Murine leukemia virus-derived retroviral vector has differential integration patterns in human cell lines used to produce recombinant factor VIII.

OBJECTIVE: Nowadays recombinant factor VIII is produced in murine cells including in Chinese hamster ovary (CHO) and baby hamster kidney cells (BHK). Previous studies, using the murine leukemia virus-derived retroviral vector pMFG-FVIII-P140K, modified two recombinant human cell lines, HepG2 and Hek293 to produce recombinant factor VIII. In order to characterize these cells, the present study aimed to analyze the integration pattern of retroviral vector pMFG-FVIII-P140K.
METHODS: This study used ligation-mediated polymerase chain reaction to locate the site of viral vector integration by sequencing polymerase chain reaction products. The sequences were compared to genomic databases to characterize respective clones.
RESULTS: The retroviral vector presented different and non-random profiles of integration between cells lines. A preference of integration for chromosomes 19, 17 and 11 was observed for HepG2FVIIIdB/P140K and chromosome 9 for Hek293FVIIIdB/P140K. In genomic regions such as CpG islands and transcription factor binding sites, there was no difference in the integration profiles for both cell lines. Integration in intronic regions of encoding protein genes (RefSeq genes) was also observed in both cell lines. Twenty percent of integrations occurred at fragile sites in the genome of the HepG2 cell line and 17% in Hek293.
CONCLUSION: The results suggest that the cell type can affect the profile of chromosomal integration of the retroviral vector used; these differences may interfere in the level of expression of recombinant proteins.

Introduction

Hemophilia A is an X-linked bleeding disorder caused by the absence or abnormality of factor VIII (FVIII), a cofactor necessary to generate fibrin. Deficiency of FVIII is the most common coagulation disorder with an incidence of approximately one in 5000 men; it currently affects about 400,000 people worldwide.1, 2 Intravenous administration of plasma-derived or recombinant FVIII (rFVIII) protein is the most accepted treatment for bleeding episodes in hemophilic patients.

The production of recombinant FVIII by biopharmaceutical industries mostly uses murine cell lines such as Chinese hamster ovary cells (CHO) and baby hamster kidney cells (BHK) with many studies having shown the efficacy and safety of these strains.3, 4, 5 However, there are some human pattern post-translational modifications which the murine cell lines are not able to emanate as these cells do not present the necessary cellular machinery.6, 7 An alternative would be to use human cell lines capable of performing the post-translational modifications that occur in the human body.

FVIII is one of the most complex proteins produced by recombinant DNA technology. Although recombinant coagulation factor products have been available in the market for 20 years,8, 9, 10 the heterologous expression of FVIII still presents some problems such as: gene size, low mRNA expression, and low protein secretion as most is trapped in the endoplasmic reticulum. Moreover, all currently approved recombinant products are produced using murine cells that have a non-human pattern of post-translational modifications.

This work generated two human cell lines: HEK293, derived from kidney, and HepG2, derived from liver tissues which were genetically modified using a new retroviral vector derived from the murine leukemia virus (MLV). The pMFG-FVIII-P140K vector contains the human FVIII gene with the B domain deleted and the selection marker gene (P140K) that confers resistance to the drugs, O6-benzylguanine and temozolomide. With this system, high levels of recombinant FVIII were produced in HepG2 cells (8 IU/mL).

Selecting a human cell line with high productivity of recombinant FVIII is the first step in the future production of a recombinant protein more similar to that produced by the human body with a resulting low risk of inhibitory antibody development, thereby providing a better quality of life for hemophiliac patients.

The goal of this study was to characterize and map the integration sites of this new vector in two human cell lines. Highly sensitive ligation-mediated polymerase chain reaction (LM-PCR) was combined with sequencing to identify integration patterns of a MLV vector carrying rFVIII. In this work, a detailed mapping of integration sites of the human cell expression of recombinant FVIII was performed using a very stringent selection method. Analyses are very important to understand the behavior of these viral vectors and whether their integration sites can influence the expression of recombinant proteins.

Methods

Cell lines and retroviral vector

Human hepatocellular carcinoma cell (HepG2 – ATCC number HB-8065) and human embryonic kidney cell lines (Hek293 – ATCC number CRL-1573) were transduced with a retrovirus vector (pMFG-P140K) that contains elements derived from the Moloney murine leukemia virus (MoMuLV). This is a bicistronic vector, which can express the P140K resistance gene and the human B-domain-deleted FVIII coagulation factor. Stable recombinant human FVIII-producing cell lines were obtained by highly stringent O6-benzylguanine and temozolomide treatment. The rFVIII expressing cells were named HepG2FVIIIdB/P140K and Hek293FVIIIdB/P140K.

