Available via license: CC BY 4.0
Content may be subject to copyright.
Accelerated growth increases the somatic
epimutation rate in trees
Frank Johannes
Technical University of Munich
Ming Zhou
Technical University of Munich https://orcid.org/0009-0005-0564-976X
Gerhard Schmied
Technical University of Munich https://orcid.org/0000-0003-2424-7705
Binh Thanh
Technical University of Munich
Monika Bradatsch
Technical University of Munich
Giulia Resente
University of Torino
Rashmi Hazarika
Technical University of Munich
Ioanna Kakoulidou
Technical University of Munich
Maria Costa
Technical University of Munich
Michele Serra
Technical University of Munich
Richard Peters
Technical University of Munich
Enno Uhl
Bavarian State Institute of Forestry (LWF)
Robert Schmitz
University of Georgia https://orcid.org/0000-0001-7538-6663
Torben Hilmers
Technical University of Munich https://orcid.org/0000-0002-4982-8867
Astor Caicoya
Technical University of Munich
Alan Crivellaro
University of Turin https://orcid.org/0000-0002-1307-3239
Hans Pretzsch
https://orcid.org/0000-0002-4958-1868
Article
Keywords: Fagus sylvatica, DNA methylation, epimutation rate, somatic epigenetic drift, somatic
evolution, somatic epimutation, tree epigenomics
Posted Date: March 7th, 2025
DOI: https://doi.org/10.21203/rs.3.rs-5925337/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: There is NO Competing Interest.
1
Accelerated growth increases the somatic epimutation rate in trees 1
2
Zhou M1#, Schmied G2#, Thanh Vo B1, Bradatsch M2, Resente G3,4, Hazarika R1, Kakoulidou I1, 3
Costa M1, Serra M1, Peters RL2, Uhl E5, Schmitz RJ6, Hilmers T2, Toraño Caicoya A2, Crivellaro 4
A3, Pretzsch H2*, Johannes F1* 5
6
7
1 Plant Epigenomics, Technical University of Munich, Freising, Germany 8
2 Tree Growth & Wood Physiology, Technical University of Munich, Freising, Germany 9
3 Department of Agricultural, Forest and Food Sciences, University of Torino, Grugliasco (TO), 10
Italy 11
4 Department of Botany and Landscape Ecology, University of Greifswald, Greifswald, Germany 12
5 Bavarian State Institute of Forestry (LWF), Bavarian State Ministry of Food, Agriculture and 13
Forestry, Freising, Germany 14
6 Department of Genetics, University of Georgia, Athens, GA, USA 15
16
*Correspondence: hans.pretzsch@tum.de, f.johannes@tum.de 17
# Contributed equally to this work 18
19
Trees are integral to ecosystems and hold considerable economic importance. Their 20
exceptional longevity and modular structure also make them valuable models for studying 21
the long-term accumulation of somatic mutations and epimutations in plants. Empirical 22
evidence indicates that the annual rate of these stochastic events correlates negatively 23
with generation time, suggesting that species with long lifespans have evolved 24
mechanisms to mitigate the build-up of deleterious somatic variants. It has been 25
hypothesized that this reduction is achieved by slowing growth and minimizing the 26
number of cell divisions per unit time, thereby reducing errors associated with DNA 27
replication. However, a direct test of this “mitotic-rate hypothesis” remains technically 28
challenging. Here we took advantage of a 150 year-old experiment in European beech to 29
show that a thinning-induced growth acceleration increases the annual rate of somatic 30
epimutations in main stems and lateral branches of trees. We demonstrate that this effect 31
is accompanied by a proportional increase in the rate of cell divisions per unit time. These 32
findings support the notation that life-history constraints on growth rates in trees are not 33
merely a trade-off between resource allocation and structural stability but also a strategy 34
to preserve genetic and epigenetic fidelity over extended lifespans. 35
Keywords: Fagus sylvatica, DNA methylation, epimutation rate, somatic epigenetic drift, somatic 36
evolution, somatic epimutation, tree epigenomics 37
38
Trees are among the longest-living organisms on earth. They have critical ecosystem functions, 39
and continue to be of major economic importance 1. Perhaps due to their longevity, sessile life-40
style and modular nature, many tree species have evolved a remarkable degree of phenotypic 41
plasticity in response to environmental stressors 2. These plastic responses are at least in part 42
2
driven by epigenetic mechanisms, including DNA methylation and histone modifications 3. One 43
example is priming in Norway spruce (Picea abies), where hormonal exposure of seedlings can 44
induce an epigenetically-encoded stress memory that confers more effective resistance to insect 45
attacks later in life 4. Although such epigenetic memories are typically lost with passage into the 46
next generation 3, their ecological consequences are nonetheless relevant, given the ontogenetic 47
time-scales involved 5. 48
In addition to transient epigenetic effects, there are also more stable epigenetic changes 49
that occur stochastically during development and aging 6,7. A well-characterized form of such 50
changes are accidental gains and losses of DNA cytosine methylation, a phenomenon that has 51
been termed “spontaneous epimutation” 6. Epimutations at CG dinucleotides in plants are 52
particularly important because they are inherited not only during somatic cell division (mitosis), 53
which allows the changes to be passed on within an individual, but also during the formation of 54
reproductive cells (meiosis), ensuring that the changes can be passed on to the next generation 55
8–20. It is believed that they originate mainly from errors made by CG methyltransferases during 56
the maintenance of DNA methylation at cell division 19. When such epimutations occur in the shoot 57
apical meristem (SAM) - a small population of stem cells at the shoot apex - they often become 58
fixed in the cell lineages that differentiate into aerial structures such as stems and branches 21. 59
