Available via license: CC BY 4.0
Content may be subject to copyright.
No evidence for a link between childhood (6-10y) cellular aging and brain
morphology (12y) in a preregistered longitudinal study.
Evie Henrike Brinkman (Donders Institute for Brain Cognition and Behaviour, Centre of
Cognitive Neuroimaging & Radboudumc)
Roseriet Beijers (Behavioural Science Institute, Radboud University, Donders Institute for
Brain Cognition and Behaviour, Centre of Cognitive Neuroimaging & Radboudumc)
Anna Tyborowska (Behavioural Science Institute, Radboud University, Nijmegen & Donders
Institute for Brain Cognition and Behaviour, Centre of Cognitive Neuroimaging)
Karin Roelofs (Behavioural Science Institute, Radboud University, Donders Institute for
Brain Cognition and Behaviour, Centre of Cognitive Neuroimaging)
Simone Kühn (Lise Meitner Group for Environmental Neuroscience, Max Planck Institute for
Human Development, Berlin, Germany & University Medical Center Hamburg-Eppendorf,
Germany)
Rogier Kievit (Donders Institute for Brain Cognition and Behaviour, Centre of Cognitive
Neuroimaging & Radboudumc)
Carolina de Weerth (Donders Institute for Brain Cognition and Behaviour, Centre of
Cognitive Neuroimaging & Radboudumc)
Corresponding author:
Evie H. Brinkman
evie.brinkman@radboudumc.nl
Radboud University Medical Center
Kapittelweg 29, 6525 EN Nijmegen, The Netherlands
Highlights (3-5):
- Investigation of cellular aging in relation to brain morphology in a community sample
(N=95)
- Epigenetic aging and telomere shortening were not associated with brain structure
- Exploratory Bayesian Analyses reveal moderate to strong evidence for null findings
- No association was found between cellular aging and white matter volume
Abstract
Animal studies show that early life environmental factors, such as stress and trauma, can have
a significant impact on a variety of bodily processes, including cellular aging and brain
development. However, whether cellular wear-and-tear effects are also associated with
individual differences in brain structures in humans, remains unknown. In this pre-registered
study in a community sample of children (N=94, Mean age=12.71 years), we prospectively
investigated the predictive value of two markers of cellular aging in childhood (at age 6 and
10) for brain morphology in early adolescence (age 12). More specifically, we associated
buccal cell telomere length and epigenetic age in childhood to individual differences in
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
2
adolescent whole-brain grey matter volume (GMV) including volumes of three regions of
interest that have been found to be sensitive to effects of early life stress (i.e. amygdala,
hippocampus, (pre)frontal cortex -PFC). Multiple regression analyses revealed no significant
associations between childhood cellular aging (at 6 and 10 years) and early adolescent brain
morphology. Exploratory Bayesian analyses indicated moderate to strong evidence for the
null-findings. These results suggest that although our sample is modest, the associations
between middle childhood cellular aging and early adolescent brain morphology are, if they
do exist, likely not particularly large in community children. Future work should investigate
whether these effects are similarly absent in large samples, in samples with a higher risk
profile and in samples characterized by different age ranges.
Keywords: cellular aging, telomere length, epigenetic age, brain structure, paediatric
imaging
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
3
1. Introduction
1
Experiences early in life can shape the development of the body and the brain. Environmental
2
factors, such as stress, toxins, and poor nutrition, can impact the wear-and-tear in cells as
3
reflected by changes in markers of cellular aging. Notably, many studies have suggested
4
similar associations between environmental risks and differences in brain morphology,
5
suggesting that the wear and tear effects may ultimately manifest at the level of brain structure
6
too (Colich et al., 2020). The goal of the current pre-registered study is to investigate whether
7
markers of cellular aging (i.e. telomere length and epigenetic age) in childhood can predict
8
brain morphology in early adolescence. Individual differences in adolescent brain morphology
9
have been shown to predict later phenotypic outcomes such as aggression, emotion regulation,
10
and episodic memory (see meta-analysis by Colich et al., 2020; or Dufford et al., 2019; el
11
Marroun et al., 2016; Ghetti & Bunge, 2012; Hanson et al., 2010; Tyborowska et al., 2018).
12
Given their roles in stress-related symptomatology, we will focus on the amygdala,
13
hippocampus, and the PFC as specific regions of interest, in addition to whole brain
14
morphology.
15
16
One marker of cellular aging is telomere length (Darrow et al., 2016; Harley et al.,
17
1992). At the end of each chromosome is the telomere, a cap which protects the
18
chromosome’s ends from degradation (Stewart et al., 2012).With each cell division the
19
telomere shortens, resulting in telomere attrition with cellular age. Over time and with
20
chronological aging, telomeres are steadily shortened, varying in rate throughout the lifespan
21
(Vaiserman & Krasnienkov, 2021). Therefore, the length of the telomeres can be seen as a
22
biological marker of cellular aging, with shorter telomere lengths reflecting older ages (Turner
23
et al., 2019). Eventually, the telomeres will become too short and the cell will go into
24
reproductive senescence (Sikora et al., 2021; Stewart et al., 2012). Importantly, telomeres in
25
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
4
senescent state show a characteristic secretion pattern known as the senescence-associated
26
secretory phenotype (SASP), including many cytokines and growth factors (Sikora et al.,
27
2021). The secretion of these cytokines leads to an inflammatory state in the body, which has
28
been extensively positively related to brain development as well as to accelerated aging
29
(Sikora et al., 2021)
30
Indeed, first evidence from a cross-sectional study in 389 children (age 6-14) suggests
31
that telomere lengths are associated with spontaneous activity in two main hubs of the default
32
mode network (DMN): the posterior cingulate cortex (PCC) and the medial prefrontal cortex
33
(mPFC) (Rebello et al., 2019). Moreover, previous research has linked activity in the DMN
34
and telomere lengths to internalizing and externalizing problems (Davis et al., 2022; Farina et
35
al., 2018; Sato et al., 2015). Shorter telomeres at age 6 were found to predict more self-
36
reported internalizing and externalizing problems at age 10 in the sample of 193 community
37
children from the current study (Beijers & Daehn, et al., 2020). These finding suggest an
38
association between telomere length and properties of brain structure and function, but studies
39
investigating this link in humans are scarce.
40
41
Another marker of cellular aging is epigenetic age. Epigenetic processes can be seen
42
as the interplay between an individual’s environment and molecular biology (Hoare et al.,
43
2020), and refer to the regulation of genome activities and gene expression by molecular
44
modifications on the DNA. Literature shows that aging has an effect on the genome-wide
45
DNA regulations, especially on DNA methylation levels. Therefore, the pattern of DNA
46
methylation can estimate the age of the DNA source, not only reflecting the chronological age
47
but also the biological age. This biological age can be captured in the form of an ‘epigenetic’
48
clock, a tool used to determine biological (i.e. epigenetic) age through determination of the
49
methylation patterns of the DNA (Horvath, 2013; McEwen et al., 2020). To calculate the
50
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
5
epigenetic clock in paediatric samples, the Paediatric Buccal cell Epigenetic (PedBE) clock is
51
used, as it is the most accurate in predicting epigenetic age in children (McEwen et al., 2020).
52
53
Research has suggested that epigenetic processes, especially the methylation of DNA,
54
play a mechanistic role in neurodevelopment and cell differentiation (Moore et al., 2013;
55
Unnikrishnan et al., 2019). Additionally, there is growing evidence for the impact of early life
56
events on methylation (Hoare et al., 2020; Horvath & Raj, 2018; Marini et al., 2020). For
57
examples, the large study on 973 adults by Marini et al. (2020) showed that early life
58
adversity may alter these normal methylation processes, leading to accelerated aging of cells.
