Expression profiling of interindividual variability following xenobiotic exposures in primary human hepatocyte cultures

https://doi.org/10.1016/j.taap.2008.04.024Get rights and content

Abstract

To examine the magnitude of human variability across the entire transcriptome after chemical challenge, we profiled gene expression responses to three different prototypic chemical inducers in primary human hepatocyte cultures from ten independent donors. Correlation between basal expression in any two hepatocyte donors ranged from r2 values of 0.967 to 0.857, and chemical treatment tended to negatively impact correlation between donors. Including anticipated target genes, 10,812, 8373, and 7847 genes were changed in at least one donor by Aroclor 1254 (A1254), di(2-ethylhexyl) phthalate (DEHP), and phenobarbital (PB), respectively. A subset of these gene targets (n = 41) were altered with a high level of reproducibility in at least 9 donors, gene responses that correlated well with literature-reported mechanism of action. Filtering responses to the level of gene subsets clarified the biological impact associated with the respective chemical effectors, in lieu of substantial interindividual variation among donor responses. In these respects, the use of hierarchical clustering methods successfully grouped seven of the ten donors into chemical-specific rather than donor-specific clusters. However, at the whole-genome level, the magnitude of conserved gene expression changes among donors was surprisingly small, with fewer than 50% of the gene responses altered by a single chemical conserved in more than one donor. The use of higher level descriptors, such as those defined by the PANTHER classification system, may enable more consistent categorization of gene expression changes across individuals, as increased reproducibility was identified using this method.

Introduction

Primary human hepatocyte cultures are recognized as the most appropriate in vitro system with which the in vivo liver can be modeled for studies of xenobiotic metabolism and biotransformation (Gomez-Lechon et al., 2003, LeCluyse, 2001). Although advantages of the human primary hepatocyte model, such as elimination of metabolism extrapolation across species and increased predictability potential for idiosyncratic toxicity, are attractive, interindividual variability is problematic, as widespread differences in metabolism among donors and hepatocyte preparations have been cited as a limitation for data reproducibility (Liguori et al., 2005, Slatter et al., 2006). Issues of genetics as well as technical factors such as the surgical, storage and transfer conditions and pathological state of the donor liver tissue may all additionally contribute to variability in response (Baccarani et al., 2003, Fisher et al., 2001, Lloyd et al., 2004).

In studies that examine genome-wide transcriptional responses to chemical treatment, the potentially complicating nature of interindividual variability has led to difficulty in how to construe differences manifested at the transcriptome scale. In an effort to minimize variability across samples, some studies have pooled RNA from multiple donors (Huang et al., 2007, Piton et al., 2005), while others have only used hepatocytes from a single donor (Keum et al., 2006, Ryu et al., 2006). Variability across donors complicates the determination of which genes are regulated, in that some studies report findings in each individual donor (Thum and Borlak, 2007), some follow a stringent criteria such that genes of interest must be changed in all donors (Harris et al., 2004, Liguori et al., 2005, Radaeva et al., 2002), some apply a filtering system so that genes of interest are those changed in a particular number of donors (Richert et al., 2003), while others report average expression across donors (Li et al., 2007, Rae et al., 2001). Overall, a limited number of donors are generally included (n = 2–6), and surprisingly few genes survive the filtering conditions: Richert et al. parsed a 38 gene dataset based on the > 12,000 genes on their array; Thum and Borlak, 44 of 302 possible genes; Liguori et al., 142 of > 22,000; Harris et al., 867 of 31,000; and Radaeva et al., 53 of 12,000. The apparent trend emerging from these studies is that, regardless of the nature of the chemical of interest, only small sets of responsive genes are regulated reproducibly by xenobiotic exposures among hepatocytes across individual donors, whereas large subsets of genes are uniquely regulated within individual donors. Despite this trend, there is often little consideration given to the apparent vast magnitude of genes changed in a donor-specific manner.

To broaden these analyses regarding the magnitude of human variability existing in whole-transcriptome responses to chemical challenge, we performed whole-genome expression profiling analyses following treatments with three different prototypic chemical inducers using hepatocytes from ten independent donors, maintained under primary culture techniques that largely preserve highly differentiated character endpoints associated with mature hepatocytes (Olsavsky et al., 2007, Page et al., 2007). Comparisons among donors were performed both at basal conditions and after treatments with Aroclor 1254 (A1254), an environmental contaminant that consists of a mixture of polychlorinated biphenyls (PCBs), di(2-ethylhexyl) phthalate (DEHP), a plasticizer agent belonging to the peroxisome proliferator class of chemicals that activate the rodent PPARα receptor, and phenobarbital (PB), a barbiturate drug used as an anti-seizure agent. These chemicals were selected both for their diverse mechanisms of action and for the wealth of information available on the effects that these chemicals have in a variety of other model systems. Using ten human hepatocyte donors, we report findings in agreement with those from previous studies, that large subsets of genes were regulated according to individual donors, whereas a relatively small set of target genes was regulated consistently in response to individual chemical treatment across donors. However, we show that by using a PANTHER-based functional category analysis rather than a gene-level approach, more consistent identification of transcriptional patterns is detected across donors treated with common agents, patterns that are concordant with known mechanisms of action for the respective agents and indicative of biological themes identified in other model systems.

