ABSTRACT
Large offspring syndrome (LOS) in cattle is a phenomenon characterized by fetal overgrowth that is largely associated with calves derived from in vitro produced embryos. Here it was tested whether occurrence of LOS as well as two other adverse pregnancy outcomes, stillbirth and abortion, could be predicted using changes in the maternal blood transcriptome early in gestation. Blood samples were collected from Holstein and crossbred heifers at 40–50 days of gestation following embryo transfer of in vitro produced embryos. A total of 203 samples were selected for RNA‐sequencing and used to compare recipients of embryos resulting in live‐born calves of normal birth weight (control; n = 96) with recipients of embryos resulting in LOS calves (n = 15), stillbirths (n = 43), and abortion before 210 days of gestation (n = 49). Large offspring syndrome was defined based on birth weights greater than 2 standard deviations from the mean. The number of differentially‐expressed genes (p < 0.01) was 32, 176, and 100 for comparisons of control versus LOS, stillbirth and abortion, respectively. Biomarkers were identified in a randomly selected, balanced training dataset using Random Forest after filtering for differential expression. Markers were then tested in the held‐out dataset through 6 machine learning models. The best biomarkers consisted of 26, 24, and 28 genes for LOS, stillbirth and abortion, respectively. The accuracy of prediction ranged from 0.91 to 0.965 for LOS, 0.73 to 0.98 for stillbirth, and 0.62 to 0.96 for abortion. Despite the relatively small sample size, the model demonstrated high sensitivity for LOS classification, supporting that the maternal blood transcriptome in early pregnancy reflects molecular alterations associated with fetal overgrowth that can be predictive of the condition.
Keywords: abortion, gestation, large offspring syndrome, machine learning, pregnancy, stillbirth, transcriptome
Blood samples from pregnant heifers carrying in vitro‐produced embryos were analyzed for RNA‐sequencing to test whether maternal blood transcriptome could be predictive of large offspring syndrome, stillbirth, and abortion compared with controls. Differentially‐expressed genes were identified in each group, and there were only a few genes that overlapped among groups. Biomarker genes predicted outcomes with accuracies ranging from 0.62 to 0.98. These findings support that maternal blood transcriptome reflects molecular alterations associated with adverse pregnancy outcomes.
1. Introduction
The environment in which the preimplantation embryo undergoes development can lead to physiological changes that ultimately modify the phenotype of the resultant offspring [1, 2]. In cattle, production of embryos through procedures involving in vitro maturation, fertilization, and culture to the blastocyst stage results in an increased incidence in the birth of large calves (fetal overgrowth) that may also exhibit other developmental abnormalities including organ alterations (i.e., larger hearts, head malformations, abdominal wall defects, enlarged tongue) and congenital malformations such as cerebellar hypoplasia and abnormal limb development [3, 4, 5, 6, 7, 8]. This collection of phenotypes is termed abnormal or large offspring syndrome (LOS). Although LOS‐like phenotypes can rarely occur in pregnancies from natural conceptions [8] or artificial insemination [4, 8, 9], the incidence is higher for calves derived from embryo production in vitro, with reported incidence varying depending on factors such as culture conditions and the definition of LOS [5, 7, 8, 10, 11].
Large Offspring Syndrome in cattle has similarities to Beckwith‐Wiedemann Syndrome (BWS) in humans [12]. Shared features include macrosomia, enlarged tongue, abdominal wall defects, and ear malformation. Beckwith‐Wiedemann Syndrome has been associated with alterations in the expression of several imprinted genes controlling growth, specifically in regions of chromosome 11 [13]. Children conceived with the use of assisted reproductive technologies have a higher risk of BWS [13]. Similarly, in cattle, epigenetic mechanisms associated with LOS include dysregulation of imprinted genes [12, 14] and altered transcript levels of developmentally important genes such as IGF1R [15].
Although severe cases of LOS in cattle are characterized by the aforementioned malformations, many cases may present only with increased birth weight, while affected calves remain otherwise phenotypically normal. Thus, major consequences of LOS are higher risks of dystocia, cesarean sections, stillbirth, and perinatal mortality [8, 9, 16]. Early identification of pregnant females destined to give birth to a calf with LOS would enable producers to mitigate these adverse consequences. To date, no validated diagnostic tools exist for LOS in cattle [8], emphasizing the need to identify molecular markers capable of predicting the syndrome.
One potential approach for identifying biomarkers of LOS is to assess the transcriptome of circulating maternal blood cells. Pregnancy is associated with changes in peripheral immune function in several species including cattle [17, 18, 19, 20]. Temporal expression patterns of gene clusters in the transcriptome of maternal blood cells from Days 21 to 105 of gestation have been observed in cattle [21], suggesting that the maternal blood gene expression changes in parallel with fetal development throughout gestation. Indeed, fetal size and weight influenced the blood molecular profile in recipients as early as 42 days of gestation [22]. Furthermore, RNA‐seq of blood cells was used to identify alterations in the transcriptome of pregnant cows associated with LOS [7] and 17–26 biomarker genes expressed in blood cells at Day 65 of gestation were predictive of calf birth weight [21].
Here it was hypothesized that molecular alterations in maternal blood early in pregnancy caused directly or indirectly by the fetus or placenta can be used to diagnose future adverse pregnancy outcomes, such as LOS, stillbirth, and abortion. The goal was to identify changes in the transcriptome of maternal blood cells between Days 40 and 50 of gestation that were associated with the development of LOS and other abnormalities in pregnancies after transfer of an in vitro produced embryo and to identify biomarkers predicted of such adverse gestational outcomes.
2. Methods
2.1. Recipients and Blood Sample Collection
The study was performed on a private farm located in Lubbock, Texas from February to April of 2024. A total of 1051 blood samples were collected by caudal venipuncture from pregnant Holstein and crossbred (Holstein × Jersey) heifers between Days 40 and 50 of gestation. Heifers were recipients of a single, vitrified Holstein embryo produced commercially by Genus PLC using conventional or X‐sorted semen. The total number of sires used to produce embryos for the study was 46 sires, of which the majority (1016 embryos produced from 36 sires) were used with X‐sorted semen. Only expanded blastocysts of Grade 1 quality [23] were transferred. Each embryo was thawed, placed in a new straw, and transferred to a recipient at Day 7 after observed estrus.
Blood samples were directly collected into Tempus Blood RNA tubes (Applied Biosystems, CA, USA) which contain a stabilizing reagent. Immediately after each blood collection, blood tubes were manually gently inverted end‐over‐end for 10 s to ensure that the stabilizing reagent made uniform contact with the sample. Samples were placed on a cooler with ice packs until being placed in a refrigerator at 4°C. Samples were maintained at 4°C for a maximum of 48 h and shipped overnight with ice packs to the University of Florida. Upon arrival, blood samples were stored at −80°C until RNA extraction was performed.
