Peripartum hypophosphatemia is associated with a hindgut-centered microbiota-metabolite-host axis in transition dairy cows

Abstract

The transition period in dairy cows is accompanied by profound shifts in mineral homeostasis and gut microbial ecology. While endocrine regulation of hypocalcemia has been extensively characterized, adaptive responses to hypophosphatemia-and the potential involvement of the gut microbiota-have received far less attention. Twenty-four Holstein dairy cows were randomly assigned to control or low-phosphorus groups. Hypophosphatemia was induced by dietary supplementation with 300 g/d synthetic zeolite from 21 days prepartum to 3 days postpartum. Blood and feces samples were collected at −21, −7, 0, 1, and 3 d relative to calving for longitudinal analysis of physiology, hindgut microbiome and plasma metabolomics to investigate host-microbiome adaptation to peripartum hypophosphatemia in dairy cows. Cows with hypophosphatemia exhibited pronounced compositional remodeling of their hindgut microbiota and extensive, persistent alterations in their plasma metabolome, with glycerophospholipid metabolism being a consistently affected pathway. Integrated correlation and mediation analyses revealed close associations between hindgut microbial variation, host metabolic reprogramming, and circulating phosphorus dynamics. In addition, a plasma feature putatively annotated as α-methyl-m-tyrosine (AMT) was identified as a candidate statistical mediator associated with the observed relationships between Lachnospiraceae_NK3A20_group abundance with systematic phosphorus concentrations. Collectively, these findings indicate that peripartum hypophosphatemia in dairy cows is accompanied by coordinated host metabolic and hindgut microbial remodeling, supporting a hindgut-centered host-metabolite-microbiome framework for understanding phosphorus adaptation during early lactation.

Introduction

The peripartum period represents a critical physiological transition marked by abrupt shifts in mineral demand, energy metabolism, and endocrine regulation. To support fetal skeletal development, calving, and the onset of lactation, dairy cows must rapidly mobilize and redistribute mineral reserves, rendering this window particularly vulnerable to disturbances in mineral homeostasis^1,2. While hypocalcemia during the peripartum period has been extensively studied due to its acute clinical consequences, adaptive responses to hypophosphatemia have received far less attention, despite phosphorus being equally essential for cellular energetics, membrane integrity, and metabolic regulation^3–7.

Calcium and phosphorus homeostasis are tightly interconnected through complex endocrine networks involving parathyroid hormone (PTH), vitamin D metabolites, and fibroblast growth factor 23 (FGF23). During the substantial mineral demands of the peripartum period, systemic regulation appears to prioritize the acute maintenance of calcium homeostasis^8,9. Consequently, hypophosphatemia may arise independently of hypocalcemia and persist beyond the immediate transition period, reflecting a distinct adaptive state rather than a secondary consequence of calcium dysregulation^10,11. However, the physiological implications of this calcium-phosphorus decoupling, particularly with respect to host metabolism and gut microbial ecology, remain largely unexplored.

The gastrointestinal tract plays a central role in mineral sensing and absorption and harbors a complex microbial ecosystem capable of modulating host metabolic responses¹². The hindgut constitutes a substantial microbial niche and is increasingly recognized as an active site for metabolic interactions between transition dairy cows and their microbiome^13,14. While research on mineral-microbiome interactions in ruminants has historically focused on the rumen due to its dominant fermentation capacity, this distal compartment of the gut exhibits distinct relevance to phosphorus (P) adaptation for several reasons. First, the hindgut of cattle and sheep actively absorbs variable quantities of phosphorus, establishing it as a functional site of mineral recovery^15–17. Second, a bidirectional relationship exists between phosphorus availability and hindgut microbial function. Phosphorus limitation has been shown to alter microbial fermentation patterns and suppress fibrolytic taxa in the gastrointestinal tract^18–21. Critically, within the hindgut specifically, lower intestinal P availability directly compromises the fibrolytic activity of the resident microbiome, while specific taxa—such as the genus Desulfovibrio in the caprine cecum and jejunum—exhibit significant correlations with high host phosphorus digestibility^22–24. Some microbial lineages have been reported to exhibit functional flexibility under low-phosphorus conditions, indicating that microbial responses are not uniformly inhibitory but may involve selective enrichment of taxa adapted to nutrient stress^20,25. Whether systemic mineral perturbations during the transition period preferentially influence hindgut microbial structure and function, independent of ruminal regulation, remains unknown.

Multi-omics approaches integrating microbiome profiling with metabolomics provide a powerful framework for dissecting host-microbiome interactions under physiological stress. In the context of phosphorus deficiency, alterations in host lipid metabolism, particularly glycerophospholipid remodeling, may represent a conserved strategy to redistribute phosphorus from structural pools toward essential metabolic functions, a response that has been documented across diverse organisms under phosphorus-limited conditions^20,25. Importantly, microbial remodeling is unlikely to represent a purely unidirectional consequence of host metabolism alone. Rather, shifts in microbial composition and metabolic output may reciprocally influence host mineral dynamics, establishing a bidirectional host-metabolite-microbiome interaction axis.

In the present study, we employed a longitudinal design to investigate how peripartum hypophosphatemia influences systemic mineral status, host metabolic pathways, and gut microbial communities in dairy cows. Longitudinal profiling revealed a more pronounced and sustained response in phosphorus metabolism, prompting a focused investigation of hypophosphatemia as the dominant adaptive signal. By integrating plasma metabolomics, hindgut microbiome analysis, we aimed to elucidate a host-metabolite-microbiome axis underlying peripartum phosphorus adaptation and to redefine the role of the hindgut in mineral homeostasis during this critical physiological peripartum.

