Endometrium on-a-chip in the transcriptome and proteomic secretome

The molecular interactions between the maternal environment and the developing embryo that are key for early pregnancy success and are influenced by factors such as maternal metabolic status. Our understanding of the mechanism(s) through which these individual nutritional stressors alter endometrial function and the in utero environment for early pregnancy success is, however, limited. Here we report, for the first time, the use of an endometrium-on-a-chip microfluidics approach to produce a multi-cellular endometrium in vitro . Isolated endometrial cells (epithelial and stromal) from the uteri of non-pregnant cows in the early-luteal phase (Day 4-7), were seeded in the upper chamber of the device (epithelial cells; 4-6 10 4 cells/mL) and stromal cells seeded in the lower chamber (1.5-2 10 4 cells/mL). Exposure of cells to different concentrations of glucose (0.5, 5.0 or 50 mM) or insulin (Vehicle, 1 or 10 ng/mL) were performed at a flow rate of 1µL/min for 72 hr. Quantitative differences in the cellular transcriptome and the secreted proteome of in vitro -derived uterine luminal fluid (ULF) were determined by RNA-sequencing and Tandem Mass Tagging Mass Spectrometry (TMT-MS), respectively. High glucose concentrations altered 21 and 191 protein-coding genes in epithelial and stromal cells, respectively (p<0.05), with a dose-dependent quantitative change in the protein secretome (1 and 23 proteins). Altering insulin concentrations resulted in limited transcriptional changes including transcripts for insulin-like binding proteins that were cell specific but altered the quantitative secretion of 196 proteins. These findings highlight one potential mechanism by which changes to maternal glucose and insulin alter uterine function. and is susceptible to reprogramming events (6). We used RNA sequencing to determine how these metabolites alter the cell-specific transcriptional response in the endometrium to these nutritional stressors. We further demonstrate how these changes alter the proteomic content of in vitro derived ULF secreted from the endometrial epithelium. Collectively, these data highlight one potential mechanism by which changes to maternal glucose and insulin concentrations alter uterine function. We propose that these are candidate proteins that can modify the developmental potential of embryos. μm cell 2 culture UK). After 5-7 days culture, cell populations were further purified using their differential plating times. characterised the cellular populations, cells were in 1x10 6 cells each cell type permeabilised the & PERM as the Primary antibodies were the cells following secondary antibody and incubated following 15 min. Cells were then CytoFLEX of and cellular ( mitochondrial protein-tyrosine kinase ( ATP binding transcripts involved in the biological processes of collagen fibril organization, blood vessel development, regulation of interferon  and  production, positive regulation of MAPK and ERK1/2 and positive regulation of energy homeostasis as well as the molecular functions of ECM structural constituent, platelet-derived growth factor binging. Platelet activation is a complex signaling pathway positively dependent on several components as well as glycoprotein (GP) Ib-IX-V complex (GPIb-IX), phosphoinositide 3-kinase (PI3K-Akt), immunoreceptor tyrosine-based activation motif (ITAM), mitogen-activated protein kinase (MAPK), extracellular signal-regulated kinases 1 and 2 (ERK1/2), among others (49). Interactions between cells such as epithelial and stromal cells of the endometrium, involve mechanisms and ECM constituents, including cell surface receptors (integrins) and receptors for fibronectin, collagen, and laminin. Stimulation of suspended platelets is an event dependent of collagen and thrombin and increase in intracellular Ca 2+ is a key element in this process. Some of these events were connected with the GO terms involved with the DEGs when the endometrial cells were exposed to physiological extremes of glucose. Exposure to different concentrations of glucose not only altered DEGs in endometrial cells, but also altered the protein composition of in vitro ULF. This included proteins involved in Platelet Activation and Lysosome pathways.


