Tumor-derived CCL2 drives tumor growth and immunosuppression in IDH1-mutant cholangiocarcinoma

Abstract

Background and aims:

Isocitrate dehydrogenase 1 (IDH1)-mutant cholangiocarcinoma (CCA) is a highly lethal subtype of hepatobiliary cancer that is often resistant to immune checkpoint inhibitor therapies. We evaluated the effects of IDH1-mutations in CCA cells on the tumor immune microenvironment and identify opportunities for therapeutic intervention.

Approach and results:

Analysis of 2,606 human CCA tumors using deconvolution of RNA-sequencing data identified decreased CD8 T cell and increased M2-like tumor-associated macrophage (TAM) infiltration in IDH1-mutant compared to IDH1-wild type tumors. To model the tumor immune microenvironment of IDH1-mutant CCA in vivo, we generated an isogenic cell line panel of mouse SB1 CCA cells containing a heterozygous IDH1 R132C (SB1^mIDH1) or control (SB1^WT) cells using CRISPR-mediated homology directed repair. SB1^mIDH1 cells recapitulated features of human IDH1-mutant CCA including D-2-HG production and increased M2-like TAM infiltration. SB1^mIDH1 cells and tumors produced increased levels of CCL2, a chemokine involved in recruitment and polarization of M2-like TAMs compared to wild type controls. In vivo neutralization of CCL2 led to decreased M2-like TAM infiltration, reduced tumor size, and improved overall survival in mice harboring SB1^mIDH1 tumors.

Conclusions:

IDH1-mutant CCA is characterized by increased abundance of M2-like TAMs. Targeting CCL2 remodels the tumor immune microenvironment and improves outcomes in preclinical models of IDH1-mutant CCA, highlighting the role for myeloid-targeted immunotherapies in the treatment of this cancer.

Graphical Abstract

Introductory statement

Cholangiocarcinoma (CCA) is an aggressive cancer that arises from biliary epithelial cells (1). CCAs exhibit diversity in their underlying genetic alterations, tumor microenvironments (TMEs), and responses to treatment (2). Prior analyses of CCA have identified associations between molecular subtypes of CCA and patterns of immune cell infiltration in the TME (3,4). Isocitrate dehydrogenase 1 (IDH1) is recurrently mutated in CCA, with IDH1 mutations found in approximately 20% of intrahepatic CCAs (5). These mutations result in neomorphic enzymatic activity of IDH1 that leads to the conversion of α-ketoglutarate to D-2-hydroxyglutarate (D-2-HG) (6). D-2-HG, which can accumulate at high levels inside cancer cells (7,8), has been implicated in the dysregulation of multiple cellular processes including epigenetic regulation, activity of signaling pathways, and metabolism (9). In addition, D-2-HG-independent actions of mutant IDH1 have been shown to alter properties of cancer cells to promote tumorigenesis (10).

The presence of IDH1 mutations has previously been associated with an immunosuppressive TME and may predict for inferior responses to immune checkpoint inhibitor therapies used in the management of advanced CCA (11,12). Prior work has implicated mutant IDH1 in diminishing T cell recruitment and responses (4,13), and our recent work (3) associated mutant IDH1 in CCA with the presence of M2-like tumor-associated macrophages (TAMs), an immunosuppressive, tumor promoting TAM subtype (14). However, the mechanisms which promote M2-like TAM accumulation in IDH1-mutant CCA remain to be elucidated. One potential limitation of prior studies of IDH1-mutant CCA is a relative lack of experimental models utilizing immunocompetent mice, thereby hindering the ability to directly examine the impact of IDH1 mutations on immune cells.

In this study, we report the generation of Idh1-mutant isogenic mouse CCA cell lines and investigate how mutant IDH1 shapes the TME. We show that IDH1-mutant CCA cells release cytokines, including CCL2, that promote M2-like TAM recruitment/polarization, and that Idh1-mutant tumors are sensitive to immunotherapies targeting CCL2.

Experimental Procedures

Cell Lines

SB1 (15) cells were generously provided by Dr. Gregory J. Gores (Mayo Clinic, Rochester, MN). SNU-1079 (16) cells were generously provided by Dr. Gregory B. Lesinski (Emory University, Atlanta, GA). Cells were maintained in RPMI 1640 Medium with L-glutamine (Thermo Fischer Scientific 11-875-093) supplemented with 10% heat-inactivated fetal bovine serum (Gemini 100–106), and 100U/mL penicillin/streptomycin (Thermo Fisher Scientific, 15070063). Cells were maintained at 37°C with 5% CO2 in a humidified incubator and routinely tested for Mycoplasma (Johns Hopkins Genetic Resources Core Facility). Cell stocks were authenticated via short tandem repeat profiling through the Johns Hopkins Genetic Resources Core Facility or the ATCC.

Estimation of immune cell proportions in human CCA tumors

The study cohort consisted of 2,606 patients diagnosed with CCA, for which DNA and RNA sequencing data were available. Patients were retrospectively selected from the Tempus real-world multimodal database, and had undergone Tempus xT assay for DNA and the latest version of the Tempus RNA-sequencing assay (xR) (17). Patients were stratified based on the presence (Mutant) or absence (WT) of IDH1 pathogenic or likely pathogenic mutations. Additional details in Supporting Methods.