2.2. Diagnosis of Pregnancy and Abortion and Pregnant Recipient Management
Pregnancy diagnosis was performed by a trained technician using transrectal ultrasound three times at approximately (1) 40–50 days of gestation (i.e., before blood sampling), (2) 70–80 days of gestation, and (3) 210 days of gestation. A pregnant recipient was considered to experience abortion if diagnosed non‐pregnant at a subsequent pregnancy check. The period of abortion was thus estimated as occurring between 40–50 and 70–80 days of gestation or between 70–80 days and 210 days of gestation.
2.3. Collection of Calf Data
Recipients confirmed pregnant at approximately 210 days of gestation were transported to South Dakota for calving. Calving dates ranged from September to December of 2024. Data were recorded for 695 calves while information for the other 260 pregnancies was missing either because of late abortions, sale of recipients, or other unexpected events. The following information was recorded: (1) date of induction of parturition when applicable, (2) days between induction of parturition and birth of the calf, (3) gestation length, (4) calf viability [1 = born alive, 2 = born dead (i.e., stillbirth), 3 = born alive but died within 48 h], (5) calf sex (male or female), and (6) calf birth weight. Induction of parturition was performed following standard operating procedures established by the farm.
2.4. Selection of Samples for RNA Sequencing
Once calving data were obtained, heifers were divided into 4 predefined groups based on pregnancy outcomes: (1) Controls—recipients with a calf having a birth weight within 0.2 standard deviations (SD) from the mean for that breed group (Holstein or Holstein × Jersey) or recipients with a calf having a birth weight between 1 and 2 SD from the mean for that breed group, (2) LOS—recipients with a calf whose body weight was greater than 2 SD above the mean for that breed group, (3) stillbirth—recipients whose calf was born dead, and (4) abortion—recipients that experienced pregnancy loss before 210 days of gestation.
Blood samples for RNA sequencing were selected from among the total number of samples for which data on the calves were available. Several criteria were used to minimize variation. Because X‐sorted semen was used during in vitro fertilization, approximately 90% of the calves were expected to be female. Thus, only blood from recipients carrying female calves were selected to minimize potential confounding effects of sex and avoid unbalanced distribution of samples between sexes. In addition, only heifers collected on the same days of gestation as for the LOS group (40–49 days for Holsteins and 43–46 days for crossbreds) were included for other groups. The difference between the expected day of birth (based on a gestation length of 280 days) and actual birth was calculated. The range of differences was from 2 days before expected calving and 31 days before expected calving. Those heifers that calved at a time representing the lowest quantile (more than 12 days before expected calving date) were excluded (n = 45 from a total of 563 calves). Birth weight and standard deviation calculations were applied after the exclusion. Only female calves were included for control, LOS, and stillbirth groups. Sex of fetus was unknown for the abortion group.
Of the remaining heifers, all (if the number of samples was < 40) or a randomly selected cohort of 40 samples in each group for each breed were selected for RNA sequencing. Since the control group had two birth weight ranges, a cohort of up to 40 samples for each range was randomly selected. The distribution of samples per group and the total number of samples used for RNA sequencing was 203 (Table 1).
TABLE 1.
Number of heifer recipients and samples used for RNA sequencing based on calf group.
| Group | Recipients | Blood samples for RNA sequencing | ||||
|---|---|---|---|---|---|---|
| Holsteins | Crossbreds | Total | Holsteins | Crossbreds | Total | |
| Controls a | 125 | 17 | 142 | 80 | 16 | 96 |
| LOS b | 12 | 4 | 16 | 11 | 4 | 15 |
| Stillbirth | 45 | 3 | 48 | 40 | 3 | 43 |
| Abortions | 73 | 9 | 82 | 40 | 9 | 49 |
Birth weights between 39.8 and 42.4 kg for Holstein (39.9 kg ≤ birth weight ≤ 42.3 kg) and 38.0 and 40.4 kg for crossbreds (38.1 kg ≤ birth weight ≤ 40.3 kg) or birth weights between 47.2 and 53.5 kg for Holstein (47.3 kg ≤ birth weight ≤ 53.4 kg) and 44.6 and 49.8 kg for crossbreds (44.7 kg ≤ BW ≤ 49.7 kg).
Birth weights greater than 53.4 kg for Holstein and 49.8 kg for crossbreds.
2.5. RNA Extraction
Whole blood RNA was extracted using Tempus Spin RNA Isolation Kit (Applied Biosystems, CA, USA) as described earlier [24]. Briefly, Tempus tube blood lysates were thawed at room temperature, diluted with phosphate‐buffered saline in the kit and centrifuged (3000 × g, 30 min, 4°C). The supernatant was discarded, and the RNA‐containing pellet was resuspended in 400 μL of RNA Purification Resuspension Solution. Samples were then transferred to RNA purification filters and subjected to sequential centrifugation and washing steps using washing solutions provided in the kit. The RNA purification filters were then dried by centrifugation and RNA was eluted with 100 μL of Nucleic Acid Purification Elution Solution using two sequential elution steps. Approximately 90 μL of purified RNA was collected, pipetted into an RNase‐free tube, aliquoted into a PCR plate and stored at −80°C. Extracted RNA samples were then transported on dry ice to the University of Florida Interdisciplinary Center for Biotechnology Research Gene Expression Core.
2.6. Library Preparation and Sequencing
Earlier [25], it was found that globin mRNA in bovine blood cells was lowly abundant based on RNA sequencing data. We found similar results in a preliminary test and concluded that globin mRNA removal was not required prior to library preparation. Accordingly, RNASeq libraries were constructed at the University of Florida Interdisciplinary Center for Biotechnology Research Gene Expression Core using a standard protocol (https://biotech.ufl.edu/gene‐expression‐genotyping/, RRID:SCR_019145). Sequencing was performed at the Interdisciplinary Center of Biotechnology Research NextGen Sequencing Core (https://biotech.ufl.edu/next‐gen‐dna/, RRID:SCR_019152). First, 500 ng of total RNA was used for mRNA isolation using the NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs, Ipswich, MA, USA, catalog #E7490). Then, RNA library construction was performed with the NEBNext Ultra II Directional RNA Library Prep Kit (New England Biolabs, catalog #E7760) with 2/3 of recommended poly A enriched RNA and 1/4 recommended reaction volume using SPT Labtech mosquito LV liquid handling instrument (Tecan, Männedorf, Switzerland). Briefly, RNA was fragmented in NEBNext First Strand Synthesis Buffer by heating at 94°C for the recommended time. This step was followed by first strand cDNA synthesis using reverse transcriptase and oligo dT primers. Synthesis of double stranded cDNA was performed using the second strand master mix provided in the kit, followed by end‐repair and adaptor ligation. Finally, the library was enriched by 8 cycles of amplification, and purified by Agencourt AMPure beads (Beckman Coulter, Brea, CA, USA, catalog # A63881). Each of the 203 individual libraries was pooled using equimolar concentrations and sequenced by Illumina NovaSeqx 25B run with 2X150 cycles in a total of 3 lanes (Illumina Inc., San Diego, CA, USA).
