Abstract
Many types of cancers represent different forms of blocked differentiation, but the relationship with major, currently identified cancer driver mutational and epigenetic events is not well defined. Normal cellular differentiation depends upon hierarchically-organized sequences of control and synthetic events coupled with cellular proliferation. An ability to identify and possibly correct or otherwise modify defective cancer stem or daughter cell differentiation, or even non-malignant precursor cell differentiation occurring prior to the effect of oncogene and suppressor genomic activity should provide many new approaches to circumventing the oncogenic process. The use of whole-genome sequencing and related procedures coupled with computer-based analyses of developmental event trees, and with the numerous reports from the ENCODE (Encyclopedia of DNA elements) project that has uncovered the large number of potential underlying regulatory and other functions in the junk DNA of cells should eventually define the temporal progression of normal and aberrant differentiating events as they modulate cellular proliferation. This information should identify many new differentiation sites and their coupling with cellular proliferation as new potential targets susceptible to therapeutic manipulation.
A current notion of the oncogenic process is that core events required for malignant change include driver mutations, accompanied by passenger mutations; many if not most of the latter do not provide a growth advantage (1). A slightly different formulation assigns primacy to founder or public mutations present in all cancer cells, semi-private mutations in a smaller fraction of developing cancer cells and private ones found in only a few of them (2). Public mutations occur between the zygote and the first transformed cells, presumably reflecting certain types of hereditary cancer rather than DNA mutational or other events subsequently occurring in somatic cells. Not all driver or public events represent alterations in primary DNA structure, but may also include fusion of genomic DNA and changes in gene expression due to epigenetic modification of genomic readouts.
From one point of view, major (primary) driver events originate from discrete genetic events that can include a broad repertoire of additional secondary genomic outcomes, the passenger events, only a few of which function as proximal activating or inhibiting events germane to an evolving oncogenic process. The implications of the large number of recently reported potential regulatory sites residing in junk DNA have yet to be incorporated into this scheme. The ability of micro-RNAs to function as oncogenes or as tumor suppressors suggests some of the likely complexity to be encountered (3).
A related question is whether eradication of a carcinoma is simply a question of controlling its cell division or if manipulation of more fundamental aspects of the state of differentiation could prove to be therapeutically superior. In fact, there is a long history of considering cancer as the product of aberrant differentiation (4).
Epithelial Cancer Develops Over a Number of Years
In addition, many types of solid epithelial cancers develop over a number of years. They include carcinoma of the colon, breast, lung, prostate and pancreas (5). For example, essentially all males over, say age 65 who have followed a Western lifestyle have the pathological hallmarks of prostate cancer cells in situ. The great majority of these men will die of some other cause before they develop clinically-overt prostate cancer.
In one recent study of pancreatic cancer, there was considered to be at least a 10-year delay between the detection of the initiating mutation and the appearance of a non-metastatic founder cell, followed by further delay of five or more years before evolution of a metastatic clone (6). In a study of intra-tumoral heterogeneity and branched evolution of renal cancer, between 63 to 69% of all somatic mutations were not detected across all tumor samples. The mammalian target of rampamycin kinase, mTOR kinase and multiple tumor-suppressor genes leading to loss of function were observed and good or poor prognosis gene signatures were present in different regions of the same tumor (7). Both studies included findings consistent with evolution of a complex intra-tumoral heterogeneity and phylogeny; the first study suggesting a linear development, the latter study revealing a more branched unfolding, with many non-synonymous mutations. Agrawal et al. suggest the difficulties encountered, using whole-exome copying and gene copy number analyses. Thirty-two squamous cell head and neck carcinomas of various origins were analyzed and six genes were analyzed in 88 similar carcinomas (8). In addition to mutations in TP53, CDKN2A, PIK3CA and HRAS, mutations in FBXW7 and NOTCH1 were found. Some of the mutations in NOTCH1-expressing cells were likely to truncate the gene product, and led to it functioning as a tumor suppressor rather than an oncogene. In a companion article by Stransky et al., whole-exome sequencing of 74 head and neck carcinomas was performed (9). In addition to genes previously involved in these cancers, including TP53, CDKN2A, PTEN, PIK3CA and HRAS, a number of genes not previously reported to be associated with this type of cancer were observed, and 30 % of the samples contained genes including mutated NOTCH1, IRF6 and TP63 genes, considered to be involved in squamous cell differentiation and consistent with this as a major driver of head and neck squamous cell carcinogenesis. One implication of these results is that once identified, major genomic events specifying early to later differentiation events [imagined as nodes connected by links (10)] in the decision tree that delineates final cellular specificity, should provide new regulatory sites for potential therapeutic interventions.
