Complementary hepatic metabolomics and proteomics reveal the adaptive mechanisms of dairy cows to the transition period

The transition period from late pregnancy to early lactation is a vital time of the lifecycle of dairy cows due to the marked metabolic challenges. Besides, the liver is the pivot point of metabolism in cattle. Never-theless, the hepatic physiological molecular adaptation during the transition period has not been elucidated, especially from the metabolomics and proteomics view. Therefore, the present study aims to investigate the hepatic metabolic alterations in transition cows by using integrative metabolomics and proteomics meth-ods. Gas chromatography quadrupole-time-of-flight mass spectrometry-based metabolomics and data-independent acquisition-based quantitative proteomics methods were used to analyze liver tissues collected from 8 healthy multiparous Holstein dairy cows 21 d before and after calving. In total, 44 metabolites and 250 proteins were identified as differentially expressed from 233 metabolites and 3,539 proteins detected from the liver biopsies during the transition period. Complementary functional analysis of different metabolites and proteins indicated the upregulated gluconeogenesis, tricarboxylic acid cycles, AA degradation, fatty acid oxidation, AMP-activated protein kinase signaling pathway, peroxisome proliferator-activated receptor signaling pathway, and ribosome proteins in postpartum dairy cows. In terms of the metabolites and proteins, glucose-6-phosphate, fructose-6-phosphate, carnitine palmitoyltransferase 1A, and phosphoenolpyruvate carboxykinase played a significant role in these pathways. The upregulated oxidative status may be accompanied by the pathways mentioned above. In addition, the upregulated glucagon and insulin signaling pathways also indicated the significant requirement for glucose in postpartum dairy cows. These outcomes, from the view of global metabolites and proteins, may present a better comprehension of the biology of the transition period, which can be helpful in further developing nutritional regulation strategies targeting the liver to help cows overcome this metabolically challenging time.


INTRODUCTION
The transition period from late gestation to early lactation is the most physically, physiologically, and metabolically challenging time during the whole lifetime of dairy cows (Grummer, 1995;Goff and Horst, 1997;Drackley, 1999).Maternal cows experience changes such as DMI decrease, fetal development, parturition, and the onset of lactation (Grummer et al., 2004;Huang et al., 2014;Zhang et al., 2015).Moreover, cows are the most susceptible to diseases over the transition period (Wankhade et al., 2017;Ceciliani et al., 2018).Thus, the transition period is critical to the future production, health, and sustainable profitability of cows and is seen as the final frontier of the biology of dairy cows (Drackley, 1999;Bobe et al., 2004).
Maternal metabolic responses, nutritional status, and physiological conditions undergo marked changes in the transition from late pregnancy to early lactation (Ceciliani et al., 2018;Luo et al., 2019).As a central metabolic organ within the whole body, the activities of the liver are intensively influenced by many nutritional and physiological factors (Jiang et al., 2008;Luo et al., 2019;Zhang et al., 2019b).The associated processes are mainly regulated by variations of genes, proteins, enzymes, and metabolites (Ceciliani et al., 2018;Shi et al., 2018;Zhang et al., 2019b).Adapting the critical metabolic pathways in the liver, to a certain extent, determines whether the cows can survive the transition period smoothly (Drackley et al., 2001).The physical compression of the rumen by the fetus and increased reproduction hormone levels during the late pregnancy period were 2 of the factors causing decreased DMI (Grummer et al., 2004;Loor, 2010), which then induces negative energy balance (NEB) in transition dairy cows (Huang et al., 2014;Zhang et al., 2015).Furthermore, the imbalance between large amounts of energy required for milk synthesis and secretion and lower DMI after calving exacerbates the NEB.The intensified NEB facilitates body lipid as well as protein mobilization and mainly results in higher uptake of nonesterified fatty acids (NEFA) by the liver (Kuhla et al., 2011;McCabe et al., 2012;Schaff et al., 2012).The excessive NEFA in the liver will be transformed into ketone bodies like BHB or reesterified triglycerides (TG) and further increase the risk of ketosis and fatty liver, respectively, both of which are, in turn, potentially detrimental to the health of the cows (Goff and Horst, 1997;Drackley, 1999;McCabe et al., 2012).Given the complexity of metabolic changes during the transition period, we still lack a complete explanation of the physiological mechanisms.Therefore, a better understanding of the global hepatic metabolites and protein profiles during the transition period may be beneficial in reducing the risk of metabolic disease and increasing the profitability of cows.
To further investigate the complex changes during the transition period, the omics approaches, which feature high-throughput and large-scale data, appear to be ideal tools.In most previous studies, quantitative PCR, microarray, or RNA-seq-based transcriptomics methods were used to investigate the hepatic physiological mechanisms associated with the transition (Loor et al., 2005;Loor, 2010;Ha et al., 2017).However, the changes in transcriptome cannot guarantee the subsequent phenotypic variation due to translation efficiency, post-transcriptional regulation, and protein half-life (Gatto et al., 2016;Ceciliani et al., 2018;Schatton and Rugarli, 2018).With the fast development in bioinformatics tools and technology, using metabolomics and proteomics in livestock science is widely accepted (Ceciliani et al., 2018;Luo et al., 2019;Zhang et al., 2019b).Owing to the high sensitivity in measuring small molecular metabolites, metabolomics is used in diverse biological systems (Zhang et al., 2017;Ceciliani et al., 2018;Zhang et al., 2019b).Quantitative proteomics, which achieves high accuracy and precision of the quantification and includes a description of posttranslational protein modifications, has been accepted in most functional proteome studies (Ceciliani et al., 2018;Muntel et al., 2019).Thus, in this exploratory and hypothesis-generating work, we aimed to perform a complementary bioinformatic analysis of the global hepatic metabolites and proteins by using untargeted metabolomics and data-independent acquisition (DIA)based quantitative proteomic on liver samples from transition dairy cows.From this work, we expected to provide a comprehensive view of the hepatic adaptation to the transition period in dairy cows.

