸ Later reports demonstrated that Vpx could be packaged into HIV-1–based vectors directly, either when Vpx was expressed by itself^26,29,30 or when expressed as a fusion with the HIV-1 accessory protein Vpr.³¹ Yet, other reports have demonstrated that efficient packaging of Vpx requires the generation of a chimeric HIV-SIV vector, in which the p6 portion of Gag from SIVmac is incorporated into the Gag of HIV-1.³²

Based on these conflicting reports from the literature, we assessed the packaging efficiency of SIVmac Vpx into vector particles when this protein was expressed either by itself or as a fusion protein with HIV-1 Vpr. N-terminal hemagglutinin (HA)-tagged versions of Vpx or of Vpx fused to Vpr (Vpx-Vpr) were cloned. Each of these constructs was transfected into 293T cells in combination with the four other vector component plasmids to produce either integration-competent (int+) or integration-deficient (int−) E1001 pseudotyped vector particles. As a negative control, int− vector was produced without either Vpx construct. Vector supernatant was concentrated and analyzed via western blot with antibodies toward HA (anti-HA) or HIV-1 p24 (anti-p24). As demonstrated in Figure 4a, both Vpx and Vpx-Vpr were packaged in vector particles and this packaging was independent of whether the vector was integration-competent or integration-deficient. Interestingly, the HA-tagged Vpx protein was packaged at greater efficiency than the Vpx-Vpr fusion, even though these two proteins were expressed at comparable levels in the 293T producer cells (data not shown) and equivalent amounts of vector were loaded on the western blot, as indicated by p24 protein content (Figure 4a, lower panel). We next tested whether Vpx from SIVmac could be detected using an antibody directed against HIV-2 Vpx, and whether removing the HA-tag would still result in Vpx protein being efficiently packaged into vector particles. ID-VP02 was produced with the inclusion of a plasmid encoding untagged Vpx and the resulting concentrated vector was analyzed via western blot using antibody toward HIV-2 Vpx (anti-Vpx). These data show that the untagged version of SIVmac Vpx was packaged into ID-VP02 and that this protein was detectable using the anti-HIV-2 Vpx antibody (Figure 4b). These data demonstrated that when full-length, untagged Vpx was expressed during vector production, it was efficiently packaged into ID-VP02 and that no further manipulations were needed on either the vector capsid or the Vpx protein to facilitate this.

Figure 4.

ID-VP02 packages Vpx from SIVmac. (a) Vpx from SIVmac was cloned by itself or as a fusion with Vpr from HIV-1 (Vpx-Vpr) and contained an HA tag at its N-terminus. Vpx or Vpx-Vpr was introduced by transfection during production of E1001-pseudotyped lentivector encoding GFP that is either integration-competent (int+) or integration-deficient (int−). As a control, one vector prep was made without either Vpx or Vpx-Vpr. Vector harvests were concentrated by ultracentrifugation and analyzed on a SDS-PAGE followed by immunoblotting with either anti-HA antibody (upper panel) or anti-p24 antibody (lower antibody). (b) Vpx was cloned without the HA tag and was tested for packaging into ID-VP02 by inclusion of a Vpx expression plasmid during production of E1001-pseudotyped lentivector encoding GFP. As a control vector was prepared without Vpx (Vpx−). Vector harvests were concentrated by ultracentrifugation and analyzed on a SDS-PAGE followed by immunoblotting with anti-Vpx antibody that is specific to HIV-2 Vpx.

We next asked if the high-mannose modified E1001 envelope could target ID-VP02 to human DCs and whether SIVmac Vpx was required to successfully transduce these cells. Human monocyte-derived DCs (MDDCs) were generated and were shown to be 93% double positive for the DC markers CD11c and DC-SIGN (Figure 5a). These cells were then incubated with increasing amounts of E1001 pseudotyped integration-deficient lentiviral vectors encoding GFP that were generated in the presence or absence of kifunensine and that were produced with or without Vpx. The reverse transcriptase inhibitor nevirapine (RT inhibitor) was used on the highest vector dose to demonstrate that the de novo production of GFP within the target cells was the direct result of transduction. As shown in Figure 5b, E1001 pseudotyped integration-deficient lentiviral vectors were able to transduce human DCs only if they contained Vpx and only if they were produced in the presence of kifunensine. To assess donor-to-donor variability, five additional donor MDDCs were evaluated for transduction with ID-VP02 with or without Vpx and produced in the presence of kifunensine. Four of these additional MDDCs were transduced by ID-VP02 in a Vpx-dependent manner (Supplementary Figure S5). One donor was not transduced by either ID-VP02 or by VSV-G pseudotyped vector, regardless of Vpx being present, suggestion of an unknown block to transduction, hence was excluded from the analysis (data not shown). In combination with the donor used in Figure 5b, all five donor MDDC transduction results were graphed together and mean ± standard deviation was calculated (Supplementary Figure S6). Next, we assessed the requirement for DC-SIGN for the transduction of human DCs with our vector by using an antibody toward human DC-SIGN (anti-DC-SIGN) as a specific blocking agent. Human MDDCs were pretreated with anti-huDCSIGN and then incubated with increasing amounts of integration-deficient lentivector containing Vpx that was either pseudotyped with VSV-G or with a high-mannose modified E1001 envelope. Both vectors were able to transduce human DCs in the absence of any blocking agent; however, only VSV-G pseudotyped vectors (which do not require DC-SIGN for entry) were able to transduce human DCs in the presence of anti-DC-SIGN antibody (Figure 5c).

