Content uploaded by Emilie Mamessier
Author content
All content in this area was uploaded by Emilie Mamessier
Content may be subject to copyright.
Research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3609
Human breast cancer cells enhance
self tolerance by promoting evasion
from NK cell antitumor immunity
Emilie Mamessier,1 Aude Sylvain,1 Marie-Laure Thibult,1 Gilles Houvenaeghel,2
Jocelyne Jacquemier,2 Rémy Castellano,1 Anthony Gonçalves,2 Pascale André,3
François Romagné,3 Gilles Thibault,4 Patrice Viens,2 Daniel Birnbaum,1,2
François Bertucci,1,2 Alessandro Moretta,5 and Daniel Olive1,2
1Centre de Recherche en Cancérologie de Marseille, INSERM UMR U891, Marseille, France. 2Institut Paoli-Calmettes, Marseille, France.
3Innate Pharma, Route de Luminy, Marseille, France. 4UMR CNRS 6239 GICC Immuno-Pharmaco-Génétique des Anticorps thérapeutiques,
Université François Rabelais de Tours, Tours, France. 5Dipartimento di Medicina Sperimentale (DIMES), University di Genova, Genoa, Italy.
NK cells are a major component of the antitumor immune response and are involved in controlling tumor
progression and metastases in animal models. Here, we show that dysfunction of these cells accompanies
human breast tumor progression. We characterized human peripheral blood NK (p-NK) cells and malignant
mammary tumor-infiltrating NK (Ti-NK) cells from patients with noninvasive and invasive breast cancers.
NK cells isolated from the peripheral blood of healthy donors and normal breast tissue were used as controls.
With disease progression, we found that expression of activating NK cell receptors (such as NKp30, NKG2D,
DNAM-1, and CD16) decreased while expression of inhibitory receptors (such as NKG2A) increased and that
this correlated with decreased NK cell function, most notably cytotoxicity. Importantly, Ti-NK cells had more
pronounced impairment of their cytotoxic potential than p-NK cells. We also identified several stroma-derived
factors, including TGF-β1, involved in tumor-induced reduction of normal NK cell function. Our data there-
fore show that breast tumor progression involves NK cell dysfunction and that breast tumors model their envi-
ronment to evade NK cell antitumor immunity. This highlights the importance of developing future therapies
able to restore NK cell cytotoxicity to limit/prevent tumor escape from antitumor immunity.
Introduction
Breast cancer (BC) is the primary cause of cancer deaths in women.
The main cause of this mortality is the metastatic spread to other
organs (1). Metastasis occurs when tumor cells acquire invasive
features (2) and the ability to escape from antitumor immunity (3,
4). Defects in antitumor immunity may also facilitate BC occur-
rence. Indeed, mice deficient in IFN-γ production spontaneously
develop mammary tumors (5). Breast tumor cells transplanted into
NOD/SCID mice (which lack adaptive immunity) form noninvasive
tumors, whereas the same cells transplanted into NOD/SCID/γ-cnull
mice (no adaptive immunity and no NK cells) form invasive tumors
that metastasize rapidly (6). This effect is strictly dependent on NK
cells (7). Similarly, in a highly metastatic model, BC metastasized to
the lung only after elimination of NK cells by Tregs (8).
Advanced BC patients show defects in antitumor immunity, such
as alterations of DC maturation (9) and an increase in Treg infil-
trates (10). Major impairment of peripheral blood NK cell matura-
tion and cytotoxic functions has also been reported in metastatic
BC (11). Several gene expression profiling studies have shown that
a better outcome is associated with a strong cytotoxic infiltrate
containing NK cells (12–15). These data suggest that BC progres-
sion is linked to antitumor immunity efficiency and particularly
to NK cells. However, the precise relationships between NK cells
and BC progression in humans have not been studied so far.
NK cells are innate immune cells that have the natural abil-
ity to distinguish normal cells from “modified” cancer cells
(16). Once activated, NK cells eliminate their target through
the release of cytotoxic enzymes (perforin 1, granzymes, granu-
lysin) and/or soluble factors (chemokines and inflammatory
cytokines), which, in turn, recruit and/or activate other effectors
(17). Activating and inhibitory receptors present on NK cells are
triggered during target cell recognition and induce a positive
or a negative cell signaling pathway, respectively. The integra-
tion of these opposite signals determines NK cell activation
(18). The main activating receptors or coreceptors of NK cells
are NKG2D, the natural cytotoxicity receptors (NCRs) NKp30
and NKp46, DNAM-1, CD2, NKp80, 2B4, and NTBA (19–21).
These molecules recognize various ligands usually upregulated
upon cellular stress (22). NK cells also express the Fc immuno-
globulin fragment low-affinity receptor or CD16, which, when
cross-linked, induces a powerful reaction called antibody-depen-
dent cellular cytotoxicity (ADCC). Inhibitory receptors include
the killer immunoglobulin receptors (KIRs), NKG2A, CD85j,
and LAIRs (23–26). They are specific for different HLA–class I
molecules. Accordingly, NK cells can kill target cells that have
lost (or express low amounts of) HLA–class I molecules, which
is frequently the case for tumor cells, including breast tumor
cells (27). However, tumor cells also have the ability to impair
NK cell “visibility” through the modulation of their receptors
(28, 29). Recent studies have shown that several molecules, nota-
bly inhibitory factors often found in the tumor microenviron-
ment, such as IDO1 and TGF-β1, can sharply impair NK cells’
phenotype and functions (30, 31).
Conflict of interest: Pascale André and François Romagné are full-time employees at
Innate Pharma, a biopharmaceutical company, and Alessandro Moretta is a founder
of and a shareholder in Innate Pharma.
Citation for this article: J Clin Invest. 2011;121(9):3609–3622. doi:10.1172/JCI45816.
research article
3610 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
We show here that mechanisms of escape from NK cell–mediat-
ed immunity are at play in BC patients. In a cohort of BC patients
sampled at different stages of the malignant process, we found
that breast tumors have altered NK cell phenotype and function
and that invasive tumors build strong inhibitory microenviron-
ments to escape NK cell antitumor immunity.
Results
p-NK cell phenotype is altered in invasive BC patients
We prospectively enrolled patients with different stages of BC at
diagnosis. Based on tumor pathologic tumor-node-metastasis
(pTNM) classification and tumor margins, groups were composed
of noninvasive (in situ) BCs (Tis) (n = 8) and invasive BCs (n = 113)
including localized (LOC) (n = 55), locally advanced (LA) (n = 26),
and metastatic stages (M) (n = 32). This classification, proposed by
the American Joint Committee on Cancer (AJCC), is based on the
histoclinical extension of disease, which represents the most impor-
tant prognostic factor of BC, with decreasing survival from the Tis
group to the M group (32). Clinical characteristics of the patients
are summarized in Table 1. Due to the prospective nature of our
study, the clinical follow-up of the patients was too short for sur-
vival analysis. Patients with benign mammary tumors (B) (n = 19)
and healthy donors (HD) (n = 22) were included as control groups.