INTRODUCTION
Successful establishment of pregnancy in placental mammals requires bilateral interactions between the developing embryo and the maternal endometrium. While direct contact with the maternal environment is not strictly required, i.e. viable embryos can be successfully produced in vitro, interactions with the maternal tract substantially enhances the quality of the embryo (1)(2)(3).
This increased developmental competency is mediated, in part, via the transport and secretion of endometrial-derived molecules (including proteins, amino acids, metabolites, lipids and RNA species encapsulated in extracellular vesicles (EVs)) that are taken up by the embryo and support development prior to establishment of the placenta (4,5). The composition of uterine luminal fluid (ULF) and the molecular interactions between the mother and developing embryo are known to be influenced by maternal factors such as the metabolic status of the mother as well as the quality of the embryo present (reviewed by (6,7)). Exposure to adverse conditions, such as nutritional insults, at specific developmental time points can alter an individual's susceptibility to disease in later life (7). Despite a significant volume of literature describing the composition of ULF (8,9), efforts to supplement culture media with known components have not substantially improved development suggesting there are likely still unknown components of ULF yet to be discovered.
The uterine epithelium, at least for a few key days, is potentially the most critical maternal tissue in the establishment of a healthy pregnancy (10). Thus, exposure of the endometrium to stressors can alter the developmental or epigenetic programming of the foetus. In the dairy cow, the early post-partum period is frequently associated with nutrition-associated metabolic stress as cows cannot take in sufficient dietary energy to off-set the demands of peak milk production. This induces a maternal metabolic environment characterized by high non-esterified fatty acids (NEFA), betahydroxybutyrate (BHB), and low insulin, IGF-I, and glucose (11,12). Given the livestock production cycle, this postpartum altered environment typically occurs at the same time as which the next pregnancy is being established. There is a growing body of evidence to suggest that this metabolic A c c e p t e d M a n u s c r i p t 4 stress compromises the ability of the reproductive tract of the lactating dairy cow to support early development (11,12) associated with alterations in global gene expression in the embryo/conceptus, the oviduct (9,13), and endometrium (13)(14)(15). Conceptuses derived from early in lactation are less developmentally competent (metabolic stress) compared to late stages of lactation (16). Even if a high-quality embryo is produced, exposure to a suboptimal uterine environment such as that in high-producing lactating cows, can compromise developmental potential (16,17).
Our understanding of the mechanism(s) through which individual metabolites alter endometrial function and the in utero environment is relatively limited. While traditional static cell culture models have been used to address the issue of conceptus-maternal interaction, they do have limitations; for example, they do not mimic the dynamic nature of these metabolic components to which the endometrium is exposed. Nor do they allow for assessment of how these metabolic extremes alter the interactions between the heterogenous cells types of the endometrium and the ULF that is produced as a consequence. Advances in microfluidics and organ-on-a-chip technologies in reproductive systems have facilitated the study of embryo development as well as cervical (18), ovarian (19), endometrial (20), and placental function (21,22) in humans and mice. Such systems have been used to mimic the bovine oviduct environment (23) and as well as the menstrual cycle in vitro using a combination of human (fallopian tube, endometrium, ectocervix, liver) and murine (ovary) components (24). However, the power of these systems has not yet been exploited to investigate how maternal nutritional stressors alter the uterine environment to which the embryo is exposed, either at the level of the transcriptomic or proteomic secretome.
Here, we report for the first time the use of microfluidics to produce a multi-cellular endometrium in vitro, which was exposed to glucose and insulin concentrations associated with maternal metabolic stressors. We specifically have focussed on recapitulating days 4-7 of pregnancy when the embryo enters the uterus in vivo, transitions between the morula and blastocyst stages A c c e p t e d M a n u s c r i p t 5 and is susceptible to reprogramming events (6). We used RNA sequencing to determine how these metabolites alter the cell-specific transcriptional response in the endometrium to these nutritional stressors. We further demonstrate how these changes alter the proteomic content of in vitroderived ULF secreted from the endometrial epithelium. Collectively, these data highlight one potential mechanism by which changes to maternal glucose and insulin concentrations alter uterine function. We propose that these are candidate proteins that can modify the developmental potential of embryos.

MATERIALS AND METHODS
Unless otherwise stated, all chemical and reagents were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO). The in vitro experimental procedures were conducted in humidified incubators maintained at 38.5°C with 5% CO 2 in air.

