CD147 mediates intrahepatic leukocyte aggregation and determines the extent of liver injury

Background Chronic inflammation is the driver of liver injury and results in progressive fibrosis and eventual cirrhosis with consequences including both liver failure and liver cancer. We have previously described increased expression of the highly multifunctional glycoprotein CD147 in liver injury. This work describes a novel role of CD147 in liver inflammation and the importance of leukocyte aggregates in determining the extent of liver injury. Methods Non-diseased, progressive injury, and cirrhotic liver from humans and mice were examined using a mAb targeting CD147. Inflammatory cell subsets were assessed by multiparameter flow cytometry. Results In liver injury, we observe abundant, intrahepatic leukocyte clusters defined as ≥5 adjacent CD45+ cells which we have termed “leukocyte aggregates”. We have shown that these leukocyte aggregates have a significant effect in determining the extent of liver injury. If CD147 is blocked in vivo, these leukocyte aggregates diminish in size and number, together with a marked significant reduction in liver injury including fibrosis. This is accompanied by no change in overall intrahepatic leukocyte numbers. Further, blocking of aggregation formation occurs prior to an appreciable increase in inflammatory markers or fibrosis. Additionally, there were no observed, “off-target” or unpredicted effects in targeting CD147. Conclusion CD147 mediates leukocyte aggregation which is associated with the development of liver injury. This is not a secondary effect, but a cause of injury as aggregate formation proceeds other markers of injury. Leukocyte aggregation has been previously described in inflammation dating back over many decades. Here we demonstrate that leukocyte aggregates determine the extent of liver injury.


Results
In liver injury, we observe abundant, intrahepatic leukocyte clusters defined as �5 adjacent CD45 + cells which we have termed "leukocyte aggregates". We have shown that these leukocyte aggregates have a significant effect in determining the extent of liver injury. If CD147 is blocked in vivo, these leukocyte aggregates diminish in size and number, together with a marked significant reduction in liver injury including fibrosis. This is accompanied by no change in overall intrahepatic leukocyte numbers. Further, blocking of aggregation formation occurs prior to an appreciable increase in inflammatory markers or fibrosis. Additionally, there were no observed, "off-target" or unpredicted effects in targeting CD147. PLOS

Introduction
The classical hallmark of liver injury is the deposition of abnormal/fibrotic extracellular matrix (ECM) and the eventual development of cirrhosis, which is mediated by the activated hepatic stellate cell (HSC). Chronic inflammation drives ongoing HSC activation and fibrosis [1]. Chronic liver inflammation can be regarded as commencing with an initial innate immune response and progressing with an ongoing insult, into a chronic injury with both sustained innate and adaptive immune components [2][3][4][5][6]. T-cell responses are clearly pivotal to the development of chronic immune-mediated hepatic injury and it has now been shown that Bcells are essential for the development of intrahepatic fibrosis leading to cirrhosis [7][8][9][10][11]. Importantly, we have previously identified a number of novel pathways of liver injury [12][13][14]. Arising from these studies, we have shown that hepatocytes remodel extracellular matrix in liver injury via the production of active matrix-metalloproteinases [15].
Leukocyte aggregations of either lymphoid or myeloid origin are thought to exacerbate inflammation and in turn, the severity of disease. In acute viral infection, TLR stimulation and TNF signalling induced myeloid cell aggregation in the liver [51]. These myeloid aggregates induced localized CD8+ T cell proliferation. Myeloid cell aggregation was not observed in chronic viral infection. It is yet to be determined if immune cell aggregation occurs in nonviral induced liver injury. The molecules associated with immune aggregation are also unknown. Aggregation of leukocytes within the liver tissue during disease has been most observed in Hepatitis C Virus (HCV) infection and alcoholic liver disease (ALD), though their role in pathogenesis and mode of development is still unclear [52]. We have previously shown that CD147 is increased in cirrhotic liver and expressed by hepatocytes (not HSC) during progressive liver injury [15]. Herein, we report CD147 upregulation on multiple immune cell subsets during liver injury. This manuscript is the first description of a novel CD147-mediated mechanism of immune cell aggregation in liver injury. Specifically, we have shown that CD147-mediated immune cell aggregation is independent of cell proliferation and anti-CD147 intervention reduces immune cell aggregation.

