Engineering Next-Generation BET-Independent MLV Vectors for Safer Gene Therapy.

Retroviral vectors have shown their curative potential in clinical trials correcting monogenetic disorders. However, therapeutic benefits were compromised due to vector-induced dysregulation of cellular genes and leukemia development in a subset of patients. Bromodomain and extraterminal domain (BET) proteins act as cellular cofactors that tether the murine leukemia virus (MLV) pre-integration complex to host chromatin via interaction with the MLV integrase (IN) and thereby define the typical gammaretroviral integration distribution. We engineered next-generation BET-independent (Bin) MLV vectors to retarget their integration to regions where they are less likely to dysregulate nearby genes. We mutated MLV IN to uncouple BET protein interaction and fused it with chromatin-binding peptides. The addition of the CBX1 chromodomain to MLV INW390A efficiently targeted integration away from gene regulatory elements. The retargeted vector produced at high titers and efficiently transduced CD34+ hematopoietic stem cells, while fewer colonies were detected in a serial colony-forming assay, a surrogate test for genotoxicity. Our findings underscore the potential of the engineered vectors to reduce the risk of insertional mutagenesis without compromising transduction efficiency. Ultimately, combined with other safety features in vector design, next-generation BinMLV vectors can improve the safety of gammaretroviral vectors for gene therapy.

Introduction

Stable integration of retroviral vectors encompassing a therapeutic transgene enables gene correction of severe blood and immune disorders. Over the past 25 years, murine leukemia virus (MLV)-based vectors have shown therapeutic benefit in gene therapy studies for primary immunodeficiencies (PIDs), such as X-linked severe combined immunodeficiency (SCID-X1), adenosine deaminase deficiency-severe combined immunodeficiency (ADA-SCID), and Wiskott-Aldrich syndrome (WAS).1, 2, 3, 4 MLV-based vectors were successfully used in the first clinical trials for ADA-SCID.5, 6, 7 This led to the recent European approval of a retrovirus-based gene therapy product (Strimvelis; GSK GlaxoSmithKline Pharmaceuticals) to treat patients that lack a suitable human leukocyte antigen (HLA)-matched related stem cell donor.8, 9 However, in clinical trials for other PIDs, several patients developed leukemia or myelodysplasia, raising concerns about the safety of gene therapy.10, 11, 12, 13, 14 These side effects have been directly attributed to the integration pattern and vector design. Insertional mutagenesis occurred as a consequence of vector integration preference in proximity of proto-oncogenes and activation by strong viral promoter and enhancer elements in the long terminal repeat (LTR) of retroviral vectors. To prevent insertional mutagenesis, self-inactivating (SIN) vectors with deleted enhancer sequences were designed. The lack of promoter/enhancer activity is compensated by weak heterologous promoters to drive transgene expression, such as the elongation factor 1 α short (EFS) and phosphoglycerate kinase (PGK) promoters.15, 16 Additionally, introduction of genetic insulator sequences has improved the safety of viral vectors.17, 18, 19, 20, 21, 22 The efficacy of these modified vectors was confirmed in pre-clinical studies and they are now in phase I/II clinical trials for several PIDs.19, 23, 24, 25, 26, 27

A complementary approach to improve the safety of gene therapy is to alter the integration pattern, directing integration away from potentially unsafe regions. Gammaretroviral integration is not random, but rather is dictated by host cellular cofactors, such as the bromodomain and extraterminal domain (BET)-containing family of proteins (BRD2, BRD3, and BRD4) that serve as anchors on the host chromatin.28, 29 A motif in the unstructured C-terminal tail of MLV integrase (IN) interacts with the extraterminal (ET) domain of BRDs, where the latter tethers the retroviral pre-integration complex (PIC) to chromatin regions enriched in BET proteins and thereby defines the integration profile.28, 29, 30 Deletion of the C-terminal domain (Δ23 amino acids [aa], IN1–380) or a single substitution (INW390A) uncouples the BET interaction, resulting in BET-independent (Bin) MLV vectors that transduce target cells at wild-type (WT) efficiency but with diminished integration in the vicinity of retroviral integration markers.

Here, we developed next-generation BinMLV vectors with a potentially safer integration profile and lower propensity to activate nearby genes in an effort to alleviate the risk of insertional mutagenesis by interfering with the chromatin-tethering process. We linked chromatin binding peptide sequences to the C-terminal end of BinMLV IN and demonstrated that fusion of these peptides to BinMLV IN generates vectors that produce at high titers and transduce cells at wild-type efficiency. The addition of the chromodomain of CBX1 to MLVIN_W390A efficiently retargeted integration away from gene regulatory elements. More importantly, the retargeted vector transduced human CD34+ hematopoietic stem cells (HSCs) at wild-type efficiency, while genotoxicity assays revealed reduced transformation potential.

Results

Efficient Transduction and Integration of Next-Generation BinMLV Vectors

To direct BinMLV integration away from potentially unsafe chromosomal regions, we tailored the chromatin-tethering process by fusing tethering peptides (between 16 and 61 aa long) to the C-terminal end of INW390A in the MLV packaging plasmid. We opted for peptides that bind histone markers that are widely spread across the chromatin (Figure 1A; Table 1). On one hand, we used peptides derived from cellular proteins that bind specific epigenetic histone modifications, such as the chromodomain of heterochromatin-binding protein 1β (CBX1, aa 20–73) and the chromodomain of Y-like protein (CDYL; aa 1–60),32, 33 giving rise to INW390A-CBX and INW390A-CDYL, respectively. Alternatively, we fused virus-derived peptides, such as the tethering domain of the human papilloma virus (HPV8) E2 protein (aa 240–255) and the N-terminal end of Kaposi sarcoma’s latency associated nuclear antigen (LANA; aa 1–31), which bind to core histone 2A and 2B, resulting in INW390A-E2 and INW390A-LANA, respectively.

