N6-methyladenosine (m6A) has been reported as an important mechanism of posttranscriptional regulation. Programmed death-ligand 1 (PD-L1) is a primary immune inhibitory molecule expressed on tumor cells that promotes immune evasion. Here we report ALKBH5 as an important m6A demethylase that orchestrates PD-L1 expression in intrahepatic cholangiocarcinoma (ICC). Regulation of PD-L1 expression by ALKBH5 was confirmed in human ICC cell lines. Sequencing of the m6A methylome identified PD-L1 mRNA as a direct target of m6A modification whose levels were regulated by ALKBH5. Furthermore, ALKBH5 and PD-L1 mRNA were shown to interact. ALKBH5 deficiency enriched m6A modification in the 3′UTR region of PD-L1 mRNA, thereby promoting its degradation in a YTHDF2-dependent manner. In vitro and in vivo, tumor-intrinsic ALKBH5 inhibited the expansion and cytotoxicity of T cells by sustaining tumor cell PD-L1 expression. The ALKBH5-PD-L1–regulating axis was further confirmed in human ICC specimens. Single-cell mass cytometry analysis unveiled a complex role of ALKBH5 in the tumor immune microenvironment by promoting the expression of PD-L1 on monocytes/macrophages and decreasing the infiltration of myeloid-derived suppressor-like cells. Analysis of specimens from patients receiving anti-PD1 immunotherapy suggested that tumors with strong nuclear expression patterns of ALKBH5 are more sensitive to anti-PD1 immunotherapy. Collectively, these results describe a new regulatory mechanism of PD-L1 by mRNA epigenetic modification by ALKBH5 and the potential role of ALKBH5 in immunotherapy response, which might provide insights for cancer immunotherapies.

Significance:

This study identifies PD-L1 mRNA as a target of ALKBH5 and reveals a role for ALKBH5 in regulating the tumor immune microenvironment and immunotherapy efficacy.

Intrahepatic cholangiocarcinoma (ICC) is the second most common primary malignant tumor of liver cancers. The recurrence rate of ICC is much higher than that of hepatocellular carcinoma (HCC; >50% in the first year after surgery; ref. 1), and the overall median survival time of ICC patients is less than 1 year (1). For patients with advanced or inoperable cholangiocarcinoma, gemcitabine–cisplatin combination is recommended as systemic therapy but with only modest improvement in overall survival, and there is still no standard second-line therapy (2). Thus, it is urgent to develop new strategies to prevent tumor progression and improve the prognosis of patients with ICC.

Programmed death-ligand 1 (PD-L1, also named B7-H1), a main inhibitory immune checkpoint molecule, help tumor cell to evade immune surveillance by binding to programmed death receptor-1 (PD-1) on T cells to induce T-cell apoptosis, anergy and functional exhaustion (3). Anti-PD-1/PD-L1 immunotherapy to block this interaction and reactivate T-cell immunity has gained outstanding results in various types of cancer (4). The anti-PD-1 antibody pembrolizumab has been approved by the FDA for the treatment of patients with unresectable or metastatic mismatch repair deficient and/or microsatellite-high solid tumors that progressed after prior therapy, which includes cholangiocarcinoma (5). The objective response rate of pembrolizumab is still low and better response was reported to be associated with high PD-L1 expression in cholangiocarcinoma (6). The complete understanding of the regulation mechanism underlying PD-L1 expression is emphasized for developing efficient combination immunotherapeutic strategies (7). It has been reported that PD-L1 expression was regulated by genetic alterations (e.g., rearrangements in the 3′UTR PD-L1 mRNA), tumor-intrinsic oncogenic pathways (e.g., RAS-MEK-ERK, CDK5, PI3K-AKT, HIF-1α), and posttranslational modifications (e.g., CMTM6/CMTM4, HIP1R, CUL3), whereas the regulation mechanism of PD-L1 on RNA epigenetic level is still unclear (8).

RNA N6-methyladenosine (m6A) is the most abundant internal modification in mRNAs, which is widely present in mammal cells and mainly present in the last exon of mRNA (9). RNA m6A methylation is regulated by a series of “Writer,” “Eraser,” and “Reader” proteins (10). M6A methyltransferase complex is mainly composed of METTL3 (methyltransferase-like 3), METTL14 (methyltransferase-like 14) and WTAP (Wilms tumor 1-associated protein), which act as "Writer" to catalyze methylation process. Whereas, FTO (fat mass and obesity-associated protein) and ALKBH5 (alkB homolog 5) act as demethylase (“Eraser”). The ultimate fate of m6A methylated mRNA depends on the “reader” (e.g., YTHDF1, YTHDF2, YTHDF3) that recognizes them, which may affect mRNA translation, stability, splicing, and nuclear transportation (10). Numerous targets of m6A has been found to involve in cell tissue development and stem cell self-renewal and differentiation (11), circadian rhythm regulation (12), T-cell homeostasis (13), mouse fertility (14), postnatal development of the mouse cerebellum (15), innate immune response (16), ultraviolet-induced DNA damage response (17) and dendritic cell antigen presentation (18). However, the possible role of m6A modification in the regulation of PD-L1 expression is still not reported, and its role in tumor immune microenvironment as well as in ICC is still unclear.

Here, we mainly reported an epigenetic regulating mechanism of PD-L1 on mRNA level by ALKBH5 depending on m6A modification. We further unveiled a complex role of ALKBH5 in immune microenvironment and demonstrated that tumors with ALKBH5 strong nuclear expression pattern are more sensitive to anti-PD-1 immunotherapy. Our results reported a new regulation mechanism of PD-L1 and unveiled a role of ALKBH5 in immunotherapy, which might shed new light on future immunotherapy clinically.

Cell culture and pharmacologic drug treatment

RBE and HCCC-9810 are from National Laboratory Cell Resource Sharing Center (Beijing, China). HUCCT1 is from JCRB Cell Bank (Japanese Collection of Research Bioresources). TFK1 is from DSMZ (German Collection of Microorganisms and Cell Cultures). LICCF, LIPF155c and LIPF178c are from China Center for Type Culture Collection (Wuhan, China). Cells were cultured at 37° in an atmosphere containing 5% CO2 in RPMI-1640 (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin–streptomycin solution (Gibco). The genetic identity of RBE, HCCC-9810, TFK1 and HUCCT1 were confirmed by short tandem repeat profiling. The cell LICCF, LIPF155c, LIPF178c is on patent and its characteristics have not been disclosed (http://www.cellresource.cn/). All the cell lines are routinely tested for Mycoplasma contamination every 3 months.

Cycloheximide (CHX, protein synthesis inhibitor; MCE, catalog no. HY-12320), MG132 (proteasome inhibitor; Selleck, Shanghai, China, catalog no. S2619) and actinomycin D (an inhibitor of DNA transcription and replication; MCE, catalog no. HY-17559) were used at a final concentration of 10 μg/mL, 50 μmol/L, and 5 μg/mL, respectively. 3-Deazaadenosine (a global methylation inhibitor; MCE, catalog no. HY-W013332) was used at a final concentration of 0–50 μmol/L. Atezolizumab (anti-PD-L1 blockade antibody; Selleck, catalog no. A2004) or control IgG (BioXcell, catalog no. BE0297) was used at 10 μg/mL for cell treatment.

PD-L1 detection on cell surface

Cells with or without IFNγ treatment for 24 hours were digested and collected. After washing with staining buffer, 1 × 106 cells were suspended in 100 μL staining buffer and incubated at 4° for 30 minutes with either PE Mouse Anti-Human PD-L1 (MIH1, BD Biosciences, catalog no. 557924) or APC Mouse Anti-Human PD-L1 (MIH1, BD Biosciences, catalog no. 563741). After washing with staining buffer, the samples were subjected to flow cytometry analysis (BD LSRFortessa). The data were further analyzed with FlowJo vX.07 software (Tree Star).

M6A sequencing and RNA sequencing

M6A sequencing (M6A-seq) was performed according to the reported protocol by Dominissini and colleagues (9). Briefly, total RNA was extracted from RBE shCtrl and shALKBH5 cells. Then, mRNA sequencing and m6A-seq were simultaneously performed (Genergy).

For m6A seq, polyadenylated RNAs was enriched from total RNA and was further fragmented into approximately 100-nucleotide-long oligonucleotides (as “input”). The fragmented RNA was incubated with anti-m6A affinity purified antibody (Synaptic Systems). The mixture was then immunoprecipitated by incubating with protein-A beads (Repligen) at 4°C for additional 2 hours. After extensive washing, bound RNA was eluted and precipitated and used for library generation. High-throughput m6A and “input” RNA sequencing (RNA-seq) of samples was performed on the NovaSeq 6000 sequencing system (Illumina, Inc.) according to the manufacturer's instructions. Library preparation, high-throughput sequencing and following analysis were performed by Genergy. Sequencing reads were mapped to human genome version 38 (Homo_sapiens.GRCh38.86) and unique mapping reads were used for further analysis (STAR softwareSTAR_2.5.2b). The m6A peak calling was done by MACS2 software (MACS2_2.1.0). Peaks were identified when P < 0.0001 and adjusted q value ≤ 0.01 and were visualized by IGV software and subjected to following analyses; (2) RNA-seq: Differentially expressed genes (DEG) analysis was performed using the DESeq2 (Deseq2_1.26.0). DEG with P < 0.05, absolute changing fold ≥ 2 were subjected to Kyoto Encyclopedia of Genes and Genomes pathway and Gene Ontology (GO) enrichment analysis.

