Senescence of multicellular individuals: imbalance of epigenetic and non-epigenetic information in histone modifications

and non-epigenetic information in histone modifications Felipe A. Veloso ∗ Facultad de Ciencias, Universidad Mayor, Santiago, Chile. (Dated: April 27, 2018) Cellular aging has been progressively elucidated by science. However, aging at the 1 multicellular-individual level is still poorly understood. A recent theory of individuated multicellularity 2 describes the emergence of crucial information content for cell differentiation. This information is 3 mostly conveyed in the non-epigenetic constraints on histone modifications near transcription start 4 sites. According to this theory, the non-epigenetic content emerges at the expense of the information 5 capacity for epigenetic content. However, it is unclear whether this “reassignment” of capacity 6 continues after adulthood. To answer this question, I analyzed publicly available high-throughput 7 data of histone H3 modifications and mRNA abundance in human primary cells. The results show 8 that the “reassignment” continues after adulthood in humans. Based on this evidence, I present a 9 falsifiable theory describing how continued “reassignment” of information capacity creates a growing 10 epigenetic/non-epigenetic information imbalance. According to my theoretical account, this imbalance 11 is the fundamental reason why individuated multicellular organisms senesce. 12


=
C Y (X 1 , . . . , X n , Y ) Epigenetic histone crosstalk (total correlation of X1, . . . , Xn that is explicitly related to Y ), depends on DNA-nucleosome interactions (Ref. 17), and conveys epigenetic information content + C(X 1 , . . . , X n |Y ), Non-epigenetic histone crosstalk (total correlation of X1, . . . , Xn that is explicitly unrelated to Y ), depends on protein/RNA-nucleosome interactions (Ref. 17), and conveys hologenic information content (1) where X 1 , . . . , X n are random variables representing n propagators (symbolized by F → N ) in the theory [17]. 163 These F → N molecules are predicted to be, in a given Cor  Schematic for proof-of-concept hypotheses and computational analysis for testing. Publicly available ChIP-seq (chromatin immunoprecipitation followed by high-throughput DNA sequencing) and RNA-seq (transcriptome high-throughput sequencing) data for human primary cell samples allowed the computation, for each TSS, of position-specific histone H3 modification levels (at every 200bp) and its associated mRNA abundance level (a). After log-transforming these levels and taking into account all TSSs, the TSS-adjacent histone H3 crosstalk (triad-wise crosstalk depicted here) was represented as a total correlation [24] or information capacity in bits, which in turn was decomposed as the sum of two measurable and explicitly unrelated components: one epigenetic (explicitly related to transcriptional changes) and the other non-epigenetic (explicitly unrelated to said changes) (b). Taking into account all samples, the log-ratio of non-epigenetic to epigenetic histone H3 crosstalk magnitude was hypothesized to be positively correlated with cell donor age in normal cells (c, left) and also to be uncorrelated with cell donor age in cancer cells (c, right). The subsequent rejection of the statistical null hypothesis in (c, left) and the failure to reject the statistical null hypothesis in (c, right) provided proof of concept for the theory of senescence proposed in this paper.

RESULTS
To test the two proof-of-concept hypotheses, i.e., C(X i , X j , X k |Y ) and C Y (X i , X j , X k , Y ) for  The log-ratio (base 2) between the non-epigenetic and 241 epigenetic histone H3 crosstalk magnitudes was thus 242 computed as the dimensionless quantity Importantly, total correlation C captures all possible 244 associations in the set of variables {X i , X j , X k } that may 245 exist starting from the pairwise level.

246
The log-ratio of non-epigenetic to epigenetic histone H3 crosstalk magnitude is positively correlated with cell donor age in normal cells      The log-ratio of non-epigenetic to epigenetic histone H3 crosstalk magnitude does not correlate with cell donor age in cancer cells When I analyzed the log-ratio of non-epigenetic to

364
To test this possibility, the correlation value r between 365 cell donor age and total information capacity (in bits) of 366 TSS-adjacent histone H3 crosstalk, computed as was obtained for all 551,300 triads of position-specific 368 histone H3 modifications for normal cells.