Figure 1

Transduction Efficiencies of Next-Generation BinMLV Vectors

(A) Schematic representation of MLV-based vector production. Structure of INWT, INW390A, and the peptides fused to INW390A are highlighted. The N-terminal HHCC zinc binding domain, the catalytic core domain (CCD), and the C-terminal domain (CTD) are indicated. Red arrowheads indicate the position of the W390A point mutation. The size of the fused peptides is proportionally represented. (B) FACS analysis of SupT1 cells transduced with equal RTUs of the indicated vectors at different MOIs. Three days post-transduction, the percentage of EGFP-positive cells was determined. Average values and standard deviations of triplicate measurements are shown. Data represent measurements from a representative experiment. (C) Mean fluorescence intensity of SupT1 cells transduced with the indicated next-generation BinMLV vectors 3 days post-transduction. Average values and standard deviations of triplicate measurements are shown. Data represent measurements from a representative experiment. ψ, packaging signal; CMV, cytomegalovirus promoter; GAG, group-specific antigen; IN, integrase; LTR, long terminal repeat; PolyA, polyadenylation signal; Pol, polymerase; PR, protease; RT, reverse transcriptase; VSV-G, vesicular stomatitis virus glycoprotein G.

Table 1

Overview of the Tethering Domains Fused to BinMLV Vectors

Protein	Accession Number	Selected Peptide	Binding Site	Sequence
CBX1	P83916	CD1 CBX	H3K9me2	EYVVEKVLDRRVVKGKVEYLLKWKGFSDEDNTWEPEENLDCPDLIAEFLQSQKT
			H3K9me3
CDYL	Q9Y232	CD1 CDYL	H3K9me2	LMTFQASHRSAWGKSRKKNWQYEGPTQKLFLKRNNVSAPDGPSDPSISVSSEQSGAQQPPA
			H3K27me2
			H3K27me3
HPV8 E2	P06422	E2	H2A/H2B	QTETKGRRYGRRPSSR
KSHV LANA	E5LC01	LANA	H2A/H2B	MAPPGMRLRSGRSTGAPLTRGSCRKRNRSPE

Overview of the aa sequences and binding sites of CBX1 and CDYL chromodomains, HPV8 E2, and KSHV LANA used in this study. The aa position of each peptide in the respective protein is indicated in superscript. Protein sequences were downloaded from the UniProt database.CBX1, chromobox homolog 1; CD, chromodomain; CDYL; chromodomain protein Y-like; HPV, human papilloma virus; KSHV, Kaposi sarcoma human virus; LANA, latency-associated nuclear antigen.

The respective packaging plasmids were subsequently used to produce vesicular stomatitis virus glycoprotein G (VSV-G) pseudotyped MLV-based vectors encoding an LTR-driven EGFP reporter (MLVIN_W390A-CBX, MLVIN_W390A-CDYL, MLVIN_W390A-E2, and MLVIN_W390A-LANA; Figure 1A). In line with previous results, transduction efficiencies in SupT1 cells were at similar levels for MLVIN_W390A and MLVIN_WT (Figure 1B). The addition of peptide sequences to the C-terminal end of MLVIN_W390A resulted in BinMLV vectors that transduced as efficiently as MLVIN_WT at different MOIs (Figure 1B) and resulted in comparable expression levels (measured as mean fluorescence intensities [MFIs] at day 3; Figure 1C). Transduction efficiencies and MFIs were corroborated at 10 days post-transduction, underscoring stable expression for the respective integrated vectors (Figures S1A and S1B). Similar data were obtained following transduction of HeLa cells (data not shown). Collectively, these results indicate that the addition of peptide sequences at the C-terminal end of MLVIN_W390A does not impair vector integrity nor transduction efficiency compared to MLVIN_WT and MLVIN_W390A.

MLVIN_W390A-Peptide Fusions Redistribute Bin Vector Integration

Next, we set out to investigate whether the respective IN chimeras redistribute MLV vector integration. Integration sites were amplified and sequenced, yielding a total of 43,676 unique sites and their computationally generated matched random control (MRC) sites. Initially, we evaluated integration frequencies relative to transcription start sites (TSSs), CpG islands (typically enriched in/near housekeeping gene promoters), and DNase hypersensitive sites (DHSs), both surrogate markers for open areas of active chromatin. In accordance with our previous work, MLVIN_W390A integration near TSSs and CpG islands decreased ∼2-fold (Figure 2A; Table S1A). Whereas integration frequencies for MLVIN_W390A-CDYL and MLVIN_W390A-E2 did not differ from MLVIN_W390A near TSSs or CpG islands, the integration frequencies for both MLVIN_W390A-CBX and MLVIN_W390A-LANA were ∼4-fold and ∼2-fold lower when compared to MLVIN_WT and MLVIN_W390A, respectively (Figure 2A; Table S1A; p < 0.001, compared to MLVIN_W390A). The detargeting effect of MLVIN_W390A-CBX and MLVIN_W390A-LANA was also evident near DHSs. 33% of MLVIN_W390A integrations occurred in a 2-kb window around DHSs (∼13% less than MLVIN_WT), whereas only 25.66% and 27.84% of MLVIN_W390A-CBX and MLVIN_W390A-LANA integrations occurred in this window (Figure 2A; Table S1A). For comparison, integration datasets of prototype foamy viral vectors (FVs) and HIV-derived lentiviral vectors (LVs) were juxtaposed (Figure 2A). FVs are known to have a lower tendency to integrate near promoter regions compared to MLV vectors. MLVIN_W390A integration near TSSs was comparable to that of FVs (10.17% and 10.3%, respectively), whereas MLVIN_W390A-CBX and MLVIN_W390A-LANA integration occurred ∼2-fold less near TSSs (Figure 2A; Table S1A) yet more frequently than LVs. Similar results were obtained for larger window sizes (4-kb window, data not shown). Together, the data confirm that fusion of the CBX1 chromodomain and LANA peptide to MLV IN shifts vector integration away from the traditional markers associated with MLV integration.

