Mammary Gland Transcriptome and Proteome Modifications by Nutrient Restriction in Early Lactation Holstein Cows Challenged with Intra-Mammary Lipopolysaccharide

The objective is to study the effects of nutrient restrictions, which induce a metabolic imbalance on the inflammatory response of the mammary gland in early lactation cows. The aim is to decipher the molecular mechanisms involved, by comparing a control, with a restriction group, a transcriptome and proteome, after an intra-mammary lipopolysaccharide challenge. Multi-parous cows were either allowed ad libitum intake of a lactation diet (n = 8), or a ration containing low nutrient density (n = 8; 48% barley straw and dry matter basis) for four days starting at 24 ± 3 days in milk. Three days after the initiation of their treatments, one healthy rear mammary quarter of 12 lactating cows was challenged with 50 µg of lipopolysaccharide (LPS). Transcriptomic and proteomic analyses were performed on mammary biopsies obtained 24 h after the LPS challenge, using bovine 44K microarrays, and nano-LC-MS/MS, respectively. Restriction-induced deficits in energy, led to a marked negative energy balance (41 versus 97 ± 15% of Net Energy for Lactation (NEL) requirements) and metabolic imbalance. A microarray analyses identified 25 differentially expressed genes in response to restriction, suggesting that restriction had modified mammary metabolism, specifically β-oxidation process. Proteomic analyses identified 53 differentially expressed proteins, which suggests that the modification of protein synthesis from mRNA splicing to folding. Under-nutrition influenced mammary gland expression of the genes involved in metabolism, thereby increasing β-oxidation and altering protein synthesis, which may affect the response to inflammation.


Introduction
Milk is synthesized in mammary glands (MG) involving a large number of genes, the expression of which is modulated at a nutritional level [1] and by the health status [2]. Mastitis is the inflammatory response of the mammary gland to pathogens. This pathology is one of the most prevalent disease and has considerable economic impact due to decreased milk production, discarded milk, cost of body weight (BW), plasma metabolite, and insulin concentrations. Feeding the ration containing 48% of straw (restricted group: REST) induced an immediate depression of dry matter intake (DMI) and decreased the energy balance from 5.2 ± 8.9 to −67.2 ± 18.9 MJ/day one day before (corresponding to day 23), and on the last day of restriction (day 27), respectively. The plasma concentrations of glucose and insulin decreased, whereas non-esterified fatty acids (NEFA) and beta-hydroxy butyrate (BHBA) increased dramatically in REST (Supplementary Table S1). The 96 h of nutrient restriction decreased milk yield from 37.9 to 22.4 kg/day (p < 0.001) and milk protein yield from 1.12 to 0.62 kg/day (p < 0.005) in REST. Milk fat percentage increased during feed restriction in REST from 4 to 5.5% (p < 0.001) and returned to pre-restriction concentrations on the same day of re-feeding a regular diet. In CONT cows, milk fat content was greatest only during the 48 h following LPS injection (p < 0.05) compared with all other DIM. These variables were unchanged in the control (CONT) cows [25]. Within 2 to 6 h after injection with lipopolysaccharide, we noticed the edema of the challenged quarter of all cows, an increase in rectal temperature up to 39.5 • C (temperature increment was +2.1 ± 0.15 • C). The effect of inflammation also was confirmed by milk somatic cell count (SCC). The day before the LPS challenge, whole udder composite milk (from PM and AM milking) SCC was 78 000/mL and 92,000/mL (p = 0.51) in CONT, and in REST, respectively. The SCC response was greater in REST compared to CONT cows (6919 versus 1956 × 1000 per mL, respectively) in composite milk samples from the two milkings that followed LPS injection. SCC returned to pre-LPS counts, within less than 7 days post-LPS challenge and biopsy. Moreover, quarter milk IL-8, IL1-β, TNF-alpha, and CXCL3 at time zero (before LPS challenge) did not differ between CONT and REST, but their concentrations increased in response to LPS in both groups. This data shows that these indicators of mammary inflammation did not differ between CONT and REST before or after the challenge.

