Lentiviral vectors escape innate sensing but trigger p53 in human hematopoietic stem and progenitor cells

We performed a time‐course RNA‐Seq analysis on cord blood (CB)‐derived CD34⁺ cells pre‐stimulated with early‐acting cytokines for 24 h and then exposed to either research‐ or clinical‐grade VSV‐g pseudotyped (SIN) LV at a high multiplicity of infection, matching current clinical vector dose requirements. As controls, cells were exposed to Poly(I:C), non‐infectious LV particles lacking the VSV‐g envelope (Bald) or heat‐inactivated vectors to control for contaminants co‐administered with the LV or kept in culture untreated (Fig 1A). The greatest transcriptional variance within our dataset was time in culture, as samples clustered in three distinct temporal groups following principal component analysis (PCA), independently of the treatment group (Fig EV1A). The mere culture of HSPC in the presence of growth‐promoting cytokines resulted in the transcriptional modulation of around 6,000–9,000 genes over time for all treatment categories (Appendix Table S1). For untreated HSPC, the most enriched pathway was the MAPK signaling (Fig EV1B), in accordance with growth factor and cytokine‐induced stimulation (Geest & Coffer, 2009). Poly(I:C)‐exposed HSPC strongly up‐regulated innate immune responses, significantly mobilizing a total of 2691 genes (nominal P < 0.05) as compared to untreated controls (Appendix Table S1). We performed term enrichment analysis of the significantly modulated genes to highlight the biological processes affected by Poly(I:C) exposure in HSPC. Regulation of immune responses, NF‐κB signaling, antiviral responses, programmed cell death, and antigen processing were among the most represented categories (Fig EV1B and data not shown). Remarkably, LV‐exposed HSPC showed much milder responses, modulating 645 and 397 genes (nominal P < 0.05) with research‐ and clinical‐grade vectors, respectively, when comparing to untreated cultured cells (Appendix Table S1). In a more stringent comparison with the respective transduction controls (Bald and inactivated purified LV), the number of genes modulated upon transduction further decreased to 321 and 281 genes (nominal P < 0.05) with research‐ and clinical‐grade vectors, respectively (Appendix Table S1), converging significantly into the DDR and in particular the p53 signaling pathway (Fig 1B). No evidence of significant innate immune activation related to TLR‐signaling or activation of NF‐kB/interferon‐stimulated gene (ISG) transcription could be detected (Appendix Table S2). Instead, analysis of the differentially expressed genes with a nominal P < 0.01 in the LV‐exposed HSPC compared to Bald‐exposed controls revealed the up‐regulation, around 18 h post‐transduction, of a cluster of genes mapping to the KEGG p53 signaling pathway (Fig 1C and D). Up‐regulation of some of the most significantly modulated genes involved in p53 signaling was confirmed by Taqman (Fig 1E) and using a clinical‐grade LV (Fig EV1C).

To further investigate how LV induces p53 signaling in HSPC, we tracked p21 mRNA induction as a marker of p53 activation. Induction of p21 was dose‐dependent and transient as levels normalized between control and transduced cells by day 5 in culture (Fig EV1D and E). It was also specific to human HSPC as CD4⁺ T cells, Lin^‐ murine HSPC, and different human cell lines of hematopoietic origin did not up‐regulate p21 upon LV exposure (Fig EV1F). Overall, LV‐induced p21 expression to a similar extent in all CD34⁺ subpopulations, although the most primitive CD34⁺ CD133⁺CD38⁻ fraction showed higher levels of induction, possibly correlating with the higher transduction levels reached in this cell fraction (Figs 1F and G, and EV1G). Both integration‐competent and integration‐defective LV (IDLV) led to a similar induction of p21 in CB as well as in bone marrow (BM)‐derived CD34⁺ HSPC (Fig 1H) at comparable vector DNA input (Fig EV1H). Of note, triggering was not due to vector stock contaminants, LV particle entry alone nor exposure of cells to viral cores lacking the genome, as neither the Bald vector nor the genome‐less Empty LV lead to up‐regulation of p21 (Fig 1H). In agreement with the IDLV‐mediated up‐regulation, p21 induction still occurred in HSPC transduced in the presence of the integrase inhibitor Raltegravir (Fig 1I), despite an efficient block in LV integration (Fig EV1I). Integration‐independent activation of p53 signaling was confirmed by Western blot, FACS and indirect immunofluorescence (IFI) staining in terms of phosphorylation of p53 at Serine 15, increase in basal p53 levels as well as induction of p21 by both LV and IDLV also in mobilized peripheral blood (mPB)‐derived CD34⁺ HSPC (Figs 1J and K, and EV1J–L). Furthermore, no changes in phosphorylated histone 2AX (ɣH2AX) foci were observed, in line with a DNA break‐independent induction of p53 (Fig EV1L). Finally, p53 signaling was abrogated in the presence of the reverse‐transcriptase inhibitor 3TC (Figs 1I and EV1I and J), suggesting that lentiviral DNA synthesis is required for p53 signaling to occur in HSPC.

