Plasma and milk metabolomics profiles in dairy cows with subclinical and clinical ketosis

Ketosis, a commonly observed energy metabolism disorder in dairy cows during the peripartal period, is distinguished by increased concentrations of β-hydroxybutyrate (BHB) in blood. This condition has a negative impact on milk production and quality, causing financial losses. An untargeted metabolomics approach was performed on plasma samples from cows between 5 and 7 DIM diagnosed as controls (CON, BHB <1.2 mM, n = 30), subclinically ketotic (SCK, 1.2 < BHB <3.0 mM, n = 30), or clinically ketotic (CK, BHB >3.0 mM, n = 30). Cows were selected from a commercial farm of 214 Holstein cows (average 305-d yield in the previous lactation of 35.42 ± 7.23 kg/d; parity, 2.41 ± 1.12; body condition score, 3.1 ± 0.45). All plasma and milk samples (n = 90) were subjected to Liquid Chromatography-Mass Spectrometry (LC-MS)- based metabolomic analysis. Statistical analyses was performed using the Graph Pad Prism 8.0, Metabo-Analyst 4.0 and R packages (version 4.1.3). Compared with the CON group, both SCK and CK groups had greater milk fat, freezing point, and fat-to-protein ratio and lower milk protein, lactose, solids-nonfat, and milk density. Within 21 d after calving, compared with CON, the SCK group experienced a reduction of 2.65 kg/d in milk yield, while the CK group experienced a decrease of 7.7 kg/d. Untargeted metabolomics analysis facilitated the annotation of a total of 5,259 and 8,423 metabolites in plasma and milk. Differentially affected metabolites were screened in CON vs. SCK, CON vs. CK, and SCK vs. CK (unpaired t -test, False discovery rate <0.05; and absolute value of log(2)-fold change >1.5). A total of 1,544 and 1,888 differentially affected metabolites were detected in plasma and milk. In plasma, glycerophospholipid metabolism, pyrimidine metabolism, tryptophan metabolism, sphingolipid metabolism, amino sugar and nucleotide sugar metabolism, phenylalanine metabolism, steroid hormone biosynthesis were identified as significant pathways. Weighted gene co-expression network analysis (WGCNA) indicated that tryptophan metabolism is a key pathway associated with the occurrence and development of ke-tosis. Increases in 5-Hydroxytryptophan and decreases in kynurenine and 3-indoleacetic acid in SCK and CK were suggestive of an impact at the gut level. The decrease of most glycerophospholipids indicated that ketosis is associated with disordered lipid metabolism. For milk, pyrimidine metabolism, purine metabolism, pantothenate and CoA biosynthesis, amino sugar and nucleotide sugar metabolism, nicotinate and nicotinamide metabolism, sphingolipid metabolism, fatty acid degradation were identified as significant pathways. The WGCNA indicated that purine and pyrimidine metabolism in plasma was highly correlated with milk yield during the peripartal period. Alterations in purine and pyrimidine metabolism characterized ketosis, with lower levels of these metabolites in both milk and blood underscoring reduced efficiency in nitrogen metabolism. Our results may help to establish a foundation for future research investigating mechanisms responsible for the occurrence and development of ketosis in peripartal cows.


INTRODUCTION
The peripartal period begins approximately 3 weeks before calving and extends up to 3 weeks after calving, during which dairy cows often experience negative energy balance (NEB) (Wathes et al., 2007).To compensate for NEB, the cow mobilizes body fat reserves and initiates a metabolic process called lipolysis, releasing fatty acids into the bloodstream (Wathes et al., Plasma and milk metabolomics profiles in dairy cows with subclinical and clinical ketosis 2007).Circulating fatty acids are oxidized in both the liver and skeletal muscle tissues or are utilized for the production of milk triglycerides in the mammary gland.When the metabolic demands for energy production from fatty acids exceed the liver's ability to oxidize them effectively, an accumulation of TAG occurs (Duffield 2000;LeBlanc 2012).In cows suffering from severe NEB there is a deficiency in the liver's ability to maintain a balance between triacylglycerol (TAG) output in the form of VLDL and TAG production (Van der Kolk et al., 2017).Excess fatty acids are also oxidized to produce carbon dioxide and ketone bodies (acetoacetic acid, BHB, and acetone).Ketosis is often associated with decreased milk production (Hubner et al., 2022), impaired immune function (Horst et al.2021), impaired reproductive performance (Alemu et al., 2023), and an increased risk of other metabolic disorders (Rodriguez et al., 2022).According to some researchers, the postcalving changes in healthy cows to energy metabolism reflect normal processes for maximizing milk synthesis, and therapeutically reducing ketones during subclinical ketosis could do more harm than good (Horst et al., 2021).Thus, elucidating the pathogenesis of ketosis is beneficial for further understanding the role that ketosis plays during the periparturient period.