Figure 2

Integration Site Distribution of Next-Generation BinMLV Vectors

(A) Murine leukemia virus (MLV)-based vector integration sites obtained from SupT1 cells and their genomic distribution. Integration percentages in 2-kb windows around TSSs, CpG island midpoints, and DHSs are listed. For comparison, integration of computer-generated match random controls (MRCs) for MLVIN_W390A, prototype foamy viral vectors (FVs), and HIV-derived lentiviral vectors (LVs) were included. P values show significant departures (***p < 0.001, pairwise Fisher test) from MLVIN_W390A. (B and C) Heatmaps summarizing the relation between vector integration site frequency and different genomic (B) or epigenetic features within a10-kb interval (C) in SupT1 cells. Evaluated vectors are indicated above the columns. Features analyzed are shown to the left of the corresponding row of the heatmap. Tile colors indicate whether a particular feature is favored or disfavored for integration of the respective datasets relative to their MRCs, as detailed in the colored ROC area scale at the bottom of the panel. Number of integration sites is indicated below each column for the respective vector. P values show the significance of departures from MLVIN_W390A integration sites in SupT1 cells (***p < 0.001, Wald statistics referred to χ2 distribution). CpG, CpG-rich island; DHS, DNase I-hypersensitive site; TSS, transcription start site.

In a next step, we analyzed integration preferences relative to a wider set of genomic features to evaluate overall vector integration (represented by heatmaps, Figure 2B). The tile color depicts the correlation for an integration dataset with the respective genomic feature (left) relative to the MRC, as quantified by the area under the receiver operating characteristic (ROC) curve. Asterisks indicate statistical significance of the integration site distributions of the respective vectors relative to that of MLVIN_W390A. MLVIN_W390A-CDYL and MLVIN_W390A-E2 integration profiles were similar to that of MLVIN_W390A (Figure 2B), whereas MLVIN_W390A-CBX and MLVIN_W390A-LANA showed significant differences. The addition of CBX1 chromodomain to MLV IN_W390A (MLVIN_W390A-CBX) significantly shifted integration for most of the genomic features (p < 0.001, compared to MLVIN_W390A), resulting in redistributed integration with a more random pattern (Figure 2B, compare tile color of MLVIN_W390A and MLVIN_W390A_CBX; tile colors shift toward gray). On the other hand, MLVIN_W390A-LANA only showed significant effects in smaller window sizes for the typical determinants of MLV vector integration, such as TSSs, CpG islands, and DHSs.

Since some of the peptides recognize specific chromatin marks, we also analyzed integration preferences near a collection of epigenetic features (Figure 2C). In line with previous data, uncoupling of BET interaction (MLVIN_W390A) yields a more random integration pattern (compare colors between MLVIN_WT and MLVIN_W390A; tile colors shift toward black). Nonetheless, MLVIN_W390A integration still correlates with histone marks associated with open and transcriptionally active chromatin, such as H3 acetylation, H3K4 mono-, di-, and trimethylation, and H3K36 mono- and trimethylation albeit to a lesser extent (tile colors shift to darker blue; integration is enriched compared to MRC), while disfavoring transcriptionally silent regions or heterochromatin, such as di- and trimethylation of H3K9, H3K27, and H3K79 (tile colors shift to yellow; integration is depleted compared to MRC). Fusion of the CBX1 chromodomain and LANA peptide to INW390A shifted integration away from markers correlating with transcriptionally active open chromatin (blue tiles overall shift toward darker blue; p < 0.001, compared to MLVIN_W390A), whereas the addition of the other peptides had no effect (MLVIN_W390A-CDYL and MLVIN_W390A-E2, Figure 2C). CBX1 is known to bind H3K9me2/3 epigenetic marks via its chromodomain. Interestingly, MLVIN_W390A-CBX shifts integration more into transcriptionally silent heterochromatin regions, which is generally disfavored for integration, marked by di- and/or tri-methylation of H3K9 and H3K27 (p < 0.001, compared to MLVIN_W390A; yellow tiles are darker). Together, these data show that the fusion of peptide tethers to the C-terminal end of MLV INW390A effectively retargets integration. As integration is shifted away from traditional MLV integration markers, known to associate with insertional mutagenesis in gene therapeutic trials, a potentially safer integration site profile might be obtained.

Addition of Small Peptides to MLVIN_1–380 Rescues Its Transduction Defect

Apart from W390A substitution, deletion of the C-terminal tail of MLV IN (Δ23 aa, IN1–380) similarly detargeted MLV integration, but with a transduction efficiency for MLVIN_1–380 that was ∼3-fold lower than for MLVIN_WT. Therefore, we fused the respective peptides as an alternative C-terminal tail to IN1–380 to assess whether transduction efficiency could be improved (MLVIN_1–380-CBX, MLVIN_1–380-CDYL, MLVIN_1–380-E2, and MLVIN_1–380-LANA; Figure 3A). Indeed, the addition of all peptides improved transduction efficiency to near wild-type levels at different MOIs (Figures 3B and S1C) and resulted in similar expression levels (Figures 3C and S1D), underscoring the importance of the C-terminal tail.

