Cancer as an evolutionary reversal—time for a paradigm shift

Introduction

The currently prevailing theory of the origin and development of cancer is the somatic mutation theory (SMT). According to the SMT, cancer arises from the accumulation of critical gain-of-function mutations in oncogenes and/or loss-of-function mutations in tumor suppressor genes that are an inevitable feature of ageing and the interaction of our cells with the environment (e.g., ionizing radiation and carcinogens). As these mutations occur more or less stochastically across the nuclear genome and most mutations are neutral (passenger mutations), the emergence of cancer within the SMT has a probabilistic account of causation to it. However, the SMT is inconsistent with many experimental findings of which I will briefly list just a few. Firstly, many cancer cells brought into an embryonic environment stop behaving like cancer cells despite having the same mutations in their nuclear genome (1). Second, similar findings have been made for cancer cells in which the cytosol has been replaced by the cytosol from healthy normal cells, but the nucleus was left intact (2). This is evidence for a tumor-suppressing effect of healthy mitochondria regardless of the mutational landscape of nuclear DNA. Third, spontaneous remission of tumors in cancer patients without any treatment is a phenomenon that is rare, but real and contradicts the ultimate role of (irreversible) genetic mutations as drivers of cancer (3). Finally, tumor gene expression profiles could apparently be shifted towards those of normal tissue through the intention of gifted human healers and their “biofields” (4). To integrate these findings with the presence of cancer-characteristic mutations, it is necessary to look beyond the presence of particular oncogenic mutations and to consider the complete network of genes and their expression that can depend on external influences such as those described above.

The totality of gene expression, not individual mutations, determines the cellular phenotype

The phenotype of a cell is not so much determined by individual genes, but by the expression of all genes and their interactions within the gene regulatory network. The latter can be conceptualized mathematically as a high-dimensional state space spanned by the expression levels of individual genes (5). Each dimension of the state space thus corresponds to the expression level of a single gene, and at any given time, a cell occupies one point in this space determined by the expression levels of all of its genes. The interactions between these genes constrain the possible trajectories that a cell can follow within this state space over time. Distinct cellular phenotypes correspond to distinct attractor regions within the state space, which are separated by inaccessible and mostly instable regions. Figure 1 shows a simple example for a two-gene network. Here, gene A (x-axis) inhibits the expression of gene B (y-axis) and vice versa. For this system, the x-y plane defines the state space. Overlain in Figure 1 are contours of a quasi-potential that is defined such that for a given expression level of gene A and B the cell will follow a trajectory which leads downhill until a stable attractor with minimum quasi-potential is reached (6). Once a cell has occupied a certain attractor defining its phenotype (e.g., “quiescent and differentiated”, Figure 1A), it usually resides there as long as the overall gene expression profile does not change significantly. In the simple gene interaction network example shown in Figure 1 such a significant change can be achieved by constitutive overexpression of gene A, which forces the cell to switch its phenotype (Figure 1B). In reality, with many thousands of genes, the expression changes of only a few genes due to genetic mutations would usually not be sufficient to drive a cell out of its attractor state. However, mutations may soften the attractor barriers so that cells can eventually gain access to other attractor regions previously unreachable to them if they experience a major trajectory change triggered by, e.g., external stimuli such as cytokines.

