Main

The clinical classification of breast carcinoma mainly on the basis of morphology is limited at present. In fact, patients with the same diagnosis can have different clinical outcomes, and this has been recently highlighted by several gene-expression studies. The classification in four main molecular subtypes supports their distinct biology. Furthermore, this information enables clinicians to tailor individual treatments.1, 2 Nevertheless, there is a need to redefine their molecular characteristics as they are heterogeneous.3

In patients with early breast carcinoma treated conservatively, local recurrence occurs between 6 and 20%, which correlates with increased distant metastasis and shorter survival.4, 5, 6, 7, 8 Therefore, research of new factors to identify patients at higher risk is necessary. Besides, the impact of breast carcinoma subtypes on local or distant control has not been evaluated extensively.9, 10, 11, 12

Recent epidemiological and clinico-pathological data have supported the role of the insulin-like growth factor-1 (IGF1R) signaling system on tumor development and progression.13, 14, 15, 16, 17 In breast carcinoma, high levels of IGF1R have been detected in 30–82%,18, 19 but its prognostic value is controversial.20, 21, 22, 23, 24, 25, 26, 27 Emerging experimental and clinical data suggest that the IGF1R and ER/PR pathways are interactive.24 About two-third of breast carcinomas have an ER-positive status, but >50% of patients respond to anti-estrogen treatment that is related to PI3K/Akt signaling upregulation by ERα. The sustained activation/inactivation of the IGF1R, PI3K/Akt and ERK signaling pathways in which Bcl2 levels are relevant might also be responsible for the response.28 Furthermore, some studies suggest that IGF1R is an emerging therapeutic target in cancer.13, 29 The aim of our study was to determine the influence of IGF1R expression and gene status in immunophenotypes of early breast carcinoma.

Materials and methods

Case Retrieval

Patient characteristics have been described previously.30 Table 1 shows the distribution of tumors in relation to clinico-pathological features and immunophenotypes. In brief, a series of 197 lymph-node-negative breast carcinoma patients in the early stage (ie, stage I–II) treated with conserving surgery and radiation therapy were retrieved from the archives of the Department of Pathology at the University General Hospital of Alicante (January 1990–December 2001). All patients had undergone a complete gross excision of the primary tumor and axillary lymph-node dissection, with a minimum of five nodes removed. Our Institutional Review Board approved the study.

Table 1 Clinico-pathological features and IGF1R data in relation to immunophenotypes
Full size table

After surgery, radiation therapy of a total dose of 50 Gy to the tumor bed was given, including a boost (median 10 Gy). Patients with ER- and/or PR-positive tumors received tamoxifen for 2–5 years and those with high risk factors (such as young age, high grade or ER/PR-negative) were additionally treated with six cycles of cyclophosphamide, methotrexate and 5-fluoroucil chemotherapy or with four cycles of doxorubicin plus cyclophosphamide. None of the patients received taxanes, trastuzumab or aromatase inhibitors. Patients’ average age was 52 years (range 23–88 years) and the median follow-up was 124 months. There have been 33 (17%) in-breast relapses, including 7 with distant relapses as well, 20 (10%) with only distant relapses, 17 (9%) with contralateral breast carcinoma (1 synchronous and 16 metachronous) and 29 (15%) deaths.

Immunohistochemistry

Immunohistochemical staining was performed on paraffin-embedded tissues obtained from tissue microarrays using standard techniques, with antibodies and conditions detailed in Table 2. The staining was scored positive as follows: ER/PR ≥10%, Bcl2 ≥50%, p53 and Ki67 ≥20%, any degree for CK5/6 (cytoplasmic) or EGFR (membranous) and HER2 according to the scoring guidelines (>30% 3+). Both α- and β-phospho/active-IGF1R subunits (membrane and/or cytoplasm) were semi-quantitatively scored according to the percentage of positive cells and intensity (from 0 to 3+) (score 0–300).

