Differential Induction of Ly6G and Ly6C Positive Myeloid Derived Suppressor Cells in Chronic Kidney and Liver Inflammation and Fibrosis

CD11b+Gr1+ myeloid derived suppressor cells (MDSC) are known to be very potent suppressors of T cell immunity and can be further stratified into granulocytic MDSC and monocytic MDSC in mice based on expression of Ly6G or Ly6C, respectively. Here, using these markers and functional assays, we aimed to identify whether MDSC are induced during chronic inflammation leading to fibrosis in both kidney and liver and whether additional markers could more specifically identify these MDSC subsets. In an adenine-induced model of kidney inflammation/fibrosis suppressive Ly6Gpos MDSC were induced. The suppressive function within the Ly6G+ MDSC population was exclusively present in IFNγRβ expressing cells. In contrast, in chronic inflammation in the liver induced by bile duct ligation, suppressive capacity was exclusively present in the Ly6Cpos MDSC subset. Gene expression analyses confirmed the differential origins and regulation of those MDSC subsets. Additionally, depletion of MDSC in either kidney or liver fibrosis enhanced fibrosis markers, indicating a protective role for MDSC in organ fibrosis. Thus, our data demonstrate that during liver inflammation and kidney fibrosis MDSC with similar function arise bearing a distinct marker profile and arising from different cell populations.

Organ fibrosis is a result of chronic inflammation and is accompanied by the infiltration of pro-inflammatory monocytes, macrophages, neutrophils and T cells. These inflammatory conditions go hand in hand with wound healing processes, which lead to continued replacement of dying parenchymal cells with connective tissue or extracellular matrix [14]. Organ fibrosis leads to severe functional damage of the organ and is one of the leading reasons for morbidity and mortality with growing prevalence in end-stage liver or kidney disease. During chronic inflammation many factors (e.g. IL-1β, TNF, IFNγ, DAMPs) are released, which may promote accumulation, activation or induction of MDSC in the inflamed organ [15]. These MDSC may then prevent immune-mediated damage and reduce the harmful effects of prolonged inflammation by switching off pro-inflammatory immune cells. However, specific identification of MDSC in chronically inflamed fibrotic organs is challenging, as pro-inflammatory monocytes, neutrophils and macrophages, expressing similar markers and effector molecules but lacking suppressive function, also infiltrate the inflamed tissue. In addition to their suppressive function, MDSC can be subdivided into two major populations that either express Ly6C or Ly6G [1,16] More specifically, monocytic Ly6C pos MDSC express CD11b, Gr-1, Ly6C but no Ly6G and the granulocytic/neutrophilic Ly6G pos MDSC express CD11b, Gr-1, Ly6G, but are low in Ly6C [15]. Many additional markers, such as B7H1, IL4Rα or IFNγRβ, are suggested to more specifically identify MDSC [17]. Myeloid cells stratified according to these markers can fulfil distinct suppressive functions in different diseases such as cancer, infection or autoimmunity [1,18,19]. However, it is not clear if these MDSC subpopulations play a suppressive role in organ fibrosis due to chronic inflammation.
In the kidney a suppressive role for MDSC has been described in renal cell carcinoma and renal transplantation ( [20,21]), but their role in kidney fibrosis has not been addressed so far. In the liver MDSC are known to accumulate in mouse models of acute immune-mediated liver injury ( [22]) as well as in patients with chronic inflammatory liver disease, like hepatitis C ( [23]), or hepatocellular carcinoma ( [24]), which is thought to arise in part as a result of chronic liver inflammation and fibrosis ( [25]). A well-established experimental animal model for liver fibrosis is bile duct ligation (BDL) in rodents, in which hydrophobic bile acid mediated liver injury leads to chronic inflammation, fibrosis and ultimately hepatic cirrhosis. In adenine-induced tubulointerstitial nephritis, excessive insoluble adenine causes tubular cell damage also leading to chronic inflammation and end-stage fibrosis. In this study, we aimed to identify the MDSC subsets that arise during chronic inflammation leading to fibrosis in both the kidney and the liver and further investigated whether additional markers could more specifically identify these MDSC subsets. Here, we describe that during adenine-induced tubulointerstitial nephritis suppressive capacity resides within the Ly6G pos MDSC subset that also expresses IFNγRβ. In contrast, during chronic hepatic inflammation after bile duct ligation, suppressive capacity was exclusively present in the Ly6C pos MDSC subset and did not correlate with IFNγRβ expression. These data indicate that during liver and kidney fibrosis suppressive MDSC with similar function, but different phenotype are induced.

