Antigen Ii-E1E2E6E7 was generated as previously described (28). The cassette was adapted to human codon usage using the GeneOptimizer Algorithm (Thermo Scientific), excluding canonical vaccinia transcription termination signal T₅NT (32) and the known instability motifs - runs of G₅ and C₅ (23). Approximately 700 bp flanking regions of the DelIII-locus were considered during the optimization run, so that the algorithm not only reduces homology within the antigen but also between antigen and flanking vaccinia sequence.

Adherent AGEI.CR pIX (CR pIX) cells were cultivated in DMEM/F-12 containing 5% fetal calf serum (FCS) as previously described (31). CR pIX cells in suspension were cultivated as previously described (31). Modified AGEI.CR pIX (CR pIX PRO) suppressor cells were generated by transduction of CR pIX cells with a VSV-G pseudotyped lentiviral vector at MOI 1 in the presence of 8 µg/ml polybrene (hexadimethrine bromide, Cat. H9268, Sigma-Aldrich, Germany) coding for the ProVector expression cassette. In detail, the ProVector expression cassette contains Plin2 shRNA under control of a U6 promoter, the tetracyclin repressor gene (tetR) controlled by a CMV promoter, and a puromycin resistance marker gene under the EF1α core promoter. CR pIX PRO cells were cultivated in respective growth medium additionally containing 1 µg/ml puromycin for selection purposes.

Recombinant MVA-CR19 vectors were generated as previously described (33). Briefly, MfPV3 Ii-E1E2E6E7 (28) or GFP was cloned into the shuttle vector SP-CR19III, suitable for integration into MVA’s deletion site III (DelIII) under the control of the MVA E/L-promoter [SSP with one point mutation (33)] or cloned into the shuttle vector pLZAW1, suitable for integration into MVA’s thymidine kinase locus (TK) under the control of the MVA SSP-promoter (34). The shuttle vectors were modified further by inserting two copies of the tetracycline operator sequence (tetO) directly downstream of the MVA promoter and by adding the Plin2 target sequence (P2TS) at the 3’-UTR of the GOI. Recombinant MVA encoding the different genes-of-interest (GOI) were generated by homologous recombination in adherent parental CR pIX or CR pIX PRO suppressor cells that prevent the undesirable expression of the GOI during generation and propagation of the recombinant MVA. To this end, the culture monolayers seeded in a six-well plate were infected with MVA-CR19 with an MOI of 0.05 and transfected with 2 µg of the individual shuttle vector by using the Effectene Transfection Reagent (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The infected/transfected culture was harvested 2-3 days post-infection/transfection, sonicated, and used for infection of a cell monolayer in a six-well plate format. Resulting foci were validated by PCR and an iterative focus purification procedure was initiated until MVA without the correct GOI expression cassette were absent (usually within 5–8 rounds of focus purification).

For the propagation of focus-purified rMVA, the cell harvest material was sonicated by using a Vial Tweeter (set to 20 s of 100% cycle and 90% amplitude, Hielscher, Germany), and CR pIX PRO cells [grown in suspension at 2×10⁶ cells per ml in 1:1 mixtures of CD-U4 and CD-VP4 media (Merck-Millipore, Darmstadt, Germany)] were inoculated with the individual recombinant MVA vectors at MOI 0.05. Finally, rMVA were harvested 48 h - 72 h post-infection and the TCID50 titer was determined. 3 propagation cycles were needed to generate the viral stocks.

Virus titers were measured by using the tissue-culture-infectious-dose 50 (TCID50) assay. Briefly, 3×10⁴ CR pIX PRO suppressor cells per well were seeded in 96-well plates. Virus-containing material was serially diluted 1:10 in 8 technical replicates and added to the cells. 72 h post-infection the cells were screened for CPE by optical inspection under the microscope. TCID50 was calculated with the Spearman-Karber formula (35).

