the tumor microenvironment shapes lineage, transcriptional ...€¦ · daniel r. goldstein4,5, paul...

Research Article

The Tumor Microenvironment Shapes Lineage,Transcriptional, and Functional Diversity of InfiltratingMyeloid Cells

Kutlu G. Elpek1, Viviana Cremasco1, Hua Shen4,5, Christopher J. Harvey1, Kai W. Wucherpfennig1,Daniel R. Goldstein4,5, Paul A. Monach2, and Shannon J. Turley1,3

AbstractMyeloid cells play important regulatory roles within the tumor environment by directly promoting tumor

progression and modulating the function of tumor-infiltrating lymphocytes, and as such, they represent apotential therapeutic target for the treatment of cancer. Although distinct subsets of tumor-associated myeloidcells have been identified, a broader analysis of the completemyeloid cell landscape within individual tumors andalso across different tumor types has been lacking. By establishing the developmental and transcriptomicsignatures of infiltrating myeloid cells from multiple primary tumors, we found that tumor-associated macro-phages (TAM) and tumor-associated neutrophils (TAN), while present within all tumors analyzed, exhibitedstrikingly different frequencies, gene expression profiles, and functions across cancer types.We also evaluated theimpact of anatomic location and circulating factors on themyeloid cell composition of tumors. Themakeup of themyeloid compartment was determined by the tumor microenvironment rather than the anatomic location oftumor development or tumor-derived circulating factors. Protumorigenic and hypoxia-associated genes wereenriched in TAMs and TANs compared with splenic myeloid-derived suppressor cells. Although all TANs had analtered expression pattern of secretory effector molecules, in each tumor type they exhibited a unique cytokine,chemokine, and associated receptor expression profile. One such molecule, haptoglobin, was uniquely expressedby 4T1 TANs and identified as a possible diagnostic biomarker for tumors characterized by the accumulation ofmyeloid cells. Thus, we have identified considerable cancer-specific diversity in the lineage, gene expression, andfunction of tumor-infiltrating myeloid cells. Cancer Immunol Res; 2(7); 655–67. �2014 AACR.

IntroductionThe tumor microenvironment contains a multiplicity of

stromal cells of hematopoietic and nonhematopoietic develop-mental origin, such as T cells, B cells, natural killer (NK) cells,myeloid cells, fibroblasts, pericytes, adipocytes, and endothelialcells,which collectively shape thedisease course (1–4). Althoughspecific roles have been identified for discrete stromal subsets,factors controlling their recruitment, expansion, and function indifferent tumors remain enigmatic. Therefore, a more completecharacterization of these subsets and a better understanding of

how they are recruited to and expand within growing tumorsand metastases are of utmost importance to developing noveltherapies and improving existing ones against cancer.

Tumor growth is associated with the accumulation of avariety of myeloid cell types (2). Common myeloid cell pro-genitors in the bone marrow can give rise to myeloid cells withimmunosuppressive potential, often referred to as myeloid-derived suppressor cells (MDSC). Monocyte-like CD11bþ

Gr1low and granulocyte/neutrophil-like CD11bþGr1hi subsetsof MDSCs have been reported to accumulate in the spleen,liver, blood, and bone marrow during tumor progression.Within tumors, myeloid cells with similar phenotypes arereferred to as tumor-associated macrophages (TAM) or neu-trophils (TAN), possibly reflecting a more differentiated iden-tity. In vitro studies have shown that differentiation of bonemarrow progenitors into MDSCs requires a combination ofcytokines, particularly interleukin (IL)-6 and granulocyte col-ony-stimulating factor (G-CSF) or granulocyte macrophagecolony-stimulating factor (GM-CSF), and the transcriptionalregulator CCAAT/enhancer-binding protein (C/EBP; ref. 5).Although splenic MDSCs are considered a reservoir for tumor-infiltrating myeloid cells (6), the exact relationship betweenthese cells remains elusive. Accumulating evidence indicatesthat MDSCs, whether in the spleen or in the tumor, have directsuppressive effects on cytotoxic leukocytes. In addition to

Authors' Affiliations: 1Department of Cancer Immunology and AIDS,Dana-Farber Cancer Institute; 2Boston University School of Medicine;3Department of Microbiology and Immunobiology, Harvard MedicalSchool, Boston, Massachusetts; Departments of 4Internal Medicine and5Immunobiology, Yale School of Medicine, New Haven, Connecticut

Note: Supplementary data for this article are available at Cancer Immu-nology Research Online (http://cancerimmunolres.aacrjournals.org/).

Current address for K.G. Elpek: Jounce Therapeutics, Inc., Cambridge, MA02138.

Corresponding Author: Shannon J. Turley, Dana-Farber Cancer Institute,44 Binney Street, D1440a, Boston, MA 02115. Phone: 617-632-4990; Fax:617-582-7999; E-mail: [email protected]

doi: 10.1158/2326-6066.CIR-13-0209

�2014 American Association for Cancer Research.

CancerImmunology

Research

www.aacrjournals.org 655

on March 25, 2021. © 2014 American Association for Cancer Research. cancerimmunolres.aacrjournals.org Downloaded from

Published OnlineFirst March 31, 2014; DOI: 10.1158/2326-6066.CIR-13-0209

http://cancerimmunolres.aacrjournals.org/

MDSCs, conventional and plasmacytoid dendritic cells (DC)may exert immunoregulatory effects in tumors (2) using avariety of mediators such as indoleamine 2,3-dioxygenase(IDO), inducible nitric oxide synthase (iNOS), and arginase Ito suppress T-cell proliferation, cytotoxicity, and effectorcytokine production.

Given that myeloid cells with protumorigenic and immu-nomodulatory functions have been observed in multiple ani-mal tumor models and in patients with cancer, they representimportant targets for immunotherapy. Efforts are underway toidentify myeloid-focused strategies. Approved chemotherapyagents, such as gemcitabine (7), 5-fluorouracil (8), and suni-tinib (9), can eliminate or prevent the accumulation of MDSCs,especially in lymphoid organs, and retard tumor progression.Likewise, agents that block myeloid recruitment to tumors,such as CSF1R inhibitors (10), hold clinical promise. However,to improve current strategies and identify new universaltargets for therapeutic intervention, it is essential to under-stand how each myeloid cell subset from one tumor relates tothe same population in other tumor types.

In this study, we analyzed myeloid subsets in multiplemurine tumors to study how phenotype, frequency, and tran-scriptional profiles relate within different tumors, using triple-negative 4T1 breast cancer, Her2þ breast cancer, and B16melanoma as models. Strikingly, each tumor type containeda distinct myeloid cell landscape, with TAMs, TANs, and DCsrepresented in all tumors but at markedly different ratios,while systemic MDSC accumulation was exquisitely tumor-specific. Our data suggest that tumor type, rather than ana-tomic location, dictates myeloid composition in the tumorlesion. Furthermore, although each subset exhibits similartranscriptional signatures associated with its identity in dif-ferent tumors, our study demonstrates that functional differ-ences exist across myeloid subsets from different tumors. Inparticular, our data suggest that haptoglobin may represent abiomarker for tumors characterized by systemic accumulationof myeloid cells. Thus, our study provides important insightsinto the identity and functional characteristics of tumor-asso-ciated myeloid subsets. Perhaps more importantly, our datasupport that in-depth transcriptomic analysis of tumor-infil-trating myeloid cells may reveal novel therapeutically attrac-tive targets for cancer.

Materials and MethodsEthics statement

All animal work has been carried out in accordance with U.S.NIH guidelines. This study is reviewed and approved by theDana-Farber Cancer Institute Animal Care andUse Committee(protocol IDs: 04-025 and 07-038).

MiceSix-week-old, sex-matched C57BL/6, Balb/c, or CD45.1þ (B6.

SJL-PtprcaPepcb/BoyJ and CBy.SJL(B6)-Ptprca/J) mice werepurchased from The Jackson Laboratory. F1 (Balb/cxC57BL/6) mice were bred in-house. RIP1-Tag2 mice were obtainedfrom Mouse Model of Human Cancer Consortium [NationalCancer Institute (NCI)]. Mice with spontaneous invasive pan-creatic ductal adenocarcinoma (PDA; Pdx1-Cre LSL-KrasG12D

PtenL/þ) were kindly provided by Dr. DePinho [while at theDana-Farber Cancer Institute (Boston, MA); currently at theMD Anderson Cancer Center (Houston, TX); ref. 11].

Cell lines and tumor modelsB16, EL4, 4T1, A20, and CT26 cells were obtained from

American Type Culture Collection. Pan02 cells were obtainedfrom the NCI Developmental Therapeutics Program reposito-ry. The 4T07 cells were obtained from Dr. J. Lieberman (Har-vard University, Cambridge, MA); they were expanded, ali-quoted, and banked in liquid nitrogen. No additional authen-tication was performed on these cells. The Her2 cell line wasderived from a spontaneous breast tumor in a Balb/c Her-2/neu–transgenic mouse that was found to lose Her-2/neuexpression in vitro. Transgenic rat Her-2/neu sequence wasinserted into a pMFG retroviral vector to obtain a stablyoverexpressing cell line. Cell lines were maintained in Dulbec-co'sModified EagleMedium (B16 and EL4) or RPMI (4T1, Her2,A20, Pan02, CT26, and 4T07) supplemented with 10% FBS and1� penicillin/streptomycin.