Figure 1 The epigenetic or “Waddington-like” landscape for a simple gene regulatory network of two genes A and B that mutually inhibit each other. The x-y-plane defines the state space and the white contours indicate the surfaces of a quasi-potential defined by Bhattacharya et al. (6). White arrows point towards the direction of decreasing potential that is defined to be zero in an attractor state. The red lines indicate exemplary trajectories that a cell would follow from an initial (white circle) to its final (red dot) state. (A) Physiological state in which the attractor corresponding to high gene A (proliferation) and low gene B expression is separated from the quiescent, differentiated phenotype attractor corresponding to high gene B and low gene A expression. The differential equations governing this system are given as dxAdt=αKnKn+xBn−γxA and dxBdt=αKnKn+xAn−γxA, with x_A and x_B denoting the gene expression levels of genes A and B, respectively; α the maximum production rate of gene A or gene B when the other gene is absent; K the inhibitor concentration at which the production is half-maximum (dissociation constant); n the Hill coefficient; γ the degradation rate of both genes. For producing this Figure, I chose: α =20, K =10, n =3, γ =1. (B) Transformed epigenetic landscape by simulating constitutive overexpression of gene A through a “boost” term such that dxAdt=αKnKn+xBn−γxA+boostAwithboostA=5.

A test of dedifferentiation towards stemness versus evolutionary reversal

The metaphor of a distinct cancer attractor within state space is very useful, because it explains the convergence of cancer evolution towards similar cancer-specific phenotypes despite mutational heterogeneity within and across tumors. However, there are two competing hypotheses concerning the nature of such a cancer attractor: one postulates that it is an embryonic/mesenchymal attractor, so that cancer represents a “dedifferentiation” of cells (5); the other postulates that it is a unicellular attractor that has been conserved during the evolution from unicellular prokaryotes to multicellular eukaryotes, so that cancer represents an evolutionary reversal of cells (7,8).

A direct comparison between both hypotheses is not straightforward, as both would correspond to partly overlapping gene activation patterns and similar hallmarks such as fast proliferation or aerobic glycolysis. However, in a seminal paper published in the International Journal of Cancer, Alexander Vinogradov and Olga Anatskaya from the Russian Academy of Sciences directly contested both hypotheses using single-cell transcriptomic features of a large variety of cells (9). The researchers obtained transcriptome data for more than 18,000 cells which they categorized into normal stem versus normal differentiated cells (pairwise data for both of these categories had to come from the same experiment) and cancer versus normal cells (pairwise data for both of these categories had to come from the same patient). The authors then built a set of logistic regression models to predict the normal stem versus differentiated cell category and the cancer versus normal cell category, respectively. As predictors, the following three transcriptome-derived variables were included in each model: (I) a unicellular signature defined as the expression ratio of genes dating back to the earliest two phylostrata (Prokaryota and unicellular Eukaryota) to all other genes; (II) a dedifferentiation/stemness signature called PluriNet, taken from the Molecular Signatures Database (https://www.gsea-msigdb.org/gsea/msigdb); (III) a cell cycle signature to adjust for the fact that both unicellularity and stemness are associated with cell proliferation.

For the normal stem versus normal differentiated cell comparison, 18 predictive models were built to classify 18 paired sets of transcriptomic data differing by the type of stem and differentiated cell, respectively. All models revealed the PluriNet signature as a significant positive predictor of stemness. In contrast, the unicellular signature was generally not significant as a predictor and its regression coefficient frequently took on a minus sign. These analyses thus confirmed that the PluriNet signature is a specific predictor of dedifferentiation/stemness. For the cancer versus normal cell comparison, 20 logistic regression models were built to compare 20 different paired datasets that either corresponded to a particular cancer type and normal cells from the same organ or to comparisons between primary and metastatic cancer cells. In all 20 models, the unicellular signature was positively and significantly associated with cancer or metastatic cancer, respectively. In contrast, except for one case, the regression coefficients for the PluriNet signature were either negative or not significant. All these results were confirmed qualitatively when probit instead of logistic regression models were used. They thus provided compelling evidence for the hypothesis that cancer is more driven by an evolutionary reversal towards unicellularity than by dedifferentiation.

