Lack of CD151/integrin α3β1 complex is predictive of poor outcome in node-negative lobular breast carcinoma: opposing roles of CD151 in invasive lobular and ductal breast cancers.

BACKGROUND: The proposed involvement of CD151 in breast cancer (BCa) progression is based on findings from studies in invasive ductal carcinoma (IDC). The IDC and invasive lobular carcinoma (ILC) represent distinct disease entities. Here we evaluated clinical significance of CD151 alone and in association with integrin α3β1 in patients with ILC in context of the data of our recent IDC study.
METHODS: Expression of CD151 and/or integrin α3β1 was evaluated in ILC samples (N=117) using immunohistochemistry. The findings were analysed in relation to our results from an IDC cohort (N=182) demonstrating a prognostic value of an expression of CD151/integrin α3β1 complex in patients with HER2-negative tumours.
RESULTS: Unlike in the IDCs, neither CD151 nor CD151/α3β1 complex showed any correlation with any of the ILC characteristics. Lack of both CD151 and α3β1 was significantly correlated with poor survival (P=0.034) in lymph node-negative ILC N(-) cases. The CD151(-)/α3β1(-) patients had 3.12-fold higher risk of death from BCa in comparison with the rest of the ILC N(-) patients.
CONCLUSIONS: Biological role of CD151/α3β1 varies between ILC and IDC. Assessment of CD151/α3β1 might help to identify ILC N(-) patients with increased risk of distant metastases.

The tetraspanin protein CD151 (Tspan24) has recently emerged as a new candidate indicator of tumour cell invasiveness (Boucheix and Rubinstein, 2001; Hemler, 2005; Zoller, 2009). Elevated expression of CD151 protein and its involvement in tumour progression have been observed in various human malignancies (Romanska and Berditchevski, 2011). In breast cancer (BCa), in particular, high expression of tetraspanin CD151 was shown to correlate with axillary lymph node involvement and patient poor overall survival (Sadej ; Kwon ; Novitskaya ). The role of CD151 in tumour invasive and metastatic progression is thought to be relying on its ability to form complexes with laminin-binding integrin receptors (i.e., α6β1, α3β1, α6β4) (Sadej ) and its involvement in the regulation of cell–cell and cell–matrix interactions (Johnson ). The underlying signalling pathways are likely to depend on adhesion-dependent and coordinated activation of small Rho and Ras GTPases, c-Akt and p38 (Novitskaya ). It was also proposed that the CD151/α3β1 complex is directly linked to cadherin and catenin, thus regulating E-cadherin-mediated cell–cell adhesion (Chattopadhyay ). A suppressive role of the CD151/α3β1 complex, recently demonstrated in ovarian cancer, has been linked to stabilisation of integrin α3β1- and E-cadherin-mediated cell–cell contacts (Baldwin ). In addition to its role in regulation of adhesion-dependent outside-in signalling pathways, CD151 might contribute to mammary tumourigenesis via ErbB2/HER2 (Deng ; Novitskaya ). Interestingly, clinical analyses of BCa patients showed that the elevated expression of CD151 correlated with poor overall survival in patients only with HER2-negative (luminal A and quintuple-negative) tumours (Kwon ). Moreover, an impact of CD151/α3β1 on invasive ductal carcinoma (IDC) patient survival was shown to be inversely correlated with the level of HER2 expression (Novitskaya ).

Existing knowledge of CD151 involvement in mammary tumourigenesis is almost exclusively based on clinical and in vitro studies of the IDC. However, it is becoming increasingly apparent that CD151 might have diverse, even opposing roles, in different biological contexts and its clinical significance should be investigated in relation to the phenotypical and histological variants of the tumour (Voss ).