This fixation results from somatic drift driven by precursor cell sampling during organ formation 22. 60
Somatically fixed epimutations therefore appear at high frequencies in many plant tissues, and 61
can be detected using bulk sequencing approaches 17 (Methods). 62
Recent evidence shows that the yearly rate of fixed somatic epimutations correlates 63
negatively with generation time, being about ~2 orders of magnitude lower in long-lived trees (e.g. 64
Populus tricocarpa: ~0.08 x 10-4, CG per year) compared to annual plant species (e.g. Arabidopsis 65
thaliana: ~9.3 x 10-4 per CG per year) 23. Similar observations have been made at the level of 66
somatic genetic mutations 13,24–31. This rate reduction may serve as a protective mechanism in 67
long-lived species to delay the accumulation of potentially deleterious (epi)mutations during aging 68
32. Multiple lines of evidence indicate that this is achieved by slowing growth and the number of 69
stem cell divisions per unit time 22,33,34. However, experimental tests of this hypothesis remain 70
challenging, as they require a direct manipulation of these developmental parameters coupled 71
with a long-term assessment of their epimutational consequences. Moreover, in vivo 72
measurements of stem cell division would need to rely on live imaging of intact meristems in 73
mature trees - an approach that remains technically challenging. 74
Here we overcome many of these challenges, and provide a compelling test of the “mitotic-75
rate hypothesis” applied to mature trees growing under natural conditions. Using European beech 76
(Fagus sylvatica L.) as a model, we took advantage of one of the oldest continuously measured 77
experimental plots in the world. The plot contains an even-aged beech stand that was established 78
in 1822 and monitored for growth at regular intervals until the present. Starting about 150 years 79
ago, different thinning strategies were applied to subplots of this experiment, which has resulted 80
in differences in the average stem growth rates of trees among subplots. We show that thinning-81
induced growth acceleration significantly increased the annual rate of somatic epimutations in 82
both stems and lateral branches, and that this effect is accompanied by a proportional increase 83
in the rate of cell divisions of SAM-derived cell lineages. Our results lend support to the “mitotic-84
rate hypothesis” as a key explanation for the delay in (epi)mutational meltdown in long-lived plants. 85
Since somatic CG epimutations can be inherited across generations, our work further illustrates 86
3
how developmental processes within individual trees can impact epigenetic diversity at the 87
population level over extended timescales. 88
89
Results 90
Experimental stand thinning leads to accelerated growth 91
The Fabrikschleichach 15 (FAB 15) long-term thinning experiment in European beech is one of 92
the oldest continuously monitored experimental plots (Methods). The ~0.4 ha sized unthinned, 93
moderately thinned, and heavily thinned plots of FAB 15 were established in 1870/1871 in a 48-94
year-old, even-aged European beech stand that originated from natural regeneration by 95
shelterwood cutting of a beech forest in 1822 35. The managed plots were thinned 12 times at 96
regular intervals since their establishment, with a significant impact on stand growth dynamics 36. 97
In 2020, 138 trees remained in the unthinned, 70 in the moderately thinned, and 40 in the heavily 98
thinned plot. This resulted in different average stem growth rates of 8.1 cm² yr−1 (unthinned), 12.2 99
cm² yr−1 (moderate), and 16.4 cm² yr−1 (heavy) within the period from 1822 to 2020, which differed 100
significantly from each other (P ≤ 0.0001 for all three comparisons with pairwise Wilcoxon rank 101
sum test and Bonferroni correction). 102
103
Accelerated growth is accompanied by increased cell counts in stems 104
We felled two representative trees from the moderately thinned plot (tree 109) and the heavily 105
thinned plot (tree 171) to further study the impact of differential thinning on growth characteristics 106
(Fig. 1 A and B). The stem growth rates of these two trees were 12.2 cm² yr−1 (tree 109, 107
moderately thinned) and 21.8 cm² yr−1 (tree 171, heavily thinned). By the time of the last 108
measurement in 2020, tree 171 was only slightly taller than tree 109 (42.8 m vs. 42.4 m), but was 109
~1.3 times thicker at diameter breast height (DBH at 1.3 m: 72.3 cm vs. 55.1 cm) (Fig. 1C), and 110
had a 2.83 fold larger crown (crown projection area: 102 m² vs. 36 m²). This indicates that tree 111
171 had significantly more foliage and branches, despite both trees germinating in the same year. 112
The differential stem growth rates were reflected in the historical growth trajectories of both trees 113
(Fig. 1D). To evaluate whether differences in stem growth rate are the result of cell size expansion 114
or increased cell proliferation, we performed cell count assays of xylem vessels, fiber cells, and 115
parenchyma cells in stem discs (Methods). The cumulative number of cells was highly correlated 116
with the cumulative growth curve of the trees (r = 0.96). On average, tree 171 had significantly 117
more and larger cells per annual ring than tree 109 (P ≤ 0.0001, Fig. 1 E-G and Fig. S1), 118
suggesting that growth acceleration is accompanied by an increased cell division rate per unit 119
time. 120
121
Accelerated growth affects epimutation accumulation in the main stem 122
Cells in the outer cambium are the endpoints of cell lineages that have their origin in a few SAM-123
derived ancestor cells at the stem core 37–39. The above cell count data implies that the depth of 124
these cell lineages is higher in tree 171 than in tree 109. Under the assumption that spontaneous 125
epimutations arise with every cell division, we expected methylation divergence between 126
cambium cells sampled from opposite sides of the stem to be higher in trees from the heavily 127