N6-methyladenosine RNA immunoprecipitation quantitative RT-PCR of fragmented mRNA assay

The enrichment of m6A-modified mRNA was the same as described in “m6A-seq and RNA-seq.” Both the immunoprecipitated m6A modified mRNA and “input” mRNA were subjected to quantitative RT-PCR with indicated primers. Fold enrichment was determined as the ratio of m6A modified mRNA to “input” mRNA.

Dual-luciferase reporter assay

Cells were seeded in 48-well plate and were transfected with indicated reporter plasmid. The activities of Firefly luciferase and Renilla luciferase in each well were assessed by the Dual-Luciferase Reporter (DLRTM) Assay System (Promega, catalog no. E1910) according to the manufacturer's instructions. Both the Firefly and Renilla luciferase acitvity were measured. The relative luciferase activity was the ratio of Firefly luciferase activity to Renilla luciferase activity.

RNA immunoprecipitation

RNA immunoprecipitation assay was conducted with the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, catalog no. E1910) according to the manufacturer's instructions. Briefly, 2 × 107 RBE or LIPF178c cells were subjected to 400 μL RIP Lysis Buffer. Dispense 450 μL each of the RIP lysate into nuclease-free microcentrifuge tubes and store at −80° C for 1 hour. Magnetic beads were coated with 5 μg of Rabbit IgG Purified or Rabbit anti-ALKBH5 (ABE547, Millipore). After quick thawing, equal amount of 200 μL RIP lysate supernatant was incubated with either Rabbit IgG-coated beads or anti-ALKBH5-coated beads overnight at 4°C, with another 10 μL RIP lysate supernatant set as “input” control. Wash the beads-antibody-protein-RNA complex total six times with 500 μL of cold RIP. The protein was then digested and the RNA was purified by the solution of phenol: chloroform: isoamyl alcohol. The quality of RNA was evaluated by NanoDrop 2000 (Thermo Fisher Scientific). The RNA interacted with ALKBH5 was assessed by quantitative RT-PCR. The “input” was subjected to Western blot analysis of ALKBH5 expression.

mRNA stability detection

Cells were cultured overnight and then treated with actinomycin D 5 μg/mL at 4, 3, 2, 1, 0.5 and 0 h before trypsinization collection. The total RNA was extracted by TRIzol (Invitrogen). Quantitative RT-PCR was conducted to determine the relative level of indicated mRNA. The mRNA decay rate was determined by nonlinear regression curve fitting (one phase decay model, Y = (Y0 − Plateau) * exp(−K*X) + Plateau) using GraphPad Prism 8.0 (GraphPad Sowtware, Inc, www.graphpad.com). Goodness of fit was quantified with R2. The halftime (t1/2) of mRNA was calculated by the fitted curve in GraphPad Prism 8.0.

Isolation and activation of peripheral blood mononuclear cells

The peripheral blood from health donors were provided by Central Laboratory and Department of Laboratory Medicine, Shanghai Tenth People's Hospital (Shanghai, China). The PBMC was isolated from the peripheral blood with Lymphoprep (STEMCELL, catalog no. 7801) according to the manufacturer's instructions. Briefly, the diluted blood was carefully added on the top of Lymphoprep. After centrifuging, the mononuclear cell layer at the plasma:Lymphoprep interface was isolated and washed once with Hanks' Balanced Salt Solution. Typically, 1 × 106 PBMC was activated with 25 μL ImmunoCult Human CD3/CD28 T Cell Activator (STEMCELL, catalog no. 10971) in 1 mL ImmunoCult-XF T Cell Expansion Medium (STEMCELL, catalog no. 10981) supplemented with 10 ng/mL IL2 (Peprotech, catalog no. 200–02). For further expansion, adjust the viable cell density to 1 × 106 cells/mL every 2–3 days by adding fresh complete ImmunoCult-XF T Cell Expansion Medium supplemented with IL2 to the cell suspension.

T-cell killing assay in vitro

T-cell killing assay in vitro was conducted as previously reported (19). The preactivated peripheral blood mononuclear cells (PBMC) were cocultured with tumor cells. After incubation, the viability of tumor cells was measured by Cell Counting Kit-8 (CCK8; Dojindo) and the apoptosis of tumor cells was detected by Annexin V-PE Apoptosis Detection Kit (Beyotime Biotechnology, catalog no. C1065S). For CCK8 assay, the ICC tumor cells were seeded into a 96-well plate and cultured overnight. The activated PBMCs were added to coculture with tumor cells at indicated ratio. After incubation for indicated period, the suspension cells (mainly PBMCs and dead cells) were removed and the adherent tumor cells were further washed twice. The cell viability of the adherent tumor cells was measured using CCK8 assay according to the manufacturer's instructions. The results were normalized to the viability of the no PBMC adding control. All experiments were performed in triplicate. For apoptosis detection, the ICC tumor cells were seeded into a 12-well plate and cultured overnight. The activated PBMCs were added to coculture with tumor cells at ratio of 4:1 for 24 hours. After incubation, both the suspension cells (mainly PBMCs and dead cells) and adherent cells were collected and subjected to Annexin V-PE staining according to the manufacturer's instructions. After washing, the samples were subjected to flow cytometry analysis (BD LSRFortessa). Both the percentage of intact tumor cells (larger cells identified by the FSC and SSC) and apoptotic cells were analyzed. The cells without PBMC coculturing were set as control.

Mouse xenograft models and T-cell killing assay in vivo

All animal experiments were conducted in conformity with NIH guidelines and approved by the Ethics Committees of Fudan University Shanghai Cancer Center. Adult female NCG mice (NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl; 6–8 weeks) were purchased from Model Animal Resource Information Platform (Nanjing, China) and randomly assigned into experiment groups. ICC tumor cells (LIPF178c-shCtrl/shALKBH5) of 5 × 106 were injected into the right flank of NCG mice. Tumor volume was calculated by the formula: volume = ab2/2 (a, the longer axis; b, the shorter axis). T-cell killing assay in vitro was conducted as previously reported (20). PBMCs from healthy donors were activated and expanded as described above. The day before tumor cell injection, PBMC (i.v. 1 × 107 cells) was adoptively transferred to NCG mice via the tail vein. At the end, the PBMC was isolated and subjected to flow cytometry for detecting T-cell percentage.

Patients and tissue specimens

Tissue specimens were obtained from first surgical resection without previous treatment at Eastern Hepatobiliary Surgery Hospital (Shanghai, China) from 2011–2013. Written informed consent was obtained from patients. The procedure of human specimen collection was approved by the Ethics Committee of Eastern Hepatobiliary Surgery Hospital. 133 pairs of ICC and paired paracancerous tissues used for constructing tissue microarray, among which, 89 paired samples with complete clinicopathologic information and follow-up data and successful ALKBH5 detection by IHC were listed in Supplementary Table S1. Other 38 sets of ICC tumor tissues were subject to protein extraction for Western blot analysis.

Another 6 specimens were obtained from surgical resections of patients before receiving immunotherapy at Mengchao Hepatobiliary hospital of Fujian Medical University, and their clinicopathologic information were listed in Supplementary Table S2. Written informed consent was obtained from patients. The procedure of human specimen collection was approved by the Institution Review Board of Mengchao Hepatobiliary Hospital of Fujian Medical University. The therapeutic responses were evaluated according to Response Evaluation Criteria in Solid Tumors (version 1.1) 3 months after first administration.

Ethics approval and consent to participate

All animal experiments were conducted in conformity with NIH guidelines and approved by the Ethics Committees of Fudan University Shanghai Cancer Center. For experiments using human samples, written informed consent was obtained from patients. The procedure of human specimen collection was approved by the Ethics Committee of Fudan University Shanghai Cancer Center.

Statistical analysis

Unless otherwise noted, the statistical tests were performed as two-sided and the values were expressed as means ± SD. The analysis was performed in GraphPad Prism8 software. Unpaired Student t test or paired Student t test was used to compare between two groups of experiments. Cumulative survival time was estimated by the Kaplan–Meier method, and the log-rank test was applied to compare groups. P < 0.05 was considered statistically significant. No animal data was excluded.

Details of other regular experiments were provided in Supplementary Materials and Methods.

Data availability

Raw Sequence reads are deposited in BioProject (NCBI, Accession PRJNA732121). Sheets of analyzed m6A-seq and mRNA-seq results are included in the supplementary files (Supplementary Table S5–S11).

PD-L1 is regulated by ALKBH5

We found that transient knockdown of ALKBH5 by 3 specific siRNAs obviously downregulated PD-L1 protein level in human ICC cell lines (RBE, HCCC-9810; Supplementary Fig. S1A), whereas transient knockdown of FTO, METTL3, or METTL14 had no obvious effect on PD-L1 protein level (Supplementary Fig. S1B–S1D). Together with the fact that ALKBH5 knockdown did not change the protein level of FTO, METTL3, or METTL14 (Supplementary Fig. S1E), we inferred that PD-L1 was specifically regulated by ALKBH5 but not FTO, METTL3, or METTL14.