Animals and Diets
In the present study, requirements and regulations of Instructive Notions with Respect to Caring for Experimental Animals, Ministry of Science and Technology of China were followed in detail.This protocol was passed by the Institutional Animal Care and Use Committee of China Agricultural University (Beijing, P. R. China, permit no.AW03039102-2).For the experimental animals, 12 healthy multiparous Holstein dairy cows with similar 305 d milk yields (total milk yield of 9,210 to 10,870 kg of the last lactation period), age (57.42 ± 6.79 mo), BCS, and calving date (difference less than 2 wk) from Sunlon Livestock Jinyindao Farm (Daxing County, Beijing, China) were kept in a free-stall barn with free access to fresh water.During the far-off and early lactation period, the dry cow and lactating cow TMR was delivered twice daily at 0730 and 1400 h.During the close-up period, the prepartum (PREP) TMR was delivered once daily at 1400 h.The detailed ingredients and chemical composition of the TMR can be found in Supplemental Table S1 (https: / / doi .org/ 10 .5281/zenodo .6792970;Zhang et al., 2022).

Feed Intake, Blood, and Milk Sample Collection and Measurement
Individual feed intake was recorded daily by the Roughage Intake Control System (RFID, Zhenghong Company), which can identify the cow ID before opening the trough and measure the feed weight before and after cow eating, as described by Gaowa et al. (2021).Blood samples of all the cows were collected from the coccygeal vein into evacuated serum tubes on d −21, −7, 7, and 21 (7 and 21 d both before and after calving) at 0600 h.All the tubes were centrifuged at 3,500 × g at 4°C for 15 min to obtain serum and stored at −20°C for further analysis.Serum total cholesterol (TC), TG, BHB, and NEFA concentrations were analyzed on a Hitachi 7600 automated biochemistry analyzer (Hitachi Co. Ltd.) using kits from Nanjing Jiancheng Bioengineering Institute (Nanjing, China).Cows were milked 3 times each day after calving at 0700, 1300, and 2100 h, and milk production was recorded daily.

Liver Sample Collection
Eight cows were randomly selected to have liver biopsies on d −21 and 21 just after the second feeding.According to our previous protocol (Shi et al., 2018), 500-1,000 mg liver samples were obtained by biopsy, rinsed, and frozen in liquid nitrogen for further determination.Samples were divided into 2 groups according to sampling day: 21 d PREP and 21 d postpartum (POSP).The sample size was decided based on previous studies with similar designs and methods on dairy cows (Skibiel et al., 2018;Zhang et al., 2019b), and a formal a priori sample size calculation was not performed for this study.

Metabolomics Analysis
The GC quadrupole TOF MS analysis and metabolites extraction were similar to a previous study (Zhang et al., 2019b).Briefly, 450 μL of methanol/chloroform (volumetric ratio = 3:1) was added to 50 mg samples from PREP and POSP to extract metabolites.Equal aliquots of extract liquid from all experimental samples were pooled as quality control (QC) specimens.Adonitol was utilized as an internal standard.To perform the following GC TOF MS analysis of all samples, an Agilent 7890 GC system was used along with a Pegasus HT TOF mass spectrometer in splitless mode (LECO Corporation).For 1 min, the initial temperatures were maintained at 50°C, then incremented to 310°C at a rate of 10°C min −1 and kept at 310°C for 8 min.The ion source, injection, and transfer line temperatures were 250, 280, and 280°C, respectively.The MS data were obtained in a full-scan mode after a solvent delay of 6.33 min with the m/z range of 50-500 at a rate of 12.5 spectra per second.The Chroma TOF 4.3X software built-in with the LECO-Fiehn Rtx5 database (LECO Corporation) was used to preprocess and annotate the metabolomics data.The peaks detected less than 50% of QC specimens or relative standard metabolomics data deviation of more than 30% in QC specimens were eliminated (Dunn et al., 2011).