Human ethics
Human tissue samples were obtained from Royal Prince Alfred Hospital, Sydney with approval of the Human Research Ethics Committee (Proposal: X10-0072). Human tissue used in this study was previously utilized for research [12,15]. Informed written consent was obtained from all participants. The ethics committee waived the need for written consent for use of donor tissue. In Australia, the ethics of human research is governed by the National Statement on Ethical Conduct in Human Research (2007) issued by the National Health and Medical Research Council (NHMRC). Under these guidelines, all research involving humans requires ethical approval. Non-diseased donor (n = 4) and end-stage cirrhotic liver tissues (n = 15) were collected from patients attending Prince Alfred Hospital, Sydney during liver transplantation.

Mice
Eight-week old male, wild-type mice (BALB/c and C57Bl/6) (n = 5-8 per group) were sourced from Animal Resources Centre (Canning Vale, WA, Australia) and housed at the Centenary Institute Animal facility (Camperdown, NSW, Australia) in compliance with the Animal Care and Ethics Committee at the University of Sydney (Prospectively approved protocol:K75/10-2008/3/4801). Liver injury was induced by either thioacetamide (TAA) (300mg/L) (MP Biomedicals, Ohio, USA) in drinking water ad libitum for four, eight or twenty weeks, or by carbon tetrachloride (CCl 4 ) (Univar, Ajax Chemicals, Sydney) via intraperitoneal injection (i.p) with 100μl of 12% CCl 4 in paraffin oil mixture once (one day) or twice weekly for four weeks. Control mice were injected with paraffin oil alone (PO). Both treated and age-matched control mice were sacrificed by CO 2 asphyxiation at conclusion of treatment. CD147 antibody was administered by i.p injection twice weekly (100μg) and control mice received IgG2a (100μg) (HB-189, ATCC). The rat anti-mouse IgG2a CD147 blocking antibody (mAb RL73.2) was produced by hybridoma cells and purified as described [53].
For peripheral blood mononuclear cell (PBMC) isolation, peripheral blood from the inferior vena cava was collected into 500μl of ice-cold Alsever's solution (Sigma-Aldrich) prior to erythrocyte removal, washing and resuspension to form a single cell suspension for flow cytometric analyses. Further aliquots were pelleted by centrifugation for snap frozen storage at -80˚C prior to downstream RNA extraction. Similarly, for splenocyte isolation, whole tissue was mechanically disrupted and passed through a 50μm sieve, prior to erythrocyte removal, washing and resuspension to form a single cell suspension for flow cytometric analyses.

Liver function tests
Blood was collected from the inferior vena cava into MiniCollect Serum Tubes (Greiner Bio-One, Kremsmünster, Germany). Serum was isolated by centrifugation at 3000 relative centrifugal force (RCF) for 10 minutes and the supernatant was collected. Serum was diluted 1:3 in PBS and tested for the activity of enzymes alkaline phosphatase (ALP), alanine transaminase (ALT) and aspartate transaminase (AST) by the Sydney South West Pathology Service. All results are measured in international units per litre (U/L).

Cell cluster quantitation
Liver sections stained for CD45 by immunofluorescence were viewed using Leica DM RBE Fluorescence Microscope (Leica Microsystems) and photographed using Leica DFC 500 camera (Leica Microsystems) in random fields of view (FOV) (n = 3-10). Fluorescent cells were quantified blind using Image-J (National Institute of Mental Health, MD, USA). Cell clusters were defined as greater than five cells in contact.

Statistics
Statistical analysis was performed using Prism 6 (GraphPad, San Diego, USA). Mann-Whitney U t-test was performed to compare against control unless otherwise specified. Significance was accepted at p<0.05. Data are shown as mean ± SEM. As indicated some data are expressed as fold change from control.

