Safer, Silencing-Resistant Lentiviral Vectors: Optimization of the Ubiquitous Chromatin-Opening Element through Elimination of Aberrant Splicing

Abstract

Gammaretroviral and lentiviral vectors have been used successfully in several clinical gene therapy trials, although powerful enhancer elements have caused insertional mutagenesis and clonal dysregulation. Self-inactivating vectors with internal heterologous regulatory elements have been developed as potentially safer and more effective alternatives. Lentiviral vectors containing a ubiquitous chromatin opening element from the human HNRPA2B1-CBX3 locus (A2UCOE), which allows position-independent, long-term transgene expression, are particularly promising. In a recently described assay, aberrantly spliced mRNA transcripts initiated in the vector A2UCOE sequence were found to lead to upregulation of growth hormone receptor gene (Ghr) expression in transduced murine Bcl-15 cells. Aberrant hybrid mRNA species formed between A2UCOE and a number of other cellular genes were also detected in transduced human PLB-985 myelomonocytic cells. Modification of the A2UCOE by mutation or deletion of recognized and potential cryptic splice donor sites was able to abrogate these splicing events and hybrid mRNA formation in Bcl-15 cells. This modification did not compromise A2UCOE regulatory activity in terms of resistance to CpG methylation and gene silencing in murine P19 embryonic carcinoma cells. These refined A2UCOE regulatory elements are likely to improve intrinsic biosafety and may be particularly useful for a number of clinical applications where robust gene expression is desirable.

INTRODUCTION

Gene replacement therapy via ex vivo modification of hematopoietic stem cells (HSCs) using gammaretroviral vectors has proved a successful treatment for a number of severe inherited diseases (9, 20). However, clonal cell dysregulation leading to leukemia or myelodysplasia caused by insertional mutagenesis has been reported in several clinical trials in which the gammaretroviral long terminal repeat (LTR) was used to promote transgene expression (10, 11, 21). In these instances, the gammaretroviral LTR elements enhanced expression of an adjacent protooncogene.

Studies on vector integration profiles and rates of insertional mutagenesis have suggested that self-inactivating (SIN) lentiviral vectors based on HIV-1 may represent a safer alternative to LTR-based gammaretroviral vectors (15, 18). As they may also be more efficient for transducing certain specific cell types, they are being developed for increasing numbers of clinical trials (6, 7). For example, a lentiviral vector used to treat X-linked adrenoleukodystrophy incorporated a SIN HIV LTR and an internal MND (myeloproliferative sarcoma virus enhancer, negative-control region deleted dl587rev primer binding site substituted) promoter. No mutagenic events have been observed in patients treated to date, and there has been early evidence for retardation of disease progression (6). Similarly, a SIN LV incorporating a human HBB locus control region mini-HBB cassette has recently been reported to ameliorate the clinical symptoms of beta-thalassemia in one patient, although this effect was at least in part due to vector-mediated clonal erythroid expansion as a consequence of integration into the HMGA2 locus. This specific integration event resulted in aberrant splicing of the cellular mRNA transcript into the vector backbone, premature polyadenylation of important suppressive regulatory micro-RNA sequences, and therefore upregulation in expression of a truncated (exons 1 to 3) HMGA2 gene transcript (7). In other nonclinical contexts, RNA transcripts initiated within a vector and spliced into adjacent cellular genes, forming vector-cellular fusion mRNA that can be translated into full or truncated cellular proteins, have also been detected in vitro (1) and in vivo (17). These studies suggest that aberrant splicing could therefore be a significant risk for vector-mediated toxicities. To address this problem, we have developed an assay to detect insertional mutagenesis by lentiviral vectors through splicing of vector-initiated mRNA transcripts to the growth hormone receptor gene (Ghr) in the interleukin 3 (IL-3)-dependent cell line Bcl-15 (4, 13).

