Neoadjuvant Chemotherapy Induces IL34 Signaling and Promotes Chemoresistance via Tumor-Associated Macrophage Polarization in Esophageal Squamous Cell Carcinoma

Esophageal cancer is one of the most aggressive gastrointestinal malignancies, and the majority of the cases in Asian countries are esophageal squamous cell carcinoma (ESCC), while adenocarcinoma is dominant in Western countries (1). Neoadjuvant chemotherapy (NAC) with cisplatin plus 5-fluorouracil (5-FU) followed by curative surgery is a treatment option to treat locally advanced ESCC and regarded as the standard treatment modality for patients with stage II/III ESCC in Japan based on the clinical trial by the Japan Clinical Oncology Group (2). NAC has several potential advantages such as shrinkage of the primary tumor and control of circulating tumor cells, while, in the patients who are insensitive to NAC, there are some disadvantages in that the tumor may progress before surgery and the treatment period is prolonged. To overcome this problem, a better understanding of cellular and molecular biology within the tumor microenvironment (TME) related to NAC in ESCC should be required.

Tumor-associated macrophages (TAM) are crucial cell components of the TME and play a crucial role in solid tumor progression (3). TAMs dominantly present as an immunosuppressive (M2) phenotype, which expresses CD163, a scavenger receptor, (4, 5) and trigger multiple processes of tumor progression such as suppression of antitumor immunity, induction of regulatory T cells, activation of tumor cell migration and invasion, and induction of angiogenesis (3, 5, 6). Previous studies reported that TAM infiltration was significantly associated with poor prognosis in esophageal cancer (7–9). The meta-analysis including a total of 16 studies involving 2,292 patients with esophageal cancer revealed that CD163⁺ TAMs were related to patient poor prognosis with esophageal cancer (10). In addition, TAM infiltration might be also related to poor response to chemotherapy undergoing NAC for esophageal cancer (11, 12). Sugimura and colleagues reported that high infiltration of CD163⁺ TAMs was significantly associated with poor response to chemotherapy in esophageal cancer, both clinically and pathologically (11). Therefore, CD163⁺ TAMs might be a crucial determinant for tumor progression and chemo-sensitivity in esophageal cancer.

Colony-stimulating factor 1 (CSF-1; also known as macrophage CSF) is one of the most important cytokines responsible for the attractant, growth, and survival of monocytes and their differentiation into macrophages by binding to the CSF-1 receptor (CSF-1R; ref. 13). The expression of CSF-1 at the TME promotes TAM infiltration, leading to new vessel formation, tumor progression, and metastasis (14, 15). The expression of CSF-1 was associated with the frequencies of TAMs and correlated with poor prognosis in various types of cancers including esophageal cancer (16–20). Regarding IL34, it was newly identified as the second ligand of CSF-1R and plays similarly to CSF-1 in the functions of monocytes/macrophages as well as TAMs (21–23). The expression of IL34 in tumor tissues was also correlated with poor prognosis in several cancers including breast cancer (24), osteosarcomas (25), hepatocellular carcinoma (26), colorectal cancer (27), lung cancer (28), and ovarian cancer (29). Most interestingly, Seino and his colleagues recently found that IL34 but not CSF-1 was produced by chemoresistance cells and contributed to not only TAM recruitment but also chemoresistance in lung cancer cells (30). Supernatant from doxorubicin-resistant or cisplatin-resistant lung cancer cells triggered human monocyte differentiation into M2-polarized CD163⁺ macrophages (30). Thus, IL34 might have a crucial role in regulations of TAM infiltration and chemo-sensitivity in tumor tissues. However, the role of IL34 in TAM infiltration and the efficacy of NAC on ESCC remains unknown.

In this study, we quantified the expression of CD163, CSF-1, and IL34 in ESCC specimens before and after NAC, and evaluated the association between the expression of these markers and chemo-sensitivity to investigate the effect of NAC on the TME in ESCC. We also examined whether and how anticancer drugs including 5-FU and cisplatin triggered the expression of CSF-1 and IL34 in human ESCC cells. Moreover, we tested whether 5-FU/cisplatin-treated ESCC cells could trigger the expression of CD163 on human peripheral blood monocytes (hPBM) in vitro.