Metabolomics techniques have proven to be valuable tools in the study of disease in medicine and biology.By profiling and quantifying metabolites, researchers can gain insights into the biochemical pathways and mechanisms underlying diseases (Sun et al., 2016;Ceciliani et al., 2018;Hao et al., 2021).In recent years, the application of metabolomics in the study of ketosis has attracted extensive attention.Sun et al. (2014) utilized H-1 nuclear magnetic resonance-based metabolomics and identified 25 differentially affected metabolites between ketosis and healthy cows, mainly associated with energy, amino acid, lipid, and carbohydrate metabolism.This discovery not only aids in the early diagnosis of ketosis, but also assists veterinarians in developing adequate therapies and aid in better clinical management.Zhang et al. (2013), using gas chromatography/ mass spectrometry, identified 2-piperidinecarboxylic acid and cis-9-hexadecenoic acid closely associated with metabolic perturbations during ketosis.Furthermore, metabolomics of cow plasma (Zhang et al. 2021b;Li et al., 2022), ruminal fluid (Eom et al. 2021a;Eom et al. 2021b), urine (Zhang et al. 2021a;Eom et al., 2022), and milk (Eom et al. 2021a;Eom et al. 2021b;Zhu et al., 2021) indicated that this technique is a powerful tool for studying metabolic changes and metabolite markers in cows.
The utilization of on-farm screening tools enables the immediate diagnosis and treatment of ketosis (Tatone et al., 2016).In comparison to the laboratory spectro-photometric assay, milk point-of-care tests have been developed to provide a faster and more cost-effective diagnosis of ketosis, with the added benefit of being less invasive than blood tests.Milk sampling allows for multiple samples to be collected and the assessment of ketosis at the herd level, providing a more dynamic data of BHB levels that could guide interventions at the herd level to improve overall cow health and productivity.Serrenho et al. (2022) conducted a study and discovered that the detection of BHB in milk allowed for the diagnosis of ketotic cows approximately 2 d later compared with the blood test.The milk test demonstrated a sensitivity of 61% and a specificity of 91% (Serrenho et al., 2022).However, compared with blood BHB, cowside milk tests generally exhibit poor sensitivity toward SCK and are not recommended as tests for herd-based monitoring (Oetzel et al., 2004).In addition to elevated blood ketones, ketotic cows experience various metabolic changes, which likely can be assessed in the milk.In comparison with ketotic cows, healthy cows exhibit elevated levels of glycerophosphocholine (GPC) and high ratios of GPC to phosphocholine (PC) in milk (Klein et al., 2012).Another study reported an increase in maleate, 3-hydroxybutyrate, acetoacetate, galactonate, and 3-hydroxykynurenine in the milk of subclinically ketotic cows (Eom et al. 2021b).Thus, by utilizing metabolomics screening, it is possible to identify biomarkers other than ketones that can accurately identify cows at risk of ketosis in its early stages and even predict the likelihood of its occurrence before current diagnostic tools.
The present study can contribute to a better understanding of the relationship between blood and milk BHB levels and the underlying metabolic changes associated with ketosis.As such, this knowledge can inform dairy management practices, allowing for earlier detection and targeted interventions to prevent and manage ketosis in dairy cows.

Animal handling and management
All procedures, including blood and milk collection, were approved by the Ethics Committee on the Use and Care of Animals at Northwest A&F University (Ethical Approval number: 2021048).All experiments were carried out at the experimental farm of Northwest A&F University (Shaanxi Province, China) in Western China (106°55′57″E, 34°48′41″N) in September -November 2021.The cow population of the farm at the time of the study was 10,175, including 5,151 lactating animals.The average parity of the cows was 2.6, with an average milk production of 11,915 kg (milk fat, 3.82%; milk protein, 3.39%).During the experimental period, a total of 2,398 cows calved, and the incidence rate of postpartum clinical ketosis was 7.63%.The experiment consisted of 214 Holstein cows (average 305-d yield from the previous lactations of 35.42 ± 7.23 kg/d; parity, 2.41 ± 1.12; body condition score, 3.1 ± 0.45) and their corresponding data records (cow identification number, herd code, calving date).These cows were raised in the same environment, for example, they were fed the same diet, had free access to freshwater, and were exposed to the same management.The cows were fed a TMR twice a day during the 60-d dry period (1100 and 1900).Following calving, cows were milked thrice daily (0500, 1300 and 2200 h) and fed a TMR thrice daily (0600, 1400 and 2300 h), primarily consisting of corn silage.The diet ingredients and composition are presented in Table S1.