Prior to RNA sequencing analysis, RNA integrity numbers were assessed and ranged from 7.9 to 9.9. One sample from the LOS group was excluded due to insufficient RNA yield resulting on a total of 203 samples. The average number of mRNA sequencing reads was ~52 million and was largely consistent across samples. During transcriptome analysis, reads were mapped to the reference genome, and de novo assembly was not performed. Therefore, the identification of potential novel genes or isoforms was not explored on this study.
2.7. Processing of RNA Sequencing Data
The genome of Bos taurus ( Bos taurus .ARS‐UCD2.0.dna.toplevel.fa) was used as the reference for RNA‐seq read alignment. The reference genome was indexed using STAR (v2.7.11b) prior to alignment. Raw sequencing reads were first assessed for quality using FastQC (v0.12.1), and adapter sequences as well as low‐quality bases were removed using fastp (v0.23.4). Processed reads were aligned to the Bos taurus reference genome using STAR (v2.7.11b) in splice‐aware mode, guided by the corresponding GTF annotation file. Only uniquely mapped reads were retained for downstream analysis. Gene‐level expression quantification was performed using featureCounts from the Subread package, with the parameters “‐T 20 ‐p ‐B ‐C ‐t exon,” generating the count matrix.
2.8. Data Analysis
All the analyses of the transcriptomic data were performed with core functions and functions from different packages for the R software (v4.5.0).
2.9. Differential Gene Expression Analysis
Differential gene expression analysis was performed in R using the DESeq2 package [26]. Raw gene counts were normalized to account for differences in library size, and differential expression was assessed using a generalized linear model based on a negative binomial distribution. Statistical significance was evaluated with the Wald test, and p‐values were adjusted for multiple testing using the Benjamini‐Hochberg method. Genes were considered differentially expressed after pairwise comparisons between groups when the p‐value was < 0.01. Genes with a false‐discovery rate (FDR) < 0.05 were also identified. Entrez IDs corresponding to DEG were then used for functional enrichment analysis with ShinyGO v0.741. Overrepresented KEGG pathways were considered significant at FDR < 0.05.
2.10. Strategy for Biomarker Discovery
Before analysis, variance in the transcriptome associated with each of the following factors was determined using the variancePartition package: gestational age quartile at the time of blood sampling, group, gestation length, induction of parturition (yes or no), and recipient breed. For biomarker discovery, samples in the control group were considered the normal reference group and were compared with samples from LOS, stillbirth, or abortion. For each biomarker discovery procedure, the analysis pipeline consisted of several steps. First, a balanced held‐out test dataset was randomly selected and excluded from all feature‐selection, preprocessing‐parameter estimation, model‐training, and model‐tuning procedures. Secondly, a balanced training dataset was randomly selected, batch‐corrected for gestational age, and used for feature selection and model tuning. Transcriptomic variation associated with gestational age at blood collection was removed in the training dataset using the sva package. The testing dataset was then normalized with the training dataset using the bapred package to remove the batch effect.
Biomarkers were identified in the randomly selected training dataset first by filtering the top 100 DEG based on p‐values as determined by a negative binomial Wald test implemented with the DESeq2 package, and then by a Random Forest‐based feature selection with the Boruta package [27]. Predictive models were trained on the training dataset using multiple algorithms, implemented with the Caret package, which included support vector machines, Random Forest, penalized regression (glmnet), XGBoost, k‐nearest neighbors, and naïve Bayes. Hyperparameters, including regularization where applicable, were optimized within the training data. Model performance was subsequently evaluated on the independent held‐out test set, after normalization of the test data using parameters learned from the training dataset to assess generalization and reduce overfitting. The number of samples included in the training and held‐out test sets differed between groups and is specified in the Results section.
These steps were repeated across 10 independent random splits. After comparing model performance across these splits, the training dataset associated with the best‐performing model (naïve Bayes) was used to predict outcomes in the remaining samples. Model robustness was further assessed by repeatedly testing two randomly selected samples per iteration (one per class) over 1000 repetitions. Final class assignment was based on majority agreement across runs. To further increase classification confidence, samples were retained only if they were correctly classified in more than 55% of iterations, or if they showed agreement with the opposite class in 100% of iterations, indicating consistent disagreement between the molecular prediction and the recorded phenotype.
3. Results
3.1. Incidence of Large Offspring Calves, Stillbirths, and Abortion
The mean body weight of the newborn calves was 41 ± 6 kg for Holstein recipients (n = 417) and 39 ± 5 kg for crossbred (n = 101) recipients, respectively. Based on birth weight, calves were classified into two groups. The control group consisted for calves within ± 0.2 SD of the mean birth weight (n = 87) or calves within ± 1 and 2 SD of the mean birth weight (n = 55). The LOS group were calves with birth weight more than two standard deviations from the mean birth weight (n = 16). Using this criterion, the incidence of LOS was 3.1% (16 calves out of 518 female calves born live). There were 51 stillbirths in the 569 pregnancies in which the expected day of gestation was no more than 12 days early (9.0%). From these stillbirth cases, blood samples were randomly selected for transcriptomics. There were an additional 11 stillbirths occurring more than 12 days before expected day of gestation so the total incidence of stillbirth was 10.7% (62/580). It is possible that some stillbirth calves include cases of LOS but birth weight was not recorded. Incidence of abortion before 210 days was 11.9% (96 cases from 809 pregnancies for which data were available).
3.2. Explanation of Variance in the Transcriptome
An initial analysis was performed to determine the contribution of various factors [gestational day at blood sampling classified in quartiles, group, gestation length, induction of parturition (yes/no), and recipient breed] to variance in the transcriptome. Of the explained variation (11.5% of total), gestational day quartile was the greatest explanatory variable which accounted for 37.9% of the variance (Figure 1). Accordingly, subsequent analyses were performed after controlling the effect of gestational age at blood collection.
FIGURE 1.
Contribution of various factors including gestation day quartile, group, gestation length, induction to parturition (yes or no), and recipient breed to the explained variance in the transcriptome collected from pregnancy recipients from Days 40 to 50 of gestation after embryo transfer.
