Relationship Between Topoisomerase 2A RNA Expression and Recurrence after Adjuvant Chemotherapy for Breast Cancer

Abstract

Purpose

To perform an exploratory analysis of the relationship between gene expression and recurrence in operable hormone receptor (HR)-positive, HER2-normal breast cancer patients treated with adjuvant doxorubicin-containing chemotherapy.

Experimental Design

RNA was extracted from archived tumor samples derived from 378 patients with stage I–III HR-positive, HER2-normal breast cancer and analyzed by RT-PCR for a panel of 374 genes, including the 21 gene Recurrence Score (RS). Patients were randomized to receive adjuvant doxorubicin plus cyclophosphamide or docetaxel in trial E2197, with no difference in recurrence seen in the treatment arms. All available recurrent cases were selected plus a non-recurrent cohort. Cox proportional hazard models were used to identify relationships between gene expression and recurrence.

Results

TOP2A expression exhibited the strongest association with increased recurrence risk (p=0.01), and was significantly associated with recurrence (p=0.008) in a multivariate analysis adjusted for clinicopathological features. Elevated TOP2A expression above the median was associated with a 2.6-fold increase (95% confidence intervals [CI], 1.3, 5.2 p=0.008) in risk of recurrence if the RS was less than 18, and a 2.0-fold increase (95% CI, 1.2, 3.2, p=0.003) if there was an intermediate RS of 18–30.

Conclusions

In patients with HR-positive, HER2-normal breast cancer, a population known to have a low incidence of TOP2A gene alterations thought to be predictive of anthracycline benefit, there is a range of TOP2A RNA expression that is strongly associated with recurrence after adjuvant anthracyclines which provides information complementary to RS, indicating that it merits further evaluation as a prognostic and predictive marker.

INTRODUCTION

Adjuvant cytotoxic chemotherapy, including anthracyclines and taxanes, reduces the relative risk of recurrence in women with operable breast cancer.^{1, 2} Anthracycline therapy has been associated with more acute toxicity, and serious delayed toxicities such as congestive heart failure and leukemia.³ Other adjuvant systemic therapies that reduce recurrence include endocrine therapy for hormone receptor (HR)-positive disease⁴, and trastuzumab for disease that overexpresses the human epidermal growth factor receptor 2 (HER2) protein.⁵ A metaanalysis including nine randomized trials indicates that only patients whose tumors overexpress HER2 protein benefit from anthracycline therapy⁶, suggesting that the 75–80% of patients with HER2-normal disease may be subjected to the risks of anthracyclines without deriving incremental benefit. It is believed that the preferential sensitivity of HER2-overexpressing disease may be due to co-amplification of the topoisomerase 2 gene (TOP2A), which resides on the same amplicon as HER2, and serves as the therapeutic target for anthracyclines.⁷ Several studies have indicated that TOP2A amplifications and/or deletions are associated with sensitivity to anthracycline therapy; although they occur in approximately 40% of HER2-overexpressing tumors⁸, they are present in less than 10% of tumors that are HER2-normal and hormone receptor (HR)-positive.⁹ Moreover, a recent trial demonstrated that the combination of docetaxel and cyclophosphamide (TC) reduced the risk of recurrence and death compared with the standard combination of doxorubicin and cyclophosphamide (AC)¹⁰, whereas another trial in a similar population showed no benefit for the doxorubicin-docetaxel combination compared with AC^{11, 12}, raising additional questions regarding the effectiveness of anthracyclines in HR-positive, HER2-normal disease.

In the current study that is the subject of this report, we evaluated the relationship between recurrence and a panel of 374 genes, including TOP2A and 21 genes included in the Oncotype DX^® Recurrence Score (Genomic Health, Inc., Redwood City, CA). ¹³ The study population included 378 patients with HR-positive, HER2-normal breast cancer and 0–3 positive axillary nodes who were treated with doxorubicin plus either cyclophosphamide or docetaxel in addition to standard hormonal therapy.¹² We restricted our analysis to only patients with HR-positive, HER2 normal breast cancer because this is precisely that population that is believed not to derive benefit from adjuvant anthracycline therapy. After finding that TOP2A RNA expression exhibited the strongest association with recurrence in our initial exploratory analysis, we sought to further evaluate the relationship between TOP2A expression and classical clinicopathologic features. We also evaluated whether it provided information that is complementary to the 21 gene assay, which does not include TOP2A. We chose to focus on TOP2A not only because of the strong association with recurrence demonstrated in our exploratory analysis, but also because multiple previous studies have suggested a relationship between response to adjuvant doxorubicin-containing therapy and TOP2A protein expression or gene amplifications and/or deletions.⁹ Moreover, in contrast to previous reports that evaluated protein expression or gene alterations, this report describes use of quantitative measurement of TOP2A RNA by qRT-PCR (quantitative reverse transcriptase polymerase chain reaction), which has been shown to be a reliable, reproducible, and quantitative method for evaluating gene expression.¹⁴ We report herein not only a strong relationship between TOP2A RNA expression and recurrence in HR-positive, HER2-normal breast cancer treated with anthracycline-containing therapy, but also evidence that TOP2A RNA expression provides information that is complementary to both classical clinicopathologic features and the 21 gene assay. Moreover, our study provides validation of a recent independent report in another robust data set demonstrating a significant relationship between TOP2A RNA expression evaluated by microarray and recurrence in HR-positive breast cancer. ¹⁵

