Taurodeoxycholate Increases the Number of Myeloid-Derived Suppressor Cells That Ameliorate Sepsis in Mice

Bile acids (BAs) control metabolism and inflammation by interacting with several receptors. Here, we report that intravenous infusion of taurodeoxycholate (TDCA) decreases serum pro-inflammatory cytokines, normalizes hypotension, protects against renal injury, and prolongs mouse survival during sepsis. TDCA increases the number of granulocytic myeloid-derived suppressor cells (MDSCLT) distinctive from MDSCs obtained without TDCA treatment (MDSCL) in the spleen of septic mice. FACS-sorted MDSCLT cells suppress T-cell proliferation and confer protection against sepsis when adoptively transferred better than MDSCL. Proteogenomic analysis indicated that TDCA controls chromatin silencing, alternative splicing, and translation of the immune proteome of MDSCLT, which increases the expression of anti-inflammatory molecules such as oncostatin, lactoferrin and CD244. TDCA also decreases the expression of pro-inflammatory molecules such as neutrophil elastase. These findings suggest that TDCA globally edits the proteome to increase the number of MDSCLT cells and affect their immune-regulatory functions to resolve systemic inflammation during sepsis.


INTRODUCTION
Bile acids (BAs) are amphiphilic surfactant molecules synthesized from cholesterol in the liver (1). There are various BAs in mammals (2). Cholic acid (CA) and Chenodeoxycholic acid (CDCA) are primary BAs, and CDCA gives rise to taurochenodeoxycholate (TCDC) and glycochoenodeoxycholate (GCDC) after conjugation with taurine and glycine, respectively (1). Intestinal bacteria produce secondary BAs, such as lithocholic acids (LCA) and deoxycholic acid (DCA), from primary BAs (1). The diversity of the BA pool is further increased following modification by various liver enzymes during the enterohepatic circulation of BAs (1).
In addition to the complex interaction between BAs and receptors, the promiscuity of BA-mediated immune regulation mechanisms has increased because many studies were conducted using BAs at concentrations that are not attainable under pharmacological conditions or pathological conditions in vivo. For example, 10∼30 µM of LCA was necessary to inhibit activation of the NLRP3 inflammasome (10). However, the median plasma level of LCA in healthy individuals and in pathological conditions is 10 ∼ 35 nM (18), which suggests that approximately 1,000∼3,000 × the physiological or pathological concentration is necessary to inhibit NLRP3 activation by LCA. Furthermore, whether this high plasma concentration could be attainable pharmacologically in vivo has not been investigated.
Among the BA receptors, TGR5 has received substantial attention because of the many studies that suggest the crucial roles of TGR5 in immune regulation (19). For example, various TGR5 agonists inhibit inflammation of the stomach (20) and brain (21). Functional impairment of TGR5 incurs more severe inflammation than wild-type mice in response to LPS (22) and contributes to autoimmune diseases (23). TGR5 agonists also negatively modulate NF-κB (24), and the TGR5-AKT-mTOR1 pathway inhibits macrophage chemotaxis (25).
In this study, we used taurodeoxycholic acid (TDCA) to investigate the mechanism of immune modulation rather than other BAs because taurine-conjugated BAs activate the TGR5 pathway better than unconjugated BAs and glycine-conjugated BAs (26,27). In addition, taurine-conjugated BAs exhibit less cytotoxicity than unconjugated BAs and glycine-conjugated BAs (28). TLCA exhibited a lower EC50 in TGR5 pathway activation; however, TLCA is more cytotoxic than TDCA (27,29). For this reason, we evaluated the mode of immune regulation by TDCA, which activates the TGR5 pathway (30).
In this study, TDCA increased the number of CD11b + Gr1 hi granulocytic myeloid-derived suppressor cells (gMDSCs) at a pharmacologically attainable plasma concentration, which were proteogenomically different from gMDSCs obtained from septic mice without TDCA treatment and ameliorated systemic inflammation (26).

Reagents and Cells
TDCA was purchased from New Zealand Pharmaceuticals Ltd. (Palmerston North, New Zealand). LPS from Salmonella enterica serotype enteritidis was obtained from Sigma-Aldrich (St. Louis, MO). Fetal bovine serum, L-glutamine and 2mercaptoethanol, penicillin, streptomycin and gentamicin were obtained from GibcoBRL (Waltham, MA). RPMI was obtained from Welgene (Gyeongsan-si, Korea). Mouse B-cell and T-cell isolation kits were obtained from Miltenyi Biotec for MACS (Bergisch Gladbach, Germany). IL-10-producing MICK-2 cells were obtained from BD Biosciences (San Jose, CA) and were used as positive controls for the FACS analysis of IL-10.

Mice
C57BL/6N mice (B6, Shizuoka, Japan), C57BL/6-IL10 tm1Cgn mice (IL-10KO, The Jackson Laboratories, Bar Harbor, ME) and C57BL/6-Gpbar1 tm1(KOMP)Vlcg mice (TGR5 KO, KOMP Repository, The Knockout Mouse Project, University of California, Davis, CA) were housed in the Seoul National University animal facility in a specific pathogen-free environment. Eight-to Twelve-week-old female mice were used for the experiments. The Institutional Animal Care and Use Committee (IACUC) of the Biomedical Research Institute in Seoul National University Hospital (AAALAC) approved all animal experiments (SNU 10-0331). The mice were monitored every 24 h for survival and other clinical signs (ruffled fur, diarrhea, lethargy, and loss of body weight) for 14 day after sepsis induction.

