Xiaojing Chu,Yu Zhang,Sijin Cheng

Changping Laboratory,Beijing 102206,China

Abstract Tumor microenvironment (TME) is highly heterogeneous and composed of complex cellular components,including multiple kinds of immune cells.Among all immune cells in TME,tumor-infiltrating myeloid cells(TIMs) account for a large proportion and play roles as key regulators in a variety of functions,ranging from immune-mediated tumor killing to tumor immune evasion.Understanding the heterogeneity of TIMs will provide valuable insights for new therapeutic targeting of myeloid cells.Single-cell genomic technologies deciphering cell composition and gene expression at single-cell resolution have largely improved our understanding of the cellular heterogeneity of TIMs and highlighted several novel cell subtypes contributing to the variation of patient survival and treatment response.However,these cell subtypes were defined based on limited data without a concordant nomenclature,which makes it difficult to understand whether they exist in different studies.Thus,in this review,we comprehensively summarized the common agreements and current different opinions on the heterogeneity of TIMs gained from single-cell studies;evaluated the feasibility of current myeloid cell targets at single-cell level and proposed a uniform nomenclature for TIM subsets.

Keywords: Single-cell genomics; tumor-infiltrating myeloid cells; tumor microenvironment; tumor immunotherapy

Introduction

Tumors are complex ecosystems with highly heterogeneous cellular components.Immunotherapy has become one of the most promising new cancer treatments,which unleashes a patient’s immune system to attack cancer.The success of the immune-checkpoint blockade (ICB) drugs such as anti-PD1 antibodies demonstrated the great potential of modulating the function of T cells in the tumor.However,the ICB treatment efficacies are limited to 20%-30% of cancer patients (1),indicating the presence of other immune evasion mechanisms within the tumor microenvironment (TME).Myeloid cells,which include granulocytes,plasmacytoid dendritic cells (pDCs),conventional dendritic cells (cDCs),monocytes and macrophages,represent the most abundant immune cells that infiltrate into tumors and play vital roles in both immune-mediated tumor killing and tumor immune evasion (2).

Macrophages are a major component of tumorinfiltrating myeloid cells (TIMs) and their effects on tumor progression are controversial.Macrophages activatedin vitrocan be categorized into two distinct groups: classically activated macrophages (M1) and alternatively activated macrophages (M2),which play antitumor and pro-tumor roles,respectively (2).Macrophagesin vivowere also reported to engage in a dual relationship with tumors although they did not fit into such a simple polarization system.Dendritic cells (DCs) play a pivotal role in tumor immunity and the dysfunction of DCs leads to failures in immune surveillance and the elimination of tumor cells(3,4).Mast cells have been observed with a highly variable composition across different cancer types (5).They express vascular endothelial growth factor A (VEGFA) and many cytokines supporting their functions of pro-angiogenesis and immune modulation in TME (6,7).Neutrophils have been associated with tumor heterogeneity and immunotherapy efficacy (8,9).Several therapeutic approaches targeting TIMs are ongoing in pre-clinical and clinical studies,although their functional heterogeneity remains to be fully characterized (10).

With the aid of single-cell technologies,our understanding of the heterogeneity of TIMs has largely improved in recent years.In this review,we reviewed the major findings of the single-cell studies upon TIMs,and summarized the agreement and differences presented in these independent studies,including the functional models of tumor-associated macrophages (TAMs),maturation of DCs and composition of mast cells and neutrophils in TME.Then by illustrating the limitations of these studies and current clinical applications,we proposed uniform guidance of nomenclature for TIM subsets to avoid discordant characterizations in future studies.

Functional heterogeneity of TAMs

TAMs represent the most abundant myeloid cells in tumors,which were reported to support tumor growth by promoting angiogenesis,immune escaping and metastasis(11).However,instead of a common phenotype,TAMs exhibit strong plasticity and intercell heterogeneity,which make them can be either pro-or antitumor.Single-cell studies provided us with a great opportunity to acquire a deep understanding of TAM heterogeneity,reflecting their different functional statusin vivo.Here we noticed that several functionally characterized TAM subsets have been revealed in multiple studies (Figure 1).