METHODS

Study Population and Treatment

The study utilized tumor specimens and clinical information from patients enrolled on trial E2197 (ClinicalTrials.gov identifier NCT00003519), coordinated by the Eastern Cooperative Oncology Group (ECOG), details of which have been reported elsewhere.¹² Patients were randomly assigned to receive four 3-week cycles of doxorubicin 60 mg/m2 and cyclophosphamide 600 mg/m2 (AC) or docetaxel 60 mg/m2 (AT), plus endocrine therapy for 5 years or longer if hormone receptor (HR) positive. Tamoxifen (20 mg daily for 5 years) was recommended for HR-positive disease beginning after completion of chemotherapy when the trial was initiated, although about 40% eventually took an aromatase inhibitor at some point before or after 5 years when it was shown that these agents were more effective than tamoxifen. After a median follow-up of 76 months, there was no significant difference between the AC and AT arms in disease-free survival (the primary study endpoint; defined as recurrence or contralateral breast cancer), relapse free interval (the endpoint used in this analysis; defined as recurrence), or overall survival in the entire study group, and in the population included in this analysis. All patients included in this analysis provided written informed consent for participation in the trial and utilization of their tissue specimens that was approved by the local institutional review board.

Because of the low overall recurrence rate in E2197, a stratified sampling design was used, with recurrences sampled more heavily than non-recurrences. Patients were sampled separately within groups defined by recurrence status, hormone receptor (HR) status (as determined by local laboratories), axillary nodal status (positive vs. negative) and treatment arm (AC vs. AT), giving 8 sampling strata with separate sampling from recurrences and non-recurrences in each stratum. E2197 included 2952 patients with operable breast cancer, of whom 1579 patients were potentially eligible for inclusion in this analysis, with reasons for exclusion summarized in the CONSORT diagram shown supplemental Figure 1. Of the 1579 potential patients eligible for the sampling and analysis, 191 (12%) had a recurrence and 1388 (88%) did not have a recurrence. All patients with recurrence were included in the sampling (defined as “case” sample), plus a randomly selected sample of 641 patients without recurrence (defined as the “control” sample, based upon the planned ratio of 1:3.5 for the case-control sampling), yielding a total of 832 patients for the analysis. Although the sampling stratification was based upon HR expression determined in local laboratories, the final classification of HR status in this analysis is based upon central HR expression testing. Of the 832 patients identified, samples from each patient were sent from the ECOG Pathology Coordinating Office (PCO) to Genomic Health. The case and control sample was selected by the coordinating statistician (RG), and all specimens were processed by the ECOG PCO and Genomic Health without knowledge of the recurrence status and clinicopathologic variables. Within the set of 383 subjects with centrally confirmed HR-positive, HER2-normal disease analyzed for the full set of 374 genes, 324 genes have complete values and 47 are missing values for a small number of cases (median = 1, range = 1 to 5), including 5 cases that were missing values for TOP2A, yielding 378 patients included in this analysis. The missing values for an individual gene were excluded from the computation of the statistic for that gene. There is thus some small variation in the case set for the individual gene statistics.

Regulatory Approvals and Manuscript Development Process

The E2197 protocol was approved by the institutional review boards of all participating institutions and was carried out in accordance with the Declaration of Helsinki, current Food and Drug Administration Good Clinical Practices, and local ethical and legal requirements. ECOG designed and coordinated the study and was responsiblefor all aspects of the data collection and analysis. Other members of the North American Breast Cancer Intergroup participated and contributed patients to the study, including the Southwest Oncology Group (SWOG), Cancer and Acute Leukemia Group B (CALGB), and the North Central Cancer Treatment Group (NCCTG). Only patients who gave consent for future research of their tumor specimen were included in the analysis. The use of specimens for this project was approved the North American Intergroup Correlative Science Committee and by the Northwestern University Institutional Review Board (which oversees the ECOG Pathology Coordinating Office, where the specimens were banked and evaluated).

