Temporal changes in global gene expression

For a full characterization of the impact of culture conditions on the behavior and functionality of hepatocytes over time, transcriptional changes were analyzed globally across the complete dataset. Therefore, fold-changes and statistically significance were calculated in relation to the particular starting points of the culture, which was defined as the reference sample (Appendix 3 and Appendix 4). For the short term culture methods, such as liver slices and suspension cultures, reference samples were defined as the 0 h time point after isolation, which means freshly cut liver slices or freshly isolated hepatocytes, respectively. The latter was used to eliminate the background of gene expression changes due to the lack of other cell types.

As previously described, the process of isolating hepatocytes caused a large number of gene expression changes which, at least in part, can be considered as common and therefore are present in all types of cultures. Consequently, the 1 d time point after plating was defined as the starting point for the longer term culture methods. The initial

changes were thereby excluded from the analysis and evaluated separately. Due to the fact that the human hepatocytes were prepared and plated in France, the first time point analyzed, 2 d after perfusion, was used as the reference sample.

Short term cultures generally showed a high correlation to their reference experiments (Figure 46). This is true not only for the liver slices, which still contain all liver-typical cells, but also for the hepatocyte suspension cultures. Major effects were first detected after 6 h for liver slices and suspension cultures and after 1 day in culture, clear differences were seen. After one day, the gene expression in both cultures was measurably different to controls correlating with the decline in viability observed for these cells.

Figure 46: Heat maps of the correlation coefficients of rat cell culture experiments compared to the reference system over time. Each square in a column or line represents the gene expression correlation of a given sample at a certain time point (arrow) relative to the reference experiments. The intensity-changes in global gene expression were used as the basis for the calculation of the correlation. Long-term experiments were split: the upper part of the square shows the correlation of the experiments to freshly isolated cells, the lower part indicates the

The longer term cultures were compared to FC as well as to cells in culture for one day. Shown in the lower part of Figure 46 are the heat maps visualizing the correlations between each time point and FC (above white line in each square) and day one of culture (below white line in each square). A reduction of the correlation coefficient, visualized by a shift from red to black to green, indicates significant changes in global gene expression in comparison to the references. For all cultures, by day one a reduced correlation was seen, although it was most pronounced in ML+FCS. As shown in the previous chapter, the initial changes introduced by the elimination of other cell types and the initial adaptation processes are likely to cause similar changes in all types of cultures.

The correlation coefficient of ML+FCS cultured cells decreased over time when compared with gene expression in FC and with cells one day in culture, reflecting the advancing dedifferentiation processes. This result perfectly correlates to the morphological analyzes described before with no stabilization of gene expression detected.

The removal of FCS from the culture media and the addition of Dex improved the correlations. The extent of initial changes was reduced and processes moving the cells away from hepatocyte-like gene expression were significantly slowed down, at least globally, after two days in culture. Until the end of culture, the gene expression of these cells showed more stability. As the aim of toxicogenomics is the detection of gene expression changes caused by compound treatment, it is important, especially for in

vitro models, to reduce the background of genes changing due to other factors, such as

the culturing, to a minimum.

Cells cultured in SW in the presence of serum showed the initial changes which were less pronounced compared to ML+FCS. Globally, the cells remained in this state until day four. Afterwards, a reduction of the gene expression correlation coefficient was detected. This process was intensified at later time points indicating the onset of dedifferentiation in these cells.

In SW culture without serum the addition of Dex had additionally positive effects on global gene expression. The gene expression changes due to the isolation process and adaptation to the culture environment, although still quite high, were least pronounced and global gene expression over time was most stable of all cultures tested. From two days in culture until the end of the culture, an increase in correlation to FC was observed suggesting some regenerative processes were taking place in these cells.

Figure 47: Heat map of genes transiently deregulated one day after perfusion in sandwich culture without serum.