Primary Endometrial Cell Isolation and Culture
Endometrial cell isolation was carried out as previously described (25). Briefly, uteri from non-pregnant cows (Bos taurus), early in the luteal phase (Day 4-7 approximately) were selected on the basis of corpus luteum morphology as previously described (26). This stage of the cycle was chosen as it is when the embryo is present in the uterus and undergoes the transition from morula to blastocyst a key developmental timepoint where is can be susceptible to extremes in the maternal environment. Endometrial tissue was dissected from the underlying myometrium and incubated in 25 mL digest solution containing bovine serum albumin (1 mg/mL, BSA), trypsin EDTA A c c e p t e d M a n u s c r i p t 6 and centrifuged at 700 g for 7 min. The resulting cell pellet containing stromal cells was resuspended in RPMI 1640 culture medium containing 10% FBS, streptomycin (50 μg/mL), and penicillin (50 IU/mL) amphotericin B (2.5 μg/mL). The 40 μm strainer was inverted and flushed with culture medium to recover epithelial cells. The cell populations were seeded at 1×10 5 cells/mL into 75 cm 2 culture flasks (Greiner BioOne, Gloucestershire, UK). After 5-7 days of culture, cell populations were further purified using their differential plating times. To characterised the cellular populations, cells were washed in PBS, and 1x10 6 cells from each cell type fixed and permeabilised using the FIX & PERM kit as per the manufacturers protocol (ThermoFisher Scientific). Primary antibodies were added to the cells following permeabilisation at the concentrations recommended by supplier, incubated for 15 min, washed, secondary antibody added and incubated for a following 15 min. Cells were then immediately analysed on a CytoFLEX S (Beckman Coulter) with appropriate gating to remove clumps of cells and cellular debris. To the epithelial cells specifically, the cells were stained with anti-keratin 18 rabbit IgG primary antibody and anti-rabbit IgG secondary antibody. The 640nm lazer was used to analyse the samples. The stromal cells were stained with anti-vimentin primary IgG and anti-mouse IgG secondary antibody. The 488nm lazer was used on the stromal cell samples ( Figure 1).

Cell Seeding into the Microfluidic Device
All cells were seeded into the devices in RPMI 1640 medium as described above ( Figure 2).
Stromal-enriched cells were seeded in the lower chamber of the device (10 µm-slide membrane, IBIDI), using a 1 mL syringe, at concentration of 1.5-2×10 4 cells/mL in a final volume of 300 µL. All devices were inverted for 2 hr to allow stromal cells to adhere to the underside of the porous membrane. Devices were placed in the normal orientation and epithelial cells seeded at 4-6×10 4 cells/mL in a final volume of 55 µL into the upper chamber. Cells were left to become 60% confluent over two days with one medium change (48 hr) before beginning the flow perfusion. For the glucose experiment, on the day of experimentation medium was changed and 5 mL of medium without

RNA extraction and sequencing
Total RNA was extracted from epithelial and stromal cells using the Mini RNeasy kit (Qiagen, Crawley, UK) following the manufacturer's recommendations. Cell samples were homogenized in 700 µL of Qiazol via vortexing for 1 min at RT. On-column DNase digestion was performed (15 min at room temperature (RT)) and RNA was eluted in 14 µL of RNase/DNase free water from the spin column membrane following centrifugation for 2 min at full speed and this step was performed twice.
RNA sequencing was performed as previously described (25) with minor modifications.
Briefly, RNA quality and quantity were confirmed using the Agilent Bioanalyzer system, and all samples had an RNA integrity number of >7.9. Stranded RNA sequencing libraries were constructed A c c e p t e d M a n u s c r i p t 8 using the Encore Complete RNA-Seq library system of NuGEN. All libraries were sequenced on NextSeq generating 75 bp single-end reads. The raw FASTQ files were inspected using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), the adapter sequences were trimmed using Cutadapt (30) and additional quality control steps taken by fastq_quality_filter program as part of FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). The mapping process was performed using align function in Rsubread (31) package by aligning the clean fastq files against the cow reference genome retrieved from Ensembl release 96 (32) (Bos taurus) and only uniquelymapped alignments were recorded. The resulting BAM files were sorted and indexed by SAMtools (33). The reads were summarized at the gene level by means of featureCounts (34). DESeq2 (35) in a paired sample design was used to identify differentially expressed protein-coding genes based on the cut-offs of a p value < 0.05 and a log 2 Fold Change > 0.1 or < -0.1. Variance stabilizing transformations were applied to the genes that had at least 10 reads in total for all samples.
Heatmaps were created using the transformed read counts based on a pool of differentially expressed protein-coding genes for each experiment. PCA analysis was carried out for the proteincoding genes that have RPKM >= 1 in at least one sample and two normalisation approaches such as log2(RPKM+1) and quantile normalization were conducted prior to PCA analysis.

