Blood or tissue samples were collected from 155 healthy volunteers and 165 patients with HCC who underwent surgical resection between January 2014 and October 2017 at The Affiliated Drum Tower Hospital of Nanjing University Medical School (Nanjing, Jiangsu, China). The blood samples were used for isolation of peripheral monocytes or tissue-infiltrating macrophages or for immunostaining. None of the patients had undergone anticancer therapy before surgery, and individuals with concurrent autoimmune disease, HIV infection, or syphilis were excluded. The patients’ clinical characteristics were classified according to the guidelines of the Union for International Cancer Control (TNM staging). All experiments were performed in compliance with government policies and the Helsinki Declaration. The individuals were informed about the study and gave consent prior to specimen collection. The clinical features of all involved patients in this study are presented in online supplementary table 1.

Mouse (Hep1-6) HCC cell lines were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS; Gibco, USA). Human monocyte cell lines (U937, RRID:CVCL_0007 and THP-1, RRID:CVCL_0006) were maintained in RPMI 1640 medium (Invitrogen) supplemented with 10% FBS. All cell lines were purchased from the cell bank of the Chinese Academy of Sciences. All cell lines were tested for mycoplasma contamination prior to culture and use. All human cell lines have been authenticated using STR profiling.

For the macrophage stimulation, U937 and THP-1 cells were treated with 100 ng/mL phorbol myristate acetate (Sigma-Aldrich, St. Louis, MO, USA) for 2 days. The following inhibitors were purchased from Selleckchem: AZD 1480 (JAK2 inhibitor, S2162), 10058-F4 (c-MYC inhibitor, s7153), and bosutinib (c-Src inhibitor s1014). Human and mouse cytokines were purchased from Abcam (Cambridge, MA, USA), including human IL-6 (ab9627), mouse IL-6 (ab198572), human IFN-γ (ab119140), and mouse IFN-γ (ab9922). Human serum IL-6 levels were measured using a human IL-6 ELISA kit (ab46042; Abcam).

Human monocytes were obtained from healthy donors and patients with HCC by a two-step gradient centrifugation procedure using Ficoll and Percoll. Non-adherent cells were discarded, and the adherent cells were further isolated using human CD14 magnetic beads (130-050-201; Miltenyi Biotec, CA). The isolated cells were induced to form macrophages by incubation for 7 days in RPMI 1640 medium supplemented with 10% fetal calf serum and 100 ng/mL of recombinant human GM-CSF protein (ab88382; Abcam).

TAMs were isolated from liver cancer tissue specimens that were cut into small pieces and digested in collagenase IV (50 IU/mL). Dissociated cells were filtered through a 75 µm cell strainer and separated by Ficoll centrifugation, and the mononuclear cells were washed and resuspended in RPMI 1640 supplemented with 10% FBS (Gibco). The macrophages were subsequently isolated using human CD14 magnetic beads (130-050-201; Miltenyi Biotec).

Different genes and miRNAs were inserted into the lentiviral vector pLV (Clontech, CA, USA) to generate expression vectors. These expression vectors were mixed with lentiviral packaging Δ8.91 and envelope-expressing VSV-G plasmids to generate lentiviral particles in 293 T cells, as described previously.17 Viral particles were concentrated by ultracentrifugation and expression vector titers were determined. The isolated monocytes and macrophages were maintained in RPMI 1640 supplemented with 10% FBS (Gibco) for further analysis.

Total RNA was isolated with TRIzol reagent. The cDNA was synthesized from total RNA using the iScript cDNA Synthesis Kit (1708890; Bio-Rad). The miR-25–3 p was detected by synthesizing its cDNA and measuring the transcription level in triplicate with the TaqMan microRNA assay kit (cat. no. 4427975; ThermoFisher, CA; ABI 7900; Life Technologies, CA, USA). Expression levels were calculated relative to U6. The primer pairs used in this study are listed in online supplementary table 2.