To explore NK cell cytotoxic potential, we first determined the
expression of 22 NK cell receptors on cells from fresh blood sam-
ples. The MFI of NKp30, NKG2D, DNAM-1, and 2B4 were lower
in patients with invasive breast tumors (LOC, LA, and M) than in
control groups and patients with Tis tumors (Figure 1, A–D). The
expression of CD16 was lower in patients with invasive BC than in
control groups (Figure 1E). The expression of other activating mol-
ecules did not vary with tumor progression (Figure 1F). In contrast,
2 inhibitory receptors, NKG2A and CD85j, were upregulated in the
M group compared with control groups and the Tis group (Figure
1, G and H). Histograms are shown in Supplemental Figure 1A (sup-
plemental material available online with this article; doi:10.1172/
JCI45816DS1). We submitted the normalized expression values of
the 22 receptors to TMEV software analysis based on unsupervised
hierarchical clustering. The clustering emphasized the similarities
between the control groups and Tis tumors on one side and the 3
invasive tumor groups on the other side (Figure 1I). The degree of
alterations clearly progressed with tumor stage, notably in the inva-
sive tumors, with underexpression of activating receptors and over-
expression of inhibitory receptors in more advanced stages.
p-NK cell functions are altered in patients with invasive BC
We next determined whether these modifications of expres-
sion were associated with altered NK cell functions. We assessed
peripheral blood NK (p-NK) cells’ functionality through their
ability to kill a target cell (cytotoxic activity), to degranulate
(CD107 positivity, associated with both helper and cytotoxic
functions), to produce IFN-γ and TNF-α (helper functions), and
to mediate ADCC. We found that effective killing (Figure 2A) and
degranulation efficiency (Figure 2B) were altered in the invasive
groups compared with the noninvasive ones. IFN-γ and TNF-α
levels were also reduced in the M group compared with the nonin-
vasive groups (Figure 2, C and D). Highly responsive NK cells can
exert multiple functions with increased potency compared with
monofunctional cells (33). We used the Boolean gating strategy
to identify this multifunctional potential of NK cells. p-NK cells
from the M group were mostly monofunctional (either CD107
or IFN-γ or TNF-α), whereas p-NK cells from the other groups
Table 1
Clinical features of BC patients
Groups B Tis LOC LA M
n = 19 8 55 26 32
Mean age (years) 56 51 57 52 52
Invasive margin No No Yes Yes Yes
Sample uptake Before surgery Before surgery Before surgery Before surgery Before therapy
Scarff-Bloom-Richardson (SBR) Grade I ND ND 37.0% 3.8%
II ND ND 38.8% 34.6%
III ND ND 24.0% 53.8%
pN+ No No 34.0% 81.0% 100.0%
Median pT (mm) ND 20.2 26.4 57.1 50.0
Perivascular invasion No No 16.6% 26.9% 9.3%
Lympho-plasmocytic stroma Absent ND ND 37.5% 47.7% ND
Moderate ND ND 56.2% 38.1% ND
High ND ND 6.3% 14.2% ND
IHC expression PR+ ND ND 62.9% 53.8% 46.8%
ER+ ND ND 75.9% 69.2% 40.6%
erbB-2+ ND ND 7.4% 15.3% 43.7%
HR–erbB-2– ND ND 18.5% 15.3% 18.7%
Histological type Ductal ND 100% intraductal 57.4% 65.4% 82.2%
Lobular ND 18.5% 19.2% 6.2%
Tubular ND 14.8% 0.0% 0.0%
Other ND 9.2% 15.3% 12.5%
PR, progesterone receptor; ER, estrogen receptor; HR, hormone receptor, i.e., PR and /or ER+; ND, not determined.
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3611
Figure 1
Phenotype of p-NK cell receptors in patients with dif-
ferent stages of BC at diagnosis and controls. (A–H)
Receptors were significantly altered among the dif-
ferent control groups and BC patients. The individual
mean fluorescence intensities (MFI) (unimodal expres-
sion) or percentages of cells positive for given markers
(bimodal expression) were graphed for the following
receptors: (A) NKp30 (MFI); (B) NKG2D (MFI); (C)
2B4 (MFI); (D) DNAM-1 (MFI); (E) CD16 (MFI); (F)
NKp46 (MFI); (G) NKG2A (percentage of cells positive
for NKG2A); and (H) CD85j (MFI). For each scattered
plot, the horizontal bar represents the mean value. (I)
Summary of the median expression of the 22 NK cell
receptors tested in the peripheral blood of controls and
BC patients. MFI or percentage of positive cell median
values were submitted to a hierarchical clustering pro-
gram to obtain a global view of receptor expression
in the different groups. HD (n = 22), B (n = 19), Tis
(n = 7), LOC (n = 55), LA (n = 26), M (n = 32). Statisti-
cal analyses were done using nonparametric unpaired
Mann-Whitney U test and Kruskal-Wallis (KW) ANOVA.
*P < 0.05; **P ≤ 0.005; ***P ≤ 0.0005.
research article
3612 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
exerted the 3 functions simultaneously (Figure 2E). Finally, we
measured the ability of p-NK cells to mediate ADCC against
SK-BR-3, a BC cell line sensitive to trastuzumab, an anti–erbB-2
monoclonal antibody known to induce ADCC (3). We were not
able to obtain enough p-NK cells to perform this test in the M
group. In the other groups, we observed a massive enhancement
of cytotoxicity in the presence of increasing doses of trastuzumab,
especially in the noninvasive groups, with more than 80% of cells
responding (Figure 2F and Supplemental Figure 2A). NK cells
from B and Tis patients were more efficient in mediating ADCC
at the lowest doses of trastuzumab than NK cells from LOC and
LA patients, which is coherent with their different levels of CD16
Figure 2
p-NK cell functions are altered in invasive BC patients. p-NK cells isolated from the different groups of patients were exposed to K562 cells in a
direct cytotoxic assay. (A) Effective killing of K562 cells. (B) Percentage of NK cells positive for CD107. (C) Percentage of NK cells positive for
IFN-γ. (D) Percentage of NK cells positive for TNF-α. (E) The multipotentiality of p-NK cells was determined from the number of functions (degranu-
lation as measured by CD107 expression, production of IFN-γ and/or TNF-α) that each p-NK cell was able to simultaneously accomplish against
K562 target cells. The E/T ratio was 1:1 in cytotoxic experiments performed against K562 cells. (F) ADCC efficiency was measured against the
SK-BR-3 BC cell line preincubated without or with increasing therapeutic doses (D1 to D7) of trastuzumab. Activation of NK cells was measured
by the expression of CD107. The E/T ratio was 2:1. The numbers of included patients per group were the following: B (n = 10), Tis (n = 7), LOC
(n = 16), LA (n = 16), M (n = 12, except for ADCC experiments where we did not obtain enough cells to perform the test). The statistical differences
between groups were established using nonparametric Mann-Whitney U test. *P < 0.05; **P ≤ 0.005. Data are represented as mean ± SEM.
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3613
expression. The lower responses observed in these patients were
not due to a decreased proportion of the VV polymorphism, the
CD16 allotype with the highest affinity for trastuzumab (Supple-
mental Figure 2B). To evaluate specific effects of the major pheno-
typic alterations, we performed a number of redirected cytotoxic
assays involving NKp30 and CD16. We found that the expression
of a given receptor directly correlated with the receptor-triggered
cytotoxicity (Supplemental Figure 2, C and D).
Thus, phenotypic alterations of p-NK cells in invasive BC corre-
late with altered functions. However, even if p-NK cells alterations
are associated with disease progression, they might not reflect the
situation within the tumor. We thus next studied tumor-infiltrat-
ing NK (Ti-NK) cells.
Ti-NK cells are functionally impaired compared with Mt-NK cells
Despite their major role in antitumor immunity and their presence
in close vicinity to cancer cells (Figure 3A), the phenotype of mam-
mary-infiltrating NK cells remains unknown. To identify altera-
tions that are tumor induced rather than tissue specific, we com-
pared NK cells isolated from breast tumors (n = 24, Ti-NK) with
paired p-NK cells (n = 11, p-NK) and/or with paired tumor-free
NK cells isolated from symmetric normal breast samples (n = 4,
mammary tissue–NK cells [Mt-NK]). Because of a low lymphocyte
extraction yield, we limited our study to NK cells isolated from the
largest LOC tumors and LA tumors. The proportion of Ti-NK cells
was similar to that found in paired tumor-free mammary tissues
(7.9% vs. 12.5%, respectively; Figure 3B). However, because the pro-
portion of CD45+ cells was lower in tumors than in healthy mam-
mary tissues, the absolute number of NK cells per million cells was
severely reduced in the malignant tissues (12,320 ± 1700 vs. 43,
265 ± 10 225 in healthy tissue; Figure 3C). Most importantly, the
phenotype of NK cells differed in malignant and healthy tissues, at
least with respect to their CD56Bright and CD56Dim phenotype: there
were more CD56Bright cells infiltrating the tumor (33.9% ± 14.8%)
than in the healthy mammary tissue (14.5% ± 6.1%). This latter
level was more comparable to that seen in P blood (8.9 ± 6.5%)
(Figure 3D). This increase of Ti-CD56Bright cells suggested either
the existence of 2 different maturation stages, as CD56Bright cells
are usually recognized as more immature NK cells with poor cyto-
toxic activity, or 2 activation stages, as the upregulation of CD56
expression is also a characteristic of activated NK cells, indepen-
dent of their function.