Effects of Under-nutrition on Expression of Genes Involved in Inflammation Response by RT-qPCR Analysis
RT-qPCR analysis was performed to quantify candidate genes (CCL5, LAP, RBP4, IL8, IL1, STAT3, CD36, and TAP) chosen on the basis of their implication in inflammation [26] Expression of these genes in mammary gland (MG) did not differ between REST and CONT (p ≥ 0.1) 24 h after the LPS challenge, except for the defensin Tracheal Antimicrobial Peptide (TAP) gene which tended to decrease in REST (p = 0.07). The expression of INSIG1, and CSN2 genes, which are involved in the biosynthesis of milk components and linked to MG metabolism did not differ between CONT and REST ( Figure 1). Figure 1. Effects of nutrient restriction and intra-mammary lipopolysaccharide (LPS) challenge on gene mRNA expression quantified by RT-qPCR and presented as ∆ CT . Comparison of the gene expression does not show a difference between control (CONT; n = 6) and restricted (REST; n = 6) Holstein cows (p ≥ 0.1). The expression of the TAP gene tended to differ (p = 0.07). UXT2, CLN3, and EIF3K were used as housekeeping genes.

Microarray Analysis
Mammary gene expression analyzed by a microarray assay allowed the identification of 33 differentially expressed genes, including 25 known genes (corrected p adj < 0.05), between CONT and REST, 24 h after the inflammatory challenge by LPS (Table 1). The expression increased for 19 and decreased for 6 genes in REST compared with CONT. All these DEG presented a fold change (FC) greater than 1.4, with two genes (PDK4 and SLC25A34) presenting an FC greater than 4. Gene ontology and function analyses revealed that most DEG are involved in metabolism, including the regulation of fatty acid (FA) oxidation, glucose, and protein metabolism, and in immune responses ( Figure 2). The results obtained, using Pathway Studio ® software, were consistent with those from Panther software. We focused on the most represented functions, in particular, those involved in metabolism and immune response. We identified DEG in FA and glucose metabolism (CPT1A, PDK4, PFKFB4), carnitine shuttle (SLC25A20, CPT1A, SLC25A34), regulation of cellular ketone metabolic process (PDK4), and the key genes in those processes. A number of genes involved autophagy (PFKFB4, DNNED) and immune function (PGLYRP3, KLF13, PLEKHA2, WC7, TRIB2, CXCR7, and MBP) processes also were altered.

Discussion
This study assesses the effects of undernutrition and the resulting metabolic imbalance on the mammary gland (MG) inflammatory response in early lactation cows using a nutrigenomic approach. The effects of dietary treatments are confirmed by decreased intake, milk yield, energy balance in underfed (REST) cows, and by changes in blood metabolite and insulin concentrations (an increase in plasma NEFA and BHBA and a decrease in insulin and glucose concentrations; Supplementary  Table S1; [25]). The inflammatory response to intra-mammary lipopolysaccharide is confirmed by clinical parameters, such as milk SCC, rectal temperature, and other classical clinical symptoms [25]. The effects of nutrient restriction and metabolic imbalance on the inflammatory response, at the RNA and protein levels, were evaluated using transcriptomic and proteomic analyses. These analyses were performed using MG samples were obtained by biopsies performed 24 h after the LPS challenge. We performed a single biopsy to avoid the potential interference of repetitive biopsies on the inflammatory response. Also, the adjacent quarter may not constitute a good control for the LPS challenged quarter, as inflammation cytokines may exert local effects and influence adjacent quarters [19]. However, one limitation of this design is that the present experimental design does not allow a kinetic data to follow the establishment of inflammation.