3.3. Differential Gene Expression Associated With Gestation of a Fetus That Developed Large Offspring Syndrome
Gene expression in blood samples of 96 pregnant females that subsequently gave birth to calves of normal body weight (control group) was compared to gene expression in blood samples of 15 pregnant females that subsequently gave birth to a LOS calf (LOS group). Differentially‐expressed genes (DEG) were defined as those with p < 0.01. From a total of 8973 genes identified, 32 were differentially‐expressed, of which 17 were upregulated, and 15 downregulated in heifers with LOS calves (Figure 2A). A list of DEG is shown in File S1, Table S1. There was only one significant DEG, an unannotated gene upregulated in LOS pregnancies, if a more stringent criterion, false discovery rate (FDR) < 0.05, was used.
FIGURE 2.
Differentially expressed genes in pregnant recipients comparing pregnancy outcomes of large calf offspring (A), stillbirth (B), or abortion (C) versus control calves. Upregulated genes are represented by orange dots and downregulated genes are represented by blue dots. The dots above the black dashed line on the y‐axis reflects the differentially expressed genes based on p < 0.01. Venn diagram (D) representing the overlap of differentially‐expressed genes for upregulated genes (left) and downregulated genes (right).
Gene ontology enrichment analysis of upregulated DEG revealed pathways associated with extracellular matrix organization, extracellular matrix assembly, and basement membrane organization (Figure 3A), suggesting alterations in extracellular matrix structure and tissue remodeling in blood cells of mothers carrying LOS conceptuses. There were no pathways associated with the downregulated DEG.
FIGURE 3.
Lollipop plots showing the top 10 enriched ontological terms for differentially expressed genes when comparing recipients of large offspring syndrome calves (A), stillbirth calves (B), and abortions (C) compared to control calves. Analyses were conducted separately for upregulated and downregulated genes. There were no significant ontologies for downregulated genes for LOS or upregulated genes for abortion.
3.4. Differential Gene Expression Associated With Gestation of a Fetus That Subsequently Experienced Stillbirth
Gene expression in blood samples of 96 pregnant females that subsequently gave birth to calves of normal body weight (control group) was compared to gene expression in blood samples of 43 pregnant females that subsequently gave birth to a stillborn calf (Figure 2B). A total of 176 DEG were identified, of which 81 were upregulated and 95 downregulated in heifers with stillbirth (File S1, Table S2). Using FDR < 0.05 as a cutoff, there was one upregulated gene (CD109) and five downregulated genes (ACTC1, LMOD2, MYBPC1, MYH7, and MYOZ2). Gene ontology enrichment analysis of upregulated DEG identified pathways associated with cell cycle regulation, chromosome organization, and chromosome segregation, whereas downregulated DEG were enriched for pathways related to muscle structure development (Figure 3B).
3.5. Differential Gene Expression Associated With Gestation of a Fetus That Subsequently Aborted
Gene expression in blood samples of 96 pregnant females that subsequently gave birth to calves of normal body weight was compared to gene expression in blood samples of 49 pregnant females that subsequently aborted before 210 days (Figure 2C). A total of 100 DEG were identified, of which 60 were upregulated and 40 downregulated in heifers that aborted (File S1, Table S3). There were only two DEG using FDR < 0.05 (GALNT14 and an unannotated gene), with both being upregulated in blood cells from the abortion group. Gene ontology analysis indicated no pathways associated with the upregulated DEG, whereas downregulated genes were enriched for immune‐related pathways, including cytokine‐mediated signaling, interferon signaling, and defense response to viral infection (Figure 3C).
3.6. Concordance of Differential Gene Expression Between Groups With Various Negative Outcomes After Embryo Transfer
A Venn diagram analysis was performed to assess whether any DEG identified using the control group were shared among the LOS, stillbirth, and abortion groups. Shared genes were considered as those that exhibited the same direction of change in expression across groups, being either consistently upregulated or consistently downregulated (Figure 2D). Overall, there were only a few genes that overlapped among groups. The upregulated genes that overlapped between abortion and stillbirth groups were CAPN13, EZH2, FAM180A, and IQGAP3. There was one upregulated gene that overlapped between LOS and abortion (unannotated gene). There was a total of four downregulated genes (AOX4, ATP8B4, PRKN, and RSAD2) shared between the stillbirth and abortion. There was only one upregulated gene (OR7D2) and no downregulated genes shared between LOS and the stillbirth group and no gene between LOS and abortion.
3.7. Biomarker Discovery for Large Offspring Syndrome
Samples randomly assigned to the training set (n = 10 controls and n = 10 LOS) were used for biomarker discovery and model development. The split yielded 26 gene‐signatures that classified calves in the held‐out dataset (n = 5 controls and n = 5 LOS) as LOS with a 90% accuracy, 100% sensitivity, and 80% specificity. One of the five classified normal samples was incorrectly predicted as LOS. This sample corresponded to the calf with the highest birth weight for the normal group. None of the LOS calves were classified as normal. The distribution of samples according to whole‐transcriptome expression, or according to biomarker‐gene expression within the training samples, is shown in Figure 4A.
FIGURE 4.
Principal component analysis. Graphs on the left represent distribution of all samples based on the whole transcriptome. Graphs on the right represent the training samples according to expression of the biomarker genes. Data are shown for (A) large offspring syndrome, (B) stillbirth, and (C) abortion.
Identified biomarker genes are shown in Table 2. Note that 18 biomarker genes were positively associated with birth to a LOS calf and 8 were negatively associated. Ontologies for biomarker genes that were positively associated included cell surface, system development, and multicellular organism development while gene ontologies for biomarker genes that were negatively associated were immune‐related.
TABLE 2.
List of biomarker genes for large offspring syndrome.