LPS Injection Model of Sepsis
The survival rate of the female mice was determined after i.p. injection of LPS (20 mg/kg), followed by the i.v. infusion of 200 µl of PBS or TDCA for 20 min (0.5 mg/kg, unless otherwise indicated) using a Medfusion 2001 system (Medex, Dublin, OH) at 30 min (unless otherwise indicated) after LPS injection. For the protection assay using IL-10 KO mice, 5 mg/kg LPS were injected i.p. For the adoptive transfer experiments, B6 mice were injected i.v. with 100 µl of purified cells. The mice were treated with LPS 24 h prior to adoptive transfer, unless otherwise specified.

CLP-Induced Sepsis Model
Female B6 mice were anesthetized, and a small abdominal midline incision was made. The cecum was ligated below the ileocecal valve and punctured 3 times using a 23-gauge needle. The abdominal incision was closed with an auto-metal clip (Mikron Precision, Inc., Ontario, Canada). The same procedure was applied to the sham-operated animals, with the exception of the ligation and puncture of the cecum. The mice were subsequently infused with 200 µl of PBS or TDCA i.v. at 2 h after CLP.

Hematoxylin and Eosin Staining
The tissues were fixed in 10% neutral buffered formalin solution (Sigma-Aldrich, St. Louis, MO) at room temperature (RT) for no less than 2 weeks and embedded in paraffin. The sections were stained with hematoxylin and eosin.

PAS Staining
Tissue sections in paraffin were deparaffinized using xylene for 10 min 4 times and were subsequently washed with distilled water for 5 min, followed by oxidization in 0.5% periodic acid solution for 15 min. After rinsing with distilled water, the sample was placed in Schiff reagent for 30 min and washed with running water for 5 min, followed by counterstaining with Harris hematoxylin for 5 min. After briefly washing with distilled water, the sample was dehydrated with 1% HCl alcohol for 10 min and cleared for microscopic observation.

Blood Chemistry
Mouse blood was collected at 48 h after LPS injection, and serum samples were harvested via centrifugation (2,000 × g, 30 min) following incubation of the blood at RT for 30 min. Alanine amino transferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN) and creatinine were measured using an automatic chemistry analyzer (Hitachi 7070, Hitachi, Tokyo, Japan).

Blood Pressure
The blood pressure was determined using the tail-cuff method with a 6-channel CODA High-Throughput NIBP Acquisition system (Kent Scientific Corporation, Torrington, CT). All measurements were recorded at each time point and averaged.

Cytokine Measurements With the Cytometric Bead Array (CBA)
Blood samples were collected in heparin-treated tubes (Sarstedt, Niimbrecht, Germany) at the indicated times after sepsis induction. The cytokine concentrations in the sera were determined using CBA according to the manufacturer's instructions (BD Biosciences). The concentrations of cytokines were calculated by a regression curve determined by known amounts of provided standard cytokines.

FACS
Leukocyte subpopulations from the spleen and bone marrow were analyzed using a FACS Canto II (BD Bioscience) and Flowjo software (Treestar, Ashland, OR) after antibody staining. The absolute numbers of spleen cells were counted by Trypan blue exclusion, and the number of CD11b + Gr1 hi cells was calculated based on the % of total splenocytes after FACS analysis.
To enumerate splenic T cells, the mice were infused with TDCA (or PBS) i.v. via the tail vein at 30 min after i.p. LPS (or PBS) injection. Splenocytes were harvested at 6, 24, or 48 h after LPS injection and stained with anti-CD4, anti-CD8 and anti-CD25 antibodies, followed by intracellular staining with Abs reactive with CTLA4 or Foxp3. The staining with anti-FoxP3 or anti-CTLA4 antibody was performed according to the manufacturer's protocol (Foxp3 Fixation/Permeabilization Concentrate and Diluent kit, eBioscience). The cells were surface-stained with FITC-conjugated anti-CD4 mAb and Percp Cy5.5-conjugated anti-CD25 mAb, fixed, and permeabilized for subsequent intracellular staining. For cytoplasmic staining, the permeabilized cells were blocked on ice for 30 min with blocking buffer and stained using PECy7-conjugated anti-FoxP3 and PEconjugated anti-CTLA4 mAbs. The cells were stained with 7amino actinomycin D (7AAD) (BD Bioscience) to determine the cell viability. IL-10 production by the cells was determined using FACS. Splenocytes were obtained from 4 groups of mice: the PBS + PBS group, PBS + TDCA group, LPS + PBS group or LPS + TDCA group. The splenocytes were stimulated with phorbol 12-myristate 13-acetate (50 ng/ml, Sigma) plus ionomycin (1 µg/ml, Sigma) for 4 h. The splenocytes were stained with Abs against CD11b and Gr1, followed by cytoplasmic staining with Abs against IL-10 after permeabilizing the cells with a one-step BD Cytofix/Cytoperm kit (BD Biosciences). Isotype control antibodies were included in all staining sets to evaluate nonspecific antibody binding.