Figure 3

Effects of Fusing the Chromatin Binding Peptides in MLVIN_1–380 Backbone

(A) Schematic representation of INWT, IN1–380, and peptides fused to IN1–380. The N-terminal HHCC zinc binding domain, the catalytic core domain (CCD), and the C-terminal domain (CTD) are indicated. The size of the fused peptides is proportionally represented. (B and C) TE (B) and mean fluorescence intensity (C) in SupT1 cells 3 days post-transduction at different MOIs. Average values and standard deviations of triplicate measurements are shown. Data represent measurements from a representative experiment.

Next, we amplified 28,607 unique integration sites (Table S1B) to evaluate whether these IN chimeras also redistributed vector integration. In line with earlier data, MLVIN_1–380 integration was decreased near TSSs, CpG islands, and DHSs compared to MLVIN_WT (Figure S2A; Table S1B). Fusion of CDYL or E2 peptides to MLVIN_1–380 did not alter integration site distribution, whereas fusion of the CBX1 chromodomain and LANA peptide redistributed integration sites in a similar fashion as for the MLVIN_W390A fusions (compare Figures 2A and S2A; compare Tables S1A and S1B). Additionally, MLVIN_1–380-CBX integration redistributed similar to MLVIN_W390A-CBX for a wide range of genomic features (Figure S2B) and histone modifications (Figure S2C), highlighting the specificity of the detargeting effects achieved by fusion of the CBX1 chromodomain or the LANA peptide to MLVIN_W390A and MLVIN_1–380.

Addition of Peptide Tethers Does Not Alter the Local MLV Integration Site Sequence

Retroviral INs show weak but discernable target sequence preferences surrounding the site of integration. This local integration site sequence is mainly determined by IN contacts with the (nucleosomal) DNA template.40, 41 To assess whether the addition of alternative peptide tethers to MLVIN_W390A influenced the local integration site sequence, we constructed sequence logos (Figure S3). Results indicate that the local integration site sequence preferences remained unaffected. Similar results were obtained for MLVIN_1–380 peptide fusions (data not shown).

Evidence for Safer Integration of Next-Generation BinMLVIN_W390A-CBX Vector

Comparative integrome analysis identified MLVIN_W390A-CBX and MLVIN_W390A-LANA as next-generation BinMLV vectors with substantial detargeting from traditional MLV markers without compromising transduction efficiency. Initially, we sought to assess whether the altered integration profiles of next-generation BinMLV vectors are potentially safer. We determined integration frequencies near a set of previously defined genomic regions43, 44 that should be avoided to prevent cellular transformation, such as regions proximal to TSSs (<50 kb TSS), cancer-related oncogenes (<300 kb AllOnco), and microRNA (miRNA) coding regions (<300 kb miRNA), or within transcription units (TUs) and ultraconserved regions (UCRs). For each integrome dataset, the percentage of integrations near these features was determined (Table S2). When assessing the individual safe harbor criteria, MLVIN_W390A-CBX integrated less frequently near each of these features when compared to MLVIN_WT, MLVIN_W390A, and LV integration. Taken together, this computational analysis suggests an overall safer integration profile that is less likely to disturb nearby genes.

Therapeutic Potential of Next-Generation BinMLV Vectors

An essential factor in engineering new viral vectors includes evaluating their therapeutic potential to transduce clinically relevant cells. Therefore, we challenged primary human CD4+ T cells and CD34+ HSCs with MLVIN_WT, MLVIN_W390A, MLVIN_W390A-CBX, and MLVIN_W390A-LANA vectors carrying a more relevant SIN gammaretroviral vector genome (MLV.SIN) driving reporter gene expression from a spleen focus-forming virus (SF) enhancer/promoter,21, 45 referred to as MLV.SIN.SFIN_WT, MLV.SIN.SFIN_W390A, MLV.SIN.SFIN_W390A-CBX, and MLV.SIN.SFIN_W390A-LANA, respectively (Figure S4A) and monitored transduction efficiency (percentage of eGFP+ cells) over time (MOI of 4.5, Figures 4A and 4B, respectively). MLV.SIN.SFIN_W390A-CBX and MLV.SIN.SFIN_W390A-LANA transduced CD4+ T cells and CD34+ HSCs to the same extent as MLV.SIN.SFIN_WT and MLV.SIN.SFIN_W390A and this was sustained over time (fluorescence-activated cell sorting [FACS] plots, Figures 4A and 4B, respectively; compare transduction efficiency (TE) and MFI at day 3 and day 10), indicating the absence of silencing effects due to the retargeted integration preference of next-generation BinMLV vectors. Additionally, we performed colony-forming unit (CFU) assays for CD34+ HSCs harvested from three different donors to determine whether the retargeted next-generation BinMLV vectors efficiently transduced hematopoietic progenitor cells (Figure 4C). After 14 days in culture, the number of colonies derived from CD34+ HSCs transduced with MLV.SIN.SFIN_W390A, MLV.SIN.SFIN_W390A-CBX, and MLV.SIN.SFIN_W390A-LANA was in line with MLV.SIN.SFIN_WT, indicating that BinMLV vectors do not affect the functionality of CD34+ HSCs (Figure 4C). FACS analysis of the respective pooled transduced CFUs revealed comparable percentages of EGFP-positive cells (Figure 4D) and no difference in MFIs could be observed (Table S3) for the different BinMLV vectors, confirming the lack of increased transgene silencing.

