hypoxia-inducible factor-dependent signaling between triple ...hypoxia-inducible factor-dependent...

Hypoxia-inducible factor-dependent signaling betweentriple-negative breast cancer cells and mesenchymalstem cells promotes macrophage recruitmentPallavi Chaturvedia,b, Daniele M. Gilkesa,b, Naoharu Takanoa,b, and Gregg L. Semenzaa,b,c,1

aVascular Program, Institute for Cell Engineering, bMcKusick-Nathans Institute of Genetic Medicine, and cDepartments of Pediatrics, Medicine, Oncology,Radiation Oncology, and Biological Chemistry, Johns Hopkins University School of Medicine, Baltimore, MD 21205

Contributed by Gregg L. Semenza, April 14, 2014 (sent for review March 20, 2014)

Intratumoral hypoxia induces the recruitment of stromal cells, suchas macrophages and mesenchymal stem cells (MSCs), which stim-ulate invasion and metastasis by breast cancer cells (BCCs). Pro-duction of macrophage colony-stimulating factor 1 (CSF1) by BCCsis required for macrophage recruitment, but the mechanismsunderlying CSF1 expression have not been delineated. Triple-negative breast cancers have increased expression of genesregulated by hypoxia-inducible factors (HIFs). In this study, wedelineate two feed-forward signaling loops between humanMDA-MB-231 triple-negative BCCs and human MSCs that drivestromal cell recruitment to primary breast tumors. The first loop,in which BCCs secrete chemokine (C-X-C motif) ligand 16 (CXCL16)that binds to C-X-C chemokine receptor type 6 (CXCR6) on MSCsand MSCs secrete chemokine CXCL10 that binds to receptor CXCR3on BCCs, drives recruitment of MSCs. The second loop, in whichMSCs secrete chemokine (C-C motif) ligand 5 that binds to C-Cchemokine receptor type 5 on BCCs and BCCs secrete cytokineCSF1 that binds to the CSF1 receptor on MSCs, drives recruitmentof tumor-associated macrophages and myeloid-derived suppressorcells. These two signaling loops operate independent of eachother, but both are dependent on the transcriptional activity ofHIFs, with hypoxia serving as a pathophysiological signal that syn-ergizes with chemokine signals from MSCs to trigger CSF1 genetranscription in triple-negative BCCs.

HIF-1 | mammary fat pad | orthotopic implantation |lymph node metastasis | lung metastasis

Breast cancer metastasis transforms a local disease that iscured by surgical excision into a systemic disease that re-sponds poorly to available therapies and is the major cause ofpatient mortality (1). Although somatic mutations have beencataloged in hundreds of human breast cancers and many genesthat promote or suppress metastasis have been identified, theanalysis of genetic alterations cannot reliably distinguish meta-static from nonmetastatic cancers (1–3). Multiple stromal celltypes, including mesenchymal stem cells (MSCs) and tumor-associated macrophages (TAMs), are recruited to the tumor mi-croenvironment and promote metastasis (4, 5). In mouse models,MSCs produce chemokines, including chemokine (C-C motif) li-gand 5 (CCL5) and chemokine (C-X-C motif) ligand 10 (CXCL10),which bind to their cognate receptors, chemokine receptor type5 (CCR5) and C-X-C chemokine receptor type 3 (CXCR3),respectively, on breast cancer cells (BCCs) to stimulate invasionand metastasis (6–9). TAMs are abundant in breast cancerand outnumber the BCCs in some cases (10). The density ofTAMs in primary breast cancer biopsies is correlated with me-tastasis and patient mortality (11–13). In mouse models, mac-rophage colony-stimulating factor 1 (CSF1) and the chemokineCCL2 are secreted by BCCs and bind to their cognate receptors,CSF1 receptor (CSF1R) and CCR2, on TAMs, leading to theirrecruitment to the tumor microenvironment, where they produceEGF and other secreted proteins that promote invasion andmetastasis (14–19).

Intratumoral hypoxia is another major microenvironmentalfactor that is associated with invasion, metastasis, and patientmortality (20–22). Cancer cells respond to the hypoxic micro-environment through the activity of hypoxia-inducible factors(HIFs), which are heterodimeric transcription factors composedof an O2-regulated HIF-1α or HIF-2α subunit and a constitu-tively expressed HIF-1β subunit (23). In primary tumor biopsies,elevated HIF-1α or HIF-2α protein levels are associated with anincreased risk of metastasis and mortality that is independent ofbreast cancer grade or stage (24–28). HIF-1α, HIF-2α, or bothare required for the transcriptional activation of a battery ofhypoxia-inducible genes whose protein products are required fordiscrete steps in the process of breast cancer invasion and me-tastasis (29–34). High expression of HIF target genes is com-monly observed in triple-negative breast cancers (TNBCs), whichlack estrogen receptor, progesterone receptor, and human epi-dermal growth factor receptor 2 (HER2) expression and respondpoorly to chemotherapy (2).Both MSCs and TAMs are recruited to the hypoxic breast

tumor microenvironment (9, 35), although the underlying mech-anisms are not fully understood. In the present study, we hy-pothesized that the presence of MSCs in the primary breasttumor may facilitate TAM recruitment. Our studies of humanMDA-MB-231 TNBC cells in immunodeficient mice revealedthat HIFs regulate the hypoxia-induced expression of CXCL16in BCCs, which was required for MSC recruitment. CCL5→CCR5signaling between MSCs and BCCs was required for CSF1expression by BCCs, which was also induced by hypoxia. Ex-pression of CSF1 and CCR5 by BCCs was required for TAMrecruitment and BCC metastasis. HIF-dependent recruitment of

Significance

The recruitment of host stromal cells, such as macrophages andmesenchymal stem cells (MSCs), to the primary tumor is a crit-ical step toward cancer malignancy. We have identified signalsthat are exchanged between breast cancer cells (BCCs) andMSCs. This signaling increases the recruitment of both MSCsand macrophages to primary tumors and increases metastasisof BCCs to lymph nodes and lungs. Reduced oxygen levels(hypoxia) in breast cancers are associated with increased risk ofmetastasis and decreased patient survival. We show thathypoxia stimulates signaling between BCCs and MSCs due tothe activity of hypoxia-inducible factors (HIFs). Drugs that blockHIF activity prevent signaling and macrophage recruitment,which suggests that they may be useful additions to breastcancer therapy.

Author contributions: G.L.S. designed research; P.C., D.M.G., and N.T. performed research;P.C. and G.L.S. analyzed data; and P.C. and G.L.S. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental.

E2120–E2129 | PNAS | Published online May 5, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1406655111

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1406655111&domain=pdf&date_stamp=2014-05-12mailto:[email protected]://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1406655111

TAMs was also demonstrated after implantation of mouse 4T1TNBC cells into the mammary fat pad (MFP) of immunocom-petent mice. Taken together, these results delineate molecularmechanisms by which intratumoral hypoxia regulates the re-cruitment of MSCs and TAMs, and their interaction with TNBCs,to stimulate invasion and metastasis.

ResultsCXCL16 Expression by BCCs Stimulates MSC Recruitment. We previ-ously demonstrated that hypoxia-induced expression of placentalgrowth factor (PGF) by MDA-MB-231 BCCs provides a signalfor the recruitment of MSCs to primary breast tumors andstimulates MSCs to express CXCL10, which binds to CXCR3on BCCs (CXCL10MSC→CXCR3BCC) to stimulate invasion andmetastasis (9). We also showed that CXCR3 expression by BCCswas required for CXCL10 expression by MSCs, but the re-sponsible signal from BCCs to MSCs was not identified (9).CXCL16 expression by prostate cancer cells was shown to pro-mote MSC recruitment (36), and CXCL16 expression by BCCshas been reported (37). We cocultured GFP-expressing humanMDA-MB-231 BCCs with human MSCs at a 1:1 ratio for 48 hunder 20% or 1% O2, and flow cytometry was performed usingGFP and CD105 immunofluorescence to sort for BCCs andMSCs, respectively. RNA was isolated from the sorted cells andRT-quantitative real-time PCR (qPCR) revealed that CXCL16expression by BCCs was induced following coculture and thathypoxia further enhanced expression (Fig. 1A). The expressionof CXCR6, a receptor for CXCL16, was induced by hypoxia inBCCs, as previously reported (38), as well as in MSCs, but co-culture did not enhance its expression (Fig. 1B).We hypothesized that CXCL10MSC→CXCR3BCC signaling might

stimulate CXCL16BCC→CXCR6MSC reciprocal signaling. To testthis hypothesis, we analyzed BCC subclones stably transfected withshRNAs that inhibit CXCR3 expression (9). CXCL16 expression inCXCR3-deficient BCCs was not induced by coculture with MSCsor by hypoxia (Fig. 1C). We next added neutralizing antibody(NAb) against CXCL10 or control IgG to cocultures of MSCs andBCCs. CXCL16 mRNA expression was significantly decreased inthe presence of CXCL10 NAb (Fig. 1D).We also hypothesized that CXCL16BCC→CXCR6MSC signal-

