to study the effects of drugs and other environmental toxicants on the liver, many believe that the use of hepatocyte-derived cultures offers advantages over studies in the liver of intact animals. Cell cultures can be manipulated more easily, and they are far less expensive to maintain than a laboratory animal or transgenic mouse colony. Indeed, this is the theme of this section (Cell Physiology) of the American Journal of Physiology. Do signaling pathways and metabolism in hepatocytes in culture, however, reflect those in the intact animal? Can drugs and other environmental toxicants elicit the same subcellular response in isolated or cultured hepatocytes, or in hepatoma established cell lines, as in the liver of the intact animal?
BACKGROUND AND HISTORY
Immortalized cell lines have their problems; many cell lines dedifferentiate and lose differentiated cell functions—not unlike cancer cells. Primary cell cultures show similar problems (29, 32). My laboratory noted (30) that in established cell lines and in primary and secondary fetal cell cultures 1) cytochrome P-450 (CYP) enzyme activities that are normally induced by polycyclic aromatic hydrocarbons (PAHs) in the intact animal are abnormally induced by phenobarbital and many other unpredicted chemicals, 2) CYP activities that are normally induced by phenobarbital are lost in cell culture, 3) constitutive CYP activities also are extinguished, and 4) phenylalanine hydroxylase activity disappears during the first few minutes of trypsinization of mammalian liver—before hepatocytes are even plated out on dishes. We thus concluded that most results from cell culture (stable established lines or primary/secondary cells) cannot easily be extrapolated to the intact animal. Attempts at growing cells (especially hepatocytes) on various matrices (34), suspensions (36), and three-dimensional continuous-flow systems (41) have been described over the intervening three decades, but the same conclusion remains.
How can gene expression in hepatocyte cultures be generated to resemble more closely that of the intact animal? The “tet-off” system (12) was a significant advance in the field; however, this required the tetracycline “inducer” (usually doxycycline, dox) to be present to prevent (repress) the trans-regulated gene from being expressed. Removal of dox then turned on the gene of interest. The “reverse tet-off” (also called “tet-on”) system (13) represented a solution to that problem; here, the inducer dox is added to upregulate the trans-regulated gene being studied. Still, the high background expression of some genes in the absence of dox (8, 9) and the recalcitrance—particularly of hepatocytes—to the reverse tet-off system (10, 11) have left many investigators frustrated. To address these problems, there is now the “second-generation reverse tet-off,” abbreviated rtTA2S-M2 (45), which shows a greater trans-activation potential and a lowered affinity for the target gene promoter in the absence of dox; rtTA2S-M2 has been shown to be successful in transgenic animals (23, 47). In the present article in focus, Goldring et al. (Ref. 11; see p. C104 in this issue), have now extended similar studies to human and mouse hepatoma cell lines—with exciting results.
Five genes were selected (11), based on relevance to their expression in liver and their response to certain drugs and other environmental toxicants: rabbit CYP2E1 and human glutathione S-transferase P1 (GSTP1) represent the so-called “phase I” and “phase II” drug-metabolizing enzymes, respectively; human nuclear factor (erythroid-derived-2)-like-1 (NFE2L1), mouse NFE2L2, and human nuclear factor (p105) of κ light polypeptide gene enhancer in B-cells-1 (NFKB1) represent three transcription factors that respond to various environmental stressor signals. This was an excellent choice of five genes, which covers the two major phases of metabolism (of drugs and other environmental toxicants) plus three transcription factors that are well known to respond to the reactive metabolites generated by such metabolism.
The rtTA2S-M2, controlling each of these five genes, was studied in three hepatoma cell lines: human HepG2 (35), human HuH7 (33), and mouse Hepa1c1c7 (1). Stable clones with CYP2E1, GSTP1, and NFE2L2 and transiently transfected NFE2L1 and NFKB1 exhibited high levels of functional expression (experimentally demonstrated by Western blot of protein and/or enzyme activity) in a time- and dose-dependent manner (11). Moreover, these variously expressed genes appeared not to cause any detectable changes in the HepG2 proteome; to my knowledge, this is the first example of a cell culture study in which possible changes in the proteome were carefully examined, as a function of artificially enhancing the expression of only one gene.
Therefore, for any of these genes that might have been extinguished because of cell culture conditions, the levels of gene product can be restored (enhanced or diminished) by rtTA2S-M2 action, to “more physiologically relevant” levels or to whatever level the investigator wishes. The ability to control the levels of each of these genes (and gene products) by titrating the dox concentration should allow the Goldring group and others to examine cellular responses to drugs, other environmental toxicants, and even endogenous homeostatic signals. Moreover, interindividual variations in the expression of any of these genes in human populations can be dramatically different (25), and such genetic (or ethnic) differences (28) could also be amenable to study in various rtTA2S-M2-expressing cell lines.