59
Consequently, this alteration of methylation patterns can lead to a deviation of the biological
60
age from the chronological age, also known as epigenetic age acceleration (EAA) (Horvath,
61
2013; McEwen et al., 2020)
62
There is a dearth of studies on the association between epigenetic age acceleration and
63
brain morphology. One recent study in young adolescents (N=44) from low income
64
households found accelerated epigenetic age to be associated with alterations in brain
65
morphology. More specifically, accelerated epigenetic age was associated with decreases in
66
regional cortical thickness (Hoare et al., 2020). Another study in 4.5-year-old children
67
(n=158), both with and without maltreatment experiences in the first 6 months of life, found
68
that accelerated epigenetic aging was related to internalizing disorders and exposure to
69
maltreatment (Dammering et al., 2021). Moreover, a study with a sample of 193 community
70
children from the current study, epigenetic age acceleration was shown to be associated with
71
internalizing behaviour, such that internalizing behaviour in 2.5-year old children predicted
72
EAA at age 6 years, which in turn predicted internalizing behaviour at age 10 years (Tollenaar
73
et al., 2021a). Both Dammering et al. (2021) and Hoare et al. (2020) argue that the stress
74
response could underlie the association between accelerated epigenetic aging and brain
75
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
6
morphology and functioning. They hypothesized that adversity early in life triggers a
76
dysregulation of the stress system, which leads to dysregulated production of the stress
77
hormone cortisol. Dammering et al. (2021) showed that epigenetic age acceleration was likely
78
caused by such glucocorticoid dysregulation, as the CpG sites (DNA sequences where methyl
79
groups bind) used for the PedBE clock are highly sensitive to glucocorticoids, which implies
80
that glucocorticoids have an impact on the methylation of the DNA. This stress regulation of
81
the epigenetic age would then cause a more rapid maturation of the brain (Dammering et al.
82
(2021)).
83
84
In this study we will prospectively assess the predictive value of telomere length and
85
epigenetic age acceleration in childhood (age 6 and 10) on grey matter brain volume (GMV)
86
in early adolescence (age 12). At age 12, individual differences in GMV development are
87
likely to be particularly pronounced, as only some children will have reached their peak GM
88
volumes (Bethlehem et al., 2021). Brain regions that continue to develop into middle
89
childhood, namely the amygdala, hippocampus, and prefrontal cortex, are especially
90
vulnerable to effects of early life negative experiences (Romeo, 2017; Tyborowska et al.,
91
2018). Indeed, both cross sectional as well as longitudinal studies have associated early life
92
stress to reductions of these regional GM volumes (Romeo, 2017; Tyborowska et al., 2018).
93
94
In sum, while there is accumulating evidence linking adverse early life events to both
95
cellular aging and brain morphology (Colich et al., 2020; Rebello et al., 2019), it remains
96
unclear whether signs of cellular aging can also be linked to brain morphology. Therefore, the
97
aim of the current preregistered study is to investigate potential associations between two
98
biomarkers of cellular aging (i.e. telomere length and epigenetic age), and brain structure at
99
age 12. Specifically, we will look at telomere length and epigenetic aging at ages 6 and 10,
100
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
7
which reflect the cellular aging differences between individual, as well as at how they change
101
over time, from age 6 to 10, reflecting increased or decreased cellular aging within
102
individuals. Brain structure will be determined in terms of whole-brain GMV (using SPM), as
103
well as a closer investigation of three regions of interest hypothesized to be especially
104
relevant as they have been associated to early life stress (i.e. amygdala, hippocampus, and
105
(pre)frontal cortex -PFC). We hypothesized that shorter telomere lengths and higher
106
epigenetic age will be associated with smaller GMV, on the whole-brain level and particularly
107
for the amygdala, hippocampus, and PFC. Additionally, the analyses mentioned above will be
108
done with white matter volume and different brain sub-regions as outcomes measures.
109
Because of a lack of previous literature, these analyses are exploratory in nature and therefore
110
no specific directional hypotheses have been formulated.
111
112
1 Materials and Methods
113
2.1 Participants and procedure
114
Following recommendations about research practices (Nosek et al., 2018; Wagenmakers et
115
al., 2012), this study was preregistered at AsPredicted:
116
https://aspredicted.org/blind.php?x=kv8cb9 (632397). This study used data from an ongoing
117
longitudinal project, the BIBO study (Basale Invloeden op de Baby Ontwikkeling; Dutch for
118
Basal Influences on Infant Development) project, which aims to investigate the influence of
119
early environmental factors and individual characteristics on child development. The BIBO
120
study originally comprised 193 healthy, community, mother-child dyads, that have been
121
followed since pregnancy (see (Beijers et al., 2011), for information about the original study).
122
Since 80.9% of the mothers attended college or university we consider the sample low-risk.
123
When children were 6 and 10 years old, buccal swabs were collected by researchers, to obtain
124
genetic material (telomere length and DNA methylation) (Asok et al., 2013; Beijers & Daehn, et
125
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
8
al., 2020; Beijers & Hartman, et al., 2020; Drury et al., 2014; McEwen et al., 2020). For the 12-year
126
BIBO collection wave, 159 children were still participating in the study and were invited for
127
an fMRI scan. Children with braces were excluded from participation (N=30), and several
128
children did not participate for other reasons (e.g., too busy, no interest; N=31), resulting in a
129
group of 97 children taking part in the visit. Markers of cellular aging did not differ between
130
the 23% of the children that did not participate and the 77% that did (see Table 1).
131
Table 1. Differences in cellular aging between children not participating in the MRI-scan and
132
children participating in the MRI-scan
133
Mean [SD]
(non-MRI
subsample; 23%)
Mean [SD]
(MRI
subsample;
77%)
t-value
p-value
Telomeres at 6 years
1.12 [0.61]
1.11 [0.53]
-.0.122
.903
Telomeres at 10 years
0.58 [0.33]
0.62 [0.33]
-1.521
.130
Epigenetic age at 6 years
6.72 [0.65]
6.72 [0.67]
-.018
.985
Epigenetic age at 10 years
11.15 [1.00]
11.02 [1.64]
.033
.947
134
A mock scanner was used to familiarize the children with the scanning environment before
135
the MRI session. MRI data of 3 children were excluded from further analysis due to poor
136
quality, resulting in a final study sample of 94 children, of which descriptive sample
137
characteristics and study variables are summarized in Table 2. The children had an average
138
chronological age of 12.71 (SD=0.3) at the MRI visit. The average chronological age at the
139
time of buccal swab collection was 6.09 (SD= 0.24) years, and 10.09 (SD=0.29) years. This
140
study was approved by the ethical committee of the Faculty of Social Sciences of the
141
Radboud University and the local medical ethics committee (CMO region Arnhem –
142
Nijmegen). The children participated voluntarily and gave oral assent, and their parents
143
provided written informed consent prior to participation.