Gene Expression Changes at mRNA Level
The study at the mRNA level was performed using two complementary approaches. The first one was a targeted approach to focus our attention on the inflammatory response, then a global analysis was performed using microarray study. We investigated the effects on candidate genes by RT-qPCR, the majority, which are involved in the immune response of interest for inflammation. The expression of candidate genes did not differ between REST and with the control (CONT) group ( Figure 1). This result suggests that the expression of genes considered important in the inflammatory response [11,[26][27][28] are not altered by the nutrient restriction in MG 24 h after LPS administration. This experimental design does not allow the ability to evaluate the modification of the expression of these genes during early inflammatory response. A tendency for decreased expression of the TAP gene in REST was observed. The product of TAP gene is a member of the family of small cationic peptides that have widespread antimicrobial activity; TAP is expressed by bovine mammary epithelial cells [29] and has a broad-spectrum activity against different strains of bacteria, including E. coli [30]. The upregulation of TAP gene expression in REST may constitute a protective mechanism against pathogens.
To complete the candidate gene analyses, a global gene expression approach, using a bovine microarray was used to assess the molecular mechanisms underlying metabolic and inflammatory MG responses, potentially affected by undernutrition and negative energy balance. Transcriptomic analysis revealed 33 differentially expressed genes in MG 24 h after LPS challenge in REST compared to CONT. The number of DEG detected in our study are small compared to the research assessing the effects of inflammation on MG transcriptome [18,19,31]. This study does not compare normal versus inflamed MG, rather it aims to evaluate the effects of undernutrition during the inflammatory response, and therefore both REST and CONT were challenged with LPS. The DEG were classified in six functional classes. In this discussion, we mainly focus our attention on genes that play a role in the metabolic processes (FA, glucose and protein metabolism) and on those involved in immune function.

DEG Involved in Metabolism
The classification of genes by bioinformatics analyses indicate that metabolic process (FA oxidation, glucose, and protein metabolisms) is the class most altered by undernutrition after an LPS challenge conditions. Among the genes presenting the highest fold change (between 2.7 and 6.8) are PDK4, CPT1, and SLC25A34, which are involved in glucose and FA metabolism. PDK4 plays a key role by inhibiting the pyruvate dehydrogenase complex. This inhibition prevents the formation of acetyl-coenzyme A from pyruvate [32], resulting in an expected decrease in glucose and an increase in fat utilization in response to prolonged undernutrition [33]. The upregulation of PDK4 was also found in leucocytes of underfed ewes, however, its expression was downregulated during an intra-mammary inflammatory challenge [34]. The large increase of PDK4 expression in REST is in agreement with decreased insulinemia and with the upregulation of ESRRA. The gene ESRRA upregulates PDK4 expression [35]. Both genes spare glucose and promote FA β-oxidation, therefore, their upregulation in REST would allow MG to shift the metabolic pathways from glycolysis to β-oxidation. The upregulation of CPT1 gene expression would also promote the β-oxidation [36,37], since it is a rate-limiting step of FA entry in mitochondria [38]. This is in line with the increased expression of the CPT1 gene observed in whole blood transcriptome of underfed dairy sheep [34]. The increase in β-oxidation is further supported by an upregulation of SLC25A20 and SLC25A34 in REST, which are two members of SLC25 mitochondrial carrier family. SLC25A20 transports carnitine and carnitine-FA complexes across the inner mitochondrial membrane. SLC25A34 is supposed to act in a similar way, but its exact function still is not known totally [39]. MG seems to spare glucose (downregulating glycolysis) and promote FAs as an energy source (upregulating β-oxidation; Figure 4) in order to adapt to underfeeding. The mammary expression of genes involved in lipid metabolism is also modified in comparison with NEB (induced by caloric restriction) and the positive energy balance of cows after the peak in lactation [40]. Interestingly, the downregulation of genes linked to fat metabolism (FA biosynthesis) is observed 24 h after E. coli infection in MG of lactating cows [18]. This suggests that inflammation downregulates the FA biosynthesis and may increase the use of preformed FA, derived from other sources, such as from the mobilization of adipose tissue. These results suggest that energy metabolism modifications, in response to inflammation, are more marked in REST than in CONT, probably due to the expected limited availability of nutrients to support an acute inflammation in REST.

