The essentials of the current concepts of cancer growth are three fold: (1) Cancer originates via mitotic division of normal cells with DNA damage 7; (2) Cancer cells originate from mutant stem cells called cancer stem cells (CSCs), which have the potential to self-propagate by asymmetric division yielding one CSC and one differentiating into tissue-specific tumor cell; 891011121314 and (3) since CSCs retain the properties of stem cells, they are immortal and have unlimited division potential without being subjected to the phenomenon of aging. Emerging new evidence hints at a major revision of these current concepts about cancer.

Current concept of cancer is based on the belief that tumor cells arise after about 13 mitotic divisions of the initiated cell 7. In addition, cancer cells also multiply by mitotic division. Therefore, the conventional non-surgical cancer treatments, consisting of chemotherapy and radiotherapy target only the mitotic populations of tumor cells. They do not differentiate between proliferating normal and tumor cells. This results in undesirable side effects that limit the level of usable dose-intensity of these treatments and restrict their application to only the fittest of patients, this approach results in the growth of resistant tumor cells in the place of originally responsive tumor often in a matter of months by a mechanism that has not been clearly understood. Development of novel strategies to improve current status of cancer therapy will require identification and exploitation of yet unrecognized differences between normal and tumor cells with respect to propagation, evolution and development of resistance to conventional treatments 1516.

The common belief that cancer cells are immortal probably originated from the early attempts at mammalian cell culture in vitro. Gey et al. 17 grew normal epithelial cells and also cultured cells from invasive colon carcinoma. They established one of the oldest cancer cell cultures, HeLa cells with continuous division potential. This resulted in the notion that cells are immortal, while the whole organisms were mortal.

With the advent of sterile culture conditions and improved culture media, Hayflick published his well known study on normal embryonic diploid human lung fibroblasts. Hayflick suggested that there was a "finite limit to the cultivation period of diploid cell strains" and that this was "attributable to the intrinsic factors which are expressed as senescence at the cellular level," while transformed or cancer cells were capable of unlimited mitotic division potential and, therefore, were considered 'immortal' 18. It has been now well established that this Hayflick limit and the phenomenon of senescence are due to the limited mitotic division potential of normal human somatic cells. At the end of their limited mitotic life span (MLS), normal cells enter a permanent non-proliferative senescence phase, characterized by a large, flat morphology, a high frequency of nuclear abnormalities and often consisted of multinucleate and/or polyploid giant cells (MN/PG cells). Senescent cells display positive stain for senescence associated β-galactosidase at pH 6.0 (SA-β-gal) 19. Such cells display mitotic crisis and are thought to die by mitotic catastrophe 20. Onset of replicative senescence in human cells is caused by the exhaustion of mitotic potential due to telomere attrition as a function of aging 212223. Telomere shortening beyond a certain limit triggers DNA damage response 242526.

Stem cells are characterized by (1) the potential to undergo continuous self-renewal and extensive proliferation, (2) the maintenance of a constant pool of the undifferentiated stem cells through the life time of the host and (3) the potential to undergo differentiation into multiple cell types upon receiving the appropriate stimulus from the microenvironment. The primordial stem cell is the fertilized oocyte or zygote, which gives rise to embryonic, germinal and adult stem cells during development. Embryonic stem cells (ESCs) isolated from the inner cell mass of 5–6 day old mammalian embryos are able to differentiate into different cell types derived from all three germ layers and are, therefore, considered pluripotent; ESCs undergo symmetric and logarithmic expansion by mitosis, both daughter cells being identical and retaining pluripotency. Germinal stem cells (GSCs) are confined to the gonads. After mitotic amplification, followed by a terminal meiotic division, they produce the egg or the sperm for sexual reproduction. During the early embryonic development, as the germ layers are formed from the ESCs, the process of cellular determination and organogenesis are initiated. The stem cells begin to undergo asymmetric mitotic division, where one daughter cell maintains the stem cell pool specific for different organs and tissues, and the other starts the process of differentiation through transit amplifying stages and becomes the progenitor of somatic cells. These are the adult stem cells (ASCs) or resident tissue-specific stem cells that are involved in tissue homeostasis and are capable of tissue regeneration. Thus, during development, the ESCs eventually give rise to about 200 different cell types of the adult organism, as would be expected of their pluripotency 627.

Studies on the susceptibility of fish and mammalian embryos to tumorigenecity by carcinogens showed that early embryos in vivo were resistant to carcinogenic agents. They became increasingly susceptible to tumorigenic mutations by carcinogens, after the development of tissue specific adult stem cells or ASCs and the process of organogenesis was initiated 28293031 The following two different conclusions can be drawn from these data 6:

(1) Adult stem cells were prone to tumorigenic mutations, while the transit amplifying and differentiating cells were not and the embryonic stem cells were also resistant to tumorigenic mutations due to their innate nature;

(2) Determined adult stem cells (ASCs) and their committed transit amplifying cells and derivatives on the way to differentiation were highly susceptible to tumor initiating mutations, while the embryonic stem cells were relatively resistant.

By this time, the Hayflick's concept of normal somatic cells had limited mitotic division potential 18 and the concept of 'immortality' of tumor cells 17 were well known. It was also demonstrated that stem cells with some mutation might play a role in the clonal origin of some cancers 303132. The concept of stem cell origin of cancers was postulated during the later part of the 19^th century 14. The first experimental proof was published by Till and McCullach 33. Based on teratocarcinoma studies, Pierce 34 proposed that cancer was due to maturation arrest of stem cells. Thus, the scientific atmosphere was ripe to interpret these data on embryonic carcinogenesis induction 303135 to mean that stem cells (ASCs) in regenerating tissues could be susceptible to epigenetic changes, while they still retained their unlimited division potential and, therefore, immortality. The proposal that transformation of normal cells into neoplastic cells with unlimited mitotic division potential was due to genetic and epigenetic alterations in ASCs, with the retention of their unlimited division potential 303135 seemed a reasonable assumption. Thus, the suggestion that mutant ASCs formed during the regenerative stages gave rise to mutant stem cells, and have retained their unlimited division potential was generally accepted. Based on the striking similarities between the ASCs and cancer cells, the 'immortality' of cancer cells has been attributed to a minority of tumor cells that can not only self-renew but also can give rise to a full fledged tumor tissue consisting of stem-like cells and abnormally differentiated cells, thus resulting in the hierarchical nature of tumor cell population 3637. These events lead to the formation of a cell mass consisting of a minor percentage of cancer stem cells, proliferating progenitor tumor cells and non-dividing, incompletely differentiated tumor cells making up the bulk of the tumor tissue. Thus, the concept that cancer is a disease of unlimited mitotic division has gained acceptance and the tumor-initiating cancer cells with self-renewal potential have been termed "Cancer Stem Cells" (CSCs). Therefore, akin to the ASCs, CSCs are thought to constitute a constant pool of mutated adult stem cells, and by virtue of their asymmetric division potential, are responsible for the continuous growth of tumor tissue, that consists of CSCs and differentiated tumor cells 8131437.