Some regulatory elements, including the gammaretroviral LTR, may be prone to CpG dinucleotide methylation, leading to loss of therapeutic gene expression (21). Future design of self-inactivating (SIN) gammaretroviral or lentiviral vectors will require the identification of suitable promoters that provide more consistent and stable expression of the therapeutic transgene. To this end, ubiquitous chromatin-opening elements (UCOEs) have emerged as interesting candidates, as they provide stable expression of transgenes regardless of integration site (2). The UCOE from the human HNRPA2B1-CBX3 locus (A2UCOE) has been shown to be more resistant to silencing than other commonly used promoters, such as EF-1α, cytomegalovirus (CMV), and spleen focus-forming virus (SFFV) (23, 24). It is devoid of classical enhancer function, based on experiments employing standard transient-transfection assays in four markedly different cell lines (24), which in principle may reduce the risk of enhancer-mediated mutagenesis. It has also been shown to provide relatively efficient therapeutic correction of disease in mouse models (19, 24). The A2UCOE is also able to confer resistance to silencing to an adjacent promoter (22), without necessarily altering any tissue specificity (5), and therefore may have broad application. As the sequence of the A2UCOE used in expression vectors extends into the first intron of either HNRPA2B1 or CBX3 or both (2, 5, 22–24), it has the potential to cause insertional mutagenesis through aberrant splicing. In this study, we tested SIN lentiviral vectors containing the A2UCOE to determine whether this can indeed take place. We show that both the inherent splice donor sites and potential cryptic sites that become activated on native donor site mutation are active from A2UCOE-initiated transcripts. However, these can be prevented from functioning by selected mutations without compromising the ability to regulate silencing-resistant gene expression.

MATERIALS AND METHODS

PCR and mutagenesis primers.

All primers used for vector construction, integration analyses, RT-PCR for mRNA analyses are listed in the Table S1 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012.

Lentiviral vectors.

pHV (12), SFFV gamma C, UCOE gamma C, UCOE eGFP (enhanced green fluorescent protein) (24), and UrMEW (5) vectors are as previously described. PEW, UfPEW, and UrPEW, which shares the vector backbone with UrMEW, were obtained by replacing the UrM element with the hPGK promoter via EcoRI-BamHI digestion, followed by insertion of the 1.5-kb UCOE sequence via blunt cloning of an EcoRV-BamHI fragment into the EcoRI site.

UCOE-MAU4 was derived by cloning a 2.2-kb A2UCOE sequence (2) with a mutated CBX3 splice donor (SD) site and an open reading frame (ORF) coding for the first 10 amino acids of human WASP into UCOE gamma C in place of the WT A2UCOE sequence, using SpeI (New England BioLabs) and AvrII (New England BioLabs, Ipswich, MA) sites. This was constructed by commercially synthesizing (GeneArt, Life Technologies Ltd., Paisley, United Kingdom) a 283-bp fragment extending from the SpeI site 132 bp upstream of the 5′-most CBX3 transcriptional start site, inserting the WASP ORF at the end of the first alternative first exon, and terminating at the AvrII 42 bp downstream of the first alternative first exon. This fragment was isolated as a SpeI-AvrII fragment from the provided pMA plasmid clone and inserted into the UCOE gamma C lentiviral vector, which had been digested with the same restriction enzymes, thereby replacing the wild-type sequence. 2SD-MAU1 was derived from UCOE-MAU4 by mutating the second CBX3 SD site by a PCR-based site-directed mutagenesis by using the UCOE-mut-RC primer.

CBX3 NSD was constructed by a three-piece ligation: the backbone was obtained by digesting UCOE gamma C with NotI (Promega, Madison, WI) and SpeI (New England BioLabs); one insert consisted of vector sequences from the NotI site to the beginning of the 5′ end of UCOE promoter, amplified by PCR using primers NotI-F and UCOE-end-RC; and the other insert consisted of the first CBX-3 exon, which was made by PCR amplification of UCOE gamma C with primers CBX3-NSD-F and UCOE-mut-RC.

MA1082 contains the 2.2-kb A2UCOE driving expression of an eGFP reporter gene (24), but the native splice donor sites of the CBX3 moiety as well as potential cryptic splice donor sites, which conform to the consensus sequence GTXXG, within the exons and intronic regions have been changed to CGXXG. A 1,448-bp genomic fragment extending from the SpeI site at position 132 upstream from the 5′-most CBX3 transcriptional start site to the BamHI site at position 1277 downstream from the start site of CBX3 within intron 2 containing these mutations was synthesized commercially (GeneArt). This mutant A2UCOE fragment was isolated from the provided pMA plasmid by SpeI-EcoRI digestion and used to replace the equivalent region in the wild-type A2UCOE-eGFP lentiviral vector.