Figure 4

Translational Potential of Next-Generation BinMLV Vectors

(A and B) TE of next-generation BinMLV vectors in primary CD4+ T cells (A) and CD34+ HSCs (B) at different time points. Percentage of EGFP-positive cells and mean fluorescence intensities (MFIs) are indicated 3 days and 10 or 13 days post-transduction for CD4+ T cells and CD34+ cells, respectively. Data represent measurements from a representative experiment. (C) Colony-forming unit (CFU) assay of human CD34+ HSCs harvested from three donors and transduced with the indicated vectors. The number of colonies was scored after 14 days. (D) TE (percentage of EGFP-positive cells) of CFU colonies (described in C) at 17 days post-transduction from three donors.

Reduced Transformation Potential of Next-Generation BinMLVIN_W390A-CBX Vector

The intrinsic integration preference of MLV-based vectors has been shown previously to be one of the driving factors of vector-mediated genotoxicity that occurred when the integrated vector dysregulated host genes, leading to oncogenic transformation. Therefore, we evaluated the genotoxic potential of next-generation BinMLV vectors carrying a MLV.SIN-vector genome with an internal SF enhancer/promoter, which is known to trigger insertional transformation events21, 45 (Figure S4A) in the in vitro immortalization assay (IVIM) and the transformational incidence in a murine CFU assay (Figure S4B).47, 48 To ensure the chance of immortalization, at least 55% of cells were transduced, corresponding to a mean vector copy number (VCN) of > 2. Murine hematopoietic lineage-depleted (lin−) bone marrow cells were transduced with the indicated vectors at different MOIs in six independent transductions in three IVIM assays (Figure S4C). Transduction efficiencies and integrated VCNs were comparable for the different vectors (Figures 5A and 5B). The IVIM assay revealed immortalization for all MLV vectors carrying the MLV.SIN.SF vector architecture (Figures 5C and S4C). The replating frequency is a measure for the fitness of clones, while the number of positive assays reflects the incidence of immortalization events. MLV.SIN.SFIN_W390A and MLV.SIN.SFIN_W390A-LANA resulted in replating clone numbers in line with MLV.SIN.SFIN_WT, whereas a reduction in clone numbers was observed for MLV.SIN.SFIN_W390A-CBX (mean number of positive wells; mean replating frequency of 2.7 × 10−3, Figure S4C). MLV.SIN.SFIN_W390A-CBX also displayed a reduced replating frequency/copy number (1.4-fold or ∼30% lower) compared to MLV.SIN.SFIN_WT, although it was not statistically significant (p > 0.05, compared to MLVIN_WT, Mann-Whitney U test, Figures 5C and S4C).

Figure 5

Assessment of Genotoxicity Profile of Next-Generation BinMLV Vectors

(A and B) TE at day 4 (A) and mean vector copy number at day 5 post-transduction (B) of next-generation BinMLV vectors in murine hematopoietic lineage marker-negative (lin−) bone marrow cells at increasing MOIs. MOI 20 was performed only once. (C) Replating frequencies corrected for the mean vector copy number, measured in DNA of mass cultures at day 5; horizontal lines indicate the median for a given vector design. (D and E) Serial CFUs assay. The number of colonies per 5,000 cells plated in methylcellulose is shown. Each dot represents the number of colonies formed for an independently transduced culture. A negative control represents non-transduced lin− cells, while a positive control represents lin− cells transduced with an LTR-SFFV-driven MLV-based vector. After the first round (D), the number of colonies was scored and cells from positive assays were harvested and re-seeded at 5,000 cells for second round colony formation (E). Colony counts > 200 colonies are shown as 220. The experimental setup is shown in Figure S4B. VCN, vector copy number.

The serial replating CFU assay is a method used to confirm cellular anchorage-independent growth in vitro. The assay provides a stringent method for the detection of the tumorigenic potential of transformed murine HSCs in semi-solid medium.47, 48 In a parallel approach, we sought to assess the serial colony-forming capacity of the transduced lineage marker-negative cells 2 weeks after expansion in the IVIM assay. We included cells transduced with an LTR-driven MLV-based vector as an additional positive control. Cells were plated at a density of 5,000 cells/well in semi-solid medium (Figure S4B). After 10 days in culture (first round, Figures 5D and S4B), the number of colonies obtained for all MLV.SIN.SF vectors was not significantly different from that of non-transduced cells (p > 0.05, compared to negative control, Mann-Whitney U test, Figure 5D), whereas the number of colonies for the positive control was significantly higher (p < 0.05, compared to negative control, Mann-Whitney U test, Figure 5D). After isolating the cells from the colony assays, 5,000 cells were replated (second round, Figures 5E and S4B). MLV.SIN.SFIN_WT- and MLV.SIN.SFIN_W390A-transduced cells formed numbers of colonies comparable to the positive control (p > 0.05, compared to positive control, Mann-Whitney U test, Figure 5E), whereas the number of colonies from MLV.SIN.SFIN_W390A-CBX transduced cells was significantly reduced (p < 0.05 compared to MLV.SIN.SFIN_WT, Mann-Whitney U test, Figure 5E). Taken together, these data indicate that engineering the MLV vector configuration by modifying the IN protein (INW390A-CBX) reduced the outgrowth of replating clones, indicating a potentially safer profile.