Whole cell lysates were prepared as previously described.18 Equal amounts of protein were boiled, separated on 10% SDS-PAGE, transferred onto a PVDF membrane, and visualized using an enhanced chemiluminescence kit (Millipore, MA, USA). Antibodies against PTPRO (ab150834), STAT3 (Y705) (ab76315), STAT3 (S727) (ab30647), STAT3 (ab68153), STAT2 (Tyr690) (#88410; Cell Signaling), STAT2 (ab134192), STAT1 (Y701) (ab30645), STAT1 (ab31369), JAK2 (Y1007+Y1008) (ab32101), JAK2 (ab108596), c-MYC (s62) (ab51156), c-MYC (ab39688), human PD-L1 (ab205921), mouse PD-L1 (ab233482), and β-actin (ab6276) were purchased from Abcam. All uncropped western-blot figures are presented in online supplementary figures s13‒16.

Paraffin-embedded and formalin-fixed HCC samples were immunostained as previously described,19 using antibodies against human PD-L1 (ab205921), CD68 (ab955), CD3 (ab16669), TIM3 (ab242080), IL-1b (ab9722), p-JAK2 (Y1007+Y1008) (ab32101), p-c-SRC (Y418) (ab4816), STAT3 (ab68153), and p-c-myc (s62) (ab51156), followed by staining with Alexa-Fluor-488–conjugated anti-mouse IgG (1:500, ab150117), Alexa-Fluor-555–conjugated anti-rabbit IgG (1:1000, ab150074), and Alexa-Fluor-647–conjugated anti-rabbit IgG (1:500, ab150075) (Abcam) antibodies. Positive cells were detected by confocal microscopy (Zeiss, Oberkochen, Germany) and quantified using Image-Pro Plus software (Media Cybernetics, MD, USA).

Macrophages isolated from wild-type (WT) or Ptpro KO mice were cocultured as a mixture with T cells (Cd3) isolated from WT mice and a mouse Hep1-6 HCC cell line at a ratio of 1:1:5 in a 3D petri dish (Micro-Tissues, RI, USA). The cells were treated differently and collected after 4 days. The macrophages and T cells were isolated from microtissues using Ficoll and further analyzed by flow cytometry and western blotting.

Human peripheral monocytes and TAMs were isolated using anti-CD14 or anti-CD68 magnetic beads and stimulated with lipopolysaccharide (LPS) for 4 hours, followed by flow cytometry. The mouse macrophages were isolated as described previously20 and treated with LPS for 4 hours, followed by flow cytometry. Cells were stained with surface markers, fixed and permeabilized with IntraPre reagent (Beckman Coulter, Fullerton, CA), and finally stained with intracellular markers. Data were acquired on a FACSVantage SE instrument and analyzed with CellQuest software. The fluorochrome-conjugated antibodies used are described in online supplementary table 3.

Human peripheral monocytes were isolated using anti-CD14 magnetic bead from three patients with HCC and healthy controls, respectively. Small RNA library sequencing was performed on the Illumina Hiseq 2500/4000 by Gene Denovo Biotechnology (Guangzhou, China). The raw data were submitted to ArrayExpress (accession code E-MTAB-8998).

The wide type and c-MYC binding site mutated promoter of miR-106b-25 cluster, 3′UTR of PTPRO containing the WT or mutant potential target site for miR-25–3 p was synthesized by Genescript (Nanjing, China) and inserted into the pGL4.10[luc2] vector (Promega, WI). For luciferase assays, stimulated U937 and THP-1 cells were co-transfected with pGL4-miR-25-PRO WT, and pGL4-miR-25-PRO MU, pGL4-PTPRO-WT 3′UTR, pGL4-PTPRO FL-WT 3′UTR or pGL4-PTPRO FL-MU 3′UTR, pGL4-PTPRO-MU 3′UTR, with miR-25–3 p mimics, or controls (GenePharma, China) using Lipofectamine 2000 (ThermoFisher, CA). Cells were harvested 48 hours after transfection and subjected to luciferase assays using a Dual-Luciferase Reporter Assay System (Promega, WI).

An orthotopic mouse model of liver cancer was established as previously described, using Hep1-6 cells.21 One week before transplantation, adoptive cell transfer was performed, and this transfer was repeated once a week after orthotopic transplantation. A cell mixture containing 5×10⁶ macrophages transduced with lentiviral particles was transplanted into six WT mice via the tail vein. The tumor volume was calculated (length×width²×0.5). All mice were killed 5 weeks after transplantation, and organs were collected for further analysis.