For subcutaneous tumors, 1 to 2.5 � 105 cells were injectedinto the flanks of na€�ve mice (4T1, Her2, CT26, and 4T07 intoBalb/c; B16 and Pan02 into C57BL/6; A20 into Balb/cxCD45.1;and EL4 into C57BL/6xCD45.1), and tumor growth was moni-toredusing calipers until tumor size reached0.5 to 1 cmdiameter.For transgenic models, tumors were monitored and analyzed atdifferent time points based on tumor growth (RIP-Tag2 at 11–12weeks, and PDA and Her2 transgenic at 11–24 weeks).

Cell preparation and flow cytometrySpleens and tumors were processed by gentle mechanical

disruption with forceps followed by enzymatic digestion using0.2 mg/mL collagenase P (Roche), 0.8 mg/mL dispase (Invitro-gen), and 0.1mg/MLDNase I (Sigma). Tissueswere incubated indigestion medium at 37�C for 30 minutes; released cells werecollected and filtered, and the remaining tissue was further pro-cessed. Blood was collected into tubes containing 2 mmol/LEDTA, and bone marrow cells were isolated by flushing mediathrough tibias and femurs. Red blood cells were lysed usingammonium–chloride–potassium solution. Cells were resus-pended in fluorescence-activated cell sorting (FACS) buffercontaining 1% FBS and 2 mmol/L EDTA.

Cells were stained in FACS buffer containing FcR-blockingantibody (2.4G2) with combinations of fluorochrome-conju-gated or biotinylated antibodies against CD11b (M1/70),CD11c (N418), CD45 (30.F11), Gr1 (RB6-8C5), MHC class II(M5/114.15.2), CD80 (16-10A1), CD86 (GL-1), CD40 (HM40-3),CD26 (H194-112), CD103 (2E7), cKit (2B8), Flt3 (A2F10), orisotype controls purchased from BioLegend and BD Bios-ciences. Cells were analyzed using BD FACSAria II or BDFACSCalibur (BD Biosciences), and FlowJo Software (TreeStar,Inc.). For cell sorting from spleens, myeloid cells were enrichedby depletion of CD3eþ, CD19þ, and CD49bþ cells using bio-tinylated antibodies and anti-biotin beads (MACS; MiltenyiBiotec). In tumor samples, CD45þ cells were enriched bypositive selection using a biotinylated antibody and anti-biotinbeads (EasySep; STEMCELL Technologies). After enrichment,cells were stained with fluorochrome-conjugated antibodies

Elpek et al.

Cancer Immunol Res; 2(7) July 2014 Cancer Immunology Research656




and propidium iodide (Sigma) to exclude dead cells. Cells weresorted using a BD FACSAria II equipped with a 100-mm nozzlerunning at 20 psi. After an initial sort to verify purity, a secondsort was performed to collect myeloid cell populations of 95%to 100% purity directly into TRIzol reagent (Invitrogen). Sam-ple size for each population was as follows: CD11b�/int

Gr1�CD11cþMHCIIþ (N ¼ 4), CD11bþGr1lowCD11cþMHCIIþ

(N ¼ 3), CD11bþGr1lowCD11c�MHCII� (N ¼ 3), and CD11bþ

Gr1hiCD11c�MHCII� (N ¼ 2) from B16 tumors; CD11bþ

Gr1lowCD11cþMHCIIþ (N¼ 3), CD11bþGr1lowCD11c�MHCII�

(N ¼ 3), and CD11bþGr1hiCD11c�MHCII� (N ¼ 2) from4T1 tumors; CD11bþGr1lowCD11cþMHCIIþ (N ¼ 3) andCD11bþGr1hiCD11c�MHCII� (N ¼ 3) from Her2 tumors;CD11bþGr1lowCD11c�MHCII� (N ¼ 2) and CD11bþ

Gr1hiCD11c�MHCII� (N ¼ 3) from spleens of 4T1 tumor-bearing mice.

RNA isolation and microarray analysisTotal RNA was prepared from TRIzol using chloroform ex-

traction according to the manufacturer's protocol, and 100 ngof RNA from each sample was used for amplification, labeling,and hybridization by Expression Analysis, Inc. Mouse Gene ST1.0 chips (Affymetrix) were used formicroarray analysis. All dataare Minimum Information About a Microarray Experiment(MIAME)-compliant, and the raw data generated as part ofthe Immunological Genome Project (ImmGen) have beendeposited in a MIAME-compliant database [National Centerfor Biotechnology Information (NCBI) Gene Expression Omni-bus (GEO) data repository; record no: GSE15907 andGSE37448].Various modules included in GenePattern platform (BroadInstitute, Cambridge, MA) were used for data analysis. Addi-tional myeloid cell populations from healthy tissues analyzedwithin the same microchip batch of ImmGen were used as acomparison [F4/80 macrophages from peritoneal cavity (PC;n ¼ 2), Ly6Cþ DCs from bone marrow (n ¼ 2), CD24�Sirpaþ

DCs from bone marrow (n ¼ 3), CD24þSirpaþ DCs from bonemarrow (n ¼ 3), CD4 DC from spleen (n ¼ 1), CD11c�CD103þ

DC from small intestine (n ¼ 1), neutrophil from bone marrow(C57BL/6; n¼ 2), and neutrophil from bone marrow (Balb.c; n¼2)]. Raw data were normalized with ExpressionFileCreator mod-ule using the Robust Microarray Average (RMA) method. TheMultiPlot module was used for dataset comparisons, includingfold-change analysis and statistical filtering. In each analysis, wepreprocessed the dataset for populations included in specificanalyses. In fold-change analyses, an arbitrary cutoff value of 2was used, in combination with coefficient of variation (CV) < 0.5for replicates. The t test P < 0.05 were considered statisticallysignificant for each probe. After filtering, we selected all probeswith a mean expression above 100 in at least one of the popula-tions analyzed, which indicates actual expression with 95%confidence level. For hierarchical clustering, Pearson correlationwas used with datasets following log2 transformation, row cen-tering, and row normalization in HierarchicalClustering module.Heatmaps were constructed using HeatmapViewer module.Pathway enrichments were analyzed by Ingenuity Pathway Anal-ysis. DC- and macrophage-associated genes were determinedpreviously (12, 13), and a similar analysis was performed toidentify neutrophil-associated genes (unpublished data; K.G.

Elpek and S.J. Turley). Principal component analyses were con-ducted using the PopulationDistances module created by Imm-Gen (probes with mean expression value >120 selected, datasetlog2 transformed, probes with top 15% variability selected).

Cytokine arraysCell culture supernatants from 4T1, Her2, and B16 cells were

collected in two separate batches when cells were approxi-mately 80% confluent and analyzed with Mouse CytokineAntibody Array 3 (RayBiotech, Inc.) according to the manu-facturer's protocol. Spot densities weremeasured using ImageJ(NIH), and relative amounts were calculated on the basis ofpositive and negative controls. Only cytokines and chemokineswith a positive signal are shown.

ELISA for haptoglobinSerum was obtained from na€�ve or tumor-bearing mice.

Human sera were obtained from patients with metastaticbreast carcinoma upon enrollment into Institutional ReviewBoard (IRB)–approved vaccine trials at the Dana-Farber Can-cer Institute. Informed consent was obtained for each partic-ipant about usage of collected samples for research purposes.Samples were collected by routine procedures before begin-ning study treatment. Haptoglobin levels were quantified usingMouse or Human Haptoglobin ELISA (ICL, Inc.) according tothe manufacturer's protocol. Concentrations were calculatedusing a four-parameter logistics curve.

Statistical analysis for nonmicroarray dataData are expressed as mean � SD and were analyzed using

one-tailed, unpaired Student t test. ELISAswere analyzed usingANOVA and Tukey test. P < 0.05 was considered statisticallysignificant.