Methods

This study was approved by the Committee of Animal Welfare and Animal Experimental Ethical Inspection of China Agricultural University (Protocol No. AW31603202-1-1).

Animal treatments and experimental diets

The present study was conducted from November 6 to December 8, 2023. A total of 24 healthy Holstein dairy cows, which were randomly assigned to either a low-phosphorus group (LPG) or a control group (CON) 21 days prior to the expected calving date, with 12 cows allocated to each group. Cows that calved prematurely (between 28 and 14 days before the expected calving date) or developed diseases during the experiment were excluded from the study. Consequently, each group ultimately contained 9 cows (average parity: CON, 3.33 ± 1.00; LPG, 3.22 ± 0.97).

Hypophosphatemia was induced through dietary supplementation with synthetic zeolite. The zeolite used in this trial was a sodium aluminum silicate with a particle size ranging from 0.4 to 1.0 μm. Based on previous studies, a supplementation rate of 300 g/d was selected to effectively reduce phosphorus availability while minimizing potential adverse effects on dry matter intake (DMI)^26–31. During the prepartum period, cows in the CON received a basal diet (TMR), whereas cows in the LPG received synthetic zeolite mixed into the same basal diet. Following calving, all cows were fed the same diets as during the prepartum period. Detailed daily ration formulations are provided in Supplementary table 1. A schematic overview of the experimental design is provided in Fig. 1.

Fig. 1. Schematic diagram of the experimental design.

Multiparous dairy cows were assigned to either a control diet (CON) or a low-phosphorus diet supplemented with 300 g/day synthetic zeolite (LPG) from 21 d before calving until parturition. After calving, all cows received a common postpartum diet. Plasma and hindgut content samples were collected at 7 days before calving, calving (0 d), and at 1 and 3 d postpartum. Plasma samples were analyzed for biochemical indices and untargeted metabolomics, while hindgut contents were subjected to 16S rRNA gene amplicon sequencing.

Sample collection and measurements

No anesthesia or euthanasia procedures were performed in this study. Rumen fluid, fecal and blood samples were collected by trained personnel using routine animal handling and sampling procedures approved by the institutional animal care and use committee. All procedures were conducted in accordance with relevant animal welfare and ethical guidelines, and all efforts were made to minimize animal stress and discomfort during sample collection. Blood samples were collected via the coccygeal vein into evacuated tubes. Samples were collected at days −21 d, −7 d, 0 d (immediately after calving), 1 d, and 3 d relative to parturition. Lithium heparin tubes were used for plasma separation. The samples were then centrifuged at 3000 × g for 10 min. The plasma fractions were divided into 2.0 mL aliquots and stored frozen at -80°C until analysis. The plasma concentrations ionized calcium (iCa) were determined using the electrode method (RAPIDPoint500, Siemens). The plasma total calcium concentrations (tCa) in plasma were determined using the arsenazo III colorimetric method. The plasma phosphorus concentrations (P) were analyzed using the Ultraviolet (UV) spectrophotometry method. The plasma magnesium concentrations (Mg) were determined using MTB colorimetric method. Plasma parathyroid hormone (PTH), 1,25-dihydroxyvitamin D₃ (1,25(OH)₂D₃), calcitonin (CT), carboxy-terminal cross-linked telopeptide of type Ⅰ collagen (CTX-Ⅰ), and osteocalcin (OC) concentrations were analyzed using commercial ELISA kits using the following methods (Beijing Gersion Biotechnology Co.)^32,33. The plasma glucose concentrations were determined using the glucose oxidase phenol 4-Aminoantipyrine Peroxidase (GOD-PAP) method. The plasma beta-hydroxybutyrate (BHBA) concentrations were measured using BHBCheck Plus blood ketone strips (PortaCheck). The plasma nonesterified fatty acid (NEFA) using a fully automatic biochemical analyzer combined with commercial kits (Beijing Gersion Biotechnology Co.).

Rumen fluid samples (prepartum 7 d, calving, postpartum 1 d and 3 d) were collected prior to morning feeding using a flexible oral stomach tube and a 50 mL syringe. The device underwent thorough cleaning with fresh warm water between sample conditions, and the initial 50 mL of rumen fluid was discarded to reduce saliva contamination. Subsequently, 100 mL of rumen fluid was obtained from each animal, filtered through four layers of sterilized gauze. Fecal samples (prepartum 7 d, calving, postpartum 1 d and 3 d) were collected rectally, then quickly divided into portions using sterile instruments and frozen at −80 °C for 16S rRNA.

16S rRNA gene amplicon sequencing of the fecal microbial community

Genomic DNA was extracted from fecal samples using the FastDNA® Spin Kit for Soil (MP Biomedicals, CA, USA) following the manufacturer’s instructions. DNA quality and concentration were assessed by 1.0% agarose gel electrophoresis and a NanoDrop2000 spectrophotometer (Thermo Scientific, United States), and samples were stored at −80 °C. The V3-V4 hypervariable region of the bacterial 16S rRNA gene was amplified with primer pairs 338 F (5’-ACTCCTACGGGAGGCAGCAG-3’) and 806 R (5’-GGACTACHVGGGTWTCTAAT-3’)³⁴. PCR amplification cycling conditions were as follows: initial denaturation at 95 °C for 3 min, followed by 27 cycles of denaturing at 95 °C for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 45 s, and single extension at 72 °C for 10 min, and end at 4 °C. PCR products were purified from 2% agarose gels using the PCR Clean-Up Kit (YuHua, Shanghai, China) and quantified with Qubit 4.0 (Thermo Fisher Scientific, USA).