The concept that resident adult tissue stem cells turn into CSCs with their own specific surface markers has gained wide acceptance. Stem cell origin of cancers has been recognized in hematopoietic malignancies decades ago 814333839 and more recently in solid tumors such as breast tumors, brain tumors 910111213, melanoma 40 and prostate cancer 41, each having its own specific surface markers. The identification of such CSCs as a target for anti-tumor therapy is currently being actively investigated with the aim of using them as specific targets for cancer therapy 111213. However, this concept is still controversial and implies, but has not yet been unequivocally proven, that CSCs are capable of asymmetric division, i.e., one daughter cell maintaining the CSC pool and the other differentiating into tumor cells 6424344. Recently, the theoretical and technical difficulties of the CSC hypothesis have been reported 42. In addition, it is likely that such tissue specific CSC markers are not unique to CSCs, but may also be shared by normal stem cells and, probably, to a lesser degree by their transit amplifying somatic cell intermediates. Further, since the CSCs are thought to be immortal, this, in turn, implies that they are not subject to aging and senescence, which will inevitably lead to telomere attrition due to senescence-checkpoint and rejuvenation via neosis to turn cancerous (6, Also see below).

Weismann 45 popularized the concept of a complete separation in metazoans between a potentially immortal germ line and a mortal soma that transfers the germ line to the next generation and then senesces. This concept has led to the conviction that somatic cells are subject to senescence, while the germ cells are not. However, it should be pointed out that even the germ cells are also subject to aging along with its host 46; however, they are rejuvenated by sexual reproduction at the beginning of each generation. According to this definition, adult stem cells would also be considered to be somatic cells, since these are probably committed to differentiate, although they might be more primitive compared to transit amplifying cells 627.

In spite of the obvious difference in the division potentials between ASCs (potential to self-renew and unlimited MLS) and differentiating somatic cells (limited MLS), it is becoming increasingly clear that ASCs may lose division potential as they approach their differentiated state, and share several properties of the transit amplifying and differentiating cells. These include: (1) responsiveness of ASCs to both intracellular and extracellular factors to stop cycling and enter differentiation pathways or to enter cycling state and play a role in tissue homeostasis 2747; (2) an age-dependent decrease of telomere length in human hematopoietic stem cells (hHSCs), probably contributing to a reduction in their proliferation potential 4849; (3) limited division potential of adult human mesenchymal stem cells (hMSCs) in vitro that can differentiate into multiple cell lineages including bone, cartilage, adipose and muscle tissues, due to a lack of telomerase 505152; (4) the loss of tumor suppressor function of p16^Ink4a/p14^Arf (or senescence checkpoint controls) confers 'immortality' to hMSCs; (5) extension of population doubling of hMSC by transduction of the telomerase gene hTERT in vitro; and (6) the emergence of some sublines of the cells with extended MLS, loss of contact inhibition, anchorage independent growth potential and formation of mesenchymal tumors in 10/10 mice by acquired loss of p16^Ink4a/p21, due to epigenetic inactivation by methylation of DBCCRI gene purportedly involved in the onset of senescence 53

Conditional telomerase expression caused proliferation of hair follicle stem cells in vivo 54, while overexpression of mTERT in basal keratinocytes resulted in increased epidermal tumors and increased wound healing in transgenic mice in vivo 55. These events are similar to the events after telomerase transduction in non-stem somatic cell populations, respectively 6212223485657585960616263. In addition, telomerase is required for retarding the shortening of telomeres and to extend the replicative life span of HSCs during serial transplantation 64. Expression of hTERT in HIV-specific CD8+ T cells showed both an enhanced and sustained capacity to inhibit HIV-1 replication and enhanced antiviral functions accompanied by an increase in proliferative potential and telomere length stabilization 65. As in the case of somatic differentiating cells, stem cells also undergo telomere attrition during aging, which is followed by breakage-fusion-bridge cycle, giving rise to self-propagating mechanism for increasing the level of genomic instability via aneuploidy 485263656667.

Stem cells are attractive candidates to initiate cancer due to their pre-existing capacity for self-renewal and 'unlimited' proliferative potential, and their likelihood of accumulating mutations through the age of the host 1447. However, cancer causing mutations also arise in the more committed progenitors, the mitotic derivatives of stem cells at any state in the differentiation pathways in various tissue cell types 86869. In fact, most of the in vitro studies since 1970s on cancer initiation, promotion, and progression have been carried out not in stem cells, but in differentiating somatic, non-stem cells such as fibroblasts, epithelial cells, myoblasts, chondrocytes etc., of various mammalian species including mouse, rat, hamsters and humans; and the resultant transformed cells isolated using the transformed focus assay do form tumors in the appropriate hosts 570717273. Consistent with the concept of origin of tumor growth from the differentiating population beyond the initiation of determination stage is the fact that almost ~200 different types of human cancers are known, just as many as the number of differentiated cell types 27. This correlates well with ~200 different microRNAs that are known to be involved in the process of differentiation 747576. Recent findings that the microRNA profiles of cancer cells reflect the developmental lineage and differentiation stage of tumors 77 suggest that ASCs and the transit amplifying cells and not simply the 'immortal' ASCs alone are prone to carcinogenic mutations. Even differentiated cells may be prone to turn carcinogenic, when infected with tumor viruses such as acute transforming viruses (Eg. RSV), Human T cell leukemia virus, SV40, human papilloma viruses HPV-5, -8, -16, -18 and -31, human hepatitis virus B, and Burkitt's lymphoma virus etc., since these viruses carry their own transforming genes, that have potential to inactivate the tumor suppressor genes required for the onset of senescence and often carry growth promoting genes 78.

