2.1.1 Animals
All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee at Purdue University prior to beginning the study. Animals used were a subset of a larger previously described study (8). Briefly, gilts (n=2146) born on a commercial farm in central Indiana were enrolled in a longitudinal observational study (Figure 1). Initial individual weights were recorded between d 0 and 2 postnatal and are referred to hereafter as birth weights. Piglets were weighed and processed on postnatal d 2-3 (ear tag, 200 mg iron, tail docking, antibiotic administration) as per routine management on the farm. Cross-fostering between litters of similar aged piglets occurred after piglets were given an ear tag (litter size standardized to 14 ± 1 piglets). Data on farrowing date, birth litter size, size of gilt’s litter at her weaning, birth weight, and weaning weight were entered into MetaFarms (MetaFarms, Inc., Burnsville, MN) database for each gilt.
Gilts (n=1084) were weaned at 21 ± 4 d, selected as potential replacement sows, and moved to the farm’s onsite nursery (Figure 1). Gilts that were not selected as replacement females were transferred to an offsite wean to finish facility for market production. Immediately following selection into the onsite nursery, swabs of the anterior vagina were taken for lipidomic analysis. Gilt’s vulva was sprayed with ethanol and wiped clean with gauze before the swab was taken. A human pap smear brush (Rovers® EndoCervex-Brush®, Oss, Netherlands) was placed into the vagina as far as possible and then rotated clockwise to get a representative scraping of the anterior vagina. Swabs were taken in duplicate and then placed in 15 mL polypropylene conical tubes (Corning™ Falcon™, Corning, NY) and immediately placed on ice, and transferred to the Purdue laboratory and stored at -80ºC until analysis.
Gilts (n=400) approaching 25 weeks of age were selected to move from the on-site nursery into the gilt development unit (GDU) to receive daily, full-contact boar exposure to induce puberty. Following observance of a second estrus, gilts were moved to a gestation crate to be bred using artificial insemination (AI) on their third estrus. Gilts failing to show estrus by 28 weeks of age were administered a full dose of P.G. 600 (Intervet America, Inc., Millsboro, DE) and bred as soon as they came into estrus. Gilts that never showed estrus were removed from the selection pool. Breeding date, age of first estrus, and induction of estrus with P.G. 600 were entered into the Metafarms database.
2.1.2 Categorization of Fertility Groups
Performance data from breeding herd animals (n=400) were extracted from Metafarms reports which included date of birth, dates of estrus detection, herd removal (cull) date, reason for removal, farrowing date and litter information (total born, number born alive, preweaning mortality). Gilts that were culled from the breeding herd for non-reproductive reasons were removed from data set. Gilts that remained (n=353) were divided into fertile and infertile groups (Figure 1). Fertile animals were classified into one of three subclasses: sows with at least 26 PSY were defined as High Fertility (HF; n=82), 20-25 PSY were classified as Middle Fertility (MF; n=43), and less than 20 PSY were classified as Low Fertility (LF; n=54). PSY was chosen as it encompasses multiple facets of fertility and prolificacy, such as ovulation rate, wean to estrus interval, farrowing rate, and embryo survival rate (17-19). Infertile animals were divided into two subclasses based on whether they exhibited estrus following routine boar exposure or P.G. 600 administration. Infertile-Estrus (IFe; n= 36) were gilts that showed signs of estrus but did not become pregnant and Infertile-No Estrus (IFno; n=138) were gilts that never exhibited signs of estrus. A subset of the most fertile (HF; n=28) and infertile animals that did not exhibit estrus (IF; n=34) were randomly selected for lipidome analysis.
2.1.3 Statistical analysis of phenotypic traits
Analysis of phenotypic traits such as birth litter size, nursing litter size, average daily gain from birth to weaning, birth weight, weaning weight, and vulva width by fertility category was performed using the GLM procedure of SAS 9.4 (SAS Inst. Inc., Cary, NC).