Agents that Can Influence Differentiation
Both hematopoietic and solid carcinomas have been considered a result of disordered differentiation, and various cytotoxic agents, doxorubicin, hydroxyurea, adriamycin etc, and non-cytotoxic agents including vitamin D, retinoids, steroids, arsenic trioxide, dimethylsulfoxide and hypomethylating agents can induce limited differentiation of at least some cancer cells (11). All trans retinoic acids (ATRA) for therapy of acute promyelocytic leukemia (PML) represents an early paradigm of such therapy. A chimeric fusion protein from a 15/17 translocation [PML gene with growth inhibitory transcription factor fused with retinoic acid alpha (RARa) myeloid differentiation gene] circumvents the developmental arrest at the promyelocytic stage (12). Gleevec (imatinib) inhibits the 9/22 translocated BCR-ABL associated tyrosine kinase of acute myelogenous and related leukemias, resulting in more functionally normal cells (13) Attempts to differentiate stem-like cells from solid tumors, in one laboratory study employing glioma cells treated with ATRA to induce differentiation in vitro and in vivo have been undertaken (14).
The relationship between and among proto-oncogenes (potentially including growth factors, tyrosine and serine/threonine kinases, GTPases, and transcription factors, suppressor genes, other drivers and epigenetic events supporting various malignancies) and normal or aberrant differentiation are obscure. The underlying developmental programs generating malignant cells are not known but must include a very large number of other genes and their regulatory components as well as proto-oncogenes and their oncogenes. Of even greater importance for eventually teasing-out the control of differentiation pathways and their defective expressions are the current and future results from whole-genome, whole-exome (coding) and triptome (transcribed RNA) analyses, coupled with computer modeling of developmental trees, especially from ENCODE, the Encyclopedia of DNA Elements Project (15). Some four million potential DNA-regulatory but non-coding sites have been ascribed to the DNA previously denoted as junk, comprising about 80% of the biologically relevant cell DNA. A majority of potential regulatory circuits underlying the developmental options of diverse cell types remain to be identified, and once done, the means to activate or suppress those most associated with disease must be devised.
Gene Regulatory Networks and Cellular Differentiation
While much is known about the control of gene function, less information is available about how groups of genes are programmed to interact with other potential developmentally-related groups of genes. Understanding these relationships should yield many potential sites for re-directing cellular differentiation. A fusion of concepts from electronic circuitry (16), computing (10) and biology (17, 18) yields notational relationships between different levels of temporal functionality. Concepts include cis and trans regulatory circuitry, modules, nodes and edges, in addition to many more arcane ones. Were flow charts outlining the developmental options and their times of appearance inherent in stem cells and their progeny available, it seems likely that a number of sites for potential modification of an individual clone's developmental history would be identified. Recent developments in the sequencing of DNA and the identification of a role for what was previously identified as junk DNA referred to above (15), should make this an attainable goal.
For our purposes, a developmental node can be viewed as a regulatory focal point with multiple inputs and outputs connecting it to upstream and downstream developmental assemblies containing programs (circuitry) encoded as distinct developmental modules (cassettes). As the progeny of pluripotent stem cells express more differentiated phenotypes, the number of developmental choices expressed by these daughter cells in the transit-amplifying compartment become more and more restricted. as they lose the ability to proliferate.