FOLR2/TREM2 model

A study on breast cancer classifiedAPOE+TAMs into two distinct subclusters based on expression patterns ofFOLR2andTREM2.FOLR2+macrophages are tissue-resident macrophages,located in the tumor stroma and highly expressFOLR2andMRC1.FOLR2+macrophages are reported to be rarely infiltrating and reside in perivascular niches.Functionally,FOLR2+macrophages are reported to prime effector CD8+T cells and be beneficial to patients’survival.In contrast toFOLR2+macrophages,TREM2+macrophages are reported as recruited macrophages (12)and an experiment performed onTREM2deficient mice showed enrichment and activation of T cells and nature killer (NK) cells,which furtherly enhanced ICB therapy (13).

Myeloid-derived suppressor cells (MDSCs)-like/TAM-like models

Early studies of TME of hepatocellular carcinoma (HCC)and ovarian cancer revealed a subset of immature macrophages that enrich MDSCs signals and highly expressMARCOandTREM1(14,15).Interestingly,this cluster of immature macrophages is enriched in TME lacking CD8+T cells,with relevance to pro-tumor functions.In contrast to MDSCs-like macrophages,TAM-like macrophages,which are marked by high expression ofTREM2,enrich in tumor tissues with CD8+T cells infiltrated or excluded.TAM-like macrophages can be further classified into two subclusters:CD169+macrophages andCX3CR1+macrophages,and the former highly express CXCL9/10/11,ligands interacting with CXCR3 receptor expressed by T cells (15).

C1QC/SPP1 model

Based onC1QCandSPP1expression patterns,several studies classified TAMs into two major subtypes:C1QC+macrophages andSPP1+macrophages (16,17).Among them,C1QC+macrophages enrich phagocytosis and antigen presentation function and highly expressAPOE. As reported in a colon cancer study (18),C1QC+macrophages might be derived fromFCN1+monocytes and enriched in adjacent zones spatially.SPP1+macrophages express genes associated with angiogenesis,such asVEFGA(5),and populate in the tumor core section in colon cancer (18).Importantly,SPP1+macrophages are enriched in patients not responding to PD-L1 blockade in both triple-negative breast cancer (19) and colorectal cancer (20).Moreover,SPP1+macrophages can interact with tumor-specificFAP+fibroblasts,which are considered to promote tumor progression (20).

Considering that,many studies have reportedSPP1as one of the marker genes ofTREM2+macrophages,SPP1+macrophages andTREM2+macrophages might largely be overlapped with each other.This was supported by a finding thatTREM2+macrophages identified by Mulderet al.(21) were clustered closely with anSPP1+macrophage subset identified in colon cancer (22).Moreover,a recent pan-cancer integrative analysis of TAMs revealed that theSPP1/C1QCmodel could be used to annotate TAMs in multiple cancer types,suggesting that this model was more likely to reflect TAM functional heterogeneityin vivo.

Figure 1 Classification models of TAMs.To characterize the transcriptional heterogeneity,TAMs have been classified into subsets based on the expression patterns of different marker genes in different studies.Classification models include FOLR2/TREM2 model,MDSClike/TAM-like model and C1QC/SPP1 model.TAM,tumor-associated macrophage;TRM,tissue-resident macrophage;NK,natural killer;MDSC,myeloid-derived suppressor cell;TME,tumor microenvironment.

Other less characterized TAM subsets

Instead ofC1QC+andSPP1+TAMs,anISG15+TAM cluster has been identified in several cancer types,which are enriched in TME,and highly express several interferon-inducible genes (5).Moreover,another study has identified an immunosuppressiveIL4L1+TAM subtype(21).Interestingly,this cluster of macrophages shows a similar expression pattern toISG15+macrophages andLAMP3+DCs,which is a mature DC subtype enriched in TME.

Current therapeutic targeting of TAMs

Figure 2 Reported TAM-target genes and their expression in TAMs.(A) Three major strategies and reported target genes of therapeutic targeting TAMs;(B) Percentages of cells expressing target genes in different cancer types.X-axis indicates percent and colors indicate cancer types;(C) Percentages of cells expressing target genes in different macrophage subclusters.X-axis indicates percent and Y-axis indicates subclusters identified in Cheng et al (5).Different cancer types are indicated by colors.TAM,tumor-associated macrophage;TLR,toll-like receptor.