Material and Methods
Mice, diet, treatments C57BL/6J mice were bred in the central animal facility in Bonn according to the Federation of European Laboratory Animal Science Association guidelines and maintained under SPF conditions. Mice were fed an adenine-enriched diet (10g/5kg) (Sniff, Soest, Deutschland) to induce kidney fibrosis [26] Mice were injected with 10 μg LPS and 1 μg IFNγ every second day for 2-3 times to induce MDSC in the spleen [2]. Retinoic acid (Tretinoin, Roche, Grenzach-Wyhlen, Germany) was provided at 1g/liter in the drinking water of mice from day 7 onwards after bileduct ligation or from day 7 on after the start of adenine feeding. This corresponds to a daily dose of approx. 0,5 mg/mouse. All efforts were taken to minimize suffering. Mice were sacrificed by cervical dislocation. All animal experiments were approved by the Animal Care Commission of Nordrhein-Westfalen (84-02.04.2013.A014 and 84-02.04.2013.A129).

Bile duct ligation
Preoperatively, mice were injected with 5mg/kg carprofen. Mice were anesthetized by isoflurane inhalation during surgery. The abdomen was opened by a midline laparotomy and the bile duct was carefully mobilized and ligated with 5-0 silk. In sham-operated controls, the bile duct was mobilized but not ligated. The incision was closed with 5-0 silk. Mice were sacrificed at the indicated times after surgery and non-parenchymal liver cells were isolated and analyzed by flow cytometry.

Isolation of non-parenchymal liver and kidney cells
Isolation of liver non-parenchymal cells was performed as described before [27]. Shortly, livers were perfused and mechanically dissociated. Non-parenchymal cells were collected from the interface after density centrifugation. Cells were analysed by flow cytometry or sorted using the gating strategy as indicated. Isolation from the kidney was performed as described earlier [28] Briefly, kidneys were perfused, mechanically dissociated and digested in RPMI 1640 (Gibco, Life Technologies, Darmstadt, Germany) containing collagenase and DNase. After letting tubular cells sediment for some minutes from homogenized cell suspensions the supernatant with non-parenchymal cells was taken to perform flow cytometry analysis or sort using the gating strategy as indicated.

Real-time PCR
Liver and kidney tissue samples were homogenised and total RNA was isolated using the NucleoSpin RNA kit (Macherey-Nagel, Düren)). RNA was transcribed into cDNA using the High-Capacity cDNA Reverse Transcription kit (Gibco, life technologies) following the manufacturers instructions. Real-time PCR was performed using the SybrGreen PCR Master mix on a light cycler 480 instrument II (Roche, Switzerland) Primers used: α-SMA fw: 5 0 -TCCAGAGTC-CAGCACAATACCAGT-3 0 , rv: 5 0 -TGACAGAGGCACCACTGAACC-3 0 , TGF-β fw:  , anti-B7-H1 (MIH-5) and CD45.2 (104). Antibodies were purchased from eBioscience or Biolegend. Dead cells were excluded with Hoechst-33258 (Sigma). All stainings were performed in the presence of 10 μg/ml Fc-Block (clone 2.4G2). Isotype matching fluorochrome-labelled antibodies were used as indicated. For the enumeration of total cell numbers per organ an equal amount of counting beads was added to each sample. Samples were acquired using an LSR Fortessa (BD Bioscience, Heidelberg, Germany), cell sorting was performed with Aria III (BD Bioscience, Heidelberg, Germany), and data was analysed using FlowJo software (TreeStar, Inc, Ashland, OR).

Inflammatory gene expression analysis
Non-parenchymal cells were isolated from individual livers and kidneys 14 and 10 days after BDL and the start of adenine feeding, respectively. Cells were stained with antibodies against CD11b, Ly6G and Ly6C in the presence of Fc-block (clone 2.4G2). Cells were washed and subsequently sorted into CD11b+Ly6C+ and CD11b+Ly6G positive fractions from each organ. 20.000 cells were analysed for their inflammatory gene expression profile using the nCounter Mouse Inflammation Gene Expression CodeSet (Nanostring Technologies, Seattle, USA) according to the manufacturer's instructions.

Statistics
All experiments were performed at least three times with groups of 3 mice unless otherwise stated. Results are expressed as mean ± SEM. Statistical significance was calculated using ANOVA ( Ã p 0.05, ÃÃ p 0.01, ÃÃÃ p 0.001).