Reagents used in this study were acquired from the indicated suppliers: 5-FU, KU-55933, and VE-821 (Merck Sigma-Aldrich); cisplatin (FUJIFILM Wako Pure Chemical Corporation); anti-human IL34 antibody [1D12] (ab101443), anti-M-CSF antibody [EP1179Y] (ab52864; Abcam); Novocastra Liquid Mouse Monoclonal Antibody CD163 (NCL-L-CD163; Leica Biosystems); anti-human CD3 antibody (IR503), anti-human CD8 (M7103), and E-cadherin mAb (NCH-38; Dako/Agilent Technology); anti-human CD105 antibody (28117-1-AP; Proteintech); vimentin mAb (SP20; Nichirei Bioscience); antibody APC/Cyanine7 anti-human CD14 [HCD14] (325619), PE anti-human CD163 [GHI/61] (333605; BioLegend); 7-AAD (559925; BD Biosciences); PLX-3397 [PLX; HY-16749], BLZ945 [BLZ; HY-12768] (MedChemExpress).

We recruited 45 patients with ESCC who received NAC [NAC (+)] (n = 33) or did not receive NAC [NAC (−)] (n = 12) at Fukushima Medical University School of Medicine (Fukushima, Japan) between February 2008 and February 2019 (Table 1). Median times from biopsy to surgery were 62.5 days (range: 20–141 days) in NAC (−) and 89 days (range: 39–176 days) in NAC (+) (P = 0.0433), respectively. Tumor—node–metastasis staging was determined according to the Japanese Classification of Esophageal Cancer (11th edition). Histopathologic evaluation of response to NAC was assessed by pathologists using FFPE sections of surgically resected ESCC specimens. On the basis of the percentage of viable residual tumor cells after NAC, the curative effect was classified as follows: grade 0, no recognizable histologic NAC effect; grade Ia, viable residual tumor cells account for more than two-thirds of ESCC tissue; grade Ib, viable residual ESCC cells account for more than one-third but less than two-thirds of ESCC tissue; grade II, viable residual tumor cells account for less than one-third of ESCC tissue. Written informed consent was obtained from all the patients. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Fukushima Medical University (Fukushima, Japan; Reference No. 2329).

Paraffin-embedded 4-μm ESCC sections fixed in 10% formaldehyde were deparaffinized in xylene and rehydrated in ethanol. Endogenous peroxidases were blocked with 0.3% hydrogen peroxide in methanol. The sections were then incubated with epitope retrieval solution (Dako) for 10 minutes at 120°C. After washing with PBS, the sections were stained with anti-human CD163, anti-human CD3, anti-human CD8, anti-human CD105, anti-human E-cadherin, anti-human vimentin, anti-human IL34, or anti-human CSF-1 at 4°C overnight. The sections were washed and incubated with horseradish peroxidase–conjugated anti-mouse or anti-rabbit secondary antibodies (EnVision+System; Dako/Agilent Technologies). Peroxidase was visualized with diaminobenzidine (Dojindo Molecular Technologies), and nuclei were counterstained with Mayer Hematoxylin Solution (FUJIFILM Wako Pure Chemical Corporation). Images were obtained using a batch slide scanner NonoZoomer-SQ (Hamamatsu Photonics).

