Introduction

Triple-negative breast carcinomas (TNBC) have been shown to be a distinct entity displaying aggressive features, both biologically and clinically [1]. These tumours are typically oestrogen receptor (ER)-negative, progesterone receptor (PR)-negative and lack the over-expression or amplification of human epidermal growth factor receptor (HER2) and hence therapeutic options other than cytotoxic treatments are often limited. They account for between 10% and 20% of all breast cancer patients but patients with this phenotype tend to fare worse prognostically, with shorter disease-free and overall survival [2]. Although this may be accounted for by the paucity of suitable treatments, it has been suggested that triple-negative breast cancers are themselves a separate biological disease [1]. Gene expression profiling has identified five major molecular subtypes of breast cancer: luminal A and B (both ER+), normal breast-like, basal-like and HER2+ (all ER-) [3–5]. Although most TNBC possess a basal phenotype, they are not synonymous with basal-like breast cancers, with a discordance of up to 30% described in the literature [2, 6]. Their behaviour is contrary to that of the more favourable prognostic group of ER+/PR+/HER2- breast cancers.

Functional magnetic resonance imaging techniques such as dynamic contrast-enhanced (DCE) and dynamic susceptibility contrast (DSC)-enhanced MRI have the ability to document tumour vascularisation and interrogate the extracellular extravascular space, with particular MRI-derived parameters reflecting different aspects of the tumour microenvironment [7]. These non-invasive methods are particularly attractive with regard to being able to probe the vascularity of triple-negative breast cancers and to identify potential biological markers suitable for specific targeted therapies. Previous imaging studies have evaluated the morphological and metabolic characteristics of TNBC in comparison with ER+/PR+/HER2 cancers with both MRI and PET-CT respectively, but as far as we aware, there is a relative lack of data regarding both the tumour stromal environment and the vasculature [8, 9].

The purpose of this study was to investigate whether breast cancers displaying the triple-negative phenotype exhibit unique vascular characteristics compared with the more common and prognostically more favourable ER+PR+HER2- group, and to quantify these differences using DCE-MRI and DSC-MRI.

Materials and methods

Study design

Eighty-three women (median age 46 years; range, 26–72) with newly diagnosed, histologically-proven primary breast cancer underwent MRI before neoadjuvant chemotherapy (NAC) as part of two prospective studies investigating the role of MRI in response to NAC [10]. All research was performed in accordance with the Helsinki Declaration. Approval was sought from the local institutional ethics review board (West Hertfordshire Hospitals protocol number EC2001-26) and informed consent obtained from all patients. We retrospectively identified ER-/PR-/HER2- or ER+/PR+/HER2- cancer cohorts from pre-treatment core biopsy specimens. Clinical staging at diagnosis was performed in accordance with the TNM (Tumour, Node, Metastasis) staging classification of malignant tumours as described by the International Union against Cancer Criteria [11].

Histopathological scoring of hormone and HER2 receptor status

Histological information was acquired from pre-treatment, diagnostic core biopsy specimens and graded according to the Bloom-Richardson scoring system [12]. Low and moderate grade tumours were grouped together as there were too few low-grade lesions for analysis. Oestrogen receptor (ER) and progesterone receptor (PR) scoring was performed according to the Allred score, a composite of the percentage of cells that stained (scored from 0 to 5) and the intensity of their staining (rated as 1–3), using immunohistochemical methods [13]. The threshold for hormone receptor positivity was taken as a composite score of 3 and above. HER2 (human epidermal growth factor receptor) status was deemed to be positive if 3+ and negative if 0 or 1+ on immunohistochemistry. Tumours which were 2+ were sent for further testing and classified as positive if amplified (≥2.1 copies) on FISH (fluorescence in-situ hybridisation) testing.