To strengthen this observation, we studied other NK cell mark-
ers. The increased expression of CD56 together with NKp44 and
CD69 (Figure 4, A and B, and Supplemental Figure 3A) suggested
that Ti-NK cells were indeed activated NK cells. Moreover, the
increased expression of NKG2A and CD27 and the decreased
expression of activating receptors (NKp30, NKG2D, DNAM-1,
CD16) and cytotoxicity-related molecules (CD57, PRF1, GZMB,
and TRAIL) on Ti-NK cells compared with p-NK cells and Mt-
NK cells suggested that Ti-NK cells have a poor cytotoxic poten-
tial. We then submitted our NK cells’ profiles to TMEV software
analysis. BC aggressiveness was evaluated with the Nottingham
Prognostic Index (NPI), a prognostic index based on 3 patho-
logical parameters (tumor size, tumor grade, and axillary LN
status) in nonmetastatic invasive BC (34). Hierarchical cluster-
ing underlined the similarity of the profiles between Mt-NK
cells and p-NK cells (Figure 4C). Ti-NK cells profiles were het-
erogeneous, with the exception of those from a group of patients
with poor prognosis (high NPI) that showed underexpression
of nearly all NK cell receptors. NK cell markers clustered into
3 groups: group I was composed of the cytotoxic-related effec-
tors, group II, the receptors involved in NK cell activation, and
group III, the molecules related to NK cell maturation markers.
Figure 3
NK cells infiltrating breast tumors
(Ti-NK cells) compared with healthy
mammary tissue (Mt-NK cells). (A)
Immunohistochemistry staining of
Ti-NK cells on a paraffin section of
a breast tumor. NK cells were iden-
tified with an anti-CD56 antibody
(brown staining). Other lympho-
cyte populations, mostly T cells,
displayed condensate dark blue
nucleus, while tumor cells are dis-
tinguishable by their larger size. (B)
FACS analysis of the percentage
of CD45+CD56+CD3– cells found
in tumor and paired healthy mam-
mary tissue. (C) Absolute number of
CD45+CD56+CD3– cells per million
cells found in tumor compared with
healthy mammary tissue. (D) Per-
centage of CD56Bright and CD56Di m
NK cells in the peripheral blood,
tumor, and healthy mammary tissue
of BC patients.
research article
3614 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
Figure 4
Phenotype of NK cells infiltrating healthy mammary tissue (Mt-NK cells), tumors (Ti-NK cells), and comparison with peripheral blood (p-NK
cells) profile. (A) Monoparametric histograms of the most important NK cell cytotoxic receptors in paired compartments, respectively, Mt-
NK cells and Ti-NK cells. (B) Monoparametric histograms of NK cell receptors involved in NK cell maturation and/or cytotoxicity in paired
compartments, respectively, Mt-NK cells and Ti-NK cells. (C) Hierarchical cluster representation of NK cell receptors expressed on Ti-NK,
Mt-NK, and p-NK. The phenotypes of 24 Ti-NK cells, 4 paired Mt-NK, and 11 paired p-NK cells were submitted to TMEV, and data were
normalized by row; then the hierarchical clustering was applied to both NK cell markers and patient samples. Markers are represented
horizontally while each patient is graphed vertically. Patients’ NPI is shown below the clustering. The following color code was used: green,
NPI < 3.4 (good-to-excellent prognosis); yellow, 3.4 < NPI < 5.4 (moderate prognosis); blue, NPI > 5.4 (poor prognosis). (D) Result of the
contingency data for the group II metamarker. The Fisher exact test was significant (P = 0.013), and the strength of association was as
follows: relative risk: 0.1477, 95% CI: 0.02170 to 1.006; odds ratio, 0.0625; 95% CI: 0.006019 to 0.6490.
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3615
Downregulation of group I molecules was observed in all Ti-NK
cells compared with Mt-NK and p-NK cells, suggesting that it
represents an alteration induced by the tumor. We searched for
correlations among the expression of these 3 groups (gene clus-
ters) and tumor aggressiveness using a metagene approach (35).
The metagene expression value was determined by calculating
the mean of the normalized expression values of all genes in the
respective gene groups (metagenes I, II, and III). The only clear
trend that emerged was that a high expression of metagene II was
associated with a good-to-excellent prognosis (NPI value < 3.4)
(Figure 4D). Together, our results indicate that most Ti-NK cells
are CD56BrightNKG2Ahi CD16loKIRlo cells with poor cytotoxic
potential (PRF1loGZMBloCD57lo), associated with poor outcome
in BC. With the exception of the KIRs, Ti-NK cell alterations were
very similar to those found in p-NK cells, but dramatically more
pronounced. Indeed, we found positive correlations between the
MFI of some markers expressed on Ti-NK cells with paired p-NK
cells (NKp30 [R = 0.645, P = 0.03], CD16 [R = 0.638 P = 0.03],
NKG2A [R = 0.774, P = 0.003], and NKG2D [R = 0.531, P = 0.07])
(Supplemental Figure 3B). The alterations detected in peripheral
blood might thus come from recirculating Ti-NK cells and/or be
induced by soluble factors secreted by the tumor.
We next asked whether compromised Ti-NK cells are neverthe-
less able to exert a helper function, as suggested by their CD56bright
phenotype. CD107 functional assays comparing paired Ti-NK and
p-NK cells showed that Ti-NK cells were activated more slowly
(CD69 as early activation marker), had a decreased degranulation
potential (CD107), and displayed strong alterations of IFN-γ and
TNF-α production (Figure 5, A and B). Moreover, Ti-NK cells were
unable to fulfill their ADCC function, which was not surprising
considering the low proportion of CD16posCD56Dim cells within
the tumor (Figure 5C).
In conclusion, we showed that NK cells are strongly altered
within a breast tumor compared with the equivalent tumor-
free environment. Alterations detected in peripheral blood are
less marked but nevertheless likely to reflect what occurs in the
Figure 5
Functionality of NK cells infiltrating tumors (Ti-NK cells) compared with paired p-NK cells. (A) Isolated NK cells from malignant tissue (red) or
peripheral blood (black) from paired sampled were used in direct cytotoxic assays on 5 paired samples from BC patients. (B) Dot plot represen-
tation of CD69, CD107, IFN-γ, and TNF-α expression in NK cells after incubation with K652 cells, according to a 1:1 E/T ratio. Isolated NK cells
were incubated overnight in medium complemented with suboptimal concentrations of IL-2 and IL-15 before incubation with K562 cells for 4
hours. (C) Dot plot representation of CD69, CD107, INF-γ, and TNF-α expression in p-NK versus Ti-NK cells after exposure to SK-BR-3 cells in
the presence of trastuzumab. The E/T ratio was of 2:1. These results were obtained on paired samples (n = 3).
research article
3616 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
tumor. These changes are indicative of a sustained escape from
NK cell antitumor immunity, probably induced by the malig-
nant cells. This latter assumption will be assessed below.
NK cells are modulated by the tumor microenvironment
Cancer cells express ligands for NK cell receptors. NK cell alterations
represent a mechanism of tumor escape or tumor immuno-edit-
ing only if the affected receptors are involved in cancer cell rec-
ognition. We thus measured the expression of NK cell ligands
on breast tumor cells. We found that breast tumors express
heterogeneous levels of ligands for NK cell receptors, notably
regarding HLA members and NKp30-ligand (L) (Figure 6, A
and B). Breast tumors also frequently expressed high levels of
DNAM-1–L and NKG2D-L (Figure 6, A and B). Most impor-
tantly, we found deregulated transcriptional expression of NK
cell ligands in a large set of BC (n = 250) compared with healthy
mammary tissues (n = 5 pools of 5) previously profiled using
DNA microarrays, suggesting that an editing process was indeed
involved (Supplemental Figure 4).