Results and DiscussionIdentification and enumeration of tumor-associatedmyeloid cells reveal tumor-specific diversity

Thus far, it remains uncertain how tumor origin, localization,and history influence the myeloid cell landscape of differentsolid malignancies. To address these questions, we initiallysought to compare the composition of myeloid cell infiltratesamong 4T1 and Her2, two distinct murine tumors sharing thesame mammary tissue origin. Tumors were established subcu-taneously in the flanks of Balb/c mice and analyzed when theyreached different tumor sizes. We characterized multiple mye-loid cell populations isolated from each tumor type on the basisof myeloid-specific surface markers CD11b, Gr1, and CD11c bycytofluorimetry. In both types of tumors and throughout tumorhistory, the CD45þ hematopoietic compartment was dominat-ed by CD11bþ cells (>75%; Fig. 1A and Supplementary Fig. S1A).The intratumoral CD11bþ cells comprised three subpopula-tions based on Gr1 and CD11c expression (Fig. 1A). In 4T1tumors, the hematopoietic compartment contained approxi-mately 20% CD11bþGr1hiCD11c� cells and >45% CD11bþ

Gr1low; with >60% of the latter expressing CD11c. Comparedwith 4T1, Her2 tumors were infiltrated by a relatively smallpopulation of CD11bþGr1hiCD11c� cells (3%), while themajority(>95%) of myeloid cells were CD11bþGr1lowCD11cþ. We also

Heterogeneity of the Tumor Myeloid Cell Landscape

www.aacrjournals.org Cancer Immunol Res; 2(7) July 2014 657




A

0 102

103

104

105

0

103

104

105

62

21

15

Gr1

CD

11

bCD45+ CD11b+Gr1low CD11b+Gr1hi CD11b–Gr1–

0 102

103

104

105

0

103

104

105

4T1

Her2

82

3

13

0 102

103

104

105

0

20

40

60

80

100

0 102

103

104

105

0

20

40

60

80

100

0 102

103

104

105

0

20

40

60

80

0 102

103

104

105

0

20

40

60

80

100

30838618

C

100

CD11b+Gr1low

CD11c+

CD11b+Gr1low

CD11c–

CD11b+Gr1hi

CD11c–

CD11b–Gr1–

CD11c+

100

101

102

103

104

100

101

102

103

104

2.91.6

100

101

102

103

104

100

101

102

103

104

3.7 1.7

100

101

102

103

104

100

101

102

103

104

451.2

CD

11

b

Naïve Her24T1

CD11c

Spleen

CD45+

0 102

103

104

105

0

20

40

60

80

100

0 102

103

104

105

0

20

40

60

80

0 102

103

104

105

0

20

40

60

80

100100

92 4 96

4T1

Her2

MHCII

0 102

103

104

105

0

20

40

60

80

100

96

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

15 85 20 80 25 75

Gr1

CD11b+

CD11c–

D

CD11b–CD103+ DC SI

Ly6C+PDCA1+ DC BM

CD4+ DC SplCD24+Sirpa– DC BMCD24–Sirpa+ DC BMCD11b–CD11c+ B16 Tmr

F4/80hi Mac PCGr1lowCD11c– 4T1 SplGr1lowCD11c– 4T1 TmrGr1lowCD11c– B16 TmrGr1lowCD11c+ B16 TmrGr1lowCD11c+ 4T1 TmrGr1lowCD11c+ Her2 Tmr

Neutro B6 BMNeutro Balbc BMCD11b+Gr1hi 4T1 TmrCD11b+Gr1hi 4T1 SplCD11b+Gr1hi Her2 TmrCD11b+Gr1hi B16 Tmr

E

Expression (log2)

3.27 13.50

Dendritic cellMacrophageNeutrophil

Signature gene

0

10

20

30

40

100

101

102

103

104

94 6

100

101

102

103

104

0

1

2

3

4

5

928

CD11c

0

20

40

60

100

101

102

103

104

595

100

101

102

103

104

0

20

40

60

34 66

0

3

6

9

12

100

101

102

103

104

8218

100

101

102

103

104

0

5

10

15 96 4

BtlaFlt3KmoP2ry10Hmgn3Bri3bpNapsaCcr7Klri1H2−Eb2Gpr68Traf1Gpr82KitGpr68Zbtb46Slamf7Rab30Dpp4Jak2Bcl11aHaaoAp1s3Spint2Tbc1d8Gpr114Pvrl1Gpr132Ass1Runx3AnpepH2−Q6CiitaH2−DMb2H2−AaH2−Ab1H2−EbCd164A930039A15RikTmem1951810011H11RikPecrComt1Pon3Mr1Fcgr1Camk1Myo7aMertkAbca1Cd14Pla2g15Plod1CtslTpp1Sepp1BlvrbGlulTspan14Tlr7Lamp2Tcn2Pld3Slc48a1Pla2g4aNlnPcyox1Tbxas1Akr1b10Amica1Cnn2Pstpip1Cbfa2t3Adam19Gm5068Fgl2Amica1Stfa2l1G0s2Fam122aGm5416Spatc1Prok2ArsgFert2Fgd4SqrdlPtplad2Tom1B430306N03RikCtsdFcgr3Clec5aTlr4Slc16a3Slfn1Csf3rSlfn4HdcRetnlgSgms2Cd300lfGrinaArg2Pdlim7Dok3Slc27a4TirapKlhl2Mboat7PxnPfkfb4Ppp1r3dPhospho1Sfxn5A530064D06Rik2010002M12RikDhrs9Gm99494732465J04RikFam123aKrt86RlfChi3l1MgamAsprv1Il1f9Mrgpra2aSlc2a3Abtb1Ankrd22Ceacam10Msl1Slc22a15Ccpg1Slc35a5A530023O14RikFoxd4E430024C06Rik

CD

11

b–C

D1

03

+ D

C S

IC

D24

+S

irpa

– D

C B

MC

D2

4–S

irp

a+ D

C B

MC

D4

+ D

C S

pl

CD

11

b–C

D1

1c

+ B

16

Tm

rL

y6

C+P

DC

A1

+ D

C B

M

Ne

utr

o B

6 B

MN

eu

tro

Ba

lbc B

MC

D1

1b

+G

r1h

i 4T

1 T

mr

CD

11

b+G

r1h

i 4T

1 S

pl

CD

11

b+G

r1h

i He

r2 T

mr

CD

11

b+G

r1h

i B1

6 T

mr

F4

/80

hi M

ac P

C

Gr1

lowC

D1

1c

– 4

T1

Sp

l

Gr1

lowC

D1

1c

– B

16

Tm

rG

r1lo

wC

D1

1c

+ B

16

Tm

rG

r1lo

wC

D1

1c

– 4

T1

Tm

rG

r1lo

wC

D1

1c

+ 4

T1

Tm

rG

r1lo

wC

D1

1c

+ H

er2

Tm

r

CD

11

b

B

Elpek et al.





found a Gr1�CD11cþ population among CD11b� cells, whichcomprised <1% of all hematopoietic cells in both 4T1 and Her2tumors. Further analysis showed thatCD11bþGr1lowCD11cþ andCD11b�Gr1�CD11cþ cells in the tumor were largely MHCIIþ,whereas CD11bþGr1lowCD11c� and CD11bþGr1hiCD11c� cellswere MHCII� (Fig. 1B). A marked accumulation of splenicCD11bþGr1low and CD11bþGr1hi cells (Fig. 1C) was observedin 4T1-bearing mice, whereas the splenic numbers in micewith Her2 tumors were similar to na€�ve mice. These datasuggest that each tumor type may contain a distinct myeloidcell infiltrate.

Lineage determination of tumor-infiltratingmyeloid cellsubsets based on transcriptional profilingWe next sought to determine the lineage of each tumor-

infiltrating myeloid cell population using transcriptomic anal-ysis. To this end, myeloid subsets from 4T1 and Her2 breastcarcinomas and B16 melanoma were sorted to high purity,according to the ImmGen standard operating procedure.Transcriptional profiling data were generated on AffymetrixST1.0 microarrays adhering to profiling and quality controlpipelines of ImmGen (14). We used hierarchical clustering ofapproximately 13.5 K probes, after excluding those withCV < 0.5 andmean expression values <120 to compare popula-tions of tumor-infiltrating myeloid cells with representativemyeloid cell subsets (DCs,macrophages, andneutrophils) from

lymphoid and nonlymphoid tissues of healthy animals. Den-dogram analysis based on gene clustering depicted similaritiesbetween neutrophils and tumor-derived Gr1hi cells, macro-pahges and Gr1low cells (regardless of CD11c expression), andconventional DCs and CD11b�Gr1�CD11cþ cells (Fig. 1D).Hierarchical clustering of 139 probes corresponding to signa-ture genes of DCs, macrophages (12, 13), and neutrophils(unpublished data; K.G. Elpek and S.J. Turley) demonstratedthat tumor-infiltrating myeloid cell populations are indeedCD11cþ and CD11c� TAMs, TANs, and DCs (Fig. 1E andSupplementary Table S1).