Since senescent cells do not respond to growth factors, and display terminal mitotic crisis, the phenomenon of senescence is thought to constitute a tumor suppressor program 798081 and is considered equivalent to the programmed cell death by apoptosis 82. The relevance of senescence as a tumor suppressor mechanism has been recently demonstrated unequivocally in different tumor systems in vitro and in vivo 8183 and constitutes a fail-safe, although not perfect, mechanism to inhibit cancer growth.

While senescent cells do not enter the mitotic cycle even in the presence of growth factors, they are alive and remain metabolically active in culture for several years 84. This cell cycle arrest is due to the tumor suppressor action of the genes involved in the p53/pRb/p16Ink4 pathway collectively termed the senescence checkpoint control. Abrogation of this pathway by mutation, epigenetic mechanisms or viral inactivation of any one of these genes bypasses senescence checkpoint and the cells grow beyond their normal intrinsic MLS. After an additional 20 – 30 population doublings, the cells enter a terminal mitotic crisis. During this second mitotic crisis period the cells display apoptotic death due to gross chromosomal abnormalities, some rare cells continue to proliferate yielding an 'immortal' cell line, by an unknown mechanism. These two mitotic crisis phases in the normal cell life span have been termed M1 and M2 5885.

Accelerated premature senescence in the absence of telomere attrition can be induced by different non-lethal conditions that cause acute genetic duress in primary cells in vitro and in vivo. Such cells also display similar characteristics such as non-responsiveness to growth factors, large, flat cell morphology with nuclear abnormalities, and stain positive for SA-β-gal and senescence associated heterochromatin formation (SAHF) [Reviewed in 1984]. DNA-damage induced repair response is activated in such cells, just as in the case of telomere attrition-induced senescent cells. Such acute genetic stress-inducing factors include unfavorable culture conditions 86, or the addition of aberrant oncogenic and mitogenic signals such as activated H-RAS 87. In addition, as can be expected, chemotherapeutic agents that induce DNA double strand breaks 88, and mitotic spindle toxins 8990 are also known to effect accelerated senescent phenotype.

Oncogene-induced premature senescence will terminate a pre-malignant condition before a fully transformed cell can develop from primary cells 8091 and is associated with the p53, p16INK4a, pRB pathway. Using a mouse model, in which the oncogene Ras was activated in the hematopoietic cells of the bone marrow, Braig et al. 92 have demonstrated that cellular senescence phenotype can efficiently block the development of lymphoma. pRB-mediated silencing of growth promoting genes by SAHF was shown to be formed via methylation of histone H3 lysine 9 (H3K9me) 93. The histone methyltransfearse Suv39h1 protein methylates histones and physically binds with pRB tumor suppressor protein 9495. Suv39h1 was shown to be required for oncogene-induced premature senescence due to the introduction of Ras in lymphocytes 92. Proliferation of primary lymphocytes was stalled by a SuV39h1-dependent H3Kme-related senescent growth arrest in response to oncogenic Ras, resulting in the inhibition of the initial step of lymphomagenesis. In Suv39h1-deficient lymphomas, RB was unable to promote senescence 92. Similarly, using conditional oncogene K-rasV12 in a mouse model for human cancer initiation, Collada et al. 96 have shown that both in lung and in pancreas, premalignant tumor tissues displayed extensive senescent cells as indicated by the expression of different senescent markers including p16INK4a, p15INK4b, SA-β-gal, Dec1, DcR2 and SAHF, while these markers were rare or absent in the same tumors after they turned malignant; this could prove useful both in prognoss and diagnosis of cancer. Similarly, chemically induced premalignant skin papillomas with H-Ras oncogenic mutation displayed senescence marker in vivo 96.

Inactivation of the tumor suppressor PTEN (a lipid phosphatase that negatively regulates PI3 Kinase-AKT/PKB survival pathway) produces hyperplasticity in mice prostate epithelial cells similar to precancerous lesions in human prostate epithelium. Expression of senescence markers was reported 97 in these lesions, which was associated with inhibition of the development of malignancy. Absence of p53 prevents the senescence response to loss of PTEN and loss of both p53 and PTEN leads to invasive prostate carcinoma in mice 97. In another study, using a conditional transgenic model with the mitogenic E2F3 transcription factor, Denchi et al., 98 have shown that E2F3 expression induced a burst of initial pituitary hyperplasia followed by a cessation of cell proliferation accompanied by expression of senescence markers. Michaloglou et al., 99 have reported that in cultures of human melanocytes and naevi, the benign precursors of malignant melanoma, an oncogenic allele BRAF, a protein kinase downstream of Ras, derived from human melanomas can induce sustained cell cycle arrest and senescence in fibroblasts and melanocytes, accompanied by the expression of p16INK4a and the common senescence marker, SA-β-gal. Congenital naevi in vivo were invariably positive for both SA-β-gal and spotty induction of p16INK4a expression, indicating that factors other than p16INK4a may cooperate with the mutant BRAF in bringing about senescence phenotype. The senescence phenotype was not brought about by telomere attrition, supporting the fact that oncogene-induced senescence is a genuine case of protective physiological process. After an initial cell proliferation, which results in the formation of naevi, such lesions typically remain static and benign.

Loss of Ku86 involved in chromosomal metabolism induces early onset of senescence in mice 100. Telomere fusions responsible for breakage fusion bridge formation can be caused by mutations in the terminal region of telomeric DNA 101102. It has been recently shown that psychological stress, both perceived and chronic, is significantly associated with higher oxidative stress, lower telomerase activity, and shorter telomere length, which are known determinants of cell senescence and longevity, in peripheral blood mononuclear cells from healthy premenopausal women 103

In general, the senescence checkpoint pathway genes such as MAPK 104105, and overexpression of p53 [reviewed in 106] and the genes that regulate p53 including ARF (p14^ARF in human or p19^ARF in mice), p33ING1 107, PML 108109110, nucleoplasmin or NPM 111 and PTEN 97 are all involved in the senescence-induced tumor suppression program [Reviewed in 1984].