144
Table 2. Overview of the children’s demographic characteristics and descriptive of the (raw)
145
variables
146
N
Mean or %
SD
Range
Gender % Girls
94
45
47.9
-
-
Education of mother %
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
9
Secondary education
College/University
18
77
19
81
-
-
-
-
Age Age wave 6 years
Age wave 10 years
Age wave 12 years
85
92
94
mean age
6.8
10.06
12.71
0.23
0.23
0.30
5.2-7.3
8.9-10.6
12.2-13.5
zBMI Age wave 6 years
Age wave 10 years
Age wave 12 years
82
92
94
mean zBMI
0.1
0.3
-0.2
0.9
0.9
1.1
-1.6-2.7
-1.7-2.7
-2.6-3
Telomere length
Age wave 6 years
Age wave 10 years
Telomere erosion
94
94
94
mean telomere length
1.13
0.66
0.92
0.53
0.33
1.31
0.3-3.4
0-1.7
-4.6-3.6
Epigenetic age
Age wave 6 years
Age wave 10 years
EAA 6
EAA 10
94
94
94
94
mean epigenetic age
6.73
11.02
-0.64
-0.99
0.58
1.26
0.58
1.22
6-9.8
2-14.8
-2.2-0.7
-3.9-5.9
Brain volumes TIV
GMV
WMV
CSF
94
94
94
94
mean volumes
1.46e6
8.05e5
4.38e5
2.13e5
1.2e5
6.4e4
4.9e4
3.7e4
1.2e6-1.7e6
Notes: EAA=epigenetic age acceleration; chronological age – estimated epigenetic age, TIV=
147
total intracranial volume. Telomere lengths are in line with previous papers using telomere
148
length (Beijers et al., 2020a, 2020b).
149
150
2.2 Measures
151
2.2.1 Telomere length.
152
DNA was extracted from buccal epithelial cells collected at age 6 (M=6.09, SD=0.24) and at
153
age 10 (M=10.09, SD=0.29) using QIAamp DNA Mini Kit (Qiagen, Germany), and was
154
quantified using Quant-iT PicoGreen reagent (Thermo Fisher Scientific, Qiagen). A
155
quantitative PCR protocol was used to perform telomere length assays (see Beijers & Daehn et
156
al., (2020a) and Beijers & Hartman et al., (2020b) .
157
Telomere length is operationalized using the formula
, where ET/S is the
158
efficiency of exponential amplification for reactions targeting the telomere single-copy gene
159
respectively, and CqT/S is the cycle at which a given replicate targeting telomeric content or
160
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
10
the single-copy gene reaches the critical threshold of fluorescence quantification. The same
161
threshold was used for all assays (36B4 and telomere). As samples were run in triplicate, the
162
mean telomeric content
and mean genome copy number
across replicates
163
was used for calculating the T/S ratio. The mean was recalculated using two replicates when
164
the replicates deviated from the mean telomeric content or mean genome copy number with
165
more than 1.5% and was considered an outlier. Inter-assay variability was controlled for in
166
line with Beijers & Daehn et al., (2020a) and Beijers & Hartman et al., (2020b).
167
Telomere lengths were corrected for age differences at the time of data collection by
168
creating residuals. These were derived from regressing the telomere lengths of each
169
assessment moment on the child’s age in months at that time point (Beijers & Daehn, et al.,
170
2020; Beijers & Hartman, et al., 2020). Negative residuals indicate accelerated aging, as telomere
171
lengths are shorter than expected, whereas positive residuals indicate slower aging, as
172
telomere lengths are longer than expected (Fig. 1A-C).
173
Furthermore, a measure of telomere erosion between age 6 and 10 years was created
174
(see Beijers & Daehn et al., (2020a) and Beijers & Hartman et al., (2020b). Here, positive values
175
reflect telomere erosion acceleration between age 6 and 10 (Fig.1E). Negative values reflect
176
telomere erosion deceleration between age 6 and 10.
177
178
2.2.2 Epigenetic age acceleration
179
DNA was extracted from buccal cells collected age 6 (M=6.09, SD=0.24) and at age 10
180
(M=10.09, SD=0.29) using QIAamp DNA Mini Kit (Qiagen, Germany). The Infinium
181
MethylationEPIC array (Illumina, USA) was used for the description of the genome wide
182
DNA methylation, necessary for the determination of the epigenetic age. Thereafter, the Minfi
183
package in R was used for signal extraction, data quality, and pre-processing of the raw data.
184
Epigenetic age was calculated using the newly developed Paediatric-Buccal-Epigenetic
185
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
11
(PedBE) clock (McEwen et al., 2020). The PedBE clock is derived from DNA methylation at
186
94 CpGs, sharing 1CpG with the Horvath clock (McEwen et al., 2020).
187
For validation of the age estimation of the PedBE clock in our sample, a Pearson
188
correlation was conducted between estimated epigenetic age and chronological age and found
189
a correlation of r=.225 (p<0.05) at age 6 and a correlation of r=.210 (p<0.05) at age 10. In
190
addition, epigenetic age at age 6 was significantly correlated with epigenetic age at age 10
191
(r=.568, p=<0.01), suggesting that our quantification of epigenetic age is relatively consistent.
192
The epigenetic age acceleration at age 6 and age 10 were operationalized as the residuals from
193
a linear model regressing PedBE-derived estimates of epigenetic age on chronological age in
194
months at the moment of data collection. A positive value reflects higher than expected
195
epigenetic age, thus epigenetic age acceleration (EAA), whereas a negative value reflects
196
lower than expected epigenetic age, thus epigenetic age deceleration (Fig. 1B-D).(McEwen et
197
al., 2020; Tollenaar et al., 2021b).
198
The pace of epigenetic age acceleration between age 6 and age 10 was operationalised
199
as the difference in raw DNA methylation estimates over time (T2-T1) (Wolf et al., 2019).
200
Values greater than 1.0 suggest an accelerated pace of epigenetic aging relative to the
201
chronological aging, while values less than 1.0 suggest slower pace of epigenetic aging
202
relative to chronological aging (Fig. 1F).
203
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
12
204
205
Figure 1. Distribution of telomere length and epigenetic age in the sample. A. The variance
206
of residualized telomere length at age 6. B. The variance of epigenetic age acceleration at age
207
6 C. The variance of residualized telomere length at age 10. D. The variance of epigenetic age
208
acceleration at age 10. E. The variance of telomere erosion between age 6 and 10 years. F.
209
The variance of epigenetic aging pace between age 6 and 10 years. A-D The bluer colours
210
indicate accelerated cellular aging, and the more yellow colours indicate decelerated cellular
211
aging. E+F The more yellow colours indicate faster cellular aging pace, and the bluer colours
212
indicate slower cellular aging pace.
213
Note: The outliers did not affect further analyses
214
215
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
13
2.2.3 Brain data.
216
Brain MRI data was acquired using a 3T MAGNETOM PrismaFit MR scanner (Siemens AG,
217
Healthcare Sector, Erlangen, Germany) with a 32 channel-coil. The children were scanned in
218
the supine position. An MPRAGE sequence (TR = 2300 ms, TE = 3.03 ms, 192 sagittal slices,
219
voxel size = 1.0 x 1.0 x 1.0 mm, FOV = 256 x 256 mm) was used to acquire whole brain T1-
220
weighted images.
221
222
2.3 Data pre-processing
223
2.3.1 Biological data pre-processing.
224
We used the Markov Chain Monte Carlo procedure in SPSS to impute data of 9 children who
225
were missing data from either the 6-year or 10-year measurement waves. The remaining
226
participant was missing data of more than one predictor and was excluded from further
227
analysis.
228
The six biological variables (telomere length at age 6 and 10, telomere erosion
229
between age 6 and 10, epigenetic age at age 6 and 10, and epigenetic pace between age 6 and
230
10) were checked for violations of normality and outliers. Outliers were identified using the
231
Multivariate Mahalanobis Distance (MD). In accordance with Tabachnick & Fiddell (2007),
232
data of one participant was considered as an outlier, as it exceeded the critical chi-square
233
value (degrees of freedom, df=6; the number of predictor variables in the model) at a critical
234
alpha value of .001. This participant was excluded prior to further analysis.