MADE2 was derived from MA1082 by making an internal deletion around the second CBX3 SD site. A PCR-based deletion was carried out: PCR amplicons from A2UCOE using the primer pairs MA-10-F–MA-10-RC and MA-Psp-F–MA-Psp-Rc (MA-10-F and MA-Rsp-Rc have 11 bp of sequence complementary to each other) were mixed and subjected to PCR using MA-10-RC–MA-10-F. This amplicon was cloned in to MA1082 using PspXI and PstI sites.

MAS4 was made from MADE2, by internally deleting a region around the first CBX3 SD site. A three-piece ligation was carried out with the MADE2 backbone (digested with PspXI and ClaI), the MADE2 insert (digested with PspXI and AvrII), and the AvrII-CBX3-GFP PCR product using the primer pair GT140–Avr-CBX3-F (digested with ClaI and AvrII).

All vector preparations were produced by cotransfection of the vector construct, HIV Gag-Pol/Rev, and vesicular stomatitis virus G protein (VSV-G) expression plasmids as previously described (4, 13).

Bcl15 cell insertional mutagenesis (IM) assay.

An IM assay specifically selecting for mutants with the vector integration in Ghr locus and subsequent mutant analyses are described in references 4 and 13. Briefly, approximately 4 × 10⁷ Bcl-15 cells were transduced with each vector at a multiplicity of infection (MOI) of 10 in the presence of IL-3. After 48 to 96 h expansion the cells were replated in supernatant containing 1 μg/ml bGH (Prospec, East Burnwick, NJ), in the absence of IL-3 to select IL-3-independent mutants. Vector integration sites were identified by linker-mediated PCR (LM-PCR) and RT-PCR assays to detect Ghr mRNA, A2UCOE-Ghr fusion mRNAs and IL-3 and mouse beta-actin mRNA were set up using primers listed in Table S1 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). RT-PCR amplicons for A2UCOE-Ghr junctions were gel purified, cloned into PGEM T-easy (Promega), and sequenced.

UrMgpW integration and its effect on cellular gene expression.

X-CGD PLB985 cells were transduced with the vector UrMgpW. Single-cell clones were generated by limiting-dilution cloning. Vector integration sites (IS) were identified by LM-PCR. The transcription levels of genes next to the IS were analyzed by qRT-PCR (see the overview in Fig. S1 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). Transcript levels were compared with those obtained from nontransduced cells. The structure of fusion mRNAs between HNRPA2B1 exon 1 and cellular genes was determined by RT-PCR sequencing.

P19 cell assays for long-term transgene expression and CpG methylation.

P19 cells were transduced with eGFP coding vectors, and the eGFP expression time course up to day 36 and the CpG methylation profile at day 21 posttransduction were examined as described in reference 23.

RESULTS

Selected A2UCOE-based lentiviral vectors were tested in an insertional mutagenesis assay using IL-3-dependent Bcl-15 cells (Fig. 1 and Table 1) (4, 13). This assay selects mutants that grow in the culture medium without IL-3 but supplemented with bovine growth hormone, so that most selected mutant clones have lentiviral vector integration in the growth hormone receptor gene (Ghr) locus. As previously reported (4), a vector containing the wild-type LTR and SFFV internal promoter (pHV), but not its self-inactivating (SIN) version (SFFV eGFP), gave rise to cytokine-independent mutant clones (Table 1). In these mutants, Ghr is activated by expression of fusion mRNA by aberrant splicing from the HIV major splicing donor (SD) site to Ghr exon 2 (the first coding exon). Although another control vector, PEW (an A2UCOE-less version of UfPEW and UrPEW), gave rise to two clones, both of these were negative for vector integration in Ghr and for its activation but positive for IL-3 mRNA (Table 1). Such clones have been interpreted as background mutants in this assay (4, 13). In contrast, all three vectors with a 1.5-kb core A2UCOE fragment (22) linked upstream of a heterologous promoter and an A2UCOE gamma C vector with a 2.2-kb A2UCOE sequence (2) directly driving expression of an IL2RG cDNA (24) produced mutant clones and most of them expressed Ghr RNA (Table 1).

Fig 1.