We checked PD-L1 as well as ALKBH5, FTO, METTL3, or METTL14 expression in 8 human ICC cell lines. We employed ICC cell lines (RBE, HCCC-9810, LIPF178c and HUCCT1) with high/moderate PD-L1 expression for further study (Supplementary Fig. S1F). ALKBH5 stable knockdown downregulated PD-L1 expression on both protein and mRNA levels in RBE, HCCC-9810, LIPF178c and HUCCT1 (Fig. 1A–D). Because IFNγ is a major stimulator for PD-L1 expression in tumor microenvironment and it can provoke PD-L1 mRNA de novo synthesis (21), we examined whether ALKBH5 could regulate IFNγ-induced PD-L1 expression. IFNγ-stimulated expression and membranous distribution of PD-L1 was significantly blunted upon ALKBH5 knockdown in ICC cell lines (Fig. 1E–H). In cell lines with moderate expression level of PD-L1 and ALKBH5 (RBE and HCCC-9810), exogenous expression of ALKBH5 increased the protein level of PD-L1 (Fig. 1I–K). At last, we confirmed the regulation of PD-L1 by ALKBH5 using sgRNA (Supplementary Fig. S1G). These data convinced that PD-L1 expression was regulated by ALKBH5.

Figure 1.

ALKBH5 regulates PD-L1 expression in human ICC cell lines. A–C, Western blot analysis for PD-L1 and ALKBH5 expression in indicated ICC cell lines with stable ALKBH5 knockdown. The statistical analysis of the efficiency of ALKBH5 knockdown (B) and the level of PD-L1/GAPDH (C) are shown. n = 3. D, Quantitative RT-PCR detected the mRNA level of PD-L1. n = 3. E–H, The expression of PD-L1 was detected in indicated ICC cell lines with IFNγ 20 ng/mL treatment for 24 hours. Representative Western blot images and statistical analysis of PD-L1 protein level (E and F). FCM detected the cell membranous PD-L1 level (G). Quantitative RT-PCR detected the mRNA level of PD-L1 (H). n = 3. I–K, Western blot analysis for PD-L1 and ALKBH5 expression in indicated ICC cell lines with stable ALKBH5 overexpression. The statistical analysis of the efficiency of ALKBH5 overexpression (J) and the relative level of PD-L1/GAPDH (K) are shown. n = 3. ns, no significant, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Figure 1.

ALKBH5 regulates PD-L1 expression in human ICC cell lines. A–C, Western blot analysis for PD-L1 and ALKBH5 expression in indicated ICC cell lines with stable ALKBH5 knockdown. The statistical analysis of the efficiency of ALKBH5 knockdown (B) and the level of PD-L1/GAPDH (C) are shown. n = 3. D, Quantitative RT-PCR detected the mRNA level of PD-L1. n = 3. E–H, The expression of PD-L1 was detected in indicated ICC cell lines with IFNγ 20 ng/mL treatment for 24 hours. Representative Western blot images and statistical analysis of PD-L1 protein level (E and F). FCM detected the cell membranous PD-L1 level (G). Quantitative RT-PCR detected the mRNA level of PD-L1 (H). n = 3. I–K, Western blot analysis for PD-L1 and ALKBH5 expression in indicated ICC cell lines with stable ALKBH5 overexpression. The statistical analysis of the efficiency of ALKBH5 overexpression (J) and the relative level of PD-L1/GAPDH (K) are shown. n = 3. ns, no significant, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Close modal

PD-L1 is a direct target of m6A modification

The fact that ALKBH5 knockdown did not obviously affect the degradation rate of PD-L1 protein was clarified by cycloheximide (CHX, a protein synthesis inhibitor) and MG132 treatment (MG132, a proteasome-dependent degradation pathway inhibitor; Supplementary Fig. S1H and S1I). Furthermore, we employed a global methylation inhibitor 3-deazaadenosine (DAA), which obviously reversed ALKBH5 knockdown-restrained PD-L1 protein level (Fig. 2A and B). Together with the observation that the overexpression of wild-type ALKBH5 but not mutant H204A (losing m6A demethylase activity) significantly upregulated PD-L1 protein and mRNA level (Fig. 2C; Supplementary Fig. S1J and S1K), it indicated that ALKBH5 manipulates PD-L1 expression depending on its demethylation activity.

Figure 2.

PD-L1 is a potential target of m6A modification regulated by ALKBH5. A, Western blot analysis for PD-L1 expression in RBE-shCtrl/shALKBH5 cells upon DAA 0–25 μmol/L treatment for 24 hours. B, Western blot analysis for PD-L1 expression in RBE-shCtrl/shALKBH5 cells with DAA 0 or 25 μmol/L treatment for 24 hours in presence of IFNγ 20 ng/mL stimulation. C, Western blot (left) and quantitative RT-PCR (right) analysis for PD-L1 expression in RBE-shALKBH5 cells transiently transfected with vector, ALKBH5, or H204A (a mutant of ALKBH5 losing m6A demethylase activity) plasmid. The successful transfection and expression of ALKBH5 was confirmed by detecting Flag and ALKBH5 protein level. Statistical analysis is shown. n = 3. ****, P < 0.0001; unpaired t test. D, Distribution of m6A peaks across mRNA in RBE-shCtrl/shALKBH5. E, Number of m6A peaks (left) and m6A modified genes (right) identified by m6A-seq in RBE-shCtrl/shALKBH5 cells. F, Pie graphs of m6A total, common and unique peak distribution in indicated regions in RBE-shCtrl/shALKBH5 cells. G, Pipeline of ALKBH5 downstream targets analysis.

Figure 2.

PD-L1 is a potential target of m6A modification regulated by ALKBH5. A, Western blot analysis for PD-L1 expression in RBE-shCtrl/shALKBH5 cells upon DAA 0–25 μmol/L treatment for 24 hours. B, Western blot analysis for PD-L1 expression in RBE-shCtrl/shALKBH5 cells with DAA 0 or 25 μmol/L treatment for 24 hours in presence of IFNγ 20 ng/mL stimulation. C, Western blot (left) and quantitative RT-PCR (right) analysis for PD-L1 expression in RBE-shALKBH5 cells transiently transfected with vector, ALKBH5, or H204A (a mutant of ALKBH5 losing m6A demethylase activity) plasmid. The successful transfection and expression of ALKBH5 was confirmed by detecting Flag and ALKBH5 protein level. Statistical analysis is shown. n = 3. ****, P < 0.0001; unpaired t test. D, Distribution of m6A peaks across mRNA in RBE-shCtrl/shALKBH5. E, Number of m6A peaks (left) and m6A modified genes (right) identified by m6A-seq in RBE-shCtrl/shALKBH5 cells. F, Pie graphs of m6A total, common and unique peak distribution in indicated regions in RBE-shCtrl/shALKBH5 cells. G, Pipeline of ALKBH5 downstream targets analysis.

Close modal

We further applied RNA-seq and m6A-seq to unearth the underlying regulating mechanism. RNA-seq identified 104 genes differentially expressed by at least 2-fold between shALKBH5 and shCtrl cells (P < 0.05, Supplementary Fig. S2A-S2D). ALKBH5 knockdown increased m6A enrichment primary near the stop codon and in the 3′UTR region (Fig. 2D). Sequence of m6A methylome identified 11,938 m6A peaks from 6,230 m6A-modified genes in shCtrl cells, and 9,609 m6A peaks from 5,894 m6A-modified genes in shALKBH5 cells (Fig. 2E). ALKBH5 knockdown altered the distribution of common peak (m6A peaks appearing in both shCtrl and shALKBH5 cells) and unique peak (m6A peaks only appearing in shCtrl or shALKBH5; Fig. 2F). Among the 104 differentially regulated genes identified by RNA-seq, CD274 (gene encoding PD-L1) was found to have unique m6A peaks in shALKBH5 cells and have common peaks with fold change larger than 1.2 in shALKBH5 group compared to shCtrl group [log2-fold change (log2FC)>0.35; Fig. 2G]. These data indicated that m6A modification existed on PD-L1 mRNA and the level of m6A on PD-L1 mRNA was regulated by ALKBH5.

Based on m6A-seq data, three m6A peaks were identified in PD-L1 mRNA, with a common peak (Peak-1) on 3′UTR region and two unique peaks (Peak-2/3) on coding sequence region (Supplementary Table S3). The m6A status of Peak-1, Peak-2, and Peak-3 was further checked by N6-methyladenosine RNA immunoprecipitation (MeRIP) quantitative RT-PCR of fragmented RNA with specific primers #13/14, #15/16 and #18/19 respectively (Fig. 3A; Supplementary Table S3). Primer #9 for non-m6A modification region in PD-L1 mRNA was set as a negative control. The m6A level of Peak-1 evaluated by primer #13/14 was significantly increased in ALKBH5-deficient cells (Fig. 3A). Consistent with our findings, a previous m6A profiling dataset (GSE87515) also identified m6A modification on PD-L1 mRNA with a common peak within 3′UTR region in glioma (motif: GGACT; Supplementary Table S4; ref. 22). The signal of common peak (Peak-1) from m6A-seq was shown (Fig. 3B; Supplementary Fig. S2E).

Figure 3.

ALKBH5 regulates PD-L1 expression via targeting m6A modification on 3′UTR region of PD-L1 mRNA. A, The m6A fold enrichment of the peak region detected by MeRIP quantitative RT-PCR of fragmented mRNA with indicated primers. n = 3. B, The signal of common peak (Peak-1) on PD-L1 mRNA. Both the blue and red slots represent the signal of IP sample minus the corresponding input sample located on the common peak of PD-L1. C, Relative luciferase activity of wild-type PD-L1 3′UTR-fused dual-luciferase reporter in indicated cells. ALKBH5 wild-type, H204A mutant, or vector plasmid were transiently transfected. n = 3. D, The graphic design of PD-L1 mutant 3′UTR (top). Relative luciferase activity of PD-L1 3′UTR wild-type/3′UTR-M/3′UTR-T-fused dual-luciferase reporter in RBE transiently transfected with ALKBH5 wild-type, H204A-mutant plasmid, or vector control. n = 3. E, RNA immunoprecipitation assay detected the interaction between ALKBH5 and indicated RNA in RBE and LIPF178c cells. Enrichment of the indicated RNA was measured by quantitative RT-PCR. RIP with nonspecific IgG was set as control. Western blot of ALKBH5 showed equal amount of input ALKBH5 protein in two groups. n = 3. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Unpaired t test.

