Novel endogenous simian retroviral integrations in Vero cells: implications for quality control of a human vaccine cell substrate

Abstract

African green monkey (AGM)-derived Vero cells have been utilized to produce various human vaccines. The Vero cell genome harbors a variety of simian endogenous type D retrovirus (SERV) sequences. In this study, a transcriptome analysis showed that DNA hypomethylation released the epigenetic repression of SERVs in Vero cells. Moreover, comparative genomic analysis of three Vero cell sublines and an AGM reference revealed that the genomes of the sublines have ~80 SERV integrations. Among them, ~60 integrations are present within all three cell sublines and absent from the reference sequence. At least several of these integrations consist of complete SERV proviruses. These results strongly suggest that SERVs integrated in the genome of Vero cells did not retrotranspose after the establishment of the cell lineage as far as cells were maintained under standard culture and passage conditions, providing a scientific basis for controlling the quality of pharmaceutical cell substrates and their derived biologics.

Figure 1

Effects of the AzaC treatment on the transcriptome of Vero cells. The mRNA-enriched fractions obtained from AzaC-treated and -untreated Vero 0111 cells were subjected to whole-transcriptome sequencing as described in the Methods. (a) The transcriptome analysis was performed using CLC Genomics Workbench 10.1 software. Significant open-reading frames (ORFs) were considered to be those with a False Discovery Rate (FDR)-normalized p-value less than 0.0001, and visualized with a volcano plot. (b) Vero cells were cultured with (lower panel) or without (middle panel) AzaC. SRV-related short reads from all paired-end short reads were collected with 15 complete genome sequences of SRV as reference sequences as described in the Methods. The short reads obtained were assembled, extending at a 100% identity cut-off and gap-closing between contigs. A single reasonable contig (9.2 kb) was obtained, followed by gene assignment and an LTR finding analysis, which matches an 8367-bp full-size consensus SERV sequence identified in the genome of the Vero 0111 cell subline (GenBank: AB935214). All short reads obtained were remapped to the SERV of the Vero genome sequence (which is shown in the upper part of the panels), followed by the extraction of genetic variations, as described in the Methods. As a comparison, the SERV assembly collected from the DNA-seq analysis described in our previous study is also shown (upper panel). Read depth is shown in light gray, while the relative levels of SNVs mismatching with the consensus SERV sequence (AB935214) are shown in the following colors (A: light green, T: red, G: orange, C: dark blue).

Figure 2

The landscape of CNV and LOH in three Vero cell lines. Each dot represents the estimated copy number in a 50 kb-length window. X- and y-axes denote chromosomal coordinates and copy numbers, respectively. The chromosome number is labeled under the panel. The boxes colored in pink represent the regions of LOH.

Figure 3

Structural comparison of SERV from different Vero sublines and different SVL. (a) Validation of Vero cell-specific SVL by genomic PCR. Nested PCR experiments were performed using genomic DNA from the three Vero cell sublines and AGM PBMC as described in the Methods. A schematic picture of nested PCR with the primer-binding sites is shown at the top. SVL IDs are specified under the panel. The black arrowheads indicate SERV-integrated PCR fragments, and the white arrowheads SERV-unintegrated PCR fragments. The arrow indicates an unknown band. Note that the sizes of smaller bands are consistent with the sizes expected for SERV-unintegrated genome sequences. 1st rd: 1st round PCR, 2nd rd: 2nd round PCR. All gel images shown include full-length without grouping. (b) A comparative analysis among twelve Vero cell-specific SVL regions. The colored bars show that the homologous region had at least 80% identity between the consensus SERV sequence found in the Vero 0111 genome (GenBank: AB935214) and each SERV region. Genetic features including mutations for respective SVLs are shown on the right side. Identity with 5′-LTR and 3′-LTR sequences of the AB935214 reference in the BLASTN homology search are: 94.3 and 73.4% for SVL13c, 72.0 and 99.0% for SVL15f, 95.4 and 74.0% for SVL22e, and 94.7 and 74.2% for SVL26, respectively. Due to the high identities with both LTRs of the reference, both 5′- and 3′-LTR regions of these SVL integrations are depicted in the panel, although the SVLs with no intervening sequence (13c, 15f, 22e, and 26) are the solo LTRs as highlighted on the right side of the panel.

Figure 4

Phylogenetic tree of SRV and SERV sequences. The SRV and SERV sequences formed two distinct clades. The arrow indicates SVL3b and 4e sequences. Bootstrap values <50% were not shown in the tree. The sequence ID at each tip represents the name and chromosome/scaffold of the SERV sequence: MFA: M. fascicularis, MMU: M. mulatta, PAN: P. anubis, and CSA: C. sabaeus. SERV23.1 and SERV25.2 are the sequences originally reported as SERVs, identified in P. cynocephalus.