The following antibodies were used in this study: monoclonal mouse anti-myc antibody (9B11, 1:1000 in western blot, 1:500 in flow cytometry, Cell Signaling Technologies, Danvers, USA), polyclonal rabbit anti-vaccinia (1-VA003-07, 1:5000 in western blot, 1:1000 in flow cytometry and immunostaining, Quartett, Berlin, Germany), monoclonal mouse anti-tubulin (DM1α, 1:1000 in western blot, Santa Cruz Biotechnologies, Heidelberg, Germany), polyclonal goat anti-mouse-HRP (115-036-003, 1:5000, Jackson, West Grove, USA), polyclonal goat anti-rabbit-HRP (P0448, 1:2000 in western blot and immunostaining, Dako, Santa Clara, USA), polyclonal goat anti-mouse-PE (550589, 1:200 in flow cytometry, BD, Franklin Lakes, USA), goat anti-rabbit Alexa Fluor 647 (A21244, 1:200 in flow cytometry, Life Technologies, Eugene, USA).

Western blot analysis was performed as previously described (36). Briefly, cells of interest were lysed in TDLB buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40 and 0.5% sodium deoxycholate) supplemented with protease inhibitors (Complete Mini, Roche, Basel, Switzerland). Total protein concentration of the supernatants was measured by the Bradford method (Protein Assay, BioRad, Feldkirchen, Germany). The proteins were separated on 10% SDS-PAGE under reducing conditions and blotted on a nitrocellulose membrane for western blot analysis. Targets were probed with primary and secondary antibodies as listed above. HRP-labeled secondary antibodies and enhanced chemiluminescence substrate or Femto ECL (Thermo Fisher, Waltham, USA) were used for detection in a Chemilux Pro device (Intas, Göttingen, Germany).

Intracellular staining of antigens was performed by using standard methods (36). Cells were fixed and permeabilized with cytofix/cytoperm-Buffer (4% PFA, 1% saponine, in PBS). All washing steps were done with perm/wash-buffer (PBS containing 0.1% saponine). All antibodies were diluted in perm/wash-buffer and incubated for 30 min each. Flow cytometry was performed by using an Attune NxT device (Thermo Fisher, Waltham, USA) with 488 nm and 638 nm excitation and 574/26 nm and 670/14 nm emission filters. Cells were gated on stained, uninfected, and stained mock-MVA-infected cells. Evaluation of data was performed by using Attune NxT software.

3x10⁴ CR pIX and CR pIX PRO cells were seeded and infected with the indicated MVA strain at an MOI of 1. After 6 and 24 hpi, cells were harvested into PBS. Total RNA was prepared with RNeasy Mini kit (Cat. 74106, Qiagen, Hilden, Germany) according to manufacturer’s instructions. Impurities from genomic DNA were digested with Turbo DNA-free kit (Cat. AM1907, Thermo Fisher, Waltham, USA). Reverse transcription quantitative PCR (RT-qPCR) was performed with Luna Universal Probe One-Step RT-qPCR kit (E3006L, NEB, Ipswich, USA) according to manufacturer’s instructions. 2 µl of total RNA extract were used on a StepOnePlus qPCR cycler (Thermo Fisher, Waltham, USA) with the following protocol: Initial reverse transcription at 55°C for 10 min, followed by initial denaturation at 95°C for 1 min, followed by 40 cycles of denaturation for 15 s at 95°C and annealing/extension for 30 s at 60°C.

Following primer and probes were used: To amplify the transgene, E2 forward primer GATACAGGCTGGGACAAAGTG and E2 reverse primer GATCACTGTTCTGCCGATATGC were used together with the E2 probe Fam-CCTGTACTATGTGCTGCACGGCCT-BHQ1. For normalization on virus infection MVA128L was amplified with MVA128L forward primer CGTTTTGCATCATACCTCCATCTT and the MVA128L reverse primer GCGGGTGCTGGAGTGCTT, together with the MVA128L probe Fam-AGGCATAAACGATTGCTGCTGTTCCTCTGT-BHQ1.

For relative quantification of the transgene expression, Ct values were normalized according the following equation (37), using the primer efficiency calculated by the StepOnePlus software:

Immunostaining was used to visualize foci, as described previously (38). 3×10⁵ CR pIX or CR pIX PRO cells per well were seeded in 24-well plates. Cells were infected with an MOI of 0.01 in 1 ml DMEM/F-12 without additives. 2 h post-infection, medium was exchanged to DMEM/F-12 with 5% FCS. 48 h post-infection, cells were fixed with ice-cold acetone/methanol solution (1:1, v/v) and blocked with blocking buffer (PBS containing 3% BSA) overnight at 4°C. Cells were sequentially stained by using an anti-vaccinia antibody and anti-rabbit-HRP, both incubated for 1 h at RT with gentle agitation. All wash steps and antibody dilution steps were conducted in blocking buffer. Finally, foci were stained with TrueBlue Peroxidase Substrate (5510-0030, Seracare, Milford, USA) until foci were visible (usually 5-10 min); subsequently the reaction was stopped with water. Pictures were taken with an inverted microscope (BZ-9000, Keyence, Frankfurt, Germany).

