ICOS deficient (ICOS^-/- ; strain B6.129P2-Icos^tm1Mak/J) and ICOSL deficient (ICOSL^-/- ; B6.129P2-Icosl^tm1Mak/J) mice in C57BL/6 background were obtained from The Jackson Laboratories (Bar Harbor, Maine, USA). C57BL/6 wild-type (WT), ICOS^-/- and ICOSL^-/- mice were housed in pathogen-free conditions and fed ad libitum with standard chow diet and water. Liver injury was induced by injecting intra-peritoneally eight-week-old male mice with carbon tetrachloride (CCl⁴) (0.6 ml/kg in olive oil). Control animals received an injection with olive oil alone. Murine recombinant ICOS bound to the human IgG1 Fc portion (ICOS-Fc) was prepared as previously described (17). ICOS-Fc (100 µg in sterile saline) was administered to mice by intraperitoneal injection 24 hours after CCl₄ and then every 12 hours up to 48 hours. Control animals received a similar amount of saline alone. Hepatic CD8⁺ T cell ablation was obtained by injecting WT mice with rat IgG2b anti-mouse CD8β monoclonal antibody (clone YTS 156.7.7). The animals received YTS 156.7.7 mAb (100 µg) three days before and immediately after CCl₄ challenge. Control animals received the same amount of isotype matched IgGs. YTS 156.7.7 anti-CD8β monoclonal antibody was a kind gift by Dr. Stephen Cobbold, William Dunn School of Pathology, University of Oxford, (Oxford, UK). All animals were euthanized 24-72 hours after CCl₄ administration.

Livers were rapidly removed, rinsed in ice-cold saline, and cut into five pieces. Aliquots were immediately frozen in liquid nitrogen and kept at −80°C until analysis. Two portions of the left lobe from each liver were fixed in 10% formalin for 24h and embedded in paraffin. 4 µM tick liver sections were stained with hematoxylin/eosin using a Roche Ventana HE 600 automatic staining system (Roche Diagnostics International AG, Rotkreuz, Switzerland) and microphotographs were taken using a Nikon Eclips CI microscope fitted with a DSR12 camera (Nikon Europe BV, Amsterdam, Netherlands) using the NIS-Elements F4.60.00 acquisition software. The extension of necrotic areas was assessed morphologically in hematoxylin/eosin-stained liver sections using the ImageJ 1.53e software (National Institute of Health, Bethesda, MD, USA). The immunostaining for the proliferation marker Ki67 and ICOSL was performed in paraffin fixed section using anti-Ki67 monoclonal antibody (clone 30-9) and anti-ICOSL goat polyclonal antibodies (CD275 cod PA5-47161) (Thermo Fisher Scientific, Milano, Italy) in an automated staining system (BenchMark ULTRA IHC/ISH System, Roche Diagnostics International AG, Rotkreuz, Switzerland). In some experiments the staining with anti-ICOSL antibodies was also evidenced in frozen liver sections by immunofluorescence using Alexafluor^®-546 donkey anti-goat IgG (H+L) (Thermo Fisher Scientific, Milano, Italy) and Leica DM 5500B fluorescence microscope and Leica Application Suite X (Leica Microsystems (Buccinasco, Italy). Plasma ALT levels were determined by a spectrometric kit supplied by Gesan Production SRL (Campobello di Mazara, Italy).