235
Pearson’s correlations between the study variables and zBMI, and buccal cell count
236
were evaluated to check the possibility of the latter two acting as confounding factors. No
237
significant correlations were found between zBMI and the covariates of interest. Therefore,
238
zBMI was not included as a covariate of no interest in further analyses. A significant
239
correlation was found between buccal cell count at age 10 and epigenetic pace between age 6
240
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
14
and 10 (0.577, p=<.001, df= 92). Therefore, a residual of epigenetic pace was calculated using
241
a regression of buccal cell count at age 10 on epigenetic pace.
242
243
2.3.2 MRI data pre-processing.
244
Raw structural T1-weighted images were checked for anatomical abnormalities, movement
245
artefacts, and alignment to the anterior commissure. We performed the following pre-
246
processing steps using Statistical Parametric Mapping 12 (SPM12), which is implemented in
247
MATLAB (version 2019a). First, we used the Diffeomorphic Anatomical Registration
248
Through Exponential Lie (DARTEL) algorithm to segment images into grey matter (GM),
249
white matter (WM), and cerebrospinal fluid (CSF), and inter-subject registration of the GM
250
images to a group average template image. Subsequently, GM images were normalized into
251
Montreal Neurological Institute (MNI) space using a customized paediatric tissue probability
252
map, which was created using the Template-O-Matic (TOM; version 1;
253
http://141.35.69.218/wordpress/software/tom) toolbox. As a last step, grey matter images
254
were smoothed using an 8x8x8 mm FWHM Gaussian kernel. Data quality after pre-
255
processing was checked using the “Check Sample Homogeneity” function of the
256
Computational Anatomy Toolbox 12 (CAT12), which indicated data of six children as
257
potential outliers, of which data of two subjects was excluded from further analysis, after
258
visual inspection. Total Intracranial Volume (TIV) was calculated as the sum of GM, WM,
259
and CSF.
260
261
2.4 Statistical analyses
262
2.4.1 Main analyses.
263
To investigate the associations between accelerated cellular aging at age 6 and whole-brain
264
GMV, a whole-brain multiple regression analysis was performed in SPM12 including the
265
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
15
predictors at age 6, that is telomere length and epigenetic age. Additionally, a second
266
multiple-regression analysis was performed including telomere erosion between 6 and 10
267
years, and epigenetic pace between 6 and 10 years, and whole-brain GMV at age 12 as
268
outcome variable, to investigate the association between the longitudinal changes of the
269
accelerated cellular aging markers and GMV. In both whole-brain multiple regressions, age,
270
sex, and Total Intracranial Volume (TIV) were entered as confounders. To assess whole-brain
271
statistical inference, the Threshold-Free Cluster Enhancement (TFCE) Toolbox in SPM12 was
272
used to perform non-parametric permutation tests. For these permutation tests, a threshold of
273
p<0.05 corrected for family wise error at a whole-brain level was used. The TFCE approach is
274
especially advantageous for VBM data as it aims to enhance spatially contiguous signal
275
without being dependent on threshold-based clustering. TFCE values at each voxel
276
represented both spatially distributed cluster size and height information (Li et al., 2017;
277
Smith & Nichols, 2009). Our ROIs (left and right amygdala, left and right hippocampus, and
278
the PFC) were selected using the marsbar-AAL tool in SPM12 described by Tzourio-Mazoyer
279
et al., (2002) The associations between accelerated cellular aging, both at age 6 as well as
280
between age 6 and 10 years, and three ROIs were analyzed using multiple regression analyses
281
in SPM12.
282
283
2.4.2 Exploratory analyses
284
To exploratorily investigate the associations between accelerated cellular aging at age 10 and
285
whole-brain GMV, a whole-brain multiple regression analysis was performed in SPM12
286
including the biomarkers of cellular aging at age 10, that is telomere length and epigenetic
287
age.
288
Additionally, to exclude the possibility that the biomarkers of accelerated cellular
289
aging are associated with brain regions that were not part of our a priori selection, all main
290
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
16
analyses described above were repeated in an exploratory fashion with the following brain
291
regions as outcome variables: the superior-, middle-, and inferior-gyrus of the frontal lobe, the
292
precentral gyrus, the rectus, the inferior-, middle-, and superior-gyrus of the temporal lobe,
293
the lingual gyrus, the fusiform gyrus, the insula, Heschl’s gyrus, the parahippocampal gyrus,
294
the inferior- and superior-gyrus of the parietal lobe, the supramarginal gyrus, the postcentral
295
gyrus, the precuneus, the inferior-, middle-, and superior-gyrus of the occipital lobe, the
296
calcarine sulcus, the cuneus, the cingulum, the caudate, the putamen, the globus pallidus, and
297
the thalamus. These brain regions were acquired using the SPM12 marsbar AAL-tool, as
298
described by Tzourio-Mazoyer et al., (2002).
299
To explore whether accelerated cellular aging is linked to brain volume through a link
300
with WMV, two multiple regression analyses were performed. The first included the
301
predictors at age 6, that is telomere length and epigenetic age. The second analysis was
302
performed including telomere erosion between 6 and 10 years, and epigenetic pace between 6
303
and 10 years, and WMV at age 12 years as the outcome variable, to investigate the association
304
between the longitudinal changes of the accelerated cellular aging markers and WMV.
305
Lastly, for all multiple regressions Bayesian analyses were performed, to quantify the
306
evidence in favor, or against, the regression models as compared to a null model. Results are
307
expressed as a Bayes factor, which represents the relative likelihood of one model compared
308
to another given the data and a prior expectation. This prior expectation was set as a default,
309
noninformative JZS prior. The BayesFactor package from the open-source software package
310
R was used to compute Bayes Factors (Morey & Rouder, 2015). For the interpretation of the
311
evidential strength the description by Jeffreys (1961) was used, where a Bayes Factor <1/10
312
indicates strong evidence for the null-hypothesis, Bayes Factors >10 indicate strong evidence
313
for H1, and a bayes factor of 0 indicates no evidence for either one (Jeffreys, 1961).
314
315
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
17
2.5 Deviations from the pre-registered study
316
In addition to the analyses described above, the pre-registration of this study describes the
317
BrainAge, a model developed by Franke et al., (2010) that uses whole-brain neuroimaging
318
data to reliably estimate the multidimensional aging pattern into one single value. The goal
319
was to use BrainAge as a variable representing general aging of the brain. Therefore, a
320
BrainAge model was created in R by regressing GMV (M=8.05e5 mm3, SD=6.4e4 mm3),
321
white matter volume (M=4.38e5 mm3, SD=4.9e4 mm3), CSF (M=2.13e5 mm3, SD=3.7e4 mm3),
322
and total intracranial volume (M=1.46e6 mm3, SD=1.2e5 mm3) onto chronological age. The
323
BrainAge per individual was operationalised as the residual of this regression, in which
324
positive values indicate older brains relative to the model-predicted age in this sample,
325
whereas negative values indicate younger brains relative to the model-predicted age for an
326
individual. As the variance in chronological age was relatively small in our sample, there was
327
no correlation between GMV and age (.085). Therefore, the predicted BrainAge and GMV
328
were significantly correlated (.482**). For this reason, BrainAge was excluded from further
329
analyses.
330
331
3.Results
332
3.1 Descriptive analyses
333
Descriptive statistics are presented in Table 2 (untransformed data). Epigenetic age at
334
age 6-years (M=7.76, SD= 0.69) and at age 10 (M= 12.44 SD=0.29) were significantly higher
335
than the chronological age (t=-21.3 p=.001; t=-14.0 p=.001, respectively). Telomere length at
336
6 years of age (M= 1.11, SD= 0.56) and at 10 years of age (M= 0.62, SD=0.33) did not differ
337
from each other (t=-.44, p=0.658), suggesting that between the ages 6 and 10 telomere length
338
did not change in the same way for the group as a whole. Telomere length at age 6 and the
339
pace of change were significantly correlated (r= -.664 p<0.01), which suggests that shorter
340
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
18
lengths of the telomere at age 6 predict a higher pace of telomere erosion (see Supplemental
341
Figure 3).