ing might stimulate CXCL10MSC→CXCR3BCC reciprocal sig-naling. To test this hypothesis, we generated BCC subclones thatwere stably transfected with shRNAs that inhibit CXCL16 ex-pression (Fig. S1). CXCL10 expression in MSCs cocultured withCXCL16-deficient BCCs was significantly decreased and was notinduced by hypoxia (Fig. 1E).To investigate whether CXCL16 secretion from BCCs stimu-

lated the motility of MSCs, we isolated conditioned medium(CM) from CXCL16-deficient BCCs or BCCs expressing a non-targeting control (NTC) shRNA. CM from CXCL16-deficientBCCs induced significantly lessMSCmigration in aBoyden chamberassay, and the augmented effect of CM from hypoxic control BCCswas not observed with CXCL16-deficient BCCs (Fig. 1F). To de-termine whether CXCL16 secretion from BCCs promotes re-cruitment ofMSCs to the primary tumor, BCCs were orthotopicallyimplanted in the MFP of female SCID mice. When the tumorsreached a volume of 250 mm3, MSCs that were originally derivedfrom a male donor were injected via the tail vein. Primary tumorswere harvested 16 h later, and MSC recruitment was determinedby qPCR assay of genomic DNA using Y chromosome-specificprimers (9). CXCL16 deficiency in BCCs significantly decreasedthe recruitment of MSCs to the primary tumor (Fig. 1G). Takentogether, these results delineate CXCL16BCC→CXCR6MSC andCXCL10MSC→CXCR3BCC reciprocal signaling, which creates afeed-forward loop between BCCs and MSCs that drives MSC re-cruitment to the primary tumor (Fig. 1H).

CXCL16 Promotes Metastasis of BCCs to Lymph Nodes and Lungs.Because CXCL16BCC→CXCR6MSC signaling was requiredfor CXCL10MSC→CXCR3BCC signaling, which stimulates in-vasion and metastasis (9), we hypothesized that CXCL16 ex-pression was required for efficient metastasis. To test thishypothesis, control or CXCL16-deficient BCCs were ortho-topically implanted in the MFP of female SCID mice. CXCL16deficiency had no effect on primary tumor growth (Fig. 2A).However, mice bearing CXCL16-deficient tumors had signifi-cantly decreased numbers of circulating tumor cells (Fig. 2B),

0

0.5

1

1.5

2

2.5

3

BCCs MSCs BCCs+

CXC

R6

mR

NA

20% O2 1% O2

20% O2 1% O2

0 0.5

1 1.5

2 2.5

3 3.5

4

CXCL

16 m

RNA

IgG CXCL10 NAb

0 1 2 3 4 5 6 7

-MSCs + MSCs

CXCL

16 m

RNA

NTC shCR3-1 shCR3-3 NTC shCR3-1 shCR3-3

0 5

10 15 20 25 30 35 40 45

- MSCs +MSCs

CXC

L10

mR

NA

NTC shCX16-3shCX16-1NTC shCX16-3shCX16-1

A

D

E

**

##*

# #

**

## ##

*

**

## ##

0 0.5

1 1.5

2 2.5

3 3.5

4

Mig

ratio

n of

MSC

s

*

# #

NTC shCX16-1 shCX16-3

0

5

10

15

20

25

30

BCCs MSCs BCCs MSCs

20% O2 1% O2

coculture

B

C9

CXCL

16 m

RNA

F20% O2 1% O2

20% O2 1% O2

20% O2 1% O2

BCC MSC

CXCL10

CXCL16

CXCR3

CXCR6

MSChoming to BCC

8

G H

##

*

*

*

MSCs

0

0.2

0.4

0.6

0.8

1

1.2


* *

SR

Y c

opy

num

ber

##

*

(rel

ativ

e to

20%

NTC

)

Fig. 1. CXCL16 expression by MDA-MB-231 BCCs is induced by hypoxia andis required for MSC recruitment. (A) GFP-expressing BCCs were coculturedwith MSCs at 20% or 1% O2 for 48 h and then flow-sorted based on GFPfluorescence for BCCs and CD105 immunofluorescence for MSCs. CXCL16mRNA levels were determined in flow-sorted BCCs and MSCs from coculturesand in BCCs or MSCs cultured alone. RNA levels were normalized to resultsfor BCCs cultured alone at 20% O2. *P < 0.05 vs. 20%;

##P < 0.001 vs. BCCsalone. (B) BCCs, MSCs, and BCCs +MSCs were incubated at 20% or 1% O2 for48 h, and CXCR6 mRNA levels were determined and normalized to BCCsat 20%. *P < 0.05 vs. 20%. (C) MDA-MB-231 subclones expressing shRNAagainst CXCR3 (shCR3-1 and shCR3-3) or an NTC shRNA were cultured aloneor with MSCs and exposed to 20% or 1% O2 for 48 h. CXCL16 mRNA levelswere normalized to NTC − MSCs at 20%. *P < 0.05; **P < 0.001 vs. 20%;#P < 0.05 vs. NTC; ##P < 0.001 vs. NTC. (D) BCCs + MSCs were treated withCXCL10 NAb or IgG and exposed to 20% or 1% O2. CXCL16 mRNA levelswere normalized to IgG at 20%. **P < 0.001 vs. 20%; ##P < 0.001 vs. IgG. (E)MDA-MB-231 subclones expressing shRNA against CXCL16 (shCX16-1 andshCX16-3) or NTC shRNA were cultured alone or with MSCs and exposedto 20% or 1% O2. CXCL10 mRNA levels were normalized to NTC at 20%.*P < 0.05; **P < 0.001 vs. NTC − MSCs; ##P < 0.001 vs. NTC + MSCs. (F) Mi-gration of MSCs in response to CM from BCC subclones cultured at 20% or1% O2 was determined and normalized to NTC at 20%. *P < 0.01 vs. 20%NTC; #P < 0.01 vs. 1% NTC. (G) BCC subclones were implanted in the MFP,and MSC recruitment to the primary tumor was analyzed by qPCR for Ychromosome sequences and normalized to NTC. *P < 0.01 vs. NTC. (H) Bi-directional signaling between BCCs and MSCs generates a feed-forwardloop that stimulates MSC recruitment.

Chaturvedi et al. PNAS | Published online May 5, 2014 | E2121

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF1

metastatic cancer cells in the lungs by qPCR (Fig. 2C) andmetastatic foci in the lungs by histology (Fig. 2 D and E), andmetastatic cancer cells in the ipsilateral axillary lymph node (Fig.2 F and G). Taken together, these results demonstrate thatCXCL16 expression by BCCs plays a significant role in pro-moting breast cancer metastasis.

HIF-Dependent BCC–MSC Interactions Promote TAM and Myeloid-Derived Suppressor Cell Recruitment. To investigate whether in-teraction of MSCs with BCCs facilitates the recruitment ofTAMs to primary breast tumors, we cocultured MSCs with BCCsfor 48 h, followed by MFP implantation of 1 × 106 coculturedcells. Controls included mice implanted with 1 × 106 BCCs aloneor BCCs mixed with MSCs immediately before implantation(Fig. 3A). When primary tumors reached 450 mm3, they wereexcised and single-cell suspensions were analyzed by flow cy-tometry for the presence of TAMs, which express CSF1R andF4/80 on their cell surface, and myeloid-derived suppressor cells(MDSCs; TAM progenitors), which express CD11b and Ly6C.CSF1R+F4/80+ and CD11b+Ly6C+ cells were increased in tu-mors derived from BCC +MSC coculture compared with tumorsderived from BCCs alone or BCC + MSC coinjection (Fig. 3 Band C and Fig. S2 A and B).Recent studies have shown that exposure of macrophages to

levels of hypoxia that are comparable to what is observed intumors leads to induction of HIF-1α and HIF-2α, which, in turn,activate a broad array of genes with proangiogenic, proinvasive,and prometastatic functions (39, 40). TAM numbers are gener-ally higher in tumors containing high overall levels of hypoxia,as seen in primary human breast carcinomas (41) and various

animal tumors (42). To investigate whether HIF activity in BCCsplays a role in TAM recruitment, we used the MDA-MB-231double-knockdown (DKD) subclone, which is stably trans-fected with vectors encoding shRNAs that inhibit HIF-1α andHIF-2α, and the empty vector (EV) subclone (9). The BCCsubclones were injected into the MFP of female SCID mice,and tumors were harvested for flow cytometric analysis whenthey reached a volume of 450 mm3 (Fig. 3D). The recruitmentof TAMs and MDSCs to tumors derived from DKD cells wassignificantly decreased compared with EV tumors (Fig. 3 E andF and Fig. S2 C and D).We next examined the effect of HIFs on macrophage re-

cruitment by treating tumor-bearing mice with digoxin to inhibitHIF activity (30, 31). When tumors reached 200 mm3, mice re-ceived daily i.p. injections of either saline or digoxin (2 mg/kg)for 7 d, followed by tumor excision and flow cytometric analysis(Fig. 3G). Recruitment of TAMs and MDSCs was significantlydecreased in primary tumors of digoxin-treated mice (Fig. 3 Hand I and Fig. S2 E and F). Taken together, these results dem-onstrate that HIFs play an important role in recruiting TAMsand MDSCs to primary breast tumors.