Generation of an isogenic R132C IDH1-mutant (SB1^mIDH1) or wild type control (SB1^WT) cell line panel

CRISPR homology-directed repair

2x10⁵ cells were transfected by nucleofection using a Lonza 4D-Nucleofector system with Lonza SF buffer (Lonza V4XC-2032) using code DS-130. A ribonucleoprotein complex comprised of an sgRNA targeting IDH1 (5’-TCATTGGCCGACATGCATAT-3’) and Alt-R S.p. HiFi Cas9 enzyme (IDT 1081060), Alt-R Cas9 Electroporation Enhancer (IDT 1075915), and single-stranded oligodeoxynucleotides (ssODNs) to serve as homology-directed repair templates for the desired genome editing outcome. ssODNs contained non-coding mutations to alter the PAM sequence were introduced. ssODN sequences for R132C mutation repair were (R132C A 5’-ACAAACATAGATATTGGAGCTAAAGGTCTGTGAAAAATAAATACGGAAAAAATCAAATTATGCTCACTCAAATGACTTACTTGGTCGCCATATGCATGACAGCCAATGATGATGGGTTTTACCCAGCCTGTCACTAGCCGGGGGATATTTTTGCAGATAATAGCTTCCCTGAAGACAGTGCCACCCAGAATGTTTCGGAT-3’ and R132C B 5’-ACAAACATAGATATTGGAGCTAAAGGTCTGTGAAAAATAAATACGGAAAAAATCAAATTATGCTCACTCAAATGACTTACTTGGTCTCCATATGCATGACAGCCAATGATGATGGGTTTTACCCAGCCTGTCACTAGCCGGGGGATATTTTTGCAGATAATAGCTTCCCTGAAGACAGTGCCACCCAGAATGTTTCGGAT-3’) and for R132R repair were (R132R A 5’-AACAAACATAGATATTGGAGCTAAAGGTCTGTGAAAAATAAATACGGAAAAAATCAAATTATGCTCACTCAAATGACTTACTTGGTCGCCATATGCATGCCGGCCAATGATGATGGGTTTTACCCAGCCTGTCACTAGCCGGGGGATATTTTTGCAGATAATAGCTTCCCTGAAGACAGTGCCACCCAGAATGTTTCGGA-3’ and R132R B 5’-AACAAACATAGATATTGGAGCTAAAGGTCTGTGAAAAATAAATACGGAAAAAATCAAATTATGCTCACTCAAATGACTTACTTGGTCTCCATATGCATGCCGGCCAATGATGATGGGTTTTACCCAGCCTGTCACTAGCCGGGGGATATTTTTGCAGATAATAGCTTCCCTGAAGACAGTGCCACCCAGAATGTTTCGGA-3’). For further details see Supporting Methods.

Cellular proliferation assays

SB1^mIDH1 or control SB1^WT cells were plated in 96-well plates at 5x10³ cells/well. Proliferation on replicate plates was determined using the alamarBlue HS cell viability reagent (Thermo Fisher A50100) per manufacturer’s protocol at the indicated time points.

In vitro AG-120 and neutralizing CCL2 treatments and cell viability assays

For measurement of survival of SB1^mIDH1 or SB1^WT clones treated with AG-120 (Cayman Chemical Company 19894), cells were plated at a density of 1x10³ cells/well in a 96 well and after 24 hours, media was changed to contain the AG-120 or DMSO vehicle control. After 72 hours, viability was assayed using CellTiter-Glo 2.0 (Promega G9242). For in vitro treatment of cells with a CCL2 neutralizing antibody (Clone 2H5, BioXCell BE0185) or control Armenian hamster IgG (BioXCell BE0091), a similar procedure was followed. For measurement of D-2-HG, cells were washed with PBS, pelleted, and snap frozen in liquid nitrogen prior to further analysis.

Animal Experimentation

Male C57BL/6J mice purchased from Jackson Laboratories and were 8 weeks of age at the start of experiments. Male mice were utilized in this study as the SB1 cell line was isolated from a male mouse, and we sought to minimize immune responses due to sex mismatched cells. Mice were maintained in accordance with the Institutional Animal Care and Use Committee guidelines, fed a standard diet, and not fasted prior to the initiation of an experiment/assessment. All interventions were performed during the light cycle.

Subcutaneous tumor implantation

1x10⁶ cancer cells were inoculated into the lower left flank of mice as previously described and measured at regular intervals (18). Tumor volume was calculated as (major axis x minor axis²)/2.

Orthotopic liver injections

1x10⁶ cancer cells in 40μL of growth factor reduced Matrigel (Corning 354230):PBS were injected into the livers of mice as previously described (19).

In vivo treatments

AG-120 treatment: Once tumors reached 50-100mm³, mice were assigned to receive either a vehicle of 0.5% Methyl Cellulose (cP 400) (Sigma Aldrich M0262-500G) + 0.2% Tween 80 (Sigma Aldrich P1754) in water or 150mg/kg AG-120 twice daily via oral gavage. For orthotopic tumor experiments treatment was started 7 days post-tumor implantation.

CCL2 neutralization: 200μg of a CCL2-neutralizing antibody (Clone 2H5, BioXCell BE0185) or Armenian hamster IgG (BioXCell BE0091) was administered twice weekly via intraperitoneal injection.

Survival analyses

Mice were monitored and euthanized if they exhibited signs of poor condition per institutional guidelines. Kaplan-Meyer analysis of survival was used to visualize survival with significant differences between groups determined by log-rank testing.

Immunohistochemistry (IHC) and tissue staining

Human CCA tumor specimens were obtained in conjunction with the Johns Hopkins University School of Medicine Department of Pathology. H&E and Masson’s trichrome staining were performed by the Johns Hopkins Oncoloty Tissue Services (OTS) core facility. IHC staining of mouse tumor sections was performed as previously described (20). Multiplex staining of human CCA tumor sections was performed by the Johns Hopkins OTS core facility using a Ventana Discovery Ultra autostainer (Roche Diagnostics) using CC1 buffer and a 32 minute antigen retrieval. Antibodies are listed in Supporting Methods.

