Transcriptomic profiling as a tool to characterize CHO cell processes

Global healthcare demands for affordable and accessible biological medicines are rapidly increasing as a result of an increasing number of patients. Biosimilars products have a potential to help decrease the price and increase the availability of the biological medicines. Monoclonal antibodies are a special class of biopharmaceutical proteins that are used to treat life-threatening diseases like cancer. Mammalian cells and more specific Chinese Hamster Ovary cells are the main production platform for these monoclonal antibodies. Many of the first products, which were developed in the early 1990s are going off patent opening the possibility to develop biosimilars. Nowadays, for the development of biosimilars and also completely new biopharmaceuticals it is required for pharmaceutical companies to include (aspects of) Quality by Design (QbD). This means that they need to have a good scientific understanding of the relation between critical process parameters (CPPs) and critical quality attributes (CQAs) of the products as wells as with key performance indicators (KPIs), like the cell concentration. These relations are usually statistical relations that are not based on detailed knowledge of biological mechanisms, although they are a good basis to acquire such knowledge. Transcriptomics is a state of the art technique to measure the global gene expression of a cell population and can give insight into the biological mechanisms that are behind the relation between CPPs, CQAs and KPIs. Furthermore, the transcriptome is a good representation of the physiological state of the cells and can thus be used for comparison of for example reactors of different scales, different processes, or comparison of a new bioreactor production run to past runs.

The aim of this thesis is to study the value of transcriptome analysis for process development and the quality by design approach for the upstream cultivation process. All experiments in this thesis use the same CHO cell clone producing a monoclonal antibody. All transcriptome measurements are done using commercially available Affymetrix CHO Gene 2.1 ST arrays.

Early process development for biopharmaceuticals is often partly done in small scale systems in a high throughput approach. Transcriptome analysis is used in this thesis to assess whether the small scale systems are representative for the production scale reactor.

The focus of Chapter 2 is to have a better understanding of the differences in cell performance between uncontrolled (shake flask) and controlled (bioreactor) systems. In this chapter, we evaluated differences in gene expression profiles between shake flask and bioreactor cultures at three different time points during the exponential and stationary phase of a batch cultivation. Both systems showed a similar large variation in gene expression over time, meaning the differences between two cultivation systems are small. However, the small gene expression difference between the two systems occurred during a short period of cultivation during batch cultivation and could be directly linked to the absence of control of some of the process cultivation parameters such as pH and DO in the shake flask.

For fed-batch processes larger differences are expected to be present between controlled and uncontrolled systems as compared to batch due to the higher cell densities, bolus addition of concentrated feeds and longer process duration. Therefore, we investigate in Chapter 3, the differences in gene expression between, shake flasks and 1 L bioreactors for a fed-batch cultivation process. The shake flask cultures grew faster than the bioreactor cultures and went earlier into the stationary and death phase. Using PCA the difference in time development was represented in PC1, while also a time independent difference was observed and represented by PC2. Although transcriptome data can also identify differentially expressed pathways, which may lead to the root cause of the differences, here we were not able to identify this root cause.

In Chapter 4, transcriptome analysis was used to evaluate the down scale of a CHO cell fed-batch process from a 10 L bioreactor to an ambr 15® (ambr) system. In this case both systems have control of pH and DO. The results of this comparative transcriptomic study showed that the variation in gene expression was less than 6% based on PCA. Moreover, the gene and pathway analysis did not reveal a direct relation of this difference with scale differences or specific process conditions. To be able to obtain an objective meaning to this 6% difference more transcriptome experiments need to be done. In this Chapter we also show that differences in gene expression in the preculture quickly disappear in the culture systems, showing that the preculture did not have an effect on the comparison.

In Chapter 5, we explore the use of transcriptomic profiling for monitoring long term production campaigns such as occur using a perfusion cultivation mode. The results show that transcriptome data can be linked to specific phases of a perfusion process like a period characterized by a cell size increase and a period with a specific nutrient limitation. More importantly, quantitative transcriptome data visualized using PCA. show a clear separation between the steady state data points and the other data points and can thus be used to identify the steady state.

In Chapter 6, transcriptomic profiling as a tool for quality by design in the cultivation process of antibody biosimilars is discussed. Comparative transcriptomic analysis could play a role in validation of the scale down system by demonstrating similar responses of biological processes for both systems. Likewise it can be an useful tool to detect process deviations, evaluate process modifications, and process characterization. Furthermore, in combination with other analyses it can give insight into biological processes that link CCPs to CQAs. Thus,  transcriptomic profiling can be part of process performance indicators and combined with other bioanalytical measure give a better understanding and faster development of the design space. When the analysis can be done routinely and sufficient fast it can possibly also be used at line to evaluate and control running production runs.