Invasive lobular carcinoma (ILC) accounts for ∼10% to 15% of newly diagnosed breast carcinomas. Classical ILCs are characterised by an outgrowth of small, uniform discohesive neoplastic cells that invade the stroma in a single-file pattern (Sikora ). The majority of ILC tumours are of low histological grade and preferentially display a luminal A phenotype (ER+/PR+/HER2−) (Weigelt ). Classical ILCs lack E-cadherin expression that is considered a defining feature of this BCa type (Sikora ). Recent studies demonstrated that ILCs differ from grade- and molecular subtype-matched IDCs in the transcriptomic profiles related to cell adhesion, cell–cell signalling and metastatic pattern, indicating that IDCs and ILCs represent distinct disease entities (Weigelt ; Sikora ).

Because of the relatively low incidence of ILC as well as the paucity of available research models (Sikora ), the mechanisms underlying its pathophysiology are poorly understood. It is likely that the role of CD151 in progression of E-cadherin-inactivated ILCs, and hence its impact on patient outcome, differs from those documented in the more common IDCs. To date, CD151 expression has been studied only in a few ILC cases, a small part (2.9% 25 out of 886) of a large cohort analysed according to the phenotypical and not the histological subtypes (Kwon ). More information is available regarding expression of integrins in ILC. Most reports are in agreement that malignant ILC phenotype is associated with decrease of expression of a range of integrin subunits. However, independently, the integrins seem to be of limited clinical value as no correlation between their expression, histological grade, nodal involvement, proliferative activity or, above all, the overall survival has ever been found (Gui ; Lanzafame ; Gonzalez ; Gonzalez-Angulo ).

The α3β1 integrin forms a stoichiometric complex with the tetraspanin CD151 and the interaction with CD151 is functionally important in α3 integrin-dependent matrix remodelling and cell spreading (Kazarov ). The aim of our study was to assess clinical significance of CD151 alone and in association with integrin α3β1 in a group of patients with ILC. The findings were analysed in context of our recent IDC study demonstrating a prognostic value of an expression of integrin CD151/α3β1 complex in patients with HER2-negative tumours (Novitskaya ).

Materials and methods

Patient selection and samples

The study included radical surgery specimens of primary ILC from 117 patients treated between 2000 and 2008 at three sites: (1) the Oncology Department of Copernicus Memorial Hospital in Łódź, Poland, (2) the UHB NHSFT, Birmingham, UK and (3) the Leeds Teaching Hospitals NHS Trust, Leeds, UK. All samples were obtained according to the local ethical regulations (project ethics licence: RNN/174/11/KE). The characteristics of the study population are summarised in Table 1. Follow-up period was defined as the time from surgery to the last observation for censored cases or death for complete observations.

Table 1

Patient characteristics

Feature	ILC	IDC
Number of patients	117	182
Age (years)
<50	21	56
⩾50	96	127
Disease stagea
I	34	44
II	60	93
III	13	45
T statusb
T1	52	61
T2	48	112
T3	15	1
T4	1	8
Gradec
1	21	105
2–3	94	77
Nodal statusd
Negative	64	92
Positive	44	90
Steroid receptor statuse
Negative	11	79
Positive	102	103
HER2 statusf
Negative	102	151
Positive	4	31

Abbreviations: IDC=invasive ductal carcinoma; ILC=invasive lobular carcinoma.

In the ILC group, data are available for a107 patients, b116 patients, c115 patients, d108 patients, e113 patients and f106 patients.

Immunohistochemistry

Serial 5 μm paraffin sections of formalin-fixed blocks were processed for immunohistochemistry for CD151 (mouse anti-human; 1 : 100; Novocastra, Newcastle upon Tyne, UK) and α3β1 (INTA3) (goat anti-human; 1 : 200, Santa Cruz, Wembley, UK) using protocols described previously (Novitskaya ). Immunostaining for E-cadherin (mouse anti-human, 1 : 50; Novocastra) was used to confirm the initial pathological diagnosis of ILC (E-cadherin-positive samples were excluded from the study). As a negative control for the immunostaining, primary antibodies were replaced by non-immune sera.

