In this study, we investigated a suite of eight 'housekeeping' genes that represent different functional classes and gene families for use as internal controls to normalize real-time PCR data in experimental treatments with an estrogen (EE2) in fish. Specifically, these genes comprised 18S rRNA, rpl8, ef1a, g6pd, bactin, gapdh, hprt1, and tbp. These 'housekeeping' genes have been validated previously for use as internal controls in many experimental systems, particularly in mammals (18S rRNA: e.g. [10-12]; ribosomal proteins: e.g. [13-15]; ef1a: e.g. [13,14,16]; g6pd: e.g. [17,18]; bactin: e.g. [10,19]), gapdh: e.g. [19,20]; hprt1: e.g. [19,21-23]; tbp e.g. ), but have not been validated for work on estrogens in fish. The development of the required real-time PCR assays for some of the target 'housekeeping' genes, whose sequences were not publicly available in the study species (18S rRNA, rp18, ef1a), first required the cloning of the full- or partial-length sequences for these genes.
A wide range of 'housekeeping' genes have been used as internal controls in expression profiling following exposure of fish to estrogens, including bactin (e.g. 17β-estradiol [E2] [24,25]; EE2 [26-28]; bisphenol A [BPA] ; nonylphenol [NP] ; alpha-zearalenol ), 16S rRNA (e.g. EE2 ), 18S rRNA (e.g. E2 [30,31]; NP ), ef1a (e.g. EE2 ; NP ), and ribosomal proteins (including S2, S3, S8, S15, S27, L4, L5, L8, L13, L21, and L28; e.g. E2, NP, and 1,1-dichloro-2 2-bis (p-chlorophenyl) ethylene [p,p'-DDE]: ). However, very few of those studies provided any validation of these genes for use as internal controls prior to their application.
Our data clearly show that the expression levels of some 'housekeeping' genes are regulated by estrogen in fathead minnow. In particular, ef1a, g6pd, bactin, and gapdh were found to be strongly regulated by estrogen, and this is consistent with data published for other teleosts. Down-regulation of bactin was shown to occur in array analyses conducted on liver tissue from male sheepshead minnow (Cyprinodon variegates) exposed to the environmental estrogens E2, EE2, diethylstilbestrol (DES), NP, and methoxychlor . This concurs with the observations that E2 and several other estrogenic compounds disrupt cytoskeletal compounds in vitro [34-36]. In contrast, bactin expression has been shown to be increased in liver of plaice (Pleuronectes platessa) and zebrafish (Danio rerio) exposed to EE2 [29,37], and in pituitary of Atlantic salmon (Salmo salar) exposed to E2 and NP . Expression of gapdh in livers of plaice and rainbow trout (Oncorhynchus mykiss) has also been shown to be repressed by various environmental estrogens, including EE2 [28,29,39]. Further evidence of estrogen-regulation of bactin and gapdh coming from mammalian studies [40-43] strengthens the case that their use as internal controls for studies for quantification of gene expression with estrogen treatment is probably inappropriate. The down-regulation of hepatic ef1a and g6pd by EE2 observed in the present study also concurs with data from zebrafish embryos , where two ef1a isoforms were down-regulated by estrogen, and from the European flounder (Platichthys flesus), where E2 inhibited activity of the G6pd enzyme in hepatocytes by 30% in males and 80% in females . In fact, estrogen-regulation of g6pd provides a likely explanation for differences we observed in hepatic g6pd expression between the sexes in this study (lower expression in females; data not shown), since female fish have higher circulating levels of endogenous estrogens than males. The effect of estrogen on gapdh and g6pd expression observed here likely results from their involvement in metabolism, since estrogens, like other sex steroids, are well known to have roles in the regulation of the metabolic processes associated with altered energy demands during gonad development and reproduction in fish [45-47]. Moreover, several other metabolic enzymes are also known to be controlled by sex steroids including estrogens in fish [27,46,48].
Interestingly, for some of the candidate 'housekeeping' genes, the effects of estrogen were tissue- and/or gender-specific. For example, ef1a and g6pd expression were only altered by EE2 in liver and not in gonad. Moreover, the expression levels of gapdh in gonad and bactin in liver were only down-regulated by EE2 in males and not in females. Complex effects of estrogens (including tissue-, gender-, and/or developmental-stage-specific effects) have been demonstrated for many estrogen-regulated genes, including 'housekeeping' genes, in fish and higher vertebrates (e.g. [4,41]) and this suggests that these genes may have different roles in different tissues, and different roles in males compared with in females in these tissues. The tissue- and gender-specific nature of these effects of EE2 on 'housekeeping' gene expression emphasises the need for caution if extrapolating data from one tissue/gender to another when choosing an internal control gene.