Purified amplicons were pooled in equimolar amounts and sequenced on an Illumina NextSeq 2000 platform (2 × 250 bp paired-end mode) (Illumina, San Diego, USA) according to standard protocols by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). Raw sequencing reads were deposited in the NCBI Sequence Read Archive (SRA) database under BioProject accession numbers PRJNA1434142.

For bioinformatics analysis, raw FASTQ files were de-multiplexed using an in-house Perl script based on exact barcode matching, while allowing up to 2 nucleotide mismatches in primer matching. Subsequent quality filtering was performed with fastp (version 0.19.6)³⁵ using the following criteria: reads were truncated at any site where the average quality score over a 10 bp sliding window dropped below 20; additionally, truncated reads shorter than 50 bp and those containing any ambiguous characters (N bases) were discarded. The filtered paired-end reads were then merged using FLASH (version 1.2.7)³⁶, requiring a minimum overlap length of 10 bp. The maximum mismatch ratio allowed in the overlap region was 0.2, and reads that could not be assembled were discarded. Finally, samples were distinguished according to the barcode and primer sequences, and the sequence direction was adjusted. High-quality merged sequences were denoised into amplicon sequence variants (ASVs) at single-nucleotide resolution using the DADA2 plugin³⁷ within the QIIME2 pipeline (version 2020.2)³⁸. Specifically, no length truncation was applied during this step, as strict quality trimming and length filtering had already been performed upstream using fastp. The sequences were first dereplicated to remove exact duplicates. Denoising was then performed with the maximum expected error threshold (maxEE) set to 2. Singletons (ASVs with an abundance of 1) were discarded. Finally, chimeric sequences were identified and removed using the ‘consensus’ method in DADA2. Regarding the filtering of low-abundance taxa, no arbitrary prevalence-based filter was applied. Instead, the exclusion of spurious reads was achieved entirely through DADA2’s rigorous error modeling combined with the aforementioned singleton removal. To ensure comparable sequencing depth for downstream alpha and beta diversity analyses, all samples were rarefied to a uniform depth of 11,408 sequences per sample. This depth was selected based on the minimum number of sequences obtained across all samples, thereby ensuring that all successfully processed samples were retained for subsequent analysis while eliminating potential biases caused by uneven sequencing efforts. Taxonomic assignment of ASVs was performed using the Naive Bayes consensus taxonomy classifier implemented in QIIME2, based on the SILVA 16S rRNA database (v138.2) with a 99% similarity identity. The classification was executed with a confidence threshold of 0.7 to ensure reliable taxonomic identification at various levels. Finally, some samples were exclued due to contamination or unsuccessful DNA extraction at specific sampling points.

Non-targeted plasma metabolomics

Plasma metabolites were extracted from 100 μL samples. Protein precipitation was initiated by adding 400 μL of cold acetonitrile: methanol (1:1, v: v) containing 0.02 mg/mL L-2-chlorophenylalanine (internal standard). Samples were vortexed (30 s), sonicated (30 min, 5 °C, 40 KHz), and incubated at −20°C for 30 min to facilitate protein precipitation. After centrifugation (13,000 × g, 4 °C, 15 min), the supernatant was collected and dried under a nitrogen stream. The dried extract was reconstituted in 100 μL acetonitrile: water (1: 1), followed by 5 min low temperature ultrasonication (5 °C, 40 KHz) and a final centrifugation (13,000 × g, 4 °C, 10 min). The resulting supernatant was transferred to sample vials for LC-MS/MS analysis.

Metabolomic profiling was performed using a Thermo UHPLC-Q Exactive HF-X system coupled with mass spectrometry, equipped with an ACQUITY HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm; Waters, USA) at Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). The injection volume was 3 μL for all LC-MS/MS analyses.

Raw LC-MS/MS data were processed and aligned using Progenesis QI software (Waters Corporation, Milford, USA). Metabolite identification was achieved by matching exact mass, retention time, and fragmentation patterns against major public databases (HMDB: http://www.hmdb.ca/, Metlin: https://metlin.scripps.edu/) and the Majorbio internal database. For metabolites assigned as Level B(ii), identification relied on high-confidence matching of experimental MS/MS spectra to computationally predicted or public reference spectra. Authentic chemical standards were not used for confirmation of metabolite identities in this untargeted metabolomics study. Internal standard peaks and known contaminants (e.g., noise, column bleed) were excluded. The resulting data matrix was further processed on the Majorbio cloud platform (https://cloud.majorbio.com). A metabolite feature was retained if detected in at least 80% of samples within any group. Missing values below the lower limit of quantification were imputed using the minimum observed value for that feature. Data were then normalized by the sum of intensities for each sample to account for variations in sample preparation and instrument stability. Quality control (QC) samples were used to filter out variables with a relative standard deviation (RSD) > 30%. Finally, the processed data were log10 transformed for subsequent statistical analysis.