T-Cell Proliferation Assay
T cells were purified from mouse spleen using MACS with a pan T-cell isolation kit (Miltenyi Biotec). A total of 2 × 10 5 normal splenic T cells were stimulated with 1 µg/ml of anti-CD3 (eBioscience, Waltham, MA) and 10 µg/ml of anti-CD28 (BD Bioscience) antibodies in RPMI 1640 medium containing 10% heat-inactivated FBS and 2 mM glutamine. FACS-sorted CD11b + Gr1 hi cells from the LPS + PBS group or the LPS + TDCA group were mixed with T cells at a final concentration of 4 × 10 4 /well (E:T = 1:5) in a 96-well flat-bottom plate (Nunc, Roskilde, Denmark) and cultured for 96 h at 37 • C in a humidified 5% CO 2 atmosphere. T cells cultured without CD11b + Gr1 hi cells served as a negative control. The cells were pulsed with 1 µCi [ 3 H] methyl-thymidine (Perkin Elmer, Waltham, MA) for 18 h. The cells were harvested with a Filtermate Harvester (Perkin Elmer), and the isotope incorporation was measured using a MicroBeta Plate Counter (Perkin Elmer). The data are expressed as the counts per minute (cpm) ± standard error of the mean (SEM).
Adoptive Transfer of CD11b + Gr1 hi Cells PBS or TDCA was infused i.v. via the tail vein at 30 min after LPS injection i.p. into B6 mice. The cells were isolated from the spleen at 24 h after LPS injection. Following incubation of the cells in blocking buffer for 30 min on ice, the cells were stained with a biotin-conjugated anti-mouse Gr1 antibody (clone RB6-8C5), followed by staining with anti-biotin microbeads (Miltenyi Biotec) at 4 • C for 15 min. After washing with MACS buffer (0.5% BSA and 2 mM EDTA in PBS), the cells were positively selected using an LS column (Miltenyi Biotec). The Gr1 + cells presorted with MACS were further stained with mAbs against CD11b (conjugated to PE) and F4/80 (conjugated to FITC) for FACS sorting. The CD11b + Gr1 hi F4/80 int cells were sorted using a FACSAria (BD Bioscience). A total of 1 × 10 5 cells were injected i.v. via the tail vein into B6 mice. The recipient mice were injected i.p. with LPS 24 h prior to adoptive transfer.

Microarray and RT-PCR
CD11b + Gr1 hi cells were purified by FACS. Total RNA was extracted from the cells using an RNeasy Mini kit according to the manufacturer's instructions (QIAGEN). The differential expression was assessed using a GeneChip Scanner 3000 7G (Affymetrix, Mouse 430_2, genome version mm10). A total of 14,074 genes (p < 0.05) were filtered, and 818 genes that exhibited >8-fold changes (p < 0.05) were analyzed using DNASTAR R (DNASTAR Inc.). A hierarchical clustering method (distance metric by Euclidean) was used to generate heatmaps. Genes with corrected p-values < 0.05 (one-way ANOVA, unpaired) and fold changes >2 or <-2 were considered significantly regulated. The microarray data have been deposited in the Gene Expression Omnibus under accession number GSE92948. For RT-PCR analysis, total RNA was isolated as previously described. cDNA was synthesized with an Omniscript RT KIT (QIAGEN) and PCR-amplified (MyCycler TM Thermal Cycler, BIO-RAD, Hercules, CA) using the Plantinum R PCR SuperMix (Thermo Fisher Scientific, Bremen, Germany). The primer sequences and PCR conditions are summarized in Table S1. Ingenuity Pathway Analysis (IPA R , QIAGEN, Redwood City) was used for the pathway analysis.

In-Solution Digestion of Cell Lysates
The CD11b + Gr1 hi cells were FACS-sorted from 2 groups of mice (LPS+PBS and LPS+TDCA). The lysates of the FACSsorted CD11b + Gr1 hi splenocytes were prepared using 8 M urea buffer, and the protein concentrations were determined by the BCA assay (Micro BCA Protein Assay Kit, ThermoFisher Scientific, Bremen, Germany). Dithiothreitol was added to the lysate (3 mM) and incubated at room temperature for 1 h. The cell lysates were mixed with iodoacetamide (5 mM) and incubated in a dark room for 1 h. One part of the lysates was mixed with 10 parts of 50 mM ammonium bicarbonate and digested with trypsin (1/50 × total protein amount of cell lysate, Promega, Madison, WI) at 37 • C for 16 h. The samples were subsequently desalted using a Macro Spin Column (C-18; Harvard Apparatus, Holliston, MA). The column was activated with 0.1% trifluoroacetic acid (TFA) in 80% acetonitrile in advance and subsequently equilibrated with 0.1% TFA in water (pH < 3.0). The samples were loaded into the column and centrifuged at 1,000 × g for 2 min at room temperature. The column was washed with 0.1% TFA in water, and the peptide fraction was eluted with 0.1% TFA in 80% acetonitrile. The peptide samples were dried with a CentriVap R benchtop vacuum concentrator (Labconco, Kansas City, MO), and the peptide concentration was determined using a BCA kit.