Figure 5.

ID-VP02 transduces human dendritic cells. (a) PBMC-derived human DCs (generated as described in the Materials and Methods section) were analyzed for the expression of CD11c and DC-SIGN by flow cytometry. Representative dot plots are shown for one donor. Number in right plot is the percentage of cells within the double positive gate. (b) PBMC-derived human DCs described in a were incubated with varying amounts of E1001-pseudotyped LV (5, 10, or 20 ng of p24) constructs that encoded GFP as a reporter gene and which either did or did not contain Vpx or were produced in the in the presence or absence of kifunensine. Transduction events were measured by assessing percent of GFP-positive cells within the CD11c-positive population. The reverse transcriptase inhibitor Nevirapine (RT inhibitor) was used on the highest vector dose (20 ng p24) to demonstrate the de novo production of GFP within the target cells. Numbers in plots are the percentage of cells within the GFP⁺ gate. (c) PBMC-derived human DCs were pretreated with a monoclonal antibody against human DC-SIGN (indicated by + or −) or left untreated then transduced with varying amounts (5, 10, or 20 ng of p24) of E1001-pseudotyped and kifunensine-treated LV (E1001+Kifu, right half), or of VSV-G-pseudotyped LV (VSV-G, left half), both containing Vpx. Transduction events were measured by gating on cells that were positive for CD11c (y-axis), and assessing percent of cells positive for GFP (x-axis). Numbers in plots are the percentage of the cells that are GFP positive. Nevirapine (RT inhibitor) was used on the highest dose of LV (20 ng).

Recent reports have disagreed on whether or not Vpx is sufficient to overcome the antiviral state induced by interferons (IFNs), and whether or not IFNs upregulate SAMHD1.^{29,33,34,35,36} We therefore pretreated human DCs with IFN-γ, known to upregulate SAMHD1,³⁵ followed by incubation with our lentiviral vector. We found that IFN-γ resulted in reduced surface expression of DC-SIGN. Furthermore, IFN-γ treatment resulted in an antiviral state that was not overcome by Vpx (Supplementary Figure S7). Collectively, these results demonstrate that the transduction of human DCs with E1001 pseudotyped integration-deficient lentiviral vectors requires (i) high-mannose structures on the E1001 glycoprotein to mediate binding and entry in a DC-SIGN specific manner and (ii) Vpx to overcome the natural SAMHD1 mediated restriction to HIV-1–based vectors that is present in human DCs. We therefore have included Vpx as a design element of ID-VP02.

Human DC subsets differentiate from various lineages. We therefore assessed the ability of ID-VP02 to transduce human DCs derived from hematopoietic progenitors instead of peripheral blood mononucleated cell (PBMC)-derived monocytes. We differentiated human CD34⁺ bone marrow cells into DC-SIGN expressing DCs (Materials and Methods) followed by incubation with ID-VP02 encoding GFP. We found that ID-VP02, similar to its ability to transduce monocyte-derived DCs, is also capable of transducing bone marrow–derived human DCs in a Vpx-dependent manner (Supplementary Figure S8) confirming DC-tropism of ID-VP02.

After demonstrating that ID-VP02 could effectively transduce human DCs and that transduction occurred in a DC-SIGN–specific manner, we next assessed how these vectors performed within the context of a heterogenous population of potential target cells, rather than a homogenous population of human DCs. To address this question, human PBMCs were placed in culture in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin (IL)-4 for three days to generate a pool of primary cells that included a sufficient number of MDDCs expressing DC-SIGN. On day 3, 20 ng p24 of ID-VP02 encoding GFP that were produced in the presence of kifunensine and that contained Vpx were added to the culture. In a separate experiment, we confirmed that 3 days are sufficient to generate MDDCs (Supplementary Figure S9) and that culturing PBMCs longer resulted in cell viability issues following transduction of PBMCs (data not shown). Three days after the introduction of the vector into the culture, cells were analyzed for the expression of GFP as a measure of transduction within the major populations of cells present at the time of analysis: DCs (CD11c^pos) 6%, B cells (CD11c^neg, CD19^pos) 10%, and T cells (CD11c^neg, CD3ε^pos) 80%. As shown in Figure 6, 42% of the cells within the CD11c^hi, DC-SIGN⁺ population were transduced compared to 0.1% for both the B- and T-cell populations present within the culture. Transduction was completely ablated in all cell populations when the reverse transcriptase inhibitor nevirapine (RT inhibitor) was included in the culture. To assess donor-to-donor variability, four additional donor PBMCs were evaluated for transduction with ID-VP02 with or without Vpx and produced in the presence of kifunensine. All of these additional PBMCs demonstrated transduction of their DCs by ID-VP02 in a Vpx-dependent manner (Supplementary Figure S10). In combination with the donor used in Figure 6, all five donor PBMC DC transduction results were graphed together and mean ± standard deviation was calculated (Supplementary Figure S11). In a separate experiment, we incubated human macrophages or CD14⁺ monocytes (both of which lack the DC-SIGN receptor) with ID-VP02 and confirmed that ID-VP02 does not transduce these cells (Supplementary Figure S12). These results demonstrate that in a heterogeneous population of human cells of which DCs are a minority, ID-VP02 specifically transduces DC-SIGN expressing DCs.