DEG Involved in Immune Response
The objective of this study was to identify the effects of an aggravated NEB on MG transcriptomic responses to inflammation. The experiment was not designed to compare gene expression in normal versus LPS challenged MG, which have been previously reported [2,11,18,31]. Microarray analysis shows that, the immune response together with the metabolic processes, are the main biological processes modified by undernutrition after LPS challenge (Figure 2), with modifications of genes different from those selected for RT-qPCR analyses. However, immune response was not the main biological process to be affected by restriction during inflammation. Among the upregulated genes is PGLYRP3, which belongs to a family of innate immunity pattern recognition molecules that are activated by LPS and bactericidal and bacteriostatic properties and are activated by LPS [41]. When the invading bacteria survive, neutrophil infiltration is replaced with T and B lymphocytes and monocytes [42]. The upregulation in REST of KLF13, PLEKHA2, WC7, and MBP, involved in the immune response by activating T and B lymphocytes, therefore, is in line with the expect recruitment of leukocytes by MG ( Figure 5). Additionally, the upregulation of PLEKHA2 in REST, a gene involved in the cell adhesion process [43], suggests that there is an increased migration of B leucocytes. Together, the upregulation of these genes suggests a different nature or a higher response to LPS stimuli in REST compared with CONT. Nevertheless, the deregulation of TRIB2 and CXCR7 suggests that the immune process could be impaired. Indeed, both genes participate in the activation of immune cells and influence IL-8 production, a chemokine upregulated in response to infection [10], where the concentration increased in milk, within 4 h after LPS infusion [ (25], and within 16-24 h after experimental infection with different strains of E.coli or LPS infusion [44,45]. The upregulation of TRIB2 and the downregulation of CXCR7 genes, however, suggests a potential IL-8 production alteration in response to inflammation in REST. These results contrast with the expected inflammatory response and could be a sign of deficient immune function under exacerbated NEB. Taken together, differences in gene expression might suggest a modified resolution of inflammation in response to LPS, due to aggravated NEB. During the course of an experiment with LPS, the inflammatory response usually declines within 24 h [44]. The REST cows might have experienced difficulties in restoring the MG homeostasis by 24 h after LPS challenge, due to the metabolic changes inherent in nutrient deficiency. However, this conception needs more detailed investigation, with a kinetic analysis, to be confirmed.

Proteins Involved in Protein Synthesis
Among the 53 differentially expressed proteins (DEP) in REST compared to CONT, 43 were downregulated. Most of these are involved in RNA and protein metabolism, with roles that vary from RNA splicing to translation. The downregulation of proteins involved in the splicing process, such as HNRNPH1, HNRPC, HNRNPA3, PCBP2, YBX1, SNRPA1, and DHX9, suggests that splicing is impaired in the MG of REST cows. This could explain in part the reduced synthesis and secretion of milk protein in REST compared with CONT [25]. Moreover, altered splicing and translation mechanisms might have a profound influence on protein biochemical properties and, ultimately, alter immune response to pathogens. Among the four proteins belonging to the HRNPs family (HNRNPH1, HNRPC, HNRNPA3, PCBP2), the first three are RNA binding proteins associated with pre-mRNAs in the nucleus, influencing pre-mRNA processing as well as other aspects of mRNA metabolism and transport. The dysfunction of HRNPs is linked to different proliferative and degenerative diseases [46], but the role of these proteins in the inflammatory response is still not fully understood. Some members of this family are reported to ensure resolution of inflammation [47]. However, the role of HRNPs could be associated with a mammary remodeling due to the restriction. In our study, YBX1 and DHX9 are downregulated in REST and the downregulation of these two genes is associated with impaired inflammatory responses [48]. Additionally, the loss of PP2AC function causes severe immunological disorders in Treg cells [49]. Thus, the downregulation of all these proteins in REST ( Figure 6) suggest a modified inflammatory response in underfed early lactation cows. A number of proteins involved in translation are downregulated in REST ( Figure 6; RPS27A, RPS15, RPS2, RPL10, EIF3H, RPN2, RPN1, FARSB, and NACA). Four riboprotein family members (3 RPS and 1 RPL) are part of a ribosome. Interestingly, protein building ribosomes alone are shown to affect the other cell processes outside the ribosome like development, apoptosis, and aging during their altered expression levels [50]. Additionally, the decrease of RPN1 and RPN2, which catalyze co-translational N-glycosylation, suggests an impaired post-translational protein modification process in the REST group. This process may play an important role in the immune system by creating the glycans on an immune cell s surface that helps migration of the cell or by glycosylating the various immunoglobulins [51]. Elsewhere, it is reported that this process can be defective during glucose deficit, leading to a reduction of protein glycosylation and harmful accumulation of unfolded proteins [52]. The decrease in RPN1 and RPN2, observed in the current study, could potentially lead to the creation of misfolded proteins in MG of REST cows.
Additionally, protein folding and its control might be modified in REST, due to the downregulation of chaperone proteins such as PDIA3, PDIA4, and CCT4 ( Figure 6). PDIA3 and PDIA4 are part of a larger super-family of a disulfide isomerase family of endoplasmic reticulum proteins that catalyze protein folding [53]. PDIA3 contributes to the correct folding of glycoproteins [54]. The loss of PDI activity and the consequent accumulation of misfolded proteins are associated with chronic inflammation [54,55]. Moreover, PDIA3 is a structural component required for the stable assembly of the peptide-loading complex of the major histocompatibility complex class I pathway. Its activity seems to play a role in lymphocyte T and B function [56]. Added to its role in folding, PDIA4 promotes the immunoglobulin G intermolecular disulfide bonding and antibody assembly in vitro [57]. Because CCT4 assists in the folding of newly translated polypeptides, this function might have been altered in REST [58]. Overall, proteomic data strongly suggest that protein synthesis is impaired by undernutrition at different levels (translation, folding, and post-translation modifications). The modifications of protein metabolism might partially explain the lower milk protein yield from 1.12 to 0.62 kg/ observed during restriction.