CD147 gene expression increases in the liver during injury
We have previously observed higher CD147 expression in primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC) and hepatitis C virus (HCV)-induced liver injury in patients [12,55]. We chose to investigate the expression of CD147 in a TAA-induced mouse model of liver injury to further establish the source and function of CD147 in liver injury. During the development of TAA-induced liver injury, we observed a progressive increase in the expression of CD147 mRNA in whole liver tissue (7-fold) ( Fig 1A). When hepatocytes and leukocytes were analysed separately, a significant increase in CD147 mRNA expression was observed in leukocytes (30-fold) ( Fig 1B) and minimal CD147 upregulation was observed on hepatocytes (2-fold) ( Fig 1B). This data shows that CD147 expression is upregulated in a mouse model of liver injury and has predominantly increased expression on leukocytes. This demonstrates that while hepatocytes are the dominant cell type in liver tissue, it is the less abundant leukocytes that have the greatest contribution to CD147 upregulation in whole liver, seen by the larger fold-change in CD147 expression on leukocytes compared to hepatocytes (30-fold vs. 2-fold for leukocytes and hepatocytes, respectively). Further, we examined leukocyte CD147 expression in blood, liver and spleen, and demonstrated an increase from control to 20 weeks in circulating and intrahepatic leukocytes (Fig 1C).

CD147 surface expression increases on liver leukocyte subsets during injury
We next characterized the expression of CD147 on various leukocyte subsets during the course of liver injury (Fig 2). Eight colour flow cytometry was utilized to gate liver CD4 + T cells, CD8 + T cells, B cells, NK cells, NKT cells, eosinophils, neutrophils and, macrophages. In untreated controls, leukocytes expressed minimal CD147 with the exception of macrophages, neutrophils and eosinophils, which all had high basal levels of CD147. CD147 expression increased on all liver leukocytes, with peak expression at 8 weeks of injury. Interestingly, as liver injury progressed, we observed a significant loss of NKT cells and macrophages, and a significant increase of CD8 + T cells, eosinophils and neutrophils from the liver.
To establish whether similar CD147 expression patterns were occurring in an extrahepatic lymphoid organ, the spleen was also analysed by flow cytometry in the same manner as

CD147 surface expression on leukocyte subsets isolated from livers in progressive liver injury.
Flow cytometry was performed on liver leukocytes from control and TAA-treated mice. Each leukocyte subset was analysed for expression of CD147 and data recorded as the percentage of cells positive for CD147 (%CD147+) or mean fluorescence intensity (MFI). Representative histogram plots are shown for each subset, where the black histogram represents the indicated liver subset and grey histogram indicates isotype control. Bar graphs represent mean ± SEM Mann-Whitney test was performed to assess significance from control where � p<0.05, �� p<0.01, ��� p<0.001 (n = 5-8 per group).
https://doi.org/10.1371/journal.pone.0215557.g002    Donor PSC ETOH AIH HCV Donor PSC ETOH AIH HCV described above. The observed trend was true for all sub-populations examined, with the exception of NK cells, which showed equally high expression at both 8 and 20 weeks. Interestingly, some splenic sub-populations showed a similar trend, however, CD147 expression was not nearly as high as in the liver and no values were significantly up-regulated compared to untreated mice. Within each treatment group, CD147 expression levels were relatively similar amongst all leukocyte subsets, with the exception of macrophages, which displayed expression levels up to 6-fold higher than any other cell type. B cells and NK cells showed the lowest relative expression of CD147, displaying levels half that of the T cell subsets (Fig 2).

Aggregation of lymphocytes during liver injury in mouse models
To determine how lymphocytes localized during chronic liver injury, CD45 staining was conducted on a range of injury models including, TAA treatment in C57Bl/6, CCl 4 in C57Bl/6 and CCl 4 in BALB/c mice (Fig 3). In uninjured control livers, lymphocytes dispersed evenly within the liver tissue (Fig 3A). In contrast, in both TAA ( Fig 3B) and CCl 4 ( Fig 3C) injury models, not only were total CD45 + cells numbers elevated (Fig 3D) but CD45 + cells were found to aggregate (defined as >5 CD45 + cells in contact). Importantly the majority of aggregation occurred around portal triads where fibrosis first appears. Peak infiltrate was observed after 8 weeks of TAA and 4 weeks of CCl 4 , which correlated with peak immune cell aggregation (Fig 3E). Thus, injury is shown to increase immune cell infiltrate and immune cell aggregation. Apoptosis and proliferation within aggregates were assessed by cleaved caspase-3 and Ki-67, respectively (S1 Fig). This was assessed in both normal and injured liver tissue. In C57Bl/6 mice, after 4 weeks of CCl 4 administration compared to untreated animals, apoptosis went from being most undetected to occasionally seen, which was a significant increase (p = 0.05), but no significant change in proliferation was observed (p = 0.08). This is consistent with other publications [56,57].