A transient return to multicellularity in polyploid giant cancer cells

In a recently published follow-up study, Vinogradov and Anatskaya extended their analyses to polyploid cancer cells, which can constitute more than 50% of metastatic tumor cells (10). By comparing the transcriptome profiles of about 10,000 samples of polyploid and diploid cancer cells, they found that in common polyploid cancer cells, genes of unicellular origin and pluripotency, but not multipotency, were upregulated compared to diploid cancer cells, while those originating from late multicellular strata (including mammalia) were downregulated. These trends were mirrored in a complementary analysis of the protein interactome, which revealed that proteins with a unicellular signature had more direct (measured as “local centrality”) and indirect (measured as global “betweeness” and “stress”) connections to other proteins than those with a pluripotency signature. In other words, the protein interaction network was primarily concentrated around genes of unicellular origin, confirming the previous analyses discussed above (9) and providing further evidence for the presence of a unicellular attractor.

Next, both researchers applied similar analyses to study a special type of polyploid cancer cells, namely polyploid giant cancer cells (PGCCs) that are associated with therapy resistance and cancer recurrence. To this aim, they utilized data from 15 experiments in which PGCCs were induced by treating initial diploid cancer cells with either ionizing radiation or chemotherapeutic drugs. To their surprise, they found that 13 of these experiments showed that PGCCs differed from their diploid progenitor cells by a transient downregulation of unicellular and pluripotent genes and a clear upregulation of multicellular genes. In the PGCC protein interactome, four clusters were identified including intercellular communication proteins and stress resistance proteins such as cytochrome P450 detoxification enzymes and small heat shock proteins (10). This shift in gene expression and protein interactions was transient, because it was confined to early progeny PGCCs, while late progeny PGCCs again followed the general trend of unicellular gene overexpression. In addition, for two early progeny PGCC lines derived from human ovarian adenocarcinoma cells by treatment with the PARP inhibitor olaparib, the same trend of unicellular gene overexpression and multicellular gene downregulation as in ordinary polyploidy cancer cells was observed.

Time for a paradigm shift

Collectively, these two papers by Anatskaya and Vinogradov lead to the conclusion that a paradigm shift is warranted to conceptualize cancer not as a genetic disease based on more or less random somatic mutations, but as a gradual and coordinated epigenetic and mutational shift towards a unicellular attractor. This is consistent with previous studies showing that genes typically upregulated in cancer (“oncogenes”) date back to the first unicellular organisms on Earth, while those downregulated (“tumor suppressor genes”) originated during the emergence of metazoans (11-14). Importantly, the loss-of-function of multicellular genes by somatic mutations is typically more common than gain-of-function of unicellular genes (12) and consistent with an evolutionarily conserved stress-response program that cancer cells obtain access to via epigenetic mechanisms (13). Wu et al. have shown that stress-induced up- or downregulation of genes can occur independent of somatic mutations, which may also implicate epigenetic modifications and/or polyploidy (15). In the framework of the epigenetic landscape concept shown in Figure 1, the gradual disruption of interactions between genes of unicellular and multicellular origin during oncogenesis (14) leads cells on a trajectory towards a unicellular attractor. However, as the example of PGCCs shows, cancer cells may be able to temporarily leave their trajectory to exploit other possibilities offered by genes of multicellular origin in order to find quick solutions when confronted with severe toxic stress. Interestingly, Vladimir Niculescu, a retired scientist from Germany, points out the homology between human PGCCs and multinucleated giant cells formed by species of the protist Entamoeba as a survival strategy under severe cell stress (16,17). Although Niculescu may overstate the role of PGCCs in cancer initiation, given their low number in tumors in general and in early stage tumors in particular (18), the findings of a short-term return towards multicellularity in early progeny PGCCs with increased cell interactions (10) is interesting in light of the Entamoeba-cancer analogy. This is because Entamoeba multinucleated giant cell formation first requires the aggregation of multiple cells, which points to increased intercellular signaling (19). Also other unicellular organisms are able to form multicellular aggregates and change their gene expression profile when exposed to toxic cell stress (20,21).