thinned compared to the moderately thinned plot (Fig. 2A). To test this hypothesis, we obtained 128
cambium samples from additional trees, four trees from the moderately thinned and three from 129
the heavily thinned plot (Methods). For each tree, three cambium samples were collected: two 130
4
neighboring replicate samples from the same side of the stem and one sample from the polar 131
opposite side (Fig. 2A). The replicates were treated as controls, as their methylation divergence 132
is expected to be minimal due to shared cell lineage ancestry close to the cambium 40,41 (Fig. 2A). 133
We generated whole genome bisulfite sequencing (WGBS) data for all samples (N=18 WGBS 134
samples in total)(Methods). Consistent with our hypothesis, methylation divergence between 135
samples from opposite sides of the stems was 2.64-fold higher in trees from the heavily thinned 136
plot compared to trees from the moderately thinned plot (Fig. 2 B and C). By contrast, methylation 137
divergence between replicate samples was much lower and not significantly different between 138
plots. These findings indicate that accelerated growth, accompanied by increased cell division 139
rates, is associated with a significant increase in epimutation accumulation in the main stem. 140
141
Accelerated growth affects epimutation accumulation along lateral branches 142
Assuming that the relationship between growth rate, cell divisions, and epimutations is a broad 143
effect across the tree, not limited to the main stem, we expected to observe a similar pattern in 144
the lateral branches throughout the crown. The cell lineages that initiate leaf formation on distal 145
terminal branches derive from a common ancestor stem cell approximately at the last shared 146
branch point 22. Leaves that are more distal from each other in the branching topology are 147
separated by more stem cell divisions 22 and therefore expected to have accumulated more 148
epimutations between them 17. One way to assess this is to employ an intra-organismal 149
phylogenetic method that relates DNA methylation divergence among distal leaves to their 150
pairwise branching distance (in years) 17 (Fig. 3A). Application of this method to tree 171 and 109 151
should reveal growth-related differences in epimutation accumulation along lateral branches per 152
unit time. To test this, we first generated whole-genome bisulfite sequencing (WGBS) data from 153
leaf samples collected from tree 171 (N=10 WGBS samples) and tree 109 (N=7 WGBS samples). 154
The leaves were carefully selected to provide a broad representation of the three-dimensional 155
branching architecture of each tree (Fig. 3A). 156
To be able to detect epimutations that emerge at individual CG dinucleotides as well as in 157
larger (~100-200 bp) genomic regions, we identified single methylation polymorphisms (SMPs) 158
and differentially methylated regions (DMRs) between leaf samples using jDMR 12 (Methods). 159
Unsupervised clustering of the samples based on their SMP or DMR profiles recapitulated the 160
known branching topology of each tree (Fig. 3A and Fig. S2, Table S2), indicating that our 161
SMP/DMR calling approach was successful at reconstructing the branching (i.e. cell lineage) 162
history of the detected CG epimutations. In addition to DNA methylation profiling, we also 163
determined the ages of the sampled branches from branch disks (Methods). Together, these 164
data allowed us to calculate DNA methylation divergence as a function of divergence time 165
between all leaf pairs (Fig. 3 B and C). Consistent with recent work in P. trichocarpa 13,17, CG 166
methylation divergence increased gradually with divergence time in both trees. This was true at 167
the level of individual CGs as well as at the level of regions (Fig. 4D). However, divergence was 168
visibly more rapid in tree 171 than in 109 (Fig. 4D), suggesting that tree 171 accumulated 169
epimutations at a faster rate per unit time. 170
To obtain direct estimates of these rates, we employed AlphaBeta 17. At the genome-wide 171
scale we found that the spontaneous mCG gain rate (ɑ) and loss rate (β) were ~1.22 times higher 172
in tree 171 than in 109 on average (tree 171: ɑ = 4.69×10-6 (SE = 0.38×10-6), β = 6.50×10-6 (SE 173
= 0.52×10-6) per CG per haploid genome per year; tree 109: ɑ = 3.83×10-6 (SE = 0.41×10-6), β = 174
5
5.36×10-6 (SE = 0.57×10-6) per site per haploid genome per year (Fig. 4B and Table S3). A similar 175
picture emerged for region-level epimutation rate estimates (tree 171: ɑ = 6.79×10-6 (SE = 176
0.46×10-6), β = 9.45×10-6 (SE = 0.64×10-6) per 100 bp region per haploid genome per year; tree 177
109: ɑ = 4.99×10-6 (SE = 0.74×10-6), β = 6.96×10-6 (SE = 1.03×10-6) per 100 bp region per haploid 178
genome per year (Fig. 4B and Table S3). Interestingly, despite these epimutation rate 179
differences, genome-wide steady-state CG methylation levels were very similar between the two 180
trees (Fig. 4 A and C). This observation can be attributed to the nearly proportional increase in 181
gain and loss rates in tree 171 relative to 109, which theory predicts to result in unchanged steady-182
state methylation 10,17,42(Methods). 183
The above genome-wide epimutation rate analysis may have included subsets of CG 184
dinucleotides that are redundantly targeted by de novo methylation pathways 43. The methylation 185
gain and loss dynamics underlying our rate estimates may therefore be independent of the 186
number of cell divisions along the branches. We sought to test if our conclusions still hold when 187
focusing on CG sites that are exclusively targeted by METHYLSTRANSFERASE 1 (MET1), 188
whose maintenance activity is known to be restricted to DNA replication 43. CG sites within gene 189
body methylated (gbM) genes provide an excellent framework to test this. In plants, gbM genes 190
are an evolutionary conserved subset of genes that feature high CG methylation and virtually no 191