Counting CD163⁺ cells was performed using ImageJ software in four fields of the hotspots of the cells. Briefly, the images were saved as 8-bit color-scale JPEG files. The open-source software package ImageJ 1.43 (NIH, Bethesda, MD) was used for measuring the number and intensities of CD163⁺ cells. Briefly, (i) color-scale images were split into the red, green, and blue images, and CD163⁺ cells were selected from the blue images using the threshold tool. (ii) color-scale images were also split into hue, saturation, and brightness images, and the saturation images were combined with the images from (i), and the number and density values for CD163⁺ cells were measured using the analysis tool (Supplementary Fig. S1). Because Image J software recognizes the overlapped CD163⁺ cells as a single CD163⁺ cell, the number of CD163⁺ cells was calculated on the basis of nuclear staining if they were overlapped. For calculating the intensity of CD163 expression, overlapped cells were considered as a single CD163⁺ cell, and the total amount of the intensities of CD163 expression were divided by the number of CD163⁺ cells. The expression level of CSF-1 and IL34 was assessed using IHC score (H-score; 0–300), which took into account the percentage of cells (0%–100%) as well as each staining intensity category (0–3+; Supplementary Fig. S2). The medians were adopted for determining the IL34 low and IL34 high, or CSF-1 low and CSF-1 high, respectively. The medians (H-score) were 120 for IL34 and 30 for CSF-1. For the assessment of CD3⁺ and CD8⁺ tumor-infiltrating lymphocytes (TIL), the invasive front region of the tumor was reviewed in four independent areas at a magnification of ×400, as described previously (31, 32). For quantification of CD105⁺ microvessel density, three tumor areas with the greatest number of microvessels were selected and counted at a magnification of ×400, as described previously (32, 33). The expression level of E-cadherin and vimentin was assessed using a staining score (0–12), which was based on the multiplication of the scoring percentages of positively stained cells (0, <5%; 1, 6%–25%; 2, 26%–50%; 3, 51%–75%; 4, 76%–100%) and the staining intensity score (0, absent; 1, week; 2, moderate; 3, strong) from individual slides, as described previously (34, 35).

The human ESCC cell lines were obtained from the Japanese Collection of Research Bioresources Cell Bank on April 7, 2020 for the TE series and RIKEN BioResource Research Center on May 28, 2020 for the KYSE series. The cells were maintained in RPMI1640 (Sigma-Aldrich) supplemented with 10% heat-inactivated FBS and penicillin/streptomycin in a 5% CO₂ atmosphere at 37°C. Cisplatin-resistant TE-1 cell line (TE-1/cisplatin) was established by repeated subcultures in the presence of 1 μmol/L cisplatin. After the final subculture with cisplatin, TE-1/cisplatin cells were cultured in a drug-free medium for a few days and subjected to qPCR analysis of IL34. Cells were passaged no more than 15 times and then a new frozen aliquot of cells was used. All analyses were performed within 6 months of obtaining the cells from the cell banks.

To assess half inhibition concentration (IC₅₀) of the ESCC cell lines, the cells were seeded at 5,000 cells per well in a 96-well plate and treated with different concentrations of 5-FU or cisplatin for 48 hours. Cell viability of each cell line was determined by the Cell Counting Kit-8 (Dojindo Molecular Technologies), and IC₅₀ was calculated using Graph pad prism 8 (GraphPad Software).

Peripheral blood mononuclear cells (PBMC) were isolated from the peripheral blood of healthy volunteers using Ficoll-Paque (GE Healthcare). Human CD14⁺ monocytes were isolated from PBMCs using the MACS magnetic cell separation system (Basophil Isolation Kit II, human; Miltenyi Biotec).

hPBMs were co-cultured indirectly with ESCC cell lines treated with or without IC₅₀ of 5-FU or cisplatin for 48 hours. Indirect co-cultures were performed by adding 5-FU/cisplatin-treated ESCC cells into 0.4-μm pore size cell culture inserts (CORNING) and placing them into 12-well culture plates containing hPBMs.

Total RNA was isolated from ESCC cell lines using the TRIzol reagent (QIAGEN). RNA was quantified on a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). cDNA was synthesized using the PrimeScript RT-PCR Kit (Takara). qPCR was performed on a 7300 Fast Real-Time PCR System (Applied Biosystems) using PrimeTime Gene Expression Master Mix (Integrated DNA Technologies) with specific primers and probes against human IL34, CSF1, and GAPDH (Integrated DNA Technologies). qPCR data were normalized against the corresponding levels of GAPDH mRNA.

hPBMs were stained with antibodies specific for APC/Cyanine7 anti-human CD14 and PE anti-human CD163 in the presence of human Fc blocking reagent (Miltenyi Biotec). After washing with PBS, the stained cells were incubated with 7-AAD in 1× binding buffer for 10 minutes at room temperature and analyzed on a BD FACSCanto II flow cytometer (BD Biosciences). Flow cytometry data were analyzed using FlowJo software (FlowJo).