Immunoblotting

Immunoblotting was performed as previously described (20). Antibodies include CCL2 (Novus Biologicals NBP1-07035, 1:1,000), GAPDH (Cell Signaling Technologies 5174S, 1:10,000), and HSP90 (Cell Signaling Technologies 4877S, 1:10,000) and an HRP-conjugated secondary antibody (Cell Signaling 7074S).

Quantitative real-time PCR

RNA was extracted using RNeasy Mini kit (Qiagen 74104) and cDNA was prepared using iScript cDNA synthesis kit (Bio-Rad 1708890). Gene expression was quantified with Taqman gene expression assays (Thermo Scientific 4453320; assay Mm00441242_m1 - Ccl2, assay Mm99999915_g1 - Gapdh) on a StepOne Real-Time PCR system (Applied Biosystems). Samples were normalized against Gapdh.

Determination of cytokine secretion

Conditioned media (CM) from of SB1^mIDH1 and SB1^WT clones was collected after 72 hours and filtered through a 0.45μm PES filter. CM volume was normalized to the cell count at the end of conditioning and assessed for cytokine secretion using the Proteome Profiler Mouse XL Cytokine Array (R&D Systems ARY028) and CCL2 by ELISA (R&D Systems DY479-05). Relative cytokine intensities were normalized in comparison to reference spots on the same membrane and same development times between membranes. Expression ratios between SB1^mIDH1 and control SB1^WT CM-exposed membranes were calculated by comparing the intensities for each cytokines using FiJi (21).

Measurement of D-2-HG

D-2-HG was quantified using an in-house mass-spec based assay. See Supporting Methods for full details.

Quantification and Statistical Analysis

Statistical analysis was performed with Prism 10 software using unpaired Student’s t-test, repeated measures one-way ANOVA, repeated measures two-way ANOVA, and logrank testing. ANOVAs were followed by the Holm-Šídák multiple comparisons test. P values < 0.05 were considered significant. Data are represented as mean ± SD unless otherwise noted. Significance was designated as follows: *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001.

Results

IDH1-mutant CCAs have a T cell suppressing tumor immune microenvironment with increased M2-like macrophages compared to IDH1-wild type tumors

Mutant IDH1 is associated with an immunosuppressive tumor phenotype (4,22,23). As prior studies were at least in part limited by the number of patient samples analyzed in this rare disease, we performed a larger-scale analysis with a focus on understanding immunogenomic differences between IDH1-wild type and IDH1-mutant CCAs. We utilized the Tempus database that contains comprehensive genomic profiling and RNA-sequencing information for tumor biopsy or resection samples from patients who have undergone genomic and transcriptomic profiling as part of their clinical care. We retrospectively analyzed de-identified transcriptomic data from either IDH1-wild type (WT; n = 2289) or IDH1-mutant (Mutant; n = 317) CCAs (Figure 1A). We performed estimations of immune cell type proportions within the tumors by leveraging a machine learning algorithm (24) to deconvolute bulk RNA-sequencing data, resulting in the largest analysis of immune infiltration patterns for IDH1-mutant CCA to date. We found multiple differences between the WT and IDH1-mutant groups (Figure 1B, Supplemental Figure S1A) including significantly reduced total immune cells and CD8 T cells in IDH1-mutant tumors . There was no difference in the proportions of regulatory T cells, neutrophils, or monocytes between WT and Mutant tumors. However, despite a decrease in total immune cells, we observed a significant increase in the proportion of M2-like tumor-associated macrophages (TAMs) in the IDH1-mutant group (Figure 1B) which correlated with decreased CD8 T cell infiltration (Supplemental Figure S1B). As deconvolution of RNA-sequencing data only estimates immune cell proportions, we performed IHC for CD206, a marker of M2 macrophages (25), using FFPE tumor samples from with CCA. Consistent with the transcriptomic data, patients with IDH1-mutant tumors had significantly more M2-like TAMs compared to patients with IDH1-wild type tumors (Figure 1C). M2-like TAMs are associated with tumor growth, metastasis, and immune escape in multiple cancers including CCA (14,26,27). Our findings suggest increased M2-like TAMs infiltration is a key feature of an immunosuppressive TME in IDH1-mutant CCA. A limitation of this human tissue analysis is the known enrichment of ARID1A and PBRM1 alterations in IDH1-mutant CCA alongside mutual exclusivity with alterations in TP53 and BAP1, among others (3). Therefore, it is possible that associations between IDH1-mutant CCA and patterns of immune infiltration are mediated by interactions with other genomic alterations rather than a direct consequence of the IDH1 mutation. We aimed to further confirm a direct link between IDH1 mutations and increased M2-like TAMs in a preclinical model.

Figure 1. The tumor immune microenvironment differs between IDH1-wild type and mutant cholangiocarcinoma.

A) Workflow for establishing estimates of infiltrating immune cell proportions in a cohort of 2606 human cholangiocarcinoma tumor samples. B) Violin plots representing proportion of total cells for immune cell types stratified by IDH1 mutation status. n = 2289 IDH1-wild type and n = 317 IDH1-mutant samples. The width of the violins reflects the density of data points at different expression levels. Log2 transformed fold change of proportions for IDH1-mutant vs. WT samples along with Wilcoxon-Test p-values are presented for each immune cell type. C) Representative images of IHC staining for CD206 and quantification of CD206+ cells/high powered field (HPF) in IDH1-wild type (WT) and IDH1-mutant (MUT) human CCA samples. n = 8 WT and n = 5 MUT samples. Scale bar = 100μm.