non-CG methylation (i.e. methylation in context CHG and CHH (where H = A, T, C)) 44–47. These 192
genes are enriched in housekeeping functions, are moderately expressed, and display low 193
transcriptional variability across cells and tissues 48–51 (extensively reviewed in Refs. 52,53). We 194
identified a liberal set of ~10,000 gbM genes out of the 65,000 annotated genes and pseudogenes 195
in the current European Beech (Fagus sylvatica L.) reference assembly (Methods). Using CGs 196
extracted from gbM genes, we repeated our epimutation rate estimation. We found that the rate 197
difference between trees 171 and 109 became even more pronounced, with CG epimutation rates 198
being ~1.66 times higher in tree 171 than in tree 109 on average (tree 171: ɑ = 9.66×10-6 (SE = 199
0.96×10-6), β = 5.62×10-6 (SE = 0.56×10-6) per CG per haploid genome per year; tree 109: ɑ 200
=5.94×10-6 (SE = 1.02×10-6), β = 3.32×10-6 (SE = 0.57×10-6) per site per haploid genome per year) 201
(Fig. 4B and Table S3). These results provide further, albeit indirect, evidence that the 202
epimutation rate differences are likely coupled with elevated cell division rates in the SAM and/or 203
in cell lineages leading up to the formation of leaf primordia. 204
205
Discussion 206
The "mitotic-rate hypothesis" proposes that the accumulation of somatic mutations (and 207
epimutations) is predominantly governed by the rate of growth and the number of cell divisions 208
per unit time, particularly within the stem cell compartment of the shoot apical meristem (SAM) 33. 209
Our findings support this view: we observed that experimentally induced acceleration of tree 210
growth correlates with increased rates of epimutations in both the main stem and lateral branches. 211
By focusing on epimutations at CG sites - whose methylation patterns are maintained during DNA 212
replication 43 - these results imply that the increased rates are driven by a higher frequency of cell 213
divisions. Our cell-count assays confirmed this conclusion. However, we acknowledge that these 214
assays were conducted exclusively in hardwood, where SAM-derived cell lineages leave a 215
traceable historical record that can be retrospectively analyzed. As such, these assays provide 216
only an indirect proxy for estimating cell division rates in other tissue types. Directly measuring 217
stem cell division rates in the SAM remains experimentally infeasible due to the technical 218
6
challenge of live imaging in intact meristems of mature trees 22. However, wood formation 219
observations (i.e., xylogenesis sampling) could help to at least better grasp the exact cell division 220
dynamics 54. 221
The "mitotic-rate hypothesis" explains our findings and accounts for broader patterns, 222
such as the negative relationship between generation time and per-year somatic (epi)mutation 223
and substitution rates observed in tree species 23,33,55. In long-lived species, slower growth and 224
reduced annual cell divisions delay the accumulation of potentially deleterious somatic variants 225
22,33, affecting both mutation and epimutation rates equally. In contrast, mammalian studies argue 226
that long-lived species have evolved more efficient DNA repair mechanisms 56. However, this 227
argument is less applicable to CG epimutations, which occur approximately four orders of 228
magnitude more frequently than DNA mutations 10–12,18, show no clear link to DNA damage, and 229
largely arise from errors by methyltransferases during cell division 19. Therefore, enhanced DNA 230
repair alone is unlikely to explain the variation in epimutation rates among plants. 231
An alternative perspective to the "mitotic-rate hypothesis" proposes that the accumulation 232
of somatic variants is determined solely by chronological age rather than developmental time 27. 233
This “age-related hypothesis” 57 implies that replication-independent mechanisms are the primary 234
drivers of somatic variations. Recent evidence supporting this view comes from a comparison of 235
two tropical tree species, Shorea laevis and Shorea leprosula, where age, rather than growth rate, 236
appeared to be the key determinant underlying mutation rate differences 27,57. Although the 237
influence of age on mutation rates have been documented in mammals - for instance, in oocytes 238
- this hypothesis would have made incorrect predictions about the outcome of our study, as fast 239
and slow growing trees had the same age but differing epimutation rates per unit time. Additionally, 240
the hypothesis fails to account for the nearly 35 to 110-fold variation in per-year (epi)mutation 241
rates observed across plant species 23. If age were the sole factor, per-year rates should be 242
constant across species, with only per-generation rates varying due to differences in lifespans. 243
Instead, the opposite is observed: when adjusting per-year rates by generation time differences, 244
rates in trees are within a mere ~3-fold of those observed in the short-lived annual plants like A. 245
thaliana 23. 246
This latter observation suggests that the net generational burden of (epi)mutations is 247
evolutionarily constrained across species 23. This constraint also predicts that accelerated growth 248
is offset by early growth cessation, senescence or mortality, which has some support in data 58–
249
60. Given the increasing anthropogenic pressures on forests and climate-driven alterations in tree 250
growth patterns, understanding the trade-offs between growth rate and genome stability has 251
profound implications for forest management and conservation. Our findings contribute novel 252
insights into the evolutionary constraints shaping plant longevity and genome maintenance. 253
254
Methods 255
Study site and sampling design 256
We relied on the long-term thinning trial Fabrikschleichach 15 (FAB 15) in a European beech 257
(Fagus sylvatica L.) forest in Central Europe (Fig. 1A) to obtain trees with exactly the same age, 258