Scoring of immunostaining for CD151 was based on Hofmann's method for membranous reactivity of ErbB2 (Hofmann ) and modified as follows: (1) 0/negative: no reactivity or only partially membranous reactivity in ⩽10% of tumour cells; (2) 1+/negative: faint membranous or partially membranous in ⩾10% of tumour cells; (3) 2+/positive: weak to moderate complete membranous in ⩾10% of tumour cells; and (4) 3+/positive: strong complete membranous in ⩾10% of the tumour cells. Immunoreactivity for α3β1 was scored as follows: (1) 0/negative: no reactivity; (2) 1+/positive: faint membranous and/or cytoplasmic staining in ⩾10% of tumour cells; (3) 2+/positive: weak to moderate membranous and/or cytoplasmic staining in ⩾10% of tumour cells; and (4) 3+/positive: strong membranous and/or cytoplasmic staining in ⩾10% of the tumour cells. Immunohistochemical staining was evaluated and scored independently by two observers (HMR and R Kordek or HMR and SC). The agreement on staining intensity was >90%. Where there was disagreement, intensity was determined by consensus. Dichotomisation of the final scores into: (1) ‘negative' and (2) ‘positive' for CD151/0–2; INTA3/0 and CD151/3 and INTA3/1–3, respectively, was guided by intensity of immunostaining in positive controls recommended by the manufacturer.

Statistical analysis

Overall survival was calculated from the date of surgery to the date of death or the last follow-up, as recommended by the Kaplan–Meier method. Differences in survival distributions were compared using log-rank test. Data for patients who died from other causes than BCa were censored at the time of death. Univariate analysis of overall survival was performed using the Cox proportional hazards regression model. Parameters showing a significant correlation (P<0.05) were included in the multivariate logistic regression analysis. Pearson's χ2 test or Fisher's exact test were used to assess the associations between expression of CD151 and α3β1 alone and their coexpression and clinicopathological variables. The results were considered statistically significant when two-sided P-value was <0.05. The analyses were performed using the Statistica 9.1 and Statistica 10 (StatSoft Inc., Tulsa, OK, USA) softwares.

Data on cancer recurrence were available in only four cases that precluded the DFS analysis.

Results

An impact of CD151/α3β1 on tumour biology differs between ILC and IDC

In normal gland, both CD151 and α3β1 showed moderate to strong, predominantly membranous immunoreactivity, confined to the basal and lateral surfaces of the myoepithelial cells, with no or very weak staining in luminal epithelial cells (Figure 1A). Similarly, in ILCs, CD151 and α3β1 were localised mainly to the membrane of tumour cells and there was a significant correlation (P=0.012) between levels of their expression (Table 2A). We observed four distinct patterns of immunoreactivity for CD151/α3β1: (1) CD151+/α3β1+ in 55 out of 117 (23.94%) cases; (2) CD151+/α3b1− in 28 out of 117 (23.94%) cases; (3) CD151−/α3β1− in 20 out of 117 (17.09%) cases and (4) CD151−/α3β1+ in 14 out of 117 (11.96%) cases (Figure 1B–E). Interestingly, the level of CD151 expression (but not expression of α3β1) on cancer cells was similar to that observed on endothelial cells of intratumoural vessels (Figure 2).

Figure 1

Expression of CD151 and (A) normal gland ( × 100; insets × 400). Both CD151 and α3β1 show moderate to strong, predominantly membranous immunoreactivity, confined to the basal and lateral surfaces of the myoepithelial cells, with no or very weak staining in luminal epithelial cells. (B) The ILC cells of similar areas of the tumour display various levels of CD151/α3β1 immunoreactivity representing four predominant phenotypes: (B) CD151+/α3β1+ (55 out of 117; 49.01%); (C) CD151+/α3β1− (28 out of 117; 23.94%) (D) CD151−/α3β1− (20 out of 117; 17.09%) and (E) CD151−/α3β1+ (14 out of 117; 11.96%).

Table 2

Association between CD151 and/α3β1 expression and clinicopathological features

(A)
	P-value
	ILC
Feature	α3β1 (high: n=69)	CD151 (high: n=83)	CD151/α3β1 (high: n=55)
α3		0.012
CD151	0.012
Size	0.574	0.047	0.381
Nodes	0.464	0.277	0.101
Stage	0.656	0.019	0.387
Grade	0.093	0.343	0.019
ErbB2	0.092	0.881	0.249
ER/PR	0.694	0.471	0.425

Abbreviations: ER/PR=oestrogen receptor/progesterone receptor; IDC=invasive ductal carcinoma; ILC=invasive lobular carcinoma. The bold values are statistically significant.

Figure 2

Expression of CD151 in intratumoural vessels ( × 400). Level of expression in endothelial cells of intratumoural vessels (asterisks) is consistent with that in epithelial cells of adjacent tumour (arrows) (A, high; B, moderate; C, negative).

There was no significant correlation between the level of α3β1 expression and any of tumour characteristics (Table 2A). On the other hand, expression of CD151 assessed alone was inversely associated with tumour size (P=0.047) and stage (P=0.019), thus indicating that CD151 might have an ability to suppress proliferation of cells and progression of disease. Interestingly, positivity for both CD151 and α3β1 (CD151+/α3β1+) was significantly associated only with grade (P=0.019). These results suggest that phenotypic dedifferentiation and adopting of anaplastic features by epithelial cells in ILC are associated with CD151 acting in a complex with α3β1.

We have previously reported that in IDC, in contrast to ILC, CD151 expression showed a positive association with stage (P=0.030) and inverse with tumour grade (P=0.041). We have also demonstrated that, when assessed in combination, positivity for CD151 and/or α3β1 correlated closer than CD151 alone with stage of disease (P<0.001 vs P=0.03 for CD151/α3β1 vs CD151, respectively) (Novitskaya ). Here we have expanded our previous analyses and evaluated coexpression of CD151 and α3β1 in the context of histopathological and clinical characteristics of the IDC cohort. The results showed that in contrast to ILCs, combined positivity for both CD151 and α3β1 showed no correlation with any of the clinicopathological features in IDCs (Table 2B). Taken together, these results suggest that the involvement of CD151 and its principal transmembrane partner, the integrin α3β1, in tumour development and progression is likely to differ between histological subtypes of BCa.

In ILC, in contrast to IDC, neither CD151 nor α3β1 hold prognostic value

Neither CD151 nor α3β1 analysed alone were of any prognostic value in ILCs (Table 3A). In contrast, in IDC, as shown previously, CD151-positive patients had 1.88-fold higher risk of death from BCa in comparison with CD151-negative patients. Moreover, a multivariate statistical analysis identified CD151 as an independent marker (P=0.0172) of poor prognosis in IDC (Table 4). In neither ILC nor IDC, coexpression was significantly associated with patient survival. Furthermore, although in the IDC cohort, clinicopathological characteristics commonly recognised as independent prognostic factors (tumour size, lymph node status, stage) were significantly associated with poor survival, none of them held any prognostic value in our ILC study population (Table 3A and B).

Table 3

Univariate analysis of prognostic factors

(A)
Factor	HR	95% CI	P-value
	ILC
α3β1	0.80	0.39–1.61	0.750
CD151	0.56	0.28–1.14	0.109
CD151/α3β1	0.72	0.35–1.49	0.380
(T1 vs T2–4)	1.78	0.84–3.78	0.165
Nodal status	1.61	0.78–3.31	0.195
Stage (I vs II vs III)	2.08	0.85–5.11	0.111
ER/PR	0.46	0.17–1.19	0.110
HER2	NAa	NAa	NAa
Grade (G1–2 vs G3)	1.46	0.55–3.87	0.441
(B)
	IDC
α3β1	1.10	0.60–2.03	0.750
CD151	1.88	1.15–3.08	0.012
CD151/α3β1	0.98	0.48–1.98	0.952
(T1 vs T2–4)	1.77	1.30–2.40	<0.001
Nodal status	3.01	1.78–5.10	<0.001
Stage (I vs II vs III)	1.81	1.27–2.57	<0.001
ER/PR	0.53	0.33–0.87	0.011
HER2	2.07	1.18–3.64	0.012
Grade (G1–2 vs G3)	1.24	0.76–2.01	0.383

Abbreviations: CI=confidence interval; ER/PR=oestrogen receptor/progesterone receptor; HR=hazards ratio; IDC=invasive ductal carcinoma; ILC=invasive lobular carcinoma; NA=not analysed.

Only 4 were HER2 positive. The bold values are statistically significant.

Table 4

Multivariate analysis of prognostic factors

	IDC
Factor	HR	95% CI	P value
CD151	1.92	1.12–3.31	0.0172
(T1 vs T2-4)	1.98	0.94–4.15	0.0711
Nodal status	4.11	2.04–8.29	<0.0001
Stage (I vs II vs III)	0.68	0.38–1.23	0.2063
ER/PR	0.53	0.32–0.91	0.0198
HER2	1.20	0.71–2.04	0.1927

Abbreviations: CI=confidence interval; ER/PR=oestrogen receptor/progesterone receptor; HR=hazards ratio; IDC=invasive ductal carcinoma. The bold values are statistically significant.

Lack of CD151/α3β1 expression is associated with poor survival in node-negative ILC

As CD151, acting in partnership with α3β1, is thought to affect invasive spread of tumour cells, we analysed prognostic values of CD151 and/or α3β1 in the ILC cohort in relation to the lymph node status. The results presented in Table 5 demonstrate that neither CD151 nor α3β1 assessed alone or in combination had any significant prognostic value in any of the subgroups. Only a trend towards statistical significance (P=0.082) was seen for CD151 in node-negative patients (ILC N(−)). Furthermore, in the ILC N(−) subgroup, the data were suggestive that the presence of either CD151 or α3β1 could be favourable to the prognosis but their coexpression had no additive effect on patient survival. Thus, we next looked at the relationships between CD151 and/or α3β1 and tumour characteristics in the lymph node-negative subgroup (characteristics of the group are summarised in Table 6A). Table 6B demonstrates that, as in the whole group, significant correlations between (1) expression of CD151 and α3β1 (P=0.018); (2) CD151 and tumour size (P=0.035); and (3) CD151/α3β1 and grade (P=0.029) were maintained. Univariate analysis in ILC N(−) cases showed that lack of combined expression of CD151 and α3β1 was significantly correlated with poor patient survival (P=0.034) and was the only prognostic factor in this group (Table 6C). The CD151/α3β1− negative patients had 3.12-fold higher risk of death from BCa in comparison with the rest of the patients with no lymph node involvement and the 5-year estimated survival rates were 64% vs 91%, respectively (Figure 3).

Table 5

Univariate analysis of prognostic value of CD151 and/or α3β1 in relation to lymph node status

	ILC N(+)			ILC N(−)
Factor	HR	95% CI	P-value	HR	95% CI	P-value
α3β1	1.28	0.45–3.60	0.641	0.46	0.16–1.31	0.146
CD151	0.64	0.22–1.87	0.412	0.40	0.14–1.22	0.082
CD151/α3β1	1.02	0.35–3.00	0.970	0.72	0.35–1.49	0.187

Abbreviations: CI=confidence interval; HR=hazards ratio; ILC=invasive lobular carcinoma.

Table 6A

Lymph node-negative subgroup. Patient characteristics

Feature	ILC
Number of patients	64
Age (years)
<50	10
⩾50	54
Disease stagea
I	34
II	28
III	1
T statusa
T1	34
T2–T4	29
Gradeb
1	13
2–3	49
Steroid receptor statusc
Negative	5
Positive	58
HER2 status
Negative	64

Abbreviation: ILC=invasive lobular carcinoma.

Data available for 63 patients, 62 patients and 63 patients.

