Altered bile acid and correlations with gut microbiome in transition dairy cows with different glucose and lipid metabolism status

The transition from pregnancy to lactation is critical in dairy cows. Among others, dairy cows experience a metabolic stress due to a large change in glucose and lipid metabolism. Recent studies revealed that bile acids (BA), besides being involved in both the emulsification and solubilization of fats during intestinal absorption, can also affect the metabolism of glucose and lipids, both directly or indirectly by affecting the gut microbiota. Thus, we used untargeted and targeted metabolomics and 16S rRNA sequencing approaches to investigate the concentration of plasma metabolites and BA, the composition of the rectum microbial community, and assess their interaction in transition dairy cows. In Experiment 1, we investigated BA and other blood parameters and gut microbiota in dairy cows without clinical diseases during the transition period, which can be seen as well adapted to the challenge of changed glucose and lipid metabolism. As expected, we detected an increased plasma concentration of β-hydroxybutyrate (BHBA) and nonesteri - fied fatty acids (NEFA) but decreased concentration of glucose, cholesterol, and triglycerides (TG). Untargeted metabolomic analysis of the plasma revealed primary BA biosynthesis was one of the affected pathways, and was consistent with the increased concentration of BA in the plasma. A correlation approach revealed a complex association between BA and microbiota with the host plasma concentration of glucose and lipid metabolites. Among BA, chenodeoxycholic acid derivates such as glycolithocholic acid, taurolithocholic acid, lithocholic acid, taurochenodeoxycholic acid, and taurodeoxycholic acid were the main hub nodes connecting microbe and blood metabolites (such as glucose, TG, and NEFA). In Experiment 2, we investigated early postpartum dairy cows with or without hyperketonemia (HPK). As ex - pected, HPK cows had increased concentration of NEFA and decreased concentrations of glucose and triglycerides. The untargeted metabolomic analysis of the plasma revealed that primary BA biosynthesis was also one of the affected pathways. Even though the BA concentration was similar among the 2 groups, the profiles of taurine conjugated BA changed significantly. A correlation analysis also revealed an association between BA and microbiota with the concentration in plasma of glucose and lipid metabolites (such as BHBA). Among BA, cho - lic acid and its derivates such as taurocholic acid, tauro α-muricholic acid, and taurodeoxycholic acid were the main hub nodes connecting microbe and blood metabolites. Our results indicated an association between BA, intestinal microbe, and glucose and lipid metabolism in transition dairy cows. These findings provide new insight into the adaptation mechanisms of dairy cows during the transition period.


INTRODUCTION
In animal production, ruminants are capable of producing high-quality human edible products by converting human-inedible feedstuffs, which is beneficial for alleviating the increasing demand for food by the continued expansion of the global population.The transition from late gestation to early lactation is the most challenging period for high-producing dairy cows due to a series of stressors, including a large metabolic change, increase oxidative stress, immune dysfunction, and inflammatory-like conditions as well as a change in gastrointestinal microbe community (Drackley et al., 2001, Huang et al., 2014, Lopreiato et al., 2020).In addition, cows are also most susceptible to diseases during the transition period (Wankhade et al., 2017, Ceciliani et al., 2018).Thus, a smooth transition period is critical to the future production and health of the cows and the profitability of dairy producers (Drackley, 1999, Overton et al., 2017, Zhang et al., 2023a).Typical of early postpartum dairy cow is the increase in lipolysis of the adipose tissue caused by a negative energy balance (Grant and Albright, 1995).However, excessive lipolysis is generally associated with hyperketonemia, hepatic lipidosis, and immunosuppression, which are associated with other metabolic diseases such as fatty liver, mastitis, and metritis (Duffield et al., 2009, Hubner et al., 2022, Gu et al., 2023).To some extent, a stable and controlled lipid and glucose metabolism is key to prevent physiological maladaptation during the transition period (Luo et al., 2019, Gao et al., 2021, Zhang et al., 2023a).
Recent studies reported that bile acids (BA) not only are essential for the digestion and absorption of fat, via emulsification and solubilization of dietary fats, but are also involved in lipid, glucose, and energy metabolism mainly through the modulation of farnesoid X receptor and G protein-coupled receptor 5 and their downstream intermediates (Keitel et al., 2019, Shin and Wang, 2019, Zheng et al., 2021).After the synthesis in hepatocytes from cholesterol, BA, especially cholic acid (CA) and chenodeoxycholic acid (CDCA), are conjugated with either taurine or glycine.The BA is stored into the gallbladder and secreted into the intestinal lumen in response to food ingestion.In the intestine, the BA are further biotransformed into secondary BA by microbes (Guzior and Quinn, 2021).Most BA (about 95%) are then reabsorbed in the ileum and enter the portal vein and are rapidly taken up by hepatocytes and re-secreted into bile in human and mice modes (Nie et al., 2015, Winston andTheriot, 2020).This synthesis, secretion, bio-transformation, and reabsorption process is called enterohepatic circulation.To maintain the BA pool homeostasis, the amount of BA synthesized in the liver must equal the amount of BA excreted in feces (about 5%).The reabsorbed BA provides the needed feedback regulation on the hepatic BA synthesis via the farnesoid X receptor (Jia et al., 2021).
Besides being regulated by BA in amount and composition, the primary BA undergo a series of biotransformation such as deconjugation/re-conjugation, dehydrogenation/re-hydrogenation, and dihydroxylation/rehydroxylation into secondary BA by intestinal microbes (Guzior and Quinn, 2021).Bile salt hydrolases (BSH) mediated deconjugation of primary BA occurs rapidly.BSH is widespread in the members of the gut microbiota, especially in Clostridium, Bacteroides, Lactobacillus, Bifidobacterium, and Enterococcus in the phyla of Firmicutes, Bacteroidetes, and Actinobacteria (Jones et al., 2008, Trabelsi et al., 2015).Thus, changes in the abundance and composition of these bacteria by physiological status and diet can affect the amount and composition of BA, as recently demonstrated (Gu et al., 2023, Yin et al., 2024).Meanwhile, BA can directly or indirectly alter gut microbes, and even can select BA-tolerant bacteria (Begley et al., 2005, Devkota et al., 2012, Nie et al., 2015).Taken together, BA can affect the communal structure of intestinal microbes, in turn, intestinal microbes can alter BA.The cross-talk between BA and intestinal microbes can play an important role in prevention of metabolic diseases, as observed in monogastric species (Lin et al., 2019, Collins et al., 2023).
The above information leads us to infer that BA and its interaction with intestinal microbe might play an important role in the regulation of glucose and lipid metabolism in transition dairy cows.Thus, our hypotheses are that: 1) the BA profile changes during the transition period in dairy cows and 2) there is an association between BA profile, intestinal microbes, and glucose and lipid metabolism in dairy cows.The hypotheses were tested by collecting blood and feces from the animals.Due to the BA enterohepatic circulation, it is possible to assess the various BA in blood.In dairy cows, the fecal microbe can be used as a proxy of the hindgut microbiota to some extent (Mao et al., 2015, Ji et al., 2018).
To have a comprehensive view of the BA-microbiome associated with host metabolism status, we used 2 typically different glucose and lipid metabolism statuses in the transition dairy cow in 2 experiments.In Experiment 1, we investigated dairy cows from prepartum (PREP) to postpartum (POSP) periods, and these cows had no clinic disease which indicates that they can well adapt to the challenge of changed glucose and lipid metabolism during the transition period.In Experiment 2, we investigate a group of POSP dairy cows with or without hyperketonemia (HPK), and the HPK cows with greater blood β-hydroxybutyrate (BHBA) concentrations which is a consequence of high concentration of circulating nonesterified fatty acids (NEFA) driven by a severe energy deficiency mainly due to glucose utilization for copious milk synthesis by the mammary gland.