Striking heterogeneity in myeloid cell infiltrates acrossdifferent tumor types

Among the three tumor types analyzed, Her2 tumors wereuniquely characterized by the presence of a large CD11cþ

TAM population, whereas 4T1 tumors contained a relativelylarge TAN population (Fig. 2A). In B16 tumors, myeloid cellscomprised only 40% of tumor-infiltrating leukocytes com-pared with >75% in 4T1 and Her2 tumors (Fig. 2A). Theaforementioned differences in myeloid abundance alsoreflected variation in the abundance of other leukocytesinfiltrating tumors, including T, B, NK, NKT cells, and pDCs.For example, B16 tumors were enriched in CD3þ T cellscompared with 4T1 and Her2 tumors, consistent with a moreimmunogenic environment in B16 melanoma (Fig. 2B). To

C

0

20

40

60

80

100

CD

45

+ in tum

or

A

%S

ubsets

within

CD

45

+ in tum

or

0

20

40

60

80

100

Other/

nonmyeloidDCsTANsCD11c+

TAMsCD11c–

TAMs

4T1

Her2

B16

Other/

nonmyeloidDCsTANsCD11c+

TAMsCD11c–

TAMs

4T07 Pan02 EL4 CT26 PDA RIP-Tag2 A20

*** *

*** ***

******

* **

*

*****

4T1 Her2 B160

10

30

20

****

%C

D3

+ w

ith

in C

D4

5+ in

tu

mo

r

B

%S

ubsets

within

Figure 2. Abundance ofmyeloid cellsubsets across different tumors. A,Whisker box plots summarizing therelative frequency of myeloidsubsets among hematopoieticcells within 4T1, Her2, and B16tumors. B, frequency of CD3þ Tcells within hematopoietic cells in4T1, Her2, and B16 tumors. C,frequency of myeloid cells in 4T07,Pan02, EL4, CT26, PDA, RIP-Tag2,and A20 tumors (n ¼ 3–7).�, P < 0.05; ��, P < 001; ��,P < 0.001.

Figure 1. Multiple myeloid cell subsets are present within tumors of different origin. A, analysis of myeloid cell subsets within subcutaneous 4T1 and Her2tumors by flow cytometry. Hematopoietic cells were analyzed for the expression of CD11b and Gr1, and gated populations in the dot plots werefurther analyzed for CD11c expression. B,MHCII expression by indicated subsets. C, accumulation of CD11bþGr1þ cells in the spleens of 4T1 tumor-bearingmice (top) and Gr1 expression (bottom). Numbers indicate percentage of cells for each gate or region. D, dendrogram analysis based on hierarchicalclustering of tumor-associated and steady-state myeloid cells (BM, bone marrow; SI, small intestine; Spl, spleen; Tmr, tumor). Sample sizes are indicated inMaterials and Methods. E, hierarchical clustering of the same populations based on a list of 139 genes associated with DCs, macrophages, or neutrophils.






further ascertain the degree of myeloid heterogeneity, weextended our analysis to five additional transplantedtumors, including 4T07 breast cancer, CT26 colon carcino-ma, EL4 T lymphoma, A20 B lymphoma, and Pan02 PDA, andthree autochthonous mouse models, including RIP-Tag2(insulinoma), Her-2/neu (mammary carcinoma), andPdx1-Cre LSL-KrasG12D PtenL/L (PDA) mice. The samefour major populations were present within each tumormodel but at significantly different frequencies across thetumors analyzed (Fig. 2A–C).

In all tumors, DCs were present at very low frequencies, withthe highest levels in B16, EL4, and RIP-Tag2 tumors (4%–10%; Fig. 2). Given that DCs account for only a small percentageof hematopoietic cells within the tumors, and that the majorsubset expressing CD11c in tumors is TAMs, our study alsosuggests that CD11c alone is an unreliable marker for DCs inneoplastic lesions. Further analysis comparing DCs in B16tumors with TAMs indicated that while TAMs expressed avariety of endocytic and regulatory receptors, including Pilrb,Sirpa, Lilrb4, and Clec5a (Supplementary Fig. S1B), tumor DCsexpressed signature genes such asDpp4, Flt3, andKit, as well asCD8/CD103 DC-associated genes (13), including Itgae, Batf3,Tlr3, Ifi205, and Xcr1, identifying them as CD11b�CD103þ

tissue-resident DCs. Flow cytometric analysis confirmed thatDCs were the myeloid cells expressing high levels of CD26(Dpp4), Flt3, cKit, and CD103 (Itgae) protein in these tumors(Supplementary Fig. S1C; and data not shown). On the basis ofthese distinctions, DCs may be considered as a referencepopulation in future studies to identify unique targets onTAMs and TANs to modulate their tumor-promoting func-tions. For example, DCs lack the surface receptorClec5a (MDL-1), whereas expression in TAMs, TANs, and neutrophils isrelatively high (>18-fold higher than in DCs), suggesting thatit may represent an attractive target for antibody-mediateddepletion of CD11bþGr1þ cells in tumors (15).

Differences in Gr1low and Gr1hi cells across several diversetumors were not restricted to the tumor lesion itself. Forexample, a substantial increase in MDSCs was observed inthe spleen and blood of 4T1 tumor-bearing mice throughouttumor progression but not in mice bearing other tumor types(Supplementary Fig. S1D and S1E).

Differences in myeloid cell composition arise fromspecificity of the tumor microenvironment

Themyeloid composition of different tumors is likely shapedby a combination of tumor-specified growth factors. Indeed,4T1, Her2, and B16 cell lines produce unique combinations ofcytokines and chemokines, suggesting that each tumor maycreate a distinctive microenvironment (Supplementary Fig.S2A). These differences may also explain why MDSCs prefer-entially accumulate in certain tumors. For example, high levelsof G-CSF and LIX (CXCL5) expression by 4T1 cells maycontribute to systemic accumulation of MDSCs in vivo (16,17). To assess whether specific tumor microenvironmentsdirectly influence the myeloid cell content, we analyzed mye-loid cells within the same tumor type growing at differentanatomic locations. Comparison of myeloid cell compositionin subcutaneous Her2 tumors and the parental spontaneous

tumor in mammary glands revealed that both lesions con-tained high frequencies of CD11cþ TAMs (50%–72%) andrelatively low frequencies of TANs (<5%) among hematopoieticcells (Fig. 3A and Supplementary Fig. S2B). A strong similaritywas observed between the myeloid cell composition of sub-cutaneous 4T1 tumors and spontaneous metastatic lesions inthe peritoneal cavity, occurring within 2 to 3 weeks after tumorchallenge. Similar to primary 4T1 tumors, thesemetastases areenriched with TAMs (�50%, of which >60% are CD11cþ) andTANs (15%–21%) among hematopoietic cells (Fig. 3A andSupplementary Fig. S2C). Furthermore, these results werecorroborated in a B16 metastatic model. When B16 cells areinjected intravenously, tumor nodules grow in lungs andkidneys within 2 to 3 weeks and contain a similar myeloidlandscape to that of subcutaneous tumors: approximately 30%TAMs (�45% expressing CD11c) and 4% to 9% TANs amonghematopoietic cells (Fig. 3A and Supplementary Fig. S2D).Altogether, these data provide evidence that tumor type ratherthan its anatomic location shapes the myeloid infiltrate.

Next, we tested the possible dominant effect of the micro-environment in dictating myeloid cell composition by analyz-ing myeloid cells in mice with two tumors growing simulta-neously. For this purpose, we compared Her2 and 4T1 or Her2and B16 tumors, as each of these tumors has distinct myeloidinfiltrates (Figs. 1 and 2). Balb/c mice were inoculated withHer2 tumors under one flank and simultaneously with 4T1tumors under the contralateral flank (Fig. 3B and Supplemen-tary Fig. S2E). Similarly, we established Her2 tumors in thepresence of B16 tumors on the opposite flank in F1 (Balb/c xC57BL/6) mice (Fig. 3C and Supplementary Fig. S2F). In bothmodels, we did not observe a difference in growth rate of eithertumor when compared with tumors grown alone (data notshown). Likewise, the myeloid content of each tumor, eitheralone or in the presence of a second tumor in the samehostwassimilar (Fig. 3B and C, left and Supplementary Fig. S2E andS2F), even though there was pronounced systemic accumula-tion of MDSCs associated with 4T1 tumor growth (Fig. 3B andC, right). Overall, these results suggest that the signals attract-ing specificmyeloid cells into tumors are driven by cancer cellsin the primary lesion, regardless of anatomic location orpresence of distant signals from other microenvironments.

Tumor-infiltrating myeloid cells exhibit a hypoxia-associated signature compared with their spleniccounterparts

As shown in Fig. 1, MDSCs accumulate in the spleens of 4T1tumor-bearing mice at high frequencies. As splenic MDSCsmay serve as reservoirs for tumor-infiltrating cells (6), it isimportant to understand how they relate to their counterpartswithin tumors. Thus, we compared the gene expression profilesof CD11bþGr1þ cells from the spleens and tumors of 4T1tumor-bearing mice. As shown in Fig. 4A, 263 probes wereupregulated in TAMs (both CD11c� and CD11cþ subsets)within 4T1 tumors comparedwithGr1lowMDSCs from spleens,whereas 152 probes were enriched in splenic cells. In addition,201 probeswere enriched in TANs comparedwith splenic Gr1hi

MDSCs, while 647 probes were enriched in Gr1hi MDSCs (Fig.4B). Next, we compared expression of probes upregulated in

Elpek et al.





Gr1

CD11c

CD

11b

Primary

(s.c.)