Specimen Processing and Genomic Analysis

All specimens underwent analysis for tumor grade, and for estrogen, progesterone, and HER2 expression in a central lab as previously described.¹⁶ Briefly, all formalin-fixed, paraffin embedded tissue (FPET) specimens were processed by the ECOG Pathology Coordinating Office (PCO) and Reference Laboratory at the Robert H. Lurie Comprehensive Cancer Center of Northwestern University (Chicago, IL). One section at 5 micron was cut, followed by three 10 micron sections using sterile conditions with fresh blade and water per tissue block. The 5 micron slide section was stained with hematoxylin and eosin (H&E), and reviewed by Genomic Health pathology using the Oncotype DX breast cancer assay criteria for acceptance and macrodissection.¹³ Tumor specimens were evaluated for histologic grade using the modified Bloom-Richardson score by a single pathologist (FLB) using the H&E stained tissue sections, and by two pathologists (FLB, SB) for estrogen, progresterone, and HER2 expression. RNA was extracted from FPET specimens by cutting three 10 micron thick sections without dissection, or from macro-dissected tumor from six 10 micron sections. Cut sections were placed into 1.5 ml RNase/Dnase free microfuge tubes (PGC Scientific, Frederick, MD), labeled with a bar-code (blinding the sample), and shipped at ambient temperature to the reference laboratory at Genomic Health, Inc.

Cases with no cancer (depleted by prior tissue studies) or with cancer cells occupying <5% of the section area were excluded from the study. Macro-dissection to obtain enriched tumor tissue was performed using a safety blade cleaned with RNaseZAP (Ambion, Austin, TX) on sections having non-tumor elements (such as smooth muscle, fibrosis, hemorrhage, normal breast stromal tissue, but not DCIS or LCIS or necrosis) that were both sufficiently localized to be amenable to macro-dissection and constituted >50% of the overall tissue area of the section.

A total of 20 tissue microarray (TMA) blocks were constructed by the ECOG PCO using two 1.0 mm cores per each patient tumor. Thin sections (4um) from TMAs were immunostained by the ECOG PCO using DakoCytomation EnVision+ System (Dako, Carpinteria, CA) in a two-step technique. After deparaffinization, sections were rehydrated, and endogenous peroxidase was blocked with 1% H202 in methanol. Following heat-induced epitope retrieval (BioCare pressure cooker; citrate buffer, pH 6.0), sections were incubated with anti-ER antibody (clone 1D5, dilution 1:100; Dako, Carpinteria, CA) or anti-PR antibody (clone 636, dilution 1:200; Dako, Carpinteria, CA) at room temperature. The reaction was visualized using Envision + kit (Dako, Carpinteria, CA) and 3,3′-diaminobenzidine as chromogen followed by light counterstaining with hematoxylin. Positive and negative controls were used in each staining run. ER and PR expression was determined on TMA sections by two pathologists (FLG, SB) simultaneously designating a proportion score (PS, range = 0 – 5), intensity score (IS, range = 0 – 3) and Allred score (AS = PS + IS, range = 0 – 8) for each case; an AS of greater than 2 was defined as positive as previously described. ¹⁷ HER2 expression was defined if there was intense membrane staining in at least 30% of cells using the DAKO Herceptest consistent with ASCO-CAP guidelines.¹⁸ Only patients with HER2-normal disease (negative or equivocal by ASCO-CAP guidelines). were include this analysis. Seven cases in the total sample were not evaluable on central IHC; ER and PR status for these cases was determined using local results and HER2 status using genomic results.

Quantitative RNA expression levels were measured by real time reverse transcriptase polymerase chain reaction (RT-PCR) using gene-specific primers.¹⁴ All specimens were also analyzed for the 21 genes in the Oncotype DX Recurrence Score (in triplicate) as previously described¹³, plus a panel of 353 other genes (run in single wells), including TOP2A. Gene amplification/deletions analysis was not conducted being already demonstrated from other studies that the incidence is very low in this phenotype (< 5–10%). The genes selected were assembled by searching the published literature, genomic databases, pathway analysis, and microarray-based gene expression profiling experiments performed in fresh frozen tissue to identify genes likely to be associated with prognosis or response to chemotherapy (see supplemental Table 1, includes the Human Gene Organization [HUGO] name)

Case and control selection, end points, and statistical analyses

Because there was no difference in the recurrence rate or overall survival between the two treatment arms, the analysis used samples from both arms. The primary endpoint for this analysis was recurrence-free interval (RFI), defined as the time from trial entry to the first evidence of breast cancer recurrence (which included invasive breast cancer in local, regional or distant sites, including the ipsilateral breast, but excluded new primary breast cancers in the opposite breast). ¹⁹

