Acknowledgment
This study was supported, in part, by the Knowledge Cluster Initiative of the Ministry of Education, Culture, Sports, Science and Technology.
A 95-gene classifier (95-GC) recently developed by us can predict the risk of relapse for ER-positive and node-negative breast cancer patients with high accuracy. This study investigated association of risk classification by 95-GC with response to neoadjuvant chemotherapy (NAC). Tumor biopsy samples obtained preoperatively from 72 patients with ER-positive breast cancer were classified by 95-GC into high-risk and low-risk for relapse. Pathological complete response (pCR) rate was numerically higher for high-risk (15.8%) than low-risk patients (8.8%) although the difference was not statistically significant. Pathological response evaluated in terms of the pathological partial response (pPR) rate (loss of tumor cells in more than two-thirds of the primary tumor) showed a significant association (P = 0.005) between the high-risk patients and a high pPR rate. Besides, external validation study using the public data base (GSE25066) showed that the pCR rate (16.4%) for high-risk patients (n = 128) was significantly (P = 0.003) higher than for low-risk patients (5.7%) (n = 159). These results demonstrate that the high-risk patients for relapse show a higher sensitivity to chemotherapy and thus are likely to benefit more from adjuvant chemotherapy.
One of the most urgent needs for breast cancer treatment practice is the development of a prognostic classifier for ER-positive and node-negative breast cancer patients which is more accurate than the conventional pathological prognostic factors. Recent advances in molecular technology has enabled the development of the classifiers based on multigene expression, such as Oncotype DX [1] and [2], MammaPrint [3] and [4], and genomic grade index (GGI) [5], which can provide valuable information on prognosis which is not obtainable with conventional pathological examination. Oncotype DX in particular is most widely used in practice for prediction of prognosis of ER-positive and node-negative patients. Very recently, however, we have also been able to develop a 95-gene classifier (95-GC) which can classify ER-positive and node-negative breast cancer patients into high-risk and low-risk with high accuracy [6].
Oncotype DX was originally developed as a predictor of prognosis for ER-positive and node-negative breast cancer patients, but interestingly, later studies have shown that the patients identified as high risk by Oncotype DX are more likely to benefit from adjuvant chemotherapy. It is reported that there was a significant improvement in disease-free survival in high-risk patients treated with adjuvant chemo-hormonal therapy than those with adjuvant hormonal therapy alone, but such a significant improvement was not observed in the low- and intermediate-risk patients [7]. A more direct association between Oncotype DX and sensitivity to chemotherapy has been investigated for breast cancer patients treated with neoadjuvant chemotherapy and it was reported that high-risk patients determined by Oncotype DX showed a significantly higher pathological complete response (pCR) to neoadjuvant chemotherapy than low-risk patients [8]. Similar results have also been reported for MammaPrint [9], [10] and [11] and GGI [12] and [13], namely, that high-risk breast cancer patients identified by these classifiers are more sensitive to neoadjuvant chemotherapy.
The reason for this association between high-risk patients determined by these classifiers and sensitivity to neoadjuvant chemotherapy is thought to be attributable, at least in part, to the fact that all these classifiers include expression of various genes related to cell proliferation, and significant association between high-risk patients and cell proliferation verified with the Ki67 marker has been reported [6]. Since the 95-GC classifier also includes several genes related to cell proliferation, it is speculated that high-risk patients identified by 95-GC are also more sensitive to chemotherapy. In order to investigate this possibility, we conducted the study reported here, in which the association between risk classification by 95-GC and response to chemotherapy was compared in the neoadjuvant setting.
Seventy-two patients with stage II and III primary breast cancer who were treated with neoadjuvant chemotherapy and subsequent surgery (mastectomy or breast conserving surgery) between 2003 and 2010 were retrospectively recruited for this study. Neoadjuvant chemotherapy consisted of paclitaxel 80 mg/m2weekly for 12 cycles followed by a combination of 5-FU [500 mg/m2], epirubicin [75 mg/m2], and cyclophosphamide [500 mg/m2] every 3 weeks for 4 cycles [P-FEC]. Before neoadjuvant chemotherapy, all the patients underwent tumor biopsy with a vacuum-assisted core-biopsy instrument (Mammotome 8G HH; Ethicon Endosurgery Inc., Cincinnati, OH) under ultrasonographic guidance for histological examination and gene expression analysis. Tumor samples for histological examination were fixed in 10% buffered formaldehyde, and tumor samples for gene expression analysis were snap frozen in liquid nitrogen and stored at −80 °C until use. Prior to the tumor biopsy, informed consent regarding the study was obtained from each patient.