Tumor cells synthesize soluble factors affecting Ti-NK cells. To deter-
mine whether the malignant cells were responsible for the
observed NK cell alterations, we exposed normal NK cells to
breast tumor supernatants. Tumor supernatants, but not super-
natants from normal mammary tissue, induced alterations of
NK cell receptors concordant with the phenotype observed on
Ti-NK cells. The most prominent effect was observed on GZMB
(P < 0.05), NKG2D (P < 0.05), and NKG2A (P < 0.005) (Figure
7A). When we exposed p-NK cells to these supernatants and then
used them in cytotoxic assays, we observed a profound inhibition
of all NK cell functions, particularly IFN-γ secretion, compared
with nonexposed p-NK cells (Figure 7B). These results suggest
that soluble factors released by tumors (epithelial or stromal
components) affect both NK cell phenotype and functions. We
determined the levels of soluble factors known to be increased in
other cancers and to alter lymphocyte functions, such as TGF-β1,
PGE2, LGALS3, and sMICA. We also measured ADAM17, a metal-
loproteinase involved in the shedding of MICA and other mole-
cules modulating NK cell functions (Supplemental Figure 5). We
correlated these factors with the respective expression of recep-
tors on mammary NK cells. The resulting R coefficients were sub-
mitted to TMEV to obtain a correlation matrix (36). The levels
of soluble inhibitory factors positively correlated with NKG2A
expression (Figure 7C), whereas they negatively correlated with
molecules related to NK cell cytotoxicity (GZMB, PRF1, CD57,
KIR, CD16). The highest negative correlations were obtained
with TGF-β1 and PGE2. TGF-β1 also displayed the highest corre-
lation with the NPI, showing that more aggressive breast tumors
contain more TGF-β1–producing cells (Figure 7D). Blocking
TGF-β1 in tumor supernatants partially restored NK cell func-
tionality, by 20% on average (Figure 7E).
Reversion of p-NK cell alterations under remission
As a final proof that NK cell alterations are tumor induced, we
established the phenotype of p-NK cells from 7 former patients
with invasive BC (Ex-BC) who had undergone surgery more than
5 years ago and had not relapsed since. We matched these cases
for age and pTNM classification with the BC patients from our
series. Patients from the benign group (B group) were used as the
Figure 6
Ligands of NK cell receptors are expressed by breast tumor cells.
(A) Epithelial cells isolated from malignant mammary tissue were
phenotyped for the ligands of the main altered NK cell receptors.
2 examples representative of each ligand expression are illustrat-
ed here. Horizontal bars indicate the percentage of epithelial cells
positive for the marker of interest (solid line, gray histogram) in com-
parison with the respective control isotype on the same population
(dashed line, white histogram). (B) Summary of MFI obtained for
each of the NK cells ligands in 5 independent experiments. Data are
represented as mean ± SEM.
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3617
control for a nonmalignant situation. NKp30, CD16, NKG2D,
and NKG2A receptors were expressed at similar levels in the B
group and Ex-BC patients, respectively up- and downregulated
compared with matched NK cells isolated from the BC groups,
indicating that a normal NK cell phenotype was restored in the
Ex-BC patients after tumor removal (Figure 8). In conclusion, our
data show that BC cells can shape a strong inhibitory microenvi-
ronment able to inhibit NK cell antitumor immunity.
Discussion
Understanding how the immune system affects the development
and progression of solid tumors remains one of the most challenging
questions in immunology. It has been previously observed that mice
with defects in IFN-γ signaling or antitumor immune cells are more
likely to develop primary tumors (5, 37, 38). Accordingly, antitumor
immunity is frequently deficient in patients with solid tumors (28, 29).
We have confirmed here that p-NK cell activity, notably IFN-γ pro-
Figure 7
Impact of breast tumor stroma on p-NK cells. (A) p-NK cells were cultured for 48 hours in complete medium, dissociation supernatant from tumor-
free samples (H, n = 4), or dissociation supernatant from malignant samples (T, n = 12). The main NK cell receptors were then phenotyped. The
resulting expressions were submitted to TMEV. The significance of the observed variation was measured with a Kruskal-Wallis (KW) test. (B)
Alterations of NK cell functions were measured from 10 different tumor supernatants. CD69 expression, CD107 degranulation, IFN-γ, and TNF-α
synthesis and percentage of absolute dead K562 cells were measured from p-NK cells cultured in complete medium (white bars) and p-NK
cells exposed to tumor supernatant (gray bars). (C) The correlation matrix of the R coefficients was obtained with a Spearman’s test between
the quantitative levels of tissue-associated soluble factors and paired NK cell receptors expression. (D) Correlation between TGF-β1 levels
found in breast tumor supernatants and the respective NPI. (E) Partial restoration of NK cell functions after preincubation of tumor supernatants
(n = 10) with blocking TGF-β1 antagonist antibody. Variations were evaluated with a nonparametric Mann-Whitney U test. *P < 0.05; **P ≤ 0.005;
***P ≤ 0.0005. Data are represented as mean ± SEM.
research article
3618 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
duction, is strongly impaired in invasive breast tumors (39, 40),
especially when the stage of disease is advanced. In situ, the immune
infiltrate has long been neglected, but we now know that it is not the
quantity but the quality of the immune infiltrate, mostly composed
of cytotoxic cells, that can predict tumor outcome of several cancers,
such as early-stage colorectal cancer, gastrointestinal tumor, pulmo-
nary adenocarcinoma, and, more recently, BC (41–43).
Our data show that at least 4 major mechanisms are associated
with an escape from NK cell antitumor immunity in invasive breast
tumors. First, we showed that breast tumor cells alter NK cell func-
tions through the modulation of their surface receptors, a mechanism
observed in several other malignancies (28, 29, 44–47). These altera-
tions are associated with invasive characteristics and a poor progno-
sis. Among invasive BCs, the cases in which NK cell infiltrates still
displayed high expression of NCRs, NKG2D, and DNAM-1 receptors,
and so which could still be considered as activated cells, have a good
prognostic index. NK cells were our primary population of interest
considering that invasive tumors preferentially develop in an NK-defi-
cient mouse background (6, 7). However, some of the affected recep-
tors, such as DNAM-1, NKG2D, and NKG2A, are also expressed by
specific populations of T cells. We found that these receptors are also
altered on tumor-infiltrating T cells, but we did not detect any associa-
tion with the acquisition of invasive characteristics (data not shown).
Data obtained on large cohorts have demonstrated a beneficial effect
on BC patients’ survival of the presence in the tumor of cytotoxic
T cells at diagnosis, suggesting that these cells also play a major role
in breast tumor outcome. It would thus be interesting to look at the
functional behavior and/or regulation of other antitumor immune
cells. By inducing the downregulation of activating receptors and the
upregulation of inhibitory receptors on lymphocytes, the tumor might
become “invisible” to antitumor immunity.
Second, we looked at the immunogenicity of breast tumors
and found that both the protein and mRNA profiles of NK cells’
ligands are different from the profiles seen in healthy mammary
tissues. This is also a phenomenon frequently observed in several
tumors (28, 48, 49). The heterogeneity of the profiles suggested
that the phenotype of tumor cells at diagnosis is the result of a
more or less successful immuno-editing (28, 50).
Third, the alterations of Ti-NK cells, and their reversibility in
BC patients in long-term remission, suggested that the tumor
induces its own tolerance from NK cell antitumor immunity
(51). We tried to identify how the tumor might influence the
expression of activating and inhibitory ligands. Based on previ-
ously published data, we showed that the alterations of NK cell
receptors could be induced by the breast tumor microenviron-
ment. TGF-β1 was an obvious candidate because it is a power-
ful inhibitor (52–54) elevated in advanced breast tumors (55).
Accordingly, blocking TGF-β1 in mice suppresses the occurrence
of metastases and restores NK cell activity (56). Furthermore,
TGF-β1 is also one of the key factors promoting epithelial-to-
mesenchymal transition (EMT), a major mechanism active at
the invasive front of solid tumors and responsible for disease
progression, metastatic potential, and chemotherapeutic resis-
tance (57–60). Mesenchymal stem cells, which support cancer
growth in situ or at the site of BC metastasis, can protect BC
cells from immune recognition directly by producing TGF-β1 or
indirectly by increasing the recruitment of Tregs (61, 62). This
is in line with the correlations we observed between the in situ
levels of TGF-β1 or Treg infiltrate and NK cells’ impaired phe-
notype and functions (refs. 31, 63, 64; Figure 7 and Supplemen-
tal Figure 6). However, because TGF-β1 only partially explained
this effect, we did not limit our study to TGF-β1 and Tregs as
Figure 8
Reversibility of p-NK cell phenotype in patients in remission. The phenotypes of NKp30, CD16, NKG2D, and NKG2A receptors were restored
in former BC patients compared with matched (age, group, and TNM classification) BC patients at diagnosis. BC (n = 7); B (n = 12); Ex-BC
(n = 7). Data are represented as mean ± SEM.