Blood samples were collected from a tail vessel in tubes containing lithium heparin for isolation of plasma.Plasma tubes were kept below 4°C until plasma was separated by centrifugation at 2,000 × g at 4°C for 10 min.Plasma aliquots were stored at −80°C for quantification of blood analytes.Milk samples were collected at 1300 after plasma BHB was tested.The milkers performed teat and udder cleaning before manual sampling, followed by discarding 3 handfuls of milk to check for abnormalities.Four quarters of milk were collected using a 60mL sample cup and immediately analyzed with the Milko Scan 4000 (Foss, USA) automatic milk analyzer for milk composition (milk fat, protein, lactose, SNF(solid nonfat), density and total solid freezing point).At least 3 technical replicates were run for each sample.Preserve 10 mL of milk in 10 mL centrifuge tube at −80°C for future metabolomics analysis.

Metabolomics by LC-MS/MS
Twenty μL plasma or milk were mixed with 300 μL precooled acetonitrile, vigorously vortexed and placed on ice for 15 min.The solution was then centrifuged at 12,000 rpm and 4°C for 15 min.The obtained supernatant was passed through a membrane with a pore size of 0.22 μm and injected into a liquid chromatography tandem mass spectrometry (LC-MS/MS) instrument.The LC/MS system for metabolomics was composed of a Waters Acquity I-Class PLUS ultra-high performance liquid tandem Waters Xevo G2-XS QTof high resolution mass spectrometer.The column used was a Waters Acquity UPLC HSS T3 column (1.8 μm 2.1*100 mm).Conditions during positive ion mode were: mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 0.1% formic acid acetonitrile.Conditions during negative ion mode were: mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 0.1% formic acid acetonitrile.The auto-sampler temperature was 6°C, and the injection volume was 1 μL.
The Waters Xevo G2-XS QTOF high resolution mass spectrometer can collect primary and secondary mass spectrometry data in MSe mode under the control of the acquisition software (MassLynx V4.2, Waters).In each data acquisition cycle, dual-channel data acquisition can be performed on both low collision energy and high collision energy at the same time.The low collision energy is 2V, the high collision energy range is 10~40V, and the scanning frequency is 0.2 s for a mass spectrum.The parameters of the ESI ion source were as follows: capillary voltage: 2000V (positive ion mode) or −1500V (negative ion mode); cone voltage: 30V; ion source temperature: 150°C; desolvent gas temperature 500°C; backflush gas flow rate: 50L/ h; Desolventizing gas flow rate: 800L/h.

Data preprocessing and annotation
The raw data collected using MassLynx V4.2 was processed with Progenesis QI software for peak extraction, peak alignment and other data processing operations.It was based on the Progenesis QI software online METLIN database and Biomark's self-built library for identification.Theoretical fragment identification and mass deviation were all within 100 ppm.Missing values exist widely in LC-MS-based metabolomics data and affect the normality and variance of the data.At present, the KNN method is considered a useful tool to deal with missing values (Kokla et al., 2019) and was used in the present study for such purpose.

Data analysis
Descriptive statistics of yield traits and of quality traits and coagulation properties of milk samples were analyzed using GraphPad Prism 8.0 (GraphPad Software Inc., USA).The data were analyzed with Tukey's Huang et al.: METABOLOMICS OF KETOSIS multiple comparisons test and results expressed as means ± SD.The data were log2-transformed and Pareto-scaled before analysis.Multivariate analysis was performed using MetaboAnalyst 4.0 (http: / / www .metaboanalyst.ca)comprising the unsupervised principal component analysis (PCA) and the pathway enrichment analyses.Differentially affected metabolites were screened in CON vs. SCK, CON vs. CK, and SCK vs. CK (unpaired t-test, False discovery rate (FDR) < 0.05; and absolute value of log(2)-fold change >1.5).Heatmap visualization was performed using the R package "pheatmap" and the clustering method used "complete" (version 4.1.3,http: / / www .R -project .org).The boxplot was generated via the boxplot function in R and the differences analyzed using R (LSD.test function in the "agricolae" package).Spearman correlations between metabolites in plasma and milk were analyzed in R using "corr.test,"and the FDR was used to adjust P values.Weighted gene co-expression network analysis (WGCNA) was performed and visualized with the "WGCNA" R package (Langfelder and Horvath 2008).

