Abstract
Background: Psychosis is one of the leading causes of disability
worldwide. Individuals with early-onset psychosis (EOP) tend to
experience a worse prognosis and shorter life expectancy. The etiology
of EOP remains unclear, but epigenetic mechanisms are known to serve as
the interface between environmental exposures and biological processes
to better understand its etiology. Objectives: We characterized the
sociodemographic and clinical characteristics, as well as genome-wide
epigenetic markers, in Mexican patients with EOP. Methods: We estimated
epigenetic age, performed an epigenome-wide association study, and
finally developed an epigenetic risk score (MRS) to predict
manifestations of psychosis. Results: We found that patients with EOP
have a higher epigenetic age using Wu’s clock (p = 0.015). Moreover,
accelerated epigenetic age was correlated with chronological age (PedBE
clock, p = 0.046), global functioning (Wu’s clock, p = 0.027), and
psychiatric admissions (DNAmTL, p = 0.038). In addition, we observed
that a reduction in years of schooling is associated with an increase
on epigenetic age (Levine’s clock, β = 5.07, p = 0.001). In our
epigenome-wide association study, we identified eight CpGs associated
with EOP. Noteworthy, a psychosis-methylation risk score (EOP-MRS) was
associated with panic disorder (β = 1.36, p = 0.03), as well as
auditory (β = 1.28, p = 0.04) and visual (β = 1.22, p = 0.04)
hallucinations. Conclusions: Years of education have an impact on
epigenetic age. Additionally, our study suggests associations of DNA
methylation with EOP. Finally, we developed an MRS that associates
clinical manifestations of psychosis.
Keywords: early-onset psychosis, epigenome-wide association study,
methylation risk score, Mexico, adolescents
1. Introduction
Early-onset psychosis (EOP) is a mental disorder characterized by the
onset of psychosis before the age of 18 years [[52]1]. Compared with
adult-onset psychosis, those affected by EOP have shorter life
expectancy, and poor treatment response [[53]2,[54]3]. The prevalence
of psychotic symptoms has been reported to be higher in adolescents
than in adults, with an estimated of 8–17% in children and adolescents
[[55]4,[56]5,[57]6], but there still need more evidence to support
this.
The etiology of EOP is still unknown. Authors have proposed that
abnormal neurodevelopment and early neurodegeneration explain the
emergence of psychosis in the young population [[58]7,[59]8].
Furthermore, the literature suggests that environmental factors such as
bullying, cannabis use, tobacco use, low birth weight, and childhood
trauma interact with biological factors in the development of psychosis
[[60]3,[61]9,[62]10,[63]11,[64]12]. On the other hand, DNA methylation
(DNAm), proposed to be a mediator between environmental exposures and
biological effects, is the most studied epigenetic mechanism and could
provide a better understanding of biological mechanisms underlying
psychotic disorders [[65]13]. New evidence from epigenome-wide
association studies (EWASs), a comparison of DNAm sites across the
genome, revealed associations with schizophrenia and first episode of
psychosis (FEP) [[66]14,[67]15]. Additionally, DNAm is associated with
psychotic symptoms in adults and the risk of neuropsychiatric disorders
during childhood [[68]8,[69]16]. However, an EWAS of clinically defined
EOP has not been performed.
The development of novel biomarkers derived from DNAm provides a new
approach to understanding disease risk and biological and etiological
mechanisms, such as aging [[70]17,[71]18,[72]19]. For example,
epigenetic clocks are excellent biomarkers to estimate biological age,
also referred to as epigenetic age [[73]7]. To date, three studies have
shown evidence that epigenetic age correlates with the severity of
psychosis, and accelerated biological age is associated with psychotic
disorders [[74]20,[75]21,[76]22,[77]23]. Furthermore, the development
of methylation risk scores (MRSs), representing the sum of an
individual epigenetic risks derived from EWAS results, a similar
construct of polygenic risk score, have been associated with
schizophrenia [[78]24], FEP [[79]17], and neuroimaging changes in
individuals with psychosis [[80]8,[81]25]. Studies using MRS showed
that it could be used to differentiate individuals affected by
psychosis and mediation effects of childhood adversity to develop
psychosis risk [[82]15]. It is noteworthy that, like in many biomedical
research areas, there is still a lack of diversity in EWASs [[83]26].
The current understanding of EOP is still limited and there is scare
evidence of epigenetic biomarkers, to advance in these field, the
current study aimed to characterize for the first time sociodemographic
and clinical characteristics, with a further comprehensive evaluation
of genome-wide epigenetic markers (including 11 epigenetic clocks and
EOP-MRS) in Mexican children and adolescents with EOP.
2. Results
2.1. Sample Description
We deeply characterized clinical and sociodemographic features of a
total of 23 children and adolescents, including 12 psychiatric patients
with psychotic symptoms (EOP group) and 11 psychiatric patients without
psychotic symptoms (non-EOP group). We observed that the EOP group was
older (mean = 15.00, p = 0.030) and had higher years of education (mean
= 9.3, p = 0.023) ([84]Table 1), nevertheless was close to the Mexican
population mean (9.73 years) [[85]27]. Furthermore, patients with EOP
had higher hospital psychiatric admissions (p = 0.027) and higher
prevalence of anxiety and stress disorders (p = 0.036). The EOP group
had a lower functional score (p = 0.00003), and severe GAF score
(global assessment of functioning) (p = 0.017) ([86]Table 1). Our
results shows that EOP had a higher comorbidity and lower
functionality.
