Department of Animal Sciences, University of Illinois, Urbana, Illinois
ABSTRACT
Long-term molecular adaptations in liver from high-producing dairy cows are virtually unknown. Liver from five Holstein cows was biopsied at 65, 30, 14, +1, +14, +28, and +49 days relative to parturition for transcript profiling using a microarray consisting of 7,872 annotated cattle cDNA inserts. More than 5,000 cDNA elements represented on the microarray were expressed in liver. From this set we identified 62 differentially expressed genes related to physiological state, with a false discovery rate threshold of P = 0.20. The dominant expression pattern consisted of upregulation from day 30 through day +1, followed by downregulation through day +28. There was a threefold decrease from day 65 through day +14 in expression of IGFBP3, GSTM5, and PDPK1. These genes mediate IGF-I transport, oxidative stress, and glucose homeostasis, respectively. IGFBP3, EIF4B, and GSTM5 mRNA levels were positively correlated with blood serum total protein. Correlation analysis showed positive associations between serum nonesterified fatty acids and mRNA expression for SAA1, CPT1A, ACADVL, and TFAP2A. Transcript levels of ACSL1, PPARA, and TFAP2A were positively correlated with serum ß-hydroxybutyrate. Expression patterns for certain genes (e.g., IGFBP3, HNF4A, GPAM) revealed adaptations commencing well ahead of parturition, suggesting they are regulated by factors other than periparturient hormonal environment. Results provide evidence that hepatic inflammatory responses occurring near parturition initiate or augment adipose catabolism. In this context, cytokines, acute-phase proteins, and serum nonesterified fatty acids are key players in periparturient cow metabolism. We propose a model for integrating gene expression, metabolite, and liver composition data to explain physiological events in placenta, adipose, and liver during the periparturient period.
microarray; lactation; inflammation
THE PERIPARTURIENT, or "transition period," may be the most critical phase of the lactation cycle for dairy cows because the incidence of metabolic and infectious diseases centers disproportionately on this time (11). Despite initial efforts to quantify aspects of periparturient liver metabolism (38), knowledge on hepatic metabolite [e.g., ammonia, nonesterified fatty acids (NEFA)] fluxes remains limited. Genomic technologies provide a complementary approach to the measurement of metabolic fluxes in liver. Expression of genes that play important roles in liver development can be measured en masse using microarrays (26).
The liver performs essential functions in the body through the expression of genes encoding plasma proteins, clotting factors, and enzymes involved in detoxification, gluconeogenesis, glycogen synthesis, and metabolism of glucose, lipid, and cholesterol (22). Adult liver is a quiescent organ exhibiting minimal proliferative capacity, in which mitosis is observed in 1 of every 20,000 hepatocytes (7). Hepatocytes, however, begin to proliferate in a synchronized fashion in response to chemical, nutritional, vascular, and/or pathogen-induced stimuli (i.e., compensatory regeneration or hyperplasia). The coordinated expression of a large number of genes is required for hepatocyte differentiation and function of the adult liver (8). Hepatic gene expression is primarily controlled through transcription factors that respond to environmental, autocrine, or paracrine signals (8). Complex interactions of hormones, growth factors, and signal transduction pathways ultimately lead to expression of developmentally regulated genes.
The genomics approach may help identify regulatory mechanisms in liver during the dry period and early stages of lactation. A previous study examined gene expression in bovine liver at various stages of pregnancy using microarray technology; however, there were few biological replicates, and a relatively small microarray consisting of 2,675 genes was employed (21). We report here temporal expression profiling of >6,300 unique genes in liver of five dairy cows fed to current National Research Council (33) recommendations throughout the dry period and early lactation using a cattle-specific, high-density cDNA microarray (15). Transcriptional changes observed indicate a complex and previously unrecognized adaptive program in bovine liver.
MATERIALS AND METHODS
Animals and management.