Aggregation of lymphocytes in injured human liver
To confirm that immune cell expression of CD147 and immune cell aggregation was not exclusive to mouse models we examined human liver tissue from healthy donors ( Fig 4A) and patients with PSC, alcohol-induced liver damage (EtOH), autoimmune hepatitis (AIH) and HCV (Fig 4B). Interestingly, we observed CD147 expression on hepatocytes and all CD45+ immune cells in both healthy and injured livers. Total CD45 + immune cells were significantly increased in diseased liver tissue, irrespective of whether the disease was immune-mediated (AIH, PSC, HCV) or not (EtOH) (Fig 4C). Inflammatory cells were shown to aggregate with liver injury (Fig 4B) and significantly more immune aggregates were observed in diseased livers compared to healthy donor livers (Fig 4D). In healthy donors, although CD45 + cells expressed some level of CD147, immune cells did not aggregate. Therefore, this data confirms that in diseased human liver, CD147 is expressed on CD45 + immune cell infiltrate and immune cells aggregate during disease.

CD147-dependent leukocyte aggregation in liver injury
To determine whether CD147 was important for the formation of immune cell aggregates, the effects of anti-CD147 intervention were studied in CCl 4 induced liver injury models in both Representative images of liver sections stained with CD45 (red) and CD147 (green) from (A) healthy control and (B) diseased patient liver (autoimmune hepatitis) on either a 20x (left panels) or 63x (right panels) objective. Yellow indicates co-expression of CD147 and CD45. Scale bar represents 100μM. Quantification of (C) total CD45+ cells and (D) immune aggregates in healthy, PSC, EtOH, AIH and HCV livers. Bars represent mean+SEM. Unpaired t-test was performed to assess significance from control where � p<0.05 and �� p<0.01 (n = 4-15 per group). Donor = healthy control liver, PSC = primary sclerosing cholangitis, ETOH = alcohol-induced liver damage, AIH = Autoimmune hepatitis, HCV = hepatitis C virus liver injury. C57Bl/6 and BALB/c mice. Anti-CD147 did not significantly reduce the number of CD45 + cells in the liver (Fig 5E and 5F). The percentage of immune cells that formed immune cell aggregates was reduced with anti-CD147 intervention in both mouse backgrounds (Fig 5G  and 5H). In C57Bl/6 mice, there was a non-significant (<1.5-fold) increase in ALT with liver injury and no significant change was seen with anti-CD147 intervention ( Fig 5I). However, in BALB/c mice, which are known to be more sensitive than C57Bl/6 mice to hepatotoxins such as CCl 4 [58], a significant increase in ALT was observed in liver injury compared to control and anti-CD147 intervention significantly reduced serum ALT levels ( Fig 5J). Thus, CD147 inhibition appeared to significantly reduce the formation of immune cell aggregates and reduce liver injury.

Specific leukocyte subsets within aggregates
We have just shown that anti-CD147 intervention reduces the percentage of immune cells in aggregates. To determine what inflammatory cells were aggregating in CCl 4 induced liver injury and to determine whether this is altered during anti-CD147 intervention, mouse liver samples were immunostained with CD45, F4/80, B220, Gr1 and CD3. F4/80 + Macrophages, B220 + B cells, Gr1 + granulocytes and CD3 + T cells (including NKT cells) were all found in aggregates during liver injury (Fig 6). After anti-CD147 treatment, we observed a decrease in the number of aggregates with at least one F4/80 + , B220 + and CD3 + cell. The number of Gr1 + cells in aggregates was not altered after anti-CD147 treatment (Fig 6). Thus, anti-CD147 intervention was shown to target specific immune cell subsets and reduce immune cell aggregation.

