Supplements of vitamins B9 and B12 affect hepatic and mammary gland gene expression profiles in lactating dairy cows

Background A combined supplement of vitamins B9 and B12 was reported to increase milk and milk component yields of dairy cows without effect on feed intake. The present study was undertaken to verify whether this supplementation positively modifies the pathways involved in milk and milk component synthesis. Thus, by studying the transcriptome activity in these tissues, the effect of supplements of both vitamins on the metabolism of both liver and mammary gland, was investigated. For this study, 24 multiparous Holstein dairy cows were assigned to 6 blocks of 4 animals each according to previous 305-day milk production. Within each block, cows were randomly assigned to weekly intramuscular injections of 5 mL of either saline 0.9 % NaCl, 320 mg of vitamin B9, 10 mg of vitamin B12 or a combination of both vitamins (B9 + B12). The experimental period began 3 weeks before the expected calving date and lasted 9 weeks of lactation. Liver and mammary biopsies were performed on lactating dairy cows 64 ± 3 days after calving. Samples from both tissues were analyzed by microarray and qPCR to identify genes differentially expressed in hepatic and mammary tissues. Results Microarray analysis identified 47 genes in hepatic tissue and 16 genes in the mammary gland whose expression was modified by the vitamin supplements. Gene ontology (GO) categorizes genes in non-overlapping domains of molecular biology. Panther is one of the online GO resources used for gene function classification. It classifies the 63 genes according to Molecular Function, Biological Process and Protein Class. Most of the biological processes modulated by the vitamin supplements were associated to developmental process, protein metabolic process, transport and response to inflammation. In the liver, most of the genes modulated by the vitamin treatments involved protein metabolic process while developmental process appeared to be more affected by the treatments in mammary gland. Out of 25 genes analysed by qPCR, 7 were validated. Conclusion The results indicate that several metabolic processes were modulated by the supplementation of vitamins in early-lactating dairy cows. In addition, the results suggest that the vitamin supplements promoted liver regeneration and reduced catabolism of lipids in early lactation. Electronic supplementary material The online version of this article (doi:10.1186/s12864-016-2872-2) contains supplementary material, which is available to authorized users.


Background
Bacteria present in rumen synthesize B vitamins in generally sufficient amounts to meet their host's requirements [1]. Notwithstanding, high-producing dairy cows could benefit from vitamin B9 and B12 supplements, especially during the critical period around calving and in early lactation [2][3][4]. Vitamin B9 plays a major role in DNA synthesis and in de novo formation of methyl groups required for the methylation cycle. On the other hand, vitamin B12 is involved in two metabolic pathways: the remethylation cycle and as coenzyme of the methylmalonyl-CoA mutase. The former, closely related to folate metabolism, is required for the regeneration of methionine and tetrahydrofolate, whereas the later allows the entry of propionate in the Krebs cycle and gluconeogenesis [5].
Results from two studies suggest that supplementary vitamin B9 might improve efficiency of nutrient utilization, especially for milk protein synthesis [6,7]. This observation is supported by the observation that in vitro hormonal stimulation of milk protein synthesis by mammary gland explants of dairy cows increased expression of 28 genes; among them, 2 genes related to folate metabolism, FOLR1 and ALDH1L1 [8]. The former is a folate transporter allowing the entry of 5-methyl-tetrahydrofolate into the cells whereas the latter is involved in purine synthesis. As vitamin B9 supplement increases milk protein synthesis, it would be interesting to investigate whether similar transcriptomic process takes place in vivo when vitamin supplements are provided to lactating cows.
A combined supplement of vitamins B9 and B12 given during the peripartum period and in early lactation altered energy partitioning during the first weeks of lactation as compared to control cows [3,4,9]. However, the mode of action of this supplement is not fully elucidated. We observed that a combined supplement of vitamins B9 and B12 increases whole-body rate of appearance of glucose which is the sum of glucose from portal absorption, glycogenolysis, and gluconeogenesis [3]. Propionate originating from rumen fermentation is the major precursor of glucose in cows [10,11] and contributes up to 60 % of glucose flux rate [12,13]. For ruminants, the glucose, essential for synthesis of milk lactose, is mostly provided by liver (up to 90 %) through gluconeogenesis [14].
