A constraints-based theory of the primary cause of senescence: imbalance of epigenetic and non-epigenetic information in histone crosstalk

  1. Felipe A. Veloso

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Cellular aging has been progressively elucidated by science. However, the fundamental cause of senescence—i.e., why organisms age at the multicellular-individual level—remains unclear. A recent theory of individuated multicellularity describes the emergence and growth of crucial information content for cell differentiation. This information is mostly conveyed in the non-epigenetic (i.e., transcription uncorrelated) histone crosstalk near transcription start sites. According to this theory, the non-epigenetic content emerges and grows at the expense of the information capacity for epigenetic content. If this “reassignment” of information capacity continues after adulthood, it may explain the senescence phenomenon. Here, I present a novel, falsifiable theory describing an uninterrupted growth of capacity for non-epigenetic information at the expense of that for epigenetic information not only during ontogeny but also throughout adulthood. As a byproduct, this continuous “reassignment” of capacity effectively creates an information imbalance in histone crosstalk, which in turn overregulates transcriptional levels. This overregulation is to be understood as transcriptional levels becoming more and more accurate but also less and less precise with respect to the needs of the multicellular individual—up to the point of dysfunctionality. This epigenetic/non-epigenetic information imbalance is proposed to be the primary reason why individuated multicellular organisms senesce.

Article activity feed

  1. Version published to 10.1101/310300v3 on bioRxiv
  2. PeerRef

    Discussion, revision and decision

    Decision - Verified with reservations

    Charles A. Schumpert: Verified manuscript

    Max Shokhirev: Verified with reservations

    Author response and revisions

    Author response to reviewer Max Shokhirev

    Dear Dr Shokhirev,

    Thank you so very much for having reviewed my paper; your comments have been very helpful for me to improve it. Please find attached two identical copies of my revision of the paper (changes are highlighted in one of the copies) and a supplementary file that relates to one of your comments. Please see below my response to your comments when applicable.

    Does the work cite relevant and sufficient literature? Some, but seems to be very limited in terms of biological literature.

    I recognize and acknowledge this issue. It will be addressed in some of the specific responses below.

    Are the conclusions adequately supported by the results? No

    I understand this question and its answer as the theory presented in the paper (the conclusions) not providing new, proof-of-principle evidence (the results). That is correct, the theory is based only on existing evidence (most importantly, the already described age-dependent "transcriptional noise"), it is consistent with it and, importantly, it provides (in the new revision) three additional, experimentally testable predictions.

    The author has laid out a theoretical argument for senescence as a tradeoff between information capacity between epigenetic and non-epigenetic content.“A constraints-based theory of senescence: imbalance of epigenetic and non-epigenetic information in histone crosstalk.” This work is interesting, but is based on a superficial understanding of the biology underlying senescence/aging, makes several dangerous oversimplifications and assumptions, and does not provide any data or analysis to support the theory.

    I would argue the theory is not based on a superficial understanding of the biology of senescence as it is currently established—in fact it is not based on the current biology of senescence at all. The theory is actually based on my previous theoretical work where I show (supported by real data analysis) how the constraints on transcription start site-adjacent histone crosstalk that are explicitly uncorrelated with transcriptional levels associate strongly with cell differentiation states—even more strongly than those constraints correlated with transcriptional levels (an association which is expected and already described in previous work). This in turn suggests the existence of an additional, higher-order type of biologically meaningful information conveyed by histone crosstalk, with this information is by definition uncorrelated with the information for precise epigenetic control of transcriptional levels. I called this information "hologenic" because it associates to cell differentiation trajectories necessary for the development of the multicellular individual as as a whole while being explicitly uncorrelated with the transcriptional levels. My previous work showed that the capacity for hologenic information in histone crosstalk grows at the expense of that for epigenetic information (this is necessary for the development of the multicellular individual). I emphasize that the theory proposed here relies on only two assumptions: (i) the overall histone crosstalk remains statistically constant in magnitude throughout adulthood and (ii) the hologenic component of the overall histone crosstalk increases at the expense of the epigenetic component throughout adulthood (with the exceptions depicted in Fig. 1b). These assumptions are of course not to be taken for granted, so in the new revision of the paper they are presented as predictions that can falsify the theory. I will further elaborate on these assumptions in my specific responses below.

    All the above being said, the version of the paper you reviewed apparently gives the wrong impression that the theory is a comprehensive description of the senescence process with all its complexity. It is not. It is not surprising then that the theory appears to oversimplify the explanatory power of mechanisms known to be part of the senescence process when in reality they are claimed—this is a big theoretical claim I am making—to be the consequence of the primary cause of senescence. It is only this primary cause what the proposed theory is all about. For these reasons and thanks to your observation, I stated this distinction explicitly in the new subsection 1.2 "Scope" and also modified the title of the paper accordingly.

