of March 20, 2014.This information is current as

Dendritic Cells in Human Breast CancerProducing−βAssociated with Lymphotoxin

Tumor-Infiltrating Lymphocytes Are High Endothelial Venule Blood Vessels for

Philippe Rochaix, Ignacio Garrido and Jean-Philippe GirardLudovic Martinet, Thomas Filleron, Sophie Le Guellec,

http://www.jimmunol.org/content/191/4/2001doi: 10.4049/jimmunol.13008722013;

2013; 191:2001-2008; Prepublished online 3 JulyJ Immunol 

MaterialSupplementary

2.DC1.htmlhttp://www.jimmunol.org/content/suppl/2013/07/03/jimmunol.130087

Referenceshttp://www.jimmunol.org/content/191/4/2001.full#ref-list-1

, 13 of which you can access for free at: cites 29 articlesThis article

Subscriptionshttp://jimmunol.org/subscriptions

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/ji/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/cgi/alerts/etocReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists, Inc. All rights reserved.Copyright © 2013 by The American Association of9650 Rockville Pike, Bethesda, MD 20814-3994.The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

by D

iane Razanajaona-D

oll on March 20, 2014

http://ww

w.jim

munol.org/

Dow

nloaded from

The Journal of Immunology

High Endothelial Venule Blood Vessels for Tumor-InfiltratingLymphocytes Are Associated with Lymphotoxin b–ProducingDendritic Cells in Human Breast Cancer

Ludovic Martinet,*,†,‡ Thomas Filleron,‡ Sophie Le Guellec,‡ Philippe Rochaix,‡

Ignacio Garrido,*,†,‡,x and Jean-Philippe Girard*,†

Blood vessels and tumor angiogenesis are generally associated with tumor growth and poor clinical outcome of cancer patients.

However, we recently discovered that some blood vessels present within the tumor microenvironment can be associated with

favorable prognosis. These vessels, designated tumor high endothelial venules (HEVs), appear to facilitate tumor destruction by

allowing high levels of lymphocyte infiltration into tumors. In this study, we investigated the mechanisms regulating HEV blood

vessels in human breast cancer. We found that lymphotoxin b was overexpressed in tumors containing high densities of HEVs

(HEVhigh) and correlated to DC-LAMP, a marker of mature DCs. DCs were the main producers of lymphotoxin b in freshly

resected HEVhigh breast tumor samples, and the density of DC-LAMP+ DCs clusters was strongly correlated with the density of

tumor HEVs, T and B cell infiltration, and favorable clinical outcome in a retrospective cohort of 146 primary invasive breast

cancer patients. Densities of tumor HEVs and DC-LAMP+ DCs were strongly reduced during breast cancer progression from in

situ carcinoma to invasive carcinoma, suggesting that loss of tumor HEVs is a critical step during breast cancer progression.

Finally, an increase in the infiltration of regulatory T cells was observed in HEVhigh breast tumors, indicating that tumor HEVs

can develop in the presence of regulatory T cells. Together, our results support a key role for DCs and DC-derived lymphotoxin in

the formation of tumor HEVs. These findings are important because novel therapeutic strategies based on the modulation of tumor

HEVs could have a major impact on clinical outcome of cancer patients. The Journal of Immunology, 2013, 191: 2001–2008.

Blood vessels and tumor angiogenesis are generally as-sociated with tumor growth and poor clinical outcome ofcancer patients (1, 2). However, we recently found that

some blood vessels present within the tumor microenvironmentcan be associated with favorable prognosis (3). These specializedblood vessels, designated tumor high endothelial venules (HEVs),are normally found in lymph nodes where they mediate the ex-travasation of large numbers of lymphocytes from the blood (4–6).In human breast tumors (3) and primary melanomas (7), thedensity of tumor HEVs was highly correlated with the density oftumor-infiltrating CD3+ T cells, CD8+ cytotoxic T cells, and

CD20+ B cells, indicating that, like in lymph nodes, HEVs mayfunction as major gateways for lymphocyte infiltration into solidtumors (3, 7–9). In contrast, we found no correlation betweentumor HEVs and density of blood vessels (HEVhigh and HEVlow

tumors had the same numbers of CD34+ tumor blood vessels),indicating that differences in the density of tumor HEVs are notrelated to differences in tumor angiogenesis (3).By allowing infiltration of naive, memory, cytotoxic, and acti-

vated T cells (3, 8, 9), tumor HEVs may facilitate destruction oftumor cells and generation of memory T cells that limit metastasisof tumor cells at distant sites. Indeed, in breast cancer, we foundthat high densities of tumor HEVs were associated with longerdisease-free and metastasis-free survival of the patients (3). TumorHEVs were also associated with good clinical characteristics inprimary melanomas, including thin Breslow thickness and lowClark level of invasion and tumor regression (7). Therefore, al-though blood vessels are generally believed to promote tumorgrowth, our studies showed that the phenotype of blood vesselsis important and that some types of blood vessels present in thetumor microenvironment (i.e., tumor HEVs) can contribute totumor suppression rather than tumor growth. Tumor HEVs havebeen observed in many different types of human solid tumors,including primary and metastatic melanomas, breast, lung, ovar-ian, and colon carcinomas (3, 7, 10, 11). A better understanding oftumor HEVs could thus have broad applications for human cancer.Interestingly, tumor HEVs have been detected in murine B16

melanoma tumors after targeting of lymphotoxin a (LTa) withinthe tumor (12, 13), and more recently in methylcholanthrene-induced fibrosarcomas upon depletion of regulatory T cells (Tregs)in FoxP3-DTR transgenic mice (14). In both cases, the presence ofHEVs in the tumors was associated with T cell infiltration andtumor regression (12–14). It is currently unknown whether Tregs

*Centre National de la Recherche Scientifique, Institut de Pharmacologie et de Biol-ogie Structurale, F-31077 Toulouse, France; †Universite de Toulouse, Universite PaulSabatier, Institut de Pharmacologie et de Biologie Structurale, F-31077 Toulouse,France; ‡Institut Claudius Regaud, F-31052 Toulouse, France; and xPlastic and Re-constructive Surgery Unit, Centre Hospitalier Universitaire Toulouse Rangueil,F-31059 Toulouse, France

Received for publication April 1, 2013. Accepted for publication June 12, 2013.

