Tumor immune microenvironment in cancer patients with leukocytosis

Abstract

Tumor-related leukocytosis (TRL) is correlated with poor survival in various types of cancers, but the microenvironment of TRL-associated human tumors has not been fully elucidated. Here, we aimed to characterize the immune microenvironment of cancer patients with TRL. The transcriptional signatures of tumor tissues obtained from cervical cancer patients with (TRL^pos) and without TRL (TRL^neg) were compared. As a surrogate for TRL diagnosis, a leukocytosis signature (LS) score was derived using genes differentially expressed between TRL^pos and TRL^neg tumors. The immunological profiles of patients in the TCGA database with high (LS^high) or low LS scores were compared. TRL^pos tumors were transcriptionally distinct from TRL^neg tumors, exhibiting up-regulation of radioresistance and down-regulation of adaptive immune response-related genes. In the TCGA cervical cancer cohort (n = 303), patients with high LS had inferior survival rates compared to those with low LS (P = 0.023). LS^high tumors were enriched in radioresistance, wound healing, and myeloid-derived suppressor cell (MDSC) signatures and had a higher infiltration of M2 macrophages and a lower infiltration of M1 macrophages and lymphocytes. LS^high tumors also expressed higher levels of CXCR2 chemokines, CSF2, and CSF3. In the pan-cancer cohort (n = 9984), LS^high tumors also exhibited poor survival, signatures of a suppressive immune microenvironment, and higher expression of CXCR2 chemokines. Our data provide evidence for a suppressive immune microenvironment in patients with TRL and suggest promising targets, such as the CXCR2 axis, for its therapeutic intervention.

Electronic supplementary material

The online version of this article (10.1007/s00262-020-02545-4) contains supplementary material, which is available to authorized users.

Introduction

Growing evidence suggests that there is a link between cancer and inflammation [1]. Inflammatory response is considered as one of the hallmarks of cancer and plays central roles at different stages of tumor development and progression [2, 3]. Cancer-related inflammation is known to have tumor-promoting effects by aiding survival of cancer cells, promoting angiogenesis, and suppressing anti-tumor immunity [1]. In 10–20% cancer patients, such inflammatory responses are reflected in peripheral blood as increases in leukocytes referred to as tumor-related leukocytosis (TRL). TRL is associated with a poor response to radiotherapy or chemotherapy and, accordingly, poor survival [4–7]. Despite the dismal prognosis of patients with TRL, there is no effective therapy for such patients.

Myeloid cells, such as myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs), are key components of tumor microenvironment inflammatory responses [8, 9]. MDSCs and TAMs aid tumor development by promoting the cellular stemness of cancer cells, angiogenesis, and epithelial–mesenchymal transition and inhibiting anti-tumor T cell responses [10]. Preclinical studies have shown that increased G-CSF secretion from tumor cells can increase both circulating and intra-tumoral MDSCs [4, 11]. In humans, TRL is associated with a higher frequency of circulating MDSCs and higher G-CSF concentrations [4, 11, 12] and negatively correlated with tumor T-cell infiltration [7, 13]. However, comprehensive analysis of the immune microenvironment of patients with TRL has not been performed.

In the present study, we compared the immune landscapes of tumors from cervical cancer patients with and without TRL using RNA sequencing. Furthermore, we utilized TCGA data to extend our findings to a larger cohort of TRL patients with cervical and non-cervical cancer with the aim of discovering promising TRL therapeutic targets.

Materials and methods

Patients and sample preparation

Patients with pathologically confirmed squamous cell carcinoma of the uterine cervix without distant metastasis who received definitive chemoradiotherapy or radiotherapy were included in the study. TRL was defined as a white blood cell count exceeding 9000/μL at baseline without any evidence of infection [5]. RNA was extracted from formalin-fixed paraffin-embedded (FFPE) tissue sections with an RNeasy FFPE Kit (Qiagen, Hilden, Germany), and only samples that passed the quality control tests were used for RNA sequencing. RNA quality was assessed using an Agilent 2100 bioanalyzer and an RNA 6000 Nano Chip (Agilent Technologies, Palo Alto, CA), and RNA quantification was performed using an ND-2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). This study was approved by the Institutional Review Board (4-2015-0454), and all patients provided informed consent before inclusion in the study.