RESULTS
Compared with the CON group, both the SCK and CK groups had greater (P < 0.05) milk fat, freezing point, and fat-to-protein ratio (Table 1).In contrast, milk protein, lactose, solids-nonfat milk, and milk density were lower (P < 0.05) in the SCK and CK groups.Within 21 d after calving, compared with CON, the SCK group experienced a reduction (P = 0.33) of 2.65 kg/d in milk yield, while the CK group experienced a larger decrease (P < 0.001) of 7.7 kg/d (Table 1 and  Figure 1A).Notably, when examining the 305-d milk yield, both the SCK and CK groups had higher milk production compared with the CON group during peak lactation, which occurred around 40 d after calving (Figure 1B).This trend persisted until the end of lactation.Specifically, the 305-d milk yield increased by 4.09 kg/d (P = 0.11) for the SCK group and 2.29 kg/d (P = 0.50) for the CK group.
The application of untargeted metabolomics analysis facilitated the annotation of a total of 5,259 and 8,423 metabolites in plasma and milk, respectively (Figure 2A and B). Figure 2C and D portray the outcomes of principal component analysis (PCA).In plasma, although the 3 groups exhibited some overlap, a discernible pattern of metabolic alterations from CON to SCK, and subsequently to CK, was evident.In regards to milk, the results revealed that metabolomics profiles of both CON and SCK cows were almost fully covered, while both had distinct clusters of metabolomes shared with CK.
Differentially affected metabolites were identified using the unpaired t-test (FDR < 0.05) and the difference multiple (absolute value of log(2)-fold change > 1.5).These results are depicted in the volcano map (supplementary Figure 1 and 2).In plasma, there were 897, 1366 and 222 differentially altered metabolites between CON and SCK, CON and CK, and SCK and CK, respectively (Figure 3A).For milk, there were 400, 1615 and 889 differentially altered metabolites between CON and SCK, CON and CK, and SCK and CK, respectively (Figure 3B).
To explore the functional differences of metabolomics differences between CON, SCK and CK cows, a KEGG enrichment analysis was performed with untargeted metabolomics data.In plasma, glycerophospholipid metabolism, pyrimidine metabolism, tryptophan metabolism, sphingolipid metabolism, amino sugar and nucleotide sugar metabolism, phenylalanine metabolism, steroid hormone biosynthesis were identified as significant pathways (Figure 3C).For milk, pyrimidine metabolism, purine metabolism, pantothenate and CoA biosynthesis, amino sugar and nucleotide sugar metabolism, nicotinate and nicotinamide metabolism, sphingolipid metabolism, fatty acid degradation were identified as significant pathways (Figure 3D).The variation of differentially altered metabolites within these major metabolic pathways are depicted in the form of heat maps (Figure 4 and 5).To further investigate the impact of metabolites on ketosis and milk production, we utilized WGCNA analysis, a powerful tool that can identify metabolite sets with highly synergistic changes (Figure 6).In plasma, the metabolites were clustered into 7 modules and those that did not cluster to any module are colored in gray (Figure 6B).Each co-expression module was associated with BHB and milk production via Pearson correlation coefficient analysis (Figure 6C).
Among these modules, the turquoise module (contained 1260 metabolites) exhibited the highest correlation with BHB, and its distribution of metabolites is depicted in Figure 7A.These results indicated that the alterations in these metabolites were strongly associated with the onset and development of ketosis.We then conducted pathway enrichment analysis for the metabolites within the turquoise module.This analysis revealed the presence of tryptophan metabolism, amino sugar and nucleotide sugar metabolism, and pyrimidine metabolism as significant pathways (Figure 7B).The alterations in plasma levels of metabolites within the tryptophan pathway can be observed in Figure 8.There was an elevation in the concentration of 5-hydroxy-Ltryptophan in the ketotic cows.Conversely, cows with ketosis exhibited significant reductions in the levels of L-tryptophan, kynurenine, cinnavalininate, tryptamine, 3-indoleacetic acid, and 5-hydroxyindoleacid.The WGCNA analysis method was also employed to examine milk samples, but the model and grouping effect yielded unsatisfactory results (see supplementary Figure 3).Consequently, further investigation in this regard was abandoned.