Preparing Sample and Data-Independent Acquisition Mass Spectrometry for Proteomic Analysis
The total protein in liver tissues from PREP and POSP were extracted by grinding method in the presence of extraction buffer, which contained 8 M urea, 2 M thiourea, 2% 3-[(3-cholamidopropyl)dimethylammonio]-1-propane sulfonate, and a proteasome inhibitor, as described in a previous study (Zhang et al., 2019b).The protein concentrations were determined using a bicinchoninic acid Assay Kit (Dingguo Changsheng) according to the manufacturer's instructions.
The protein digestion was performed following the filter-aided sample preparation protocol (Wisniewski et al., 2009).Briefly, after being reduced with 20 mM dl-dithiothreitol, protein samples were alkylated with 50 mM iodoacetamide.Then, samples were transferred onto filters and digested by 2% trypsin at 37°C for 12 h.The peptide samples were collected for the following MS analysis.
The DIA MS technique utilized 20 m/z isolation windows from 400 to 800, 30 m/z isolation windows from 800 to 1,000, and 50 m/z isolation windows from 800 to 1,000.First, a full scan at 30,000 full widths at half maximum (FWHM) resolving power (at 200 m/z) was conducted after sequential high-energy collisional dissociation-MS/MS scans at a normalized collision energy of 30 and resolution of 15,000 FWHM.The ranges between 400 and 1,200 m/z were measured by such tests, with the highest injection times of 55 min for MS and the auto setting for MS/MS.Acquired at variable resolutions, values were adjusted to 3 × 10 6 for MS and 1 × 10 6 for MS/MS.The MS/MS scan range was adjusted to 400-1,200 m/z.

Western Blot Analysis
The protein specimens utilized in the western blot were the same as those in proteomic analysis.Protein samples were boiled for 10 min at 95°C.Then, by resolving 20 μg of total protein per lane via SDS PAGE, they were conveyed to a polyvinylidene fluoride membrane (0.45 μm, Merck Millipore) through the semidry transfer assembly (Bio-Rad Laboratories).

Data Processing, Bioinformatics, and Statistical Analysis
The DMI, milk production, and serum metabolites data were first checked for normality and analyzed using the PROC MIXED procedure of SAS version 9.4 (SAS Institute Inc.) with sampling time (week or day) as fixed effect and cows within time as a random effect.Results were reported as least squares means.Significant differences were declared at P ≤ 0.05, and trends were reported at 0.05 < P < 0.10.
The metabolomics analysis process was similar to previous studies (Dunn et al., 2011;Zhang et al., 2020).Briefly, the raw data were converted into Chroma TOF4.3X software (LECO) with a built-in LECO-FiehnRtx5 database.Then peaks extraction, peak alignment, peak identification, deconvolution analysis, and integration of the peak area were performed.The missing value (metabolites that were not detected in some samples) in the original data was simulated by using a numerical simulation method that fills half of the minimum value.The limit of detection (LOD) was determined by the signal-to-noise ratio (S/N) of the corresponding peaks, and the peaks with S/N less than 3.0 were considered noise.The peaks detected in less than 50% of original and QC samples, less than 400 similarities, relative standard deviation greater than 30% in QC samples, or beyond the interquartile range to filter data were removed.Data were standardized by peak area normalization methods.The unit variance scaling was selected as the data scale conversion mode.The maximal covariance between response variables and measured data was obtained for metabolomics analysis using principal component analysis (PCA) and orthogonal projections to latent structures-discriminant analysis (OPLS-DA) in SIMCA.Significantly differently produced metabolites (DPM) between treatments were recognized using variable importance in projection (VIP) scores (VIP no less than 1.0) obtained from the OPLS-DA model and P-values (P value less than 0.05).The metabolic pathways analysis of 44 DPM were processed using MetaboAnalyst 4.0 with default parameters and selecting Bos taurus as a pathway library (Chong et al., 2018).
The proteomics analysis process was adapted from a previously published protocol (Egertson et al., 2015).Briefly, all the data-dependent acquisition MS data were thoroughly searched against the database of the UniProtKB (Bos taurus; data of access 01.05.2021) for peptide identification and quantification by using Proteome Discoverer Version 2.2 (Thermo Scientific).A file for the results was created using raw data for each experimental set searched in a single batch.The Proteome Discoverer's outputs provide a set of files utilized as the reference spectra library containing peptide sequences, modifications, charge states, confidence scores, retention times, and the equivalent fragment ions intensity and m/z.Then, DIA data processing spectral and library generation were conducted utilizing Skyline Version 3.5 (MacLean et al., 2010;Egertson et al., 2015).No statistical analysis or calculation was performed using the missing values.The raw data have been deposited to the ProteomeXchange Consortium (http: / / proteomecentral .proteomexchange.org)via the iProX partner repository (Du et al., 2017) with the data set identifier PXD025564.
Protein differences between treatments were compared, and P values were determined utilizing the Student's t-test.A fold change (FC) of 1.5 and false discovery rates (FDR)-adjusted P-value <0.05 (q ≤ 0.05) were used as the threshold for identifying differently synthesized proteins (DSP).To increase the Zhang et al.: HEPATIC METABOLIC CHARACTERISTICS OF TRANSITION DAIRY COWS robustness of our study, the DSP presented in at least half of the samples were used in the following analysis.The overall DSP were examined to enrich Gene Ontology (GO) terms, cellular component (CC), molecular function (MF), and biological process (BP), as well as Kyoto Encyclopedia of Genes and Genomes database (KEGG) pathways.Considerably enriched GO terms and KEGG pathways were identified as q ≤ 0.05.Furthermore, the protein-protein interaction (PPI) networks of upregulated and downregulated DSP were built and graphically visualized utilizing the searching device for the Retrieval of Interacting Genes (STRING) V.11.0 with the default parameters (Szklarczyk et al., 2019).For western blot data, using a t-test, parameter differences between treatments were compared with calculate P values.