The mRNA expression z-score of genes (RNA-Seq V2 RSEM normalized, RNA-Seq data) was obtained from The Cancer Genome Atlas (TCGA) ESCC dataset (n = 95). The relationship between mRNA expressions of CD163 and CSF1 in TCGA cohorts was analyzed through cBioPortal (36).

Data are expressed as means ± SEM. The statistical analyses were performed using Graph pad prism 8 or R. Pairs of groups were compared by the unpaired Student t test, Mann–Whitney test, or Wilcoxon matched-pairs test, as appropriate. For multigroup comparisons, we applied one-way ANOVA with post hoc testing or the Kruskal–Wallis test with Dunn post hoc test, as appropriate. Categorical variables were compared by the χ² or Fisher exact test. The Spearman correlation test was used to analyze the association in each experiment. The Cox proportional hazards regression model was used for univariate and multivariate survival and recurrence-free survival analyses. A value of P < 0.05 was considered to be significant.

To first assess the effect of NAC on TAMs in ESCC, we analyzed CD163 expression in formalin-fixed, paraffin-embedded biopsy and surgically resected ESCC specimens from patients who received NAC [NAC (+)] (n = 33) or did not receive NAC [NAC (−)] (n = 12; Fig. 1A). IHC staining of CD163 revealed that TAM infiltration was observed in ESCC specimens of all patients in this study. The number of CD163⁺ TAMs did not differ between biopsy and surgically resected specimens in patients with NAC (−) (Fig. 1B and C). On the other hand, the number of CD163⁺ TAMs were dramatically increased in surgically resected specimens as compared with biopsy specimens in patients with NAC (+) (Fig. 1B and C). Moreover, the number of CD163⁺ TAMs was comparable between NAC (−) and NAC (+) in biopsy specimens, while it was significantly higher in surgically resected specimens in NAC (+) than those in NAC (−) (Fig. 1D). Consistent with the increased frequencies of CD163⁺ TAMs, the intensities of CD163 expression were markedly increased in ESCC specimens after NAC as compared with those before NAC (Fig. 1B, E, and F), suggesting that NAC might shift the phenotype of TAMs toward more immunosuppressive M2 phenotype in ESCC. These results suggest that NAC triggers TAM polarization into M2 type in ESCC.

To investigate whether M2-polarized TAM by NAC stimulated immunosuppression, EMT, and angiogenesis in ESCC, we examined the correlations between the number of CD163⁺ cells and T-cell infiltration (the number of CD3⁺ and CD8⁺ TILs), EMT status (the expression of E-cadherin and vimentin), or the degree of angiogenesis (CD105⁺ microvessel counts) in surgically resected ESCC specimens from NAC (+). Among these markers, CD105⁺ microvessel counts were significantly correlated with the number of CD163⁺ cells in ESCC after NAC (Fig. 1G and H), suggesting that an increased number of CD163⁺ M2 macrophages by NAC might contribute to angiogenesis, which enhances tumor growth and reduces chemosensitivity.

Next, we determine the dominant cytokine which triggers M2-like TAM polarization by NAC in ESCC. Because CSF-1 is known as a primary cytokine that regulates infiltration, differentiation, and polarization of TAMs at the tumor sites (14), and TCGA dataset showed that mRNA expression of CD163 was significantly associated with mRNA expression of CSF-1 in ESCC tissues (Supplementary Fig. S3), we investigated the expression of CSF-1 in ESCC. As shown in Fig. 2A, CSF-1 mainly expressed in tumor cells but not in stromal cells in ESCC specimens (Fig. 2A). The expression of CSF-1 was not changed before and after NAC in ESCC (Fig. 2B), and the expression level of CSF-1 was comparable between NAC (−) and NAC (+) in surgically resected ESCC specimens (Fig. 2C), suggesting that CSF-1 might not be involved in M2-like TAM polarization by NAC in ESCC.