Discussion

Retroviral vector technology offers great potential to treat genetic disorders and is a powerful tool for long-term correction of genetic defects in a variety of severe hematological disorders.1, 50, 51 Despite the initial success, a subset of patients developed serious adverse events, such as leukemia or myelodysplasia, that could be directly related to the design and characteristics of the viral vector used.10, 11, 12, 13, 14 Understanding the mechanisms of retroviral vector genotoxicity is, therefore, essential to engineer improved viral vectors with reduced genotoxic potential.

Vector-mediated genotoxicity is defined by (1) the specific integration profile and (2) the design of the integrating proviral genome. Each retroviral family displays a specific integration profile. MLV integration is significantly enriched near TSSs and active enhancer regions and thus potentially causes insertional mutagenesis. Significant efforts have been made to develop next generations of retroviral vectors with reduced genotoxic potential, such as SIN vectors and weaker internal promoters.19, 21, 52, 53, 54, 55, 56, 57, 58 Nonetheless, their integration is still targeted to gene regulatory regions,38, 59 where they have the potential to disrupt or dysregulate the transcription of nearby genes by other mechanisms. Thus, the development of viral vectors that integrate away from genes may be safer for clinical applications. One approach is the development of other viral vector platforms with more favorable genomic distributions, like lentiviral, foamy, and alpharetroviral vectors.25, 60, 61, 62 Alternatively, retroviral vectors can be re-engineered to obtain a more desirable integration pattern that is detargeted from its traditional chromosomal locations and ultimately only occurs at the preferred sites of the host-cell chromosome.

Retroviral integration site selection is dictated by the interaction between the viral IN as part of the PIC and cellular cofactors. Previously, we re-engineered the MLV- and HIV-cellular tethering cofactors (BET and LEDGF/p75, respectively) and demonstrated efficient redistribution of retroviral integration without compromising transgene expression.29, 63, 64, 65 However, this approach requires the introduction (at least transient) of artificial anchors in target cells prior to application of the therapeutic vectors,63, 64 which is not always desirable in a clinical setting. A more straightforward strategy is to directly engineer vector particles to contain proteins with adapted or unique binding domains to direct integration. Other groups attempted to redistribute MLV integration through modifications of the MLV Gag protein by fusion of tethering peptides to rescue a mutated MLV p12 protein.66, 67 However, integration site distribution of the engineered MLV p12 chimeras was not altered,66, 67 suggesting that the primary role of MLV p12 is tethering the virus/vector to the condensed host-cell chromatin rather than targeting the genomic integration (reviewed in Rein). Analysis of MLV IN-BET interaction enabled us to generate Bin MLV vectors. Interaction with BET proteins was uncoupled either by deletion of the C-terminal tail of MLV IN (MLVIN_1–380) or by a single point mutation (MLVIN_W390A).

Here, we successfully modified the MLV integration profile by fusing alternative chromatin binding peptides to BinMLV IN, resulting in a significantly different integration site pattern (Figure 2). The addition of the CBX1 chromodomain or LANA peptide redistributed MLV integration to the same extent relative to the typical MLV markers (TSSs, DHSs, and CpG islands), with a more than 3-fold reduction in integration frequency near TSSs and CpG islands compared to wild-type MLV (Figures 2A and S2A; Table S1). Comparative integrome analysis near a range of genomic and epigenetic features corroborated that INLANA vectors (MLVIN_W390A-LANA and MLVIN_1–380-LANA) and INCBX vectors (MLVIN_W390A-CBX and MLVIN_1–380-CBX) redistribute along the chromatin, although the effect was much more pronounced for INCBX vectors (Figures 2B and 2C and S2B and S2C). A similar strategy was employed by Hocum et al., where the full CBX1 protein was fused to the IN of foamy virus and additional modifications in the FV Gag protein were required to achieve significant effects on retargeting FV integration. The newly generated foamy retroviral vectors (FV) integrated ∼2-fold less frequent near genes and proto-oncogenes. In the case of MLV, fusion of a single CBX1 chromodomain to the C-terminal end of MLV INW390A was sufficient to shift integration toward epigenetic markers for transcriptionally silent regions (H3K9me2/3 and H3K27me3), known to be bound by CBX1 (Figure 2C; Table S1).

Next to the retargeting effect, next-generation BinMLV vectors can be produced at high titers and efficiently transduce clinically relevant cells such as primary CD4+ T cells and CD34+ HSCs without any apparent transgene silencing (Figure 4), highlighting the translational potential of these vectors. Finally, we assessed the newly engineered Bin MLV vectors (MLVIN_W390A-CBX and MLVIN_W390A-LANA) in the IVIM assay as well as in a murine serial replating CFU assay to predict the genotoxic profile (Figure S4B). BinMLV and wild-type MLV-based vectors were produced carrying the same SIN gammaretroviral vector genome (MLV.SIN.SF) with an internal SF enhancer/promoter to drive EGFP expression (Figure S4A), a design known to trigger insertional transformation.21, 45 Whereas MLV.SIN.SFIN_WT, MLV.SIN.SFIN_W390A, and MLV.SIN.SFIN_W390A-LANA displayed a comparable frequency of IVIM clones, immortalization events were slightly reduced for MLV.SIN.SFIN_W390A-CBX (Figures 5C and S4C) but were not significantly different. The results of the IVIM assay in our experiments confirmed that the strong spleen focus-forming virus (SFFV) enhancer/promoter present in all vector genomes produced replating clones with similar efficiency in all cultures independent of the IN, while the low replating in two assays of MLVIN_W390A-CBX may be attributed to the detargeted integration pattern. Hence, at least for SFFV, promoter strength has a greater impact on the generation of insertional mutants than integration site preference in the IVIM assay.

