Potential genotoxicity from integration sites in CLAD dogs treated successfully with gammaretroviral vector-mediated gene therapy

Summary

Integration site analysis was performed on six dogs with canine leukocyte adhesion deficiency (CLAD) that survived greater than one-year after infusion of autologous CD34+ bone marrow cells transduced with a gammaretroviral vector expressing canine CD18. A total of 387 retroviral insertion sites (RIS) were identified in the peripheral blood leukocytes from the six dogs at 1 year post-infusion. A total of 129 RIS were identified in CD3+ T-lymphocytes and 102 RIS in neutrophils from two dogs at 3 years post-infusion. RIS occurred preferentially within 30 kb of transcription start sites, including 40 near oncogenes and 52 near genes active in hematopoietic stem cells. Integrations clustered around common insertion sites more frequently than random. Despite potential genotoxicity from RIS, to date there has been no progression to oligoclonal hematopoiesis and no evidence that vector integration sites influenced cell survival or proliferation. Continued follow-up in disease-specific animal models such as CLAD will be required to provide an accurate estimate of the genotoxicity using gammaretroviral vectors for hematopoietic stem cell gene therapy.

Although the theoretical risk of retroviral-mediated oncogene activation through insertional mutagenesis had been considered to be low, this risk has been shown to be significant in recent human gene therapy clinical trials using gammaretroviral vectors.^1–3 The importance of the individual variables responsible for genotoxicity in hematopoietic stem cell gene therapy trials using gammaretroviral vectors remains uncertain. Human clinical trials, as well as small and large animal studies, using gammaretroviral vectors have implicated the vector, transgene, and retroviral insertion sites (RIS) in the development of serious adverse events (SAEs). In each case, the gammaretroviral vector integration itself has been thought to contribute to the genotoxicity observed, but may not be the only factor responsible for clonal outgrowth.

Successful hematopoietic stem cell gene therapy in humans using gammaretroviral vectors have generally relied upon an in vivo selective growth advantage for the transduced cells provided by a therapeutic transgene, such as the interleukin-2 receptor gamma chain (IL2Rg) in X-linked severe combined immunodeficiency disease (SCID-X1) and adenosine deaminase (ADA) in SCID-ADA.^4–6 Initial excitement in the early reports from these trials was subsequently tempered by SAEs which occurred in four of the eight children with therapeutic levels of in vivo transgene expression in the French SCID-X1 trial² and in one of ten children in the UK SCID-X1 trial.⁷ The selective advantage conferred by the transgene was thought to not only contribute to the observed clinical success but also to genotoxic clonal outgrowth.⁶

In the recent non-myeloablative gene therapy trial for chronic granulomatous disease (CGD), in which no selective advantage was provided by the CYBB (gp91phox) transgene, two patients developed clonal dominance of retrovirally marked cells, accompanied by multiple RV insertions in EVI1-MDS1, PRDM16, and SETBP1.³ Thus, factors other than the transgene, such as RIS, appear to play a major role in clonal dominance.

Previous animal studies, including mice, dogs, and non-human primates, also point to genotoxic risk from gammaretroviral vectors. Insertional activation of Evi-1 by gammaretroviral vectors was first described in mice.⁸ A study of 14 non-human primates had integrations in EVI-1 after gammaretroviral gene transfer, with one animal developing a granulocytic sarcoma.⁹ A recent study examining RIS in dogs marked with a gammaretroviral vector also revealed a bias of RIS occurring in and near proto-oncogenes.¹⁰

In the current study we conducted integration site analysis in six dogs with canine leukocyte adhesion deficiency (CLAD) that had been successfully treated with gammaretroviral vector-mediated ex vivo gene therapy.¹¹ CLAD is analogous to leukocyte adhesion deficiency (LAD) in children, and thus represents a preclinical large-animal model for LAD. The characteristics of the six treated dogs with CLAD, including age at infusion, conditioning regimen, and transduced cell dose infused, are shown (Supplementary Table 1). In this study, the therapeutic CD18 transgene did not confer a strong selective growth advantage to the transduced cells, and a clinically applicable non-myeloablative conditioning regimen was used to facilitate engraftment of the transduced cells. The percentages of CD18+ leukocytes in the peripheral blood in the 6 dogs ranged from 0.7 to 11.2% at 12–36 months follow-up and represent therapeutic levels (Supplementary Table 1).

To evaluate the diversity of clones contributing to long-term hematopoiesis after autologous transplantation of CD34+ BM transduced with a gammaretroviral vector carrying canine CD18, we identified integration sites by linear-amplification-mediated (LAM) PCR from all six treated animals (Figure 1). LAM-PCR, performed at 6 month intervals using DNA from peripheral blood leukocytes demonstrated multiple banding patterns indicative of polyclonal hematopoiesis in all six dogs (Figure 1a). Similar polyclonal hematopoiesis was evident for both myeloid and lymphoid lineages in dogs B1 and C1 examined at 36 and 31 months post-infusion, respectively (Figure 1b).