Figure 6.

ID-VP02 targets human dendritic cells. Human peripheral blood mononucleated cells were treated with granulocyte-macrophage colony-stimulating factor and IL-4 for 3 days, then incubated with 20 ng p24 of ID-VP02 encoding GFP. At the time of vector addition, the culture consisted primarily of DCs, B cells, and T cells (data not shown). Three days after transduction, cells were analyzed for surface markers and categorized as either DCs (CD11c^pos) 6%, B cells (CD11c^neg, CD19^pos) 10%, or T cells (CD11c^neg, CD3ε^pos) 80%. Transduction events were measured by assessing percent of cells positive for GFP within each population of cells. Numbers in plots are the percentage of cells within the GFP⁺ gate.

Discussion

In this report, we have described the development of a novel integration-deficient lentiviral vector that can successfully deliver antigen-encoding nucleic acids specifically to human DCs. This technology, termed ID-VP02, is the result of significant advances that have been made to a prototoype vector described previously.⁶ With the development of ID-VP02, we have generated a vector for the purposes of direct in vivo delivery into patients and have demonstrated that (i) ID-VP02 incorporates safety features, in addition to having a split-genome, that are designed to eliminate integrase-dependent integration events with two redundant mechanisms, (ii) ID-VP02 overcomes limitations in upstream vector production yields of integration-deficient versions of the prototype, thus enabling manufacturing for the evaluation of the vector in clinical settings, and (iii) ID-VP02 specifically transduces human DCs by packaging Vpx and by utilizing a genetically modified Sindbis virus envelope (E1001) that is posttranslationally modified to contain a terminal high-mannose carbohydrate structure.

ID-VP02 has been engineered to be integration-deficient by the incorporation of two specific elements into the design. The first is the use of a pol gene that encodes a mutant integrase that has been rendered noncatalytic by point mutation of one of the key amino acids in its active site (D64V).^37,38,39 Catalytic site mutations within HIV-1 integrase that make the enzyme inactive severely impair the ability of these mutant viruses to integrate into host chromatin, although integration is not completely ablated. Previous studies have demonstrated that D64V and related integrase catalytic domain point mutants decrease integration of HIV-1 virus or HIV-1–based lentiviral vectors by 10³- to 10⁴-fold, in comparison to a matched control encoding a WT integrase.^11,37,40 With our vector design and experimental methods, we detected an ~2 log decrease in integration of a lentivector encoding the D64V point mutation (Figure 1a; D64V/703). It is likely that the observed residual integration events result from enzyme-independent integration. This hypothesis is based on published work with HIV-1 D64V virus, in which sites of integration lacked the canonical target site duplications which are the signature of HIV-1 integrase enzymatic activity.⁴⁰

The second design element for integration deficiency is the deletion from the vector backbone of the polypurine tract that is proximal to the 3′LTR (704). This deletion favors the formation of single-LTR dsDNA circles (episomes) during plus-strand DNA synthesis (as opposed to linear molecules, which are the primary product of Wild Type lentivirus reverse transcription and DNA synthesis). Unlike for linear molecules, although the sequences remain intact the viral attachment (att) sites present within single-LTR circles are not accessible to facilitate integrase-mediated chromosomal integration.⁹ Consistent with this model, we observed that the integration rate of a ΔPPT lentiviral vector (704) was decreased by ~2 logs compared to a lentiviral vector encoding a wild-type genome, as measured by Alu-PCR. The combination of the ΔPPT deletion with the D64V point mutation in integrase resulted in an even greater impact on vector integration rate than either of these two mutations alone, consistent with a previously published report,⁹ suggesting that these two mutations have a synergistic effect on decreasing integration rate. In contrast, no such synergy has been observed between the D64V integrase point mutation and mutation of the viral att sites.¹¹ In conclusion, the results from the studies reported here indicate that the integration rate of ID-VP02 is 2–3 logs below that of the standard, integration-competent third-generation lentiviral vectors, as demonstrated by three independent measurements of vector integration rate. For clinical applications in which prolonged antigen expression is not required, such as for vaccines and antigen-directed immunotherapies, integration-deficient lentiviral vectors clearly have a safety advantage over their integration-competent analogs. The 2–3 log decrease in integration rate observed for ID-VP02 considerably decreases the potential risk of insertional mutagenesis, which has been observed in clinical trials among patients that have received integration-competent γ retrovirus-based vectors, specifically those based on murine leukemia virus.^41,42