Factor VIII activity quantification

Supernatants were harvested and analyzed for FVIII expression using a one-stage clotting assay (activated partial thromboplastin time – aPTT). The FVIII concentrations are given as international units (IU) as defined by the World Health Organization; 1 IU/mL FVIII is equivalent to 200 ng/mL. Briefly, a test sample of 100 μL of FVIII-deficient plasma (Biomerieux, Durham) was incubated at 37 °C with 100 μL of the APTT reagent (Platelin® LS – Biomerieux). Clotting was initiated by the addition of 100 μL 25 mM CaCl2 and clotting times were determined using the COAG-A-MATE device (Organon Teknika).

Ligation-mediated polymerase chain reaction

Integration sites were cloned by LM-PCR as previously described. Briefly, genomic DNA was extracted from 3 × 106 cells digested with MseI and SacI to prevent amplification of internal 5′ long terminal repeat (LTR) fragments, and bound to a MseI double-strand linker. LM-PCR was performed with nested primers specific for the LTR and the linker. PCR products were shotgun-cloned by the TOPO TA cloning kit (Invitrogen) into libraries of integration junctions which were sequenced to saturation.

Sequence analysis

The sequences generated were subjected to an initial analysis using the ChromasPro computer program. In this first stage, only the sequence that contained the adapter sequence (linker) at one end and the sequence related to the retroviral LTR at the other end were selected. The sequences were converted into Fasta format and sent to the Human Genome BLAT database to analyze the homology of the human genome. Only sequences that showed over 95% homology with sequences stored in the database were considered.

Then, sequences were submitted to analysis using the public Quickmap tool, which allows detailed mapping of local integration of the retroviral vector. This tool analyzes a wide range of data on genomic features (genes, transcription start sites, CpG islands, and TFBS, among others) in a window of ±250 kb around the vector integration site. Moreover, the strands on either side were analyzed separately and the individual distance of each feature to the base of integration was calculated in base pairs.

Statistical analysis

As described by Laufs et al. in 2006, in order to test whether the number of integrations is equally distributed along the chromosomes, the chi-squared goodness-of-fit test was used to analyze whether the observed number of integrations (oi) arose from a multinomial distribution with specified expected integrations (ei) for the 24 chromosomes (22 autosomes as well as the sex chromosomes, X and Y). The differences obtained were highly significant as evidenced by the retroviral insertion site in chromosome score (RISC) defined as: (oi − ei)2/ei values [oi < ei then RISC score × (−1)].15, 16 For the detection of preferred genomic integration sites a cut-off of (oi − ei)2/ei = 3 was set.

Fisher's exact test was used to determine whether there is an association between the integrations and the TFBS or CpG island regions in the two cell lines (p-value ≤ 0.05).

Results

Activity analysis of biologically active recombinant factor VIII

For the characterization of recombinant cell lines, rFVIII activity levels were determined as previously detailed.

As shown in Table 1, the HepG2FVIIIdB/P140K and Hek293FVIIIdB/P140K cell lines showed high secretion of biologically active recombinant FVIII (about 7.9 IU/mL and 2.1 IU/mL, respectively).

Table 1

Biological activity of recombinant FVIII in HepG2FVIIIdB/P140K and Hek293FVIIIdB/P140K cell lines.

Sample	IU/mL
DMEM medium	0.01
Supernatant HepG2 (non-transduced)	0.04
Supernatant Hek293 (non-transduced)	0.12
Supernatant HepG2FVIIIdB/P140K	7.90
Supernatant Hek293FVIIIdB/P140K	2.10

Chromosomal integration pattern

A total of 422 clones were sequenced of which 201 were derived from HepG2FVIIIdB/P140K and 221 from Hek293FVIIIdB/P140K cells. Of these, 302 (71.56%) sequences were found in the human DNA database with identities ≥98% (123 from HepG2 and 179 from Hek293 cell lines). The integration sites were sequenced to saturation; these sites began to repeat subsequent to the integrations described here. DNA sequences were submitted to an analysis using the Quickmap tool, which classifies sequences as ‘true’ when they are homologous to only one place in the human genome, or ‘false’ when sequences are ambiguous, i.e. they are homologous to more than one location in the genome. Of the 123 HepG2FVIIIdB/P140K sequences, 73 were considered true whereas the remaining 50 were ambiguous. For the 179 Hek293FVIIIdB/P140K sequences, 64 were true and 115 were ambiguous.

