High level recombinant protein expression in mammalian cells not only relies on potent transcription promoting elements and optimal design of the expression cassette but also crucially depends on appropriate chromosomal sites that support the incoming expression cassette upon integration. Sophisticated chromosomal engineering approaches based on site specific recombinases allow the precise/controlled integration of expression cassettes into tagged chromosomal sites (reviewed in [17
]). A first-generation targeting system with a single recombinase target site has recently been evaluated for protein production. Comparable (consistent) antibody expression levels could be achieved upon targeting at defined integration sites in CHO cells using Cre-[33
] or Flp-mediated integration [34
]. The potential of this technology was further exploited for the production of a human polyclonal anti-RhD antibody [24
] by integrating 25 individual antibody expression cassettes into a defined FRT tagged integration site in CHO cells (Flp-In™ cell line; Invitrogen). Irrespective of this success first generation targeting systems are limited by fact that the tagging sequences cannot be eliminated. Usually, complete vectors including bacterial sequences are co-integrated which have been shown to decrease transgene expression from neighboring promoters (e.g. [35
]). Also, since excision of the targeted cassette is favored, the targeting efficiency can be unsatisfactory. In this respect, exchange of tagging cassettes via RMCE seems to be the method of choice. While proof of principle has been given for production of retroviral vectors [20
] so far, a systematic approach evaluating RMCE for protein production has been missing.
This report describes a comparative approach to identify potent integration sites that support stable protein production and to exploit these sites by Flp RMCE to target expression cassettes of choice. For this purpose we used CHO-K1 and HEK293 cells - the most relevant cell lines for protein production. Motivated by the notion that any tagging procedure has an intrinsic bias for specific patterns/types of integration sites we employed random tagging strategies based on retroviral or plasmidic transfer of screening vectors. For monitoring expression we employed eGFP (either in a single expression unit or in a fusion to a selection marker) or an IgG molecule. With these screening approaches we established a set of tagged master cell lines that stably express the respective reporter gene(s). Importantly, both plasmidic and retroviral tagging proved to be appropriate for RMCE since they lead to a high-percentage of single copy integrants ranging from 30 to > 90% - a prerequisite for targeted integration.
Evaluation of the performance of cassette exchange in the different master cell clones derived from HEK293 and CHO-K1 cells using various targeting vectors proved to be highly efficient with ≥85% correctly targeted daughter clones. In certain applications, this high efficiency might overcome the need for subsequent sub-cloning. Further, analysis of the production levels could confirm that site directed integration significantly reduces the variations of clonal expression levels. This is expected from isogenic clones and is in accordance to previous reports evaluating this method for production of retroviral vectors [20
Further, we investigated the flexibility of the tagged integration sites with respect to supporting other promoters. Unexpectedly, in the master clones HP-GFP and HRV-GFP targeting of SV40 based antibody cassettes failed to provide significant levels of antibody expression (Figure ). In contrast, high level antibody expression was obtained in these integration sites upon targeting either an MPSV/CMV chimeric promoter or a bidirectional composite promoter, resulting in 3.38 and 1.69 pcd, respectively. Interestingly, the opposite situation was observed in CHO-K1 cell clones: while targeting of the SV40 promoter cassettes resulted in high level antibody expression in the range of 2.5-4.4 pcd, performance of promoters such as the composite MPSV/CMV and bidirectional AdMLP/pEF promoters was significantly impaired (Figure ).
Several reports and also results from our lab (e.g. [37
]; data not shown) give evidence that the SV40 promoter is a potent, although not the most favourable promoter in both CHO and HEK293 cells. This rules out that this differential performance obtained upon targeting is a consequence of a diverse set of transcription factors differentially supporting in these two cell lines. Rather, it seems to be that the nature/composition of the initial tagging vector would define the capacity of the integration site with respect to supporting promoters: the master cell clones HP-GFP and HRV-GFP which were incompatible with SV40 based targeting vectors were tagged and screened for high level GFP expression from a PGK promoter and an MSCV promoter, respectively. In contrast, the master cell clones CP-HTG and CP-ABHTG which showed high level of SV40 based expression upon targeting were initially screened to support an SV40 promoter driven tagging vector. Together, these data indicate that specific promoters show preferential performances in certain integration sites.
Our data from the screening for high expression clones show that chromosomal sites that support high level expression can be identified with both promoters. However, it seems that the nature of the integration site specifically defines the final strength of a given promoter. This interpretation is not immediately compatible with the general believe that certain promoters are particularly strong in certain cell lines. It is important to note that the data that led to this conclusion are derived from transient expression experiments or from experiments in which pools of transfectants were analyzed. For transient expression the composition of soluble (transcription) factors might indeed constitute the dominant level of promoter strength [38
]. However, upon stable integration into the host genome the influence of the surrounding chromatin might be dominant over the influence of the soluble factors, given that the composition of the promoter allows expression at all.
This suggests that the strength/potential of a specific integration site is linked to a certain promoter - and is not necessarily supporting any integrated expression cassette. For application of the tag-and targeting approach it indicates that the molecular composition of targeting vectors and chromosomal integration site go hand-in-hand. Thus, it will be important to consider the specific requirements of a particular integration site relating to the maximum level of recombinant protein production that can be achieved.
Various types of chromosomal elements have been identified that contribute and modulate individual expression cassettes upon integration [2
]. In the last years increasing evidence has been provided showing that not only specific genetic elements but also complex epigenetic mechanisms can be involved. We employed RMCE to test if expression from a weak integration site can be increased by chromosomal engineering of the integration site. However, neither the cHS4 element nor a potent S/MAR could significantly (more than 2 fold) increase the level of expression (data not shown). This indicates that the mere integration of chromosomal elements into specific loci is not of benefit per se
but would require certain prerequisites. At the same time this gives evidence that our knowledge about the chromosomal elements and their influence on transgene expression is still rudimental. This might explain why the rational construction of synthetic expression domains providing per se
all the needs for position independent and high expression is not straight forward and is still in its infancy. With the technologies now available for targeting transgenes to pre-defined loci our understanding of mechanism governing the crosstalk of chromosomal elements should be broadened.