Generation of an isogenic Idh1-mutant CCA mouse cell line

While prior studies have utilized cell lines derived from genetically engineered mouse models of IDH1-mutant CCA (4), there were no readily available syngeneic mouse CCA cell lines with which to directly compare the effects of a heterozygous Idh1 mutation to those of wild type within the same CCA cell line background. Therefore, we utilized CRISPR-mediated homology-directed repair (HDR) to generate an isogenic cell line panel utilizing the previously characterized SB1 mouse CCA cell line (15). SB1 cells underwent nucleofection to enable delivery of HiFi Cas9, an sgRNA specific to Idh1, and DNA repair templates containing either an IDH1 R132C or IDH1 R132R silent mutation coding sequence (Figure 2A). Clones were screened for the heterozygous incorporation of either the R132C Mutant (SB1^mIDH1) or WT (SB1^WT) sequences at the gDNA level (Figure 2B) and for expression of these sequences as determined by cDNA sequencing (Supplementary Figure S2A). SB1^WT and SB1^mIDH1 mutant clones maintained expression of IDH1 at the protein level (Supplementary Figure S2B). Using this process, we established a panel of multiple independently derived SB1^WT and SB1^mIDH1 clones. Proliferation of SB1^mIDH1 and SB1^WT clones was similar in vitro (Figure 2C). SB1^mIDH1 clones produced significantly more D-2-HG compared to SB1^WT clones (Figure 2D). Treatment of SB1^mIDH1 clones with the mutant IDH1 inhibitor, AG-120 (28), in vitro resulted in a significant reduction in D-2-HG (Figure 2D). Treatment with AG-120 did not reduce cell viability in either the SB1^WT or SB1^mIDH1 clones, similar to prior studies of other non-isogenic Idh1-mutant CCA mouse cell lines (Supplementary Figure S2C). (4). Together, these data show that SB1^mIDH1 clones have key molecular characteristics expected of Idh1-mutant cells, including production of D-2-HG and molecular responses to mutant IDH1 inhibitors in vitro.

Figure 2. Development and characterization of an isogenic murine cell model of Idh1 R132C-mutant cholangiocarcinoma.

A) Schematic of CRISPR homology-directed repair editing strategy for the generation of IDH1 R132C mutant clones (SB1^mIDH1) and IDH1 R132R wild type (SB1^WT) SB1 clones. B) Representative genomic DNA sequencing traces for SB1 parental, SB1^WT, and SB1^mIDH1 mutant clones illustrating successful editing of the Idh1 genomic locus in SB1^mIDH1 and SB1^WT clones. C) Proliferation, represented as normalized fold change relative to day 1 measurements for each group, of the parental SB1 cell line, SB1^WT, and SB1^mIDH1 mutant clones as measured by an Alamar Blue assay at the indicated time points. n = 5 clones per genotype. All lines were assayed in technical triplicate per time point. D) Amount of intracellular D-2-HG in the presence of 1μM AG-120 or vehicle control for 72 hours as measured by LC-MS-based assay. n = 3 clones per genotype per treatment condition. Two-way ANOVA used to compare across genotype and AG-120 treatment groups. E) Weight of orthotopic SB1^WT and SB1^mIDH1 tumors collected at day 35 post-implantation. n = 9 SB1^WT and 6 SB1^mIDH1 derived tumors with 2 clones per group. F). Amount of D-2-HG per gram of tumor from orthotopic IDH1 SB1^WT and SB1^mIDH1 tumors at day 32 post-implantation. n = 4 SB1^WT and 6 SB1^mIDH1 tumors. G) Representative images of H&E stained tissue sections from orthotopic SB1^WT and SB1^mIDH1 tumors. Scale bar = 100μm. H) Tumor volume measurements of mice harboring subcutaneous SB1^mIDH1 tumors treated with either AG-120 or vehicle control. n = 6 vehicle control and 10 AG-120 treated mice. Comparison of curves made using non-linear fit analysis. I) Amount of D-2-HG per gram of tumor from subcutaneous SB1^mIDH1 tumors treated with AG-120 or vehicle for 3 days prior to tumor collection. n = 5 tumors per group.

We next characterized the SB1^WT and SB1^mIDH1 isogenic cell line panel in vivo. There was no difference in tumor growth after subcutaneous injection of SB1 parental, SB1^mIDH1, or SB1^WT lines into syngeneic C57BL/6 mice (Supplementary Figure S3A). As there can be differences between subcutaneous and orthotopic tumors that affect the growth of hepatobiliary tumor models (18,29), we also performed orthotopic injection of SB1^mIDH1 and SB1^WT clones into the livers of C57BL/6 mice and found no difference in tumor growth (Figure 2E) or overall survival (Supplemental Figure S3B) between mice with SB1^mIDH1 and SB1^WT tumors. As expected, SB1^mIDH1 tumors contain significantly more D-2-HG compared to SB1^WT tumors (Figure 2F). Histopathologic analysis of SB1^WT and SB1^mIDH1 orthotopic tumors did not reveal significant differences between tumors arising from the various clones by H&E (Figure 2G) or Masson’s Trichrome (Supplemental Figure S3C) staining. Treatment with AG-120 reduced SB1^mIDH1 tumor growth compared to vehicle control and significantly decreased D-2-HG levels (Figure 2H–I). Collectively, these results demonstrate that the knock-in of a heterozygous IDH1 R132C mutation into the SB1 cell line results in a syngeneic cell line model of Idh1-mutant CCA that recapitulates major features of IDH1-mutant CCAs. These SB1^mIDH1 mutant clones can be directly compared to clones from the same cell line background harboring a control, silent mutation in both in vitro and in vivo experiments, allowing for further study of the effects of mutant IDH1 in CCA.