Unlike the standard case-cohort design, only a subset of the recurrences from the E2197 study are included in the sample here. To estimate the magnitude of effects in the full E2197 population, sampling weights for each of the 16 groups in the sample are defined by the number of patients in the E2197 study in that group divided by the number in the sample. In the weighted analyses, contributions to estimators and other quantities, such as partial likelihoods, are multiplied by these weights. If the patients included in the case-control sample are randomly selected from the recurrences and non-recurrences within each stratum, then the weighted estimators give consistent estimates of the corresponding quantities from the full E2197 sample. Since availability and analyzability of tissue samples was a factor in the selection, the possibility of systematic bias in the selection cannot be completely ruled out, but the comparisons in Table 1 of Goldstein et al ¹¹ using this same dataset suggests that for many purposes, weighted analysis of the case-control sample should be representative of the full E2197 study. The weighted partial likelihood computed in this fashion is used for estimating hazard ratios and testing effects. The variance of the partial likelihood estimators is estimated using the general approach of Lin²⁰, which leads to a generalization of the variance estimator from Borgan et. al²¹ to allow subsampling of cases. Weighted averages, with proportions estimated using weighted averages of indicator variables, are also used for estimating the distribution of factors and for comparing the distributions between the overall E2197 study population and the genomic sample. Tests comparing factor distributions are based on asymptotic normality of the difference in weighted averages. The adjusted variance based on Lin was used in the score statistics²⁰, and additional details of the weighted analysis methods are described by Gray.²²

Weighted Cox proportional hazards model score tests were used to rank genes by their individual significance for predicting recurrence risk by one of the collaborating statisticians (RG). The significance level was determined from Cox model score statistics using the available data for each gene as a single linear covariate without adjustment for other factors. Adjusted p-values controlling the false discovery proportion (FDP) at ≤ 10% were computed using algorithm B* in Korn et al.²³, using 500 permutations. The p-values were applied in a ‘step-down’ fashion, which was the only approach guaranteed to give the stated level of control in the simulations of Korn et al.²⁴ The adjusted p-values give the level of confidence that the FDP is less than or equal to 10%, in the sense that the p-value is the proportion of experiments where the true FDP is expected to exceed the stated rate. Genes were ranked by significance levels, or equivalently by values of the test statistics. The test statistics give a ranking of strength of association that is invariant to differences in scale for different genes and they directly relate to the strength of evidence of whether the genes affect the hazard rate for recurrence, although this criterion is somewhat arbitrary and not the only possible scale invariant measure of association Weighted analysis of proportional hazards models was used to estimate hazard ratios and test for effects in joint models, and weighted Kaplan-Meier estimators of RFI distributions were used to estimate recurrence rates; the weighting algorithm corrects for any potential bias in sampling of the non-relapse cohort population. All p values are two-sided.

RESULTS

Characteristics of Study Population

There were 378 patients who had centrally confirmed HR-positive, HER2-normal disease and had genomic information available for TOP2A expression, of whom 77 had recurrences; their characteristics are shown in Table 1. The weighted distribution of Recurrence Scores was 51.0% in the RS<18 group, 32.4% in the RS 18–30 group and 16.6% in the RS≥31 group. The median followup for the study population was 6.3 years, and the five-year relapse free interval and overall survival rates were 90.4% and 93%, respectively.

Table 1.

Patient characeristics

Characteristics	Result
AT Arm	50.8 % (1.2)
AC Arm	49.2% (1.2)
Age ≤ 45	23.2 % (2.2)
Age 46 – 65	63.9 % (2.6)
Age > 65	12.9 % (1.8)
Premeno	40.8 % (2.6)
Postmeno	59.2% (2.6)
Tumor ≤ 2.0 cm	53.3% (2.7)
Tumor 2.1 – 5.0 cm	42.9% (2.7)
Tumor > 5.0 cm	3.9% (1.1)
Node neg	53.4% (1.2)
1 pos Node	25.1% (1.8)
2 pos Nodes	14.9% (1.6)
3 pos Nodes	6.6% (1.2)
Low Grade	25.4% (2.4)
Intermed Grade	48.8% (2.7)
High Grade	25.8% (2.3)
5-year RFI	90.4% (0.8)
5-year OS	93.0% (1.2)
Median follow-up	6.3 years

