Whole transcriptome characterization of aberrant splicing events induced by lentiviral vector integrations

Abstract

Gamma-retroviral/lentiviral vectors (γRV/LV) with self-inactivating (SIN) long terminal repeats (LTRs) and internal moderate cellular promoters pose a reduced risk of insertional mutagenesis when compared with vectors with active LTRs. Yet, in a recent LV-based clinical trial for β-thalassemia, vector integration within the HMGA2 gene induced the formation of an aberrantly spliced mRNA form that appeared to cause clonal dominance. Using a method that we developed, cDNA linear amplification-mediated PCR, in combination with high-throughput sequencing, we conducted a whole transcriptome analysis of chimeric LV-cellular fusion transcripts in transduced human lymphoblastoid cells and primary hematopoietic stem/progenitor cells. We observed a surprising abundance of read-through transcription originating outside and inside the provirus and identified the vector sequences contributing to the aberrant splicing process. We found that SIN LV has a sharply reduced propensity to engage in aberrant splicing compared with that of vectors carrying active LTRs. Moreover, by recoding the identified vector splice sites, we reduced residual read-through transcription and demonstrated an effective strategy for improving vectors. Characterization of the mechanisms and genetic features underlying vector-induced aberrant splicing will enable the generation of safer vectors, with low impact on the cellular transcriptome.

Figure 1. cLAM-PCR procedure for the retrieval of LV cellular fusion transcripts.

(A) Scheme of the experimental procedure for cLAM-PCR. Total mRNA is retrotranscribed into double-stranded cDNA (ds cDNA) using oligo-dT primers. Linear PCR uses a biotinylated primer located upstream/downstream of a known LV splice site, allowing extension into vector or an unknown cellular portion of a chimeric transcript. Single-stranded product is purified by streptavidin-coupled magnetic beads, double stranded using Klenow enzyme, and cut using restriction enzymes (REs). A linker cassette compatible with the restriction enzyme cut is ligated, and 2 sequential nested PCRs are performed. The final PCR products are then sequenced. ss cDNA, single-stranded cDNA. (B) FACS plots showing the percentage of GFP⁺ in JY cells and CD34⁺ HSPCs after SIN.LV.PGK transduction. The VCN and the MOI are indicated. Numbers in the graph indicate the percentage of GFP⁺ cells. (C) Representative band pattern of cLAM-PCR performed on mRNA from SIN.LV.PGK-transduced cells. Retrotranscribed mRNA (RT+) and negative controls (RT–) were used. By sequencing, bands in retrotranscribed mRNA samples corresponded to aberrant transcripts or unspliced internal control sequences. Rare faint bands in negative controls corresponded to oligonucleotide dimers or concatemers. H₂O, negative PCR control from the linear amplification reaction to the second exponential phase. M, marker.

Figure 2. cLAM-PCR procedure for the retrieval of LV cellular fusion transcripts.

Cryptic splice sites identified by cLAM in the LV backbone are shown in parentheses: SA1, SD4, SA7, SA3, SA4, and SD5. cLAM-PCR primer sets, UPLVSD and DWLVSA as well as UPcrypSD and DWcrypSA, are indicated. LV.exon_1, LV.exon_1a, LV.exon_1b, LV.exon_2, and LV.exon_3, as defined by their boundary splice sites, are indicated.

Figure 3. Examples of chimeric LV/cellular gene/genome transcripts.

Chimeric sequences are aligned on the human genome sequence using BLAT and shown on the UCSC genome browser. Sequences aligned to exonic sequences (black boxes) of know transcripts (chromosomal coordinates and size interval are shown above each panel). Orientation of vector and genes with respect to genome is indicated by orientation of triangles and arrows, respectively. Vector position and size are arbitrary. The 10 bases surrounding the vector/genomic junction are indicated: black text on white background indicates vector sequence, white text on black background indicates genomic sequence. In the 3 top panels, LV integrations in the same gene transcriptional orientation involved the canonical vector splice donor site SD1 sequence fused downstream of the SA site of a gene exon (i.e., RPL22, top panel); the vector splice acceptor sequence SA1 fused to cellular exons upstream (i.e., BLNK, second panel); and, in some cases, junctions with a splice site in an unannotated exon within gene introns were found (i.e., USP49, third panel). In some cases, fusion transcripts aligned discontinuously to genomic portions without annotated transcripts were identified (bottom panel). Chr, chromosome.

Figure 4. Representation of aberrant splicing events within the LV backbone and quantification of transcription levels of LV backbone portions.

(A) Schematic representation of the position of the recoded splice sites within the LV backbone. The different mutations were distributed in 3 different vector constructs (indicated as MutSD, Mut1_13, and Mut14_15). RRE, rev-responsive element; wPRE, woodchuck hepatitis posttranscriptional regulatory element. (B) Titers of the 3 different recoded vectors. The titer is defined as number of transducing units per milliliter (TU/ml) of vector preparation. (C) Representation of the positions of the 4 TaqMan primer sets on SIN.LV.PGK and LV.SF.LTR vectors. U3RU5 recognizes the portion from the LV.LTR to the SD1, encompassing the cryptic splice acceptor SA1; LV.FUSION recognizes the internally spliced transcript (SD1 to SA2); SA-PPT recognizes the sequence downstream of the canonical splice acceptor SA2, encompassing the cryptic donor SD4; and GFP recognizes the GFP transgene sequence. (D–G) RT-qPCR results of transcription levels of different LV backbone portions performed on JY cells transduced with SIN.LV.PGK or LV.SF.LTR at MOI 0.1 (white bars) or MOI 10 (black bars). ΔCt values were obtained using β₂ microglobulin (B2M) as normalizer to measure the relative expression levels with respect to the housekeeping cellular gene (vs. B2M). ΔCt values obtained using GFP as normalizer to measure the relative expression levels with respect to transgene expression (vs. GFP). (H) ΔCt values obtained using GFP as normalizer from JY cells transduced with SIN.LV.PGK. The recoded vectors are indicated. Probe sets used are indicated. Statistical evaluation was performed by 1-way ANOVA with Bonferroni’s correction (*P < 0.001; **P < 0.0001).