Table 2 Panel of antibodies for the immunohistochemical analysis
Full size table

In Situ Hybridization Analysis

HER2 gene status was confirmed by fluorescence in situ hybridization (FISH) (Dako pharmaDxTM) or chromogenic in situ hybridization (CISH) (Spot light; Zymed) in nondefinitive cases (2+ and <30% 3+ cells).31

Quantitative RT-PCR and DNA Sequencing

DNA isolation method, PCR conditions, primers, as well as mutational and quantitative analyses of IGF1R have been described previously.30 In brief, we extracted genomic DNA using the QIAmp DNA mini-kit, and isolated total RNA by handling the RNeasy formalin-fixed paraffin-embedded kit (Qiagen, Hilden, Germany) from 2–3 paraffin-embedded cylinders (1-mm thick) from preselected tumor areas with at least 30% tumor cell content. Areas of necrosis and hemorrhage were excluded. A total of 150 primary breast carcinomas had sufficient formalin-fixed paraffin-embedded tissue available for RNA extraction, with sufficient quantity and quality. These included 30 with local recurrence, 14 with distant metastasis and 106 without recurrence. RT-PCR was performed in 20 μl reaction volumes with random hexamer priming, 10 μl of extracted RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems Inc., Santa Clara, CA, USA). TaqMan Gene-Expression Assays of IGF1R hCG1640727, β-glucuronidase (GUSB) hCG18478 and glyceraldehyde-3-phosphate dehydrogenase hCG2005673 were performed in the ABI Prism 7500 System (Applied Biosystems) in triplicate. RT minus controls and two commercial positive controls (total human and breast carcinoma RNA) were included. IGF1R quantification was normalized to GUSB, and the relative changes in gene expression were calculated by the ΔΔCT method.

Primer sets for exons 19 and 21 were obtained from RefSec DNA NM000875.2, designed using Primer3 (TIB Molbiol Syntheselabor, Berlin). PCR amplicons were double-strand sequenced using the ABI Prism BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) and the ABI Prism 310 automated sequencer.

Statistical Analysis

The χ2 or Fisher tests were used to determine the distribution of characteristics among the immunophenotypes; the nonparametric Spearman's correlation test was used to determine the relationship between the IGF1R protein and mRNA; and the Mann–Whitney or Kruskal–Wallis tests were used to determine the expressions’ differences between pathological features. Receiver operating characteristic curves and area under the curve were generated to determine a cutoff value of the IGF1R gene expression and the potential clinical utility to predict prognosis (recurrence and death). Kaplan–Meier survival plots and log-rank tests were performed for comparison of survival curves, and for multivariate analysis, the Cox proportional hazards model was applied. P-values <0.05 were considered statistically significant.

Results

Tumors were classified as luminal A (ER/PR ≥50% and Bcl2 ≥50%, p53 <20% or Ki67 <20%) in 40% (78/197) cases, as luminal B (ER and/or PR <50% and Bcl2 <50%, p53 ≥20% or Ki67 ≥20%) in 24% (48/197), as HER2-positive (≥30% 3+ or amplified by FISH or CISH) in 19% (37/197) and as basal/triple-negative (ER/PR/HER2-negative±CK5/6±EGFR) in 17% (34/197) cases. The luminal A subtype was more frequently present in older patients (43%) as a unilateral mass (42%) of ≤2 cm in size (44%), low grade (67%), without tumor necrosis (50%) and negative vascular invasion (40%) (Table 1). However, there were no statistically differences regarding immunophenotypes and local recurrence (13% for luminal A, 23% for luminal B, 22% for HER2 and 12% for basal/triple-negative tumors) or distant metastases (8% of luminal A, 8% of luminal B, 11% of HER2 and 18% of basal/triple-negative; P=NS). Interestingly, among HER2-positive patients, local recurrence was developed in none with PR-positive tumors, but in 60% of ER-positive/PR-negative tumors and in 40% of ER/PR-negative tumors (P=0.004).