Hypoxia and Coculture Induce HIF-Dependent CSF1 and CSF1RExpression. Previous studies demonstrated reciprocal paracrineinteractions in which BCCs secrete CSF1 and sense EGF,whereas TAMs sense CSF1 and secrete EGF (16–18), but thetrigger for CSF1 expression was not determined. CSF1 binds toits cognate receptor, CSF1R, which is expressed by TAMs;however, CSF1R is also expressed in >50% of breast tumors,

0

200

400

600

800

1000

1200

21 28 35 42 49

Tum

or v

olum

e (m

m)

Days

0

0.2

0.4

0.6

0.8

1

1.2


Circ

ulat

ing

tum

or c

ells

0

0.2

0.4

0.6

0.8

1

1.2

3



*

* * *

A B C

D ENTC shCX16-1 shCX16-3

0 1 2 3 4 5 6 7 8


** **

No.

of l

ung

foci

F

0 10 20 30 40 50 60 70 80 G


****

BCC

s (%

of L

N a

rea)NTC shCX16-1 shCX16-3

Met

asta

tic B

CC

’s in

lung

s

Fig. 2. CXCL16 promotes lung and lymph node metastasis of MDA-MB-231BCCs. BCC subclones were implanted in the MFP of SCID mice. (A) Primarytumor volumes were determined serially. (B) Circulating tumor cells in pe-ripheral blood were determined by qPCR using primers specific for human18S rRNA and normalized to NTC. *P < 0.01 vs. NTC. (C) Lung DNA wasanalyzed by qPCR with primers specific for human HK2 sequences and nor-malized to NTC. *P < 0.01 vs. NTC. (D) Photomicrographs of H&E-stained lungsections. (Scale bars: 2 mm.) (E) Lung sections were scored for metastatic fociper field. **P < 0.001 vs. NTC. (F) Lymph node sections were stained with anantibody that specifically recognizes human vimentin. (Scale bars: 0.5 mm.) (G)Stained area was quantified by image analysis. **P < 0.001 vs. NTC.

0 0.5

1 1.5

2 2.5

3 3.5

4 A

M

M

BCCs orBCCs+MSCs (Mix or Coculture )

*

**

BCCs + MSCsBCCs BCCs + MSCs(Mix) (Coculture)

0

2

4

6

8

10

12

**

BCCs + MSCsBCCs BCCs + MSCs(Mix) (Coculture)

CB

D

0 1 2 3 4 5 6 7 8 9

10

**

DKDEV

E

0

0.5

1

1.5

2

2.5

3

**

DKDEV

F

G H I

Saline Digoxin

**

0

1234567

0

1

2

3

4

**

Saline Digoxin

% C

SF1R

+ F4/

80+ c

ells

% C

SF1R

+ F4/

80+ c

ells

% C

SF1R

+ F4/

80+ c

ells

% C

D11b

Ly6

C c

ells

++

% C

D11b

Ly6

C c

ells

++

% C

D11b

Ly6

C c

ells

++

MFPEV or DKD

Tumor FACS

F4/80CSF1R

CD11bLy6C

MonocyticMDSCs

MFPMDA-231

MonocyticMDSCs

M

F4/80CSF1R

CD11bLy6C

FACSTumor

Saline Digoxinor

HIF

7 days(200 mm )3

(450 mm )3

(450 mm )3

MFP

Tumor FACS

F4/80CSF1R

CD11bLy6C

MonocyticMDSCs

Fig. 3. Macrophage recruitment is dependent on HIF activity. (A) 1 × 106

MDA-MB-231 BCCs alone, 1 × 106 BCCs mixed with 1 × 106 MSCs immediatelybefore injection, or 1 × 106 BCCs cocultured with 1 × 106 MSCs for 48 h wereimplanted in the MFP of SCID mice. The percentage of CSF1R+F4/80+ (B) andCD11b+Ly6C+ (C) cells in 450-mm3 primary tumors was determined by fluo-rescence-activated cell sorting (FACS). *P < 0.05; **P < 0.01 vs. BCCs. (D)MDA-MB-231 EV or DKD cells were implanted in the MFP of SCID mice. Thepercentage of CSF1R+F4/80+ (E) and CD11b+Ly6C+ (F) cells in 450-mm3 pri-mary tumors was determined by FACS. **P < 0.01 vs. EV. (G) Mice bearing200-mm3 MDA-MB-231 tumors were treated with saline or digoxin for 7 d.The percentage of CSF1R+F4/80+ (H) and CD11b+Ly6C+ (I) cells in primarytumors was determined. **P < 0.01 vs. saline.

E2122 | www.pnas.org/cgi/doi/10.1073/pnas.1406655111 Chaturvedi et al.

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF2www.pnas.org/cgi/doi/10.1073/pnas.1406655111

which suggests that CSF1 mediates both autocrine and paracrinesignaling (14–18). Coculture of GFP-expressing human MDA-MB-231 BCCs with human MSCs and flow sorting of GFP+

BCCs and CD105+ MSCs revealed that CSF1 mRNA expressionwas induced in BCCs (Fig. 4A), whereas CSF1R mRNA ex-pression was induced in both MSCs and BCCs (Fig. 4C). Ex-pression of both CSF1 and CSF1R was augmented in BCCssubjected to hypoxia (Fig. 4 A and C).To investigate the role of HIFs in these phenomena, MSCs

were cocultured with the DKD or EV subclone of BCCs. Ex-pression of CSF1 and CSF1R mRNA was significantly decreasedin DKD + MSC, compared with EV + MSC, cocultures at both20% and 1% O2 (Fig. 4 B and D). Levels of secreted CSF1protein were also significantly decreased in CM isolated fromDKD + MSC, compared with EV + MSC, cocultures (Fig. S3A).Pharmacological inhibition of HIF activity using acriflavine, whichblocks the dimerization of HIF-1α or HIF-2α with HIF-1β (43),inhibited coculture- and hypoxia-induced CSF1 and CSF1RmRNA expression (Fig. 4 E and F). Taken together, these resultsindicate that cross-talk between BCCs and MSCs is mediatedthrough HIF-dependent CSF1→CSF1R signaling.

CSF1 Promotes Macrophage Recruitment and Metastasis. We gen-erated MDA-MB-231 subclones that were stably transfected withvector encoding either of two different shRNAs targeting CSF1.Efficient knockdown of CSF1 mRNA expression and proteinsecretion was confirmed by RT-qPCR and ELISA, respectively(Fig. S3 B and C). Coculture-induced CSF1 secretion was alsoabrogated (Fig. S3C). CSF1 deficiency had no effect on primarytumor growth after MFP implantation (Fig. 5A). Mice bearingtumors derived from CSF1-deficient BCCs had significantly de-creased numbers of circulating tumor cells (Fig. 5B), metastaticcancer cells (Fig. 5C) and metastatic foci (Fig. 5 D and E) in thelungs, metastatic cancer cells in the ipsilateral axillary lymphnode (Fig. 5 F and G), and CSF1R+F4/80+ TAMs (Fig. 5H) andCD11b+Ly6C+ MDSCs (Fig. 5I) recruited to the primary tumor.These data demonstrate that CSF1 expression in BCCs playsa significant role in promoting macrophage recruitment and me-tastasis to both lymph nodes and lungs.

CCL5→CCR5 Signaling Between MSCs and BCCs Is Required for CSF1Expression. Previous studies showed that CCL5→CCR5 signalingstimulates breast cancer metastasis (6) and that coculture andhypoxia induce the expression of CCL5 in MSCs and its cognate

0 5

10 15 20 25 30 35 40 45

CSF

1 m

RN

A

EV DKD EV

MSCs

DKD

MSCs

MSCs++

0

5

10

15

20

25

Veh ACF Veh ACF

CSF1

mRN

A

0

2

4

6

8

10

12

14

CSF

1R m

RN

A

20% O2 20% O2 1% O2 1% O2

0

2

4

6

8

10

CSF

1R m

RN

A

20% O2

1% O2

MSCsMSCs

Veh ACF Veh ACF

*

#

##

*#

**

##

* *** **

A

D

E

B

0

50

100

150

200

**

##

0

10

20

30

40

50

BCCs MSCs BCCs MSCsCoculture

CSF

1R m

RN

A

20% O2

1% O2

EV DKD EV

MSCs

DKD

MSCs

MSCs+ +

**

**

20% O2

1% O2

20% O2

1% O2

CSF

1 m

RN

A

C

F

BCCs MSCs BCCs MSCsCoculture

**

BCCs

BCCsBCCs

+ MSCsBCCs + MSCs

Fig. 4. HIFs mediate coculture- and hypoxia-induced expression of CSF1 andCSF1R. (A and C) GFP-expressing MDA-MB-231 BCCs were cultured in-dividually or cocultured with MSCs at 20% or 1% O2 for 48 h, and GFP

+ BCCsand CD105+ MSCs were then sorted by FACS. CSF1 (A) and CSF1R (C) mRNAlevels were determined and normalized to BCCs at 20% O2. **P < 0.001 vs.20% BCCs; ##P < 0.001 vs. 1% BCCs. (B and D) EV BCCs, DKD BCCs, MSCs, EV +MSCs or DKD + MSCs were cultured at 20% or 1% O2 for 48 h. CSF1 (B) andCSF1R (D) mRNA levels were determined and normalized to EV at 20%. *P <0.05 vs. 20% EV; **P < 0.001 vs. 1% EV; #P < 0.01 vs. 1% EV; ##P < 0.001 vs.1% EV +MSCs. (E and F) BCCs, MSCs, or BCCs +MSCs were treated with 1 μMacriflavine (ACF) or DMSO vehicle (Veh) and exposed to 20% or 1% O2 for48 h. CSF1 (E) and CSF1R (F) mRNA levels were normalized to BCC Veh at20%. *P < 0.05 vs. Veh; **P < 0.001 vs. Veh.

