Fax: (011) 44 131 537 2449
Predicting response and resistance to endocrine therapy†
Profiling patients on aromatase inhibitors
Version of Record online: 10 DEC 2007
Copyright © 2007 American Cancer Society
Volume 112, Issue S3, pages 689–694, 1 February 2008
How to Cite
Miller, W. R., Larionov, A., Anderson, T. J., Walker, J. R., Krause, A., Evans, D. B. and Dixon, J. M. (2008), Predicting response and resistance to endocrine therapy. Cancer, 112: 689–694. doi: 10.1002/cncr.23187
Presented at Endocrine and Targeted Manipulation of Breast Cancer: Proceedings of the Sixth Cambridge Conference, April 30–May 1, 2007, Cambridge, Massachusetts.
- Issue online: 18 JAN 2008
- Version of Record online: 10 DEC 2007
- Manuscript Accepted: 2 OCT 2007
- Manuscript Revised: 26 SEP 2007
- Manuscript Received: 30 JUL 2007
- genetic expression;
- neoadjuvant therapy;
- clinical response/resistance;
- breast neoplasm;
- estrogen receptor
Selection for endocrine therapy requires the identification of markers that accurately predict response/resistance. In this report, the authors review their published work and abstract results from an unpublished study to illustrate the potential of RNA microarrays from sequential tumor biopsies from patients who were offered neoadjuvant endocrine therapy treatment to identify the molecular signatures associated with tumor sensitivity/resistance. Clinical response was assessed by serial ultrasound measurements in postmenopausal women with large, primary, estrogen receptor-rich breast cancers who received neoadjuvant treatment with letrozole for 3 months. Tumor RNA from biopsies that were taken before and after 14 days of treatment was hybridized on Affymetrix U133A chips to determine expression profiles. Classic estrogen-dependent genes and markers of proliferation were changed with treatment in most tumors but were poorly associated with clinical response (they frequently were changed in letrozole-resistant tumors). Differential expression patterns could be used to identify heterogeneity in clinically resistant tumors. The results indicated that molecular profiling of early changes with letrozole treatment offers the opportunity to distinguish between clinically responsive and nonresponsive tumors and provides important information about the heterogeneity of endocrine resistance. Cancer 2008. © 2007 American Cancer Society.
Third-generation aromatase inhibitors, such as letrozole, have an established role in the treatment of hormone-sensitive postmenopausal breast cancer.1–7 The drugs are specific, in that their only primary endocrine effect is to inhibit the aromatase enzyme8 and reduce endogenously synthesized estrogen.8-10 In hormone-dependent breast cancers, estrogen deprivation may cause cell death, reduced proliferation, and tumor regression. However, this does not occur in all tumors, and there is a need to identify molecular markers of response so that treatment may be given selectively. In this report, we review some published work and abstract results from an unpublished study from our own research to define early changes in the expression of estrogen deprivation/proliferation markers.
MATERIALS AND METHODS
Patients were postmenopausal women who presented to the Edinburgh Breast Unit with large, primary, estrogen receptor (ER)-positive breast cancer. All provided informed consent to be included in the study, which had been approved by the local ethics committee (2001/W/BU/09 and 2001/W/BU/10). A consecutive series of 80 patients was recruited excluding patients who had known multifocal tumors or tumors of special histological type. Patients received neoadjuvant treatment for 3 months with letrozole (Femara; 2.5 mg daily). Core biopsies were taken with a 14-gauge needle before and after 10 to 14 days of treatment. Clinical response was based on reductions in tumor volumes, as determined from ultrasound measurements over 3 months. Reduction in volume >50% was taken as evidence of clinical response.
Tissue Processing and RNA Extraction
Biopsies were snap-frozen immediately and stored in liquid nitrogen. Frozen sections were taken to confirm the presence of cancerous tissue. Only samples in which the malignant component comprised ≥20% of the section area were used in further analyses. Biopsies were pulverized using U2 microdismembranator U (Braun Biotech). Total RNA was extracted by using TRI-reagent (Sigma) and was purified further on RNeasy mini columns (Qiagen). RNA quantity and quality were verified on a Bioanalyser 2100 (Agilent).