Developmental Plasticity Inherent in the Evolution of Cellular Differentiation
Introduction or activation of selected genes into a cell can redirect differentiation in unexpected ways. Gurdon and Melton first demonstrated that adult frog nuclei could reprogram irradiated or enucleated frog eggs and generate a viable adult (19). Differentiated embryonic or adult mouse fibroblasts transfected with Oct3/4, Soc 2, cMyc and Klf4 have been converted to cells exhibiting properties of embryonic stem cells (20). There is evidence that some differentiating cells in the transit-amplifying compartment can revert to cells with cancer stem cell markers (21, 22), providing an example of reversion of a more differentiated cell to an earlier precursor. For example, a c-Met receptor tyrosine kinase, dependent upon downstream transcription factors associated with stem cells such as Sox2, Oct14 and Nanog, re-programmed glioblastoma cells towards a cancer stem cell-like phenotype (21).
There are numerous examples of directed differentiation of human-induced pluripotent stem cells as diverse as active motor neurones, (23), mature airway epithelia-expressing (CFTR) protein (24), cardiomyocytes (25) and renal cancer stem cells (26), among others. Eight growth factors, (bFGF), (TGF-B1), activitv-A, bone morphogenic protein-4, (HGF), (EGF), (BNGF) and retinoic acid directed human embryonic stem cells cultured as embryoid bodies into a variety of cells with different epithelial and mesenchymal morphologies consistent with epithelial, mesodermal and endodermal properties (27). Recently, normal mouse pups were derived from embryonic stem cells and from induced pluripotent stem cells (28).
In addition, proteins related to oncogenes and tumor suppressor genes such as BCL-2, ERB2, c-MYC, P53 and MN23, detected by immunohistochemistry appear during fetal gonadal development during the first trimester (29). Such proteins underwrite normal development, presumably provided no changes in primary structure or in their normal regulation have occurred.
A further contribution to the phenotypic mosaic presented by cancer cells in different locations of a cancer from transit amplifying cells that de-differentiate to putative cancer stem cells has already been mentioned elsewhere (20, 21).
A Simple Hierarchic Example
To consider a hypothetical example, assume a discrete defect in any of four primary hierarchically-organized developmental nodes activated sequentially for normal cellular differentiation denoted A→B→C→D, each coupled with a control of cell proliferation, is required for the eventual development of cancer cell clone D, capable of metastasis. Defects might directly affect the node itself, its coupling with cell proliferation, or some other downstream relationship with the succeeding node. Presumably such nodes are under the control of elements, some of which are located in regions of what was previously denoted as non-functional junk DNA. There does not seem to be an overall understanding of any given mammalian program for cellular differentiation, the options at sites which normally channel its development or can arrest its development, details of coupling with cellular proliferation at these sites, or of activation of the downstream developmental program. Lacking at least the outlines of such information, correcting any developing or extant malignant change or genetic or other inborn errors could remain somewhat hit or miss. One of the potential results of the ENCODE program would seem to be fine-grained descriptions of developmental event trees of interest.
Some Questions about the Properties of Hierarchic Developmental Programs
To what extent would the progression of later developmental events, as in the hypothetical example of developmental nodes B through D, become independent of the preceding ones? Would later events have achieved functional autonomy? If so, interdiction of an earlier node or attempted redirection to a different developmental program could have little effect on later development of a cancer. Ideally, successful interruption or other modulation of an earlier stage of oncogenesis would abort continued evolution of a malignant clone. At the very least, it might prevent development of new predecessor oncogenic stem cells, despite leaving those subsequent to the intervention unaffected. Redirection of a repaired clone to include cells expressing an earlier, more benign, pre-malignant lineage, or even induction of cellular quiescence, senescence, or even various forms of programmed cell death could have unexpected effects on malignant cells located developmentally downstream.
Does uncoupling of cellular proliferation relative to a physiological requirement for cell numbers in the local tissue niche occur with each affected stage? If the biological consequences associated with an identified developmental node were expressed through a driver mutation or an epigenetic event, their relationship might be obscured by oncologically-irrelevant noise of events that did not affect differentiation or replication, Driver events which contribute to dys-regulated proliferation have provided therapeutic targets that tend to regularize differentiation and control excess proliferation.