Lentiviral vectors containing the A2UCOE. (a) A2UCOE in the human genome consists of dual divergently transcribed promoters of HNRPA2B1 and CBX3. To date, three transcription initiation sites (CBX3-1, -2, and -3) and two alternative first exons (boxes indicating SD sites at their left ends) for CBX3 and a single transcription initiation and first-intron SD site for HNRPA2B1 (SD at the right end) have been reported. The two versions of the A2UCOE (1.5 kb and 2.2 kb) (2, 22) that were used in this study are depicted. The 1.5-kb fragment is missing the second CBX3 SD site, while the 2.2-kb fragment is missing the HNRPA2B1 SD site. The dotted and striped boxes represent HNRPA2B1 and CBX3 sequences, respectively. (b) A non-SIN vector, pHV, and a SIN vector with a SFFV LTR enhancer/promoter (SFFV gammaC) were used as positive and negative controls, respectively, for insertional mutagenesis assays as described in reference 4. Vectors with the 1.5-kb A2UCOE (UfPEW, UrPEW, and UrMEW) share the backbone of the PEW vector (5). The proviral forms after transduction are depicted. Triangles above the LTR boxes indicate the U3 deletion in SIN vectors.

Table 1.

Frequency of insertional mutagenesis by Ghr activation

Vector	Expt	No. of:			Frequency of:		GHR fusion mRNA (no. of positive clones/ no. tested)b
		Cells screened	Integrants screened	Mutants	Cellsa	Integrants
pHV	1	3.6 × 107	1.9 × 108	13	3.61 × 10−7	6.84 × 10−8	+c
SFFV gamma C	1	3.6 × 107	2.1 × 108	0	<2.78 × 10−8	<4.76 × 10−9	NA
PEW	6	3.6 × 107	3.15 × 108	2	5.56 × 10−8	6.36 × 10−9	0/2d
UrPEW	6	3.6 × 107	2.12 × 108	32	8.89 × 10−7	1.51 × 10−7	5/5
UfPEW	6	3.6 × 107	7.49 × 107	7	1.94 × 10−7	9.35 × 10−8	5/5
UrMEW	6	3.6 × 107	2.39 × 108	22	6.11 × 10−7	9.22 × 10−8	4/7
UCOE gamma C	1	2.2 × 108	2.7 × 107	27	1.23 × 10−7	1 × 10−6	4/4
UCOE gamma C	3	3.6 × 107	5.36 × 107	4	1.11 × 10−7	7.46 × 10−8	4/4
UCOE gamma C	4	5.4 × 107	1.24 × 108	7	1.3 × 10−7	5.64 × 10−8	NT
UCOE gamma C	5	5.4 × 107	2.05 × 107	13	2.41 × 10−7	6.34 × 10−7	NT
UCOE gamma C	6	3.6 × 107	1.19 × 108	7	1.94 × 10−7	5.89 × 10−8	NT
UCOE MAU 4	2	3.6 × 107	1.91 × 107	2	5.56 × 10−8	1.05 × 10−7	2/2
CBX3 NSD	3	3.6 × 107	9.32 × 107	0	<2.78 × 10−8	<1.07 × 10−8	NA
2SD MAU 1	4	1.8 × 108	3.24 × 108	6	3.33 × 10−8	1.85 × 10−8	5/6
MA1082	5	5.4 × 107	3.19 × 107	3	5.56 × 10−8	9.42 × 10−8	0/3d
MADE2	5	5.4 × 107	3.94 × 107	0	<1.85 × 10−8	<2.54 × 10−8	NA
MAS4	5	5.4 × 107	2.07 × 107	1	1.85 × 10−8	5.01 × 10−8	0/1d

^aFor the two groups of data in bold, P = 0.036 by the Mann-Whitney U test.

^bNA, not applicable; NT, not tested.

^cNot tested in this particular experiment; the majority of previously tested mutants were positive (4, 13).

^dPEW and MAS4 mutants and two of three MA1082 mutants were IL-3 mRNA positive and likely to be background mutants by spontaneous IL-3 activation; the mechanism for the remaining MA1028 mutant is unknown.

In order to determine the mechanism of Ghr activation by A2UCOE vectors, integration sites in a number of mutant clones were characterized (Fig. 2a), as previously described (4, 13). The structure of the fusion Ghr mRNA was also examined by sequencing of reverse transcription-PCR (RT-PCR) amplicons using primer pairs with one in the A2UCOE and the other in Ghr exon 2 (Fig. 2b). While UfPEW was integrated in the same orientation as the Ghr gene upstream of exon 2, the other vectors were arranged in the opposite orientation (Fig. 2a). Amplification of fusion transcripts was consistently successful for certain configurations of the vector and A2UCOE: HNPRA2B1- and Ghr-specific primers generated amplicons from clones containing the 1.5-kb A2UCOE fragment (UfPEW, UrPEW, and UrMEW); CBX3- and Ghr-specific primers generated amplicons from the integrated 2.2-kb A2UCOE fragment. Sequencing of these amplicons demonstrated mRNA splicing events as shown in Fig. 2b and indicated that both the inherent SD site of HNPRA2B1 and the first SD site of CBX3 were used. Overall, these studies identified a potential risk of the A2UCOE to cause activation of neighboring genes through aberrant splicing.

