Cell Bank Origin of MDCK Parental Cells Shapes Adaptation to Serum-Free Suspension Culture and Canine Adenoviral Vector Production

Phenotypic variation in cultured mammalian cell lines is known to be induced by passaging and culture conditions. Yet, the effect these variations have on the production of viral vectors has been overlooked. In this work we evaluated the impact of using Madin–Darby canine kidney (MDCK) parental cells from American Type Culture Collection (ATCC) or European Collection of Authenticated Cell Cultures (ECACC) cell bank repositories in both adherent and suspension cultures for the production of canine adenoviral vectors type 2 (CAV-2). To further explore the differences between cells, we conducted whole-genome transcriptome analysis. ECACC’s MDCK showed to be a less heterogeneous population, more difficult to adapt to suspension and serum-free culture conditions, but more permissive to CAV-2 replication progression, enabling higher yields. Transcriptome data indicated that this increased permissiveness is due to a general down-regulation of biological networks of innate immunity in ECACC cells, including apoptosis and death receptor signaling, Janus kinase/signal transducers and activators of transcription (JAK/STAT) signaling, toll-like receptors signaling and the canonical pathway of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling. These results show the impact of MDCK source on the outcome of viral-based production processes further elucidating transcriptome signatures underlying enhanced adenoviral replication. Following functional validation, the genes and networks identified herein can be targeted in future engineering approaches aiming at improving the production of CAV-2 gene therapy vectors.


Introduction
Canine adenoviral vectors type 2 two (exhibit a preferential tropism for transducing neurons and present high levels of retrograde transport. These features make them an excellent tool in neurobiology and neuropathophysiology studies, and a powerful gene transfer vehicle for brain gene therapy [1,2]. In addition to gene therapy, these vectors have been used in vaccinations and in oncolytic virotherapy

Adaptation of MDCK Cells to Serum-Free Suspension Growth
MDCK were adapted to two serum-free culture media: Adenovirus Expression Medium AEM and SFM4BHK21. These media were selected because AEM demonstrated to enable appreciable cell growth and CAV-2 production with the MDCK.SUS2 cell line [9] from ECACC, while the SFM4BHK21 medium showed to be suitable to adapt MDCK parental cells from ATCC to grow in single-cell suspension cultures [18].
To adapt ATCC cells, we started from serum-containing adherent cultures growing in static monolayer, and employed two approaches: (i) direct transfer, where cells growing in static monolayer in serum-containing medium were placed directly into stirred cultures with serum-free culture medium (AEM or SFM4BHK21); or (ii) stepwise transfer, where cells were first adapted to the new serum-free culture medium (AEM or SFM4BHK21) in static cultures prior to being placed into stirred cultures with serum-free culture medium. The results showed that ATCC cells could be directly adapted to grow in suspension with SFM4BHK21 (Figure 2A), while the best adaptation strategy with AEM had to follow the stepwise approach ( Figure 2B).
Despite several attempts, MDCK cells from ECACC could not be adapted to grow in suspension

Adaptation of MDCK Cells to Serum-Free Suspension Growth
MDCK were adapted to two serum-free culture media: Adenovirus Expression Medium AEM and SFM4BHK21. These media were selected because AEM demonstrated to enable appreciable cell growth and CAV-2 production with the MDCK.SUS2 cell line [9] from ECACC, while the SFM4BHK21 medium showed to be suitable to adapt MDCK parental cells from ATCC to grow in single-cell suspension cultures [18].
To adapt ATCC cells, we started from serum-containing adherent cultures growing in static monolayer, and employed two approaches: (i) direct transfer, where cells growing in static monolayer in serum-containing medium were placed directly into stirred cultures with serum-free culture medium (AEM or SFM4BHK21); or (ii) stepwise transfer, where cells were first adapted to the new serum-free culture medium (AEM or SFM4BHK21) in static cultures prior to being placed into stirred cultures with serum-free culture medium. The results showed that ATCC cells could be directly adapted to grow in suspension with SFM4BHK21 (Figure 2A), while the best adaptation strategy with AEM had to follow the stepwise approach ( Figure 2B). Therefore, MDCK.SUS2 cells already growing in suspension as aggregates in the AEM medium [9], were used as ECACC representatives and adapted to grow in suspension in SFM4BHK21 medium. Similar to what was described above, cells were either directly transferred to the SFM4BHK21 medium or stepwise adapted by gradually increasing the percentage of this medium in each subculture passage. When directly transferred to SFM4BHK21 medium, MDCK.SUS2 cells showed no considerable growth during the adaptation phase. Moreover, from day 20 to day 29 there was a slight but continuous reduction in the cumulative cell number indicating cell death ( Figure 2C). This contrasted with what was observed with the stepwise approach, in which cells grew continuously from day one of the adaptation phase ( Figure 2D). In the case of MDCK.SUS2, 'direct transfer' applies to SFMBHK21 medium only since these cells were already growing in suspension in AEM medium; one AEM culture was kept as control in each adaptation campaign and is shown for comparison purposes. Stepwise adaptation was performed by increasing the percentage of the SFMBHK21 in the final culture medium used to maintain Erlenmeyer cultures. Figures A1 and A2 show the morphology and Figure A3 shows flow cytometry analysis of adapted cells to better characterize macroscopic changes induced by the adaptation (Appendix A, Supplementary Material). Figure A4 summarizes published MDCK cells derivations (from ATCC or ECACC parental strains) in the context of bioprocess production, including those reported in this work (Appendix A, Supplementary Material). Stepwise adaptation was performed by first adapting cells to the new serum-free medium in static monolayer cultures and then transfer to suspension cultures. (C,D) Cumulative number of MDCK.SUS2 cells (from ECACC) following a direct transfer (C) or stepwise transfer (D) strategy. In the case of MDCK.SUS2, 'direct transfer' applies to SFMBHK21 medium only since these cells were already growing in suspension in AEM medium; one AEM culture was kept as control in each adaptation campaign and is shown for comparison purposes. Stepwise adaptation was performed by increasing the percentage of the SFMBHK21 in the final culture medium used to maintain Erlenmeyer cultures. Figures A1 and A2 show the morphology and Figure A3 shows flow cytometry analysis of adapted cells to better characterize macroscopic changes induced by the adaptation (Appendix A, Supplementary Materials). Figure A4 summarizes published MDCK cells derivations (from ATCC or ECACC parental strains) in the context of bioprocess production, including those reported in this work (Appendix A, Supplementary Materials).
Despite several attempts, MDCK cells from ECACC could not be adapted to grow in suspension in any of these formulations (AEM or SFMBHK21) starting from serum-containing static conditions. Therefore, MDCK.SUS2 cells already growing in suspension as aggregates in the AEM medium [9], were used as ECACC representatives and adapted to grow in suspension in SFM4BHK21 medium. Similar to what was described above, cells were either directly transferred to the SFM4BHK21 medium or stepwise adapted by gradually increasing the percentage of this medium in each subculture passage.
When directly transferred to SFM4BHK21 medium, MDCK.SUS2 cells showed no considerable growth during the adaptation phase. Moreover, from day 20 to day 29 there was a slight but continuous reduction in the cumulative cell number indicating cell death ( Figure 2C). This contrasted with what was observed with the stepwise approach, in which cells grew continuously from day one of the adaptation phase ( Figure 2D).