Figure 3.

ALKBH5 regulates PD-L1 expression via targeting m6A modification on 3′UTR region of PD-L1 mRNA. A, The m6A fold enrichment of the peak region detected by MeRIP quantitative RT-PCR of fragmented mRNA with indicated primers. n = 3. B, The signal of common peak (Peak-1) on PD-L1 mRNA. Both the blue and red slots represent the signal of IP sample minus the corresponding input sample located on the common peak of PD-L1. C, Relative luciferase activity of wild-type PD-L1 3′UTR-fused dual-luciferase reporter in indicated cells. ALKBH5 wild-type, H204A mutant, or vector plasmid were transiently transfected. n = 3. D, The graphic design of PD-L1 mutant 3′UTR (top). Relative luciferase activity of PD-L1 3′UTR wild-type/3′UTR-M/3′UTR-T-fused dual-luciferase reporter in RBE transiently transfected with ALKBH5 wild-type, H204A-mutant plasmid, or vector control. n = 3. E, RNA immunoprecipitation assay detected the interaction between ALKBH5 and indicated RNA in RBE and LIPF178c cells. Enrichment of the indicated RNA was measured by quantitative RT-PCR. RIP with nonspecific IgG was set as control. Western blot of ALKBH5 showed equal amount of input ALKBH5 protein in two groups. n = 3. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Unpaired t test.

Close modal

To test whether ALKBH5-regulated PD-L1 expression was dependent on the Peak-1 region, we constructed PD-L1 Peak-1 region (3′UTR) that spanned from 943 bp (immediately after the stop codon) to 1461 bp on PD-L1 transcript variant 1 into Dual-Luciferase Reporter plasmid. PD-L1 3′UTR reporter was significantly augmented in RBE-ALKBH5 cells compared to RBE-Ctrl cells (Fig. 3C, left). The relative luciferase activity of PD-L1 3′UTR was significantly lower in cells with exogenous ALKBH5-mutant H204A expression compared to that with ALKBH5 wild-type expression (Fig. 3C, middle). Transient ALKBH5 and its mutant H204A overexpression both raised the endogenous ALKBH5 protein level (Supplementary Fig. S3). For this reason, ALKBH5-mutant H204A overexpression still increased the relative luciferase activity in wild-type or control cells but failed to increase it in ALKBH5-deficient cells (Fig. 3C, right).

To further address the m6A modification site in PD-L1 transcripts, we constructed a mutant PD-L1 3′UTR reporter plasmid (3′UTR-M) by replacing adenosine (A) with thymine (T) within identified m6A motifs (GSE87515, Supplementary Table S4), and a truncate PD-L1 3′UTR reporter plasmid that spanned from 1158 to 1461 bp on PD-L1 mRNA (3′UTR-T; Fig. 3D). Compared with 3′UTR fused reporter, the relative luciferase activity of 3′UTR-M or 3′UTR-T significantly decreased in ALKBH5 or its mutant H204A overexpression cells but not in vector control, implying the regulation of 3′UTR depending on ALKBH5 (Fig. 3D). Of note, the relative luciferase activity of 3′UTR-M was higher than that of 3′UTR-T, implying that unknown mechanism might be involved in ALKBH5-induced PD-L1 expression besides those m6A modification sites (Fig. 3D). At last, the interaction between ALKBH5 and PD-L1 mRNA was confirmed by RIP assay (Fig. 3E). FOXM1 pre-mRNA was previously reported to interact with ALKBH5 (22) and was set as a positive control (Fig. 3E). Taken together, these data highlighted that ALKBH5 interacted with PD-L1 mRNA and regulated the level of m6A modification within PD-L1 3′UTR region.

ALKBH5 deficiency promotes PD-L1 mRNA degradation

It has been reported that m6A modification affects mRNA stability, splicing, nuclear export, and/or translation of target mRNAs, depending on the distinct proteins that recognize them (e.g., YTHDF1, YTHDF2, YTHDF3; ref. 10). Because the level of all variants decreased, it suggested ALKBH5 deficiency–induced decreased PD-L1 expression was not due to alternative splicing (Fig. 4A; Supplementary Fig. S4A). Also, ALKBH5 did not affect PD-L1 promoter activity as determined by promoter reporter assay (Fig. 4B; Supplementary Fig. S4B). Measuring the change of PD-L1 mRNA after blocking new RNA synthesis with actinomycin D revealed that ALKBH5-stable overexpression significantly enhanced PD-L1 mRNA stability while ALKBH5 knockdown accelerated PD-L1 mRNA degradation (Fig. 4C; Supplementary Fig. S4C and S4D). Further, ALKBH5 deficiency significantly decreased and ALKBH5 overexpression significantly increased the level of PD-L1 mature mRNA in both cytoplasmic and nuclear fraction (Fig. 4D). The PD-L1 pre-mRNA level was also significantly decreased upon ALKBH5 knockdown and increased upon ALKBH5 overexpression (Fig. 4E; Supplementary Fig. S4E). Figure 4F showed that the half-life (t1/2) of PD-L1 pre-mRNA in control and ALKBH5 knockdown cells was the same, suggesting the stability of PD-L1 pre-mRNA was not affected by ALKBH5. And ALKBH5 knockdown did not affect PD-L1 promoter activity (Fig. 4B), suggesting the synthesis of PD-L1 pre-mRNA was barely affected by ALKBH5. Therefore, the reduced PD-L1 pre-mRNA level in ALKBH5 knockdown cells might be a result of increased transition to mature mRNA due to the depleted nuclear/cytoplasmic mature mRNA pool. Taken together, these data demonstrated that ALKBH5 mainly regulated PD-L1 mRNA stability, exerting overwhelmed effects on nuclear/cytoplasmic PD-L1 mRNA pool.

Figure 4.

ALKBH5 deficiency accelerates the degradation of PD-L1 mRNA dependent on YTHDF2. A, Quantitative RT-PCR detected the level of PD-L1 transcript variants in indicated ICC cells. n = 3. B, Relative luciferase activity of the PD-L1 promoter fused Gluc and SEAP dual reporter in RBE Ctrl/ALKBH5 cells. n = 3. C, The stability of PD-L1 mRNA was detected by quantitative RT-PCR upon actinomycin D 10 μg/ml treatment. D, Quantification of PD-L1 mature mRNA level in total, nuclear, and cytoplasmic fraction by quantitative RT-PCR in RBE-shCtrl/shALKBH5 cells or RBE-Ctrl/ALKBH5 cells. The nuclear and cytoplasmic fraction was checked by Western blot analysis. E, Quantitative RT-PCR detected PD-L1 pre-mRNA level. n = 3. F, The stability of PD-L1 pre-mRNA was detected by quantitative RT-PCR upon actinomycin D 10 μg/mL treatment. G, The successful knockdown of YTHDF2 by siRNA#1/#2 was confirmed by Western blot analysis. The stability of PD-L1 mRNA was detected by quantitative RT-PCR in RBE cells transfected with siRNA#1/#2 targeting YTHDF2 upon actinomycin D 10 μg/mL treatment. H, The level of PD-L1 mRNA was detected by quantitative RT-PCR in RBE-shCtrl/shALKBH5 cells transfected with siRNA#1/#2 targeting YTHDF2 with IFNγ 20 ng/mL stimulation for 24 hours. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Figure 4.

ALKBH5 deficiency accelerates the degradation of PD-L1 mRNA dependent on YTHDF2. A, Quantitative RT-PCR detected the level of PD-L1 transcript variants in indicated ICC cells. n = 3. B, Relative luciferase activity of the PD-L1 promoter fused Gluc and SEAP dual reporter in RBE Ctrl/ALKBH5 cells. n = 3. C, The stability of PD-L1 mRNA was detected by quantitative RT-PCR upon actinomycin D 10 μg/ml treatment. D, Quantification of PD-L1 mature mRNA level in total, nuclear, and cytoplasmic fraction by quantitative RT-PCR in RBE-shCtrl/shALKBH5 cells or RBE-Ctrl/ALKBH5 cells. The nuclear and cytoplasmic fraction was checked by Western blot analysis. E, Quantitative RT-PCR detected PD-L1 pre-mRNA level. n = 3. F, The stability of PD-L1 pre-mRNA was detected by quantitative RT-PCR upon actinomycin D 10 μg/mL treatment. G, The successful knockdown of YTHDF2 by siRNA#1/#2 was confirmed by Western blot analysis. The stability of PD-L1 mRNA was detected by quantitative RT-PCR in RBE cells transfected with siRNA#1/#2 targeting YTHDF2 upon actinomycin D 10 μg/mL treatment. H, The level of PD-L1 mRNA was detected by quantitative RT-PCR in RBE-shCtrl/shALKBH5 cells transfected with siRNA#1/#2 targeting YTHDF2 with IFNγ 20 ng/mL stimulation for 24 hours. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Close modal

YTHDF2 was reported to specifically recognize and bind m6A-containing RNAs, regulating the stability of mRNA (23). As shown in Fig. 4G, the successful knockdown of YTHDF2 by siRNA (#1, #2) was verified by Western blot, and YTHDF2 knockdown significantly increased the stability of PD-L1 mRNA. Knockdown of YTHDF2 increased PD-L1 mRNA level in RBE cells and RBE-shCtrl/shALKBH5 (Supplementary Fig. S5A-S4C). Exogenous expression of YTHDF2 significantly decreased PD-L1 mRNA level in RBE-shCtrl cells but not in RBE-shALKBH5 cells (Supplementary Fig. S5D). Moreover, YTHDF2 knockdown completely reversed the downregulation of PD-L1 mRNA level upon ALKBH5 knockdown with IFNγ treatment (Fig. 4H). These data suggested that ALKBH5 regulated the stability of PD-L1 mRNA in an YTHDF2-dependent manner.