MRI data acquisition and analysis

The MRI examinations were performed on a 1.5 T Siemens Symphony (Siemens Medical Systems, Erlangen, Germany) using a dedicated bilateral breast coil. Initial diagnostic coronal and axial T1-weighted and T2-weighted MRI were obtained through both breasts. Proton density-weighted gradient recalled echo images were acquired sagittally (TR 350 ms, TE 4.7 ms, flip angle (α) 6˚) for four slices (three through tumour and one through the contralateral normal breast). Dynamic T1-weighted (relaxivity-weighted) images (TR 11 ms, TE 4.7 ms, α 35°, 256 × 256 matrix) matched to the proton density images were then acquired at the same slice positions every 12 s for a total imaging time of 8 min (40 sets) for the purposes of T1 mapping [14, 15]. Intravenous gadopentetate dimeglumine [0.1 mmol/kg body weight Gd-DTPA, Magnevist, Bayer-Schering] was injected at 4 ml/s during the fifth acquisition time point followed by a 20 ml saline flush at 4 ml/s. Dynamic T2*-weighted (susceptibility-weighted) images (TR 30 ms, TE 20 ms, (α) 40°, 128 × 128 matrix, central slice) using 0.2 mmol/kg IV Gd-DTPA injected at 4 ml/s after 20 s, were then acquired every 2 s over 2 min (60 sets).

Images were analysed using specialist MRI software (MRIW version 4.3, Institute of Cancer Research, London, UK) [16] with whole tumour regions of interest (ROI) drawn on DCE-MRI T1-weighted subtraction images by a single observer. DCE-MRI analysis was performed using pharmacokinetic modelling of contrast kinetics according to the 2-compartment Tofts model [17] and a modified Fritz-Hansen assumed arterial input function [18–20]. Values for quantitative kinetic parameters Ktrans (inflow transfer constant; min−1), ve (leakage space; %), kep (outflow rate constant; min−1) and semi-quantitative parameter IAUGC60 (initial area under the gadolinium-time curve over 60 s; mM/s−1) as well as T2*-weighted quantitative parameters rBV (relative blood volume; arbitrary units (AU)), MTT (mean transit time; s) and rBF (relative blood flow; AU) were calculated pixel-by-pixel.

Ktrans is a transfer constant measuring the rate of transport of contrast medium from plasma to the extravascular extracellular space (EES) and provides a measure of vascular permeability and blood flow; ve is the fraction of the tumour volume occupied by the EES, whilst kep describes the outflow rate of contrast medium from the EES back to plasma. These parameters are related by the equation:

$$ {{\hbox{k}}_{\rm{ep}}} = {{\hbox{K}}^{\rm{trans}}}/{{\hbox{v}}_{\rm{e}}} $$

Dynamic T2*-weighted parameters are derived from the changing tissue signal intensities corresponding to the first passage of a bolus of contrast agent in the vascular space [16]. By fitting a gamma variate model function, relative blood volume (rBV), mean transit time (MTT) and relative blood flow (rBF) can be measured. MTT is the mean time for contrast agent to perfuse a region of interest. According to tracer dilution theory, rBV is the area under the fitted curve and MTT is the full width half maximum of the fitted peak, with these constants related by the following central volume theorem:

$$ {\hbox{BF}} = {\hbox{BV}}/{\hbox{MTT}} $$

Statistical analysis

Statistical analysis was performed using the StatsDirect software package (Sale, Cheshire, UK). Median values for baseline parameters were calculated from pixel histograms of kinetic parameters because of their non-normal distributions. MRI-derived kinetic parameters were compared across receptor status and with other known prognostic variables (age, grade, tumour stage, nodal status) using the Mann-Whitney U test. Analysis of Receiver Operating Characteristic (ROC) curves for each parameter was used to elucidate the parameter best able to identify the triple-negative phenotype.

Results

In total, 37 patients were suitable for analysis, 16 of whom were identified as ER-/PR-/HER2- and 21 as ER+/PR+/HER2-. 12 patients did not undergo their first MRI, one patient’s tumour was not visible on MRI, full receptor status was not available in 11 and 22 patients with other receptor phenotypes were excluded. TNBC comprised 19% of the total study population with a median age of 42.5 years (range 34–57) and median tumour size of 60 mm (range 40–100), with most being of the invasive ductal subtype (88%). In comparison, in the ER+/PR+/HER2- group, the median age was 49 years (range 26–70) and median tumour size 50 mm (range 25–150) with 71% exhibiting ductal histology. There were significantly more high-grade tumours in the TNBC group (p = 0.006). The characteristics of each group are provided in Table 1.