Five multiparous Holstein cows were randomly selected from a group of twelve control cows enrolled in a large experiment designed to assess the effects of plane of nutrition during the "far-off" and "close-up" dry periods on prepartum metabolism and postpartum metabolism and performance. Cows were housed in individual tie stalls throughout the experiment and were allowed to exercise daily in an outside lot from 0700 to 1000. All cows underwent normal parturition and were free from health disorders during the study. All procedures were conducted under protocols approved by the University of Illinois Institutional Animal Care and Use Committee.
Liver tissue collection and RNA extraction.
Liver was sampled via puncture biopsy (12) from cows under local anesthesia at 0700 on days 65 (dry-off), 30, 14, +1, +14, +28, and +49 relative to parturition. Tissue (1.01.5 g) was weighed postbiopsy, placed in 1015 ml of ice-cold TRIzol reagent (Invitrogen, Carlsbad, CA), homogenized, and RNA extracted according to the manufacturer's protocol. RNA was resuspended in RNA storage buffer (Ambion, Austin, TX) and stored at 80°C until use.
Liver tissue composition, blood serum metabolites, body condition score, and energy balance.
A portion of liver tissue collected for RNA extraction was stored in liquid N for analysis of total lipid, triacylglycerol, and glycogen. Blood serum was assayed for concentrations of NEFA, total protein, urea nitrogen, ß-hydroxybutyrate (BHBA), glucose, and insulin as described previously (10). Body weight and body condition score were determined for each cow weekly from day 65 to day +56 relative to parturition. Three persons assigned body condition scores independently at each time of scoring throughout the experiment. Energy balance was calculated (33) individually for each cow (see Supplemental Material; available at the Physiological Genomics web site).1
.
Microarrays.
A 7,872-element cDNA microarray spotted in duplicate on amino silane-coated glass slides (15) was used for transcript profiling. Annotation was based on similarity searches (January 2005) using sequential Basic Local Alignment Search Tool (BLAST)N and TBLASTX against human and mouse UniGene databases and the human genome, as previously described (15). The 7,872-element array represents >6,300 unique genes. Gene Ontology (GO) terms were parsed from LocusLink to functionally annotate cDNA sequences. RNA (20 µg) from liver tissue and a reference standard derived from a mixture of tissues not including liver (15) were used to make aminoallyl-labeled cDNA followed by incorporation of Cy3-ester and Cy5-ester (Amersham, Piscataway, NJ) (44). Each experimental sample was cohybridized with the reference standard, which allowed the treatment of fluorescence ratios as measurements of relative expression across all samples and time points. Probes were hybridized to the array for 2 days at 42°C.
All samples were hybridized to duplicate slides for a total of four spots per cDNA element. Slides were scanned for both dye channels with a Scanarray 4000 (GSI-Lumonics, Billerica, MA) dual-laser confocal scanner, and images were processed using GenePix 4.0 (Axon Instruments). Microarray data are deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (accession no. GSE2692).
Quantitative real-time RT-PCR.
Total RNA was purified with the RNeasy Mini Kit and residual DNA removed using the RNase-Free DNase Set (Qiagen, Valencia, CA). RT-PCR reactions were carried out in triplicate according to the manufacturer's instructions (Invitrogen). Primers were designed to specifically amplify 16 target cDNAs and an endogenous control gene, bovine ß-actin (ACTB) (Primer Express Software v2.0; Applied Biosystems, Foster City, CA). For relative quantification of target cDNA, samples were run in triplicate in a 384-well plate using an ABI Prism 7900 HT SDS instrument (Applied Biosystems). To control reproducibility of the essay, each primer set was run in triplicate with cDNAs originated from three separate RT-PCR reactions. Details regarding reaction conditions, primer design, and relative quantification of target cDNAs are presented in the Supplemental Material (Supplemental "Materials and Methods" and Supplemental Table S1).
Data analyses.