369
The analysis revealed that the correlation coefficients r The notable exceptions to be made for the prediction 468 above are a few species able to undergo reverse 469 developmental processes from adult to juvenile stages.

470
One such species is the jellyfish Turritopsis nutricula [30], 471 which is predicted to display an analogous negative 472 correlation in the processes, i.e., "reassignment" in The log-ratio of non-epigenetic to epigenetic histone crosstalk magnitude increases during development as the embryo grows (black f (t) in darker blue area). After the organism reaches its mature form (yellow area), the log-ratio continues to increase (orange f (t))-with a few notable exceptions (blue and magenta f (t)). This continuous increase in turn creates an increasing dysfunctional imbalance of information contents that translates into senescence and, eventually, into death.
extrinsic attempt to correct for the hologenic/epigenetic 547 content imbalance. This problem resides in that hologenic 548 constraints, whose growth in magnitude has senescence 549 as a byproduct, are the very constraints preventing an 550 otherwise likely onset of cancer [17].
where P (x, y) is the joint probability of X=x and Y =y.  P (x, y) log b P (x, y) P (x)P (y) Note that if and only if X and Y are statistically used in this work. The first is interaction information [50] 764 or co-information [51], also symbolized by I, which is 765 defined analogously to Eq. 8 for a set V of n discrete 766 random variables as where |U | is the cardinality (in this case, the number To specifically quantify the magnitude of histone H3 782 crosstalk, the second multivariate generalization of mutual 783 information used in this work was total correlation [24] 784 (symbolized by C) or multiinformation [25], which is 785 defined as (12) i.e., as the sum of the marginal uncertainties of the random 787 variables {X 1 , . . . , X n } minus their joint uncertainty. I(X i ; X j ) + i,j,k I(X i ; X j ; X k ) + . . . + I(X 1 ; . . . ; X n ). (13) This expression for total correlation C as a sum of 797 interaction information quantities I along with the sum 798 decomposition of I in Eq. 11 allows C to be decomposed 799 also as a sum: 800 C(X 1 , . . . , X n ) = C Y (X 1 , . . . , X n , Y )+C(X 1 , . . . , X n |Y ), (14) where C Y (X 1 , . . . , X n , Y ) is the sum (analogous to that 801 of Eq. 13) of all interaction information quantities I but abundance levels (such power has already been used 825 to predict mRNA levels with high accuracy [23]).

826
The uncertainty coefficient U [52] is defined as i.e., U (Y |X 1 , . . . , X n ) is the relative decrease in i.e., a phenomenon known as synergy of a set of 862 predictor variables [53] (see Table 2).
which is known to follow a Student's t-distribution 911 with n−2 degrees of freedom, and where n is the 912 number of data pairs [56]. For the hypothesized 913 positive correlation between the non-epigenetic/epigenetic 914 histone H3 crosstalk log-ratio and age, the statistical null 915 hypothesis was tested against the alternative hypothesis 916 that the correlation is greater than zero (i.e., one-sided 917 Student's t-test). For the hypothesized non-significant 918 correlation between the overall histone H3 crosstalk 919 magnitude and age, the statistical null hypothesis 920 was tested against the alternative hypothesis that the 921 correlation is greater or less than zero (i.e., two-sided 922 Student's t-test).

923
On the other hand, the distribution of correlation 924 coefficients (r) is known to be non-Gaussian [57], which 925 can be easily appreciated in Fig. 3 Table 3. Metadata for each primary cell sample analyzed.
Note: Age entries originally tabulated as 90+ were entered as 90 into the computational analysis. Metadata source: CEEHRC.

ACKNOWLEDGMENTS
I wish to thank Angelika H. Hofmann at SciWri Services for editing this paper into an English I could only hope to write.