Table 1 Characteristics of ER-/PR-/HER2- and ER+/PR+/HER2- patient groups
Full size table

Overall significantly higher values for kep (0.70 vs 0.56, p = 0.044), and lower values for ve (0.33 vs 0.39, p = 0.001) and MTT (44.27 vs 47.69, p = 0.007) were observed in TNBC (Table 2; Fig. 1). Both rBV and rBF were higher in TNBC but these results were not statistically significant. ROC curve analysis revealed ve to be the best identifier of triple negativity (sensitivity 81%, specificity 76%, area under ROC curve 0.80). Representative images with corresponding histograms comparing ve and kep in TNBC with ER+/PR+/HER- are shown in Fig. 2.

Table 2 Comparison of dynamic MRI kinetic parameters in triple negative breast cancers with ER+/PR+/HER2- breast cancers
Full size table
Fig. 1
figure 1

Box and whisker plots of MRI kinetic parameters (median, interquartile range, 5th and 95th percentile values) illustrating significant differences between the TNBC and ER+PR+HER2- patient groups. a TNBC: ve 0.33 (0.28–0.35), ER+PR+HER2-: ve 0.39 (0.37–0.44); p = 0.001. b TNBC: kep 0.70 (0.60–0.93), ER+PR+HER2-: kep 0.56 (0.48–0.66); p = 0.044. c TNBC: MTT 44.27 (38.31–46.13), ER+PR+HER2-: MTT 47.69 (45.74–49.69); p = 0.007

Full size image
Fig. 2
figure 2

Representative parametric maps with corresponding whole tumour histogram depictions of: a lower ve values in (i) TNBC compared with (ii) ER+/PR+/HER2- BC b higher kep values in (i) TNBC compared with (ii) ER+/PR+/HER2- BC in two patients matched for tumour size (T2), lymph node status (N0), grade (G3) and histological subtype (invasive ductal carcinoma)

Full size image

No correlation was found between age and any MRI kinetic parameters. When stratified according to tumour grade within each cohort, kep values were higher in high-grade lesions compared with low-grade TNBC (0.82 vs 0.51, p = 0.019) as were rBV (276.14 vs 79.25, p = 0.008) and rBF (6.16 vs 1.87, p = 0.014) (Table 3) but not within ER+/PR+/HER2- cancers. Comparisons were also made between TNBC and ER+/PR+/HER2- groups. Values for ve (0.33 vs 0.40, p = 0.037) and MTT (40.49 vs 47.55, p = 0.008) remained significantly lower but only in low-grade TNBC. This effect was not seen in high-grade tumours (Mann-Whitney U test; p > 0.05).

Table 3 Comparison of MRI kinetic parameters in Triple negative breast cancers with ER+/PR+/HER2- breast cancers according to grade
Full size table

For tumour stage comparisons, T3 and T4 breast cancers were grouped together as there were too few T4 tumours for separate analysis. Values for ve were significantly lower across all T stages in TNBC (T2 0.30 vs 0.38, p = 0.035; T3/4 0.34 vs 0.41, p = 0.012) compared with the ER+PR+HER2- group. Although values for kep were higher and MTT shorter in TNBC across all T stages, these results did not reach statistical significance.

Values for ve were also significantly lower for node-negative TNBC (0.33 vs 0.41, p = 0.004) and in node-positive breast cancer, MTT was shorter in the TNBC group (43.96 vs 47.58, p = 0.008) (Table 4). kep was once again higher in the TNBC group across all nodal stages but results did not reach statistical significance.

Table 4 Comparison of MRI kinetic parameters in Triple-negative breast with ER+/PR+/HER2- breast cancers according to nodal status
Full size table

Discussion

Triple-negative breast cancers constitute only 10% to 20% of all breast cancers but treatment is limited by the lack of effective targeted therapies and by their aggressive clinical behaviour, which is typically characterised by earlier relapses and worse outcomes [5]. The clinical characteristics of the TNBC population in this study are similar to those previously reported, in that women tend to be younger, have an increased likelihood of lymph node involvement and higher grade tumours in comparison to hormone receptor-positive, HER2 receptor-negative patients.