The transduction of human DCs via HIV-1–based lentiviral vectors occurs via a multistep process. The first step involves binding and entry of the vector particles to human DCs. Our results demonstrate two specific modifications incorporated into the ID-VP02 platform design that enable the binding and entry of the vector particles into human DCs. The first is the envelope glycoprotein E1001 which has binding specificity to the human DC-SIGN membrane protein. Derived from native SINV, an arbovirus known to infect dermal DCs via the DC-SIGN receptor,^5,18 E1001 has been genetically modified from the prototype to increase its DC tropism and prevent binding to ubiquitous heparan sulfate receptors.¹⁴ Importantly, E1001 was also modified from the prototype to include a functional furin cleavage site between the E3 and E2 regions of the glycoprotein. The criticality of this genetic element was highlighted upon the transition from an integration-competent to integration-deficient vector design, likely due to previously reported findings that, irrespective of the nature of the envelope glycoprotein, integration-defective vectors are more difficult to manufacture at high titers when compared to their integration-competent counterparts.⁴³

In addition to these genetic modifications, ID-VP02 is produced in the presence of the mannosidase I inhibitor, kifunensine.⁴⁴ Production of ID-VP02 in this manner results in the generation of ID-VP02 envelope glycoproteins with terminal high-mannose residues. These genetic and glycobiological modifications of the ID-VP02 envelope facilitate use of human DC-SIGN as a receptor for binding and entry. The ID-VP02 glycoprotein modified with high-mannose in order to target DC-SIGN may mimic the high-mannose state of SINV glycoprotein produced naturally from its insect host. While the presence or absence of mannosidase I in insects has been debated, insect cells lack the processing machinery required to remove the terminal mannose residues on the SINV envelope glycoprotein.^21,45 Upon infecting a host through an insect bite, SINV gets delivered to a site that is rich in DC-SIGN expressing dermal DCs and thus can take advantage of its natural high-mannose moieties to target DCs. In addition to SINV, many viruses have evolved to infect new hosts via routes that would make them accessible to DCs (reviewed in ref. ⁴⁶). Considered sentinels of the immune system, DCs capture antigens and traffic them to lymph nodes to present them to T cells. Viruses have exploited this pathway in many ways by using DC-SIGN directly or indirectly to infect DCs. For example, HIV-1 targets DC-SIGN to facilitate infection of CD4 T cells,⁴⁷ RSV targets DC-SIGN to downregulate signaling within DCs,⁴⁸ and Measles virus entry into DCs is also DC-SIGN dependent.^49,50 By making ID-VP02 tropic toward DC-SIGN, we have tried to use DC targeting to our advantage in order to maximize the potential to generate a therapeutic immune response.

The second step required for the transduction of human DCs by HIV-1–based lentiviral vectors involves overcoming an infection block that is naturally present in human DCs and macrophages toward lentiviruses. This restriction is encoded by a cellular factor called SAMHD1 that prevents the reverse transcription of lentiviral genomes in the cytoplasm after viral entry.²⁵ To overcome this block, we incorporated SIVmac Vpx into the ID-VP02 platform as an accessory protein. Vpx is an accessory protein found naturally in the HIV-2/SIVsm (SIV from sooty mangabeys) lineage but is absent from HIV-1. Vpx inclusion overcomes restriction in target DCs by promoting the degradation of SAMHD1.^24,25 Our results show that the ability of ID-VP02 to transduce human DCs requires both the presence of high-mannose on the E1001 envelope glycoprotein and the incorporation of Vpx into the vector particles. Curiously, recent reports have conflicting evidence on IFNs upregulating SAMHD1 and whether Vpx can overcome an antiviral state induced by IFNs.^{29,33,34,35,36} We, too, observed that monocyte-derived DCs, when treated with IFN-γ, resulted in an antiviral state that was not overcome by Vpx, suggesting activation of other antiviral restriction mechanisms.

Host restriction factors like SAMHD1 have evolved to block viral infections at various steps of the viral life cycle once a virus has entered a target cell (reviewed in ref. ⁵¹). For example, Trim5α blocks reverse transcription of retroviruses by targeting the viral capsid for degradation⁵² and Apobec3G induces G to A hypermutations within the retroviral genome during reverse transcription.⁵³ SAMHD1 works by depleting the nucleotide pool in the cell that lentiviruses need to complete reverse transcription.⁵⁴ Interestingly, viruses have, in return, evolved factors to deal with these restriction factors. Retroviral capsids have evolved to escape recognition from Trim5α⁵⁵ and lentiviruses evolved Vif and Vpx to counteract Apobec3G and SAMHD1, respectively.^25,53 As viral vectors continue to be developed for the clinic, different challenges of overcoming restriction factors in hosts will present themselves based on the species and the tissue being targeted. In order to overcome SAMHD1 restriction in target DCs, we incorporated Vpx from SIVmac into ID-VP02. Interestingly, although there have been conflicting reports on the need to generate chimeric capsids in order to package Vpx into virions,^29,30,32 we found that Vpx appeared to be packaged into ID-VP02 without us needing to modify the capsid of the vector.