RNA sequencing and analysis

Four patients with TRL (TRL^pos) and four patients without TRL (TRL^neg) were eligible for RNA sequencing. Detailed methods of RNA sequencing are provided in Supplementary materials. Gene counts were normalized by library size, and differential gene expression of genes was analyzed using DESeq2 (version 1.24.0). Differentially expressed genes (DEGs) were determined as those with an adjusted P < 0.05 and a log₂fold change > 1. Gene set enrichment analysis (GSEA) was performed using Broad Institute software (http://software.broadinstitute.org/gsea/index.jsp). Gene ontology (GO) analysis of the gene expression data obtained from MsigDB (http://www.gsea-msigdb.org/gsea/index.jsp) was performed using GSEA and visualized by the Enrichment Map tool in Cytoscape version 3.7.1 (https://cytoscape.org). The enrichment score for gene sets was calculated using a gene set variation analysis (GSVA) with the GSVA R package (version 1.32.0). Heatmaps were generated using the pheatmap package (version 1.0.12).

The cancer genome atlas (TCGA) data acquisition and definition of subgroups

RNA sequencing data and patient clinical information from the TCGA database were obtained for 9984 patients comprising 32 solid tumors using FireBrowse (Broad Institute). A “leukocytosis signature (LS) score” was calculated for each individual using GSVA and was defined as the enrichment score for the gene set derived from the genes differentially expressed between TRL^pos and TRL^neg tumors. In detail, the GSVA score for significantly down-regulated genes in the TRL^pos tumor was subtracted from the GSVA score of significantly up-regulated genes in TRL^pos tumor. Patients within the upper quartile of the LS score were defined as LS^high, with the remaining patients being defined as LS^low.

Immune signature and tumor immune infiltrate analysis

The signature scores for five immune expression signatures, namely the interferon (IFN)-γ response, the transforming growth factor (TGF)-β response, macrophage/monocytes, lymphocyte infiltration, and wound healing, were obtained from a previous paper [14]. These five immune signatures were selected from 160 immune expression signatures that robustly reproduced the co-clustering of the immune signature sets, and signature scores were calculated using GSVA. Data for the fraction of immune cell subsets infiltrating the tissue from the TCGA cohort were calculated using CIBERSORT [15]. The suppressive MDSC gene signature was derived from a previous study that identified the top 100 differentially expressed genes between suppressive MDSCs and monocytes sorted from peripheral blood mononuclear cells [16]. The T cell-inflamed gene expression profile, which is known to predict tumor response to anti-programmed death-1 (PD-1) therapy was calculated by GSVA using a previously reported gene set [17]. The neoantigen load for each patient from the TCGA cohort was obtained from a previous paper [14], which were determined from single-nucleotide variants and indel mutations that were predicted to result in Major histocompatibility complex binding peptides. The T cell receptor (TCR) and B cell receptor (BCR) diversities of patients from the TCGA cohort, which were measured using the Shannon entropy, were obtained from a previous report [14].

Statistical analysis

A Student’s t-test was performed to compare continuous variables between two groups. Survival curves were generated using Kaplan–Meier curves and compared using a log-rank test. A Cox regression was performed to determine the hazard ratio (HR) for the LS score on survival. A Pearson correlation analysis was performed to determine the correlation between two continuous variables. Two-sided P-values < 0.05 were considered significant. All statistical analyses were performed in R version 3.5.1 (http://www.r-project.org) or GraphPad Prism version 6.0 (GraphPad Software Inc., San Diego, CA).

Results

The characteristics of the four TRL^pos patients and the four TRL^neg patients are summarized in Table 1. TRL^pos patients tended to have lower age, larger tumor, and more advanced tumor stage compared to TRL^neg patients. In addition, TRL^pos patients showed predominance of neutrophils among their leukocytes and a higher neutrophil-to-lymphocyte ratio (NLR) compared to TRL^neg patients. All patients had received radiotherapy or chemoradiotherapy. Three of the four TRL^pos patients experienced recurrence, while TRL^neg patients experienced no recurrence. At last follow-up, two patients were alive among the four TRL^pos patients and all TRL^neg patients were alive.