This work was supported by the Institut National du Cancer (Grant 2012-039), FondationReseau Innovations Therapeutiques en Cancerologie, Region Midi-Pyrenees, FondationARC pour la Recherche sur le Cancer (Programme ARC number SL220110603471;Equipment ARC number ECL2010R00650), Ligue Nationale contre le Cancer (post-doctoral fellowship to L.M.), and the Association “Le Cancer du Sein, Parlons-en!”

Address correspondence and reprint requests to Dr. Jean-Philippe Girard, Institut dePharmacologie et de Biologie Structurale, Centre National de la Recherche Scienti-fique, Universite de Toulouse, 205 Route de Narbonne, 31077 Toulouse, France.E-mail address: jean-philippe.girard@ipbs.fr

The online version of this article contains supplemental material.

Abbreviations used in this article: CT, cycling threshold; DC, dendritic cell; DCIS,ductal carcinoma in situ; DFS, disease-free survival; HEV, high endothelial venule;ICR, Institut Claudius Regaud; IDC, invasive ductal carcinoma; LTa, lymphotoxin a;LTb, lymphotoxin b; LTbR, lymphotoxin b receptor; OS, overall survival; qRT-PCR,quantitative RT-PCR; Treg, regulatory T cell.

Copyright� 2013 by The American Association of Immunologists, Inc. 0022-1767/13/$16.00

www.jimmunol.org/cgi/doi/10.4049/jimmunol.1300872

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

regulate HEV differentiation directly. In contrast, CD11c+ den-dritic cells (DCs) were recently shown to play a critical role in thisprocess (15). In vivo depletion of DCs in adult mice resulted ina profound downregulation of the HEV phenotype in mouse lymphnodes. Coculture experiments revealed that lymphotoxin b receptor(LTbR) signaling is involved in the dialog between DCs and HEVendothelial cells (15). DCs expressed lymphotoxin ligands forLTbR, and DC-derived lymphotoxin was found to be importantfor HEV-mediated extravasation of lymphocytes in lymph nodes(15). DCs and LTbR signaling have also been implicated in theregulation of HEVs in murine inflamed lymph nodes and chronicallyinflamed nonlymphoid tissues (16–23). Despite these importantadvances in mouse models, little is known yet about the mecha-nisms governing the development of HEVs in human solid tumors.In this study, we investigated the mechanisms regulating HEV

blood vessels in human breast cancer. We show that high densitiesof tumor HEVs in breast tumors were associated with high ex-pression levels of lymphotoxin b (LTb) and high densities of DC-LAMP+ DC clusters. LTb expression was strongly correlated withexpression of DC-LAMP, and CD11c+ DCs were found to be themajor producers of membrane-bound LTb (LTa1b2) in the breasttumor microenvironment. Tumor HEVs were often surrounded byFascin+ and DC-LAMP+ DCs within breast tumor stroma, and thedensity of DC-LAMP+ DC clusters was strongly correlated withthe density of tumor HEVs, T and B cell infiltration, and favorableclinical outcome of breast cancer patients. Strikingly, a progres-sive loss of tumor HEVs and DC-LAMP+ DCs was observed duringbreast cancer progression from in situ to invasive carcinoma. Fi-nally, densities of Tregs and tumor HEVs were correlated in hu-man breast tumors. However, the Tregs/CD3+ T cells ratio wassignificantly reduced in HEVhigh breast tumors.

Materials and MethodsPatients

Fresh, frozen, and paraffin-embedded breast tumor samples were all ob-tained from breast cancer patients undergoing surgery at the InstitutClaudius Regaud (ICR Biological Resource Center, Toulouse, France).Approval of the study was obtained from the Scientific Review Boardof the ICR. A written informed consent was obtained from the patientsbefore inclusion in the study. The retrospective study was conducted witha cohort of 146 unselected primary invasive breast cancer patients operatedat ICR between 1997 and 1998 (3). Patient characteristics have beenpreviously described (3) and are summarized in Supplemental Table I.Experiments also involved tissue samples from 110 patients operated atICR between 2004 and 2012, including invasive ductal carcinomas (IDCs;n = 66), ductal carcinomas in situ (DCISs; n = 29), and nonmalignantbreast tissues (n = 15).

Immunohistochemistry and immunofluorescence staining

Immunohistochemistry was performed on 5-mm consecutive sections fromRCL2, formalin- or Duboscq-fixed, paraffin-embedded tumor blocks usinga Techmate Horizon slide processor (Dako, Trappes, France) as previouslydescribed (3). Details of the Abs, fixatives, and Ag retrieval methods usedare provided in Supplemental Table II. For immunofluorescence detection,tumor slides were incubated with fluorochrome- or biotin-coupled sec-ondary Abs diluted in PBS, BSA 1% for 30 min at room temperature andcounterstained with DAPI. When necessary, AF350-, AF488-, or AF546-coupled streptavidin (Invitrogen) was added for 20 min. Images were ac-quired using a Nikon Eclipse 90i microscope (Nikon, Champigny-sur-Marne,France) operated with Nikon NIS Elements BR software.

Methods for cell quantification

Tumor slides stained with MECA-79, anti-CD3, anti-CD20, anti-DC-LAMP, or anti-Foxp3 Abs were scanned with a high-resolution scanner(NDP slide scanner, Hamamatsu and Panoramic 250 Flash; 3Dhistech). Thedensity of tumor HEVs (HEV/mm2) was calculated and patients witha high and a low density of tumor HEVs were separated with the followingcutoff (0.19 HEV/mm2) as previously described (3). CD3+, CD20+, andFOXP3+ cells densities were evaluated by automatic cell count with

ImageJ software (National Institutes of Health, Bethesda, MD) as de-scribed previously (3). We evaluated the density of DC-LAMP+ infiltratingcells by optical quantification of the number of DC-LAMP+ DC clusterswithin tumor stroma (Supplemental Fig. 1A). An optical grading (grade0, 1, 2, and 3 for none, low, intermediate, and high number of tumor-infiltrating DC-LAMP+ cells) of DC-LAMP marker was also performedto verify the results obtained with DC-LAMP quantification. We obtainedsimilar results with the two different methods; therefore, only the resultsobtained with the quantitative method are presented in this article. We usedthe following cutoff (highest tercile versus two lowest terciles) to discrimi-nate tumors with a high and low density of DC-LAMP+ cells. Quantificationof blood vessels and scoring were performed by three independent observers(L.M., I.G., S.L.G.) who were blinded to the clinical outcome.