Livers were digested by type IV collagenase from Clostridium histolyticum (Sigma-Aldrich, St. Louis, MO, USA), and intrahepatic leukocytes were isolated by multiple differential centrifugation steps according to (19). The cell preparations were then subjected to red cell lysis by RBC Lysis Buffer (eBioscience, Thermo Fisher Scientific, Milano, Italy) and stained using combinations of the following monoclonal antibodies: CD45 (Clone 30-F11, Cat. 12-0451-82), CD3 (Clone 17A2, Cat. 17-0032-82) CD4 (Clone GK1.5, Cat. N. 56-0041-80) CD8 (Clone 53-6.7, Cat. 11-0081-82), Ly6C (Clone HK1.4, Cat. 53-59-32-80), NK1.1 (Clone PK136, Cat. 12-5941-81), Ly6G (Clone RB6-8C5, Cat. 47-5931-82), MHCII (Clone M5/114.15.2, Cat. 56-5321-80), CD103 (Clone 2E7, Cat. 12-1031-81), CD69 (Clone H1.2F3, Cat. 12-0691-82), CD107a (Clone eBio1D4B, Cat. N. 12-1071-81), CD206 (Clone MR6F3, Cat. 25-2061-80) MerTK (Clone DS5MMER, Cat. N. 56-5751-80 eBioscience, Thermo Fisher Scientific, Milano, Italy), CD11b (Clone M1/70, Cat. 101212), ICOS (Clone 15F9, Cat. 107705), ICOSL (Clone HK5.3, Cat. 107405), F4-80 (Clone BM8, Cat. 123113, Biolegend, San Diego, CA, USA), TREM-2 (Clone 78.18, Cat. MA5-28223, Thermo Fisher Scientific, Milano, Italy). In some experiments, macrophage viability was evaluated by cell staining using the rh-annexin V/FITC kit (Bender Med Systems, Vienna, Austria). Sample analysis was performed using the Attune NxT flow-cytometer (Thermo Fischer Scientific, Waltham, MA, USA) and data were elaborated with FlowJo™ Software (BD Biosciences, San Jose, CA, USA).

mRNA was extracted from snap-frozen liver fragments using the TRIzol^® Reagent (Thermo Fischer Scientific, Milano, Italy) as previously reported (20). cDNA was generated from 1 µg of mRNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems Italia, Monza, Italy) in a Techne TC-312 thermocycler (TecneInc, Burlington NJ, USA). Real-Time PCR was performed in a CFX96™ Real-time PCR System (Bio-Rad, Hercules, California, USA) using TaqMan Gene Expression Master Mix and TaqMan Gene Expression probes for mouse TNF-α (Mm99999068_m1), CCL2 (Mm00441242_m1), CD11b (Mm00434455_m1), TREM-1 (Mm01278455_m1), CD206 (Mm01329362_m1), TREM-2 (Mm04209422_m1), MerTK (Mm00434920_m1), CX₃CR1 (Mm00438354_m1) and beta-actin (Cat. N 4352663, Applied Biosystems Italia, Monza, Italy). All samples were run in duplicate, and the relative gene expression was calculated as 2^-ΔCt over that of β-actin gene.

Statistical analyses were performed by SPSS statistical software (SPSS Inc. Chicago IL, USA) using one-way ANOVA test with Tukey’s correction for multiple comparisons or Kruskal-Wallis test for non-parametric values. Significance was taken at the 5% level. Normality distribution was assessed by the Kolmogorov-Smirnov algorithm.

According to previous studies (20, 21), acute liver injury caused by mice poisoning with the hepatotoxic agent carbon tetrachloride (CCl₄) leads to centrilobular necrosis and an extensive inflammatory response which peaked after about 24 hours and rapidly declined thereafter with an almost complete recovery after 72 hours ( Figures 1A, B ). Beside liver infiltration by phagocytes, hepatic inflammation also involved the recruitment of both CD4⁺ helper and CD8⁺ cytotoxic T-lymphocytes ( Figure 1C ). Flow cytometry analysis showed that liver infiltrating T-lymphocytes expressed ICOS and that the overall prevalence of CD3⁺/ICOS⁺ cells followed the evolution of the inflammatory process, being maximal around 24 hours and declining in the recovery phase, mainly in relation to the lowering in CD4⁺ T-cell prevalence ( Figure 2A ). However, liver healing did not affect the prevalence of ICOS-positive CD8⁺ T-cells infiltrating the liver after 72 hours, which instead further increased as compared to early time points ( Figure 2B ). These CD8⁺ T lymphocytes comprised both CD103⁺ and CD103^- cells and expressed the activation markers CD69 and CD107a ( Supplementary Figure 1 ). In the same animals, we also observed an early up regulation in the hepatic expression of ICOSL, which mainly involved CD11b^high/F4-80⁺ MoMFs, ( Figure 2C ). The fraction of ICOSL-positive MoMFs was well appreciable at the peak of inflammation and further increased during the recovery phase ( Figure 2C ). Consistently, immunostaining with an anti-ICOSL polyclonal serum of liver sections from CCl₄-treated at 72 hours showed selective labelling of infiltrating MoMFs in centrilobular areas, while no staining was appreciable in control livers ( Supplementary Figure 2 ).