Data are presented as the mean±SEM. The χ² test and Student’s t-test analysis of variance were used to evaluate differences in demographic and clinical characteristics. Kaplan-Meier survival curves were generated, and the log-rank test was used to investigate the significance of various variables for survival. All the expression experiments conducted in vitro were repeated at least three times with triplicate samples. Pearson’s correlation analysis was used to analyze the relationship of associated factors. Statistical analysis was performed using STATA V.9.2 and presented with the GraphPad Prism (CA, USA). In all cases, p value <0.05 was considered significant.

We studied the expression of PTPRO and PD-L1 expression in monocytes isolated from the peripheral blood using magnetic beads labeled with anti-CD14 antibody. Significantly decreased expression PTPRO but increased PD-L1 transcription levels revealed in monocytes from patients with HCC when compared with monocytes from control subjects (figure 1A,B). Linear correlation analysis revealed negative correlations between PTPRO expression and PD-L1 in human HCC monocytes (PTPRO vs PD-L1, r=−0.4934, p<0.0001; figure 1C). We also found that the level of monocyte PTPRO was significantly associated with patient’s age, hepatitis B virus infection, tumor size and tumor number (online supplementary table 1).

We had follow-up data for the 5-year survival for 129 of the 165 patients with HCC. The patients with HCC were subdivided into high and low expression subgroups according to the gene transcription medians of PD-L1 and PTPRO in the monocytes. Kaplan-Meier curves revealed that patients with either PD-L1^High or PTPRO^Low had significantly lower post-surgery overall survival (OS) and relapse-free survival (RFS) when compared with PD-L1^Low or PTPRO^High patients (figure 1D,E) (Monocyte PD-L1^High vs Monocyte PD-L1^Low: p=0.0018, HR 2.144, 95% CI 1.341 to 3.426 for OS; p=0.0069, HR 2.099, 95% CI 1.209 to 3.645 for RFS; Monocyte PTPRO^High vs Monocyte PTPRO^Low: p=0.0017, HR 2.162, 95% CI 1.348 to 3.467 for OS; p<0.0001, HR 3.332, 95% CI 1.911 to 5.811 for RFS, by the log-rank test). Multiple-variant analysis revealed that Monocyte PD-L1^Low+Monocyte PTPRO^High patients had significantly higher OS and RFS compared with Monocyte PD-L1^High+Monocyte PTPRO^Low patients (figure 1F) (Monocyte PD-L1^Low+Monocyte PTPRO^High vs Monocyte PD-L1^High+Monocyte PTPRO^Low p=0.0002, HR 3.019, 95% CI 1.348 to 1.775 for OS; and p=0.0010, HR 3.018, 95% CI 1.592 to 5.719 for RFS). These findings indicated that monocyte PTPRO and PD-L1 expressions are valuable indicators for predicting the prognosis of patients with HCC. However, the underlying mechanism requires further exploration.

The PD-L1 expression in HCC-infiltrating macrophages of the PTPRO^Low and PTPRO^High groups (n=20 for each group) was examined in macrophages extracted from fresh tumor tissues and further purified with CD 68 magnetic beads. We first used western blotting to compare the expression of PD-L1 in the macrophages from the PTPRO^Low and PTPRO^High expression groups (n=20), and we found a significantly higher expression of PD-L1 in the PTPRO^Low group (figure 2A). The expressions of PD-L1 in the macrophages were also detected by flow cytometry following stimulation with brefeldin A (10 µg/mL; 5 hours at 37°C) (n=50 for each group). The results were similar to those obtained with western blotting, indicating a higher number of PD-L1–positive macrophages (CD68) in the PTPRO^Low group than in the PTPRO^High group (figure 2B).

We then investigated the relationship between PD-L1–expressing macrophages and exhausted tumor-infiltrating T cells (TILs) (n=50 for each group). Significantly more PD-L1–positive macrophages were detected in the tumor tissues of the PTPRO^Low patients, and this was positively related to significantly more Tim3-positive (one of the markers for T-cell exhaustion) TILs (r=0.606, p<0.0001) (figure 2C–F). These results suggested an association between monocyte/macrophage PTPRO expression and immunosuppression in human HCC.