ALKBH5 suppresses antitumor T-cell immunity in vitro and in vivo

PD-L1 on tumor cells binds to PD-1 on activating T cells, leading to the exhaustion and apoptosis of T cells subsequently, which has been identified as a key process in tumor cell–mediated immune escape (24). To test whether ALKBH5 could regulate antitumor T-cell immunity via regulating PD-L1, we performed T-cell–mediated killing assay in vitro by employing a coculture system in which activated PBMC from healthy donors were cocultured with human ICC cell lines (RBE, LIPF178c). PBMC was activated and expanded with anti-CD3/CD28 antibody and IL2 before coculturing with tumor cells. The percentage of CD3+ T cells in lymphocytes used for experiments varies and some (PBMC#11#16#17) achieved 90% (Supplementary Fig. S6A). ICC cells with ALKBH5 knockdown were more vulnerable against T-cell killing, and vice versa (Fig. 5A and B; Supplementary Fig. S6B-S6D). The number of apoptotic tumor cells was increased in ALKBH5-deficient cells upon PBMC coculturing (Fig. 5C).

Figure 5.

ALKBH5 suppresses antitumor T-cell immunity in vitro. A, CCK8 assay detected the killing of tumor cells by indicated activated PBMCs. ICC cells were cocultured with or without PBMCs for 24 hours. Data were normalized to their respective no PBMC controls. B, LIPF178c-shCtrl/shALKBH5 cells were cocultured with or without activated PBMC#13 for 3 days, with the ratio of PBMC to tumor cell number as 4. Representative micrographs of cells or DAPI staining are shown. C, Activated PBMCs (#11, #12, #13) were cocultured with LIPF178c-shCtrl/shALKBH5 cells for 24 hours at the ratio of PBMC to tumor cell number as 4. The PBMCs were collected and stained with PE-Annexin V, then subjected to FCM analysis. The percentage of apoptotic cells was analyzed. D and E, Activated PBMCs (#1, #3, #4) were cocultured with RBE-shCtrl/shALKBH5 (D) and activated PBMCs (#5, #8, #10) were cocultured with RBE-Ctrl/ALKBH5 (E) for 3 days at the ratio of PBMC to tumor cell number as 4. The percentage of CD3 in PBMCs were detected by FCM. Representative plots of CD3+ T cells are shown. Data were normalized to control group. n = 3. F and G, Quantitative RT-PCR was performed to detect PRF1 (perforin-1), GZMB (granzyme), GNLY (granulysin), and IFNG (IFNγ) in activated PBMCs cocultured with LIPF178c-shCtrl/shALKBH5 (F) for indicated time or with RBE-Ctrl/ALKBH5 (G) cells for 48 hours. The ratio of PBMCs to tumor cell number was 4. n = 3. H and I, FCM detected the intracellular IFNγ (H) and IL2 (I) level of CD3+CD8+ T cells in PBMCs after coculturing with RBE-shCtrl/shALKBH5 cells for 3 days. n = 3. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Figure 5.

ALKBH5 suppresses antitumor T-cell immunity in vitro. A, CCK8 assay detected the killing of tumor cells by indicated activated PBMCs. ICC cells were cocultured with or without PBMCs for 24 hours. Data were normalized to their respective no PBMC controls. B, LIPF178c-shCtrl/shALKBH5 cells were cocultured with or without activated PBMC#13 for 3 days, with the ratio of PBMC to tumor cell number as 4. Representative micrographs of cells or DAPI staining are shown. C, Activated PBMCs (#11, #12, #13) were cocultured with LIPF178c-shCtrl/shALKBH5 cells for 24 hours at the ratio of PBMC to tumor cell number as 4. The PBMCs were collected and stained with PE-Annexin V, then subjected to FCM analysis. The percentage of apoptotic cells was analyzed. D and E, Activated PBMCs (#1, #3, #4) were cocultured with RBE-shCtrl/shALKBH5 (D) and activated PBMCs (#5, #8, #10) were cocultured with RBE-Ctrl/ALKBH5 (E) for 3 days at the ratio of PBMC to tumor cell number as 4. The percentage of CD3 in PBMCs were detected by FCM. Representative plots of CD3+ T cells are shown. Data were normalized to control group. n = 3. F and G, Quantitative RT-PCR was performed to detect PRF1 (perforin-1), GZMB (granzyme), GNLY (granulysin), and IFNG (IFNγ) in activated PBMCs cocultured with LIPF178c-shCtrl/shALKBH5 (F) for indicated time or with RBE-Ctrl/ALKBH5 (G) cells for 48 hours. The ratio of PBMCs to tumor cell number was 4. n = 3. H and I, FCM detected the intracellular IFNγ (H) and IL2 (I) level of CD3+CD8+ T cells in PBMCs after coculturing with RBE-shCtrl/shALKBH5 cells for 3 days. n = 3. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test.

Close modal

In addition, the PBMCs cocultured with RBE-shALKBH5 cells showed larger cell colonies than that cocultured with RBE-shCtrl cells, implying that the inhibitory effect on T cells exerted by tumor cells was attenuated by ALKBH5 stable knockdown (Supplementary Fig. S7A). The CD3+ cell proportion of PBMCs was increased upon co-culturing with RBE-shALKBH5 cells compared with RBE-shCtrl cells, and were decreased upon coculturing with RBE-ALKBH5 cells compared with RBE-Ctrl cells (Fig. 5D and E).

Moreover, the mRNA levels of PRF1 (Perforin 1), GZMB (Granzyme B), GNLY (Granulysin), IFNG in PBMCs were increased upon coculturing with ALKBH5 knockdown cells, and decreased upon coculturing with ALKBH5 overexpressed cells (Fig. 5F and G). The IFNγ and IL2 levels of CD8+ T cells detected by flow cytometry were also increased upon coculturing with ALKBH5-deficient cells compared to the corresponding control cells (Fig. 5H and I). Taken together, these data suggested that the deficiency of ALKBH5 in tumor cells enhanced T-cell–mediated antitumor activity and vice versa.

To verify the potential effect of ALKBH5 on ICC progression in vivo, we first inoculated subcutaneous xenograft in immunodeficient NCG mice with LIPF178c-shCtrl/shALKBH5 and found nonsignificant difference (Fig. 6AC), indicating that ALKBH5 exerted a minor effect on ICC cell growth in vivo. We also examined the direct biological effects of ALKBH5 in ICC in vitro, without considering immune mechanisms. It showed that ALKBH5 knockdown promoted the malignancy of ICC cells including PD-L1 membranous-positive tumor cells (RBE/HCCC-9810/LIPF178c) as well as PD-L1 membranous-negative tumor cells (LIPF155c), indicating the existence of other targets (Supplementary Fig. S8A-S8D). The PD-L1 protein level was obviously downregulated by ALKBH5 knockdown in xenografts (Fig. 6D). Since the lack of murine ICC cell line, we further inoculated subcutaneous xenograft of LIPF178c-shCtrl/shALKBH5 in PBMC-transferred immunodeficient NCG mice. The human PBMC was in advance activated and expanded with anit-CD3/CD28 antibody and IL2 for 12 days before transfer, and the percentage of CD3+ T cells reached 80%. The tumor size of LIPF178c-shALKBH5 group was significantly smaller than that of shCtrl group in PBMC-transferred NCG mice (Fig. 6E–G). As expected, the percentage of human CD3+ cell in PBMCs was found to be much higher in shALKBH5 group than that in shCtrl group (Fig. 6H).

Figure 6.

ALKBH5 suppresses antitumor T-cell immunity in vivo. A–C, LIPF178c-shCtrl/shALKBH5 cells were injected subcutaneously into the right flank of NCG mice to obtain tumor xenografts. Tumor volume (B) and tumor weight (C) were measured. n = 5. ns, nonsignificant. Unpaired t test. D, Western blot analysis detected PD-L1 level in the subcutaneous xenografts described in A-C. E–H, LIPF178c-shCtrl/shALKBH5 was injected subcutaneously into the right flank of NCG mice, which were adoptively transferred with activated PBMC. E–G, Images of subcutaneous xenografts (E), tumor volume (F), and tumor weight (G) are shown. n = 5. H, FCM detected the percentage of human CD3+ T cells in PBMCs of NCG mice. n = 5. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 compared with shCtrl with PBMC transfer. Paired t test.

Figure 6.

ALKBH5 suppresses antitumor T-cell immunity in vivo. A–C, LIPF178c-shCtrl/shALKBH5 cells were injected subcutaneously into the right flank of NCG mice to obtain tumor xenografts. Tumor volume (B) and tumor weight (C) were measured. n = 5. ns, nonsignificant. Unpaired t test. D, Western blot analysis detected PD-L1 level in the subcutaneous xenografts described in A-C. E–H, LIPF178c-shCtrl/shALKBH5 was injected subcutaneously into the right flank of NCG mice, which were adoptively transferred with activated PBMC. E–G, Images of subcutaneous xenografts (E), tumor volume (F), and tumor weight (G) are shown. n = 5. H, FCM detected the percentage of human CD3+ T cells in PBMCs of NCG mice. n = 5. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 compared with shCtrl with PBMC transfer. Paired t test.