0

200

400

600

800

1000

1200

1400

21 28 35 42 48 56

Days

NTC shCSF1-1 shCSF1-5

Tum

or v

olum

e (m

m3 )

0 10 20 30 40 50 60 70

BCCs

(% o

f LN

area

)



No.

of l

ung

foci

0 1 2 3 4 5 6 7 8 9

0 0.5

1 1.5

2 2.5

3 3.5

4 4.5

NTC shCSF1-1 shCSF1-5 NTC shCSF1-1 shCSF1-5


0

0.2

0.4

0.6

0.8

1

1.2


Circ

ulat

ing

tum

or c

ells

NTC shCSF1-1 shCSF1-5 0 2 4 6 8

10 12

A B

D E

F G

H I

** **

0

0.2

0.4

0.6

0.8

1

1.2 C


* *

** **

****

***

***

% C

SF1R

+ F4/

80+ c

ells

% C

D11b

Ly6

C c

ells

++

Met

asta

tic B

CC

s in

lung

s

Fig. 5. CSF1 promotes metastasis of MDA-MB-231 BCCs. MDA-MB-231subclones expressing shRNA against CSF1 (shCSF1-1 and shCSF1-5) or NTCshRNA were implanted in the MFP of female SCID mice. (A) Primary tu-mor volumes were determined serially. (B) Circulating tumor cells weredetermined by qPCR and normalized to NTC. *P < 0.01 vs. NTC. (C )Metastatic BCCs in the lungs were determined by qPCR and normalizedto NTC. **P < 0.001 vs. NTC. (D) Photomicrographs of H&E-stained lungsections. (Scale bars: 0.5 mm.) (E ) Lung sections were scored for metastaticfoci per field. **P < 0.001 vs. NTC. (F) Lymph node sections were stained withhuman-specific vimentin. (Scale bars: 0.5 mm.) (G) Stained lymph node (LN)area was quantified by image analysis. **P < 0.001 vs. NTC. (H and I) Thepercentage of CSF1R+F4/80+ (H) and CD11b+Ly6C+ (I) cells in primary tumorswas determined by FACS. *P < 0.01 vs. NTC; **P < 0.001 vs. NTC.


MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF3

receptor CCR5 in BCCs (9, 44). However, neither the upstreamtrigger nor the downstream effector of CCL5→CCR5 signalingwas delineated. To investigate whether CCL5→CCR5 signalingregulates CSF1 expression, CCL5 NAb was added to coculturesof MSCs and BCCs. Expression of both CCL5 (Fig. 6A) andCCR5 (Fig. 6B) mRNA was significantly decreased in the pres-ence of CCL5 NAb. Blocking CCL5→CCR5 signaling also sig-nificantly decreased CSF1 mRNA levels (Fig. 6C). Next, wegenerated MDA-MB-231 subclones stably transfected with vec-tor encoding either of two different shRNAs, which inhibitedexpression of CCR5 mRNA (Fig. S4A) and cell surface protein(Fig. S4B). CCR5 deficiency in BCCs blocked the induction ofCSF1 mRNA expression by coculture or hypoxia (Fig. 6D). Re-markably, CSF1 deficiency in BCCs abrogated hypoxia- and co-culture-induced CCL5 mRNA expression by MSCs (Fig. 6E).Taken together, these results demonstrate that CSF1 expressionby BCCs is regulated by CCL5MSC→CCR5BCC signaling and thatCCL5 expression by MSCs is regulated by CSF1BCC→CSF1RMSCsignaling, indicating the existence of a second feed-forward loop,which, in this case, promotes the recruitment of TAMs andMDSCs to primary breast tumors (Fig. 6F).

CCR5 Promotes Macrophage Recruitment and Metastasis. To in-vestigate the role of CCR5 in breast cancer pathogenesis further,CCR5-deficient BCCs were implanted in the MFP of femaleSCID mice. CCR5 deficiency had no effect on primary tumorgrowth (Fig. 7A). However, mice bearing CCR5-deficient tumorsshowed significantly decreased numbers of circulating tumor cells(Fig. 7B), metastatic cancer cells (Fig. 7C) and metastatic foci(Fig. 7 D and E) in the lungs, metastatic cancer cells in lymphnodes (Fig. 7 F and G), and TAMs (Fig. 7H) and MDSCs (Fig.7I) recruited to the primary tumor. These results indicate thatCCR5 expression by BCCs promotes macrophage recruitmentand metastasis.

CSF1 and CCR5 Are HIF Target Genes. Analysis of the human CSF1gene sequence revealed two candidate HIF binding sites located0.6 kb (site 1) and 2.5 kb (site 2) 5′ to the transcription start site(Fig. 8A). To determine whether HIFs bind to these sites, ChIPassays were performed with MDA-MB-231 BCCs, which dem-onstrated hypoxia-induced binding of HIF-1α (Fig. 8B), HIF-1β(Fig. 8C), and HIF-2α (Fig. 8D) to site 1 and hypoxia-inducedbinding of HIF-1α (Fig. 8E) and HIF-1β (Fig. 8F), but not HIF-2α (Fig. 8G), to site 2.

- MSCs +MSCs

0 2 4 6 8

10 12 14 16 18 20

0

5

10

15

20

25

-MSCs +MSCs

0

0.5

1

1.5

2

2.5

0

0.5

1

1.5

2

2.5

CCL5 NAb CCL5 NAb

20% O2 1% O2

IgG IgG

CC

R5

mR

NA

CS

F1 m

RN

A

20% O2 1% O2

NTC shCCR5-4 shCCR5-5 NTC shCCR5-4 shCCR5-5 NTC shCSF1-1 shCSF1-5 NTC shCSF1-1 shCSF1-5

CS

F1 m

RN

A

CC

L5 m

RN

A20% O2 1% O2

20% O2 1% O2

B

0

1

2

3

4

5

CC

L5 m

RN

A

CCL5 NAbIgG

20% O2 1% O2

**

##

*

*

# #

**

## ##

**

## ##*

# #

E

A C

D

BCC MSC

CCL5

TAM

CSF1

CSF1R

CCR5

homing to BCC

CSF1R

F

TAM

Fig. 6. CCL5-CCR5 signaling between MSCs and BCCs is required for CSF1expression. (A–C) MDA-MB-231 BCCs +MSCs were treated with CCL5 NAb orIgG and exposed to 20% or 1% O2 for 48 h. CCL5 (A), CCR5 (B), and CSF1 (C)mRNA levels were normalized to 20% IgG. *P < 0.01 vs. 20% IgG; **P < 0.001vs. 20% IgG; #P < 0.01 vs. 1% IgG; ##P < 0.001 vs. 1% IgG. (D) MDA-MB-231subclones expressing NTC shRNA or shRNA against CCR5 (shCCR5-4 andshCCR5-5) were cultured alone or with MSCs and exposed to 20% or 1% O2for 48 h. CSF1 mRNA levels were normalized to 20% NTC. *P < 0.05 vs. 20%NTC; **P < 0.001 vs. 20% NTC; #P < 0.05 vs. 1% NTC; ##P < 0.001 vs. 1% NTC +MSCs. (E) MDA-MB-231 subclones expressing NTC shRNA or shRNA againstCSF1 (shCSF1-1 and shCSF1-5) were cultured alone or with MSCs and exposedto 20% or 1% O2 for 48 h. CCL5 mRNA levels were normalized to 20% NTC.**P < 0.001 vs. 20% NTC + MSCs; ##P < 0.001 vs. 1% NTC + MSCs. (F) Bi-directional signaling between BCCs and MSCs generates a feed-forward loopthat stimulates tumor-associated macrophage (TAM) recruitment.