As postoperative adjuvant therapy, tamoxifen (20 mg/day), goserelin (3.75 mg every 4 weeks) plus tamoxifen (20 mg/day), leuprorelin (3.75 mg every 4 weeks) plus tamoxifen (20 mg/day), anastrozole (1 mg/day), and letrozole (2.5 mg/day) were given to 12, 3, 7, 16 and 9 patients, respectively. Tamoxifen, anastrozole, or letrozole was administered typically for 5 years or until recurrence within 5 years, and goserelin or leuprorolin for 2 years or until recurrence within 2 years. Trastsuzumab (6 mg/kg every three weeks for one year) was given to 10 patients with HER2-positive tumors. Patient characteristics are listed inTable 1.
Abbreviations: PR, progesterone receptor; HER2, human epidermal growth factor receptor 2.
Determined in tumor biopsy samples before neoadjuvant chemotherapy.
Fisher’s exact test.
Trizol (Invitrogen, Carlsbad, CA) or Qiagen RNeasy mini kit (QIAGEN Sciences, Germantown, MD) was used to extract RNA from tumor biopsy samples obtained before neoadjuvant chemotherapy. The presence of tumor cells in the biopsy samples was estimated by histological confirmation of their presence in the adjacent tumor biopsy samples. RNA (50 ng) was subjected to gene expression analysis using a DNA microarray (Human Genome U133 Plus 2.0 Array; Affymetrix, Santa Clara, CA) according to a previously described method [14].
The pathological response to P-FEC was evaluated by using the surgical specimens obtained at surgery. The surgical specimens were cut into 5-mm slices, and hematoxylin- and eosin-stained sections (3-μm) were prepared to determine the presence or absence of tumor cells. A complete loss of invasive tumor cells and lymph node-negative status were defined as pCR irrespective of the presence or absence of ductal carcinoma in situ components. In addition, we defined pPR (pathological partial response) as loss of tumor cells in more than two-thirds of the primary tumor. pPR thus was equal to the histologically determined effect of grade IIa + IIb + III according to the General Rules for Clinical and Pathological Recording of Breast Cancer 2005 [15].
ER, PR, and Ki67 levels in tumor biopsy samples obtained before neoadjuvant chemotherapy were determined by immunohistochemistry according to a previously described method [14]. Cut-off values for ER, PR, and Ki67 were 10%, 10%, and 20%, respectively. HER2 amplification was determined by fluorescence in situ hybridization (FISH) using the PathVysion HER-2 DNA Probe Kit (Vysis/Abbott Molecular Inc., Chicago, IL). A tumor was determined as HER2-amplified if the FISH ratio was >2.0.
DNA microarray data were used for analysis with the 95-GC classifier of tumors from 72 ER-positive patients into high-risk and low-risk according to a previously described method [6]. The median follow-up was 3 years. In addition, we used public databases for the following external validation studies. The gene expression data of ER-positive tumors (n = 299) treated with neoadjuvant chemotherapy consisting of weekly paclitaxel × 12 cycles or triweekly docetaxel x 4 cycles followed by FAC (5-FU/doxorubicin/cyclophosphamide) × 4 cycles were extracted from the public data base GSE25066 [16], and the patients were classified into high-risk and low-risk with 95-GC. In addition, the gene expression data of ER-positive and node-positive tumors treated with adjuvant hormonal therapy alone were extracted from public data bases GSE2990 [5], GSE4922 [17], GSE6532 [18], and GSE9195 [19], followed by classification into high-risk and low-risk patients by 95-GC. Intrinsic subtyping of breast tumors was done as previously described [20]. Hierarchical clustering analysis combined with Spearman’s rank correlation coefficient and Ward’s method was performed for visualization by means of Partek Genomics Suite 6.5 (Partek Inc., St. Louis, MO).
Patients were also classified into the high-, intermediate-, and low-risk groups by recurrence score (≧31, 18–30, and <18, respectively) of Oncotype DX 21-gene classifier (21-GC), which was determined using “Recurrence Online” (http://www.recurrenceonline.com/) that was developed by Győrffy et al. [21]. “Recurrence Online” can compute 21-GC-based recurrence score using expression data obtained by Affymetrix microarray.