We have reported that PTPRO can suppress hepatocarcinogenesis in mouse hepatocytes via suppression of Jak2/Stat3 and c-Src/Stat3 activation.14 We also previously reported that deletion of PTPRO truncated (PTPROt) decreased the quantity and quality of cytotoxic T lymphocytes (CTLs) in a mouse HCC model.22 Therefore, we doubted that the deletion of PTPRO in macrophages would suppress CTLs via PD-L1–induced T-cell exhaustion. We therefore analyzed PD-L1 expression in mouse HCC tissues in the present study, and we found a significant enhancement of PD-L1 in HCC tissues in PTPRO KO mice when compared with WT control, and especially in TAMs (figure 2G,H). We also found a significant increase in the percentage of exhausted T cells (Tim3 positive) in HCC tissues from PTPRO KO mice when compared with WT controls (figure 2G‒I). Moreover, the percentage of exhausted T cells was positively correlated to the percentage of Pd-L1 positive macrophage (r=0.756, p=0.0004) (figure 2J). We excluded the effects caused by varied PD-L1 expression due to different HCC tissues by constructing a 3D culture system that included the mouse Hep1-6 HCC cell line, macrophages isolated from PTPRO KO and WT control mice, and T cells isolated from WT mice. We found that deletion of PTPRO in the macrophages could increase T-cell exhaustion by enhanced production of PD-L1, and that the T-cell exhaustion could be significantly restored by treatment with PD-1 antibody (figure 2K,L). We found that decreases in PTPRO could promote PD-L1 expression in macrophages/monocytes and induce T-cell exhaustion in both human and mouse HCC.

We explored the molecular mechanism underlying PTPRO effects by overexpression and knockdown of PTPRO in THP-1–derived and U937-derived macrophages, respectively (online supplementary figure s1), and following further treatment with or without tumor conditioned medium (CM). The expression of PD-L1 significantly decreased in the PTPRO-overexpressing THP-1 and U937 cells but significantly increased in PTPRO KO cells. However, the PD-L1 expression did not differ unless the cells were stimulated with CM (figure 3A), indicating that PTPRO can suppress in vitro PD-L1 expression in macrophages with certain stimulations.

Our previous studies indicated that PTPRO suppresses JAK2 activation in HCC cells, and other research has also indicated that activation of c-MYC, a downstream regulator of STAT3,23 can enhance PD-L1 expression.24 Therefore, we postulated that PTPRO might suppress PD-L1 expression by suppression of both the JAK2/STAT1 and JAK2/STAT3/c-MYC signaling pathways. We tested this hypothesis by evaluating STAT1 and STAT3 activation in TAMs from patients with HCC. We observed a stronger activation of JAK2/STAT1 (Y701) and JAK2/STAT3 (Y705), as well as increased c-MYC expression, in the PTPRO^Low group than in the PTPRO^High group (n=20 for each group) (figure 3B). Taken together with the PD-L1 expression shown in figure 2A, these data suggested that PD-L1 expression might be associated with JAK2/STAT1 and JAK2/STAT3/c-MYC activation in human HCC in monocyte or TAM.

We investigated the potential mechanism of PTPRO suppression on PD-L-1 using inhibitors that specifically targeted JAK2 (AZD1480) and c-MYC (10074-G5). Suppression of JAK2 and c-MYC decreased the PD-L1 transcription and expression. Culture of monocytes isolated from healthy controls (n=3) in CM increased the transcription and expression of PD-L1 and further increases were obtained by inhibiting JAK2 and c-MYC. Blockade of JAK2 signaling, in particular, decreased PD-L1 expression to levels similar to those observed in the medium control (figure 3C,D). As was observed in human macrophages, PD-L1 expression also significantly decreased when JAK2 and c-MYC were blocked in THP-1-PTPRO shRNA and U-937-PTPRO shRNA cells; this effect was particularly evident following JAK2 suppression (figure 3E,F).