IL34 is another ligand of CSF-1R, which might be induced upon chemotherapy (28, 30, 37, 38). We, therefore, examined whether NAC induced the expression of IL34 in ESCC. Like CSF-1, the expression of IL34 was dominantly observed in tumor cells but not in stromal cells in ESCC (Fig. 2D). The expression of IL34 was comparable between biopsy and surgically resected specimens in NAC (−), while it was markedly increased in surgically resected specimens as compared with biopsy specimens in NAC (+) (Fig. 2D and E), suggesting that NAC triggers the expression of IL34 in ESCC. Interestingly, a positive correlation was obtained between the number of CD163⁺ TAMs and IL34 expression in tumor cells in ESCC after NAC, while a similar trend was not observed in ESCC before NAC (Fig. 2F and G). These data suggest that IL34 might be the dominant factor to polarize TAMs into M2 phenotype upon NAC.

It has been reported that IL34 was highly produced by chemoresistance lung cancer cell lines (30). We, therefore, examined whether IL34 was highly expressed in NAC-nonresponsive patients with ESCC in comparison with NAC-responsive patients with ESCC based on pathologic evaluation for surgically resected specimens. As shown in Fig. 3A, the NAC-nonresponsive group (grade 0-Ia, n = 26) exhibited higher levels of IL34 expression in both post-NAC and pre-NAC compared with the NAC-nonresponsive group (grade Ib-II, n = 7; Fig. 3A), suggesting that higher expression of IL34 might be involved in poorer NAC sensitivity, and the level of IL34 expression in biopsy might be a potential biomarker for response to NAC. Moreover, although increments of IL34 expression [(H-score of Post-) – (H-score of Pre-)] tended to be higher in the NAC-nonresponsive group than in the NAC-responsive group (Fig. 3B), it was not significant (P = 0.1166). When dividing NAC (+) patients into IL34-low group (n = 17) or IL34-high group (n = 16), the percentage of NAC-nonresponsive patients was significantly higher in IL34-high group (93.4%) than in IL34-low group (64.7%; Fig. 3C). Notably, we found that the higher expression of IL34 was significantly associated with poorer overall survival (OS) and recurrence-free survival (RFS) after NAC in ESCC (Fig. 3D and E). Although a univariate analysis indicated that the level of IL34 was significantly associated with OS and RFS, it did not reach statistical significance after multivariate analysis incorporating known clinical prognostic factors (Supplementary Table S1).

We also examined whether CSF-1 expression was associated with sensitivity of NAC and OS/RFS in patients with ESCC. However, we could not find any associations between CSF-1 expression and NAC sensitivity in ESCC (Supplementary Fig. S4A and S4B). In addition, OS and RFS after NAC did not differ between the CSF-1–high group and CSF-1–low group in patients with ESCC (Supplementary Fig. S4C–S4E).

Taken together, our data suggest that IL34, but not CSF-1, is highly produced by chemotherapy in NAC-nonresponsive patients with ESCC, which might contribute to the worse prognosis.

On the basis of the above results, IL34 might be the key factor in the change in cytokine expression before and after NAC. To further evaluate the effect of drugs on the expression of IL34 in human ESCC cells, 14 human ESCC cell lines were treated with IC₅₀ of 5-FU or cisplatin and subjected to qPCR analysis to evaluate mRNA expression of IL34 and CSF-1. Increased mRNA expression of IL34 by 5-FU treatment was significantly observed in 11 among 14 ESCC cell lines (Fig. 4A, top). However, significant induction of mRNA expression of CSF-1 by 5-FU treatment was observed in only three among 14 ESCC cell lines (Fig. 4A, top), and the fold changes of mRNA expression of IL34 was much higher than those of mRNA expression of CSF-1 in ESCC cell lines (Fig. 4B, left). Similar trends were also observed in cisplatin-treated ESCC cell lines (Fig. 4A, bottom and 4B, right). These results suggest that 5-FU and cisplatin specifically induce mRNA expression of IL34 in human ESCC cells.

Anticancer drugs including 5-FU and cisplatin cause DNA lesions including double-strand breaks, DNA-protein cross-links, and intrastrand cross-links, leading to lethal effects on tumor cells (39–41). In response to the DNA lesions, potential sensors of DNA damage such as ataxia telangiectasia mutated (ATM) and ataxia telangiectasia and rad3 related (ATR) are activated and consequently suppress the function of cyclin-dependent kinase to initiate cell-cycle arrest (39). Therefore, we tested whether DNA damage signaling pathways regulated by ATM and ATR were involved in the regulation of IL34 expression by 5-FU and cisplatin in human ESCC cells. mRNA expressions of IL34 were significantly increased by 5-FU and cisplatin in TE-1, TE-4, and TE-6 cells, and these inductions were partially inhibited by the inhibitors of ATM and ATR (Fig. 4C), suggesting that 5-FU– and cisplatin-induced mRNA expression of IL34 is mediated at least in part through the activation of ATM- and ATR-mediated DNA damage signaling pathways in human ESCC cells.