RNA Amplification and Reverse Transcription
RNA (500 ng) was subject to 2 rounds of amplification.11 Briefly, the first round of amplification used nonbiotinylated ribonucleotides and the Ambion MegScript kit (Ambion, Austin, Tex). This complementary RNA (cRNA) was converted to double-stranded DNA, and biotinylated cRNA was generated by using the Enzo kit.
Microarray Hybridization and Data Normalization
Biotinylated cRNA was fragmented and added to Affymetrix HG_U133A chips (Affymetrix, Santa Clara, Calif) as described in the standard protocol outlined in the Gene Chip Expression Analysis Technical Manual (Affymetrix). After hybridization, microarrays were washed with a custom GNF chip washer and were scanned with an Affymetrix 3000 laser scanner. Expression values as read out from Affymetrix' CEL files were normalized by using Robust Multichip Average methodology.12-14
Derivation of Database
One of 80 tumors was a protocol violation, because it was an ER-negative tumor on review. In 11 of the remaining 79 tumors, 1 or both biopsies had either <20% malignancy or failed to yield good-quality RNA. Microarray analysis was performed on the remaining 68 tumors (84%). Analyzable results were obtained in paired biopsies in 58 tumors (73%). The database from these tumors was subjected to bioinformatic analyses.
Changes in Gene Expression Associated With Therapy
Three approaches were used to characterize changes in gene expression that occurred with therapy: 1) to examine consistency of change by identifying genes that were up-regulated or down-regulated in most tumors irrespective of the magnitude of effect, 2) to identify genes that showed the greatest magnitude of change irrespective of the number of tumors, and 3) significance analysis of microarrays15 to identify the genes that changed most significantly with therapy. The resultant gene lists have recently been published.16 These revealed changes in genes classically associated with estrogen regulation (KIAA0101, trefoil factor 3 [TFF3], SERPINA3, IRS1, TFF1) and cell proliferation (CDC2, Cyclin B1, CKS2, TYMS, PCNA). Details are summarized in Table 1.
|Gene||Cases (58 patients)||Decrease (median fold)||SAM (P value)|
|KIAA0101||54 (2nd)||2.28 (3)||<10−8 (2)|
|TFF3||45 (43rd)||1.71 (14)||0.005 (200+)|
|SERPINA 3||45 (43rd)||1.96 (6)||0.0002 (129)|
|IRS-1||48 (14th)||1.50 (27)||0.001 (176)|
|TFF1||50 (60th)||2.61 (1)||6.1 × 10−5 (96)|
|CDC-2||43 (74rd)||1.82 (9)||0.005 (149)|
|Cyclin B1||49 (8th)||1.69 (16)||<10−8 (11)|
|CKS-2||50 (6th)||1.86 (7)||8.4 × 10−6 (58)|
|TYMS||47 (22nd)||1.42 (43)||8.1 × 10−7 (39)|
|PCNA||44 (59th)||1.37 (62)||0.001 (113)|
Gene Profiles and Clinical Response/Resistance
Of 58 tumors that were analyzed for genetic changes, 6 tumors were excluded because of inconsistencies in assessment of clinical response. Seventy-one percent of the remaining tumors (37 of 52) were classified as responders (>50% reduction in tumor volume by ultrasound), and 29% (15 of 52 tumors) were classified as nonresponders (<50% reduction in tumor volume).
Markers of estrogen sensitivity and proliferation
Changes in the markers identified in Table 1 are illustrated in Figure 1 and are subdivided according to clinical response. The predominant effect was a decrease in expression, and this was observed irrespective of whether tumors were classified as clinical responders or nonresponders. Examination of individual molecular patterns in the nonresponding group (Fig. 2) suggested heterogeneity in the patterns of change that occurred with treatment. Thus, it was possible to identify a group of tumors in which 1) both estrogen-dependent and proliferation genes were decreased, 2) markers were changed only marginally, and 3) estrogen-dependent genes were decreased but proliferation genes were increased.
Blockade of estrogen synthesis using aromatase inhibitors increasingly occupies a central role in the management of postmenopausal women with ER-positive tumors. Agents like letrozole, anastrozole, and exemestane have great potency and specificity8, 17, 18 and can be used to examine the molecular effects of estrogen deprivation in breast cancers in vivo. Although investigations have been performed on cell lines19 there is no substantial investigation on primary breast cancers.