The abundance of TAMs is usually associated with poor prognosis in cancer patients,and the unique transcriptional profiles of TAMs make them good targets for antitumor therapy.Treatments on TAMs mainly aim to block the recruitment of macrophages,promote macrophage apoptosis and induce the transformation of macrophages from pro-tumor phenotype to anti-tumor phenotype (23)(Figure 2A).Technological advances in single-cell genomics allow us to evaluate the expression patterns of those wellknown target genes in TAMs.Using the TAM expression profiles collected by Chenget al.(5),we first compared the expression level of target genes in TAMs from different cancer types (Figure 2B),and observed thatCXCR4,CSF1RandSIRPawere highly expressed by TAMs in different cancer types,indicating such targets can be utilized for multiple cancer types.In contrast,PIK3CGandLILRB1,targets for transcriptional reprogramming of macrophages,show high expression variation in different cancer types,suggesting that the clinical application of such targets should be tailored towards different tumor types.Moreover,we also compared the expression of these target genes in different TAM subsets (Figure 2C),providing valuable insights into the susceptibility of TAM subpopulations for each target gene.For example,we observed a higher expression level ofCSF1RinC1QC+macrophages compared to that in other subclusters such asSPP1+macrophages,suggesting anti-CSF1R therapy was more likely to affect theC1QC+macrophages.

DC maturation and their regulatory roles in TME

The ability to initialize adaptive immune response makes DCs potential immunotherapy targets.DCs account for a small,heterogeneous but relatively constant proportion of TME.In general,DCs are classified into conventional DCs(cDCs) and plasmacytoid DCs (pDCs).cDCs can be further classified into two major subtypes cDC1s and cDC2s.After stimulation,cDC1s promote Th1 response and activate CD8+T cells and fewer cDC1s were observed in tumor tissues compared to normal tissues (24).Additionally,the abundance of cDC1s is associated with the immunotherapy response (25).Thus,cDC1 is usually considered an antitumor cell type (26).The functions of cDC2s and pDCs are controversial.cDC2s initialize Th2 and limit Th17 responses and interact with CD4+T cells,which are usually considered to be pro-tumor effects (27-29).However,it is also reported that cDC2 activates the antitumor CD4+conventional T cells (Tconvs) in the absence of Treg (30).pDC infiltration has been reported in TME and shows immunosuppressive effects.On the other hand,IFNα/β and TNFα produced by pDCs suggest an antitumor potential (31).Considering the complexity of DC populations in TME,many studies have focused on DCs,and instead of the classic DC subtypes (cDC1s,cDC2s and pDCs),we summarized some important subclusters and their potential functions identified in single-cell studies of tumors.

LAMP3+ DCs and mregDC

Different from cDC1 and cDC2,a new DC mature state calledLAMP3+DC (14) or mregDC (32) was reported recently in diverse tumors. This new DC state is characterized by specific high expression ofLAMP3,immune regulators [PD-L1 (CD274)andFAS],maturation genes (CD40andCD80) and migratory genes (CCR7andFSCN1).TheLAMP3+DCs are enriched in tumor tissues,both intratumorally and at the invasive margin (33),and work as an interacting hub with T cells in TME (34),suggesting their indispensable roles in TME.

A single-cell study on bladder urothelial carcinoma revealed a high expression ofCCL17,CCL19andCCL22and suggest a function of recruiting Tregs ofLAMP3+DCs(35).Besides,coculture with tumorLAMP3+DCs isolated from an ECSC patient can significantly suppress CD8+T cell proliferation and their release of IL-2 and IFN-γ (36).Importantly,tumor-derivedLAMP3+DCs exhibit higher expression ofPD-L1compared to that from non-cancer tissues and anti-PD-L1 treatment could rescue the immunosuppressive effect ofLAMP3+DCs on CD8+T cells,suggesting thatLAMP3+DCs suppress CD8+T cells self-renew and function through PD-L1/PD-1 pathway(36).TheLAMP3+DCs were reported to originate from both cDC1 and cDC2,maintaining the capacity of interacting with CD8+T cells from cDC1 and interacting with Tregs from cDC2.We think that the diverse functions ofLAMP3+DCs could be characterized based on their developmental origins andIL12Bcould be used as a gene marker of cDC1-derivedLAMP3+DCs (5).

