Discussion

The production of viral vectors for gene therapy, such as recombinant AAV, is complex and usually demands coordinated expression of multiple genes within a cell to produce a packaged and functional virus (Srivastava et al., 2021). Typically, multiple plasmids and/or adenovirus infection (mammalian systems), or various baculoviruses (insect cell system) are needed, resulting in significant levels of defective product (e.g., empty or non-infectious particles) and other process-related contaminants (e.g., helper virus) (reviewed in Merten, 2016; Penaud-Budloo et al., 2018). Insect cells with the dual baculovirus expression vector system, herein used, have shown high titers for different AAV serotypes (Cecchini et al., 2011; Pais et al., 2019; Smith et al., 2009), although implications for final product quality remain a concern (e.g. genome packaging, ratio of capsid proteins and their correct folding). The host cell transcriptional response to virus infection, as well as the heterogeneity reported in clonal CHO cells (Tzani et al., 2021) and virus populations (Sun et al., 2020), could suggest that such production processes are highly heterogeneous and can potentially impact product titers and/or quality. Thus, to assess the characteristics of the dual baculovirus system, as well as how Sf9 cells responds to baculovirus infection by transcriptome alterations, single-cell RNA-sequencing was employed.
In this study, we observed heterogeneity in non-infected Sf9 insect cells, similar to what has been reported in clonally derived cell lines (Tzani et al., 2021). While this highlights the individuality of cells within the population, the most dominant influence observed herein was associated to cell cycle, with cells identified in G2/M phase showing distinct clustering compared to others, as reported elsewhere (Tirosh et al., 2016; Tzani et al., 2021). Moreover, baculovirus infection has been hypothesized to arrest cells in a “pseudo” S phase (Rohrmann, 2019), which we could confirm in our study, i.e. we observed an increase in S phase association to infected cells. The high association to G1 phase in late infection samples was however unexpected. Upon further evaluation of the cell cycle scoring, it was observed that cell cycle phase association was biased, as high baculovirus gene expression at later infection stages masks cell cycle genes. In this scenario, cells cannot be associated to either S or G2M phases and thus are attributed to G1 by default. Despite being useful to decipher heterogeneity in non-infected samples, cell cycle scoring was not considered as a correct approach to describe heterogeneity in samples in which the host cell transcriptome is overwhelmed by foreign virus replication. An additional source of variation observed in a small sub-population of non-infected cells was correlated to the activation of stress response mechanism. While more stringent quality control parameters might exclude this cluster, this would not be possible in this system as the impact of baculovirus on the transcriptome limits the regression of some quality control parameters (i.e., number of detected genes).
Baculovirus infection has been shown to follow a random Poisson distribution (Palomares et al., 2002). However, in a dual baculovirus system this tends to be more complex, as interference and synergistic effects of both viruses can be observed (Mena et al., 2007), as well as differences in virus replication and/or infection kinetics could occur (Galibert et al., 2021). Indeed, along infection cells had highergfp expression when compared to rep and/or cap , highlighting possible differences in infection kinetics between both rBACs. While this could arise from the promoters regulating the expression of each transgene (cmv is an earlier promoter than the later viral promoters polh/p10 ), other possible sources of asynchronous infection and replication of both baculoviruses include titer determination and random variations in infection kinetics, emphasizing a potential need for customized infection strategies.
The heterogeneity of infected cells was linked to the overexpression of baculovirus genes, infection progression and host cell transcriptome responses. However, it was also probably augmented because of the low MOI infection strategy employed here. While high MOI processes are less desirable due to challenges associated with generation of master virus stocks (e.g., larger production scale are needed), these could prove more useful if a synchronous infection between two baculoviruses is desired. Another limitation of low MOI, dual-baculovirus based processes is the fact that in case one of these baculoviruses infects and replicates faster than the other, cells infected with the more replicative virus might be unable to receive the other, in a process called super-infection exclusion (i.e., previously infected cells cannot be re-infected) (Beperet et al., 2014; Folimonova, 2012). The possibility of baculovirus reinfection has however been reported (Gotoh et al., 2008; Mena et al., 2007), nevertheless it is still unknown how long infected cells are susceptible to new viruses entry and how efficient is the expression of the newly arrived transgene (Sokolenko et al., 2012). A recent report showed the limitations of re-infection, as a maximum of 40% of cells were found infected with both rBACs in a dual baculovirus system using similar conditions (Galibert et al., 2021), highlighting the need to assess possibilities to increase this number to improve product quality and titer.
The production of fully packed AAV particles is dependent on the presence of both recombinant baculoviruses carrying their respective transgenes in the same cell. In our study, at 24 hpi, although all cells are infected, only 29.4% of cells were shown to have all the necessary transgenes expressed to produce packaged AAV particles. While this does not necessarily correlate to subsequent protein expression levels, as AAV proteins have been shown to undergo post-transcriptional and translational regulation (Virag et al., 2009), this data raises the possibility of a potential production bottleneck. Nevertheless, the number of cells showing transcriptional capacity for producing packed AAVs could have been underestimated here, since both rep andcap transgenes are expressed using late viral promoters and thus expression of these genes in cells that have been infected in the second infection cycle (occurring between 18 and 24 hpi) might not yet be detectable at 24 hpi.
Baculovirus infection has been shown to significantly impact the host cell machinery, activating stress response, cell cycle arrest and reorganization of the cytoskeleton and cell nucleus, while shutting down cell growth and protein folding capacities among others (as reviewed in Monteiro et al., 2012 and Rohrmann, 2013). Here, similar biological processes to those previously reported were shown to vary along infection, such as stress response mechanisms (e.g., heat shock protein 68 ) (Chen et al., 2014; Koczka et al., 2018). This might arise due to the response to unfolded proteins, as folding capacity has been found impacted (Koczka et al., 2018) and could result in reduced product quality along infection. Additionally, energy metabolism alterations along infection have been reported (Bernal et al., 2009; Bernal et al., 2010; Monteiro et al., 2017) and also be associated with mitochondrial function (Chen et al., 2014; Xue et al., 2012), which were also found altered along infection here. Alteration of the glutamate dehydrogenase and glutamate synthase genes predicted to encode proteins involved in the ammonia recycling system (Bernal et al., 2009; Doverskog et al., 2000), identified herein, has been shown in bulk analysis (Virgolini et al., 2022) and further confirms the impact of infection on host cell machinery.
Overall, 75% of enriched gene ontology terms identified herein were also found in our previous bulk transcriptomics analysis (Virgolini et al., 2022). Nonetheless, the majority of the GO biological processes identified in previous bulk RNA-seq data were only found at later stages of infection, suggesting that transcriptomic alterations of a sub-population of cells might be masked in this analysis. This further underlines the added benefit of single-cell analysis in heterogeneous production systems, as it is able to dissect early transcriptional alterations due to stress and/or infection in a sub-population of cells, thus having the potential to predict subsequent population response.