We also examined the correlations between IC₅₀ of 5-FU or cisplatin and basal levels of IL34 or IL34 levels after drug treatment in ESCC cell lines. Chemo-sensitivity (IC₅₀) was not significantly correlated with both basal levels of IL34 and levels after drug treatment in ESCC cell lines (Supplementary Fig. S5A and S5B). However, we found that fold changes between basal levels of IL34 and levels after drug treatment was significantly correlated with chemo-sensitivity to cisplatin, but not 5-FU in ESCC cell lines (Supplementary Fig. S5C and S5D), suggesting that higher fold change of IL34 expression after drug treatment might be involved in chemo-sensitivity to cisplatin in ESCC cells, which seems to be consistent with our clinical data that increments of IL34 expression might be associated with poorer chemo-sensitivity in ESCC (Fig. 3B).

To further examine whether IL34 expression is enhanced in chemoresistant ESCC cells, we established cisplatin-resistant TE-1 cells (TE-1/cisplatin). We demonstrated that the IC₅₀ of TE-1/cisplatin cells was significantly higher than that of parental TE-1 cells (Supplementary Fig. S6A). mRNA expression of IL34 was markedly increased in cisplatin-resistant TE-1/cisplatin cells as compared with that in the parental cisplatin-sensitive TE-1 cells (Supplementary Fig. S6B), suggesting that IL34 expression is enhanced in the chemoresistant ESCC cells as compared with the chemo-sensitive ESCC cells, which is consistent with our clinical data (Fig. 3).

We finally tested whether 5-FU/cisplatin-treated ESCC cells have the potential to induce protumorigenic monocytes/macrophages in vitro. To test this, we established an in vitro co-culture system using hPBMs and human ESCC cell lines (Fig. 5A). hPBMs were co-cultured with or without TE cells or 5-FU/cisplatin-treated TE cells in the absence or presence of CSF-1R kinase inhibitors (PLX or BLZ), which can block the activation of CSF-1R axis by IL34, for 4 days and subjected to flow cytometry analysis to detect the surface expressions of CD14, which is highly expressed in IL34-induced immunosuppressive monocytes/macrophages, and CD163. The expressions of both CD14 and CD163 in hPBMs were increased by the co-cultured with TE-1 cells or TE-4 cells, and the expression was further enhanced by the co-cultured with 5-FU/cisplatin-treated TE-1 cells or TE-4 cells (Fig. 5B and C). Moreover, the increased expressions of CD14 and CD163 in hPBMs by co-culture with 5-FU/cisplatin-treated TE cells were significantly attenuated by the treatment with CSF-1R inhibitors (Fig. 5B and C). These results indicate that 5-FU/cisplatin-treated ESCC cells could induce monocytes/macrophages polarization into CD163⁺ immunosuppressive phenotype via the activation of the CSF-1R axis.

In this study, we described the role of IL34 on M2-TAM infiltration by NAC, the efficacy of NAC, and patients' OS and RFS after NAC in ESCC. We found that NAC dramatically enhanced TAM polarization into M2 phenotype in ESCC. A positive correlation between the frequencies of CD163⁺ TAMs and the expression of IL34 on tumor cells in ESCC was observed after NAC. We also demonstrated that 5-FU and cisplatin specifically increased mRNA expression of IL34 on ESCC cells. In addition, hPBMs co-cultured with 5-FU/cisplatin-treated ESCC cells increased the expression of CD14 and CD163 through the CSF-1R axis. These data suggest that tumor cell–derived IL34 by NAC enhances M2-like TAM polarization, and the higher IL34 expression is significantly associated with poorer NAC sensitivity of ESCC.