Discussion
This study has demonstrated that with progressive inflammation-associated tissue injury, immune cells aggregate and directly contribute to the severity of the injury. Our novel discovery is that with liver injury, CD147 is principally upregulated on the surface of leukocytes and mediates cell-cell aggregation that determines the extent of liver injury. We have also shown that if CD147 is blocked with a mAb, then the numbers of infiltrating CD45 + cells in the liver remain unchanged but leukocytes are no longer found in aggregates. Importantly, we have already reported that there is a significant reduction in liver injury seen with anti-CD147 mAb [15]. Further, CD147-mediated leukocyte aggregation appears to cause (or significantly exacerbate) injury as aggregate formation proceeds the development of significant fibrosis [15]. Therefore, this is not just a reduction in aggregation and inflammatory markers (AST/ALT) but also a reduction in resultant fibrosis. All intrahepatic leukocyte subpopulations (CD4 + , CD8 + , NK, B-cell and macrophages) rapidly increase CD147 surface protein and total mRNA expression during liver injury. Therefore, this data shows that following liver injury, circulating and intrahepatic leukocytes increase Anti-CD147 antibody intervention in CCl 4 induced liver injury in C57Bl/6 (left panels) or BALB/c (right panels) mice. CD45+ leukocytes were stained in frozen tissue sections in control, treatment (CCl 4 ), treatment with anti-CD147 antibody intervention, or treatment with IgG2a isotype antibody controls. Representative immunofluorescent images of CD45 + clusters defined as �5 cells per cluster (white circles) are shown in C57Bl/6 (A) and BALB/c mice (B) treated with IgG2a. Representative immunofluorescence images of CD45 + clusters following αCD147 intervention in C57Bl/6 (C) and BALB/c (D) mice. Total CD45 + leukocytes in C57Bl/6 (E) and BALB/c (F) mice. (G) Percentage of CD45 + cells in aggregates for each treatment group in C57Bl/6 (G) and BALB/c (H) mice. ALT levels (U/L) as a measure of liver damage in C57Bl/6 (I) and BALB/c (J) mice, calculated as a fold change from control. Bar graphs represent mean ± SEM. Mann-Whitney test was performed to assess significance from control where � p<0.05, �� p<0.01, ��� p<0.001 (n = 5-8 per group). Control = untreated, CCl 4 injury = injury alone, αCD147 = anti-CD147 mAb in CCl 4 injury, IgG2a = Isotype mAb control in CCl 4 injury.  CD147 expression. Subsequently, it has been shown leukocytes undergo activation [59] and form aggregates that contribute to injury. It is unknown if leukocyte activation occurs concurrent or prior to aggregation. The recruitment of inflammatory cells from the periphery to sites of liver injury is important in the pathogenesis of liver injury [59] but its relationship to aggregation and dependence on CD147 expression was unrecognised. Based on these results, CD147 is likely pivotal to the development of inflammation in liver injury, partly through the previously unrecognized role in leukocyte aggregate formation.
In liver injury, anti-CD147 intervention leads to a reduction in immune cell aggregation, characterised by a reduction in serum transaminases and total tissue MMP activity [15]. Importantly, there is no significant change in the total number of CD45 + cells infiltrating the liver. This suggests that the cells are not proliferating to form clusters but rather they aggregate with injury. Further, we have shown that the clusters that diminish with anti-CD147 intervention in injury contain B-cells (B220) and macrophages/Kupffer cells (F4/80), whereas the isotype antibody controls have the same phenotype as wild-type animals. Importantly, we have shown these leukocyte clusters form in progressive human liver injury, irrespective of disease etiology.
Supporting information S1 Fig. Apoptosis and proliferation within immune cell aggregates. In CCl 4 treated C57Bl/6 mice and healthy controls, liver tissue was stained with cleaved caspase-3 and Ki-67 to assess apoptosis and proliferation within aggregates, respectively. (A) Representative images of apoptosis (left panels) and proliferation (right panels) from healthy control mice (top panels) and CCl 4 treated mice (bottom panels). (B) The number of positive events per frame where bar graphs represent mean ± SEM Mann-Whitney test was performed to assess significance from control. (n = 5-7 per group). (EPS)

Fig 6. Anti-CD147 intervention reduces aggregation of macrophages, B cells, T cells and granulocytes during liver injury.
CCl 4 treated C57Bl/6 mice liver tissue was stained for CD45 colocalisation with (A) F4/80, (B) B220, (C) Gr1 or (D) CD3, to identify immune subset specific aggregates. Representative images (left panel) and quantified immune cell-specific aggregates (right panel; calculated as at least one cell of interest per aggregate) in mice treated with CCl 4 alone or CCl 4 in conjunction with anti-CD147 intervention. Bar graphs represent mean ± SEM. Mann-Whitney test was performed to assess significance from control where � p<0.05, �� p<0.01 (n = 5-7 per group). CCl 4 = injury alone, αCD147 = anti-CD147 in CCl 4 injury.