Identication of novel candidate factor SNX5 specic for pregnancy failure in Holstein heifers

Pregnancy loss predominantly occurs during periods between blastocyst hatching and conceptus (embryo plus extraembryonic membranes) implantation to the endometrium in cattle. Insucient biochemical communication between conceptus and endometrium has been suspected as the primary cause for early embryonic losses. If molecules regulating this communication were identied, molecular mechanisms associated with early pregnancy success or loss could be better understood. To identify novel factors as detection markers of non-pregnant or females undergoing embryonic loss, blood sera from embryo-transferred heifers on day 7 (day 0 = day of estrus) were collected on day 17, 20, or 22, which were subjected to metabolome and global proteome iTRAQ analyses. On each sample, the metabolome analysis partly divided serum components into pregnant or not. In the iTRAQ analysis, heatmap analysis with 25 unique proteins separated into pregnant or not on day 20 or 22. Furthermore, receiver operating characteristic curve (ROC) analysis identied ve candidate proteins detecting non-pregnant heifers, of which SNX5 in day 22 sera had the highest area under the curve (AUC), 0.983. We also detected SNX5 in day 22 sera from non-pregnant heifers using western blotting. These results suggest that high SNX5 in day 22 sera could predict early pregnancy loss in heifers.


Introduction
In cattle, pregnancy loss predominantly occurs during the period between blastocyst hatching and conceptus (embryo plus extraembryonic membranes) implantation to the endometrium 1 . Insu cient biochemical communication between conceptus and endometrium has been suspected as the primary cause for early embryonic losses. Numerous global analyses have been conducted to identify key molecules that govern feto-maternal communication, which could aid in establishing an optimal uterine environment for conceptus implantation to the endometrium [2][3][4][5][6][7][8][9] . If such molecules were identi ed, the molecular mechanisms associated with pregnancy success or loss during this period could be elucidated.
In our previous cattle study of the peri-implantation period, we identi ed a series of events required for the progression of conceptus implantation to placentation 10 : lymphocyte homing, the epithelial to mesenchymal transition (EMT), and dynamic changes in the production of interferon tau (IFNT), a cytokine required for the process of maternal recognition of pregnancy in ruminants. In our co-culture system, the treatment of bovine trophoblast cells and endometrial epithelial cells with uterine ushing uids (UFs) obtained from pregnant day 17, 20 or 22 (day 0 = day of estrus, day of embryo implantation is initiated on days 19-19.5) induced lymphocyte homing and EMT of trophoblast cells and in utero gene expression pattern including IFNT on day 17, 20 or 22, respectively 11,12 . This indicates that the proper uterine environment necessary for the establishment of conceptus implantation could be de ned biochemically. Although several global analyses during the peri-implantation period have also demonstrated changes in intrauterine transcripts and protein levels [12][13][14][15] , factors that initiate and/or drive conceptus implantation to the endometrium have not yet been determined.
To discriminate pregnancy losses from cows successfully conceived and carrying pregnancy is pivotal for pro tability in the cattle industry. We propose that earlier detection of pregnancy failure can allow for sooner preparation of the uterus/dam for the next pregnancy opportunity, resulting in the reduction of calving intervals. Several studies have explored to identify a factor speci c for pregnancy establishment, but except pregnancy associated glycoproteins (PAGs), a de nitive candidate has not yet been determined. Previously, IFNT has been considered as a strong candidate for the determination of pregnancy success. Except for inside the sheep uterine vein, IFNT has not been identi ed in the peripheral blood. Instead of IFNT, IFN-stimulated gene products have been examined to see whether one or two of those could be used for pregnancy diagnosis; however, factors speci c to or a method of convenient pregnancy testing have not yet been determined or developed.
Pregnancy establishment requires a series of dynamic physiological events and the conceptus goes through not only its elongation, but also attachment, adhesion to the uterine endometrium, followed by placentation. Because of dynamic changes in ruminants' conceptuses in utero, some changes in transcripts and/or proteins could also be seen in the peripheral blood. We, therefore, hypothesized that the presence or absence of conceptuses in the uterus could be identi ed through changes in the peripheral blood compositions, which could be minor changes in proteins and/or metabolites. To test this hypothesis, blood samples taken from Holstein heifers, which went through embryo transfer (ET) procedure on day 7, and later the status of pregnancy or non-pregnancy (NP) were evaluated. Serum samples from the control heifers, which did not go through ET procedure, were also obtained. All serum samples obtained during estrous cycle and peri-implantation periods were subjected to protein global analysis and metabolome analysis, from which their expression pro les were evaluated.