Distant relapse-free survival (DRFS) was calculated with the Kaplan–Meier method and evaluated with the log-rank test. Association of the 95-GC-classified high- or low-risk group with the various clinicopathological parameters was determined with the chi-square test or Fisher’s exact test. All statistical analyses were two-sided and P < 0.05 was judged to be significant.
Patients (n = 72) with ER-positive tumors were classified into high-risk (n = 38) and low-risk (n = 34) by means of 95-GC and the associations of their classification with the various clinicopathological parameters determined in tumor biopsy samples before neoadjuvant chemotherapy is shown in Table 1. Results of hierarchical cluster analysis of the 72 tumors are shown as a heatmap in Fig. 1. Low-risk patients were significantly more likely to have low-Ki67 tumors (P = 0.008) and low histological grade tumors (P = 0.02). All these tumors were also classified into intrinsic subtypes, and their relationship with the 95-GC-classified risk groups is shown in Fig. 2. Low-risk patients were significantly more likely to have luminal A tumors and less likely to have luminal B tumors (P = 0.002) than high-risk patients. There was a significant (P < 0.0001) association between risk classification determined by 95-GC and 21-GC (Oncotype DX).
Response to neoadjuvant chemotherapy was evaluated pathologically and compared with the risk groups classified by 95-GC (Fig. 3). pCR rate was numerically higher for high-risk (15.8%) than low-risk patients (8.8%) although the difference was not statistically significant. Pathological response evaluated in terms of the pPR rate (loss of tumor cells in more than two-thirds of the primary tumor) showed a significant association (P = 0.005) between the high-risk patients and a high pPR rate.
In addition, we used the public data base GSE25066 to investigate the relationship between the 95-GC-classified risk groups and pCR rates. The gene expression data of ER-positive tumors (n = 299) extracted from this data base were subjected to risk classification by 95-GC, resulting in the identification of high-risk (n = 132) and low-risk tumors (n = 167). Of these 299 tumors, information on pathological response was obtainable for 287. The pCR rate (16.4%) for high-risk patients (n = 128) was significantly (P = 0.003) higher than for low-risk patients (5.7%) (n = 159).
First, we tried to validate the 21-GC-based risk classification determined by “Recurrence Online” using our previously reported cohort of 105 ER-positive and node-negative patients treated with adjuvant hormonal therapy alone[6]. Patients were classified into the high-risk (n = 40), intermediate-risk (n = 11), and low-risk patients (n = 54) (Fig. 4a). Distant relapse-free survival was significantly (P = 0.0003) better in low-risk patients than high-risk patients (Fig. 4a).
Then, response to neoadjuvant chemotherapy was compared between the risk groups. Seventy-two patients were classified into the high-risk (n = 41), intermediate-risk (n = 8), and low-risk patients (n = 23). Since the number of intermediate-risk patients was so small (n = 8) that intermediate-risk patients and low-risk patients were combined in the following analysis. pCR rate was numerically higher for high-risk patients (19.5%) than intermediate- plus low-risk patients (3.2%) although the difference was not statistically significant. pPR rate was significantly higher (P = 0.0004) for high-risk patients (70.7%) than intermediate- plus low-risk patients (29.0%).
95-GC was originally constructed to predict the prognosis of ER-positive and node-negative breast cancer patients treated with adjuvant hormonal therapy alone. We herein applied this classifier to the ER-positive and node-positive breast cancer patients (n = 252) who were treated with adjuvant hormonal therapy alone and included in the public databases GSE2990, GSE4922, GSE6532, GSE9195. As shown in Fig. 5a, there was a significant (P < 0.0001) difference in distant relapse-free survival (DRFS) between the high-risk and low-risk patients.
DRFS of the high-risk and low-risk patients among the ER-positive patients treated with neoadjuvant chemotherapy, i.e., the patients in the present study (n = 72, Fig. 5b) and those in the MD Anderson Cancer Center series (GSE25066) (n = 299; Fig. 5c), was compared. There was no significant difference in DRFS between the high-risk and low-risk patients in either cohort.