Orthotopic injection of SB1^mIDH1 clones recapitulates features of the tumor immune microenvironment of IDH1-mutant CCA

To examine the immune effect of Idh1 mutations, we collected SB1^WT and SB1^mIDH1 SB1 orthotopic tumors and performed IHC staining for immune cell markers. CD45+ cell infiltration, representing the total number of immune cells in the tumors, did not differ between SB1^mIDH1 and SB1^WT tumors (Figure 3A). There were no significant differences in the abundance of CD8+ (cytotoxic T cells), CD4+ (helper T cells), and FoxP3+ (regulatory T cells) cells between SB1^WT and SB1^mIDH1 tumors, though there was a trend toward fewer CD8+ T cells in SB1^mIDH1 tumors (Figure 3B–D). PD-L1 expression appeared similar between SB1^WT and SB1^mIDH1 tumors (Supplemental Figure S4A). Staining for F4/80, a pan-macrophage marker (30), and CD68, a marker often upregulated in M1-like macrophages but that is expressed by all macrophages (25), revealed no difference in total macrophage infiltration into SB1^mIDH1 and SB1^WT tumors (Figure 3E, Supplemental Figure S4B–C). There was an increase in CD206+ cells in SB1^mIDH1 tumors (Figure 3F), similar to our analysis of immune infiltration in human CCAs. Treatment of SB1^mIDH1 tumors with AG-120 resulted in a trend toward increased CD8+ (Supplementary Figure S4D–E) T cell infiltration, but did not alter CD206+ (Supplementary Figure S4F–G) cell infiltration compared to vehicle control. These findings suggest that the isogenic cell line panel established in this study recapitulates immunologic features of IDH1-mutant CCA, including increased M2-like TAMs.

Figure 3. SB1^mIDH1 tumors have a distinct tumor immune microenvironment with an increase in M2-like tumor associated macrophages.

IHC staining was performed on orthotopic SB1^WT and SB1^mIDH1 tumor sections. Representative images (left) and quantification of staining (right) for A) CD8, B) CD4, C) FoxP3, D) F4/80, E) CD68, and F) CD206. n ≥ 4 tumors per tumor genotype for each IHC stain. Scale bar = 100μm except for panel D where scale bar = 50μm.

IDH1-mutant CCA cells produce and secrete elevated levels of CCL2

The immune differences between tumors from SB1^mIDH1 and SB1^WT clones led us to hypothesize that intrinsic cancer cell factors driven by the presence of mutant IDH1 could promote the recruitment of TAMs or polarization toward an M2-like phenotype. One way in which tumor cells impact the recruitment and polarization of macrophages is through secreted ligands (31). We collected conditioned media from SB1^mIDH1 and SB1^WT clones and performed a multiplex cytokine array to examine differences in secreted ligands that may impact macrophages in CCA. We identified a number of differentially secreted factors with putative effects on macrophage recruitment and polarization. SB1^mIDH1 cells had increased levels of factors implicated in augmenting anti-inflammatory effects of macrophages including CCL20 (32), GDF-15 (33), and CCL2 (34), but no change in M-CSF (Figure 4A, Supplementary Table S1). Notably, CCL2 has a well-established role in attracting myeloid cells, but also influences polarization in a context dependent manner. Prior studies have highlighted CCL2’s ability to promote tumor progression through the actions of TAMs (35). Studies of mutant IDH1 in CCA have suggested CCL2 expression is linked with active mutant IDH1 (4). Additionally, multiple therapeutic strategies targeting the CCL2/CCR2 signaling axis are being explored as part of multimodal treatment approaches aimed at improving antitumor immune responses or overcoming resistance to other anticancer therapies (36–38). Understanding the role of CCL2 in IDH1-mutant CCA may provide a rationale for the use of these CCL2/CCR2 targeted agents in IDH1-mutant CCA. Therefore, we investigated the increased expression of CCL2 in SB1^mIDH1 clones and the role of CCL2 in IDH1-mutant CCA.

Figure 4. CCL2 protein expression is increased in IDH1 mutant cells.

A) Representative cytokine array blot exposed to conditioned media derived from a SB1^WT or SB1^mIDH1 cell line (left) with heat map of relative pixel intensity (arbitrary units) for selected cytokines. B) Immunoblot for CCL2 in SB1 SB1^WT and I SB1^mIDH1 clones. GAPDH was used as a loading control. C) Quantification of CCL2 band intensity relative to GAPDH normalized to the average of SB1^WT clones. n = 3 clones per genotype. D) Concentration of CCL2 in conditioned media normalized to cell count from SB1^WT and SB1^mIDH1. n ≥ 2 clones per genotype performed in technical duplicate. E) Immunoblot for CCL2 in SB1^mIDH1 mutant clones treated with 1μM AG-120 (+) or vehicle control (−) for 72 hours. HSP90 was used as a loading control. F) Representative images of CCL2 staining by IHC in SB1^WT or SB1^mIDH1 orthotopic tumors. Scale bar = 100μm. G) Representative images of IHC staining for CCL2 (brown) and αSMA (purple) in IDH1-wild type (WT) and IDH1-mutant human cholangiocarcinoma tumors. n = 5 IDH1 Mutant and n = 6 WT samples. Scale bar = 100μm. Each image is from a unique patient sample.