LOOKING AT THE BIGGER PICTURE: USING CYP1 ENZYMES AS AN EXAMPLE
Mammalian CYP1A1, CYP1A2, and CYP1B1 genes—encoding the enzymes CYP1A1, CYP1A2 and CYP1B1, respectively—are regulated by the aromatic hydrocarbon receptor (AHR). The three CYP1 enzymes are well known to participate in both the metabolic activation and detoxication of numerous PAHs (e.g., benzo[a]pyrene, BaP), halogenated hydrocarbons, and aromatic amines present in combustion products, as well as a number of benzoflavones and indole derivatives occurring in cruciferous plants. Many substrates for the CYP1 enzymes are also AHR ligands. More than 15-fold differences in AHR affinity among inbred mouse strains reflect variations in CYP1 inducibility and clearly have been shown to be associated with differences in risk of toxicity or cancer caused by PAHs and arylamines. At least 12-fold variability in the human AHR affinity also exists (24), but—for ethical reasons—differences in human risk of toxicity or cancer related to AHR activation are difficult to study and remain unproven.
The role of CYP1A1 in BaP-induced toxicity was first most elegantly demonstrated in the Hepa1c1c7 hepatoma cell line (16). BaP-treated Hepa-1 cells grew only rarely as resistant variants; such colonies were used to complement the “resistance” phenotype in other colonies, which led to the discovery of many complementation groups, of which three have been examined in the most detail. These variants were ultimately defined as the mouse genes encoding Cyp1a1, Ahr, and the AHR's dimerization partner, the AHR nuclear translocator, Arnt (15). Thus CYP1A1 activates BaP to become toxic, and the AHR and ARNT are necessary for Cyp1a1 inducible expression; these experiments (in cell culture as well as others in microsomal suspensions in vitro) show that CYP1A1 is a primary determinant for BaP toxicity.
One therefore presumes that CYP1A1 is likely to be responsible for BaP-mediated toxicity in the intact animal, and, indeed, many studies have shown a correlation between the high-affinity AHR, inducible CYP1A1, and risk of PAH-induced skin or lung cancer in the rodent. On the basis of such results, we reasoned that knocking out the Cyp1a1 gene would protect the animals against BaP; however, we were wrong. When given oral BaP, all Cyp1a1(−/−) mice died within 30 days, whereas Cyp1a1(+/+) wild-type mice survived for a year. BaP-DNA adducts were unexpectedly much higher in the gastrointestinal tract, liver, spleen, and marrow of Cyp1a1(−/−) than in the tissues of wild-type mice. Immunotoxicity occurred in Cyp1a1(−/−) mice but not in wild-type mice. BaP pharmacokinetic studies showed that adducts accumulate to high levels in Cyp1a1(−/−) mice, despite much lower rates of BaP metabolism in the genetic absence of CYP1A1 (44). Thus, without CYP1A1 present to detoxify and eliminate the parent BaP substrate, the accumulated BaP and its slow oxidative metabolism by other enzymes is far worse for the intact mouse than it is to have CYP1A1 present.
We found a similar surprise in studies of CYP1A2. Substantial constitutive CYP1A2 activity exists in mammalian liver. There are >60-fold differences in hepatic CYP1A2 between humans (6, 26), yet no mutations in or near the CYP1A2 gene have unequivocally been demonstrated to account for the striking interindividual differences in levels of this constitutive expression. CYP1A2 substrates include one to two dozen drugs, plus many environmental amines: N-heterocyclic amines found in charcoal-grilled food—such as 2-amino-3-methylimidazo[4,5f]quinoline (IQ) and 2-amino-1-methyl-6-phenylimidazo[4,5b]pyridine (PhIP)—and arylamines such as the human urinary bladder carcinogen 4-aminobiphenyl (ABP).
Metabolic activation of ABP by CYP1A2 in vitro is well known to cause enhanced ABP-DNA adducts and toxicity (discussed in Ref. 43). Injection of 4-hydroxy-ABP (a major product of CYP1A2-mediated metabolism of ABP), but not the ABP parent compound itself, into a dog's bladder produces ABP-DNA adducts and toxicity. The most common route of administration for ABP in humans is through the skin and lung. We hypothesized that a knockout of the Cyp1a2 gene would protect the mouse against ABP; again, we were wrong: Cyp1a2(−/−) mice treated topically with ABP had higher levels of ABP-DNA adducts in the liver and urinary bladder and more hepatic oxidative stress than Cyp1a2(+/+) wild-type mice (43)—presumably due to the accumulation of abnormally high amounts of ABP and then due to its oxidative metabolism by enzymes other than CYP1A2. A similar “contradiction” has been seen in ABP-induced hepatocellular carcinomas and preneoplastic foci (22) and ABP-induced methemoglobinemia (37). Further paradoxical responses have also been observed with the food mutagens IQ and PhIP on DNA adducts in liver, kidney, mammary gland, and colon (38) and the effect of PhIP on the incidence of several types of malignancies (21).