To analyze virus replication, a multi-step growth curve was conducted as described previously (5). 3×10⁵ CR pIX and CR pIX PRO cells per well were seeded in 24-well plates. Cells were infected at an MOI of 0.05 in 200 µl DMEM/F-12 without additives. 45 min post-infection, cells were carefully washed once with PBS and incubated with a medium containing 5% FCS. Cells and supernatant were harvested at 0, 24, 48, and 72 h after adsorption, freeze-thawed thrice, sonicated for 1 min, and titrated as mentioned above.

For analysis of genetic stability of rMVA, the viruses were passaged on parental CR pIX and CR pIX PRO suppressor cells, similar to the protocol of Kremer and colleagues (38). 1.2×10⁶ cells per well were seeded in a 6-well plate 24 h before infection. Cells were infected at an MOI of 0.05 in 1 ml DMEM/F-12 without additives. 2 h post-infection, medium was exchanged to DMEM/F-12 with 5% FCS. 48 h post-infection, cells and supernatant were harvested (=passage 1), freeze-thawed three times, and additionally sonicated three times for 1 min in a cup sonifier. New cells were again infected with 1 ml of a 1:1000 dilution of the virus material from the previous passage. This was repeated to a maximum number of 10 passages. All obtained virus material was stored at -80°C.

To obtain single foci, cell monolayers were infected with different dilutions and covered with growth medium containing 0.8% low melting agarose. 72 h post-infection, foci were picked in 300 µl DMEM/F-12, freeze-thawed three times, sonicated for 1 min, and amplified by infection of 3×10⁵ cells per well in 24-well plates.

For genotyping of transgene insertion loci, PCR was utilized. Genomic DNA of either focus-purified or bulk material of virus-infected cells or virus stocks was prepared with a Quick-DNA MiniPrep kit (D3025, Zymo, Freiburg, Germany) according to the manufacturer’s instructions. PCR was conducted with Q5 High-fidelity DNA polymerase (M0491L, NEB, Ipswich, USA) according to standard protocols. Primers were taken from literature (33, 38) and are listed in Table 1 . Analysis was done by electrophoresis in 0.5%, 0.8% or 1.5% agarose TBE gels, depending on the amplicon size.

For Sanger sequencing analysis, PCR amplicons of bulk material were purified from agarose gel by using QIAquick Gel Extraction Kit (28706X4, QIAGEN, Hilden, Germany), ligated into pJET1.2/blunt (K1231, Thermo Fisher Scientific, Waltham, USA) and used for transformation of E. coli. Plasmids were isolated from single clones and sequenced.

Exogenous cellular gDNA was depleted from virus-containing material as described previously (30). Briefly, virus stock or virus-containing material was precipitated by the addition of polyethylene glycol to a final concentration of 8% (w/w), incubated for 30 min on ice, and subsequently centrifuged at 6600 g for 1 h. The translucent pellet was dissolved in PBS and exogenous gDNA was digested with 8 units of TurboDNAse for 1 h, followed by adding 20 mM EDTA and heat-inactivation at 80°C for 10 min. Viral gDNA was prepared with a Quick-DNA MiniPrep kit (D3025, Zymo, Freiburg, Germany) according to the manufacturer’s instructions. 10 µl of each gDNA were barcoded for NGS with the Nextera XT DNA library Prep kit (FC-131-1096, Illumina, San Diego, USA) according to the manufacturer’s instructions. NGS was performed with the NextSeq500 system by using a NextSeq500/550 High Output Kit v2.5 with 300 cycles (20024905, Illumina, San Diego, USA). FastQ files were evaluated with CLC Genomics Workbench 22 (Qiagen, Hilden, Germany). Obtained sequences were assembled by using derivatives of the MVA-CR19 genome sequence (GenBank accession number KY633487.1) with Ii-E1E2E6E7 transgene integrated into DelIII or TK locus. Normalized transgene coverage was calculated with the formula:

We have demonstrated previously that a fusion protein comprising the early antigens E1, E2, E6 and E7 of Macaca fascicularis papillomavirus type 3 (MfPV3) are well suited to induce potent T cell responses in outbred CD1 and OF1 mice when delivered via DNA or adenoviral vectors (28). Repeated attempts to generate a recombinant MVA (rMVA-CR19) expressing the corresponding Ii-E1E2E6E7 fusion protein for booster immunization purposes was hampered by the stepwise loss of transgene positive foci alongside the focus selection process. This led us to hypothesize that expression of the transgene Ii-E1E2E6E7 imposed a disadvantage on rMVA propagation.

Hence, we herein aimed to develop a system to suppress transgene expression during the generation, selection and amplification of the rMVA. For this purpose, we first modified the AGEI.CR pIX cell line (CR pIX) by means of lentiviral transduction to constitutively express the tetracycline repressor (TetR) together with a rat perilipin-2 (Plin2) shRNA yielding AGEI.CR pIX PRO (CR pIX PRO). Complementary, two tetracycline operator sequences (tetO) and a Plin2 shRNA target sequence (P2TS) were integrated into the 5’ and the 3’ UTR of the transgene expression module of the transfer plasmid used to generate rMVA via in vitro recombination (IVR). According to these design features, residual mRNAs possibly resulting from incomplete TetR/tetO-mediated transcription repression will be degraded via binding of Plin2 shRNA to the 3’ PT2S on the transcript, collectively resulting in a sustained knock-down of toxic transgene expression. The CR pIX PRO suppressor cell line used in conjunction with the tetO/Plin2 transfer plasmids for IVR, rMVA focus selection and subsequent rMVA amplification is in the following referred to as ProVector system ( Figure 1A ).

Performing the IVR with the transfer plasmids encoding the presumably toxic MfPV3 Ii-E1E2E6E7 fusion protein, now flanked by the tetO/P2TS control elements, in the CR pIX PRO suppressor cells and applying the ProVector system for subsequent rMVA selection, we eventually succeeded to generate two rMVA: MVA-CR19-TK (short CR19 M-TK) and MVA-CR19-DelIII (CR19 M-DelIII) with the MfPV3 Ii-E1E2E6E7 transgene sequence integrated either into the thymidine kinase (TK) locus or, alternatively, in the deletion III (DelIII) locus, respectively ( Supplementary Figure 1 ). Virus stocks for the subsequent experiments were produced starting from single foci via 3 rounds of expansion on CR pIX PRO suppressor cells.

Specific knock-down of Ii-E1E2E6E7 expression by CR pIX PRO suppressor cells was initially proven via western blot analysis and quantified by flow cytometry using a C-terminally fused myc-tag for expression monitoring of the fusion protein, due to unavailability of MfPV3 early protein-specific antibodies and HPV-antibodies are not cross-reactive ( Figures 1B, C ). Expression of Ii-E1E2E6E7 was clearly reduced in CR pIX PRO suppressor cells as compared to parental CR pIX cells, when infected with CR19 M-TK and CR19 M-DelIII, respectively, suggesting successful suppression of transgene expression ( Figure 1B ). GFP expression was comparable in parental CR pIX cells and CR pIX PRO suppressor cells following infection with rMVA-GFP lacking tetO and P2TS, respectively. This demonstrates that the presence of the TetR and Plin2 shRNAs do not impact transgene expression per se. Flow cytometry analysis confirmed approximately 4.3-fold and 6-fold suppression of Ii-E1E2E6E7 expression in CR pIX PRO suppressor cells compared with the parental CR pIX cells 48 hpi with CR19 M-TK and CR19 M-DelIII ( Figure 1C ; Supplementary Figure 2 ). Ii-E1E2E6E7 mRNA levels were determined by RT-qPCR and shown to be reduced in CR pIX PRO suppressor cells by a factor of 5.8 (CR19 M-DelIII) or 1.98 (CR19 M-TK) at 6 hpi, respectively, and by a factor of 11.43 (CR19 M-DelIII) or 2.96 (CR19 M-TK) after 24 h ( Supplementary Figure 3 ). The observed reduction of Ii-E1E2E6E7 transcripts level in the CR pIX PRO suppressor cells might be underestimated, and mainly capture the impact of TetR-mediated suppression of transcription, as exact quantification of siRNA-mediated mRNA degradation is hard to achieve by means of RT-qPCR (39).