Table 1 summarizes the clinico-pathological features and IGF1R data in relation to immunophenotypes. Total IGF1R correlated with pospho-IGF1R (active) and IGF1R mRNA (Spearman's correlation coefficient: r=0.289, P=0.001; r=0.204, P=0.013, respectively). The median with percentiles 25 and 75 (P25–P75) for each analyzed factor is given in Table 3. For the purpose of the study, primary tumors were classified as having α- or pospho-IGF1R low vs high expression, defined using the median scores as cutoffs.24 A high expression of α-IGF1R (Figure 1a) was more frequently seen in luminal A (49%) cases, followed by HER2-positive (22%), luminal B (20%) and basal/triple-negative cases (9%) (P=0.01). Similarly, pospho-IGF1R (Figure 1b) was predominantly detected in luminal A (49%) cases, and to a lesser extent among HER2-positive (27%), luminal B (15%) and basal/triple-negative cases (9%) (P=0.047). The median relative IGF1R mRNA for all analyzed primary tumors (n=150) was 2.32 (P25–P75=1.23–3.91). Overall, 50% of tumors showed a ≥2-fold gene expression relative to the calibrator. Higher median levels were detected in luminal A, followed by luminal B and HER2 and basal/triple-negative (P<0.000; Kruskal–Wallis test). The cutoff points with the highest sensibility and specificity for defining high and low IGF1R mRNA were set at 2.50 and 2.61, respectively, derived from receiver operating characteristic curves to predict recurrence or death. The receiver operating characteristic curve to predict disease-free survival had an area under the curve (95% CI) of 0.58 (0.47–0.68) (P=0.14). The receiver operating characteristic curve to predict death had an area under the curve of 0.63 (0.52–0.75) (P=0.043). A direct correlation was found between the HR-positive status (77%) and increased expression of α-IGF1R (63%; P<0.001), mRNA (47%; P=0.001) or active IGF1R (64%; P=0.07).

Table 3 Median values with percentiles (P25–P75) of α-IGF1R, p-IGF1R
Full size table
Figure 1
figure 1

Representative immunohistochemical stains for (a) α-IGF1R and (b) phosphorylated IGF1R (membrane±cytoplasm).

Full size image

Among the 85 tumors analyzed for IGF1R mutations in the β-subunit (TK domain at positions Tyr1131, Tyr1135 and Tyr1136; and C-terminal domain at positions Tyr1250 and Tyr1251), we only detected mutations in Tyr1131 (A3532G) in six primary tumors: three with local recurrence (one luminal B, one HER2 and one basal/triple-negative) and in three that did not recur (one luminal A, one HER2 and one basal/triple-negative).

Survival analyses showed shorter disease-free survival in patients with high-grade tumors (70 vs 90%; P=0.003) or vascular invasion present (70 vs 78%; P=0.056). Similarly, poorer overall survival was observed in patients with high-grade tumors (82 vs 100%; P=0.001) with vascular invasion (75 vs 91%; P=0.001), necrosis (74 vs 92%; P=0.001) or high content (cutoff ≥2.61) of IGF1R mRNA (90 vs 78%; P=0.048). However, no differences, neither for the IGF1R protein (total or active) nor for the mutation status, were found for disease-free and overall survival (P=NS) (Kaplan–Meier; log-rank test).

With regard to immunophenotypes, patients with luminal A tumors had a trend toward a better disease-free survival (81%) than did those with luminal B (69%), HER2-positive (68%) or basal/triple-negative tumors (71%) (P=0.06). Similarly, this was observed for overall survival, with 94% of patients with luminal A alive, 83% for luminal B, 78% for HER2-positive and 76% for those with basal/triple-negative tumors (P=0.05). Stratification of data according to IGF1R expression results, showed no significant correlation for disease-free survival neither with protein levels (total and active) nor with mRNA levels (P=NS). Nevertheless, patients whose tumors expressed ≥2.61-fold IGF1R mRNA had poorer overall survival among luminal A (100 vs 85%), luminal B (100 vs 72%) and basal/triple-negative groups (80 vs 50%), and similar for HER2 (73 vs 78%) (P=0.004). However, only a trend was found for active IGF1R results: luminal A (100 vs 88%), luminal B (100 vs 84%), basal/triple-negative groups (75 vs 50%) and HER2 (60 vs 82%) (P=0.12); and no significant associations were detected for α-IGF1R levels or mutation status (P=NS).