However, the finding of a transient activation of multicellular protein network hubs in PGCCs appears to contradict the notion of Lineweaver et al. that “cancer tends to revert, irreversibly, toward phylogenetically earlier states” (22). In the gene-centric model of Lineweaver et al., this is a consequence of accumulating genetic mutations leading to a serial loss of phylogenetically younger capabilities that could only be regained by highly improbable “reverse” mutations (22). However, the concept of a unicellular attractor in the epigenetic landscape as proposed by Anatskaya and Vinogradov and schematically shown in Figure 1 does not necessarily depend on somatic mutations to be accessible, as it is the complex interplay of gene regulation within the whole gene network that defines the topology of the state space potential (5). The examples of tumor reversion that I have listed in the first paragraph of this Editorial also speak against the irreversibility of the loss of normal cellular behavior.

Links with the metabolic theory of cancer

This leads to the question about the ultimate cause of cancer: why do cells start their trajectory towards the unicellular attractor? One possibility, mentioned by Vinogradov and Anatskaya, is that oncogenesis can “begin as dedifferentiation caused, for instance, by injury healing and the corresponding epithelial-mesenchymal transition”, with a gradual further shift towards unicellularity (9). Besides a role of tumor suppressor genes in terminating the inflammatory wound healing process (23), effective wound healing has much to do with the coordinated social behavior, or cooperation, between many different cell types such as macrophages, neutrophils and fibroblasts. Thus, an alternative view is that the failure of cooperation between cells triggers oncogenesis. I have generalized this idea to a transdisciplinary evolutionary theory according to which it is a breakdown of cooperation on a certain level of human organization that is the ultimate cause of oncogenesis (8,24). As outlined in Table 1, the different levels can be traced back to different major evolutionary transitions that were enabled by new types of cooperation. In addition, the different levels causally influence each other through top-down and bottom-up causation, so that the breakdown of cooperation at higher levels can ultimately induce a breakdown of cooperation at lower levels. At the genome level, this provides access to the unicellular attractor. An example for downward causation is the cellular level where cooperation between the mitochondria and the nucleus takes place. Healthy mitochondria are potent tumor suppressors, but if their cooperation with the nucleus fails this could lead to malignant transformation. Thomas Seyfried has provided compelling evidence that dysfunctional or damaged mitochondria can induce genomic instability, an upregulation of unicellular genes (a.k.a. oncogenes) and downregulation of tumor suppressor genes through retrograde signaling (25,26). Guha et al. have shown that reducing mitochondrial DNA (mtDNA) content in human mammary epithelial cells initiates the epithelial-mesenchymal transition through Calcineurin-dependent retrograde signaling and produces breast cancer stem cells; this process is reversible by restoring the mtDNA content (27). In addition, damaged or dysfunctional mitochondria reduce the available ATP yield; to compensate, cells may upregulate cytosolic glucose fermentation (the Warburg effect) and mitochondrial substrate level phosphorylation (glutaminolysis), both of which are metabolic hallmarks of cancer (28). According to Bhat et al., acquisition of a Warburg phenotype can alter the epigenetic landscape such that it places the cell on a new trajectory towards a cancer attractor (29). Finally, mitochondrial dysfunction is consistent with the loss of capabilities that emerged after the evolution of mitochondria and could underlie metabolic inflexibility of cancer cells as well as a sensitivity to an elevation of oxygen concentrations in the microenvironment. These two weaknesses of cancer cells could be exploited therapeutically through ketogenic metabolic therapy (30) and hyperbaric air/hyperbaric oxygen therapy (31), respectively.