HepG2/FVIIIdBP140K clone sequences showed that integrations occurred in 18 of 23 chromosomes with preference for specific chromosomes. For these cells, 19% of the integrations occurred within chromosome 19, followed by 16% within chromosome 17 and 9.6% within chromosome 11. For the Hek293FVIIIdB/P140K cells, insertions affected 20 of the 23 chromosomes, with high frequencies of integrations occurring within chromosomes 9 (14%) and X (9%) (Figure 1A).

Figure 1

(A) Integration pattern of retroviral vector pMFG-FVIII-P140K in the 23 chromosomes of two cell lines producing recombinant factor VIII. (B) RISC score – number of chromosomal integrations observed versus number of expected integrations of the PMFG-FVIII-P140K vector in the HepG2FVIIIdB/P140K and Hek293FVIIIdB/P140K cell lines (cut-off = 3).

Statistical analysis of the number of integrations observed versus expected number of integrations is given by the RISC score. Using a cut-off of 3, the profiles of integration of retroviral vectors are not random and differ between the cells studied. For the HepG2 cell line there was a preference for chromosomes 19 (RISC = 94.28), 17 (RISC = 49.28) and 11 (RISC = 3.44), and for the HEK293 cell line, a preference for chromosome 9 (RISC = 10.67) was observed as shown in Figure 1B.

Frequency of integration near transcription factor binding sites and CpG islands

In this study, a total of 140 TFBS were affected by integrations that occurred in HepG2FVIIIdB/P140K cells. Of these, five integrations occurred in areas within 5 kb, nine integrations within 10 kb and 39 within 30 kb of TFBS. In Hek293FVIIIdB/P140K kidney cells, a total of 117 TFBS were affected, and only two insertions occurred within 5 kb, one integration within 10 kb and nine within 30 kb of TFBS. The integration profiles of TFBS regions showed no statistically significant differences between the two cell lines according to Fisher's exact test (p-value = 0.606) as shown in Figure 2A.

Figure 2

(A) Number of retroviral vector integrations observed in regions near to transcription factor binding sites (TFBS) (p-value = 0.606). (B) Number of retroviral vector integrations observed in regions near to CpG island (p-value = 0.5462).

The distance of integration sites from the next CpG islands up- or down-stream was also quantified. Of 73 integrations of the retroviral vector, 106 CpG islands were affected in the HepG2 cell line. Of these, 17 occurred in areas within 5 kb, eight integrations within 10 kb and 25 within 30 kb of CpG islands. Of 64 integrations of the retroviral vector in Hek293FVIIIdB/P140K kidney cells, 81 CpG islands were affected with six of these insertions occurring within 5 kb, only one integration within 10 kb and seven within 30 kb of CpG islands. As shown in Figure 2B, there was no statistical differences between cell lines related to the integration profiles of the retroviral vector pMFG-FVIII-P140K near CpG islands (Fisher test; p-value = 0.5462).

Analysis of targeted genes

Of the 73 sequences analyzed from HepG2/FVIIIdBP140K cells, 38 (52%) were located in reference sequence (RefSeq) genes. Of all integrations, only 6% were in exons and 94% were in intronic regions. For the 64 clones of the Hek293/FVIIIdBP140K cell line, 28 (44%) were integrated into RefSeq genes; 10% were in exons and 90% in regions of introns. The frequency of retroviral vector insertions in RefSeq genes in HepG2 cells (52%) was higher than the results generated by a set of 10,000 random integrations simulated by the Quickmap program (about 40%). Hek293 cells showed a similar frequency to that expected randomly (44%).

For HepG2 cells, of the 38 integrations in RefSeq genes, the gene with the largest number of insertions (32%) was the collagen type 1 (COL1A1) gene, followed by the Down syndrome cell adhesion molecule (DSCAML1) gene (13%). These genes are located on chromosomes 17 and 11 respectively. Hek293 cells had a total of 28 integrations in RefSeq genes with 29% occurring in the PPAPR3 gene, which is part of the lipid phosphatases family and is located on chromosome 9.

Integration in fragile sites

When retroviral integrations in fragile sites were analyzed, the HepG2/FVIIIdBP140K cells showed 20% of integrations in fragile sites whereas Hek293/FVIIIdBP140K cells showed around 17%. Fragile sites are classified into rare and common according to their frequency and characteristics on chromosomes. Integrations were observed in both groups. As shown in Figure 3, the most affected fragile sites of HepG2 cells were FRA11G (5.35%) and FRA11B (5.35%), whereas the most affected sites for Hek293 were FRA2B (3.77%) and FRA1H (3.77%). When these results are compared with a set of 10,000 random integrations simulated by the Quickmap computer program, there was a difference in the percentage of integration as the integration was random; expected frequencies would be 0.18% of fragile sites for HepG2 and 0.44% for Hek293.