Our results indicated that NAC significantly promoted the expression of CD163⁺ TAMs and their further polarization into the M2 phenotype, as judged by the intensities of CD163 expression, in ESCC (Fig. 1). CD163 is a well-known marker for M2 macrophages as well as TAMs, and CD163⁺ TAMs negatively regulate antitumor immune responses through the production of anti-inflammatory mediators such as IL10 and TGFβ (42, 43). In addition, a previous study reported that CD163 is related to the production of protumor cytokines such as IL6 and CXCL2 in TAMs, and TAM-derived IL6 was preferentially associated with tumor cell proliferation (44), suggesting that CD163⁺ TAMs also have the potential to enhance tumor progression by supporting tumor cell proliferation. Therefore, enhanced polarization of CD163⁺ TAMs by NAC might inhibit antitumor immune responses and promote tumor cell proliferation, leading to tumor development and progression in patients with ESCC.

We revealed that IL34 was a dominant cytokine that attracted M2-TAMs in ESCC after NAC. The expression of IL34 was significantly associated with increased frequencies of CD163⁺ TAM (Fig. 2), which likely to be consistent with a previous report that enhanced expression of IL34 was correlated with the frequencies of CD163⁺ macrophages in refractory melanoma that acquired resistance to anti-PD-1 immunotherapy (45). It is likely that in untreated, basal state of tumors, CSF-1 but not IL34 might be continuously produced by tumor cells through activating of the IFNγ-mediated signaling pathway, while IL34 but not CSF-1 might be highly produced from tumor cells by the stimulations of several anticancer agents or therapies such as radiotherapy. Although IL34 and CSF-1 showed an equivalent ability to support monocyte growth, survival, and differentiation, the binding affinity of IL34 to human CSF-1R was 9.5- to 13-fold higher than that of CSF-1 to human CSF-1R (46), and IL34 could induce stronger tyrosine phosphorylation of CSF-1R and higher production of downstream mediators in comparison with CSF-1 (22, 23), suggesting that IL34 might have a superior potential to differentiate monocytes them into immunosuppressive M2 phenotype. Thus, it is likely that CD163⁺ TAMs accumulated by IL34 after NAC might have more critical pathogenic roles in tumor progression and chemo-insensitivity.

Enhanced expression of IL34 in ESCC contributed to worse OS and RFS in patients after NAC (Fig. 3). It has been reported that IL34 induced the differentiation of monocytes into IL10^high IL12^low immunoregulatory macrophages, which maintain locally restrained inflammation, resulting in the enhancement of angiogenesis and metastasis (47). In addition, other reports demonstrated that IL34 promoted the survival of chemoresistant cells by activating the CSF-1R/Akt axis (30, 38). Interestingly, chemoresistant lung cancer cells expressed CSF-1R, and autocrine CSF-1R signaling by IL34 drove chemoresistance (30). Therefore, a worse prognosis in patients with IL34-high ESCC after NAC might be sustained by not only the accumulation and differentiation of immunosuppressive CD163⁺ TAMs but also activation of tumor cell proliferation by the IL34/CSF-1R axis.

Besides IL34, IL6 has also been reported to promote chemoresistance in ESCC. IL6 derived from cancer-associated fibroblasts protected ESCC cells from apoptosis by upregulating C-X-C motif chemokine receptor 7 (CXCR7) expression through signal transducer and activator of transcription 3/NFκB pathway, and the high expression of CXCR7 and IL6 significantly contributed to worse OS and PFS in ESCC (48). Moreover, Dijkgraaf and colleagues demonstrated that platinum-based anticancer drug-induced IL6 expression from cancer cells differentiated monocytes into tumor-promoting M2 macrophages (49). Thus, IL6 might be also involved in M2 macrophage polarization and poorer sensitivity to NAC in ESCC.

According to the results of our in vitro co-culture experiments, 5-FU/cisplatin-stimulated ESCC cells could trigger the expression of CD163 in hPBMs through the CSF-1R axis (Fig. 5). Previous reports demonstrated that both CSF-1 and IL34 have the potential to induce the expression of CD163 in hPBMs (50). In response to 5-FU and cisplatin, mRNA expression of IL34 but not CSF-1 was significantly induced in TE-1 and TE-4 cells (Fig. 4A and B). Therefore, CD163 expression in hPBMs co-cultured with 5-FU/cisplatin-treated TE cells might be dominantly mediated by the IL34/CSF-1R axis but not the CSF-1/CSF-1R axis.