342
Table 3 shows the Pearson correlations between the study variables. Total Intracranial
343
Volume (TIV) was significantly correlated with gender but insignificantly correlated with
344
telomere length and epigenetic age. Interestingly, telomere length and epigenetic age were not
345
correlated.
346
347
Table 3. Correlogram representing the Pearson correlations between all study variables.
348
Colors indicate different values of the correlation coefficient. The size of the circle is
349
proportional to the correlation coefficients.
350
351
352
Notes: TIV = total intracranial volume.
353
354
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
19
3.2 Main analyses
355
The analyses described under 3.2.1 to 3.3.2 were all carried out using the TFCE threshold of
356
p<0.05 corrected for Family Wise error (FWE).
357
358
3.2.1 Biomarkers of cellular aging at age 6 years.
359
The whole-brain VBM analysis with telomere lengths and epigenetic age at 6 years did not
360
yield any significant associations with whole-brain GMV (all corrected p-values above
361
0.998), nor with subregions of the brain thought to be associated with early life adversity, i.e.,
362
the amygdala, hippocampus, and PFC.
363
364
3.2.2 Biomarkers of cellular aging at age 10 years
365
With respect to telomere lengths and epigenetic age at 10 years of age, the whole-brain VBM
366
analysis did not yield any significant associations with these biomarkers and whole-brain
367
GMV (all corrected p-values above 0.998), nor with subregions of the brain thought to be
368
associated with early life adversity, i.e., the amygdala, hippocampus, and PFC.
369
370
3.2.3 Changes in biomarkers of cellular age between 6 and 10 years of age.
371
Regarding the changes in telomere length and epigenetic age between the age of 6 and 10, the
372
whole-brain VBM analysis did not yield any significant associations with whole-brain GMV
373
(all corrected p-values above 0.998), nor with the three subregions of interest (no
374
suprathreshold values were found).
375
376
3.3 Exploratory analyses
377
3.3.1 Exploratory brain regions
378
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
20
All the main analyses described above were repeated in an exploratory fashion with the brain
379
regions described in the method section, acquired using the SPM12 marsbar AAL-tool
380
(described by: Tzourio-Mazoyer et al. (2002)) as outcome variables. These analyses, using a
381
threshold of p<0.05 corrected for Family Wise error at whole-brain level, also did not find
382
significant associations.
383
384
3.3.2 White matter volumes
385
Regarding both the telomere lengths and epigenetic age at 6 years of age, as well as the
386
longitudinal changes in telomere length and epigenetic age between the age of 6 and 10,
387
multiple regression analyses did not yield any significant associations with WMV (t=-1.312,
388
p=0.193 and t=-1.148, p=0.254 respectively).
389
390
3.3.3 Bayesian analyses
391
To be able to quantify whether the frequentist absence of effects was indeed evidence in favor
392
of the null hypothesis, rather than an absence of precision or evidence either way, the main
393
analyses were tested using Bayesian analyses. These analyses yielded Bayes Factors
394
indicating a moderate to strong evidence for the null hypothesis (see Table 4). The full model
395
with the predictors, telomere length and epigenetic age at age 6 was approximately 70
396
(bf=69.8) times less likely than the model including only the covariates of no interest (age,
397
TIV, and gender), which suggests that evidence was found that indicates that it is unlikely that
398
telomere length and epigenetic age predict GMV. Similarly, the full model with the predictors
399
at age 10 was approximately 70 (bf=68.3) times less likely than the model including only the
400
covariates. For both models including the predictors at age 6 or 10 years, the Bayes Factor
401
indicated moderate evidence for the null hypothesis (bf= 0.172, and bf= 0.120 respectively).
402
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
21
The model best predicting GMV, included only TIV (see Supplementary Material for all
403
possible models).
404
405
Table 4. Bayes factor for selected models
406
Telomere length and epigenetic age at age 6
Bayes factor
Full model
5.888e+31
Telomere length + epigenetic age
.172
Telomere length
.361
Epigenetic age
.373
TIV, age, gender
4.111e+33
Telomere length and epigenetic age at age 10
Full model
7.017e+31
Telomere length + epigenetic age
.070
Telomere length
.217
Epigenetic age
.217
TIV, age, gender
4.111e+33
Telomere erosion and epigenetic pace between 6 and 10
Full model
6.0239e+31
Telomere erosion + epigenetic pace
.120
Telomere erosion
.284
Epigenetic pace
.270
TIV, age, gender
4.111e+33
Notes: Full model= GMV~Telomere length+epigenetic age+TIV+ age+gender. GMV= grey
407
matter volume, TIV= Intracranial Volume, age= age at 12 year measurement round.
408
409
4.Discussion
410
The current study investigated whether markers of cellular aging assessed in middle
411
childhood, namely telomere length and epigenetic age acceleration, were associated with
412
brain morphology in early adolescence in a low-risk sample. We hypothesized that shorter
413
telomere length and higher epigenetic age acceleration at age 6 and 10 would be associated
414
with smaller whole-brain grey matter, particularly in the three regions of interest (i.e.
415
amygdala, hippocampus, and PFC) at age 12. Contrary to our expectations, we found no
416
evidence for associations between the cellular aging markers and brain structure. These results
417
were supported by exploratory Bayesian analyses, revealing Bayes Factors indicating
418
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
22
moderate to strong evidence for the null findings. Finally, exploratory analyses inspecting
419
associations between the two markers of cellular aging and both white matter as well as other
420
subregions of the brain, not previously described in relation to the topic, delivered null
421
findings as well.
422
One explanation for these findings points to the fact that the neural effects were tested
423
two years after measuring cellular aging. Potentially, the associations between cellular aging
424
and brain morphometry are only short-lived. The brain develops rapidly and is known to be
425
vulnerable to environmental factors, particularly during its development in adolescence. It is
426
possible that cellular aging processes at ages 6 and 10 did lead to short-term changes in the
427
brain but that because of the brain’s plasticity at these ages, potential temporary impacts of
428
cellular aging (f.i. through inflammation or glucocorticoid dysregulation) were reversed in the
429
subsequent months or years. Such reversal mechanisms may particularly function in
430
community samples in which the levels of stress, and hence the impact on cellular aging
431
processes, are not as high as in high-risk samples where, in turn, effects on the brain may be
432
less reversible and more cumulative in nature. To clarify these issues, longitudinal studies
433
investigating both cellular aging as well as brain morphology over time are needed.
434
Alternatively, it is possible that the effects of cellular aging processes on brain development
435
may be more permanent, as they could possibly be considered programming effects.
436
Accordingly, a possible alternative explanation for our null-results could be that the period
437
from 6-10 years may be a period in which stress has less impact on the brain. Potentially,
438
stress early in life may impact cellular aging which in turn affects brain developmental
439
trajectories.
440
Another explanation for our null-results is related to sensitivity of group versus
441
individual analyses; group-level analyses of alterations in brain structure may be less sensitive
442
than analyses that account for individual brain development. Early developmental studies
443
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
23
focused on deterministic models of brain development, assuming brain development proceeds
444
via a prescribed blueprint that is innately specified in all individuals. Genome-wide
445
association studies have shown that volumetric brain changes are heritable, and associated
446
with substantial variability in brain volumes across different children of the same age (Brown,
447
2017). Besides, developmental studies show a link between puberty onset, ranging between
448
the ages of 8 and 14, and brain maturation, such that higher pubertal developmental scores are
449
associated with more mature brain development (Beck et al., 2023; Dehestani et al., 2023;
450
Holm et al., 2023). Potentially, the (small) effects of cellular aging on the maturation of the
451
brain are masked by the greater effect of inter-individual variation on brain maturation.