Figure 3

Integration frequency of pMFG-FVIII-P140K retroviral vector in fragile sites of both HepG2FVIIIdB/P140K and Hek293FVIIIdB/P140K cell lines compared with an expected random integration.

Discussion

The present study aimed to characterize the integration pattern of the retroviral vector pMFG-FVIII-P140K in two cell lines producing recombinant B-deleted FVIII. The strains used were grown in artificial in vitro conditions using culture and selected by very stringent treatment using chemotherapeutic drugs (temozolomide and O6-benzylguanine). The integration profile of the retrovirus described is specific to these populations. The cell lines used in this work serve as a model that may not accurately represent the biology of normal primary cells.

Through an analysis in the integration sites of the retroviral vector pMFG-FVIII-P140K, a non-random pattern was observed for both cells since there was a preference for insertion into specific chromosomes (19 and 17 in HepG2 and 9 in Hek293) in a pattern that was not proportional to the size of the chromosomes. The distribution of integration sites can be influenced by various characteristics of the cell population including the fact that these cells were selected with high stringency and have a difficult karyotype to define. The differences in the integration frequencies between chromosomes are in part a function of gene density. For example, Mitchell et al. in 2004 described that the gene-rich Chromosome 19 showed more integration than expected by chance, whereas the gene-poor Chromosome 18 showed less integration. The intracellular location of the chromosomes can also influence integration, as it has been suggested that this is relatively fixed for each cell type but differs between cell types.18, 19, 20 Moreover, retroviral vector integration into chromosomes with high gene density could explain why, when the two cell lines were compared in this study, the expression of rFVIII in HepG2 was almost four times higher than for HEK293 cells.

The analysis of insertion in genomic regions showed that for the HepG2 cell line, 52% of the integrations were in genes encoding proteins, whereas for HEK293 this figure was 44%. These results corroborate those of other authors who reported that the integration of viral vectors derived from MLV are greater than 45% in cells such as the hematopoietic stem cells of rhesus monkeys, CD34+ cells and HeLa cells. It is worth mentioning that despite the high frequency of integration within protein-coding genes, more than 90% occurred in intronic regions in both cell lines. Ustek et al., using next generation sequencing technology, showed a high frequency of intragenic integration sites in 293T cells using the lentiviral vector pLVTHM.

In regard to the integration into genomic regions, such as CpG islands and TFBS, the HEK293 and HepG2 cell lines showed similar integration patterns. However it is interesting to note that when the distance of retroviral vector integration was analyzed, the results for integration within 1 kb of CpG islands and TFBS are not in agreement with the literature11, 17, 18, 24 as most of the integrations occurred at a distance of between 30 and 60 kb. This finding may be related to the fact that in most published studies the authors used infected cells and analyzed the integration within a maximum of 48 h and did not use any selection strategy, unlike the present study in which the cells remained in a culture after transduction of the retroviral vector and underwent highly stringent treatment to select a cell clone producing FVIII.

A study by Russo-Carbolante et al., using the Sk-Hep and 293T cell lines and Green Fluorescent Protein (GFP) for selection, showed that the lentiviral vector had different integration profiles for the two strains. However, when analyzing insertion to specific regions of the genome, such as CpG islands and transcriptional start sites (TSS), both cells showed similar integration patterns with over 50% of the integrations occurring at a distance of within 30 kb, thus in accordance with the present study. Since the cells and viral vectors used were different and the integration pattern was similar between cells, this suggests that it is possible that the selection strategies used may have an influence on the identification of cell clones with similar integration sites.

Conclusion

These results show that the vector integration profiles of both cells were different from those described in the literature possibly because of the use of selection systems. There is evidence that there is no preferential integration site of the vector when the integration profiles into genomic regions are analyzed. However, the vector showed non-random patterns of chromosomal integration which differ between the studied cell lines.

HepG2FVIIIdB/P140K cells expressed higher levels of FVIII compared to Hek293FVIIIdB/P140K, possibly due to a higher number of integrations in chromosomes with high gene density (chromosomes 17 and 19).

Together, these results suggest that cell type can affect the profile of chromosomal integration of the retroviral vector used and such discrepancies may interfere in the level of expression of recombinant proteins.

Conflicts of interest

The authors declare no conflicts of interest.