To further address whether detargeted BinMLV vectors may be less genotoxic, we employed serial replating CFU assays of expanded murine HSCs47, 48 to assess vector integration-related genotoxicity. The murine CFU assays demonstrated that MLV.SIN.SFIN_WT and MLV.SIN.SFIN_W390A potently transformed murine HSCs, whereas fusion of the CBX1 chromodomain to MLV INW390A (MLV.SIN.SFIN_W390A-CBX) resulted in a significantly reduced transformational incidence (p < 0.05, Figure 5E).

In conclusion, we demonstrate the potential to engineer MLV-based vectors that detarget from unsafe regions by fusing peptide fragments to the C-terminal end of MLV IN. The mere addition of the CBX1 chromodomain (and LANA peptide, to a lesser extent) was sufficient to detarget integration preference away from the traditional markers of MLV integration. In order to translate these findings into suitable vectors for the (pre)clinical field, the performance of the new Bin MLV should be evaluated in more relevant pre-clinical safety assays and bone marrow/HSC transplantation assays to validate their improved safety profile. Together, our findings will help achieve better control of MLV-based vector integration preferences. Combining next-generation BinMLV packaging constructs with next-generation SIN gammaretroviral vector architectures that incorporate a weaker physiological promoter less likely to dysregulate nearby genes19, 70 will lead to an additional reduction in genotoxicity.

Materials and Methods

Plasmids

BinMLV IN was cloned as previously described. Chromodomians of CBX and CDYL fusions were cloned with gBlocks (IDT) in PacI- and NotI-digested pcDNA3.MLV.gp packaging plasmid, a kind gift from Prof. Axel Schambach. HPV8 E2 and LANA peptide fusions were introduced into the indicated vectors by an oligonucleotide annealing strategy using the same restriction sites. Oligonucleotide sequences are listed in Table S4. All enzymes were purchased from Thermo Fisher Scientific. The integrity of all plasmids was verified by DNA sequencing.

Cell Culture

SupT1 cells were cultured in RPMI-1640 medium (Gibco-BRL/Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (Gibco-BRL) and gentamicin (50 μg/mL; Gibco-BRL). HeLa cells were cultured in DMEM (Gibco-BRL) supplemented with 8% heat-inactivated fetal bovine serum and gentamicin. All cells were grown in a humidified atmosphere with 5% CO2 at 37°C.

Primary Cell Purification

Human peripheral blood mononuclear cells (PBMCs) were purified from buffy coats of three different donors, obtained from the Red Cross blood transfusion center, using density-gradient centrifugation (Lymphoprep; Axis-Shield). Primary CD4+ T cells were selectively enriched using bi-specific monoclonal antibody (mAb) CD3.8 (0.5 μg/mL, NIH AIDS Reagents Program; https://www.aidsreagent.org) for 5 days. CD4+ T cells were cultured in RPMI medium supplemented with 15% FBS, gentamycin, interleukin (IL)-2 (100 U/mL; Peprotech), and MEM Non-Essential Amino Acids Solution (MEM NEAA) (50 μg/mL; Gibco-BRL), referred to as T-cell medium (TCM). CD34+ HSCs were positively selected with anti-CD34-conjugated microbeads according to the manufacturer’s instructions (MACS; Miltenyi Biotec) and stimulated for 48 hr in StemSpan SFEMII medium containing CC100 Cytokine Cocktail (STEMCELL Technologies).

Retroviral Vector Production and Transduction

Viral vectors were produced as previously described. Briefly, MLV-based vectors were produced by a triple polyethylenimine (PEI)-based or Ca-phosphate transfection of 293T cells with a pVSV-G envelope, pcDNA3.MLV.gp packaging plasmids or their derived fusions (see above), and p450-GFP transfer plasmid (kindly provided by F.D. Bushman) encoding an LTR-driven EGFP reporter. For the IVIM assay, the transfer plasmid pSRS11.SF.GFP.pre (referred to here as MLV.SIN.SF.EGFP.pre) was used, which was kindly provided by Axel Schambach.

Produced vectors were concentrated by tangential flow filtration and normalized based on RT units (RTUs; non-functional titration) by the SYBR Green I product-enhanced reverse transcriptase assay (SG-PERT). Subsequently, functional transducing titers were determined in SupT1 cells reaching titers > 107 TU/mL. For transduction of laboratory cell lines, SupT1 cells (12 × 104/well) and HeLa cells (2 × 104/well) were seeded in 96-well plates and transduced with a MOI of 1 and 3 of the respective vectors.

Seventy-two hours post-transduction, 50% of the cells were harvested for FACS analysis, while the remaining 50% were cultured for 10 days post-transduction for a second FACS analysis and to perform integration site sequencing. Prior to primary cell transduction, CD4+ T cells (25.104/well) and CD34+ HSCs (10.104/well) were pre-stimulated for 5 days in TCM and 2 days in StemSpan medium enriched with CC100 Cytokine Cocktail, respectively. An MOI of 4.5 of the different vectors was applied by spinoculation (2 hr, 1,200 g). Cells were analyzed for EGFP expression by flow cytometry at the indicated time points.

Genomic DNA Isolation and qPCR

Genomic DNA (gDNA) isolation and qPCR were performed as previously described. Briefly, 2 million cells were pelleted and genomic DNA was extracted using a Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich). Samples corresponding to 700 ng genomic DNA were used for analysis. Each reaction contained 12.5 μL iQ Supermix (Bio-Rad), 40 nM forward and reverse EGFP primer, and 40 nM EGFP probe in a final volume of 25 μL. RNaseP or β-actin was quantified as the endogenous control (TaqMan RNaseP control reagent; Applied Biosystems). Samples were run in triplicate for 3 min at 95°C followed by 50 cycles of 10 s at 95°C and 30 s at 55°C in a LightCycler 480 (Roche Applied Science). Analysis was performed using the LightCycler 480 software supplied by the manufacturer.