Our data suggest that the ATM/ATR-mediated DNA damage signaling pathways partially contribute to 5-FU/cisplatin-induced mRNA expression of IL34 in ESCC cells (Fig. 4C). ATM and ATR are the key mediators of DNA damage response (DDR), and inhibiting the DDR has become an attractive therapeutic concept in cancer therapy because (i) increased DDR signaling might be associated with resistance to genotoxic therapies, and (ii) many cancers are highly dependent on not defected, remaining DDR pathways for survival. Our findings raise a novel concept for the inhibition of DDR singling in cancer therapy, which may block chemotherapy-induced IL34 production from tumor cells, resulting in the improvement of the TME via suppression of M2-like TAM polarization.

The limitations of this study are as follows:

• (i) The time from biopsy to surgery in NAC (+) was significantly longer than that in NAC (−). However, we believe that roughly 20 days difference in the time from biopsy to surgery probably has little effect on the TME clinically and biologically.

• (ii) Biopsy specimens are a small part of tumor tissues and have heterogeneity. Thus, there might be a possibility that the observation of biopsy specimens does not reflect the observation of whole tumor tissues.

• (iii) As the number of cases in this study is small, further investigations in a larger number of cases should be needed.

In summary, this study is the first report to show that IL34 derived from tumor cells promotes M2-like TAM polarization, and consequently reduces the efficacy of NAC in patients with ESCC. M2 TAM depletion and repolarization toward an M1-like state have been considered as an effective therapeutic strategy to synergistically enhance the effect of cancer treatments including chemotherapy, radiotherapy, and checkpoint blockade immunotherapy, and the clinical trials using both antibodies and small molecules against CSF-1R have been conducted (3). However, the occurrence of severe adverse events has been reported due to the unwanted disruption of perivascular macrophages in nontarget tissues and organs (51). Our findings suggest a possibility that IL34 may become an alternative target to CSF-1R for combination therapy with lesser side effects. Depletion and repolarization of TAMs into M1-like phenotype by the blockade of IL34 may improve the efficacy of current standard chemotherapy to ESCC. Moreover, the most recent report showed that tumor-derived IL34 also inhibited the therapeutic effects of checkpoint blockade immunotherapy (52). ESCC is one of the tumors which should be anticipated to respond to immune checkpoint blockade therapy (53, 54). Thus, IL34 blockade may be applicable to broad ranges of combination therapies in ESCC.

No disclosures were reported.

S. Nakajima: Conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, supervision, validation, writing–original draft, writing–review and editing. K. Mimura: Conceptualization, funding acquisition, investigation, methodology, supervision. K. Saito: Investigation, methodology. A.K. Thar Min: Investigation, methodology. E. Endo: Investigation, resources. L. Yamada: Investigation, resources. K. Kase: Investigation, resources. N. Yamauchi: Investigation, resources. T. Matsumoto: Investigation, resources. H. Nakano: Investigation, resources. Y. Kanke: Funding acquisition, investigation, resources, supervision. H. Okayama: Funding acquisition, investigation, resources, supervision. M. Saito: Funding acquisition, investigation, resources, supervision. P. Neupane: Investigation. Z. Saze: Investigation, resources, supervision. Y. Watanabe: Investigation, resources. H. Hanayama: Investigation, resources. S. Hayase: Investigation, resources, supervision. A. Kaneta: Investigation, resources. T. Momma: Investigation, resources, supervision. S. Ohki: Investigation, resources, supervision. H. Ohira: Resources. K. Kono: Conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, writing–original draft, writing–review and editing.

We are grateful to all the patients who contributed to this study. We also thank Masayo Sugeno, Sakino Arai, Eri Takahashi, Hideko Taguchi, and Ayumi Hirose-Nakajima for excellent technical assistant and helpful secretarial assistance. This work was supported by Japan Society for the Promotion of Science KAKENHI grant Nos. 19K09128 and 20K17695.

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.