Flow cytometry and cell sorting

Freshly resected breast tumor samples were reduced in small fragments andincubated 30 min at 37˚C in sterile RPMI 1640 containing collagenase IV(1 mg/ml; Sigma-Aldrich). The samples were classified “HEVhigh” whenhigh densities of tumor HEVs were detected on adjacent sections by im-munostaining for MECA-79. Total cells were then extracted by mechanicaldispersion and incubated for 30 min at 4˚C with fluorochrome-conjugatedmAbs directed against different immune cell markers or their isotype-matched controls (Supplemental Table II). LTb expression was deter-mined by incubating PBMCs or total cells isolated from HEVhigh breasttumor samples (n = 9) with primary Ab directed against human LTb(MAB1684; R&D, Minneapolis, MN) for 30 min at 4˚C, followed by in-cubation with FITC-conjugated goat anti-mouse IgG secondary Abs(Jackson Immunoresearch, Suffolk, U.K.) for 30 min at 4˚C. Fluorochrome-conjugated mAbs directed against CD3, CD20, CD56, CD11c, and CD45were then added for 30 min at 4˚C. Analyses were performed on a six-colorfluorescence-activated cell sorter (LSRII; Becton Dickinson) with Diva(Becton Dickinson) and FlowJo softwares (Tree Star, Ashland, OR).

For cell sorting, total cells from freshly resected HEVhigh breast tumorsamples (n = 7), extracted as described earlier, were pooled together and

FIGURE 1. Expression of LTb is increased in HEVhigh breast tumors and

highly correlated with expression of HEV-associated chemokines. (A)

Graphs showing the relative mRNA levels of LTbR signaling pathway genes

in HEVhigh (n = 10, black bars) and HEVlow (n = 11, white bars) breast tumor

samples. (B) Graphs showing the correlation between the relative mRNA

levels of LTB (coding for LTb) and HEV-associated chemokines (CCL19,

CCL21, or CXCL13) in breast tumor samples, and the absence of correlation

between expression of LTB and expression of chemokines CCL2, CXCL8, or

CXCL12. *p , 0.05, **p , 0.01, Mann–Whitney U test.

2002 TUMOR HEVs AND DENDRITIC CELLS IN HUMAN BREAST CANCER

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

incubated for 30 min at 4˚C with fluorochrome-conjugated mAbs directedagainst different immune and tumor cell markers. Tumor cells (EpCAM+

CD452), T lymphocytes (CD45+CD3+), B lymphocytes (CD45+CD20+),NK cells (CD32CD56+), and DCs (CD45+, lin2, HLA-DR+, CD11c+,CD1A+) were isolated by flow cytometry using a FACSAria II cellsorter.

Quantitative RT-PCR

An RNeasy isolation kit was used to isolate total RNA (Qiagen, Valencia,CA) from 22 cryopreserved breast tumor samples (12 HEVlow versus 10HEVhigh, identified by immunohistochemistry with MECA-79) and cellsisolated by cell sorting as described earlier. The integrity and the quantityof the RNA were evaluated using a bioanalyzer-2100 (Agilent Technolo-gies, Palo Alto, CA). cDNA was prepared by reverse transcription usingsuperscript VILO cDNA Synthesis Kit (Invitrogen, Paisley, U.K.). Quan-titative RT-PCR (qRT-PCR) experiments were performed using PowerSYBR Green mix with an ABI PRISM 7300HT (Applied Biosystems,Warrington, U.K.) according to manufacturer’s instructions. All reactionswere done in triplicate and normalized to the expression of GAPDH. Heatmap representation of gene expression in tumor tissues containing a high(HEVhigh) or a low density of HEVs (HEVlow) was performed with the useof Genesis software (Institute for Genomics and Bioinformatics, Graz,Austria) (24, 25). The expression of each gene was obtained by the Dcycling threshold (CT) method as 2(2DCTsample), and the relative change inexpression between HEVhigh and HEVlow tumor samples was calculated as22(DCTsample 2 average DCT from HEVlow tumors).

Statistical analysis

Categorical variables were reported by frequencies and percentages; con-tinuous variables were presented by median and range. Comparisons be-tween groups were performed using the Mann–Whitney rank sum test forcontinuous variables and x2 or Fisher exact test for categorical variables.

Correlations between continuous variables were evaluated using Spearmanrank correlation test.

We analyzed two main end points: disease-free survival (DFS), whichwas defined as time from surgery to any recurrence (local or regional ordistant metastasis) or death, and overall survival (OS), which was definedas time from surgery to death from any cause. The Kaplan–Meier product-limit estimator was used to display time-to-event curves for the two endpoints. Comparisons between groups were performed using log rank test.Cox regression model was applied to determine whether a factor was anindependent predictor of survival in multivariate analysis (with backwardvariable elimination). All p values were two sided, and statistical sig-nificance was defined as p , 0.05. Statistical analyses were performedusing the STATA 11.0 (STATA Corp, College Station, TX) software.