DEP Involved in Immune Response
Undernutrition downregulated FARSB protein expression. The decrease of this protein is linked with impaired acute inflammation responses in mice [59], suggesting an impairment in immune system function. In contrast, there was an upregulation of proteins such as SERPINA3, SERPINA3-5, F1MLW8, and Q1RMN8. SERPINA is an acute-phase protein, whose concentration can rise during acute and chronic inflammation [60]. F1MLW8 and Q1RMN8 proteins are similar to immunoglobulin lambda and typical for B-cells, and are important for its maturation from pre-B cells to mature ones [61]. The increase of these four proteins in REST MG, suggests that the resolution inflammation process was delayed in REST, compared to CONT 24 h after LPS challenge, whereas it could be considered that, potentially, it has already resolved in CONT. This is in line with the reported peak in SCC at 12 h that declines 24 h after LPS challenge [44].
The decreased translation process and post-translational protein modification (folding and glycosylation), that is observed at the protein level, might alter protein synthesis and activate an unfolded protein response [52]. This role could also be suggested to affect the proteins involved in the immune response.

Ethics Statement, Treatments and Sampling
The cows were housed at the Herbivore Research Unit of INRA Research Center of Auvergne-Rhone-Alpes. Animal procedures were performed in compliance with Regional Animal Care Committee guidelines CEMEAA: Auvergne, French Ministry of Agriculture and European Union guidelines for animal research C2EA-02. All procedures were approved by the regional ethics committee on animal experimentation (APAFIS #2018062913565518). The animals were in their second to the fourth days of lactation, with a body condition score (BCS) of 2.0 to 2.2 (0 to 5 scale), a week before feed restricted diet. All animals were observed for uterine disease and did not present any signs of abnormality. Additionally, the health history of each animal was inspected and only those without any health problems, within the last 6 months before calving, were chosen. At 24 ± 3 days in milk, sixteen multiparous Holstein cows were allowed ad libitum intake of a lactation diet CTRL, n = 8, 7.1 MJ/kg DM NEL, 17.4% Crude Protein. Their diet constituted of corn (24.2% dry matter), corn silage (29%), grass silage (25.5%), soybean meal (16.9%), and complemented with vitamins and minerals (0.9%). The underfed (REST) group received a ration diluted with barley straw (48% DM) for 96 h (RES, n = 8; 5.16 MJ/kg DM NEL, 12.2% CP). Therefore, the ratio of forage to concentrate differed from 58.0/42.0 in control (CONT) group to 79.2/20.8 in REST group [25]. Dry matter intake, milk yield, energy balance, plasma insulin, glucose, non-esterified fatty acids (NEFA) and BHB (β-hydroxybutyrate) concentrations did not differ between CONT and REST immediately before underfeeding (21.8 kg/day, 39.0 kg/day, -5.6 MJ/day, 22 µIU/mL, 3.78, 0.415, and 0.66 mM, respectively, at day -1), but were significantly altered in REST at 72 h of underfeeding (Supplementary Table S1). Following 72 h of restriction or control diet, one healthy rear mammary quarter was injected with 50 µg of lipopolysaccharide E. coli 0111:B4; (LPS-EB Ultrapure, InvivoGen, San Diego, CA, USA) diluted in 10 mL of sterile saline (CDM Lavoisier, Paris, France), containing 0.5 mg/mL BSA cell culture grade, endotoxin free, A9576, (Sigma-Aldrich, St. Louis, MO, USA), using a sterile disposable syringe fitted with a sterile teat cannula. Mammary biopsies were performed 24 h after the LPS injection, as previously described [62], corresponding to 96 h of feed restriction or control diet. Tissue samples were immediately frozen in liquid nitrogen and stored at −80 • C prior to RNA and protein analyses.