| Entrez ID | Symbol | Gene name | Entrez ID | Symbol | Gene name |
|---|---|---|---|---|---|
| Positively associated | |||||
| 505687 | PLS3 | Plastin 3 | 507054 | IL33 | Interleukin 33 |
| 100299 080 | SAPCD2 | Suppressor APC domain containing 2 | 534326 | ADGRG6 | Adhesion G protein‐coupled receptor G6 |
| 280733 | BGN | Biglycan | 536153 | GPC6 | Glypican 6 |
| 523755 | ITGA11 | Integrin subunit alpha 11 | 615651 | ZBED3 | Zinc finger BED‐type containing 3 |
| 533803 | TCEA3 | Transcription elongation factor A3 | 281733 | CXADR | CXADR Ig‐like cell adhesion molecule |
| 519654 | ADAM20 | ADAM metallopeptidase domain 20 | 540106 | PDGFB | Platelet derived growth factor subunit B |
| 781282 | SVEP1 | Sushi, von Willebrand factor type A, EGF and pentraxin domain containing 1 | 535750 | ALS2 | Alsin Rho guanine nucleotide exchange factor ALS2 |
| 614786 | WWTR1 | WW domain containing transcription regulator 1 | 112446 907 | LOC112446907 | U6 spliceosomal RNA |
| 614940 | PRSS35 | Serine protease 35 | 782075 | IL26 | Interleukin 26 |
| Negatively associated | |||||
| 520709 | IL1RL1 | Interleukin 1 receptor like 1 | 512732 | STMN4 | Stathmin 4 |
| 511023 | STAT2 | Signal transducer and activator of transcription 2 | 338039 | CASP4 | Caspase 4, apoptosis‐related cysteine peptidase |
| 535490 | IFIH1 | Interferon induced with helicase C domain 1 | 531838 | CHRNA2 | Cholinergic receptor nicotinic alpha 2 subunit |
| 514373 | HERC5 | HECT and RLD domain containing E3 ubiquitin protein ligase 5 | 527520 | HERC6 | HECT and RLD domain containing E3 ubiquitin protein ligase family member 6 |
The selected model was then applied to the remaining 91 samples, comprising 86 normal and five LOS calves. Eight of the 86 normal samples were misclassified as LOS, whereas all LOS cases (n = 5) were correctly identified, leading to an overall 91% accuracy and specificity. The LOS phenotype appears biologically consistent, as reflected by a sensitivity of 100%, indicating no false negatives. The high negative predictive value (NPV) of 100% reflects strong predictive performance for identifying normal samples, although the positive predictive value of 38% indicates overlap between phenotypes, reflecting both the low prevalence of LOS in the dataset and the presence of intermediate molecular profiles between phenotypic groups.
When the analysis was restricted to samples for which at least 55% of the repeated predictions agreed on the assigned class, 86 samples met this confidence threshold. Of these, three of the 81 calves with a recorded normal phenotype were predicted as LOS, whereas all five LOS calves were correctly identified. Under this criterion, accuracy increased to 96.5% and specificity to 96.3%, while sensitivity remained 100%.
3.8. Biomarkers of Stillbirth and Abortion
Using an approach similar to that applied for LOS, 24 and 28 candidate biomarker genes were identified for the prediction of stillbirth (Table 3) and abortion (Table 4), respectively. In the held‐out test sets (n = 10 per group), model performance showed accuracy, sensitivity, and specificity values ranging from 80% to 90%, with higher performance observed for the stillbirth group. The distribution of samples according to whole‐transcriptome expression, or according to biomarker‐gene expression within the training samples for stillbirth and abortion, is shown in Figure 3B,C, respectively. When the models were applied to all the remaining samples not included in the corresponding training sets, performance decreased to 72%–73% for stillbirth and 62%–63% for abortion, which may reflect greater heterogeneity within these phenotypic groups. Restricting the analysis to samples with ≥ 55% agreement across repeated predictions increased classification consistency for both groups, suggesting that the biomarker signals may be concentrated in a subset of samples with more consistent molecular profiles, potentially reflecting phenotypic heterogeneity rather than a predictive signature shared by all affected samples.
TABLE 3.
List of biomarker genes for stillbirth.
| Entrez ID | Symbol | Gene name | Entrez ID | Symbol | Gene name |
|---|---|---|---|---|---|
| Positively associated | |||||
| 508869 | RND1 | Rho family GTPase 1 | 537366 | CDPF1 | Cysteine rich DPF motif domain containing 1 |
| 525579 | OTUD7B | OTU deubiquitinase 7B | 777770 | KIF3C | Kinesin family member 3C |
| 615501 | SH3RF3 | SH3 domain containing ring finger 3 | 533996 | PHLDB1 | Pleckstrin homology like domain family B member 1 |
| 615490 | SMOC2 | SPARC related modular calcium binding 2 | 767586 | KANK2 | KN motif and ankyrin repeat domains 2 |
| 509963 | ANGPTL4 | Angiopoietin like 4 | 538340 | MAP3K9 | Mitogen‐activated protein kinase kinase kinase 9 |
| 613639 | TSPAN12 | Tetraspanin 12 | 518441 | WDR53 | WD repeat domain 53 |
| 534043 | CD109 | CD109 molecule | 516661 | LGI3 | Leucine rich repeat LGI family member 3 |
| 519892 | NUAK1 | NUAK family kinase 1 | 524990 | PLVAP | Plasmalemma vesicle associated protein |
| 281489 | SLAMF1 | Signaling lymphocytic activation molecule family member 1 | 616564 | SSBP2 | Single stranded DNA binding protein 2 |
| Negatively associated | |||||
| 540357 | CHST7 | Carbohydrate sulfotransferase 7 | 536741 | MME | Membrane metalloendopeptidase |
| 100295 476 | EREG | Epiregulin | 515700 | IL1R2 | Interleukin 1 receptor Type 2 |
| 516323 | DYNLT4 | Dynein light chain Tctex‐Type 4 | 767965 | TEX33 | Testis expressed 33 |
TABLE 4.
List of biomarker genes for abortion.
| Entrez ID | Symbol | Gene name | Entrez ID | Symbol | Gene name |
|---|---|---|---|---|---|
| Positively associated | |||||
| 614603 | SARDH | Sarcosine dehydrogenase | 112447918 | LOC112447918 | Small nucleolar RNA SNORA7 |
| 514435 | DMRT2 | Doublesex and mab‐3 related transcription factor 2 | 538340 | MAP3K9 | Mitogen‐activated protein kinase kinase kinase 9 |
| 789769 | SH2D5 | SH2 domain containing 5 | 522921 | HBEGF | Heparin binding EGF like growth factor |
| 112443 642 | LOC112443642 | Small nucleolar RNA SNORD82 | 530616 | TCF3 | Transcription factor 3 |
| 533379 | MLF1 | Myeloid leukemia factor 1 | 282484 | SLC34A2 | Solute carrier family 34 member 2 |
| 515594 | PRMT5 | Protein arginine methyltransferase 5 | 540602 | NANS | N‐acetylneuraminate synthase |
| 616664 | MSH4 | MutS homolog 4 | 540863 | CCDC18 | Coiled‐coil domain containing 18 |
| 535401 | EFCAB6 | EF‐hand calcium binding domain 6 | 783548 | TNKS1BP1 | Tankyrase 1 binding protein 1 |
| 615507 | GSTM3 | Glutathione S‐transferase mu 3 | 618738 | SULT4A1 | Sulfotransferase family 4A member 1 |
| 112443 561 | LOC112443561 | Small nucleolar RNA SNORA7 | 510557 | HSF2 | Heat shock transcription factor 2 |
| 540500 | GALNT14 | Polypeptide N‐acetylgalactosaminyltransferase 14 | |||
| Negatively associated | |||||
| 539420 | ELF5 | E74 like ETS transcription factor 5 | 539695 | OLIG2 | Oligodendrocyte transcription factor 2 |
| 524375 | ALPK1 | Alpha kinase 1 | 508202 | KRT79 | Keratin 79 |
| 768233 | ELMOD1 | ELMO domain containing 1 | 507280 | TM4SF18 | Transmembrane 4 L six family member 18 |
| 618114 | PJVK | Pejvakin | |||
4. Discussion
The major finding of this study was that maternal blood‐cell transcriptomic profiles during early gestation could predict the birth of an LOS calf. In particular, a machine learning approach identified a total of 26 biomarker genes predictive of LOS with 90% accuracy, with no samples from the LOS group being misclassified. Thus, the predictive model showed strong sensitivity and overall discriminatory ability for LOS. Models were less predictive of whether the pregnancy would result in stillbirth or abortion. The fact that models were less reliable than those developed for LOS likely reflects the biological heterogeneity associated with stillbirth and abortion. For example, incidence of both conditions is affected by factors such as infectious diseases, nutrition, or genetics [28], all of which may differentially affect the maternal blood transcriptome.