Quantitative proteomic analysis of in vitro-derived uterine luminal fluid recovered from the upper chamber
Medium (n=3 samples per group) from the upper chamber (in vitro-derived ULF) following glucose or insulin exposure were subjected to albumin depletion according to the manufacturer's instructions (Thermo Fisher Scientific, Loughborough, UK). Individual samples were digested with trypsin (2.5 µg trypsin; 37 °C, overnight), labelled with Tandem Mass Tag (TMT) ten plex reagents according to the manufacturer's protocol (Thermo Fisher Scientific) and pooled. The pooled sample was evaporated to dryness, resuspended in 5% formic acid, and then desalted using a SepPak A c c e p t e d M a n u s c r i p t 9 cartridge according to the manufacturer's instructions (Waters, Milford, Massachusetts, USA). Eluate from the SepPak cartridge was again evaporated to dryness and resuspended in buffer A (20 mM ammonium hydroxide, pH 10) prior to fractionation by high pH reversed-phase chromatography using an Ultimate 3000 liquid chromatography system (Thermo Scientific). In brief, the sample was loaded onto an XBridge BEH C18 Column (130Å, 3.5 µm, 2.1 mm X 150 mm, Waters, UK) in buffer A and peptides eluted with an increasing gradient of buffer B (20 mM Ammonium Hydroxide in acetonitrile, pH 10) from 0-95% over 60 min. The resulting fractions were evaporated to dryness and resuspended in 1% formic acid prior to analysis by nano-LC MSMS using an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific).