Genes Associated with Increased Recurrence

Continuous RS, without considering other variables, was a highly significant predictor of recurrence (p=0.004 overall). When genes were considered individually, without adjusting for effects of other factors, there were 10 genes significantly (adjusted p-values < 0.05, controlling the false positive rate at 10%) associated with increased risk of recurrence (Table 2), including TOP2A, DEPDC1, NUSAP1, AURKB, KIFC1, GAPDH, BUB1B, BIRC5, TYMS, and PLK1. TOP2A exhibited the strongest association with increased risk of recurrence (unadjusted P value = 0.000006, Korn’s adjusted P value =0.01). There were 40 genes in the 374 gene set with correlations with TOP2A that were > 0.5 (see supplemental Table 2). The highest correlations were for proliferation related genes, including MKI67, BIRC5, CCNB1, MYBL2, and AURKA, which comprise all five of the genes included in the proliferation group of the Oncotype DX RS. The correlation coefficient between TOP2A and the proliferation group was 0.84, suggesting that some of the relationship between TOP2A and increased recurrence risk may be due to the association with proliferation. Furthermore, of the 9 genes other than TOP2A for which increased expression correlated with an increased recurrence risk, expression of all but one gene (GAPDH) correlated with TOP2A expression, with 6 of the remaining 8 genes being highly correlated (r≥ 0.70). This high degree of overlap between genes associated with recurrence and genes whose expression correlated with TOP2A expression provides further support for our approach of focusing on TOP2A expression in our analysis.

Table 2.

Genes for which increased expression was associated with increased recurrence risk

HUGO Gene Symbol	Rank	Gene Name	EntrezGene ID	Chromosome	Nominal P Value	Kon’rs Adjusted P Value	Hazard Ratio	Correlation with TOP2A
TOP2A	1	Topoisomerase II alpha	7153	17q21-q22	5.67E-06	0.01	1.55	—
DEPDC1	2	DEP domain containing 1	55635	1p31.2	8.39E-06	0.01	1.67	0.7
NUSAP1	3	nucleolar and spindle associated protein 1	51203	15q15.1	9.58E-06	0.01	1.64	0.56
AURKB	4	aurora kinase B	9212	17p13.1	1.21E-05	0.01	1.59	0.74
KIFC1	5	kinesin family member C1	3833	6p21.3	6.75E-05	0.0299	1.59	0.71
GAPDH	6	Glyceraldehyde 3 phosphate dehydrogenase	2597	12p13	8.07E-05	0.0339	2.55	<0.5
BUB1B	7	BUB1 budding uninhibited by benzimidazoles 1 homolog beta	701	15q15.1	0.00025	0.0399	6.69	0.65
BIRC5	8	baculoviral IAP repeat-containing 5 (survivin)	332	17q25	0.0003	0.0339	1.4	0.8
TYMS	9	Thymidylate synthase	7298	18p11.32	0.00041	0.0439	1.59	0.71
PLK1	10	polo-like kinase 1	5347	16p12.1	0.00042	0.0439	1.45	0.74

Relationship Between TOP2A Expression as a Continuous Variable and Recurrence

The relationship between TOP2A expression and recurrence was evaluated as a continuous variable using a flexible family of curves (a natural spline with 3 degrees of freedom) to model the effect of TOP2A in a proportional hazards model for recurrence. The estimated curve, with 95%CI, shows substantial increasing trend over most of the range between TOP2A expression and increasing recurrence risk (Figure 1).

Figure 1.

Log hazard ratio for recurrence risk as a function of TOP2A expression. No adjustment for other factors. Solid lines are at ± 2 standard errors. (P = 0.0003 overall, P = 0.28 for nonlinearity.)

Relationship Between Clinical and Genomic Features and TOP2A Expression

The relationship between clinical and genomic features, recurrence, and TOP2A expression was also evaluated as a categorical variable. The median split (value=6.23) for TOP2A was used since it appeared that the slope of the estimated recurrence log hazard ratio function was near its maximum at the median value (Figure 1). The range of TOP2A values in the low expression group was 2.90 to 6.23 and in the high expression group the range was 6.23 to 10.41, with 189 cases in each group. Clinical variables examined included age, nodal status, centrally determined tumor grade, and tumor size (Table 3). TOP2A expression was significantly more likely to be high in tumors associated with poor grade (p<0.0001) and high RS (p<0.0001). The relationship between TOP2A expression and centrally determined estrogen receptor (ER) and progesterone receptor (PR) expression quantitated by the Allred score was also evaluated. Tumors with a very low ER and PR Allred Scores of 0–2 were significantly more likely to have high TOP2A expression (p=0.004), whereas tumors with a high PR score of 8 (p=0.003) were more likely to exhibit low TOP2A expression (full data set not shown).

Table 3.