ResultsLTb is overexpressed in HEVhigh breast tumors

Previous studies in mouse have demonstrated that maintenance ofHEVs depend on LTbR signaling initiated by its cell-associatedligand LTa1b2, an heterotrimer formed by the membrane-associatedsubunit LTb, and LTa, which is critical for cell-surface expressionof the complex (15, 16, 18). We thus analyzed the mRNA levels ofLTBR and its ligands LTA (encoding LTa), LTB (encoding LTb),and LIGHT (an alternative LTBR ligand) by qRT-PCR in HEVhigh

and HEVlow breast tumor samples. Levels of LTBR and LIGHTmRNAwere not significantly different between HEVhigh and HEVlow

tumors. A small but significant increase in LTA was detected inHEVhigh tumors (Fig. 1A), whereas LTB was specifically andstrongly overexpressed (.10-fold) in HEVhigh breast tumors (Fig.1A). Interestingly, LTB relative expression within breast tumor

FIGURE 2. LTb is produced by DCs

within the breast tumor microenviron-

ment. (A) Representative histograms

showing the staining with anti-LTb Ab

(filled histogram) or isotype control

(open histogram) at the cell surface of

CD452 tumor cells, CD3+ T cells, CD20+

B cells, CD32CD56+ NK cells, and

CD11c+ DCs from HEVhigh breast tu-

mor samples. (B) Mean (+SD) percent-

age of CD3+ T cells, CD20+ B cells,

CD32CD56+ NK cells, and CD11c+

DCs among LTb-expressing cells in

HEVhigh breast tumor samples (n = 9).

(C) Graph showing the mean (+SD)

relative LTB mRNA expression in

Epcam+ tumor cells, CD3+ T cells,

CD20+ B cells, CD11c+CD1A+HLADR+

DCs, and CD32CD56+ NK cells, isolated

from HEVhigh breast tumor samples by

cell sorting (n = 7). Relative expression to

B cells is shown. (D) Graph showing the

correlation between the relative mRNA

expression levels of DC-LAMP and LTB

in breast tumor samples (n = 21). (E)

Graphs showing the correlation between

the relative mRNA expression levels of

DC-LAMP and CCL19, CCL21, or

CXCL13 in breast tumor samples (n =

21), and the absence of correlation

between expression of DC-LAMP and

expression of chemokines CCL2, CXCL8,

or CXCL12. **p , 0.01, ***p , 0.001,

Mann–Whitney U test.

The Journal of Immunology 2003

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

samples was highly correlated to expression of chemokines as-

sociated with HEV-mediated lymphocyte extravasation (4, 10, 14,

26, 27), including CCL21 (Spearman r = 0.78, p , 0.001), CCL19

(Spearman r = 0.91, p, 0.001), and CXCL13 (Spearman r = 0.78,

p , 0.001) (Fig. 1B). In contrast, LTB mRNA levels were not

correlated (p . 0.05) to mRNA levels of chemokines implicated

in myeloid cell, neutrophil, and lymphocyte migration indepen-

dently of HEVs (CCL2, CXCL8, CXCL12; Fig. 1B).

DCs are the major producers of LTb within the breast tumormicroenvironment

To identify the cells that produce membrane-associated LTb(LTa1b2) within breast tumors, we used flow cytometry to ana-lyze cell-surface expression of LTb on cells isolated from freshlyresected breast tumor samples containing high densities of tumorHEVs and large numbers of tumor-infiltrating lymphocytes (HEVhigh

breast tumors). Control experiments were performed on bloodmononuclear cells from healthy donors stimulated or not with anti-CD3/anti-CD28–coated beads to induce LTb expression (Supple-mental Fig. 2). In HEVhigh breast tumors, ∼30% of CD11c+ DCsexpressed membrane-bound LTb (Fig. 2A, 2B). In contrast, veryfew CD452 tumor cells, CD32CD56+ NK cells, CD3+ T cells, andCD20+ B cells had detectable LTb expression. To confirm theseresults, we used qRT-PCR to analyze LTb mRNA expression intumor cells, NK cells, T cells, B cells, and DCs isolated fromHEVhigh breast tumor samples by cell sorting (Supplemental Fig. 2).We found that LTB mRNA levels were significantly higher in DCsthan in other immune cell populations (Fig. 2C) and were stronglycorrelated to the levels of mature DC marker DC-LAMP in breasttumor samples (Fig. 2D). Similarly to LTB, DC-LAMP expressionin breast tumor samples was correlated (p, 0.001) to expression ofHEV-associated chemokines CCL19, CCL21, and CXCL13, but not

FIGURE 3. Tumor HEVs are as-

sociated with DC-LAMP+ DCs clus-

ters in human breast tumors. (A) DC-

associated genes are upregulated in

HEVhigh breast tumors. Expression

levels of the indicated genes were

determined by quantitative PCR and

compared between 22 human breast

tumor samples (10 HEVhigh versus

12 HEVlow). Heat map representa-

tions of the DC and “angiogenesis”

gene clusters are shown. Genes are

plotted from the minimal level of

expression (green) to the maximal

level (red). The Mann–Whitney U

test was used to compare the expres-

sion levels of each gene between

tumor groups. n.s., p . 0.05. (B and

C) Immunofluorescence staining of

breast tumor sections with Abs against

HEV (MECA-79) and mature DC

markers Fascin (B) and DC-LAMP

(C). DNAwas stained with DAPI (B)

and T cells with an anti-CD3 Ab (C).

Original magnifications 3100. (D)

Immunohistochemical detection of

DC-LAMP in breast tumors con-

taining either low (DC-LAMPlow) or

high (DC-LAMPhigh) densities of

DC-LAMP+ cells. (E and F) Serial

breast tumor sections (n = 146) were

stained with Abs directed against DC-

LAMP or HEV (MECA-79), and the

density of the two cell types was

calculated as described in Materials

and Methods. Representative pictures

of the spatial association between

HEVs and DC-LAMP+ DCs clusters

are shown (E). The density of tumor

HEVs is significantly higher in breast

tumors with a high density of DC-

LAMP+ DC clusters (F). Original

magnification (D) 320; (E) 310.

***p, 0.001, Mann–Whitney U test.

2004 TUMOR HEVs AND DENDRITIC CELLS IN HUMAN BREAST CANCER

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

(p. 0.05) to that of chemokines CCL2, CXCL8, and CXCL12 (Fig.2E). Together, these results indicated that CD11c+ DCs representthe main source of membrane-associated LTb within human breasttumors.