Ly6C and Ly6G positive myeloid cells accumulate during inflammation and fibrosis in liver and kidney
In order to characterise MDSC arising in vivo after chronic inflammation we induced liver fibrosis via bile-duct ligation (BDL) [29] and kidney fibrosis by feeding mice an adenine rich diet [26]. Furthermore, as a positive control, mice were either injected with LPS/IFNγ, as this has been shown to induce suppressive Gr-1 positive MDSC in the spleen, or in vitro generated BM-derived MDSC were used [2]. Although under steady state conditions, MDSC-like myeloid cells are present in the liver, kidney and spleen (Fig. 1A), upon BDL, adenine feeding or LPS/ IFNγ treatment, the numbers of CD11b + Gr-1 int/high myeloid cells significantly increased in the respective organs (Fig. 1A). We then determined the relative contribution of monocytic myeloid cells (CD11b + Ly6C high Ly6G neg ) and granulocytic myeloid cells (CD11b + Ly6C int Ly6G high ) within the Gr-1 positive myeloid population (Fig. 1B, C). Both subsets were present in liver and kidney, even under steady-state conditions, but monocytic myeloid cells were significantly increased after BDL whereas granulocytic myeloid cells were enriched in fibrotic kidneys after adenine feeding (Fig. 1B, C). In the spleen of LPS/IFNγ-treated mice and in the bone-marrow cultures both monocytic and granulocytic myeloid cells were present, although the proportion of granulocytic myeloid cells was increased in both cases (Fig. 1B, C). Not only did the proportion of CD11b + Ly6C + and CD11b + Ly6G + cells increase in the liver and kidney, respectively, also absolute numbers of both cell types were increased substantially (Fig. 1D, E) Together, these data show that during liver and kidney fibrosis myeloid cells with a different MDSC phenotype accumulate. Furthermore, by separately analysing Ly6G and Ly6C expression, populations of MDSC-like myeloid cells reported to be Gr-1 positive, show differential accumulation depending on the anatomical location.
Monocytic MDSC accumulate in the liver whereas granulocytic MDSC are induced in the kidney during organ inflammation and fibrosis Suppressive activity has been reported for both monocytic Ly6C high expressing and Ly6G expressing myeloid cells [1,16]. As we found that Ly6G and Ly6C expressing myeloid cells are induced with differential preference in organ fibrosis, cytokine injection in vivo and in vitro culture of bone marrow cells with CSF2 ( Fig. 1), we analysed which of these subsets constituted MDSC and thus possessed suppressive capacity. As the development of MDSC is proposed to be IFNγ-dependent and that IFNγ is needed MDSC effector function [12,13], we sorted CD11b + myeloid cells according to their expression of Ly6C, Ly6G and IFNγRβ ( Fig. 2A) and cocultured them with CFSE-labelled T cells in the presence of anti-CD3ε/CD28 coated beads. Interestingly, IFNγRβ positive Ly6C/G expressing myeloid cells were readily detected in the inflamed spleen and kidney, but not in the chronically inflamed liver after bile duct ligation or the bone marrow cultures. In bile duct ligated mice, Ly6C pos myeloid cells possessed suppressive function (Fig. 2B), whereas in adenine-fed mice suppressive capacity was present within the renal Ly6G pos myeloid subset (Fig. 2C), In both the in vivo LPS/IFNγ-induced and bone marrow-derived myeloid cells, suppressive activity was present in both Ly6G pos and Ly6G pos subsets (Fig. 2D, E). Moreover, in the kidney suppressive capacity was restricted to IFNγRβ expressing myeloid cells (Fig. 2C). In contrast, in the liver IFNγRβ was not expressed on suppressive myeloid cells (Fig. 2B), nor did the presence or absence of IFNγRβ on splenic myeloid cells influence their suppressive capacity (Fig. 2D). Thus, IFNγRβ expression does not seem to constitute a reliable or unique marker for the identification of functionally active MDSC. Gene expression analysis of a large panel of inflammatory genes revealed that the Ly6G pos and Ly6C pos MDSC from liver and kidney are not similarly regulated. Most genes that were differentially expressed by these two subsets, were either part of a Ly6C-signature (also present in Ly6C pos cells from an adenine-fed kidney), or of a Ly6G-signature in both liver and kidney (S1 Fig.). A subset of genes was specifically regulated in kidney Ly6G pos MDSC, of which several seem to be involved in the continuous attraction and activation of neutrophillic granulocytes/myeloid cells. Interestingly, neither liver nor kidney MDSC expressed arginase 1 (arg1) or iNOS (nos2) mRNA, indicating that suppressive function in both liver and kidney is not achieved via depletion of arginine or production of ROS. Together, these data indicate that MDSC accumulating during chronic inflammation and fibrosis in liver and kidney can originate from the monocytic