Taken together, these observations indicate that reverse‐transcribed unintegrated LV DNA specifically triggers p53 rather than innate immune signaling in human HSPC, despite cells can rapidly up‐regulate these pathways upon Poly(I:C) exposure.

Besides LV, the other viral vectors most often used to introduce genetic material into cells are integrating gamma‐retroviral vectors (ɣRV) and DNA‐based non‐integrating Adeno‐associated vectors (AAV). We tested the capacity of both ɣRV and AAV‐6, that efficiently transduces human HSPC (Wang et al, 2015), to induce p21 in HSPC. AAV‐6‐exposed HSPC showed robust p21 induction similar to LV, while VSV‐g pseudotyped ɣRV only weakly affected p21 expression at routinely used vector doses (Figs 1L and EV1M and N), potentially due to the lower transduction efficiency of ɣRV and given the strong linear correlation between VCN and p21 responses (Fig EV1D and N). Both HIV‐1‐derived LV and AAV‐6 actively import the vector DNA into the nucleus of target cells (Bushman et al, 2005; Nonnenmacher & Weber, 2012). To test whether nuclear import of vector DNA is involved in p53 signaling in HSPC in the context of LV, we generated a vector devoid of the central polypurine tract (cPPT) required for efficient nuclear import of the pre‐integration complex (Follenzi et al, 2000; Zennou et al, 2000). Up to threefold lower nuclear import of the ΔcPPT LV compared to unmodified LV was verified in both 293T cells and HSPC (Fig EV1O and P), and the former vector induced twofold less p21 mRNA compared to control LV in CD34⁺ cells (Fig 1M), suggesting that efficient nuclear import of vector DNA is required to activate p53 signaling in HSPC.

As p53 signaling has recently been linked with type I IFN pathways (Hartlova et al, 2015; Yu et al, 2015), we also examined expression of several ISG in human HSPC after exposure to the different vectors. In agreement with our RNA‐Seq dataset, no activation of ISG could be evidenced in human HSPC upon LV or AAV‐6 transduction (Fig 1N). Conversely, ɣRV triggered significant up‐regulation of ISG independently of reverse‐transcription and integration, as ISG induction occurred also in the presence of the reverse‐transcriptase inhibitor azidothymidine (AZT) or Raltegravir (Figs 1N and EV1Q). ɣRV‐mediated ISG induction was inhibited when the type I IFN receptor signaling was blocked (Fig EV1Q), but it does not explain lack of p53 signaling upon ɣRV exposure, as pre‐treatment of HSPC with type I IFN did not prevent induction of p21 by LV (Fig EV1R). Noteworthy, both LV and ɣRV readily triggered ISG expression in murine Lin⁻ cells (Fig EV1S), indicating that the capacity of LV to avoid type I IFN activation is specific to human cells.

p53 has a pivotal role in regulating HSC quiescence and homeostasis during both steady‐state hematopoiesis and under replicative stress (Liu et al, 2009; Lane & Scadden, 2010; Milyavsky et al, 2010; Mohrin et al, 2010). The p21 induction observed upon LV transduction (Fig EV2A) led to a slight but significant delay in cellular proliferation rates, in particular in the most primitive CD34⁺CD133⁺CD90⁺ fraction (Figs 2A and EV2B). Similar but sustained growth delay was observed in p21 overexpressing HSPC used as a positive control for this assay (Figs 2A and EV2A and B). In agreement, the proportion of cells that had undergone fewer divisions (group C1 in Fig EV2C) tended to be more represented in the LV‐exposed condition as compared to control cells in the bulk CD34⁺ HSPC (Fig EV2D) and transduced HSPC displayed a higher fraction of cells in the G₀ cell cycle phase as compared to controls (Fig 2B). Furthermore, within the transduced population, a stronger proliferation delay was observed for the GFP^high fraction, likely harboring more vector copies, compared to the GFP^low ones (Fig EV2D). In agreement, higher p21 induction was detected by FACS in the more transduced BFP^high cells (Fig EV1J).