In conclusion, ID-VP02 has been designed with the goal of developing licensed antigen-directed active immunotherapeutics that can be safely administered directly to patients and that would deliver antigen-encoding nucleic acids to DCs in order to induce antigen-specific effector T cells populations. At Immune Design, we are in the process of conducting in vivo preclinical platform proof-of-concept as well as investigational new drug (IND)-enabling activities that will permit the testing of the safety and immunogenicity of a first ID-VP02–based clinical candidate in oncology settings.

Materials and Methods

Vector components. Similar to third-generation lentiviral vectors, ID-VP02 has a split genome and contains a self-inactivating (SIN) deletion in the 3′LTR (ΔU3). ID-VP02 is generated via a 5 plasmid system in which the transfer vector (lentiviral vector genome), a modified gag/pol transcript (RI gag/pol), accessory protein Rev from HIV-1, accessory protein Vpx from SIVmac, and E1001 envelope are each encoded on a separate plasmid. The ID-VP02 genome contains cis-elements that are derived from HIV-1 important for packaging (psi-packaging signal), for splicing (splice donor and acceptor sites) and the Rev-responsive element. A CMV promoter transcribes the genome which is flanked by a 5′LTR (R and U5) and a 3′LTR (ΔU3, R, and U5). Downstream of the antigen insert is a modified version of the woodchuck hepatitis posttranscriptional element which is present for the purposes of increased antigen expression.⁵⁶ The modified version of the woodchuck hepatitis posttranscriptional element prevents the translation of the open reading frame of the woodchuck hepatitis virus X protein, thereby eliminating chances of its expression in transduced cells.⁵⁷ Rev is included as an accessory protein as it allows for the export of unspliced messages of the ID-VP02 genome from the nucleus to be packaged into vector particles.^58,59 Features unique to ID-VP02 include the following: ID-VP02 is pseudotyped with a novel envelope, E1001, that has been modified to optimize tropism toward the receptor DC-SIGN, and to minimize tropism toward the more ubiquitous heparan sulfate receptor (Supplementary Figure S2). Enhanced DC-SIGN tropism results from a glycine substitution at amino acid 160 of the E2 protein.¹⁸ Reduced heparan binding results from a glutamate substitution at amino acid 159 of E2^18,19 and from a glutamate substitution at amino acid 70 of E2.¹⁷ E1001 includes the Sindbis virus RSKR cleavage motif between E3 and E2 proteins to preserve entry into target cells (reviewed in Leung et al.⁶⁰). The antigen promoter is the human Ubiquitin-C promoter that has been modified to have its natural intron deleted (ΔUbiC). This deletion minimizes the chance of splice variants that may cause heterogeneity in vector lots. The ΔU3 of the LV genome has an extended deletion of the usual SIN to cover and also delete the 3′-poly purine tract,⁶¹ referred to in the text as 704. The deletion of this portion of the vector backbone provides one of the two mechanisms which render ID-VP02 integration deficient: this deletion favors the formation of single-LTR reverse transcribed episomal dsDNA circles that are not capable of chromosomal integration (as opposed to double-LTRs in wild-type lentivirus).⁹ The second mechanism to achieve integration deficiency is the D64V catalytic mutation in the integrase enzyme encoded by the pol gene.⁶² The design of ID-VP02 includes a codon optimized gag/pol plasmid for production that is devoid of the Rev-responsive element. The removal of the Rev-responsive element and codon optimization of the gag/pol gene minimizes the chance of psi-gag recombination and provides an added measure of safety by reducing the opportunity for the formation of a replication-competent Lentivirus during vector production.^63,64,65 Incorporated in the design of ID-VP02 is the inclusion of Vpx from SIVmac as an additional accessory protein. Vpx inclusion overcomes restriction in target human DCs by promoting the degradation of SAMHD1.^24,25