but with strongly divergent growth rates. The region is characterized by a mild climate with an 259
average annual temperature of 7.5 °C and an annual precipitation of 820 mm, with the natural 260
vegetation being submontane European beech-sessile oak forests. FAB 15 is one of the longest 261
continuously measured forest experimental sites worldwide and consists of three differently 262
7
managed plots, each with a size of ~0.4 ha: unthinned, moderately thinned, and heavily thinned. 263
The beech forest, now over 200 years old, originated from a shelterwood cutting in 1822, and the 264
various treatments and continuous measurements began in 1870/1871. The unthinned plots 265
remained essentially untouched, apart from the occasional removal of dying or dead trees to 266
prevent possible damage to the stand from fungal or insect infestation. In contrast, the managed 267
plots were moderately or heavily thinned to reduce stand density by removing mainly suppressed 268
trees, but also tall trees, especially in the case of the more intense treatment 36. In total, the 269
managed plots have been thinned 12 times with different intensities since their establishment, 270
resulting in different tree sizes and stem growth rates (Fig. 1 B and C). We selected a 271
representative, average sample tree from each thinned plot for further analyses, which was felled 272
and precisely measured. Both selected trees had a particularly low branching point to allow for 273
distinct differences in branch divergence time within the same tree. After felling, branches along 274
the crown periphery were selected (distributed across lower, middle, and upper crown) and leaves 275
were sampled from the ends of these branches (Fig. 3A). All leaf samples were kept frozen at -276
80 °C for subsequent DNA methylation analysis. For a correct assignment of branches within the 277
tree architecture, we reconstructed the tree topology and only selected those branches for 278
sampling whose branching path could be clearly traced back. Furthermore, we obtained stem and 279
branch disks from different positions along the stem axis and crown topology to allow for age and 280
stem growth rate estimations. In more detail, a stem disk was extracted from each tree at 1.3 m 281
height, while branch disks from leaf-sampled branches were taken 10 cm before and after each 282
branching position. We followed the sampling methods described in several published studies 61–
283
64 to collect cambium samples at 1.3 m height from moderately thinned and heavily thinned plots. 284
All samples were immediately washed with Ethanol 70% and stored at -80 for DNA extraction. 285
286
Growth rate analysis 287
We used the stems disks to measure the annual tree ring widths from four cardinal directions (N, 288
E, S, W; see Fig. 1E) to the nearest of 1/100 mm using a digital positioning table for stem disk 289
measurements (Digitalpositometer, Biritz and Hatzl GmbH, Austria) after sanding the disks with 290
progressively finer sandpaper (up to 800 grit). Tree ring widths were transformed to basal area 291
increments (bai) using the formula
, where represents the tree’s radius at 292
1.3 m height and the respective year . We defined the stem growth rate as the average annual 293
basal area increment, considering all four series per stem disk. 294
295
Branch age determination 296
For determining the age of the different branches, we utilized the derived branch disks, which 297
were sanded to enhance the visibility of annual tree ring borders (see also stem disk preparation). 298
Tree rings of all branch disks (from four cardinal directions per disk) were measured with the 299
digital positioning table Lintab 5 and the software TSAPWin (both Rinntech, Heidelberg, 300
Germany). We visually cross-dated all tree ring series, taking into account distinctive growth 301
patterns across the different samples to ensure the correct dating of the individual tree rings 65. 302
We counted all tree rings from bark to pith as an estimation of the age of each branch. 303
304
Cell count assays 305
8
We relied on wood anatomical methods to derive estimates of annually produced xylem cells for 306
both sample trees. Stem disks were cut with a small circular saw into ~1cm wide sections, 307
capturing all tree rings from the pith to the bark. For both trees, we used a section without any 308
cracks or damages for further wood anatomical analyses. The cross sections were cut into smaller 309
segments of 3-5 cm before taking transverse micro sections of 10-20 µm thickness with a sliding 310
GSL-1 microtome (Schenkung Dapples, Zürich, Switzerland). Sample preparation followed a 311
protocol by Gärtner & Schweingruber 66. In more detail, all samples were bleached and washed 312
with distilled water, before they were double-stained with safranin and astrablue for 5 minutes 313
(mixture of 1:1). Subsequently, the samples were again rinsed with distilled water, ethanol of 314
increasing purity (80 %, 96 %, and anhydrous ethanol) for dehydration, and xylene before being 315
permanently embedded on a glass slide with Canada balsam and dried out in an oven at 60°C 316
for 24h. We used an optical microscope (Zeiss Axio Imager Z2) at × 200 magnification, equipped 317
with an integrated camera (Zeiss Axiocam 305 color) to capture micrographs of wood anatomical 318
features. The micrographs were stitched together using the software ZEN 3.2 blue edition (all 319
three by Carl Zeiss Microscopy Deutschland GmbH, Oberkochen, Germany). Finally, we utilized 320
CARROT, a software based on Deep Convolutional Neural Network (DCNN) algorithms, 321
specifically programmed and trained for quantitative wood anatomy analyses (Fig. S3). The 322
software was employed to recognize and segment all types of cells for each ring from the obtained 323
micrographs 67. We assessed the number of xylem cells per tree ring along a 5000 pixel wide 324