The above studies indicate that an initial burst of cell division due to activation of an oncogene expression in primary diploid cells results in the formation of premalignant tumor growth arrest with senescent morphology, and such lesions often remain benign without any further proliferation, until it progresses to a malignant state by additional mutations 112 and epimutations. This type of senescent phenotype appears to be caused by a telomere attrition-independent or proliferative history-independent mechanism and is termed premature or accelerated senescence which is likely to be favored by the absence of telomerase in somatic cells. However, cell types which maintain telomere length due to endogenous telomerase activity also display a senescent phenotype in vitro. An example is rodent fibroblasts with long telomeres and telomerase expression that can be induced to undergo premature senescence due to culture conditions 85 or after exposure to genotoxins 5. Human epithelial cells when grown on plastic will undergo senescence, but, when cultured on a feeder layer, they proliferate indefinitely without any sign of senescence 113. These data demonstrate that even in the absence of intrinsic mechanism of limited life span due to telomerase expression, such cells will respond to extrinsic factor(s)-induced cumulative damage to DNA by entering premature senescence phase.

It is clear that the senescence phenotype can be induced under different conditions that might cause impediment to normal mitosis creating a mitotic crisis, including: (1) intrinsic ageing induced senescence (M1); (2) proliferative history dependent telomere attrition induced mitotic crisis (M2); (3) spontaneous, cumulative DNA damage induced senescence; (4) oncogene-induced accelerated senescence; and (5) genotoxin-induced premature senescence.

As described above, in primary cells, induction of an accelerated senescent phase acts as an efficient tumor suppressor mechanism; and DNA damaging agents or genotoxins also elicit an accelerated senescence-like phenotype [Reviewed in 114115116] by activating DNA damage response pathways 88. Tumor cells in vivo also display the phenomenon of senescence in response to exposure to genotoxins [reviewed in 1984]. Such cells may either undergo cytostasis and may never undergo mitosis again, even in the presence of growth factors, or they may die via apoptoss or mitotic catastrophe, Some cells may escape senescence and give rise to resistant tumor growth. (See Fig. 1 for the possible different fates of cells after exposure to genotoxins.)

Since senescence appears to be a tumor suppressor mechanism, it appears attractive to induce senescence in tumors in vivo in order to create a cytostatic state, where the tumor may not be completely eliminated, but can be maintained in a 'harmless' (non-proliferative) state 1984. This approach is especially attractive, since initiation of the senescent phenotype is accompanied by suppression of mitotic genes and overexpression of mitosis inhibitor genes [Reviewed in 84117]. Accordingly, te Poele et al., 118 have reported that in the archival breast tumors from patients who had undergone chemotherapy, when sectioned and stained for the senescent marker SA-β-gal, 41% of the tumors stained positive. In untreated controls only 10% of the tumors showed sporadic positive staining for SA-β-gal. Prof. E. Sikora, [Personal communication] has obsrved that SA-β-gal positive cells can undergo neosis, and tumor cells are apparently defective in senescent checkpoint control(s), and chances are that eventually some of these senescent cells might escape senescence and give rise to resistant tumor growth via neosis (see below). Therefore, induction of senescence as an anti-cancer therapy should be approached with caution for the following reasons:

1. Most of the genotoxins are carcinogens.

2. Tumor cells already have mutations in the senescent checkpoint pathway. Therefore, the chances of some cells escaping senescence are very high, especially in advanced tumors.

3. Senescent cells secrete factors that can promote tumor progression 81119.

4. Senescent cells facilitate tumorigenesis in adjacent cells 119.

5. Ageing senescent cells may accumulate additional mutations due to oxidative damage and escape senescent phase via neosis and may result in recurrence of resistant tumor growth (See below).

6. Since most anti-cancer chemicals are carcinogenic, and the tumor tissue is a mixture of normal and tumor cells, chemotherapeutic drugs may facilitate tumorigenic transformation of normal or preneoplastic cells 19.

Therefore, the approach of controlling cancer by inducing senescence in vivo, although tempting, may in the long run increase the chances of resistant tumor growth, or facilitate origin of new tumors. One has to better understand the molecular events that regulate senescence and the mode of escape from senescence in tumors, the longevity and fate of senescent cells in vivo, before one can design effective anti-cancer treatment strategies based on senescence.

Neosis is the newly defined mode of cell division that occurs only in senescent, polyploid cells, and has not been observed in normal diploid cells. Neosis is a parasexual, somatic, reduction division displayed by a subset of multinucleate and/or polyploid giant cells or MN/PG cells formed during the spontaneous senescent phase of normal cells at the end of their MLS or genetic stress-induced accelerated senescence phase in tumor cells. It is characterized by: (1) chromosome distribution to daughter cells via nuclear budding in the presence of an intact nuclear envelope, (2) followed by asymmetric cytokinesis, giving rise to an indefinite number of small, aneuploid, mitotically active cells termed Raju cells (Raju meaning King in Telugu language), after which the polyploid neosis mother cell (NMC) dies. Raju cells display transient stem cell-like properties, and mature into tumor cells with extended MLS, finally arriving at a secondary/tertiary senescent phase. Neosis is interspersed with the extended MLS of Raju cells and their senescent phase and is repeated several times during tumor growth in a progressively non-synchronous fashion. Further, neosis is also responsible for the outgrowth of resistant tumor cells after exposure to genotoxins. Neosis appears to be a mode of origin and continuous growth of different tumor types, including hematological malignancies, carcinoma and sarcomas. The significant role of neosis appears to be fine tuning the damaged genome by epigenetic modulation in order to escape death via stress-induced senescence 56.