This paradoxical response has not been seen in Cyp1b1(−/−) or Ahr(−/−) knockout mouse lines (24). CYP1B1 metabolizes numerous PAHs, as well as many N-heterocyclic amines, arylamines, and amino azo dyes, and several other carcinogens (14). Unlike CYP1A1, CYP1B1 often shows substantial constitutive levels. CYP1B1 expression is high in vascular endothelial cells, squamous epithelial cells, white blood cells and myeloid precursors, breast, prostate, uterus, various types of tumors, adrenal cortex, and other tissues. As one might have predicted from in vitro studies, the Cyp1b1(−/−) mouse exhibits increased protection against 7,12-dimethylbenz[a]anthracene (DMBA)-induced lymphomas (3), DMBA-induced marrow toxicity and preleukemia (31), and dibenzo[a,l]pyrene-induced tumors (2). Hence, if CYP1B1 is not present in the Cyp1b1(−/−) knockout mouse to activate these environmental chemicals, less toxicity or malignancy is seen.
The Ahr(−/−) knockout mouse exhibits a lack of constitutive and inducible expression of CYP1A1, CYP1A2, and CYP1B1. As one might also have predicted from in vitro studies, the Ahr(−/−) mouse is highly resistant to 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced toxicity (7), topical BaP-induced skin tumors (38), and benzene-induced hemotoxicity (46).
Thus, in the context of hepatoma cells or microsomal metabolism in vitro, CYP1A1 is the primary determinant of BaP-mediated toxicity and DNA adduct formation, whereas CYP1A2 is the primary determinant of arylamine-mediated toxicity and DNA adduct formation. In contrast, in the context of the intact animal, CYP1A1 and CYP1A2 can be protective. This dual role has not been seen with CYP1B1. An attempt to explain this apparent discrepancy has been proposed (24) and is summarized in Fig. 1. In microsomes, in 9,000 g supernatant (S9) fractions commonly used for the Ames test, or in hepatoma cell lines: the complete absence of, or a loose coupling between, the phase I CYP1 and the phase II enzymes would result in enhanced adduct formation, oxidative stress, and toxicity. In gastrointestinal epithelial cells or hepatocytes, however, it is possible that CYP1A1 and CYP1A2 are tightly coupled, resulting in efficient detoxication, rather than the release of significant amounts of reactive metabolites, thereby leading to toxicity or cancer. In immune cells, it is possible that CYP1B1 is not tightly coupled to phase II metabolism, or that phase II metabolism is low or absent, resulting in enhanced BaP-DNA adduct formation and toxicity in spleen, thymus, and bone marrow. An additional likely factor is the level of the CYP1 enzyme: gastrointestinal and hepatic CYP1A1 and CYP1A2 levels are very high in the paradoxical systems described above, whereas CYP1B1 content, in relative terms, is not high in immune cells.
CELLULAR AND SUBCELLULAR CONTEXT AND PHARMACOKINETICS
To reiterate, cells in culture might not retain their cell-cell context, and perhaps even the context of their subcellular organelles (in subcellular expression, signaling, and organization), with regard to the same cell type in the intact animal. In all the Cyp1 knockout mouse studies, the role of CYP1 in detoxication vs. activation (to cause toxicity or cancer) is likely to depend on the subcellular content and location of the CYP1 enzymes, the amount of phase II metabolism, the degree of coupling of the CYP1 phase I to the phase II enzymes, and cell type- and tissue-specific context. One or more of these factors could result in abnormally high accumulations of drugs or other environmental toxicants in the intact animal, which leads to profound pharmacokinetic effects (route of administration, target organ) of the drug or chemical under study (24).
The notion that CYP1A1 is causative in PAH-mediated toxicity and carcinogenesis (or CYP1A2 causative in ABP-, IQ- or PhIP-mediated toxicity and malignancy) may not be warranted, and, in fact, the contrary may be true. These findings underscore the difficulties in using results collected in cell cultures and in vitro to extrapolate to the in vivo situation. In vitro data have been invaluable in helping determine the catalytic specificities of CYP1 enzymes; from this perspective, there can be little doubt that CYP1B1 and CYP1A1 represent major cellular activities toward PAH metabolism or that CYP1A2 carries out arylamine metabolism. The roles of CYP1 in causing, preventing, or not participating in PAH- or arylamine-mediated toxicities in the intact animal will therefore require further studies.