Dry Matter Intake, Milk Production, and Serum Metabolites
From PREP to POSP, the DMI and serum TC content increased 1.63-and 1.87-fold (P ≤ 0.01), respectively, and TG content decreased 1.34-fold (P < 0.01; Figure 1).Milk production increased (P < 0.01) with the time after calving.Serum NEFA and BHB contents were similar from PREP to POSP.

Metabolite Profiles of Liver Samples and Data Analysis
In total, 553 valid peaks and 233 metabolites were identified in these 2 groups, and both were in the same metabolite classes (Supplemental Table S2; https: / / doi .org/ 10 .5281/zenodo .6792970;Zhang et al., 2022).The 2 groups of specimens were well separated, and samples in the same group were well aggregated in the PCA score plot (Supplemental Figure S1A; https: / / doi .org/ 10 .5281/zenodo .6792970;Zhang et al., 2022).The QC specimens were well overlapped in the PCA score plot, indicating that the metabolomics method used was robust, highly repeatable, and stable.The equivalent R 2 Y value of the OPLS-DA model was 0.995 in POSP vs. PREP, and the intercept of permutation tests was 0.91, revealing the good effectiveness of the model for identifying the difference between the 2 treatments (Supplemental Figure S1B).Besides, the Q 2 value and intercept were 0.75 and −0.48, respectively, indicating that the OPLS-DA model had good predictability and no overfitting.The specimens in the OPLS-DA score plots were within Hotelling's T2 ellipse of 95% (Supplemental Figure S1C).

Differential Metabolite and Pathway Analysis
Under the criterion of VIP > 1.0 and P < 0.05, we identified 44 DPM from PREP to POSP, of which 30 increased and 14 decreased in the POSP group (Table 1; Supplemental Figures S2A, S3A, https: / / doi .org/ 10 .5281/zenodo .6792970;Zhang et al., 2022).According to the pathway analysis of DPM, 15 pathways with q < 0.05 and impact ≥0.04 profoundly changed from PREP to POSP (Figure 2).The AA, lipid, energy, and nucleotide metabolism were mainly involved in these pathways.
In total, 250 DSP were identified from PREP to POSP, of which 169 were upregulated and 81 were downregulated in the POSP group compared with the PREP group (Supplemental Figure S2B).Based on the protein abundance data of the 250 DSP, the 2 groups' clusters were well separated (Supplemental Figure S3B).
For the 5 selected proteins, the FC among treatments in WB were in line with those in the DIA data, and 3 proteins possessed the same significance in the WB platform as in the proteomic platform (Figure 3).
There were 2 tensive networks of upregulated DSP in the PPI network (Figure 5A).The first network featured lipid and carbohydrate metabolism, including

Functional Annotations and Interaction Network of the Downregulated Differently Synthesized Proteins (Postpartum Versus Prepartum)
By enrichment of 81 downregulated DSP from POSP to PREP into 710 GO terms, they were categorized in terms of their BP (70.3%),MF (15.8%), and CC (13.9%), but none of them were as significant.These proteins were enriched into 39 pathways through KEGG pathway analysis, of which 10 paths were significant (q < 0.05; Figure 6), namely vitamin B6 metabolism, nicotinate and nicotinamide metabolism, platinum drug resistance, glutathione metabolism, metabolism of xenobiotics by cytochrome P450 (CYP), drug metabolism-other enzymes, drug metabolism-CYP, steroid hormone biosynthesis, serotonergic synapse, and chemical carcinogenesis.In addition, 21 DSP were mapped into these 10 pathways (Table 2).
There was no clear network of downregulated DSP from POSP to PREP in the PPI network (Figure 5B).Only several glutathione S-transferase Mu (GSTM) and CYP, including GSTM1, GSTM2, GSTM3, GSTM4, cytochrome P450 2C18 (CYP2C18), CYP2C19, and CYP2D14 have more interactions than other DSP and may have critical roles in oxidative status regulation.