6B

Association of CD151 and/or α3β1 and tumour characteristics

	α3β1 (high: n=38)	CD151 (high: n=47)	CD151/α3β1 (high: n=32)
α3β1		0.018
CD151	0.018
Tumour size (T1 vs T2–4)	0.312	0.035	0.892
Grade	0.079	0.342	0.029
HER2	NAa	NAa	NAa
ER	0.377	0.999	0.192
PR	NAb	NAb	NAb

Abbreviations: ER/PR=oestrogen receptor/progesterone receptor; NA=not analysed.

All tumours were HER2 negative.

Lack of data for 45 cases. The bold values are statistically significant.

6C

Univariate analysis of prognostic factors

	ILC N(−)
Factor	HR	95% CI	P-value
α3β1	0.462	0.16–1.31	0.146
CD151	0.402	0.14–1.22	0.082
CD151/α3β1	0.724	0.35–1.49	0.187
CD151−/α3β1−	3.125	1.09–8.99	0.034
Tumour size (T1 vs T2–4)	1.780	0.61–5.18	0.290
Grade	1.361	0.36–5.14	0.650
ER	0.372	0.08–1.72	0.206

Abbreviations: CI=confidence interval; ER=oestrogen receptor; HR=hazards ratio; ILC=invasive lobular carcinoma. The bold values are statistically significant.

Figure 3

Kaplan–Meier curves of overall survival for patients with invasive lobular breast cancer. Overall survival of patients with tumours negative for both CD151 and α3β1 in relation to the rest of the cohort.

Discussion

An overexpression of CD151/Tspan24 has been repeatedly reported as a negative predictor of overall survival in patients with invasive ductal breast carcinoma (Yang ; Sadej ; Kwon ; Novitskaya ) but nothing is known about its clinical significance in invasive lobular BCa. Here we demonstrate that, in contrast to IDC, loss of CD151 expression in complex with that of its principal molecular partner, the integrin α3β1, is significantly associated with poor survival in patients with lymph node-negative ILC. There have been a few attempts to establish an expression signature of a risk of distant metastasis in primary lymph node-negative IDC (Mirza ; Wang ; Tutt ). To the best of our knowledge, this is the first study providing information on prognostic factors in a subgroup of lymph node-negative ILC patients.

The results of our study provide support for the notion of biological diversity of CD151 in human cancer. Acting alone and/or in complex with laminin-binding integrins, CD151 has been linked with various aspects of carcinogenesis (Palmer ; Sadej ). Although its pro-migratory and pro-invasive functions are well established, it was also reported that in certain cell types the presence of CD151 is associated with suppression of motility. For example, its downregulation induced by hypoxic stress at the primary site in colon cancer was shown to decrease adhesion of cells to the extracellular matrix and enhance cell motility (Chien ). Similarly, both up- and downregulation of integrin α3β1 expression have been correlated with tumour invasion and poor prognosis (Adachi ; Fukushima ; Gustafson-Wagner and Stipp, 2013). Furthermore, the CD151/α3β1 complex has been recently shown to suppress ovarian tumour growth (Baldwin ). Although in our study no prognostic value could be ascribed to either CD151 or α3β1 assessed alone, their independent suppressor-like functions cannot be excluded. Further mechanistic investigation would be required to evaluate an actual impact of a loss of the CD151/α3β1 complex on ILC tumour behaviour.

Increasing evidence suggests that ILCs and IDCs are biologically different. This is reflected in disparities in morphological and phenotypic characteristics, transcriptomic profiles related particularly to cell–cell and cell–matrix interactions (Weigelt ), as well as responsiveness to neoadjuvant therapy (Korkola ). As demonstrated by several genomic studies, the only consistent finding characterising ILC tumours is inactivation of CDH1 (E-cadherin, cadherin-1) (Bertucci ; Weigelt , 2010; Sikora ) and, indeed, lack of E-cadherin expression is considered a hallmark of ILC. Although the role of E-cadherin in tumour onset and progression is still largely unknown, its inactivation alone is clearly not sufficient to induce neoplastic growth. It has been shown that combined loss of E-cadherin and p53, but not E-cadherin alone, in murine mammary epithelial cells induces metastatic carcinomas that resemble human ILC (Derksen ).