Vector production. Vector was produced in either large scale (CF10) or small scale (T25). For large scale production, 293T cells were seeded at 5E8 cells/1L in a 10-layer cell factory (Nunc, catalog#140400) in Dulbecco's modified Eagle's medium media containing 5% serum, L-glutamine, and antibiotics. Three days later, cells were transfected using polyethylenimine (PEI) (stock 1 mg/ml) and total plasmid DNA at a ratio of 3:1 (ml PEI:mg DNA). Per 10-layer cell factory, 1 mg of vector genome plasmid, and 0.5 mg of remaining plasmids (gag/pol, Rev, Vpx, and envelope) were used. Five hours later, media was replaced with 1 l of serum-free media (Transfx-293 media, Hyclone catalog# SH30860.LS). For kifunensine (Glycosyn, catalog# FC-034) treatment, kifunensine was included at a final concentration of 1 µg/ml, or at varying doses as indicated in Figure 2. For DMNJ treatment, DMNJ was included in the culture media at a final concentration of 400 µg/ml (Tocris, catalog# 1259). For Swainsonine treatment, swainsonine was included in the culture media at a final concentration of 10 µg/ml (Tocris, catalog# 3208). Vector was harvested 2 and 3 days after transfection. Harvests were clarified using a prefilter and 0.45 µm stericup filter (Millipore, Billerica, MA). Vector was concentrated by spinning in a 1 l centrifuge bottle at 16,000g for 5 hours. Pellet from each L harvest was either resuspended in 1 ml of Hanks' balanced salt solution (HBSS) and aliquoted for storage at −80 °C, or was resuspended in 1 ml of buffer for benzonase treatment (50 mmol/l Tris-HCL pH7.5, 1 mmol/l MgCl₂, 5% v/v sucrose). Benzonase nuclease was added at a final of 250 U/ml and incubated overnight at 4 °C in order to degrade any left-over plasmids from the transfection. Benzonase-treated viral preps were reconcentrated using a sucrose cushion (30% sucrose top, 5% sucrose bottom) and centrifuged at 116,000g in an ultracentrifuge for 1.5 hours at 4 °C. Vector pellet was resuspended in 1 ml HBSS, aliquoted, and stored in −80 °C. For small scale vector production, 293T cells were seeded in a T25 flask and transfected using PEI similar to as described above, but with 4 µg of vector genome plasmid and 2 µg of remaining plasmids. Five hours later, media was replaced with 4 ml of Dulbecco's modified Eagle's medium media containing 5% serum, L-glutamine, and antibiotics. For kifunensine treatment, kifunensine was included at a final concentration of 1 µg/ml. Vector was harvested 2 and 3 days after transfection and was clarified using 0.45 µm filter. Vector was either frozen as crude supernatant or concentrated by centrifuging at 116,000 x g in an ultracentrifuge for 1.5 hours at 4 °C, resuspended in an appropriate volume and stored −80 °C.

Vector quantification: p24 assay. Quantification of p24 was performed using the HIV-1 p24 ELISA kit by Advanced Bioscience Laboratories (Rockville, MD), following the manufacturer's directions.

Vector quantification: green fluorescence units assay and targeting ratio. Either 293T cells stably expressing DC-SIGN or HT1080 cells stably expressing DC-SIGN were plated at 2E5 cells/well (293T) or 4E4 cells/well (HT1080) in a 12-well plate in 1 ml Dulbecco's modified Eagle's medium media containing 5% serum, L-glutamine, and antibiotics. Twenty-four hours later, cells in each well were transduced with twofold dilutions of vector encoding GFP. Each amount of vector is prepared in a 1 ml final volume in DMEM complete media. For crude vector 5, twofold serial dilutions are prepared starting from 200 µl of vector per well. For concentrated vector 5, twofold serial dilutions are prepared starting from 1 µl of concentrated vector per well. As a control to rule out pseudotransduction, 10 µmol/l of the reverse-transcriptase inhibitor nevirapine was used with the highest volume of vector in a parallel well. Forty-eight hours after transduction, cells were analyzed for GFP expression on a Guava machine (Guava technologies, now Millipore). Green fluorescence units per ml was calculated by using a best fit (least squares) linear regression model based on the volumes of vector and the resulting percent GFP values in order to predict the number of GFP-positive cells per ml of vector using the FORECAST function in EXCEL. Events that resulted in less than 1% of GFP-positive cells were set as the limit of quantification.

Vector quantification: genomes assay. Genomic RNA was isolated from vector particles using the QIAamp Viral RNA Mini kit (Qiagen, Valencia, CA). To eliminate contaminating DNA, the extracted nucleic acid was then digested with DNAseI (Invitrogen, Carlsbad, CA) following the manufacturer's directions. Two dilutions of each DNAseI-treated RNA sample were then analyzed by quantitative reverse transcription polymerase chain reaction (RT-PCR) using the RNA Ultrasense One-Step Quantitative RT-PCR System (Invitrogen) and previously described vector-specific primers and probe.⁶⁶ The RNA genome copy number was calculated in reference to a standard curve comprised of linearized plasmid DNA containing the target sequences, diluted over a 7-log range (1E1–1E7 copies). The genome titer as expressed here reflects the number of physical vector particles, calculated based on genomes, with each vector particle predicted to contain two single-stranded copies of genomic RNA.

Cell lines. Receptors were cloned individually into a retroviral (Clontech, Mountain View, CA) or lentiviral expression system containing puromycin resistance. Vector was prepared in small scale as described above. Target cell lines were transduced at a high moi. Twenty-four hours after transduction, media was replaced with puromycin-containing media.

Anti-SINV western blot and gel shift western blot. LVs were produced as described above, in the presence or absence of varying amounts of kifunensine or DMNJ (AGScientific #D-1155). Concentrated vector was treated with either PNGaseF (NEB, cat#P0704S) or EndoH (NEB cat#P0702S) according to manufacturer's instructions or left untreated. Preps were denatured in loading dye containing dithiothreitol (Novex; Invitrogen) and separated on a 10% Bis-Tris gel. Gels were transferred onto nitrocellulose and probed with anti-SINV antibody (ATCC, cat#VR-1248AF).