Integrating Metabolomics and Proteomics Analyses
The DSP and DPM in significantly changed KEGG pathways were mapped together using the KEGG Mapper tool (Kanehisa et al., 2017) to the KEGG pathway.The mapped pathways included metabolic pathways, biosynthesis of AA, carbon metabolism, PPAR signaling pathway, peroxisome, ribosome, fatty acid (FA) degradation, glutathione metabolism, tyrosine metabolism, valine, leucine and isoleucine degradation, glycine, serine and threonine metabolism, cysteine and methionine metabolism, arginine and proline metabolism, alanine metabolism, AMP-activated protein kinase (AMPK) signaling pathway, pyruvate metabolism, TCA cycles, glycolysis/gluconeogenesis, tryptophan metabolism, and glycerolipid metabolism.These critical pathways mapped with DSP and DPM were mainly clustered into AA metabolism, lipid metabolism, carbohydrate metabolism, and oxidative status.Ten DPM and 34 DSP were primarily involved in these pathways and identified as key components.These crucial DPM and DSP with mapped pathways were manually linked together (Figures 7 and 8).

DISCUSSION
Due to the dramatic changes from late pregnancy to early lactation, the transition period is critical in a dairy cow's lifecycle.The imbalance between energy requirement and energy intake may induce severe NEB in dairy cows, which increases the susceptibility to both metabolic and infectious diseases.To cope with the  challenges, comprehensive adaptive mechanisms, including the metabolic, endocrine, and immune system, should be accomplished.Thus, this study used metabolomics and proteomics procedures to reveal an overview of physiological alterations in the liver of dairy cows during the transition period, which should provide a better understanding of the adaptation mechanism and further benefit cows to overcome this challenging time.
To our knowledge, this study was one of the only studies that have investigated the liver samples of transition dairy cows by using metabolomics or proteomics methods.We also recognized that more sampling time points using the same dairy cows might be helpful to capture a data set on the whole dynamic adaptation of the liver to lactation.Moreover, further studies involving the regulation and coordination of metabolic interaction among other sections, such as the nervous system, adipose tissue, skeletal muscle, gut, and mammary gland, are also crucial components for adaptations to lactation.Part of our results revealed by the metabolomics and proteomics methods are in accordance with previous works based on transcriptomic analysis (Ha et al., 2017;Gao et al., 2021), which strengthens the importance of the findings presented in this work and also confirms that some necessary adaptions simultaneously occur in the mRNA, protein, and metabolite levels.
On the other hand, the complementary results from multi-omics can help us to get a more comprehensive understanding of this adaptation process.

Carbohydrate Metabolism
The most prominent feature of the transition period in dairy cows is the imbalance between energy requirement and energy intake, which can induce NEB.During the POSP period, requirements for glucose and metabolizable energy increase 2-to 3-fold more than in the PREP period (Drackley et al., 2001).To meet the energy requirement for maintenance and lactation, the body has to accelerate the carbohydrate metabolism, especially gluconeogenesis, to produce more energy and glucose.A discrepancy of nearly 500 g/d of glucose exists between predicted glucose from digestible energy intake and estimated glucose in POSP dairy cows, which must be made up by increased gluconeogenesis (Drackley et al., 2001).Consistent with former studies (Drackley et al., 2001;Loor, 2010;Laguna et al., 2017), the upregulated rate-limiting enzymes, PCK1 and GK, and increased important intermediates, glucose-6-phosphate (G6P) and fructose-6-phosphate (F6P), indicated upregulated gluconeogenesis in the liver of dairy cows after calving in our study.By using the transcriptomic method, a previous study also identified increased hepatic PCK1 and G6P gene expression and gluconeogenesis to adapt to the transition period in dairy cows (Gao et al., 2021).
By isocitrate dehydrogenase (IDH), the oxidative decarboxylation of isocitrate is catalyzed along with the production of α-ketoglutarate and CO 2 .Three isoforms of IDH exist, namely IDH1, IDH2, and IDH3, all of which localize to the mitochondrion and peroxisome as well as cytosol (Corpas et al., 1999).All of the IDH identified in our study were located in mitochondrion.ATP citrate synthase catalyzes the reversible reaction from phosphate, ADP, acetyl-CoA, and oxaloacetate to ATP, citrate, and CoA (Lill et al., 1982).The upregulated IDH and ACLY, as well as increased essential substrates, fumarate and malate, indicated an increased TCA cycle in the liver after calving, which was in line with previous studies (Da Poian and Castanho, 2015;Luo et al., 2019).Upregulated TCA cycles in the liver and plasma were found in dairy cows immediately to 28 d POSP based on the transcriptomic and metabolomics methods (Luo et al., 2019;Gao et al., 2021;Scharen et al., 2021).
Pyruvate carboxylase catalyzes the physiologically irreversible carboxylation of pyruvate to create oxaloacetate (Da Poian and Castanho, 2015).Pyruvate can be generated from AA metabolism and then be converted to acetyl-CoA, which is also an end product of lipid metabolism.Thus, pyruvate and acetyl-CoA are critical intermediates in carbohydrate, lipid, and AA metabolism (Guo et al., 2016).The upregulated PC, fumarate, and malate were consistent with the augmented pyruvate metabolism and TCA cycle and might indicate the flux of substrates from AA metabolism into the TCA cycle in our study.In accordance, previous studies reported an increased abundance of mRNA for PC around calving (Drackley et al., 2001;Loor et al., 2006;Gao et al., 2021).Reynolds et al. (2003) also emphasized the important role of PC in converting alanine and lactate to glucose in early lactation dairy cows.Consistent with our study, Luo et al. (2019) also found decreased valine, proline, and isoleucine in dairy cows after calving, which served as precursors of the TCA cycle.Amino acids are also substrates for gluconeogenesis, such as alanine, valine, proline, and isoleucine, which can contribute up to 60% glucose in ruminants (Seal and Reynolds, 1993;Da Poian and Castanho, 2015).The increased amplitude of converting alanine to glucose was even greater than converting propionate to glucose in the liver tissue isolated from early lactation dairy cows (Drackley et al., 2001).

Lipid Metabolism
The PPAR signaling pathway is one of the most significant changed paths in our study with 11 DSP, including acyl-CoA synthetase long-chain family member 1 (ACSL1), APOA1, APOA5, CPT1A, CPT2, CYP7A1, GK, matrix metalloproteinase 1 (MMP1), PCK1, perilipin 4 (PLIN4), and SLC27A2, which were mapped in this pathway.Peroxisome proliferator-activated receptors are identified initially as novel members of the nuclear receptors involved in activating the acyl-CoA oxidase gene (ACOX1) promoter encoding the main enzymes of peroxisomal long-chain fatty acids (LCFA: 10-18 carbons long) β-oxidation in ruminants (Bionaz et al., 2013).Peroxisome proliferator-activated receptors also contribute to metabolism pathways such as lipid transport, FA transport, FA oxidation, cholesterol metabolism, adipocyte differentiation, and gluconeogenesis (Bionaz et al., 2013;Hong et al., 2019).Specifically, the abovementioned 11 DSP, which were all upregulated in our study, were involved in cholesterol transport, bile acids synthesis, FA β-oxidation, extracellular matrix breakdown, gluconeogenesis, and lipid storage (Wolins et al., 2003;McCabe et al., 2012;Shi et al., 2018), indicating upregulated lipid metabolism, especially PPAR signaling pathway.A previous study also confirmed the pivotal role of the PPAR signaling pathway in hepatic adaptation to the early POSP period in dairy cows by using the transcriptomic method (Gao et al., 2021).Among the 3 isotypes of PPAR, the mRNA expression of PPARA (PPARα) is predominant in the liver of ruminants (Bionaz et al., 2013).Previous studies showed that activating PPARα controls the catabolism of FA, and the expression of PPARA in the liver of dairy cows increases during the transition period (Loor et al., 2005;Schlegel et al., 2012).In POSP dairy cows, elevated NEFA, especially LCFA, might active PPARs and lead to increased oxidation and decreased esterification of FA in the liver (Grummer, 1995;Drackley, 1999).In addition, it is also reported that PPARs can be activated by glucose in the ruminants (Bionaz et al., 2013).In this study, 9 upregulated DSP, including ACSL1, ACOT8, HAO2, MLYCD, NUDT12, NUDT19, peroxisomal biogenesis factor 11 gamma (PEX11G), phytanoyl-CoA dioxygenase (PHYH), and SLC27A2, were mapped in peroxisomes, which were involved in the α-, β-, and other-oxidation processes of FA.During the transition period, LCFA was the most affected FA in the plasma of dairy cows and was the energy source of the cells (Contreras et al., 2010;Contreras and Sordillo, 2011;Da Poian and Castanho, 2015).The upregulated ACSL in our result was in agreement with the important role of ACSL in the oxidation of LCFA in both peroxisomes and mitochondria.With the amount of NEFA entering the liver increased by multiple times, the peroxisomal pathway is induced as an auxiliary pathway to mitochondrial β-oxidation.Being critical players in the carnitine shuttle system, CPT1A and CPT2 were also upregulated in this study, which was similar to results from previous studies showing that hepatic CPT1 mRNA expression or protein activity increased after calving relative to late pregnancy in dairy cows (Dann and Drackley, 2005;Loor et al., 2005;Loor et al., 2006).This indicated the increased oxidation of FA after calving in both peroxisome and mitochondria in our study.Similarly, a previous study also found that the active expression of PPARA in the liver of transition dairy cows resulted in downstream activation of genes, such as ACSL1, ACOX1, CPT1A, and PCK1, which have key functions in FA oxidation and gluconeogenesis (Loor et al., 2005).
As a sensor and regulator of energy, the AMPK signaling pathway can increase hepatic lipid oxidation by regulating the expression of PPARα and sterol regulatory element-binding protein 1c (SREBP-1c) and then help relieve the NEB in transition dairy cows (Li et al., 2013;Shen et al., 2019).In this work, CPT1A, PCK1, G6P, and F6P were also mapped in the AMPK signaling pathway, indicating the upregulated lipid oxidation, which was consistent with the former results.In accordance with our study, previous studies also reported the upregulated AMPK signaling pathway and suggested its activation effect on the PPAR signaling pathway in transition dairy cows (Li et al., 2013;Ha et al., 2017;Gao et al., 2021).However, the activators of the AMPK signaling pathway still needed further investigation in our study and previous transcriptomic studies.

Ribosome Proteins
Another significantly changed pathway was ribosome with 13 DSP, including RPL7, RPL7A, RPL8, RPL13, RPL19, RPL21, RPL21L1), RPL28, PRL35, RPS6, RPS9, PRS11, and RPS13, which were mapped in this pathway.Ribosomes consist of 2 major components, the small and large ribosomal subunit, and are often associated with the endoplasmic reticulum serving as the site of biological protein synthesis (translation) (Da Poian and Castanho, 2015).All of these 13 DSP were upregulated, which indicated upregulated protein synthesis in this study.As mentioned above, most metabolic processes were upregulated after calving, which requires large amounts of enzymes to participate in these reactions.Given that most enzymes are proteins, there is no wonder that protein synthesis was upregulated after calving.In addition, the number of upregulated DSP after calving was about 2-fold that of upregulated DSP before calving, which was in line with the upregulated protein synthesis after calving.Similarly, previous studies also reported a substantially increased fractional protein synthetic rate in the liver of POSP dairy cows compared with PREP ones (Bell, 1995;Gao et al., 2021).Meanwhile, the increased liver mass (about 9%) might also be a reason for upregulated protein synthesis in POSP dairy cows (Drackley et al., 2001).

Oxidative Status
Even though the lipid and protein mobilization can provide energy-generated substrates to transiently meet the energy requirement of lactation and maintain that requirement in POSP dairy cows, this process may simultaneously produce some reactive oxidative species (ROS).The ROS, including superoxide (O 2 -) and hydrogen peroxide (H 2 O 2 ), are mainly produced during oxidative phosphorylation, the TCA cycle, or intracellular FA oxidation; particularly, peroxisomal β-oxidation leads to considerable quantities of ROS (Schaff et al., 2012;Surai et al., 2019).The β-oxidation of certain types of FA can produce a significant amount of ROS, especially valid for palmitic acid (Contreras and Sordillo, 2011).It was found that GST family members can remove ROS from the liver and that the CYP family, especially the CYP1-3 family enzymes, account for up to 80% of oxidative metabolism (Uehara et al., 2011;Zhang et al., 2019b).According to other studies (Sharma et al., 2011;Abuelo et al., 2013), the downregulated CYP2C18, CYP2D14, GSTM1, GSTM2, GSTM3, and GSTM4 in our study might indicate increased oxidative status and decreased antioxidative defense ability in POSP dairy cows.The increased arachidonate and upregulated ACSL1 were probably involved in ROS generation in this study.Previous studies even showed that severely imbalanced redox would cause oxidative stress in the transition dairy cows, which is related to impaired immune function and subsequently increased susceptibility to production diseases and other health problems (Sordillo and Mavangira, 2014;Abuelo et al., 2019;Liang et al., 2019).

Other Important Metabolic Pathways
Except for mapping to the AMPK signaling pathway, CPT1A, PCK1, and F6P were also mapped to both glucagon signaling and insulin signaling pathways in this study.The glucagon signaling pathway mainly helps glucagon to exert its contribution to increasing blood glucose by the conversion of liver glycogen into glucose (De Koster and Opsomer, 2013;Cardoso et al., 2020).Insulin can elicit different effects on the carbohydrate, lipid, and protein metabolism in dairy cows (De Koster and Opsomer, 2013;Zhang et al., 2019a;Wu et al., 2020).In the liver, insulin has a suppression effect on ketogenesis, gluconeogenesis, glycogenolysis, and protein degradation (De Koster and Opsomer, 2013).In early lactating dairy cows, a transient state of insulin resistance can guarantee glucose enters the mammary gland by limiting glucose used by peripheral tissues such as skeletal muscles and adipose tissue to support lactation (De Koster and Opsomer, 2013;Wu et al., 2020).The simultaneously upregulated glucagon signaling and insulin signaling pathways indicated the significant demand for glucose in POSP dairy cows, which was consistent with former works (Drackley et al., 2001;Zhang et al., 2015;Huang et al., 2017).