Recovery of Integration Sites and Analysis of Integration Site Distributions

Recovery of integration sites was performed as previously described. Briefly, linkers were ligated to restriction enzyme-digested (MseI) genomic DNA isolated from transduced cells and virus-host DNA junctions were amplified by nested PCR. Samples were individually barcoded with the second pair of PCR primers to generate 454 libraries. PCR products were purified and sequenced using 454/Roche pyrosequencing (titanium technology). Reads were quality-filtered by requiring perfect matches to the LTR linker, barcode, and flanking LTR and were subsequently mapped to the human/mouse genome. All sites were required to align to the reference genome within 3 bp of the LTR edge. To control for possible biases in the datasets due to the choice of the MseI restriction endonuclease in cloning integration sites, random control sites were generated computationally and matched to experimental sites with respect to the distance to the nearest MseI cleavage site (MRC).74, 75 To do so, each experimental integration site was paired with three random control sites in the genome with respect to the distance to the nearest MseI cleavage site in the genome. A more detailed explanation can be found in the supplemental guidelines included in Ocwieja et al.

Analyses were carried out as described in Marshall et al. A detailed account of the statistical methods used and the methods for forming and analyzing heatmaps using ROC curves can be found in Brady et al. Consensus sequence analysis at the point of integration was performed using WebLogo3 (http://140.114.98.75/weblogo/). For association with specific genomic features, the distance of each integration site (in kilobases) to the respective genomic feature was calculated (midpoint of the CpG island or DHS, and the X5-end of genes as a measure for the TSS). Integration sites upstream of the genomic feature were given negative kilobase values, while downstream integration sites were calculated as positive. For heatmaps, comparisons were carried out over three different interval sizes surrounding each integration site (5 kb, 10 kb, and 50 kb), since previous studies have shown that the interval sizes chosen for comparison can influence the conclusions. In this study, results were in line for each interval size examined (data not shown). Only the data for the 10-kb interval are shown. In the heatmap, the distribution of experimental MLV sites is normalized to that of the MRC sites, as a control for recovery bias due to cleavage by restriction enzymes (in our case, MseI).74, 75 Results of statistical tests comparing the distributions of integration sites to the reference dataset are summarized as asterisks on each tile of the heatmap. Datasets used in the safe harbor analysis were retrieved from the Ensembl and/or UCSC databases (TxDB knownGenes, miRNA biotype, UCR; hg19) using BioMart. The AllOnco list was used for oncogenes as published in Sadelain Analysis was performed according to the parameters defined in Papapetrou.

CD34+ HSC CFU Assay

For human CD34+ HSC CFU assays, freshly purified cells were seeded at 5,000 cells/well in human methylcellulose medium (methocult H4230; STEMCELL Technologies) supplemented with CC100 Cytokine Cocktail. Cells were plated in 35-mm petri dishes and cultured in a fully humidified atmosphere with 5% CO2 at 37°C for 14 days. The number of colonies was scored after 14 days. For FACS analysis of the CFUs, colonies were harvested, washed twice with PBS, and evaluated for EGFP expression.

IVIM Assay

The IVIM assay was performed as described earlier. Briefly, murine lineage marker-negative bone marrow cells were isolated from the tibias and femurs of C57BL/6 mice and enriched for stem and progenitor cells (mouse lineage cell depletion kit; Miltenyi Biotec) and frozen in aliquots. After thawing and 48 hr of prestimulation, 1 × 105 lin− cells were transduced on 2 consecutive days on RetroNectin-coated (Takara; Clontech) wells with a MOI of 5, 10, or 20. TE was analyzed by flow cytometry 4 days thereafter. Cells were expanded for 2 weeks in IMDM, 10% FCS, 1 mM glutamine, 1% (v/v) penicillin/streptomycin, murine stem cell factor (mSCF) (50 ng/mL), human FMS-like tyrosine kinase 3 ligand (hFlt3L) (100 ng/mL), murine IL (mIL)-3 (20 ng/mL), and human IL (hIL)-11 (100 ng/mL; all cytokines purchased from Peprotech) and diluted to a cell density of 500,000 cells/mL approximately twice a week. Cells were the seeded on 96-well suspension plates at a density of 100 cells per well (48 wells seeded from each culture). Replating clones were detected by microscopic scoring. The replating frequency (according to Poisson distribution) was calculated with L-calc (STEMCELL Technologies) and normalized by VCN as determined 5 days post-transduction.

Murine Serial CFU Assay

Murine lineage marker-negative bone marrow (BM) cells were purified and transduced as in the IVIM assay. After 2 weeks of expansion, cells were plated (5,000 cells/well) in methylcellulose (HSC006; R&D Systems) supplemented with 20 ng/mL interleukin-3, and 50 ng/mL murine stem cell factor. The number of colonies was scored after 10 days. For serial replating, colonies were harvested and replated in fresh methylcellulose medium at the same density of 5,000 cells/well for the subsequent round.

Author Contributions

S.E.A., D.V.L., and R.G. designed the experiments. S.E.A., D.V.L., and R.G. wrote the manuscript. S.E.A., D.V.L., J.D.R., F.S., U.M., and R.G. performed experiments and analyzed data. J.D. and L.V. performed bioinformatics for integration site analysis. R.G. and Z.D. supervised the project. All authors read and approved the final manuscript.