Close modal

ALKBH5 suppresses antitumor T-cell immunity in a PD-L1–dependent manner

Flow cytometry detected high membranous distribution of PD-L1 on LIPF178c and no distribution on LIPF155c cells (Supplementary Fig. S9A). And nonsignificant change of the tumor cell viability was observed between LIPF155c-shALKBH5 and LIPF155c-shCtrl cells in coculturing with PBMCs (Supplementary Fig. S9B). Together with the observation that the difference of cell vulnerability to T-cell killing between LIPF178c-shCtrl and LIPF178c-shALKBH5 was counteracted with the blockade of PD-L1 by atezolizumab (Supplementary Fig. S9C and S9D) our data suggested that PD-L1 is necessary for ALKBH5-mediated tumor resistance against antitumor T-cell immunity.

The positive correlation between ALKBH5 and PD-L1 in clinical ICC specimens

We further checked the clinical correlation between ALKBH5 and PD-L1 expression. The positive correlation between ALKBH5 and PD-L1 protein level was confirmed in human ICC tumor tissues by Western blot (Pearson r = 0.7447, P < 0.0001; Fig. 7A and B). The correlation between ALKBH5 and PD-L1 transcripts was explored in The Cancer Genome Atlas datasets on the online platform “GEPIA” (25), displaying significant positive correlation in various tumors (Supplementary Fig. S10A and S10B), indicating the regulation of PD-L1 by ALKBH5 might be ubiquitous.

Figure 7.

ALKBH5 positively correlates with PD-L1 expression in clinical ICC specimens. A and B, Western blot analysis detected ALKBH5 and PD-L1 protein level in clinical ICC tumor samples. The Pearson correlation between ALKBH5 and PD-L1 protein level was calculated (r = 0.7447, P < 0.0001; B). C, Top, representative ALKBH5 expression pattern the tumor area of ICC tissue microarray detected by IHC. Bottom, the PD-L1 expression of the same specimen is shown. D, The frequency of PD-L1/CD3/CD4/CD8–positive/negative samples in ALKBH5-positive/negative samples. The percentage of PD-L1/CD3/CD4/CD8–positive samples in each group is annotated above the column. Fisher exact test.

Figure 7.

ALKBH5 positively correlates with PD-L1 expression in clinical ICC specimens. A and B, Western blot analysis detected ALKBH5 and PD-L1 protein level in clinical ICC tumor samples. The Pearson correlation between ALKBH5 and PD-L1 protein level was calculated (r = 0.7447, P < 0.0001; B). C, Top, representative ALKBH5 expression pattern the tumor area of ICC tissue microarray detected by IHC. Bottom, the PD-L1 expression of the same specimen is shown. D, The frequency of PD-L1/CD3/CD4/CD8–positive/negative samples in ALKBH5-positive/negative samples. The percentage of PD-L1/CD3/CD4/CD8–positive samples in each group is annotated above the column. Fisher exact test.

Close modal

We further performed IHC of ALKBH5 as well as CD3, CD4, CD8, and PD-L1 on ICC tissue microarrays. ALKBH5 and PD-L1 expression levels on tumor cells were used for analysis, and representative positive and negative expression of ALKBH5, PD-L1, CD3, CD4, CD8 were displayed (Fig. 7C; Supplementary Fig. S11A). Comparing the expression of ALKBH5 in ICC tumor tissues and paired peritumor bile canaliculus cells, it found that ALKBH5 was mainly nuclear-enriched in ICC tumor tissues whereas cytoplasm/membrane-enriched in paired peritumor bile canaliculus cells, indicating a role of ALKBH5 in ICC development (Supplementary Fig. S11B). ALKBH5 positive expression was negatively related to tumor number and plasma CEA and CA125 level (Supplementary Table S1). ALKBH5 is mainly reported to be detected in nucleus (22, 26). However, in our ICC tissue microarrays, apart from the majority displaying nuclear enrichment, we noticed 12% (11/89) specimens displayed strong ALKBH5 expression on tumor cell membrane (membrane enrichment pattern, Fig. 7C). Because ALKBH5 colocalizes with nuclear speckles to exert its demethylase function in nuclear (14) and m6A methylation and demethylation is essentially completed upon the release of mRNA into the nucleoplasm (27), ALKBH5 expression on tumor cell membrane might not function as m6A demethylase and some unknown aspects of ALKBH5 need further exploration.

Studies on ICC tissue microarrays further found that PD-L1–negative expression and CD3-positive expression predicted better prognosis for ICC patients, whereas ALKBH5, CD4, or CD8 did not show significant influence on the prognosis (Supplementary Fig. S11C). Besides, significant correlation was found between PD-L1 and ALKBH5 expression (Fig. 7D).

Taken together, these data suggested the existence of ALKBH5–PD-L1 regulating axis in ICC. However, the fact that the percentage of CD3/CD4/CD8–positive tissues was comparable in ALKBH5-negative and positive tissues indicated that other indistinct mechanisms might be involved.

The complex role of ALKBH5 in ICC tumor immune microenvironment

Because of the lack of murine ICC cell line, we only studied the role of ALKBH5 in PBMC-transferred NCG mice (primarily T cells due to CD3/CD28 stimulation), which neglected non-T immune cell. Therefore, CD34+ humanized mice with an intact immune system were applied to study the role of ALKBH5 in ICC immune microenvironment comprehensively. We still found that the tumor size and weight of LIPF178C-shALKBH5 xenografts was significantly smaller than that of LIPF178C-shCtrl xenografts (Fig. 8A). Mass cytometry “cytometry by time of flight” (CyTOF) was applied for immune cell subsets analysis, and the antibody panel comprised mainly non-T-cell phenotyping markers and lacked markers for T-cell subtypes. Analysis of CD45+ cells identified 6 classic immune cell subsets including T cells, monocytes/macrophages, NK cells, conventional dendritic cells, plasmacytoid dendritic cells, and B cells (Supplementary Fig. S12A and S12B). T cells account for the largest part of TILs followed by monocyte/macrophage, and no significant change of T-cell percentage was found (Fig. 8B), in consistence with the findings on ICC tissue microarray (Fig. 7D). We noticed that the components of non-T cells varied dramatically on the tSNE map (Supplementary Fig. S12B, bottom). Therefore, random sampling of CD45+CD3 cells was passed to further analysis and 17 clusters were identified, including monocyte/macrophage (cluster-1, 2, 3, 4, 8, 10, 14 and 15), NK (cluster-5 and 13), conventional dendritic cell (cDC, cluster-17), plasmacytoid dendritic cell (pDC, cluster-16) and B cell (cluster-12; Supplementary Fig. S12C and S12D). Cluster-8 was characterized by CD11b+CD33+CD14+HLA-DRlowCD68lowCD86CD64, resembling myeloid-derived suppressor cell (MDSC) cells (28), and its percentage was significant higher in shALBH5 group as compared to shCtrl group (Fig. 8C; Supplementary Fig. S12D). Moreover, we found significantly increased expression of PD-L1 on monocyte/macrophage cells in shCtrl group compared with that in shALKBH5 group (Fig. 8D and E). Taken together, these data suggested that besides tumor-intrinsic ALKBH5–PD-L1 regulating axis, ALKBH5 can regulate tumor immune microenvironment by promoting the expression of PD-L1 on monocyte/macrophage cells and decreasing the infiltration of MDSC-like cells.

Figure 8.

The role of ALKBH5 in ICC immune microenvironment. A, LIPF178c-shCtrl/shALKBH5 was injected subcutaneously into the right flank of CD34+ humanized mice. Images of subcutaneous xenografts, tumor volume, and tumor weight are shown. n = 5. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test. B, Statistical analysis of the percentage of cell types identified in CD45+ cells. n = 4. Nonsignificant difference by unpaired t test. C, Statistical analysis of the percentage of clusters identified in CD45+CD3 cells. n = 3. *, P < 0.05; unpaired t test. D, The expression level of PD-L1 and cell density is displayed on tSNE plots. The dotted lines indicate obvious difference in monocyte/macrophage composition between shCtrl and shALKBH5 groups. E, Violin plot together with boxplot of PD-L1 expression in monocyte/macrophage cells from shCtrl/shALKBH5 xenografts. ****, P < 0.0001; Welch two-sample t test. F, ALKBH5 and PD-L1 expression in the tumor area of indicated ICC tissues detected by IHC. PD, progressive disease; SD, stable disease.

Figure 8.

The role of ALKBH5 in ICC immune microenvironment. A, LIPF178c-shCtrl/shALKBH5 was injected subcutaneously into the right flank of CD34+ humanized mice. Images of subcutaneous xenografts, tumor volume, and tumor weight are shown. n = 5. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired t test. B, Statistical analysis of the percentage of cell types identified in CD45+ cells. n = 4. Nonsignificant difference by unpaired t test. C, Statistical analysis of the percentage of clusters identified in CD45+CD3 cells. n = 3. *, P < 0.05; unpaired t test. D, The expression level of PD-L1 and cell density is displayed on tSNE plots. The dotted lines indicate obvious difference in monocyte/macrophage composition between shCtrl and shALKBH5 groups. E, Violin plot together with boxplot of PD-L1 expression in monocyte/macrophage cells from shCtrl/shALKBH5 xenografts. ****, P < 0.0001; Welch two-sample t test. F, ALKBH5 and PD-L1 expression in the tumor area of indicated ICC tissues detected by IHC. PD, progressive disease; SD, stable disease.