MDSC influence fibrosis progression in both liver and kidney inflammation
Although we find MDSC with potent ex vivo inhibitory function to accumulate during both liver and kidney fibrosis (Fig. 2), it was unclear whether the presence of MDSC would influence fibrosis progression. To test this, we treated bile duct ligated and adenine-fed mice with alltrans-retinoic acid (ATRA) in their drinking water or not, which is known to change MDSC functionality leading to their inability to exert suppressive function ( [30]). As a control we sorted CD11b + Gr-1 pos cells from the liver of ATRA-treated bile-duct ligated mice and found viable (Hoechst negative), non-parenchymal cells stained with CD11b and Gr-1 (A) or viable, CD11b pos cells stained for Ly6G and Ly6C (B). Representative (A, B) and cumulative (C) data of 4 (liver, spleen) or 3 (kidney, bone marrow) independent experiments are shown (n>9). Absolute numbers of CD11b + Ly6C + and CD11b + Ly6G + meyloid cells in the liver (D) and kidney (E) at the indicated time-points after BDL or adeninefeeding, respectively. Data are depicted as mean +/-SEM. Significance was calculated by ANOVA. *p 0.05, **p 0.01, ***p 0.001. doi:10.1371/journal.pone.0119662.g001 that these did not posses suppressive capacity towards T cells in a proliferation assay (data not shown). In both bile-duct ligated and adenine-fed ATRA-treated mice, fibrosis markers like α-SMA, TGF-β, Collagen IV and vimentin were significantly increased (Fig. 3), indicating that the induction of MDSC during chronic inflammation dampens ensuing fibrosis in both the kidney and the liver.

General MDSC specific surface markers not found
So far, one of the major difficulties in the study of MDSC function in inflammation-associated diseases is the absence of a reliable surface marker that is specifically expressed on myeloid cells with suppressive function. Although IFNγRβ expression defined Ly6G pos MDSC in the setting of kidney fibrosis, we excluded the IFNγRβ as a general marker, since it was not universally expressed on monocytic or granulocytic MDSC (Fig. 4), as it was not expressed on liverderived Ly6C pos MDSC. Several other markers have been proposed as markers for MDSC in recent years. Thus, we performed a comprehensive phenotypic analysis to possibly identify a common additional marker for MDSC besides CD11b and Ly6G or Ly6C (Fig. 4 and S2 Fig.). Many of these already proposed markers are indeed expressed on several MDSC populations. For instance, IL-4Rα expression is induced during liver and kidney fibrosis and in the spleen after LPS/IFNγ treatment (Fig. 4). However, in liver and kidney the IL-4Rα is not only expressed on the suppressive subpopulation of myeloid cells, but also on the non-suppressive myeloid subpopulation, indicating that IL-4Rα expression cannot serve as a common marker for suppressive MDSC [33]. Furthermore, B7H1 expressed by MDSC can have a functional role in mediating suppression [17,31]. In the liver, B7H1 expression was apparently restricted to the Ly6C pos MDSC subset (Fig. 4). However, in the kidney Ly6C pos and Ly6G pos myeloid cells expressed B7H1, excluding B7H1 as a general MDSC specific marker. In order to clarify whether in principle liver and kidney MDSC are functionally similar or not we analysed mRNA expression levels of a panel of inflammation-associated genes (S1 Fig.). These analyses showed that these expression patterns were organ-independently either Ly6C-or Ly6G-associated but did not associate with suppressive function. Only a small panel of inflammatory genes were exclusively expressed in kidney Ly6G pos MDSC, known to be involved in the chemotactic attraction of neutrophillic granulocytes (S1 Fig.). However, taken together these data do not identify a universal MDSC marker and supports the notion that such a marker may not exist, although within particular disease entities such additional markers can be identified.