Neosis-like events have been reported in the literature sporadically for more than a century 120 under different names [Reviewed in 6]. Although to date, these events were not generally correlated with any functional significance, it is becoming clear that such a process is involved in the escape of senescent cells into neoplastic cells [Reviewed in 6]. The earliest report that suggested a connection between neosis-like events with escape from the senescent primary cells was by Zitcer and Dunnabecke 121. These authors reported that human senescent amniocytes bypassed senescence via nuclear budding followed by cellularization. Zybina et al., 122123 have reported spontaneous polyploidization via endomitosis and endoreduplication during the differentiation of mammalian, extraembryonic placental trophoblast giant cells in vivo. Polyploidization was followed by fragmentation of the nucleus via nuclear budding into small individual aneuploid nuclei, resulting in synsytium-like post-budding multinucleate giant cells. Near the end of pregnancy, small individual cells were formed by asymmetric cytokinesis and cellularization. However, as in the case of p53+/+ MEF/MGB cells 5, these individual cells disintegrated without any further mitotic division. Based on the similarities between the steps involved in the maturation of trophoblasts and the process of neosis (Table 2), it was suggested 6 that gestational trophoblastic cancers 124125 originate via neosis due to compromise of some tumor suppressor function at different stages of trophoblast development. It is apparent that such a phenomenon is involved in the origin of tumor cells, due to inadvertent expression of neotic gene(s) during DNA damage-induced loss of p53 or related senescent checkpoint genes.

Solari et al., 126 have reported continuous production of small mononucleate Raju-like cells by nuclear budding from multinucleate giant polyploid cells formed by fusion of avian senescent peripheral blood mononuclear cells in vitro, even in the presence of an inhibitor of mitosis. These cells resembled Raju cells in their morphology and fused again initiating secondary nuclear budding to yield the next generation of Raju-like cells. Although the authors did not discus the relationship of their observations to senescence, immortalization or transformation of cells, these events are very similar to the primary (P-neosis) and secondary/tertiary or S/T-neosis reported by us 56. Walen in a series of articles 127128129 has reported similar nuclear budding giving rise to viable cells with continuous division potential in different cell systems. Romanov et al., 130 have reported the immortalization of senescent human breast epithelial cells in vitro; although the exact events of this process has not been described; they mention a process of micronucleation, which could be the process of nuclear budding related to neosis 5 or the nuclear fragmentation that occurs during mitotic catastrophe 131.

While the above reports described the P-neosis-like events in aging primary cells, recently S/T-neosis-like events have been described in human cancer cell lines treated with genotoxns. Upon exposure of human tumor derived HT1080 cells to thiophosphomidium, cells reached a rapid senescence state and produced 'macro cells' (polyploid giant NMCs), which yielded small mononuclear cells (Raju cells), by nuclear budding, a process termed 'sporosis' 132. In a detailed study on Burkitt's lymphoma cells with mutant p53, Erenpreisa and her co-workers 133134135136 have described sequential events that occur during radiation-induced neosis-like events that resulted in the production of mitotically active Raju-like cells. Makaravskiy et al., 89 reported that exposure of PC3 prostate cancer cells to docetaxel resulted in growth arrest, multinucleation and giant cell formation, which gave rise to docetaxel-resistant clones; this resistance was associated with transient expression of a β-tubulin isoform and was independent of P-glycoprotein, bcl2 and bcl-xl expression. The recent literature showing neosis-like events are listed in Table 1. In every case, senescent MN/PGs yield mononuclear Raju-like cells that display extended MLS.

In 1998, using computerized video time-lapse microscopy, we started to study the cellular process involved in transformed focus formation 5, which is considered to be the in vitro equivalent of in vivo tumorigenicity 707172. This involved: (1) video documentation of the transformation process at the level of individual cells, (2) isolation and cloning of the individual transformed foci without being contaminated by surrounding non-transformed cells, and (3) study the fate of these individual neotic clones (Raju cells derived from a single neosis mother cell or NMC) by serial cultivation under standard culture conditions and under anchorage independent conditions. These studies suggested that transformed focus formation occurs via the novel process termed neosis 5 and not via mitosis as previously thought.

Exposure of C3H10T1/2 cells to carcinogens induced accelerated formation of large senescent MN/PG cells that are supposed to be permanently arrested from undergoing mitosis and eventually die via mitotic catastrophe 1981137138. Around 14–21 days post-exposure to carcinogens, a minor subset of these senescence-like MN/PGs (hereafter termed the neosis mother cells or NMCs), before they died, produced several small mononuclear cells by karyokinesis via nuclear budding in the presence of the intact nuclear membrane. Each nuclear bud was immediately loaded with genomic DNA and surrounded by a small fragment of the cytoplasm delimited by plasma membrane by asymmetric cytokinesis. Each NMC produced about 10+/- 2 Raju cells. The latter were very small (~6–8 μm in diameter), had very high N/C ratio and immediately after birth, resumed symmetric mitotic division, inherited aneuploidy and genomic instability; they grew in soft agar, and displayed genotype and phenotype different from the mother cell. Presumably, they have reactivated telomerase expression 139140141 that resulted in the extension of their mitotic life span (MLS). Thus neosis and not mitosis is the mode of cell division resulting in tumorigenesis.

If neosis were the mode of transformation of normal cells into neoplastic cells, the spontaneous transformation of p53-/- mouse embryo fibroblasts should also involve senescent phase followed by neosis. In fact this is what we observed, when we studied the mode of spontaneous transformation of p53-/- cells using a similar approach. The primary cultures of p53-/- MEF/MGB cells entered senescent phase at the end of its diploid mitotic life span (about 6–7 passages), and almost all of these MN/PGs underwent spontaneous neotic division, each giving rise to 50 or more Raju cells, which by repetitive mitosis gave rise to spontaneously transformed cell lines. All the NMCs died within a month leaving the transformed cells in the Petri plate. The isogenic control wild type p53+/+ MEF/MGB cells underwent neosis yielding fewer number of Raju cells, but these Raju cells died without yielding a transformed cell line, probably due to the presence of p53, which is detrimental to cells with genomic instability 56142 (see additional files 1, 2, 3, 4, 5, 6).

We also observed certain degree of plasticity in neosis, where the nuclear budding during karyokinesis may remain constant, but cytokinesis may follow immediately to form Raju cells sequentially or be delayed to form post-budding synsytium-like multinucleate cell, followed by cytokinesis yielding several Raju cells simultaneously 6.

