Estrogen (17β-estradiol) is a circulating steroid hormone that binds to intracellular estrogen receptors (ER) after passively diffusing through the plasma membrane (Thomas et al. [1]). Estrogen frequently plays a crucial role in breast tumorigenesis by promoting cellular proliferation, and decreasing apoptosis ([2–4]; Russo et al. [5]; Yager et al. [6]). ER expression in breast cancers is used as a prognostic and predictive tool that reliably correlates with the clinical progression of disease and its response to hormonal therapies.

Although the ER status of breast carcinomas is typically expressed as simply positive or negative, there is frequently considerable heterogeneity of ER expression among cells of the same tumor. In fact, typical classification of a tumor as ER positive requires only 1% of the cells expresses ER (Hammond et al. [7]). There is now evidence [8, 9] that the prevalence of ER expression within cells in the same tumor correlates with the degree and duration of response to anti-estrogen therapy.

Our goal here is to investigate the evolutionary and ecological forces that govern heterogeneity of ER expression in breast cancers. Recent studies have demonstrated substantial heterogeneity in cells within the same tumor as a result of intratumoral evolution [10–14]. Generally, this heterogeneity is viewed as a genetic process in which stochastic mutations generate new populations in an unpredictable if not chaotic process. We note, however, that genetic changes are simply one component of evolution and that intratumoral Darwinian dynamics emerge fundamentally from environmental selection forces that promote phenotypic (not genotypic) adaptations [15]. Furthermore, we acknowledge that a large body of work exists which addresses the complex dynamics of ER expression in vitro [16, 17] and in vivo (Shipitson et al. [18]). We embrace these works and do not suggest that phenotypic adaptation alone is sufficient explain variation in ER expression.

Instead, we propose that intratumoral cellular heterogeneity represents a predictable process driven by variations in environmental selection forces leading to predictable and reproducible adaptive strategies. The most obvious source of environmental selection is blood flow which, in most cancers, is spatially and temporally heterogeneous resulting in regions of necrosis in poorly perfused regions.

We propose that ER expression will be observed if it provides an adaptive advantage. Specifically, we propose that ER will be expressed only when estrogen is present in the microenvironment. When estrogen is absent, ER expression represents a needless expenditure of resources and will be selected against. Since the source of estrogen in the breast is typically (although not always) interstitial fluid and moves from the vessels into the cell by a simple reaction diffusion model identical to oxygen, nutrients, etc. [19], we propose the hypothesis that ER + cells will be found in regions of high blood flow while ER- cells will be present in regions of poor blood flow. This results in the prediction that the prevalence of ER + cells will generally follow the distribution of blood flow. To test this hypothesis we examined regional distribution of ER + and ER- cells compared to vascular density and regional necrosis within 24 clinical breast cancers of variable ER status and tumor grade.

ER expression is a useful predictive and prognostic biomarker in breast cancer, however, is often extremely variable even within the same tumor. Similar to cellular heterogeneity found in other cancers [10–14], this variation in ER staining is a consequence of intratumoral evolutionary dynamics (Figure 5). Here we address the Darwinian dynamics that might govern cellular ER expression. There are two general components of ER evolution. The first, which cannot be answered with our current study, is the variability of ER expression in the same tumor. Can breast cancer cells adjust ER expression (and other growth factor receptors) so that ER + and - cells in the same tumor represent a single generalist population that phenotypically adapts to various environments? Alternatively the ER expression could be a relatively fixed property and then ER + and – cells represent separate, specialist populations. The second component of intratumoral Darwinian dynamics is the environmental factors that are selection forces that define phenotypic fitness. This is the focus of our current work.

Here we propose that spatial heterogeneity in ER expression is the sequela of intratumoral evolution driven not by random mutations but by variations in environmental selection forces and predictable cellular adaptive strategies. We specifically hypothesize that ER expression will correlate with local concentrations of estrogen. Since estrogen diffusion from blood vessels is spatially limited by reaction–diffusion kinetics similar to oxygen and glucose, we predict a correlation between vascularization, necrosis and ER expression (Teschendorff et al. [29]; [30–32]). In order to test cell density as a plausible barrier of diffusion we evaluated the mean cell number per mm² and did not find any significant difference in the 24 samples evaluated. This suggests cell density alone does not correlate with ER status.