100

101

102

103

104

100

101

102

103

104

59

7

100

101

102

103

104

100

101

102

103

104

696

4 96

Mammary gland

(transgenic)

CD45+

TAMs

100

101

102

103

104

100

101

102

103

104

44 12

100

101

102

103

104

100

101

102

103

104

60 21

Peritoneal met

(spont.)

100

101

102

103

104

100

101

102

103

104

333.8

100

101

102

103

104

100

101

102

103

104

314

100

101

102

103

104

100

101

102

103

104

283

Lung met

(i.v.)Kidney met

(i.v.)

Gr1

CD11c

A Her2 4T1 B16

B

100

101

102

103

104

100

101

102

103

104

72

2

100

101

102

103

104

100

101

102

103

104

322.3

100

101

102

103

104

100

101

102

103

104

831.3

100

101

102

103

104

100

101

102

103

104

3310

100

101

102

103

104

100

101

102

103

104

713.7

100

101

102

103

104

100

101

102

103

104

5817.3

100

101

102

103

104

100

101

102

103

104

77.411.4

100

101

102

103

104

100

101

102

103

104

5821

C

CD

11b

Her2 4T1

Gr1

CD11c

Her2 4T1

Single tumor Double tumor

CD45+

TAMs

100

101

102

103

104

0

20

40

60

100

101

102

103

104

0

30

60

90

120

298

100

101

102

103

104

0

10

20

30

40

50

25

75

100

101

102

103

104

0

10

20

30

40

50

29

71

100

101

102

103

104

0

5

10

1556 44

100

101

102

103

104

0

5

10

15

20

25 61 39

100

101

102

103

104

0

5

10

15

20

25

62 38

4

0

20

40

60

80

100

101

102

103

10 100

101

102

103

104

0

20

40

60

80

100

100

101

102

103

104

0

10

20

30

40

793

2674

0

20

40

60

100

101

102

103

104 10

010

110

210

310

4

0

10

20

30

40

51 49

100

101

102

103

104

0

20

40

60

36 64694

100

101

102

103

104

0

50

100

150

5 95

100

101

102

103

104

0

20

40

60

80

4 96 54 46

Primary

(s.c.)

CD11c

Gr1

Primary

(s.c.)

CD

11b

Gr1

CD11c

CD45+

TAMs

Her2 B16 Her2 B16

Single tumor Double tumor

%C

ells

in C

D45

+ N.S.N.S.

Her2 4T1 Double0

20

30

40

10

Gr1low MDSC Gr1hi MDSC

***

*

Spleen

%C

ells

in C

D45

+ N.S.

N.S.

Her2 B16 Double0

4

6

8

2

Gr1low MDSC Gr1hi MDSC

Spleen

Figure 3. Tumor type dictates the composition of myeloid cell subsets within tumors. A, myeloid cells within subcutaneous and mammary gland oftransgenic Her2 tumors (n ¼ 4–7); within primary subcutaneous and spontaneous peritoneal metastatic 4T1 tumors (n ¼ 3); and within subcutaneousB16 tumors or lung and kidney metastases (n ¼ 3–4). B, left, analysis of myeloid subsets within tumors from Balb/c mice with single (Her2 or4T1; n ¼ 5) or double (Her2 and 4T1 on opposite flanks; n ¼ 8) tumors. Right, frequency of myeloid cells in the spleens of tumor-bearing mice. C, left,analysis of myeloid cell subsets within tumors from F1 (Balb/c � C57BL/6) mice with single (Her2 or B16; n ¼ 3) or double (Her2 and B16 onopposite flanks; n ¼ 3) tumors. Right, frequency of myeloid cells in the spleens of tumor-bearing mice. Numbers indicate percentage of cells for eachgate or region. N.S., not statistically significant; �, P < 0.05; ��, P < 0.001.






4T1 tumorswith that ofmyeloid cells originated in other tumormodels. Hierarchical clustering indicated that TAMs, TANs,and splenic MDSC populations clustered separately, confirm-ing differences between tumor and splenic myeloid cells.Notably, similar profiles were obtained from equivalent popu-lations originated from different tumor types (Fig. 4C), sup-porting that transcriptomic analysis may illuminate universalfeatures of tumor myeloid cells. Indeed, this analysis identifiedseveral genes that highlight the protumorigenic roles of tumor-infiltrating myeloid cells. For example, genes involved inextracellular matrix remodeling (Mmp12, Mmp13, Mmp14,Adam8, and Tgm2), immunomodulation (Arg1, Nos2, Cd274,Ptgs2, and Spp1), and hypoxia regulation (Vegfa, Hif1a, andSlc2a1) were markedly upregulated in TAMs and TANs com-pared with splenic MDSCs. Interestingly, expression of thesegenes was similar between CD11b�CD103þ tumor DCs andDCs from tumor-free mice. This result suggests that hypoxicresponses are not uniform within the samemicroenvironmentand may be tightly regulated among tumor-infiltrating leuko-cytes. Immunosuppression and tissue remodeling are criticalfunctions for promoting tumor progression, and hypoxia,through hypoxia-inducible factor-1a (Hif-1a; ref. 18), canpromote transcriptional changes that facilitate the selection

of aggressive tumor cells. Expression analysis of Hif-1a–reg-ulated hypoxia-associated genes (19) revealed that these genesare indeed enriched in TAMs and TANs compared with splenicMDSCs and steady-state populations, indicating reprogram-ming in the tumor environment (Fig. 4D and SupplementaryTable S2). Hypoxia and Hif-1a have been implicated in pro-longed survival and reduced oxidative burst in neutrophils(20, 21), and as suchmay also contribute to the accumulation ofthese short-lived cells.

Neutrophils in the tumor environment areprotumorigenic and exhibit an altered functional profile

Our finding that the myeloid cell landscape of each tumortype is distinctive raises the possibility that different tumormicroenvironments impart unique functional attributes in themyeloid cells therein. Although TAMs in different tumors havebeen extensively characterized by functional and transcriptionalprofiling approaches (22–26), our understanding of TANs islimited. Recently, Fridlender and colleagues compared steady-state neutrophils to splenic MDSCs and TANs in a mesotheli-oma model by transcriptional profiling and identified strikingdifferences (27). However, this study focused only on a singletumor and it is not known if such characteristics of TANs are

1 100.1Fold change

(4T1 CD11c+ TAM vs Gr1low MDSC)

263

152

1

0.2

10

Fold

cha

nge

(4T1

CD

11c– T

AM

vs

Gr1

low M

DS

C)A

1 100.2Fold change (4T1 TAN vs Gr1hi MDSC)

100

10–1

10–2

10–3

10–4

10–5

647 20110–6

10–7

KdrCar9Pdk1Cdh1Ccnd1PdgfbIgf2Igfbp2TgfaSerpine1Flt1Cxcl12Mmp2MetCdkn1aPfkfb3Egln3Nos2Slc2a1Arg1VegfaHif1aHk2Hk1Gys1LdhaPgk1Pkm2Eno1AldoaMmp9Slc2a3Kdm5bCcng2Cxcr4PlaurCrebbpCrebbp

B

C

Gr1

low M

DS

C 4

T1G

r1hi M

DS

C 4

T1C

D11

c– TA

M 4

T1C

D11

c– TA

M B

16C

D11

c+ TA

M B

16C

D11

c+ TA

M 4

T1C

D11

c+ TA

M H

er2

TAN

4T1

TAN

B16

TAN

Her

2

CD

11b– C

D10

3+ DC

SI

Ly6C

+ PD

CA

1+ DC

BM

CD

4+ DC

Spl

CD

24+ S

irpa– D

C B

MC

D24

– Sirp

a+ DC

BM

DC

B16

Tm

r

F4/8

0hi M

ac P

C

Neu

tro B

6N

eutro

Bal

bcG

r1hi M

DS

C 4

T1G

r1lo

w M

DS

C 4

T1TA

N 4

T1TA

N B

16TA

N H

er2

CD

11c– T

AM

4T1

CD

11c– T

AM

B16

CD

11c+ T

AM

B16

CD

11c+ T

AM

4T1

CD

11c+ T

AM

Her

2

D

Expression (log2)3.87 13.68

Expression (log2)4.74 13.54

Mmp14

Mmp12Spp1

Cd274

Ptps2

Arg1Tgm2

VegfaNos2

Adam8Slc2a1

Hif1a

Arg1

Cd274

Tgm2

VegfaSlc2a1

t tes

t P v

alue

(log

10)

Abca1Cxcl16

Serpinb9Mgl2

Havcr2

Cxcl10

Il1a

Niacr1Arg1Egln3

Cstb Tnfrsf26

Hypoxia-associated genes

Figure 4. Tumor-infiltrating myeloid cells express immunomodulatory genes. A, comparison of gene expression profiles of CD11cþ and CD11c� TAMs toGr1low MDSCs from 4T1 tumor-bearing mice based on 2-fold change (fold-change vs. fold-change plot). B, comparison of gene expression profilesof TANs and Gr1hi MDSCs from 4T1 tumor-bearing mice (fold-change vs. t test P value plot). C, heatmap generated by hierarchical clustering usingprobes identified in A and B for all tumor myeloid subsets. D, heatmap showing the expression of hypoxia-related genes by tumor-associated andsteady-state myeloid cells. Selected genes associated with the populations are indicated on the plots.