Table 1.

Patient characteristics of those with TRL (n = 4) and without TRL (n = 4)

Patient	Age/sex	Histology	Tumor size (cm)	FIGO stagea	TNM stageb	Leukocyte count (/μL)	ANC (/μL)	ALC (/μL)	NLR (/μL)	Treatment	Recurrence	PFS (m)	OS (m)	Last f/u status
TRLpos 1	37/F	Squamous cell ca., non-keratinizing type	7.0	IIB (IIIB)	T2bN1	15,770	11,380	2650	4.29	CCRT	None	57.8	57.8	Alive
TRLpos 2	35/F	Squamous cell ca., keratinizing type	6.8	IVA (IVA)	T4N1	9900	8470	530	15.98	CCRT	Lung metastasis	7.8	58.4	Alive
TRLpos 3	66/F	Squamous cell ca., non-keratinizing type	6.6	IVA (IVA)	T4N1	11,850	9450	1660	5.69	RT	Pelvic LN metastasis	3.5	8.3	Dead
TRLpos 4	28/F	Squamous cell ca., non-keratinizing type	7.3	IIA (IIIB)	T2a2N1	12,960	8250	2200	3.75	CCRT	Lung metastasis	7.5	16.5	Dead
TRLneg 1	76/F	Squamous cell ca., non-keratinizing type	1.9	IIA1 (IIA1)	T2a1N0	6090	4200	1430	2.94	CCRT	None	57.8	57.8	Alive
TRLneg 2	56/F	Squamous cell ca., non-keratinizing type	5.0	IB2 (IIIB)	T1b2N1	8010	3710	3340	1.11	CCRT	None	55.7	55.7	Alive
TRLneg 3	75/F	Squamous cell ca., non-keratinizing type	5.6	IIB (IIIB)	T2bN1	8370	5300	2400	2.21	CCRT	None	60.8	60.8	Alive
TRLneg 4	54/F	Squamous cell ca., non-keratinizing type	3.1	IIB (IIB)	T2bN0	7320	2970	3890	0.76	CCRT	None	69.4	69.4	Alive

TRL, tumor-related leukocytosis; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; NLR, neutrophil-to-lymphocyte ratio; PFS, progression-free survival; OS, overall survival; CCRT, concurrent chemoradiotherapy; RT, radiotherapy; LN, lymph node

^a2019 FIGO staging is provided in parenthesis

^bAJCC 8th edition

Next, we performed RNA sequencing of tumors obtained from four TRL^pos and four TRL^neg patients. Principal component analysis of TRL^pos and TRL^neg tumor gene expression showed distinct transcriptional landscapes (Fig. 1a). There were 591 differentially expressed genes in TRL^pos tumors (log₂fold change > 1, adjusted P < 0.05), of which 227 genes were significantly up-regulated and 364 genes were significantly down-regulated (Fig. 1b; Supplementary Table 1). Unsupervised hierarchical clustering confirmed that the TRL^pos and TRL^neg tumor gene signatures were distinct (Fig. 1c). Genes up-regulated in TRL^pos tumors included those related to cancer–testis antigens (MAGEA1, MAGEA3, MAGEA6, MAGEC2, ACTL8), neutrophil chemotaxis (CXCL3, CXCL8), inflammatory cytokines (IL1B), and epithelial–mesenchymal transition (LOXL2, PFN2) [18], whereas down-regulated genes included those related to T cells (CD8A, CD8B), T-cell immune response (HLA-E, IDO1), and the complement pathway (C7) (Fig. 1b, Supplementary Table 1). GSEA using GO gene sets showed that TRL^neg tumors exhibit enrichment in gene sets related to lymphocyte-mediated immunity, cellular response to IFN-γ, and T cell activation (Fig. 1d), while TRL^pos tumors exhibit enrichment in gene sets related to DNA repair, replication, and cell cycle progression, implying the possibility of intrinsic radioresistance in TRL^pos tumors (Fig. 1d). Supporting this, TRL^pos tumors also showed enrichment in radioresistance-related genes (Fig. 1e) [19]. We further assessed the radioresistance signature in the TCGA cohort of uterine cervical cancer patients. Due to the lack of blood count data in the TCGA cohort and as indicative of TRL diagnosis, we calculated the LS score for each tumor, and defined LS^high and LS^low tumors as TRL^pos and TRL^neg tumors, respectively. The radioresistance signature score, derived from the radioresistance-related genes [19] by GSVA, showed a significant correlation with the LS (Fig. 1f), such that LS^high tumors exhibited a significantly higher radioresistance signature score than LS^low tumors (Fig. 1g). Patients with LS^high tumors had significantly lower survival rates than patients with LS^low tumors (Fig. 1h). Although TRL^pos or LS^high tumors exhibited signatures for radioresistance, a high LS score was still associated with a poor prognosis in patients who did not receive radiotherapy (Fig. 1i), implying that the poor prognosis of patients with TRL may stem not only from the innate radioresistance of tumor cells but also from other factors.