LV‐exposed HSPC showed a slight but significant dose‐dependent increase in apoptotic cells (Figs 2C and EV2E). The combined effects of lower proliferation and apoptosis of LV‐transduced HSPC were reflected also in terms of total cell counts over time in culture (Fig 2D). Apoptosis occurred to similar extent in all CD34⁺ subpopulations (Fig 2E), and lower CFU‐GM colony output was observed in LV‐exposed total HSPC (Fig 2F) with the CD133⁺CD38⁻ and CD133⁺CD38^int fractions showing significantly lower CFU‐GM counts as compared to controls (Fig EV2F), correlating with higher p53 activation (Fig 1G). Increased apoptosis followed also exposure of HSPC to AAV‐6 and clinical‐grade LV. Other p53‐independent mechanisms may also lead to apoptosis in human HSPC, as also ɣRV‐exposed cells showed similar percentages of apoptotic cells (Fig EV2G and H). Inhibition of reverse‐transcription, but not of integration, completely inhibited LV‐mediated induction of apoptosis in HSPC, correlating with their impact on p21 induction (Figs 1I and 2G).

To further determine how LV transduction affects human HSC function, we transplanted immunocompromised NSG mice with equal and limiting numbers of CB‐derived CD34⁺ HSPC exposed to either a transducing LV or a genome‐less “Empty LV” control vector. To investigate an eventual selective disadvantage of transduced HSPC, a group of mice received a mix of LV and Empty LV‐exposed cells in a 3:1 ratio (Fig 3A). The level of transduction, p53 activation, and consequent impact on cell viability and clonogenicity of transplanted HSPC was verified in vitro to be comparable to our prior results (Fig EV3A–E). Despite equal cellular input, LV‐exposed HSPC showed a significantly lower engraftment at all time‐points compared to controls (Fig 3B). Decreased engraftment was confirmed also in mPB‐derived HSPC transduced according to the current clinical standard protocol based on two subsequent rounds of transduction with a VSV‐g pseudotyped clinical‐grade LV (Figs 3C and EV3F). Of note, no significant differences in the numbers of CD34⁺ cells retrieved from the bone marrow of NSG mice 16 h after transplantation could be detected between transduced and control HSPC (Fig 3D), suggesting that LV transduction does not alter HSPC homing capacity. Once engrafted, no selective disadvantage of transduced HSPC over controls could be seen in the mix condition. Accordingly, the percentages of transduced GFP⁺ cells remained constant over time (Fig EV3G). LV transduction did not alter lineage composition of human cells in periphery (Fig EV3H), but analysis of the bone marrow at 12 weeks post‐transplantation reflected the levels of human cells in the peripheral blood and confirmed significantly lower engraftment of LV‐exposed HSPC (Fig 3B). Nevertheless, no significant differences in the percentages of CD34⁺ cells could be observed between the different groups, and equal frequency of more primitive CD34⁺CD38⁻ and committed CD34⁺ CD38⁺ cells was seen in the bone marrow (Fig 3E). Within the more primitive CD34⁺CD38⁻ fraction, the proportion of HSC, immature lymphoid progenitors (MLP) and multipotent progenitors (MPP; Doulatov et al, 2012; Notta et al, 2016) was also conserved between groups (Fig 3F) and no differences in lineage composition could be observed in the spleen of primary recipients (data not shown). Upon secondary transplantation, all mice were successfully repopulated independently of the treatment group (Fig 3G). Although a trend toward lower engraftment of LV‐exposed HSPC was still observed in secondary recipients, the differences compared to the control group were no longer significant. Absence of selective advantage of untransduced HSPC was confirmed in this setting as the level of engraftment and percentage of GFP⁺ cells remained stable over time in the mix condition (Figs 3G and EV3I). At the end of the experiment, no major differences in the bone marrow composition were seen in terms of total CD34⁺ HSPC and frequency of more primitive CD34⁺CD38⁻ and committed CD34⁺ CD38⁺ cells (Figs 3H and EV3J). In agreement, a limiting dilution assay (LDA) further confirmed that LV transduction does not significantly alter the long‐term repopulating stem cell frequencies (Fig 3I), although engraftment levels were again slightly lower in the LV‐exposed group due to lower numbers of viable cells infused from matched treatment doses (Fig EV3K).