Close modal

The fact that ALKBH5 regulates PD-L1 level on both ICC tumor cells and monocyte/macrophage cells motivate us to explore whether ALKBH5 expression is related to anti-PD-L1/PD1 immunotherapy response. Six specimens from patients before receiving anti-PD1 immunotherapy were collected, and the therapeutic responses were evaluated 3 months later [4/6 Stable Disease (SD), 2/6 Progressive Disease (PD); Supplementary Table S2]. IHC detection of ALKBH5 and PD-L1 found that SD specimens displayed high ALKBH5 expression with strong nuclear enrichment pattern and PD-L1–positive expression on either tumor or immune cells, and PD specimens displayed ALKBH5 expression with membrane enrichment and PD-L1-negative expression (Fig. 8F). These data indicated that strong nuclear enrichment expression of ALKBH5 might be an indicator for anti-PD-L1/PD1 immunotherapy response.

The low responsiveness of anti-PD-1/PD-L1 immunotherapy highlights the requirement of complete understanding of PD-L1 regulation mechanism. Here, we reported for the first time that PD-L1 was regulated by m6A modification on mRNA level, and the main m6A enzyme orchestrating PD-L1 mRNA level was ALKBH5. The significant correlation between ALKBH5 and PD-L1 was confirmed in clinical ICC specimens by both Western blot and IHC. Moreover, the significant correlation between ALKBH5 and PD-L1 transcripts was found in more than 10 types of cancer in The Cancer Genome Atlas database, indicating that the regulation is not limited in ICC.

Importantly, we proved the interaction between ALKBH5 protein and PD-L1 mRNA and the increased m6A modification on the 3′UTR region of PD-L1 mRNA in ALKBH5-deficient cells. It has been documented that m6A modification affects mRNA stability, splicing, nuclear export, and/or translation of target mRNAs, depending on the distinct proteins that recognize them (e.g., YTHDF1, YTHDF2, YTHDF3; ref. 10). Here, we did not find that ALKBH5 affected the transcript splicing or promoter activity of PD-L1, indicating that the mRNA level decreased by ALKBH5 knockdown was not due to decreased transcription or alternative splicing. We mainly illustrated that the decreased PD-L1 mRNA level by ALKBH5 knockdown was mainly due to the accelerated degradation of PD-L1 mRNA depending on m6A reader protein YTHDF2, thereby decreasing PD-L1 protein level as a result. However, whether m6A modification affecting the translation efficiency of PD-L1 was not included in this study.

ALKBH5 was reported to promote cancer progression by enhancing the renewal and proliferation of cancer stem cells in glioblastoma and breast cancer (22, 29). In contrast, ALKBH5 was reported to inhibit the pancreatic cancer malignancy by regulating long non-coding RNA methylation (30), suggesting the complicated biological function of m6A in different cancer types. In our studies, we demonstrated that tumor-intrinsic ALKBH5-PD-L1 axis inhibited T cell–mediated cytotoxicity. The lack of murine ICC cell line really restricted the study on tumor immune microenvironment and optimizing immunotherapy in ICC. Therefore, we firstly employed a coculture system in which activated PBMCs from healthy donors were cocultured with human ICC cells in vitro and in vivo to illustrate the inhibitory effect of ALKBH5 on nonspecific T-cell immunity was PD-L1-dependent. However, studies on ICC tissue microarrays demonstrated that T-cell infiltration was not really affected by ALKBH5 expression, suggesting the limitation of our models. For deeper understanding of ALKBH5′s role in ICC tumor immune microenvironment, we employed CD34+ humanized NCG mice models, which have the same composition of immune cell subsets especially regarding myeloid cells, which is considered to be a better preclinical model (31, 32). By applying CD34+ humanized NCG mice and single-cell mass cytometry analysis, we further found that ALKBH5 promoted the expression of PD-L1 on monocyte/macrophage cells and decreased the accumulation of MDSC-like cells, unveiling a complicated role of ALKBH5 in tumor immune microenvironment. More importantly, we found that tumors with ALKBH5 strong nuclear expression pattern are more sensitive to anti-PD1 immunotherapy.

Gene Oncology-Biological Process items monocyte chemotaxis, regulation of fatty acid biosynthetic process, and regulation of fatty acid metabolic process were significantly enriched in RBE-shALKBH5 cells (Supplementary Fig. S13A). Fatty acid metabolism plays an important role in MDSCs induction and MDSCs rely on fatty acid as the major metabolic fuel for the production of inhibitory cytokines to support their immunosuppressive functions (33–35). Fatty acid metabolism also correlates with immune excluded/desert microenvironment (31). Increased fatty acid metabolism in ALKBH5-deficient tumor cells might be a reason for the accumulation of MDSCs and help remodeling a suppressive immune microenvironment. In addition, Tseng and colleagues reported that the PD-L1 expression on pleural effusion tumor cells was associated with the PD-L1 expression on macrophages in lung cancer (36), indicating that PD-L1 high expression on both tumor cells and macrophages might not be occasionally existed in our models. The further exploration on these findings will be interesting.

Looking for candidate biomarker for choosing suitable patient population for specific type of immunotherapy is now underlined. Although utilities of PD-L1 IHC assays for predicting anti-PD-1/PD-L1 immunotherapy response was approved in some cancers (e.g., pembrolizumab in non-small cell lung cancer; ref. 37), the majority showed inconsistent between PD-L1 immunohistochemistry readout and response (38). PD-L1 predictive role is confounded by technical issues and by discrepancies in scoring systems (38, 39). N-linked glycosylation of PD-L1 impedes its recognition by IHC, and Lee and colleagues improved PD-L1 detection method by removing glycosylation (40). Combination of PD-L1 detection with other biomarkers (e.g., tumor mutational burden, tumor-infiltrating lymphocytes, cytotoxic gene signature) also shows promise (37). Here, we suggested that the addition of ALKBH5 might have addictive predictive value beyond PD-L1 IHC alone.

In conclusion, our study demonstrates that PD-L1 is a direct target of ALKBH5 as well as m6A modification. Mechanically, the lack of ALKBH5 enhances the m6A modification on 3′UTR region of PD-L1 mRNA and mainly accelerates the degradation of PD-L1 mRNA in a YTHDF2-dependent manner, thereby reducing PD-L1 protein level consequently. ALKBH5 can inhibit antitumor T-cell immunity through ALKBH5-PD-L1 axis in ICC. Besides, we also find that ALKBH5 exerts complicated influence on the tumor immune microenvironment. ALKBH5 promotes the expression of PD-L1 on monocyte/macrophage cells and decreasing the infiltration of MDSC-like cells. At last, studies on specimens from patients before receiving anti-PD1 immunotherapy suggested that tumors with ALKBH5 strong nuclear expression pattern might be more sensitive to anti-PD1 immunotherapy. Taken together, our study here unveils the molecular mechanism of PD-L1 expression on RNA epigenetic level, and extends the understanding of ALKBH5 in tumor immune microenvironment and immunotherapy.

No disclosures were reported.

X. Qiu: Conceptualization, software, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, writing–review and editing. S. Yang: Conceptualization, software, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft. S. Wang: Conceptualization, validation, investigation, writing–original draft. J. Wu: Validation, investigation. B. Zheng: Software, validation, investigation, methodology. K. Wang: Software, methodology. S. Shen: Methodology. S. Jeong: Resources, methodology. Z. Li: Resources, software. Y. Zhu: Software, methodology. T. Wu: Methodology. X. Wu: Resources, methodology. R. Wu: Resources, methodology. W. Liu: Resources, methodology. H.-Y. Wang: Conceptualization, resources, funding acquisition. L. Chen: Conceptualization, resources, supervision, funding acquisition.