Fig. 1.

TRL^pos tumors exhibit a gene expression profile distinct from that of TRL^neg tumors. a Principal component analysis of tumors from patients with (TRL^pos; n = 4) and without (TRL^neg; n = 4) TRL. b Volcano plot demonstrating the differentially expressed genes between the TRL^pos and TRL^neg tumors. The blue and red dots indicate significantly down- or up-regulated genes in TRL^pos tumors, respectively (log₂fold change > 1, adjusted P < 0.05). c Unbiased hierarchical clustering analysis of differentially expressed genes in TRL^pos and TRL^neg tumors. d Expression levels of gene ontology (GO) gene sets in TRL^pos and TRL^neg tumors. A gene set enrichment analysis (GSEA) was used to identify positive (red) and negative (blue) enrichment of GO gene sets in TRL^pos tumors. The size of the nodes represents the number of genes in each pathway, and the links represent genes that are shared by two given pathways. Networks were generated using cytoscape. e GSEA of the radioresistance gene signature in TRL^pos (n = 4) and TRL^neg (n = 4) tumors. f Correlation plot between the LS score and radioresistance gene signature score in the TCGA uterine cervical cancer cohort (n = 303). Signature scores were calculated using a gene set variation analysis (GSVA). g Radioresistance gene signature scores in LS^high and LS^low tumors. The lines in the boxplot indicate median values, the boxes indicate the IQR values, and the whiskers extend to 1.5 × the IQR values. h, i Overall survival of patients with LS^low tumors and those with LS^high tumors in the whole TCGA cervical cancer cohort (h; n = 303) and patients who did not receive radiotherapy (i; n = 176). Student’s t-test (g), log-rank test (h, i). ****, P < 0.0001

We next investigated the tumor immune microenvironment in uterine cervical cancer patients of the TCGA cohort based on the LS score. The LS score was significantly and negatively correlated with the macrophage/monocyte activation signature and the lymphocyte infiltration signature score and was significantly positively correlated with the wound healing signature score (Fig. 2a). A clustering analysis of LS^high and LS^low tumors also showed that the LS^high tumors were enriched in the cluster with a high wound healing signature score and low macrophage/monocyte activation, IFN-γ response, TGF-β response, and lymphocyte infiltration signature scores (Fig. 2b). However, there were some LS^high tumors that exhibited high IFN-γ response, TGF-β response, and lymphocyte infiltration signature scores, implying heterogeneity in the LS^high tumors (Fig. 2b). We next evaluated the degree of MDSC activation in the TRL^pos and TRL^neg tumors. The suppressive MDSC gene signature was significantly enriched in TRL^pos tumors (Fig. 2c). In the TCGA cohort, the suppressive MDSC gene signature score was significantly correlated with the LS (Fig. 2d), and LS^high tumors exhibited higher suppressive MDSC gene signature scores than LS^low tumors (Fig. 2e). The LS^high tumors had a higher fraction of M2 macrophages, mast cells, and eosinophils and a lower fraction of M1 macrophages and lymphocytes (Fig. 2f). Among lymphocytes, the fractions of non-Treg CD4 T cells, B cells, and plasma cells were significantly higher in TRL^neg tumors (Fig. 2f).