Growth of MDCK Cells in Suspension Cultures
The growth profile of suspension-adapted cells was analyzed in AEM and SFM4BHK21 media. For each medium, the cells selected to be further evaluated were those showing faster adaptation times and higher cumulative cell numbers. When using SFM4BHK21 medium, this corresponded to direct transfer for ATCC cells and stepwise transfer for ECACC cells. With the AEM medium, it corresponded to stepwise transfer for ATCC cells and to what is designated 'direct transfer' for the ECACC cells, which is not an actual transfer since we started from MDCK.SUS2 cells already growing in this formulation. ECACC and ATCC cell lines showed similar growth profiles for the same culture medium although, although ATCC cells achieved higher values of maximum cell concentrations in the AEM medium ( Figure 3). Overall, AEM was the culture medium in which both MDCK cell lines presented better growth performances.

Growth of MDCK Cells in Suspension Cultures
The growth profile of suspension-adapted cells was analyzed in AEM and SFM4BHK21 media. For each medium, the cells selected to be further evaluated were those showing faster adaptation times and higher cumulative cell numbers. When using SFM4BHK21 medium, this corresponded to direct transfer for ATCC cells and stepwise transfer for ECACC cells. With the AEM medium, it corresponded to stepwise transfer for ATCC cells and to what is designated 'direct transfer' for the ECACC cells, which is not an actual transfer since we started from MDCK.SUS2 cells already growing in this formulation. ECACC and ATCC cell lines showed similar growth profiles for the same culture medium although, although ATCC cells achieved higher values of maximum cell concentrations in the AEM medium ( Figure 3). Overall, AEM was the culture medium in which both MDCK cell lines presented better growth performances. . Growth profile of MDCK cells from ATCC or ECACC adapted to AEM or SFM4BHK21 media. Growth profile of selected MDCK cells in suspension cultures with (A) AEM and (B) SFM4BKH21 media. Cells selected for evaluation were those showing lower adaptation times and a higher cumulative cell number. When using the AEM medium, this corresponded to stepwise transfer for ATCC cells and direct transfer for ECACC cells [9] while in the SFM4BHK21 medium, this corresponded to direct transfer for ATCC cells and stepwise transfer for ECACC cells.

CAV-2 Prodution and Infection Progression in Suspension Cultures with AEM Medium
CAV-2 production was evaluated in suspension cultures of ECACC and ATCC cells growing in the AEM medium. Cells were infected at a concentration of 0.8-1 × 10 6 cells/mL, with medium exchange at infection through centrifugation. The results showed that 72 h post infection was the harvesting time with the highest titers for both cell lines ( Figure 4A). CAV-2 titers obtained with ECACC cells were approximately three-fold higher than those obtained with ATCC cells, with a cell specific productivity of 1870 (±651) IP/cell and 827 (±55) IP/cell, respectively.
Since reduced productivity performance of ATCC cells had already been observed in serumcontaining static adherent cultures ( Figure 1B), we hypothesized that these cells were either less prone to virus entry, or more restrictive in allowing infection progression. To evaluate virus entry, transduction efficiency was estimated by quantifying the percentage of GFP-positive cells after infection with CAVGFP vector. This is an E1-deleted non-replicative CAV-2 viral vector harboring a GFP expression cassette as transgene. Therefore, it does not replicate in these cells reporting only on transduction efficiency directly and, thus, on virus entry indirectly. Adherent cells were also included as control. The results showed less GFP-positive cells when infecting ATCC cells in serum-containing Cells selected for evaluation were those showing lower adaptation times and a higher cumulative cell number. When using the AEM medium, this corresponded to stepwise transfer for ATCC cells and direct transfer for ECACC cells [9] while in the SFM4BHK21 medium, this corresponded to direct transfer for ATCC cells and stepwise transfer for ECACC cells.

CAV-2 Prodution and Infection Progression in Suspension Cultures with AEM Medium
CAV-2 production was evaluated in suspension cultures of ECACC and ATCC cells growing in the AEM medium. Cells were infected at a concentration of 0.8-1 × 10 6 cells/mL, with medium exchange at infection through centrifugation. The results showed that 72 h post infection was the harvesting time with the highest titers for both cell lines ( Figure 4A). CAV-2 titers obtained with ECACC cells were approximately three-fold higher than those obtained with ATCC cells, with a cell specific productivity of 1870 (±651) IP/cell and 827 (±55) IP/cell, respectively. The progression of CAV-2 infection was evaluated by monitoring the expression of the capsid protein pIX (fused with a GFP), one of the last viral products to be expressed right before the generation of viral particles. Additionally, here, serum-containing adherent cultures were included as control. In static serum-containing adherent cultures, flow cytometry analysis revealed low levels of GFP-positive cells in ATCC cells. CAV-2 infected cells in suspension cultures originating from ATCC achieved higher levels of GFP-positive cells at later time points compared to ECACC ( Figure  5A). In addition, ATCC cells showed an atypical CAV-2 replication profile, where the percentage of infected cells seemed to progress more slowly at early time points, relative to ECACC, to then increase abruptly at later time points ( Figure 5A). These results also suggest reduced expression levels of the viral proteins, namely pIX, in ATCC cells compared to ECACC's.
To gather further evidence on potential restrictions to infection progression in ATCC cells, we analyzed cell size variations since an increase in cell size is a known indicator of infection cycle progression during adenovirus replication [20]. Prior to infection, ATCC and ECACC cells in the same culture system showed similar cell volume ( Figure 5B), although cells from suspension cultures were 2.4-fold smaller than those in static cultures. This suggests that adaptation to suspension favored the survival of smaller cells, which is corroborated by flow cytometry analysis ( Figure A3, Appendix A, Supplementary Material). After infection, MDCK cells from ECACC increased in volume in both cultures while no relevant changes were observed for ATCC cells ( Figure 5B). These results support that MDCK cells from ATCC and ECACC react differently to CAV-2 infection, with ECACC cells showing more evidence of enhanced infection progression. Together, our analyses corroborate that ECACC cells were less restrictive to infection progression of CAV-2 than ATCC's, more accentuated in adherent than in suspension cultures. Since reduced productivity performance of ATCC cells had already been observed in serum-containing static adherent cultures ( Figure 1B), we hypothesized that these cells were either less prone to virus entry, or more restrictive in allowing infection progression. To evaluate virus entry, transduction efficiency was estimated by quantifying the percentage of GFP-positive cells after infection with CAVGFP vector. This is an E1-deleted non-replicative CAV-2 viral vector harboring a GFP expression cassette as transgene. Therefore, it does not replicate in these cells reporting only on transduction efficiency directly and, thus, on virus entry indirectly. Adherent cells were also included as control. The results showed less GFP-positive cells when infecting ATCC cells in serum-containing static control cultures although, in suspension cultures, the percentage of GFP-positive cells was similar between ATCC and ECACC ( Figure 4B).
The progression of CAV-2 infection was evaluated by monitoring the expression of the capsid protein pIX (fused with a GFP), one of the last viral products to be expressed right before the generation of viral particles. Additionally, here, serum-containing adherent cultures were included as control. In static serum-containing adherent cultures, flow cytometry analysis revealed low levels of GFP-positive cells in ATCC cells. CAV-2 infected cells in suspension cultures originating from ATCC achieved higher levels of GFP-positive cells at later time points compared to ECACC ( Figure 5A). In addition, ATCC cells showed an atypical CAV-2 replication profile, where the percentage of infected cells seemed to progress more slowly at early time points, relative to ECACC, to then increase abruptly at later time points ( Figure 5A). These results also suggest reduced expression levels of the viral proteins, namely pIX, in ATCC cells compared to ECACC's.

Transcriptome Analysis
To further investigate the differences between MDCK parental cells from ATCC and ECACC, we conducted whole-genome transcriptome analysis. We analyzed infected and non-infected cells from both ATCC and ECACC, cultured either in static and serum-containing medium or in suspension and AEM medium, totaling eight datasets. We started by applying principal component analysis (PCA), an unsupervised data analysis technique that reduces the data dimensionality by creating new variables (principal components, PCs) based on orthogonal transformation that maximizes the variance in the dataset [21]. The first principal components (typically, PC1 to PC5, depending on the dataset size and complexity) capture most of the variance of the dataset, thus providing a good indication of the experimental factors responsible for the differences across samples. Principal component analysis showed that variability is maximized when static serumcontaining cells are compared with suspension serum-free cells, separable immediately by PC1, regardless of cell bank origin or infection ( Figure 6A): PC1 in the x-axis separates static and serumcontaining cultures (on the left side of the axis) from suspension and serum-free cultures (on the right side of the axis). The second layer of variability arises from cell bank origin ( Figure 6A, PC2, y-axis) followed by CAV-2 infection ( Figure 6B, PC3, y-axis).
To identify biological pathways relevant in infection, we compared CAV-2 infected with noninfected cells for all matching pairwise comparisons. Differentially expressed genes were analyzed by Ingenuity® Pathway Analysis (IPA). A total of 157 IPA canonical pathways were identified (Supplementary File 1) and assigned to 7 categories ( Figure 6C). From these, two categories called our attention: (i) 'cell stress and injury' and (ii) 'immune response'. First because these comprised about 40% of significantly enriched pathways; second because these two categories contain the pathways related to progression of infection, which, according to our previous results and hypothesis ( Figure 5) would be a major factor to understand the differences between the two parental cells regarding CAV-2 productivity. To gather further evidence on potential restrictions to infection progression in ATCC cells, we analyzed cell size variations since an increase in cell size is a known indicator of infection cycle progression during adenovirus replication [20]. Prior to infection, ATCC and ECACC cells in the same culture system showed similar cell volume ( Figure 5B), although cells from suspension cultures were 2.4-fold smaller than those in static cultures. This suggests that adaptation to suspension favored the survival of smaller cells, which is corroborated by flow cytometry analysis ( Figure A3, Appendix A, Supplementary Materials). After infection, MDCK cells from ECACC increased in volume in both cultures while no relevant changes were observed for ATCC cells ( Figure 5B). These results support that MDCK cells from ATCC and ECACC react differently to CAV-2 infection, with ECACC cells showing more evidence of enhanced infection progression. Together, our analyses corroborate that ECACC cells were less restrictive to infection progression of CAV-2 than ATCC's, more accentuated in adherent than in suspension cultures.

Transcriptome Analysis
To further investigate the differences between MDCK parental cells from ATCC and ECACC, we conducted whole-genome transcriptome analysis. We analyzed infected and non-infected cells from both ATCC and ECACC, cultured either in static and serum-containing medium or in suspension and AEM medium, totaling eight datasets. We started by applying principal component analysis (PCA), an unsupervised data analysis technique that reduces the data dimensionality by creating new variables (principal components, PCs) based on orthogonal transformation that maximizes the variance in the dataset [21]. The first principal components (typically, PC1 to PC5, depending on the dataset size and complexity) capture most of the variance of the dataset, thus providing a good indication of the experimental factors responsible for the differences across samples. Principal component analysis showed that variability is maximized when static serum-containing cells are compared with suspension serum-free cells, separable immediately by PC1, regardless of cell bank origin or infection ( Figure 6A): PC1 in the x-axis separates static and serum-containing cultures (on the left side of the axis) from suspension and serum-free cultures (on the right side of the axis). The second layer of variability arises from cell bank origin ( Figure 6A, PC2, y-axis) followed by CAV-2 infection ( Figure 6B, PC3, y-axis). To investigate how the pathways in these two categories differ in cells from the two sources, a second pathway analysis was conducted comparing ECACC to ATCC. Among the 296 significantly enriched pathways, 41 corresponded to pathways in the category of 'cell stress and injury' or 'immune response' with predominance of sub-pathways related to 'pathogen influenced signaling', 'cytokine signaling' and 'apoptosis' (Supplementary File 2). Within these, we further analyzed differential gene expression patterns in the comparison 'ECACC vs. ATCC' cells. These patterns are To identify biological pathways relevant in infection, we compared CAV-2 infected with non-infected cells for all matching pairwise comparisons. Differentially expressed genes were analyzed by Ingenuity ® Pathway Analysis (IPA). A total of 157 IPA canonical pathways were identified (Supplementary File 1) and assigned to 7 categories ( Figure 6C). From these, two categories called our attention: (i) 'cell stress and injury' and (ii) 'immune response'. First because these comprised about 40% of significantly enriched pathways; second because these two categories contain the pathways related to progression of infection, which, according to our previous results and hypothesis (Figure 5) would be a major factor to understand the differences between the two parental cells regarding CAV-2 productivity.
To investigate how the pathways in these two categories differ in cells from the two sources, a second pathway analysis was conducted comparing ECACC to ATCC. Among the 296 significantly enriched pathways, 41 corresponded to pathways in the category of 'cell stress and injury' or 'immune response' with predominance of sub-pathways related to 'pathogen influenced signaling', 'cytokine signaling' and 'apoptosis' (Supplementary File 2). Within these, we further analyzed differential gene expression patterns in the comparison 'ECACC vs. ATCC' cells. These patterns are summarized in Supplementary File 3 and were mapped into four pathways ( Apoptosis and death receptor pathways are involved in programmed cell death in response to external stimuli, including viral infections [22]. Transcriptome data showed that, although ECACC cells had an up-regulation in death receptors (e.g., FAS or TNFRSF12A) and their signaling partners (e.g., TRADD), they presented general down-regulation of apoptotic effectors, namely caspases ( Figure 7). In addition, anti-apoptotic proteins such as BIRC, CFLAR or BCL2 and its homologs, were found to be highly up-regulated in ECACC cells relative to ATCC cells, three to five-fold increase (Supplementary File 3).  Apoptosis and death receptor pathways are involved in programmed cell death in response to external stimuli, including viral infections [22]. Transcriptome data showed that, although ECACC cells had an up-regulation in death receptors (e.g., FAS or TNFRSF12A) and their signaling partners (e.g., TRADD), they presented general down-regulation of apoptotic effectors, namely caspases ( Figure 7). In addition, anti-apoptotic proteins such as BIRC, CFLAR or BCL2 and its homologs, were found to be highly up-regulated in ECACC cells relative to ATCC cells, three to five-fold increase (Supplementary File 3).
The JAK/STAT pathway is a central communication system mediating signaling transduction to link response to external stimuli with gene expression. This communication cascade is involved in cellular development, homeostasis and immunity, including a direct link to the interferon signaling pathway [23]. Transcriptome data revealed a general down-regulation of genes in the branch related to interferon response (Figure 8, left side of the pathway) and an up-regulation of the cell survival branch (Figure 8, right side of the pathway) in ECACC cells, relative to ATCC's. The JAK/STAT pathway is a central communication system mediating signaling transduction to link response to external stimuli with gene expression. This communication cascade is involved in cellular development, homeostasis and immunity, including a direct link to the interferon signaling pathway [23]. Transcriptome data revealed a general down-regulation of genes in the branch related to interferon response (Figure 8, left side of the pathway) and an up-regulation of the cell survival branch (Figure 8, right side of the pathway) in ECACC cells, relative to ATCC's. The toll-like receptor (TLR) signaling is one of the mechanisms of the innate immune system to sense extracellular pathogens [24]. Transcriptome data showed down-regulation of many toll-like receptors in ECACC relative to ATCC cells, most notably TLR4 and TLR6 at the cell surface and known to recognize viral proteins [25]. Important mediators of TLR signaling and its connection to NF-κB canonical pathways were also found to be down-regulated (Figure 9, Supplementary File 3). Additionally, down-regulation of more than three-fold was found for the CD14 antigen, which has also been implicated in the activation of antiviral defense response [26]. The toll-like receptor (TLR) signaling is one of the mechanisms of the innate immune system to sense extracellular pathogens [24]. Transcriptome data showed down-regulation of many toll-like receptors in ECACC relative to ATCC cells, most notably TLR4 and TLR6 at the cell surface and known to recognize viral proteins [25]. Important mediators of TLR signaling and its connection to NF-κB canonical pathways were also found to be down-regulated (Figure 9, Supplementary File 3). Additionally, down-regulation of more than three-fold was found for the CD14 antigen, which has also been implicated in the activation of antiviral defense response [26]. Finally, NF-κB pathway was also one of the most highly enriched when comparing ECACC with ATCC cells. The NF-κB pathway is a central transcription regulator of genes involved in immunity, inflammation and cell survival. The pathway is generally subdivided into the canonical pathway and the non-canonical pathway; a third sub-pathway is often considered (alternative/atypical) [27]. The canonical pathway is directly linked to inflammation and the immune response while the others are more connected to cell proliferation and survival. Transcriptome data indicated a down-regulation of genes involved in inflammation and immune response mechanism (canonical pathway) and an up-regulation of genes involved in the cell survival and proliferation mechanisms (non-canonical and atypical) (Figure 10, Supplementary File 3).
Overall, the mapping of differential gene expression in the pathways more enriched in our analysis (Figures 7-10, Supplementary File 3) showed expression patterns that point to enhanced resistance to apoptosis and a general down-regulation of innate immunity of ECACC cells relative to ATCC cells. Finally, NF-κB pathway was also one of the most highly enriched when comparing ECACC with ATCC cells. The NF-κB pathway is a central transcription regulator of genes involved in immunity, inflammation and cell survival. The pathway is generally subdivided into the canonical pathway and the non-canonical pathway; a third sub-pathway is often considered (alternative/atypical) [27]. The canonical pathway is directly linked to inflammation and the immune response while the others are more connected to cell proliferation and survival. Transcriptome data indicated a down-regulation of genes involved in inflammation and immune response mechanism (canonical pathway) and an up-regulation of genes involved in the cell survival and proliferation mechanisms (non-canonical and atypical) (Figure 10, Supplementary File 3).
Overall, the mapping of differential gene expression in the pathways more enriched in our analysis (Figures 7-10, Supplementary File 3) showed expression patterns that point to enhanced resistance to apoptosis and a general down-regulation of innate immunity of ECACC cells relative to ATCC cells.

Influenza Virus Production in MDCK Cells Suspension Cultures with AEM Medium
This work was focused on MDCK cells as hosts for the production of CAV-2. However, MDCK cells are highly relevant in the context of cell-culture based vaccines for influenza virus. We further questioned whether the differences in productivity observed for CAV-2 would occur when cells are used for the production of influenza virus. Therefore, cells were infected at ~2.5 × 10 6

Influenza Virus Production in MDCK Cells Suspension Cultures with AEM Medium
This work was focused on MDCK cells as hosts for the production of CAV-2. However, MDCK cells are highly relevant in the context of cell-culture based vaccines for influenza virus. We further questioned whether the differences in productivity observed for CAV-2 would occur when cells are used for the production of influenza virus. Therefore, cells were infected at~2.5 × 10 6 cells/mL with two influenza virus strains: A/PR/8/34 (RKI) and A/Uruguay/716/2007 (NIBSC). For the A/PR/8/34 (RKI) strain, virus release dynamics assessed by hemagglutination (HA) assay were similar in ECACC and ATCC cells ( Figure 11A). Accordingly, both cell lines produced similar maximum infectious titers at the same time (~30 hpi) as assessed by 50 % tissue culture infective dose (TCID 50 ) assay (Appendix A, Table A2, Supplementary Materials). The HA and TCID 50 Figure 11B and Appendix A,

Discussion
Phenotypic variation of cultured cell lines has long been acknowledged and is known to be induced by passage number accumulation, certain culture conditions or even by operator handling [28]. Therefore, the cell bank origin is an expectable source of variability. However, it has been largely overlooked since most researchers conduct their experimental activities using cells from a single source, often, from either ATCC or ECACC cell banks. Previous work from our group to adapt MDCK cells to serum-free suspension growth [9,19] and difficulties to obtain the phenotypes reported by others [18], led us to hypothesize that the cell bank origin was a determinant factor in the final outcome of this adaptation process. Moreover, such differences could be extended to virus production performance, an aspect that was yet to be analyzed. Therefore, in this work, we evaluated MDCK parental cells from ATCC and ECACC for CAV-2 production in adherent and suspension cultures.
A morphological analysis of both cells during the first two to three passages from the original cell bank vials revealed higher heterogeneity of ATCC cells ( Figure 1A). This suggests that ATCC cells may be closer to the original population, given that heterogeneity is expected to be reduced with passage number, particularly when fast growing clones overgrow the slower ones. Indeed, although the cells are described as 'parental' in both cases, MDCK were deposited first at ATCC and only later at ECACC [16]. Thus, it is expectable that ATCC cells contain subsets functionally similar to ECACC's. In line with our results, these subsets of cells should be more permissive to virus replication and less likely to survive serum weaning. These functional differences in virus replication have already been demonstrated by sub-cloning ATCC's MDCK cells, which resulted in clones with different abilities to replicate influenza virus [17]. Heterogeneity differences were also reflected in macroscopic morphological parameters by flow cytometric analysis (Appendix A, Figure A3, Supplementary Material) although there was no evidence for a link between the morphological and functional differences.
The impact of higher heterogeneity when adapting to serum-free suspension culture is important. In respect to cell physiology, serum removal and anchorage-independent growth is a

Discussion
Phenotypic variation of cultured cell lines has long been acknowledged and is known to be induced by passage number accumulation, certain culture conditions or even by operator handling [28]. Therefore, the cell bank origin is an expectable source of variability. However, it has been largely overlooked since most researchers conduct their experimental activities using cells from a single source, often, from either ATCC or ECACC cell banks. Previous work from our group to adapt MDCK cells to serum-free suspension growth [9,19] and difficulties to obtain the phenotypes reported by others [18], led us to hypothesize that the cell bank origin was a determinant factor in the final outcome of this adaptation process. Moreover, such differences could be extended to virus production performance, an aspect that was yet to be analyzed. Therefore, in this work, we evaluated MDCK parental cells from ATCC and ECACC for CAV-2 production in adherent and suspension cultures.
A morphological analysis of both cells during the first two to three passages from the original cell bank vials revealed higher heterogeneity of ATCC cells ( Figure 1A). This suggests that ATCC cells may be closer to the original population, given that heterogeneity is expected to be reduced with passage number, particularly when fast growing clones overgrow the slower ones. Indeed, although the cells are described as 'parental' in both cases, MDCK were deposited first at ATCC and only later at ECACC [16]. Thus, it is expectable that ATCC cells contain subsets functionally similar to ECACC's. In line with our results, these subsets of cells should be more permissive to virus replication and less likely to survive serum weaning. These functional differences in virus replication have already been demonstrated by sub-cloning ATCC's MDCK cells, which resulted in clones with different abilities to replicate influenza virus [17]. Heterogeneity differences were also reflected in macroscopic morphological parameters by flow cytometric analysis (Appendix A, Figure A3, Supplementary Materials) although there was no evidence for a link between the morphological and functional differences.
The impact of higher heterogeneity when adapting to serum-free suspension culture is important. In respect to cell physiology, serum removal and anchorage-independent growth is a major change. It mimics an epithelial-to-mesenchymal transition [29,30] which is one of the most complex and multilayer-leveled processes that cells can undergo [31]. In fact, it is not by chance that principal component analysis identified 'serum-containing adherent vs. serum-free suspension' cultures as the first source of variability in the dataset, before cell bank repository origin or CAV-2 infection ( Figure 6A). During adaptation to serum-free suspension growth, substantial reduction in heterogeneity is expected since only a fraction of the initial cell population will endure the selection process. In this context, starting from higher heterogeneity will always be an advantage to survive in response to external variations and this is probably the reason for which ATCC cells adapted easier and faster. The existence of a plateau during which no variation occurred in the cumulative cell number for at least 10 days, supports that such transition and the corresponding selection process could be taking place in ATCC cells (Figure 2A,B). This 'selection plateau' was absent in ECACC cells, which were found to be extremely difficult to adapt and did not grow when directly transferred to new media ( Figure 2C). This finding is supported by Bissinger and colleagues, also reporting cell death when directly transferring MDCK.SUS2 (ECACC) to Xeno medium [32]. MDCK parental cells from ECACC seem to require complex culture strategies, such as the already described 10 weeks of serum-weaning plus 10 weeks adaptation in pendulum spinner flasks [19], and longer adaptation processes with stepwise medium variations ( Figure 2D). The transcriptome datasets generated in this study, can also be used to better understand the easier adaptation process of ATCC cells to serum-free suspension conditions. However, a direct comparison of 'ATCC vs. ECACC' with respect to the adherent-to-suspension transition should be conducted carefully. Because MDCK parental cells from ECACC could neither be adapted to grow in AEM nor in SFM4BHK21 from serum-containing conditions, we had to start from cells already growing in suspension in another serum-free formulation (MDCK.SUS2 [19]). This introduces an intermediate step that does not exist in the adaptation of ATCC cells and, for this reason, we ruled out exploring such analysis herein. Yet, such analysis is possible and can be guided by the proteomics comparison analysis of MDCK.SUS2 and its parental MDCK cell line reported by Kluge et al. [29].
Apart from the serum-free adaptation process, the main purpose of this study was to evaluate the impact of the cell bank origin of MDCK parental cells in CAV-2 production. Differences in productivity between the two parental cells were evident in serum-containing adherent cultures ( Figure 1B) and remained in serum-free suspension adapted cells ( Figure 4A). These differences seemed to occur with influenza A virus (IAV) production as well, although to a lesser extent ( Figure 11). The studies with influenza virus were conducted not only because of the relevance of MDCK in cell-based influenza vaccine production [33,34] but also to evaluate whether the differences in virus production performance between the two cell sources would be maintained. Nevertheless, IAV and CAV-2 are substantially different viruses and direct comparisons should be avoided. This is particularly relevant in the case of transcriptome data of CAV-2 infected cells that does not allow predicting changes in the transcriptome of IAV-infected cells.
The studies on monitoring CAV-2 infection progression ( Figure 5) pointed to increased restrictions in viral cycle infection progression in ATCC cells. Further understanding the determinants underlying such differences required a systems-level approach since the production of viruses and viral vectors in mammalian cells is a complex process involving many biological networks [35]. Hence, we used whole-genome transcriptome data and took a pathway analysis approach to identify biological pathways over-represented in the context of virus production. Within these pathways, the main differences between the two parental cells were investigated.
The first pathway analysis, comparing infected with non-infected cells, revealed a predominance of signaling pathways ( Figure 6C). This is a characteristic transcriptome signature of adenovirus infection [36], matching fast cellular resources hijacking typical of acute infections, and contrasts with the outcomes of previous analysis of viruses producing chronic infections, such as retroviruses, where metabolic pathways dominate [37]. Among the identified pathways, those related to 'cell stress and injury' and 'immune response' were the most noteworthy since these categories were likely to contain the biological players related to viral infection progression. Indeed, sub-pathways herein included 'pathogen influenced signaling', 'cytokine signaling' and 'apoptosis'. These sub-pathways were explored in-depth and gene expression patterns were mapped for the most relevant ones (Figures 7-10).
The control protein E1A (E1A) of adenoviruses has a strong pro-apoptotic effect and infected cells will inevitably undergo apoptosis. This is also the case for E1A of CAV-2 [36]. Although adenoviruses carry their own anti-apoptotic agents [38], including the control protein E1B which is a BCL2 apoptosis regulator homolog, host cells with naturally increased robustness to apoptosis can achieve higher titers by delaying death and enabling more productive use of intracellular resources for the generation and assembly of virions. Therefore, gene expression patterns associated with increased resistance to apoptosis in ECACC cells (Figure 7) are likely related to the higher productivities of these cells relative to ATCC's. Increased cell survival is in line with the up-regulation of the JAK/STAT pathway branch related to cell proliferation ( Figure 8) and the up-regulation of the NF-κB non-canonical pathway ( Figure 10). Whether the patterns of increased resistance to apoptosis synergize with or are a consequence of the up-regulation in these other two pathways is unclear.
In addition to increased resistance to apoptosis, the most relevant pathways for the differences in virus production between ECACC and ATCC cells seemed to be those relating to innate immunity systems. Gene expression patterns in these pathways pointed to a reduction in mechanisms of sensing and, most importantly, fighting viral infections. These included a down-regulation of genes in: (i) the JAK/STAT pathway branch related to interferon response (Figure 8), (ii) those involved in toll-like receptors signaling (Figure 9), and (iii) genes in the NF-κB inflammation and immune response canonical pathway ( Figure 10). All of these innate immunity-related systems (interferon response, toll-like receptors and NF-κB canonical pathway) provide key mechanisms in fighting adenoviral infections [39]. Interestingly, we found no obvious differences in other well-known innate immunity systems activated by adenoviral infection such as DNA cytosolic sensing mechanisms [39]. This may be related to an insufficient interferon-induced antiviral state reported in MDCK cells since most of these mechanisms depend, at least in part, on interferon signaling [40]. The evident down-regulation of these networks in ECACC cells, relative to ATCC's, is most likely responsible for the differences in CAV-2 productivity. Moreover, these networks and the identified genes are important target candidates, which, following functional validation, can be used for future genetic engineering approaches to improve the production of CAV-2 for gene therapy. This can be particularly useful in the production of helper-dependent vectors where bioprocess titers are still low.

Viruses
CAV-2pIX-GFP and CAVGFP are derived from the CAV-2 strain Toronto A 26/61, GenBank J04368. CAV-2pIX-GFP is a replication competent virus with a C-terminal fusion of eGFP in protein IX [41] and CAVGFP [4] is an E1-deleted (∆E1) vector containing an eGFP expression cassette. Viral vectors stocks were prepared and purified by CsCl gradients as described previously [4,7].

Adaptation to Serum-Free Media and Suspension and Growth Assays
Cells were adapted to grow in suspension with serum-free medium using a direct or a stepwise approach. Direct adaptations were performed by placing cells from adherent cultures in DMEM supplemented with FBS directly in final serum-free medium in shaker flasks. Cells from adherent cultures were pelleted by centrifugation (300× g, 10 min, 4 • C) and suspended in serum-free medium at 1 × 10 6 cells/mL. Culture medium was replaced every 2-3 days until cell growth became evident. Once cell concentration reached 2 × 10 6 cells/mL, 0.5 × 10 6 cells/mL were seeded in fresh medium. Cells were considered adapted when a maximum cell concentration of 2 × 10 6 cells/mL was attained in 3 days without medium exchange in at least 3 subsequent passages.
Stepwise adaptations were performed by first adapting cells to serum-free medium (SFM) in adherent cultures and then transfer these SFM-adapted cells to shaker flasks. Adaptation in adherent cultures was performed by gradually increasing the ratio of serum-free-to-serum-containing medium during 3 to 6 subsequent passages. SFM percentage in culture medium was increased when cells in adaptation presented similar growth of adherent cells in DMEM kept in parallel cultures. When cell growth became slower, percentage of SFM was maintained in two consecutive passages. Cells were considered adapted when 90% confluence was reached within 3-4 days in 3 consecutive passages in 100% SFM. Suspension cultures of SFM-adapted cells were maintained as mentioned above.
Growth of suspension cells was assayed using an inoculum of 0.5 × 10 6 cells/mL in 500 mL shaker flasks with a working volume of 150 mL. Growth was monitored for 10 days by determining cell concentration and viability. All parameters were determined at least in 24 h intervals.

Cell Permissiveness to CAV-2 Infection
To estimate the infection efficiency of CAV-2 in the different cells, cells were infected with CAVGFP using a multiplicity of infection (MOI) of 10 IP/cells with medium exchange at the time of infection. GFP-positive cells were determined at 24 h post infection through flow cytometric analysis (CyFlow Space, Partec, Münster, Germany). The infection assays in adherent cultures were performed in 6-well plates. For these, MDCK cells from ECACC and ATCC were seeded at 1.5 × 10 4 cells/cm 2 and 3 × 10 4 cells/cm 2 , respectively, to be infected the day after with a confluence of~80% in a working volume of 2 mL/well. Infection assays in suspension cultures were performed in 125 mL shaker flasks, seeding 5 × 10 5 cells/mL in 20-25 mL working volume. Cells were infected 24-36 h later, when cell concentration was at 0.8-1 × 10 6 cells/mL.

Progression of Infection and Production of CAV-2
The production of CAV-2pIX-GFP was used to evaluate virus productivity. The expression of capsid protein IX after CAV-2pIX-GFP infection was used to also evaluate the progression of CAV-2 infection by flow cytometry (CyFlow Space, Partec, Münster, Germany). Adherent and suspension cultures for these assays were prepared as described above. Adherent cultures were infected at MOI 5, while suspension cultures at MOI 1. Sampling and cell monitoring were performed at 24 h intervals. At each time point, parallel non-infected cultures of each of the corresponding cells were used as gating controls. Viruses were collected by disrupting cells with lysis buffer (Tris/HCl 10 mM, pH 8.0 and 0.1% (v/v) Triton-X 100). The resulting sample was clarified at 3000× g for 10 min at 4 • C and stored at −85 • C until further analysis.

Production of Influenza Virus
Suspension MDCK cells cultivated in AEM were grown to a cell concentration of 2.5 × 10 6 cells/mL in 150 mL vented shake flasks (75 mL working volume). For infection, cells were first pelleted (150× g, 5 min) and resuspended in a mixture of 95% (v/v) fresh and 5% (v/v) conditioned medium containing the virus seed at MOI 10 −5 virions/cell. To obtain optimal infection conditions, trypsin concentration had to be adjusted for each cell. For MDCK ATCC and ECACC cells, trypsin was added at 5 × 10 −6 units/cell and 2.5 × 10 −6 units/cell, respectively. Shake flasks were sampled twice per day for cell counting. Supernatant was clarified by centrifugation (300× g, 10 min) and stored at −85 • C until further analysis.

Cell Concentration and Size
Cells were counted using a Fuchs-Rosenthal hemocytometer chamber and viability determined by the trypan blue exclusion method in the majority of the assays. For influenza virus experiments, trypan blue exclusion was automated using a ViCELL™XR device (Beckman Coulter, Indianapolis, IN, USA). In adherent cultures, cells were counted directly from cell suspension after trypsin detachment. For suspension cell lines growing in aggregates, trypsin was used to obtain single cell suspensions. Briefly, from 1 mL sample, cells were centrifuged for 10 min at 300× g and the supernatant removed (900 µL). Then, same volume of trypsin was added, cells were incubated at 37 • C until evident detachment, suspended by pipetting up and down and finally counted. Suspension cultures with single cells were counted directly from the sample taken from shaker flasks.
Cell diameter and volume was determined using a CASY ® Cell Counter (Schärfe Systems, Reutlingen, Germany). The appropriate program was established according to the instructions of the manufacturer. For measurement, 50-100 µL of cell suspension were transferred to a CASY ® cup containing 10 mL CASY ® ton, mixed by inverting three times and placed in the CASY ® Cell Counter.