Section snippets

Cell culture

 Primary human hepatocytes were obtained from the University of Pittsburgh, through the Liver Tissue Cell Distribution System, NIH Contract #N01-DK-7-0004/HHSN267200700004C. Available donor information has been described in detail previously (Olsavsky et al., 2007). Hepatocytes were plated on collagen-coated T25 flasks, and within 48 h a dilute overlay of Matrigel (225 μg/ml; BD Biosciences, San Jose, CA) was added dropwise to the cultures in William's Media E supplemented with 1%

Correlation between hepatocyte donors

To assess the magnitude of variation between donors across the entire transcriptome, our initial analysis consisted of obtaining correlation coefficients as a global measure of similarity between donors. A comparison of mean correlation coefficients revealed a high level of correlation between eight of the ten donors, with mean correlation values ranging from r2 = 0.917 ± 0.022 to r2 = 0.932 ± 0.028 (standard deviation; Fig. 1A). The remaining two donors, D and E, had significantly lower correlation

Discussion

In this study, global transcriptional responses were examined in primary hepatocyte cultures from ten independent donors. Basal gene expression profiles as well as responses to three prototypic chemical inducers were assessed, with the intent of evaluating both trends in reproducibility across donors and mechanisms that may be conserved among donors, unique for each chemical class. Overall, the data obtained revealed that only small sets of genes were regulated consistently across donors,

Acknowledgments

This research was funded by a Toxicogenomics Research Consortium grant from the National Institutes of Environmental Health Sciences, U19 ES11387 and a grant from the National Institute of General Medical Sciences, GM66411.

References (44)

  • OlsavskyK.M. et al.

    Gene expression profiling and differentiation assessment in primary human hepatocyte cultures, established hepatoma cell lines, and human liver tissues

    Toxicol. Appl. Pharmacol.

    (2007)
  • ParkS. et al.

    Characterization of anti-apoptotic action of TCDD as a defensive cellular stress response reaction against the cell damaging action of ultra-violet irradiation in an immortalized normal human mammary epithelial cell line, MCF10A

    Toxicology

    (2006)
  • RadaevaS. et al.

    Interferon-alpha activates multiple STAT signals and down-regulates c-Met in primary human hepatocytes

    Gastroenterology

    (2002)
  • RichertL. et al.

    Effects of clofibric acid on mRNA expression profiles in primary cultures of rat, mouse and human hepatocytes

    Toxicol. Appl. Pharmacol.

    (2003)
  • SaitoM.

    Polychlorinated biphenyls-induced lipid peroxidation as measured by thiobarbituric acid-reactive substances in liver subcellular fractions of rats

    Biochim. Biophys. Acta

    (1990)
  • WuM. et al.

    AMID, an apoptosis-inducing factor-homologous mitochondrion-associated protein, induces caspase-independent apoptosis

    J. Biol. Chem.

    (2002)
  • YamanakaT. et al.

    Chemotherapeutic agents augment TRAIL-induced apoptosis in human hepatocellular carcinoma cell lines

    Hepatology

    (2000)
  • YuehM.F. et al.

    Nrf2-Keap1 signaling pathway regulates human UGT1A1 expression in vitro and in transgenic UGT1 mice

    J. Biol. Chem.

    (2007)
  • BammlerT. et al.

    Standardizing global gene expression analysis between laboratories and across platforms

    Nat. Methods

    (2005)
  • BivikC. et al.

    Hsp70 protects against UVB induced apoptosis by preventing release of cathepsins and cytochrome c in human melanocytes

    Carcinogenesis

    (2007)
  • CarauxG. et al.

    PermutMatrix: a graphical environment to arrange gene expression profiles in optimal linear order

    Bioinformatics

    (2005)
  • FisherR.L. et al.

    Human liver quality is a dominant factor in the outcome of in vitro studies

    Cell Biol. Toxicol.

    (2001)
  • Cited by (0)

    View full text