This work was supported by the National Research Program of China (2017YFA0505803, 2017YFC0908100), the State Key Project for Infectious Diseases (2018ZX10732202–001, 2018ZX10302207–004), National Natural Science Foundation of China (81790633, 81830054, 81902412, 81672860, 61922047, 82002921, and 81802853), National Natural Science Foundation of Shanghai (17ZR143800), and Youth Foundation of Fudan University Shanghai Cancer Center (YJQN201934).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Razumilava
N
,
Gores
GJ
. 
Cholangiocarcinoma
.
Lancet
2014
;
383
:
2168
79
.
2.
Sirica
AE
,
Gores
GJ
,
Groopman
JD
,
Selaru
FM
,
Strazzabosco
M
,
Wang
WX
, et al
Intrahepatic cholangiocarcinoma: continuing challenges and translational advances
.
Hepatology
2019
;
69
:
1803
15
.
3.
Chen
L
. 
Co-inhibitory molecules of the B7-CD28 family in the control of T-cell immunity
.
Nat Rev Immunol
2004
;
4
:
336
47
.
4.
Wolchok
JD
. 
PD-1 blockers
.
Cell
2015
;
162
:
937
.
5.
Rizvi
S
,
Khan
SA
,
Hallemeier
CL
,
Kelley
RK
,
Gores
GJ
. 
Cholangiocarcinoma - evolving concepts and therapeutic strategies
.
Nat Rev Clin Oncol
2018
;
15
:
95
111
.
6.
Ahn
S
,
Lee
JC
,
Shin
DW
,
Kim
J
,
Hwang
JH
. 
High PD-L1 expression is associated with therapeutic response to pembrolizumab in patients with advanced biliary tract cancer
.
Sci Rep
2020
;
10
:
12348
.
7.
Zou
W
,
Wolchok
JD
,
Chen
L
. 
PD-L1 (B7-H1) and PD-1 pathway blockade for cancer therapy: mechanisms, response biomarkers, and combinations
.
Sci Transl Med
2016
;
8
:
328rv4
.
8.
Sun
C
,
Mezzadra
R
,
Schumacher
TN
. 
Regulation and function of the PD-L1 checkpoint
.
Immunity
2018
;
48
:
434
52
.
9.
Dominissini
D
,
Moshitch-Moshkovitz
S
,
Schwartz
S
,
Salmon-Divon
M
,
Ungar
L
,
Osenberg
S
, et al
Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq
.
Nature
2012
;
485
:
201
6
.
10.
Deng
X
,
Su
R
,
Weng
H
,
Huang
H
,
Li
Z
,
Chen
J
. 
RNA N (6)-methyladenosine modification in cancers: current status and perspectives
.
Cell Res
2018
;
28
:
507
17
.
11.
Batista
PJ
,
Molinie
B
,
Wang
J
,
Qu
K
,
Zhang
J
,
Li
L
, et al
m(6)A RNA modification controls cell fate transition in mammalian embryonic stem cells
.
Cell Stem Cell
2014
;
15
:
707
19
.
12.
Fustin
JM
,
Doi
M
,
Yamaguchi
Y
,
Hida
H
,
Nishimura
S
,
Yoshida
M
, et al
RNA-methylation-dependent RNA processing controls the speed of the circadian clock
.
Cell
2013
;
155
:
793
806
.
13.
Li
HB
,
Tong
J
,
Zhu
S
,
Batista
PJ
,
Duffy
EE
,
Zhao
J
, et al
m(6)A mRNA methylation controls T cell homeostasis by targeting the IL-7/STAT5/SOCS pathways
.
Nature
2017
;
548
:
338
42
.
14.
Zheng
G
,
Dahl
JA
,
Niu
Y
,
Fedorcsak
P
,
Huang
CM
,
Li
CJ
, et al
ALKBH5 is a mammalian RNA demethylase that impacts RNA metabolism and mouse fertility
.
Mol Cell
2013
;
49
:
18
29
.
15.
Ma
C
,
Chang
M
,
Lv
H
,
Zhang
ZW
,
Zhang
W
,
He
X
, et al
RNA m(6)A methylation participates in regulation of postnatal development of the mouse cerebellum
.
Genome Biol
2018
;
19
:
68
.
16.
Rubio
RM
,
Depledge
DP
,
Bianco
C
,
Thompson
L
,
Mohr
I
. 
RNA m(6) a modification enzymes shape innate responses to DNA by regulating interferon beta
.
Genes Dev
2018
;
32
:
1472
84
.
17.
Xiang
Y
,
Laurent
B
,
Hsu
CH
,
Nachtergaele
S
,
Lu
Z
,
Sheng
W
, et al
RNA m(6)A methylation regulates the ultraviolet-induced DNA damage response
.
Nature
2017
;
543
:
573
6
.
18.
Han
D
,
Liu
J
,
Chen
C
,
Dong
L
,
Liu
Y
,
Chang
R
, et al
Anti-tumour immunity controlled through mRNA m(6)A methylation and YTHDF1 in dendritic cells
.
Nature
2019
;
566
:
270
4
.
19.
Li
M
,
Liu
F
,
Zhang
F
,
Zhou
W
,
Jiang
X
,
Yang
Y
, et al
Genomic ERBB2/ERBB3 mutations promote PD-L1-mediated immune escape in gallbladder cancer: a whole-exome sequencing analysis
.
Gut
2019
;
68
:
1024
33
.
20.
Wang
X
,
He
Q
,
Shen
H
,
Xia
A
,
Tian
W
,
Yu
W
, et al
TOX promotes the exhaustion of antitumor CD8(+) T cells by preventing PD1 degradation in hepatocellular carcinoma
.
J Hepatol
2019
;
71
:
731
41
.
21.
Lee
SJ
,
Jang
BC
,
Lee
SW
,
Yang
YI
,
Suh
SI
,
Park
YM
, et al
Interferon regulatory factor-1 is prerequisite to the constitutive expression and IFN-gamma-induced upregulation of B7-H1 (CD274)
.
FEBS Lett
2006
;
580
:
755
62
.
22.
Zhang
S
,
Zhao
BS
,
Zhou
A
,
Lin
K
,
Zheng
S
,
Lu
Z
, et al
m(6)A demethylase ALKBH5 maintains tumorigenicity of glioblastoma stem-like cells by sustaining FOXM1 expression and cell proliferation program
.
Cancer Cell
2017
;
31
:
591
606
.
23.
Wang
X
,
Lu
Z
,
Gomez
A
,
Hon
GC
,
Yue
Y
,
Han
D
, et al
N6-methyladenosine-dependent regulation of messenger RNA stability
.
Nature
2014
;
505
:
117
20
.
24.
Sanmamed
MF
,
Chen
L
. 
A paradigm shift in cancer immunotherapy: from enhancement to normalization
.
Cell
2018
;
175
:
313
26
.
25.
Tang
Z
,
Li
C
,
Kang
B
,
Gao
G
,
Li
C
,
Zhang
Z
. 
GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses
.
Nucleic Acids Res
2017
;
45
:
W98
W102
.
26.
Tang
B
,
Yang
Y
,
Kang
M
,
Wang
Y
,
Wang
Y
,
Bi
Y
, et al
m(6)A demethylase ALKBH5 inhibits pancreatic cancer tumorigenesis by decreasing WIF-1 RNA methylation and mediating wnt signaling
.
Mol Cancer
2020
;
19
:
3
.
27.
Ke
S
,
Pandya-Jones
A
,
Saito
Y
,
Fak
JJ
,
Vagbo
CB
,
Geula
S
, et al
m(6)A mRNA modifications are deposited in nascent pre-mRNA and are not required for splicing but do specify cytoplasmic turnover
.
Genes Dev
2017
;
31
:
990
1006
.
28.
Wynn
TA
. 
Myeloid-cell differentiation redefined in cancer
.
Nat Immunol
2013
;
14
:
197
9
.
29.
Zhang
C
,
Samanta
D
,
Lu
H
,
Bullen
JW
,
Zhang
H
,
Chen
I
, et al
Hypoxia induces the breast cancer stem cell phenotype by HIF-dependent and ALKBH5-mediated m(6)A-demethylation of NANOG mRNA
.
Proc Natl Acad Sci U S A
2016
;
113
:
E2047
56
.
30.
He
Y
,
Hu
H
,
Wang
Y
,
Yuan
H
,
Lu
Z
,
Wu
P
, et al
ALKBH5 inhibits pancreatic cancer motility by decreasing long non-coding RNA KCNK15-AS1 methylation
.
Cell Physiol Biochem
2018
;
48
:
838
46
.
31.
Hegde
PS
,
Chen
DS
. 
Top 10 challenges in cancer immunotherapy
.
Immunity
2020
;
52
:
17
35
.
32.
Zhao
Y
,
Shuen
TWH
,
Toh
TB
,
Chan
XY
,
Liu
M
,
Tan
SY
, et al
Development of a new patient-derived xenograft humanised mouse model to study human-specific tumour microenvironment and immunotherapy
.
Gut
2018
;
67
:
1845
54
.
33.
Al-Khami
AA
,
Zheng
L
,
Del Valle
L
,
Hossain
F
,
Wyczechowska
D
,
Zabaleta
J
, et al
Exogenous lipid uptake induces metabolic and functional reprogramming of tumor-associated myeloid-derived suppressor cells
.
Oncoimmunology
2017
;
6
:
e1344804
.
34.
Veglia
F
,
Tyurin
VA
,
Blasi
M
,
De Leo
A
,
Kossenkov
AV
,
Donthireddy
L
, et al
Fatty acid transport protein 2 reprograms neutrophils in cancer
.
Nature
2019
;
569
:
73
8
.
35.
Ostrand-Rosenberg
S
. 
Myeloid derived-suppressor cells: their role in cancer and obesity
.
Curr Opin Immunol
2018
;
51
:
68
75
.
36.
Tseng
YH
,
Ho
HL
,
Lai
CR
,
Luo
YH
,
Tseng
YC
,
Whang-Peng
J
, et al
PD-L1 expression of tumor cells, macrophages, and immune cells in non-small cell lung cancer patients with malignant pleural effusion
.
J Thorac Oncol
2018
;
13
:
447
53
.
37.
Cottrell
TR
,
Taube
JM
. 
PD-L1 and emerging biomarkers in immune checkpoint blockade therapy
.
Cancer J
2018
;
24
:
41
6
.
38.
Diggs
LP
,
Hsueh
EC
. 
Utility of PD-L1 immunohistochemistry assays for predicting PD-1/PD-L1 inhibitor response
.
Biomark Res
2017
;
5
:
12
.
39.
Hirsch
FR
,
McElhinny
A
,
Stanforth
D
,
Ranger-Moore
J
,
Jansson
M
,
Kulangara
K
, et al
PD-L1 immunohistochemistry assays for lung cancer: results from phase 1 of the blueprint PD-L1 IHC assay comparison project
.
J Thorac Oncol
2017
;
12
:
208
22
.
40.
Lee
HH
,
Wang
YN
,
Xia
W
,
Chen
CH
,
Rau
KM
,
Ye
L
, et al
Removal of N-linked glycosylation enhances PD-L1 detection and predicts anti-PD-1/PD-L1 therapeutic efficacy
.
Cancer Cell
2019
;
36
:
168
78
.

Supplementary data