Nano-LC Mass Spectrometry
High pH reversed-phase (RP) fractions were further fractionated using an Ultimate 3000 nano-LC system in line with an Orbitrap Fusion Lumos mass spectrometer (Thermo Scientific).
Peptides in 1% (vol/vol) formic acid were injected onto an Acclaim PepMap C18 nano-trap column (Thermo Scientific). After washing with 0.5% (vol/vol) acetonitrile 0.1% (vol/vol) formic acid peptides were resolved on a 250 mm × 75 μm Acclaim PepMap C18 reverse phase analytical column (Thermo Scientific) over a 150 min organic gradient, using 7 gradient segments (1-6% solvent B over 1 min, 6-15% B over 58 min, 15-32% B over 58 min, 32-40% B over 5 min, 40-90% B over 1min, held at 90% B for 6 min and then reduced to 1% B over 1 min) with a flow rate of 300 nl min −1 . Solvent A was 0.1% formic acid and Solvent B was aqueous 80% acetonitrile in 0.1% formic acid. Peptides were ionized by nano-electrospray ionization at 2.0kV using a stainless-steel emitter with an internal diameter of 30 μm (Thermo Scientific) and a capillary temperature of 275 °C. All spectra were acquired using an Orbitrap Fusion Lumos mass spectrometer controlled by Xcalibur 4.1 software (Thermo Scientific) and operated in data-dependent acquisition mode using an SPS-MS3 workflow. Fourier transform mass analysers 1 (FTMS1) spectra were collected at a resolution of 120 000, with an automatic gain A c c e p t e d M a n u s c r i p t 10 control (AGC) target of 200 000 and a max injection time of 50 ms. Precursors were filtered with an intensity threshold of 5000, according to charge state (to include charge states 2-7) and with monoisotopic peak determination set to Peptide. Previously interrogated precursors were excluded using a dynamic window (60 s +/-10 ppm). The MS2 precursors were isolated with a quadrupole isolation window of 0.7 m/z. ITMS2 spectra were collected with an AGC target of 10 000, max injection time of 70ms and CID collision energy of 35%. For Fourier transform mass analysers 3 (FTMS3) analysis, the Orbitrap was operated at 50 000 resolution with an AGC target of 50 000 and a max injection time of 105ms. Precursors were fragmented by high energy collision dissociation (HCD) at a normalised collision energy of 60% to ensure maximal TMT reporter ion yield. Synchronous Precursor Selection (SPS) was enabled to include up to 5 MS2 fragment ions in the FTMS3 scan.
The raw data files were processed and quantified using Proteome Discoverer software v2.1 (Thermo Scientific) and searched against the UniProt Bos taurus database (downloaded June 2019: 46309 entries) using the SEQUEST algorithm. Peptide precursor mass tolerance was set at 10 ppm, and MS/MS tolerance was set at 0.6Da. Search criteria included oxidation of methionine (+15.9949) as a variable modification and carbamidomethylation of cysteine (+57.0214) and the addition of the TMT mass tag (+229.163) to peptide N-termini and lysine as fixed modifications. Searches were performed with full tryptic digestion and a maximum of 2 missed cleavages were allowed. The reverse database search option was enabled, and all data was filtered to satisfy false discovery rate (FDR) of 5%. Differences in protein abundance amongst groups were determined using an unpaired T-test following the FDR filtration step. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD024218. To perform and visualization of the networks of the signalling pathways was used the software platform Cytoscape version 3.7.2 (37). The signalling pathways were presented by the Fold Enrichment each networking and the GO terms were arranged by the Enrichment Score (-log 10 [Pvalue] ), respectively.  Functional annotation analysis determined n = 8 and n = 14 overrepresented biological processes, n = 6 and n = 5 overrepresented cellular components and n = 4 and n = 6 overrepresented molecular functions, for epithelial and stromal cells exposed to 50 mM glucose, respectively ( Figure 4; Table 1).

Exposure to high concentrations of glucose alters the proteomic content of in vitro-derived ULF
PCA ( Figure 5A) revealed that exposure of endometrium on-a-chip to physiological extremes of glucose (0.5 mM and 50 mM) altered the overall composition of proteins in the in vitro-derived ULF. The highest concentration of glucose (50 mM) changed the abundance of 23 proteins compared to controls (5 mM), the majority of which were increased (p < 0.05: Figure 5B). When the lower concentration (0.5 mM) was compared with the control (5 mM), only one protein was altered.
Finally, when the physiologic extremes (0.5 mM vs 50 mM) were compared, eight proteins were found to be differentially abundant in in vitro-derived ULF. Functional annotation analysis revealed the proteins up-regulated following glucose treatment were involved in Platelet and Lysosome pathways ( Table 2). There were no over-represented pathways associated with the proteins that were decreased in abundance following glucose exposure (P>0.05).  Figure   8. In early development, mammalian embryos use glucose as the main energy source to synthesize glycogen, nucleic acids, proteins, and lipids (47,48). It is critical, therefore, that sufficient glucose is transported into the uterine lumen to support embryo development; however, glucose