    I recognize and acknowledge that the paper does not present new data or experimental results in direct support of the theory—something that arguably makes it less compelling to be considered let alone to be tested experimentally. But I do wish to point out that a scientific theory, must "only" (i) effectively explain the phenomena it is aimed to explain (in this case, the primary cause of senescence), (ii) be consistent with existing observations/results (most importantly, with the well described age-dependent "transcriptional noise" in this case), and (iii) provide non-trivial, experimentally testable predictions that can falsify it. I have tried to make up for the lack of preliminary supporting evidence by adding three new straightforward, experimentally testable predictions that can falsify it. Importantly, two of said predictions (C and D in subsection 2.6) relate to precisely how senescence can be slowed down, even stopped, or accelerated—and with associated effects in terms of resistance/propensity to carcinogenesis—in any non-human species, because the direct testing requires genome editing. In this context, I have no problem granting that the prior probability of predictions C and D being verified experimentally is exceedingly small. On the other hand, the experimental verification of predictions C and D would arguably be a game-changing result in terms of the fundamental understanding of the senescence process, however unlikely this scenario is a priori. In this context, I submit to you that the extraordinary nature of predictions C and D in both theoretical and practical terms (I reiterate, this is not to say predictions C and D will be verified) more than makes up for the lack of preliminary evidence for the theory. The prospect of slowing down or even stopping the senescence process may catch the attention of at least some research groups with genome-editing capabilities, given that the gene edits described in predictions C and D are very specific. (You can find more about predictions C and D in my response to your comments related to the subsection 2.4 of the paper you reviewed).

    Sections 1.1-1.3 The author only mentions the Hayflick limit as a biological reference for senescence. There is a very rich body of literature on senescence and aging that is completely overlooked here. The author should include additional references to reviews for senescence and aging to orient the reader to the complexity of these biological processes (e.g. PMC8658264, PMC7846274).

    Thank you for pointing this out. I have included the suggested review articles as references and, most importantly, I tried now to clarify the scope of the theory in a new subsection (1.2 Scope). The scope of the theory is in a sense, very limited: it is indented to explain only the beginning of the causal chain of age-related changes we identify as aging at the multicellular-individual level. In other words, it is about what fundamentally triggers the process. On the other hand, such a scope (i.e., describing the first cause) is quite ambitious in the sense that it should allow, at least in principle, for the manipulation of the process (either slowing it down or accelerating it). I will come back to this point later in my response.

    Please clarify what you mean by senescence vs aging for both cells and individuals. Senescence is a natural biological process that cells/organisms use to turn off cell replication due to damage (e.g. telomere shortening, double-stranded breaks, etc.). Other cells can also facilitate this process through signaling (e.g. immune cells or contact inhibition).

    I am not a native English speaker, and one the first things I did when addressing this problem was to study the associated terminology in the literature. Unfortunately, this terminology is not particularly monolithic. In some articles, the age-dependent, progressive dysfunction undergone by multicellular individuals once they reach their mature form is referred to as "senescence" (PMID 1677205, 6776406, 12940353, 22884974, among others), "biological aging" (PMID 31833194, 33982659, 34700008, among others), or even simply as "aging" (PMID 24862019, 34990845, 31173843, among others). In this context, I decided to stick to "senescence" mainly because (i) it is only one word and (ii) unlike "aging", "senescence" directly and unambiguously implies time-dependent dysfunction or decay. At any rate, to distinguish the term from cellular senescence/cellular aging I created a Glossary in the paper where these terms and others are clarified to avoid confusion.

    Aging is typically thought of as an organismal phenomenon, which is still poorly understood but is theorized to include tradeoffs (as you describe in section 1.2). It is also accepted that aging is cell, tissue, and organism specific. Since you talk about senescence and aging across both biological scales, it is important to define exactly what your theory pertains to.

    I am glad we agree that senescence is poorly understood (especially in comparison to cellular senescence). Unfortunately, some colleagues in the community interpret this as saying the research been done on the topic is worthless—it is not—when it is really pointing out the phenomenon largely lacks falsifiable theories, let alone an already tested falsifiable theory (with experiments failing to falsify it).

    Section 1.4

    The author posits that senescence is an imbalance in information contents of histone post-translational modifications around transcription start sites. This is just one level of regulation, albeit an important one. The author seems to completely overlook many other types of regulation (e.g. microRNA, lincRNAs, metabolic/energetic constraints, non-proximal regulation at enhancers, higher ordered structure of the chromatin, post-translational regulation of proteins, and etc.). How can all of these other important levels of regulation fit into this theory? All have been implicated in senescence/aging in some form or another.