It has been shown that at the onset of lactation, the liver as well as the mammary gland undergo numerous adaptations to support milk synthesis [15]. In early lactation, feed intake is not sufficient to meet the nutrient demand for milk production, leading to a negative energy balance [15]. Furthermore, the liver undergoes extensive physiological and biochemical changes mediated by significant alterations in hepatic gene expression in an attempt to re-establish metabolic homeostasis and to counteract the adverse effects of negative energy balance [16]. In addition, during this period, the mammary gland is actively remodeling. In fact, by investigating the transcriptional response of the mammary gland during early lactation, Connor et al. [17] observed changes in mammary expression of genes involved in cell proliferation, cellular remodeling, and nutrient transport.
Therefore, the aim of the present project was to profile the genes for which expression undergoes major changes in hepatic and mammary tissues of lactating dairy cows according to vitamins B9 and B12 supply.

Results
At week 9 of lactation, there was no treatment effect (P ≥ 0.1) on dry matter intake, milk production, milk total solid yields as well as milk component contents. Dry matter intake, milk production, milk total solid yields, milk fat, protein and lactose contents averaged 18.8 (SE 1.9) kg/d, 34.7 (SE 2.9) kg/d, 3.92 (SE 0.33) kg/d, 36.0 (SE 2.0) g/kg, 30.2 (SE 1.4) g/kg, and 46.9 (SE 6.0) g/kg, respectively. During the seven weeks following calving, body condition score losses tended to be reduced (interaction vitamin × time, P = 0.10) in cows receiving the vitamin B9 supplements, alone or combined with vitamin B12. Plasma concentrations of non-esterified fatty acids and β-hydroxybutyrate were also lower (P ≤ 0.06), averaging 175 and 243 μM (SEM 17) and 0.70 and 0.77 mM (SEM 0.03) for cows receiving or not receiving supplementary vitamin B9, respectively.
The microarray analysis revealed that, as compared to control cows receiving no vitamin supplement, the vitamin treatments significantly changed (FDR ≤ 0.05) the expression of some genes by more than 2-fold in hepatic and mammary tissues ( Fig. 1 a and b, respectively). The expression of 47 genes in hepatic tissue was modified by at least one of the vitamin treatment compared to control cows (Fig. 1a). From this panel of 47 genes, 41 could be assigned to a biological process using the Panther classification system (Tables 1 and 2). Out of these 47 genes, expression of 14 genes was analyzed by qPCR (Fig. 2). Although nine false positive genes were observed, five genes were confirmed to be differentially expressed in the group of cows supplemented with vitamins.
The expression of two genes was significantly modified in liver by the B9 treatment as illustrated in the Venn diagram (Fig. 1a). One gene that also significantly downregulated by the B12 treatment was confirmed by qPCR for these cows. This repression of DLK1 in liver was down-regulated by the three vitamin treatments (Fig. 2). The B12 treatment had the greatest impact on the liver with 22 genes (totalizing 39 significant probes; Fig. 1a) whose expression level differed by more than 2-fold compared to control (Table 2).
Among the 22 genes affected by the B12 treatment, four genes were also influenced in cows receiving the B9 + B12 treatment: the methallothioneins 1A (MT1A) and 1E (MTIE), the transmembrane glycoprotein (GPNMB), and an uncharacterized protein (MGC126945) ( Table 1). Expression of two of these genes (MT1A and GPNMB) was studied by qPCR (Fig. 2). This analysis confirmed the repression of MTIA in liver of cows receiving B12 and B9 + B12 treatments. Results from the microarray analysis indicated an up-regulation of the expression of GPNMB with both B12 and B9 + B12 treatments. The qPCR analysis showed a numerical but not statistically significant increase with these treatments probably due to the large variation in the expression for this gene among the limited number of animals of this study (Fig. 2). Among the other 17 genes affected by the B12 treatment, the expression of seven genes were studied by qPCR but no effect were confirmed for 6 of them (HERC6, IF127, ISG15, MEP1B, PPP1R3B and SFRP1). Only the expression of the G0/G1switch 2 gene (G0S2) was up-regulated in liver of cows receiving the B12 treatment as compared to control (Fig. 2).