Elpek et al.





common across different tumors. By comparing TANs from 4T1,B16, and Her2 tumors to Gr1hi MDSCs from 4T1 tumor-bearingmice and neutrophils from the bone marrow of na€�ve mice (Fig.4), we identified a conserved expression pattern of immuno-modulatory and hypoxia-related genes across these tumors. Wealso compared genes highly expressed in both Gr1hi MDSCs andTANs from 4T1 tumors to their expression in steady-stateneutrophils (Fig. 5A, left) and found 214 probes selectivelyenriched in 4T1-associated neutrophils. Interestingly, many ofthese genes were similarly enriched in TANs from B16 and Her2tumors compared with steady-state neutrophils (170 of 214passed CV < 0.5 filter; Fig. 5A, right). Heatmap analysis alsoindicated that these probes were highly expressed by all TANsand Gr1hi MDSCs compared with steady-state neutrophils (Fig.5B and Supplementary Table S3). Importantly, these genes areinvolved in pathways critical for tumor growth and immuneresponses, such as iNOS, IL-10, IL-6, and peroxisome prolifera-

tor-activated receptor (PPAR) signaling (Fig. 5C). Collectively,our data support that tumor-associated myeloid cells exhibit aprotumorigenic transcriptional program.

To further interrogate molecular programs in TANs acrossdifferent tumors, a side-by-side comparison of TANs from4T1, Her2, and B16 tumors was performed. We identified 43probes enriched in 4T1 TANs, 61 in Her2 TANs and 93 in B16TANs (Fig. 6A). Genes involved in cell trafficking or differ-entiation, such as Cxcr1, Foxo3, and Slc28a2, were enriched2- to 5-fold in 4T1 TANs and may regulate the markedaccumulation of TANs in this cancer type (28). Given thatour data point to a direct, tumor cell–specific involvement inthe recruitment and expansion of TANs, we analyzed theexpression levels of cytokines, cytokine receptors, chemo-kines, and chemokine receptors by TANs, as these mayaccount for the tumor–myeloid cell cross-talk regulatingcell infiltration (Fig. 6B and Supplementary Table S4).

A

1 100.1

Fold change (4T1 TAN vs. naïve neutrophil)

214

1

0.1

10

Fo

ld c

ha

ng

e

(4T

1 G

r1h

i MD

SC

vs. n

aïv

e n

eu

tro

ph

il)

1 100.1

Fold change (Her2 TAN vs. naïve neutrophil)

170

1

0.1

10

Fo

ld c

ha

ng

e

(B1

6 T

AN

vs. n

aïv

e n

eu

tro

ph

il)

B

0 1 2 3 4 5 6 7 8 9

IL10 signaling

Hepatic fibrosis/hepatic stellate

cell activation

LXR/RXR activation

Hepatic cholestasis

iNOS signaling

VDR/RXR activation

IL6 signaling

PPAR signaling

AHR signaling

Role of macrophages/fibroblasts/

endothelial cells in RA

Glucocorticoid receptor signaling

MIF regulation of innate immunity

Eumelanin biosynthesis

LPS/IL1–mediated inhibition

of RXR function

Acute phase response signaling

Molecular mechanisms of cancer

Production of NO and ROS

in macrophagesIL12 signaling and production

in macrophages

HGF signaling

PPARα/RXRα activation

Ingenuity canonical pathway

Top 20 –log10

(P )

Gr1

hi M

DS

C 4

T1

TA

N 4

T1

TA

N B

16

TA

N H

er2

Ne

utr

o B

6N

eu

tro

Ba

lb

C

Expression (log2)

4.23 13.73

Figure 5. Expression ofprotumorigenic genes by TANs. A,comparison of gene expressionprofiles of TANs and Gr1hi MDSCsfrom 4T1 tumor-bearing mice tonaïve neutrophils (left, fold-changevs. fold-change, 2-fold change).Expression of these probes wasfurther analyzed in fold-changeversus fold-change plotscomparing TANs from B16 andHer2 tumors with naïve neutrophils(right); 170 of the 214 probespassed the CV < 0.5 filter. B,heatmap showing the expressionof the 170 probes identified in Aacross neutrophils. C, top 20canonical pathways enriched forTANs and Gr1hi MDSCs comparedwith naïve neutrophils analyzed byIngenuity Pathway Analysis. ROS,reactive oxygen species.






Strikingly, we identified different expression profiles for eachTAN population, which may reflect the differences observedin the frequencies of these populations in each tumor. Inparticular, TANs from 4T1 tumors expressed significantlygreater levels of Ltb4r1 (BLT1), a leukotriene B4 chemotacticfactor for neutrophils (29, 30). Interestingly, we also foundenrichment of genes in the anti-inflammatory PPARa sig-naling pathway (Fig. 5D), which is also a target for leuko-

triene B4. We further analyzed the expression of PPARatarget genes associated with inflammation (31) in neutrophilsubsets (Fig. 6C and Supplementary Table S5). Althoughmany of these genes are expressed at higher levels in TANscompared with Gr1hi MDSCs and steady-state neutrophils,neutrophils from 4T1-bearing mice expressed higher levelsof Stat1, Stat2, and Stat3, among other genes. Stat3 is amediator of CXCR2 expression in neutrophils upon G-CSF

Ccl12Ccl2Ccl5Ccl4Ccl7Ccl9Ccl22Ccl8Ccl3Ccl6Ccl27aCxcl2Cxcl3Ccl1Cx3cl1Ccl17Cxcl1Cxcl14Cxcl16Xcl1Cxcl10Cxcl9Ccl19Ccl11Cxcl17Ccl25Cxcl11Ccrl2

BA

Ifna1Ifna9IfnzFlt3lIfna7Il1bIfna5Il17aIl31Il13Il25Il16LtbIfna2MifIl12aIl15Il18Il1f9Il17cCsf3Ifna14Il23aTn fIl12bIl1aTgfb1

Il2raIl2rbIl7rIl2rgIl10rbIl12rb1Il9rIl17rbIl11ra1Il6raIl17raIl13ra1Il28raIl15raIl27ra

Ccr2Ccr5

Ccr1Ccr10Ccr7Xcr1Cxcr3Cxcr4Cxcr1Cxcr2Ltb4r1

Fold change (4T1 vs. B16 TANs)

Fold

change (

4T

1 v

s. H

er2

TA

Ns)

TA

N 4

T1

TA

N B

16

TA

N H

er2

TA

N 4

T1

TA

N B

16

TA

N H

er2

TA

N 4

T1

TA

N B

16

TA

N H

er2

Expression (log2)

5.73 13.37

Expression (log2)

5.53 11.34

Expression (log2)

5.67 13.68

Expression (log2)

5.91 12.47

B6

Balb/cMDSC

4T1TAN

B16TAN

Her2TAN

C

TA

N B

16

TA

N 4

T1

TA

N H

er2

1 100.1

1

0.1

10

Ltb4r1

Slc28a2Foxo3Cxcr1

6143

93

Ccgn2Gbp1

Irgm2 Irf47

Irf1

Plxdc2

ArsbItgb5

Axl

Ccr2

Ccl8Trim12a

Cd36

Ly6a

Thbs1

Ccl7

LifrEmr1Vcam1Il18Pla1aCcl2Il1rnMt2Mt1Ccl3Cxcl10NfkbiaIcam1Ifi47 7CebpbIl1b bBirc3Irgm2Stat2Stat3Il1rapSteap4Stat1Il6raOrm2Lcn2Traf2Nfkb1

Cx3cr1

Gr1

hi M

DS

C 4

T1

TA

N 4

T1

TA

N B

16

TA

N H

er2

Neutr

o B

6N

eutr

o B

alb

D

Expression (log2)

5.77 13.87

–0.2

0

0.2

0.4

–0.20

0.20.4

−0.4

−0.2

0

0.2

0.4

PC2 (14.6%)

PC1

(66.4%)

PC3

(5.9%)

Cytokine ChemokineCytokine

receptor

Chemokine

receptor

Figure 6. Tumor type–specific transcriptional profiles of TANs. A, fold-change versus fold-change plot comparing 4T1 TANswith TANs inHer2 andB16, basedon 2-fold change. Selected genes associated with the populations are indicated on the plots. B, chemokine, chemokine receptor, cytokine, and cytokinereceptor expression profiles of TANs. C, PPARa-associated gene expression profiles of neutrophils. D, principal component analysis of neutrophilpopulations.