Statistical analysis

Independent samples t-tests were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) to analyze energy metabolism indicators. To evaluate the overall longitudinal effects on continuous variables (including plasma metabolic indicators) across the entire experimental period, a Linear Mixed-Effects Model (LMM) was employed. In this model, ‘Treatment’, ‘Time’, and the ‘Treatment × Time’ interaction were assigned as fixed effects, whereas individual animal ID was modeled as a random effect to account for repeated sampling. Based on the previously rarefied ASV data (uniform depth of 11,408 sequences per sample), alpha diversity indices (Chao 1 and Shannon) were calculated from ASV information using Mothur (version 1.30.2)³⁹. These two indices were selected to provide a complementary evaluation of the gut microbiota: the Chao1 index estimates total species richness with high sensitivity to rare taxa, whereas the Shannon index accounts for both species richness and community evenness, reflecting overall diversity⁴⁰. Prior to evaluating differences between groups, continuous variables (including plasma metabolic indicators, alpha diversity indices, and taxonomic abundances) were assessed for data normality using the Shapiro-Wilk test and for homogeneity of variance using Levene’s test. For data meeting both assumptions, such as plasma indicators and alpha diversity indices at days 0, 1, and 3, independent samples Student’s t-tests were performed. Conversely, for data exhibiting skewed distributions or unequal variances—specifically, the Chao1 index at day -7 and the taxonomic abundances across the four time points—the non-parametric Mann-Whitney U test was applied. Beta diversity was assessed by Principal Coordinate Analysis (PCoA) based on Bray-Curtis dissimilarity, implemented with the Vegan (version 2.5-3) package in R, with the Adonis method employed to test for differences between groups. Furthermore, to assess the differences between groups at specific individual time points, data were first assessed for normality using the Shapiro-Wilk test and for homogeneity of variance using Levene’s test. To identify significantly abundant bacterial taxa (phylum to genus level) among groups, Linear Discriminant Analysis Effect Size (LEfSe)⁴¹ (http://huttenhower.sph.harvard.edu/LEfSe) was performed, applying thresholds of an LDA score > 3 and P < 0.05. For differential metabolite analysis, multivariate statistical methods, including Principal Component Analysis (PCA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), were applied using the ‘ropls’ R package (version 1.6.2). Model stability was verified through 7-cycle interactive validation. Metabolites were considered significantly different if they exhibited a Variable Importance in Projection (VIP) score > 1 from the OPLS-DA model and a P-value < 0.05 from a two-tailed Student’s t-test. Metabolic pathway enrichment analysis for differential metabolites was conducted using the KEGG database (http://www.genome.jp/kegg/), with statistical enrichment calculated via Python’s ‘scipy.stats’ package. Correlation heatmaps were generated to visualize relationships between microbial relative abundances and phenotypic data (blood mineral and hormone concentrations) or the relative abundance of differential metabolites. Further integrative analyses were performed using the ‘Deep MetOrigin Analysis’ module on the MetOrigin 2.0 web platform (https://metorigin.met-bioinformatics.cn/home/)⁴². For data preprocessing, the missing value filtering method was set to ‘High Frequency’ for both microbial and metabolomic data. Differential taxa and metabolites input for this module were filtered using the Mann-Whitney U test with a P-value cutoff of < 0.05. Subsequently, Origin Analysis, Pathway Analysis, Correlation Analysis, and Mediation Analysis were sequentially conducted. Specifically, Spearman’s rank correlation was employed for Correlation Analysis (significance denoted as *P < 0.05, **P < 0.01), while Mediation Analysis was executed with customized thresholds of an absolute correlation coefficient (|R | ) > 0.5 and P < 0.001.

Results

Hypophosphatemia is decoupled from systemic calcium dysregulation during the peripartum period

During the peripartum period, cows subjected to a hypophosphatemia exhibited a sustained reduction in plasma phosphorus concentrations from late gestation to early lactation, compared with control cows (Fig. 2A). In contrast, plasma magnesium concentrations did not differ significantly between groups, indicating that homeostasis of other electrolytes was not disturbed. Circulating ionized calcium concentrations were consistently higher in the low-phosphorus group, accompanied by elevated levels of 1,25(OH)₂D₃, indicating activation of endocrine mechanisms supporting calcium homeostasis. Total calcium concentrations showed only modest temporal variation between groups, while calcitonin levels remained largely unchanged, suggesting the absence of counter-regulatory suppression of calcium mobilization. Parathyroid hormone concentrations displayed a transient postpartum increase in hypophosphatemia cows, consistent with adaptive endocrine compensation rather than pathological dysregulation.

Fig. 2. Effects of hypophosphatemia on plasma mineral, hormone, bone metabolism, and energy metabolism indicators in transition dairy cows.

A Longitudinal changes in plasma phosphorus, magnesium, total calcium, ionized calcium, 1,25-(OH)₂D₃, parathyroid hormone and calcitonin concentrations from late gestation (-21 d) to early lactation (3 d) in low-phosphorus (LPG) and control cows (CON). B Plasma markers of bone metabolism at calving, including osteocalcin (OC) and C-terminal telopeptide of type I collagen (CTX-I). C Circulating indicators of energy metabolism at calving, including BHBA, glucose and NEFA.

Markers of bone metabolism further supported this adaptive mineral profile (Fig. 2B). Plasma osteocalcin concentrations were higher in cows under hypophosphatemia, whereas CTX-I levels did not differ significantly between groups, indicating that maintenance of circulating calcium was not primarily associated with by excessive bone resorption during the early transition period. Importantly, circulating β-hydroxybutyrate (BHBA), non-esterified fatty acids (NEFA), and glucose concentrations were comparable between groups (Fig. 2C), excluding systemic energy imbalance as a confounding factor.

Together, these results demonstrate that peripartum hypophosphatemia induces a phosphorus-dominant adaptive state characterized by preserved calcium homeostasis and coordinated endocrine responses, rather than generalized mineral or metabolic dysregulation.