Figure 1.

Evaluation of polyclonality by linear amplification-mediated PCR (LAM-PCR). Genomic DNA was collected from six CLAD dogs (A1, A2, A3, B1, B2 and C1) treated by gammaretroviral vector PG13/Mscv-cCD18¹¹ at designated months post-infusion of autologous transduced CD34+ cells. DNA was isolated from (a) peripheral blood leukocytes and (b) flow sorted CD3+ lymphocytes and neutrophils using either the Blood and Cell Culture DNA Kit (Qiagen, Inc., Valencia, CA, USA) or the Wizard Genomic DNA purification Kit (Promega Corp., Madison, WI, USA). Cells were sorted by immunostaining for T-cells using an antibody to canine CD3 (CA17.2A12; AbD Serotec, Raleigh, NC, USA) conjugated to AlexaFluor 647 by using a Zenon labelling kit, and for neutrophils using an antibody to canine neutrophils (CADO48A; VMRD, Inc., Pullman, WA, USA) conjugated to phycoerythrin. Labelled cells were sorted to purities of 91.6% and 87.2% for CD3+ cells in dogs B1 and C1, respectively, with <0.5% neutrophil contamination. Neutrophils were sorted to purities of 97.8% and 93.5% for B1 and C1, respectively, with 0.1% CD3+ cell contamination. Cells were sorted using a FACSVantage SE with DiVa software (BD Biosciences, San Jose, CA, USA) equipped with 488 nm and 633 nm excitation lasers. Linear-Amplification-mediated (LAM) PCR was performed on extracted DNA, as previously described.¹¹ LAM-PCR samples were electrophoresed on a high-resolution Spreadex gel (Elchrom Scientific AG, Cham, Switzerland) and photographed using a gel documentation system (Biochemi system, UVP, Upland, CA, USA). Bands represent insertion sites. M indicates size markers; bp, base pairs; -C, negative water control; N, neutrophils; L, lymphocytes.

To confirm the polyclonal nature of the samples, linker-mediated (LM)-PCR was used to search for potentially over-represented clones or RIS within genes and gene regions. When LM-PCR amplicons were sequenced, the integration sites in all six dogs were polyclonally derived (complete list of integration sites in Supplementary Table 2) with no predominant clone(s) observed. We obtained a total of 618 retroviral insertion sites (RIS) from the treated CLAD dogs by sequence analysis of LM-PCR amplicons: 387 RIS from peripheral blood leukocytes from all six dogs at 5–21 months post-infusion (36 to 88 RIS per dog), 129 RIS from sorted CD3+ lymphocytes and 102 RIS from sorted neutrophils from dogs B1 and C1 at 31–36 months post-infusion. We identified 535 unique RIS after adjustment of 83 RIS that overlapped between the cell types and timepoints examined (see Supplementary Table 2). RIS isolated separately from lymphocytes and neutrophils confirmed polyclonal insertion sites in each lineage. In addition, 28 RIS were shared between the sorted lymphocytes and neutrophils RIS, indicating that a common hematopoietic precursor was transduced. The positions corresponding to human sequences listed in the Reference Sequence database (RefSeq) that mapped to the dog genome were examined for RIS that occurred within 30 kilobases (kb) upstream or downstream of the RefSeq genes. Comparison of the RIS to a computer-generated dataset of 1200 random positions in the dog genome revealed a modest propensity of RIS to be located within the RefSeq genes (43.4%), although this was not significant compared to the random dataset (39.3%).

Further examination of the RIS showed a tendency to occur at common insertion sites (CIS).¹² CIS may represent either insertional hotspots or loci that confer some selective advantage. A large number of CIS were present among the RIS, when compared to that expected from a random dataset: 39 CIS with two insertions within 30 kb (expected: 3.7 CIS), 13 CIS with three insertions within 50 kb (expected: 0.035 CIS) and 6 CIS with four insertions within 100 kb (expected: 0.001 CIS) (Supplementary Table 3). Several genes were found near CIS, including PTP4A2, ABHD1, TCF23, NID1, NACA, and HSD17B6. Overall, 71 RIS (13.3%) of the 535 unique RIS were found within CIS, similar to the percentage observed by other investigators with gammaretroviral vectors.^1,7

RIS located near oncogenes were determined by comparison to a list of human cancer genes compiled from the University of New South Wales (UNSW) Embryology DNA-Tumor Suppressor and Oncogene Database (http://embryology.med.unsw.edu.au/DNA/DNA10.htm) and from the Cancer Gene Consensus database at the Wellcome Trust Sanger Institute (http://www.sanger.ac.uk/genetics/CGP/Census/). Forty RIS (7.5%, expected: 3.2%, P < 0.001) were located in, or within, 30kb of 32 oncogenes (Figure 2). None of the RIS were near the genes implicated in clonal proliferation in the gene therapy trial for CGD (EVI1-MDS1, PRDM16 and SETBP1). Although no RIS in our study were present near the major gene implicated in leukemia in several SCID-X1 children treated by gene therapy, LMO2,² several of the RIS were near other prominent cancer-related genes such as ABL1, BCL2, ETV6, FLI1, and NF1, including some that were also associated with a CIS (Supplementary Table 3).