By multivariate analyses for disease-free survival, only histological grade (P=0.01) emerged as a significant predictor of relapse, with a trend for the presence of vascular invasion (P=0.12). With regard to overall survival, histological grade (P=0.03) and levels of IGF1R mRNA (P=0.01) showed an independent value, with a trend for vascular invasion (P=0.06) and subtype (P=0.15; Cox regression) (Table 4).

Table 4 Multivariate analysis (overall survival) of histological and biological factors (Cox model)
Full size table

Discussion

The current research is focused on the comprehensive analysis of IGF1R, a potential therapeutic target gene that may help to identify patients at different risk of recurrence among early breast carcinoma subtypes. In these series, luminal A was the most frequent subtype, associated with good prognostic factors and patients’ outcome. In contrast, patients with basal/triple-negative tumors developed distant metastasis and therefore, poor prognosis, more frequently compared with previous reports.9 Nguyen et al10 in a series of breast carcinoma patients with similar characteristics also showed a low proportion of local recurrence in luminal A subtype (0.8%), but increased distant metastases in luminal B and basal subtypes. It is noteworthy that the high proportion of our patients with local failure (17%) might be attributed to longer follow-up (median 124 months). However, results of the more recently published series might not be comparable as several factors might have influenced the outcome, such as better preoperative breast imaging or more effective adjuvant local or systemic treatments. Moreover, it is unknown whether the differences seen in the recurrence rate of the HER2 group would have been observed if those patients had received trastuzumab.

IGF1R expression was detected in a higher proportion of luminal A tumors, followed by luminal B and HER2-positive tumors, with the lowest rate in basal/triple-negative, in line with previous investigators who reported co-expression of IGF1R and ER signaling systems.19, 24, 25, 32 The presence of IGF1R in tumors with higher content of ER/PR support the development of a subset of breast carcinoma subtypes along these two distinct pathways, mainly luminal A and B. It is plausible that together with microenvironment interactions and/or additional gene alterations, it might contribute to tumor heterogeneity.3 IGF1R overexpression is not a requirement for cellular transformation, but its presence even at low levels, is required for the activation of the main substrate IRS1, which in turn activates the PI3K and Shc/Ras/ERKs pathways.29

Our results partially agree with a recent immunohistochemical study.27 Their analysis of 226 breast carcinomas (181 luminal, 31 triple-negative and 14 HER2-positive) showed activated IGF1R regardless of the subtypes (luminal in 48.1%, triple-negative in 41.9% and HER2-positive in 64.3%), no correlation with ER status or activated Akt, but associated with poor prognosis, whereas total IGF1R was not. However, the clinical characteristics and treatment of their patients were not specified, which is difficult for comparison with our data as we included only lymph-node-negative breast carcinomas with conservative local treatment. It is interesting that our further analysis according to systemic therapy groups showed no differences regarding the total recurrence rate.30

Despite the significant correlations between the IGF1R protein and gene expression, we found that only increased IGF1R mRNA levels predicted poor survival. More interesting, however, were the survival results after stratification according to tumor subtypes and IGF1R mRNA expression, which showed ≥2.5-fold change in the different poor prognostic groups.

An association of IGF1R with less aggressive tumors and HR-positive status has been reported previously.20, 21, 22, 25, 32, 33 Anti-estrogens are clinically useful in the treatment of ER-positive breast carcinomas, and they have major effects on the IGF signaling pathways. On the other hand, IGFs exert powerful mitogenic effects in tamoxifen-responsive cells and act to support estrogen/ER-promoted growth. Owing to this close cross-talk, tamoxifen reduces IGF1R signaling at various levels as part of its response mechanisms.34 Nonetheless, emerging data suggest the role of the IGF1R network in resistance to endocrine therapy, but with contradictory results. Chan et al35 demonstrated in the rat mammary glands that tamoxifen-induced growth inhibition was associated with decreased basal phosphorylation of IGF1R, IRS1 and p85 subunit of PI3K. A novel experimental study in breast carcinoma cell lines has shown that the growth-inhibitory effect of anti-estrogens is due to the additional functional interactions with the PI3K/Akt and ERK pathways, as well as Bcl2.28 In contrast, Brockdorff et al36 found an association between anti-estrogen resistance and reduced α-IGF1R expression. Gee et al24 reported in a small (n=64) series of breast carcinomas that the α-IGF1R expression was higher in ER/PR-positive tumors and that those patients showed tamoxifen response (increased time to progression). In contrast, only our subgroup of patients with high IGF1R mRNA content tumors exhibit shorter survival, suggesting the potential relevance of increased IGF1R signaling in tamoxifen-resistance disease. Nevertheless, IGF1R may have some relevance in a subgroup of aggressive basal/triple-negative tumors.2, 37, 38