From these data, we investigated whether interfering with ICOS/ICOSL signalling might affect the evolution of damage-associated hepatic inflammation. Time course analysis of the liver transcripts for pro-inflammatory markers in wild-type (WT) and ICOS knockout (ICOS^-/- ) mice receiving CCl₄ showed the same pattern in the up-regulation and decline of tumor necrosis factor-α (TNF-α), CCL2 and the leukocyte integrin αM (CD11b) mRNAs, even though the down-modulation of these markers was more rapid in ICOS^-/- mice ( Supplementary Figure 3 ). In line with these findings, also the expression of the Triggering Receptor Expressed on Myeloid cells 1 (TREM-1), a membrane receptor involved in regulating the pro-inflammatory activity of neutrophils and macrophages (22, 23), was comparable in the two strains ( Supplementary Figure 2 ). Consistently, flow cytometry analysis of liver MoMFs did not evidence appreciable difference in the hepatic recruitment of Ly6C^high/CD11b^high/F4-80⁺ pro-inflammatory MoMFs between WT and ICOS^-/- mice 24 hours after CCl₄ administration ( Supplementary Figure 4A ). ICOS absence also did not affect the functional maturation of MoMFs with the expression of high levels of Class II Mayor Histocompatibility Complex (MHCII) ( Supplementary Figure 3A ) as well as the up regulation of the fractalkine receptor CX₃CR1 ( Supplementary Figure 4B ) (11, 21).

In line with the pattern of liver inflammation, we observed that after 24 hours from CCl₄ challenge, the extent of hepatic injury was similar in WT and ICOS^-/- mice, despite circulating alanine aminotransferase (ALT) was slightly higher in these latter ( Figure 3A ). However, while in WT mice hepatic circulating ALT appreciably declined after 48 hours from CCl₄ poisoning and liver damage almost completely recovered within 72 hours, ALT elevation persisted in ICOS^-/- mice at 48 hours ( Figure 3A ). Liver histology confirmed that ICOS^-/- mice failed to clear centrilobular necrosis during the recovery from CCl₄ intoxication ( Figures 3B, C ). Furthermore, 5 out 13 (38%) of CCl₄-treated ICOS^-/- mice died by acute liver failure between 48 and 72 hours ( Supplementary Figure 5A ). At 72 hours the ICOS^-/- mice who survived showed serum ALT about two folds higher than WT animals and still evident centrilobular necrotic areas ( Figures 3A–C ). In these animals, the immunostaining for the proliferation marker Ki67 did not evidence significant differences in the labelling of hepatocyte nuclei as compared to the wild-type littermates ( Supplementary Figure 5B ), suggesting that interfering with ICOS/ICOSL signaling does not affect hepatocyte regeneration but it might affect the responses of non-parenchymal cells.

From the observation that the absence of ICOS affected the recovery from acute hepatic injury, we investigated whether this might involve the switching of MoMF phenotype occurring during the healing response. The analysis of the transcripts for markers of reparative MoMFs such as the mannose receptor (CD206) and the efferocytosis receptor c-Met Proto-Oncogene Tyrosine Kinase (MerTK) (9), evidenced that the expression of both markers was dramatically reduced during acute inflammation and recovered to control values during liver repair ( Figure 4 ). Similarly, hepatic healing was associated with the up-regulation in the Triggering Receptor Expressed on Myeloid cells 2 (TREM-2) ( Figure 4 ), a plasma membrane receptor involved in the damping MoMF inflammatory response (24, 25). In these settings, ICOS^-/- mice were less efficient than WT animals in up-regulating CD206, MerTK and TREM-2 ( Figure 4 ). Furthermore, by evaluating the distribution of CD11b⁺/F4-80⁺ MoMFs 72 hours after CCl₄ poisoning, we observed that the prevalence of MoMFs was lowered by about 30% in the livers of ICOS^-/- mice ( Figure 5A ) and that such a loss involved the fraction of CD206/TREM-2/MerTK positive reparative MoMFs ( Figure 5B ). Along with the reduction in reparative MoMFs and the impaired clearance of necrotic areas, CCl₄ challenged ICOS^-/- mice also showed the persistence of a sustained hepatic infiltration by Ly6G^high/CD11b⁺/F4-80^- granulocytes (3.2 ± 0.7% vs 8.6 ± 2.5% of CD45⁺ cells; p<0.01), while the prevalence of other myeloid cells was not affected.