Tumor HEVs are associated with DC-LAMP+ DCs withinbreast tumor microenvironment

When we compared the expression of genes related to immunesubpopulations and angiogenesis, by qRT-PCR in cryopreservedbreast tumor samples containing either low (HEVlow) or high(HEVhigh) densities of tumor HEVs (12HEVlow versus 10HEVhigh),we observed that the “DCs” cluster was significantly upregulated inHEVhigh breast tumors (Fig. 3A). In contrast, the “angiogenesis”cluster was not significantly different between tumors with a lowor a high density of tumor HEVs (Fig. 3A). To further define theassociation between tumor HEVs and DCs within the breast tu-mor microenvironment, tissue sections were double stained withthe HEV-specific mAbMECA-79 and Abs against two markers ofmature DCs, Fascin and DC-LAMP. These immunofluorescenceanalyses revealed that tumor HEVs were often surrounded bymature Fascin+ DCs (Fig. 3B) and DC-LAMP+ DCs (Fig. 3C).

We then analyzed DC-LAMP and MECA-79 markers by immu-nohistochemistry in a retrospective cohort of 146 primary inva-sive breast cancer patients, previously used to study the impact oftumor HEVs on clinical outcome (3). The density of DC-LAMP+

cells was variable within breast tumor microenvironment, allowingus to define tumors containing either low (DC-LAMPlow) or high(DC-LAMPhigh) densities of DC-LAMP+ cells (Fig. 3D). TumorHEVs were mainly found in DC-LAMP+ cell–rich areas within tu-mor stroma (Fig. 3E), and significantly higher densities of tumorHEVs (Fig. 3F) were observed in DC-LAMPhigh breast tumors. Weconcluded that tumor HEVs were associated with DC-LAMP+ DCsclusters within breast tumor stroma.

DC-LAMP+ DCs correlate with density of tumorHEVs, lymphocyte infiltration, and favorable clinical outcomeof breast cancer patients

Significantly higher densities of CD3+ T cells and CD20+

B cells were found in breast tumors containing high densitiesof DC-LAMP+ DC clusters (Fig. 4A, Supplemental Fig. 1B).Indeed, the density of DC-LAMP+ DC clusters was stronglycorrelated with the density of tumor HEVs (Spearman r = 0.63,

FIGURE 4. Density of DC-LAMP+ DCs

correlates with HEV density, lymphocyte infil-

tration, and favorable clinical outcome of breast

cancer patients. (A) Representative images from

immunohistochemistry with DC-LAMP, MECA-

79, CD3, and CD20 Abs showing the spatial as-

sociation between DCs, HEVs, T cells, and B

cells within breast tumor stroma. Original mag-

nification 310. (B) The density of DC-LAMP+

DC clusters is correlated with the density of

HEVs, CD3+ T cells, and CD20+ B cells in breast

tumor samples. Serial breast tumor sections (n =

146) were stained with Abs directed against DC-

LAMP, HEV (MECA-79), CD3, and CD20, and

the density of the different cell types was calcu-

lated as described in the Materials and Methods.

(C) Kaplan–Meier curves for DFS and OS rates of

146 patients with primary breast cancer, accord-

ing to the density of DC-LAMP+ DCs. The rapid

decrease in the DC-LAMPhigh DFS curve after

144 mo is due to the small number of patients at

risk at this time.

Table I. Multivariate analysis using Cox proportional hazard model

DFS OS

HR 95% CI p (Wald) HR 95% CI p (Wald)

DC-LampLow 3.56 1.42–8.94 0.007 4.14 1.28–13.32 0.017High 1 1

Tumor size, cm,2 1 1$2 3.32 1.82–6.06 ,0.001 3.18 1.58–6.38 0.001

Nodal status (N)N2 1 1N+ 2.15 1.22–3.79 0.005 2.38 1.23–4.61 0.010

HR, Hazard ratio.

The Journal of Immunology 2005

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

p , 0.001), CD3+ T cells (Spearman r = 0.72, p , 0.001), andCD20+ B cells (Spearman r = 0.58, p , 0.001; Fig. 4B), con-sistent with the link between tumor HEVs and tumor-infiltratinglymphocytes (3).We then analyzed the prognostic value of DC-LAMP marker in

the retrospective cohort. In univariate analysis, we observed thata high density of DC-LAMP+ cells was significantly associatedwith a longer DFS (p , 0.02) and OS (p , 0.02) as comparedwith tumors with a low density of DC-LAMP+ cells (Fig. 4C,Supplemental Table I). DC-LAMP density also showed a signifi-cant correlation with DFS and OS in multivariate analysis afteradjusting on prognostic factors previously identified (tumor size,grade, estrogen receptor status, HER2 expression). The adjustedhazard ratios of DFS and OS rates for patients with DC-LAMPlow

tumors versus DC-LAMPhigh tumors were 3.56 (95% CI, 1.42–8.94, p = 0.007) and 4.14 (95% CI, 1.28–13.32, p = 0.017), re-spectively (Table I). We concluded that a high density of DC-LAMP+ DCs correlates with the presence of HEV blood vesselswithin breast tumors and constitutes an independent factor of goodclinical outcome for breast cancer patients.

Progressive loss of DC-LAMP+ DCs and tumor HEVs duringbreast cancer progression

The tumorigenesis and progression of breast cancer constitutea multistep process that is influenced by many factors includinggenetic composition, age, hormonal status, and immune envi-ronment. IDC is thought to derive from a series of intermediatehyperplasic and neoplastic stages including DCIS. To betterunderstand the role of tumor HEVs during breast tumor devel-opment, we performed MECA-79 immunohistochemistry on 15nonmalignant breast tissues, 29 pure DCISs, and 173 IDCs (Fig.5A). HEVs were present in breast tumor tissues, but not innormal breast tissue samples (Fig. 5A), and their density was sig-nificantly higher in DCIS (7.467 6 1.324) than in IDC (0.6363 60.1132; Fig. 5B). To evaluate the role of tumor HEVs in the pro-gression of DCIS to invasive carcinomas, we compared the densityof MECA-79+ HEVs in matched IDC and DCIS components of 31cases of invasive carcinomas with a DCIS component (Fig. 5C).We found a preferential localization of HEVs around in situ tumorcomponents (Fig. 5C). A strong reduction in the density of tumorHEVs between DCIS components and IDC components was ob-served for all tumor samples analyzed (Fig. 5D). Given the clearassociation between DCs and HEVs in peripheral lymph nodes(15) and breast tumor tissues (this study), we then analyzed thedensity of DC-LAMP+ DCs in invasive carcinomas containingmatched invasive and DCIS components (Fig. 5E). Similar to thedensity of tumor HEVs, we found that the density of DC-LAMP+

DCs was significantly reduced between DCIS and IDC compo-nents (Fig. 5F). Finally, we analyzed the density of CD3+ T cellsin matched DCIS and IDC tumor areas (Fig. 5G), and we foundthat it was significantly reduced in IDC compared with DCIS areas(Fig. 5H). Together, these results indicate that breast cancer inva-siveness is associated with a progressive loss of HEV blood vessels,DC-LAMP+ DCs, and CD3+ T cells around DCIS structures.