Fig. 2.

LS^high tumors have a more suppressive immune microenvironment than LS^low tumors. a Correlation between the LS score and signature scores for the IFN-γ response, the TGF-β response, macrophage, lymphocyte infiltration, and wound healing in the TCGA uterine cervical cancer cohort (n = 303). b Unbiased hierarchical clustering analysis for the signature scores for the IFN-γ response, the TGF-β response, macrophage, lymphocyte infiltration, and wound healing signatures in LS^high (n = 75) and LS^low (n = 228) tumors. c GSEA of the immunosuppressive MDSC gene signature in TRL^pos (n = 4) and TRL^neg (n = 4) tumors. d Correlation plot between the LS score and the immunosuppressive MDSC gene signature score (n = 303). e Immunosuppressive MDSC gene signature scores in the LS^high (n = 75) and LS^low (n = 228) tumors. The lines in the boxplot indicate median values, the boxes indicate the IQR values, and the whiskers extend to 1.5 × the IQR values. f The fraction of different immune cell subsets in the LS^high (n = 75) and LS^low (n = 228) tumors. The fractions of cells were derived from CIBERSORT. Bar graphs represent mean and s.e.m. Student’s t-test (e, f). ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

Given the importance of chemotaxis in the recruitment of immune cells to peripheral tissues, we next examined the expression of chemokines. An unsupervised hierarchical clustering of 42 chemokines in 303 uterine cervical tumors demonstrated the clustering of chemokines related to CXCR2, a chemokine receptor that mediates both neutrophil and MDSC trafficking (Fig. 3a). Moreover, the expression of the CXCR2 chemokines CXCL1, CXCL2, CXCL3, CXCL5, and CXCL8 was significantly higher in LS^high tumors (Fig. 3b). In addition to chemokines, we also examined the expression of growth factors for neutrophils or monocytes, such as CSF1 (M-CSF), CSF2 (GM-CSF), and CSF3 (G-CSF), and found that LS^high tumors had a significantly higher expression of CSF2 and CSF3, but not CSF1 compared to LS^low tumors (Fig. 3c). Thus, LS^high tumors showed the potential for enhanced chemotaxis of neutrophils or MDSCs to the tumor site and higher expression of neutrophil- or MDSC-inducing cytokines.

Fig. 3.

Chemokine expression in LS^high and LS^low tumors. a Unbiased hierarchical clustering analysis for gene expression of 42 types of chemokine in the LS^high (n = 75) and LS^low (n = 228) tumors in the TCGA uterine cervical cancer cohort. b Gene expression of selected CXCR2-related chemokines in LS^high (n = 75) and LS^low (n = 228) tumors. c Gene expression of CSF1, CSF2, and CSF3 in LS^high (n = 75) and LS^low (n = 228) tumors. The lines in the boxplot indicate median values, the boxes indicate the IQR values, and the whiskers extend to 1.5 × the IQR values. Student’s t-test (b, c). ns, not significant; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

Considering the role of MDSCs in disrupting the effectiveness of immune checkpoint blockade, we next investigated the potential of TRL^pos and TRL^neg or LS^high and LS^low tumors to respond to anti-PD-1 therapy. The T cell-inflamed gene signature, which has been shown to predict responses to anti-PD-1 therapy in multiple types of cancer [17], was found to be significantly enriched in TRL^neg tumors (Fig. 4a). The T cell-inflamed gene signature score was also higher in LS^low tumors, with a borderline significance (P = 0.0581; Fig. 4b). The neoantigen number, which is also known to be a predictive marker for anti-PD-1 therapy [20], was similar in the LS^high and LS^low tumors (Fig. 4c). TCR and BCR diversity was significantly lower in the LS^high tumors (Fig. 4d).