Labeling Peptides for iTRAQ
Isobaric tags for relative and absolute quantitation (iTRAQ) were used to compare the proteomes of CD11b + GR1 hi cells from 2 groups (LPS + PBS and LPS + TDCA). One hundred micrograms of peptide from each group were labeled according to the manufacturer's protocol for the iTRAQ reagent kit (Sciex, MA). Briefly, the peptide samples were reconstituted with 500 mM triethylammonium bicarbonate (TEAB) buffer, sonicated and vortexed. The 4-plex iTRAQ reagent dissolved with ethanol was added to the peptide samples and incubated at room temperature for 1 h. To one part of the sample, 3 parts of 0.05% TFA was added and incubated at room temperature for 30 min. The iTRAQ reagent-labeled peptides were pooled and subsequently concentrated to 300 µl using a CentriVap R benchtop vacuum concentrator (Labconco). The samples were then mixed with 1 ml of 50 mM triethylammonium bicarbonate (TEAB).

Q Exactive TM Hybrid Quadrupole-Orbitrap TM Mass Spectrometry
The fractionated peptide samples were subsequently loaded onto trap (C18, 3 µm, 0.7 cm, Thermo Fisher Scientific) and EASY-Spray columns (C18, 2 µm, 100 Å, 50 cm, Thermo Fisher Scientific). Easy nano II Ultra Performance Liquid Chromatography and Q-Exactive Mass Spectrometry systems (Thermo Fisher Scientific) were used to separate the peptides. The peptides were separated at a flow rate of 250 nl/min with a gradient (2-35% acetonitrile in 0.1% formic acid) for 65 min, followed by washing with 90% acetonitrile in 0.1% formic acid for 10 min. For Q-Exactive, a top 10 method was used. The Orbitrap mass analyzer was used to acquire full MS scans (m/z 300-1,600 range; resolution, 70,000). The AGC target value was 3.0E + 6. The ten most intense peaks with charge states ≥2 were fragmented in the HCD collision cell (normalized collision energy of 32%). The tandem mass spectrum was acquired in the Orbitrap mass analyzer (resolution, 17,500; AGC target value, 1.0E + 5; the intensity threshold, 8.3E + 3; the maximum allowed ion accumulation times, 20 ms for full MS scans and 120 ms for tandem mass spectrum). The peaks with 1 and 5 more charged states were excluded. The same precursor ions were also excluded after 20 s using the dynamic exclusion function.

Annotation of Peptide Sequences
The Trans Proteomic Pipeline (Seattle Proteomic Center, Seattle, WA, USA) was used to convert the mass data files into mzXML files. Peptide masses were searched using a concatenated forward and reverse mouse international protein index (IPI) database (decoy ipi.MOUSE.v3.80 database, 54285 entries) (31) with the SEQUEST-Sorcerer platform (Thermo Fisher Scientific, Sage-N Research, Milpitas, CA). Sorcerer (Sage-N Research, Milpitas, CA) was used to estimate the false discovery rate (FDR). All searches were performed based on the trypsin specificity, allowing two missed cleavages. Carbamidomethylation of cysteine was set as a fixed modification, and the oxidation of methionine, N-term and lysine iTRAQ modifications were set as variable modifications. The precursor ion mass tolerance and the fragment ion mass tolerance were set to 10 ppm and 1.0 Da, respectively.
Scaffold Q+ (Proteome Software, Portland, OR) was used to compare spectral counts, validate MS/MS-based peptides, identify proteins (FDR < 1% in at least 2 peptides), and calculate log2-fold changes (FC) and p-values (Student's t-test). Proteins with redundant peptides and multiple isoforms that could not be differentiated based on MS/MS spectra were grouped separately (= primarily assigned protein). Differentially expressed proteins (FC > 1.5) were analyzed using IPA R .

MPO and NE Activity
The myeloperoxidase (MPO) activity of the lysed CD11b + Gr1 hi cells was determined using a fluorometric MPO activity assay kit (Abcam, Cambridge, MA) according to the manufacturer's protocol. The fluorescent signal was detected with a multi-detection microplate reader (Cytation 3, Biotek Instruments, Winooski, VT, USA). The NE activity of the lysed CD11b + Gr1 hi cells was determined using a mouse neutrophil elastase (NE) activity ELISA kit (Cusabio Corporation, Wuhan, China) according to the manufacturer's instructions.