TWO NOVEL EXCITING BREAKTHROUGHS IN THE FIELD, AS I SEE IT
Recently, mice have been generated with exclusively human liver function. Human hepatocytes are transplanted into urokinase-type plasminogen activator-transgenic severe combined immunodeficiency mice (uPA/SCID), which otherwise are immunodeficient and undergo liver failure; these chimeric mice no longer undergo liver failure and allow human hepatocytes to propagate in their livers and retain normal human pharmacological responses, such as drug metabolism by phase I (18, 42) and phase II (20) enzymes and inducibility by rifampicin and a PAH (19). With this system, of course, each mouse will also reflect the liver profile and genetic makeup of the human who has donated the hepatocytes—meaning that interindividual differences and ethnic haplotypes of drug-metabolizing enzymes, other enzymes, receptors, transporters, and any other drug target located in human liver might be effectively studied in these types of mouse lines.
More recently, “humanized” hCYP1A1_1A2 mouse lines (having the mouse orthologous Cyp1a1 or Cyp1a2 gene genetically disrupted) have been reported (4, 17). Human CYP1A1 and CYP1A2 are oriented head to head, 23.3 kb apart; the entire CYP1A1_CYP1A2 locus of <40 kb, in the middle of a 200-kb bacterial artificial chromosome, appears to exhibit physiological tissue-specific expression and induction of both human CYP1A1 and CYP1A2 mRNAs and enzymes in the absence of either mouse orthologous gene (17). For theophylline clearance, the humanized hCYP1A1_1A2 “knock-in” mouse exhibits a human, rather than a mouse, profile of urinary theophylline metabolites, and the Cyp1a2(−/−) knockout mouse shows greatly impaired theophylline clearance (5). A second hCYP1A1_1A2 mouse line has been made independently, and this mouse exhibits a human, rather than a mouse, profile for PhIP metabolism (4). These hCYP1A1_1A2 mouse lines should be invaluable in human risk assessment for various environmental toxicants and carcinogens, present in combustion processes, which cannot be tested in clinical populations for ethical reasons (17).
STANDARDIZATION OF GENE NOMENCLATURE
Finally, as a long-standing member of the International Advisory Committee of the Human Genome Organization-sponsored (HUGO) Gene Nomenclature Committee (HGNC; http://www.gene.ucl.ac.uk/nomenclature/), I am obliged to comment on the “NRF1, NRF2” gene nomenclature used in the paper by Goldring et al. (11). In the published literature, these mammalian genes (along with a third family member) are almost uniformly called “NRF1, NRF2, and NRF3,” encoding the gene products “NRF1, NRF2, and NRF3.” These names have become very popular over the past several years, just like the erroneous terms “COX-1” and “COX-2” for cyclooxygenase-1 and -2 have predominated over the official HGNC nomenclature of the PTGS1 and PTGS2 genes (27). The HGNC website indicates that “nuclear respiratory factor-1” should be named the NRF1 gene, and this decision occurred before characterization of the oxidative stress-related “NRF1, NRF2, and NRF3” genes. The official names for these three “nuclear factor (erythroid-derived-2)-like” genes therefore became NFE2L1, NFE2L2, and NFE2L3, respectively, because they show homology to the NFE2 gene on human chromosome 12q13. The NFE2L1, NFE2L2, and NFE2L3 genes are located on human chromosomes 17q21.3, 2q31, and 7p14-p15, respectively.
Why is it important to agree on one specific name for each gene? As multiple genomes are sequenced, it is imperative to consider the complexity of genes, genetic architecture, gene expression, gene-gene and gene-product interactions, and evolutionary relatedness across species. For example, fly geneticists have enjoyed using wonderfully whimsical names such as daughterless, groucho, hedgehog, mad (mothers against decapentaplegic), faint sausage, current bun, clootie dumpling, lunatic fringe, indy (I’m not dead yet), pokemon plutonium, and saxophone. These quaint names have potentially serious consequences, because fly genes generally have homologs in the human, and, in trying to maintain consistency with this nomenclature, functional information can be lost. Furthermore, if a doctor must explain to the parent of a child who has a disorder due to a mutated sonic hedgehog gene (one of three Drosophila orthologs in humans), this may be uncomfortable because of the gene's humorous name (27). It is therefore mandatory that we all agree on a particular (and serious) gene name. Standardized nomenclature across all species not only makes one's own research easier, and facilitates understanding for those outside the field to understand quickly a gene or gene family, but it will also be helpful to the future generations of graduate students and postdoctoral fellows who might venture into genomics research.
This work was supported in part by National Institute of Environmental Health Sciences Grant P30-ES-06096.
- Copyright © 2006 the American Physiological Society