Not all ER-positive tumors respond to aromatase inhibitors, and there is a need for molecular markers that predict for response/resistance to treatment to optimize patient management and identify molecular pathways associated with resistance. By combining a neoadjuvant protocol in which tumors are sequentially biopsied before and after 14 days of treatment and RNA microarray analysis on the paired biopsies it is possible to: 1) identify early changes in gene expression produced by treatment and 2) correlate gene expression profiles with tumor response/resistance to treatment.
Gene Expression Changed With Letrozole Treatment
Among the down-regulated genes were KIAA0101, insulin-like growth factor 1 receptor, SERPINA3, TFF1, and TFF3. These are classic examples of estrogen-regulated genes in which it has been demonstrated that expression is either induced by estrogen or reduced by estrogen deprivation/antiestrogens in experimental systems.19–23 In addition, down-regulated probes included those for the genes associated with proliferation, for example, CDC2, cyclin B1, CKS2, TYMS, and PCNA. Again, it has been demonstrated that these genes are down-regulated by estrogen deprivation in estrogen-responsive systems in vitro.20, 24–27 The observations in the current study also are consistent with reductions in immunohistochemical staining for the proliferation marker Ki67 observed in many of the same tumors.29
Clinical Response and Resistance
Classic estrogen-regulated genes, such as TFF1, are unlikely to be accurate predictors of response to therapy in individual breast cancers, because they frequently are changed with treatment in both clinically responding and nonresponding tumors. This is consistent with our previous report that the expression of steroid receptors (and their change with treatment) did not differ significantly between responding and nonresponding tumors.28
Similar observations apply to changes in proliferation indices that occur more frequently than clinical response. We reported this previously with regard to immunohistochemical staining for Ki6728 but now demonstrate the same phenomenon with regard to RNA expression of other markers of proliferation in the same tumor population. The challenge remains to explain why the marked reductions in proliferation indices do not translate into clinical response.
Bioinformatic analysis also indicated that nonresponsive tumors exhibited dissimilarities between each other with regard to these markers, indicating heterogeneity in tumors with a primary resistance to aromatase inhibitors. This is in keeping with the diverse mechanisms which are suggested for tumor resistance to aromatase inhibitors in particular8, 29 and to endocrine therapies in general.8, 30, 31
It also can be deduced that the absence of a clinical response is not always associated with the lack of molecular changes. Therefore, primary resistance to letrozole should not be equated with hormone-insensitivity. Indeed, it is possible that compensatory molecular changes with treatment may be a cause of clinical resistance.
The questions and discussion below follow from the oral presentation given at the Sixth Cambridge Conference on Endocrine and Targeted Manipulation of Breast Cancer and do not correspond directly to the written article, which is a more general review.
Dr. Paul Goss: We know that the aromatase inhibitors drop estrogen levels in and around the primary tumor. You can see a set of genes that are clearly regulated, presumably almost across all of the tumors, because they all have estrogen-regulated genes present—including the normal tissues. So you give the drug, and you see this family of estrogen genes drop or respond, but those genes don't predict the response of the tumor to the therapy. Instead, another set of genes, which are not estrogen-regulated, do predict the response. The families of genes that predict response are similar in responders but not in the non-responders. What does that mean? Is it that the drug is working on different pathways from the estrogen pathway as we currently think? Also, is there a difference in this effect between the different aromatase inhibitors?
Dr. Miller: In terms of different aromatase inhibitors, we don't know, because we've only looked at letrozole, but what I'm saying is that if you look at those classic estrogen-regulated genes, they clearly go down. We know that estrogen deprivation is affecting the tumor, because the estrogen-regulating genes are going down. But there is also another group of genes, which are not classic estrogen-regulated genes, which change, and that those are influential in predicting between responders and nonresponders.
Dr. Coombes: I suppose you could argue that the most suitable molecular targets are the ones which, at 2 weeks, have not altered. You could therefore hypothesize that you've already knocked down your estrogen-regulated targets, but the ones that you haven't affected are the ones that you need drugs against.
Dr. Miller: You could say the molecular endpoint is a signature which allows you, at 14 days, to determine whether a patient is going to respond or not respond. The other reason to do this is to understand something about the biology. The whole idea is to identify potential targets. Some of the genes do change with treatment and may be even up-regulated. Maybe the treatment induces these pathways and thereby achieves a resistance phenotype. If you know this, then you can add in treatments that circumvent those pathways.