There are several reports suggesting that an association of CD151 with integrin α3β1 might be important for regulation of E-cadherin-dependent cell–cell interactions. In mouse kidney, the CD151/integrin α3β1, acting as both an organiser and a component of a large multimolecular complex containing E-cadherin–β-catenin, promoted its association with the actin cytoskeleton and cadherin-mediated cell–cell adhesion. Deletion of integrin α3β1 in this system was found to disturb E-cadherin localisation and function (Chattopadhyay ). In highly expressing E-cadherin human keratinocytes (HaCat cell line), blocking of CD151 supported cell dispersal (Chometon ), whereas CD151 overexpression enhanced carcinoma cell–cell association (Shigeta ). Similarly, in A431 epithelial carcinoma cells, near total silencing of CD151 destabilised E-cadherin-dependent cell–cell junctions. However, it was not the disruption of the E-cadherin regulatory complex but an excessive RhoA activation and disorganisation of actin fibres at the cell–cell junctions, induced by loss of CD151, that led to the enhancement of cell migration (Johnson ). Through stabilisation of E-cadherin based cell–cell interactions, CD151 was suggested to counteract the metastatic progression of endometrial cancer (Voss ). Activation of Rho/ROCK signalling axis triggered by loss of E-cadherin was recently demonstrated to be responsible for induction of anoikis resistance and invasive phenotype in a mouse model of human ILC (Schackmann ). Interestingly, results of our own study have shown that in IDC cells, depletion of CD151 and α3β1 caused increase of active RhoA (Novitskaya ). Taken together, these findings seem to suggest that although loss of E-cadherin is a prerequisite of the lobular phenotype, other factors, including, perhaps, the CD151-α3β1 partnership, also contribute to invasive behaviour of cancer cells.

The results of our study demonstrated that unlike IDC, lack of both CD151 and integrin α3β1 correlated with decreased survival of patients with lymph node-negative ILCs. Unlike IDC, ILC tumour cells are deprived of E-cadherin expression and the function and significance of the CD151/α3β1 complex in biology of E-cadherin-negative cells are poorly characterised. However, it is conceivable that in this particular biological context, as in the settings described above, the CD151/α3β1 complex is controlling cell–cell adhesion and loss of E-cadherin contributes but is not decisive in destabilisation of cell–cell contacts and enhancement of cell migration.

The pathogenesis of IDC and ILC seems indeed to be governed by distinct mechanisms. Not only the high expression of CD151, but also the lymph node status, one of the most clinically important indicators of poor prognosis in IDC, was not correlated with survival of patients in our ILC study population. Instead, our results showed that the combined absence of CD151 and α3β1 was a negative prognostic factor in patients with no lymph node involvement. This unexpected finding gives a new insight into the biology of ILC and possible contribution of CD151 and α3β1 to disease progression. CD151 is highly expressed in endothelial cells and is thought to regulate pathologic angiogenesis. It has also been demonstrated that CD151 plays an important role in maintaining integrity of endothelial layer (Zhang ). Our immunohistochemical analysis showed that the level of CD151 expression on endothelial cells was correlated with that seen on tumour-associated endothelium. Although the mechanisms underlying this ‘phenotypic unity' remains unknown, it is possible that the compromised integrity of CD151-negative endothelium would facilitate intravasation of CD151-negative cells. Whatever the underlying mechanisms, the results of our study indicate that loss of CD151-α3β1 may serve as a potential prognostic marker of poor survival in a subgroup of patients deemed to carry a low risk of cancer-related death (ie, lymph node negative). This could have important clinical implications by helping to identify patients likely to benefit from adjuvant therapy.

In summary, the results of our comparative analysis of clinical significance of CD151/integrin α3β1 in ILC and IDC highlight considerable differences in biology between these two BCa types. Furthermore, the study suggests that evaluation of the level of CD151/α3β1 expression might provide important information on behaviour of ILC tumours and identify patients with increased risk of distant metastases, commonly considered ineligible for routine adjuvant therapy.