Raju cells are defined as the nascent daughter cells of neotic division, before they undergo their first symmetric mitosis. When the extended MLS of individual neotic clones was studied by serial subculturing, around the 20^th passage they spontaneously underwent a senescent phase and displayed mitotic crisis. Some of these MN/PGs spontaneously underwent secondary neosis, each NMC producing ~10 +/- 2 secondary Raju cells, only to repeat the cycle of EMLS followed by senescence, neosis, and production of the next generation of Raju cells. 5.

We and others have reported neosis occurring in several human and rodent tumor cell systems 56. In tumor cell cultures, neosis was repeated several times each neotic division interspersed with extended, but limited, MLS of the neotic offspring followed by a senescent phase. Thus, during three years of continuous subculturing, human metastatic neuroblastoma HTB11 cells underwent three episodes of spontaneous senescence followed by S/T-neosis and yielded new populations of mitotically active Raju cells in a progressively non-synchronous fashion 5, (see additional files 1, 2, 3, 4, 5, 6). The first and second episodes of mitotic crisis lasted for about 7 to 8 weeks (with medium change twice a week) during which time the cells remained in the senescent phase. The crisis period was succeeded by neosis, giving rise to small, mononucleate Raju cells with extended MLS. However, when HTB11 cells were subjected to carcinogen exposure, the senescent phase lasted for only two weeks before the appearance of Raju cells, reducing the duration of senescent phase and mitotic crisis, indicating that genetic stress was high enough to induce neosis within a shorter duration. This implies that during spontaneous senescence and prolonged mitotic crisis (up to 7–8 weeks), the cells may slowly be accumulating additional mutations (probably due to oxidative stress) raising the level of genomic instability sufficient to induce spontaneous neosis.

We, therefore, hypothesize that self-renewal of cancer growth is made feasible by the transient re-expression of some stem cell properties in Raju cells. This, probably, occurs after epigenetic modulation of the non-viable polyploid genome of NMC prior to the neotic S phase or S_N phase. Unlike in the mitotic S phase or S_M, where DNA is replicated only once before each division, during the S_N phase, DNA is replicated several times and the newly synthesized DNA is preferentially transported to the nuclear bud, followed by asymmetric cytokinesis. Thus, the NMCs display stem cell behavior in that the newly synthesized genomic DNA is asymmetrically segregated to the daughter Raju cells 56. The nascent Raju cell genome often underwent mitotic division even before cellularization while still within the cytoplasm of the NMC 6133.

Raju cells are unique in that they transiently display certain stem cell-like properties such as extended, but limited MLS, expression of telomerase, and potential to differentiate. They entered symmetric mitotic division immediately after they became independent cells. The mitotic derivatives of Raju cells grew larger in size and matured into tumor cells in a couple of days, while undergoing mitosis 5. We interpret the increase in cell size during the course of mitotic division as an embryonic cell property that incorporates the G1 phase of the cell cycle 6143. As they undergo symmetric mitotic proliferation, they mature into tumor cells and probably undergo defective differentiation and gradually lose stem cell properties. At the end of their extended, but limited MLS, the tumor cells spontaneously entered a senescent phase. Since these cells have already lost some senescent checkpoint pathway gene(s) function, some of these senescent cells stand a good chance of undergoing S/T-neosis, thus contributing to the continuity of tumor cell lineage. Thus, Raju cells behave like committed stem cells immediately after birth and slowly acquire somatic cell properties during the prolipherative phase. Given in the box below are some of the stem cell properties of Raju cells and the somatic cell properties of their mitotic derivatives (Table 3).

Chemotherapy and radiation therapy that target DNA as well as the process of cell division have been used with partial success as the main treatments of a variety of human tumors 144. In normal cells, cell cycle checkpoints protect the cells from accumulating errors in the genome by blocking cells at different points in the cell cycle by enforcing the dependency of late events on the completion of early events 145. Exposure of cells to genotoxins causes interruption of the progression of cell cycle due to these different cell cycle checkpoint controls. The individual cellular response to chemotherapeutic agents will depend upon the nature of the agent used, the position of the cell in the mitotic cell cycle, and the severity of the cellular damage. Cells with acute lethal damage may undergo immediate necrosis, or some of them will initiate the process of programmed cell death or apoptosis, which may lead to immediate cell suicide or delayed death by a day or two. In rare occasions, the cells may complete several mitotic divisions and the whole clone of cells might die simultaneously at a later date. Almost all the surviving cells will stop synthesizing DNA immediately. A major fraction of the surviving cells recover from the shock after several hours of quiescence and reenter cell cycle, probably after repair (or misrepair) of DNA damage. Most of such cells do not show any alterations in their morphology or growth behavior at least immediately after recovery [Fig. 1].

Of all the different checkpoint controls the most important one is the mitotic checkpoint or the spindle checkpoint, which is considered the primary defense against aneuploidy and ensures accurate chromosome segregation in order to produce genetically identical daughter cells 90146. During karyokinesis in mitosis and meiosis, the nuclear envelope is dismantled at which time the spindle checkpoint is activated 146147. Often the senescent cells form nuclear envelopes around fragments of the genome, a process termed micronucleation 131148149. Thus, unable to maintain G2 arrest, they enter mitosis and after being arrested for several hours at metaphase, they eventually die without successfully completing mitosis. This process is known as mitotic catastrophe or mitotic death 150151154155.

Since most cancer cells are deficient in the tumor suppressor function of the p53/pRB/p16INK4a signal transduction pathway, the G1, and G2 checkpoint functions are lost. After uncoupling apoptosis from G1 and S, tumor cells with mutant p53 arrive at G2/M interphase, at which time decisions regarding cell survival and death may be initiated. Cells that did not repair the damage to DNA arrive at G2 with point mutations and double strand breaks. Some of the double strand breaks are repaired by homologous recombination and non-homologous end rejoining during the G2 arrest 152153. Most of the surviving cells with unrepaired chromosomal lesions activate the senescent program via the DNA damage response. Senescent cells are characterized by SA-β-gal expression, SAHF, and are usually very large due to endomitosis and endoreduplication and are generally unable to divide again even in the presence of growth factors. Further, the senescent cells downregulate mitotic genes and upregulate anti-mitotic genes 198184. Under these circumstances, the senescent cancer cells will face inevitable death, unless they adapt and evade cell death, by eliminating the mitotically non-viable genome, and multiply by producing mitotically viable genome in order to be able to maintain the continuity of tumor cell lineage. All these conditions are fulfilled by neosis 5. In this connection, it is interesting to note that senescent SA-β-gal positive cells formed after exposure of human colon carcinoma HT116 cells to doxorubicin underwent nuclear budding and formed Raju cells via neosis [Prof. E. Sikora, personal communication].