Developmental noise from passenger events that do not promote the development of cancer cells or modulate their proliferation obscure more fundamental events. This suggests the comment (paraphrased) ascribed to Winston Churchill that in war, Truth should be surrounded by a bodyguard of lies.
One way to visualize the phenotypic variations of malignantly-transformed somatic cells and their progeny is as a collection of interconnected, branching cellular arborizations, resulting in what amounts to cones of bifurcating progeny. Over time, the expected number of cells in such a developing cone might be of the order of AAAA>BBB>CC>D, depending upon circumstances (e.g. rates of growth, transition times, genomic stability and mutation rates) as these transitions occur. A cross-section through such a tumor would contain a varied collection of nests of cells at differing stages of malignant development, contributing to the lack of developmental synchonicity and the observed phenotypic variation with differing marker expression reported in comparative studies of different regions in the same tumor (7, 8). To what extent regional variations in gene expression within a tumor could thwart response to a targeted therapy would depend upon circumstances.
Opportunities from Manipulating Events within Developmental Programs
Use of analytical techniques alluded to in (15) gives promise of identifying events responsible for the development over time of cellular specificity, a flow chart of major determinative events of cellular differentiation with identification of the potential opportunities expressed or left dormant. Application of these procedures implies the possibility of redirecting cellular differentiation for the alteration of genetic and somatic disease. Presumably, events associated with particular developmental nodes and their event trees may include elements defined as pro-oncogenes, suppressor genes, molecules controlling DNA synthesis, a variety of small RNAs (3), some originating from DNA previously labeled as junk, epigenetic events involving covalent modification of proteins such as histones and others, all dependent upon contextual details expressed in each cell type, at a particular time and in a particular cellular niche, all likely to differ over time within and between cells of a related clone. Presumably some regulatory events would be proximate to developmental nodes and others biochemically more distant. It would be expected that some of the mechanistic details associated with such nodes would include new configurations of regulatory and synthetic events.
Currently, some of the less complicated strategies for modifying differentiation with therapeutic intent resemble those given to medical students of an earlier era studying dermatology; if the lesion is wet, dry it and if dry, wet it. For loss of function, restore or mimic the normal event; for a gain of some aberrant expression, including more distal control of an otherwise normal process, block it. Perhaps even better, activate a program representing an alternative pathway which precludes the oncogenic program for which it is substituted, or induce some form of lethality: senescence, apoptosis, autophagy or necrosis. Would activation of an earlier developmental program over-ride later oncogenic events? The developmental outcome of directly blocking a developmental node should be more widespread than simply interfering with an associated event such as cell proliferation. Blocking a consequence of the BCR-ABL oncogene fusion protein in chronic myelogenous leukemia with the tyrosine kinase inhibitor imtinib mesylate, represents a paradigmatic example of inhibiting a driver event.
Taken together, the information available to the interested non-expert seems to suggest a potential utility for identifying developmental driver or public mutation or non-mutational (epigenetic) oncogenic differentiation-related events, presumably responsible for impaired differentiation underlying the evolution of cancer cells. Not much information seems available about the true developmental wiring of controlling elements in normal cells as they interact with elements such as drivers essential for delineating their developmental programs. While viewing cancer as a defect in differentiation certainly is not new (4, 30), the ability to employ ENCODE as a tool for parsing-out differentiation pathways should provide a powerful means to do so.
Increasing knowledge on how cellular differentiation is controlled is likely to identify developmental defects responsible for malignant transformation of cells and suggest new targets for potential therapeutic intervention.
Acknowledgements
We thank the Seidel Family Trust and Dr. Jules Harris for their support. Thanks are also due to Dr. Peter Hart, Chairman of Nephrology, Cook County Hospital in Chicago, Illinois and his staff for their support.
- Received November 17, 2012.
- Revision received December 8, 2012.
- Accepted December 11, 2012.
- Copyright © 2013 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved