Chinese hamster ovary (CHO) cells are the most commonly used expression host of choice for manufacturing biopharmaceuticals. Development of recombinant CHO (rCHO) cell lines is critical for the biopharmaceutical industry for the production of recombinant therapeutic proteins (Walsh, 2018). Since the first use of CHO cells, stable cell line development (CLD) has been performed by untargeted random integration of transgenes followed by a selection and large-scale clone screening process. Although this traditional CLD method is easy to perform, the issue of clonal variation, particularly in transgene expression patterns, arises because of uncontrolled transgene integration sites, copy number variation, and transgene rearrangement upon random integration (Lee et al., 2018; Lee et al., 2019).
Site-specific integration of transgenes into designated genomic sites has been proposed as an alternative CLD method, as it can mitigate the high levels of heterogeneity seen in clonal rCHO cell lines, thus reducing CLD timelines (Lee et al., 2015b; Lee et al., 2019). The advent of programmable nuclease-mediated genome editing technology has accelerated targeted modification of the CHO genome by inducing a site-specific DNA double-strand break (DSB), which results in the activation of DNA damage repair pathways. Among various genome editing tools, CRISPR/Cas9 technology, which is based on the RNA-guided engineered nuclease, has been rapidly adopted in the CHO community due to its simple composition and high targeting efficiency (Lee et al., 2015a). Notably, CRISPR/Cas9 was successfully implemented to insert transgenes in the target CHO genomic sites in a precise manner with homology-directed repair (HDR) upon the creation of DSBs (Lee et al., 2015b; Lee et al., 2016).
The targeted integration (TI)-based CLD methods are proven to be effective in both academia and industry as they achieve reproducible transgene expression levels and stability (Grav et al., 2018; Zhang et al., 2015). Despite its many advantages, the low targeting efficiency of the current TI technology hinders multiplexed gene knock-ins (KIs), as well as high-throughput screening for identifying genomic hot spots yielding high and stable transgene expression levels (Lee et al., 2019). Given the major clinical and commercial success of monoclonal antibody therapeutics and the emergence of next-generation difficult-to-express antibody products, including multi-specific antibody formats and fusion proteins (Spiess et al., 2015; Walsh, 2018), the demand for multiple KI based CLD will continue to grow because efficient expression of antibodies requires not only the balanced expression of heavy and light chains but also the co-expression of various engineering targets (Ho et al., 2013; Pybus et al., 2014). Selection-based strategy has previously been shown to target several single (Lee et al., 2015b) and multiple loci (Gaidukov et al., 2018) in CHO cells; however, sequential integration along with the use of different selection markers is not ideal for an industrial setting and is also time-consuming. We previously demonstrated a selection marker-free strategy to improve TI through fluorescent enrichment of cells with HDR-mediated genome editing; however, its targeting efficiency was low when challenged with integrating large transgenes (Lee et al., 2016).
In this study, we describe a simple optimized strategy that allows efficient CRISPR/Cas9- mediated TI of transgenes in a transient setup in CHO-K1 cells. To measure targeting efficiency, we used a promoter-trap based KI monitoring CHO-K1 cell line in which an enhanced green fluorescent protein (EGFP) expression cassette without a promoter region was integrated at the specific genomic site (Figure 1a). Co-transfection of a Cas9/single guide RNA (sgRNA) expression vector and donor plasmid, allowed TI of the promoter into the upstream of EGFP coding sequences, and this then restored EGFP expression (Figure 1a). Because the EGFP+ cells represent TI of the promoter, the percentage of HDR events can be quantitatively measured by flow cytometry. To select an efficient sgRNA target site, we designed two sgRNAs, and used sgRNA1 which showed higher editing efficiency that was detected by the T7E1 assay (Figure S1).  
To test and optimize TI with the monitoring system, we determined KI efficiency of three exogenous constitutive promoters, including short variant of CMV (148 bp), human EF1α (1167 bp) and Chinese hamster EF1α promoter (1665 bp), 3 and 6 days after electroporation. (Figure 1b). The EGFP+ populations were visible 48 hours after transfection regardless of the promoter lengths. However, the percentage of EGFP+ populations was marginal when transfected with the Cas9 vector and individual donor plasmid. After subculturing on day 6, the EGFP+ populations were not detectable without sgRNA expression; however, they were noticeable with sgRNA expression (Figure 1b). Both electroporation and lipofection achieved detectable TI, although higher overall KI efficiency was seen using electroporation (Figure S2). Therefore, unless otherwise stated, we proceeded to use electroporation in the subsequent experiments.
The mean fluorescence intensity (MFI) of the EGFP+ populations showed different strengths of three exogenous promoters among which the EF1α promoters exhibited higher EGFP expression levels (Figure 1c). This result is consistent with the previous study where EGFP expression levels were compared in targeted integrants (Lee et al., 2018). Genomic DNA based 5′ Junction PCR analysis of the EGFP+ sorting pool further verified TI (Figure 1d). In addition, we successfully acquired EGFP expressing clonal cell lines upon limiting dilution of transfected pools of cells (Supplementary Table S1).  The above results demonstrated that the monitoring system enables the detection of true positive KI events in a transient system.
With the promoter-trap based monitoring cell line, TI of three promoters was achieved in a transient setup. The KI efficiencies were comparable among promoters from 5 to 8% (Figure 1b). However, the efficiency was lower than that seen from the selection-based strategy (Lee et al., 2015b). Given the expectation that the efficiency should be lower when applying multiplex gene KIs or inserting longer transgenes, we modified the HDR donor vector design to enhance transient KI efficiency. Compared with a conventional circular donor (CCD), the modified donor, termed double cut donor (DCD), contained two sgRNA recognition sequences, which were identical to the genomic site targeting sgRNA/PAM sequence and flanked homology arms in the donor vector (Figure 2a). By simply adding sgRNA recognition sequences, genomic DSBs and in vivo linearization of target promoter sequences flanked by homologous sequences can be synchronized. In vivo cleavable donors were used to insert large gene of interests (GOIs) in a homology-independent manner (Cristea et al., 2013; He et al., 2016). However, non-homologous end-joining based KI approaches often cause incomplete integration with mutagenic junctions and the insertion of backbone fragments; thus, HDR-based KI approaches incorporating cleavable donors have been developed (Yao et al., 2017; Zhang et al., 2017). Although previous studies have shown that the DCD functions efficiently in human cells (Zhang et al., 2017), its effect on transient KI efficiency remains unknown in CHO cells. To address this, we transfected Cas9/sgRNA1 expression vectors with either CCD or DCD of the three different promoters in the CHO-K1 monitoring cell line. The use of DCDs resulted in increased KI efficiencies of up to about 13-18%, which were seen 6 days after transfection, in the case of adherent cell line (Figure 2b). It corresponded to a 2.9–4.5 fold increase in the KI populations compared with CCDs. Interestingly, the effect of DCD on KI efficiency was more pronounced in the case of serum-free adapted suspension monitoring CHO cell line than that seen in the serum adherent monitoring cell line, corresponding to a 3.9-10.1 fold increase in the KI populations compared with CCDs (Figure 2c).
To maximize the targeting efficiency, we determined an optimal ratio of the DCD to Cas9/sgRNA expression vector. To test this, KI efficiencies were assessed in cells transfected with five different ratios of constructs (Figure S3). We found that the use of approximately equal amounts of two constructs, applied in Figure 2b and 2c, led to the highest KI efficiency, while increasing the level of either Cas9/sgRNA expression vector or DCD had negative effects on KI efficiency.
Based on the significant increase in KI efficiencies with DCDs, we postulated whether or not multiplexed gene KIs could be achieved in CHO cells without the need of an additional enriching process. To monitor and quantify double KI events, we established a double-KI monitoring CHO-K1 cell line by inserting a promoter-less TagRFP657 expression cassette into another locus, C1GALT1C1 (COSMC), in the parental single KI monitoring cell line (Figure S4a). A sequence from the mouse Rosa26 locus (Chu et al., 2015) was included upstream of the TagRFP657 coding sequence as an sgRNA target site. 5’/3’ Junction PCR and out-out PCR was conducted to evaluate the correct integration at the COSMC locus (Figure S4b). We then evaluated the impact of DCDs to enable double TI by introducing two Cas9/sgRNA expression vectors and two human EF1α promoter donor plasmids harboring 650 – 800 bp homology arms (Figure 3a). This homology arm length corresponds to the truncated TagRFP657 and EGFP coding sequences, as well as the range of the highest HDR with double cut donor, which leads to the transgene integration mediated by almost all complete HDRs (Zhang et al., 2017). In parallel, we tested CCDs for double TI. For the generation of double KI integrants, we rationally designed two strategies, simultaneous and sequential KI (Figure 3b). In the simultaneous KI strategy, all vector constructs needed for the TI were simultaneously introduced by a one-step transfection. Although simultaneous KI strategy is simple and fast, efficiency can be reduced as a result of the maximum quantity of vector required for the transfection being divided at individual sites. In contrast, the sequential KI strategy targets two loci one by one through two rounds of sequential transfection. In the sequential KI strategy, second transfection was conducted at passage 2 (time interval day 6). Regardless of the target site order, either site 1 followed by site 2 or site 2 followed by site 1, both EGFP+/TagRFP657+ populations were detected in a range of approximately 1–2% when using DCDs on day 12, in the adherent monitoring cell line (Figure 3c). Conversely, when using CCDs, EGFP+/TagRFP657+ populations were rarely detected, demonstrating the limitation of the conventional KI method. The KI efficiency at the COSMC locus was lower than that at the non-coding region, which highlights how genomic location affects KI efficiency. In the simultaneous KI strategy, double KI events were also distinctly detected only when using DCDs (Figure 3d). Although KI efficiency at individual sites was lower than that of sequential KI in the adherent monitoring cell line, the level of double-KI integrants was comparable at 1–2%. In the serum-free adapted suspension double-KI monitoring CHO-K1 cell line, the use of DCDs led to significant increase in KI efficiency, 7.3 and 36 fold increase in EGFP+ and TagRFP657+ populations respectively, compared with CCDs, which results in detection of double-targeted integrants about 2-4% on day 6 (Figure 3d).
In summary, we provide an efficient CRISPR/Cas9 mediated KI strategy that can be successfully applied to integrate single or double GOIs into target genomic sites in CHO cells. The promoter-trap based single or double KI monitoring CHO-K1 cell lines were established to detect and quantify KI events in a simple manner. Depending on the culture mode, making a change in donor DNA template design by adding sgRNA recognition sequences increased targeting efficiency by 2.9–10.1 fold in the non-coding region and 13.5-36 fold at the COSMC locus (Figure 2 and 3). This then led to simultaneous double KIs in a transient set up. The simple modification of donor plasmid can be more general than changing the widely used Cas9/sgRNA vector; thus it will be easily applicable to existing CRISPR/Cas9 mediated TI for generating rCHO cell lines. In this way, we generated rCHO cell lines by TI without the need of selection and enriching processes and were also able to avoid the sequential CLD process for multiple KIs. Ultimately, this platform can significantly reduce CLD timelines and save efforts to engineer CHO cell lines.