Western blots to verify Vpx packaging. Vector with different Vpx constructs was prepared as described above. Vpx constructs were either a fusion of SIVmac Vpx and HIV-1 Vpr (with an HA tag at the N-terminus of the construct) or SIVmac Vpx (with an HA tag at the N-terminus) or SIVmac Vpx without an HA tag. Concentrated vector was then analyzed via sodium dodecyl sulfate–polyacrylamide gel electrophoresis by loading 50 or 100 ng of p24 per well of denatured vector using 4–12% NuPAGE Bis-Tris precast gels (Invitrogen, catalog # NP0321PK2) and transferring onto a nitrocellulose membrane. Blots were then probed with either anti-HA antibody (Covance, Princeton, NJ; catalog # MMS-101P) or anti-p24 antibody (Abcam, Cambridge, UK; catalog # ab9071) or anti-Vpx antibody (AIDS reagent, catalog # 2710).

Donor information for human PBMCs and primary cells. Donor #018_20110413 PBMCs (obtained Immune Design) were used in Figure 5a,b; Figure 6; Supplementary Figure S6, S11, and S12a, as well as the experiment culturing PBMCs for 6 days (data not shown). Donor PBMCs obtained from CTL, Shaker Heights, OH, were used for Figure 5c. Multiple donor MDDC experiments were performed on monocytes obtained from StemCell Technologies, Vancouver, BC, (lot numbers 93124065, 776277065, 948287065, 511218065, 013284065) and were used in Supplementary Figures S5 and S6. Multiple donor PBMC experiments were performed on PBMCs obtained from Immune Design (lot numbers 010_20110609, 019_20110609, 003_20100218, 007_20100218) and were used in Supplementary Figures S10 and S11. Donor monocytes (obtained from StemCell Technologies) were used to assess transduction of monocytes (Supplementary Figure S12c), and were used to differentiate into DCs that were pretreated with IFN-γ prior to transduction (Supplementary Figure S7). Donor macrophages (StemCell Technologies) were used to assess transduction of macrophages (Supplementary Figure S12b). Donor CD34⁺ bone marrow cells (Astarte Biologics, Redmond, WA) were used to differentiate into DCs and transduce with ID-VP02 (Supplementary Figure S6).

Generation of human PBMC-derived DCs. Thawed human PBMCs were incubated in antiaggregate solution (CTL, cat# CTL-AA-005) for 15 minutes at room temperature. Cells were then counted and diluted to make 5E7/ml in a final volume of 4 ml. From these cells, CD14⁺ monocytes were enriched using a negative selection kit (StemCell Technologies, cat#19059) following the manufacturer's instructions. Resulting cells were plated in a 10 cm plate in Roswell Park Memorial Institute medium containing 10% human serum, human GM-CSF (100 ng/ml), and human IL-4 (50 ng/ml). Three days later, cells were replenished with human GM-CSF and IL-4 (100 and 50 ng/ml, respectively). Two days later, half of the media was replaced with fresh media; human GM-CSF and IL-4 were added for a final of 100 and 50 ng/ml, respectively. Two days later, cells were either transduced with vector, or stained with antibodies and analyzed for surface markers. For transduction, cells were removed from the plate using a cell-lifter, counted, and plated in a U-bottom 96-well at 1E5 cells/100 µl/well in media consisting of RPMI containing 10% human serum. Nevirapine control wells were pretreated with 10 µmol/l nevirapine for 1 hour at 37 °C. Twofold dilutions of vector were prepared in a total volume of 100 µl media, 20 ng to 5 ng of p24 of vector/100 µl. Each 100 µl vector solution was added to 100 µl of cells in the 96-well plate to bring the total to 200 µl. Virus was spinoculated at 1,200 x g for 20 minutes at room temperature, followed by incubation for 2 hours at 37 °C. Cells were spun down, vector containing media was removed, and fresh media was added to cells. 10 µmol/l nevirapine was included for nevirapine control wells. Three days after transduction, cells were stained for surface markers and analyzed for GFP expression.

Generation of human DCs from CD34⁺ bone marrow cells. Human CD34⁺ bone marrow cells were obtained from Astarte Biologics (catalog #1015) and were cultured using the “standard procedure for DC differentiation” described by Canque et al.⁶⁷ These cells were cultured in Roswell Park Memorial Institute medium 10% fetal bovine serum with antibiotics media containing FLT3-Ligand (50 ng/ml), stem cell factor (50 ng/ml), tumor necrosis factor-α (10 ng/ml) and GM-CSF (20 ng/ml) for 9 days, replacing with fresh media every 2–4 days. At this point, they were profiled for surface markers and were shown to be high for class II and medium for CD11c, but negative for DC-SIGN (data not shown). These “pre-DCs” were then differentiated into DCs using GM-CSF (100 ng/ml) and IL-4 (50 ng/ml) for 6 more days at which point surface staining of DC-SIGN was confirmed (data not shown) followed by incubation with ID-VP02.