Discussion
Currently, the identification of suppressive MDSC in for instance murine tumour models is based on the co-expression of CD11b and Gr-1 on myeloid cells. Such CD11b + Gr-1 + myeloid cells are then tested for their suppressive capacity towards innate or adaptive immune cells to formally prove their MDSC status. Pro-inflammatory molecules such as IL-1β [32], IL-6 [9] and PGE 2 [33] are thought to promote the induction of MDSC in tumours, whereas proinflammatory TNF signalling can drive the accumulation of MDSC in tumours [34]. Using CD11b and Gr1 surface molecules to identify suppressive myeloid cells, these cells are found in infection [35,36] and autoimmunity [37]. In this study, we investigated whether in local chronic inflammation MDSC are induced to suppress ongoing inflammatory immune reactions and their resulting organ damage. We now demonstrate, using well-established models of kidney or liver inflammation and fibrosis, that during chronic inflammation, CD11b + Gr1 + myeloid cells expand. Further staining for Ly6C and Ly6G, both of which are recognised by anti-Gr1 antibodies, showed differential expansion of Ly6G pos and Ly6C pos myeloid cells. The first preferentially accumulate in the inflamed kidney, whereas the second accumulated significantly in the livers of bile duct-ligated mice, not only as percentage within CD11b + myeloid cells but very prominently also in absolute numbers. In the inflamed kidney, suppressive MDSC expressed the IFNγRβ in addition to Ly6G. In contrast, in livers of bile duct ligated mice, suppressive MDSC expressed Ly6C but did not express IFNγRβ. In contrast, BM-derived MDSC did not express IFNγRβ at all, although these cells possessed suppressive capacity. Moreover, although IFNγ was necessary to induce MDSC in the spleen, expression of the IFNγRβ receptor was not required for suppressive activity of these splenic MDSC. The IFNγRα chain binds to IFNγ and the IFNγRβ chain transduces signals to the nucleus [38]. Although the IFNγRα chain is expressed highly on the membranes of B-, T-and myeloid cells, the IFNγRβ is only highly expressed only on the membrane of myeloid cells [39]. We now show that the IFNγRβ is not generally expressed on MDSC, which may explain the contrasting findings reported in literature on the importance of IFNγ for the induction and function of MDSC [10,12,13].
In order to classify MDSC induced in a specific organ, we decided to compare them directly to a defined MDSC "standard", i.e. bone marrow derived and LPS/IFNγ induced splenic MDSC. Surprisingly, compared to BM and splenic MDSC, in the kidney and liver suppressive activity was only present in either Ly6G pos or Ly6C pos myeloid cells, respectively. We have recently reported that MDSC can be generated from monocytic cells by activated hepatic stellate cells (HSC) [36], indicating that monocytes arriving in the inflamed liver can locally be differentiated into suppressive MDSC during fibrosis when HSC become activated and differentiate into myofibroblasts. In the kidney such myofibroblasts, important for fibrosis progression, can arise from various subsets of renal cell populations, including fibroblasts, epithelial and endothelial cells [37], indicating that in different organs the cell population and their mechanisms of induction of MDSC may differ considerably.
Extensive phenotypic analysis of all myeloid cell subsets did not reveal a potential candidate for a liver MDSC specific marker. In tumours, IL4Rα expression not only correlates with suppressive function [40], but can also be used to specifically target MDSC [41]. However, in all of the conditions we tested, IL4Rα is expressed on both Ly6G pos and Ly6C pos subsets and cannot be used to specifically distinguish suppressive from non-suppressive myeloid subpopulations in kidney or liver inflammation and fibrosis. In livers of bile duct-ligated mice, B7H1 is expressed on the suppressive Ly6C pos subset. Although in the kidney, spleen and bone marrow such a correlation cannot be detected, we and others have reported that hepatic stellate cells can induce MDSC [42,43], which repress immune responses via B7H1 [31].
Gene expression analysis of a large panel of inflammatory genes revealed that both Ly6C pos and Ly6G pos MDSC were similarly regulated as their non-suppressive counterparts. For instance, Ly6C pos MDSC in the liver express high levels of CCR2 and also produced the ligand CCL2, similar to for instance Ly6C pos inflammatory monocytes ( [44]). Ly6G pos MDSC from the kidney did specifically induce the expression of several chemokines and chemokine receptors, involved in the attraction and functional modulation of granulocytes. This may reflect that MDSC in the kidney and liver utilize different signalling modes for further attraction of myeloid cells, which may then also be modulated into becoming MDSC.
Although we could not identify a general marker for MDSC, specific markers for MDSC, like the IFNγRβ in the kidney, may exist in specific disease entities. Thus, our study shows that during chronic inflammation leading to organ fibrosis in the liver and kidney myeloid-derived suppressor cells are efficiently induced capable of inhibiting potentially harmful immune responses.  20.000 cells were analysed. Average gene-expression of mean-centered data is shown for hepatic CD11b + Ly6C + (n = 3), CD11b + Ly6G + (n = 3) and renal CD11b + Ly6C + (n = 1) and CD11b +Ly6G+ (n = 2) myeloid cells. High and low expression of individual genes is indicated by a colour code.