Elpek et al.





stimulation (32). In agreement with this, G-CSF is highlyexpressed by 4T1 cells (Supplementary Fig. S2A) and Cxcr2 isexpressed at higher levels by 4T1 TANs (Fig. 6B). Overall,these signaling pathways may incite the marked accumula-tion of TANs and Gr1hi MDSCs in the 4T1 model. Therelationship between neutrophil populations from differenttumors was also confirmed by the principal componentanalysis to highlight functional differences. Neutrophils fromna€�ve mice clustered separately and more closely to MDSCson PC2 (Fig. 6D). On the other hand, while TANs clusteredseparately from neutrophils and MDSCs, variability on PC1reflected tumor-to-tumor differences among TANs.

Haptoglobin levels are associated with myeloid cellaccumulationDuring acute inflammatory responses, neutrophils exocy-

tose effector molecules stored in specific granules (33).Recently, Fridlender and colleagues showed that some ofthese molecules are actually downregulated in TANs com-pared with those in steady-state neutrophils and Gr1hi

MDSCs (27). To understand how this critical function isinfluenced by different tumors, we compared the expressionof these effector molecules in steady-state neutrophils, in

splenic Gr1hi MDSCs from 4T1 tumor-bearing mice, and inTANs from 4T1, Her2, and B16 tumors. Heatmap analysisand hierarchical clustering indicated that all TANs sharedan altered expression profile compared with na€�ve granulo-cytes (Fig. 7A and Supplementary Table S6). Interestingly,Gr1hi MDSCs expressed the majority of the effector mole-cules. Among the molecules associated with neutrophilgranules is the acute-phase responder haptoglobin (Hp; ref.34). Haptoglobin is known to be produced in the liver;however, recent studies indicate that it is also expressedby neutrophils and packaged into secretory granules (35).Analysis of haptoglobin expression by tumor-associatedmyeloid cells revealed that haptoglobin is not expressed byDCs or CD11cþ TAMs in tumors, and its level is moderate inCD11c� TAMs and in TANs from B16 and Her2 tumors (Fig.7B). However, haptoglobin expression is high in na€�ve gran-ulocytes, Gr1hi MDSCs, and TANs from 4T1 tumors. Accord-ingly, serum haptoglobin levels were elevated in mice bear-ing advanced 4T1 tumors but not in Her2 or B16 tumor-bearing hosts (Fig. 7C). Indeed, serum haptoglobin levelscorrelated with the frequency of Gr1hi MDSCs in the blood of4T1 tumor-bearing mice (Fig. 7D). This select pattern sug-gests that haptoglobin expression may be associated with

B

A

C

Mpo

Ela

ne

Cts

gP

rtn3

Bpi

Slc

11a1

Fcgr3

Mm

p2

Cd14

Lyz2

Lyz1

Cyba

Ltf

Cam

pM

mp25

Itgb2

Itgam

Mm

p8

Mm

p9

Fpr1

Lcn2

Hp

Mean e

xpre

ssio

n v

alu

e

0

200

400

600

800

S M L S M L

4T1 Her2Naive

L

B16

Seru

m H

P c

oncentr

ation (

μg/m

L)

Gr1hi MDSC 4T1

TAN 4T1TAN B16

TAN Her2

Neutro B6Neutro Balb

Neutr

o B

6

Neutr

o B

alb

0

2,000

4,000

6,000

8,000

Gr1

hi M

DS

C 4

T1

TA

N 4

T1

TA

N H

er2

TA

N B

16

Gr1

low M

DS

C 4

T1

CD

11c

- TA

M 4

T1

CD

11c

- TA

M B

16

CD

11c

+ T

AM

4T

1

CD

11c

+ T

AM

B16

CD

11c

+ T

AM

Her2

DC

B16

***

Exp

ressio

n (lo

g2 )

5.4

31

3.9

0

D

%G

r1h

i in b

lood

0

50

100

100 200 300 400 5000

Serum HP concentration (μg/mL)

R 2 = 0.71

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

Seru

m H

P c

oncentr

ation (

μg/m

L)

Healthy

donor

Breast cancer

patient

E**

Figure 7. Altered expression profileof secretory granule molecules inTANs. A, expression of granulareffector molecules by neutrophils.B, mean expression value ofhaptoglobin (Hp, HP) acrossneutrophils and tumor-associatedmyeloid cells. C, serumhaptoglobin levels in mice withsmall (S), medium (M), large (L)tumors and naïve mice. D,correlation between the frequencyof Gr1hi cells in blood and theserum haptoglobin concentration.E, haptoglobin levels in serum fromhealthy donors (n¼ 4) and patientswith breast cancer (n ¼ 10).�, P < 0.05; ��, P < 0.01.






the considerable myelopoiesis observed in the 4T1 setting(Supplementary Fig. S1E). Although further studies arerequired to confirm a direct relationship between neutro-phils and haptoglobin levels, we detected elevated serumhaptoglobin in patients with breast cancer (Fig. 7E), con-sistent with reports for patients with glioblastoma andovarian cancer (36, 37). Thus, haptoglobin may serve as adiagnostic marker indicative of cancers with myeloid cellaccumulation.

Tumors require a sustained influx of myeloid cells tosupport angiogenesis, remodel peritumoral stroma, andsuppress antitumor immunity. Therefore, tumor-associatedmyeloid cells represent potential targets to improve immu-notherapy as well as chemotherapy against cancer. Althoughthere are limited strategies available, identifying novel tar-gets for elimination or functional differentiation of thesecells requires a deeper understanding of their development,trafficking, and phenotypic and functional characteristics.Moreover, to ascertain whether a broader therapeutic appli-cation is feasible, it is critical to determine whether keymolecular attributes are conserved among specific myeloidpopulations from tumors of different origins. A betterunderstanding of the functional differences among discretemyeloid subsets in tumor lesions, as well as secondarylymphoid organs, may lay the groundwork for developmentof novel therapeutics. In this regard, our study providesfundamental insights into functional differences of tumor-infiltrating myeloid cells, and identifies potential molecularpathways that could be investigated for therapies aimed atreducing infiltration and/or reprogramming function oftumor myeloid cells.

Disclosure of Potential Conflicts of InterestK.W.Wucherpfennig reports serving as a consultant/advisory board member

for Novartis and Ipsum. No potential conflicts of interest were disclosed by theother authors.

Authors' ContributionsConception and design: K.G. Elpek, S.J. TurleyDevelopment of methodology: K.G. Elpek, K.W. Wucherpfennig, S.J. TurleyAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): K.G. Elpek, V. Cremasco, C.J. Harvey, D.R. Goldstein,P.A. Monach, S.J. TurleyAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis):K.G. Elpek, V. Cremasco, C.J. Harvey, K.W.Wucherp-fennig, P.A. Monach, S.J. TurleyWriting, review, and/or revision of the manuscript: K.G. Elpek, V.Cremasco, D.R. Goldstein, P.A. Monach, S.J. TurleyAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): H. Shen

AcknowledgmentsThe authors thank the Immunological Genome Project team for technical

help and discussions; Drs. Dobrin Draganov and Glenn Dranoff for the Her2 cellline; and Dr. Ronald DePinho for PDA mice.

Grant SupportThis study was supported by research funding from NIH T32 grant to K.G.

Elpek, NIH T32 CA 070083-15 and Postdoctoral Fellowship from the CancerResearch Institute to V. Cremasco, NIH grants (AG028082, AG033049, AI098108,and AI101918) and an Established Investigator Award (0940006N) from Amer-ican Heart Association to D.R. Goldstein, NIH grants (ROI DK074500, PO1AI045757, and R21 CA182598), American Cancer Society Research Scholar Grant,and Claudia Adams Barr Award for Innovative Cancer Research to S.J. Turley.

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

Received November 30, 2013; revised February 19, 2014; accepted March 12,2014; published OnlineFirst March 31, 2014.

References1. Bussolati B, Grange C, Camussi G. Tumor exploits alternative strat-

egies to achieve vascularization. FASEB J 2011;25:2874–82.2. Gabrilovich DI, Ostrand-Rosenberg S, Bronte V. Coordinated regula-

tion of myeloid cells by tumours. Nat Rev Immunol 2012;12:253–68.3. Rahir G, Moser M. Tumor microenvironment and lymphocyte infiltra-

tion. Cancer Immunol Immunother 2012;61:751–9.4. Tlsty TD, Coussens LM. Tumor stroma and regulation of cancer

development. Annu Rev Pathol 2006;1:119–50.5. Marigo I, Bosio E, Solito S, Mesa C, Fernandez A, Dolcetti L, et al.

Tumor-induced tolerance and immune suppression depend on the C/EBPbeta transcription factor. Immunity 2010;32:790–802.

6. Cortez-Retamozo V, Etzrodt M, Newton A, Rauch PJ, Chudnovskiy A,Berger C, et al. Origins of tumor-associated macrophages and neu-trophils. Proc Natl Acad Sci U S A 2012;109:2491–6.