Distribution of clinical and genomic features in low and high TOP2A expression groups

	Low TOP2A	High TOP2A	P value
	(n=189)	(n=189)
Age
≤ 45 years	23%	23%	0.94
45–65 years	62%	67%	0.34
> 65 years	15%	10%	0.2
Nodal Status
0 Positive	55%	51%	0.48
1 Positive	26%	24%	0.58
2–3 Positive	19%	25%	0.14
Grade
Well differentiated	39%	10%	<0.0001
Moderately differentiated	51%	47%	0.46
Poorly differentiated	10%	44%	<0.0001
Tumor Size
≤ 2cm	56%	50%	0.22
2–5 cm	41%	45%	0.41
> 5 cm	3%	5%	0.31
Standard Definitions
RS < 18	63%	36%	<0.0001
RS 18–30	33%	34%	0.79
RS ≥31	4%	30%	<0.0001
TAILORx Definitions
RS < 11	24%	13%	0.005
RS 11 – 25	63%	44%	0.0005
RS > 25	13%	43%	<0.0001

Multivariate Models Evaluating Relationship Between TOP2A Expression and Clinical and Genomic Features

Cox proportional hazards models were fit to examine the joint effects of factors on recurrence rates (Table 4). The models included the factors of age, nodal status, centrally determined tumor grade, and tumor size. In Model I, which did not include TOP2A expression, features associated with an increased risk of recurrence included young age (p=0.05), 2–3 positive axillary lymph nodes (p=0.001), and poor grade (p=0.002). Model II added TOP2A as a continuous linear variable to Model I; TOP2A x +5 vs. x was used for the hazard ratio, where x is an arbitrary value of TOP2A (comparable to the analysis of RS as continuous variable in the report by Paik et al¹³). In this model, TOP2A expression was a highly significant predictor for increased recurrence (hazard ratio 5.01, 95% CI 1.53, 16.43; p=0.008). Model III added TOP2A (high vs. low, using the median split) to Model I. In this model, there was a very strong trend of high TOP2A expression being associated with an increased risk of recurrence (hazard ratio 1.76, 95% CI 0.98, 3.15; p=0.06). The corresponding models with RS were also fit as both a continuous and categorical variable, as RS is used for clinical decision making using both scenarios. The models including RS yielded similar results, with TOP2A expression being significantly associated with recurrence when analyzed as a continuous variable, and showing a strong trend if evaluated as categorical variable (see supplemental Tables 3 and 4).

Table 4.

Estimated hazard ratios and 95% confidence intervals from joint models for recurrence rates

	Model I	P Value	Model II	P Value	Model III	P Value
Age 45–65 vs. >65	1.06 (0.45,2.49)		1.02 (0.44, 2.39)		1.02 (0.43,2.41)
Age ≤ 45 vs. >65	2.04 (0.82,5.09)	0.05	1.97 (0.80, 4.86)	0.05	2 (0.80,5.01)	0.05
Nodes 1 vs. 0	0.91 (0.51,1.62)		0.93 (0.52, 1.66)		0.89 (0.50,1.60)
Nodes 2–3 vs. 0	2.31 (1.46,3.65)	0.001	2.32 (1.46, 3.69)	0.002	2.21 (1.39,3.52)	0.003
Grade Mod. vs. Well	1.51 (0.75,3.03)		1.2 (0.59, 2.43)		1.28 (0.63,2.61)
Grade Poor vs. Well	3.14 (1.53,6.42)	0.002	1.95 (0.88, 4.32)	0.17	2.20 (0.98,4.93)	0.1
Tumor size >2 vs. ≤ 2 cm	1.15 (0.69,1.92)	0.6	1.08 (0.64, 1.80)	0.78	1.13 (0.68,1.88)	0.65
TOP2A x+5 vs. x			5.01 (1.53,16.43)	0.008
TOP2A High vs. Low					1.76 (0.98,3.15)	0.06

Impact of TOP2A Expression and Risk in Low, Intermediate, and High RS

The relationship between RS risk groups and TOP2A expression is shown in Table 5, including the standard and TAILORx risk group definitions.²⁵ The 5 year risk of recurrence for all patients was 4.8%, 14.6%, and 14.5% for patients with low, intermediate and high RS, respectively using the classical definitions. The relative risk (RR) of recurrence was significantly higher in the high vs. low risk (relative risk 3.0, p<0.0001), and intermediate vs. low (relative risk 3.0, p<0.0001), but not high vs. intermediate. Similar trends were noted if the TAILORx definitions were used. Thus, in this chemotherapy treated population, recurrence rates were significantly higher irrespective of the categorical risk definition for high or intermediate for low RS (but not when comparing high vs. intermediate RS).

Table 5.