Human breast tumor HEVs develop in the presence of FOXP3+

Tregs

Tumor HEVs have recently been found to be induced in mousetumor models upon depletion of Tregs (14). To better define thelink between Tregs and HEV blood vessels within breast tumorstroma, we analyzed MECA-79, FOXP3, and CD3 markers byimmunohistochemistry for 40 invasive breast cancer patients (Fig.6A). We found a significant increase in the density of tumor-infiltrating FOXP3+ Tregs (Fig. 6B) and CD3+ T cells (Fig. 6C)

in tumors containing a high density of HEV blood vessels (p ,0.01), and a significant correlation between the density of FOXP3+

Tregs and tumor HEVs (Spearman r = 0.54, p , 0.003). However,we also observed a significant reduction in the FOXP3+ Tregs/CD3+ T cell ratio in tumors with a high density of tumor HEVs(p , 0.01; Fig. 6D). These findings demonstrate that HEV bloodvessels can fully develop in the presence of Tregs within breasttumor stroma, but that their density is influenced by the Tregs/T cells ratio.

DiscussionThe current dogma in the field of tumor angiogenesis is that bloodvessels contribute to tumor growth, and they are thus generally

FIGURE 5. Loss of tumor HEV and DC-LAMP+ DC clusters during

breast cancer progression from in situ to IDC. (A) Representative images

from digitized breast tissue slides stained with MECA-79 Ab showing that

tumor HEVs are more frequently observed around DCISs than IDCs, and

are absent from normal breast tissues. (B) Graph showing the density of

MECA-79+ HEVs in normal breast tissue (n = 15), DCIS (n = 29), and IDC

(n = 173). (C–H) Serial breast tumor sections containing both in situ and

invasive components were stained with Abs directed against HEV (MECA-

79), DC-LAMP, and T cells (CD3), and the density of the different cell

types was calculated in matched in situ and invasive components. Repre-

sentative images of the preferential location of MECA-79+ HEV blood

vessels (C, box), DC-LAMP+ DCs clusters (E), and CD3+ T cells (G) in

DCIS areas and their relative absence in IDC areas are shown (C, E, G).

The density of tumor HEVs (D, n = 31), DC-LAMP+ DCs clusters (F, n =

31), and CD3+ T cells (H, n = 16) was decreased between in situ (DCIS)

and invasive (IDC) matched samples. Original magnification (A) 35, (C)

32.5, (E)35, (G)32.5. **p, 0.01, ***p, 0.001, Mann–Whitney U test.

2006 TUMOR HEVs AND DENDRITIC CELLS IN HUMAN BREAST CANCER

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

associated with poor prognosis. Recently, we proposed the novelconcept that “tumor blood vessels are not all the same and thatsome types of blood vessels found in the tumor microenvironment(i.e., tumor HEVs) can be associated with favorable clinical out-come” (3). Tumor HEVs were frequently observed in the stromaof human solid tumors and appeared to contribute to tumor sup-pression by allowing high levels of lymphocyte infiltration (in-cluding CD8+ cytotoxic T cells infiltration) into tumors (3, 7). Abetter understanding of tumor HEVs and their regulation is im-portant because it could allow, in the future, the manipulation oftumor blood vessel phenotype to transform regular tumor bloodvessels into tumor HEVs.The lymphotoxin pathway has been shown to regulate HEV

blood vessels in both lymphoid organs and chronically inflamedtissues (16, 18–20, 22, 23). We demonstrated previously that DCsand DC-derived lymphotoxin are critical for maintenance of HEVblood vessel phenotype in mouse lymph nodes (4, 15). In thisstudy, we asked whether similar mechanisms regulate tumor HEVsin human breast tumors. We found that LTa and LTb were bothoverexpressed in HEVhigh tumors, and DCs were the major pro-ducers of membrane-associated LTb (LTa1b2) within breasttumors. Whereas most T and B cells expressed LTb after poly-clonal activation of PBMCs, we observed only a few T andB lymphocytes expressing LTb in breast tumors. In contrast,a considerable fraction (∼30%) of CD11c+ DCs from HEVhigh

breast tumors expressed membrane-bound LTb. In addition, DC-LAMP expression was strongly correlated to LTb expression inbreast tumor samples. These results suggested that DCs maycontribute to the formation of HEVs in human breast tumorsthrough LTb production. In support of this possibility, tumorHEVs were often surrounded by clusters of mature Fascin+ andDC-LAMP+ DCs, and high densities of DC-LAMP+ DCs clusterswere observed in HEVhigh breast tumors. The density of DC-LAMP+ DCs clusters was strongly correlated with the densityof tumor HEVs, T and B cell infiltration, and favorable clinicaloutcome (longer DFS and OS) in a retrospective cohort of 146primary invasive breast cancer patients. Together, these resultsindicated that, similar to mouse lymph nodes (15), DCs and

lymphotoxin pathway may be critical regulators of HEV bloodvessels in human breast tumors.We observed that densities of DC-LAMP+ DCs and tumor HEVs

were significantly higher in DCIS than in IDC. Interestingly, ina series of 31 invasive breast carcinomas containing matched DCISand IDC components, the density of tumor HEVs and DC-LAMP+