It is now emerging that triple-negative breast cancers have distinct clinical and pathological features. Histopathological characterisation has revealed these tumours to be of higher grade, with higher mitotic indices and Ki67 counts. Other features have included scant stromal content, central necrosis, pushing margins of invasion, a stromal lymphocytic infiltrate and multiple apoptotic cells [1, 21, 22]. Molecular profiling has revealed the differential expression of genes associated with cell division and proliferation and intracellular DNA repair between both subtypes [23]. To date, there has been a paucity of studies evaluating the functional imaging characteristics of TNBC. Studies have revealed distinct morphological features on mammography and MR imaging [8, 24–26] with fluorine-18 fluorodeoxyglucose (FDG)-positron emission tomography (PET) demonstrating a higher sensitivity in detecting these tumours because of enhanced FDG uptake [9]. Increased FDG uptake is likely to be related to upregulation of glucose transporters which may be related to constitutive or hypoxia-mediated upregulation of HIF-1 alpha [27]. Activation of HIF-1 also results in increased tumour angiogenesis which in turn results in a chaotic tumour vascular bed with arterio-venous shunts as discussed below.

We have demonstrated lower values of ve in TNBC with DCE-MRI, which values of ve in TNBC, which describes the fraction of the tumour volume occupied by the EES. This is consistent with a more compact, tightly packed and a highly cellular microenvironment. Values for ve also remained significantly lower in low-grade TNBC in comparison with ER+PR+HER2- tumours despite there being a higher number of high-grade tumours in TNBC. Although ve was lower in high-grade TNBC in comparison with the ER+PR+HER2- arm, these results were not statistically significant which may be partly accounted for by the smaller patient numbers in the latter. Furthermore, ve was reduced in the triple-negative phenotype for all T stages and node-negative patients. Although this was not seen in the node-positive population, lymph node positivity, a reflection of a more aggressive primary breast tumour and corresponding underlying tumour hypercellularity, may be an overriding influence over hormonal and HER2 status on ve values.

In preclinical breast cancer models, tumour-derived fibroblasts in the stroma are able to induce greater cell proliferation because of higher basic fibroblast growth factor (bFGF) levels in hormone receptor-negative tumours. Gene expression profiling has also revealed that stromal expression can vary amongst breast carcinomas, as it is dependent on the differential expression of a set of extracellular matrix-related (ECM 1-4) genes [28]. ECM1 tumours were overrepresented by basal-like tumours and highly associated with lymphoid infiltration, which in turn, can stimulate fibroblastic production of matrix components [28]. Furthermore, patients with an ECM1 tumour profile had worse outcomes suggesting that stromal composition may be a factor in the poorer prognosis of TNBC.

Preclinical evidence supports not only increased cellularity but also increased neovascularisation in TNBC [29]. Greater intratumoural levels of the angiogenic mediator, VEGF (vascular endothelial growth factor) have been reported, with levels up to three times greater than that seen in ER+/PR+ tumours [30]. Other studies have reported p53 mutations in up to 80% of TNBC compared with 20% in other breast cancers, which are associated with higher plasma and intratumoural VEGF levels [31, 32]. Wild-type p53 downregulates the VEGF promoter and is involved in the regulation of thrombospondin-1, an inhibitor of angiogenesis. Higher bFGF levels have also been implicated in increased angiogenesis via VEGF upregulation in breast cancer cells [33]. The consequent hypervascularisation results in structurally and functionally abnormal tumour vessels, multiple arterio-venous shunts and loss of blood flow regulation [34]. In this study, the shorter mean transit times in TNBC are in keeping with hyperdynamic blood flow in part due to blood shunting, with proportionally larger increases in blood flow relative to blood volume seen within this tumour type (MTT = BV/BF). Similarly, higher kep values in the TNBC group reflect the rapid return of contrast medium into the vasculature which is consistent with increased capillary permeability.