Fig 2.

Lentiviral vector insertion in the Ghr locus and its activation. (a) Vector insertion sites in the mouse Ghr locus were determined as previously described (4, 13) for mutant clones following transduction of Bcl15 cells with UfPEW (UF clones), UrPEW (URP), UrMEW (URM), and UCOE gamma C (UgC). UF clones had vectors in the same direction as Ghr, while the others had reverse-orientation insertion. (b) Fusion mRNA was produced by splicing from the HNRPA2B1 splice donor (SD) to the Ghr exon 2 splice acceptor (SA) in mutants by vectors containing the 1.5-kb A2UCOE. (c) Fusion mRNA obtained by splicing from the first SD of CBX3 to the Ghr exon 2 SA was detected in UCOE-gamma C vector-transduced cells containing the 2.2-kb A2UCOE fragment.

In order to determine if splicing from the A2UCOE SD sites was a general phenomenon occurring at other integration sites, we transduced the human myelomonocytic cell line X-CGD PLB985 at a high MOI with the UrMgpW vector, which expresses a gp91phox transgene instead of eGFP. Transduced cells expressing the transgene were sorted for isolation of single-cell clones by limiting dilution. Several clones were characterized for vector integrations. One clone containing six integrations (Fig. 3a and Fig. S1 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012), as defined by linear amplification-mediated PCR (LAM-PCR), was selected for further studies. The level of transcription of exons located upstream and downstream of the integration sites was compared with expression levels of those in the nontransduced parental cell. We found increased transcription levels for two genes; PSMD14 on human chromosome 2 and ZFAND2A on chromosome 7. In both cases, transcripts initiated at the HNRPA2B1 promoter were in the same orientation as the cellular gene transcripts. Molecular cloning and sequencing of these transcripts revealed that in both cases, a fusion transcript was generated which included the first exon of HNRPA2B1 fused to a cellular exon (Fig. 3b and c). No other genes were found to be activated by vector insertion, including genes in the same orientation as CBX3 transcription (see Fig. S1 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012).

Fig 3.

Aberrant splicing from the 1.5-kb A2UCOE to cellular genes in human cells. X-CGD PLB985 cells were transduced with the vector URMgp91W (a derivative of URMEW containing gp91phox in place of eGFP). A single-cell clone, UrM1, was generated by limited-dilution cloning. (a) Vector integration sites and orientation compared to the reference human genome sequence at UCSC's human BLAT search site (http://genome.ucsc.edu/cgi-bin/hgBlat?command=start, version Feb. 2009, GRCg37/hg19) were determined by LM-PCR. Expression levels of nearby genes for each integration were determined by qRT-PCR for the UrM1 clone as well as the parental X-CGD PLB985 cells using primers listed in Table S1 for the regions indicated in Fig. S1 (http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). The two genes for which a significant increase in RNA levels in X-CGD PLB985 cells was observed are in bold. (b and c) Splicing patterns of fusion mRNAs generated between transcripts from the URMpg91W vector and PSMD14 (b) and ZFAND2A (c) genes.

Next we attempted to generate potentially safer A2UCOE elements devoid of the aberrant splicing activity through selective modification of the 2.2-kb A2UCOE fragment. As this fragment lacks the HNPRA2B1 SD site and therefore cannot give rise to HNPRA2B1-Ghr fusion mRNA, the CBX3 promoter-leader sequence region was modified as depicted in Fig. 4a (also see Fig. S2 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). Six modified constructs were tested for insertional mutagenesis (Table 1). Of these, two constructs, UCOE MAU4 and 2SD MAU1, gave rise to Ghr mRNA-positive mutants. Examination of Ghr expression in UCOE MAU4 mutants revealed that the second CBX3 SD was used, while in 2SD MAU1 mutants, where both CBX3 SD sites were destroyed, cryptic SD sites several nucleotides upstream of the second CBX3 SD were also activated to generate CBX3-Ghr fusion mRNA species (Fig. 4b). However, more extensive modifications were sufficient to suppress CBX3 splicing activity, as no Ghr mRNA-positive mutants were generated. In particular, the construct MA1082 and its derivatives (MADE2 and MAS4), which had all inherent and potential cryptic GTXXG consensus SD sequences mutated to CGXXG, collectively showed a mutagenesis frequency significantly smaller than that of the wild-type A2UCOE (Table 1).

