Gene expression profiling in hepatic tissue of newly weaned pigs fed pharmacological zinc and phytase supplemented diets

Background Zinc (Zn) is an essential trace element. However, Zn bioavailability from commonly consumed plants may be reduced due to phytic acid. Zn supplementation has been used to treat diarrheal disease in children, and in the U.S. swine industry at pharmacological levels to promote growth and fecal consistency, but underlying mechanisms explaining these beneficial effects remain unknown. Moreover, adding supplemental phytase improves Zn bioavailability. Thus, we hypothesized that benefits of pharmacological Zn supplementation result from changes in gene expression that could be further affected by supplemental phytase. The goal of this study was to investigate the effects of feeding newly weaned pigs dietary Zn (150, 1,000, or 2,000 mg Zn/kg) as Zn oxide with or without phytase [500 phytase units (FTU)/kg] for 14 d on hepatic gene expression. Liver RNA from pigs fed 150, 1,000, or 2,000 mg Zn/kg, or 1,000 mg Zn/kg with phytase (n = 4 per treatment) was reverse transcribed and examined using the differential display reverse transcription polymerase chain reaction technique. Liver RNA from pigs fed 150 or 2,000 mg Zn/kg (n = 4 per treatment) was also evaluated using a 70-mer oligonucleotide microarray. Results Expressed sequence tags for 61 putatively differentially expressed transcripts were cloned and sequenced. In addition, interrogation of a 13,297 element oligonucleotide microarray revealed 650 annotated transcripts (FDR ≤ 0.05) affected by pharmacological Zn supplementation. Seven transcripts exhibiting differential expression in pigs fed pharmacological Zn with sequence similarities to genes encoding GLO1, PRDX4, ACY1, ORM1, CPB2, GSTM4, and HSP70.2 were selected for confirmation. Relative hepatic GLO1 (P < 0.0007), PRDX4 (P < 0.009) and ACY1 (P < 0.01) mRNA abundances were confirmed to be greater in pigs fed 1,000 (n = 8) and 2,000 (n = 8) mg Zn/kg than in pigs fed 150 (n = 7) mg Zn/kg. Relative hepatic HSP70.2 (P < 0.002) mRNA abundance was confirmed to be lower in pigs fed 2,000 mg Zn/kg than in pigs fed 150 or 1,000 mg Zn/kg. Conclusion Results suggest that feeding pharmacological Zn (1,000 or 2,000 mg Zn/kg) affects genes involved in reducing oxidative stress and in amino acid metabolism, which are essential for cell detoxification and proper cell function.


Background
Differential gene expression is responsible for morphological and phenotypical differences among animals, since the transcriptome is dynamic [1]. Studies on how micronutrients affect gene expression will help to clarify the role of trace elements in health and metabolism and their connection to biochemical events [2]. The trace mineral zinc (Zn) is involved in gene expression in numerous ways including DNA replication, RNA transcription, through the activity of transcription factors, DNA and RNA polymerases, signal transduction, oxidative stress and playing a role in programmed cell death [3]. It was recently estimated that almost 3,000 different proteins encoded by the human genome bind zinc [4]. An example of such proteins is metallothionein (MT), which exerts a protective effect against stress by acting as an antioxidant, and as a Zn storage and transfer protein [5]. Furthermore, the transcriptional regulation of the MT gene by dietary Zn has been demonstrated in rats [6].
In developing countries Zn deficiency is caused by ingestion of high cereal protein diets rich in phytic acid, an organic form of phosphorus (P) [7]. In addition, common animal feedstuffs contain 60-80% of the total P in the phytic acid molecule, which make Zn unavailable for absorption [8]. Rats fed diets containing 3,000 mg Zn/kg and phytic acid absorb less and excrete more Zn [9]. This may be explained by the formation of mineral-phytic acid complexes, plus the negligible amount of endogenous phytase production in mammals. Therefore, the tendency to excrete phytic acid bound minerals in the feces is increased, and mineral absorption is decreased. Phytase supplementation studies in pigs demonstrate increased plasma Zn concentration [10,11], greater Zn retention [12], and improved bone mineralization [13], while adding phytase to human diets also results in increased iron absorption [14]. Thus, phytase supplementation provides a potential solution to improve Zn availability.
In humans, Zn supplementation has been shown to reduce the morbidity of diarrhea [15], respiratory disease [16], and nematode infection [17]. Moreover, high Zn supplementation to weaned pigs has been successful in increasing growth and decreasing diarrhea [18], while feeding pharmacological Zn (3,000 mg Zn/kg) to weaned pigs improves fecal consistency and also intestinal morphology by increasing villous height and reducing crypt depths of the duodenum and jejunum [19]. However, molecular analyses of the effects of pharmacological Zn to explain the favorable symptoms observed when Zn is supplemented in high quantities have not been undertaken.
We previously reported that feeding the recommended amounts of phytase (500 FTU/kg) and pharmacological Zn (1,000 -2,000 mg/kg) to newly weaned pigs increased MT mRNA abundance and protein concentrations in the liver, kidney and intestinal mucosa [11]. Thus, additional genes are also likely to be affected by this dietary intervention and serve as the basis for the improved health effects obtained with pharmacological Zn supplementation. Therefore, the objective of this experiment was to determine the identity of genes that are differentially expressed in the liver of pigs fed diets with or without pharmacological Zn, and to determine if phytase supplementation further affected gene expression by using differential display reverse transcription polymerase chain reaction (DDRT-PCR) and oligonucleotide microarray technologies. In this paper we report 61 amplicons identified by DDRT-PCR and 650 annotated genes (FDR ≤ 0.05) identified by microarray analysis with putative differential gene expression, of which 5 transcripts with roles in oxidative stress reduction and amino acid metabolism were confirmed to have differential expression in pigs fed pharmacological Zn diets.