Given these differences in tumour vascular characteristics, anti-angiogenic therapy may be an appropriate targeted treatment for TNBC. There are a number of clinical trials currently addressing the benefit of anti-VEGF therapy in women with TNBC. So far, subset analyses of patients with triple-negative cancers in completed trials of bevacizumab in metastatic breast cancer have been encouraging, with the AVADO trial of bevacizumab (7.5 mg/kg vs 15 mg/kg), in addition to docetaxel in the first-line metastatic setting, demonstrating an improvement in PFS from 6.0 to 8.1 months (HR = 0.60, 95% CI, 0.39–0.92) in the 15 mg/kg arm [35, 36]. In the E2100 trial, there were also improvements in PFS from 4.7 to 10.2 months (HR = 0.45; 95% CI, 0.33–0.61) [37]. Subset analysis in RIBBON I, a phase III trial of bevacizumab in the first- and second-line treatment of metastatic breast cancer also showed improvements in PFS regardless of chemotherapy type (4.2 to 6.1 months in the capecitabine cohort (HR = 0.72, 95% CI, 0.49–1.06); and from 8.2 to 14.5 months in the taxotere/anthracycline cohort (HR = 0.78, 95% CI, 0.53–1.15)), although these differences were not statistically significant [38]. Currently there are several studies assessing the role of bevacizumab in early TNBC including BEATRICE in the adjuvant setting and CALGB 40503 in the neoadjuvant setting.

Limitations

Given the limited study population and small patient numbers in each group, this study was not conducive to a multivariate analysis. We examined two subsets only; the triple-negative and ER+/PR+/HER2- groups as the latter confers a better prognosis, and further studies comparing the triple-negative group with other breast cancers are needed. We recognise that although multiple statistical comparisons were performed in this analysis where the sample size was small this was an exploratory study designed to be hypothesis generating. However further validation in a larger patient population is warranted. To our knowledge, findings using fully quantitative methods of MRI kinetic analysis in contrast to semi-quantitative techniques such as enhancement kinetics [25, 39] have not been widely reported in the literature. A recent study demonstrated more rapid initial enhancement and higher washout rates in ER-/PR- breast cancers (as distinct from our defined subset of ER-/PR-/HER2- tumours); however like many studies, methods of data acquisition differed with longer temporal resolution times and no accounting for first pass effects [39].

Values for Ktrans, a widely accepted DCE-MRI-derived measurement of perfusion and permeability, were not increased in TNBC in this study. However this transfer constant is not always an ideal parameter for reflecting vascularity. Any conditions influencing blood perfusion including patient factors such as cardiac output and hypertension which are difficult to control for in the clinical setting may potentially confound measurements of Ktrans. kep on the other hand, provides a robust method of describing actual tumour capillary permeability because it is disconnected from perfusion. Although we appreciate that a mathematical relationship exists between kep and ve: there is however, an underlying physiological justification for this. It is also important to mention that the compartmental pharmacokinetic modelling utilised in MRI analysis assumes that contrast agent resides and exchanges between two compartments, the vascular space and the EES, and hence may be an oversimplification of underlying tumour physiology.

Furthermore, triple-negative breast carcinomas are a heterogeneous subset of tumours and unlikely to represent a uniform entity [40]. Tumours were large in this study reflecting the neoadjuvant setting in which these patients were recruited. We recognise that although necrotic tumours may potentially influence our results, this was not the case in our study, but future studies incorporating pixel enhancement analysis and diffusion-weighted MRI may be of value in studying this further. Imaging studies investigating the underlying biological and functional features of these tumours will ultimately lead to an improved understanding of this subset of breast cancers. In particular, other multiparametric MRI techniques such as diffusion-weighted MRI which interrogates the extracellular space and intrinsic susceptibility-weighted MRI which can potentially inform on blood tumour oxygenation, may provide further information and reinforce our findings [41, 42].

Conclusions

In conclusion, the lower values of ve and MTT, and higher kep in TNBC as elucidated by DCE-MRI, are in keeping with accumulating evidence that these tumours exhibit distinct underlying pathophysiological characteristics and clinical features in comparison with ER+/PR+/HER2- breast cancers. Dynamic MRI findings from our study were consistent with increased stromal cellular density and vascular permeability, suggesting that benefit may be gained from targeted therapies, in particular those directed at disrupting tumour angiogenesis.