Conclusions

In summary, the studies of Anatskaya and Vinogradov (9,10) further confirm the theory that cancer equals an evolutionary reversal towards unicellularity, explaining why the propensity to develop cancer is present in all multicellular organisms. The fact that this reversal to unicellularity can be triggered epigenetically from different levels of organization should lead us to rethink our therapeutic and preventive strategies. To be effective, cancer prevention and treatment must simultaneously target the multiple levels of human organization including the often-neglected psychosocial-spiritual level, the mitochondria and the tumor microenvironment (24,30,32). As the evolutionary reversal brings with it the loss of certain multicellular capabilities, cancer treatment should seek to exploit such losses in the form of “target cancer’s weaknesses” strategies (22). Ultimately, we need a paradigm shift away from the SMT towards a transdisciplinary theory of cancer as an evolutionary reversal in order to implement such ideas in cancer prevention and routine treatment approaches.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Translational Breast Cancer Research. The article has undergone external peer review.

Peer Review File: Available at https://tbcr.amegroups.com/article/view/10.21037/tbcr-2026-1-0012/prf

Funding: None.

Conflicts of Interest: The author has completed the ICMJE uniform disclosure form (available at https://tbcr.amegroups.com/article/view/10.21037/tbcr-2026-1-0012/coif). R.J.K. serves as an unpaid editorial board member of Translational Breast Cancer Research from December 2025 to December 2027. He also received a royalty from a book on cancer and cooperates with MITOcare (Munich, Germany), a manufacturer of nutritional supplements. The author has no other conflicts of interest to declare.