DC3

DC3 (CD88-CD1c+CD163+DC) is a circulating subtype of DCs not originated from common DC precursors(CDP) as cDCs and pDCs (37). DC3s are proinflammatory DCs,which can trigger the expansion of tissue-resident memory CD8+T cells and enhance MHC expression,and thus the presence of DC3s is associated with increased survival of patients with melanoma metastasis (38).In addition,as an intermediated functional subtype between cDC2 and monocyte,DC3 highly produces TNFα as monocyte and IL23,IL12p70 as cDC2(37).Notably,human DC3s can be differentiated from induced pluripotent stem cells,allowing the potential application of DC3s in tumor immunotherapy (39).

Current therapeutic targeting of DCs

GM-CSF/CSF2RA,FLT3L/FLT3 and immunostimulating ligands of toll-like receptors (TLRs) have been considered in DC-targeted immunotherapies by promoting the differentiation,migration and maturation of DCs (3).Other strategies contain stimulating DCs with known tumor-associated antibodies to induce cancer vaccination and improve the efficiency of DC by inhibiting checkpoint SHP1 (also called PTPN6) (40) and silencing negative immune regulatory molecule IDO (41,42) and IL6/STAT3 signaling pathway (43,44). Besides,as DC-specific membrane receptors,DEC-205,CD40,CD209,CLEC4A,CLEC7A,CLEC9A and CLEC12A have been targeted for delivering antibodies and adjuvants,which can further enhance the cross-presentation of DCs (3).Accessing the expression patterns of these target genes in single-cell data,we found that FLT3 and IDO1 have higher expression in cDC1s andLAMP3+DCs than in cDC2,whileTLR2is enriched in cDC2,suggesting different DC subtypes may have different susceptibilities to drugs that interfere with those targets (Figure 3).

Controversial effects of mast cells in TME

Figure 3 Reported DC-target genes and their expression in DC subclusters.X-axis indicates the percent of cells expressing target genes,and Y-axis indicates DC subclusters identified in Cheng et al (5).Different cancer types are indicated by colors.DC,dendritic cell;pDC,plasmacytoid dendritic cell;cDC,conventional dendritic cell;TLR,toll-like receptor.

Mast cell has higher proportions in tumor tissues than in normal tissues in most cancer types and compared to nontumor tissues,mast cells in TME down-regulate the expression of TNF.However,the effects of mast cells on tumor progression and immunotherapy response are still controversial.The pro/antitumor effects of mast cells highly vary among different cancer types or different progression statuses.For example,the abundance of mast cells in TME was found to be positively associated with longer survival of patients with colon cancer and nasopharyngeal cancer (5,45),and with shorter survival in clear cell renal cell carcinoma and uterine corpus endometrial carcinoma (UCEC) (5,46).

Mast cells have the effects of pro-angiogenesis,supported by the observed high expression ofVEGFAin single-cell studies.Besides,mast cells can express antiinflammatory cytokines,such asIL10andTGFB,and recruit Tregs,contributing to the immune suppressive environment to support the tumor progression.Moreover,mast cells were reported to highly express PD-L1 in the early lung adenocarcinoma (47) and contribute to the resistance to anti-PD1 therapy in the high-grade serous ovarian cancer (48) and mouse melanoma model (49).At the same time,mast cells can also release chemokines such as CCL3 and CCL5 to recruit NK cells and T cells,which helps the T cell infiltrate tumor tissues.

Diverse neutrophils composition in TME

Despite their capacities of killing tumor cells and recruiting T cells,neutrophils in TME are considered to play protumor roles in general,for their expressing ARG1,reactive oxygen species,nitric oxide and immune checkpoint ligands programmed cell death ligand 1 (PD-L1) (10).Previous studies have associated neutrophils with cancer progression by forming an immunosuppressive microenvironment and helping tumor cells metastasize.A single-cell study on lung cancer has identified three subsets,conserved within human and mouse neutrophils,including a tumor-specific subset,which highly expressesCCL3,CSF1,CSTB,CTSBandIRAK2(50).Notably,another study on lung cancer reveals the difference in neutrophils between lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC).An accumulation of neutrophils was reported in LUSC,which might be driven by the overexpression of theSOX2/CXCL5pathway.At the same time,stronger interactions were observed between tumor cells and neutrophils in LUAD (51).Moreover,another study has reported an elevation of chemotaxis factors (e.g.,CXCL2andCXCL8) in TME in HCC,promoting the recruitment of neutrophils (52).