CAV-2 Quantification
Quantification of infectious CAV-2 particles (IP) was performed by monitoring the expression of GFP by flow cytometric analysis of DK cells subjected to serial dilutions of viral samples, as described previously [7]. Quantification of genome-containing CAV-2 particles (VG) was performed by quantitative real-time PCR, as described previously [42]. Briefly, viral genomes were extracted and purified by High Pure Viral Nucleic Acid Kit (Roche Diagnostics, Penzberg, Germany), and SYBR Green I dye chemistry was used to detect PCR products using LightCycler system. Primers against the GFP gene were used: forward 5 -CAGAAGAACGGCATCAAGGT-3 and reverse 5 -CTGGGTGCTCAGGTAGTGG-3 .

Influenza Virus Titer Determinations
Determination of total influenza virus concentration was performed with a hemagglutination assay as described before [43]. Read-out is log 10 HAU/100 µL with a relative standard deviation of the method of 9.3% and a maximum error of ±0.2 log 10 HAU/100 µL. This can be converted into virions/mL by: Total virus concentration = 2 × 10 7 × 10 log 10 HAU/100 µL (1) Infectious virus particle concentration was determined with a TCID 50 assay as described by Genzel and Reichl [44]. Titer calculations were performed according to the Spearman-Kärber method. The limit of the detection was 3.2 × 10 2 virions/mL and dilution error was ±0.3 log 10 . For calculations of cell-specific virus yields the maximum virus titers were divided by the maximum cell concentration.

RNA Extraction for Transcriptome Analysis
Total RNA for transcriptome analysis was extracted at 33 h post-infection with a replicative-competent CAV-2. RNeasy Mini Kit (Qiagen, Valencia, CA, USA) for total RNA extractions was used according to the instructions of the manufacturer. Total RNA pellet was eluted in 100 µL of nuclease-free water (Qiagen, Valencia, CA, USA) and stored at −85 • C until further processing. RNA yields were quantified using NanoDrop 2000 (Thermo Scientific, Waltham, MA, USA) and RNA quality was characterized by the quotient of the 28S to 18S ribosomal RNA electropherogram peak using an Agilent 2100 bioanalyzer and the RNA Nano Chip (Agilent, Santa Clara, CA, USA).

Transcriptome Data Processing and Analysis
Data pre-processing and analysis was done in Affymetrix ® Expression Console™ Software using Robust Multi-array Average (RMA)-Sketch at gene level normalization. Differentially expressed genes were considered based on 2-fold expression difference for each comparison analysis considered and evaluated for pathway analysis using Ingenuity Pathway Analysis (IPA) software (Ingenuity ® Systems, www.ingenuity.com). Normalized data was also used for principal component analysis using R software [45]. The entire microarray dataset was submitted to ArrayExpress of EMBL-European Bioinformatics Institute with the accession number E-MTAB-9379.

Statistical Analysis
Statistical analysis to compare ATCC and ECACC cells in respect to infectious viral titers was carried out using the Welch's t-test.

Conclusions
This work shows that MDCK parental cells from ECCAC are a less heterogeneous population, most likely, a sub-population of the parental cells deposited at ATCC. While this sub-population is less prone to adapt to suspension and serum-free culture conditions, it was found to be more permissive to virus replication progression. Transcriptome data indicates that this increase in permissiveness is due to a general down-regulation of biological networks of innate immunity in ECCAC cells, such as apoptosis and death receptor signaling, JAK/STAT signaling, toll-like receptors signaling and NF-κB signaling. (Instituto Gulbenkian de Ciência, Portugal) for technical and scientific support on Affymetrix Canine Gene microarray experiments.

Conflicts of Interest:
The authors declare no conflict of interest.  Figure A1. Morphology of MDCK cells from ATCC after adaptation to suspension. Cells were adapted to one of two serum-free culture media-AEM or SFM4BHK21-following a direct approach (A,C) or a stepwise approach (B,D). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 μm. cells previously adapted to suspension and growing as aggregates in AEM [9], were used as ECACC supplier representative. Direct adaptation of MDCK.SUS2 to SFM4BHK21 medium did not enable cell growth, only the stepwise approach ensured cell survival and resulted in single cell cultures (B). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 μm. Figure A1. Morphology of MDCK cells from ATCC after adaptation to suspension. Cells were adapted to one of two serum-free culture media-AEM or SFM4BHK21-following a direct approach (A,C) or a stepwise approach (B,D). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 µm. Appendix A Figure A1. Morphology of MDCK cells from ATCC after adaptation to suspension. Cells were adapted to one of two serum-free culture media-AEM or SFM4BHK21-following a direct approach (A,C) or a stepwise approach (B,D). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 μm. cells previously adapted to suspension and growing as aggregates in AEM [9], were used as ECACC supplier representative. Direct adaptation of MDCK.SUS2 to SFM4BHK21 medium did not enable cell growth, only the stepwise approach ensured cell survival and resulted in single cell cultures (B). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 μm. cells previously adapted to suspension and growing as aggregates in AEM [9], were used as ECACC supplier representative. Direct adaptation of MDCK.SUS2 to SFM4BHK21 medium did not enable cell growth, only the stepwise approach ensured cell survival and resulted in single cell cultures (B). Cells were analyzed and digitally visualized as described in Figure 1. Scale bar-100 µm.    Non-infected cells (left panels) were used as negative controls for gating and up to 2% false positives in these control samples were considered acceptable. Data corresponds to representative histograms from the 24 h post-infection time-point (refer to Figure 5A from the main manuscript). This time point was chosen as benchmark since fluorescence intensity values are comparable between samples, while later timepoints may be affected, to more or less extent, by superinfection events. Suspension cultures correspond to cells growing in AEM medium while adherent conditions correspond to cells growing in static monolayer in DMEM with 10% FBS. RFU: relative fluorescence units.