Identification of expressed sequences displayed by DDRT-PCR
A total of 66 putatively differentially expressed amplicons were cloned and sequenced. High quality nucleotide sequence data was obtained for 61 of these expressed sequence tags (ESTs) and all were submitted to the Gen-Bank database [GenBank:CB826605 -CB826610, Gen-Bank:CF106636 -CF106687, GenBank:DW177047 -DW177050]. A search of the EMBL/GenBank and TIGR databases revealed that 29 of the ESTs (48%) had significant similarities to genes with known identities, 7 were similar to mitochondrial DNA (11%), and the 25 remaining sequences (41%) had unknown identities (Table 1). Identified genes were similar to genes involved in signal transduction (48%), nucleic acid metabolism (17%), oxidative stress response (11%), protein and amino acid metabolism (10%), blood coagulation (7%), and structural roles (7%). Various mRNA abundance patterns were observed among the dietary treatment groups (Table 1) and 5 genes were selected for subsequent confirmation analyses.

Identification of differentially expressed genes by oligonucleotide microarray analysis
To further monitor gene expression changes due to pharmacological Zn supplementation, we conducted a transcriptional profiling analysis using a commercially available pig 70-mer oligonucleotide set spotted on glass slide microarrays. The hepatic transcript abundance patterns from animals fed 150 mg Zn/kg (Zn 150 ) compared to animals fed 2,000 mg Zn/kg (Zn 2000 ) revealed 683 annotated genes that were significantly altered at a false discovery rate (FDR) adjusted P ≤ 0.05, plus several hundred additional un-annotated oligonucleotides. Several of  these probes represent multiple spots of the same oligonucleotide. Taking this into account, a total of 650 unique genes were identified as differentially expressed at FDR ≤ 0.05 in the microarray analysis [see Additional file 1]. These genes have been categorized according to functional groups, and the file contains the oligonucleotide ID, fold change, raw and adjusted P values, and annotation for each gene.