Fig 4.

Modification of the CBX3 region in the 2.2-kb form of A2UCOE. (a) Vectors with either IL2RG cDNA (UCOE-Mau4, CBX3 NSD, and 2SD-MAU-1) or eGFP (MA1082, MADE2, and MAS4) and a modified 2.2-kb version of A2UCOE were constructed. Black crosses with black lines in UCOE-MAU4 and 2SD-MAU-1 represent mutations within the first 5 bases of the intron (GTAAG to GCCAC) and a 39-bp insertion following this. The black cross at the second CBX-3 SD site in 2SD-MAU-1 represents a change of the first 5 bp of the intron (GTCGG to TCCGA). Gray crosses in MA1082, MADE2, and MAS4 represent changes of reported or potential SD consensus intron sequence (GTXXG to CGXXC). The sequence alignment of this region is presented in the supplementary data (see Fig. S2 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). (b) UCOE-Mau4 mutants arose by Ghr activation by splicing from the second CBX-3 SD site. Disruption of this second SD site resulted in Ghr transcription by activation of a cryptic SD site just upstream of this site in the mutants by 2SD-Mau-1.

Our aim was to identify a version of the A2UCOE fragment that was fully functional but which did not contain active SD sites. To assess the regulatory function of four modified A2UCOE vectors devoid of splicing activity, their long-term transgene (eGFP) expression and HNRPA2B1 CpG methylation profile were examined in a mouse embryonic carcinoma P19 stem cell line which has a strong tendency toward repression of transgene expression through promoter methylation (23). These were compared with properties of a vector containing a wild-type 2.2-kb A2UCOE fragment which has previously been shown to be resistant to DNA methylation-mediated silencing (23). The P19 cells were transduced with vectors at a multiplicity of infection (MOI) of 1 (Fig. 5a), 0.5, or 2 (see Fig. S3 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012), and eGFP expression was assessed at different time points (see Fig. S4 at http://www.ucl.ac.uk/infection-immunity/). While the CBX3 NSD vector, with an extensive A2UCOE deletion within the first exons of both HNRPA2B1 and CBX3, showed a continuous decline of eGFP-positive cells, eGFP-positive cells for the other three vectors were sustained at a level similar to that observed with the wild-type 2.2-kb A2UCOE vector. The decline in eGFP expression with CBX3 NSD-transduced cells may result from either loss of vector or epigenetically (DNA methylation) mediated gene silencing. However, an assessment of vector copy number (VCN) per cell at day 21 posttransduction revealed that cells transduced with CBX3 NSD had a higher average VCN and thus a lower eGFP-positive-cell-to-VCN ratio than cells transduced with the other vectors tested (see Table S2 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). This shows that the decline in eGFP-positive cells in CBX3 NSD-transduced cells is not due to vector loss but occurs via transgene silencing, as we have observed previously with vectors lacking a UCOE function (5, 23).

Fig 5.

Long-term transgene expression and CpG methylation of A2UCOE derivatives in P19 cells. (a) P19 cells were transduced with eGFP coding vectors with the wild-type or modified A2UCOE driving eGFP at an MOI of 1 were monitored periodically for transgene expression for several weeks. The percentage of the cell population that scored positive for eGFP expression as measured by flow cytometry is shown. Experiments conducted with different vector doses (MOI = 0.5 and 2) showed similar results (see Fig. S3 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). Flow cytometry profiles of both experiments at MOIs of 0.5, 1, and 2 are shown in Fig. S4. (b) DNA from cells transduced with vectors at an MOI of 2 was extracted at day 21 posttransduction and subjected to sodium bisulfite conversion plus DNA sequence analysis to determine methylation status. The percentage of methylated CpG sites (methyl-CpGs) among all 63 CpG residues in the HNRPA2B1 region was calculated after sequencing of clones obtained following PCR of bisulfite-treated cellular DNA (see Fig. S5 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012 for sequencing results for each clone). The average percent methylated CpG sites for each vector is given. P values determined using the Wilcoxon rank sum test are shown.