On the other hand, the brown module had the highest correlation with milk yield, and its metabolite distribution is illustrated in Figure 7C.These results indicated that the alterations in these metabolites were strongly associated with milk yield.Similar to the turquoise module, we performed pathway enrichment analysis for the metabolites within the brown module.This analysis identified pyrimidine metabolism, glycerophospholipid metabolism, and purine metabolism as significant pathways (Figure 7D).Following this, our attention was directed toward examining the relationship between the intermediate substances generated during the metabolic processes of purine and pyrimidine in both plasma and milk (Figure 9).Most the metabolites exhibited a positive correlation between the blood and milk.
The screening of biomarkers can contribute to the early diagnosis of disorders and diseases.Metabolites with high gene significance (GS) and module membership (MM) were selected as potential biomarkers of the module.We selected 5 metabolites that displayed the highest GS with BHB or milk yield (Figure 10).These metabolites, which may serve as potential biomarkers for ketosis identification were identified as butylparaben, urolithin a 3-sulfate, Mumefural, ethyl brevifolincarboxylate, and L-histidinol-phosphate in blood.The 5 metabolites associated with milk yield were 1-Octanesulfonic acid, kuwanon E, pratensein 3′ -Glucoside, (Z) -methyl 3-(methylsulfinyl)-1-propenyl disulfide, and tafluprost free acid.

DISCUSSION
The correlation between ketosis and decreased milk production in dairy cows is well established (Overton et al., 2017).Our study revealed that the negative impact of ketosis on milk production occurred primarily during early lactation, with both subclinical and clinical ketosis resulting in higher milk yield during peak lactation compared with the CON group.The greater capacity for milk yield in higher-producing cows necessitates greater levels of energy intake and often exacerbates NEB and facilitates fat depot mobilization, subsequently leading to the development of ketosis and fatty liver (Chapinal et al., 2012).It has been observed that cows with a greater genetic potential for high milk production can experience more pronounced NEB, ultimately resulting in early lactation ketosis (Rutherford et al., 2016).Similar studies have reported that ketosis was associated with higher milk yield (Vanholder et al., 2015;Rathbun et al., 2017;Ruoff et al., 2017).Fatty liver also develops in those cows with a greater potential to produce milk (Gonzalez et al., 2013).Hence, the link between increased demands for milk production and the frequent occurrence of peripartal metabolic diseases is evident.2022) emphasized the importance of considering the timing of ketosis diagnosis when examining its relationship with performance outcomes.Effective management during the dry period and prompt treatment of ketosis during early lactation can decrease incidence and mitigate economic losses (Rathbun et al., 2017).In our experimental farm, ketotic cows received appropriate treatment, potentially contributing to the higher 305-d milk production in the ketosis group.
The relationship between DMI and energy metabolism in postpartal dairy cows is important for maintaining cow health and productivity.Insufficient DMI to meet postpartal energy demands predisposes cows to metabolic disorders, including ketosis and fatty liver, adversely affecting milk yield.However, in present study, because the DMI was calculated by taking the average value for the entire barn rather than measuring it individually for each cow, it was not feasible to perform statistical analysis of DMI for each group.The state of negative energy balance in cows cannot be accurately assessed by the lack of DMI data.Research on how individual cows adjust their metabolism in response to energy deficits and the onset of ketosis is hindered.Nonetheless, alternative indicators, including blood metabolite concentrations, body condition scores, and milk output, offer valuable insights into individual energy balance and metabolic health (Xu et al., 2018 and2020b).Meanwhile, after calving, reduced DMI is a widely acknowledged sign of clinical ketosis in dairy cows (Duffield 2000).
Assessing the change of a metabolite or certain types of metabolites are important means to understand the occurrence and development of diseases.In the present study, alterations in the plasma and milk metabolic profiles of cows affected by subclinical ketosis and clinical ketosis are discussed primarily around the correlation among the various vital metabolic pathways and the degree of ketosis.