Comparison of serum composition between pregnant, NP heifers and those with estrous cycle
We next investigated serum metabolic data from NP heifers, which were compared with those from the estrous cycle. First, we compared serum metabolites among cyclic day 0, 7, 14, or 17. The heatmap and PCA showed that estrous cycle data were divided into different cyclic days (Fig. 2a). In the serum metabolite comparison of day 17 pregnant or NP heifers with those of cyclic ones, components on cyclic days were separated from those of day 17 pregnant or NP heifers (Fig. 2b). However, components on day 17 were not divided from those of NP heifers. On days 20 comparison, components of cyclic days were completely separated from those of day 20 pregnant or NP heifers (Fig. 2c). Components on day 20 could be further divided into pregnant and NP groups. Similar to D20, components on cyclic days were separated from those of day 22 pregnant or NP heifers (Fig. 2d). Components on day 22 could also be divided into pregnant and NP heifers.
Identi cation of speci c proteins in day 22 blood sera detecting NP heifers We next investigated the proteome pro le of sera from day 22 pregnant or NP heifers. Like day 20 sera, PCA did not divide protein components on day 22 pregnant from those of NP heifers (Fig. 4a). Heatmap analysis with 25 unique proteins separated serum protein components in NP heifers from those of day 22 samples except for some (Fig. 4b), of which ROC analysis identi ed ve candidate proteins detecting NP heifers (P < 0.05): sorting nexin 5 (SNX5; AUC 0.98), acetyl-CoA carboxylase alpha (ACAC; AUC 0.95), cleavage and polyadenylation speci c factor 6 (CPSF6; AUC 0.95), damage speci c DNA binding protein 1 (DDB1; AUC 0.93), and serum response factor binding protein 1 (SRFBP1) (Fig. 4c). To examine whether these ve proteins could be used as markers to detect NP heifers, day 22 peripheral serum samples were subjected to western blotting analysis. SNX5 in sera from NP heifers was higher than that in the pregnant heifers (Fig. 4d). However, ACAC, CPSF6, DDB1, and SRFBP1 in sera did not differ between day 22 pregnant and NP heifers. Western blots were again executed using another antibody speci c for SNX5, resulting that SNX5 in the NP heifers was higher than that of pregnant heifers on day 22 (Fig. 4e). Moreover, the network analysis with day 22 protein pro les identi ed impacting proteins: component C3 (C3), thyroglobulin (TG), c-type lectin domain family 12 member B (CLEC12B), ankyrin repeat and SOCS box protein 17 (ASB17), and beta-2-glycoprotein1 (APOH) (Fig. 4f).