Throughout the study, milk samples were collected at 4 consecutive milkings each week before the beginning of the restriction and just before the LPS challenge and analyzed for SCC. Only healthy cows were included in the study. Additionally cows were screened for mastitis, one week before and immediately before the LPS challenge, using the California Mastitis Test (Neodis, Rambouillet, France) for all quarters, and somatic cell counts of rear quarter milk samples (Galilait, Theix, 63122 Saint Genès-Champanelle, France), one week before and immediately prior to the LPS challenge. Only cows with SCC lower than 100,000 cells/mL, in a rear quarter, were included in the study. Indeed, cows were considered healthy if the quarter SCC was inferior to 100,000 cells/mL and were free of any other signs of health problems [25]. Additionally, foremilk samples were collected from the LPS challenged quarters, immediately before the morning milking that preceded the LPS injection (time 0), and at 4, 6, 10, and 24 h after LPS injection. These quarter milk samples were analyzed for IL-8, IL1-β, TNF-alpha, and CXCL3 using Elisa [25].

RNA Preparation and Analyses
RNA and protein extractions were performed from the same mammary biopsy samples n = 16 animals, (8 CONT and 8 REST). The total RNA was extracted from 50 mg of the mammary gland (MG) by using the mirVana miRNA Isolation Kit (Thermo Fisher Sciences, Waltham, MA USA). The concentration and purity of RNA were estimated by spectrophotometry NanodropTH, (ND-1000, NanoDrop Technologies LLC, Wilmington, DE, USA), and by using the Bioanalyzer 2100 (Agilent Technologies Inc., Santa Clara, CA, USA), respectively. Once these validation steps were completed, only 12 cows (6 RES and 6 CTR) were kept for gene expression analyses at mRNA level, which presented a good and uniform quality of the samples for a microarray experiment.

RT-PCR Analyses
Reverse transcription (RT) was performed on 2 µg of total RNA using the "High Capacity RNA to cDNA" kit and following the manufacturer's recommendations (Applied Biosystems, Villebon Sur Yvette, France) in a final volume of 20 µL. In parallel, negative controls were performed without the matrices. The primers are described in Table 3. The genes, UXT2, CLN3, and EIF3K were used as housekeeping genes [63]. Real-time quantitative PCR was performed on the StepOnePlus™ PCR System (Applied Biosystems, Villebon Sur Yvette, France), using 5 µL of 50 fold-diluted single-stranded cDNA and the TFPower SYBRGreen PCR Master Mix, according to the manufacturer's instructions (Applied Biosystems, Villebon Sur Yvette, France). Subsequent to an initial denaturing step (95 • C for 10 min), the PCR mixture was subjected to the following two-step cycle, which was repeated 40 times: Denaturing for 15 s at 95 • C and annealing and extension for 45 s at 60 or 62 • C. The results were expressed as a fold change of Ct values relative to the control using the ∆ Ct method [64]. The significance was determined using a t-test with p < 0.05 considered as significant.