7. Le HK, Graham L, Cha E, Morales JK, Manjili MH, Bear HD. Gemci-tabine directly inhibits myeloid derived suppressor cells in BALB/cmice bearing 4T1 mammary carcinoma and augments expansion of Tcells from tumor-bearing mice. Int Immunopharmacol 2009;9:900–9.

8. Vincent J, Mignot G, Chalmin F, Ladoire S, Bruchard M, Chevriaux A,et al. 5-Fluorouracil selectively kills tumor-associatedmyeloid-derivedsuppressor cells resulting in enhanced T cell-dependent antitumorimmunity. Cancer Res 2010;70:3052–61.

9. Ko JS, Zea AH, Rini BI, Ireland JL, Elson P, Cohen P, et al. Sunitinibmediates reversal of myeloid-derived suppressor cell accumulation inrenal cell carcinoma patients. Clin Cancer Res 2009;15:2148–57.

10. DeNardo DG, Brennan DJ, Rexhepaj E, Ruffell B, Shiao SL, MaddenSF, et al. Leukocyte complexity predicts breast cancer survival andfunctionally regulates response to chemotherapy. Cancer Discov2011;1:54–67.

11. Ying H, Elpek KG, Vinjamoori A, Zimmerman SM, ChuGC, Yan H, et al.PTEN is a major tumor suppressor in pancreatic ductal adenocarci-noma and regulates an NF-kappaB-cytokine network. Cancer Discov2011;1:158–69.

12. Gautier EL, Shay T, Miller J, Greter M, Jakubzick C, Ivanov S, et al.Gene-expression profiles and transcriptional regulatory pathways thatunderlie the identity and diversity of mouse tissue macrophages. NatImmunol 2010;13:1118–28.

13. Miller JC, Brown BD, Shay T, Gautier EL, Jojic V, Cohain A, et al.Deciphering the transcriptional network of the dendritic cell lineage.Nat Immunol 2012;13:888–99.

14. Heng TS, Painter MW. The Immunological Genome Project: networksof gene expression in immune cells. Nat Immunol 2008;9:1091–4.

15. Cheung R, Shen F, Phillips JH, McGeachy MJ, Cua DJ, Heyworth PG,et al. Activation ofMDL-1 (CLEC5A) on immaturemyeloid cells triggerslethal shock in mice. J Clin Invest 2011;121:4446–61.

16. Toh B, Wang X, Keeble J, Sim WJ, Khoo K, Wong WC, et al. Mesen-chymal transition and dissemination of cancer cells is driven bymyeloid-derived suppressor cells infiltrating the primary tumor. PLoSBiol 2011;9:e1001162.

17. Waight JD, Hu Q, Miller A, Liu S, Abrams SI. Tumor-derived G-CSFfacilitates neoplastic growth through a granulocytic myeloid-derived suppressor cell-dependent mechanism. PLoS ONE 2011;6:e27690.

18. Shay JE, Celeste Simon M. Hypoxia-inducible factors: crosstalkbetween inflammation and metabolism. Semin Cell Dev Biol 2012;23:389–94.

19. Favaro E, Lord S, Harris AL, Buffa FM. Gene expression and hypoxia inbreast cancer. Genome Med 2011;3:55.


Elpek et al.




20. McGovern NN, Cowburn AS, Porter L, Walmsley SR, Summers C,Thompson AA, et al. Hypoxia selectively inhibits respiratory burstactivity and killing of Staphylococcus aureus in human neutrophils.J Immunol 2011;186:453–63.

21. Walmsley SR, Print C, Farahi N, Peyssonnaux C, Johnson RS, CramerT, et al. Hypoxia-induced neutrophil survival is mediated by HIF-1alpha–dependent NF-kappaB activity. J Exp Med 2005;201:105–15.

22. Duff MD, Mestre J, Maddali S, Yan ZP, Stapleton P, Daly JM. Analysisof gene expression in the tumor-associated macrophage. J Surg Res2007;142:119–28.

23. Ojalvo LS, King W, Cox D, Pollard JW. High-density gene expressionanalysis of tumor-associated macrophages from mouse mammarytumors. Am J Pathol 2009;174:1048–64.

24. Ojalvo LS, Whittaker CA, Condeelis JS, Pollard JW. Gene expressionanalysis of macrophages that facilitate tumor invasion supports a rolefor Wnt-signaling in mediating their activity in primary mammarytumors. J Immunol 2010;184:702–12.

25. Pucci F, Venneri MA, Biziato D, Nonis A, Moi D, Sica A, et al. Adistinguishing gene signature shared by tumor-infiltrating Tie2-expressing monocytes, blood "resident" monocytes, and embryonicmacrophages suggests common functions and developmental rela-tionships. Blood 2009;114:901–14.

26. Stearman RS, Dwyer-Nield L,GradyMC,Malkinson AM,GeraciMW. Amacrophagegeneexpression signature definesafield effect in the lungtumor microenvironment. Cancer Res 2008;68:34–43.

27. Fridlender ZG, Sun J, Mishalian I, Singhal S, Cheng G, Kapoor V, et al.Transcriptomic analysis comparing tumor-associated neutrophils withgranulocyticmyeloid-derived suppressor cells andnormal neutrophils.PLoS ONE 2012;7:e31524.

28. Turrel-Davin F, Tournadre A, Pachot A, Arnaud B, Cazalis MA, MouginB, et al. FoxO3a involved in neutrophil and T cell survival is over-

expressed in rheumatoid blood and synovial tissue. Ann Rheum Dis2010;69:755–60.

29. Kim ND, Chou RC, Seung E, Tager AM, Luster AD. A unique require-ment for the leukotriene B4 receptor BLT1 for neutrophil recruitment ininflammatory arthritis. J Exp Med 2006;203:829–35.

30. Saiwai H, Ohkawa Y, Yamada H, Kumamaru H, Harada A, Okano H,et al. The LTB4-BLT1 axis mediates neutrophil infiltration and sec-ondary injury in experimental spinal cord injury. Am J Pathol2010;176:2352–66.

31. Rakhshandehroo M, Knoch B, Muller M, Kersten S. Peroxisome pro-liferator-activated receptor alpha target genes. PPAR Res 2010;2010:pii:612089.

32. Nguyen-Jackson H, Panopoulos AD, Zhang H, Li HS, Watowich SS.STAT3 controls the neutrophil migratory response to CXCR2 ligandsby direct activation of G-CSF–induced CXCR2 expression and viamodulation of CXCR2 signal transduction. Blood 2010;115:3354–63.

33. Borregaard N. Neutrophils, from marrow to microbes. Immunity2010;33:657–70.

34. Wang Y, Kinzie E, Berger FG, Lim SK, Baumann H. Haptoglobin, aninflammation-inducible plasma protein. Redox Rep 2001;6:379–85.

35. Theilgaard-Monch K, Jacobsen LC, Nielsen MJ, Rasmussen T, UdbyL, Gharib M, et al. Haptoglobin is synthesized during granulocytedifferentiation, stored in specific granules, and released by neutrophilsin response to activation. Blood 2006;108:353–61.

36. Kumar DM, Thota B, Shinde SV, Prasanna KV, Hegde AS, ArivazhaganA, et al. Proteomic identification of haptoglobin alpha2 as a glioblas-toma serumbiomarker: implications in cancer cell migration and tumorgrowth. J Proteome Res 2010;9:5557–67.

37. Zhao C, Annamalai L, Guo C, Kothandaraman N, Koh SC, Zhang H,et al. Circulating haptoglobin is an independent prognostic factor in thesera of patients with epithelial ovarian cancer. Neoplasia 2007;9:1–7.






2014;2:655-667. Published OnlineFirst March 31, 2014.Cancer Immunol Res Kutlu G. Elpek, Viviana Cremasco, Hua Shen, et al. and Functional Diversity of Infiltrating Myeloid CellsThe Tumor Microenvironment Shapes Lineage, Transcriptional,

Updated version

10.1158/2326-6066.CIR-13-0209doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerimmunolres.aacrjournals.org/content/suppl/2014/04/03/2326-6066.CIR-13-0209.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerimmunolres.aacrjournals.org/content/2/7/655.full#ref-list-1

This article cites 37 articles, 14 of which you can access for free at:

Citing articles

http://cancerimmunolres.aacrjournals.org/content/2/7/655.full#related-urls

This article has been cited by 7 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerimmunolres.aacrjournals.org/content/2/7/655To request permission to re-use all or part of this article, use this link



http://cancerimmunolres.aacrjournals.org/lookup/doi/10.1158/2326-6066.CIR-13-0209

http://cancerimmunolres.aacrjournals.org/content/suppl/2014/04/03/2326-6066.CIR-13-0209.DC1

http://cancerimmunolres.aacrjournals.org/content/2/7/655.full#ref-list-1

http://cancerimmunolres.aacrjournals.org/content/2/7/655.full#related-urls

http://cancerimmunolres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerimmunolres.aacrjournals.org/content/2/7/655


the tumor microenvironment shapes lineage, transcriptional ...€¦ · daniel r. goldstein4,5, paul...

Documents