Statistical Analysis
The data are expressed as the mean ± SEM. Student's twotailed t-tests were employed to compare the test group with the control group, unless otherwise indicated. A Kaplan-Meier survival analysis and log-rank test were used to calculate the mean survival time and determine the statistical significance of the survival differences. P < 0.05 was considered statistically significant.
TDCA decreased liver and kidney damage in septic mice (Figures 1C,D and Figure S2). H&E staining of the kidney showed that the LPS-injected mice exhibited marked vacuolar degeneration of the tubules (arrows in Figure 1C). TDCA infusion almost completely ameliorated LPS-induced kidney lesions ( Figure 1C). The mucopolysaccharides on the basement membranes of the glomerular capillary loops and the tubular epithelium of the kidney were also stained with PAS (arrows in Figure 1C, right column). The loss of the brush border was remarkable in the LPS + PBS group (an arrow) and was significantly recovered by TDCA infusion (an arrow in Figure 1C). TDCA infusion normalized kidney function, liver function and hypotension from 4 h after LPS injection (Figures 1D,E, Figure S2). The production of cytokines, such as TNF-α, MCP-1, IL-6, and IL-1β, was also significantly inhibited by TDCA in both the LPS injection and CLP models (Figures 1F-H, Figure S3).
The Phenotype of CD11b + Gr1 hi Cells Increased by TDCA As previously reported (34), mice with sepsis exhibited reduced splenocyte numbers at 48 h after LPS injection and 72 h after CLP (Figure 2A, Figure S4). However, the total numbers of splenocytes were significantly increased in the LPS + TDCA group at 48 h after LPS injection and 72 h in the CLP + TDCA group. CD11b + Gr1 + cells increased both in the LPS + TDCA group and CLP + TDCA group (Figures 2B,C, Figures S4, S5). TDCA treatment did not increase the number of T cells or CD11c + cells ( Figure S6). There were no significant changes in the number of CD4 + FoxP3 + T reg cells or the expression of CTLA4 on these cells (Figure S7).

Proteogenomic Profiling of MDSC LT
The gene expression profiles were compared between MDSC LT and MDSC L (Figure 3A). The microarray results indicated that 818 genes showed more than 8-fold changes in the 95% confidence interval (p < 0.05, Figure 3B). These genes were further analyzed to decipher the association with signaling pathways using Ingenuity Pathway Analysis (IPA R , QIAGEN, Redwood City, CA). The expressions of genes for the signaling pathways necessary for immune regulation were controlled by TDCA (p < 0.05, Figures 3C,D). In particular, the genes for pro-inflammatory signaling pathways, such as chemotaxis [denoted as Rac signaling (35), leukocyte extravasation, ILK signaling (36), and PAK signaling (37)], MAPK signaling, CREB signaling (38), p70S6K signaling (39), thrombin signaling (40), and P2Y purinergic receptor signaling (41), were significantly down-regulated ( Figure 3C). The genes for the antigen presentation pathway and cell cycle control were highly up-regulated ( Figure 3D).
We further compared the proteomic profiles between MDSC LT and MDSC L cells (Figure 4). Using iTRAQ labeling, 1,643 unique proteins were identified at a protein threshold with a 1.0% false discovery rate. Among these proteins, 887 showing a peptide spectral count in more than 2 assays from triplicate assays are depicted in the heat map ( Figure 4A). The heat map shows two distinct proteome clusters that were downregulated in MDSC LT (cluster A) and up-regulated in MDSC LT (cluster B) compared with MDSC L cells. Various heterogeneous ribonucleoproteins (hnRNPs) responsible for the alternative splicing of pre-mRNA were down-regulated ( Figure 4B, nuclear cDNAs from each mouse were pooled and gene expression was profiled in duplicate. Eight hundred eighteen genes were filtered (>8-fold change, p < 0.05) from 14,074 genes (p < 0.05) and analyzed to generate a heat map and scatter plot. The gene with the lowest expression value is displayed as a highly suppressed gene in blue (signal intensity = 2 −2.9 ). The highest expression value is displayed as a highly expressed gene in red (signal intensity = 2 15.8 ). The range of expression intensity is color coded and is depicted in the right. (B) The double-log scatter plot indicates the signal intensities of all oligo probes. The 818 genes shown in red showed > 8-fold change and p-value < 0.05. The best-fit linear trend (dotted line) and the ±2-fold change lines are indicated. (C) Four hundred ninety-six genes that exhibited 5-fold down regulation (p < 0.05) were extracted, and their pathway associations were investigated. (D) Five hundred forty-four genes that exhibited 8-fold up regulation (p < 0.05) were extracted, and their pathway associations were investigated using IPA ® . The pathways overlapping with the proteomic analysis are denoted by "*". −Log (p-value) was calculated using the right-tailed Fisher's exact test to measure the likelihood that the association between a set of genes and pathway was due to random chance. Pathways with p < 0.05 are shown.
blue symbols) (42). The proteins for cytoskeletal rearrangement were down-regulated ( Figure 4B, red symbols). A set of histones (1H1C, 1H1E, H2AFX) that repress transcription by preventing transcription factor access (43) were up-regulated in MDSC LT compared with MDSC L cells ( Figure 4C, nuclear blue symbols). TDCA treatment resulted in heterogeneous ribosome composition by up-regulating the expression of a set of ribosomal proteins in MDSC LT (Figure 4C, green symbols). Specialized sets of ribosomes with different functions may be complexed because of the differential expression of ribosomal subunit proteins following TDCA treatment (44). The expression of proteome for mitochondrial oxidative phosphorylation was up-regulated ( Figure 4C, gray symbols). In addition, CAMP was up-regulated ( Figure 4C, cytosolic black symbol), which participates in inhibiting the expression of various pro-inflammatory molecules, such as TNF-α, CXCL8, IL6, CCL2, IL-1β, leukotriene B4 and nitric oxide, by regulating Erk1/2, P38 MAPK, Jnk, Akt, and NF-κB (45). In the proteomic analysis, proteins for chemotaxis and pro-inflammatory pathways were down-regulated, as indicated by microarray analysis (Figure 4D). Proteomic profiling also showed that the expression of various proteins responsible for endocytosis and immune-regulation, such as the acute phase response, EIF2 signaling (46), and LXR/RXR/FXR signaling (47), were up-regulated by TDCA treatment (Figure 4E). In both the microarray and iTraq analyses, the pathways for p70S6K signaling, P2Y purinergic receptor signaling, MAPK signaling, thrombin signaling, and leukocyte extravasation signaling were down-regulated in MDSC LT compared with MDSC L .