Microarray Analyses
Microarray analyses were performed on twelve animals (6 RES and 6 CTRL) using 100 ng of total RNA from each MG sample. The total RNA was amplified, fluorescently labeled, and hybridized to the bovine 4 × 44K microarray (Agilent Technologies, Inc. Santa Clara, CA, USA), and all the procedures described below were performed according to the manufacturer instructions (Agilent Technologies, Inc. Santa Clara, CA, USA). Briefly, for each hybridization, the total RNA was linearly amplified and labeled with Cy3 using the one-color Low Input Quick Amp Labeling Kit. Then, 1650 ng of Cy3-labeled cRNA was hybridized on the microarrays using the Gene Expression Hyb Kit. Hybridization was performed for 17 h at 65 • C in a rotating hybridization oven at 10 rpm. Following hybridization, all microarrays were washed and scanned using the Agilent Microarray Scanner G2565A. The resulting TIFF-images (Tagged Image File Format) were processed using Feature Extraction software Version 11 to obtain normalized data. Normalized with 75th percentile shift, the data were analyzed using GeneSpring software. The moderated t-test with Westfall-Young familywise error rate (FWER) correction was applied [70]. The differences were considered significant at an adjusted p < 0.05. The data were accessible through the GEO series accession number GSE114975. The classification and functional analyses of differentially expressed genes were performed using PANTHER [71] and confirmed using Pathway Studio ® software (Elsevier, The Netherlands).

Protein Preparation and Analyses
The proteins were extracted by homogenizing 80 mg MG tissue (n = 16; 8 RES and 8 CTRL) in 2 mL lysis buffer (8.3 M urea, 2 M thiourea, 2% CHAPS, 1% DDT). Following homogenization, the samples were incubated for 5 min at room temperature and centrifuged at 10,000× g for 30 min at 8 • C. The protein concentrations were measured in supernatant with Quick Start Bradford protein assay (BioRad, Marnes-La-Coquette, France), aliquoted and then stored at −20 • C, until further preparation. Sample supernatants were mixed with 1 volume of Laemmli buffer and heated at 60 • C for 5 min. Separation, by SDS-PAGE (12% acrylamide), was performed using a Mini-Protean II electrophoresis unit (BioRad, Marnes-La-Coquette, France) and 100 µg protein loaded per lane. To concentrate the samples, the gels were run at 80 V until the dye front reached the bottom of the concentration gel. Gels were stained overnight in Coomassie brilliant blue G-250. Excised lanes were reduced and alkyled before de-staining in 25 mM ammonium bicarbonate with acetonitrile (50/50 v/v). Following dehydration with 100% acetonitrile, gel pieces were dried in a Speed Vacuum and the samples were preserved at −20 • C until LC MS/MS analysis.

LC MS/MS Analysis
The proteins were hydrolyzed overnight at 37 • C, using 800 ng (80 µL) of sequence grade-modified trypsin (Promega, France) per band. Subsequent to extraction by 64 µL of acetonitrile 100% and sonication, the peptides were concentrated in a Speed Vacuum and volume was adjusted to 30 µL with an aqueous solution (99.9% H2O, 0.1% TFA). Peptide mixture (2.5 µL) was injected into the nano HPLC Ultimate 3000, (Thermo Fisher Scientific, Courtaboeuf, France) after a preliminary step of desalting and concentration in the pre-column 300 µm × 5 mm, (ThermoFisher, Courtaboeuf, France) for 6 min, and a second step of separation in an analytical C18 column 75 µm, 25 cm, (Pepmap Thermo Fisher Scientific, Courtaboeuf, France) with a 10-40% gradient (A: 0.1% FA in water, B: 0.1% FA in acetonitrile) at 450 nL/min. The eluate was electrosprayed through the CaptiveSpray ion source into the mass spectrometer QTOF impact II (Bruker, Wissembourg, France) operated in CID Data Dependent mode. Each MS analysis was succeeded by as many MSMS analysis as possible within 3 s.