452
However, due to the current study design, we cannot disentangle faster rates of maturation
453
(different starting points, same endpoint) from discrete volumetric differences during early
454
adolescence. Further longitudinal studies investigating the effects of cellular aging on
455
individual brain development trajectories are therefore needed.
456
A final possible explanation for our null results could be that the associations might be
457
very weak in a low-risk sample such as the one in this study, and we hence may not have had
458
enough power to detect them with a sample of 94 adolescents. Indeed, Rebello et al. (2020)
459
found only marginal relations between telomere length and brain connectivity in a study
460
sample of 389 individuals, which did not survive strict Bonferonni corrections. A power
461
analysis suggests that with a sample size of 94 children, an alpha of 0.05, and a power of 0.8,
462
an effect size of f2 of 0.067 with a critical t= 1.66 can be found, which is in the moderate to
463
large range for individual differences (Gignac & Szodorai, 2016). However, the results of the
464
Bayesian analyses indicate a moderate to strong evidence for acknowledging the null-
465
hypothesis given our uninformative prior. Together, the results of our study suggest that
466
although our sample size is modest, the associations between middle childhood cellular aging
467
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
24
and early adolescent brain morphology are, if they do exist, likely not particularly large in
468
community children.
469
470
4.1 Strengths and limitations
471
This pre-registered study has several strengths, including the longitudinal design and
472
measurement of two markers of cellular aging, namely telomere length and epigenetic age, of
473
which epigenetic age was determined with the PedBE clock, a new model specifically
474
developed for children (McEwen et al., 2020). T1 images were registered to a pediatric
475
specific standard space, allowing for a more accurate registration. Using Bayesian analyses,
476
we could symmetrically quantify evidence in favor of the absence of associations in our
477
sample. However, some limitations should also be acknowledged. First, cellular aging was
478
measured at ages 6 and 10, while brain morphology was measured at age 12 years. The lack
479
of cellular aging measures at age 12 could be considered a limitation, as it was not possible to
480
account for potential contemporary associations. A second potential limitation is that out of
481
the potential pool of 128 children that were eligible to participate in the 12-year measurement,
482
around 23% declined participation. However, because the cellular aging markers did not
483
differ between participating and non-participating children, this does not appear to be a
484
limitation that could have influenced the results. A final limitation is that while the markers
485
for cellular aging were assessed at two childhood ages, the whole-brain GMV were only
486
measured at 12 years, meaning we could not relate accelerated cellular aging to changes in
487
brain structure over time.
488
489
5.Conclusion
490
In conclusion, we found no significant associations between childhood cellular aging (at 6 and
491
10 years) and adolescent brain morphology. Exploratory Bayesian analyses indicated
492
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
25
moderate to strong evidence for the null-findings. These results point at the lack of a strong
493
relation between markers of cellular aging and brain volume during childhood. Future studies
494
might benefit from a longitudinal study design with cellular aging measures in early
495
development (younger ages), biological and brain measures at the same age, as well as
496
individual as opposed to group-level brain development trajectories.
497
498
6. Acknowledgements
499
CRediT roles:
500
- Conceptualization
501
- Data curation
502
- Formal analysis
503
- Funding acquisition
504
- Investigation
505
- Methodology
506
- Project administration
507
- Resources
508
- Software
509
- Supervision
510
- Validation
511
- Visualization
512
- Roles/Writing - original draft
513
- Writing - review & editing.
514
-
515
EB – Conceptualization, Formal analysis, Investigation, Roles/Writing – original draft
516
RB – Funding acquisition (BIBO), Data curation, Writing – review & editing, Data curation
517
AT – Methodology, MRI data acquisition, Data curation, Writing – review & editing
518
KR – Methodology, supervision MRI data acquisition, Writing – review & editing
519
SK – Methodology, Writing – review & editing
520
RK – Methodology, Investigation, Supervision, Roles/Writing – original draft
521
CW – Project administration, Conceptualization, Funding acquisition, Supervision,
522
Investigation, Roles/Writing – original draft
523
524
7. Data + Code statement
525
The data of this study is part of an ongoing longitudinal study of which the data is still
526
acquired and analyzed. Therefore, it cannot be made openly available in a public repository.
527
Moreover, the parents of the participating children signed an informed consent that did not
528
include the possibility of openly available data. However, for research purposes such as meta-
529
analyses, it is possible to request the anonymized data by contacting: dr. Carolina de Weerth
530
using a formal data sharing agreement where the goals of the project are outlined and the data
531
transfer and potential co-authorships are described.
532
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
26
The analyses of this study were performed on software openly accessible, namely R
533
(Download the RStudio IDE - RStudio) and SPM12 (SPM12 Software - Statistical Parametric
534
Mapping (ucl.ac.uk)).
535
536
8. Declaration of Interest
537
Declarations of interest: none
538
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
27
6.References
Asok, A., Bernard, K., Roth, T. L., Rosen, J. B., & Dozier, M. (2013). Parental
539
responsiveness moderates the association between early-life stress and reduced
540
telomere length. Development and Psychopathology, 25(3).
541
https://doi.org/10.1017/S0954579413000011
542
Beck, D., Ferschmann, L., MacSweeney, N., Norbom, L. B., Wiker, T., Aksnes, E., Karl,
543
V., Dégeilh, F., Holm, M., Mills, K. L., Andreassen, O. A., Agartz, I., Westlye, L.
544
T., von Soest, T., & Tamnes, C. K. (2023). Puberty differentially predicts brain
545
maturation in male and female youth: A longitudinal ABCD Study. Developmental
546
Cognitive Neuroscience, 61, 101261. https://doi.org/10.1016/j.dcn.2023.101261
547
Beijers, R., Daehn, D., Shalev, I., Belsky, J., & de Weerth, C. (2020). Biological
548
embedding of maternal postpartum depressive symptoms: The potential role of
549
cortisol and telomere length. Biological Psychology, 150.
550
https://doi.org/10.1016/j.biopsycho.2019.107809
551
Beijers, R., Hartman, S., Shalev, I., Hastings, W., Mattern, B. C., de Weerth, C., &
552
Belsky, J. (2020). Testing three hypotheses about effects of sensitive–insensitive
553
parenting on telomeres. Developmental Psychology, 56(2).
554
https://doi.org/10.1037/dev0000879
555
Beijers, R., Jansen, J., Riksen-Walraven, M., & de Weerth, C. (2011). Attachment and
556
Infant Night Waking: A Longitudinal Study From Birth Through the First Year of
557
Life. Journal of Developmental & Behavioral Pediatrics, 32(9), 635–643.
558
https://doi.org/10.1097/DBP.0b013e318228888d
559
Bethlehem, R. A. I., Seidlitz, J., White, S. R., Vogel, J. W., Anderson, K. M., &
560
Adamson C. (2021). Brain charts for the human lifespan.
561
Brown, T. T. (2017). Individual differences in human brain development. WIREs
562
Cognitive Science, 8(1–2). https://doi.org/10.1002/wcs.1389
563
Colich, N. L., Rosen, M. L., Williams, E. S., & McLaughlin, K. A. (2020). Biological
564
aging in childhood and adolescence following experiences of threat and deprivation:
565
A systematic review and meta-analysis. Psychological Bulletin, 146(9).