Four of the candidate 'housekeeping' genes (18S rRNA, rpl8, hprt1 and tbp) were found to be unaffected by estrogen treatment in both tissues examined in this work and are, therefore, considered valid for use as internal controls in such experiments. The high stability of rpl8 expression following estrogen treatment conforms with data from largemouth bass (Micropterus salmoides), where the expression of a suite of 11 ribosomal proteins, including rpl8, did not fluctuate appreciably (. However, the results of another study on rainbow trout demonstrated EE2-induction of the ribosomal protein ribosomal protein s3 (rps3) , emphasising that not all ribosomal proteins may be estrogen-independent. Although there is no other data currently available concerning possible estrogen regulation of hprt1 and tbp in fish, hprt has been described as an estrogen-independent 'housekeeping' gene in mammals , supporting our findings here.
The lack of regulation of 18S rRNA by EE2 in this study (21 day exposure) concurs with the findings for an exposure to EE2 conducted on zebrafish (liver, for periods of 48 and 168 hours) . In that study however, and based on array analyses, 18S rRNA was up-regulated in a concentration-dependent manner in zebrafish liver after a shorter term (24 hour) exposure to EE2, suggesting that any 18S rRNA regulation by estrogen may occur in a temporal manner. Although rRNAs have been advocated for use as internal controls in many other experimental systems and their levels are thought to be less likely to vary under conditions that affect the expression of mRNAs (e.g. [10-12,49]), there is some evidence that another rRNA, 28S rRNA, is also regulated by estrogen in fish . Moreover, some clear disadvantages in the use of rRNAs as internal controls have been highlighted by other researchers (e.g. see [17,50,51]. Most notably, rRNAs can only be used as internal controls for total RNA preparations (since mRNA purification methods randomly remove them) and cannot be used when an oligo(dT) reverse transcription has been carried out. The use of rRNAs, which constitute 80–90% of total RNA, has also been criticised due to their very high general expression level which far exceeds that for mRNAs (e.g. [51,52]). A very high general level of expression of 18S rRNA was also observed in the current study (18S rRNA had a mean Ct of approximately 11, compared to mean Cts of approximately 23–32 for the other candidate 'housekeeping' genes). It is imperative, therefore, that these issues are considered if selecting 18S rRNA as an internal control for studies on environmental estrogens in fish.
To assess the importance of selecting an appropriately validated 'housekeeping' gene for use as an internal control, in our final analyses we measured the expression of two genes of interest, vtg and cyp1a, in the livers of fathead minnow exposed to EE2 and compared the relative expression results obtained using the different housekeeping genes to those obtained when the amount of input cDNA was normalized. vtg and cyp1a were selected for this analysis because of their biological roles in oocyte development and xenobiotic metabolism, respectively, because they are well characterised as estrogen-inducible and estrogen-downregulated genes, respectively, and because they are frequently included as genes of interest in studies on the effects of environmental estrogens (e.g. [32,53,54]). When the amount of input cDNA was normalized, vtg mRNA expression in liver was highly (approximately 200-fold) up-regulated by EE2 in males (but not in females), while cyp1a mRNA expression was down-regulated by EE2 in both sexes (to between 15–28% of the control level), and these findings were as expected and in agreement with the current literature (e.g. [32,53,54]). We then alternatively normalized vtg and cyp1a expression using a 'housekeeping' gene approach, with each of the eight different 'housekeeping' genes.
Depending on the 'housekeeping' gene used for the gene of interest normalization, vtg mRNA levels were either unchanged or increased (to widely varying degrees from 250- to 8,300-fold), and cyp1a mRNA levels were either down-regulated, unchanged or increased (and, again, to different degrees). As expected, the use of appropriate 'housekeeping' genes (18S rRNA, rpl8, tbp, or hprt1; which were not affected by the EE2 treatment) for the normalizations, gave very similar results to the data generated when the input cDNA amount was normalized, demonstrating that the use of these genes is valid and effectively normalizes for differences in input cDNA. In contrast, when 'housekeeping' genes down-regulated by the EE2 treatment (ef1a, g6pd, bactin, or gapdh) were used, the expression levels of the genes of interest were overestimated, which lead to inappropriate conclusions about the manner in which these genes were regulated, or not regulated, by estrogen, and/or the degree to which they were estrogen-regulated. Likewise, if a 'housekeeping' gene which was up-regulated by EE2 had been used for normalization, the expression levels of the genes of interest would have been underestimated. These findings have major implications for studies on the effects and potential impacts of environmental estrogens in fish that use internal control genes that have not been validated for the specifics of the experimental system, and in particular for studies which require to detect small changes in the expression of the gene of interest. An interesting aside from this work has been to show that an environmental concentration of EE2 is capable of disrupting expression of a suite of genes that play key roles in general cell function and maintenance, adding further to the concern about the potential impacts of this estrogen in the aquatic environment.