band extending from the start of treatment in 1870 to the felling date in 2020. 325
326
Whole Genome Bisulfite Sequencing analysis 327
DNA was extracted from cambium tissue using Qiagen's DNeasy Plant Pro Kit (catalog no. 69204, 328
Qiagen) following the manufacturer's protocol with the addition of 100 μl of the solution PS per 329
sample due to high amounts of phenolic compounds in beech species. We used NEBNext Ultra 330
II DNA Library Prep Kit (catalog no. E7103, New England BioLabs) for sequencing library 331
preparation and EZ-96 DNA Methylation-Gold MagPrep (catalog no. D5042, ZYMO) for bisulfite 332
treatment. The protocol involved: i) end repair and 3' adenylation of sonicated DNA fragments by 333
Covaris R230 (Covaris), ii) NEBNext adaptor ligation, iii) cleanup of libraries with AMPure XP 334
Beads (catalog no. A63881, Beckman Coulter), iv) bisulfite treatment, v) PCR enrichment and 335
index ligation using KAPA HiFi Uracil+ Kit (catalog no. KK2802, Agilent) and NEBNext Multiplex 336
Oligos for Enzymatic Methyl-seq (catalog no. E7140L, New England BioLabs) for bisulfite libraries 337
(12 cycles), vi) final cleanup with AMPure XP Beads. Finally, they were sequenced on a NovaSeq 338
X Plus platform (Illumina) in a paired-end 150bp format. DNA was extracted from leaves using 339
Qiagen’s DNeasy Plant Mini Kit. WGBS libraries were constructed by BGI (Beijing Genomics 340
Institute) and then sequenced on a NovaSeq 6000 platform (Illumina) in a paired-end 150bp 341
format. Removing the low-quality sequencing samples, we had 18 cambium samples and 17 leaf 342
samples. The Fagus sylvatica L. reference genome and annotation was used from 343
http://www.beechgenome.net/ 68. The WGBS data was processed with the MethylStar pipeline 69. 344
Specifically, sequencing quality was checked with FastQC v0.11.7, and clean reads were 345
obtained using Trimmomatic v0.39 with parameters: ILLUMINACLIP:TruSeq3-PE.fa:1:30:9 346
LEADING:20 TRAILING:20, SLIDINGWINDOW: 4:20 MINLEN:36. We mapped the reads to the 347
reference genome, removed duplicates, and extracted the methylation levels using Bismark 348
v0.19.1 with default parameters. METHimpute was used for cytosine methylation state calling 70. 349
9
The mean coverage of samples was about 20X, and the mean mapping rate was about 70% 350
(Table S1). 351
352
Identification of gene body methylated genes 353
We identified gene body methylated (gbM) genes using gbMine (github.com/jlab-code/gbMine). 354
The software uses the outputs of MethylStar and an annotation (gff3) file. gbMINE provides two 355
flags, -- genomicBackground and -- exons, which allows us to identify gbM genes using four 356
different genomic features combinations. We only set -- exons, which means that we only consider 357
exonic cytosines in both the foreground and the background. Using a binomial model, gbMINE 358
classifies a gene as gbM if the proportion of mCG in the exons of that gene is statistically higher 359
than that of exons in the background, while the proportion of mCHG and mCHH is not significantly 360
different. We obtained a set of gbM genes for each sample, and used the union as a liberal set of 361
gbM genes. As a further filter, we calculated the non-CG methylation levels of introns of the union 362
set, which generated a distribution of intronic non-CG methylation levels for candidate gbM gene. 363
Finally, we only retained gbM genes whose intronic non-CG methylation levels were less than the 364
median of that distribution. 365
366
SMP and DMR calling 367
DMR calling was performed using jDMR 12 (github.com/jlab-code/jDMRgrid). Briefly, we divided 368
the genome into windows of 100 bp each. Within these windows, we calculated the number 369
methylated and total cytosines, which were used as input for Methimpute 70, a finite state Hidden 370
Markov Model (HMM) with binomial emission densities. We employed Methimpute’s three-state 371
HMM, where each 100 bp window was classified as methylated, unmethylated, or intermediate. 372
The intermediate state calls are designed to capture somatic epiheterozygotes, which are only 373
visible in WGBS data in the form of “intermediate” methylation levels. This approach yielded a 374
matrix representing the methylation state of each region and sample (Table S2). jDMR queries 375
this matrix to identify methylation state switches between samples within each 100 bp window 376
(e.g. unmethylated to intermediate methylated) and defines these as DMRs. A similar strategy 377
was employed to call SMPs, with the window size being reduced to single CG sites. Although our 378
SMP and DMR analysis was based on bulk leaf methylome data, jDMR’s HMM approach is robust 379
to measurement noise arising from the cell layer-specificity in which somatic variations often 380
originate in the SAM 71,72. This becomes evident from the clustering results in Fig. 3A. 381
382
Epimutation rate estimation 383
We estimated somatic epimutation rates using AlphaBeta 17, a computational method that infers 384
the spontaneous epimutation rate from WGBS data. AlphaBeta requires methylation data and 385
pedigree data as input. Following Shahryary et al. 17 and Hofmeister et al. 13, we treated the tree 386
branching structure as a pedigree of somatic cell lineages, where the leaves represent the lineage 387
end-points. AlphaBeta calculates the methylation divergence (D) and divergence time (Δt) 388
between all pairs of leaf samples. The model parameters are estimated using numerical nonlinear 389
least squares optimization. We estimated epimutation rates under a neutral epimutation model 390
(AlphaBeta’s ABneutralSOMA model). 391
392
Data availability 393
10
All WGBS raw data used for this article have been deposited in the NCBI SRA under accession 394
PRJNA1215776. 395
396
ACKNOWLEDGEMENTS 397
This work was partly funded by the Bundesministerium für Bildung und Forschung (Project: 398
epiSOMA). We would like to thank the Bavarian State Institute of Forestry (Forest Protection 399