Five year recurrence rates by Recurrence Score and TOP2A expression

Recurrence Score	All Patients	Low TOP2A	High TOP2A	Relative Risk of 5-year Recurrence
				High vs. Low TOP2A
Standard Risk Group Definitions
RS <18	4.80% (3.4, 6.7)	3.20% (1.9, 5.3)	8.10% (5.0,12.8)	2.6 (95% CI 1.3, 5.2, P=0.008)
RS 18–30	14.60% (11.6, 18.2)	9.90% (6.6,14.5)	19.70% (15.0,25.5)	2 (95% CI 1.2, 3.2, P=0.004)
RS ≥ 31	14.50% (10.5, 19.8)	13.70% (5.4,30.2)	14.80% (10.4,20.8)	1.1 (95% CI 0.4, 2.7, P=0.86)
TAILORx Risk Group Definitions
RS <11	3.30% (1.7, 6.4)	0.90% (0.2, 3.8)	8.60% (4.3,16.5)	9.3 (95% CI 2.0, 44.4, P=0.005)
RS 11–25	10.20% (8.1, 12.6)	7.00% (5.0, 9.8)	15.30% (11.4,20.2)	2.2 (95% CI 1.4, 3.4, P=0.0005)
RS > 25	13.20% (10.1, 17.1)	9.60% (5.0,17.8)	14.40% (10.7,19.2)	1.5 (95% CI 0.75, 3.0, P=0.25)

We evaluated the impact of TOP2A expression on recurrence rates in the RS risk groups. For patients with low TOP2A expression, the risk of recurrence for tumors with low, intermediate, and high RS using the standard RS definitions was 3.2%, 9.9%, and 13.7%, respectively. For tumors with high TOP2A expression, the corresponding risk of recurrence was 8.1%, 19.7%, and 14.8%, respectively. Therefore, high TOP2A expression was associated with a 2.6-fold increased risk (95% CI, 1.3, 5.2, p=0.008) of recurrence for tumors with a low RS of less than 18, and a 2.0-fold increased risk (95% CI, 1.2, 3.2, p=0.003) of recurrence if there was an intermediate RS of 18–30. TOP2A expression was increased in 33% of those with a low RS, and 50% of those with an intermediate RS. Notably, most tumors that had a high RS also exhibited high TOP2A expression. The results were similar if the TAILORx risk group definitions were used for the mid-range RS group of 11–25 (2.2-fold increased risk, p=0.0005) and high risk RS group of more than 25 (no difference), but provided greater discrimination of the low RS group of less than 11 (9.3-fold increased risk, p=0.005).

Relationship Between TOP2A Expression and Recurrence by Treatment Arm

There was no difference in recurrence rate between the two treatment arms of doxorubicin-cyclophosphamide (AC) and doxorubicin-docetaxel (AT) in the overall study population¹², nor in the subgroup included in this analysis. When TOP2A was analyzed as a continuous (linear) variable, then in a model with treatment, TOP2A and their interaction, the interaction was not significant (p=0.27). When TOP2A was analyzed as a categorical variable using the median split, the interaction was again not significant (p=0.08). To the extent that there was a suggestion of interaction, the treatment difference favored AC for low TOP2A and favored AT for high TOP2A expression. Alternatively, the TOP2A effect is somewhat weaker with AT than with AC.

DISCUSSION

We performed an exploratory analysis evaluating the relationship between recurrence and panel of selected genes in patients with HR-positive, HER2 normal breast cancer who received standard doxorubicin-containing chemotherapy plus hormonal therapy, and who were followed for at least five years. The approach used was not a genome-wide screen, but rather was limited to a panel of 374 genes that were selected because of their known or postulated association with prognosis or response to chemotherapy; the genes were identified by a search of the published literature, genomic databases, pathway analyses, and microarray-based gene expression profiling experiments performed in fresh frozen tissue. We found that of the 374 genes evaluated, increased expression of only 10 genes was associated with an increased risk of recurrence, with TOP2A exhibiting the strongest association, and with most of the other significant genes highly correlated with TOP2A expression. We also found that although increased TOP2A RNA expression correlated highly with poor tumor grade and high RS, its expression nevertheless remained highly associated with recurrence when adjusted for classical clinicopathologic features (including grade) and with Recurrence Score. Our findings confirm the report by Rody et al, who recently reported an analysis of Affymetrix microarray data from 1,681 breast cancer patients which revealed that higher TOP2A expression significantly correlated with tumor size, poor grade, HER2 expression, and positive lymph nodes, and was associated with a poorer survival in ER-positive, HER2 normal disease (P=0.001), but not ER negative disease.¹⁵ The association between TOP2A gene expression and recurrence was independent of whether patients were untreated or had received adjuvant therapy, and was the single most important prognostic factor in a multivariate model (HR 2.40, 95% CI 1.68–3.43, P<0.001). Our findings not only confirm this report, but also demonstrate that TOP2A RNA expression provided information that was complementary to prognostic information provided by the Oncotype DX Recurrence Score in those with a low and mid-range score.