0 0.5

1 1.5

2 2.5

3 3.5

4

0

200

400

600

800

1000

1200

21 28 35 42 49

Tum

or V

olum

e (m

m3

)

Days

*

NTC shCCR5-4 shCCR5-5

A B C

0

0.2

0.4

0.6

0.8

1

1.2

**

NTC shCCR5-4 C

ircul

atin

g tu

mor

cel

ls

0

0.2

0.4

0.6

0.8

1

1.2

shCCR5-5 NTC shCCR5-4 shCCR5-5

** **

D

0

1

2

3

4

5

6

7


H


* * **

NTC shCCR5-4 shCCR5-5 E

FG

0

2

4

6

8

10

12


** **

No.

of l

ung

foci

I


0

10

20

30

40

50

60

70


****

BC

Cs

(% o

f LN

are

a)

% C

SF1R

+ F4/

80+ c

ells

% C

D11b

Ly6

C c

ells

++

Met

asta

tic B

CC

s in

lung

s

Fig. 7. CCR5 promotes macrophage recruitment and breast cancer metas-tasis. MDA-MB-231 subclones were implanted in the MFP of SCID mice. (A)Primary tumor volumes were determined serially. (B) Circulating tumor cellswere determined by qPCR and normalized to NTC. *P < 0.05 vs. NTC. (C)Metastatic BCCs in the lungs were determined by qPCR. **P < 0.001 vs. NTC.(D) Photomicrographs of H&E-stained lung sections. (Scale bars: 0.5 mm.) (E)Lung sections were scored for metastatic foci per field. **P < 0.001 vs. NTC.(F and G) Lymph node sections were stained with human-specific vimentinantibody (F), and staining was quantified by image analysis (G). **P < 0.001vs. NTC. (Scale bars: F, 0.5 mm.) The percentage of CSF1R+F4/80+ (H) andCD11b+Ly6C+ (I) cells in primary tumors was determined by FACS. *P < 0.01vs. NTC.


Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=SF4www.pnas.org/cgi/doi/10.1073/pnas.1406655111

Analysis of the human CCR5 gene sequence revealed candi-date HIF binding sites in the 5′-flanking and 3′-untranslated regions,1,370 bp 5′ (site 1) and 8,065 bp 3′ (site 2) of the transcription startsite, respectively (Fig. 8H). ChIP assays revealed hypoxia-inducedbinding of HIF-2α and HIF-1β, but not HIF-1α, to site 1 (Fig. 8 I–K)and hypoxia-induced binding of HIF-1α and HIF-1β, but not HIF-2α, to site 2 (Fig. 8 L–N). Taken together, the ChIP data demon-strate that the human CSF1 and CCR5 genes are directly regulatedby both HIF-1 and HIF-2.

CCL5 and CSF1 Stimulate Macrophage Migration. To study macro-phage migration directly, we isolated bone marrow cells fromBALB/c mice and incubated them in the presence of CSF1 tostimulate macrophage (BM-Mϕ) differentiation, with CSF1R+F4/80+ cells representing 80% of the final population (Fig. 9A).CM from BCCs +MSCs cocultured at 20% O2 stimulated BM-Mϕmigration, and the effect was augmented when CM from hypoxiccocultures was used (Fig. 9B). The stimulatory effect of CM fromcocultures was abrogated when CCL5 NAb was added to the CM(Fig. 9C) or when CSF1-deficient BCCs were used (Fig. 9D). These

results indicate that coculture- and hypoxia-induced CCL5 andCSF1 expression stimulate BM-Mϕ migration.

HIFs Are Required for Macrophage Recruitment and Metastasis ina Syngeneic Mouse Mammary Carcinoma Model. To investigate therole of HIFs in macrophage recruitment in an immunocompe-tent model of TNBC, we used 4T1 mouse mammary carcinomacells, which form primary tumors and metastases similar to hu-man TNBC after implantation into the MFP of syngeneic BALB/cmice (45). We generated 4T1 subclones that were stably trans-fected with vectors encoding NTC shRNA or shRNA(s) thatinhibited the expression of HIF-1α (sh1α), HIF-2α (sh2α), or both

0

1

2

3

4

5

0 0.2 0.4 0.6 0.8

1 1.2 1.4 1.6

IgG HIF2

0 0.5 1

1.5 2

2.5 3

3.5 4

IgG HIF1

E1 E2 E3 E4 E5 E6 E9 E7 E8

CSF1 ~600 nt~2500 nt

*

*

Site 1Site 2

E1 E2 E3

CCR5 -1370 nt

Site 1

+8065 nt

Site 2

**

0 0.2 0.4 0.6 0.8

1 1.2 1.4

0

1

2

3

4

5

6 *

0

1

2

3

4

5 20% O2 1% O2

*

*

0

1

2

3

4

5

Bind

ing

at s

ite 2

20% O2 1% O2

IgG HIF1B

indi

ng a

t site

2*

Bind

ing

at s

ite 2

Bin

ding

at s

ite 1

Bind

ing

at s

ite 1

Bin

ding

at s

ite 1

HIF10

0.5 1

1.5 2

2.5 3

3.5

0

1

2

3

4

5

IgG HIF1

Bin

ding

at s

ite 1

Bind

ing

at s

ite 1

20% O2 1% O2

IgG

20% O2 1% O2

HIF2IgG

Bind

ing

at s

ite 1

A

B C D

E G

H

F

I J K

L M N

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 20% O2

1% O2 **

Bin

ding

at s

ite 2

Bin

ding

at s

ite 2

Bin

ding

at s

ite 2

20% O2 1% O2

*

20% O2 1% O2

20% O2 1% O2

20% O2 1% O2

IgG HIF1 IgG HIF1 IgG HIF2

IgG HIF1 IgG HIF1 IgG HIF2

20% O2 1% O2

20% O2 1% O2

0 0.2 0.4 0.6 0.8

1 1.2 1.4 1.6 20% O2

1% O2

Fig. 8. CSF1 and CCR5 are HIF target genes. (A) Candidate HIF binding siteswere identified in the 5′-flanking region of the human CSF1 gene. (B–G)MDA-MB-231 cells were incubated at 20% or 1% O2 for 16 h, and ChIP assayswere performed using IgG or antibody against HIF-1α, HIF1β, or HIF-2α.Primers flanking site 1 and site 2 were used for qPCR, and values werenormalized to IgG 20%. *P < 0.05 vs. all other conditions, Student’s t test onlog-converted values. (H) Candidate HIF binding sites were identified in the5′-flanking region (site 1) and 3′-untranslated region (site 2) of the humanCCR5 gene. (I–N) MDA-231 cells were incubated at 20% or 1% O2 for 16 h,and ChIP assays were performed using IgG or antibodies against HIF-1α, HIF-1β, or HIF-2α. Primers flanking site 1 or site 2 were used for qPCR, and valueswere normalized to IgG 20%. *P < 0.05 vs. all other conditions; **P < 0.001.

BCCs

MSCs

BCCs + MSCs

20% O2 1% O2

CM: 20% O2 1% O2

0

5

10

15

20

BCCs MSCs

CCL5 NAb

IgG Ab

CCL5 NAb

BCCs

0

5

10

15

BCCs IgG

BCCs+MSCsCCL5 NAb

20% O2 1% O2

A

B C

*

**

#

*

**

#

##

IgG Ab

CM:

NTC shCSF1-1 NTC shCSF1-1

-MSCs +MSCs

BCCs + MSCs

BCCs + MSCs

CM: 20% O2 1% O2

BCCs + MSCs

CM:

20% O2

1% O2

BM

79.5%

BM-MΦ

F4/80

0 2 4 6 8

10 12 14 16 18

NTC shCSF1-1 NTC shCSF1-1

**

#

##

*

Mig

ratio

n of

BM

-Mϕ

Mig

ratio

n of

BM

-Mϕ

Mig

ratio

n of

BM

-Mϕ 20

-MSCs +MSCs

CS

F1R

+

+

CM:

0.03%

D

Fig. 9. Coculture- and hypoxia-induced CCL5 and CSF1 expression is re-quired for migration of bone marrow-derived macrophages (BM-Mϕ). (A)Bone marrow cells were isolated from the BALB/c mice and differentiatedinto macrophages (BM-Mϕ) by supplementing the culture media with CSF1.The efficiency of differentiation was determined by FACS analysis of CSF1Rand F4/80 surface antigen expression, using antibodies conjugated to phy-coerythrin (PE-A) or allophycocyanin (APC-A). (B) BM-Mϕ were seeded in theupper compartment of a Boyden chamber, and the number of cells thatmigrated through the filter in response to CM from nonhypoxic (20% O2) orhypoxic (1% O2) MDA-MB-231 cells (BCCs), MSCs, or BCCs + MSCs in thelower compartment were counted under light microscopy after stainingwith crystal violet. (Scale bars: 0.2 mm.) Data were normalized to CM fromBCCs at 20% O2. *P < 0.05 vs. 20% BCCs; **P < 0.01 vs. 20% BCCs;

#P < 0.01vs. 20% BCCs + MSCs. (C) BM-Mϕ migration in response to CM isolated fromBCCs or BCCs + MSCs (supplemented with IgG or CCL5 NAb) was determinedand normalized to CM from BCCs at 20% O2. (Scale bars: 0.2 mm.) *P < 0.05vs. 20% BCCs; **P < 0.01 vs. 20% BCCs; #P < 0.05 vs. 20% IgG; ##P < 0.001 vs.1% IgG. (D) BM-Mϕ migration in response to CM isolated from NTC orshCSF1-1 BCCs (alone or cocultured with MSCs) was determined and nor-malized to CM from NTC cells at 20% O2. (Scale bars: 0.2 mm.) *P < 0.05 vs.20% NTC − MSCs; **P < 0.01 vs. 20% NTC − MSCs; #P < 0.05 vs. 1% NTC -MSCs; ##P < 0.001 vs. 1% NTC + MSCs.


MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

22,

202

1

0

200

400

600

800

1000

1200

5 10 15 20

* *

*

*

Tum

or V

olum

e (m

m )3

0 2 4 6 8

10 12 14 16 18 20

0

5

10

15

20

25

* *

*

*

** 0 2 4 6 8

10 12 14 16

****

*

No.

of L

N fo

ci

A

C

0

5

10

15

20

***

F

Days

0

0.5

1

1.5

2

2.5

Tum

or w

eigh

t (g)

****

B

D

No.

of l

ung

met

asta

ses

E

% C

SF1R

+ F4/

80+ c

ells

% C

D11b

Ly6

C c

ells

++

sh1sh2

sh2sh1

NTC

DKD

NTC DKD

DKD

DKDNTC

NTC

sh1

sh1

sh2

sh2

sh2sh2

DKDDKD

NTCNTC sh1

sh1

sh2 DKDNTC sh1 sh2 DKDNTC sh1

Fig. 10. HIFs are required for macrophage recruitment and metastasis in the immunocompetent 4T1 mouse mammary carcinoma model of TNBC. (A) 4T1subclones expressing NTC shRNA or shRNA against HIF-1α (sh1α), HIF-2α (sh2α), or both HIF-1α and HIF-2α (DKD) were implanted in the MFP of femalesyngeneic BALB/c mice. Primary tumor volumes were determined serially. *P < 0.05 vs. NTC or sh2α. (B) Tumor weights were measured at the end of theexperiment. **P < 0.01 vs. NTC. (C) Whole-mount inflated lungs were stained with India blue dye. The number of metastatic nodules (arrows) was counted.*P < 0.05 vs. NTC; **P < 0.01 vs. NTC. (D) Lymph node sections were stained with H&E. (Scale bars: 2 mm.) The number of metastatic foci (arrows) was counted.*P < 0.05 vs. NTC; **P < 0.01 vs. NTC. (E and F) The percentage of CSF1R+F4/80+ (E) and CD11b+Ly6C+ (F) cells in primary tumors was determined by FACS. *P <0.05 vs. NTC; **P < 0.01 vs. NTC.


Dow

nloa

ded

by g

uest

on

June

22,

202

1

www.pnas.org/cgi/doi/10.1073/pnas.1406655111

(DKD) (Fig. S5A). After orthotopic implantation in the MFP offemale BALB/c mice, the growth of primary tumors derived fromsh1α or DKD cells was significantly decreased compared withtumors derived from sh2α or NTC cells (Fig. 10 A and B). Thenumbers of metastatic nodules in lungs (Fig. 10C) and metastaticfoci in axillary lymph nodes (Fig. 10D) harvested from miceimplanted with sh1α, sh2α, or DKD cells were significantlydecreased compared those from mice implanted with NTCcells. Recruitment of TAMs (Fig. 10E and Fig. S5B) and MDSCs(Fig. 10F and Fig. S5C) was significantly decreased in primarytumors derived from sh1α and DKD cells compared with tu-mors derived from sh2α or NTC cells. Taken together, these dataindicate that HIFs are required for primary tumor growth,macrophage recruitment, and metastasis of TNBCs in immuno-competent mice.

Breast Cancer Gene Expression Data Suggest Signaling Pathways AreClinically Relevant. To determine whether the two bidirectionalsignaling loops that were delineated in BCC-MSC cocultures areclinically relevant, we analyzed microarray gene expression dataobtained from 530 primary human breast cancers, 63 samples ofadjacent normal breast tissue, and 6 metastases (2). We dem-onstrated that CXCR3 expression (and activation) in BCCs wasrequired for the induction of CXCL16 expression in BCCs inresponse to coculture with MSCs and exposure to hypoxia (Fig. 1),and Pearson’s test revealed that CXCR3 and CXCL16 mRNAlevels were highly correlated (P = 10−13) in the clinical specimens.Similarly, we demonstrated that CCR5 expression (and activation)in BCCs was required for the induction of CSF1 expression inBCCs in response to coculture and hypoxia (Fig. 6), and Pearson’stest revealed that CCR5 and CSF1 mRNA levels were highlycorrelated (P = 10−16) in the clinical specimens.

DiscussionIn this study, we have delineated two HIF-regulated feed-forwardsignaling loops between BCCs and MSCs that drive stromal cellrecruitment to primary breast tumors (Fig. 11). The first loop, in-volving CXCL16BCC→CXCR6MSC and CXCL10MSC→CXCR3BCCsignaling arms, with each arm stimulating the activity of theother, drives recruitment of MSCs. The second loop, involvingCCL5MSC→CCR5BCC and CSF1BCC→CSF1RMSC signalingarms, with each arm stimulating the activity of the other,drives recruitment of TAMs and MDSCs. These two pathwaysoperate independent of each other, but both are dependent onHIF activity, with hypoxia serving as a physiological signal thatsynergizes with chemokine signals from MSCs to trigger CSF1and CCR5 gene transcription in BCCs.The recruitment of MSCs promotes cancer progression and

metastasis (6–9). In a previous study, we showed that PGF se-cretion by BCCs facilitates the recruitment of MSCs to breasttumors and that PGF binds to VEGF receptor 1 (VEGFR1) onMSCs to stimulate CXCL10 expression (9). In the present study,we have demonstrated that in addition to PGF, CXCL16 secre-tion by hypoxic BCCs recruits MSCs to primary breast tumors.Furthermore, whereas PGFBCC→VEGFR1MSC signaling is unidi-rectional, we show that CXCL16BCC→CXCR6MSC signaling ispart of a bidirectional feed-forward loop that activates CXCR3signaling in BCCs (Fig. 11), which is proinvasive and prometastatic(9). As a result, CXCL16 deficiency markedly impaired lymphnode and lung metastasis of BCCs.Elegant studies have demonstrated that CSF1 expression by

BCCs promotes macrophage recruitment and metastasis inmouse models of breast cancer (15–18). Elevated serum levels ofCSF1 predict lymph node involvement in women with early- stagebreast cancer and decreased overall survival in postmenopausalpatients with breast cancer (46). However, the mechanisms bywhich CSF1 expression is induced in BCCs had not been de-lineated. Similarly, CCL5MSC→CCR5BCC signaling was linked toinvasion and metastasis (6, 47), but the upstream trigger anddownstream effector had not been delineated. We have nowunified these observations by showing that CCL5MSC→CCR5BCCsignaling induces CSF1 expression in BCCs, which serves torecruit CSF1R+ TAMs, as well as feeding back to stimulateCCL5 expression by MSCs (Fig. 6H).HIF activity in cancer cells plays critical roles in the pro-

duction of angiogenic growth factors and the mobilization ofbone marrow-derived angiogenic cells, which are blocked bytreating tumor-bearing mice with HIF inhibitors (43). Treatmentof mice bearing primary breast tumors with acriflavine or digoxinpotently inhibits metastatic niche formation and blocks lungand lymph node metastasis (30, 31, 48). Hypoxia induces HIF-dependent expression of genes in BCCs that activate cell motility(34), ECM remodeling (32, 33), lymphangiogenesis, and lymphnode metastasis (31), as well as extravasation and lung metastasis(30). The present study now identifies recruitment of MSCs,TAMs, and MDSCs as additional components of the metastaticprocess that are stimulated by intratumoral hypoxia in an HIF-dependent manner. Finally, both human MDA-MB-231 andmouse 4T1 cells are preclinical models for the 15% of humanbreast cancers, designated as TNBC, that do not express pro-gesterone, estrogen, or HER2 receptors and manifest the basal/claudin-low gene expression pattern. Further studies are re-quired to determine whether these same HIF-driven intercellularsignaling mechanisms are also activated by hypoxia in estrogen/progesterone receptor-positive and HER2+ breast cancers.Targeted therapies are not available for TNBCs, which are

treated with cytotoxic chemotherapy to which

digoxin, blocked signaling and recruitment of TAMs and MDSCs,suggesting that addition of HIF inhibitors to existing thera-peutic regimens may improve the clinical outcome in patientswith TNBC.