Findings from previous studies suggest that fetal size can influence the maternal blood transcriptome, even when the fetus is derived through in vitro production and is therefore genetically distinct from the surrogate mother. For example, several clusters of co‐expressed genes in maternal blood cells were correlated to fetal weight at Day 42 of gestation [22]. In addition, Rivera et al. [7] identified several genes whose expression in blood cells at Day 55 and 105 of gestation differed in dams carrying the two largest fetuses at Day 105 of gestation as compared to other females. In these studies, blood gene expression and fetal weight were assessed at the same gestational stage, either Day 42 or Day 105. Nonetheless, a more recent study identified a maternal blood transcriptomic signature at Day 63 of gestation that predicted calf birth weight [21], supporting the concept that changes in the maternal blood transcriptome during early gestation may provide information about the fetal phenotype several months later at birth.
The present study extends this concept by identifying maternal blood genes associated with LOS, stillbirth, and abortion at an even earlier stage of gestation, around Days 40–50, which corresponds to the beginning of the fetal period in cattle. The observation that all three conditions were associated with changes in the maternal blood transcriptome early in gestation suggests that conceptus dysfunction is reflected in altered maternal physiology. The DEG, although represented by a small number of genes, were associated with distinct biological pathways for each pregnancy outcome, indicating that each condition may induce a specific maternal response. Fewer genes associated with calf birth weight were identified in the present study than earlier [21]. This difference may be related to the approximately 20‐day later sampling time in that earlier study, as the maternal blood transcriptome at a more advanced stage of gestation may more closely reflect the subsequent growth trajectory and birth phenotype of the fetus. Indeed, it is now well established that the conceptus (i.e., fetal‐placental unit) can change the function of maternal blood cells throughout gestation and as early as Day 18 of pregnancy, when trophoblast‐derived interferons cause upregulation of interferon‐stimulated genes [29]. Furthermore, the proportion of white blood cells that were CD4+CD25+ cells was elevated in pregnant cows as compared to non‐pregnant cows at Days 33–34 of gestation [17].
It is unknown whether changes in maternal function reflected by alterations in the blood cell transcriptome contribute to the development of LOS, stillbirth or abortion, or are the result of fetal abnormalities alone. Among the biomarker genes for LOS were those related to pro‐inflammatory cytokines such as IL26 and IL33. Another biomarker gene was CASP4 involved in immune cell responses [30] and may indicate changes in maternal innate immune activation in response to an abnormal pregnancy environment. Perhaps, dysregulation of immune regulatory pathways contributes to abnormal fetal development. Nonetheless, it is most likely that changes in the maternal blood cell transcriptome are a response to altered fetal and/or placental development rather than a cause of LOS, given that epigenetic abnormalities are thought to play a central role in the development of this syndrome [8]. The rarity of LOS as a syndrome has made conclusions about the relationship of the syndrome to alterations in placental function uncertain [8]. One greatly enlarged fetus recovered at Day 86 of gestation had increased fetal size and placental weight [10]. Similarly, fetal weight was correlated with placental weight at Day 90 of gestation [31]. It is possible that alterations in placental function are related to early changes in the in vitro produced blastocyst caused by the artificial environment in which it is produced. These early perturbations could possibly lead to changes in the maternal immune response to the conceptus and result in changes in the blood cell transcriptome.
The criterion used to define LOS in the current study was a birth weight greater than 2 standard deviations above the average birth weight, that is, 50.3 kg for Holstein calves and 49.9 kg for crossbred calves, resulting in a 3.1% incidence of LOS. Although the reported incidence of LOS/overgrowth among in vitro‐derived calves varies considerably among studies, most reports describe frequencies ranging from approximately 4%–14%, with some studies reporting substantially higher incidences depending on embryo production and culture conditions [8]. Despite its low occurrence, LOS can have substantial consequences for both the calf and the dam including dystocia. Dystocia can increase the incidence of stillbirth [32] and affect dam longevity by increasing relative risk of culling [33] as well as subsequent milk production [34, 35] and reproductive performance of the dam [34]. Thus, there is great value in predicting LOS early in gestation because pregnancy could be terminated or special considerations made for periparturient veterinary care.
The primary limitation of this study is the relatively small sample size in the LOS group (n = 15) which probably contributed to the low positive predictive value obtained from the model using all predictions. Increasing the number of LOS samples to achieve a sample size comparable to that of normal calves is advised to strengthen the robustness of the conclusions. Moreover, future samples should be collected from multiple farms to assess the generalizability of the identified biomarkers across different housing and management conditions. As additional datasets become available, their integration with the present findings will further reinforce and validate the observed transcriptomic signals. Regarding the stillbirth and abortion groups, the main limitation is the unknown timing and causes of these outcomes, which may have contributed to the reduced predictive performance of the models. In particular, the variable timing of abortion and the multifactorial nature of stillbirth may have increased heterogeneity in gene expression profiles, potentially reducing the consistency of differential signals required for robust biomarker predictive detection.
In summary, this study demonstrates that maternal blood transcriptomic profiles during early gestation contain informative signatures associated with adverse pregnancy outcomes in cattle, including LOS, stillbirth, and abortion. Although predictive performance varied among outcomes, differential gene expression analyses revealed limited overlap in associated genes, suggesting that each condition is characterized by partially distinct molecular signatures. Overall, results support the potential use of maternal blood as a non‐invasive source of biomarkers that could be developed into a bioassay predictive of pregnancy outcome and calf phenotype.
Author Contributions
Tatiane S. Maia , Rebecca Krisher, and Peter J. Hansen conceived the study; Rebecca Krisher, Paula Rodriquez‐Vilamil, and Peter J. Hansen coordinated activities and provided resources; Tatiane S. Maia , Jeniffer Pamela J. Laguna, Jeniffer F. Escoto, Yanping Zhang, and Abigail J. Sproull collected and processed samples and collected data; Maria B. Rabaglino and Tatiane S. Maia conducted data analysis; Tatiane S. Maia wrote the first draft of the paper, and all authors reviewed, edited, and approved the final manuscript.