DCs was strongly reduced between DCIS and IDC components.These observations suggest that loss of tumor HEVs and DC-LAMP+ DCs during breast cancer progression may representa critical step in the transition from in situ carcinoma to invasivecarcinoma.Although our results strongly suggest that DCs foster HEV

development and/or maintenance within human breast tumors, wecannot exclude the alternative possibilities that DCs arriving viatissue lymphatics may preferentially be attracted to pre-existingHEVs or that DCs may enter the stroma via tumor HEVs. Func-tional analyses in mouse tumor models will be required to dem-onstrate the direct role of DCs in the induction and/or maintenanceof tumor HEVs.Surprisingly, increased densities of Tregs were observed in

breast tumors containing high densities of tumor HEVs. In contrast,it was previously reported that in mouse methylcholanthrene-inducedfibrosarcomas, formation of tumor HEVs only occurs after de-pletion of Tregs (14). Our results indicate that the absence of Tregsis not an essential prerequisite for development of HEVs in humanbreast tumors. However, the relative proportion of Tregs comparedwith other T cell populations appears to be important becauseHEVhigh breast tumors were associated with low FOXP3+ Tregs/CD3+ T cell ratios. Therefore, Tregs may limit HEV neogenesis inhuman breast tumors, similar to mouse tumor models (14). LTaand LTbwere upregulated in mouse fibrosarcomas upon depletionof Tregs (14), suggesting that Tregs may inhibit HEV develop-ment through reduction of LTa1b2 levels.Interestingly, targeting of LTa to mouse melanomas tumors has

been shown to result in formation of tumor HEVs, T cell infil-tration, and eradication of the tumors (12, 13). Overexpression ofLIGHT, the second ligand for LTbR, in mouse tumors has alsobeen shown to induce antitumor immunity and eradication of the

FIGURE 6. Tumor HEVs develop in the presence of

FOXP3+ Tregs. (A–D) Serial breast tumor sections (n =

40) were stained with Abs directed against HEV

(MECA-79), CD3, and FOXP3, and the density of the

different cell types was calculated. Representative

pictures of MECA-79+ HEVs and CD3+ and FOXP3+

T cells in HEVlow and HEVhigh breast tumors are

shown Original magnification 35. (A). The density of

FOXP3+ T cells (B) and CD3+ T cells (C) is increased

in tumors containing a high density of tumor HEVs.

The FOXP3+/CD3+ T cells ratio is decreased in

HEVhigh breast tumors (D). **p , 0.01, ***p , 0.001,

Mann–Whitney U test.

The Journal of Immunology 2007

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from

tumors (28, 29). However, it remains unclear whether tumor HEVswere generated upon overexpression of LIGHT. Although DCs inlymph nodes express both LTa/LTb and LIGHT (15), it is un-known yet whether LIGHT plays a role in the regulation of HEVs.In this study, we observed that LIGHT was not overexpressedin HEVhigh breast tumors, indicating that LTa1b2, rather thanLIGHT, is likely to be the major LTbR ligand involved in thegeneration of human breast tumor HEVs.Further characterization of the cellular and molecular mecha-

nisms regulating the formation of tumor HEVs (including the rolesof DCs and lymphotoxin) will be important because it could lead tothe development of improved therapeutic strategies for human solidtumors. For instance, it could provide means to induce the HEVendothelial cell differentiation program in tumor blood vessels.This would increase the density of tumor HEVs without increasingtumor angiogenesis (total number of tumor blood vessels) andwould suppress tumor growth through enhanced recruitment ofcytotoxic lymphocytes. Novel therapeutic strategies based on themodulation of tumor HEVs could thus have a major impact ontumor growth and clinical outcome of cancer patients.

AcknowledgmentsWe are grateful to Jean-Jacques Fournie and members of the Girard team

and ICR Anatomopathology Service for help. We thank the Biological

Resource Center of the ICR for providing breast tumor samples. We thank

Renaud Poincloux for designing the lymphocyte count macro and the

IBiSA TRI facility for flow cytometry and imaging.

DisclosuresThe authors have no financial conflicts of interest.

References1. Carmeliet, P. 2005. Angiogenesis in life, disease and medicine. Nature 438: 932–

936.2. Carmeliet, P., and R. K. Jain. 2000. Angiogenesis in cancer and other diseases.

Nature 407: 249–257.3. Martinet, L., I. Garrido, T. Filleron, S. Le Guellec, E. Bellard, J. J. Fournie,

P. Rochaix, and J. P. Girard. 2011. Human solid tumors contain high endothelialvenules: association with T- and B-lymphocyte infiltration and favorable prog-nosis in breast cancer. Cancer Res. 71: 5678–5687.

4. Girard, J. P., C. Moussion, and R. Forster. 2012. HEVs, lymphatics andhomeostatic immune cell trafficking in lymph nodes. Nat. Rev. Immunol. 12:762–773.

5. von Andrian, U. H., and T. R. Mempel. 2003. Homing and cellular traffic inlymph nodes. Nat. Rev. Immunol. 3: 867–878.

6. Miyasaka, M., and T. Tanaka. 2004. Lymphocyte trafficking across high endo-thelial venules: dogmas and enigmas. Nat. Rev. Immunol. 4: 360–370.

7. Martinet, L., S. Le Guellec, T. Filleron, L. Lamant, N. Meyer, P. Rochaix,I. Garrido, and J. P. Girard. 2012. High endothelial venules (HEVs) in humanmelanoma lesions: Major gateways for tumor-infiltrating lymphocytes.OncoImmunology 1: 829–839.

8. Martinet, L., I. Garrido, and J. P. Girard. 2012. Tumor high endothelial venules(HEVs) predict lymphocyte infiltration and favorable prognosis in breast cancer.OncoImmunology 1: 789–790.

9. Fridman, W. H., J. Galon, F. Pages, E. Tartour, C. Sautes-Fridman, andG. Kroemer. 2011. Prognostic and predictive impact of intra- and peritumoralimmune infiltrates. Cancer Res. 71: 5601–5605.

10. de Chaisemartin, L., J. Goc, D. Damotte, P. Validire, P. Magdeleinat, M. Alifano,I. Cremer, W. H. Fridman, C. Sautes-Fridman, and M. C. Dieu-Nosjean. 2011.Characterization of chemokines and adhesion molecules associated with T cellpresence in tertiary lymphoid structures in human lung cancer. Cancer Res. 71:6391–6399.