Fig. 4.

TRL^pos and LS^high tumors exhibit features of anti-PD-1 unresponsiveness. a GSEA of the T-cell-inflamed gene signature in TRL^pos and TRL^neg tumors. b T-cell-inflamed gene signature scores in LS^high (n = 75) and LS^low (n = 228) tumors. c Neoantigen number in LS^high (n = 75) and LS^low (n = 228) tumors. d T-cell receptor (TCR) and B-cell receptor (BCR) diversity in LS^high (n = 75) and LS^low (n = 228) tumors. Diversity indicates the Shannon entropy score. The lines in the boxplot indicate the median values, the boxes indicate the IQR values, and the whiskers extend to 1.5 × the IQR values. Bar graphs represent mean and s.e.m. Student’s t-test (b-d). ns, not significant; *, P < 0.05; ****, P < 0.0001

To extend our findings derived from patients with uterine cervical cancer, we compared the survival of patients with LS^high and LS^low tumors and the immune signatures of these tumors in the pan-cancer cohort of the TCGA database comprising 32 types of solid tumors and 9984 patients. Patients with LS^high tumors exhibited a significantly poorer overall survival compared to those with LS^low tumors (HR 1.57, 95% CI 1.48–1.67, P < 0.0001; Fig. 5a). LS^high tumors also had a significantly higher MDSC signature score than LS^low tumors (Fig. 5b). As seen in the TCGA uterine cervical cancer cohort, LS^high tumors exhibited a higher wound healing signature, but lower IFN-γ response, TGF-β response, macrophage/monocyte activation, and lymphocyte infiltration signatures (Fig. 5c). In terms of chemokine expression, LS^high tumors expressed significantly higher levels of CXCL1, CXCL2, CXCL3, CXCL5, and CXCL8 (Supplementary Fig. 1a). In addition, LS^high tumors expressed higher levels of CSF2 and CSF3 (Supplementary Fig. 1b). We next evaluated the prognostic impact of the LS score in the 32 types of solid tumors. A higher LS score was associated with poorer overall survival not only in uterine cervical cancer but also in adrenocortical carcinoma, bladder urothelial carcinoma, head and neck squamous cell carcinoma, kidney chromophobe, kidney renal clear cell carcinoma, kidney renal papillary cell carcinoma, liver hepatocellular carcinoma, lung adenocarcinoma, mesothelioma, sarcoma, skin cutaneous melanoma, uterine corpus endometrial carcinoma, and uveal melanoma (Fig. 5d).

Fig. 5.

Pan-cancer analysis of LS^high and LS^low tumors. a Overall survival of patients with LS^high and LS^low tumors in the TCGA pan-cancer cohort (n = 9984). b Immunosuppressive MDSC gene signature scores in the LS^high (n = 2496) and LS^low (n = 7488) tumors. c Signature scores for the IFN-γ response, the TGF-β response, macrophage, lymphocyte infiltration, and wound healing signatures in LS^high (n = 2496) and LS^low (n = 7488) tumors. d Forest plot demonstrating the hazard ratio and 95% CI of the LS in overall survival among 32 types of solid tumors. Tumor types with a significant association between overall survival and the LS are marked in red. Bar graphs represent mean and s.e.m. Student’s t-test (b, c). Cox regression analysis (d). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

Discussion

Although a large number of studies have demonstrated the prognostic impact of TRL in various cancer types and treatment settings [4–7], the microenvironment in human TRL^pos tumors has not been well studied. Here, we comprehensively characterized the immune microenvironment of TRL^pos tumors in patients with cervical cancer, as well as various other types of cancers. TRL^pos and TRL^neg tumors exhibited distinct transcriptional profiles and TRL^pos tumors exhibited increased radioresistant signature, but reduced lymphocyte activation signature. We utilized DEGs from the TRL^pos and TRL^neg tumors to derive an LS score and defined LS^high and LS^low tumors in the TCGA cohort as representative of TRL^pos and TRL^neg tumors, respectively. Further analysis revealed that LS^high tumors exhibit a more suppressive immune microenvironment than LS^low tumors, characterized by higher MDSC and wound healing signatures and lower macrophage activation and lymphocyte infiltration signatures. Furthermore, we found that TRL^pos or LS^high tumors had higher expression levels of CXCR2 chemokines than TRL^neg or LS^low tumors, suggesting the CXCR2 axis may be a promising target for patients with TRL.

