Following gene therapy of SCID-X1 using murine leukemia virus (MLV) derived vector, two patients developed leukemia owing to an activating vector integration near the LMO2 gene. We found that these integrations reside within FRA11E, a common fragile site known to correlate with chromosomal breakpoints in tumors. Further analysis showed that fragile sites attract a nonrandom number of MLV integrations, shedding light on its integration mechanism and risk-to-benefit ratio in gene therapy.

X-linked severe combined immunodeficiency (SCID-X1) is a lethal genetic T-cell deficiency caused by mutations in the common γ chain (γc) gene. Correction of SCID-X1 was previously shown by transducing the murine leukemia retrovirus (MLV)-derived vector containing the γc gene into autologous CD34+ bone marrow cells.¹ Almost 3 years after treatment, three out of the 10 treated children have developed monoclonal acute lymphoblastic leukemia-like lymphoproliferation. In the leukemic cell clones of two patients, an MLV vector integration was mapped to the vicinity of the transcription start site of the LMO2 proto-oncogene. This resulted in high levels of transcription and expression of the LMO2 gene. However, the molecular basis for the recurrent MLV integrations in the LMO2 gene remained unclear.² The MLV integration in the leukemic cells of the third patient is not yet fully characterized.³

Specific regions in the genome, defined as common fragile sites (CFSs), correlate with chromosomal breakpoints in tumors such as deletions, translocations, amplifications and integration sites of oncogenic DNA viruses such as HPV16, HPV18, HBV, EBV and AAV (reviewed by Popsecu⁴). Common fragile sites are chromosomal loci exhibiting perturbed chromatin condensation, which appear as constrictions, gaps or breaks on chromosomes from cells exposed to partial inhibition of DNA replication.⁵ About 90 CFSs were cytogenetically identified in the human genome, 13 of which were characterized at the molecular level (Supplementary Table S1). The cytogenetic expression of CFSs appears along a large genomic region up to a few mega bases. Replication analyses revealed a specifically delayed elongation of DNA replication along CFSs, which is enhanced by replication stress.^{6, 7, 8}

Here we investigated the possibility that CFSs are preferential sites for the integration of MLV retroviral vectors. We first analyzed whether the LMO2 gene resides within a CFS. The LMO2 gene has been mapped to the short arm of chromosome 11 (11p13), a cytogenetic band harboring a CFS, FRA11E, which is a moderately expressed fragile site. To analyze whether LMO2 is located within FRA11E, we molecularly characterized genomic regions within 11p13. For this, we determined the location of clones from the 11p13 region relative to FRA11E gaps and constrictions, by fluorescent in situ hybridization (FISH), on metaphase chromosomes from GM00847 cells induced by aphidicolin to express fragility. We analyzed two BAC clones, 17k7 and 786C16, mapped 2 Mb apart in 11p13; the latter contains the LMO2 gene. The analysis revealed that both clones reside within FRA11E, because on different chromosomes from the same preparation, their hybridization signals appeared proximal to, distal to or on both sides of FRA11E gaps and constrictions (Figure 1). These results indicate that LMO2 is indeed a part of FRA11E.

Examples of fluorescent in situ hybridization signals relative to FRA11E. Fluorescent in situ hybridization analysis was performed on metaphase chromosomes from GM00847 cells following aphidicolin treatment. Fluorescein isothiocyanate (FITC)-labeled 786C16 and 17K7 show signals distal and proximal to FRA11E on different chromosomes. The FRA11E probes were co-hybridized with a probe to the centromere to identify chromosome 11. For each clone, the propidium staining is to the right and the FISH with FITC-labeled probes to the left. The arrows in the propoidum images point to FRA11E gaps. The number of analyzed chromosome 11 is summarized in the table.

Full size image

We further investigated whether following infection with MLV-derived vectors, the integration sites correlate with CFSs. For this, we worked with 666 MLV integrations that have been identified in precursor CD3+ T cells of nine SCID-X patients who participated in the gene therapy trial (manuscript in preparation). The analysis was performed using linear amplification-mediated polymerase chain reaction (LAM-PCR)⁹ on DNA from pretransplant CD34+ cells and CD3+ T lymphocytes sampled from the blood of the patients 5–41 months after gene therapy. These CD3+ cells most likely derive from transduced progenitor CD34+ cells. In addition, we analyzed 820 MLV integrations in HeLa cells infected with MLV-derived vector. The latter MLV integrations were analyzed 48 h post-infection ¹⁰ and thus a selection is not expected to affect the observed repertoire of integrations.