566
https://doi.org/10.1037/bul0000270
567
Dammering, F., Martins, J., Dittrich, K., Czamara, D., Rex-Haffner, M., Overfeld, J., de
568
Punder, K., Buss, C., Entringer, S., Winter, S. M., Binder, E. B., & Heim, C.
569
(2021). The pediatric buccal epigenetic clock identifies significant ageing
570
acceleration in children with internalizing disorder and maltreatment exposure.
571
Neurobiology of Stress, 15, 100394. https://doi.org/10.1016/j.ynstr.2021.100394
572
Darrow, S. M., Verhoeven, J. E., Révész, D., Lindqvist, D., Penninx, B. W. J. H.,
573
Delucchi, K. L., Wolkowitz, O. M., & Mathews, C. A. (2016). The Association
574
Between Psychiatric Disorders and Telomere Length: A Meta-Analysis Involving
575
14,827 Persons. Psychosomatic Medicine, 78(7), 776–787.
576
https://doi.org/10.1097/PSY.0000000000000356
577
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
28
Davis, K., Hirsch, E., Gee, D., Andover, M., & Roy, A. K. (2022). Mediating role of the
578
default mode network on parental acceptance/warmth and psychopathology in
579
youth. Brain Imaging and Behavior, 16(5), 2229–2238.
580
https://doi.org/10.1007/s11682-022-00692-z
581
Dehestani, N., Whittle, S., Vijayakumar, N., & Silk, T. J. (2023). Developmental brain
582
changes during puberty and associations with mental health problems.
583
Developmental Cognitive Neuroscience, 60, 101227.
584
https://doi.org/10.1016/j.dcn.2023.101227
585
Drury, S. S., Mabile, E., Brett, Z. H., Esteves, K., Jones, E., Shirtcliff, E. A., & Theall,
586
K. P. (2014). The Association of Telomere Length With Family Violence and
587
Disruption. PEDIATRICS, 134(1). https://doi.org/10.1542/peds.2013-3415
588
Dufford, A. J., Bianco, H., & Kim, P. (2019). Socioeconomic disadvantage, brain
589
morphometry, and attentional bias to threat in middle childhood. Cognitive,
590
Affective, & Behavioral Neuroscience, 19(2), 309–326.
591
https://doi.org/10.3758/s13415-018-00670-3
592
el Marroun, H., Tiemeier, H., Muetzel, R. L., Thijssen, S., van der Knaap, N. J. F.,
593
Jaddoe, V. W. v., Fernández, G., Verhulst, F. C., & White, T. J. H. (2016).
594
PRENATAL EXPOSURE TO MATERNAL AND PATERNAL DEPRESSIVE
595
SYMPTOMS AND BRAIN MORPHOLOGY: A POPULATION-BASED
596
PROSPECTIVE NEUROIMAGING STUDY IN YOUNG CHILDREN.
597
Depression and Anxiety, 33(7). https://doi.org/10.1002/da.22524
598
Farina, B., Della Marca, G., Maestoso, G., Amoroso, N., Valenti, E. M., Carbone, G. A.,
599
Massullo, C., Contardi, A., & Imperatori, C. (2018). The Association among
600
Default Mode Network Functional Connectivity, Mentalization, and
601
Psychopathology in a Nonclinical Sample: An eLORETA Study. Psychopathology,
602
51(1), 16–23. https://doi.org/10.1159/000485517
603
Franke, K., Ziegler, G., Klöppel, S., & Gaser, C. (2010). Estimating the age of healthy
604
subjects from T1-weighted MRI scans using kernel methods: Exploring the
605
influence of various parameters. NeuroImage, 50(3).
606
https://doi.org/10.1016/j.neuroimage.2010.01.005
607
Ghetti, S., & Bunge, S. A. (2012). Neural changes underlying the development of
608
episodic memory during middle childhood. Developmental Cognitive Neuroscience,
609
2(4), 381–395. https://doi.org/10.1016/j.dcn.2012.05.002
610
Gignac, G. E., & Szodorai, E. T. (2016). Effect size guidelines for individual differences
611
researchers. Personality and Individual Differences, 102, 74–78.
612
https://doi.org/10.1016/j.paid.2016.06.069
613
Hanson, J. L., Chung, M. K., Avants, B. B., Shirtcliff, E. A., Gee, J. C., Davidson, R. J.,
614
& Pollak, S. D. (2010). Early Stress Is Associated with Alterations in the
615
Orbitofrontal Cortex: A Tensor-Based Morphometry Investigation of Brain
616
Structure and Behavioral Risk. Journal of Neuroscience, 30(22), 7466–7472.
617
https://doi.org/10.1523/JNEUROSCI.0859-10.2010
618
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
29
Harley, C. B., Vaziri, H., Counter, C. M., & Allsopp, R. C. (1992). The telomere
619
hypothesis of cellular aging. Experimental Gerontology, 27(4), 375–382.
620
https://doi.org/10.1016/0531-5565(92)90068-B
621
Hoare, J., Stein, D. J., Heany, S. J., Fouche, J.-P., Phillips, N., Er, S., Myer, L., Zar, H.
622
J., Horvath, S., & Levine, A. J. (2020). Accelerated epigenetic aging in adolescents
623
from low-income households is associated with altered development of brain
624
structures. Metabolic Brain Disease, 35(8), 1287–1298.
625
https://doi.org/10.1007/s11011-020-00589-0
626
Holm, M. C., Leonardsen, E. H., Beck, D., Dahl, A., Kjelkenes, R., de Lange, A.-M. G.,
627
& Westlye, L. T. (2023). Linking brain maturation and puberty during early
628
adolescence using longitudinal brain age prediction in the ABCD cohort.
629
Developmental Cognitive Neuroscience, 60, 101220.
630
https://doi.org/10.1016/j.dcn.2023.101220
631
Horvath, S. (2013). DNA methylation age of human tissues and cell types. Genome
632
Biology, 14(10), R115. https://doi.org/10.1186/gb-2013-14-10-r115
633
Horvath, S., & Raj, K. (2018). DNA methylation-based biomarkers and the epigenetic
634
clock theory of ageing. Nature Reviews Genetics, 19(6), 371–384.
635
https://doi.org/10.1038/s41576-018-0004-3
636
Jeffreys, H. (1961). The theory of probability (third). Oxford University Press.
637
Li, H., Nickerson, L. D., Nichols, T. E., & Gao, J.-H. (2017). Comparison of a non-
638
stationary voxelation-corrected cluster-size test with TFCE for group-Level MRI
639
inference. Human Brain Mapping, 38(3). https://doi.org/10.1002/hbm.23453
640
Marini, S., Davis, K. A., Soare, T. W., Zhu, Y., Suderman, M. J., Simpkin, A. J., Smith,
641
A. D. A. C., Wolf, E. J., Relton, C. L., & Dunn, E. C. (2020). Adversity exposure
642
during sensitive periods predicts accelerated epigenetic aging in children.
643
Psychoneuroendocrinology, 113. https://doi.org/10.1016/j.psyneuen.2019.104484
644
McEwen, L. M., O’Donnell, K. J., McGill, M. G., Edgar, R. D., Jones, M. J., MacIsaac,
645
J. L., Lin, D. T. S., Ramadori, K., Morin, A., Gladish, N., Garg, E., Unternaehrer,
646
E., Pokhvisneva, I., Karnani, N., Kee, M. Z. L., Klengel, T., Adler, N. E., Barr, R.