TRL in cervical cancer has been suggested as a distinct clinical entity with a lower response rate to radiation and a higher potential for metastasis or locoregional recurrence [5, 11]. As from the previous report [4], we found that patients with TRL tended to have lower age, larger tumor, and more advanced tumor stage than patients without TRL. Importantly, we found that TRL^pos and TRL^neg tumors have distinct transcriptional profiles. One of the features of TRL^pos tumors was enrichment of radioresistant gene signatures. Extending these findings to the TCGA cohort, we found that patients with LS^high tumors also had a higher radioresistant signature and poorer survival than patients with LS^low tumors. However, a high LS score was also associated with poorer survival in patients treated without radiation, which implies that the poor prognosis of LS^high patients may not be solely attributed to radioresistance. As a result, we focused on examining the immune microenvironment.

TRL^pos or LS^high tumors exhibited features of suppressive anti-tumor immunity. These tumors were enriched for suppressive MDSC gene signatures, implying that MDSCs are more likely to infiltrate human tumors associated with leukocytosis. In addition, not only MDSCs, but also M2 macrophages, mast cells, and eosinophils were more abundant in LS^high tumors. M2-polarized macrophages in tumors, or TAMs [9], mast cells [21], and eosinophils [22] are known to have pro-tumorigenic effects, which supports the suppressive anti-tumor immunity in patients with TRL. TAMs and mast cells are also key players in the chronic wound healing process [23], and we found that LS^high tumors exhibit an elevated wound healing process signature. The chronic wound healing process has remarkable similarities with tumor growth since during the wound healing process an immunosuppressive environment is established through the action of a diverse array of immune cells and cytokines [24]. LS^high tumors also showed a lower lymphocyte fraction, notably including non-Treg CD4 T cells, B cells, and plasma cells. A higher faction of plasma cells in tumors has been shown to be a predictor of favorable survival, which is associated with tertiary lymphoid structures (TLSs) in the cancer microenvironment [25]. TLSs are ectopic lymphoid organs that develop in non-lymphoid tissues that are composed of a T cell-rich zone and B cell follicle with germinal center characteristics, are surrounded by plasma cells, and represent a privileged site for tumor antigen presentation to T and B cells [26]. The presence of TLSs in tumors has been shown to be associated with favorable clinical outcomes [26].

Although the mechanism behind the formation of TRL in humans is yet to be revealed, preclinical studies have suggested that G-CSF secreted from tumor cells may play a role in inducing the generation of MDSCs from the bone marrow [4, 12]. Here, we also found that the expression of CSF2 (GM-CSF) and CSF3 (G-CSF) is elevated in LS^high tumors. In addition, we also examined the expression of chemokines, since chemotaxis is a crucial process for immune cell recruitment to peripheral tissues. We found that LS^high tumors have a higher expression of CXCR2 chemokines, such as CXCL1, CXCL2, CXCL3, CXCL5, and CXCL8 (IL-8). The CXCR2 chemokine axis has been shown to serve a central role in the recruitment of MDSCs leading to tumor progression, and blocking CXCR2 trafficking leads to enhanced anti-tumor effects of anti-PD-1 therapy in mouse tumor models [27, 28]. In humans, IL-8 has been suggested to be a potent chemotactic factor for MDSCs [29]. Therefore, targeting CXCR2 chemotaxis may be an effective strategy to improve the treatment outcome in patients with TRL. In human subjects, an anti-IL-8 antibody showed a favorable safety profile and reduction in serum IL-8 levels; however, there was no objective tumor response in patients with metastatic or unresectable solid tumors [30]. Although blocking the CXCR2 axis itself did not cause dramatic tumor reductions, it may be promising for combination with other treatments, such as radiotherapy, since local radiotherapy has been shown to enhance the production of IL-8 in tumors [31] and increase the frequency of MDSCs [32].

