Biomarker metabolite mating of viable frozen-thawed in vitro-produced bovine embryos with pregnancy-competent recipients leads to improved birth rates

Selection of competent recipients before embryo transfer (ET) is indispensable for improving pregnancy and birth rates in cattle. However, pregnancy prediction can fail when the competence of the embryo is ignored. We hypothesized that the pregnancy potential of biomarkers could improve with information on embryonic competence. In vitro-produced embryos cultured singly for 24 h (from d 6 to 7) were transferred to d 7 synchronized recipients as fresh or after freezing and thawing. Recipient blood was collected on d 0 (estrus; n = 108) and d 7 (4-6 h before ET; n = 107) and plasma was analyzed by nuclear magnetic resonance (1H+NMR). Spent embryo culture medium (CM) was collected and analyzed by ultra-high-performance liquid chromatography tandem mass spectrometry in a subset of n = 70 samples. Concentrations of metabolites quantified in plasma (n = 35) were statistically analyzed as a function of pregnancy diagnosed on d 40, d 62 and birth. Univariate analysis with plasma metabolites consisted of a block study with controllable fixed factors (i.e., embryo cryopreservation, recipient breed, and day of blood collection; Wilcoxon test and t-test). Metabolite concentrations in recipients and embryos were independently analyzed by iterations that reclassified embryos or recipients using the support vector machine. Iterations identified some competent embryos, but mostly competent recipients that had a pregnancy incompetent partner embryo. Misclassified recipients that could be classified as competent were reanalyzed in a new iteration to improve the predictive model. After subsequent iterations, the predictive potential of recipient biomarkers was recalculated. On d 0, creatine, acetone and l-phenylalanine were the most relevant biomarkers at d 40, d 62, and birth, and on d 7, l-glutamine, l-lysine, and ornithine. Creatine was the most representative biomarker within blocks (n = 20), with a uniform distribution over pregnancy endpoints and type of embryos. Biomarkers showed higher abundance on d 7 than d 0, were more predictive for d 40 and d 62 than at birth, and the pregnancy predictive ability was lower with frozen-thawed (F-T) embryos. Six metabolic pathways differed between d 40 pregnant recipients for fresh and F-T embryos. Within F-T embryos, more recipients were misclassified, probably due to pregnancy losses, but were accurately identified when combined with embryonic metabolite signals. After recalculation, 12 biomarkers increased receiver operator characteristic-area under the curve (>0.65) at birth, highlighting creatine (receiver operator characteristic-area under the curve = 0.851), and 5 new biomarkers were identified. Combining metabolic information of recipient and embryos improves the confidence and accuracy of single biomarkers.