Funding
This work was supported by Genus (Genus plc); L.E. “Red” Larson Endowment.
Ethics Statement
Animal use was approved by the University of Florida Institutional Animal Care and Use Committee (project IACUC202300000030).
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: List of genes found on the blood transcriptome on recipients that conceived a large offspring calf versus conceived a normal calf.
Table S2: List of genes found on the blood transcriptome on recipients that conceived a stillborn calf versus conceived a normal calf.
Table S3: List of genes found on the blood transcriptome on recipients that aborted up to 210 days of gestation versus recipients that conceived a normal calf.
Acknowledgments
The authors thank everyone at Arroyo Seco farm, especially Justin Veenendaal, Luis Jiménez and Moisés Dominguez, for facilitation of blood and data collection. The authors thank Kelsey Lockhart and Sudhir Naswa from Genus and Thiago Amaral from Genus PLC/ABS‐Brasil for providing guidance with data recording and RNA analysis. The authors also thank Ru Dai at the University of Florida Interdisciplinary Center for Biotechnology Research for assistance with RNA sequencing and Anderson Silva and Massimiliano S. Tagliamonte from the University of Florida Bioinformatics Core Facility for assistance with data processing. The graphical abstract was prepared using BioRender (https://BioRender.com/xmc29vb). Research was funded by a grant from Genus plc and the L.E. “Red” Larson Endowment. Declaration of generative AI and AI‐assisted technologies in the manuscript preparation process: During the preparation of this work, the authors used ChatGPT (OpenAI) to assist with improvement of grammar and sentence clarity. The authors reviewed and edited the output as needed and take full responsibility for the content of the published article. The graphical abstract was designed using the AI‐assisted features of BioRender. All scientific content, data interpretation and conclusions were independently developed by the authors.
Data Availability Statement
Raw data are deposited in the NCBI Gene Expression Omnibus and are accessible through GEO accession number GSE343491.
References
- 1. Hansen P. J., “Implications of Assisted Reproductive Technologies for Pregnancy Outcomes in Mammals,” Annual Review of Animal Biosciences 15 (2020): 395–413, 10.1146/annurev-animal-021419-084010. [DOI] [PubMed] [Google Scholar]
- 2. Velazquez M. A., Idriss A., Chavatte‐Palmer P., and Fleming T. P., “The Mammalian Preimplantation Embryo: Its Role in the Environmental Programming of Postnatal Health and Performance,” Animal Reproduction Science 256 (2023): 107321, 10.1016/j.anireprosci.2023.107321. [DOI] [PubMed] [Google Scholar]
- 3. Young L. E., Sinclair K. D., and Wilmut I., “Large Offspring Syndrome in Cattle and Sheep,” Reviews of Reproduction 3 (1998): 155–163, 10.1530/ror.0.0030155. [DOI] [PubMed] [Google Scholar]
- 4. van Wagtendonk‐de Leeuw A. M., Aerts B. J., and den Daas J. H., “Abnormal Offspring Following In Vitro Production of Bovine Preimplantation Embryos: A Field Study,” Theriogenology 49 (1998): 883–894, 10.1016/s0093-691x(98)00038-7. [DOI] [PubMed] [Google Scholar]
- 5. Farin P. W., Crosier A. E., and Farin C. E., “Influence of In Vitro Systems on Embryo Survival and Fetal Development in Cattle,” Theriogenology 55 (2001): 151–170, 10.1016/s0093-691x(00)00452-0. [DOI] [PubMed] [Google Scholar]
- 6. Farin C. E., Farmer W. T., and Farin P. W., “Pregnancy Recognition and Abnormal Offspring Syndrome in Cattle,” Reproduction, Fertility, and Development 22 (2010): 75–87, 10.1071/RD09217. [DOI] [PubMed] [Google Scholar]
- 7. Rivera R. M., Goldkamp A. K., Patel B. N., et al., “Identification of Large Offspring Syndrome During Pregnancy Through Ultrasonography and Maternal Blood Transcriptome Analyses,” Scientific Reports 12 (2022): 10540, 10.1038/s41598-022-14597-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Nava‐Trujillo H. and Rivera R. M., “Large Offspring Syndrome in Ruminants: Current Status and Prediction During Pregnancy,” Animal 17 (2023): 100740, 10.1016/j.animal.2023.100740. [DOI] [PubMed] [Google Scholar]
- 9. Kruip T. A. M. and den Daas J. H. G., “In Vitro Produced and Cloned Embryos: Effects on Pregnancy, Parturition and Offspring,” Theriogenology 47 (1997): 43–52, 10.1016/S0093-691X(96)00338-X. [DOI] [Google Scholar]
- 10. Siqueira L. G., Tribulo P., Chen Z., et al., “Colony‐Stimulating Factor 2 Acts From Days 5 to 7 of Development to Modify Programming of the Bovine Conceptus at Day 86 of Gestation,” Biology of Reproduction 96 (2017): 743–757, 10.1093/biolre/iox018. [DOI] [PubMed] [Google Scholar]
- 11. Crowe A. D., Doyle R. C., Lonergan P., and Butler S. T., “Gestation Length, Calf Birth Weight, Calving Difficulty, Perinatal Mortality, and Calf Health Following Timed Artificial Insemination or Embryo Transfer With Fresh or Frozen In Vitro‐Produced Embryos,” Journal of Dairy Science 108 (2025): 12759–12773, 10.3168/jds.2025-26958. [DOI] [PubMed] [Google Scholar]
- 12. Chen Z., Robbins K. M., Wells K. D., and Rivera R. M., “Large Offspring Syndrome: A Bovine Model for the Human Loss‐Of‐Imprinting Overgrowth Syndrome Beckwith‐Wiedemann,” Epigenetics 8 (2013): 591–601, 10.4161/epi.24655. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Choufani S., Shuman C., and Weksberg R., “Beckwith–Wiedemann Syndrome,” American Journal of Medical Genetics—Seminars in Medical Genetics, Part C 154C (2010): 343–354, 10.1002/ajmg.c.30267. [DOI] [PubMed] [Google Scholar]
- 14. Chen Z., Hagen D. E., Elsik C. G., et al., “Characterization of Global Loss of Imprinting in Fetal Overgrowth Syndrome Induced by Assisted Reproduction,” Proceedings. National Academy of Sciences. United States of America 112 (2015): 4618–4623, 10.1073/pnas.1422088112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Lazzari G., Wrenzycki C., Herrmann D., et al., “Cellular and Molecular Deviations in Bovine In Vitro‐Produced Embryos Are Related to the Large Offspring Syndrome,” Biology of Reproduction 67 (2002): 767–775, 10.1095/biolreprod.102.004481. [DOI] [PubMed] [Google Scholar]