Mode of Immune Regulation and Proliferation of MDSC LT
Based on the proteogenomic profile, we further characterized MDSC LT compared with MDSC L . The MPO responsible for ROS generation was higher in MDSC LT than MDSC L cells ( Figure 5A). Microarray and RT-PCR analysis of transcripts showed that iNOS levels in both MDSC LT and MDSC L cells were similar to those of normal splenocytes (data not shown). There were no significant differences in the mRNA expression of arginase-1 (p = 0.07, Figure 5B). These findings suggest that MDSC LT exerted an immune-regulatory role similar to the role played by gMDSCs using ROS (48). In addition, neutrophil elastase (NE) activity was lower in MDSC LT than MDSC L cells (p < 0.05, Figure 5C). Considering that NE plays a role in tissue destruction via extracellular matrix proteolysis, MDSC LT may lead to less tissue destruction than MDSC L cells.
We further characterized the IL-10 production induced by TDCA because IL-10 plays a crucial role in the anti-inflammatory phenotypes of MDSC (49). Interestingly, there was no significant The proteins in cluster A were further sub-clustered based on their common functions. The red symbols represent molecules that play roles in cellular trafficking. Blue symbols (alternative splicing), green symbols (protein metabolism), gray symbols (mitochondrial oxidative phosphorylation) and yellow symbols (pro-inflammatory signaling/adhesion/ cytokine molecules) were used to denote each sub-cluster. (C) The proteins in cluster B were further sub-clustered as (B). The pink symbols in the nucleus represent molecules that are potentially associated with the cell cycle and differentiation. The blue symbols in the nucleus represent histones for chromatin remodeling. The green symbols (ribosomal proteins), cytosolic pink symbols (cytoskeletal proteins), gray symbols (mitochondrial respiratory chain) and yellow symbols (acute phase response proteins) were color-coded based on their functions listed in the database of Ingenuity Pathway Analysis (IPA ® , QIAGEN). Open symbols are molecules predicted by IPA ® that were not observed in the experiments. The gray lines represent potential interactions between the molecules predicted by IPA ® . The curved blue lines and pink lines represent the cytoplasmic and nuclear membranes, respectively. Canonical signaling pathways for up-regulated proteins (D) and down-regulated proteins (E) were analyzed. *denotes signaling pathways overlapped with the microarray analysis. -Log (p-value) was calculated using the right-tailed Fisher's exact test to measure the likelihood that the association between a set of genes and pathway was due to random chance. Pathways with p < 0.05 are shown. IL-10 production by MDSC LT and MDSC L (Figure S14A). In addition, TDCA treatment conferred protection to IL-10 KO mice comparable to WT mice following LPS challenge ( Figure S14B). TDCA treatment after LPS injection did not affect the serum concentration of IL-10 of B6 mice (Figures S14C,D), which suggests that IL-10 may not be involved in the antiinflammatory effector functions of MDSC LT and TDCA.
The expression levels of oncostatin M (Osm) and lactoferrin, which play crucial roles in innate immune responses, were increased in the spleens of septic mice after TDCA infusion (Figures 5D,E).
Because the microarray data showed increased expression of CD244 in MDSC LT compared with MDSC L (fold change = 2.7 ± 1.1) and the CD244 hi MDSC population was significantly increased in tuberculosis patients and in tumor-bearing mice (50,51), we analyzed the expression of CD244 on MDSC LT . FACS analysis also showed that MDSC LT expressed higher levels of CD244 than MDSC L cells (Figures 5F,G and Figure S15). Intravenous injection of anti-CD244 antibody neutralized the protective effect of TDCA infusion in vivo (Figure 5H). Because CD244 is expressed on various cell types, we investigated the role of CD244 expressed on MDSC LT after adoptive transfer of MDSC LT into septic mice and i.v. injection of anti-CD244 antibody. Intravenous injection of anti-CD244 antibody abrogated the protective effect of adoptive transfer of MDSC LT cells in the LPS injection setting (Figure 5I).
Adoptive transfer of MDSC LT increased the number of MDSC LT in both the bone marrow ( Figure 5J) and spleen ( Figure 5K) of recipient B6 mice with sepsis, which suggests that TDCA may induce the production of an autocrine factor necessary for increasing the number of MDSC LT cells. Microarray and RT-PCR analyses showed that MDSC LT expresses prokineticin2 to a greater extent than MDSC L cells, which is crucial for MDSC proliferation (Figure 5L) (52).