SB1^mIDH1 and SB1^WT clones did not differ in their expression of the RNA transcript for Ccl2 (Supplementary Figure S5A), though there was a non-significant trend towards decreased Ccl2 RNA levels with AG-120 treatment in the SB1^mIDH1 but not SB1^WT clones. However, intracellular CCL2 levels measured by immunoblotting and secreted CCL2 measured by ELISA were significantly higher in SB1^mIDH1 compared to SB1^WT clones (Figure 4B–D). This suggests that CCL2 protein production, degradation, or secretion is differentially affected by the presence of mutant IDH1. This is not entirely unexpected given that multiple prior studies have highlighted a lack of concordance between RNA expression and protein expression across multiple human and mouse models (39–42). In addition, we found that AG-120 did not alter CCL2 RNA or protein expression in SB1^mIDH1 clones (Figure 4E, Supplementary Figure S5A–B), and did not decrease CCL2 expression in SNU-1079, an IDH1-mutant human CCA cell line (16) (Supplementary Figure S5C), suggesting that inhibition of mutant IDH1 is not sufficient to reduce CCL2. To determine if CCL2 protein expression was increased in IDH1-mutant tumors in vivo, we performed IHC staining for CCL2. We found increased CCL2 expression in cancer cells in SB1^mIDH1 compared to SB1^WT tumors (Figure 4F, Supplementary Figure S5D). We next examined CCL2 expression in human tumor samples from patients with IDH1-mutant and IDH1-wild type CCA (Figure 4G, Supplementary Figure S5E). We observed increased CCL2 staining (brown stain) in the IDH1-mutant samples, with CCL2 originating predominantly from cancer cells rather than alpha smooth muscle actin positive stromal fibroblast cells (Figure 4G). These data suggest that cells with heterozygous IDH1 R132C mutations have increased CCL2 protein expression and secretion, and that mutant IDH1 inhibitors such as ivosidenib may not be sufficient to reduce CCL2 levels in IDH1-mutant cancer cells.

CCL2 neutralization reduces tumor growth and alters the tumor immune microenvironment in IDH1-mutant tumors

We hypothesized that CCL2 may be responsible for the increased M2-like macrophage infiltration in SB1^mIDH1 tumors, and thus is a potential therapeutic target in IDH1-mutant CCA. To test this hypothesis, we established orthotopic liver tumors using SB1^mIDH1 and SB1^WT clones and treated the mice with either a CCL2 neutralizing (αCCL2) or an isotype control antibody. While treatment with αCCL2 did not alter overall survival or tumor size in mice with SB1^WT tumors, αCCL2 significantly increased overall survival and decreased tumor size in mice with SB1^mIDH1 mutant tumors (Figure 5A–B). We performed IHC staining for CCL2 in SB1^mIDH1 and SB1^WT tumors and found that, consistent with prior experiments, SB1^mIDH1 tumors had increased CCL2 tumors with control treatment, and αCCL2 treatment reduced CCL2 levels in both SB1^mIDH1 and SB1^WT tumor groups (Figure 5C). Given the lack of efficacy of αCCL2 treatment in mice with SB1^WT tumors, these findings suggest that increased CCL2 in SB1^mIDH1 tumors is a driver of tumor growth and progression in Idh1-mutant CCA in our models. This could suggest that there may be immune-driven or cancer cell directed effects of CCL2 neutralization in IDH1-mutant CCA responsible for this difference in outcome.

Figure 5. CCL2 neutralization reduces in vivo tumor growth in IDH1-mutant cholangiocarcinoma.

A) Kaplan-Meier survival analysis for mice with orthotopic SB1^WT or SB1^mIDH1 tumors treated with either an isotype control (Isotype) or a CCL2 neutralizing antibody (αCCL2). Log-rank testing used to compare Iso vs. αCCL2 curves within each tumor genotype with multiple comparisons testing adjusted for by setting total alpha among all comparisons at 0.05. n = 6-10 mice per group. B) Orthotopic tumor weights in mg 24 days post tumor implantation for SB1^WT and SB1^mIDH1 tumors treated with an isotype control or CCL2 neutralizing antibody. n = 8-10 mice per group. One-way ANOVA with post-hoc Holm-Sidak’s multiple comparison testing correction used to evaluate differences between isotype and αCCL2 treatment within each genotype. C) Representative images of CCL2 IHC staining in tumors from (B). Scale bar = 100μm. D) Representative images of IHC staining of orthotopic tumors from (B) for CD206 with quantification of staining presented in E). n = 4-5 tumors per group. Scale bar = 100μm. F) Representative images of IHC staining of orthotopic tumors from (B) for CD8 with quantification of staining presented in G). n = 4-5 tumors per group. Scale bar = 100μm.

As CCL2 plays a major role in shaping the immune milieu in cancers (43), we examined SB1^mIDH1 and SB1^WT tumors treated with either αCCL2 or isotype control for differences in infiltrating immune cells. As we previously demonstrated that IDH1-mutant tumors have increased M2-like TAM infiltration, we hypothesized that CCL2 neutralization may reduce the abundance of these immunosuppressive cells in IDH1-mutant tumors. SB1^WT tumors treated with αCCL2 or isotype control had similar numbers of CD206+ cells, while SB1^mIDH1 tumors treated with αCCL2 had significantly fewer CD206+ cells compared to isotype control (Figure 5D–E). Next, we examined these tumors for changes in CD8+ T cell infiltration, as increased CD8+ T infiltration has been associated with increased antitumor immune responses and improved likelihood of response to mutant-IDH1 targeted treatment (4). There were significantly more CD8+ T cells in SB1^mIDH1 tumors treated with αCCL2 compared to control, while there was no significant difference in the number of CD8+ T cells in the SB1^WT group (Figure 5F–G). Since we observed differences in the tumor growth and immune cell infiltration in the SB1^mIDH1 group, we performed additional analyses and found no change in CD45+ , F4/80+, FoxP3+, or CD4+ cell infiltration with αCCL2 (Supplementary Figure S6A–D). Despite these findings, there were no significant differences in cleaved caspase-3 levels among the experimental groups (Supplementary Figure S6E–F), though there was a trend toward increased cleaved caspase-3 in SB1^mIDH1tumors treated with αCCL2. We investigated effects of αCCL2 on the proliferation of SB1^WT and SB1^mIDH1 clones in vitro and observed no inhibition of SB1 clone growth (Supplementary Figure S6G). Together, these findings show a beneficial impact of CCL2 neutralization that is specific to IDH1-mutant cells and remodels the tumor immune microenvironment to decrease features of immunosuppression, highlighting a possible role for immune activation in limiting the growth of IDH1-mutant CCA.