It is noteworthy that in this study, phenotypes have been only stratified according to immunohistochemical and in situ hybridization (for HER2 tumors) results. In addition, besides the HR-positive status, Bcl2, p53 and Ki67 data were taken into consideration for classifying tumors into luminal A or luminal B, based on the important role of apoptosis and proliferation-related genes emphasized in previous studies,39, 40, 41 and also supported by our study. Our approach to classify all positive cases independently of the ER/PR status as HER2 subtype has been reinforced by recent reports.12, 42

IGF1R mutations, present only in Tyr1131 (A3532G), were associated with neither tumors that recurred nor immunophenotypes, suggesting an unrelated gene event with a specific phenotype. Previous mutational analysis studies have shown that signals derived from both regions of the IGF1R cooperate to enhance tumor metastasis.15 Moreover, a relationship between local recurrence and radioresistance due to mutations at the C-terminal domain43 could not be confirmed in our former study.30 So far, no other previous investigators have analyzed the importance of IGF1R mRNA levels or IGF1R mutation status on breast carcinoma subtypes for a comparison with our data.

Histological grade and vascular invasion were associated with recurrence, with a trend for the breast carcinoma subtype. Our patients with longer survival also had low-grade tumors, absence of vascular invasion and necrosis, luminal A phenotype and low content of IGF1R mRNA. Nevertheless, only histological grade and IGF1R mRNA revealed an independent prognostic value. In contrast to previous studies, we found no negative impact of local recurrence on outcome.4, 6 Interestingly, Smid et al11 have evidenced the preference of breast carcinoma to relapse to specific organs, sharing biological features and pathways with their distant metastatic site.

The gene-expression signature has been suggested to be more powerful to predict local recurrence after conservative treatment,44 metastasis45 and tamoxifen response40 than standard clinical-pathological criteria. However, this technology is still impractical for the routine because of the time and expense required. As shown by previous investigators and us, biomarker expression by immunohistochemistry such as ER/PR, HER2, including Bcl2, Ki67 and p53, is a valid and readily available approach to stratify breast carcinoma in subtypes, which in turn proportion useful clinical information.10 Moreover, our study suggests that analysis of IGF1R may add predictive information regarding hormone therapy response and prognostic value among patients with breast carcinoma subtypes with a HR-positive status. This is of great interest as inhibition of IGF1R signaling using antibodies and small molecules is an emerging therapeutic strategy,27 also in combination with conventional chemotherapy, radiation therapy or co-targeting multiple pathways.13, 29

Finally, results from the Nurses’ Health Study suggest that increased expression of IGF1R in the cytoplasm of the benign ductal epithelium is correlated with the risk of breast carcinoma development.46 If this was supported in further studies, new clinical trials would eventually determine whether IGF1R inhibitors have a role not only in the treatment of breast carcinoma but also in its prevention.

In conclusion, luminal A is the most frequent subtype of breast carcinoma in patients in early stage, in association with low risk factors. The involvement of IGF1R in luminal A and B tumors, as well as in a subgroup of HER2/HR-positive tumors might be related on the one hand to a positive ER/PR cross-talk. On the other hand, its association with the lack of responsiveness to tamoxifen is supported by a poor patient outcome.

Our study highlights the clinical relevance of breast carcinoma subtype classification obtained using routine immunohistochemical methods, and IGF1R as a biomarker that adds prognostic information, which serves as a potential therapeutic target.