Overall, these results indicate that although exposure to LV negatively impacts HSPC maintenance ex vivo and engraftment in vivo due to acute induction of apoptosis, it does not affect their homing, composition, lineage output, or long‐term repopulating capacity.

To test whether blocking the p53 signaling during ex vivo HSPC transduction could prevent some of the above‐described functional consequences, we first exposed HSPC to a control LV or an LV over‐expressing a dominant negative form of p53, GSE56 (Milyavsky et al, 2010; Nucera et al, 2016), that we verified to efficiently block p53 signaling in HSPC even upon strong DNA damage by camptothecin (CPT) (Fig 4A). Activation of the p53 signaling in terms of p21 induction upon a second round of LV or AAV6 transduction was completely prevented in GSE56‐expressing cells compared to control transduced counterparts (Fig 4A). In agreement, reduced apoptosis was observed in GSE56‐expressing HSPC upon a second round of transduction (Fig 4B).

p53 signaling can be induced by different upstream signal mediators such as the ATM kinase. Pharmacological inhibition of ATM during LV or IDLV exposure significantly reduced p21 mRNA up‐regulation (Fig 4C), decreased p53 phosphorylation and p21 protein levels (Fig 4D), and partially inhibited induction of apoptosis in HSPC (Fig 4E) without compromising transduction efficiencies (Fig 4F). Positive impact of ATM inhibition on apoptosis, colony output, and cell counts in culture together with reduced p21 induction was confirmed also in the more clinically relevant setting of a two‐hit transduction protocol in mPB CD34⁺ cells using a clinical‐grade LV (Fig 4G–I). Also AAV‐6 mediated p21 induction and apoptosis were prevented by ATM inhibition (Fig 4J–K) while it had no impact of γRV‐mediated induction of ISG (Fig EV3L). Of note, although ATM inhibition improved cell survival of transduced HSPC, it did not affect the observed proliferation delay potentially due to residual p21 activity (Fig EV3L). In agreement with lack of p21 induction, no proliferation delay was detected in γRV transduced HSPC (Fig EV3L). Importantly, transient inhibition of p53 signaling during the ex vivo transduction procedure improved LV‐exposed HSPC engraftment resulting in comparable levels of human CD45⁺ cells detected in the peripheral blood between transduced and Bald‐exposed cells (Fig 4L and M), without altering lineage composition or transduction efficiency in vivo (Fig EV3M and N). Also in this setting long‐term human cell engraftment was less affected by LV exposure, further indicating that activation of the p53 signaling upon transduction predominantly impacts short‐term HSPC.

These results further confirm that increased apoptosis and lower engraftment observed after LV transduction are dependent on p53 activation in human HSPC and that these responses can be at least partially prevented by transient inhibition of the upstream mediator ATM without affecting transduction efficiency.

Lentiviral vector and γRV were produced by transient transfection in 293T cells and were all VSV‐g pseudotyped and concentrated by ultracentrifugation as already described (Montini et al, 2006). After transfection and collection, the purified and clinical‐grade LV underwent two steps of column chromatography, ion exchange and gel filtration, to purify the vector from contaminants due to transient transfection (DNA and debris). Bald vector, an entry‐incompetent LV, was produced omitting the Env‐encoding plasmid during vector production. Empty LV, a genome‐less lentiviral vector was produced omitting the SIN LV PGK‐GFP transfer vector during vector production. To produce the p21 overexpressing LV we replaced the thymidine kinase cDNA of a previously described bidirectional LV expression cassette (Amendola et al, 2005), with the cDNA of the human p21 gene, that is express under the control of the human PGK promoter. The GFP is expressed from the minimal CMV promoter in opposite orientation. The GSE56 overexpressing vector is an already described bidirectional vector which expresses GSE56 and GFP in opposite orientation (Nucera et al, 2016). Purified LV was kindly provided by Molmed S.p.a (Milan, Italy). Inactivated purified LV was obtaining by heating (1h at 56°C). AAV6‐IL2RG‐eGFP was kindly provided by Sangamo Biosciences. Self‐inactivating VSV‐g pseudotyped LV, integrase‐defective LV (IDLV), and γ‐RV stocks were prepared, concentrated by ultracentrifugation, and titered as previously described (Follenzi et al, 2000; Montini et al, 2006; Lombardo et al, 2007). For clinical‐grade LV, a two‐step column chromatography, ion exchange followed by gel filtration, was performed to remove contaminants as previously described (Biffi et al, 2013). All CD34⁺ cells were pre‐stimulated with early‐acting cytokines 24 h prior to transduction at the indicated multiplicity of infection (MOI) as previously described (Petrillo et al, 2015).