As mentioned above, during the recovery from CCl₄-induced hepatocyte injury, ICOS was prevalently expressed by liver CD8⁺ T cells, while ICOSL was up regulated in MoMFs. To explore the possibility that the interaction between CD8⁺/ICOS⁺ T cells and ICOS-L⁺ MoMFs might contribute to liver healing, we depleted liver CD8⁺ T cells using an anti-CD8 monoclonal antibody (mAb). Figure 6A shows that mice treatment with anti-CD8 mAb selectively lowered circulating and hepatic CD8⁺ T-cells. Furthermore, the overall prevalence of ICOS expressing CD3⁺ T cells within the liver was halved in the same animals 48 hours after CCl₄ (3.8 ± 0.4% vs 2.2 ± 0.6% of CD45⁺ cells; p<0.02). CD8⁺ T-cell depletion also specifically reduced the fraction of CD11b^interm/F4-80⁺/CD206^high cells that were TREM-2 and MerTK positive without affecting that of CD11b^high/F4-80⁺/CD206^low MoMFs ( Figure 6B ). Such an effect was associated with a persistence of elevated ALT release and histological evidence of a delayed liver healing in CD8⁺ T cell-depleted mice 48 hours after CCl₄ challenge ( Figure 6C ).

From the above results and the observation that CD206⁺/TREM-2⁺/MerTK⁺ MoMFs expressed ICOSL ( Supplementary Figure 6 ), we postulated that the reverse signaling triggered by the interaction of ICOS expressing CD8⁺ T-cells with ICOSL present in macrophages might influence the survival of reparative MoMFs. By labelling apoptotic cells with FITC-annexin V, we observed that the prevalence of annexin V-positive MoMFs in CCl₄-poisoned ICOS^-/- mice was more than two folds higher than in similarly treated wild-type littermates ( Figure 7A ). Interestingly, the treatment of ICOS^-/- mice with murine recombinant ICOS bound to the human IgG Fc portion (ICOS-Fc) injected 24 hours after CCl₄ completely prevented the loss of hepatic MoMFs and rescued them from apoptosis ( Figure 7A ). In particular, ICOS-Fc supplementation stimulated CD206 and MerTK transcripts and maintained the fraction of CD206⁺/TREM-2⁺/MerTK⁺ MoMFs ( Figure 7B ). The possible role of ICOSL-mediated reverse signaling in regulating the survival of reparative MoMFs was further supported by the observation that mice deficient for ICOSL (ICOSL^-/- ) showed a marked loss of MoMFs 72 hours after CCl₄ poisoning ( Figure 8A ), which associated with an extensive annexin V staining ( Figure 8B ) and an impaired recovery from acute hepatic damage ( Figure 8C ).

The observation that ICOS/ICOSL dyad was required for the survival of reparative liver MoMFs prompted us to evaluate the effects of ICOS-Fc on the evolution of CCl₄-induced acute liver damage in ICOS^-/- mice. Using the same protocol of ICOS-Fc administration effective in preventing MoMF apoptosis, we observed a significant improvement of ALT release and a reduction in the extension of necrotic areas 48 hours after CCl₄ ( Figure 9A ). All ICOS^-/- animals receiving ICOS-Fc (6 out of 6) survived CCl₄ poisoning and histological evaluation revealed an almost complete recovery of liver necrosis after 72 hours ( Figure 9B ). Consistently, at this time point circulating ALT levels in ICOS^-/- mice receiving ICOS-Fc were lower than in ICOS^-/- mice receiving saline and comparable with those in WT mice ( Figure 9B ), supporting the hypothesis that ICOS-mediated signals are required for a full reparative response of liver infiltrating macrophages.