Confirmation of differential expression as affected by dietary treatment
In the present study, we focused on genes differentially expressed due to pharmacological Zn supplementation, and further evaluated if their expression was affected by phytase supplementation. A total of 7 genes were selected from the display gels and microarray hybridization results that exhibited different relative mRNA abundance in the liver of pigs fed pharmacological Zn diets without phytase (Zn 1000 , Zn 2000 ) versus pigs fed Zn 150 . In addition to these 7 genes, an oligonucleotide for MT on the microarray was observed to exhibit a 3.8-fold higher hepatic mRNA abundance in pigs fed Zn 2000 than in pigs fed Zn 150 , a result which we have already confirmed in our previous study where we reported hepatic MT mRNA abundance and protein concentration to be significantly higher in pigs fed pharmacological Zn diets than in pigs fed an adequate Zn diet [11]. To confirm differential gene expression, a combination of relative real time RT-PCR and northern blot analyses were utilized with all 7 animals from Exp. 1 (fed Zn 150 or Zn 2000 without phytase), and all 23 animals of Exp.2, which were fed the six dietary treatments.
Three oxidative stress response genes involved in the reduction of peroxides [peroxiredoxin 4 (PRDX4)], detoxification of glycating agents [glyoxalase I (GLO1)], and transfer of glutathione [glutathione transferase M4 (GSTM4)], exhibited differential expression. Northern blot analysis of PRDX4 revealed a single transcript of 0.95 kb (Figure 1), which agrees with the size reported for human PRDX4 [20]. Real time reverse transcription PCR (RT-PCR) was used to examine mRNA abundance of PRDX4. Relative abundance of PRDX4 mRNA ( Figure 2A) in liver of pigs fed pharmacological Zn diets was 2 fold higher in pigs fed Zn 1000 , and 4 fold higher in pigs fed Zn 2000 when compared to pigs fed Zn 150 (P < 0.009). This result was supported by the subsequent microarray analysis which revealed significantly higher hybridization to a PRDX4 oligonucleotide present on the array by the Zn 2000 samples than by the Zn 150 samples (data not shown).
PCR amplification efficiency plots revealed that GAPDH was not a suitable normalizing gene for GLO1, therefore northern blot hybridization was performed for confirmation of GLO1. A single transcript (~1.9 kb) was observed ( Figure 1), which was in accordance with the size of human GLO1 [21]. Relative GLO1 mRNA abundance was greater (P < 0.0007) in pigs fed pharmacological Zn diets (Zn 1000 and Zn 2000 ) when compared to pigs fed adequate Northern blots of GLO1, ACY1 and PRDX4 Figure 1 Northern blots of GLO1, ACY1 and PRDX4. Northern blot analysis of liver GLO1, ACY1 and PRDX4 mRNA of pigs fed 150, 1,000 or 2,000 mg Zn/kg (with or without phytase) for 14 d post-weaning (Exp.2). Liver total RNA (12 μg) isolated from each pig was analyzed using the cDNA clones derived from differential display gels as probes. Also shown is 18S rRNA hybridization to assess equivalence of RNA loading. Figure 2 Gene expression confirmation of DDRT-PCR and microarray transcripts. Relative mRNA abundance of selected transcripts was confirmed using real time RT-PCR or nothern blot analyses in pigs fed 150, 1,000 or 2,000 mg Zn/kg for 14 d post-weaning. A) PRDX4 mRNA Abundance. Relative real time RT-PCR was performed and fold changes relative to GAPDH and a common Zn 150 reference sample are presented. Values are means ± SEM, n = 8, (except Zn 150 , n = 7). A significant zinc effect was detected for relative PRDX4 mRNA abundance, P < 0.009. No Zn by phytase interaction, or phytase effect were detected. B) GLO1 mRNA Abundance. Northern blot analysis was performed and values are mean optical density readings ± SEM, n = 8 (except Zn 150 , n = 7). A significant zinc effect was detected for relative hepatic GLO1 mRNA abundance, P < 0.0007. No Zn by phytase interaction, or phytase effect were detected. C) ACY1 mRNA Abundance. Northern blot analysis was performed and values are mean optical density readings ± SEM, n = 8 (except Zn 150 , n = 7). A significant zinc effect was detected for relative hepatic ACY1 mRNA abundance, P < 0.01. No Zn by phytase interaction, or phytase effect were detected. D) HSP70.2 mRNA Abundance. Relative real time RT-PCR was performed and fold changes relative to Zn 2000 using 18S as normalizing gene are presented. Values are means ± SEM, n = 7, (except Zn 2000 , n = 8). A significant zinc effect was detected for relative HSP70.2 mRNA abundance, P < 0.002. No Zn by phytase interaction, or phytase effect were detected. Zn ( Figure 2B). Supplemental phytase did not affect the mRNA abundance of either PRDX4 or GLO1.

