Heterogeneity of drug target expression among metastatic lesions: lessons from a breast cancer autopsy program.

The target tissue for most cancer therapeutics is metastatic disease, either the treatment of overt metastases or the prevention of colonization by micrometastases in the adjuvant setting. Despite this, few drugs are tested in the metastatic setting in preclinical experiments, and relatively little is known about the expression of therapeutic targets in human metastases. A firestorm was ignited by several reports in which paired primary breast tumors and metastases were profiled by gene expression microarrays. Paired samples clustered together using multiple bioinformatic methods, suggesting little variation in gene expression patterns (1–3). Other reports using the same type of analysis identified alterations in gene expression between paired primary tumors and either lymph node or distant metastases (4–8). These reports are supplemented by studies at the protein level indicating a myriad of expression differences between primary tumors and matched metastases (reviewed in ref. 9, Supplementary Table S1). Whether there is a difference in molecular profile between primary tumors and their metastasis is not strictly an academic question, considering the fact that there is a growing belief that molecular target expression data should be used to enter patients onto clinical trials. Is the information provided by a primary tumor alone sufficient? Wu et al. (10) recently analyzed tissues from rapid autopsies of 10 breast cancer patients, providing a wealth of data on therapeutic targets in the metastatic setting. The patients were roughly half lymph node positive versus negative at diagnosis; the mean interval between diagnosis and death was 6.3 years (range, 2-11). Metastatic tissue was harvested from four to seven independent lesions, which represents an advantage over most other studies that have compared a primary with a singlematched metastatic lesion. Tissues were obtained within 4 h of death and should also be suitable for high-quality RNA expression profiling. Several trends observed in the expression of therapeutic targets are diagrammed on Fig. 1. With regard to estrogen receptor (ER), six of the primary tumors were positive; the remainder of the cases met criteria for triple-negative breast cancer. For the ER cases, the ER staining of the metastases was equivalent to that of the primary tumor in two cases, reduced (in terms of percentage of positive staining tumor cells) in two cases, and negative in the remaining two cases. Similar trends have been reported previously in the literature, based on immunohistochemical analyses of primary tumors compared with a lymph node metastasis, in some (11–14) but not all (15, 16) articles. Progesterone receptor expression followed similar trends in the autopsy study, decreasing in the metastases from two cases and becoming negative in an additional two cases. Although many elegant pathways have been described for tamoxifen resistance, simple loss of hormone receptor expression must be considered another contributing factor. The expression of several receptor tyrosine kinases was determined (10). Four of the ER primary tumors were positive for epidermal growth factor receptor (EGFR) by immunohistochemistry. In each case, heterogeneity was observed in EGFR levels among the metastases. Similarly, c-Met was expressed by immunohistochemistry in the four triple-negative cases. c-Met staining was consistent between primary tumors and metastases in two of these cases and heterogeneous in the remaining two cases, varying from 0 to 2+ in the metastases. Her-2 was unamplified in all primary tumors, thus limiting the analysis of this receptor tyrosine kinase. Fluorescence in situ hybridization data for the metastatic tissues were unamplified in the vast majority of cases but reached 3.0 in an omental sample from a single case. Wu et al. (10) also presented a comprehensive analysis of the DNA methylation status of seven gene promoters. Relative concordance of promoter methylation status among primary tumors and multiple metastases was found for the HIN1, Twist , and ERa genes, whereas various degrees of discordance was found for RASSF1A, Cyclin D2, APCc1 , and RARb (Fig. 1). The data suggest that some genes may be more amenable to epigenetically based therapies than others. Although the complete loss of a therapeutic target in a metastatic lesion makes sense as a potential determinant of drug response, what is the significance of a change from 2+ to 1+ intensity, for example? Few targets have been comprehensively examined for the relationship of quantitative expression to drug efficacy, yet the idea is attractive. One of the best examples is ER, where a quantitative Allred score has been developed. The Allred score reflects the percentage of positively staining tumor cells added to an average intensity score. This continuous score segregated patient disease-free survival to endocrine therapy, suggesting that intervals in expression may be biologically meaningful (17). In a neoadjuvant chemotherapy trial, the Allred score increased between biopsy and surgery in 49 of 296 breast cancer patients and was associated with improved disease-free survival but not overall survival (18). When ER was quantified by another multilevel algorithm, a shorter time to recurrence was associated with lower levels of ER in tamoxifen-treated (P = 0.07) and anastrozole-treated (P = 0.009) patients in the Arimidex, Tamoxifen, Alone or in Combination (ATAC) trial (19). For EGFR, multiple metrics Commentary