Discussion

Advanced CCA remains a devastating diagnosis, with a median survival rate of approximately one year (44). Despite efforts to subtype CCA based on genomic alterations , the first-line treatment of CCA predominantly follows a “one-size-fits-all” approach of chemotherapy combined with anti-PD1/anti-PDL1 immunotherapy (12). Understanding the relationship between specific molecular alterations and the TME, and the mechanisms driving these associations, is critically important for developing targeted and effective therapies. Prior work demonstrated that IDH1-mutant CCA clusters distinctly from other subtypes of CCA (3,45,46), and that treatment with a mutant IDH1 inhibitor can reprogram the TME of IDH1 mutant tumors (4,47). However, a direct comparison of the IDH1-mutant and wildtype TME in CCA has not, to our knowledge, been performed.

We developed an isogenic mouse cell line system that can be used to directly compare cells with or without a heterozygous Idh1 mutation and utilized in vivo in immunocompetent mice. SB1^mIDH1 cells generated in this study recapitulate major features of IDH1-mutant human CCAs, including production of D-2-HG and molecular responses to mutant IDH1 inhibitors. Using our isogenic cell line panel, we demonstrated that Idh1-mutant tumor cells promote the accumulation of M2-like TAMs in part through elevated CCL2. CCL2 neutralization remodels the TME toward a less immunosuppressive state, inhibits tumor growth, and improves survival in mice with Idh1-mutant tumors. These data add to the accumulating evidence that IDH1-mutations may benefit tumors y inducing an immune suppressive TME, and also support CCL2 inhibition as a therapeutic strategy for IDH1-mutant CCA.

Interestingly, we observed that only CCL2 protein expression was higher in IDH1-mutant patient samples and mouse cell line clones in our study, highlighting an important role for screening for effects of mutant IDH1 at both the RNA and protein levels. Our study did not fully uncover a mechanism driving a difference in CCL2 protein expression in IDH1-mutant tumors. Potential mechanisms, including effects of mutant IDH1 on regulation of translation or on secretory pathways is warranted. While we investigated the effects of CCL2 in this study, there are likely additional effects of mutant IDH1 that influence the TME and macrophages that remain to be elucidated.

While a strength of our investigation lies in the direct isolation of the effect of IDH1 mutations on the TME, a limitation is that our model was derived from the SB1 mouse cell line, which also harbors alterations in YAP and Akt. AKT1 mutations are associated with decreased M2-like TAM infiltration (48), and the patient cohort in our data did not have any patients with co-occurring IDH1 and AKT1 mutations, similar to prior CCA sequencing that suggest this combination of mutations is rare (45). Together, this suggests that mutant IDH1 is sufficient to induce the M2-like TAM enriched phenotype we have observed. Given the lack of availability of IDH1-mutant CCA cell lines suitable to study the immune microenvironment in vivo, we believe the generation of an isogenic cell line panel consisting of multiple, independently derived clones that model WT and mutant IDH1 in a CCA background with the same and thus directly comparable set of co-occurring genetic alterations represents a major strength of this study. This model recapitulates major features of the tumor immune microenvironment observed in human CCA, such as increased abundance of M2-like TAMs, supporting its validity as a tool in the preclinical exploration of IDH1-mutant CCA.

Understanding the effects of mutant IDH1 remains critically important even in the era of mutant IDH1 inhibitors. Prior work evaluating macrophage populations in IDH1-mutant gliomas has implicated D-2-HG as playing a key role in promoting immunosuppressive actions of macrophages (49), suggesting that reductions in D-2-HG levels may reverse features of immunosuppression in IDH1-mutant cancers. While treatment with ivosidenib can greatly decrease D-2-HG levels, prior studies have shown that D-2-HG levels in patients with IDH1-mutant tumors are still elevated with ivosidenib treatment compared to IDH1-wild type cancers (50). IDH1 inhibitors have produced modest overall responses for patients with IDH1-mutant CCA (51). Importantly, multiple mechanisms of acquired resistance to mutant IDH inhibitors have been reported (52,53), often resulting in increased D-2-HG production. In summary, mutant IDH1 in CCA plays a critical role in reshaping the TME through mechanisms that do not necessarily depend on direct action of D-2-HG on tumor-infiltrating immune cells. Our data suggest that further investigation of myeloid targeted immunotherapy strategies in IDH1-mutant CCA is warranted for advancing the treatment of IDH1-mutant CCA.