Jak2/Stat1 and Jak2/Stat3/c-Myc signaling was also enhanced in macrophages from PTPRO KO mice when compared with WT control mice (online supplementary figure s2). We examined the regulation of JAK2 by PTPRO in macrophages by analyzing the relative distribution of PTPRO and JAK2. PTPRO was found predominantly on the cell membrane, whereas JAK2 was co-localized with PTPRO in U937 cells, which indicated a possible binding of PTPRO to JAK2 (online supplementary figure s3). Co-immunoprecipitation (co-IP) assays revealed increased levels of JAK2 in the PTPRO co-IP assays over control levels following treatment with IL-6, which indicated that PTPRO could regulate JAK2 by direct binding (online supplementary figure s4). These findings indicated that PTPRO can suppress PD-L1 expression in monocytes and macrophages by suppressing JAK2/STAT1 and JAK2/STAT3/c-MYC. Based on a previous report,13 we also performed a luciferase gene reporter assay, using PD-L1 promoter (−500 bp), which contained binding site of STAT1/STAT3 and STAT2/STAT5. We found that INF-γ can increase the PD-L1 promoter significantly and deletion of STAT1/STAT3 binding site but not STAT2/STAT5 binding site can decrease PD-L1 promoter dramatically; this result might suggest that STAT1 and STAT3 were dominant in PD-L1 transcription in macrophage (online supplementary figure s5).

No significant increase was observed in PD-L1 expression in the monocytes/macrophages in the absence of CM (figure 3A). We used a cytokine array to find the difference in cytokine secretion between control and tumor conditional medium treated macrophages, and the result indicated that IL-6 and INF-γ were the most variable of all cytokines (online supplementary figure s6). Based on previous knowledge,6 24–26 we postulated that IFN-γ and IL-6 might be the activators for JAK2/STAT1/PD-L1 and JAK2/STAT3/c-MYC/PD-L1 activation in the monocytes/macrophages. We tested this hypothesis by treating the macrophages with IL-6 (50 ng/mL) and IFN-γ (50 ng/mL), and we found that both INF-γ and IL-6 increased the Pd-L1/PD-L1 expression in Ptpro KO and PTPRO knockdown macrophages. IFN-γ, but not IL-6, also increased PD-L1 expression in WT and normal-PTPRO–expressing macrophages 24 hours after treatment (online supplementary figure s7 A‒D). IL-6 treatment significantly increased Pd-L1/PD-L1 expression both directly and in an INF-γ–dependent manner in WT and normal-PTPRO–expressing macrophages 72 hours after the treatment. The IL-6 promotion was particularly evident for IFN-γ–induced Pd-1/PD-L1 expression in Ptpro KO and PTPRO knockdown macrophages (figure 4A,B). This result could indicate that PTPRO expression level is essential for IL-6–induced Pd-L1/PD-L1 expression in macrophages.

We also treated WT mouse macrophages and U-937–derived and THP-1–derived macrophages with IL-6 and tracked the expression of PTPRO at different times. The IL-6 treatment decreased the transcription and expression of PTPRO in both mouse and human macrophages 72 hours after treatment (figure 4C,D and online supplementary figure s8). Monitoring of PD-L1 also showed that its expression paralleled the PTPRO expression both in human and mouse macrophage (figure 4C,D). Therefore, IL-6 activated the STAT3/c-MYC/PDL-1 axis and deregulated PTPRO action in monocytes/macrophages.

We also compared the miRNA expression in the monocytes between patients with HCC and the healthy controls, and the results indicated 313 miRNAs significantly upregulated and 114 miRNAs significantly downregulated in HCC monocytes (figure 5A,B). In addition, we also queried potential binding miRNAs within the 3′UTR of PTPRO (online supplementary figure s9); among these 313 upregulated miRNAs, we found that only miR-25–3 p had the potential to bind to the 3′UTR of PTPRO. In addition, serum IL-6 levels were significantly increased in most of the patients with HCC.27 These findings led us to question whether links might exist between increased serum IL-6 levels and decreased PTPRO expression in monocytes or macrophages. We also queried whether these miRNAs might be the linkers.

Expression of miR-25–3 p was confirmed in human monocytes, and its expression was significantly higher in the HCC monocytes than in the healthy control monocytes (figure 5C). We examined the potential relationship between miR-25–3 p and PTPRO expression by linear correlation studies of the expression of miR-25–3 p and PTPRO or PD-L1 in monocytes. The miR-25–3 p expression in HCC monocytes was negatively correlated with PTPRO expression (r=−0.580, p<0.01) (figure 5D) and positively correlated with PD-L1 expression (r=0.495, p<0.01) (figure 5E) in human HCC monocytes.