Discussion
Using PCA and the heatmap with the 25 unique factors from the metabolome analysis, serum components were divided into days 17, 20, and 20 pregnant from those of NP heifers. On each pregnant day sample, serum components were partly divided into pregnant and NP groups. Notably, serum components from cyclic day 0, 7, 14, or 17 were distinctive, which also differed from those of days 17, 20, or 22 NP heifers. These indicate that serum components in NP heifers differ from those of cyclic ones and may dynamically change during the peri-implantation periods. In the proteome iTRAQ analysis, the heatmap with 25 unique proteins mostly separated into pregnant and NP heifers on day 20 or 22.
Furthermore, ROC analysis identi ed ve candidate proteins detecting heifers not pregnant, of which SNX5 in day 22 sera had the highest AUC, 0.983. Using western blotting, we detected SNX5 in day 22 sera from NP heifers. These results indicate that sera on pregnant day 22, 2 ~ 3 days after conceptus attachment to the maternal endometrium is initiated, well re ect the differences between pregnant and NP heifers during the early pregnancy period, and suggest that high levels of SNX5 in peripheral sera on day 22 could predict pregnancy failure or females undergoing embryonic loss in ET heifers.
SNX5 encodes a member of the sorting nexin family. Members of this family contain a phox (PX) domain, a phosphoinositide binding domain, and are involved in intracellular tra cking. SNX5 protein functions in endosomal sorting, the phosphoinositide signaling pathway, macropinocytosis, and micropinocytosis 16,17 . The macropinocytosis and micropinocytosis are considered essential for providing nutrients from a mother to the fetus. In humans, the trophoblast cells possess various transporters for glucose, amino acids, and fatty acid and are believed to preferentially utilize these low molecular-weight nutrients. However, maternal exposure to nutrient insu ciency will cause restricted nutrient supply to the placenta. The macropinocytosis is adapted as an alternative means of nutrient source to allow sustained fetal growth, which is greatly enhanced during amino acid shortage 18, 19 . In addition, it has been reported that SNX5 stimulated autophagy during viral infection 20 . In cattle, amino acids are essential for the survival and development of embryo 14,21 . Furthermore, amino acid concentrations in uterine uids during early pregnancy differ in fertile and subfertile dairy cows 22 . In this study, the metabolome analysis showed that the serum levels of several amino acids differed between pregnant and NP heifers. These observations suggest that the increase in amino acid usages for placental and fetal growth recruit more plasma amino acids to the uterus in pregnant animals, but in the NP heifers, which experienced early embryonic loss, these recruitments do not occur, resulting in transient increase in some amino acids and SNX5 in the plasma.
Pregnancy-associated glycoproteins (PAGs), produced by mono-, bi-or multi-nucleated trophectodermal cells, are released into maternal circulations. PAGs with a high concentration in peripheral blood during early pregnancy are the target factors for pregnancy test, which is the commercially available ELISA detection for domestic ruminants 23 . This detection method using the PAGs antibodies enables cows to discriminate pregnancy or not on day 28 after insemination. PAGs are pepsin-like aspartic proteinases expressed by mononuclear trophoblast and by binuclear trophoblast cells in cattle and sheep [24][25][26] , although PAG function as active proteases is unclear. PAGs are expressed after conceptus implantation and expression increases throughout pregnancy with peak blood concentrations at near parturition 24 .
Recently, in addition to PAGs, several studies have been attempting to identify speci c markers for discriminating success and failure of pregnancy in ruminants 22,27−32 . However, these outcomes have not yet reached practical application. This study identi ed SNX5 as a novel target for discrimination between pregnant and NP heifers. Furthermore, SNX5 is detectable using blood serum on pregnant day 22, indicating that SNX5 could be usable as a pregnancy test prior to the ELISA test with PAG antibodies. Whereas, it has not been determined where SNX5 has been expressed and how SNX5 functions during pregnancy. Further investigation is required to understand the molecular mechanisms by which SNX5 affects the success or failure of pregnancy.
In conclusion, this study shows that the results from global metabolome analysis with blood sera on days 17, 20, and 22 are partly divided into pregnant and NP heifers. The proteome iTRAQ analysis also separated serum protein components between pregnant and NP heifers on day 20 or 22. Furthermore, ROC analysis identi ed ve candidate proteins detecting pregnancy failure, of which SNX5 was detected as the novel target for discrimination between pregnant and NP heifers. These results indicate that peripheral blood on pregnant day 22 can be used for a pregnancy test, and suggest that high levels of serum SNX5 on day 22 are predictive of females undergoing embryonic loss in heifers and possibly cows.

