High level recombinant antibody production in Chinese hamster ovary (CHO) cells and characterisation of the carcinoembryonic antigen (CEA) specific human full-size IgG1 H10

摘要

Monoclonal antibodies (mAbs) are constantly gaining importance as a major class of recombinant protein therapeutics and diagnostics. Especially the field of cancer immunotherapy expanded remarkably within the last years. Nowadays, the majority of therapeutic mAbs is produced in CHO cells, which represent the current gold standard for the production of recombinant glycoproteins in the biopharmaceutical industry. To meet the market demand, it is necessary to generate production processes capable of delivering sufficient quantities of product. Thereby not only an optimised production strategy, but also the availability of high producing cell lines is essential. In the frame of this work a strategy for high level IgG production in CHO suspension cells was established based on new vector designs and improved screening and cultivation procedures. The objective was to produce the human, full-size IgG1 H10 in the dihydrofolate reductase (DHFR) deficient, suspension CHO cell line DG44. For this purpose two novel, polycistronic CHO expression vectors were designed: The tricistronic vector pAPI:H10 and the bicistronic tandem vector pAPT:H10. Both vectors allowed tightly linked antibody heavy chain (HC) and light chain (LC) expression with a ratio suitable for high level antibody production. In addition, efficient MTX-induced gene amplification was performed, resulting in a 410- to 480-fold IgG yield improvement. During a comparative study of both vector systems, the pAPT:H10 vector turned out to allow higher IgG yield production due to a more favourable antibody HC:LC ratio and a lower MTX-mediated cell growth inhibition. Further optimisation steps like the testing of different, commercially available CHO cell culture media, long-term cultivation of cell cultures and the comparison of different cultivation vessels for orbital shaking cultivation conditions promised significant yield improvement and suspension cell culture scale up. Furthermore, two commonly used MTX-based gene amplification strategies had been tested. In this context, serum-independent, high throughput automated limiting dilution with single cell regeneration documentation was performed. In addition, a comparative high throughput cultivation and gene amplification in 50 ml bioreactor tubes was carried out, allowing the parallel screening of a large number of clones under process-orientated conditions. Thereby H10 yields of to 304 μg/ml H10 (15 pg/cell/day) were achieved. Eventually the best producing, gene amplified DG44 culture, transfected with the vector pAPT:H10, achieved production levels of up to 554 μg/ml H10 (25 pg/cell/day) in a normal batch culture. Furthermore, yields of up to 312 μg/ml H10 (23 pg/cell/day) were documented here with a monoclonal, pAPI:H10 transfected, high producer. To my knowledge this is the highest antibody level ever reported with a tricistronic vector system. In the second part of this thesis, the human anti-CEA H10 antibody was analysed regarding its functional binding to the carcinoembryonic antigen (CEA) and its efficiency to induce ADCC-mediated cell target killing. CEA is a well characterised tumor marker and overexpressed in a variety of tumors, like human coloncarcinoma. The use of an anti-CEA antibody in the context of cancer immunotherapy therefore represents a promising approach to specifically localise and destroy CEA-positive cancer cells. To test the suitability of H10 as a potential therapeutic antibody it was produced in DG44 cells with two different N-glycan structures: A complex CHO glycoform (H10_CHO_WT) and a glycoengineered version bearing a bisected N-glycan (H10_CHO_GnTIII). The latter glycoform is known to induce ADCC more efficiently than the normal complex CHO glycoform. The bisected N-glycan structure was generated by the coexpression of Golgi localized β1-4-N-acetylglucosaminyltransferase III (GnTIII) in an H10_CHO_WT secreting DG44 cell line. Thereby, large amounts of up to 70% bisected N-glycan structures were achieved. Subsequent functional analysis of the H10 glycovariants showed similar activities regarding CEA antigen binding on the cell surface of CEA-positive cells, like LS174T or the recombinant HEK293T-CEA_16 cell line, which has been improved by FACS-based sorting within this work. There was no indication of cross-reactivity to the CEA related non-specific cross-reacting antigen NCA, which is often observed with antibodies recognising the N-domain of CEA, like the H10 antibody. However, cross-reactivity to intracellular, non-identified proteins in the cell extract of CEA-expressing and non-CEA-expressing human cells was found, whose binding could not be observed on intact cells. Two strategies were established to detect human CEA in human tissue via the H10 antibody without cross-reacting to human, endogenous tissue immunglobulins. Thereby, H10 was shown to specifically bind CEA and higher CEA levels were observed in coloncarcinoma sections compared to tissue of normal colon mucosa. Finally, H10 was shown to have the potential of killing target cells via ADCC, whereas the efficiency was dependent on the CEA antigen density on the cell surface. Even though high levels of bisected N-glycan structures could be achieved by glycoengineering of H10, SPR-analysis of both H10 glycoversions showed clearly that this glycomodification has no improving effect on the FcγIIIa receptor binding affinity. This indicates that no ADCC induction increase can be accomplished by the insertion of bisected N-glycans in CHO produced IgGs.