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3619
potential inhibitors of NK cell functions. We found that not
only PGE2 (65, 66), sMICA (67), LGALS3 (68), and ADAM17
(69), but also IDO1 (data not shown and ref. 70) contributed to
the immunosuppressive environment. All these molecules have
previously been implicated in tumor progression and impair-
ment of NK cell phenotypes and functions. In the MMTV/Neu
mouse model, which spontaneously develops invasive mammary
tumors (71), we observed alterations of NK cell phenotypes and
functions concordant with those found in humans (Supplemen-
tal Figure 7). It would be interesting to determine whether the
administration of blocking antibodies against the various actors
of the tumor microenvironment previously mentioned, alone or
in combination, could delay or accelerate invasive tumors and/
or metastasis occurrence in this model.
Finally, we looked at how the microenvironment influences final
maturation of NK cells in tissues. A limited amount of informa-
tion on other tissue-infiltrating NK cell phenotypes is available
(72, 73). NK cells usually show tissue-specific patterns of expres-
sion and disturbance of the cytotoxic CD56DimCD16+KIR+/– and
helper/precursor CD56BrightCD16–KIR– balance (74, 75). Notably,
human NK cells collected from nonreactive LNs are CD16–KIR–,
whereas NK cells derived from reactive/efferent LN and blood
express CD56DimCD16+KIR+ (76). The current belief is that
immature CD56BrightNK cells acquire these molecules in the LN
during inflammation and then circulate as KIR+CD16+NK cells.
CD56BrightCD16–KIR– and CD56DimCD16+KIR+/– NK cells might
thus correspond to sequential steps of differentiation, like the
CD56BrightCD16lowKIR+/– found in breast tumors (77, 78). Togeth-
er, these observations support the hypothesis that secondary
lymphoid organs, or any microenvironment, can be the site not
only of human NK cell final maturation but also of self-tolerance
acquisition during immune reaction. Several recent studies show-
ing that NK cell developmental programming is not entirely fixed
and that mature NK cells can be reeducated by their environment
support this hypothesis (79–81). We showed that BC Ti-NK cells
expressed activating molecules (CD56Bright, CD25, CD69, NKp44)
like mature activated NK cells, but the levels of all other markers
were instead characteristic of immature and nonfunctional NK
cells (82). The most intriguing feature was the disappearance of
the KIR, otherwise present on Mt-NK cells, confirming that Ti-
NK cells could not exert cytotoxic functions like noneducated
cells (83). The education process strongly depends on the engage-
ment of activating and inhibitory receptors (84), but ligands were
extremely heterogeneous in the different breast tumors, whereas
KIR loss was homogeneous. We suggest that a strong inhibitory
environment can reorient or reverse the transcriptional program
of NK cell maturation toward a nonreactive self-tolerant profile.
In conclusion, our study highlights the role of NK cells in the con-
trol of invasive breast tumors, suggesting that restoring NK cells’
antitumor efficiency (with adoptive transfer of preactivated or allo-
reactive NK cells) after surgery might help clear residual tumor cells
as demonstrated in other malignancies (85–87). Most importantly,
we showed that invasive breast tumors induce self tolerance in NK
cells, resulting in attenuated malignant cell immunogenicity and the
creation of a multifaceted immunosuppressive microenvironment
that blunts NK cells’ cytotoxicity and prevents their final matura-
tion process. NK cell subversion is thus another active mechanism,
in addition to the other breast tumor–derived mechanisms such as
increased Treg infiltrates, dendritic cell modulation, and inhibitor
secretion, that contributes to breast tumor progression. Finally,
these results are particularly interesting in the context of BC dor-
mancy and immune evasion and highlight the importance of devis-
ing future therapies able to enhance NK cell cytotoxicity efficiently
to further prevent invasive breast tumor recurrence (88).
Methods
Patients. Benign breast tumor and BC patients treated at the Institut Paoli-
Calmettes were prospectively recruited on diagnosis between January 2007
and December 2009. Blood and/or tumors were respectively sampled
before or during the surgical diagnostic or therapeutic act, before admin-
istration of any treatment related to tumor progression. Fresh samples
were extemporaneously treated, before the determination of the diagnosis
by the surgeon. After analysis of morphological tumor characteristics by
pathologists, patients were retrospectively classified into 5 groups, as fol-
lows (Table 1): patients with a benign tumor (the B group), patients with
an in situ cancer (the Tis group), patients with invasive localized cancer
(pT1N0 an pT2N0, referred to as the LOC group), patients with invasive
locally advanced cancer (pT2N1-2 to pT4, referred as the LA group), and
patients whose initial breast tumor had given rise to metastases in distant
organ(s) (referred to as the M group). We also included a control group
of former BC patients, followed in our institution, who had undergone
surgery at least 5 years ago and had not relapsed since. NPI was used to
establish the precise prognostic value in invasive tumors, based on 3 fac-
tors: tumor grade, number of LNs involved, and size of the tumor.
NK cell phenotype by flow cytometry. 100 μl of fresh whole blood or 0.7 × 106
Ti-cells were incubated with the appropriated antibodies (Supplemen-
tal Table 1) on a rocking platform for 30 minutes. Red blood cells were
lysed with OptiLyse B (Beckman Coulter). Samples were extemporane-
ously analyzed on a BD FACS Canto (BD Biosciences). Before and after
reading patient samples, fluorescence intensities of the FACS Canto were
standardized over time with PMT 7-Color Setup Beads (BD Biosciences)
to compensate for fluorescence intensity variability. The gating strategy
consisted of the elimination of the doublets based on the FCS-A/FCS-H
parameters, followed by the removal of dead cells. NK cell population was
selected based on the following phenotype: CD45posCD3negCD56pos. Treg-
cells were selected on the following expression: CD45posCD3posCD4posCD
25hiCD127lo.
NK cell isolation from peripheral blood samples. NK cells were negatively iso-
lated with the NK cell StemSep System (StemCell Technology) according
to the manufacturer’s instructions. The purity and viability of sorted cells
were established and always greater than 94%.
NK cells were incubated in RPMI, 10% FCS, complemented with sub-
optimal concentrations of IL-2 (100 U/ml; Proleukine, Chiron) and IL-15
(5 μg/ml; R&D Systems) overnight.
Isolation of epithelial tumor cells from mammary tissue. The minced mammary
tumor was incubated with 1× collagenase/hyaluronidase for 16 hours at
37°C on a rotary shaker, according to the manufacturer’s instructions
(StemCell Technology). Briefly, the liquefied fat layer and supernatant
were discarded after centrifugation. The pellet, enriched in epithelial cells,
was resuspended in prewarmed trypsin-EDTA, and then washed. Cells were
resuspended in prewarmed dispase (5 mg/ml) complement with DNase I
(1 mg/ml). After additional washes, cells were counted and resuspended in
PBS, 2% BSA, for fluorescent staining if viability exceeded 75%.
Isolation of lymphocyte infiltrates in mammary tissue. A section made within
the core of the malignant area was selected by the pathologist and extem-
poraneously cleared of pieces of fat, then weighed; an equal volume (w/v)
of medium (RPMI 1640) was added to the tumor. The tumor was mechan-
ically disrupted; then the supernatant was harvested and centrifuged at
high speed to remove residual cellular elements. This supernatant, later
referred to in the text as “supernatant of dissociation,” was aliquoted and
research article
3620 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
frozen at –20°C until further use. The mean volume of the tumor superna-
tant of dissociation was 0.640 ± 0.420 ml. After mechanical disruption and
removal of the supernatant of dissociation, the tumor was digested for 1
hour under agitation with collagenase Ia (1 mg/ml) and DNase I (50 × 103
units/ml; Sigma-Aldrich). Cell suspension was then used for flow cytom-
etry staining or NK cell isolation if viability was greater than 80%.
