CANCER AND CELLS APPEARANCE
Cells change their appearance as they become metastatic, going through a process referred to as the epithelial to mesenchymal transition (EMT). This is mainly apparent as a reorganisation of the actin cytoskeleton: the scaffold of protein filaments that gives the cell shape and facilitates movement (Fig. 1).
1. Images of the epithelial-mesenchymal transition.
The EMT is a developmental programme and can be followed by the disappearance of certain proteins (e.g. E-cadherin (CDH1) and cytokeratin 18, which are markers for epithelial cells) and the up-regulation of others (e.g. vimentin and α- smooth muscle actin) as the cells start to look more like fibroblasts (more spindle-like).
The EMT occurs during normal development and also in wound healing (Fig. 2).
2. The cycle of epithelial-cell plasticity.
However, if you take breast epithelial cells that will not form tumours and induce the EMT (e.g. by expressing the transcription factors Twist or Snail) you can show that these changes are not the only ones to occur. Most of the mobile, mesenchymal-like cells that develop have a specific pattern of expression of two surface proteins, CD44 and CD24: they are CD44high/CD24low. This is precisely the pattern found on stem cells, cells that have the capacity for self-renewal and are pluripotent, that is, they can differentiate into any cell type. Because of their high capacity for self- renewal, stem cells are considered to be prime targets for the mutational assaults that produce tumours. As stem cells become progressively more genetically unstable they may promote changes in adjacent (stromal) cells that in turn part of an expanding tumour.
The theory that cancers arise from (mutations in) stem cells or germ cells is over 150 years old. It holds, in effect, that random mutations giving rise to a cell with the potential to develop into a tumour do not occur arbitrarily in terms of the affected cell (the ‘stochastic model’ of carcinogenesis in which any cell may be targeted) but arise specifically in precursor cells that retain some of the properties of stem cells. The main observations supporting this ‘cancer stem cell hypothesis’ are that (1) the ‘initiating cells’ in acute myelogenous leukaemia bear a phenotypic resemblance to normal hematopoietic stem cells (they are CD34+/CD38−); (2) major signalling pathways controlling stem cell self-renewal (WNT, Notch, Hedgehog) can, when dysregulated, also promote tumourigenesis; and (3) when cells from a tumour are transferred between animals only a very small sub-population (0.1 to 0.0001%) forms new tumours, and the cells with that capacity carry surface proteins that are also found on stem cells. Populations defined as cancer stem cells have now been identified in cell lines derived from a range of cancers, notably brain, breast, colon and lung, on the basis of specific membrane proteins (e.g. CD133). In addition, cancer stem cells are held to resemble their normal counterparts by producing lower levels of ROS than do mature cells, thereby diminishing the extent of DNA damage they suffer. This may, in turn, contribute to their resistance to radiation and they may also show elevated drug resistance.
The question of whether stem cells (or at least cells that take on stem cell characteristics) are the critical components of developing tumours is of great importance when we come to consider the impact of drug treatment. Take a tumour that has reached the size of a 1 cm diameter sphere, which means about 109 cells: a drug treatment kills 99.9% of the tumour cells so that a tumour is undetectable unless you cut out the tissue to analyse it. Nonetheless, despite the efficiency of the drug, one million tumour cells remain: if only one of these has stem cell properties a tumour will recur. This is not, however, the only potential problem. When stem cell-like features are induced in breast epithelial cells yet another surprising change occurs: they become sensitive to a drug that effect on the precursor epithelial cells. The drug is salinomycin (which makes the cell membrane permeable to potassium) and it kills cancer stem cells but not other epithelial cells. This raises the problem that established cancer drugs, for example, paclitaxel, may be relatively ineffective against stem cells so that they produce the most undesirable of consequences: an enrichment of the stem cell pool from which tumours develop. This is a problem that has already been encountered in breast cancer patients who have been treated with conventional drugs (letrozole or docetaxel) where the residual cells carry stem cell marker proteins. The CD44high/CD24low cells that appear during the EMT have other properties that are characteristic of stem cells and this phenotype is also associated with human basal-type breast cancers, especially those arising from inherited mutations in the BRCA1 gene frequently affected in breast cancer.
The implications for drug efficacy have focused attention on the cancer stem cell hypothesis. Notwithstanding, it remains controversial and recent findings suggest that it might be less an illuminating model and more a confusing distraction. Most critically these have revealed that, far from being a fraction of one per cent, the proportion of tumour cells with tumorigenic potential may exceed 25%, depending on the precise experimental conditions. Moreover, some of the surface markers that have been proposed do not, in fact, discriminate between tumourigenic and non-tumourigenic cells. Most persuasively, transcription profiling has shown that embryonic stem cells and cancer cells share gene expression signatures. Moreover, the embryonic stem cell programme includes a distinct MYC-regulated module – recall that MYC is a ‘master transcription factor’. Thus the coincidence in signatures and, by implication, other features of cancer cell ‘stemness’, arise from the central role of MYC and the fact that it is over-expressed in a high proportion of human cancers.