DISCUSSION
Here, we report that TDCA increases the number of immuneregulatory MDSC LT cells, which are distinct from gMDSCs that previously reported, through the global editing of the proteome that increase anti-inflammatory molecules such as MPO, Osm, lactoferrin, and CD244, in addition to prokineticin 2, which are essential for the proliferation of myeloid cells (53). Global editing of the proteome by TDCA decreases pro-inflammatory functions by inhibiting the expression of tissue-degrading enzymes such as neutrophil elastase. In addition, anti-inflammatory molecule expression was increased in MDSC LT compared with MDSC L . For example, Osm is involved in anti-inflammatory responses and restores normal homoeostasis after tissue injury or infection (54). Without Osm, LPS induces exaggerated pathogenesis (55). Lactoferrin also possesses anti-oxidant, anti-inflammatory and anti-bacterial functions, which are crucial for protection from sepsis (56). Considering previously published expression profiles (57,58), immunohistochemistry data showing increased expression of Osm and lactoferrin in spleen might be due to MDSC LT population in spleen (59-63). Thus, TDCA ameliorates systemic inflammation, normalizes blood pressure, prevents kidney injury and prolongs survival in a mouse sepsis model. Considering that the median plasma concentration of TDCA is 33.9 nM in healthy individuals (64), approximately 10∼20 × the physiological plasma levels of TDCA is necessary to inhibit the systemic inflammation incurred by LPS injection or by CLP. The IC50 of TDCA that inhibits the binding of N-3 H-methylscopolamine to the M3 muscarinic receptor of acetylcholine was 170 µM (33), which suggests that the plasma Cmax after i.v. infusion of 1 mg/kg TDCA is 1/176 × the concentration necessary for antagonizing the M3 muscarinic receptor. The EC50 of deoxycholic acid (DCA) to increase cAMP production in a TGR5-dependent manner was 1 µM (26). This finding suggests that the plasma Cmax of TDCA after 0.5 mg/kg i.v. infusion may be sufficient to activate the TGR5 pathway in vivo. Considering together with that TDCA does not protect TGR5 KO mice under sepsis in this study, TDCA might play roles via TGR5 pathway to control sepsis.
TDCA activates the S1PR2 pathway (65) in addition to the TGR5 pathway (24,26,30). Activation of the S1PR2 pathway augments pro-inflammatory responses, and TDCA activates the S1PR2 pathway at more than 5 µM (66). Accordingly, it is less likely that S1PR2 plays roles in TDCA-mediated immune regulation, even in various pathological conditions where the plasma concentration of TDCA is less than 1 µM (18).
Conjugated BAs rarely activate nuclear receptors. The most potent activator of FXR, CDCA, had an EC50 of 50 µM in FXR activation, and other conjugated BAs were inactive in this assay at concentrations up to 100 µM (16). Conjugated BAs in fasting blood are typically less than 1 µM with concentrations that reach as high as 5 µM after a meal (4). For these reasons, we hypothesized that the immunomodulatory role of TDCA may be largely dependent on the activation of the TGR5 pathway in vivo; however, it needs to be ruled out for interaction with other membrane receptors, such as α1β4/α5β1 integrin, mAChRs and large conductance Ca 2+ -activated K + channels using a panel of loss-of-function or gain-of-function studies in future investigations.
When rats were injected with TDCA i.v. once every day for 28 days, the no-observed-adverse-effect level (NOAEL) was >10 mg/kg (67). Taken together with the toxic dose ranges previously reported (32), the pharmacological effects of TDCA are observed at less than 10 × the plasma concentration of healthy individuals, which is far less than the toxic dose ranges (32).
TDCA-induced CD11b + Gr1 hi CD31 − CD177 + MDSC LT cells are distinctive from the conventional gMDSC subsets present in tumors and inflammatory conditions. In the tumor microenvironment, the CD11b + Gr1 hi CD31 − gMDSC subset is less immunosuppressive than the CD11b + Gr1 int CD31 + mMDSC subset (68). In comparison, TDCA-induced CD11b + Gr1 hi CD31 − MDSC LT cells are more antiinflammatory than MDSC L cells, which increased under inflammatory conditions without TDCA (conventional gMDSCs). Furthermore, in terms of anti-inflammatory potency, TDCA-induced CD11b + Gr1 int CD31 + cells (with a surface phenotype of conventional mMDSCs) did exhibited reduced anti-inflammatory functions compared with MDSC LT or MDSC L cells. These findings suggest that the phenotype and function of the TDCA-induced MDSC LT cells are different from the gMDSC and mMDSC subsets in tumor-bearing mice and gMDSCs present in inflammatory conditions (68).
The CD244 expression on MDSC LT was substantially higher than that of MDSC L . CD244 -CD48 interactions have been reported as both activating and inhibitory, depending on the context or the isoform of CD244 engaged (69). Regardless of the contradictory roles of CD244, several reports suggest critical roles for CD244 -CD48 interactions in immune regulation. CD244 on intraepithelial lymphocytes inhibits inflammatory colitis (70). In addition, the risks of rheumatoid arthritis and systemic lupus erythematosus are increased in patients with genetic polymorphisms in CD244 (71). Moreover, mutation of the CD244 gene is closely linked to the risk of autoimmunity (72).
In this study, we identified an increase in the expression of CD244 on MDSC LT and neutralization of the regulatory activity of MDSC LT by anti-CD244 antibody treatment. Thus, CD244 on MDSC LT may play a crucial role in inhibiting the function of other inflammatory cells as an immune checkpoint. However, detailed loss-of-function studies are necessary to assess the role of CD244 expressed on MDSC LT in sepsis because CD244 also plays crucial roles in NK cell functions and NK cells contribute to antibacterial immunity via crosstalk with other immune cells (73). For these reasons, the homotypic interaction of CD244 and heterotypic CD244-CD48 interactions between various hematopoietic cells must be examined to elucidate protective roles of TDCA.
Proteogenomic analysis suggests three potential transcriptional and translational control mechanisms for global proteome editing by TDCA (Figure 6). First, increases in linker histones by TDCA may inhibit the binding of various transcription factors to chromosomes (43). Second, the expression of various hnRNPs diversified by TDCA may change of the mRNA repertoire by reprogramming alternative splicing (42). Finally, the heterogeneous ribosome composition incurred by TDCA generates new sets of functional proteomes (44). In this manner, global editing of the proteome may be responsible for the down-regulation of pro-inflammatory molecules that play roles in chemotaxis, thrombin signaling (74), IL-1 signaling and purinergic responses (75) and the up-regulation of antiinflammatory responses (notably APR, ROS generation and FXR), as well as increased hematopoiesis.
One question remains elusive. How does TDCA inhibit inflammation and yet control bacteria in mice CLP model that are infected with live bacteria? Because inhibition in a pro-inflammatory response leaves that animal at a significant disadvantage and should lead to increased sensitivity to infection. Our proteomics and immunohistochemistry data provides plausible bacterial clearance mechanisms exerted by TDCA. TDCA increases expression of MPO, lactoferrin and oncostatin M (Figure 5). These molecules play roles in bacterial clearance (59-63). MPO is essential in oxidative killing of bacteria in phagolysosomes (76). MPO generates lethal antimicrobial oxidants that react with ingested bacteria to kill them (76). Lactoferrin is a non-haem iron-binding protein (77). Thus, lactoferrin sequestrate iron in sites of infection, which deprives the microorganism of iron essential for living, thus killing bacteria (78). Lactoferrin also interact with the cellular membrane of infectious agent directly and cause bacterial lysis (78). Oncostatin M is responsible for increased levels of liver iron regulatory hormone hepcidin and decreases serum FIGURE 6 | Schematic diagram indicating the potential anti-inflammatory mode of action of TDCA in mice with sepsis. When administered, TDCA edits the transcription/translation machinery and reprograms CD11b + Gr1 hi cells to regulate inflammation and proliferation. In future studies, the respective roles of known TDCA receptors (such as TGR5, α1β4/α5β1 integrin, mAChRs and large conductance Ca 2+ -activated K + channels) must be tested in immuno-regulation in response to TDCA.
iron levels that hampers bacterial growth (79). Although many reports showed bacterial clearance by these molecules, we need to decipher exact mechanism of bacterial clearance by these molecules when we infused TDCA into mice under sepsis.
More than 2 million individuals worldwide suffer from sepsis on an annual basis (80). Because a plethora of pathogenic signaling pathways are simultaneously activated in septic patients, clinical trials targeting a single inflammatory mediator, coagulation factor or pro-inflammatory signal transducer have not shown significant survival benefits (81). In contrast to former strategies examining the blockade of pro-inflammatory pathways, targeting intrinsic immune regulatory mechanisms may be more effective for inhibiting the broad spectrum of pathways that are activated in sepsis (82). For these reasons, in vivo expansion of MDSC LT using a pharmacological dose of TDCA may be a plausible approach to inhibit the broadspectrum pathogenesis exhibited in septic patients.