There are no predictive markers that reliably identify which patients with operable breast cancer benefit from adjuvant anthracycline therapy. In HR-positive, HER2-normal breast cancer, the most common phenotype that comprises about two-thirds of all breast cancers, it has been suggested that anthracyclines are probably not beneficial because of the low incidence of TOP2A gene amplifications and/or deletions that are predictive of anthracycline benefit.⁹ In this analysis, we included only patients with HER2-normal disease as defined by immunohistochemistry in a central lab, a population expected to have a low likelihood of harboring TOP2A gene alterations. The strong association between TOP2A RNA expression with recurrence suggests that higher RNA expression of TOP2A measured by quantitative RT-PCR may be associated with resistance to anthracycline treatment in this population; alternatively, it may identify individuals who are resistant to chemotherapy in general, or destined to have a poor prognosis irrespective of therapy. In other settings such as colorectal and pancreatic cancer, high tumor expression of the target enzyme (thymidylate synthetase) has been associated with resistance to therapy (5-fluorouracil).^26–28 The results were similar in this analysis, in that high TOP2A RNA expression was associated with apparent resistance to doxorubicin-containing therapy. It is noteworthy that although TOP2A expression correlated with the Oncotype DX proliferation genes, and higher TOP2A expression was associated with resistance to doxorubicin-containing chemotherapy, high proliferation scores in other studies have been associated with resistance to tamoxifen¹³ and greater sensitivity to doxorubicin²⁹ and non-doxorubicin³⁰ containing-chemotherapy regimens. Based upon our results and other work, we hypothesize that low TOP2A expression may identify a group of tumors that may be particularly sensitive to anthracycline therapy, and that high TOP2A expression may identify tumors that are resistant to anthracyclines, and which may be more sensitive to taxanes. This hypothesis requires testing and validation in other trials.

In this analysis that included only patients with HR-positive, HER2-normal disease, we found a statistically significant association between the RS groups (i.e. Low, Intermediate and High) and risk of recurrence (Table 5). Although there was no association between RS analyzed as a continuous variable over its entire range and recurrence in the multivariate analysis (supplemental Table 3), a previously reported analysis of this same dataset indicated a significant association between RS and recurrence when evaluated as a continuous variable between a RS of 0 and 40, but not above 40.¹¹ This is consistent with previous studies demonstrating greater chemotherapy benefit in patients with an elevated RS treated with non-anthracycline³⁰ and anthracycline-containing adjuvant chemotherapy.²⁹ This also explains its lack of association with recurrence when evaluated as a continuous variable over the entire range of RS, especially in a population with HER2-normal disease (since increased HER2 expression contributes to high RS). In particular, TOP2A expression level provided information for patients with an intermediate (“classical” definition) or mid-range (TAILORx definition) RS, a range in which there is considerable therapeutic uncertainty regarding whether chemotherapy is beneficial; this uncertainty has led to initiation of the “TAILORx” trial.^{25, 31}

This analysis had a number of important strengths, including the central IHC testing for hormone receptors and HER2, the specific phenotype evaluated (HR-positive, HER2-normal), the method for evaluating TOP2A (RNA expression), and treatment administered (anthracycline therapy). There are also several important limitations. First, the median followup of 6.3 years may be of insufficient duration in patients with HR-positive disease, who frequently have late recurrences. Second, absence of a control group that did not receive chemotherapy makes it impossible to determine whether TOP2A is prognostic, predictive of therapeutic benefit, or both. Third, the absence of a non-anthracycline treatment arm makes it impossible to determine whether increased TOP2A expression is prognostic, predictive of benefit for any standard cytotoxic regimen, or predictive of benefit specifically for anthracycline-containing chemotherapy. In addition, although the exploratory nature of our analysis was pre-planned, the chosen cut off point for TOP2A expression (i.e. the median value 6.23) was not pre-specified, and the predictive accuracy may be overestimated. Although these questions cannot be answered from this dataset, they may be addressed by evaluation of archival specimens from other completed studies.

In conclusion, we have shown in this exploratory analysis that evaluation of TOP2A RNA expression in HR-positive, HER2-normal operable breast cancer treated with adjuvant anthracycline-containing therapy provides information that is complementary to classical clinicopathologic information, and to validated genomic predictors that include proliferation genes such as the Oncotype DX RS. Evaluation of TOP2A RNA expression may be particularly useful in situations whether there is therapeutic equipoise, such as patients with an intermediate RS. We view these finding as hypothesis generating, and believe that additional studies are warranted in this specific patient population and perhaps other settings.

Supplementary Material

1

Supplemental Figure 1: Consort diagram.