By the time many patients present in the clinic with a primary tumour causing overt clinical symptoms, there are already metastases in distant tissues. However, although some types of cancer cell metastasise particular tissues (e.g. secondary breast tumours often arise first in the lymph nodes and then in bone, liver and lung) how metastasising cells select their target sites is not at all clear.
For the best part of one hundred years Paget’s aphorism of ‘seed and soil’ pretty well summed up our knowledge of how metastasis worked. Eventually, the use of radio-labelled tumour cells provided conclusive evidence for dissemination through the vasculature and for metastatic development at selective sites, with the inference that these are determined by interactions between tumour cells and their host. Despite such advances, we’ve had to wait until the twenty-first century for any further, significant insight into this process.
Malignant cells may retain sufficient phenotypic features to identify the tissue from which they originated, despite their frequently abnormal appearance, and they are often described as having a de-differentiated, embryonic phenotype. Metastasis per se does not require genetic changes. In (Fig. 3), we alluded to the notion of ‘metastasis suppressor genes’ that restrain tissue do not affect primary tumour growth.
3. Genetic signals in cancer development.
These include, for example, NM23 that indirectly restricts cell division to repress the spread of melanoma, breast and colon cancer cells. Conversely, evidence that activation of a subset of genes promotes the migration of tumour cells has given rise to the concept of ‘metastatic signatures’. Many of the components of these profiles of activity affect growth- factor signalling – specifically the mitogen-activated protein kinase (MAPK) pathway. They may also include members of the interleukin family and we have already encountered IL4 and IL6 playing important roles in the tumour microenvironment. These cytokines are capable of stimulating the invasiveness of breast carcinoma cells and over-expression of another member of the family (IL11) has been associated specifically with colonisation in bone. IL11 is a regulator of bone cell proliferation and differentiation and can collaborate with another cytokine, osteopontin, which also regulates bone tissue homeostasis. Both of these genes may be further activated by TGFβ, abundantly present in the bone matrix. These and related findings are beginning to reveal cooperative interactions that enhance metastatic activity. Even so, our current picture of the molecular basis of metastatic control remains cloudy.
A major breakthrough has come from mouse studies revealing that cells in primary tumours release proteins into the circulation and these, in effect, tag what will become landing points for metastasising cells. In other words, these sites are determined before any tumour cells actually set foot outside the confines of a primary tumour (Fig. 4).
4. Metastatic book-marking.
The process is more complex than by a single type of protein released by a tumour. As well as sending a signal to the target landing strip, the tumour also releases proteins that act on cells in the bone marrow. This is the site of synthesis of circulating red cells and white cells, and the arrival of signals from a tumour causes the release of haematopoietic precursor cells (HPCs). These are to the signal from a tumour, carrying a marker in the form of a protein on their surface (in fact for one of the receptors to which VEGF binds, VEGFR1). These cells diffuse through the circulation until they reach the chosen location for metastasis, recognisable because the tumour signal tagged the landing site by making cells in that region produce a surface protein (fibronectin) to which the VEGFR1-bearing cells bind. They do this through a second cell surface protein, integrin α4β1. In other words, the target is recognised by HPCs because it provides a sticky patch for any cell with integrin α4β1 on its surface. The presence of VEGFR1 on the attached HPCs then provides a marker for tumour cells to home in on, a process that David Lyden and his colleagues have named ‘cellular bookmarking’, released by HPCs are then able to reshape the microenvironment for the tumour cells to come.
This seems an extraordinarily elaborate mechanism for directing tumour cells to a target. How might it have come about, given that tumour cells cannot evolve in the sense of getting better at being metastatic: they just have to go with what they’ve got? We don’t know but the most likely explanation is that they are taking advantage of natural defence mechanisms. We’ve commented several times that tumour cells are, in a sense, ‘foreign’. That is, they are an abnormal growth and activate the immune system. Perhaps what is happening is that the signals emitted by the tumour cells are just a by-product of the genetic disruption that characterises cancer cells. Nevertheless, they may signal that there is ‘damage somewhere in the body’ to the cells of the bone marrow. That at least would explain why the bone marrow decides to release cells that are, in effect, a response to a tumour. The second question is more difficult: why should tumours release proteins that mark specific sites? Well, we know that different types of tumour metastasis to different places. In fact, if tumour cells are cultured in vitro and the medium in which they have been grown is injected into mice – no cells, just the medium containing any proteins the cells have released – metastatic sites are generated just as if a primary tumour had been sending out signals. This must reflect the fact that specific proteins released by tumours encounter binding partners (receptors) that happen to be present on, say, bone cells or liver cells but not on most other cell types.