Tryptophan is an essential amino acid and tryptophan metabolism plays a crucial role in the regulation of immune function, inflammation, and calcium metabolism (Le Floc'h et al., 2011;Laporta et al., 2015;Agus et al., 2018).There are 3 main routes of tryptophan catabolism, i.e., (i) indole/AhR (aryl hydrocarbon receptor), (ii) kynurenine (Kyn), and (iii) serotonin (5-hydroxytryptamine [5-HT]) pathways (Agus et al., 2018).In present study, the fact that ketosis promoted tryptophan metabolism to 5-Hydroxytryptophan (5-HTP) while inhibiting Kyn and AhR metabolism was suggestive of an effect of this disease at the gut level.In nonruminants, AhR is considered a key component of the immune response at barrier sites, thus, it is crucial for intestinal homeostasis by acting on epithelial renewal (Lamas et al., 2018).
Indole, indole acetic acid, and tryptamine have been established as AhR ligands with various degrees of affinities (Cheng et al., 2015).In mice, dysregulation of the gut flora inhibits tryptophan metabolism of AhR, thereby promoting progression of alcoholic liver disease (Hendrikx et al., 2019).Indole acetic acid increased hepatic PPAR-gamma co-activator-1 α (PGC1a) expression, contributing to mitochondrial respiration improvement in non-alcoholic fatty liver disease (Zhang et al., 2022).In the non-ruminant liver, indole-3-acetate attenuated inflammatory responses caused by lipid loading and downregulated the expression of fatty acid synthase and sterol regulatory element-binding protein-1c (Krishnan et al., 2018).Thus, AHR and its metabolites can have a positive effect on the regulation of lipid metabolism in bovine liver.As such, targeting AHR could help relieve the pressure on lipid metabolism in peripartal cows.While the investigation of changes in AHR and its metabolites in dairy cows is limited, targeting this pathway may hold therapeutic potential for ketosis.
In nonruminants, the Kynurenine pathway is the main pathway of tryptophan metabolism, plays an important role in the regulation of lipolysis, inflam-mation and gluconeogenesis (Cervenka et al., 2017;Goodarzi et al., 2021).The decrease in kynurenine levels with ketosis could have been linked with tryptophan metabolism.The kynurenine to tryptophan ratio is commonly employed as an indicator of indoleamine 2,3-dioxygenase (IDO1) activity and the overall function of the kynurenine metabolic pathway (Arnone et al., 2018).In the present study, there was no significant difference in this ratio between the ketosis and control groups.Thus, the observed decline in kynurenine levels might be attributed to a concurrent decrease in the levels of its precursor substance, tryptophan.During the peripartal period, serotonin fulfills diverse functions, i.e., bone metabolism (Ducy and Karsenty 2010), energy balance regulation (Tecott 2007), mammary gland physiology regulation (Connelly et al., 2021), and immunomodulation (Herr et al., 2017).Multiple studies have demonstrated that the administration of 5-HTP leads to an elevation in serotonin levels and a temporary reduction in blood calcium levels (Laporta et al., 2015;et al., 2021).Others have observed that cows infused with 5-HTP daily starting at either 21 or 10 d before calving exhibited higher blood calcium levels after calving compared with the control group (Hernández-Castellano et al., 2017;Slater et al., 2018).These differences across studies are likely driven by the homeostatic control of calcium, which is regulated by a negative feedback mechanism.A transient hypocalcemia promotes the adaptation and regulatory capacity of calcium, while cows infused with 5-HTP experience a reduction in milk production, particularly in the production of colostrum (Weaver et al., 2016;Hernández-Castellano et al., 2017).Thus, a decrease in milk output may be a contributing factor to the observed increase in blood calcium.Serotonin has the ability to stimulate lipolysis and elevate the plasma fatty acids levels, although it does not exacerbate NEB (Laporta et al., 2015).Although serotonin was inversely associated with the severity of ketosis (Laporta et al., 2013), other studies have failed to establish a connection between serotonin levels and BHB and NEFA levels (Horst et al., 2019).Consequently, the decrease in milk production observed in ketotic cows during early lactation may be linked to the regulation of the serotonin pathway.
The impact of energy balance on estrus behavior of postpartum cows is noteworthy (Rutherford et al., 2016).Ketosis not only diminishes pregnancy success during the first `artificial insemination in cows (Walsh et al., 2007), but it also prolongs the interval from calving to first observed oestrus (Rutherford et al., 2016).The disruption in steroid hormone biosynthesis path-way in ketotic cows in the present study may be one of the causes.Support for this idea arises from data indicating that the levels of 17-hydroxyprogesterone and testosterone, which are metabolites of progesterone, were decreased in ketotic cows (Pishvaei et al., 2021).Together, the low LH pulse frequency and low levels of blood glucose, insulin and IGF-I in ketotic cows limit the production of estrogen by dominant follicles (Butler et al., 2003).Another study reported that lower circulating estradiol levels in lame cows affected the behavioral expression of estrus (Morris et al., 2011).Additionally, Dobson et al. (2007) demonstrated that insufficient progesterone priming resulted in reduced responsiveness of the hypothalamus to estradiol.Regardless of the pathological condition, for cows to establish pregnancy a state of positive energy balance Phosphatidylcholine (PC) is a glycerophospholipid that serves as a key component of the very-low-density lipoprotein (VLDL) monolayer in the liver, and play a crucial role in lipid absorption and transport, cell signaling, and lipoprotein synthesis (DeLong et al., 1999;Huang et al. 2023a;Humer et al., 2016;McFadden et al., 2020).Despite its low concentration in bovine plasma, VLDL is a primary source of fatty acids for peripheral tissues, particularly during early lactation (Bobe et al., 2004).Consequently, reduced levels of choline or PC hinder the efficient synthesis of VLDL, resulting in decreased TAG output and the accumulation of fat in the liver (Piepenbrink and Overton 2003).
Choline deficiency in vitro decreased the levels of PC, and increased the fusion rate of LD in primary hepatocytes from dairy cows leading to lipid accumulation (Lu et al., 2023).We observed significant fluctuations in plasma PC levels during the peripartal period in dairy goats (Huang et al. 2023b).Compared with clinically healthy cows, severe hepatic lipidosis is associated with lower levels of serum PC containing short-chain fatty acid components (Imhasly et al., 2014).The reduction of PC and lypc in SCK and CK cows is highly suggestive that postpartal PC supply and/or synthesis were insufficient, thus, increasing the risk of TAG accumulation in the liver.Thus, although 2 recent meta-analyses did not reveal benefits of feeding rumen-protected choline (RPC) during the periparturient period in terms of liver TAG (Arshad et al., 2020;Humer et al., 2019), the present data suggested that feeding RPC may reduce the risk of fatty liver in cows demonstrating signs of subclinical ketosis.Higher genetic merit for milk production is associated with elevated levels of milk BHB, underscoring the vulnerability of these cows to developing clinical ketosis (Koeck et al., 2014).The present study highlighted that metabolite composition of milk is altered by subclinical and clinical ketosis, providing several potential molecules that could serve as early detection markers of disease.
Relative to nonruminant species, efficiency of dietary/microbial nitrogen use for milk protein synthesis in dairy cows is much lower, leading to numerous experiments over the past several decades aimed at developing approaches to alleviate this problem (Tamminga 1992;Firkins 1996;Kohn et al., 2005).Nucleic acid metabolites, specifically purine and pyrimidine, serve as non-protein nitrogen sources that can potentially reflect metabolic capacity of ruminal microorganisms, e.g., uric acid and allantoin (Schager et al., 2003).Besides these compounds, there is limited information on the role of other nitrogen-containing compounds such as purine and pyrimidine metabolites on the nutritional physiology of ruminants.Our research indicated that alterations in purine and pyrimidine metabolism characterize ketosis, with lower levels of these metabolites in both milk and blood underscoring reduced efficiency in nitrogen metabolism.
The lower levels of lactose and protein in milk of ketotic cows were not unexpected based on published data (Yang et al., 2019).Previous studies have revealed an increase in allantoin levels in milk as protein supply increased, with milk production also exerting an influence (Gonda and Lindberg 1997).Schager et al. (2003) also observed a similar trend, noting that allantoin output increased as the percentage of concentrate in the diet increased.It is likely that the main source of allantoin is the blood, rather than the catabolism of uric acid in mammary tissue (Schager et al., 2003).The positive correlation between the excretion of allantoin in milk and the postruminal flow of microbial nitrogen in dairy cattle agrees with data from Timmermans et al. (2000).Additionally, Rocchetti et al. (2022) demonstrated that pyrimidine intermediates and degradation products in milk could serve as potential markers in corn-based diets.