During an initial screening, microarrays that did not contain at least 10,000 spots with median background-subtracted signal intensities >3 SD above background in both Cy3 and Cy5 channels were repeated. Data from a total of 68 microarrays were normalized and used for statistical analysis. Median background-subtracted signal intensities for each spot were computed before normalization. Three criteria were then implemented to filter unreliable intensity values: 1) if foreground was less than background, then signal intensity was set to 1.0; 2) only those spots that had signal intensities >3 SD above background for both Cy3 and Cy5 channels were kept for further analysis; and 3) sequences with less than 8 observations per time point (out of a potential total of 20) were removed. Data were normalized for dye and array effects (i.e., Loess normalization and array centering). A mixed-effects model with repeated measures was then fitted to the adjusted ratios (liver/reference standard) using Proc MIXED (SAS; SAS Institute, Cary, NC). The model consisted of day as a fixed effect and cow as a random variable. Probability values for the effect of day were adjusted for the number of comparisons using Benjamini and Hochberg's false discovery rate (FDR) (37). Estimates for the available genes at each time point were then computed. Differences in relative expression were considered significant at an FDR-adjusted P of 0.20, corresponding to raw P values 103. Data from quantitative real-time RT-PCR (qPCR) (relative mRNA levels calculated with a standard curve) were analyzed as described elsewhere (3), using the same statistical model described above; differences were considered significant at P
0.05. Expression patterns for the 62 differentially expressed genes identified by microarray were analyzed using k-means clustering (13) (GeneSpring 6.1; Silicon Genetics, Redwood City, CA). Normalized, log2-transformed ratios computed by GeneSpring were used for k-means clustering. These genes were classified according to their GO terms, "molecular function" and "biological process," to facilitate data mining (Supplemental Table S2). The Proc CORR procedure of SAS was used to analyze correlations among expression profiles over time for each differentially expressed gene, blood and liver metabolites, and energy balance (Supplemental Table S3).
RESULTS
Energy balance, body weight and body condition score, metabolic indicators in blood, and liver composition.
Cows were in positive energy balance for nearly the entire dry period (Fig. 1A). Energy balance was negative during the first week postpartum but became positive at week 2 and remained so thereafter. Body weight, body condition score, and energy intake increased from day 65 to day 14 (Supplemental Fig. S1). Subsequently, there was a decrease in body weight and body condition score from day 14 through day +21 relative to parturition. Serum NEFA concentration increased steadily from 1 wk before parturition through the end of week 1 postpartum (Fig. 1B). Overall, energy balance reflected the amount of energy consumed from the diet (Supplemental Fig. S1). Temporal concentrations for serum BHBA, total protein, urea nitrogen, insulin, and glucose are presented in Supplemental Fig. S2. Total lipid, triacylglycerol, and glycogen content in liver did not change appreciably on days 65, 30, and 14 relative to parturition (Fig. 1C). Liver total lipid and triacylglycerol content was markedly greater on days +1 and +14 postpartum relative to day 14 and then declined to prepartum values by day +49. In contrast, glycogen content decreased substantially between day 14 and day +1 and then increased gradually through day +49 (Fig. 1C). Changes in blood metabolites and liver composition were typical of those previously reported for periparturient dairy cows (11, 19).
|
|
|
|
|
DISCUSSION
We have conducted the first experiment to our knowledge linking the bovine liver transcriptome to metabolic indicators during the dry period and early lactation. Although transcription profiling cannot by itself explain complex physiological processes occurring in liver prepartum and during lactation, the significant correlation between transcript and protein levels (30), the wealth of information on hepatic metabolism in periparturient dairy cows (11, 19, 38), and recent studies in rodents (6, 28) allowed us to integrate our data into a general model to explain the physiological events in placenta, adipose, and liver (see Fig. 6). The model takes into account gene expression, blood metabolites, liver composition, and energy balance data. The following discussion of the significant effects of physiological state on hepatic gene expression is organized around physiological and metabolic parameters of lactation biology. While many elements of the model are only suggestive, we have used the extensive gene annotation system for our microarray and other powerful data mining tools to help understand the complex biology of periparturient dairy cows.