647
G., Letourneau, N., … Kobor, M. S. (2020). The PedBE clock accurately estimates
648
DNA methylation age in pediatric buccal cells. Proceedings of the National
649
Academy of Sciences, 117(38). https://doi.org/10.1073/pnas.1820843116
650
Morey, R. D., & Rouder, J. N. (2015). BayesFactor CRAN.
651
Nosek, B. A., Ebersole, C. R., DeHaven, A. C., & Mellor, D. T. (2018). The
652
preregistration revolution. Proceedings of the National Academy of Sciences,
653
115(11), 2600–2606. https://doi.org/10.1073/pnas.1708274114
654
Rebello, K., Moura, L. M., Xavier, G., Spindola, L. M., Carvalho, C. M., Hoexter, M.
655
Q., Gadelha, A., Picon, F., Pan, P. M., Zugman, A., Grassi-Oliveira, R., Brietzke,
656
E., Belangero, S. I., Salum, G. A., Rohde, L. A., Miguel, E. C., Bressan, R.,
657
Jackowski, A., & Sato, J. R. (2019). Association between spontaneous activity of
658
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
30
the default mode network hubs and leukocyte telomere length in late childhood and
659
early adolescence. Journal of Psychosomatic Research, 127, 109864.
660
https://doi.org/10.1016/j.jpsychores.2019.109864
661
Romeo, R. D. (2017). The impact of stress on the structure of the adolescent brain:
662
Implications for adolescent mental health. Brain Research, 1654, 185–191.
663
https://doi.org/10.1016/j.brainres.2016.03.021
664
Sato, J. R., Biazoli, C. E., Salum, G. A., Gadelha, A., Crossley, N., Satterthwaite, T. D.,
665
Vieira, G., Zugman, A., Picon, F. A., Pan, P. M., Hoexter, M. Q., Anés, M., Moura,
666
L. M., Del’aquilla, M. A. G., Amaro, E., McGuire, P., Lacerda, A. L. T., Rohde, L.
667
A., Miguel, E. C., … Bressan, R. A. (2015). Temporal stability of network
668
centrality in control and default mode networks: Specific associations with
669
externalizing psychopathology in children and adolescents. Human Brain Mapping,
670
36(12), 4926–4937. https://doi.org/10.1002/hbm.22985
671
Sikora, E., Bielak-Zmijewska, A., Dudkowska, M., Krzystyniak, A., Mosieniak, G.,
672
Wesierska, M., & Wlodarczyk, J. (2021). Cellular Senescence in Brain Aging.
673
Frontiers in Aging Neuroscience, 13. https://doi.org/10.3389/fnagi.2021.646924
674
SMITH, S., & NICHOLS, T. (2009). Threshold-free cluster enhancement: Addressing
675
problems of smoothing, threshold dependence and localisation in cluster inference.
676
NeuroImage, 44(1). https://doi.org/10.1016/j.neuroimage.2008.03.061
677
Stewart, J. A., Chaiken, M. F., Wang, F., & Price, C. M. (2012). Maintaining the end:
678
Roles of telomere proteins in end-protection, telomere replication and length
679
regulation. Mutation Research/Fundamental and Molecular Mechanisms of
680
Mutagenesis, 730(1–2), 12–19. https://doi.org/10.1016/j.mrfmmm.2011.08.011
681
Tollenaar, M. S., Beijers, R., Garg, E., Nguyen, T. T. T., Lin, D. T. S., MacIsaac, J. L.,
682
Shalev, I., Kobor, M. S., Meaney, M. J., O’Donnell, K. J., & de Weerth, C. (2021a).
683
Internalizing symptoms associate with the pace of epigenetic aging in childhood.
684
Biological Psychology, 159, 108021.
685
https://doi.org/10.1016/j.biopsycho.2021.108021
686
Tollenaar, M. S., Beijers, R., Garg, E., Nguyen, T. T. T., Lin, D. T. S., MacIsaac, J. L.,
687
Shalev, I., Kobor, M. S., Meaney, M. J., O’Donnell, K. J., & de Weerth, C. (2021b).
688
Internalizing symptoms associate with the pace of epigenetic aging in childhood.
689
Biological Psychology, 159. https://doi.org/10.1016/j.biopsycho.2021.108021
690
Turner, K., Vasu, V., & Griffin, D. (2019). Telomere Biology and Human Phenotype.
691
Cells, 8(1). https://doi.org/10.3390/cells8010073
692
Tyborowska, A., Volman, I., Niermann, H. C. M., Pouwels, J. L., Smeekens, S.,
693
Cillessen, A. H. N., Toni, I., & Roelofs, K. (2018). Early-life and pubertal stress
694
differentially modulate grey matter development in human adolescents. Scientific
695
Reports, 8(1), 9201. https://doi.org/10.1038/s41598-018-27439-5
696
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix,
697
N., Mazoyer, B., & Joliot, M. (2002). Automated Anatomical Labeling of
698
Activations in SPM Using a Macroscopic Anatomical Parcellation of the MNI MRI
699
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
31
Single-Subject Brain. NeuroImage, 15(1), 273–289.
700
https://doi.org/10.1006/nimg.2001.0978
701
Unnikrishnan, A., Freeman, W. M., Jackson, J., Wren, J. D., Porter, H., & Richardson,
702
A. (2019). The role of DNA methylation in epigenetics of aging. Pharmacology &
703
Therapeutics, 195. https://doi.org/10.1016/j.pharmthera.2018.11.001
704
Vaiserman, A., & Krasnienkov, D. (2021). Telomere Length as a Marker of Biological
705
Age: State-of-the-Art, Open Issues, and Future Perspectives. Frontiers in Genetics,
706
11. https://doi.org/10.3389/fgene.2020.630186
707
Wagenmakers, E.-J., Wetzels, R., Borsboom, D., van der Maas, H. L. J., & Kievit, R. A.
708
(2012). An Agenda for Purely Confirmatory Research. Perspectives on
709
Psychological Science, 7(6), 632–638. https://doi.org/10.1177/1745691612463078
710
Wolf, E. J., Logue, M. W., Morrison, F. G., Wilcox, E. S., Stone, A., Schichman, S. A.,
711
McGlinchey, R. E., Milberg, W. P., & Miller, M. W. (2019). Posttraumatic
712
psychopathology and the pace of the epigenetic clock: a longitudinal investigation.
713
Psychological Medicine, 49(5). https://doi.org/10.1017/S0033291718001411
714
715
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
32
7. Supplementary Material
716
717
Supplementary Figure 1. Bayesian analysis for all possible models predicting GMV at age 6,
718
with the upper model presenting the model best predicting GMV and the lowest model
719
presenting the worst model. Note: TIV= Total Intracranial Volume, AGE= age at MRI
720
measurement round, Tel6 = telomere length at age 6, Epi6 = epigenetic age at age 6. Explain
721
what the colours of the rows mean.
722
723
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
33
724
Supplementary Figure 2. Bayesian analysis for all possible models predicting GMV at age
725
10, with the upper model presenting the model best predicting GMV and the lowest model
726
presenting the worst model. Note: TIV= Total Intracranial Volume, AGE= age at MRI
727
measurement round, Tel10 = telomere erosion between age 6 and 10, Epi10 = epigenetic pace
728
between age 6 and 10. Explain colours.
729
730
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
34
731
Supplementary figure 3. Difference in telomere length between age 6 and 10 years per
732
individual. Individuals with shorter telomeres at age 6 than age 10 are depicted in dark blue,
733
and individuals with longer telomeres at age 6 than age 10 are depicted in light blue.
734
.CC-BY 4.0 International licensemade available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
The copyright holder for this preprintthis version posted August 22, 2023. ; https://doi.org/10.1101/2023.08.18.553696doi: bioRxiv preprint