Genes found to be transiently deregulated after one day in culture and returning to their original expression level (Figure 47) were mainly genes known to be involved in early stress response, inflammatory mechanisms and intracellular signalling. Networks built from these specifically deregulated genes (confidence of 95% to be only deregulated at 1 d in SW-FCS) confirmed these findings but also showed a link to the regulation of fatty acid biosynthetic processes (Table 16). The expression of the PPARs was reduced 1.8-fold. This transcription factor is known for its ability to induce gluconeogenesis and to reduce fatty acid β-oxidation.

Processes Size Target p-Value

immune system process (40.0%), V(D)J recombination (7.5%), nitric

oxide transport (5.0%) 50 10 1.05e-21

protein kinase cascade (44.0%), stress-activated protein kinase

signaling pathway (26.0%), protein amino acid phosphorylation (44.0%) 50 10 1.64e-21 fatty acid biosynthetic process (23.1%), carboxylic acid biosynthetic

process (23.1%), organic acid biosynthetic process (23.1%) 50 8 8.45e-18 regulation of biosynthetic process (20.6%), regulation of cellular

biosynthetic process (20.6%), biological regulation (79.4%) 50 9 2.35e-19 response to stress (65.1%), positive regulation of cellular metabolic

process (46.5%), positive regulation of metabolic process (46.5%) 50 8 7.23e-17

Table 16: Top five networks highly enriched with genes found to be deregulated only at day one in SW-FCS cultured hepatocytes. “Size” refers to the number of network objects contained and “Target“ is the number of affected objects contained in these networks.

Figure 48: Heat maps of the correlation coefficients of human cell culture experiments to their reference systems over time. Experiments were ordered according to the time scale (big arrow) and separated into short- and long-term experiments. The intensity-changes in global gene expression were used as the basis for the calculation of the correlation. Long-term experiments are split up; the upper part of the square shows the correlation of the experiments to freshly isolated cells, the lower part indicates the correlation to the status 1 d after plating which was defined as the reference experiment for later analyses. Red squares indicate high correlation, green indicate low correlation. The pictures show cells of each longer-term culture at day one (left) and day 10 (right) of culture.

The results of the long term cultures displayed differences from the results gained with rat hepatocytes. Due to technical reasons, the 2 d time point was the first sample to be taken and therefore this was defined as a second reference, together with FC. For ML and SW cultures, a distinctly worse correlation was detected at the initial time points. This change of gene expression was less pronounced in ML culture indicating a greater stability of these cells. Over time, both cultures demonstrated only minor changes pointing to a generally better stability of gene expression in human cells compared with the situation in rat. Another source of variance when working with primary human cells is the large inter-individual donor difference. Both the basal gene expression and the individual reactions of the cells can be remarkably different. This was confirmed by our data. Four different donors were clearly differentiated, based on their gene expression and the extent of gene expression changes over time (Figure 48 ML and SW). The correlation “in-between” donors at a certain time point was 0.97 for primary cultured human hepatocytes. For the suspension culture, the level of correlation between different time points was in the same range (0.95). Therefore, genes found to be differentially deregulated may be influenced more by donor specificity than by time in culture.

Figure 49: Plot of intra group and inter group correlation coefficients of human hepatocyte cultures.

Experiments with primary human hepatocytes should be carefully analyzed with respect to donor specific gene expression. If possible, more than three biological replicates should be included to ensure that general biological trends are visible above any individual variances.

For both rat and human, the established cell lines (FaO and HepG2, respectively) were found to vary greatly from primary cells, showing many differences in global gene expression. Another cell line, the recently established HepaRG, showed a high stability of gene expression over time. Additionally, the gene expression was closer to that of primary human hepatocytes than that of HepG2 cells. Large differences were detected compared to FC, maybe due to the lack of inter-individual differences and the increased stability of gene expression over time in culture. These cells may therefore be a suitable experimental system for toxicogenomics studies. Further analyses have to be conducted, including the monitoring of the existence of certain metabolic enzymes, which allows liver-like metabolism in these cells (chapter 3.3.6.4).