Tumors are highly heterogeneous and dynamic (2
). The cell populations, both normal and cancer, composing the tumor continuously evolve, posing a major challenge for effective cancer treatment. Thus, to understand the clinical behavior of tumors, it is essential to define the various cancer cell populations they contain and to determine how these populations behave together as a whole.
Targeting cancer cells with specific mutations, currently the most favored approach for cancer treatment, inevitably selects for resistant clones. Often these resistant cells already exist in the primary tumor at the time of diagnosis (22
) but go undetected due to the limited sensitivity of the methods currently used. Examples include the emergence of MET-amplified tumors in lung cancer patients resistant to EGFR inhibitors (25
) and the relapse of chronic myelogenous leukemia (CML) following imatinib mesylate treatment due to mutations in BCR-ABL that confer resistance (26
). Thus, there is a pressing need for the development and application of techniques that allow the quantitative definition of intratumor diversity at the single-cell level in archived clinical samples.
Here we have applied a new approach for the analysis of intratumor diversity based on FISH and immunofluorescence staining for selected markers specific for the phenotype of interest in combination with ecological and evolutionary methods for data analysis. Because both of these experimental methods are routinely used in diagnostic pathology laboratories and both data collection and its mathematical evaluation can fairly easily be automated, our approach can directly be translated into clinical practice.
Using breast cancer as an example, we have shown that application of ecological and evolutionary methods to genetic and phenotypic data collected on populations of individual cancer cells demonstrated a high degree of heterogeneity for chromosomal alterations in cancer cells homogeneous for markers associated with stem cell–like and more-differentiated epithelial cell traits and between in situ and invasive areas of the same tumors. These findings are inconsistent with the hypothesis that CD24+
more-differentiated luminal epithelial breast cancer cells are the direct progeny of CD44+
stem cell–like breast cancer cells. Thus, a simple hierarchical differentiation–based cancer stem cell model (27
) does not appear to be applicable for breast carcinomas. Furthermore, our results also show that both stem cell–like and more-differentiated cancer cell populations evolve during tumor progression. Even if CD24 and CD44 are not specific markers for stem cell–like and more-differentiated breast cancer cells, the degree of genetic diversity we observed within tumors is inconsistent with a strictly cellular differentiation status–based tumor progression model. Furthermore, recent data indicate that even homogeneous-appearing cell populations can have heterogeneous responses to physiologic stimuli due to cell-to-cell variability in protein levels and signaling pathways (28
). Thus, the clinical behavior, including therapeutic resistance and recurrence, of the tumor is determined by the combined behavior of the tumor as a whole; it must be regarded as a complex system composed of highly variable individual cells.
Intratumor heterogeneity is not limited to markers associated with differentiation states but is likely to be a general feature of all measurable characteristics, including gene expression, mutation, and epigenetic modification patterns. Indeed, prior studies analyzing the expression of various genes including estrogen receptor, cytokeratins, and HER2
in DCIS and invasive breast carcinomas have demonstrated a high degree of diversity in a subset of tumors (29
). The intratumor heterogeneity of several of these markers that are used for the categorization of breast tumors into major subtypes (e.g., luminal, HER2+
, and basal-like) reflects the imperfection of such classification schemes.
Intratumor heterogeneity has also been reported for mutations in tumor suppressor (TP53) and oncogenes (RAS, PIK3CA) in breast and other tumor types (3
), demonstrating a continuous selection process within tumors that drives their evolution. Because several of these mutant genes are targets of new molecular-based cancer therapy, assessing their heterogeneity within tumors prior to treatment is important for the design of more effective combinatorial treatment approaches.
Another interesting question pertains to the underlying mechanisms that maintain and promote intratumor heterogeneity. At this point, we can only speculate on these, since there is limited experimental evidence, especially in human tumors. With rare exceptions, tumors initiate from a single transformed cell, and multiple rounds of clonal expansion are required to produce a clinically symptomatic tumor. During this expansion phase, genetic and epigenetic instability, which is a common feature of most tumors, produces a wide range of tumor cell variants with favorable traits that natural selection acts on. In solid tumors, spatial restrictions lead to the generation of separate niches in different parts of the tumor that favor the outgrowth of cancer cells with different characteristics. This spatial restriction might also promote cooperation among tumor cell populations, as has been demonstrated in colon cancer based on mathematical modeling supported by some experimental observations (5
). Due to the importance of these issues for effective cancer treatment, further studies are required to better understand the sources of intratumor diversity and their consequences.
In summary, in this study we have demonstrated the power of analyzing tumors as ecosystems and suggest that quantitative measures of intratumor diversity might be clinically useful biomarkers predicting prognosis and response to treatment.