2.1.4 Lipidomic Extraction and Multiple Reaction Monitoring (MRM) Profiling
Lipids were extracted from swabs using the Bligh and Dyer method (20). Samples were thawed at room temperature before the start of the extraction. Swabs were rinsed with 500 µl deionized water and vortexed to remove cellular material in the same 15 ml polypropylene conical tubes. Two hundred µl of sample was transferred to a new 1.7 ml tube (Axygen®, Corning, New York). Four hundred fifty µl of methanol prepared with 1mM butylated hydroxytoluene and 250 µl of chloroform was added to the solution and vortexed. Samples were then incubated for 15 min at 4º C. An additional 250 µl of both deionized water and chloroform were added to the tube and centrifuged 10 min at 3000 rcf at 4º C. The solution was separated into three phases consisting of the polar, protein, and organic (lipid) phase. The lipid phase was removed, placed into 1.7 ml microcentrifuge tubes, and dried in a vacuum concentrator for 8 h. Dried pellets were resuspended 200 µl of acetonitrile, methanol, and ammonium acetate 3:6.65:0.35 (v/v/v). Further 10X dilution of sample in solvent was used for direct injection.
Multiple reaction monitoring (MRM) profiling was done using a two-part process beginning with a discovery phase followed by a screening phase. The discovery phase was used to determine which lipids were detectable in the samples. For this, ten µl from each sample was pooled by phenotype into a 1.7 ml tube and then dried by nitrogen flow for 8 h. Dried lipid extracts were diluted in 200 µl of acetonitrile, methanol, and ammonium acetate 3:6.65:0.35 (v/v/v). Eight µl from each pooled sample were injected into the microautosampler (G1377A) in a QQQ6410 triple quadrupole mass spectrometer (Agilent Technologies, San Jose, CA) equipped with an ESI ion source. A solvent solution containing acetonitrile with 1% formic acid at 10 µl/min was pumped between injections (CapPump G1376A, Agilent Technologies, San Jose, CA). Pure methanol was injected between samples to remove any remaining lipids from the previous injection.
During the discovery phase, pooled samples were screened for the chemical classes of acylcarnitine (AC), cholesteryl ester (CE), ceramide, free fatty acids (FFA), phosphatidylcholines (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylglycerol (PG), phosphatidylserine (PS), and triacylglycerol (TAG) using MRM profiling methods previously reported (21-24). Additionally, lipids from previous research that had been found by MRM profiling to discriminate between gilts fed different postnatal diets were also screened (25). Our studies described in Harlow et al., 2019, identified 146 lipids that discriminated between gilts that suckled colostrum versus bottle fed milk replacer the first 48 h postnatal, and are referred to as Method 1 (15). Lipids in Method 1 were primarily TAG, PE, PC, and PG. A second set of 197 lipids distinguished between gilts fed a lard based fat supplementation and unsupplemented animals, and are referred to as Method 2 (15). Lipids in Method 2 were primarily glycerolipids (15). Processing the initial chemical class data was completed using MSConvert20 which converted each set of profiling method data into mzML format. Signal intensity for ions present in the samples was obtained using an inhouse script. Ions with values >30% in at least one of the samples compared to a blank within each profiling method were selected for the screening phase.
- Due to the large number of MRMs to be screened, the samples were interrogated using four lists of MRMs, which we refer as methods:
- Phosphatidylcholines (PC) (Supplemental Table 1)
- Free fatty acids (FFA) (Supplemental Table 2)
- Lipid classes defined by Harlow et al., as Method 1 (Supplemental Table 3)
- Lipid classes defined by Harlow et al., as Method 2 (Supplemental Table 4)
2.1.4 MRM profiling data analysis in MetaboAnalyst
For analysis, relative intensity of a given MRM in each sample was calculated by class of lipid or method employed in each screen, following removal of MRM ion pairs with intensities less than 1.3-fold of the blank sample. Relative intensity of MRM ion pairs was calculated by dividing intensity by the sum of intensities of all lipids within a sample by screening method. Relative intensity of MRM ion pairs were uploaded into MetaboAnalyst 3.0 (26) and data were normalized using autoscaling. Student t-test analysis was used to identify MRMs that distinguished between fertility phenotypes, using an alpha of 0.05 of nominal P-value to identify differentially distributed lipids. Biomarker analysis was completed using classical univariate receiver operating characteristic (ROC) curve analysis with area-under-the-curve (AUC) value used to determine a lipid’s potential as a biomarker. The following AUC scale was used to evaluate lipids as potential biomarkers: excellent = 0.9-1.0; good=0.8-0.9; fair=0.7-0.8; poor=0.6-0.7; fail=0.5-0.6 (27).