After 10 days, colonies were identified as erythroid or myeloid by morphological criteria, counted, plucked, and pooled into sets of three to obtain DNA for quantification of vector content. (NSG) mice were purchased from Jackson laboratory. Human CB‐derived CD34⁺ cells were pre‐stimulated and transduced as indicated. All animal procedures were performed according to protocols approved by the Animal Care and Use Committee of the Ospedale San Raffaele (IACUC 611) and according to the Italian law. From eight up to 10 weeks of age, NSG mice were sublethally irradiated (radiation dose: 200 cGy for mice weighting 18–25 g and of 220 cGy for mice above 25 g of weight) 24 h prior to xenotransplantation. 8 × 10⁴ human CB‐derived CD34⁺ cells were injected into the tail vein of primary NSG mice 24 h after transduction. Peripheral blood was sampled at indicated times post‐transplant and analyzed as previously described (Petrillo et al, 2015). At sacrifice, the cells from the spleen and BM isolated from the primary recipients were analyzed at flow cytometry and the CD34⁺ cells were purified from the BM through positive magnetic bead selection on LD and MS columns (Miltenyi) according to the manufacturer's instructions. Purity was verified by FACS prior to pooling by condition and injection into secondary recipients. Between 9 × 10⁵ and 1 × 10⁶ CD34⁺ cells isolated from the primary hosts were injected into the tail vein of sublethally irradiated secondary NSG mice (8–10 weeks old). Peripheral blood was sampled at indicated times post‐transplant and analyzed as described above. At 14 weeks of age, all mice were sacrificed by CO₂ to analyze the BM and the spleen of secondary mice as described above. Cell counts for all in vivo experiments before and after treatment, prior to transplantation, are provided in Appendix Table S3.

Limiting dilution assay was performed by transplanting into irradiated 8‐week‐old NSG mice four different doses of CB‐CD34⁺ cells counted prior to transduction (3,000, 10,000, 30,000, or 90,000) either for Bald and LV groups. Cells were injected 24 h after the transduction. A mouse was scored as positively engrafted if it had > 0.1% engraftment in multiple lineages in the BM at the time of sacrifice (16 weeks). HSC frequency was estimated by linear regression analysis and Poisson statistics using publicly available ELDA (extreme limiting dilution analysis, http://bioinf.wehi.edu.au/software/elda/) software (Hu and Smyth, 2009).

For homing assays, 5 × 10⁵ CB‐derived CD34⁺ cells (pre‐treatment dose) were transplanted 24 h after transduction into irradiated 8‐week‐old NSG mice that were sacrificed 16 h post‐transplantation for analysis. The whole bone marrow was harvested from the lower leg of the mice. During the FACS analysis, cell count beads (Flow‐Count Fluorospheres) by BECKMAN COULTER were added in each sample to estimate the absolute number of human CD34⁺ cells per sample. Cell counts for all in vivo experiments before and after treatment, prior to transplantation, are provided in Appendix Table S3.

Sequencing was performed on the Illumina HiSeq 2000 platform using SBS 2x100PE protocol (Smyth, 2004). Pathway analysis was initially performed using EnrichrR platform (Chen et al, 2013); most of the advanced network modeling was performed using Cytoscape. The complete RNA‐Seq dataset is available at NCBI, accession number GSE92652.

FACS on HSPC was performed as previously described (Petrillo et al, 2015). The apoptosis assays were performed with the Annexin V Apoptosis Detection Kit I (BD Pharmingen) according to the manufacturer's instructions. The cell cycle analysis was performed by Ki67 (BD Pharmingen) and Hoechst (Invitrogen) staining as previously described (Lechman et al, 2012).

Data are expressed as mean ± SEM. Nonparametric Wilcoxon signed‐rank test was used to assess the fold of expression of specific genes respect to the internal control set value as 1. Comparison of multiple groups was performed with nonparametric (Kruskal–Wallis) for unpaired dataset, while the nonparametric (Friedman test) was used for matched paired observations. In both cases, Dunn's adjustment was used for multiple comparisons. Nonparametric unpaired Mann–Whitney test was used to assess the difference between two groups. Significance was considered at P < 0.05. The number of samples analyzed and the statistical test used are indicated in the legends for each figure.

For further details see the Appendix Supplementary Methods.