Genomic DNA extracted from vector-transduced P19 cells at day 21 posttransduction was treated with sodium bisulfite to convert unmethylated cytosine residues to thymine and was subjected to a nested PCR using primers to specifically amplify the A2UCOE region. PCR-generated products were cloned and sequenced (see Fig. S5 at http://www.ucl.ac.uk/infection-immunity/themes/tak_suppJVI2012). The percentage of methylated CpG residues (methyl-CpGs) among all assessed CpGs for each clone was determined (Fig. 5b). Distinctive methylation patterns between CBX3 NSD and the other vectors, including the wild-type A2UCOE vector, were revealed: compared to a high rate (average, 35%) of methylation in CBX3 NSD, the other modified vectors had lower methylation rates, ranging between 4% (MA1082) and 15% (MAS4). These results indicated that MA1082, MADE2, and MAS4 vectors are as resistant as the wild-type vector to CpG methylation and consequent transgene silencing.

DISCUSSION

The predominant mechanism of insertional mutagenesis by the gammaretroviral vectors used in early laboratory studies and clinical trials has been enhancer-mediated host gene activation. However, any integrating retroviral vector could potentially cause aberrant gene expression by dysregulated splicing, and studies have identified this phenomenon with lentiviral vectors (1, 16, 17). This can arise through either splicing from a donor site within the vector to an adjacent gene, producing a coding fusion transcript, or splicing to an acceptor site within the vector, thereby truncating a cellular transcript and dysregulating turnover or translation. The latter mechanism has caused dominant clonal expansion in one of two patients participating in a clinical trial using a lentiviral vector to treat β-thalassemia (7). Lentiviral vectors, like the parent HIV, have the potential to carry multiple splice donor and acceptor sites because the unspliced viral genomic RNA is exported from the nucleus by the viral protein Rev. Therefore, it will be important to screen candidate clinical lentiviral vectors for potential splice donor and acceptor sites that may cause inadvertent mutagenesis through this mechanism, as well as for vector enhancer-driven mutagenesis.

In this study, we report that lentiviral vectors incorporating the A2UCOE, which are resistant to gene silencing via CpG methylation (5, 23), have mutagenic potential through aberrant splicing. All three splice donor sites in the A2UCOE bidirectional transcripts (Fig. 1a) can give rise to aberrant mRNA species fused to a downstream cellular gene. Furthermore, simple point mutation of inherent splice donor sites results in activation of cryptic donor sites. This indicates that unexpected splicing events from the vector to cellular genes could potentially pose a genotoxic risk. It is also possible that different cryptic donor sites may operate in other cell types.

Our goal was to generate 2.2-kb UCOE derivatives with abrogated mutagenic activity yet retain the normal regulatory function of the element. A large truncation in the CBX3 NSD vector to delete all potential SD sites was unsuccessful, as the gene-regulatory properties of long-term transgene expression and resistance to CpG methylation were lost. The high level of CpG methylation and eGFP expression silencing of the CBX3 NSD vector supports the role of CpG methylation in A2UCOE transcription regulation, although other mechanisms, such as establishment of repressive histone modification patterns, may also play a role. The minimal A2UCOE sequence required for silencing resistance has yet to be identified. However, it is noteworthy that vectors containing the 1.5-kb A2UCOE fragment, which extends to Esp3I sites in the first introns of both HNRPA2B1 and CBX3 and thus possesses intact SD sites at both ends of this fragment, are silencing resistant (5, 22). In contrast, deletion of a core 1-kb subregion at the ends of the first exons of HNRPA2B1 and CBX3 in the case of the CBX NSD construct results in loss of function (Fig. 5a). This suggests that a specific critical functional region is removed in the generation of the CBX NSD element or that the CpG island encompassing the HNRPA2B1 and CBX3 promoters may have been reduced below a critical size (14). It is possible that association of CpG islands with a minimum number of nonmethylated-CpG-binding proteins such as KDM2A (3) and CFP1 (8), which are thought to be involved in the maintenance of an open, transcriptionally active chromatin structure, is required for stability of expression. Therefore, deleting a CpG island below a crucial size may lower binding of such factors below a point where open or active chromatin can be maintained. We have found that mutations of defined and cryptic SD sites in MA1082 derivatives, with or without small deletions around these sites, abrogated splicing activity while retaining the original A2UCOE regulatory function. These safety-modified elements will therefore be ideal for regulation of gene expression in clinical applications where robust constitutive gene expression is desirable.