DISCUSSION
can also act to modify the transcriptional response of a cell which modifies the uterine environment.
We demonstrated that exposure to altered concentrations of glucose changed the expression of Platelet activation plays a critical role in the function of platelets and it is involved in coagulation and inflammatory processes. Normally, platelet activation is induced by collagen or soluble platelet agonists that bind to G protein receptors, which stimulates the activation of platelet receptors (integrin  IIb  3 ), mediating platelet adhesion and aggregation (49). Interestingly, the Platelet-Activating Factor (PAF), a potent lipid mediator of inflammation and allergy, is involved in several reproductive processes (50) and PAF receptors are present in the oviduct of hamsters (51) and mice (52) and the oviduct and endometrium of cows (53). In humans, PFA increases vascular permeability and vasodilation, necessary processes for embryo implantation and plays an important paracrine role in stromal and epithelial cells interactions during this process (50). We have shown A c c e p t e d M a n u s c r i p t 17 components of this pathway are altered by glucose and propose that may contribute to reduced endometrial function associated with altered glucose concentrations.
The other signaling pathway related with the proteins from the in vitro ULF is the Lysosome pathway. The lysosome pathway is the primary site of cell digestion, lysosomes support cell function, recycling and providing a set of metabolites, such as amino acids, saccharides, lipids, ions and nucleobases and a key integrator and organizer of cellular adaptation and survival (54). Lysosomes link important metabolic processes encoded by AMPK (adenosine monophosphate-activated protein kinase) and GSK3β (glycogen synthase kinase 3) signaling hubs. AMPK is a primary cellular sensor for energy stress and glucose levels, promoting catabolic programs in response to low energy levels (55). AMPK has been linked to endometrial cancer in humans and depending on the context, AMPK can promote proliferation or cellular death (56). In addition, GSK3β is a kinase with apparently contradictory functions (57); for example, the presence of GSK3β in lysosomes can stimulate cell growth and survival, while its presence in the nucleus can promote cell death functions (58). In humans, GSK3 is expressed in endometrial cells and GSK-3β phosphorylated form exhibits cyclic variation. Furthermore, phosphorylation of GSK-3β is regulated by progesterone and the inhibition of GSK-3β is temporally regulated with the increased glycogen synthesis in the endometrial cells during the luteal phase (48,59). Thus, incorrect inactivation of GSK-3β could result in inadequate glycogen production and could potentially affect embryo implantation (48). In a previous study with dairy cows, increasing the circulating energy substrate, by exogenous infusion of glucose, was directly associated with a decrease in embryo development (size, width and area) (42), in contrast to what was expected. We propose that physiological glucose extremes may directly affect these important signaling pathways (Platelet activation and Lysosome) and interfering in processes such as receptivity and embryonic implantation.
Insulin plays an essential metabolic role in regulating energy homeostasis in the body and insulin-dependent signaling also has key functions in reproductive events and early development. In interferon-tau and prostaglandin metabolism (73,74). However, to the best of our knowledge, the effects of insulin during the early luteal phase in in vitro bovine endometrial cells had not previously been investigated.
We observed that components of the complement and coagulation cascade were altered when the endometrium on-a-chip was exposed to a low concentration of insulin. In humans, changes in ECM and/or in ECM-related signaling pathways are often attributed to pathological events, including premature birth, cervical incompetence, endometriosis, polycystic ovary syndrome and neoplasms in the reproductive tract (81). Thus, demonstrating that ECM- A c c e p t e d M a n u s c r i p t 20 receptor interactions pathway can be also involved in other tissues and biological events. In ruminant livestock species, it was previously reported that the the characteristic post-hatching elongation of the conceptus requires interaction between the trophectoderm and the uterine luminal epithelium that causes a mosaic of interactions between integrins and ECM, which act together to promote adhesion during implantation (82,83). While our investigations were carried out in the pre-elongation stage of development, insulin-regulated changes to ECM components of the endometrium in early development may contribute to a compromised uterine environment and pregnancy loss. We found limited correlation between the transcriptomics signature of the epithelial cells and the protein content of the secretome. This is not unusual given differences in rates of translation from mRNA into proteins as well as protein turnover and degradation rates (84).
Moreover, there is an order of magnitude in the difference of detection of transcripts via RNAseq and proteins via mass spectrometry likely contributing to the differences observed.
In summary, we are the first to report the use of microfluidics to mimic the bovine endometrium in vitro. We demonstrate that this approach allows us to determine how alterations in individual components of the maternal metabolism impact endometrial function. We specifically demonstrate at the transcriptional and proteomic levels that altered concentrations of glucose and insulin change ECM components of the endometrium. We propose that changes to these ECM components contribute to the compromised uterine environment associated with metabolic extremes in maternal circulation.

DATA AVAILABILITY
The mass spectrometry proteomics data have been deposited to the             Figure 8.