95-GC was first constructed based on the gene expression data of a large number (n = 549) of ER-positive and node-negative breast cancer patients obtained from public data bases and then validated for 105 ER-positive and node-negative patients at our institute. At that time, it had not been determined yet whether 95-GC is useful for the classification of ER-positive and node-positive breast cancer patients into high-risk and the low-risk patients. In the study presented here, we applied 95-GC to the data for the ER-positive and node-positive breast cancer patients (n = 252) listed in the public data base GSE2990, GSE4922, GSE6532, and GSE9195 and were able to demonstrate that the high-risk patients showed a significantly poorer prognosis than the low-risk patients. This observation seems to be compatible with the report that Oncotype DX, which was initially constructed based on data for ER-positive and node-negative breast cancer patients [1], can also predict prognosis of ER-positive and node-positive breast cancer patients [22].
Interestingly, Chang et al. reported that patients with a high recurrence score for Oncotype DX showed a significantly (P = 0.008) higher pCR rate for neoadjuvant chemotherapy (docetaxel) than those with a low recurrence score [8], suggesting that the recurrence score can also be useful for the prediction of response to chemotherapy. In the presents study, we have also studied the association of Oncotype DX (21-GC)-based risk and response to neoadjuvant chemotherapy, and have been able to show the consistent results that 21-based high-risk is significantly associated with a better response. Similarly, results for MammaPrint also reportedly showed a significantly higher pCR rate (P = 0.015) and near pCR (residual invasive component <2 mm) rate (P = 0.008) for anthracycline and/or taxane for patients with a poor prognosis MammaPrint signature than those with a good prognosis signature [11]. In our study we have not been able to show a significant difference in the rate of pCR to P-FEC between high-risk (15.8%) and low-risk patients (8.8%) but, when a more generous criterion for pPR was adopted, a significantly higher pPR rate could be demonstrated for the high-risk (68.4%) than the low-risk patients (35.3%). Since the pCR rate is generally low for ER-positive tumors, using pPR instead of pCR appears to be a reasonable option for evaluating chemo-sensitivity. In addition, we conducted an external validation study using the data base (GSE25066) deposited from the MD Anderson Cancer Center, and were able to demonstrate that high-risk patients show a significantly higher pCR rate (16.4%) to neoadjuvant chemotherapy than do low-risk patients (5.7%). Interestingly, pCR rates for high-risk patients in our and the MD Anderson Cancer Center series were very similar, as were pCR rates for low-risk patients, probably because essentially the same neoadjuvant chemotherapeutic regimens were used for both series. The large number of patients enrolled in the MD Anderson Cancer Center series seems to have made a statistically significant difference in the pCR rates. Putting these results together suggests that high-risk patients are more likely to respond to chemotherapy using anthracycline and/or taxane than low-risk patients.
The fact that the addition of adjuvant CAF (cyclophosphamide/doxorubicin/5-FU) to tamoxifen resulted in a significant improvement in DRFS for patients with a high, but not a low or intermediate, recurrence score determined by Oncotype DX [7] seems to be compatible with the findings that breast tumors with a high Oncotype DX recurrence score are more sensitive to neoadjuvant chemotherapy [8]. As mentioned earlier, we were able to demonstrate that there is a significant difference in DRFS between 95-GC-based high-risk and low-risk patients among ER-positive and node-positive patients (Fig. 5a). However, there was no significant difference in DRFS between high-risk and the low-risk patients when they were treated with neoadjuvant chemotherapy and adjuvant hormonal therapy (Fig. 5b and c). These findings suggest that the high-risk patients classified by 95-GC, like those by Oncotype DX or MammaPrint, are at high risk for relapse but they are more likely to benefit from adjuvant chemotherapy.
In the present study, we compared the 95-GC based risk groups with intrinsic subtypes (Fig. 2), and found that low-risk patients were significantly more likely to have luminal A tumors and less likely to have luminal B tumors (P = 0.002) than high-risk patients. 95-GC includes a significant number of the genes related to cell proliferation, and low-risk tumors by 95-GC have a lower Ki67 index than high-risk tumors (Table 1). Similarly, luminal A tumors are known to have a lower proliferation activity than luminal B tumors[23], and recently we have confirmed that in our series[24]. Thus, the reason why low-risk patients were significantly more likely to have luminal A tumors and less likely to have luminal B tumors (P = 0.002) than high-risk patients seems to be explained, at least in part, by the fact that low-risk tumors, like luminal A tumors, have a low proliferation activity and that high-risk tumors, like luminal B tumors, have a high proliferation activity.
In conclusion, we have shown that risk classification by 95-GC is applicable to ER-positive and node-positive breast cancer patients for classification into those with high risk and low risk for relapse. The patients classified as high-risk by 95-GC are considered to be more sensitive to chemotherapy and thus are more likely to benefit from adjuvant chemotherapy. However, our findings need to be verified in a future study including a larger number of patients.