Materials and MethodsCell Lines and Culture. Mycoplasma-free and molecularly authenticated hu-man MDA-MB-231 BCCs were maintained in high-glucose (4.5 mg/mL)DMEMwith 10% FBS and 1% penicillin/streptomycin. The 4T1mouse mammarycarcinoma cells (American Type Culture Collection) were maintained in RPMI-1640 with 10% FBS and 1% penicillin/streptomycin. Human bone marrow-derived MSCs (49) were obtained from the Tulane Center for Gene Therapy.MSCs were maintained in α-MEM supplemented with 20% FBS and 1%penicillin/streptomycin. Cells were maintained at 37 °C in a 5% CO2 and 95%air incubator (20% O2). Hypoxic cells were maintained at 37 °C in a modularincubator chamber (Billups–Rothenberg) flushed with a gas mixture containing1%O2, 5% CO2, and 94% N2. For coculture experiments, equal numbers of MSCsand BCCs were seeded in a 1:1 ratio of DMEM/10% FBS and α-MEM/20% FBS.

Transduction with shRNA Vectors. The pLKO.1-puro lentiviral vectors encodingshRNA targeting human CSF1 (clone ID: NM_000757), human CCR5 (clone ID:NM_000579), mouse HIF-1α (clone ID: NM_010431), and mouse HIF-2α (cloneID: NM_010137) were purchased from Sigma–Aldrich. The recombinant vectorswere cotransfected with plasmid pCMV-dR8.91 and plasmid encoding vesicularstomatitis virus G protein into 293T cells using FuGENE 6 (Roche Applied Sci-ence). Viral supernatant was collected 48 h posttransfection, filtered (0.45-μmpore size), and added to MDA-MB-231 cells in the presence of 8 μg/mL poly-brene (Sigma–Aldrich). Puromycin (0.5 μg/mL) was added to the medium ofcells transduced with pLKO.1-puro vectors for selection.

RT-qPCR. Total RNA was extracted from cells using TRIzol (Invitrogen) andtreated with DNase I (Ambion). One microgram of total RNA was used forfirst-strand DNA synthesis with the iScript cDNA Synthesis system (BioRad).qPCR was performed using human-specific primers and SYBR Green qPCRMasterMix (Fermentas). For each primer pair, the annealing temperaturewasoptimized by gradient PCR. The expression of each target mRNA relative to18S rRNA was calculated based on the threshold cycle (Ct) as 2−Δ(ΔCt), whereΔCt = Cttarget − Ct18S and Δ(ΔCt) = ΔCttest − ΔCtcontrol (22). Primer sequencesare provided in Table S1.

Animal Studies. Female 5- to 7-wk-old SCID (National Cancer Institute) orBALB/c (Charles River Laboratories) mice were studied according to protocolsapproved by the Johns Hopkins University Animal Care and Use Committeethat were in accordance with the National Institutes of Health Guide for theCare and Use of Laboratory Animals (50). Digoxin and saline for injectionwere obtained from the research pharmacy of the Johns Hopkins Hospital.BCCs were harvested by trypsinization, washed twice in PBS, counted, andsuspended at 107 cells/mL in a 1:1 solution of PBS and Matrigel (Corning).Mice were anesthetized, and 2 × 106 cells were injected into the MFP. Pri-mary tumors were measured in three dimensions (a, b, and c), and volume (V)was calculated as V = abc × 0.52. Primary tumors and the ipsilateral axillarylymph node were harvested (31). Lungs were perfused with PBS, and onelung was inflated for formalin fixation, paraffin embedding, and stainingwith H&E; the other lung was used to isolate genomic DNA for qPCR toquantify human HK2 and mouse 18S rRNA gene sequences, as previouslydescribed (30).

Circulating Tumor Cell Assay. Total RNA isolated from 0.5 mL of whole bloodwas subjected to qPCR using primers specific for human 18S rRNA (31).

MSC Recruitment Assay. 2 × 106 MDA-MB-231 cells were implanted in theMFP of female SCID mice. When the tumor reached a volume of 450 mm3,5 × 105 MSCs (derived from a male donor) were injected i.v. and the primary

tumor was harvested after 16 h. MSCs recruited to the primary tumor weredetermined by qPCR analysis of SRY copy number (9).

Flow Cytometry. Tumor tissue was minced and digested with 1 mg/mL type 1collagenase (Sigma) at 37 °C for 30 min. Digested tissues were filtered through70-μm cell strainers. Cells were incubated with Fc Block (BD Pharmingen). Toidentify TAMs, cells were stained with peridinin chlorophyll protein-conju-gated CSF1R antibody (BD Biosciences) and allophycocyanin-conjugated F4/80antibody (Novus Biologicals), and subjected to flow cytometry. To identifyMDSCs, cells were stained with allophycocyanin-conjugated CD11b antibody(BD Biosciences) and peridinin chlorophyll protein-conjugated Ly6C antibody(BD Biosciences), and subjected to flow cytometry. Unstained control andsingle-stained cells were prepared in every experiment for gating. Dead cellswere gated out by side-scatter and forward-scatter analysis.

Isolation of Mouse Bone Marrow-Derived Macrophages. Bone marrow cellswere isolated and cultured for 10 d in RPMI-1640 supplemented with 10%FBS, 1% penicillin/streptomycin, and 100 ng/mL CSF1 (R&D Systems) accordingto standard protocols (51, 52).

Migration Assays. Cells were seeded onto an uncoated filter in the uppercompartment of a 24-well Boyden chamber (8-mm pore size; Costar) andallowed to migrate for 8 h in response to CM in the lower compartment. Thecells that migrated to the underside of the filter were stained with crystalviolet and counted under bright-field microscopy.

India Ink Staining of Lungs. Mice were euthanized, and India ink (15%) wasinjected into the lungs through the trachea. The lungs were fixed in Feketet’ssolution (100 mL of 70% alcohol, 10 mL of formalin, and 5 mL of glacialacetic acid) at room temperature. After de-staining, metastatic nodules appearwhite on a black background (53).

Analysis of Lymph Node Metastasis. Immunohistochemical analyses wereperformed on formalin-fixed, paraffin-embedded lymph node sections aspreviously described (9). Staining was performed using an antibody thatspecifically recognizes human vimentin (Santa Cruz Biotechnology) and an-alyzed by ImageJ software (National Institutes of Health). The acquiredimages in red-green-blue (RGB) color were separated into different colorchannels by a color deconvolution method (54).

ChIP Assays. MDA-MB-231 cells were cross-linked with formaldehyde andlysed with SDS lysis buffer. Chromatin was sheared by sonication, and lysateswere precleared with salmon sperm DNA/protein A-agarose slurry (Millipore)and incubated with antibody against HIF-1α (Santa Cruz Biotechnology),HIF-1β (Novus Biologicals), or HIF-2α (Novus Biologicals), or with IgG (SantaCruz Biotechnology or Novus Biologicals) as previously described (9).

Statistical Analysis. All data are presented as mean ± SEM (n = 3). Differencesbetween groups were analyzed by one-way ANOVA, unless otherwise stated,before normalization of data. Level 3 gene expression data from 596 patientsin the Breast Invasive Carcinoma dataset (2) were obtained from The CancerGenome Atlas Data Portal (http://tcga-data.nci.nih.gov/tcga/tcgaHome2.jsp).GraphPad Prism software was used to calculate Pearson’s correlation coef-ficients and determine P values for coexpression.

ACKNOWLEDGMENTS. We thank Karen Padgett of Novus Biologicals, Inc.for providing IgG and antibodies against F4/80, HIF-1β, and HIF-2α. We thankDarwin Prockop for providing human MSCs from the Tulane Center for GeneTherapy, which was supported by Grant P40-RR017447 from the National Insti-tutes of Health. G.L.S. is an American Cancer Society Research Professor and theC. Michael Armstrong Professor at Johns Hopkins University School of Medicine.This work was supported by funds from the American Cancer Society and Na-tional Cancer Institute. D.M.G. was supported by a postdoctoral fellowship fromthe Susan G. Komen Foundation, and N.T. was supported by the Japan Scienceand Technology Agency.

1. Marino N, et al. (2013) Breast cancer metastasis: Issues for the personalization of itsprevention and treatment. Am J Pathol 183(4):1084–1095.

2. Cancer Genome Atlas Network (2012) Comprehensive molecular portraits of humanbreast tumours. Nature 490(7418):61–70.

3. Bohl CR, Harihar S, Denning WL, Sharma R, Welch DR (2014) Metastasis suppressors inbreast cancers: Mechanistic insights and clinical potential. J Mol Med (Berl) 92(1):13–30.

4. Joyce JA, Pollard JW (2009) Microenvironmental regulation of metastasis. Nat RevCancer 9(4):239–252.

5. Hanahan D, Coussens LM (2012) Accessories to the crime: Functions of cells recruitedto the tumor microenvironment. Cancer Cell 21(3):309–322.

6. Karnoub AE, et al. (2007) Mesenchymal stem cells within tumour stroma promotebreast cancer metastasis. Nature 449(7162):557–563.

7. Mi Z, et al. (2011) Osteopontin promotes CCL5-mesenchymal stromal cell-mediatedbreast cancer metastasis. Carcinogenesis 32(4):477–487.