The difference among groups was denoted as follows: *: P < 0.05; **: P < 0.01; ***: P < 0.001. Error bars represent SEM (n = 9).

Peripartum hypophosphatemia selectively alters hindgut microbial community structure

In conditions of peripartum low-phosphate status, the structure of the rumen microbial community remained unchanged (Supplementary fig. 1), whereas the hindgut microbial communities exhibited pronounced structural remodeling. Alpha diversity indices, including Chao1 richness and Shannon diversity, showed no significant differences between groups across the transition period (Fig. 3A), indicating that overall microbial richness and evenness were largely preserved.

Fig. 3. Effects of hypophosphatemia on alpha diversity, beta diversity, and taxonomic composition of the hindgut microbiota in transition dairy cows.

A Chao1 richness index and Shannon diversity index of hindgut microbiota at -7 d, 0 d (calving), 1 d, and 3 d relative to parturition in cows under low-phosphorus status (LPG) and control cows (CON). B Principal coordinate analysis (PCoA) based on Bray Curtis dissimilarity showing time-resolved separation of hindgut microbial community structure between LPG and CON cows at -7 d, 0 d (calving), 1 d, and 3 d. Ellipses indicate 95% confidence intervals. C Relative abundances of representative bacterial genera showing significant differences between groups across transition stages. The difference among groups was denoted as follows: *: P < 0.05; **: P < 0.01; ***: P < 0.001. Error bars represent SEM.

In contrast, beta diversity analyses revealed a clear and time-dependent separation of hindgut microbial community structure between low-phosphorus and control cows (Fig. 3B). Principal coordinate analysis demonstrated significant group clustering at -7 d, 0 d (calving), 1 d, and 3 d relative to parturition, with PERMANOVA confirming that hypophosphatemia explained a significant proportion of community variation at each time point. These results indicate that peripartum hypophosphatemia induces compositional reorganization of the hindgut microbiota without broadly disrupting microbial diversity.

Taxonomic profiling further identified specific bacterial taxa consistently enriched under low-phosphorus conditions (Fig. 3C). Members of the family Lachnospiraceae, including Lachnospiraceae_NK3A20_group and Acetitomaculum, as well as Lachnospiraceae_UCG-010, were significantly more abundant in low-phosphorus cows during the transition period. In contrast, taxa such as Phascolarctobacterium showed higher relative abundance in control cows. Several additional genera, including Bacillus, Saccharofermentans, Anaerovorax, and Family_XIII_AD3011_group, exhibited transient or time-specific responses to hypophosphatemia.

Peripartum hypophosphatemia induces sustained host metabolic reprogramming dominated by glycerophospholipid metabolism

Untargeted plasma metabolomics revealed extensive and time-resolved metabolic alterations in dairy cows subjected to hypophosphatemia across the transition period. Comparative analyses identified numerous differentially abundant metabolites (DAMs) between low-phosphorus and control cows at each time point, with both the number and direction of changes varying over time. Specifically, 218, 138, 194, and 115 metabolites were significantly reduced in the low-phosphorus group at -7 d, 0 d (calving), 1 d, and 3 d relative to parturition, respectively, whereas 93, 106, 109, and 144 metabolites were significantly increased at the corresponding stages (Fig. 4A).

Fig. 4. Effects of hypophosphatemia on the plasma metabolomic profile of transition dairy cows.

A Volcano plots showing differential plasma metabolites between cows in low-phosphorus status (LPG) and control cows (CON) at -7 d, 0 d (calving), 1 d, and 3 d relative to parturition. B Pathway enrichment and topology analyses of differential metabolites at each time point. C Upset plot and Venn diagrams illustrating the number and overlap of differential metabolites across transition stages. D Upset plot and Venn diagrams illustrating the number and overlap of differential metabolic pathways across transition stages. E Focused enrichment analysis highlighting glycerophospholipid metabolism across all time points.

Pathway enrichment and topological analyses further revealed that metabolic responses to hypophosphatemia were highly selective rather than globally distributed (Fig. 4B). Among all significantly enriched pathways, glycerophospholipid metabolism emerged as the only pathway consistently altered across all time points (Fig. 4D and E). In contrast, pathways related to central carbon metabolism, amino acid metabolism, and energy metabolism-including the tricarboxylic acid cycle and glycolysis/gluconeogenesis-exhibited transient or stage-dependent alterations.

Overlap analysis demonstrated that only a limited subset of metabolites was consistently altered across multiple stages, with 22 metabolites shared across all four time points (Fig. 4C). Notably, several metabolites associated with glycerophospholipid metabolism, including glycerylphosphorylcholine (GPC) were repeatedly detected across consecutive transition phases, with a continuous decrease in variety (Supplementary fig. 2), whereas metabolites from other pathways showed more stage-specific patterns. Together, these results demonstrate a sustained and structured host metabolic reprogramming under peripartum hypophosphatemia, dominated by glycerophospholipid metabolism rather than diffuse metabolic disruption.

Supplementary fig. 3 illustrates the differentially enriched metabolic pathways at various time points. Consequently, we were able to identify metabolic pathways that were common to two or more time points, as well as sphingolipid metabolic pathways closely associated with glycerophospholipid metabolism. To further visualize pathway-level metabolic alterations, significantly enriched pathways were mapped with their corresponding differential metabolites (Fig. 5). This pathway-based representation revealed coordinated downregulation of multiple intermediates within glycerophospholipid metabolism across transition stages. In contrast, alterations in sphingolipid metabolism, amino acid metabolism, the tricarboxylic acid (TCA) cycle, and glycolysis/gluconeogenesis were more variable and stage dependent. These patterns are consistent with enrichment and overlap analyses, reinforcing glycerophospholipid metabolism as the most persistently affected metabolic pathway under hypophosphatemia.