For surface staining, cells were resuspended in 100 µl/well of FACS buffer (phosphate-buffered saline, 2% fetal bovine serum, and 0.05% sodium azide) containing 10% rat serum as blocking reagent, and incubated on ice for 10 minutes. Antibodies were diluted in 100 µl of FACS buffer, without the blocking reagent, according to manufacturer's recommendations. Antibodies were added to the cells and incubated 15 minutes on ice, protected from light. Cells were washed with 200 µl FACS buffer, followed by fixing with 200 µl of Cytofix (BD Biosciences, Rockville, MD, catalog #554655) and incubated on ice for 10 minutes protected from light. Cells were washed with 200 µl of FACS buffer and resuspended in 200 µl of FACS buffer and analyzed.

Quantification of integration by Alu-PCR. 293T huDC-SIGN cells seeded at 5E5 cells/well in six-well plates were transduced in triplicate with 2E9 genomes per well of vector. At 48 hours posttransduction, cells were harvested and genomic DNA extracted using the DNeasy Kit (Qiagen). Genomic DNA was analyzed using an Alu-LTR–based nested-PCR assay, which amplifies only provirus sequences that have been integrated into the host genomic DNA. We introduced the following modifications into the previously published method.¹⁰ Platinum Taq (Life Technologies, Grand Island, NY) was used for the first round of amplification in a final reaction volume of 25 µl. The first-round PCR cycle conditions were as follows: a denaturation step of 2 minutes at 95 °C and then 20 cycles of amplification (95 °C for 30 seconds, 55 °C for 30 seconds, 72 °C for 90 seconds). Nested PCR was performed using EXPRESS qPCR Supermix Universal (Life Technologies) and 100 nmol/l of probe MH603⁶⁶ in a final volume of 25 µl. The nested PCR protocol began with a 2 minutes hold at 50 °C and a 10-minute denaturation step at 95 °C, followed by 40 cycles of amplification (95 °C for 15 seconds, 60 °C for 30 seconds). All amplification reactions were performed using the Bio-Rad CFX (-96 or -384 model; Bio-Rad Laboratories, Hercules, CA). The copy number of integrated provirus was calculated in reference to a standard curve generated by parallel nested Alu-PCR of a reference 293T cell line containing integrated provirus of known copy number, diluted over a 5-log range. The total genomic DNA in the standard curve was normalized by mixing with genomic DNA from nontransduced cells; each standard and unknown sample contained 100 ng total genomic DNA. Our assay allowed the detection of 58 proviruses (Experiment 1) or 4 proviruses (Experiment 2) in 100 ng of genomic DNA.

Quantification of integration by neomycin resistance. Vectors encoding GFP-T2A-NeoR antigen were independently analyzed for integration rate by neomycin resistant colony formation. HT1080 huDC-SIGN cells were transduced in six-well plates with 0.5 ml of serially diluted vector (normalized by genomes) for 2 hours, after which 2 ml of complete medium was added. At 24 hours posttransduction, cells were fed with medium containing 800 µg/ml G418 (Life Technologies). Cells were then grown without passaging for 11–13 days under G418 selection, after which colonies were visualized by staining with crystal violet (BD Biosciences). Total integration events were calculated as follows: (# of colonies) × (dilution factor) = integration events.

Quantification of integration by GFP expression. For vectors encoding GFP-T2A-NeoR antigen, relative integration rate was measured by GFP expression over time in bulk culture. HT1080 huDC-SIGN cells were transduced in six-well plates with equal amounts of WT/703 or D64V/704 vector (normalized by genomes) in 0.5 ml for 2 hours, followed by the addition of 2 ml of complete medium. Transduced cells were maintained in medium without drug selection for up to 30 days, passaging at regular intervals. During this period, cells were periodically analyzed for GFP expression by flow cytometry (Guava EasyCyte Plus; Millipore).

SUPPLEMENTARY MATERIAL Figure S1. huDC-SIGN expressing cell lines. Figure S2. SINVHR, SVGmu, and E1001 Sequence Alignments. Figure S3. Increased Transduction of 293T huDC-SIGN cells with E1001 pseudotyped vector produced in the presence of Kifuensine. Figure S4. Schematic representation of the high-mannose glycosylation state on a single SINV glycoprotein. Figure S5. ID-VP02 transduces human dendritic cells from multiple donors in a Vpx dependent manner. Figure S6. Monocyte derived dendritic cells from 5 donors transduced with ID-VP02. Figure S7. IFN-gamma treatment downregulates DC-SIGN expression. Figure S8. ID-VP02 transduces dendritic cells generated from CD34+ progenitor cells. Figure S9. 3 days is sufficient to generate MDDCs. Figure S10. ID-VP02 targets human dendritic cells in PBMCs from multiple donors in a Vpx dependent manner. Figure S11. Dendritic cells within PBMCs from 5 donors transduced with ID-VP02. Figure S12. ID-VP02 transduces dendritic cells, but not macrophages or monocytes.