Ethics approval statement
This study was performed in strict accordance with the guidelines of the Administration of Affairs Concerning Experimental Animals (Ministry of Science and Technology, China, revised 2004), and all the procedures were approved by the Institutional Animal Care and Use Committee of the Northwest A&F University.

Zhang et al.: Altered bile acid…
Experiment 1 Twenty multiparous Chinese Holstein dairy cows without clinical diseases having similar 305-d milk yields (10596 ± 1757 kg), parity (2.5 ± 0.76), body condition scores (BCS; 3.46 ± 0.10), and due date were selected from a large cohort of 3000 dairy cows in Modern Farming (Saibei) Co., Ltd.(Zhangjiakou, China) at the time when they arrived in the transition barn (about 22 d before calving).The cows were housed in free-stall barns and had free access to water.A total mixed ration (TMR) was provided daily at 0630, 1230, and 1930 h.The ingredients and chemical composition of the PREP and POSP TMR are provided in Supplemental Table S1.After calving, cows were milked 3 times per day at 0700, 1300, and 2100 h.
Experiment 2 From the same barn of Experiment 1, 14 POSP dairy cows with HPK and 14 POSP dairy cows with healthy conditions (CON) were used in Experiment 2. Blood samples of all cows were collected via the caudal vein 2 h after the morning feeding on 9 ± 2 d after calving.The BHBA concentrations were measured with a handhold BHBA Check Plus Meter and ketone strip (TaiDoc Technology Corporation, Moorestown, NJ, USA) on fresh collected blood.Cows with plasma BHBA concentrations >1.80 mmol/L were considered to have HPK.Cows with no clinical signs of ketosis (including depression and significantly decreased intake and milk production) and normal plasma BHBA concentrations (i.e., < 1.00 mmol/L) were considered to be healthy controls (Duffield et al., 2009, Sun et al., 2014, Hubner et al., 2022).The clinical signs of ketosis were confirmed by the veterinarians of the farm.All cows were 5-21 d in milk and had the same POSP TMR and management regimen as in Experiment 1.