8. El-Haibi CP, et al. (2012) Critical role for lysyl oxidase in mesenchymal stem cell-drivenbreast cancer malignancy. Proc Natl Acad Sci USA 109(43):17460–17465.

9. Chaturvedi P, et al. (2013) Hypoxia-inducible factor-dependent breast cancer-mesenchy-mal stem cell bidirectional signaling promotes metastasis. J Clin Invest 123(1):189–205.

10. Kelly PMA, Davison RS, Bliss E, McGee JO (1988) Macrophages in human breast dis-ease: A quantitative immunohistochemical study. Br J Cancer 57(2):174–177.


Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406655111/-/DCSupplemental/pnas.201406655SI.pdf?targetid=nameddest=ST1http://tcga-data.nci.nih.gov/tcga/tcgaHome2.jspwww.pnas.org/cgi/doi/10.1073/pnas.1406655111

11. Leek RD, Landers RJ, Harris AL, Lewis CE (1999) Necrosis correlates with high vasculardensity and focal macrophage infiltration in invasive carcinoma of the breast. Br JCancer 79(5-6):991–995.

12. Tsutsui S, et al. (2005) Macrophage infiltration and its prognostic implications inbreast cancer: The relationship with VEGF expression and microvessel density. OncolRep 14(2):425–431.

13. Medrek C, Pontén F, Jirström K, Leandersson K (2012) The presence of tumor asso-ciated macrophages in tumor stroma as a prognostic marker for breast cancer pa-tients. BMC Cancer 12:306.

14. O’Sullivan C, Lewis CE, Harris AL, McGee JO (1993) Secretion of epidermal growthfactor by macrophages associated with breast carcinoma. Lancet 342(8864):148–149.

15. Lin EY, Nguyen AV, Russell RG, Pollard JW (2001) Colony-stimulating factor 1 pro-motes progression of mammary tumors to malignancy. J Exp Med 193(6):727–740.

16. Wyckoff J, et al. (2004) A paracrine loop between tumor cells and macrophages isrequired for tumor cell migration in mammary tumors. Cancer Res 64(19):7022–7029.

17. Pollard JW (2004) Tumour-educated macrophages promote tumour progression andmetastasis. Nat Rev Cancer 4(1):71–78.

18. Goswami S, et al. (2005) Macrophages promote the invasion of breast carcinoma cellsvia a colony-stimulating factor-1/epidermal growth factor paracrine loop. Cancer Res65(12):5278–5283.

19. Qian BZ, et al. (2011) CCL2 recruits inflammatory monocytes to facilitate breast-tumourmetastasis. Nature 475(7355):222–225.

20. Harris AL (2002) Hypoxia—A key regulatory factor in tumour growth. Nat Rev Cancer2(1):38–47.

21. Vaupel P, Mayer A, Höckel M (2004) Tumor hypoxia and malignant progression.Methods Enzymol 381:335–354.

22. Sullivan R, Graham CH (2007) Hypoxia-driven selection of the metastatic phenotype.Cancer Metastasis Rev 26(2):319–331.

23. Semenza GL (2012) Hypoxia-inducible factors: Mediators of cancer progression andtargets for cancer therapy. Trends Pharmacol Sci 33(4):207–214.

24. Schindl M, et al.; Austrian Breast and Colorectal Cancer Study Group (2002) Over-expression of hypoxia-inducible factor 1α is associated with an unfavorable prognosisin lymph node-positive breast cancer. Clin Cancer Res 8(6):1831–1837.

25. Bos R, et al. (2003) Levels of hypoxia-inducible factor-1α independently predict prognosisin patients with lymph node negative breast carcinoma. Cancer 97(6):1573–1581.

26. Dales JP, et al. (2005) Overexpression of hypoxia-inducible factor HIF-1α predicts earlyrelapse in breast cancer: Retrospective study in a series of 745 patients. Int J Cancer116(5):734–739.

27. Generali D, et al. (2006) Hypoxia-inducible factor-1α expression predicts a poor re-sponse to primary chemoendocrine therapy and disease-free survival in primary hu-man breast cancer. Clin Cancer Res 12(15):4562–4568.

28. Helczynska K, et al. (2008) Hypoxia-inducible factor-2α correlates to distant re-currence and poor outcome in invasive breast cancer. Cancer Res 68(22):9212–9220.

29. Wong CC, et al. (2011) Hypoxia-inducible factor 1 is a master regulator of breastcancer metastatic niche formation. Proc Natl Acad Sci USA 108(39):16369–16374.

30. Zhang H, et al. (2012) HIF-1-dependent expression of angiopoietin-like 4 and L1CAMmediates vascular metastasis of hypoxic breast cancer cells to the lungs. Oncogene31(14):1757–1770.

31. Schito L, et al. (2012) Hypoxia-inducible factor 1-dependent expression of platelet-derived growth factor B promotes lymphatic metastasis of hypoxic breast cancer cells.Proc Natl Acad Sci USA 109(40):E2707–E2716.

32. Gilkes DM, et al. (2013) Collagen prolyl hydroxylases are essential for breast cancer

metastasis. Cancer Res 73(11):3285–3296.33. Gilkes DM, et al. (2013) Procollagen lysyl hydroxylase 2 is essential for hypoxia-

induced breast cancer metastasis. Mol Cancer Res 11(5):456–466.34. Gilkes DM, et al. (2014) Hypoxia-inducible factors mediate coordinated RhoA-ROCK1

expression and signaling in breast cancer cells. Proc Natl Acad Sci USA 111(3):E384–E393.35. Murdoch C, Giannoudis A, Lewis CE (2004) Mechanisms regulating the recruitment of

macrophages into hypoxic areas of tumors and other ischemic tissues. Blood 104(8):

2224–2234.36. Jung Y, et al. (2013) Recruitment of mesenchymal stem cells into prostate tumours

promotes metastasis. Nat Commun 4:1795.37. Matsumura S, et al. (2008) Radiation-induced CXCL16 release by breast cancer cells

attracts effector T cells. J Immunol 181(5):3099–3107.38. Lin S, et al. (2009) Chemokine C-X-C motif receptor 6 contributes to cell migration

during hypoxia. Cancer Lett 279(1):108–117.39. Lewis C, Murdoch C (2005) Macrophage responses to hypoxia: Implications for tumor

progression and anti-cancer therapies. Am J Pathol 167(3):627–635.40. Murdoch C, Lewis CE (2005) Macrophage migration and gene expression in response

to tumor hypoxia. Int J Cancer 117(5):701–708.41. Leek RD, Harris AL (2002) Tumor-associated macrophages in breast cancer. J Mammary

Gland Biol Neoplasia 7(2):177–189.42. De Palma M, Lewis CE (2013) Macrophage regulation of tumor responses to anti-

cancer therapies. Cancer Cell 23(3):277–286.43. Lee K, et al. (2009) Acriflavine inhibits HIF-1 dimerization, tumor growth, and vas-

cularization. Proc Natl Acad Sci USA 106(42):17910–17915.44. Lin S, et al. (2012) Chemokine C-C motif receptor 5 and C-C motif ligand 5 promote

cancer cell migration under hypoxia. Cancer Sci 103(5):904–912.45. Aslakson CJ, Miller FR (1992) Selective events in the metastatic process defined by

analysis of the sequential dissemination of subpopulations of a mouse mammary

tumor. Cancer Res 52(6):1399–1405.46. Aharinejad S, et al. (2013) Elevated CSF1 serum concentration predicts poor overall

survival in women with early breast cancer. Endocr Relat Cancer 20(6):777–783.47. Velasco-Velázquez M, et al. (2012) CCR5 antagonist blocks metastasis of basal breast

cancer cells. Cancer Res 72(15):3839–3850.48. Wong CC, et al. (2012) Inhibitors of hypoxia-inducible factor 1 block breast cancer

metastatic niche formation and lung metastasis. J Mol Med (Berl) 90(7):803–815.49. Colter DC, Class R, DiGirolamo CM, Prockop DJ (2000) Rapid expansion of recycling

stem cells in cultures of plastic-adherent cells from human bone marrow. Proc Natl

Acad Sci USA 97(7):3213–3218.50. Committee on Care and Use of Laboratory Animals (1985) Guide for the Care and Use

of Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ. No. (NIH) 85-23.51. Zhang X, Goncalves R, Mosser DM (2008) The isolation and characterization of murine

macrophages. Curr Protoc Immunol Chapter 14:Unit 14. 1.52. Weischenfeldt J, Porse B (2008) Bone Marrow-Derived Macrophages (BMM): Isolation

and Applications. CSH Protoc 2008:pdb prot5080.53. Wang Y, et al. (2011) Integrin subunits α5 and α6 regulate cell cycle by modulating the

chk1 and Rb/E2F pathways to affect breast cancer metastasis. Mol Cancer 10:84.54. Ruifrok AC, Johnston DA (2001) Quantification of histochemical staining by color

deconvolution. Anal Quant Cytol Histol 23(4):291–299.


MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

22,

202

1

hypoxia-inducible factor-dependent signaling between triple ...hypoxia-inducible factor-dependent...

Documents