    What you point out here is very important, thank you. In a remarkable piece of research, Kumar and colleagues showed that core nucleosomal histone post-translational modification (hPTM) profiles are able to predict transcript abundance levels with very high accuracy (R~0.9, ref. 33). The constraints on hPTMs underpinning this predictive power (in turn underpinned by DNA-histone octamer interactions)—as well as those constraints on hPTMs that are explicitly uncorrelated to transcriptional levels—are central to the theory proposed. As stated in the new section 1.2, the complex cascade of changes/interactions characterizing senescence escapes the scope of the theory. In this context, most of the types of regulation you mention are under this theory not actually regulation in a "teleological" (the quotes are meant to avoid alienating the reader) sense but rather types of propagation/amplification of truly regulated/dysregulated changes. One of my goals when developing this theory was to try shift attention from "molecule A-collides with molecule-B, which collides to..." into higher-order constraints and trying to explain phenomena such as the well-known, age-dependent "transcriptional noise", which under the theory presented should be understood as senescence itself. Furthermore, I maintain the relative slow progress we have made in understanding phenomena such as cancer and senescence (in spite of the abundance of high-throughput data) comes from

    The author further suggests that histone crosstalk information content can be decomposed into two unrelated components: epigenetic and non-epigenetic. The non-epigenetic component is described as “hologenic information content,” which stems from a previously published work by the author. Non-epigenetic is confusing in this context since really this is information content that stems from the synergies of individual cells to form a whole, e.g. the emergent information content that comes from many cells working together (or at least this is how I understand the underlying theory). This information content is important for the general maintenance and survival of the organism. The author should clarify this point further, since this seems to be one of the fundamental assertions being made in the paper. For example, bringing in the descriptions used in section 2, can further clarify these central points.

    In my previous work, "hologenic" information content is defined as being uncorrelated with (i.e., orthogonal to) changes in transcriptional levels, in the same way "epigenetic" information has been defined (traditionally and for good reason) as being uncorrelated with changes in the DNA reason. Hologenic information content emerges when proliferation-generated extracellular gradients of secreted molecules start to being used to perform regulatory work (after being transduced) on the histone crosstalk of each cell's nucleus.

    In addition, the author states: “ Moreover, the sum decomposition in Eq. 1 implies that the growth in magnitude (bits) of the hologenic (i.e., non-epigenetic) component must be accompanied by a decrease in magnitude of the epigenetic component.” This is not necessarily true, since signaling is a separate biological process from the regulation of gene expression. In other words, both can increase or decrease simultaneously. For example, a healthy non-senescent immune cell can upregulate very specific transcriptional programs that lead to very complex signaling and extra-cellular interactions. You can argue that both represent an increase in information content for both the epigenetic and non-epigenetic “hologenic” components. In addition, as cells naturally senesce they are programmed to turn off cell-cycling while upregulating autophagy and repair processes. They may not upregulate extracellular signaling at this time, which would seem to contradict the author’s theory/statement. In this case, the simplification that all cells are the same is dangerous because it overlooks the tradeoff of information contents between cells. It also ignores important repair pathways (senescence being one of them), to deal with cells that have dysregulated their natural processes over time. It also overlooks the important action of immune cells that work to get rid of cancer and poorly-functioning cells.

    This comment of yours (referring to complex yet specific signaling pathways and interactions) clearly shows I did a poor job (if not utterly failed) in conveying that the epigenetic and hologenic components must be understood in chromatin-wide terms. Yes, the random variables used to define both components in the sum decomposition of Eq.1 are defined with respect to a single, generic transcription start site, but these random variables take their respective values from data for all transcription start sites in the nucleus. This is why the terms Eq. 1 and the log-ratio in Eq. 2 must be understood as chromatin-wide terms. Again, this approach intends to shift attention from specific (however important) molecular mechanism to higher-order, information conveying hologenic/epigenetic constraints (whose imbalance are proposed to trigger senescence as proposed in this theory). The chromatin-wide nature of the hologenic and epigenetic components is not an obvious consideration but it is a very important one, so it is now explicit (twice) in the revised text and I thank you for bringing this to my attention.

    For a statistically invariant level of overall histone crosstalk C(X1,...,Xn) in Eq. 1, a growth in magnitude of the hologenic (i.e., non-epigenetic) component must be accompanied by a decrease in magnitude of the epigenetic component and vice versa. This chromatin-wide trade-off might not hold, as you suggest, only if the overall histone crosstalk C(X1,...,Xn) varies significantly (in particular, if it varies significantly throughout adulthood). If, in fact, the overall histone crosstalk C(X1,...,Xn) varied significantly throughout adulthood the proposed theory would make no sense whatsoever. Mathematically, C(X1,...,Xn) is finite and upper bounded by ΣH(Xi) - max H(Xi) (where H(Xi) is the marginal Shannon uncertainty of Xi), and one can further expect that the overall histone crosstalk represented nu C(X1,...,Xn) remains statistically invariant for a number of reason, chief among them the massive chromatin instability that would ensue if it indeed C(X1,...,Xn) varied. For this reason, I included the C(X1,...,Xn) time-invariance as an additional prediction for the falsifiability of the theory.