It has been hypothesized that the increase in milk production after calving is accompanied by the removal of apoptotic epithelial cells from the mammary gland (Miller et al., 2006;Monks et al., 2008).Studies have shown that the apoptotic index in the mammary gland is 4 times higher in early lactation compared with late lactation (Capuco et al., 2001).The presence of new cells in the mammary gland necessitates extensive DNA and RNA synthesis, particularly in cows experiencing more severe NEB (Huberman and Riggs 1968;Xu et al. 2020a) due to the high cellular energy requirements for cell proliferation.In the present study, the low concentrations of purine and pyrimidine metabolic intermediates in the milk suggested cows with ketosis may experience a disruption in purine and pyrimidine metabolism caused by energy deficiency.This can occur due to reduced feed intake, altered ruminal fermentation, and increased breakdown of body tissues.Together, these events not only impact mammary cell renewal, but also influence the availability of precursors for milk component synthesis, subsequently reducing milk synthesis and milk quality.
A negative correlation between intermediates involved in nucleic acid synthesis and NEB also suggested a potential link with the severity of disease (Gonda and Lindberg 1997).The evaluation of our findings regarding content of purine and pyrimidine metabolic intermediates in dairy cow milk is challenging due to the limited number of similar studies available in the literature.

CONCLUSIONS
The application of untargeted metabolomics has facilitated the detection of alterations in the metabolic profiles of both plasma and milk in dairy cows afflicted with subclinical and clinical ketosis.Tryptophan metabolism is a key metabolic pathway associated with the occurrence and development of ketosis.Purine and pyrimidine metabolism was highly correlated with milk yield during the peripartal period.Our results may help to establish a foundation for future research investigating the occurrence and development mechanism of ketosis in peripartal cows.By focusing on these aspects in future studies, we can expand our knowledge of ketosis in cows and contribute to the development of more effective preventive and management approaches.

Figure 1 .
Figure 1.Daily milk yield of dairy cows during the peripartal period (A) and whole lactation (B).The average milk production at 21 d postpartum is depicted by a dotted line.CON, control; SCK, subclinical ketosis; CK, clinical ketosis.Error bars indicate standard errors.

Figure 2 .
Figure 2. Metabolite composition and principal component analysis (PCA) in dairy cow plasma (A and C) and milk (B and D).CON, control; SCK, subclinical ketosis; CK, clinical ketosis.

Figure 3 .
Figure 3. Analysis of differentially affected metabolites in dairy cow plasma (A and C) and milk (B and D).A and B, Venn diagram of differentially affected metabolites from each experimental group.C and D, Metabolic pathway analysis using MetaboAnalyst 4.0 (http: / / www .metaboanalyst.ca).The circles represent the different metabolic pathways.The darker circles indicate significant changes for specific metabolites in the corresponding pathway whereas the size of the circle corresponds to the pathway impact score.CON, control; SCK, subclinical ketosis; CK, clinical ketosis.
Figure 7. Co-expression network of metabolites in plasma constructed via weighted gene co-expression network analysis.(A and C) A scatter plot of Gene Significance (GS) vs. Module Membership (MM).Gene significance (absolute values) represents associations of individual genes with β-hydroxybutyrate (BHB) and milk yield.Module membership represents the correlation between each module and the gene expression profile.(B and D) Metabolic pathway enrichment analysis of metabolites in modules significantly related to BHB and milk yield.Circles represent the different metabolic pathways.The darker circles indicate significant changes for specific metabolites in the corresponding pathway, whereas the size of the circle corresponds to the pathway impact score.
Figure 8.A schematic diagram depicting the tryptophan metabolic pathway using the differentially affected metabolites.Mean values with different letters (a-c) highlight statistically significant differences (LSD, P < 0.05).CON, control; SCK, subclinical ketosis; CK, clinical ketosis.

Figure 9 .
Figure 9. Heat map correlation of differentially affected metabolites associated with purine and pyrimidine metabolic pathways in plasma and milk.Red: positive correlation, blue: negative correlation.The p-value and R were determined according to Spearman's rank correlation test.

Table 1 .
Huang et al.: METABOLOMICS OF KETOSIS Descriptive statistics of yield traits and of quality traits and coagulation properties of milk samples 1 1Data are expressed as mean ± SD (n = 30/group).Mean values with different letters (a-c) in rows show statistically significant differences (P < 0.05).CON, control; SCK, subclinical ketosis; CK, clinical ketosis.