Department) and the Ecophysiology group at TUM for their support with technical equipment and 400
the Bavarian State Forests (BaySF) for their support in setting up and maintaining the underlying 401
long-term experiments. We thank W. Wanney, L. Schlegel, and R. Bhardwaj for helping collect 402
cambium samples. G. Schmied acknowledges further funding from the Bavarian State Ministry of 403
Food, Agriculture and Forestry (Project: beechGPT). M. Zhou holds a fellowship from the China 404
Scholarship Council (CSC NO. 202204910009). 405
406
407
REFERENCES 408
1. Fazan, L., Song, Y.-G. & Kozlowski, G. The Woody Planet: From Past Triumph to 409
Manmade Decline. Plants 9, (2020). 410
2. Gratani, L. Plant Phenotypic Plasticity in Response to Environmental Factors. Advances in 411
Botany 2014, (2014). 412
3. Lämke, J. & Bäurle, I. Epigenetic and chromatin-based mechanisms in environmental 413
stress adaptation and stress memory in plants. Genome Biol. 18, 124 (2017). 414
4. Mageroy, M. H. et al. Priming of inducible defenses protects Norway spruce against tree-415
killing bark beetles. Plant Cell Environ. 43, 420–430 (2020). 416
5. Amaral, J. et al. Advances and Promises of Epigenetics for Forest Trees. For. Trees 417
Livelihoods 11, 976 (2020). 418
6. Johannes, F. & Schmitz, R. J. Spontaneous epimutations in plants. New Phytol. 221, 1253–419
1259 (2019). 420
7. Vo, B. T., Mas, P. & Johannes, F. Time’s up: Epigenetic clocks in plants. Curr Opin Plant 421
Biol 81, 102602 (2024). 422
8. Schmitz, R. J. et al. Transgenerational epigenetic instability is a source of novel methylation 423
variants. Science 334, 369–373 (2011). 424
9. Becker, C. et al. Spontaneous epigenetic variation in the Arabidopsis thaliana methylome. 425
11
Nature 480, 245–249 (2011). 426
10. van der Graaf, A. et al. Rate, spectrum, and evolutionary dynamics of spontaneous 427
epimutations. Proceedings of the National Academy of Sciences 112, 6676–6681 (2015). 428
11. Denkena, J., Johannes, F. & Colomé-Tatché, M. Region-level epimutation rates in 429
Arabidopsis thaliana. Heredity 1–13 (2021). 430
12. Hazarika, R. R. et al. Molecular properties of epimutation hotspots. Nat Plants 8, 146–156 431
(2022). 432
13. Hofmeister, B. T. et al. A genome assembly and the somatic genetic and epigenetic 433
mutation rate in a wild long-lived perennial Populus trichocarpa. Genome Biol. 21, 259 434
(2020). 435
14. Jiang, C. et al. Environmentally responsive genome-wide accumulation of de novo 436
Arabidopsis thaliana mutations and epimutations. Genome Res. 24, 1821–1829 (2014). 437
15. Ganguly, D. R., Crisp, P. A., Eichten, S. R. & Pogson, B. J. The Arabidopsis DNA 438
Methylome Is Stable under Transgenerational Drought Stress. Plant Physiol. 175, 1893–439
1912 (2017). 440
16. Zheng, X. et al. Transgenerational epimutations induced by multi-generation drought 441
imposition mediate rice plant’s adaptation to drought condition. Sci. Rep. 7, 39843 (2017). 442
17. Shahryary, Y. et al. AlphaBeta: computational inference of epimutation rates and spectra 443
from high-throughput DNA methylation data in plants. Genome Biol. 21, 260 (2020). 444
18. Yao, N. et al. An evolutionary epigenetic clock in plants. Science 381, 1440–1445 (2023). 445
19. Briffa, A. et al. Millennia-long epigenetic fluctuations generate intragenic DNA methylation 446
variance in Arabidopsis populations. Cell Syst (2023) doi:10.1016/j.cels.2023.10.007. 447
20. Ibañez, V. N. et al. Environmental and genealogical effects on DNA methylation in a 448
widespread apomictic dandelion lineage. J. Evol. Biol. 36, 663–674 (2023). 449
21. Chen, Y., Burian, A. & Johannes, F. Somatic epigenetic drift during shoot branching: a cell 450
lineage-based model. Genetics 227, iyae091 (2024). 451
12
22. Burian, A., Barbier de Reuille, P. & Kuhlemeier, C. Patterns of Stem Cell Divisions 452
Contribute to Plant Longevity. Curr. Biol. 26, 1385–1394 (2016). 453
23. Johannes, F. Allometric scaling of somatic mutation and epimutation rates in trees. 454
Evolution 79, 1–5 (2024). 455
24. Duan, Y. et al. Limited accumulation of high-frequency somatic mutations in a 1700-year-456
old Osmanthus fragrans tree. Tree Physiol. 42, 2040–2049 (2022). 457
25. Hanlon, V. C. T., Otto, S. P. & Aitken, S. N. Somatic mutations substantially increase the 458
per-generation mutation rate in the conifer Picea sitchensis. Evol Lett 3, 348–358 (2019). 459
26. Perez-Roman, E., Borredá, C., López-García Usach, A. & Talon, M. Single-nucleotide 460
mosaicism in citrus: Estimations of somatic mutation rates and total number of variants. 461
Plant Genome 15, e20162 (2022). 462
27. Satake, A. et al. Somatic mutation rates scale with time not growth rate in long-lived tropical 463
trees. Elife 12, (2023). 464
28. Schmid-Siegert, E. et al. Low number of fixed somatic mutations in a long-lived oak tree. 465
Nat Plants 3, 926–929 (2017). 466
29. Wang, L. et al. The architecture of intra-organism mutation rate variation in plants. PLoS 467
Biol 17, e3000191 (2019). 468
30. Xie, Z. et al. Mutation rate analysis via parent-progeny sequencing of the perennial peach. 469
I. A low rate in woody perennials and a higher mutagenicity in hybrids. Proc. Biol. Sci. 283, 470
(2016). 471
31. Orr, A. J. et al. A phylogenomic approach reveals a low somatic mutation rate in a long-472
lived plant. Proceedings of the Royal Society B: Biological Sciences 287, (2020). 473
32. Schoen, D. J. & Schultz, S. T. Somatic Mutation and Evolution in Plants. Annu. Rev. Ecol. 474
Evol. Syst. 50, 49–73 (2019). 475
33. Lanfear, R. et al. Taller plants have lower rates of molecular evolution. Nat. Commun. 4, 476
1879 (2013). 477
13
34. Li, W. H., Ellsworth, D. L., Krushkal, J., Chang, B. H. & Hewett-Emmett, D. Rates of 478
nucleotide substitution in primates and rodents and the generation-time effect hypothesis. 479
Mol Phylogenet Evol 5, 182–187 (1996). 480
35. Burschel, P. & Huss, J. Grundriss des Waldbaus: ein Leitfaden für Studium und Praxis. 481