We postulated that during such critical time of mitotic crisis, the cells may revert back to neosis, which may be an evolutionary throw back just to tide over the crisis 5. Interestingly, this mode of cell division resembles the asexual reproduction found in parasitic protozoans and protista and has been reported to be an evolutionary phase in sexual reproduction by meiosis in higher organisms 153. However, it has been demonstrated that such a primitive mode of cell division is still being activated during the maturation of extraembryonic tropholast cells during mammalian pregnancy. Thus, it appears that the gene(s) that execute neosis are still functional in the mammalian genome, which are supposed to be active only during the trophoblast maturation in the extra embryonic tissue during pregnancy 122123, and are inadvertently expressed in the absence of certain tumor suppressor genes (p53?) due to DNA damage.

When tumor cells are exposed to conventional chemotherapy or radiation therapy, most of the accelerated senescent cells may eventually die, but some do escape by adapting to the adverse conditions by undergoing polyploidization, and karyokinesis via nuclear budding 5133. Since mitotic checkpoint or spindle checkpoint is activated only after the dismantling of the nuclear envelope 146, polyploidization followed by nuclear budding with an intact nuclear envelope is bound to protect the NMC from cell death due to mitotic checkpoint. Therefore, some of these polyploid cells successfully undergo reduction division or de-polyploidization yielding near diploid daughter Raju cells with transient stem cell properties 56.

When spindle checkpoint is defective, a minor fraction of senescent cells escape mitotic catastrophe and become tetraploid due to mitotic slippage or cytokinesis failure 154155. Such cells have been shown to undergo multipolar mitosis giving rise to aneuploid cells 154. However, it is not sure if these aneuploid cells will survive long enough to contribute to the clonal selection in tumor tissue 156. We have shown that a subpopulation of tetraploid cells can undergo neosis by nuclear budding to give rise to several daughter cells before they die 5. Since the nuclear envelope is not dismantled during neosis, spindle checkpoint is probably not activated and this might favor successful completion of neosis, often accidentally yielding aneuploid neotic offspring, which may survive and proliferate due to loss of genomic stability. During mitotic catastrophe, nuclear envelope is dismantled and spindle checkpoint is activated, leading to mitotic catastrophe via micronucleation and death 2084148149150. Thus, mitotic catastrophe and neosis are mutually exclusive phenomena displayed by tumor cells with opposite effects. While the former eliminates the tumor cells with a defective genome, the latter helps adapt them to genomic instability-induced accumulation of genetic and epigenetic alterations by reducing the GI load in the neotic offspring via epigenetic modulation and thus contributing to the continuity of tumor cell lineage.

Epigenetic modulation is probably the way the genome alters its behavior in response to the environment 157. "The genome functions like a highly sensitive organ of the cell that monitors its own activities and corrects common errors, senses unusual and unexpected events, and responds to them, often by restructuring itself" 158. This statement probably accurately fits the behavior of cancer cells. Recent studies have revealed that genetic mutations alone do not lead to cancer and that epigenetics plays a major role in the origin and progression of tumors. Epigenetic alterations – non-DNA sequence-based heritable alterations – have been shown to initiate genomic instability, even before gene mutations enter into the process 159160161162. Epigenetic mutations fall into two main categories: (1) Altered DNA methylation of CpG dinucleotides, both losses or hypomethylation (results in gene activation) and gains or hypermethylation (results in silencing the gene) and (2) altered patterns of histone modifications such as acetylation or deacetylation of lysine residues, 160161162163164165166.

Epigenetic modulation occurs during all stages of tumor growth from the initiation at the progenitor cells through tumor formation and progression 162163. During the initiation stages, epigenetic modifications mimic the effect of genetic damage by altering the expression of tumor suppressor genes, thus compromising the tumor suppressor function of the senescence program; for example, silencing of tumor suppressor genes by promoter DNA hypermethylation and chromatin hypoacetylation, which may affect the expression of diverse genes including p53, RB1, p16INK4A, Von Hippel-Landau tumor suppressor (VHL) and MutL protein homologue 1 (MLH1) 109162163168169170. Global hypomethylation of chromatin and loss of imprinting lead to chromosomal instability and increased tumor incidence both in vitro and in vivo 165166167; epigenetic activation of R-ras is responsible for gastric cancer 168 and cyclinD2 and mapsin activation in pancreatic cancer 169170.

Tumor cells constitutively express some meiotic genes called Cancer/Testes (CT) antigens 171172173174. Immediately after exposure to radiation, some of these meiotic genes were translationally upregulated in Namalwa Burkitt's lymphoma cells with p53 mutant gene 175. It has been recently demonstrated that exposure of tumor cells to genotoxins results in the translational upregulation of cMOS gene 175, which is involved in switching cells from mitosis to meiosis II and forcing the cell to undergo reduction division 172. Constitutively expressed in low levels in untreated tumor cells, expression of meiotic cohesion gene REC8 was also enhanced after irradiation of p53 mutated Namalwa Burkitt's lymphoma cells, along with other meiosis-specific genes DMC1, STAG3, SYCP1 and SYCP3. Expression of these genes reached a peak level during the mitotic arrest phase and was proportional to the endopolyploid cells 175.