Gene expression confirmation of DDRT-PCR and microarray transcripts
Microarray analysis had indicated GSTM4 mRNA abundance to be higher in pigs fed Zn 2000 than pigs fed Zn 150 . Real time RT-PCR results for GSTM4 supported a trend for Zn 2000 pigs to have higher GSTM4 mRNA abundance, although it did not reach statistical significance due to large animal variability. There was a trend for significance (P < 0.09) towards an interaction between Zn and phytase supplementation for relative GSTM4 mRNA abundance. GSTM4 mRNA abundance tended to increase when phytase was supplemented to the diets of pigs fed adequate (Zn 150 P) or the first level of pharmacological Zn (Zn 1000 P). However, when phytase was supplemented with the higher level of pharmacological Zn (Zn 2000 P) no further increase in expression was observed.
N-aminoacylase I (ACY1) and heat shock protein 70.2 (HSP70.2) are genes involved in amino acid and heat shock metabolism that also displayed differential expression. Northern blot analysis revealed the presence of a single ACY1 transcript (~1.3 kb, Figure 1), which is in accordance with the size reported for pig ACY1 [22]. Relative ACY1 mRNA abundance was increased (P < 0.01) in pigs fed pharmacological Zn (Zn 1000 and Zn 2000 ), and no phytase effect was observed ( Figure 2C). Relative abundance of HSP70.2 mRNA in liver of pigs fed Zn 150 and Zn 1000 diets was comparable and higher than in pigs fed Zn 2000 ( Figure 2D). Supplemental phytase did not affect the mRNA abundance of HSP70.2.
Based on the differential display gels, an acute phase response gene, orosomucoid 1 (ORM1), and a gene in the plasma carboxypeptidase family, carboxypeptidase B2 (CPB2), also exhibited increased mRNA abundance in pigs fed pharmacological Zn diets. However, these genes were not confirmed to be differentially expressed in pigs fed any of the dietary treatments using relative real time RT-PCR due to large animal variability (data not shown).