A statistical analysis was carried out comparing the observed frequency of MLV integrations in each human chromosome to that expected from the length of chromosome. The 1486 MLV integrations in CD3+ and HeLa cells revealed a significant difference between the observed and expected distribution of integrations in part of the chromosomes (P<10e−10) (see details in Supplementary Information). Therefore, we restricted further analysis to integrations in human chromosomes harboring molecularly characterized CFSs (Supplementary Table S1). This included 305 integrations in CD3+ T cells and 396 integrations in HeLa cells.¹⁰ The analysis showed that the frequency of MLV integration in CFSs in HeLa and CD3+ T cells is not significantly different (P=0.86), allowing us to use the combined data set of 701 integrations. We have found 36 MLV integrations in genomic regions encompassed by the molecularly cloned CFSs, 1.5 fold significantly higher than expected (P=0.016), indicating that MLV preferentially integrated in CFSs.

Interestingly, in our data set, seven MLV integrations were found in the 2.2 Mb of the cloned FRA11E region, in HeLa cells (n=2) and in CD3+ T cells of four patients (n=5). These integrations were distributed along the FRA11E characterized region: three in the vicinity of the start sites of the LMO2, two in the vicinity of the start site of RCN1 gene and two in non-coding regions, each of them >40 kb upstream to LMO2. This further emphasizes the role of the genomic region in the integration of MLV-derived vector near the LMO2 gene.

One of the earliest models for chromosomal influence on MLV integration targeting proposed that open chromatin favored integration.¹¹ Recently, analysis on the set of MLV integration in HeLa cells, which we also analyzed in this study, suggested that the distance from the transcription start site is an important parameter contributing to MLV integration intensity, and may suggest a role for open chromatin conformation in these viral integrations.^{10, 12} Analysis of MLV integrations near transcription start sites of our data set revealed no significance difference between the entire genome and fragile sites (P>0.5), pinpointing to the contribution of other factors to the preferential MLV integration in fragile sites. Although unknown, the suggested mechanism of favored integrations to open chromatin is consistent with the decondensed chromatin conformation (gaps, constrictions) and breaks at fragile sites. It is worth noting that analysis of the correlation between fragile sites and integrations of human immunodeficiency virus (HIV) in several cell lines¹² revealed no preferential integration in fragile sites (P=0.631), consistent with other differences between MLV and HIV in integration site preference in the human genome.¹² One such difference is that HIV infects non-dividing cells; hence, its integration is not expected to be affected by the chromatin condensation status.

It is important to emphasize that in this study we could analyze only 14 molecularly characterized fragile sites out of the 90 cytogenetically mapped sites (14%); hence, the number of MLV integrations into fragile sites in our data set is expected to be much higher. In summary, the results presented in this study indicate that fragile sites attract a nonrandom number of integrations, which is of substantial interest for developing and understanding the integration mechanism of MLV vectors used in gene therapy trials and their risk-to-benefit ratio. Further studies aiming at understanding the effect of viral transduction on fragile sites expression are required in order to shed light on the mechanism/s of viral integration in gene therapy trials.

We thank Professor Norman Grover for his assistance in the statistical analyses, Dr Nicoletta Archidiacono for providing centromere 11 FISH probe and Professor Steve Scherer for providing chromosome 11 BAC clones.

Affiliations

1. Department of Genetics, Silberman Life Sciences Institute, Hebrew University, Jerusalem, Israel

   • A C Bester
   • , M Schwartz
   • , N Ben-Porat
   •  & B Kerem

2. Department of Internal Medicine I, Institute of Molecular Medicine and Cell Research, Albert-Ludwigs-University, Freiburg, Germany

   • M Schmidt
   •  & C Von Kalle

3. INSERM Unit 429, Laboratoire de Cytogénétique, Paris, France

   • A Garrigue
   • , S Hacein-Bey-Abina
   • , M Cavazzana-Calvo
   •  & A Fischer

4. Department de Biotherapie Assistance Publique–Hopitaux de Paris, Paris, France

   • S Hacein-Bey-Abina
   •  & M Cavazzana-Calvo

5. Molecular and Gene Therapy Program, Division of Experimental Hematology, Cincinnati Children's Hospital Research Foundation, Cincinnati, OH, USA

   • C Von Kalle

6. Unité d'Immunologie et d'Hématologie Pédiatriques, Hôpital Necker, Paris, France

   • A Fischer

Corresponding author

Correspondence to B Kerem.

To obtain permission to re-use content from this article visit RightsLink.

Publication history

DOI

https://doi.org/10.1038/sj.gt.3302752