We found that TRL^high or LS^high tumors exhibit a significantly lower T-cell-inflamed signature which is a well-established gene signature that predicts the response to anti-PD-1 therapy [17]. In addition, LS^low tumors exhibited a lower TCR and BCR diversity than LS^high tumors. A previous study has also shown that the diversity of peripheral blood T-cell repertoire is negatively correlated with NLR [33]. Furthermore, a higher TCR diversity has been associated with the response to immune checkpoint blockade [34]. Previous studies have also demonstrated that a high absolute neutrophil count or NLR, which is tightly associated with TRL, is associated with a poor response to anti-PD-1 therapy [35–37]. Taken together, tumors associated with leukocytosis are less likely to respond to immune checkpoint blockade and further combinatorial treatments may be needed.

We performed our initial analysis of TRL^pos and TRL^neg tumors in uterine cervical cancer patients, but found that the findings could be reproduced in multiple types of solid tumors. Cancer types that showed a significant association between a higher LS score and poor overall survival in our study have also been shown to have significant association between TRL or high NLR, with poor prognosis [4–7, 38–40]. However, a higher LS score was significantly associated with poorer survival, but not in all types of cancers. We, thus, suggest that TRL may not have the same prognostic impact in all types of cancers and that the mechanisms for this differential impact in various cancers should be the subject of future investigations.

The current study had some limitations. First, the blood count data lacked in the TCGA cohort, which limited the diagnosis of TRL. As an alternative, we used the DEGs obtained from TRL^pos and TRL^neg tumors as a surrogate to define TRL in patients in the TCGA cohort. Another limitation is that RNA was extracted from FFPE tissue sections; so as to overcome this limitation, we only used RNA samples that passed quality control tests to assure the accuracy of sequencing data. Moreover, a low number of patients were included for comparison of TRL^pos and TRL^neg tumors. In addition, TRL^pos patients exhibited a heterogeneous clinical response. Although TRL^pos patients are suggested to display a distinct clinical entity [4], one of the four patients was recurrence-free for over 4 years, while the others recurred within 1 year post-treatment. Future prospective studies utilizing prospectively collected fresh tissue with blood count data in a larger cohort will better define the immune landscape of tumors associated with leukocytosis.

In conclusion, tumors associated with leukocytosis exhibit a more suppressive immune microenvironment characterized by a higher wound healing signature accompanied by a higher infiltration of MDSCs and TAMs with lower lymphocyte infiltration compared to tumors not associated with leukocytosis. Our findings merit future clinical trials applying novel therapeutics based on the understanding of the distinct immune microenvironment of tumors associated with leukocytosis to improve the treatment outcomes of cancer patients with TRL.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

BCR

B cell receptor

DEG

Differentially expressed gene

FFPE

Formalin-fixed paraffin-embedded

G-CSF

Granulocyte colony-stimulating factor

GSEA

Gene set enrichment analysis

GM-CSF

Granulocyte–macrophage colony-stimulating factor

GO

Gene ontology

GSVA

Gene set variation analysis

IFN

Interferon

LS

Leukocytosis signature

M-CSF

Macrophage colony-stimulating factor

MDSC

Myeloid-derived suppressor cells

NLR

Neutrophil-to-lymphocyte ratio

PD-1

Programmed death-1

TAM

Tumor-associated macrophage

TCGA

The cancer genome atlas

TCR

T cell receptor

TGF

Transforming growth factor

TRL

Tumor-related leukocytosis

Author contributions

KHK and YBK contributed to the study concept and design. JSC and YBK involved in data acquisition. KHK, NSS, JSC, and YBK were involved in data analysis and interpretation. KHK drafted a preliminary version of the paper, and all authors reviewed and critically revised the manuscript.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional Review Board of Severance Hospital (4-2015-0454) and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study.