AUTHOR CONTRIBUTIONS
SC contributed to the design and analysis of the FACS experiments in addition to adoptive transfer experiments. Y-HK investigated the role of TDCA in terms of blood pressure, cytokine production, and survival using the LPS injection and CLP models, performed the animal work in the adoptive transfer experiments and determined the surface phenotypes of immune cells by FACS. Y-JK determined the intracellular phenotype of immune cells by FACS and performed the splenocyte preparation and FACS staining for FACS sorting in the adoptive transfer experiments. Y-WK studied effects of TDCA using TGR5 KO mice. SM, YL, JJ, DH, and YK contributed to the proteomic analysis. H-EJ was involved in the gene expression studies using microarray and RT-PCR. J-AC and T-JK analyzed the protein expression using western blotting. T-CC and N-HC studied the tissues using confocal microscopy. H-JM supported unspecified basic laboratory work. H-RK assisted with the FACS sorting. YN and S-HS studied the tissues after staining with H&E and PAS. H-CL and E-HN were involved in the gene expression studies. HC and MC analyzed the microarray data. NM provided guidance for the neutrophil studies and the antibodies necessary for neutrophil staining. All authors contributed to the manuscript preparation. S-YS supervised the study, analyzed the data and wrote the manuscript.

ACKNOWLEDGMENTS
The authors thank Hyung Yeol Choi for assistance in cell sorting and Je-In Youn for comments regarding the MDSCs. This work was supported by a grant funded by the National Research Foundation, Ministry of Science, ICT and future planning (2012R1A5A2A44671346), a grant from the Korea Healthcare Technology R&D Project, Ministry of Health and Welfare (A062260), and a grant from Gangwon Province, Republic of Korea.