Supplementary Material

Figure S1

Supplemental Figure S1. Immune cell proportions estimated by deconvolution of RNA-sequencing data in IDH1-wild type and mutant cholangiocarcinoma. A) Violin plot representation of additional immune cell types stratified by IDH1 genetic status with 2289 IDH1-wild type and 317 IDH1-mutant samples. The width of the violins reflects the density of data points at different expression levels. Log2 transformed fold change of proportions for IDH1-mutant vs. WT samples along with Wilcoxon-Test p-values are presented for each immune cell type. B) Scatter plots of CD8 T cell against M2 macrophage infiltration estimated by deconvolution of RNA-sequencing data from the full cohort (left) or patients with IDH1-mutations (right).

Figure S2

Supplemental Figure S2. In vitro characterization of an isogenic IDH1-mutant murine cell line. A) Representative cDNA sequencing traces from SB1^WT (harboring a heterozygous silent Idh1 R132R mutation) and SB1^mIDH1 clones. B) Immunoblot of SB1^WT and SB1^mIDH1 clones for IDH1. HSP90 was used as a loading control. C) Cell Titer Glo 2.0 viability assay of SB1^WT and SB1^mIDH1 clones treated with the indicated doses of AG-120 for 72 hours. n = 3 SB1^WT and 3 SB1^mIDH1 clones with assay performed in technical triplicate per AG-120 dose.

Figure S3

Supplemental Figure S3. Characterization of an isogenic IDH1-mutant murine cell line in vivo. A) Tumor volume measurements of mice harboring subcutaneous SB1 parental, SB1^WT, or SB1^mIDH1 tumors. n = 2 SB1 parental, 10 SB1^WT, and 10 SB1^mIDH1 tumor-bearing mice. B) Kaplan-Meier survival analysis for mice with orthotopic SB1^WT or SB1^mIDH1 tumors. Log-rank testing used to compare curves. n = 8 SB1^WT and 10 SB1^mIDH1 tumor-bearing mice. C) Representative images of Masson’s trichrome stain of SB1^WT and SB1^mIDH1 orthotopic tumors. n = 3 tumors per tumor genotype. Scale bar = 100μm.

Figure S4

Supplemental Figure S4. Characterization of the tumor immune microenvironment in SB1^WT and SB1^mIDH1 tumors. Representative images of A) PD-L1 and B) CD68 IHC staining of orthotopic SB1^WT and SB1^mIDH1 tumors with quantification of CD68 staining presented in (C). n ≥ 4 tumors per tumor genotype for each IHC stain. Scale bar = 100μm. D) Representative images of IHC staining for D) CD8 with quantification in (E) and F) CD206 with quantification in (G) of established orthotopic SB1^mIDH1 tumor sections treated with vehicle control or AG-120 for 5 days. Scale bar = 100μm.

Figure S5

Supplemental Figure S5. CCL2 is produced by Idh1-mutant clones. A) SB1^WT and SB1^mIDH1 clones were treated with 1μM AG-120 or vehicle control for 72 hours. Relative Ccl2 transcript levels (compared to Gapdh) were determined by two-step quantitative real-time PCR and normalized to the average of the vehicle treated SB1^WT group. n = 3 clones per condition with quantitative real-time PCR performed in technical duplicate per clone per assay. Two-way ANOVA used to compare across groups. B) Immunoblot for CCL2 in SB1^WT and SB1^mIDH1 clones treated with 1μM AG-120 (+) or vehicle control (−) for 72 hours. GAPDH was used as a loading control. C) Immunoblot for CCL2 in SNU-1079 treated with 1μM AG-120 or vehicle control for 72 hours. GAPDH was used as a loading control. D) Representative images of CCL2 IHC in SB1^WT and SB1^mIDH1 orthotopic tumors showing the border between tumor and adjacent normal liver. Scale bar = 100μm. E) Additional representative images of IHC staining for CCL2 (brown) and αSMA (purple) in IDH1-wild type (WT) and IDH1-mutant human cholangiocarcinoma tumors. n = 5 IDH1 Mutant and n = 6 WT samples. Scale bar = 100μm. Each image is from a unique patient sample.

Figure S6

Supplemental Figure S6. Effect of CCL2 neutralization on Idh1-mutant tumors in a syngeneic mouse model. Quantification of IHC staining of SB1^mIDH1 tumors treated with Isotype control or αCCL2 for A) CD45, B) CD4, C) FoxP3, and D) F4/80. n ≥ 5 tumors per group. E) Representative images of IHC for cleaved caspase-3 in orthotopic SB1^WTand SB1^mIDH1 tumors treated with either an isotype control (Isotype) or a CCL2 neutralizing antibody (αCCL2). Staining quantified in (F). n ≥ 6 tumors per group. Scale bar = 100μm. G) Normalized relative luminescence (compared to isotype control treated cells for each biological replicate) for SB1WT and SB1mIDH1 cells treated with the indicated doses of αCCL2 for 72 hours. n = 3 clones per condition in technical duplicate (isotype control) or triplicate (αCCL2 doses).

List of abbreviations:

IDH1

isocitrate dehydrogenase 1

CCA

cholangiocarcinoma

BTC

biliary tract cancer

CRISPR

Clustered Regularly Interspaced Short Palindromic Repeats

HDR

homology-directed repair

TAM

tumor-associated macrophage

CCL2

C-C motif ligand 2

TME

tumor microenvironment

D-2-HG

D-2-hydroxyglutarate

FBS

fetal bovine serum

FFPE

formalin-fixed paraffin-embedded

ssODNs

single-stranded oligodeoxynucleotides

gDNA

genomic DNA

cDNA

complementary DNA

PCR

polymerase chain reaction

DMSO

dimethylsulfoxide

PBS

phosphate-buffered saline

H&E

hematoxylin and eosin

DAB

3,3’-Diaminobenzine

WT

wild type