Ethical Statement: The author is accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Proietti S, Cucina A, Pensotti A, et al. Tumor reversion and embryo morphogenetic factors. Semin Cancer Biol 2022;79:83-90. [Crossref] [PubMed]
2. Kaipparettu BA, Ma Y, Park JH, et al. Crosstalk from non-cancerous mitochondria can inhibit tumor properties of metastatic cells by suppressing oncogenic pathways. PLoS One 2013;8:e61747. [Crossref] [PubMed]
3. Turner KA. Spontaneous/Radical Remission of Cancer: Transpersonal Results from a Grounded Theory Study. Int J Transpers Stud 2014;33:42-56.
4. Yang P, Jiang Y, Rhea PR, et al. Human biofield therapy and the growth of mouse lung carcinoma. Integr Cancer Ther 2019;18:1534735419840797. [Crossref] [PubMed]
5. Huang S, Ingber DE. A non-genetic basis for cancer progression and metastasis: self-organizing attractors in cell regulatory networks. Breast Dis 2006-2007;26:27-54.
6. Bhattacharya S, Zhang Q, Andersen ME. A deterministic map of Waddington's epigenetic landscape for cell fate specification. BMC Syst Biol 2011;5:85. [Crossref] [PubMed]
7. Davies PC, Lineweaver CH. Cancer tumors as Metazoa 1.0: tapping genes of ancient ancestors. Phys Biol 2011;8:015001. [Crossref] [PubMed]
8. Klement RJ. Cancer as a global health crisis with deep evolutionary roots. Glob Transitions 2024;6:45-65.
9. Vinogradov AE, Anatskaya OV. "Cell dedifferentiation" versus "evolutionary reversal" theories of cancer: The direct contest of transcriptomic features. Int J Cancer 2025;156:1802-13. [Crossref] [PubMed]
10. Vinogradov AE, Anatskaya OV. Polyploid giant cancer cells (PGCC): short-term return to multicellularity. Biol Res 2025;58:71. [Crossref] [PubMed]
11. Domazet-Loso T, Tautz D. Phylostratigraphic tracking of cancer genes suggests a link to the emergence of multicellularity in metazoa. BMC Biol 2010;8:66. [Crossref] [PubMed]
12. Chen H, Lin F, Xing K, et al. The reverse evolution from multicellularity to unicellularity during carcinogenesis. Nat Commun 2015;6:6367. [Crossref] [PubMed]
13. Cisneros L, Bussey KJ, Orr AJ, et al. Ancient genes establish stress-induced mutation as a hallmark of cancer. PLoS One 2017;12:e0176258. [Crossref] [PubMed]
14. Trigos AS, Pearson RB, Papenfuss AT, et al. Altered interactions between unicellular and multicellular genes drive hallmarks of transformation in a diverse range of solid tumors. Proc Natl Acad Sci U S A 2017;114:6406-11. [Crossref] [PubMed]
15. Wu A, Zhang Q, Lambert G, et al. Ancient hot and cold genes and chemotherapy resistance emergence. Proc Natl Acad Sci U S A 2015;112:10467-72. [Crossref] [PubMed]
16. Niculescu VF. Understanding cancer from an evolutionary perspective: high-risk reprogramming of genome-damaged stem cells. Acad Med 2024;1:
17. Niculescu VF. From multicellular constraint to unicellular control: Ancient mechanisms of genome reconstruction, repair, and expansion in cancer evolution. Cancer Plus 2025;7:42-72.
18. Qu Y, Zhang L, Rong Z, et al. Number of glioma polyploid giant cancer cells (PGCCs) associated with vasculogenic mimicry formation and tumor grade in human glioma. J Exp Clin Cancer Res 2013;32:75. [Crossref] [PubMed]
19. Hazra S, Kalyan Dinda S, Kumar Mondal N, et al. Giant cells: multiple cells unite to survive. Front Cell Infect Microbiol 2023;13:1220589. [Crossref] [PubMed]
20. Tong K, Bozdag GO, Ratcliff WC. Selective drivers of simple multicellularity. Curr Opin Microbiol 2022;67:102141. [Crossref] [PubMed]
21. Zhang X, Zhang Y, Chen Z, et al. Exploring cell aggregation as a defense strategy against perchlorate stress in Chlamydomonas reinhardtii through multi-omics analysis. Sci Total Environ 2023;905:167045. [Crossref] [PubMed]
22. Lineweaver CH, Davies PC, Vincent MD. Targeting cancer's weaknesses (not its strengths): Therapeutic strategies suggested by the atavistic model. Bioessays 2014;36:827-35. [Crossref] [PubMed]
23. Shaterian A, Kao S, Chen L, et al. The candidate tumor suppressor gene Ecrg4 as a wound terminating factor in cutaneous injury. Arch Dermatol Res 2013;305:141-9. [Crossref] [PubMed]
24. Klement RJ. Krebs. Weckruf des Körpers. Krebs verstehen, vermeiden, ganzheitlich behandeln. 1st ed. Baden Baden, Germany: Deutscher Wissenschaftsverlag; 2023.
25. Seyfried TN, Shelton LM. Cancer as a metabolic disease. Nutr Metab (Lond) 2010;7:7. [Crossref] [PubMed]
26. Seyfried TN. Cancer as a mitochondrial metabolic disease. Front Cell Dev Biol 2015;3:43. [Crossref] [PubMed]
27. Guha M, Srinivasan S, Ruthel G, et al. Mitochondrial retrograde signaling induces epithelial-mesenchymal transition and generates breast cancer stem cells. Oncogene 2014;33:5238-50. [Crossref] [PubMed]
28. Seyfried TN, Lee DC, Duraj T, et al. The Warburg hypothesis and the emergence of the mitochondrial metabolic theory of cancer. J Bioenerg Biomembr 2025;57:57-83. [Crossref] [PubMed]
29. Bhat PJ, Darunte L, Kareenhalli V, et al. Can metabolic plasticity be a cause for cancer? Warburg-Waddington legacy revisited. Clin Epigenetics 2011;2:113-22.
30. Duraj T, Kalamian M, Zuccoli G, et al. Clinical research framework proposal for ketogenic metabolic therapy in glioblastoma. BMC Med 2024;22:578. [Crossref] [PubMed]
31. Klement RJ, Eckert JM, Sweeney RA. Anti-cancer effects of hyperbaric oxygen therapy in mice: A meta-analysis. Oncologie 2024;26:941-55.
32. Wallace R, Wallace D, Wallace RG. Toward cultural oncology: The evolutionary information dynamics of cancer. Open Syst Inf Dyn 2003;10:159-81.