Discussion
Differential display reverse transcription PCR is a technique that allows the simultaneous comparison of multiple treatments for the identification of differentially regulated mRNAs, and the generation of ESTs that can be compared with DNA sequences available in current databases to identify genes, including genes in non-conventional biological models such as the pig [23,24]. The disadvantage of DDRT-PCR vs. microarray approaches is clearly that it is a much more time consuming technique to perform. However, DDRT-PCR does have some advantages. Gene discovery with DDRT-PCR does not require prior knowledge of gene or EST sequences as is needed for construction of microarrays [25]. In the present study, sev-eral genes selected for confirmation from the DDRT-PCR study were not present on the microarray used for subsequent analyses despite the relatively large number of oligonucleotides on the microarray (over 13,000). In addition, DDRT-PCR allows direct comparisons to be made between more than two samples at a time and it may be more sensitive for detection of relatively low abundance transcripts. For this study, DDRT-PCR and oligonucleotide microarray technologies were used to reveal differentially expressed genes in the liver of pigs fed pharmacological Zn diets and to determine if expression was further affected when supplemental phytase was added to the diets. In total we found 61 ESTs with putative differential expression using DDRT-PCR among the various treatment groups and identified 650 differentially expressed (FDR ≤ 0.05) annotated genes with the microarray. Seven genes were selected for confirmation. This study provides evidence of a differential hepatic transcriptional profile in pigs due to consumption of diets containing basal or pharmacological concentrations of Zn.
The PRDX4 gene encodes for a ubiquitous cytosolic enzyme [EC 1.11.1.15], which reduces hydrogen peroxide via redox-active cysteine residues. In addition to hydrogen peroxides, Jin and collaborators [20] reported that peroxiredoxins can also regulate peroxide-mediated signaling cascades, and their overexpression reduces hydrogen peroxides in response to tumor necrosis factor α (TNF-α).
Overexpression of PRDX4 in mice suppresses B-cell nuclear factor of kappa light polypeptide gene enhancer (Nf-κB) activation through regulation of B-cell inhibitor of kappa light polypeptide gene enhancer (I-κB) phosphorylation [20]. Gallagher and Phelan reported that inhibition of Nf-κB leads to an increase in PRDX6 transcription in mouse hepatocytes [26]. Similarly, MT has also been implicated in inhibiting TNF-induced activation of Nf-κB DNA binding, by inhibition of I-κB degradation and subsequent NF-κB suppression [27]. An in vitro study demonstrated that Zn pretreatment increased the availability of intracellular Zn ions and abrogated LPS-induced oxidative stress and I-κB/NF-κB signaling cascade in liver Kupffer cells [28]. Kim et al. [29] suggest that MT modulates intracellular signaling molecules such as transcription factors (Nf-κB) by sequestering intracellular Zn, which is required for its DNA binding activity. Peroxiredoxins are considered redox regulators of signal transduction [20], acting like a switch to control intracellular pathways. Perhaps the increase in intracellular Zn causes an alteration in cell signaling [30], and PRDX4 upregulation occurs as a result. In addition, the suppressive effect of PRDX4 on Nf-κB may be one of the many mechanisms of how pharmacological Zn, as ZnO enhances health in the newly weaned pig. For example, studies in Caco-2 cells exposed to enterotoxigenic E. coli demonstrate that Zn as ZnO protects these cells by inhibiting bacterial adhesion and internalization into cells, thus modulating subsequent cytokine expression that might occur in response to pathogenic exposure [31].
Glyoxalase 1 [EC 4.4.1.5] is a cytosolic ubiquitous Zn metalloenzyme that catalyzes the glutathione dependent conversion of glycating agents such as methylglyoxal to S-D-lactoylglutathione via a 1,2-hydrogen transfer [32]. This enzyme works in conjunction with glyoxalase II to further convert S-D-lactoylglutathione to D-lactate [33]. Methylglyoxal can react with arginine and lysine residues in proteins, and at high concentrations it inhibits glycolytic enzymes [21], enhances lipid peroxidation, and binds DNA and RNA causing macromolecular damage [34]. Genomic analysis of the GLO1 gene revealed the presence of functional metal response elements (MREs) located 647 bp downstream of the transcription initiation site [21]. Another gene containing MREs in its promoter region is the MT gene [35]. The MREs control transcriptional activation of MT upon exposure to Zn, through the action of metal transcription factor-1 [36]. Transcriptional activity of GLO1 in transfected HepG2 cells was elevated twofold upon exposure to 25 and 75 μM ZnCl 2 for 48 hr [21]. We have previously shown that MT mRNA abundance and protein concentration are increased in liver, kidney and intestinal mucosa of pigs fed pharmacological Zn [11]. This observation was confirmed for liver MT mRNA abundance with the microarray analysis in the present study and similar expression patterns were observed for PRDX4 and GLO1.
Zinc involvement in oxidative stress is well known. Its deficiency causes oxidative DNA damage by impairment of antioxidant defense, and compromises DNA repair mechanisms [37]. However, the mechanism of how supplemental Zn exerts its antioxidant action is not well defined. Our current microarray analysis indicated that mRNA abundances of many oxidative stress molecules are affected by pharmacological Zn supplementation including GSTM4, glutathione peroxidase (GPX), NADPH oxidase (NOX1), heme oxygenase 2 (HMOX2), cytochrome C oxidase VIIB (COX7B), cytochrome P450 (CYPC27), and peroxisome proliferator activated receptor gamma-2 (PPARG) [see Additional file 1]. From these, GSTM4 was selected for confirmation. GSTM4 encodes the fourth member of the glutathione S-transferase that belongs to the mu class. The mu class of enzymes function in the detoxification of electrophilic compounds, including carcinogens, therapeutic drugs, environmental toxins and products of oxidative stress, by conjugation with glutathione [38]. Our results are in agreement with previous studies where dietary Zn affected the hepatic gene expression of GST, GPX and CYP [39]. In addition to these genes, it has been suggested that MT induction by Zn in hepatocytes creates an intracellular pool of thiols and Zn ions, thus it can act similarly to glutathione which can then react with the reactive species [40,41]. Moreover, the ability of Zn to occupy iron and copper binding sites on lipids, proteins and DNA, may also contribute towards its antioxidant action [42]. The present study shows that pharmacological Zn supplementation is associated with increased mRNA abundance of PRDX4 and GLO1, and a tendency to increase GSTM4 when combined with dietary phytase. Interestingly these genes participate in the oxidative stress response through differing mechanisms. Perhaps the increased intracellular concentrations of Zn exert an increase in Zn dependent enzymes creating an effective hepatoprotective role. Our results for hepatic GLO1 expression are in agreement with observations in GLO1 transfected HepG2 cells [21]. It is possible that Zn transcriptionally regulates GLO1 through MRE activation, in a similar fashion to the MT gene, therefore offering protection against glycating agents. However, to our knowledge, we provide the first evidence that dietary pharmacological Zn increases PRDX4 and GLO1 mRNA abundances. Additional studies will be undertaken to further clarify the influence of pharmacological Zn on GSTM4 and related oxidative stress protein activity in pigs.
Molecular chaperones or stress proteins like heat shock proteins 70.2 and 90 (Hsp70 and Hsp90), maintain the correct conformational homeostasis of proteins, helping newly synthesized proteins to be folded and transported across membranes and protecting them against a number of stresses thus playing the role of chaperones [43,44]. The expression of heat shock proteins is under complex control, but with final common pathways involving the heat shock factors (HSFs), the function of which is zincdependent [45]. Several heat shock proteins have been identified including Hsp27, Hsp47, Hsp70 and Hsp90, and are named according to their molecular weight [46]. In our present study we were able to confirm that pharmacological Zn decreases HSP70.2 mRNA abundance in the liver. Interestingly, in vitro studies [47] which examined the relationship between protein misfolding, aggregation and chaperone induction showed that oxidative stress increased expression of Hsp90. However, mild oxidative stress did not induce protein denaturation and stress response. This work also showed that adding Zn induced chaperone expression and higher Zn concentrations counteracted the heat shock and oxidative stress mediated induction of chaperones by inhibiting their expression [47]. These studies support our findings that at higher Zn concentrations, there is lower abundance of HSP70.2 mRNA.
Three additional genes were selected for confirmation. ACY1 encodes for a Zn metalloenzyme involved in peptide metabolism, ORM1 encodes for an acute phase reactant protein and CPB2 encodes for a plasma carboxypeptidase. ACY1 [EC 3.5.1.14] is a cytosolic enzyme with highest abundance in pig kidney versus the liver [22,48]. The biological role of ACY1 is to hydrolyze neutral and hydrophobic α-N-acyl-L-amino acids generated during protein degradation [49]. This Zn dependent enzyme has been identified in mammals primarily in kidney and liver [48], it contains a single Zn atom per subunit, and it is proposed to have catalytic roles [50]. However, Heese et al. [51] suggested that Zn plays a structural role, because the Zn binding site was too far from the catalytic site of the enzyme. Results of northern blot analysis in this study confirm that pharmacological Zn supplementation increases the relative abundance of ACY1 mRNA. No differences were observed in the abundance of this transcript between animals fed Zn 1000 and Zn 2000 , suggesting that at dietary concentrations greater than 1,000 mg Zn/kg, relative ACY1 abundance reaches a plateau. To our knowledge, this study provides the first evidence of Zn effects on the abundance of ACY1 mRNA, suggesting a role in protein and amino acid metabolism.
Orosomucoid 1 and CPB2 are genes that encode an acute phase reactant protein known as α-1-acid glycoprotein, and plasma pro-carboxypeptidase [EC 3.4.17.20], respectively. These plasma proteins are secreted into the blood after being synthesized in the liver [52,53]. Serum concentrations of ORM1 are high in fetal [54] and neonatal pigs, decreasing markedly to a constant level by 2 wk of age [55] and declining further by 112 d. In a similar pattern, liver ORM1 mRNA is relatively abundant in late fetal and neonatal pigs then declines rapidly after birth [56]. Alpha-1-acid glycoprotein plasma concentration rises during inflammation, thus its function is related to modulation of an inflammatory response primarily caused by interleukin 1, interleukin 6 and glucocorticoids [57]. Sequence analysis of the murine ORM1 gene revealed the presence of four MRE sequences in the 5' flanking region, and one in intron 5 [58]. Furthermore, relative liver ORM1 mRNA abundance in mice injected i.p. with the heavy metal mercury (0.5 mg Hg/kg) was increased in a time dependent manner. Zinc was also used to investigate ORM1 mRNA response upon heavy metal exposure, but its effect on ORM1 mRNA was not as marked as Hg [59]. The authors concluded that heavy metals regulate ORM1 at the transcriptional level. Interestingly, we found this gene to be affected by dietary treatment in both experiments using different anchor and arbitrary primer combinations, although the pattern of transcript abundance differed. We did not confirm a significant effect of dietary Zn on liver ORM1 mRNA abundance for pigs in either experiment using relative real time RT-PCR.
Plasma CPB2 [E.C. 3.4.17.20] is a carboxypeptidase that appears in plasma after clotting of blood [60]. Several names have been used for this enzyme according to its general catalytic mechanism, including carboxypeptidase U (U = unstable), carboxypeptidase R (R = cleaves Arg and Lys residues in C-termini), plasma carboxypeptidase B (B = basic) and most recently thrombin-activatable fibrinolysis inhibitor (TAFI) [61]. Eaton et al. [62] showed that CPB2 becomes inactive upon addition of a Zn chelating agent, 1,10-phenantroline. The presence of CPB2 has been demonstrated in the pig, rabbit and mouse, with activity ranging from 20% (mouse) to 500% (pig) compared to human serum CPB2 [63]. The physiological function of CPB2 is to inhibit fibrinolysis by removing Cterminal lysine residues from fibrin and limiting plasmin formation [64]. Fibrinogen is also secreted during an inflammatory response, suggesting that overexpression of CPB2 could be part of a mechanism to decrease inflammation.
Our display gels indicated that pharmacological Zn increases the mRNA abundance of ORM1 (for Exp. 2 pigs) and CPB2 in pig liver. However, relative real time PCR results did not confirm their differential expression. The inability to confirm the differential display observations may be due to the low number of pigs examined along with relatively high variability in transcript abundance between pigs or the increased sensitivity of the real time RT-PCR assays.

Conclusion
The EST data obtained and submitted to the GenBank database as well as the transcriptional profiling from the microarray study provide sequence and differential gene expression information for hepatic genes in the pig. In addition, this study provides important information that contributes to understanding the benefits of feeding pharmacological Zn diets to pigs. New mechanisms affected by dietary Zn were evaluated and can be further considered in future studies, including the role of Zn in reducing oxidative stress and protein metabolism, both processes that are essential for cell detoxification and function.

Animals and diets
Animals and diet composition have been previously described [11]. A smaller preliminary study was performed to optimize the technique (Exp. 1).

Sample collection and total RNA isolation
Details about euthanasia, liver sample collection and total RNA isolation were previously published [11]. Liver samples were obtained from 3 pigs per treatment for Exp. 1 (n = 6 total) and from 4 pigs per treatment for Exp. 2 (n = 24 total). RNA quality and integrity were determined by calculating the A 260/280 ratio and by agarose gel electrophoresis, respectively. Additionally, RNA quality and quantity were further confirmed with the RNA 6000 Pico LabChip ® kit using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Differential display reverse transcription polymerase chain reaction (DDRT-PCR)
DDRT-PCR experiments were performed as previously reported by our laboratory [65] using modifications of published procedures [23]. Briefly, genomic DNA contamination was minimized by treating 1 μg of total RNA with 1 U of amplification grade DNAse I ( Oligonucleotide primers for DDRT-PCR were obtained from the U.S. Pig Genome Coordination Program http:// www.animalgenome.org/pigs/resources/ddprimer.html, and those used for these experiments were randomly selected ( Denaturing loading dye (96% formamide, 2% bromophenol blue and 2% xylene cyanol) was mixed with each DDRT-PCR sample (8 μl), and samples were dried on medium heat for 5 min using a speed-vacuum (Savant Instruments Inc., Farmingdale, NY) followed by a 3 min 95°C denaturing step. The samples were electrophoresed on 0.4 mm 5.2% polyacrylamide denaturing gels. Following a 5-6 hr run at 60 W on a vertical gel box (Model S2, Invitrogen, Life Technologies/GIBCO BRL Sequence Systems, Carlsbad, CA), gels were transferred onto Whatman chromatography paper (Whatman ® , Maidstone, England) and dried at 80°C for 30 min using a slab dryer (SGD 2000, Savant). Gels were exposed to Biomax™ film (Eastman Kodak Co., Rochester, NY) overnight.