Protein Identification and Label-Free Quantitation
The raw files were loaded, at the end of each LC-MS/MS analysis, into the Progenesis QI software Non-linear Dynamics, v 4.1 (Newcastle upon Tyne, UK) and label-free quantitation was performed using a proprietary workflow alignment, peak picking, normalization, design set up, quantitation, and protein identification.
Regarding protein identification (Supplementary File S2), the Mascot V.2.5, internally licensed version (www.matrixscience.com) was used with uniprot-ref_Bos taurus database 19.840 sequences (07/2015). The following parameters were considered for the searches: Peptide mass tolerance was set to 10 ppm; fragment mass tolerance was set to 0.05 Da and a maximum of two missed cleavages was allowed. Variable modifications were methionine oxidation (M), carbamidomethylation (C) of cysteine and Deamidated (NQ). Protein identification was considered valid, if at least two peptides with a statistically significant Mascot score were assigned, with a false discovery rate (FDR) less than 1%.
Concerning label-free quantitation, all unique validated peptides of an identified protein were included, and the total cumulative abundance was calculated by summing up the abundances of all unique peptides allocated to the respective protein. Statistical analysis was performed, using the "between subject design," and the p-values were calculated by an analysis of variance, using the normalized abundances across all runs. Differential proteins were conserved for interpretation if the peptides' individual abundances showed a good correlation with protein abundance. All differential proteins were inspected manually with these correlation criteria. To extract the maximum biological information of differentially expressed proteins, PANTHER [71], Pathway Studio ® software (Elsevier, The Netherlands) and UniProt [72], were used.

Conclusion
Undernutrition affected multiple aspects of MG function, as demonstrated by modifications of milk secretion, and MG mRNA and protein expression. During this study, expression analyses were performed 24 h post-LPS challenge corresponding to the period of inflammation resolution. The effects of undernutrition on studied candidate genes, known as major genes relating to the innate immune responses, were weak. Therefore, the transcriptomic and proteomic analyses pointed out modifications of energy metabolism (fatty acid and glucose), and protein metabolism (synthesis and post-translational modification), respectively, but relatively few genes involved in immune response were affected. Our nutrigenomic analyses have suggested that undernutrition of early lactating cows modified the mammary gland metabolism. The holistic analyses of the systemic reaction in the mammary gland expands the knowledge of the effects of NEB and metabolic imbalance occurring in early lactation, during inflammation. These identified genes may be relevant for quantitative trait loci studies and genomic selection.
Supplementary Materials: Supplementary Materials can be found at http://www.mdpi.com/1422-0067/20/5/ 1156/s1. Table S1: Plasma insulin and metabolite concentration at the day of biopsy after dietary treatment and response to LPS challenge. Occurring at day 24 ± 3 of lactation, animals were assigned to a control (CONT, n = 8) or restricted (REST, n = 8) group. The REST animals received the ration diluted with barley straw (48% DM) for 4 days when cows from CONT were allowed to continue ad libitum intake of a lactation diet (7.1 MJ/kg DM NEL, 17.4% CP). Occurring on day 3, corresponding to the 27th day of lactation, the rear mammary quarter of animals from both groups was injected with 50 µg of LPS. Mammary biopsies were performed 24 h after LPS challenge. p < 0.01 for all variables. File S2: Protein analysis report.
Funding: Financial support for this research was provided by GISA meta-program of INRA (Ruminflame project).

Acknowledgments:
The authors thank the staff at Herbipole INRA (UE1414, Theix, France), Simon Collange for performing biopsies, Sebastien Bes, Arnaud Delavaud, Caroline Soulard, Martine Tourret, Didier Bany, Emile Tixier, Cyril Labonne, and Jacques Rouel for technical assistance, and for sample collection and laboratory analyses. The authors are grateful to the Galilait laboratory (Clermont-Ferrand, France) for the milk component, SCC and microbiological analyses. Special appreciation is extended to Gilles Foucras and Pascal Rainard for their helpful discussion regarding LPS and choice of candidate genes. Transcriptomic analyses were performed at the "PlateForme d'Exploration du Métabolisme" (PFEM, INRA, Theix, France).

Conflicts of Interest:
The authors declare no conflict of interest.