Sample collection and analysis of individual metabolites.
For Experiment 1, blood samples of all cows were collected via the caudal vein at −8 ± 2 d (8 d before expected calving) and 7 ± 1 d (7 d after calving) 2 h after the morning feeding (Zhang et al., 2023b).The glucose concentrations in Experiments 1 and 2 were measured with a OneTouch Verio flex ® glucometer (Johnson & Johnson Services, Inc., New Brunswick, NJ, USA) on fresh collected blood.The measurement of blood BHBA concentrations in Experiments 1 was the same as above described.Plasma was collected using EDTA as an anticoagulant, and then centrifuged at 3,000 × g for 10 min at room temperature.At the time just after blood sampling, fecal samples in Experiments 1 and 2 were collected from the rectum by using sterile swab (FS916, Swwip, Shenzhen Cleanmo Technology Co., Ltd., Shenzhen, China).The aliquots of plasma and fecal samples were flash frozen in liquid nitrogen, and then transferred to −80°C when back to the laboratory.
Plasma metabolites extraction and untargeted metabolomic analysis.After the addition of 400 μL of extract solvent (acetonitrile-methanol, 1:1, containing 0.1% formic acid) into 100 μL plasma samples, the mixture was vortexed for 30 s, sonicated for 10 min in the icewater bath followed by incubation at −40°C for 1 h and centrifugation at 13,800 × g and 4°C for 15 min.The supernatant was transferred to a new tube and dried in a vacuum concentrator at 37°C.The dried samples were reconstituted with 100 μL 50% acetonitrile by sonication for 10 min in the ice-water bath, and then centrifugation at 16,200 × g at 4°C for 15 min.The quality control sample was prepared as an equal aliquot of the supernatants from all of the samples.
The untargeted metabolome analysis was measured by an ultra-high-performance liquid tandem chromatography-quadrupole time of flight/mass spectrometry analysis (UHPLC-QTOF/MS).The UHPLC separation was carried out using an Agilent 1290 Infinity series UHPLC System (Agilent Technologies, Santa Clara, CA, USA), equipped with a UPLC BEH Amide column (2.1 × 100 mm, 1.7 μm, Waters Corporation, Milford, MA, USA).The mobile phase consisted of 25 mmol/L ammonium acetate and 25 mmol/L ammonia hydroxide in water (pH = 9.75; phase A) and acetonitrile (phase B).The elution gradient was set as follows: 0.0-1.0min, 95% B; 1.0-14.0min, 65-95% B; 14.0-16.0min, 40-65% B; 16.0-18.0Bile acids standard solution preparation and plasma bile acids target metabolomic analysis.The bile acids target metabolomic analysis was measured by a UHPLCparallel reaction monitoring-MS method by Shanghai Biotree Biotech Co., Ltd.(Shanghai, China), as described in a previous study (Han et al., 2015).Stock solutions of 41 bile acids (purchased from Shanghai Zzbio Co., Ltd., Shanghai, China or Steraloids Inc., Newport, USA with 95-100% purity) were individually prepared by dissolving or diluting each standard substance to give a final concentration of 10 mmol/L.An aliquot of each of the stock solutions was transferred to a 10 mL flask to form a mixed working standard solution.A series of calibration standard solutions were then prepared by stepwise dilution of this mixed standard solution (containing the isotopically-labeled internal standard mixture in identical concentrations with the samples).
The UHPLC separation was carried out using an UH-PLC System (Vanquish, Thermo Fisher Scientific, San Jose, CA), equipped with a UPLC BEH C18 column (150 × 2.1 mm, 1.7 μm, Waters).The mobile phase A was 1 mmol/L ammonium acetate and 0.1% acetic acid in the water, and the mobile phase B was acetonitrile.The column temperature was set at 50°C.The auto-sampler temperature was set at 4°C and the injection volume was 1 μL.A Q Exactive HFX mass spectrometer (Thermo Fisher Scientific) was applied for assay development.Typical ion source parameters were: spray voltage = +3500/-3100 V, sheath gas (N 2 ) flow rate = 40 L/min, aux gas (N 2 ) flow rate = 15 L/min at 350°C, and capillary temperature = 320°C, with no sweep gas.The parallel reaction monitoring parameters for each of the targeted analytes were optimized, by injecting the standard solutions of the individual analytes into the API source of the mass spectrometer.Since most of the analytes did not show product ion acceptable for quantification, the precursor ion in high resolution was selected for quantification.
Working standard solutions were subjected to UPLC-PRM-MS/MS analysis using the methods described above.The standard curves were built by using the peak areas ratio for the analyte: internal standard as y, and the concentration (nmol/L) for the analyte as x.The least squares method was used for the regression fitting in Excel.The optimal accuracy and correlation coefficient (R 2 ) were obtained when applied 1/x as the weight.
The lower limits of detection (LLODs) and lower limits of quantitation (LLOQs) were determined by the signal-to-noise ratios (S/N) of the corresponding analyte.The LLODs and LLOQs were defined as the analyte concentrations that led to peaks with S/N of 3 and 10, respectively, according to the guideline for bioanalytical method validation (Brodie and Hill, 2002).The values below the LLOD were considered to be 0 while the values between LLOD and LLOQ were considered to be detectable.Metabolites with >80% of samples with 0 values were excluded from further analysis (Smilde et al., 2005).The precision of the quantitation was measured as the relative standard deviation, determined by injecting analytical replicates of the quality control sample.The percent recovery was calculated as [(mean observed concentration)/(spiked concentration)] × 100%, which was also used to measure the accuracy of quantitation.
Plasma metabolomic data analysis.For both untargeted and targeted metabolomic analysis, the peak outliers were first filtered based on the coefficient of variation and null values of every peak.After normalization by total ion current, the peaks were fed into the SIMCA software package (V16.0.2,Sartorius Stedim Data Analytics AB, Umea, Sweden) for principal component analysis and orthogonal projections to latent structures-discriminant analysis (OPLS-DA).The 7-fold cross-validation, interpretability, predictability, and permutation test (n = 200) were used to measure the effectiveness of the model in OPLS-DA.The parameters of variable importance projection value >1.0 in OPLS-DA and P < 0.05 in students' t-test were used as criteria to identify differently produced metabolites (DPM) between 2 time points or groups.The pathway analysis of different metabolites was done using MetaboAnalyst 5.0 with Bos Taurus as a pathway library, and the P-value and pathway impact were obtained by topology analysis based on weighted calculation (Pang et al., 2021).
Fecal bacterial community analysis.The genomic DNA was extracted from the fecal samples of 26 cows (n = 13/time point) from Experiment 1 and 28 cows (n = 14/group) from Experiment 2 by using a PowerSoil ® DNA Isolation Kit (MoBio Laboratories, Carlsbad, CA), following the manufacturer's protocol.The DNA quality test, PCR amplification, sequencing, and analysis were as previously described (Zhang et al., 2017, Zhang et al., 2023b).Briefly, The V3-V4 region of the bacteria 16S rRNA gene was amplified using forward primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) (Zhang et al., 2020) and reverse primers 806R (5′-GTGGACTACH-VGGGTWTCTAAT-3′) (Zhang et al., 2017).Sequencing was done on an Illumina MiSeq PE300 paired-end platform by Beijing Allwegene Technology Co., Ltd.(Bei- jing, China).The sequences were analyzed using QIIME 2 (https: / / qiime2 .org)according to the inbuilt pipeline (Bolyen et al., 2019).The operational taxonomic unit was denominated according to the Silva bacteria database.One sample from Experiment 2 with quite low numbers of operational taxonomic unit was excluded from the subsequent analyses.Chao1, observed species, PD whole tree, and Shannon were used to estimate α diversity.After estimating the β diversity in QIIME2, the non-metric multidimensional scaling (NMDS) analysis derived from a Bray-Curtis dissimilarity matrix was conducted in R (version 3.6.0).The stress value in NMDS meant the representativeness of the analysis in terms of reduced dimensions with stress <0.2 as a good representation (Chong et al., 2020).Its ordination measure, analysis of similarity (ANOSIM) which tests whether within-group distances are greater or equal to between-group distances, was also conducted similarly in R. A greater similarity withingroup and greater difference between-group, which is indicated by R value, and P-value <0.05 in ANOSIM meant there is a significant difference between groups (Chong et al., 2020).Variables of microbial communities were analyzed with linear discriminant analysis effect size (LEfSe) analysis to identify significantly changed bacteria between groups using the criterion of linear discriminant analysis score >3.0 (Segata et al., 2011).
After assembling and filtering, the raw sequence was submitted to the NCBI Sequence Read Archive (SRA; http: / / www .ncbi.nlm.nih.gov/Traces/ sra/ ), under accession number PRJNA835200 and PRJNA835215.