Excision and re-amplification of DDRT-PCR products
Radiographs were developed and subsequently carefully evaluated to detect differences in transcript abundance by visually comparing bands among the different dietary treatments. Transcripts were selected that exhibited a consistent band intensity among samples within treatments and different intensities across treatments. A band intensity scale to describe the pattern of transcript abundance was created by assigning a representative number to each dietary treatment (1 = Zn 150 ; 2 = Zn 1000 ; 3 = Zn 1000 P; 4 = Zn 2000 ) observed in the display gel from highest (>) to lowest or comparable (=) abundance (Table 1). Selected bands were circumscribed from the gels, re-hydrated in 100 μl DEPC-treated water and heated at 50°C for 30 min. Re-amplification reactions included 2 μl of gel band eluate and the anchor and arbitrary primers used in the DDRT-PCR step under the same reaction conditions and cycling parameters described above, excluding the iso-

Primer identification number Primer sequence
Anchor # 2 d(T) 12   tope. To assess quality of the reactions, 1 and 2 μl of reamplified products were electrophoresed with a λ Hind III marker (Invitrogen, Life Technologies, Carlsbad, CA) in 1% agarose gels stained with ethidium bromide (Sigma-Aldrich, St. Louis, MO).  [67]. Specifically, the "normexp" function advocated by Smyth [67] was used for background subtraction followed by print-tip specific loess normalization for dye intensity bias for each array. The Cy5:Cy5 log2 ratios were further scale normalized between arrays. A simple linear model fitting these ratios as a function of an overall intercept(dye) and treatment effects was fitted for each gene using the "eBayes" option in limma. Statistical significance was based on the estimated false discovery rates (FDR) as also provided by limma. Relative quantification was determined using duplicate cDNA samples from each animal. PCR amplification efficiency plots were generated using serially diluted cDNA (4 dilutions) to confirm the use of GAPDH or 18S as normalizing genes for each primer pair. Results were recorded relative to a common liver cDNA sample from a pig fed Zn 150 after normalizing for GAPDH or 18S. Relative gene expression changes were then computed using the 2 -ΔΔCt method [68].

Independent confirmation by northern blot hybridization
When PCR amplification plots did not reveal equal amplification efficiency for GAPDH with the target gene, confirmation of DDRT-PCR results for some genes was done by northern blot analysis using the cDNA clones obtained by DDRT-PCR, labeled as described previously [11] with an 18S rRNA for normalization.

Statistical analysis
The data were analyzed using analysis of variance based on the MIXED procedure of SAS [69]. If the main effects (Zn or phytase) or their interaction were not statistically significant, the datasets were merged. For the relative real time RT-PCR analyses, the model included pig nested within the treatment interaction with the plate as random effect. Satterthwaite's approximation was used to determine the error df for test [70]. A base 2 logarithmic transformation (Log 2 ) of the target and normalization gene Cts was made to account for the exponential amplification of cDNA prior to analyzing Ct values with the MIXED proce-dure of SAS. The transformed target Cts, and GAPDH or 18S Cts modeled as covariates, were used in a regressionbased normalization of Cts in accordance with recommendations by Poehlman [71]. These values were compared to the results calculated by using the 2 -ΔΔCt method [68], and data is presented as fold changes. For northern blot analyses, the blot by treatment interaction was used in the model with pig as a random effect. The 18S rRNA values were modeled as covariates for a regression-based normalization of the genes' relative mRNA abundance. Differences were considered to be significant when P < 0.05.

Authors' contributions
MMM performed the animal study, prepared and analyzed the diets, helped in tissue collection, carried out part of experiment 2 DDRT-PCR reactions, performed most of the bioinformatics and statistical analyses, performed part of the confirmation assays and prepared the manuscript. GMH obtained the animals for this study, helped in obtaining the tissues, designed and coordinated the nutrition study as the PI, and helped draft the manuscript. NER helped in tissue collection, performed experiment 1 DDRT-PCR analysis, conducted part of the real time RT-PCR confirmation, and helped draft the manuscript. VDR performed part of the real time RT-PCR confirmation and conducted some of the real time RT-PCR statistical analyses. RJT performed statistical analysis of the microarray data and assisted with analysis of real time RT-PCR data. JEL prepared and analyzed the nutritional composition of the diets and helped with tissue collection. CJW carried out part of experiment 2 DDRT-PCR reactions. AMR carried out the microarray hybridizations. CWE helped in obtaining the tissues, designed and coordinated the  DDRT-PCR and microarray studies as the PI, and prepared the manuscript. All authors have read and approved the final manuscript.