Following successful generation of an rMVA-MfPV3-Ii-E1E2E6E7 virus stock on CR pIX PRO suppressor cells, we set out to test the hypothesis that (i) expression of certain potentially harmful transgenes can have a detrimental impact on virus replication and that (ii) replicative capacity can be restored in such cases by restricting expression of such transgenes. For this purpose, replication kinetics of CR19 M-DelIII and, for comparison, CR19 GFP, and CR19 empty were measured under restricting and non-restricting conditions. CR pIX PRO suppressor cells and non-modified parental CR pIX cells were infected with the respective rMVA at an MOI of 0.05, and a multistep growth curve was generated by harvesting infected cells together with cell supernatant after 0, 24, 48, and 72 hpi. Titration on CR pIX PRO suppressor cells showed that the titer of CR19 M-DelIII was significantly reduced by a factor of approximately 10 when propagated on CR pIX cells compared with CR pIX PRO suppressor cells at all time points ( Figure 2A ). As a reference, virus titer of CR19 M-DelIII on CR pIX PRO cells was equal to the titers of both control viruses CR19 empty and CR19 GFP that in turn displayed no differences on both cell lines. Consistent results were obtained when comparing the focus size of the rMVA at 48 hpi (MOI of 0.01) ( Figure 2B ; Supplementary Figure 4 ). Immunostaining with a vaccinia-specific antibody revealed comparable foci sizes formed by CR19 empty, CR19 GFP, and CR19 M-DelIII on CR pIX PRO suppressor cells. In contrast, smaller sized foci with less intense staining were notified following infection of the parental CR pIX cells CR19 M-DelIII expressing the Ii-E1E2E6E7 transgene, but not for CR19 GFP and CR19 empty.

Taken together, this suggests that unrestricted expression of the MfPV3 Ii-E1E2E6E7 antigen results in a replication disadvantage for CR19 M-DelIII and, consequently, reduced virus titers and mitigated virus spread. This replication deficit can be restored when Ii-E1E2E6E7 expression is knocked-down in CR pIX PRO suppressor cells.

Provided the MfPV3-Ii-E1E2E67 transgene has negative impact on the MVA replicative capacity under non-restricting conditions, the transgene might also have an influence on the genetic stability of the rMVA by imposing a strong negative selection on such transgene-expressing rMVA. To test this, we first generated a CR19 M-DelIII and CR19 M-TK seed virus stock (defined as passage 0) by expanding the originally selected, positive foci (i.e. passage -3) in three consecutive amplification steps on CR pIX PRO suppressor cells, respectively. Such seed virus stocks were then serially passaged side-by-side on the parental CR pIX cells and on CR pIX PRO suppressor cells for 10 passages as described by Kremer et al. (38). Harvested virus samples were subsequently titrated under restricting conditions on CR pIX PRO suppressor cells ( Supplementary Figure 5 ).

Titration of CR19 M-DelIII virus samples harvested after each passage on CR pIX PRO suppressor cells yielded similar titers across all passages with a trend towards marginally higher titers in later passages, which did, however, not reach statistical significance ( Figure 3A ). Starting from the same seed virus stock (passage 0), CR19 M-DelIII lost already after 1 passage on the parental CR pIX cells one order of magnitude in virus titer as compared to the same virus passaged on CR pIX PRO suppressor cells ( Figure 2A ). This observation was in line with previous findings suggesting a replication disadvantage of CR19 M-DelIII under non-restricting conditions in the parental CR pIX cells ( Figure 2 ; Supplementary Figure 4 ). Beginning with passage 3, the titers of CR pIX-passaged CR19 M-DelIII increased and reached the same level as CR-pIX-PRO-passaged CR19 M-DelIII from passage 5 onwards.

A different effect was observed when CR19 M-TK was passaged on the two cell lines: Whilst CR19 M-TK, similar to CR19 M-DelIII, experienced a significant loss in titer after a single passage on the parental CR pIX cells, the titers of CR19 M-TK did - unlike notified for CR19 M-DelIII - not increase while being passaged on the parental CR pIX cells and consistently remained approximately 100-fold below the levels obtained for CR19 M-TK passaged on CR pIX PRO suppressor cells ( Figure 3B ). In general, CR19 M-TK exhibited slightly lower virus titers as compared with CR19 M-DelIII across all passages on both cell lines.