Periparturient liver mass and gene expression.
The increase in mRNA expression level between days +14 and +28 postpartum (Fig. 3, clusters 2 and 4) for genes associated with translation initiation and protein biosynthesis (EIF4B), intracellular protein degradation (PSMC2), protein localization and transport (LMAN2), insulin signaling and glucose homeostasis (PDPK1), cellular proliferation (CRIP1), regulation of various aspects of cell growth and maintenance (IGFBP3), and regulation of transcription (HNF4A, Fig. 2) coincided not only with an increase in serum total protein (Supplemental Fig. S2) but also with greater liver mass observed by others around this time (39). Four unannotated transcripts were upregulated from day +14 through day +28 (Fig. 3, cluster 7), suggesting they may play unidentified roles in bovine liver biology. Correlations among serum total protein and expression patterns for IGFBP3, HNF4A, EIF4B, PDPK1, and CRIP1 were positive (P 0.05; Supplemental Table S3).
Environmental stresses and nutrition can reduce protein biosynthesis in mammalian cells, which are closely correlated with the concentration and activity of EIF4F and EIF4B (16). Nutrient availability (e.g., amino acids) is the primary signal that induces activation of EIF4B (16). CRIP1, found in many tissues including immune, intestinal, and liver, is a newly recognized transcription factor that has a role in cell growth, differentiation, and turnover (23). LMAN2 is a protein synthesized by liver that responds as an acute-phase reactant, activates the complement cascade, and serves a crucial role in intracellular protein transport (20). There was a gradual decline in energy balance, feed intake, serum total protein, and insulin from the last week prepartum through the end of the first week postpartum. We found the opposite response for NEFA, and others have shown marked muscle protein degradation (2) indicating a general state of catabolism in the cow. Taken together, downregulation of EIF4B, HNF4A, PSMC2, IGFBP3, LMAN2, and CRIP1 in liver around parturition (Figs. 2 and 5A) partly explains these physiological measurements (Fig. 6).
|
|
TNF- and IL-1ß stimulate the synthesis of positive acute-phase proteins such as SAA1 (36), partly explaining the nearly sixfold increase in expression of SAA1 (Fig. 3, cluster 8, and Fig. 5C) observed on day +1 relative to day 14 or +14. Importantly, qPCR analysis of SAA1 expression showed marked upregulation (
3-fold) by day 14 (Supplemental Figs. S3 and S6). SAA1 synthesis in liver is upregulated by as much as 1,000-fold during the acute-phase response (36). Its promoter is occupied by HNF4
in human hepatocytes (34), but expression of HNF4A (Fig. 2) did not quite parallel that of SAA1, indicating that only small changes in abundance of this transcription factor may be necessary to elicit drastic changes in expression for some of the genes under its control. TNFA upregulation in liver macrophages may thus act in a paracrine fashion and elicit potent upregulation of SAA1 in hepatocytes, and it has been shown that proinflammatory and signaling genes are upregulated in liver from mice induced to develop fatty liver and insulin resistance by high-fat diets (6). Our data provide evidence for a role of bovine hepatic macrophages or Kupffer cells in the etiology of periparturient fatty liver (Fig. 6).
Periparturient hepatic lipid metabolism.