Fig. 5. Peripartum hypophosphatemia on metabolic pathway alterations in plasma of transition dairy cows.

Note: Different colored blocks in the figure represent distinct metabolic pathways. Metabolites in red font are shared between two adjacent pathways. The heatmaps beside metabolites indicate up-regulation (green) and down-regulation (orange) in low-phosphorus status cows at different time points.

Origin analysis reveals a prominent host-microbiota co-metabolic signature in peripartum hypophosphatemia

To investigate the potential origins of systemic metabolic alterations, differential metabolites were annotated based on predicted source attribution, including host-derived, microbiota-derived, and host-microbiota co-metabolism-associated metabolites. Across all transition stages, a substantial proportion of differential metabolites and enriched pathways were assigned to co-metabolic processes, exceeding those exclusively attributed to host or microbiota origins.

Venn analyses further demonstrated extensive overlap between host- and microbiota-associated metabolites and pathways, with co-metabolism representing the dominant shared component across time points (Fig. 6A-D). In contrast, metabolites uniquely attributed to either host or microbiota constituted a smaller and more variable fraction, indicating that systemic metabolic responses to hypophosphatemia are largely shaped by integrated host-microbial metabolic activity rather than isolated contributions.

Fig. 6. Effects of hypophosphatemia on the origin attribution of differential plasma metabolites and enriched pathways in transition dairy cows.

A Prepartum 7 days, B Calving, C Postpartum 1 day, and D Postpartum 3 days.

At the pathway level, co-metabolism associated pathways consistently accounted for the largest proportion of enriched metabolic functions across transition stages, whereas host- or microbiota-specific pathways displayed greater temporal variability. These results suggest that peripartum hypophosphatemia induces a metabolic environment characterized by coordinated host-microbiota metabolic engagement.

Note: Top-left: Bar plots show the number of metabolites assigned to each category: host-derived (blue), microbiota-derived (green), drug-related (purple), food-related (pink), environmental (brown), and unknown metabolites (gray). Bottom-left: Bar chart presents pathway enrichment results grouped by origin attribution; bar colors denote host-derived (blue), microbiota-derived (green), or co-metabolism-associated (orange) pathways. Statistical significance is expressed as log_0.05 (P-value). Top-right: Venn diagram illustrates the overlap of metabolites attributed to host, microbiota, and co-metabolism at each time point. Bottom-right: Venn diagram illustrates the overlap of metabolic pathways attributed to host, microbiota, and co-metabolism at each time point.

Gut microbiota interplay with host physiology and metabolic profiles

Following the identification of key metabolite origins, we conducted an in-depth investigation into the extensive associations between altered gut microbial communities, host physiological parameters, and differential plasma metabolites across the peripartum period. The correlational landscape exhibited significant temporal shifts (Fig. 7 A-D), underscoring the dynamic host-microbiome interplay during this critical window.

Fig. 7. Peripartum hypophosphatemia on correlations between hindgut microbial taxa and host mineral and hormone indicators in transition dairy cows.

Heatmaps illustrate the Spearman’s rank correlations between the relative abundance of fecal microbial taxa (columns) and plasma concentrations of key host mineral metabolism indicators (rows). A Pre-partum 7 days, B Calving, C Postpartum 1 day, and D Postpartum 3 days. The size of each circle corresponds to the absolute value of the correlation coefficient, while the color indicates the direction of the correlation (purple, positive; green, negative).

A notable trend involved microbial taxa displaying opposing correlations with phosphorus, consistent with their counter-regulatory roles in host mineral homeostasis. For instance, at calving, Lachnospiraceae_NK3A20_group correlated negatively with P (Fig. 7 A-D). Extending these observations to the molecular level, subsequent correlation analyses between differential plasma metabolites and microbial taxa further highlighted Lachnospiraceae_NK3A20_group (along with Acetitomaculum) as consistently exhibiting significant negative correlations with specific metabolites across the peripartum transition. Notably, at prepartum day 7 and calving, these two microbial groups displayed negative correlations with two phenylalanine metabolism-related metabolites (phenylacetic acid and 2-hydroxyphenylacetic acid). Furthermore, during the postpartum 1 d, key glycerophospholipid metabolites (e.g., various phosphatidylcholines (PC) and lysophosphatidylcholines (LPC)) also showed significant negative correlations with Lachnospiraceae_NK3A20_group (Supplementary fig. 4). This pattern suggests a potential interplay where factors such as low phosphorus status might influence the hindgut microbiota, particularly these two genera, by modulating host glycerophospholipid metabolism.

Integrative analyses reveal metabolite-mediated associations between hindgut microbiota and host phenotypes

To explore potential metabolite-mediated linkages between host phosphorus-related phenotypes and hindgut microbiota, metabolites significantly correlated with both host phenotypes and microbial taxa were first identified (|R | ≥ 0.7, P < 0.001). A total of three, one, and thirty metabolites associated with both phenotype and microbiota were detected at 7 days before calving, on the day of calving, and on day 1 postpartum, respectively (Fig. 8A-B).

Fig. 8. Integrative analyses identify metabolite-associated host-hindgut microbiome relationships under peripartum hypophosphatemia.