We next quantified Ii-E1E2E6E7 transgene expression under non-restricting conditions in parental CR pIX cells of CR19 M-DelIII samples harvested after each passage on either CR pIX or CR pIX PRO suppressor cells, respectively, in a high-throughput flow-cytometry-based assay. Infected CR pIX cells were co-stained to monitor intracellular expression of Ii-E1E2E6E7 (via its myc-tag) and vaccinia proteins (polyclonal anti-vaccinia-antibody) 48 hpi ( Figure 4 ). As controls, uninfected cells and CR19-empty-infected cells were used to gate for CR19-infected cells (vac⁺), and CR19 M-DelIII to gate for Ii-E1E2E6E7 expression out of vac⁺ cells ( Supplementary Figure 6 ). A rapid reduction in the frequency of Ii-E1E2E6E7 expressing cells among the CR19 M-DelIII-infected cells was observed. This was already apparent with virus samples derived from passage 1 on CR pIX cells ( Figure 4A ). After only 3 passages on the parental CR pIX cells, expression of Ii-E1E2E6E7 was lost. In contrast, expression of Ii-E1E2E6E7 could be observed with CR19 M-DelIII passaged on CR pIX PRO suppressor cells throughout all 10 passages. However, even suppression of transgene expression in the ProVector system could not prevent a continuously decreasing fraction of Ii-E1E2E6E7 expressing cells amongst vaccinia virus positive cells with passages from 70% to only 10%. A rapid reduction in the frequency of Ii-E1E2E6E7⁺ cells amongst vaccinia virus antigen positive cells was also observed for CR19 M-TK passaged on CR pIX cells ( Figure 4B ). Interestingly, no significant difference in the fraction of Ii-E1E2E6E7 expressing cells was observed with passage level of CR19 M-TK grown on CR pIX PRO suppressor cells.

In conclusion, this experiment suggests a strong negative selection on rMVA expressing Ii-E1E2E6E7. Suppression of transgene expression by CR19 pIX PRO suppressor cells led to prolonged production of Ii-E1E2E6E7. Remarkably, the integration site had a major impact on the stability of transgene expression: Whereas expression of the DelIII-inserted transgene continuously declined upon passaging, even when the virus was passaged under restricting conditions, the TK-locus-inserted transgene appeared to be maintained more stably, but was strictly dependent on passaging on CR19 pIX PRO suppressor cells.

We next investigated whether the reduction of Ii-E1E2E6E7-expressing CR19-M-DelIII- and CR19-M-TK-infected cells is caused by alterations in the DNA sequence of the transgene expression cassette. PCR analyses on single foci from selected passages were performed with primers that bind in the flanking regions of the DelIII and TK locus, respectively (38). In line with the transgene expression analysis above, the fraction of correctly sized PCR amplicons for CR19 M-DelIII quickly declined with the number of passages on CR pIX cells ( Supplementary Table 1 ). Some foci, which proved positive with vaccinia-specific primers, showed no specific DelIII amplicon, neither the full length, not a shortened PCR product. This may be explained by acquisition of mutations or deletions in the primer binding sites. In contrast, the expected PCR amplicon could be detected in almost all foci of CR19 M-DelIII passaged on CR pIX PRO.

However, this analysis does not exclude short deletions, point mutations or insertions/deletions of few bases that might result in frame shifts or truncated products. Thus, next-generation sequencing (NGS) was employed as an unbiased method to deeply characterize the genetic integrity of the passaged rMVA. Predominantly viral DNA was isolated from CR-pIX- and CR-pIX-PRO-passaged CR19 M-DelIII and CR19 M-TK by depleting cellular gDNA, and subjected to Illumina NextSeq500 deep-sequencing. The resulting reads were aligned to the CR19 M-DelIII guiding sequence ( Supplementary Table 2 ). The obtained coverage maps revealed a wide deletion at the DelIII locus when CR19 M-DelIII had been passaged on parental CR pIX cells ( Figures 5A, B ; Supplementary Figure 7A ), whereas the central region of the viral genome was stable. This shows that recombination especially affects the transgene (zero coverage in P5 and P10) and adjacent regions (normalized coverage below 0.5). To quantify the fraction of the DelIII locus deletion along the passages, the mean read coverage of the transgene Ii-E1E2E6E7 was normalized to the mean read coverage of the essential MVA-DNA polymerase gene locus (MVA056L) ( Figure 5C ; Supplementary Table 2 ). The normalized Ii-E1E2E6E7 mean read coverage of CR19 M-DelIII rapidly dropped to almost 0% within three passages on CR pIX cells, whereas normalized mean read coverage of Ii-E1E2E6E7 of CR–pIX-PRO-passaged CR19 M-DelIII was overall high with a trend towards some decline at later passages.