We speculated that miR-25–3 p could regulate PTPRO expression in monocytes by targeting the 3′UTR. The miRNA 3′UTR prediction results indicated that miR-25–3 p can align with 1381–1388 bp at the 3′UTR of PTPRO (figure 5F). We verified this regulation by generating mutated PTPRO 3′UTR and sub-cloning it into the pGL4-Luc vector. Transfection of U973-derived and THP-1–derived macrophages with the mutation in the 3′UTR of PTPRO significantly decreased the transcription activity in both cell types, thereby confirming that miR-25–3 p regulated PTPRO expression (figure 5F). Moreover, we confirmed that miR-25–3 p can decrease both transcription and protein expression by targeting the 3′UTR of PTPRO in macrophages (online supplementary figure s10).

Previous studies have reported that c-MYC can regulate the miR-25–3 p cluster.28 Also, we further confirmed such regulation by using a luciferase gene reporter assay (online supplementary figure s11). We surmised that upregulation of IL-6 in HCC monocytes and macrophages might promote the expression of miR-25–3 p via activation of the STAT3/c-MYC signaling pathway. We verified this potential regulation mechanism by investigating the serum IL-6 levels in patients with HCC by ELISA (online supplementary figure s12) and by performing a linear correlation study between the serum IL-6 levels and expression of miR-25–3 p or PTPRO in HCC monocytes. We found a clear positive linear correlation between serum IL-6 and monocyte expression of miR-25–3 p or PTPRO (IL-6 vs miR-25–3 p, r=0.414, p<0.01; IL-6 vs PTPRO, r=−0.458, p<0.01) (figure 5G,H). We then treated THP-1–derived and U937-derived macrophages with 50 ng/mL IL-6 and found that miR-25–3 p increased significantly 48 hours after the treatment and that miR-25–3 p transcription was restored using a c-MYC inhibitor (10058-F4) or by treatment with c-MYC shRNA (figure 5I). This signaling pathway analysis also indicated that the decreased expression of PTPRO by IL-6 was accompanied by an increased activation of STAT3/c-MYC signaling 72 hours after treatment. Treatment of U937-derived and THP-1–derived macrophages with AZD 1480, 10058-F4, c-MYC shRNA, or an miR-25–3 p inhibitor restored the expression of PTPRO that had been reduced by IL-6 treatment (figure 5J).

We investigated TAM differentiation and the tumor growth–promoting effects of the c-MYC/miR-25–3 p/PTPRO axis using an orthotopic tumor transplantation model with adoptive cell transfer. As in our previous studies,22 we saw a significant increase in tumor volume in Ptpro KO mice when compared with the WT controls. Adoptive transfer of miR-25–3 p–overexpressing macrophages into WT mice (WT+miR-25–3 p^Mφ) significantly increased the tumor volume, though this volume was still smaller than the volume of the tumors that developed in the Ptpro KO mice. WT mice that underwent transfer of c-Myc-knockdown macrophages (WT+c Myc shRNA^Mφ) showed a dramatic decrease in tumor volume (figure 6A).

The TME was investigated by immunofluorescence staining of the Pd-L1 positive macrophages and exhausted T cells within the tumor. The percentage of Pd-L1 (+) macrophages and Tim3 (+) T cells increased in Ptpro KO and WT mice following adoptive transfer of miR-25–3 p–overexpressing macrophages. Conversely, the percentage of Pd-L1 (+) macrophages and Tim3 (+) T cells decreased significantly following transfer of c-MYC–silenced macrophages into WT mice (figure 6B). The percentage of Th1 cells and CTLs within the tumor decreased significantly with the adoptive transfer of miR-25–3 p–overexpressing macrophages into WT mice. Conversely, the percentage of Th1 cells and CTLs increased significantly following transfer of c-MYC–silenced macrophages into WT mice (figure 6C).

We also investigated the signaling pathway in TAMs. Activation of STAT3 was significantly increased following depletion of Ptpro, but STAT3 activation increased significantly in WT mice following adoptive transfer of miR-25–3 p–overexpressing macrophages (figure 6D).