A Venn diagrams showing the number of differential metabolites significantly associated with hindgut microbial taxa and host phenotypic traits across transition stages. B Bar plots illustrating metabolites associated with hindgut microbiota (yellow) and host phenotypes (green) at −7 d, 0 d (calving), and 1 d relative to parturition. C Sankey diagrams depicting representative associations among hindgut microbial taxa, differential metabolites, and host phenotypic indicators at each time point. D Mediation analyses evaluating the extent to which selected metabolites mediate the associations between representative microbial taxa and host phosphorus-related phenotypes. Values indicate the proportion of the total effect mediated by the metabolite, where |R| between microbes and metabolites ≥ 0.7, P < 0.001. E The relative abundance of the plasma feature putatively annotated as α-methyl-m-tyrosine (AMT) at multiple time points.

At 7 days prepartum, 2,3-dihydroxybenzoic acid and neryl arabinofuranosyl-glucoside were identified as intermediary metabolites linking an unclassified taxon within the Anaerovoracaceae family to plasma phosphorus concentrations. At calving, 1-phenylethanol was associated with Lachnospiraceae_NK4B4_group and plasma ionized calcium levels. On day 1 postpartum, three metabolites were identified that formed association paths between Lachnospiraceae_NK3A20_group and Acetitomaculum with plasma phosphorus concentrations (Fig. 8C).

Mediation analyses were subsequently performed to quantify the extent to which selected metabolites statistically mediated the associations between representative microbial taxa and host phosphorus-related phenotypes. Alpha-methyl-m-tyrosine (AMT) accounted for 83% of the total indirect effect linking Lachnospiraceae_NK3A20_group with plasma phosphorus concentrations (average causal mediation effect (ACME = −0.59, P_medi = 0.02). In addition, a partial mediation trend was observed for Acetitomaculum via alpha-methyl-m-tyrosine (ACME = −0.40, P_medi = 0.06), explaining 52% of the indirect effect (Fig. 8D). These findings collectively provide statistical evidence that specific hindgut microbial taxa are associated with phosphorus-related host phenotypes through intermediate plasma metabolites, while causal directionality and metabolite identity require further validation. Relative abundance of AMT in plasma from LPG cows was significantly elevated from 7 days prenatally to 3 days postpartum (Fig. 8E).

Discussion

The present study demonstrates that peripartum hypophosphatemia in dairy cows constitutes a distinct adaptive state characterized by coordinated host metabolic reprogramming and selective hindgut microbial remodeling, independent of overt hypocalcemia. While systemic mineral regulation during the transition period has traditionally been interpreted through the lens of calcium homeostasis, our findings reveal that phosphorus imbalance can emerge independently of overt hypocalcemia and persist as a primary adaptive signal extending well beyond the immediate peripartum window. Crucially, these systemic alterations were not accompanied by detectable restructuring in ruminal microbiota, underscoring the hindgut as a previously underappreciated site of microbiome-mineral crosstalk during this critical physiological transition^43–45.

Consistent with established physiological models, cows under hypophosphatemia exhibited elevated circulating 1,25-dihydroxyvitamin D₃ and parathyroid hormone concentrations, indicative of active calcium homeostasis maintenance during the peripartum period⁴⁴. Notably, throughout the study period, there were no significant differences in plasma magnesium concentrations or energy metabolism indices among the groups, suggesting that the observed microbial and metabolic remodeling was primarily attributable to the induced hypophosphatemia rather than systemic electrolyte or energy disturbances. A limitation of this study is that plasma potassium concentrations were not measured. However, two studies have found that zeolite does not cause changes in plasma potassium concentrations^46,47. This provides some support for the notion that the effects observed in this study were not driven by potassium ion disturbances. Against this endocrine backdrop, a prominent feature of this adaptive response appeared to be a selective remodeling of the hindgut microbiota. Notably, Lachnospiraceae_NK3A20_group and Acetitomaculum were enriched and displayed consistent negative associations with circulating phosphorus concentrations, positioning the hindgut as a dynamically responsive compartment during systemic mineral perturbation. Members of the Lachnospiraceae family are recognized for their capacity to produce short-chain fatty acids and close interaction with host energy metabolism⁴⁸. While direct involvement in phosphorus metabolism cannot be inferred from the present data, their selective enrichment under hypophosphatemia is consistent with ecological adaptation to the altered physiological and biochemical conditions of the hindgut.

Plasma metabolomic profiling further revealed a sustained remodeling of host metabolism under hypophosphatemia, with glycerophospholipid metabolism emerging as the only pathway consistently altered across all transition stages. Persistent decrease in phosphatidylcholines and lysophosphatidylcholines within the hypophosphatemia cows, which is the principal phosphorus-containing lipid pools, suggests that the host actively reallocates phosphorus from structural lipid reservoirs to sustain essential metabolic functions^49,50. This lipid remodeling likely alters bile acid composition and intestinal biochemical conditions, imposing selective pressures that favor bile acid-tolerant taxa, including members of the Lachnospiraceae^51–57, and thereby providing a coherent metabolic basis for the observed hindgut microbial shifts.