In addition to these four genes influenced by B12 treatment, alone and in combination with B9, 24 other genes were affected when both vitamins were administrated simultaneously to the cows; four genes were studied by qPCR. Two of them, namely the haptoprotein (HP) and lysyl oxidase-like 4 (LOXL4), were confirmed as being respectively down and up-regulated by the combined treatment. Nevertheless, the expression of LOXL4 was also up-regulated in liver of B9 cows whereas, the expression of HP was down-regulated by all vitamin treatments (Fig. 2). No difference in expression of GSTA5 and SLC22A9 could be detected.
In mammary gland, the B9 treatment modified the expression of nine genes by 2-fold as compared to cows receiving no vitamin supplement. The B12 treatment affected five genes and the expression of five genes was influenced by the B9 + B12 treatment (Fig. 1b). The B9 + B12 shared one gene with the B12 treatment. Based on results from the microarray analysis, three genes whose expression was modified by the B9 treatment (Fig. 1b), cell death-inducing DFFA-like effector a (CIDEA), the androgen binding protein beta-like (LOC785756), and the periostin (POSTN) genes were studied by qPCR in addition to some candidate genes (Fig. 3). Out of the 11 genes analyzed by qPCR, two genes were confirmed; then, the percentage of false positive in the mammary was 82 %. RAB15 and POSTN were significantly up-regulated respectively by the B12 and B9 as compared to control (Fig. 3). Out of the 16 genes affected in the mammary gland by vitamin supplements, 13 were classified using Panther classification system (Table 3).       As shown in Tables 1 and 3, some of the genes identified by microarray in liver and mammary gland were annotated and clustered into three major gene ontology groups: Protein Class, Molecular Function and Biological Process. In Tables 2 and 4, gene expression clusters are categorized within biological processes (levels 1, 2 and 3). This allows looking for statistically over-and under-represented biological process categories among the genes. Most of the genes modulated in the liver by at least one of the vitamin supplements were associated to developmental process, protein metabolic process, transport and immune response (Table 2). However, protein metabolic process was overrepresented (with 9 genes: MYOM1, HP, CDK5R1, MEP1B, LOXL4, FUT5, HERC6, LOC515676 and ISG15) in the gene list affected by treatments in the liver ( Table 2). In the mammary gland, where very few genes were affected by the vitamin treatments, the over-represented biological process was developmental process involving 4 genes: CIDEA, POSTN, GPR110 and IRX6 (Table 4).
Gene symbols were uploaded to the Panther workspace in order to classify the genes of interest by selecting the Bos taurus reference gene list based on the selected All genes modulated in the liver by the vitamins treatment, including those that were not found or classified by organism [18]. Notwithstanding, some genes in both tissues (liver and mammary gland) could not be assigned to any biological process or molecular function category by Panther classification system (Tables 1, 2, 3 and 4). It is also possible that there is no experimental data to support their biological annotation.

Discussion
Microarray analysis was used to investigate the effects of vitamin B9 and B12 supplements given alone or in combination during the peripartum early lactation period on liver and mammary gland tissues. The expression of only a limited number of genes was modulated by the vitamin treatments in both tissues, which suggests a subtle vitamin effect on the tissue metabolism that would have been better characterized using a larger population. Notwithstanding, the qPCR analyses show that, in liver, expression of keys genes, such as DLK1, LOXL4, G0S2, GSTA5, HP, MT1A, IFI27 and SFRP1 was modulated at different extents (significant effects for five genes and a strong trend for three genes) by the vitamin treatments. Delta-like 1 homolog (DLK1/Pref-1) is a surface marker of hematopoietic progenitor cells (HPCs) associated with less differentiated hepatocellular phenotypes [19] and it has been shown to act in vitro as an inhibitor of Notch signaling [20,21] to promote liver regeneration [21]. Interestingly, DLK1 is an imprinted gene which is involved in lipid metabolic reprogramming [22]. An increased concentration of this biomarker in blood serum is associated with hepatic cancer [23] whereas downregulation of DLK1 expression through an epigenetic mechanism contributes to attenuate liver disease [24]. Because vitamin B9 plays a major role in de novo formation of methyl groups and vitamin B12 is required for the remethylation cycle, we can speculate that DLK1 expression could also be repressed in the liver of the lactating cows through an epigenetic mechanism. Because DLK1 suppresses glucose production and fatty acid synthesis and oxidation in hepatocytes [25], supplementation of both vitamins B9 and B12 might increase liver metabolism through a genomic imprinting mechanism which negatively impacts the DLK1 pathway. Lysyl oxidase-like member 4 (LOXL4), a matrixremodeling enzymes, is extracellularly secreted and significantly contributes to ECM deposition [26]. Activity of lysyl oxidase (LOX) and LOX like proteins are correlated to collagen and elastin deposition and, in adult mammals, are essential to tissue maintenance [27]. Recent studies have provided compelling evidence that G0S2 is abundantly expressed in metabolically active tissues such as liver, and acts as a molecular brake on triglyceride catabolism [28]. Triglyceride hydrolase activity of adipose triglyceride lipase can be selectively inhibited by G0S2 [28]. Hence, increasing the expression of G0S2 decreased lipolysis [29] which is supported by the reduction of plasma concentrations of non-esterified fatty acids in cows receiving vitamin B9 supplements, alone or combined with vitamin B12 in the present study. Although increased IFI27 expression was not significant, the pattern was highly similar to G0S2 thus suggesting a similar B9 supplement effect on the liver for this gene. Expression of the alpha-inducible protein 27 (IFI27) is up-regulated during inflammatory wound repair process [30] and expression of this gene also alters immune response and mitochondrial function [31]. Interestingly, DLK1 locus expression is also associated with a restriction of the mitochondrial metabolism [32]. These gene expression patterns support the hypothesis that both vitamins B9 and B12 improve the hepatic function which might reduce metabolic stress during the transition period and early lactation of dairy cows. This is further supported by the marked reduction of the HP and MT1A genes. The liver is the major site for the synthesis of acute phase proteins including haptoglobin (HP) and metallothionein 1A (MT1A) [30]. During stress response, it is reported that physiological processes aimed on redistribution of energy utilization in specific organs stimulating mobilization of body reserves. In mammary gland, administration of the three vitamin treatments had a very limited effect on gene expression as described above. Interestingly, in the present study, whereas vitamin treatments had no effect on milk total solids yield and dry matter intake, vitamin B9 supplements, given alone or in combination with vitamin B12, decreased body condition score losses during the first weeks of lactation as well as plasma concentrations of non-esterified fatty acids and β-hydroxybutyrate [9] suggesting an improvement in energy balance for these cows.
In the present study, all the genes that have their expression affected by the vitamin treatments in the liver, are involved in tissue repair, resorption of inflammation and lipid metabolism although no mode of action can be clearly identified. During the first weeks of lactation, dairy cows are generally in negative energy balance because nutrient intake increases less rapidly than nutrient demand for initiation of lactation which leads to mobilization of body reserves. Cows are losing (See figure on previous page.) Fig. 2 Expression of genes measured by quantitative PCR in the liver of the dairy cows that received either no vitamin supplement: Control: no vitamin supplement; vitamin B9 supplement (B9); vitamin B12 supplement (B12) or a combined supplement of vitamins B9 and B12 (B9 + B12). Means different from the control treatment are indicated by *** when P value ≤ 0.01, ** when P values were between 0.01 and 0.05 and * when there was a trend with P values between 0.05 and 0.1 body condition score and non-esterified fatty acids are released from adipose tissues and their plasma concentrations increased. During this period, dairy cows are also prone to liver steatosis because hepatic uptake of non-esterified fatty acids is greater than the amounts oxidized or secreted by the liver [33]. Accumulation of lipids in liver affects integrity and function of hepatic cells [33]; in response, liver parenchymal cells produce an acute-phase glycoprotein haptoglobulin [34]. Because a decrease in DLK1 can improved fatty acid oxidation from hepatocytes [25], an improved β-oxidation of non-esterified fatty acids in liver could help to reduce ketone body formation and plasma concentrations of β-hydroxybutyrate. The improvement in energy balance observed in cows receiving supplementary vitamin B9, alone or combined with vitamin B12, likely reduced the liver burden caused by mobilization of body fat reserves which could explain the changes in hepatic gene expression described above. For instance, the increase of LOXL4 and G0S2 strongly support that these treatments protect body fat from catabolism. Prevention of liver damage or improved liver performance is not only important for maintaining liver function but also for general health of high-yielding dairy cows.

Conclusion
In the present study, a supplement of vitamin B9, given alone or in combination with vitamin B12, reduced mobilization of body fat reserves and hepatic lipid catabolism in early lactation. Changes in expression of genes described above support the hypothesis that hepatic tissue integrity in early lactation was improved by these vitamin supplements.

Animals and treatments
For the purpose of the present study, biopsies of hepatic and mammary tissues were taken from 24 multiparous Holstein cows from the dairy herd at the Agriculture and Agri-Food Canada Research Centre (Sherbrooke, Quebec, Canada) at the end of a larger study [9]. Care of cows followed the guidelines of the National Farm Animal Care Council (2009) [35]. Animals were kept in a tie-stall barn under 18:30 h of light per day (05:00 to 23:30 h) and milked twice daily (07:30 and 19:30 h). The experimental period began 3 weeks before the expected calving date and lasted until 9 weeks of lactation. The cows were fed ad libitum a close-up diet beginning 3 weeks before the expected date of calving until parturition and, then a lactation diet both formulated to meet or exceed the National Research Council (NRC) recommendations [36]. Long hay (0.5 kg) was given at 07:30 h and total mixed ration was served once daily at 08:30 h. Cows had free access to water.
Cows were assigned to 6 blocks of 4 animals each according to their 305-d milk production during the previous lactation. Within each block, cows were randomly assigned to one of the following treatments: weekly intramuscular injections of 5 mL of either saline 0.9 % NaCl (Control group), 320 mg of pteroylmonoglutamic acid (MP Biomedicals, Solon, OH, USA; (Vitamin B9 group), 10 mg of cyanocobalamin (5 000 μg/mL, Vetoquinol, Lavaltrie, Quebec, Canada; (Vitamin B12 group) or 320 mg of pteroylmonoglutamic acid and 10 mg of cyanocobalamin (B9 + B12 group). Thus, there were 6 animals per treatment group.

Biological material collection and tissue handling
Mammary gland and hepatic tissues were obtained from the lactating dairy cows, 64 ± 3 days after calving. Biopsies were performed under local anesthesia. The process of hepatic biopsies uses ultrasound guidance to minimize the hemorrhagic risks [4]. Both procedures were approved by the Institutional Committee on Animal Care of the Sherbrooke Research and Development Centre, Agriculture and Agri-Food Canada, Sherbrooke, QC, Canada according to the guidelines of the Canadian Coucil on Animal Care [37]. Tissues were immediately frozen into liquid nitrogen and stored at −80°C until use.

Total RNA isolation and purification
Total RNA was extracted from hepatic and mammary tissues by using a QIAzol Lysis Reagent (QIAGEN Inc., Toronto, ON, Canada) following the original manufacturer's protocol, with slight modifications. Briefly, frozen samples (100 mg of tissue) were homogenized in 2 mL of QIAzol Lysis Reagent on ice using a Tissue-Tearor. A volume of 600 μL QIAzol Lysis Reagent was added to 400 μL of homogenate; the mixture was vigorously vortexed and kept at room temperature for 5 min to promote dissociation of nucleoprotein complexes. A volume of 200 μL of chloroform was added; the mixture was shaken and left at room temperature for 3 min followed by a centrifugation at 12 000 × g for 15 min at 4°C to remove lipids. After centrifugation, the aqueous fraction (upper layer) was taken and RNA was precipitated by (See figure on previous page.) Fig. 3 Expression of genes in mammary gland measured by quantitative PCR according to treatments. Control: no vitamin supplement; B9: vitamin B9 supplement; B12: vitamin B12 supplement; B9 + B12: combined supplement of vitamins B9 and B12. Means different from the control treatment are indicated by *** when P value ≤ 0.01, ** when P values were between 0.01 and 0.05 and * when there was a trend with P values between 0.05 and 0.1  Differentially expressed genes in mammary tissue were assigned to gene ontology pathways using the Panther tool. Some genes were not found in Panther genes list. Furthermore, among the genes ID found by the PANTHER system, some were not classified to a category.  FlexArray, a Bioconductor R based software, was developed by Génome Québec to provide researchers with a user-friendly interface for the analysis of microarray experiments. Raw microarray expression intensities were corrected for background using normexp, according to Ritchie et al. [39]. Between-array normalization was performed so that the background corrected intensities have similar distributions across the arrays. Comparison of the vitamin treatments with the control tissue was performed by Cyber-T, a version of the t-test that uses a Bayesian estimate of the within treatment variance [40,41]. Up-or downregulated genes lists were analyzed on Protein ANalysis THrough Evolutionary Relationships (Panther) classification system [42].

Quantitative real time-PCR
Quantifications by real-time PCR (qPCR) following reverse transcription were performed as previously described [43] with minor modifications. The reverse transcription PCR reactions were performed with the SuperScript II reverse transcriptase (Life Technologies Inc., Burlington ON, Canada) according to the manufacturer's protocol and using 500 ng of total RNA extracted from each of the 24 animals. An equivalent quantity of cDNA is synthesized in a final reaction volume of 20 μL, giving a concentration of 25 ng/μl of cDNA. A pool of cDNA intended to estimate the efficiencies primers, was made using the cDNA of each animal. Primers were designed for each gene using the Primer Express 3 software package (Applied Biosystems, Life Technologies Corporation, Burlington, ON, Canada) using the reference sequence from the RefSeq database of the National Center for Biotechnology Information depository. Primers for a total of 34 genes for both tissues were designed. Optimizations of primers were performed for each gene by testing different concentrations of both forward and reverse primers, each ranging from 50 to 900 nM. Estimations of primer efficiencies were analyzed using the standard curves made from a serial of seven dilutions (1/7.5, 1/15, 1/30, 1/60, 1/ 120, 1/240, 1/480) of the pool of the cDNA samples (25 ng/μl of cDNA). As 3 μl of each dilution were used in a final reaction volume of 10 μl for qPCR, the concentrations used per dilution were: 1, 0.5, 0.25, 0.125, 0.0625, 0.03125 and 0.015625 (ng/μl). Additional file 1 provides experimental information and PCR amplification efficiency for all genes. The qPCR reactions (10 μL, final volume) were performed on 96 well plates using Fast SYBR Green PCR Master Mix (Applied Biosystems) in a 7500 Fast Real Time-PCR System (Life Technologies, Burlington, ON, Canada) as the manufacturer's instructions. The PCR thermal cycling conditions comprised an initial 20 s denaturation step at 95°C followed by 40 cycles at 95°C for 3 s followed by an annealing/elongation period at 60°C for 30 s. A dissociation step was included for all amplifications to confirm the presence of single discrete PCR products of the expected size. Twenty-five genes (14 from hepatic tissue data and 11 from mammary gland tissue data) were subject to qPCR validation because they were found differentially expressed by microarray as expressed by log2 of their fold change. In addition, in the mammary gland, four genes (FOLR1, ELF5, B4GALT1 and LALBA) were chosen because of their implication in the metabolic pathway involving vitamins B9 and B12. The expression of 5 putative reference genes, namely actin beta (ACTB), ubiquitously-expressed transcript (UXT), peptidylprolyl isomerase A (PPIA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), tyrosine 3-mono oxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (YWHAZ) was determined for all samples as recommended [44][45][46]. All the 24 animals were used to perform the qPCR analysis of 34 genes for both tissues. Once the more stable genes were selected, the normalization factor was calculated using a geometrical average as recommended [45]. The combination of ACTB and PPIA was appropriate to normalize the data from hepatic tissue. The appropriated combination was UXT and PPIA to normalize the data from mammary gland.

Data analysis and statistics
Using FlexArray microarray analysis, a significant result at a P value < 0.05 after false discovery rate correction with a minimum of ± two-fold change in gene expression for the respective treatment group compared to control tissues was considered biologically interesting. Gene category overrepresentation analysis consists in grouping genes into categories by some common biological property and then tested to find categories that are over represented amongst the differentially expressed genes. Gene ontology describes and categorizes gene products in three non-overlapping domains of molecular biology [47]. Panther (http:// pantherdb.org) is a visualisation browser of GO [48], using version 10 which included other organisms. Panther ranks proteins (and their genes) according to Family (and subfamily), Molecular function, Biological process and Pathway. The process of classification is extensively explained by Mi and colleagues [42]. Only significant differentially expressed genes were analysed by qPCR. Data from qPCR were analyzed using SAS Institute procedures (2008). Means were assumed to be different at P ≤ 0.05 and tended to differ at 0.05 < P ≥ 0.1. Normfinder indicated the interested combination of reference genes to normalize qPCR data.