Cytotoxic activity analysis. NK cells were tested for cytotoxic activity against
the leukemic HLA-Ineg K562 cell line (direct cytotoxicity), the FcγR-positive
P815 mastocytoma murine cell line (redirected cytotoxicity). or the erbB-
2–overexpressing BC cell line SK-BR-3 (used to measure ADCC potential
of NK cells in the presence of trastuzumab, the therapeutic mAb targeting
erbB-2). These cytotoxic tests were all done in 4-hour assays. The measured
parameters were degranulation (CD107a and CD107b) and cytokine pro-
duction (IFN-γ and TNF-α) by NK cells or the percentage of absolute num-
ber of dead targets, referred to as the cell death index (CDI). The respective
effector/target (E/T) ratios are indicated in the figure legends.
In redirected experiments, NK cells were preincubated (20 minutes at
37°C) with saturating amounts of purified antibodies (IgG1, anti-CD16,
anti-NKp46, anti-NKG2D, anti-NKG2A, anti-KIR, anti-NKp30, and anti-
DNAM-1 mAb), then used in cytotoxic assays.
In ADCC experiments, SK-BR-3 cells were preincubated (30 minutes
at 4°C) with increasing amounts of purified trastuzumab, then used in
cytotoxic assays.
For flow cytometric experiments measuring the degranulation and cytokine
production of NK cells and CD107a and CD107b mAb plus GolgiStop, all
purchased from BD Biosciences, were added in each well at the beginning of
the culture. At the end of this incubation, NK cells were stained with CD56
and permeabilized with Cytofix/Cytoperm (BD Biosciences). Intracellular
antibodies were next added (IFN-γ and TNF-α) before cell analysis.
For cytotoxic assays measuring the CDI (percentage of dead targets com-
pared with targets not exposed to NK cells), experiments were done in True-
Count Beads Tubes (BD Biosciences) to precisely quantify the number of
NK cell–mediated dead cells. Dead cells were visualized with the Live/Dead
red reagent according to the manufacturer’s instructions (Invitrogen).
For the study of alterations induced by breast supernatants, purified NK
cells were cultured with or without healthy or tumor supernatants (dilu-
tion 1:2) before staining 48 hours later or used in CD107 assays 40 hours
later (followed by 4 hours of CD107 assay).
Finally, for the study of TGF-β1 involvement in breast tumor–mediated
alterations, NK cells were cultured in the presence of breast tumor superna-
tants (dilution 1:4) preincubated with blocking anti–TGF-β1 (20 μg/ml).
Immunohistochemistry. Immunohistochemical staining was performed on
5-μm cryostat sections fixed in 4% paraformaldehyde. After neutralization
of the endogenous peroxidase and saturation, sections were layered with
the anti-CD56 mAb (Dako) or isotype-matched control mAb, revealed with
the Vectastain ABC Kit (Dako), and counterstained with Gill hematoxylin
(Merck). Although CD56 is not strictly specific for NK cells, the percentage
of CD56+CD3+ cells determined by flow cytometry staining was negligible.
Therefore, we estimated that CD56 staining gave a reasonable representa-
tion of NK cell infiltrates within the tumor.
ELISA. ELISAs were performed according to the manufacturer’s
instructions. sMICA, TGF-β1, LGALS3, and ADAM17 were purchased
from R&D Systems, while the PGE2 ELISA kit was purchased from Cay-
man Chemical (Interchim).
Microarray experiments. U133+2 pan-genomic Affymetrix data from
malignant and healthy mammary tissues were downloaded from the public
GEO data sets (GSE21653).
MMTV-Neu mice. FVB/N-Tg(MMTVneu)202 Mul/J mice (MMTV-neu
mice) were obtained from the Jackson Laboratory. All mice were main-
tained on an FVB background. Mice were palpated weekly from 5 months
of age and, once mammary gland nodules were detected, tumor diameter
was measured twice weekly with calipers. MMTV-Neu mice (n = 6) and FVB
control mice (n = 6) were compared, and a Kaplan-Meier analysis was done
to determine tumor-free survival. Monthly blood samples were collected
under isoflurane (5%) inhalation anesthesia and analyzed by f low cytom-
etry. At the end of the protocol, mice were sacrificed and breast tumors and
organs were collected.
Statistics. All statistical analyses were done with StatView software. Dif-
ferences between breast tumor patients’ groups were evaluated using
2-tailed nonparametric unpaired Mann-Whitney U test. The comparisons
between NK cell markers in paired tumor and peripheral blood samples
were done with nonparametric Wilcoxon tests. Finally, all correlations
between MFIs of NK cell markers expressed on Ti-NK and p-NK were
evaluated with nonparametric Spearman correlation tests. The equality
of the population medians among the different groups was triggered
with the nonparametric ANOVA (Kruskal-Wallis) for each marker. In all
figures with histograms, data are represented using the mean ± SEM.
P < 0.05 was considered as significant.
Human study approval. This study was approved by the Institutional Ethic
Committee Review Board (Comité d’orientation Stratégique [COS], Mar-
seille, France) from the Institut Paoli-Calmettes. During the inclusion visit,
each patient gave written informed consent for research use.
Mouse study approval. All animal procedures were performed in accordance
with protocols approved by the local Committee for Animal Experimenta-
tion (CAE) of Marseille, France.
Acknowledgments
We thank the Departments of Medical Oncology, Surgical Oncol-
ogy, Pathology and the crew of the “Hôpital de jour” at the Insti-
tut Paoli-Calmettes as well as the patients, and controls for their
contribution to this work. We also want to thank J. Ewbank for
careful editing of the manuscript. The “Institut National du Can-
cer” and ANR-RIB funded this study. E. Mamessier was funded
by the Association pour la Recherche contre le Cancer for 3 years.
A. Moretta was funded by AIRC: IG project n. 10643 and Special
Project 5x1000 n. 9962.
Received for publication November 17, 2010, and accepted in
revised form June 29, 2011.
Address correspondence to: Emilie Mamessier or Daniel Olive,
INSERM UMR 891, 27 Bd Lei Roure, 13009 Marseille, France.
Phone: 33.491.758.415; Fax: 33.491.260.364; E-mail: mamessier@
ciml.univ-mrs.fr (E. Mamessier), daniel.olive@inserm.fr (D. Olive).
Emilie Mamessier’s present address is: Laboratory Genomic Insta-
bility and Human Hemopathies, CIML, Parc Scientifique de Lumi-
ny, 13288 Marseille, France.
1. Tirona MT, Sehgal R, Ballester O. Prevention
of breast cancer (part I): epidemiology, risk fac-
tors, and risk assessment tools. Cancer Invest.
2010;28(7):743–750.
2. Joyce JA, Pollard JW. Microenvironmental regulation
of metastasis. Nat Rev Cancer. 2009;9(4):239–252.
3. Kute TE, et al. Breast tumor cells isolated from
in vitro resistance to trastuzumab remain sensi-
tive to trastuzumab anti-tumor effects in vivo
and to ADCC killing. Cancer Immunol Immunother.
2009;58(11):1887–1896.
4. Smyth MJ, Dunn GP, Schreiber RD. Cancer immu-
nosurveillance and immunoediting: the roles of
immunity in suppressing tumor development and
shaping tumor immunogenicity. Adv Immunol.
2006;90:1–50.
5. Sha nkara n V, et a l. IFN-gamma and lympho-
cytes prevent primary tumor development and
sha pe tum or imm unoge nicit y. Nature. 200 1;
410(6832):1107–1111.
6. Dewan MZ, et al. Natural-killer-cells in breast
research article
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011 3621
cancer cell growth and metastasis in SCID mice.
Biomed Pharmacother. 2005;59 suppl 2:S375–S379.
7. Dewan MZ, et al. Role of Natural-killer-cells in
hormone-independent rapid tumor formation and
spontaneous metastasis of breast cancer cells in
vivo. Breast Cancer Res Treat. 2007;104(3):267–275.
8. Olkhanud PB, et al. Breast cancer lung metas-
tasis requires expression of chemokine receptor
CCR4 and regulatory T cells. Cancer Res. 2009;
69(14):5996–6004.
9. Treilleux I, et al. Dendritic cell infiltration and
prognosis of early stage breast cancer. Clin Cancer
Res. 2004;10(22):7466–7474.