Integrative correlation and mediation analyses identified specific metabolite-mediated statistical linkages between hindgut microbial taxa and host phosphorus-related phenotypes. Among these, α-methyl-m-tyrosine (AMT) was identified as a candidate mediator linking hindgut Lachnospiraceae_NK3A20_group with host phosphorus metabolism. Although a direct mechanistic role of AMT in phosphorus regulation has not been experimentally established, several indirect observations provide a biologically plausible context for this statistical association. Dopamine is an established regulator of renal phosphate handling: activation of dopamine D1-like receptors in the proximal tubule induces acute internalization of the major sodium-phosphate cotransporter NaPi-IIa (SLC34A1), thereby increasing urinary phosphate excretion⁵⁸. Furthermore, catecholamines or related neuroendocrine pathways have been implicated in the regulation of fibroblast growth factor 23 (FGF23), a potent phosphaturic hormone that suppresses renal phosphate reabsorption⁵⁹. This suggests that AMT-associated shifts in catecholamine signaling could potentially influence systemic phosphorus handling through both dopaminergic and FGF23-dependent pathways. In the intestine, dopamine contributes to the control of ion transport, motility, and epithelial barrier function^60,61, whereas phosphate absorption depends largely on NaPi-IIb (SLC34A2), whose activity is tightly regulated by 1,25(OH)₂D₃ and dietary phosphate availability⁶². Furthermore, members of the Lachnospiraceae family are also known to participate in aromatic amino acid metabolism and the generation of phenolic intermediate⁶³.Consistent with this, two additional phenylalanine derivatives, phenylacetic acid and 2-hydroxyphenylacetic acid, were likewise altered and negatively correlated with this taxon, indicating broader perturbation of the aromatic amino acid metabolic axis under low-phosphorus conditions. Taken together, these findings support a multi-layered association among hindgut Lachnospiraceae_NK3A20_group, aromatic amino acid metabolism, and systemic phosphorus dynamics. Notably, we did not measure intestinal phosphate transporter expression, epithelial phosphate uptake, or tissue-level molecular responses in the present study, and thus, no direct AMT-transporter relationship or hindgut-to-host regulatory mechanism can be inferred from these data alone. However, the mediation effect identified here is statistical in nature, and AMT should therefore be regarded as a candidate biomarker-associated intermediary of the low-phosphorus adaptive state rather than a proven causal regulator.

Building on these observations, we propose a working model in which peripartum hypophosphatemia initiates host metabolic reprogramming centered on glycerophospholipid metabolism, remodeling the biochemical landscape of the hindgut and promoting selective microbial enrichment. Metabolite-mediated associations, particularly those involving AMT, may represent functional intermediary links between hindgut microbial variation and systemic phosphorus-related phenotypes. While experimental validation through targeted interventions remains necessary, this model offers a mechanistic framework for understanding coordinated host-metabolite-hindgut microbiome adaptation under phosphorus stress, and positions the hindgut as an active participant in, rather than a passive responder to, mineral homeostasis.

Several limitations of this study also could define the agenda for future research directions. First, it is recommended that future studies include a complete electrolyte panel (Na⁺, K⁺, Cl^-) to fully characterize off-target effects of zeolite supplementation. As an observational multi-omics investigation, the present design is well-suited for generating mechanistic hypotheses and identifying key microbial and metabolic candidates, but cannot establish definitive causality. Interventional studies, including microbial transplantation or targeted metabolite supplementation, are required to formally test the proposed axis. Additionally, while fecal sampling provides a non-invasive and field-applicable proxy for hindgut microbial ecology, it does not fully capture the spatial heterogeneity of the large intestine. Terminal studies incorporating multi-site intestinal sampling would complement our longitudinal data and refine the spatial architecture of the host-microbiome interactions. To note, the biological origin of AMT remains uncertain. Although AMT was identified as a candidate statistical mediator associated with the relationship between Lachnospiraceae_NK3A20_group abundance and plasma phosphorus concentrations.

In summary, this study establishes a hindgut-centered host-metabolite-microbiome framework for peripartum phosphorus adaptation in dairy cows. Persistent glycerophospholipid dysregulation, selective enrichment of Lachnospiraceae taxa, and the predominant mediating role of α-methyl-m-tyrosine collectively position the hindgut as an active participant in mineral homeostasis and reframe peripartum hypophosphatemia as a distinct adaptive state with coordinated metabolic and microbial dimensions. These findings suggest that hindgut-associated microbial and metabolic signatures may serve as candidate indicators of phosphorus adaptation during the periparturient transition. Collectively, our data support the existence of coordinated host–metabolite–microbiome responses under hypophosphatemic conditions. However, the biological identity and functional significance of the AMT-annotated feature remain uncertain and require targeted validation.

Author contributions

J.Y.: Writing—Original Draft, Methodology, Formal analysis, Data Curation, Visualization, Writing—Review and Editing. X.Y.Z.: Data Curation, Visualization, Writing—Review and Editing. S.Y.: Investigation, Methodology, Data Curation, Writing—Review and Editing. C.L.L.: Methodology, Data Curation, Writing—Review and Editing. Z.H.W.: Investigation, Writing—Review and Editing. Q.Q.W.: Investigation, Writing—Review and Editing. Y.Y.H.: Data Curation, Writing—Review and Editing. Y.H.: Investigation, Writing—Review and Editing. S.W.: Visualization, Data Curation, Writing—Review and Editing. F.L.K.: Visualization, Resources, Writing—Review and Editing. M.Z.: Investigation, Writing—Review and Editing. Z.J.C.: Investigation, Writing—Review and Editing. S.L.L.: Investigation, Writing—Review and Editing. W.W.: Conceptualization, Project administration, Supervision, Writing—Original Draft, Writing—Review and Editing.

Data availability

Raw sequencing reads from 16S rRNA gene amplicon sequencing have been submitted to the NCBI Sequence Read Archive (SRA) under BioProject accession numbers PRJNA1434142.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41522-026-01078-5.