Ethics statement
All animal procedures in this study were performed in accordance with the guidelines of the Committee for Experimental Animals at Zen-noh Embryo Transfer Center (Hokkaido, Japan), with the approval of the Institutional Animal Care and Use Committee of Zen-noh Embryo Transfer Center (Approval number: ZET20190628). All animals used were raised and kept at this center throughout the course of this experimentation. This study was carried out in compliance with the ARRIVE guidelines.

Collection of bovine blood samples
Estrous synchronization, superovulation, and embryo transfer (ET) processes were performed as previously described 33 . Recipient heifers (14~16 months old, n = 20) were given a single injection of 0.75 mg cloprostenol to synchronize their estrous cycles and the heifers exhibited behavioral estrus 40-48 h after the injection (n = 17, day 0 = day of estrus). For ET processes, day 7 embryos were collected from super-ovulated and arti cially inseminated (AI) Japanese black cows (3~5 years old, n = 5). Two blastocysts each derived from the superovulation/AI procedure were then transferred non-surgically into the uterine horn of Holstein heifers, ipsilateral to the corpus luteum, on day 7 of the estrous cycle. Blood samples were collected on pregnant days 17, 20, and 22. In addition, blood samples from a group of heifers (14~16 months old, n = 3), which went through the same estrous synchronization without ET procedure, were collected on days 0, 7, 14, and 17 of the estrous cycle. The diagnosis of pregnancy or non-pregnancy (NP), indicative of embryonic loss, was done by real-time B-mode ultrasonography (Convex scanner HS-1500, Honda electronics Co. Ltd., Toyohashi, Japan) on days 30 of gestation, from which blood samples were divided into pregnant or NP group. After sorting of blood samples, albumin was removed from serum samples by ProMax Albumin Removal Kit (Polysciences, Warrington, PA, USA).

iTRAQ analysis
A global analysis of proteins using iTRAQ analysis was performed as described previously 12 . Brie y, serum samples from ET-heifers on day 17, 20 or 22 and those without ET were resuspended in 30 µl iTRAQ lysis buffer (50 mM TAEB, 0.1 % SDS). Total protein (100 µg) was subjected to trypsin digestion and then reacted with appropriate iTRAQ reagent according to the manufacturer's instructions. Sample fractionation was performed with an Agilent 3100 OFFGEL Fractionator (Agilent Technologies, Santa Clara, CA, USA). Furthermore, mass spectrometry analysis was performed with a Thermo Scienti c LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scienti c, Waltham, MA, USA). Mascot software was used to simultaneously identify and quantify proteins.

Metabolome analysis
A serum metabolomics analysis was performed using GC/MS as described previously 35 with some modi cations. In brief, a sample of 50 μl of serum was mixed with 5 μl of 1 mg/ml 2-isopropylmalic acid (Sigma-Aldrich) in distilled water as an internal standard, and 250 μl of methanol-chloroform-water (2.5:1:1) mixture. Then samples were lyophilized, and added with 40 μl of 20 mg/ml methoxyamine hydrochloride (Sigma-Aldrich), dissolved in pyridine for oximation. After mixing, the samples were shaken for 90 min at 30°C. Next 20 μl of N-methyl N-trimethylsilyl-tri uoroacetamide (GL Science, Tokyo, Japan) was added for trimethylsilylation, and the mixture was incubated at 37°C for 45 min. The sample was subjected to GC/MS (GCMS QP2010-Ultra; Shimadzu, Kyoto, Japan). The Shimadzu Smart Metabolites Database (Shimadzu) was used to identify metabolites. Samples were normalized by a pooled sample from control group. A metabolic pathway analysis was performed using MetaboAnalyst 36 . Metabolites that signi cantly diffed between two groups were subjected to an enrichment analysis (http://www.metaboanalyst.ca/faces/upload/EnrichUploadView.xhtml).

Statistical analysis
All experimental data represent the results obtained from three or more independent experiments each with triplicate assays. Data were expressed as the mean ± SEM. A P-value < 0.05 was considered statistically signi cant.

Declaration of competing interest
The authors declare no con icts of interest.