CONCLUSIONS
In this work, integrative proteomics and metabolomics techniques were utilized to assess the hepatic adaptation over the transition period in dairy cows.The omics data showed enhanced AA degradation, FA oxidation, AMPK signaling pathway, and PPAR signaling pathway in POSP cows to provide energetic substrates for the TCA cycle and gluconeogenesis.The upregulated glucagon and insulin signaling pathways also indicated the large requirement for energy in POSP dairy cows.As a consequence of increased lipid mobilization and AA and carbohydrate metabolism, oxidative status was elevated, which was highly associated with metabolic and infectious diseases.In addition, the G6P, F6P, CPT1A, and PCK1 might be the critical players participating in carbohydrate and lipid metabolism in that period.Such data, from the view of metabolites and proteins, which is different from the view of previous transcripts, present an integrative comprehension of the physiological metabolics in the liver during the transition period in dairy cows.This should help develop nutritional regulation strategies to further help cows overcome this challenging time.
Figure 1.Comparison of DMI, milk production, and serum metabolites of cows during the transition period (n = 12).Total cholesterol (TC), triglycerides (TG), nonesterified fatty acids (NEFA), and BHB were measured from the serum samples.P-values were determined using the PROC MIXED procedure of SAS with sampling time (week or day) as a fixed effect and cows within time as a random effect.Error bars represent SEM.

Figure 2 .
Figure 2. The pathway analysis of differentially created metabolites recognized in the liver biopsies of dairy cows during the transition period utilizing MetaboAnalyst 4.0 (n = 8).The circles' color from white to yellow to red represents incremental fold change [−log(P)].P-values were determined utilizing the built-in statistical method of MetaboAnalyst 4.0.The circle size from small to large denotes an increase in the pathway impact.

Figure 3 .
Figure 3. Expression patterns of selected protein candidates in the liver biopsies of dairy cows during the transition period (n = 8).A, B, C, and D = western blot of selected protein candidate levels in the liver of dairy cows during the transition period.E = relative selected protein candidates; β-actin protein levels were calculated by a grayscale scan.Data are expressed as mean ± standard error of means.AU = arbitrary unit; CPT1A = carnitine palmitoyltransferase 1A; CPT2 = carnitine O-palmitoyltransferase 2; GSTM3 = glutathione S-transferase Mu 3; PCK1 = phosphoenolpyruvate carboxykinase.Prepartum (PREP) and postpartum (POSP) represent the prepartum and postpartum periods, respectively.

Figure 4 .
Figure 4.The top-most 20 Gene Ontology (GO) terms of differentially synthesized proteins in the liver biopsies of dairy cows during the transition period (n = 8).Green bars represent biological process terms; blue bars represent cellular component terms; red bars represent molecular function terms.
Figure 5. Protein-protein interaction (PPI) network analysis of the upregulated (A) and downregulated (B) differentially synthesized proteins in the liver biopsies (n = 8).Protein-protein interaction network was visualized and analyzed utilizing STRING V.11.0.The nodes in the cluster denote the proteins, and the lines between the nodes represent direct or indirect PPI modes.A purple line shows experimental evidence, a blue line suggests database evidence, and a yellow line shows text mining.

Figure 6 .
Figure 6.The Kyoto Encyclopedia of Genes and Genomes (KEGG) path improvement analysis of downregulated proteins in the liver biopsies of dairy cows during the transition period (n = 8).Only the top 20 paths are presented based on P-value.
Zhang et al.: HEPATIC METABOLIC CHARACTERISTICS OF TRANSITION DAIRY COWS

Table 1 .
Zhang et al.:HEPATIC METABOLIC CHARACTERISTICS OF TRANSITION DAIRY COWS Differentially produced metabolite screening for pathway assessment recognized by GC-TOF/MS in the liver biopsies of dairy cows during the transition period (n = 8) 1

Table 2 .
Important differentially synthesized proteins in the liver biopsies of dairy cows during the transition period (n = 8) 1 1Only the differentially synthesized proteins mapping in the significant Kyoto Encyclopedia of Genes and Genomes (KEGG) database pathways are displayed in the table.FC = fold change; PREP and POSP represent the prepartum and postpartum periods, respectively.2 Determined as the ratios for the tags in the POSP group/PREP group.3 + and −: abundance increased and decreased in the AC group, respectively.4 P-values were determined utilizing the Student's t-test between the PREP and POSP groups.