Uniform nomenclature for TIM subsets

Single-cell genomics has become a routine and informative approach for understanding tumor immunity.To annotate TIM subsets in a consistent and standardized manner facilitating cross-study analyses,we summarized 15 common TIM subclusters with their marker genes from a pan-cancer study and proposed a uniform nomenclature for future studies (Figure 4).Notably,among them,C01,C05,C09,C13,C14,C15 were found to be enriched in TME,suggesting they might play important roles in tumorigenesis and development.We also listed three types of tissue-resident macrophages (TRMs) that preferred to enrich in adjacent normal tissues.LYVE1+TRMs resemble the reported tissue-resident interstitial macrophages located adjacent to blood vessels (53).NLRP3+TRMs exhibit high expression ofIL1Band may have the ability to promote the inflammatory response.The alveolar TRMs were the only type of canonical TRMs identified in tumors and corresponding adjacent normal tissues. They maintained specific transcriptional characteristics with high expression ofPPARG,MARCO,MRC1andMSR1(11),and can be easily distinguishable from other TAMs.The relationships between different types of TRMs and TAMs have been computationally imputed by diverse trajectory analyses. However,to fully elucidate their complex developmental trajectories,rigorousin vitrocell differentiation andin vivolineage tracing studies are required.

Conclusions and perspectives

Figure 4 Nomenclature of myeloid cell clusters.Fifteen common functional myeloid cell clusters,which are marked by different representative genes,could be identified in single-cell studies of TME.TME,tumor microenvironment;pDC,plasmacytoid dendritic cell;cDC,conventional dendritic cell;DC,dendritic cell;TRM,tissue-resident macrophage;TAM,tumor-associated macrophage.

To sum up,in the review,we systematically summarized the current knowledge of the heterogeneity of TIMs from single-cell studies and revealed the expression patterns of current target genes in different subsets of TAMs and DCs at single-cell level.To guide the follow-up single-cell analyses of TIMs,we proposed a uniform nomenclature for TIM subsets.We believe that the unified nomenclature will facilitate researchers to deeply characterize the function of each TIM subset,especially for TAM subsets.Although single-cell technologies facilitate our understanding of the heterogeneity of TIMs,there remain some problems to be solved.For macrophages,we raise some points that might promote the understanding of TAM functional subclusters.Firstly,DCs,particularly,cDC2s and the newly identified DC3s (37) have relatively similar expression patterns compared to monocytes/macrophages in TME,which are reflexed by the close distance in clustering profiling,resulting in a mislabeling of TAMs,particularly in studies focusing on adaptive immune cells.Secondly,MDSCs,which are associated with tumor metastases (54),were not well annotated in single-cell sequencing-based studies.A better characterization of MDSCs might improve our understanding of the immunosuppressive tumor microenvironment (55).Thirdly,tissue-specific gene expression was observed in tissue-resident macrophages,contributing to cellular heterogeneity and difficulty in the functional identification of TAMs.

Like macrophages,DC studies were limited by cell type annotations.Because DC is a relatively rare cell type in TME and some subtypes have relatively similar gene expression patterns,such as cDC3,cDC2 and monocytes,it requires adequate resolution to identify subclusters.Thus,not every ontogenetic subtype has its responding functional subtype in single-cell studies of TME,and one function subtype could originate from two or three ontogenetic subtypes.A prefiltering by FACS might help the researchers to zoom into the specific subtype in future studies.Limited by the functional studies,from now on,none of the DC therapeutic targets were proposed or firstly identified in a single-cell-based study,remaining a lack of computational pipelines in the identification of new targets.

Mast cells and neutrophils were reported to play pathological roles in many immune-mediated diseases such as allergies and asthma (56-58).Despite their importance,only a few single-cell studies have focused on them.Due to the low abundance,it is hard to identify subclusters of these cell types in TME,resulting in double-edged effects reported and hard conclusions from the many cancer studies.

Considering the limitations of the current studies mentioned above,future studies of TIM could benefit from better cell type annotations or enriching specific cell types before sequencing and spatial transcriptomics tracking the cell location and adjacent relations.Furtherly,ideas about potential therapeutical targets might gain insights from computational approaches beyond clustering.

Acknowledgements

We thank Dr.Y.Miao and J.Wang for the discussion.The presented study is supported by Changping Laboratory.

Footnote

Conflicts of Interest: The authors have no conflicts of interest to declare.