When such defective cells reach senescent phase and undergo neosis, these polyploid NMCs give rise to near-diploid or aneuploid Raju cells; this indicates that neosis must comprise properties of meiosis, mitosis and neosis specific events. During the extended MLS, Raju cells differentiate into tumor cells, while also accumulating additional mutational and epimutational alterations. This will increase the GI load, and cells will arrive at the next senescent phase due to incomplete differentiation displaying mitotic crisis. The onset of senescence is accompanied by several changes in the gene expression profile of the cells: (1) downregulation of mitotic genes; (2) upregulation of anti-mitotic genes 1984. During neosis the following changes in the gene-expression profile must occur: (1) reexpression of the 'immortalizing enzyme' telomerase is obligatory for the extension of mitotic life span, but not sufficient for neoplastic transformation 139140141176177; (2) ectopic expression of meiotic genes collectively called cancer/testes (CT) antigens in tumor cells, which are thought to play a role in transformation 171172173174175176177178, (3) in addition to ectopic expression of stem cell self-renewal genes including microRNAs, notch, oct4, and Bmi1 77179180181182183184 and (4) multidrug resistance genes 185186. Therefore, the most significant event during neosis appears to be the alteration in gene expression profile, which is likely to be brought about by epigenetic modulation. This will in effect reduce the GI load, making it possible to produce mitotically viable genomes of Raju cells from the non-viable polyploid genome. Our data and those of others 56127128129152153175187 suggest that the genome of such polyploid cells, although not mitotically viable, may undergo epigenetic modulation in order to reduce the degree of GI load to yield mitotically active daughter Raju cells with EMLS and thus contribute to the continuity of tumor cell lineage in a non-synchronous fashion, while the mitotically non-viable polyploid genome is eliminated by the post-neotic death of NMC. Thus, the genes for neosis which are silent in the genome, and should be active only during the maturation of the extraembryonic trophoblast cells during pregnancy 122123, are reexpressed in tumor cells in the absence of p53 or related tumor suppressor genes due to DNA damage.

Neosis is repeated several times during the growth of tumors. This implies that global epigenetic modulation occurs throughout the life of the tumors, repetitively at least during each neosis, since continuous proliferation will lead to accumulation of gene mutations and epimutations, which may be often detrimental to the cell. Therefore, we propose that during neosis the cell with defective senescence checkpoint control(s) tend to undergo endoreplication/multinucleation and after restructuring the genome followed by multiple rounds of neotic S phase (S_N phase), produces daughter Raju cells with EMLS via nuclear budding and asymmetric cytokinesis. This decreases the GI load in the Raju cells, while the non-viable polyploid genome of the NMC is discarded during its post-neotic demise. In the absence of tumor suppressor gene(s), the genome appears to be highly plastic (and tolerant to DNA damage) and responds to the damage and never activates the mitotic checkpoint by keeping the nuclear envelope intact and tides over the crisis by producing several Raju cells with stem cell properties and helps the continuous growth of tumor.

The question of the source and mechanism of the origin of aneuploidy is still being debated 188. More than a century ago, Hanesmann 189 and Boveri 120, suggested that aneuploidy is the result of non-disjunction during bipolar division or multipolor mitosis, respectively. The rare occurrence of pluripolar spindles represented Boveri's paradigm for a type of abnormal mitosis that can produce a variety of random chromosomal combinations. Unbalanced bipolar divisions or pluripolar mitoses will fail to distribute the chromosomal material to the daughter cells correctly. Therefore, both of these mechanisms can potentially give rise to tumor progenitors 120189190. Accordingly, a parasexual cycle of polyploidization and segregation of chromosomes has been reported to occur in human fibroblasts, which has been assumed to involve multipolar spindle formation and chromosome non-disjunction 191192. However, it should be pointed out that the observations of both Hanesmann 189 and Boveri 120 were largely made in tumor cell populations, which led them to arrive at this conclusion. Since tumor cells have already lost genomic stability, they could tolerate errors in chromosomal distribution and continue dividing in order to survive 193. Therefore, this does not address the question of origin of aneuploidy. We propose neosis as the third and most likely mode of arriving at aneuploidy for the following reasons:

(1) It has been shown recently that structural and numerical chromosome alterations in colon cancer develop through telomere-mediated anaphase bridges and not through mitotic multipolarity. In fact, multipolarity results in uneven chromosome distribution to daughter cells that gives rise to gross genomic changes such as nullisomies and non-viable daughter cells, and therefore, rarely contributed to clonal evolution tumor cells 194.

(2) It now seems that all human aneuploidies (cells that have chromosome number other than 46) in normal diploid human cell systems that occur during development result in embryonic lethality, except certain combinations of sex chromosomes and hyperdiploid, trisomies 13, 18 and 21, which yield severe birth defects in humans 195196.

(3) Further, it has been recently shown that chromosome non-disjunction in primary human cells yields tetraploid cells rather than aneuploid cells due to failure of cytokinesis. This demonstrates that tetraploid cells do not directly give rise of aneuploidy. Therefore, there must be an intermediate step between tetraploid cells and the origin of aneuploid progeny of tumor cells 197.

(4) We observed primary neosis that gave rise to transformed cells in a binucleate giant cell with a chromosome bridge 5. The fact that this NMC is binucleate with an isthmus, indicates that this cell has undergone one unsuccessful mitotic division attended by cytokinesis failure and has, therefore, an at least tetraploid genome.

(5) Thus, neosis appears to be the intermediate step between tetraploid cells and the origin of aneuploid tumor cells. It has the potential to give rise to aneuploid or near diploid stem cell-like Raju cells via polyploidization followed by karyokinesis via nuclear budding without activating mitotic checkpoint control by keeping the nuclear envelope intact.

(6) Accordingly, p53-/- MN/PG cells spontaneously gave rise to transformed cells via neosis 5 and in p53 null mouse cells cytokinesis failure-generated tetraploid cells promote tumorigenesis in vivo 198.

(7) These cells have extended MLS in the absence of senescent checkpoint controls (e.g. P53-/- cells); but they perish in the presence of proper checkpoint controls (eg., p53+/+ cells). Thus, p53+/+ MN/PG mouse cells did not yield viable Raju cells 56 and the p53+/+ tetraploid cells did not produce tumors in vivo 198.

(8) In support of this, it is also known that in several human and rodent tumor systems, tetraploidy is the intermediate stage before the genesis of neoplastic growth 155156199200201202203.

Therefore, we propose that senescent cells with tetraploid or higher ploidy genomes have the potential to under go neosis, creating conditions for automatic onset of aneuploidy to drive malignancy 67, if all the conditions for survival of the genome are met with in the resultant Raju cells.

Polyploidy can result either due to endomitosis and endoreduplication, or by cell fusion. It has been shown that Mad2- or BubR1-depleted cells (the genes involved in spindle checkpoint control) that do not complete cytokinesis remain viable through continued cycles of DNA replicatio