Statistical analysis.
The power analysis was carried out in Experiment 1 using blood NEFA concentration as the primary response variable to obtain a power of 0.80 under the significance level of 0.05, with the estimated population variance of 0.14 and the to-be-detected difference of 0.74 mmol/L between PREP and POSP cows (Gao et al., 2021).For Experiment 2, the blood BHBA concentration was used as the primary response variable to obtain a power of 0.80 under the significance level of 0.05, with the estimated population variance of 0.15 and the to-be-detected difference of 0.40 mmol/L between CON and HPK cows (Sun et al., 2014).The concentrations of plasma routine metabolites, and concentrations and proportions of BA data were first checked for normality and outliers using the UNIVARIATE procedure of SAS (version 9.4, SAS Institute Inc., Cary, NC), and then analyzed using the PROC MIXED procedure.Covariance structures, including autoregressive type 1, compound symmetry, and unstructured, were used in the covariance structures model based on the lowest Akaike information criterion (Littell et al., 1998).In Experiment 1, the sampling time points (PREP and POSP) were used as fixed effect and cows within time as a random effect.In Experiment 2, the conditions (CON and HPK) were used as fixed effect and cows within condition as a random effect.
Significantly changed blood routine metabolites, BA (proportion level), and fecal microbiota were used for correlation analysis in R (v.4.2.0).Spearman's rank correlations and q-values were calculated using the psych packages in R. Correlations were identified with a significance threshold of |r| > 0.40 and FDR correct P-value (q-value) < 0.05 from the Spearman's correlation data.The network was built according to the correlations and visualized using the Cytoscape software (v.3.7.2).
Results were reported as least squares means.The level of statistical significance was set at P < 0.05 or q-value <0.05.A tendency for significance was declared at 0.05 ≤ P < 0.10 or 0.05 ≤ q < 0.10.

Plasma individual metabolites in Experiment 1
Compared with the PREP, the POSP had lower (P ≤ 0.02) concentrations of glucose, TC, TG, albumin, and PUN and alanine aminotransferase activity but greater (P < 0.01) concentrations of NEFA, BHBA, TBA, TBIL, and DBIL and AST activity (Table 1).

Plasma metabolites revealed by the untargeted metabolomics approach in Experiment 1
In total, 91 peaks were obtained from all 40 samples and 59 peaks were confirmed as valid peaks after filter and normalization.Data of the UHPLC-QTOF/MS spectra among the 2 groups analyzed using principal component analysis score plots are listed in Figure S1A.The interpretability of Y (R 2 Y) and predictability of OPLS-DA models in POSP vs. PREP was 0.951 and 0.897 (Figure S1B), respectively, indicating satisfactory effectiveness of the model, which can be used to identify the difference between the 2 groups.The OPLS-DA score result is shown in Figure S1C, and all the samples in the score plots were within 95% of Hotelling's T 2 ellipse.
A total of 20 DPM were obtained from the comparison, with 15 increased and 5 decreased in POSP as compared with PREP (Figure S1D), and most of them belong to amino acids, peptides, and amino acid analogs at the subclass level (Table S2).According to pathway analysis of these DPM, phenylalanine, tyrosine and tryptophan biosynthesis, taurine and hypotaurine metabolism, glycine, serine and threonine metabolism, tyrosine metabolism, glyoxylate and dicarboxylate metabolism, β-alanine metabolism, glutathione metabolism, and primary BA biosynthesis, were affected (enrichment of q < 0.05 and Zhang et al.: Altered bile acid… pathway impact ≥0.07) by the transition period in dairy cows (Figure 1A).

Plasma bile acids profiles in Experiment 1
The LLODs and LLOQs of analytes ranged from 0.12 to 7.81 nmol/L and 0.24-15.62nmol/L, respectively (Table S3).The recoveries determined were 79.5-114.2%for all the analytes, with all the relative standard deviations below 19.0% (n = 5).In total, 24 bile acids, with 6 primary and 13 secondary BA, were detected from 40 plasma samples.However, 4 BA were detectable in very few samples (3-5 samples) and were excluded from the subsequent analyses.Among the detected BA, the concentrations of 12 BA were greater (P ≤ 0.03) in the POSP than in the PREP, and the concentrations of lithocholic acid (LCA) and taurochenodeoxycholic acid (TCDCA) tended to be greater (P = 0.05 and P = 0.07, respectively; Table S4).The proportions of GCA, glycolithocholic acid (GLCA), taurolithocholic acid (TLCA), and LCA were greater (P ≤ 0.01) in the POSP than that in the PREP, while the proportion of TDCA and TCDCA had the opposite trend (P ≤ 0.03; Figure 1B and C).
The primary BA such as CA, glycocholic acid (GCA), taurocholic acid (TCA), taurodeoxycholic acid (TDCA), glycochenodeoxycholic acid (GCDCA), and CDCA made up 80.9% and 84.4% of the total detected bile acids in the PREP and POSP, respectively (Table 2 and Figure 1B and C).The CA, GCA, TCA, TDCA, and glycodeoxycholic acid (GDCA) were the 5 predominant BA in both PREP and POSP (Figure 1B and C).The concentration of all BA categories was greater (P ≤ 0.05) in the POSP than in the PREP (Table 2).The proportion of primary BA and glycine conjugated BA was greater (P ≤ 0.05) while the proportion of secondary BA was lower in POSP vs. PREP (Table 2).The proportion of other categories of BA was not different (Table 2).

Fecal bacterial community in Experiment 1
For the α diversity, the POSP had a greater observed species index value than the PREP, while Chao1, PD whole tree, and Shannon index values were not affected (Figure 2A-D).The good's coverage of all the samples was greater than 0.972.For the β diversity, there was a significant difference between PREP and POSP based on the NMDS (stress = 0.1121; Figure 2E) and ANOSIM (R = 0.552, P = 0.001).

Plasma routine metabolites in Experiment 2
Compared with the CON, the HPK had lower (P ≤ 0.04) concentrations of glucose, TG, and PUN and a tendency (0.05 ≤ P < 0.10) for a lower concentration of TC and albumin: globulin, but greater (P ≤ 0.03) concentrations of BHBA, NEFA, TBIL, and DBIL and AST activity (Table 3).

Plasma metabolites revealed by the untargeted metabolomics approach in Experiment 2
In total, 548 peaks were confirmed as valid peaks after filter and normalization from 28 samples from the positive and negative ion models.The interpretability of Y (R 2 Y) and predictability of OPLS-DA models in HPK vs. CON was 0.963 and 0.533, respectively, indicating satisfactory effectiveness of the model, which can be used to identify the difference between the 2 groups (Figure S3B).A total of 143 DPM were identified, with 48 being elevated and 95 decreased in HPK as compared with CON (Figure S3D).Most of those DPM were amino acids, peptides, and amino acid analogs, lipids and lipid-like molecules, organic acids and derivatives, and organoheterocyclic compounds at the subclass level (Table S5).
According to the pathway analysis of these DPM, 18 pathways were significantly affected (q < 0.05 and pathway impact ≥0.07) by ketosis in dairy cows (Figure 4A).Among them, the taurine and hypotaurine metabolism pathway, which were enriched by taurine and TCA, had the greatest pathway impact at 0.43.Meanwhile, taurine and TCA as well as TCDCA were mapped in primary BA biosynthesis with a pathway impact of 0.07.

Plasma bile acids profiles in Experiment 2
In total, 27 BA, with 10 primary and 14 secondary BA, were detected from 26 plasma samples.The concentrations of 3-dehydrocholic acid, TDCA, tauro α-muricholic acid (T-α-MCA), and TCA were greater (P = 0.01) while the concentration of TCDCA tended (P = 0.06) to be greater and the concentration of CDCA and chenodeoxycholic acid-3-sulfate tended (P < 0.10) to be lower in HPK vs. CON (Table S6).The proportions of 3-dehydrocholic acid, TDCA, T-α-MCA, and TCA were greater (P ≤ 0.01) in the HPK group than that in the CON, while the proportion of CA had the opposite trend (P = 0.03; Figure 4B and C).
The BA, CA, GCA, TCA, GDCA, GCDCA, 7-ketodeoxycholic acid, TDCA, and TCDCA were the 8 predominant BA in both the CON and HPK, which made up 96.3% and 96.8% of the total BA in the them, respectively (Figure 4B and C).The concentration of free BA The pathway was plotted by using MetaboAnalyst 5.0.The color of the circles from white to yellow to red denotes incremental −log(p).The size of the circles from small to large indicates an increment of pathway impact, that correspond to the perturbation of the pathway based on the summarized normalized topology measures of those perturbed metabolites in each pathway.(B and C) All bile acids that could be assessed were considered as 100% in both graphs; those bile acids with a share of less than 0.25% were classified as "others."decreased in the HPK (P = 0.03), but the concentration of taurine-conjugated BA increased (P = 0.01) and the concentration of conjugated BA and CA derivatives tended to increase (P = 0.06) compared with the CON (Table 4).The proportion of taurine conjugated BA and CA derivates (P ≤ 0.02) were higher while proportion of conjugated and free BA was lower (P = 0.05) in HPK vs. CON (Table 4).

Fecal bacterial community in Experiment 2
For the α diversity, no significant differences were found in the Chao1, PD whole tree, and Shannon index values between CON and HPK (Figure 5A-D).The good's coverage of all the samples was greater than 0.990.For the β diversity, there was a significant difference between CON and HPK based on the NMDS (stress = 0.18; Figure 5E) and ANOSIM (R = 0.11, P = 0.02).

DISCUSSION
To cope with the challenges associated with the transition period, dairy cows need to activate all the metabolic pathways particularly the glucose and lipid metabolism to alleviate the negative energy balance (Drackley et al., 2001).Previous studies have revealed a role of BA in metabolic disorders (e.g., obesity and diabetes in both mice and human) (Qi et al., 2015, So et al., 2020, Li et al., 2021).Indeed, previous studies reported changed hepatic BA biosynthesis and serum BA metabolism pathways during the transition from pregnancy to lactation in dairy cows (Ghaffari et al., 2023, Zhang et al., 2023a, Ghaffari et al., 2024).In addition, the microbiota of the digestive tract is known to influence the metabolism of the host, particularly glucose and lipid metabolism (Liu et al., 2017, Gu et al., 2023).However, the complete BA profile, BA-gastrointestinal tract (GIT) microbe interaction, and their relationship with host metabolism status in transition dairy cows are still not fully understood.In the current study, we applied untargeted and targeted metabolomic and 16S rRNA sequencing approaches to investigate the blood BA, metabolites, fecal microbiota, and their interactions in the transition dairy cows.

Experiment 1
The decreased glucose and increased NEFA and BHBA concentrations in plasma indicated that cows in our experiment experienced the typical energy deficiency and lipid mobilization of the early POSP.The significant increase in AST, TBIL, and DBIL and decreased albumin observed in our study is also somewhat typical of POSP cows (Drackley et al., 2001, Zhang et al., 2015, Ha et al., 2017), likely driven by an increase in inflammatory-like conditions (Bionaz et al., 2007).It has been previously demonstrated that the increase in TBIL and the decrease in albumin in plasma POSP is due to the acute phase reaction of the liver to inflammatory-like conditions in early POSP cows (Bertoni et al., 2008, Trevisi andMinuti, 2018).
Our analysis identified 8 pathways including the following significantly affected metabolites from PREP to POSP: tyrosine, glycine, betaine, ureidopropionic acid, GCA, and taurine.Tyrosine, which was downregulated in POSP vs. PREP, is part of the phenylalanine, tyrosine and tryptophan biosynthesis and tyrosine metabolism pathways.As one of the important ketogenic and glycogenic amino acids in non-essential amino acids, the reduced tyrosine postpartum is likely due to the use of this amino acid for energy production in the liver and for protein synthesis in the mammary gland (Luo et al., 2019, Zhang et al., 2023a).The increased glycine, GCA, and taurine enriched in primary BA biosynthesis might indicate that more BA were biosynthesized, or less BA were reabsorbed, or both after calving, which is in agreement with greater plasma TBA content in the POSP.Most of the identified BA by targeted metabolomic analysis were more abundant in the POSP than that in the PREP, and the total quantity of these 20 BA was close to the number of TBA measured by the commercial kit (Hunan Yonghe-Sun Biotechnology Co., Ltd., Changsha, China), indicating a good accuracy of BA target metabolomic analysis.The increased level of BA can play a role in modulating the glucose and lipid metabolism (Nie et al., 2015).The increased proportion of primary BA and decreased proportion of secondary BA in plasma of POSP vs. PREP may indicate that less primary BA was converted into secondary BA in the intestine as the intestinal BA will be re-absorbed into the blood (Jia et al., 2021).Interestingly, the large majority of BA with a decreased proportion were conjugated CDCA (e.g., TCDCA) and their secondary derivates (e.g., LCA, GLCA, and TCLA), indicating a decrease activity of the alternative pathway (producing the non-12α-hydroxylated primary BA, CDCA) in POSP dairy cows (Jia et al., 2021).This is of importance, as those compounds are known to affect body glucose and lipid metabolism and they can be also cytotoxic molecules leading to oxidative stress and other diseases (Winston and Theriot, 2020); however, other secondary BA, such as UCDA, can alleviate intestinal inflammation in low birth weight piglets and can confer diarrhea resistance in newborn dairy calves (He et al., 2022, Pi et al., 2023).
Bile acids can interact with intestinal microbiota to regulate body metabolism (Nie et al., 2015, Collins et al., 2023).On one hand, BA are potent antimicrobials and are known to have the ability to impact susceptible bacteria, such as Balantidium, Enterococcus, Pneumococcus, and Staphylococcus as well as members of the phylum Spirochaetes, by both bacteriostatic and bactericidal manners (Stacey andWebb, 1947, Guzior andQuinn, 2021).Primary BA disrupt bacterial membranes in a dose-dependent manner and non-conjugated BA have a greater effect on bacterial membranes compared with their conjugated counterparts (Kurdi et al., 2006, Sannasiddappa et al., 2017, Guzior and Quinn, 2021).The reduction in bacterial growth by BA has been validated in several species of Staphylococcus, Lactobacillus, and Bifidobacterium (Kurdi et al., 2006, Sannasiddappa et al., 2017, Guzior and Quinn, 2021).A previous study reported that the difference in BA composition can explain about 37% of the difference in the GIT microbial community (Zhang et al., 2018b).
Gut microbiota has 4 distinct pathways to transform BA, such as deconjugation, dehydroxylation, oxidation, and epimerization (Winston and Theriot, 2020, Guzior and Quinn, 2021, Jia et al., 2021).A previous genome-centric study revealed that over 38% of the microbial population participated in BA deconjugation, oxidation, and dehydroxylation pathways within the intestine of dairy cows, especially the large intestine (Lin et al., 2023).Microbial BSH, an enzyme that deconjugates both glycine-and taurine-bound primary BA, appears to be important in health-related BA (Guzior and Quinn, 2021).A previous metagenomic study reported functional BSH is present in all major bacterial divisions, though the activity of these enzymes varies (Jones et al., 2008).Among them, Bacteroides, Blautia, Bifidobacterium, Clostridium, Enterococcus, Eubacterium, Lactobacillus, and Roseburia were widely investigated, and approximately 59.7% of BSH gene-containing bacteria belonged to Firmicutes (Jones et al., 2008, Song et al., 2019).
In this study, we reported increased relative abundances of Bacteroides and Roseburia in POSP cows, which were also the predominant bacteria in the hindgut of dairy cows as found in previous studies (Zhang et al., 2018a, Bach et al., 2019).Previous studies suggested that Bacteroides spp.can encode 12α-hydroxysteroid dehydrogenases and play a major role in deconjugating BA (Doden et al., 2018, Harris et al., 2018, Guzior and Quinn, 2021).These might be the reason for the negative association between Bacteroides and BA such as GLCA, TLCA, and LCA in this study.Previously, Alistipes was classified as a member of the genus Bacteroides, which indicated that Alistipes may have a similar function as Bacteroides in the GIT (Zhang et al., 2018a).Currently, Alistipes are classified in the family of Rikenellaceae with the genus of Rikenellaceae RC9 gut group, and both of them were the predominant bacteria and increased in POSP vs. PREP cows.A recent genome-centric investigation showed that Alistipes was the top (7.1% of all the metagenome-assembled genomes) BSH-carrying genus in the GIT of dairy cows, which indicated that it may play an important role in the GIT as the distinct BA tolerance ability (Lin et al., 2023).
Through producing butyrate as one of the main end products, Roseburia spp.play a major role in maintaining gut health and immune defense (Tamanai-Shacoori et al., 2017).The phylogenetic analysis showed a close relationship between Roseburia and Clostridium (Tamanai-Shacoori et al., 2017).As Clostridium spp. was one of the main sources of BSH (Song et al., 2019), it is reasonable to infer that Roseburia also have BSH, and that seems one of the reasons that Roseburia was negatively associated with TLCA in our results.Moreover, Roseburia spp.could be used as a biomarker for symptomatic gallstone formation with higher fecal BA concentrations, which indicated that Roseburia might have an important role in BA metabolism (Tamanai-Shacoori et al., 2017).Belonging in the same family with Roseburia, Butyrivibrio also displayed a bile tolerance capability which might contribute to the negative association with TLCA in this study.
Recently, Akkermansia, which can ameliorate metabolic diseases, improve the gut barrier, and lead to the reshaping of the gut microbiota composition, was seen as the next-generation biotherapeutic agent (Rao et al., 2021, Kalia et al., 2022).In addition, Akkermansia was also reported to efficiently increase lipid oxidation and BA metabolism (Rao et al., 2021).A positive association between circulating NEFA and BHBA concentrations and the relative abundance of Akkermansia was revealed in a prior study suggesting that Akkermansia contributes to intensifying lipolysis in dairy cows (Luo et al., 2022), which was consistent with our study founding a positive correlation between blood NEFA concentration and GIT Akkermansia abundance.The size of the circles from small to large indicates an increment of pathway impact, that correspond to the perturbation of the pathway based on the summarized normalized topology measures of those perturbed metabolites in each pathway.(B and C) All bile acids that could be assessed were considered as 100% in both graphs; those bile acids with a share of less than 0.25% were classified as "others."The samples were taken 9 ± 2 d after calving.
Except for the BA-microbe interaction, diet and host were other main factors affecting the composition of GIT microbiota (Meale et al., 2016, Zhang et al., 2018a, Liu et al., 2019).In the POSP dairy cows, the diets contained more protein, starch, EE, and energy but less NDF than the diets in the PREP dairy cows.The enriched nonstructural carbohydrate in the diets may have led to the increased in relative abundance of Treponema and members in Prevotellaceae, consistent with previous studies (Zhang et al., 2018a, Bach et al., 2019).On the contrary, the structural carbohydrate degradation-related bacteria, such as Monoglobus, Christensenellaceae R-7 group, Oscillospiraceae NK4A214 group, Butyricicoccaceae UCG-009, and Family XIII AD3011 group, were decreased in POSP cows, which was similar with previous studies (Lima et al., 2015, Bach et al., 2019).

Experiment 2
The changed blood routine metabolites indicated that HPK cows were experiencing severe energy deficiency and lipid mobilization, especially the elevated NEFA content meaning greater lipids entered the liver to be metabolized (Nicola et al., 2022).Similar to Experiment 1, a complex network among the interaction of BA and microbiota and host glucose and lipid metabolism was revealed in Experiment 2, and BA seems to be the hub of this network.The family Christensenellaceae has been reported to be associated with intestinal health and decreased inflammation (Goodrich et al., 2014, Bach et al., 2019).A previous study also reported that obese humans lack Christensenella and mice receiving Christensenella had lower lipidosis (Goodrich et al., 2014), in agreement with those data in our study the HPK cows with high lipolysis had lower Christensenella R-7 group.The exact mechanism of the effects of Christensenella on lipid metabolism needs further investigation.
As mentioned previously, Clostridium had more BSH activity and participate in the formation of secondary BA, such as DCA from CA and LCA from CDCA (Winston and Theriot, 2020), which might contribute to the negative relationship between the relative abundance of Clostridium sensu stricto 1 and proportion of TDCA in our study.Paeniclostridium, previously known as Clostridium, might also have the BSH activity and participate in the formation of secondary BA, which can be one of the reasons that had similar association with BA as Clostridium.
Romboutsia, which is a member of the family Peptostreptococcaceae, has been previously associated with glucose and lipid metabolism pathways, such as cholesterol metabolism, regulation of lipolysis in adipocytes, insulin resistance, fat digestion and absorption, thermogenesis, and blood glucose control (Therdtatha et al., 2021, Yin et al., 2023).In our study we detected a positive association between Romboutsia abundance and plasma glucose (r = 0.44, q = 0.02) and a negative association between Romboutsia abundance and plasma BHBA (r = −0.45,q = 0.02).Previous studies also found positive correlations between Romboutsia and 7-oxo-CDCA, 3-oxo-CDCA, 3,7-dioxo-CDCA, and other primary BA, which indicate that Romboutsia might be involved in the transformation of CDCA (Therdtatha et al., 2021, Porru et al., 2022).One genomic study revealed that Romboutsia had a specific gene encoding conjugated bile acid hydrolases, which is involved in the hydrolysis of the amide linkage in conjugated bile salts, releasing primary BA (Gerritsen et al., 2017).This might explain the negative correlation between Romboutsia and TDCA (r = −0.65,q < 0.01) and TCA (r = −0.73,q < 0.01) and the positive correlation between Romboutsia and CA (r = 0.55, q < 0.01).By using quantitative trait locus analysis in mice, SLC10A2 that regulates both the abundance of Turicibacter and plasma levels of CA was identified as a locus containing the gene for the ileal BA transporter (Kemis et al., 2019).From an in vitro culture experiment, they also found that Turicibacter had an excellent deconjugate ability to transform ~96-100% TCA within 24 h, which indicated bacterial 7α-hydroxysteroid dehydrogenase activity.Consistent with our result, they also reported that Turicibacter abundance was negatively correlated with TCA levels.In addition, a previous study also revealed that host association with Turicibacter can alter intestine expression of multiple gene pathways, including those important for lipid and steroid metabolism, with corresponding reductions in host systemic triglyceride levels and inguinal adipocyte size (Fung et al., 2019).Even though our study also reported lower plasma TG levels with lower Turicibacter abundance, we cannot illustrate whether this effect on lipid metabolism was mainly through the interaction between the host and Turicibacter or not.

Limitations.
To our knowledge, this is one of the first works investigating the BA profiles and their potential interaction with the intestinal microbe and host metabolism in dairy cows.Although our investigation attempted to provide a comprehensive insight into the potential contribution of the BA-microbiome interaction on host metabolism, we also recognized that there are several limitations in our study.We used correlations to assess the interaction between metabolites, BA, and microbiota.Correlation is not causation.We used the blood BA profiles to represent the body's BA metabolism and fecal microbiota to represent the intestinal microbe community in this study which are both important limitations.As the existence of enterohepatic circulation, we cannot get a clear clue for the cause of the increase in plasmatic BA in early postpartum cows.Future work is needed to investigate the ability of BA synthesis in the liver and the reabsorbed of BA in the hindgut.More sampling during the study would also have strengthen the use of correlations, helping to identify some biological association.

CONCLUSIONS
In summary, the present study confirmed alterations of the glucose and lipid metabolism, BA profiles, and intestinal microbiota in transition cows.The correlation between BA profiles and intestinal microbiota revealed by our study suggest an association of CDCA metabolism, particularly TCDCA, GLCA, TLCA, and LCA, with some microbes and routine metabolites of the metabolism in the host (such as NEFA, TG, and glucose) in transition dairy cows.Cholic acid metabolism-related TCA and TDCA were the main hub nodes connecting microbe and host glucose and lipid metabolism routine metabolites (such as BHBA) in dairy cows with HPK.A better understanding of the metabolic and microbial adaptations and their interaction during the transition period might be helpful to find suitable regulatory targets for supporting dairy cows in adapting to the metabolic challenges related to the onset of lactation.

ACKNOWLEDGMENTS
This project was supported by the National Natural Science Foundation of China (grant numbers 32102570 and 32072761), the fellowship of China Postdoctoral Science Foundation (grant number 2021M702691), and the "Double First-Class" Funding for Animal Husbandry in China (grant number Z1010222001).The funding body has not participated in or interfered with the research.We would like to thank Shanglin Yu, Qiao Zhou, Guangfu Tang, and other members in Dr. Yao' lab for helping in sampling, measuring, visualizing, and discussion.
Figure 1.Pathway analysis of differentially produced metabolites (A) and proportions of bile acids in the plasma of dairy cows sampling on −8 ± 2 d before (B) and 7 ± 1 d after (C) calving (n = 20/time point).(A)The pathway was plotted by using MetaboAnalyst 5.0.The color of the circles from white to yellow to red denotes incremental −log(p).The size of the circles from small to large indicates an increment of pathway impact, that correspond to the perturbation of the pathway based on the summarized normalized topology measures of those perturbed metabolites in each pathway.(B and C) All bile acids that could be assessed were considered as 100% in both graphs; those bile acids with a share of less than 0.25% were classified as "others." Figure 2. Alpha and β diversity of fecal bacteria in dairy cows before and after calving (n = 20/time point).Fecal bacteria α diversity was estimated by the Chao1 value (A), Observed species (B), PD whole tree (C), and Shannon index (D).Fecal bacteria β diversity was estimated by the non-metric multidimensional scaling (NMDS; E).Boxes represent the interquartile range (IQR) between the first and third quartiles (25th and 75th percentiles, respectively), and the horizontal line inside the box defines the median.Whiskers represent the lowest and highest values within 1.5 times the IQR from the first and third quartiles, respectively.Boxes with different letters above their whiskers are different (P < 0.05) between groups.PREP and POSP, sampling on −8 ± 2 d before and 7 ± 1 d after calving, respectively.

Figure 3 .
Figure 3. Fecal bacteria at the phylum (A) and genus (B) levels and the relative abundance (%; more than 0.01%) of significantly changed taxa between POSP and PREP in Experiment 1 (C) (n = 20/time point).Only the top 15 relative abundance of bacteria at the genus level were shown in panel B. Only the relative abundance greater than 0.01% genus was shown in panel C.The phylum and family names were also shown with the corresponding genus.LDA, Linear discriminant analysis.PREP and POSP, sampling on −8 ± 2 d before and 7 ± 1 d after calving, respectively.
Figure 4. Pathway analysis of differentially produced metabolites (A) and proportions of bile acids in the plasma of dairy cows without (B) or with (C) hyperketonemia (n = 14/group).(A) The pathway was plotted by using MetaboAnalyst 5.0.The color of the circles from white to yellow to red denotes incremental enrichment (−log(p)).The size of the circles from small to large indicates an increment of pathway impact, that correspond to the perturbation of the pathway based on the summarized normalized topology measures of those perturbed metabolites in each pathway.(B and C) All bile acids that could be assessed were considered as 100% in both graphs; those bile acids with a share of less than 0.25% were classified as "others."The samples were taken 9 ± 2 d after calving.

Figure 5 .
Figure 5. Alpha and β diversity of fecal bacteria in dairy cows with or without hyperketonemia (n = 14/group).Fecal bacteria α diversity was estimated by the Chao1 value (A), Observed species (B), PD whole tree (C), and Shannon index (D).Fecal bacteria β diversity was estimated by the non-metric multidimensional scaling (NMDS; E).Boxes represent the interquartile range (IQR) between the first and third quartiles (25th and 75th percentiles, respectively), and the horizontal line inside the box defines the median.Whiskers represent the lowest and highest values within 1.5 times the IQR from the first and third quartiles, respectively.Boxes with different letters above their whiskers are different (P < 0.05) between groups.CON, healthy dairy cows; HPK, hyperketonemia dairy cows.The samples were taken 9 ± 2 d after calving.

Figure 6 .
Figure 6.Fecal bacteria at the phylum (A) and genus (B) levels and the relative abundance (%) of significantly changed taxa between HPK and CON in Experiment 2 (C) (n = 14/group).Only the top 15 relative abundance of bacteria at the genus level are shown in panel B. The phylum and family names are also shown with the corresponding genus.LDA, Linear discriminant analysis.CON, healthy dairy cows; HPK, hyperketonemia dairy cows.The samples were taken 9 ± 2 d after calving.

Figure 7 .
Figure 7.The network among significantly changed plasma bile acids proportion, blood individual metabolites, and fecal bacteria from Experiments 1 (A) and 2 (B).Each network link has an absolute spearman rank correlation ≥0.40 with FDR-corrected significance <0.05.The light red lines indicate positive correlations, and light blue lines indicate negative correlations.The thickness of the line represents the strength of the correlation, and the larger the correlation coefficient, the thicker the line.Bile acids are shown by pearl blue square, blood individual metabolites are shown by orange hexagon, and bacteria are shown by green circle.The size of a node represents how many connections it has, and the more connections, the larger the node.BHBA, β-hydroxybutyrate; CA, cholic acid; GCA, glycocholic acid; GLCA, glycolithocholic acid; LCA, lithocholic acid; NEFA, nonesterified fatty acids; TCA, taurocholic acid; TCDCA, taurochenodeoxycholic acid; TDCA, taurodeoxycholic acid; TG, total triglycerides; TLCA, taurolithocholic acid; T-MCA, tauro α-muricholic acid.

Table 2 .
Zhang et al.:Altered bile acid… Compositions of bile acid pool (calculated categories) in plasma from transition dairy cows in Experiment 1 (n = 20/time point).

Table 3 .
Zhang et al.:Altered bile acid… Blood metabolites in healthy cows and cows with hyperketonemia from Experiment 2 (n = 14/group)

Table 4 .
Zhang et al.:Altered bile acid… Composition of bile acid pool (calculated categories) in plasma from healthy cows and cows with hyperketonemia in Experiment 2 (n = 14/group)