This study was supported, in part, by the Knowledge Cluster Initiative of the Ministry of Education, Culture, Sports, Science and Technology.
A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer
N. Engl. J. Med., 351 (2004), pp. 2817–2826
A population-based study of tumor gene expression and risk of breast cancer death among lymph node-negative patients
Breast Cancer Res., 8 (2006), p. R25
Gene expression profiling predicts clinical outcome of breast cancer
Nature, 415 (2002), pp. 530–536
A gene-expression signature as a predictor of survival in breast cancer
N. Engl. J. Med., 347 (2002), pp. 1999–2009
Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis
J. Natl Cancer Inst., 98 (2006), pp. 262–272
Development of 95-gene classifier as a powerful predictor of recurrences in node-negative and ER-positive breast cancer patients
Breast Cancer Res. Treat., 128 (2011), pp. 633–641
Prognostic and predictive value of the 21-gene recurrence score assay in postmenopausal women with node-positive, oestrogen-receptor-positive breast cancer on chemotherapy: a retrospective analysis of a randomised trial
Lancet. Oncol., 11 (2010), pp. 55–65
Gene expression patterns in formalin-fixed, paraffin-embedded core biopsies predict docetaxel chemosensitivity in breast cancer patients
Breast Cancer Res. Treat., 108 (2008), pp. 233–240
The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1–3 positive lymph nodes in an independent validation study
Breast Cancer Res. Treat., 116 (2009), pp. 295–302
The predictive value of the 70-gene signature for adjuvant chemotherapy in early breast cancer
Breast Cancer Res. Treat., 120 (2010), pp. 655–661
The 70-gene signature as a response predictor for neoadjuvant chemotherapy in breast cancer
Breast Cancer Res. Treat., 119 (2010), pp. 551–558
Genomic grade index: an important tool for assessing breast cancer tumor grade and prognosis
Crit. Rev. Oncol. Hematol., 77 (2011), pp. 20–29
Genomic grade index is associated with response to chemotherapy in patients with breast cancer
J. Clin. Oncol., 27 (2009), pp. 3185–3191
Prediction of pathologic complete response to sequential paclitaxel and 5-fluorouracil/epirubicin/cyclophosphamide therapy using a 70-gene classifier for breast cancers
Cancer, 117 (2011), pp. 3682–3690
General rules for clinical and pathological recording of breast cancer 2005
Breast Cancer, 12 (Suppl) (2005), pp. S1–27
A genomic predictor of response and survival following taxane-anthracycline chemotherapy for invasive breast cancer
JAMA., 305 (2011), pp. 1873–1881
Genetic reclassification of histologic grade delineates new clinical subtypes of breast cancer
Cancer Res., 66 (2006), pp. 10292–10301
Definition of clinically distinct molecular subtypes in estrogen receptor-positive breast carcinomas through genomic grade
J. Clin. Oncol., 25 (2007), pp. 1239–1246
Predicting prognosis using molecular profiling in estrogen receptor-positive breast cancer treated with tamoxifen
BMC Genomics, 9 (2008), p. 239
Gene expression signature of TP53 but not its mutation status predicts response to sequential paclitaxel and 5-FU/epirubicin/cyclophosphamide in human breast cancer
Cancer Lett., 307 (2011), pp. 149–157
RecurrenceOnline: an online analysis tool to determine breast cancer recurrence and hormone receptor status using microarray data
Breast Cancer Res. Treat., 132 (2012), pp. 1025–1034
Prediction of risk of distant recurrence using the 21-gene recurrence score in node-negative and node-positive postmenopausal patients with breast cancer treated with anastrozole or tamoxifen: a TransATAC study
J. Clin. Oncol., 28 (2010), pp. 1829–1834
Randomized phase II neoadjuvant comparison between letrozole, anastrozole, and exemestane for postmenopausal women with estrogen receptor-rich stage 2 to 3 breast cancer: clinical and biomarker outcomes and predictive value of the baseline PAM50-based intrinsic subtype–ACOSOG Z1031
J. Clin. Oncol., 29 (2011), pp. 2342–2349
GSTP1 expression predicts poor pathological complete response to neoadjuvant chemotherapy in ER-negative breast cancer
Cancer Sci., 103 (2012), pp. 913–920