Aligning the NGS reads of CR pIX-passaged CR19 M-DelIII with the expected CR19 M-DelIII genome sequence also showed a reduction in read coverage upstream of the transgene ( Supplementary Figure 7A ), spanning from MVA157L to MVA164R, and also a complete absence of MVA165R and parts of MVA166R, both downstream of the transgene. Thus, this comprises not only the deletion of the transgene but also large sequence parts flanking the DelIII locus and extending beyond the flanking sites used for homologous recombination during generation of the recombinant virus ( Figure 5D ).

CR-pIX-PRO-passaged CR19 M-DelIII exhibited a slight reduction in read coverage in passages 5-10. Detailed analysis revealed reduced read coverage spanning from the middle of the E2 gene within the transgene until MVA165R (nucleotide position 161317 to 165771, Figure 5B ; Supplementary Figures 7B, C ), indicative for a truncation of the transgene. This may explain the continuously decreasing fraction of Ii-E1E2E6E7 expressing cells amongst vaccinia virus positive cells from 70% (passage 1) to only 10% (passage 10).

To confirm the deletions found by NGS, a PCR analysis with a primer pair spanning from MVA155R to MVA167R was done with the same DNA preparation used for NGS ( Figures 5E, F ). This PCR analysis confirmed fast deletion of the transgene from the DelIII locus when passaged on parental CR pIX cells ( Figure 5F ), resulting in bands of a lower molecular weight than the expected full-length PCR amplicon of 14169 bp (CR19 M-DelIII; black arrow, Figure 5F ). Moreover, frequent detection of bands of a lower molecular weight than the expected empty DelIII integration site were observed (< 9471 bp, white arrow). These bands resemble deletions not only of the transgene itself but also of neighboring genes upstream and downstream of the DelIII locus. This PCR analysis was repeated for an independent replicate of the passaging experiment (without depletion of cellular gDNA) with similar results ( Supplementary Figure 8 ).

In contrast to CR19 M-DelIII passaged on CR pIX, CR19 M-TK bands resembling shortened PCR products as a result of deletions represented only a minor fraction in the PCR analysis ( Figures 6A, B ; Supplementary Table 1 ). This PCR analysis was repeated for an independent replicate of the passaging experiment and the results were similar ( Supplementary Figure 9 ). This was also confirmed by the deep-sequencing analysis of the passaged CR19 M-TK that revealed no difference in normalized transgene coverage between the two cell lines (not shown). However, instead of deletions, specific mutations within the transgene could be observed ( Figures 6C, D ). Passaging of CR19 M-TK on parental CR pIX cells led to the early accumulation of a virus variant harboring a specific guanine to thymidine exchange resulting in the transition of E265 (GAG) to an early stop codon (TAG; referred to as E265*). The fraction of the E265* rMVA mutant among all rMVA peaked at passage 5, but then declined again until passage 9. Furthermore, during passaging on CR pIX, a variant carrying an insertion of guanine after nucleotide position 893 emerged. This insertion results in a frameshift beginning at L299 leading to an early translational stop at position 342 and thus to a truncated protein, referred to as L299Tfs*342. L299Tfs*342 could already be detected in about 5% of the recovered reads of passage 1 on CR pIX cells and accumulated until all recovered reads in passage 9 exhibited this mutation. Both mutations found by deep-sequencing could be verified by Sanger sequencing of pJET1.2/blunt-subcloned PCR amplicons of passage 3 and 10 ( Supplementary Figure 10 ). L299Tfs*342 could also be observed among CR19 M-TK passaged on CR pIX PRO cells, albeit only in a very small fraction of the recovered reads. The frequency of this mutant increased only gradually so that at passage 10 the fraction of reads without the L299Tfs*342 was still approximately 90%.