Our results do show that ER + tumors are associated with larger blood vessels and a lower percentage of tissue necrosis. It should however be noted that differences in CD34 staining may over or underestimate the vascularity due to tumor-specific alterations in the vascular bed such that not all of the endothethial cells may be appreciated. Our initial prediction that vessel number will increase in ER + samples was not supported. However, the interaction between vessel size and ER status was three times higher than the interactions between tumor grade and ER status and vessel area. Our second hypothesis, that ER status would be inversely correlated with necrosis, was even more strongly supported. This suggests that as ductal carcinoma in situ progresses towards invasion, 1) the larger the vasculature is early in disease progression, the lower the volume of necrosis and 2) if necrosis does not increase with the cancer progression, then ER + cells are more likely to dominate the population.

While the number of patients evaluated in this study is limited, and a larger patient population would be desirable, the number of individual vessels evaluated is on the order of 10³ to 10⁴ per patient. For this reason, our results indicate statistical significance to detect differences between ER positive and negative patients.

Furthermore, the amount of necrosis in DCIS samples was not found to be dependent on the size of the DCIS or the availability of vasculature outside of the basement membrane of the duct itself. Initially, we hypothesized in DCIS increased necrosis would correlate with ductal size and in turn the distance from the center of the gland to vascular resources. This was not the case in our results. Also of interest, in a single case vasculature was observed inside the DCIS. This case was not found to contain necrosis and was ER + (Figure 4E).

Poor vascularization and necrosis was originally hypothesized to be a proxy for hypoxia induced cell death and thus an indicator of low estrogen availability. However, there are a number of plausible explanations (many of which may be responsible in part) why necrotic regions may play a role. In our previous work, we hypothesized that tumor heterogeneity could be predictable similar to that of a riparian zone in a desert environment. Oxygenated phenotypes or relatively highly perfused regions could be equivalent to mesic species and poorly vascularized (distal from a blood supply) would be equivalent to xeric species [15]. In this regard, ER-positive phenotypes are mesic while ER-negatives are xeric phenotypes.

Xeric habitats are formed by evaporation of water and accumulation of salt which results in salty soil that select for xeric species. In our scenario, hypoxic phenotypes may be shaped by a depletion of nutrients, oxygen and metabolites. That is why ER-negative cells may be adapted to tissue of poor vascularization and higher necrosis. Of course toxification (i.e. salty soil) in such an environment may be another plausible consideration. Regardless, this unavailability of estrogen is one reasonable explanation for estrogen-independent tissue selection. This may be a testable hypothesis in vitro or using techniques including laser capture microdissection to isolate specific regions of high vascularity within patient tumors and evaluating the estrogen concentrations. This is a key future direction for this research.

A significant limitation of our analysis is our inability to measure temporal variations in blood flow. That is, the cyclical and random variations in blood flow which have been extensively observed. These variations will result in temporal variations in estrogen concentration which could alter ER expression. This imprecise link of vascular density and blood flow could result in similar variations in the correlation between vascular density and ER expression.

In conclusion, we find ER expression and metrics of vascular density and blood flow in 24 clinical breast cancers show direct correlations between ER + tumors and blood vessel size and inverse correlation with necrosis consistent with predictions. This correlation, if confirmed, suggest strategies to promote intratumoral blood flow could be explored as a somewhat counter-intuitive approach to increase the efficacy of anti-estrogen drugs.

Due to natural selection, ER- tumors could be evolving in a way that resists (adapts to) the absence of estrogen. As a future direction, we hypothesize that anti-estrogen therapy (e.g. Tamoxifen) can select for ER-independent cells. In contrast, cyclic introduction of estrogen may improve survival rate by continually altering, rather than unilaterally shifting, toward an ER- population. In other words, this theory suggests that modulation (and not eradication or extinction of certain population) may prove to be an advantageous treatment strategy.

Furthermore, it may be possible that ER + cells cluster around vasculature and effectively act as a barrier. While this is a future direction of this research and has not yet been tested, it may explain how both populations coexist spatially in a single tumor. More interestingly, it may also be possible that this spatial pattern keeps ER- cells farther from blood vessels where they might enter the bloodstream and form metastatic tumors. ER- may be more prone to metastasize if ER + cells are less successful invading novel tissues. Future work should determine the spatial relationships around vasculature and the propensity for each population to metastasize. This more in depth assessment of regional distributions of ER + and ER- cells will be important to understand whether heterogeneous ER staining correlates directly with vessel distribution (i.e. whether ER + cells congregate nearer to the vessels within a tumor).

In summary, we conclude that ER status is selected for given vascular availability which could have meaningful and exploitable therapeutic decision making implications.

Written informed consent was obtained via Institutional Review Board from the University of South Florida approval for the use of patient information for the publication of this report and any accompanying images.