10. Gob ert M, et al. Regulatory T cells recruited
through CCL22/CCR4 are selectively activated in
lymphoid infiltrates surrounding primary breast
tumors and lead to an adverse clinical outcome.
Cancer Res. 2009;69(5):2000–2009.
11. Caras I, et al. Evidence for immune defects in breast
and lung cancer patients. Cancer Immunol Immuno-
ther. 2004;53(12):1146–1152.
12. Alexe G, et al. High expression of lymphocyte-asso-
ciated genes in node-negative HER2+ breast can-
cers correlates with lower recurrence rates. Cancer
Res. 2007;67(22):10669–10676.
13. Bertucci F, et al. Gene expression profiling shows
medullary breast cancer is a subgroup of basal
breast cancers. Cancer Res. 2006;66(9):4636–4644.
14. Fin ak G, e t al. Stromal gene expression pre-
dicts clinical outcome in breast cancer. Nat Med.
2008;14(5):518–527.
15. Sabatier R, et al. A gene expression signature identi-
fies two prognostic subgroups of basal breast can-
cer. Breast Cancer Res Treat. 2011;126(2):407–420.
16. Trinchieri G. Biology of Natural-killer-cells. Adv
Immunol. 1989;47:187–376.
17. Vivier E, Tomasello E, Baratin M, Walzer T, Ugolini
S. Functions of Natural-killer-cells. Nat Immunol.
2008;9(5):503–510.
18. Morett a L, Bottin o C, Pende D, Vi tale M, Mi n-
gari MC, Moretta A. Different checkpoints in
human NK-cell activation. Trends Immunol. 2004;
25(12):670–676.
19. Lanier LL. NK cell receptors. Annu Rev Immunol.
1998;16:359–393.
20. Moretta A, et al. Activating receptors and corecep-
tors involved in human Natural-killer cell-mediat-
ed cytolysis. Annu Rev Immunol. 2001;19:197–223.
21. Bauer S, et al. Activation of NK-cells and T cells by
NKG2D, a receptor for stress-inducible MICA. Sci-
ence. 1999;285(5428):727–729.
22. Moret ta L, B ottin o C, Pe nde D, Castr icon i R,
Mingari MC, Moretta A. Surface NK receptors
and their ligands on tumor cells. Semin Immunol.
2006;18(3):151–158.
23. Moretta A , et al . Receptors for HLA-class I-mol-
ecules in human Natural-killer-cells. Annu Rev
Immunol. 1996;14:619–648.
24. Moretta A, et al. Existence of both inhibitory (p58)
and activatory (p50) receptors for HLA-C mol-
ecules in human natural-killer cells. J Exp Med.
1995;182(3):875–884.
25. Long EO. Regulation of immune response
through inhibitory receptors. Annu Rev Immunol.
1999;17:875–904.
26. Lopez-Botet M, et al. Structure and function of the
CD94 C-type lectin receptor complex involved in
the recognition of HLA class I molecules. Immunol
Rev. 1997;155:165–174.
27. Madjd Z, Spendlove I, Pinder SE, Ellis IO, Durrant
LG. Total loss of MHC class-I is an independent
indicator of good prognosis in breast cancer. Int J
Cancer. 2005;117(2):248–255.
28. McGilvray RW, et al. NKG2D ligand expression in
human colorectal cancer reveals associations with
prognosis and evidence for immunoediting. Clin
Cancer Res. 2009;15(22):6993–7002.
29. Le Mau x Chan sac B, et al. NK-cells infiltrating a
MHC class-I-def icient lung adenocarcinoma dis-
play impaired cytotoxic activity toward autologous
tumor cells associated with altered NK-cell-trigger-
ing receptors. J Immunol. 2005;175(9):5790–5798.
30. Della Chiesa M, et al. The tryptophan catabolite L-
kynurenine inhibits the surface expression of NKp46-
and NKG2D-activating receptors and regulates NK-
cell function. Blood. 2006;108(13):4118–4125.
31. Castric oni R, et al. Transforming growth factor
beta 1 inhibits expression of NKp30 and NKG2D
receptors: Consequences for the NK-mediated
killing of dendritic cells. Proc Natl Acad Sci U S A.
2003;100(7):4120–4125.
32. Yarbro JW, Page DL, Fielding LP, Partridge EE,
Murphy GP. American Joint Committee on Cancer
prognostic factors consensus conference. Cancer.
1999;86(11):2436–2446.
33. Bryceson YT, et al. Functional analysis of human
NK-c ells b y fl ow cyto metr y. Methods Mol Biol.
2010;612:335–352.
34. Todd JH, et al. Confirmation of a prognos-
tic index in primary breast cancer. Br J Cancer.
1987;56(4):489–492.
35. Rosenwal d A , et al . The use of molecular profil-
ing to predict survival after chemotherapy for dif-
fuse large-B-cell lymphoma. N Engl J Med. 2002;
346(25):1937–1947.
36. Galon J, et al. Type, density, and location of immune
cells within human colorectal tumors predict clini-
cal outcome. Science. 2006;313(5795):1960–1964.
37. Dighe AS, Richards E, Old LJ, Schreiber RD.
Enhanced in vivo growth and resistance to rejection
of tumor cells expressing dominant negative IFN
gamma receptors. Immunity. 1994;1(6):447–456.
38. Kaplan DH, et al. Demonstration of an interferon
gamma-dependent tumor sur veillance system in
immunocompetent mice. Proc Natl Acad Sci U S A.
1998;95(13):7556–7561.
39. Garner WL, Minton JP, James AG, Hoffmann CC.
Human breast cancer and impaired NK-cell func-
tion. J Surg Oncol. 1983;24(1):64–66
40. Dewan MZ, et al. Natural-killer activity of periph-
eral-blood mononuclear cells in breast cancer
patients. Biomed Pharmacother. 2009;63(9):703–706.
41. Pagès F, et al. In situ cytotoxic and memory T cells
predict outcome in patients with early-stage colorec-
tal cancer. J Clin Oncol. 2009;27(35):5944–5951.
42. Takanami I, Takeuchi K, Giga M. The pro gnostic
value of Natural-killer cell infiltration in resected
pulmonary adenocarcinoma. J Thorac Cardiovasc
Surg. 2001;121(6):1058–1063.
43. Mahmoud SM, et al. Tumor-infiltrating CD8+ lym-
phocytes predict clinical outcome in breast cancer.
J Clin Oncol. 2011;29(15):1949–1955.
44. Costello R T, et al. Defective expression and func-
tion of Nat ural- kille r cel l-tri ggeri ng rec eptor s
in patients with acute myeloid leukemia. Blood.
2002;99(10):3661–3667.
45. Lakshmikanth T, et al. NCRs and DNAM-1 medi-
ate NK-cell recognition and lysis of human and
mouse melanoma cell lines in-vitro and in-vivo.
J Clin Invest. 2009;119(5):1251–1263.
46. Epling-Burnette PK, et al. Reduced Natural-killer
function associated with high-risk myelodysplas-
tic syndrome and reduced expression of activating
NK-receptors. Blood. 2007;109(11):4816–4824.
47. Carlsten M, et al. Reduced DNAM-1 expression on
bone marrow NK-cells associated with impaired
killing of CD34+ blasts in myelodysplastic syn-
drome. Leukemia. 2010;24(9):1607–1616.
48. Bedel R , et al. Novel role for STAT3 in transcrip-
tio nal re gulat ion of NK-im mune c ell ta rget-
ing r ecept or MIC A on c ancer cells . Cancer Res.
2011;71(5):1615–1626.
49. Kloess S, et al. IL-2-activated haploidentical NK-
cells restore NKG2D-mediated NK-cell cytotoxicity
in neuroblastoma patients by scavenging of plasma
MICA. Eur J Immunol. 2010;40(11):3255–3267.
50. Swann JB, Smyth MJ. Immune surveillance of
tumors. J Clin Invest. 2007;117(5):1137–1146.
51. Fauriat C, et al. Deficient expression of NCR in
NK-cells from acute myeloid leukemia: Evolution
during leukemia treatment and impact of leuke-
mia cells in NCRdull phenotype induction. Blood.
2007;109(1):323–330.