    Also, it seems crosstalk, correlation, capacity, and content, are used interchangeably. Please clarify that these are all the same, or use one of these terms to avoid confusion.

    I went through the use of these terms in the paper and, unfortunately, I cannot reduce them to just one term or dispense with them altogether without losing rigor for the theory. Because of this I decided to include them in the Glossary, hoping that it will make any reader recognize that their respective uses in the text are actually not interchangeable.

    Section 1.5 The author provides a general approach for measuring the log of the ratio of epigenetic and non-epigenetic capacities for a particular histone modification at three positions (i,j,k), and for some measured abundance of mRNA Y. Since we typically measure abundance of a particular modification genome-wide, and the mRNA level for tens of thousands of genes, how would a realistic equation look like (i.e. one that has 10k mRNA levels, and 10k histone positions)? In addition, the author does not explain how to combine correlations across multiple histone modifications. Please expand this section to make it relevant for real-world genome-wide measurements since this will be important for falsifying the theory.

    Since public datasets are available (e.g. the aging atlas https://doi.org/10.1093/nar/gkaa894), the author should show an example of how a dataset might be used to falsify or demonstrate the theory in more detail.

    This response to this observation can be found in the new Fig. 2 and with greater detail in Supplementary File 1. In summary, the data analysis approach is basically the same used by Kumar et al. (ref. 33). That is, with tandem RNA-seq/ChIP-seq data obtained from the same cell sample a table can be constructed with the rows representing the transcription start sites (TSSs) in the genome and the columns displaying the normalized ChIP-seq signal for each hPTM in different positions relative to the respective TSS (variables X1,...,Xn in the paper) plus a column with the respective measured transcript abundance for each TSS (variable Y).

    Section 2.1 The author uses correlation of the log ratio of the epigenetic and non-epigenetic content with age as a readout of “reassignment” of crosstalk/contents, arguing that for cancer cells this correlation should be essentially zero. This seems like an oversimplification of the “reassignment” process since senescence may occur in phases across the age of a cell/organism, and since there might be both increases and decreases in the log ratio of contents due to natural biological processes and variability. Would it not be better to measure the sum of changes in the log ratio or the difference between the log ratios at different ages?

    In addition, the biological age of a cell/tissue/organism can vary. For example, stem cells may have negligent aging, while other cells might age relatively quickly. Again, the author should clarify the context of age: are we measuring strictly chronological age correlation? Should we consider different correlations for each cell/tissue in the organism? What about tradeoffs in information content between cell types and tissues? In other words, it is unclear how the theory should be applied to biological systems.

    This is a great observation, thank you. Yes, the correlation in Eq.3 relates strictly to chronological age (in other words, to time). The correlation must hold for somatic cells of the same type according to the theory; now this condition is explicit in the text. In this context, some cell types and tissue may senesce (see new Fig 1c, center) faster than others as you point out; in this case the associated slope is predicted to be steeper, whereas cell types that senesce relatively slower the slope should be gentler. Only in species displaying negligible senescence the hologenic/epigenetic log-ratio should remain constant (i.e., zero slope, as depicted in Fig. 1b, blue curve), or fluctuate significantly in species displaying "reversible" development (Fig. 1b, magenta curve).

    Section 2.2 The author argues that senescence is an emergent property of the loss of information content for epigenetic histone crosstalk and an increase in information content of “hologenic” information content (e.g. cell signaling and anti-tumor signaling). I believe this premise does not stem from the reality of biological systems (see my comments for section 1.4).

    The trade-off between capacity for hologenic and epigenetic information within a constant overall histone crosstalk magnitude—in particular, the growth of the former at the expense of the latter throughout adulthood generating a dysfunctional imbalance—is arguably the cornerstone of the theory in fundamental terms. Whatever my response was to your comments about subsection 1.4, this crucial trade-off cannot be taken for granted, however compelling the arguments are. In this context, there is no better solution than putting the hologenic/epigenetic trade-off to the test (see prediction B for falsifiability of the theory, also further detailed in the new revision of the paper). Realistically, however, I expect prediction B to be tested (and the hologenic/epigenetic log-ratio quite thoroughly examined) only if predictions C and D are verified. In that scenario, it will be interesting to see whether the hologenic/epigenetic log-ratio increase may be steeper in some tissues (which should then explain why those tissues senesce faster than others).

    Also, this section seems to be contradicting the author’s conclusions and is very confusing. The author seems to argue that there is both more AND less constraint at the multi-cellular level (organismal)? Please clarify or remove this section.

    I can see now how it seems contradictory because I was saying the capacity for hologenic information (which is about transcriptional levels being accurate for the multicellular individual as a whole) increases up to the point of being dysfunctional at the multicellular-individual level. Here I failed to convey that said dysfunctional outcome derives from the concurrent decrease of capacity for epigenetic information, not from the increase of capacity for hologenic content per se). Thank you for bringing this to my attention. I decided to remove this subsection altogether because the hologenic/epigenetic trade-off is covered in greater detail in the next subsection.

    Section 2.3

    Senescence as transcriptional overregulation is vague. Here the author is arguing that as epigenetic constraint decreases, you have a decrease in precision (e.g. loss of regulation), but then you have a competing global or hologenic increase in constraints, which constrains the expression of genes for the overall benefit of the organism. A shift toward global constraint.

    My intention here is to establish a fundamental contrast between the group of diseases we call cancer and senescence using the difference between the concepts of accuracy (average closses of the actual values to a target value) and precision (variance of the actual values, also added to the Glossary). In this context, since cancer is an almost canonical example of gene dysregulation (transcriptional accuracy is lost because capacity for hologenic information is lost), we can understand senescence as transcriptional overregulation in the sense of too much capacity for hologenic information gained over time at the expense of capacity for epigenetic information, thereby losing transcriptional precision). I added a third panel "c" to Fig. 1 to clarify the proposed contrast between cancer and senescence, in terms of impaired transcriptional regulation (i.e., inaccurate up to dysfunction in cancer and imprecise up to dysfunction in senescence).

    Section 2.4 This seems to be describing an illustrative real-world example? This section is incredibly specific and again only focuses on one possible mechanism and does not include any measured data or analysis. Please preface this section to explain that this is just one of many possible examples. Again, it will be good to provide other examples looking at other aspects of aging biology (not just histone modifications).

    I am not including examples of mechanisms propagating dysfunctional changes in the senescence process because this theory is not about adding one more possible mechanism to the collection of well-described mechanisms associated to senescence. The theoretical claim I am making here is that the sequence of steps described in subsection 2.4 constitute the primary cause that triggers senescence throughout the multicellular individual's adulthood. This is of course an extraordinary theoretical claim and, as the great Carl Sagan pointed out, as such requires extraordinary evidence (or, in this theoretical paper, an extraordinary prediction aimed to obtain said evidence).

    This is why I also added two predictions, C and D, to the now completely rewritten subsection 2.6 "Falsifiability". These predictions are extraordinary in that they provide extremely specific sufficient conditions for either slowing down or accelerating the senescence process in any non-human species. For this reason (i.e., adding predictions C and D) I have duly updated the "Competing interests" section of the paper. [Note: in a separate paper I explore in greater depth the theoretical underpinnings of predictions C and D.]

    Section 2.5-2.9,3

    This seems to be a general discussion. It would be easier to organize these sections into one discussion section for added clarity. Again, I would recommend not talking about sweeping statements like “Senesensce’s ultimate cause” and “Can senescence be stopped?” since this theory only addresses one small aspect of the biology underlying aging and senescence and does not address the heterogeneity of aging. These topics are controversial and should be addressed very carefully to avoid alienating the biological community.

    Thank you for your suggestion, I organized a Discussion section accordingly, placing the predictions for falsifiability in it. Additionally, I changed “Senescence’s ultimate cause” for “Senescence’s proposed ultimate cause”, clarifying also in the Scope subsection 1.2 that ultimate/proximate is used only in terms of the concepts of causality as introduced and named by Ernst Mayr.

    Following your advice, I removed the subsection “Can senescence be stopped?” Now, I believe this theory is bound to alienate at least some colleagues in the biological community (among those who read the paper, that is). As you point out, the theory addresses only one small aspect of the senescence process, but it is not any small aspect. It is about what triggers the process or, equivalently, what is the initial link of the causal chain leading to senescence with all its mechanistic complexity. This implies all other proposed primary causes would, simply by logical exclusion, be incorrect.

    One of the reasons I developed this theory in the first place is that I failed to find even a single falsifiable description proposing a primary cause of senescence that integrates Mayr's proximate and ultimate concepts of causality. To be clear, this is not a shortage of explanatory accounts for the senescence process; it is a shortage of experimentally falsifiable ones. There are scientific fields where falsifiability is inherently difficult or probably impossible to meet, such as paleoclimatology. But that is not the case in biology—certainly not in developmental biology. This is why I agree with you in that the senescence process is still poorly understood in fundamental terms (which by no means implies all work in the field is worthless).

    Forgive me for the following related digression/personal note: Since I started writing this paper I faced the dilemma of how many research works on this topic should I cite, given there is plenty (even considering this is not a review-type article). We scientists are only human and as such we very much like our work being cited in terms of current knowledge. That changes, however, when the explanatory account we like the most (or even worse, the explanatory account we ourselves proposed) is being cited to acknowledge its existence but the lack of predictions to falsify the "theory" is also pointed out (our reaction changes probably because unfalsifiable explanatory accounts can be always dismissed as just storytelling, however compelling the story might be). These rather petty emotional responses of ours have of course nothing to do with the advancement of science. Yet, many scientific journals—however prestigious, and particularly in the life sciences—routinely publish articles the word "theory" describing the work in the title and/or body when there is no prediction to be found for falsifiability. As a student, this is when I learned the interests of the scientific publishing industry (as we know it) and those of science are completely uncorrelated—not really surprising since their respective goals are so different. In this context, I found the PeerRef initiative/model very interesting since it aims to focus purely on the scientific content as opposed to essentially non-scientific considerations.] In the previous revision of the paper I offered only one prediction to falsify the theory, which is arguably cumbersome—and therefore arguably not too appealing—for experimentalist colleagues to test. For that reason I added three predictions (A, C, and D) which are straightforward to test in the laboratory. In this context, I am hoping the scientific community will be open to consider and to test falsifiable theories, however alienating they might be. Especially when we are dealing with a phenomenon for which paradigmatically accepted explanatory accounts is, at least to the best of my knowledge" all we currently have. Last but certainly not least, I wish to express again my gratitude for all the time you devoted to review my paper. Whether or not the theory it presents resists falsification attempts, I firmly believe the paper itself is now better than it was before thanks to your feedback.

    Reviewer response

    In general, it seems the paper is improved and includes quite a bit of additional clarification and qualifying statements.

    The scope is also much more constrained and it is clear that the author is sticking only to proposing a possible theory, which is fine with me, albeit not as exciting as it might have been if the author had gone through and provided real-world examples or evidence. It still bothers me that the work is trying to implicate a very specific mechanism for senescence (chromatic cross-talk and regulation), since now it is clear that the work is mostly a theoretical exercise without much basis in established biology. In other words, chromatin cross talk might be an example of one way that this could happen among others (e.g. other epigenetic regulation or lack thereof). As a theoretical work it is fine as long as you agree that it is sufficient for the scope of the journal

    Decision changed - Verified with reservations: The content is scientifically sound, but has shortcomings that could be improved by further studies and/or minor revisions.

    Author response to reviewer Charles A. Schumpert

    Dear Dr Schumpert,

    I wish to thank you for having reviewed my paper; your comments have been very encouraging for the effort of moving my theory forward. Please find attached two identical copies of my revision of the paper (changes are highlighted in one of the copies) and a supplementary file that relates to one of your comments. Please see below my response to your comments when applicable.

    Overall the manuscript is written brilliantly and provides excellent context to the audience about a complex theoretical biological concept. No flaws can be found, although one could argue against a few of the points in the assumptions used to construct the theory, there’s nothing illogical or irrational.

    Thank you for kind words, but I have to admit any and all brilliance that may be found in the write-up must be credited to my wonderful editor Angelika Hofmann. Regarding the assumptions, this theory relies on two critical ones, which of course cannot be taken at face value. This is why these assumptions inform prediction B (in the new revision). There are now four experimentally testable predictions to falsify the theory; hopefully some will be appealing to experimentalist colleagues.

    In your opinion how could the author improve the study? The writing of the paper makes it easy to read, which can sometimes be a challenge with theoretical biology manuscripts. Potentially adding a bit more context on the various theories of aging may help demonstrate the marriage of the ideas into the theory he constructed.

    One of the problems I faced when writing this paper was how many different explanatory accounts (there is plenty) I was going to cite provided I would have to underscore their lack of testable predictions for falsifying them—which in time becomes dangerously close of being downright unfalsifiable. As Wolfgang Pauli famously said, unfalsifiable theories are not even wrong. In other words, the problem was how many readers I was going to alienate while having no intention to do so. Navigating academia's social ocean is not an easy task, at least not to me, so I decided to compromise. To be clear, I am by no means claiming my theory is correct but underscoring that (i) it can be tested experimentally and (ii) explains the known age-dependent, cell-to-cell transcriptional noise that I argue should be regarded as senescence itself in fundamental terms.

  3. PeerRef

    Peer review report

    Reviewer: Max Shokhirev Institution: Salk Institute for Biological Studies email: maxshok@gmail.com

    Section 1 – Serious concerns

    • Do you have any serious concerns about the manuscript such as fraud, plagiarism, unethical or unsafe practices? No
    • Have authors’ provided the necessary ethics approval (from authors’ institution or an ethics committee)? not applicable

    Section 2 – Language quality

    • How would you rate the English language quality? High quality

    Section 3 – validity and reproducibility

    • Does the work cite relevant and sufficient literature? Some, but seems to be very limited in terms of biological literature.
    • Is the study design appropriate and are the methods used valid? Yes
    • Are the methods documented and analysis provided so that the study can be replicated? not applicable
    • Is the source data that underlies the result available so that the study can be replicated? Yes
    • Is the statistical analysis and its interpretation appropriate? Yes
    • Is quality of the figures and tables satisfactory? Yes
    • Are the conclusions adequately supported by the results? Yes
    • Are there any objective errors or fundamental flaws that make the research invalid? How could the author improve the study?

    The author has laid out a theoretical argument for senescence as a tradeoff between information capacity between epigenetic and non-epigenetic content.“A constraints-based theory of senescence: imbalance of epigenetic and non-epigenetic information in histone crosstalk.” This work is interesting, but is based on a superficial understanding of the biology underlying senescence/aging, makes several dangerous oversimplifications and assumptions, and does not provide any data or analysis to support the theory. I’ve laid out my comments for each section below:

    Sections 1.1-1.3

    The author only mentions the Hayflick limit as a biological reference for senescence. There is a very rich body of literature on senescence and aging that is completely overlooked here. The author should include additional references to reviews for senescence and aging to orient the reader to the complexity of these biological processes (e.g. PMC8658264, PMC7846274). Please clarify what you mean by senescence vs aging for both cells and individuals. Senescence is a natural biological process that cells/organisms use to turn off cell replication due to damage (e.g. telomere shortening, double-stranded breaks, etc.). Other cells can also facilitate this process through signaling (e.g. immune cells or contact inhibition). Aging is typically thought of as an organismal phenomenon, which is still poorly understood but is theorized to include tradeoffs (as you describe in section 1.2). It is also accepted that aging is cell, tissue, and organism specific. Since you talk about senescence and aging across both biological scales, it is important to define exactly what your theory pertains to.

    Section 1.4

    The author posits that senescence is an imbalance in information contents of histone post-translational modifications around transcription start sites. This is just one level of regulation, albeit an important one. The author seems to completely overlook many other types of regulation (e.g. microRNA, lincRNAs, metabolic/energetic constraints, non-proximal regulation at enhancers, higher ordered structure of the chromatin, post-translational regulation of proteins, and etc.). How can all of these other important levels of regulation fit into this theory? All have been implicated in senescence/aging in some form or another. The author further suggests that histone crosstalk information content can be decomposed into two unrelated components: epigenetic and non-epigenetic. The non-epigenetic component is described as “hologenic information content,” which stems from a previously published work by the author. Non-epigenetic is confusing in this context since really this is information content that stems from the synergies of individual cells to form a whole, e.g. the emergent information content that comes from many cells working together (or at least this is how I understand the underlying theory). This information content is important for the general maintenance and survival of the organism. The author should clarify this point further, since this seems to be one of the fundamental assertions being made in the paper. For example, bringing in the descriptions used in section 2, can further clarify these central points. In addition, the author states: “ Moreover, the sum decomposition in Eq. 1 implies that the growth in magnitude (bits) of the hologenic (i.e., non-epigenetic) component must be accompanied by a decrease in magnitude of the epigenetic component.” This is not necessarily true, since signaling is a separate biological process from the regulation of gene expression. In other words, both can increase or decrease simultaneously. For example, a healthy non-senescent immune cell can upregulate very specific transcriptional programs that lead to very complex signaling and extra-cellular interactions. You can argue that both represent an increase in information content for both the epigenetic and non-epigenetic “hologenic” components. In addition, as cells naturally senesce they are programmed to turn off cell-cycling while upregulating autophagy and repair processes. They may not upregulate extracellular signaling at this time, which would seem to contradict the author’s theory/statement. In this case, the simplification that all cells are the same is dangerous because it overlooks the tradeoff of information contents between cells. It also ignores important repair pathways (senescence being one of them), to deal with cells that have dysregulated their natural processes over time. It also overlooks the important action of immune cells that work to get rid of cancer and poorly-functioning cells. Also, it seems crosstalk, correlation, capacity, and content, are used interchangeably. Please clarify that these are all the same, or use one of these terms to avoid confusion.

    Section 1.5

    The author provides a general approach for measuring the log of the ratio of epigenetic and non-epigenetic capacities for a particular histone modification at three positions (i,j,k), and for some measured abundance of mRNA Y. Since we typically measure abundance of a particular modification genome-wide, and the mRNA level for tens of thousands of genes, how would a realistic equation look like (i.e. one that has 10k mRNA levels, and 10k histone positions)? In addition, the author does not explain how to combine correlations across multiple histone modifications. Please expand this section to make it relevant for real-world genome-wide measurements since this will be important for falsifying the theory. Since public datasets are available (e.g. the aging atlas https://doi.org/10.1093/nar/gkaa894), the author should show an example of how a dataset might be used to falsify or demonstrate the theory in more detail.

    Section 2.1

    The author uses correlation of the log ratio of the epigenetic and non-epigenetic content with age as a readout of “reassignment” of crosstalk/contents, arguing that for cancer cells this correlation should be essentially zero. This seems like an oversimplification of the “reassignment” process since senescence may occur in phases across the age of a cell/organism, and since there might be both increases and decreases in the log ratio of contents due to natural biological processes and variability. Would it not be better to measure the sum of changes in the log ratio or the difference between the log ratios at different ages?

    In addition, the biological age of a cell/tissue/organism can vary. For example, stem cells may have negligent aging, while other cells might age relatively quickly. Again, the author should clarify the context of age: are we measuring strictly chronological age correlation? Should we consider different correlations for each cell/tissue in the organism? What about tradeoffs in information content between cell types and tissues? In other words, it is unclear how the theory should be applied to biological systems.

    Section 2.2

    The author argues that senescence is an emergent property of the loss of information content for epigenetic histone crosstalk and an increase in information content of “hologenic” information content (e.g. cell signaling and anti-tumor signaling). I believe this premise does not stem from the reality of biological systems (see my comments for section 1.4). Also, this section seems to be contradicting the author’s conclusions and is very confusing. The author seems to argue that there is both more AND less constraint at the multi-cellular level (organismal)? Please clarify or remove this section.

    Section 2.3

    Senescence as transcriptional overregulation is vague. Here the author is arguing that as epigenetic constraint decreases, you have a decrease in precision (e.g. loss of regulation), but then you have a competing global or hologenic increase in constraints, which constrains the expression of genes for the overall benefit of the organism. A shift toward global constraint.

    Section 2.4

    This seems to be describing an illustrative real-world example? This section is incredibly specific and again only focuses on one possible mechanism and does not include any measured data or analysis. Please preface this section to explain that this is just one of many possible examples. Again, it will be good to provide other examples looking at other aspects of aging biology (not just histone modifications).

    Section 2.5-2.9,3

    This seems to be a general discussion. It would be easier to organize these sections into one discussion section for added clarity. Again, I would recommend not talking about sweeping statements like “Senesensce’s ultimate cause” and “Can senescence be stopped?” since this theory only addresses one small aspect of the biology underlying aging and senescence and does not address the heterogeneity of aging. These topics are controversial and should be addressed very carefully to avoid alienating the biological community.

    Section 4 – Decision

    Revisions required

  4. PeerRef

    Peer review report

    Reviewer: Charles A. Schumpert Institution: University of South Carolina email: schumpca@email.sc.edu

    Section 1 – Serious concerns

    • Do you have any serious concerns about the manuscript such as fraud, plagiarism, unethical or unsafe practices? No
    • Have authors’ provided the necessary ethics approval (from authors’ institution or an ethics committee)? not applicable

    Section 2 – Language quality

    • How would you rate the English language quality? High quality

    Section 3 – validity and reproducibility

    • Does the work cite relevant and sufficient literature? Yes
    • Is the study design appropriate and are the methods used valid? Yes
    • Are the methods documented and analysis provided so that the study can be replicated? not applicable
    • Is the source data that underlies the result available so that the study can be replicated? Yes
    • Is the statistical analysis and its interpretation appropriate? Yes
    • Is quality of the figures and tables satisfactory? Yes
    • Are the conclusions adequately supported by the results? Yes
    • Are there any objective errors or fundamental flaws that make the research invalid? Please describe these thoroughly.

    Overall the manuscript is written brilliantly and provides excellent context to the audience about a complex theoretical biological concept. No flaws can be found, although one could argue against a few of the points in the assumptions used to construct the theory, there’s nothing illogical or irrational.

    Section 4 – Suggestions

    • In your opinion how could the author improve the study?

    The writing of the paper makes it easy to read, which can sometimes be a challenge with theoretical biology manuscripts. Potentially adding a bit more context on the various theories of aging may help demonstrate the marriage of the ideas into the theory he constructed.

    • Do you have any other feedback or comments for the Author? No

    Section 5 – Decision

    Verified manuscript

  5. Version published to 10.1101/310300v2 on bioRxiv
  6. Version published to 10.1101/310300v1 on bioRxiv