Upregulation of SREBF1 in mice resulted in fatty liver (45). Deletion of SREBF1 in Lepob/Lepob mice lowered hepatic expression for fatty acid synthase (Fasn), Gpam, ATP-citrate lyase (Acly), and Spot 14 (S14), which in turn correlated with reduced liver triacylglycerol concentration (47). Our data on plasma NEFA and liver composition verified typical responses occurring in bovine liver during the periparturient period (11, 19). More importantly, we show that increased hepatic expression of mRNA encoding proinflammatory cytokines and acute-phase proteins (TNFA, SAA1) is positively correlated with lipid mobilization from adipose tissue (i.e., increased plasma NEFA, Fig. 1B) and fatty acid oxidation (i.e., increased plasma BHBA, Supplemental Fig. S2). Relative to day +1, the marked increase in expression of GPAM (2-fold) and SREBF1 (
3-fold) assessed by qPCR on day +14 through day +49 resembled the pattern of serum NEFA (Fig. 1B) and liver triacylglycerol concentration (Fig. 1C). These data offer support for a role of both genes in the development of periparturient fatty liver (Fig. 6). We present strong evidence that periparturient serum NEFA concentrations and thus hepatic NEFA uptake (38) are not only a primary mediator of GPAM and SREBF1 expression but also of HNF4A and PPARA (Fig. 2 and Fig. 6). Despite marked upregulation of PPARA, upregulation of SREBF1 and subsequent upregulation of GPAM are necessary responses in mice (24, 42) to accommodate the greater influx of NEFA into liver, i.e., there is an imbalance between fatty acid oxidation and lipid synthesis. The relative importance of GPAM and SREBF1 in the etiology of periparturient fatty liver remains to be determined.
PPAR- is a nuclear protein that mediates effects of NEFA on peroxisomal and mitochondrial fatty acid oxidation and upregulates genes associated with ketogenesis and ureagenesis (28). Our data (e.g., Figs. 1 and 2) strongly support the concept that PPAR-
plays a central role in these metabolic events in periparturient bovine liver (Fig. 6). PPARA upregulation also exerts anti-inflammatory responses by inhibiting expression of proinflammatory cytokines and acute-phase proteins, along with reducing hepatic amino acid catabolism (28). The marked upregulation of PPARA on day +1 along with increased serum NEFA and BHBA, and the expression patterns of CPT1A, ACOX1, ACSL1, and ACADVL (Supplemental Fig. S3), show for the first time that PPARA expression in transition dairy cows is highly associated with expression of these genes, as proposed earlier (11) (Fig. 6). The expression pattern of HNF4A is also suggestive of an important role for this gene in fatty acid oxidation and gluconeogenesis through binding (34) of the promoter region of ACADVL and ACOX1 (ß-oxidation) and PCK1 (gluconeogenesis). The gradual increase in periparturient liver triacylglycerol despite PPARA activation may be an example of an anti-inflammatory mechanism failing upon greater proinflammatory pressure. Similar to mice (42), an imbalance between PPAR-
-mediated anti-inflammatory and NF-
B-mediated proinflammatory signals in transition dairy cows may contribute to increased inflammation during fatty liver development.
Metabolic stress, insulin signaling, growth hormone/IGF-I axis, and hepatic gene expression.
Expression of GSTM5 decreased nearly 3-fold from day 30 through day +14 (Fig. 3, cluster 4) during the time when liver triacylglycerol accumulated (Fig. 1C) and hepatic oxidative activity increased 2-fold (38). Our result agrees with observations in fatty liver from mice fed high-fat diets short-term (i.e., 11 days; Ref. 24) but contrasts with the marked upregulation (>3-fold) of GST mRNAs in fatty liver from mice fed high-fat diets long-term (i.e., 12 wk) (18). Those studies and ours indicate that, in liver, there is short- and long-term transcriptional regulation of biological stress responses to oxidative stress elicited by nutrition. The decline in GSTM5 around parturition could be a factor in the higher risk for hepatic health disorders observed in dairy cows (11) (Fig. 6), because GST expression and activity has profound effects on sensitivity to physiological insults (4). Recent studies in mice support a role for growth hormone/IGF-I signaling in the regulation of antioxidant defense (4), i.e., increased systemic concentration of growth hormone results in lower liver S-adenosylmethionine, S-adenosylhomocysteine, methionine adenosyltransferase activity, and liver GST activity. These observations provide a link between elevated serum growth hormone observed by others (35) and reduced hepatic GSTM5 expression during the periparturient period.