- 16. Behboodi E., Anderson G. B., BonDurant R. H., et al., “Birth of Large Calves That Developed From In Vitro‐Derived Bovine Embryos,” Theriogenology 44 (1995): 227–232, 10.1016/0093-691X(95)00172-5. [DOI] [PubMed] [Google Scholar]
- 17. Oliveira L. J. and Hansen P. J., “Deviations in Populations of Peripheral Blood Mononuclear Cells and Endometrial Macrophages in the Cow During Pregnancy,” Reproduction 136 (2008): 481–490, 10.1530/REP-08-0218. [DOI] [PubMed] [Google Scholar]
- 18. Gomez‐Lopez N., Romero R., Hassan S. S., et al., “The Cellular Transcriptome in the Maternal Circulation During Normal Pregnancy: A Longitudinal Study,” Frontiers in Immunology 10 (2019): 2863, 10.3389/fimmu.2019.02863. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Mohapatra S. K., Panda B. S. K., Verma A. K., Kapila R., and Dang A. K., “Implantation Associated Changes in Expression Profile of Indoleamine‐2,3‐Dioxygenase 1, Th1–Th2 Cytokines and Interferon‐Stimulated Genes on Neutrophils and Peripheral Blood Mononuclear Cells of Crossbred Cows,” Journal of Reproductive Immunology 142 (2020): 103188, 10.1016/j.jri.2020.103188. [DOI] [PubMed] [Google Scholar]
- 20. de los Santos J. A., Andrade J. P. N., Cangiano L. R., Iriarte A., Peñagaricano F., and Parrish J. J., “Transcriptomic Analysis Reveals Gene Expression Changes in Peripheral White Blood Cells of Cows After Embryo Transfer: Implications for Pregnancy Tolerance,” Reproduction in Domestic Animals 58 (2023): 946–954, 10.1111/rda.14371. [DOI] [PubMed] [Google Scholar]
- 21. Thompson L., van Helvoort P. G., Duijn M., et al., “The Maternal Blood Transcriptome Reflects Changes in Fetal Growth and Is an Accurate Predictor of Birth Weight in Cattle,” FASEB Journal 39 (2025): e71201, 10.1096/fj.202502648R. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Rabaglino M. B., Sánchez J. M., McDonald M., O'Callaghan E., and Lonergan P., “Maternal Blood Transcriptome as a Sensor of Fetal Organ Maturation at the End of Organogenesis in Cattle,” Biology of Reproduction 109 (2023): 749–758, 10.1093/biolre/ioad103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Bó G. and Mapletoft R., “Evaluation and Classification of Bovine Embryos,” Animal Reproduction 54 (2013): 344–348. [Google Scholar]
- 24. McLoughlin K. E., Correia C. N., Browne J. A., et al., “RNA‐Seq Transcriptome Analysis of Peripheral Blood From Cattle Infected With Mycobacterium bovis Across an Experimental Time Course,” Frontiers in Veterinary Science 8 (2021): 662002, 10.3389/fvets.2021.662002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Correia C. N., McLoughlin K. E., Nalpas N. C., et al., “RNA Sequencing Reveals Extremely Low Levels of Reticulocyte‐Derived Globin Gene Transcripts in Peripheral Blood From Horses (Equus caballus) and Cattle (Bos taurus),” Frontiers in Genetics 9 (2018): 278, 10.3389/fgene.2018.00278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Love M. I., Huber W., and Anders S., “Moderated Estimation of Fold Change and Dispersion for RNA‐Seq Data With DESeq2,” Genome Biology 2014, no. 15 (2014): 550, 10.1186/s13059-014-0550-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Kursa M. B. and Rudnicki W. R., “Feature Selection With the Boruta Package,” Journal of Statistical Software 36 (2010): 1–13, 10.18637/jss.v036.i11. [DOI] [Google Scholar]
- 28. Mee J. F., “Investigation of Bovine Abortion and Stillbirth/Perinatal Mortality – Similar Diagnostic Challenges, Different Approaches,” Irish Veterinary Journal 73 (2020): 20, 10.1186/s13620-020-00172-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Rocha C. C., da Silva Andrade S. C., de Melo G. D., et al., “Early Pregnancy‐Induced Transcripts in Peripheral Blood Immune Cells in Bos indicus Heifers,” Scientific Reports 10 (2020): 13733, 10.1038/s41598-020-70616-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Tian G., Li Q., Niu L., et al., “CASP4 can be a Diagnostic Biomarker and Correlated With Immune Infiltrates in Gliomas,” Frontiers in Oncology 12 (2023): 1025065, 10.3389/fonc.2022.1025065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Bertolini M. and Anderson G. B., “The Placenta as a Contributor to Production of Large Calves,” Theriogenology 57 (2002): 181–187, 10.1016/s0093-691x(01)00665-3. [DOI] [PubMed] [Google Scholar]
- 32. Bicalho R. C., Galvão K. N., Cheong S. H., Gilbert R. O., Warnick L. D., and Guard C. L., “Effect of Stillbirths on Dam Survival and Reproduction Performance in Holstein Dairy Cows,” Journal of Dairy Science 90 (2007): 2797–2803, 10.3168/jds.2006-504. [DOI] [PubMed] [Google Scholar]
- 33. Morek‐Kopeć M., Zarnecki A., Ptak E., and Otwinowska‐Mindur A., “Effect of Calving Difficulties and Calf Mortality on Functional Longevity in Polish Holstein‐Friesian Cows,” Animals (Basel) 11 (2021): 2792, 10.3390/ani11102792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Dematawewa C. M. and Berger P. J., “Effect of Dystocia on Yield, Fertility, and Cow Losses and an Economic Evaluation of Dystocia Scores for Holsteins,” Journal of Dairy Science 80 (1997): 754–761, 10.3168/jds.s0022-0302(97)75995-2. [DOI] [PubMed] [Google Scholar]
- 35. Roche S. M., Ross J. A., Schatz C., et al., “Impact of Dystocia on Milk Production, Somatic Cell Count, Reproduction and Culling in Holstein Dairy Cows,” Animals 13 (2023): 346, 10.3390/ani13030346. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: List of genes found on the blood transcriptome on recipients that conceived a large offspring calf versus conceived a normal calf.
Table S2: List of genes found on the blood transcriptome on recipients that conceived a stillborn calf versus conceived a normal calf.
Table S3: List of genes found on the blood transcriptome on recipients that aborted up to 210 days of gestation versus recipients that conceived a normal calf.
Data Availability Statement
Raw data are deposited in the NCBI Gene Expression Omnibus and are accessible through GEO accession number GSE343491.