11. Cipponi, A., M. Mercier, T. Seremet, J. F. Baurain, I. Theate, J. van den Oord,M. Stas, T. Boon, P. G. Coulie, and N. van Baren. 2012. Neogenesis of lymphoidstructures and antibody responses occur in human melanoma metastases. CancerRes. 72: 3997–4007.

12. Schrama, D., P. thor Straten, W. H. Fischer, A. D. McLellan, E. B. Brocker,R. A. Reisfeld, and J. C. Becker. 2001. Targeting of lymphotoxin-alpha to thetumor elicits an efficient immune response associated with induction of pe-ripheral lymphoid-like tissue. Immunity 14: 111–121.

13. Schrama, D., H. Voigt, A. O. Eggert, R. Xiang, H. Zhou, T. N. Schumacher,M. H. Andersen, P. thor Straten, R. A. Reisfeld, and J. C. Becker. 2008. Im-munological tumor destruction in a murine melanoma model by targeted LTal-pha independent of secondary lymphoid tissue. Cancer Immunol. Immunother.57: 85–95.

14. Hindley, J. P., E. Jones, K. Smart, H. Bridgeman, S. N. Lauder, B. Ondondo,S. Cutting, K. Ladell, K. K. Wynn, D. Withers, et al. 2012. T-cell traffickingfacilitated by high endothelial venules is required for tumor control after regu-latory T-cell depletion. Cancer Res. 72: 5473–5482.

15. Moussion, C., and J. P. Girard. 2011. Dendritic cells control lymphocyte entry tolymph nodes through high endothelial venules. Nature 479: 542–546.

16. Browning, J. L., N. Allaire, A. Ngam-Ek, E. Notidis, J. Hunt, S. Perrin, andR. A. Fava. 2005. Lymphotoxin-beta receptor signaling is required for the ho-meostatic control of HEV differentiation and function. Immunity 23: 539–550.

17. Liao, S., and N. H. Ruddle. 2006. Synchrony of high endothelial venules andlymphatic vessels revealed by immunization. J. Immunol. 177: 3369–3379.

18. Drayton, D. L., X. Ying, J. Lee, W. Lesslauer, and N. H. Ruddle. 2003. EctopicLT alpha beta directs lymphoid organ neogenesis with concomitant expression ofperipheral node addressin and a HEV-restricted sulfotransferase. J. Exp. Med.197: 1153–1163.

19. Drayton, D. L., S. Liao, R. H. Mounzer, and N. H. Ruddle. 2006. Lymphoidorgan development: from ontogeny to neogenesis. Nat. Immunol. 7: 344–353.

20. Marinkovic, T., A. Garin, Y. Yokota, Y. X. Fu, N. H. Ruddle, G. C. Furtado, andS. A. Lira. 2006. Interaction of mature CD3+CD4+ T cells with dendritic cellstriggers the development of tertiary lymphoid structures in the thyroid. J. Clin.Invest. 116: 2622–2632.

21. Webster, B., E. H. Ekland, L. M. Agle, S. Chyou, R. Ruggieri, and T. T. Lu.2006. Regulation of lymph node vascular growth by dendritic cells. J. Exp. Med.203: 1903–1913.

22. Grabner, R., K. Lotzer, S. Dopping, M. Hildner, D. Radke, M. Beer,R. Spanbroek, B. Lippert, C. A. Reardon, G. S. Getz, et al. 2009. Lymphotoxinbeta receptor signaling promotes tertiary lymphoid organogenesis in the aortaadventitia of aged ApoE-/- mice. J. Exp. Med. 206: 233–248.

23. Motallebzadeh, R., S. Rehakova, T. M. Conlon, T. S. Win, C. J. Callaghan,M. Goddard, E. M. Bolton, N. H. Ruddle, J. A. Bradley, and G. J. Pettigrew.2012. Blocking lymphotoxin signaling abrogates the development of ectopiclymphoid tissue within cardiac allografts and inhibits effector antibody respon-ses. FASEB J. 26: 51–62.

24. Galon, J., A. Costes, F. Sanchez-Cabo, A. Kirilovsky, B. Mlecnik, C. Lagorce-Pages, M. Tosolini, M. Camus, A. Berger, P. Wind, et al. 2006. Type, density, andlocation of immune cells within human colorectal tumors predict clinical out-come. Science 313: 1960–1964.

25. Pages, F., A. Kirilovsky, B. Mlecnik, M. Asslaber, M. Tosolini, G. Bindea,C. Lagorce, P. Wind, F. Marliot, P. Bruneval, et al. 2009. In situ cytotoxic andmemory T cells predict outcome in patients with early-stage colorectal cancer. J.Clin. Oncol. 27: 5944–5951.

26. Cyster, J. G. 2005. Chemokines, sphingosine-1-phosphate, and cell migration insecondary lymphoid organs. Annu. Rev. Immunol. 23: 127–159.

27. Forster, R., A. C. Davalos-Misslitz, and A. Rot. 2008. CCR7 and its ligands:balancing immunity and tolerance. Nat. Rev. Immunol. 8: 362–371.

28. Yu, P., Y. Lee, W. Liu, R. K. Chin, J. Wang, Y. Wang, A. Schietinger, M. Philip,H. Schreiber, and Y. X. Fu. 2004. Priming of naive T cells inside tumors leads toeradication of established tumors. Nat. Immunol. 5: 141–149.

29. Yu, P., Y. Lee, Y. Wang, X. Liu, S. Auh, T. F. Gajewski, H. Schreiber, Z. You,C. Kaynor, X. Wang, and Y. X. Fu. 2007. Targeting the primary tumor to gen-erate CTL for the effective eradication of spontaneous metastases. J. Immunol.179: 1960–1968.

2008 TUMOR HEVs AND DENDRITIC CELLS IN HUMAN BREAST CANCER

by Diane R

azanajaona-Doll on M

arch 20, 2014http://w

ww

.jimm

unol.org/D

ownloaded from