Figure 2.

Genes identified near RIS. LM-PCR was performed as previously described,¹⁵ with slight modifications (see Supplementary methods), and amplicons were sequenced to identify the precise viral integration sites within the dog genome. Genes within 30 kb of each RIS were determined (Supplementary Table 2) and further characterized as either cancer-related, expressed in HSCs, or related to genes found near RIS in the CGD clinical trial, and are listed as three different sets. The numbers in the outer circle of the VENN diagram indicated the number of genes identified for each data set. The numbers of genes that overlap between the three gene sets are represented in the common area of the VENN diagram. The percentage of integrations expected within genes is based on random RIS analysis (Rnd) generated in silico (see Supplementary Methods). The experimental percentage of RIS close to genes in each of these sets is shown (Exp) with the corresponding statistical significance (P value).

To examine the propensity of the retrovirus to integrate near genes active in early hematopoiesis, we compared the genes near RIS to a database of genes expressed in HSCs.¹³ We found 52 RIS (9.7%; expected: 4.7%; P < 0.001) located near 47 HSC genes, indicating preferential retroviral integration (Figure 2). This increased propensity of RIS near HSC genes was not unexpected given that canine CD34+ cells were used and several HSC genes were likely activated by the G-CSF, SCF, and Flt3L growth factors used.

The non-myeloablative conditioning regimen used in both the CGD trial and our study, in conjunction with the fact that both diseases involve a transgene without a strong selective advantage, prompted us to compare our RIS profile with that of the CGD trial. Genes within 30 kb of RIS in treated CLAD dogs were compared to the genes within 30 kb of RIS found in the two CGD gene therapy patients (Figure 2).³ There was a significant overlap between the RIS profiles in both studies: 51 RIS in our study (9.5%) were close to 61 genes associated with RIS in the CGD trial (expected: 3.9%, P < 0.001). Of note, seven of these 51 (13.7%) genes shared between the two studies were also oncogenes (CIITA, FLI1, LYN, MAF, MLLT6, NF1 and SEPT9).

To determine whether RIS occurred close to genes with particular functions that may influence in vivo selection, we used gene ontology classification to identify gene families among genes within 30kb of the RIS. We observed overrepresentation (P < 0.05) of seven categories when compared to the expected distribution, however, none of these categories involved cell proliferation or differentiation (Figure 3a). Similarly, a comparison of gene families at either early (≤ 21 months) or late (> 21 months) timepoints (Figure 3b), or in either lymphocytes or neutrophils (Figure 3c) in dogs B1 and C1, revealed overrepresented families, but none appeared to be related to cell proliferation or survival.

Figure 3.

Ontological classes of genes within 30 kb of RIS. RIS from (a) pooled, unique RIS, (b) RIS from early (≤ 21 months post-transplantation) and late (> 21 months post-transplantation), or (c) sorted neutrophils and CD3+ lymphocytes were compared to an averaged random dataset (see Supplementary Methods). Percentages represent the number of genes for each particular gene class divided by the total number of genes found in or near insertion sites. The asterisk (*) indicates those classes over-represented using a modified one-tailed Fisher’s Exact test.

Despite the increased frequency of retroviral insertions near oncogenes in both the CGD and CLAD studies, to-date the animals in this CLAD study show no evidence of clonal expansion up to 3 years post-treatment, a time period representing approximately 25 percent of the canine lifespan. In comparison, clonal expansion in two CGD patients occurred relatively rapidly, with clonal dominance occurring within 5 months after gene therapy. In the SCID-X1 clinical trials, clonal expansion led to leukemia in five patients at 24 to 68 months post-infusion. In the CLAD studies, it would be expected that leukemia or lymphoproliferation would occur much earlier because dogs have a faster HSC replication rate (assuming greater similarity to cats) than humans.¹⁴

The difference in outcomes between the CGD trial and our study may have been influenced by several factors, including vector design (the highly active spleen focus forming virus LTR used in the CGD trial versus the murine stem cell virus LTR used in this study), the source of target cells (G-CSF mobilized peripheral blood CD34+ cells used in the CGD trial versus bone marrow CD34+ cells in this study), and length of time in culture (5 days in the CGD trial versus 3 days in our trial). There was also a higher transduction efficiency with subsequent increased vector MOI levels in the cells infused into the CGD patients. However, the multiple insertions near or within oncogenes that we identified raises concerns that clonal dominance remains a distinct possibility, and that long-term follow-up of these animals will be necessary to accurately estimate the risk from the gammaretroviral vector used in this study.

These results also emphasize the value of disease-specific, pre-clinical animal models in assessing risks from the complex interaction of vector, transgene, insertion site, and disease in gene therapy studies. In particular, the CLAD model enables one to identify potential interactions between the transgene and the integration site, such as the one that occurred in the SCID-X1 trial.²