(1997). 482
36. Pretzsch, H. The social drift of trees. Consequence for growth trend detection, stand 483
dynamics, and silviculture. Eur. J. For. Res. 140, 703–719 (2021). 484
37. Ha, C. M., Jun, J. H. & Fletcher, J. C. Shoot apical meristem form and function. Curr. Top. 485
Dev. Biol. 91, 103–140 (2010). 486
38. Whaley, W. G. Developmental changes in apical meristems. Proc. Natl. Acad. Sci. U. S. A. 487
25, 445–448 (1939). 488
39. Wybouw, B., Zhang, X. & Mähönen, A. P. Vascular cambium stem cells: past, present and 489
future. New Phytol. 243, 851–865 (2024). 490
40. Miodek, A., Gizińska, A., Włoch, W. & Kojs, P. Intrusive growth of initials does not affect 491
cambial circumference in Robinia pseudoacacia. Sci. Rep. 12, 7428 (2022). 492
41. Tomescu, A. M. F. & Groover, A. T. Mosaic modularity: an updated perspective and 493
research agenda for the evolution of vascular cambial growth. New Phytol. 222, 1719–1735 494
(2019). 495
42. Goeldel, C. & Johannes, F. Stochasticity in gene body methylation. Curr. Opin. Plant Biol. 496
75, 102436 (2023). 497
43. Law, J. A. & Jacobsen, S. E. Establishing, maintaining and modifying DNA methylation 498
patterns in plants and animals. Nat. Rev. Genet. 11, 204–220 (2010). 499
44. Tran, R. K. et al. DNA methylation profiling identifies CG methylation clusters in Arabidopsis 500
genes. Curr. Biol. 15, 154–159 (2005). 501
45. Niederhuth, C. E. et al. Widespread natural variation of DNA methylation within 502
angiosperms. Genome Biol. 17, 194 (2016). 503
14
46. Takuno, S., Ran, J.-H. & Gaut, B. S. Evolutionary patterns of genic DNA methylation vary 504
across land plants. Nat Plants 2, 15222 (2016). 505
47. Bewick, A. J. et al. The evolution of CHROMOMETHYLASES and gene body DNA 506
methylation in plants. Genome Biol. 18, 65 (2017). 507
48. Horvath, R., Laenen, B., Takuno, S. & Slotte, T. Single-cell expression noise and gene-508
body methylation in Arabidopsis thaliana. Heredity 123, 81–91 (2019). 509
49. Zilberman, D., Gehring, M., Tran, R. K., Ballinger, T. & Henikoff, S. Genome-wide analysis 510
of Arabidopsis thaliana DNA methylation uncovers an interdependence between 511
methylation and transcription. Nat. Genet. 39, 61–69 (2007). 512
50. Vaughn, M. W. et al. Epigenetic Natural Variation in Arabidopsis thaliana. PLoS Biol. 5, 513
e174 (2007). 514
51. Zhang, X. et al. Genome-wide high-resolution mapping and functional analysis of DNA 515
methylation in arabidopsis. Cell 126, 1189–1201 (2006). 516
52. Bewick, A. J. & Schmitz, R. J. Gene body DNA methylation in plants. Curr. Opin. Plant Biol. 517
36, 103–110 (2017). 518
53. Muyle, A. M., Seymour, D. K., Lv, Y., Huettel, B. & Gaut, B. S. Gene Body Methylation in 519
Plants: Mechanisms, Functions, and Important Implications for Understanding Evolutionary 520
Processes. Genome Biol. Evol. 14, (2022). 521
54. Cuny, H. E. et al. Couplings in cell differentiation kinetics mitigate air temperature influence 522
on conifer wood anatomy. Plant Cell Environ 42, 1222–1232 (2019). 523
55. De La Torre, A. R., Li, Z., Van de Peer, Y. & Ingvarsson, P. K. Contrasting Rates of 524
Molecular Evolution and Patterns of Selection among Gymnosperms and Flowering Plants. 525
Mol. Biol. Evol. 34, 1363–1377 (2017). 526
56. Cagan, A. et al. Somatic mutation rates scale with lifespan across mammals. Nature 604, 527
517–524 (2022). 528
57. Leroy, T. Re-evaluating the driving force behind mutations. Elife 12, (2023). 529
15
58. Pretzsch, H. The course of tree growth. Theory and reality. For. Ecol. Manage. 478, 118508 530
(2020). 531
59. Searle, E. B. & Chen, H. Y. H. Temporal declines in tree longevity associated with faster 532
lifetime growth rates in boreal forests. Environ. Res. Lett. 13, 125003 (2018). 533
60. Bialic-Murphy, L. et al. The pace of life for forest trees. Science 386, 92–98 (2024). 534
61. Colpaert, N. et al. Sampling tissue for DNA analysis of trees: Trunk cambium as an 535
alternative to canopy leaves. Silvae Genet. 54, 265–269 (2005). 536
62. Siregar, I. Z. et al. Collecting wood core samples from Macassar ebony (Diospyros celebica 537
Bakh.) for multi-purpose analysis using pickering punch. MethodsX 9, 101728 (2022). 538
63. Mangaravite, É. et al. A feasible method to extract DNA from the cambium of high-canopy 539
trees: from harvest to assessment. Acta Amazon. 50, 335–338 (2020). 540
64. Gardner, S. T. et al. Development of DNA methylation-based epigenetic age predictors in 541
loblolly pine (Pinus taeda). Mol. Ecol. Resour. 23, 131–144 (2023). 542
65. Speer, J. H. Fundamentals of Tree Ring Research. (University of Arizona Press, 2010). 543
66. Gärtner, H. & Schweingruber, F. H. Microscopic Preparation Techniques for Plant Stem 544
Analysis. (Kessel, 2013). 545
67. Resente, G. et al. Mask, Train, Repeat! Artificial Intelligence for Quantitative Wood 546
Anatomy. Front. Plant Sci. 12, 767400 (2021). 547
68. Mishra, B. et al. A Chromosome-Level Genome Assembly of the European Beech (Fagus 548
sylvatica) Reveals Anomalies for Organelle DNA Integration, Repeat Content and 549
Distribution of SNPs. Front. Genet. 12, 691058 (2021). 550
69. Shahryary, Y., Hazarika, R. R. & Johannes, F. MethylStar: A fast and robust pre-processing 551
pipeline for bulk or single-cell whole-genome bisulfite sequencing data. BMC Genomics 21, 552
479 (2020). 553
70. Taudt, A. et al. METHimpute: imputation-guided construction of complete methylomes from 554
WGBS data. BMC Genomics 19, 444 (2018). 555
16
71. Amundson, K. R. et al. Differential mutation rates in plant meristematic layers. bioRxiv 556
2023.09.25.559363 (2023) doi:10.1101/2023.09.25.559363. 557
72. Sun, H. et al. The identification and analysis of meristematic mutations within the apple tree 558
that developed the RubyMac sport mutation. BMC Plant Biol 24, (2024). 559
Figures
Figure 1
See image above for gure legend
Figure 2
See image above for gure legend
Figure 3
See image above for gure legend