Upregulation of cytosolic PCK1 mRNA abundance in liver (1) increases hepatic glucose output (38) needed to meet mammary demands for milk synthesis after parturition (2). Greater mammary uptake of glucose soon after parturition maintains low serum glucose concentration despite greater than twofold increase in hepatic glucose output (38). Insulin inhibits gluconeogenesis in mice through insulin receptor-mediated PDPK1 activation (31). Our data show that PDPK1 expression (Fig. 3, cluster 4) decreased 3-fold between day 65 and day +14. More importantly, PDPK1 downregulation during days 14 through +14 corresponded with liver glycogen depletion (Fig. 1) indicating, as in liver-PDPK1/ mice (31), that the PDPK1 signaling pathway in periparturient dairy cows plays an important role in regulating glucose homeostasis and controlling expression of insulin-regulated genes (43) (Fig. 6).
Hepatic abundance of IGFBP3 has profound implications for liver and peripheral tissues because of the longer half-life of the IGFBP3/IGF-I ternary complex (9). Periparturient dairy cows have higher plasma growth hormone, the primary regulator of IGF-I synthesis and secretion in hepatocytes, but have lower plasma IGF-I due to reduced hepatic mRNA abundance of growth hormone receptor 1A (35). Recently, the possibility has been raised that insulin regulates the efficiency of growth hormone signaling in periparturient liver (5, 40). Infusion of insulin under euglycemia increased plasma IGF-I and hepatic levels of IGF-I mRNA and growth hormone receptor protein, along with increased plasma IGFBP3 (5, 40). We observed an 3-fold gradual decrease in expression of IGFBP3 from dry-off through day +14 (Fig. 3, cluster 4), similar to the pattern of serum insulin concentration (Supplemental Fig. S2) and hepatic mRNA and plasma IGF-I observed by others around this time (35, 40). The parallel expression pattern for PDPK1 and IGFBP3 suggests that IGFBP3 actions in bovine liver occur via PDPK1 signaling pathways. We show that differential liver IGFBP3 expression indeed represents one plausible mechanism whereby insulin regulates plasma IGF-I concentrations (5, 40) (Fig. 6).
In conclusion, hepatic adaptations to the onset of parturition are characterized by gene expression changes commencing well ahead of parturition, suggesting they are regulated by other factors (e.g., nutrient availability) besides the hormonal environment characteristic of the periparturient period. We propose a model (Fig. 6) with our gene expression, metabolite, hormone, and energy balance data, in addition to data from the literature, to account for observed periparturient behavioral and hepatic adaptations. The dynamics of hepatic energy metabolism during the transition period likely vary depending on the nature and intensity of the inflammatory response, which, on the basis of our results and those of others (e.g., Ref. 25), seems to be a key signal initiating adipose and muscle tissue catabolism. Cytokines, acute-phase proteins, and the increasing NEFA pool are key players affecting biological functions in peripheral tissues and brain in transition dairy cows.
GRANTS
Gene expression profiling was supported by award no. 2001-35206-10946 from National Research Initiative Competitive Grants Program/Cooperative State Research, Education, and Extension Service/United States Department of Agriculture to J. K. Drackley, H. A. Lewin, and S. L. Rodriguez-Zas. Support for the animal work was provided by funds from the State of Illinois through the Illinois Council on Food and Agricultural Research (C-FAR).
ACKNOWLEDGMENTS
We gratefully acknowledge the help from the staff of the Univ. of Illinois Dairy Research and Teaching Unit for Animal Care. The advice of Dr. Mark R. Band (Director, Functional Genomics Unit, The W. M. Keck Center for Comparative and Functional Genomics, Univ. of Illinois) during portions of the microarray work also is greatly appreciated.
FOOTNOTES
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: J. J. Loor, Dept. of Animal Sciences, Univ. of Illinois, 1207 West Gregory Dr., 498 Animal Sciences Laboratory, Urbana, IL 61801 (e-mail: jloor{at}uiuc.edu).
10.1152/physiolgenomics.00132.2005.
1 The Supplemental Material for this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/00132.2005/DC1
REFERENCES