52. Flavell RA, Sanjabi S, Wrzesinski SH, Licona-
Limón P. The polarization of immune cells in the
tumor environment by TGF-beta. Nat Rev Immunol.
2010;10(8):554–567.
53. Yang L, Pang Y, Moses HL. TGF-beta and immune
cells: an important regulatory axis in the tumor
microenvironment and progression. Trends Immu-
nol. 2010;31(6):220–227.
54. Toomey D, et al. TGF-beta1 is elevated in breast
cancer tissue and regulates nitric oxide produc-
tion from a number of cellular sources during
hypoxia re-oxygenation injury. Br J Biomed Sci. 2001;
58(3):177–183.
55. Arteaga CL, Hurd SD, Winnier AR, Johnson MD,
Fendly BM, Forbes JT. Anti-transforming growth
factor-beta antibodie s inhibit breast ca ncer cell
tumorigenicity and increase mouse spleen Natu-
ral-killer-cell activity. Implications for a possible
role of tumor cell/host TGF-beta interactions
in human breast cancer progression. J Clin Invest.
1993;92(6):2569–2576.
56. Thiery JP, Sl eeman JP. Complex networks orch es-
trate epithelial-mesenchymal transitions. Nat R ev
Mol Cell Biol. 2006;7(2):131–142.
57. Christfori G. New signals from the invasive front.
Nature. 2006;441(7092):444–450.
58. Brabletz T, Jung A, Spaderna S, Hlubek F, Kirch-
ner T. Migrating cancer stem cells- an integrated
concept of malignant tumor progression. Nat R ev
Cancer. 2005;5(9):744–749.
59. Yang J, Weinberg RA. Epithelial-mesenchymal tran-
sition: at the crossroads of development and tumor
metastasis. Dev Cell. 2008;14(6):818–829.
60. Santisteban M, et al. Immune-induced epithelial to
mesenchymal transition in vivo generates breast can-
cer stem cells. Cancer Res. 2009;69(7):2887–2895.
61. Patel SA, Meyer JR, Greco SJ, Co rcoran KE, Bryan
M, Rameshwar P. Mesenchymal stem cells pro-
tect breast cancer cells through regulatory T cells:
role of mesenchymal stem cell-derived TGF-beta.
J Immunol. 2010;184(10):5885–5894.
62. Lin SY, et al. The isolation of novel mesenchymal
stromal cell chemotactic factors from the con-
ditioned medium of tumor cells. Exp Cell Res.
2008;314(17):3107–3117.
63. Lee JC, Lee KM, Kim DW, Heo DS. Elevated TGF-
beta1 secretion and down-modulation of NKG2D
underlies impaired NK cytotoxicity in cancer
patients. J Immunol. 2004;172(12):7335–7340.
64. Smyth MJ, Teng MW, Swann J, Kyparissoudis K,
Godfrey DI, Hayakawa Y. CD4+CD25+ T regulato-
ry cells suppress NK-cell-mediated immunotherapy
of cancer. J Immunol. 2006;176(3):1582–1587.
65. Martinet L, Jean C, Dietrich G, Fournié JJ, Poupot
R. PGE2 inhibits Natural-killer and gammad-
elta T-cell cytotoxicity triggered by NKR and TCR
through a cAMP-mediated PKA type I-dependent
signaling. Biochem Pharmacol. 2010;80(6):838–845.
66. Kundu N, Ma X, Holt D, Goloubeva O, Ostrand-
Rosenberg S, Fulton AM. Antagonism of the
prostaglandin E receptor EP4 inhibits metastasis
and enhances NK function. Breast Cancer Res Treat.
2009;117(2):235–242.
67. Madjd Z, et a l. Upregulation of MICA on high-
grade invasive operable breast carcinoma. Cancer
Immun. 2007;7:17.
68. Demotte N, et al. A galectin-3 ligand corrects the
impaired function of human CD4 and CD8 tumor-
infiltrating lymphocytes and favors tumor rejec-
tion in mice. Cancer Res. 2010;70(19):7476–7488.
69. McGowan PM, Ryan BM, Hill AD, McDermott E,
research article
3622 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 9 September 2011
O’Higgins N, Duffy MJ. ADAM-17 expression in
breast cancer correlates with variables of tumor pro-
gression. Clin Cancer Res. 2007;13(8):2335–2343.
70. Godin-Ethier J, et al. Human activated T-lymphocytes
modulate IDO expression in tumors through Th1/
Th2 balance. J Immunol. 2009;183(12):7752–7760.
71. Guy CT, Webster MA, Schaller M, Parsons TJ, Car-
diff RD, Muller WJ. Expression of the neu-protoon-
cogene in the mammary epithelium of transgenic
mice induces metastatic disease. Proc Natl Acad Sci
U S A. 1992;89(22):10578–10582.
72. Luci C, et al. Influence of the transcription fac-
tor RORgammat on the development of NKp46+
cell populations in gut and skin. Nat Immunol.
2009;10(1):75–82.
73. Castriconi R, et al. Natural-killer cell-mediated kill-
ing of freshly isolated neuroblastoma cells: critical
role of DNAX accessory molecule-1-poliovirus recep-
tor interaction. Cancer Res. 2004;64(24):9180–9184.
74. Cooper MA, Fehniger TA , Caligiuri MA. The biol-
ogy of human Natural-killer-cell subsets. Trends
Immunol. 2001;22(11):633–640.
75. Ferlazzo G, et al. Distinct roles of IL-12 and IL-15
in human Natural-killer cell activation by dendrit-
ic-cells from secondary lymphoid organs. Proc Natl
Acad Sci U S A. 2004;101(47):16606–16611.
76. Romagnani C, et al. CD56brightCD16– killer Ig-like
receptor- NK-cells display longer telomeres and
acquire features of CD56dim NK-cells upon activa-
tion. J Immunol. 2007;178(8):4947–4955.
77. Freud AG, Caligiuri MA. Human Natural-killer cell
development. Immunol Rev. 2006;214:56–72.
78. Freud AG, et a l. Evidence for discrete stages of
human Natural-killer cell differentiation in vivo.
J Exp Med. 2006;203(4):1033–1043.
79. Sun JC, Beilke JN, Bezman NA, Lanier LL. Homeo-
static proliferation generates long-lived Natural-
killer-cells that respond against viral infection.
J Exp Med. 2011;208(2):357–368.
80. Joncker NT, Shifrin N, Delebecque F, Raulet DH.
Mature Natural-killer-cells reset their responsive-
ness when exposed to an altered MHC environ-
ment. J Exp Med. 2010;207(10):2065–2072.
81. Elliott JM, Wahle JA, Yokoyama WM. MHC class-I-
deficient Natural-killer-cells acquire a licensed phe-
notype after transfer into an MHC class I-sufficient
environment. J Exp Med. 2010;207(10):2073–2079.
82. Carre ga P, et al. Natural-killer-cells infiltrating
human non-small-cell lung cancer are enriched
in CD56BrightCD16– cells and display an impaired
capability to kill tumor cells. Cancer. 2008;
112(4):863–875.
83. Jonsson AH, Yokoyama WM. Natural-killer cell
tolerance licensing and other mechanisms. Adv
Immunol. 2009;101:27–79.
84. Sun JC. Re-educating Natural-killer-cells. J Exp Med.
2010;207(10):2049–2052.
85. Berg M, et al. Clinical-grade ex vivo-expanded
human Natural-killer-cells up-regulate activat-
ing receptors and death receptor ligands and have
enhanced cytolytic activity against tumor cells.
Cytotherapy. 2009;11(3):341–355.
86. Romagné F, et al. Preclinical characterization of
1-7F9, a novel human anti-KIR receptor therapeutic
antibody that augments Natural-killer-mediated kill-
ing of tumor cells. Blood. 2009;114(13):2667–2677.
87. Ru gg er i L , et al . Donor Natural-killer cell
allorecognition of missing-self in haploidentical
hematopoietic transplantation for acute myeloid
leukemia: challenging its predictive value. Blood.
2007;110(1):433–440.
88. Kundu N, Walser TC, Ma X, Fulton AM.
Cyclooxygenase inhibitors modulate NK activities
that control metastatic disease. Cancer Immunol
Immunother. 2005;54(10):981–987.