Table 1.
Sociodemographic and clinical characteristics of psychiatric patients.
Characteristic EOP (n = 12) Non-EOP (n = 11) p
Age years ± SD 15.5 ± 1.56 13.36 ± 2.57 0.030 ^a
Gender Male, n (%) 6 (50) 7 (64) 0.680 ^c
Female, n (%) 6 (50) 4 (36)
Education years ± SD 9.3 ± 1.96 7.2 ± 2.05 0.023 ^a
Body mass
index z-score, mean ± SD 1.08 ± 1.15 0.67 ± 1.50 0.368 ^a
Psychiatric
admissions n (%) 7 (58) 1 (9) 0.027 ^c
Total, median
(min–max) 1 (0–4) 0 (0–1) 0.016 ^b
Psychiatric
comorbidity n (%) 12(100) 9 (81) 0.370 ^c
Mood disorders, n (%) 11 (91) 8 (72) 0.316 ^c
Anxiety and stress disorders, n (%) 10 (83) 4 (36) 0.036 ^c
Conduct disorders, n (%) 5 (41) 5 (45) 1 ^c
Neurodevelopment disorders, n (%) 1 (8) 4 (36) 0.155 ^c
Eating disorder, n (%) 4 (33) 0 0.093 ^c
GAF Total score, mean ± SD 43.33 ± 15.14 74 ± 6.41 0.00003 ^a
Minimal, n (%) 1 (8) 3 (27) 0.017 ^c
Mild, n (%) 1 (8) 2 (18)
Moderate, n (%) 2 (16) 0
Severe, n (%) 8 (66) 0
Epigenetic age Wu’s clock, mean ± SD 11.08 ± 0.89 10.29 ± 0.78 0.015
[87]Open in a new tab
Abbreviations: BMI = body mass index; GAF = global assessment of
functioning; SD = standard deviation; T = Student’s t-test; U =
Mann–Whitney U test; χ^2 = chi-square test; ^a = statistic from
Student’s t-test; ^b = statistic from Mann–Whitney–Wilcoxon test; ^c =
statistic from χ^2 test. Notes: Fisher’s exact test was applied when
values <5. Bold values denote statistical significance, p < 0.05.
2.2. Epigenetic Age
There is a hypothesis that patients with psychosis have higher
biological age [[88]28]. To explore this, we further characterized 11
epigenetic clocks to evaluate whether if patients with EOP have an
increased biological age. Our study found that the epigenetic age was
higher in the EOP group compared to the non-EOP group ([89]Table 1). We
identified a correlation of epigenetic age with sociodemographic and
clinical characteristics with the Wu clock, with a lower functionality
measured by the GAF scale being associated with a higher epigenetic age
([90]Figure 1). Moreover, a higher number of admissions was correlated
with an increased epigenetic age. Additionally, the same correlations
were observed with the PedBE clock, showing a similar direction of
effects. In contrast, the DNAmTL clock showed a negative correlation
with the number of admissions ([91]Table 2).
Figure 1.
[92]Figure 1
[93]Open in a new tab
Heatmap of correlations between sociodemographic, clinical
characteristics and epigenetic age in psychiatric patients. The heatmap
colors correspond to the significance (p < 0.05): green indicates a
positive correlation, and red indicates a negative correlation. Blank
spaces indicate no significant correlation. Best Linear Unbiased
Prediction clock (BLUP); DNA methylation-based telomere length
(DNAmTL); Dunedin Pace of Aging Methylation (DunedinPoAm38); Elastic
Net clock (EN); global assessment of functioning (GAF); Horvath clock
(multi-tissue, Horvath-1); Horvath clock (skin and blood, Horvath-2);
Levine clock (PhenoAge); Pediatric Buccal Epigenetic clock (PedBE).
Table 2.
Correlations between sociodemographic, clinical characteristics and
epigenetic age in psychiatric patients.
Epigenetic Calculator Age
r, p GAF
r, p Admissions
r, p
BLUP 0.37, 0.079 −0.40, 0.056 0.40, 0.056
DNAmTL −0.12, 0.555 0.27, 0.198 −0.43, 0.038
DunnedinPoAm38 0.26, 0.224 −0.17, 0.434 0.19, 0.368
EN 0.26, 0.220 −0.26, 0.217 0.40, 0.057
Hannum 0.25, 0.238 −0.23, 0.274 0.29, 0.174
Horvath-1 0.18, 0.398 0.13, 0.553 −0.07, 0.742
Horvath-2 0.03, 0.848 −0.17, 0.427 0.34, 0.107
Levine 0.35, 0.100 −0.05, 0.806 0.09, 0.672
PedBE 0.41, 0.046 −0.53, 0.008 0.56, 0.005
Wu 0.30, 0.150 −0.45, 0.027 0.49, 0.015
Zhang 0.24, 0.253 −0.21, 0.324 0.35, 0.097
[94]Open in a new tab
Abbreviations: R = Pearson’s correlation, Best Linear Unbiased
Prediction clock (BLUP); DNA methylation-based telomere length
(DNAmTL); Dunedin Pace of Aging Methylation (DunedinPoAm38); Elastic
Net clock (EN); global assessment of functioning (GAF); Horvath’s clock
(multi-tissue, Horvath-1); Horvath’s clock (skin and blood, Horvath-2);
Levine’s clock (PhenoAge); Pediatric Buccal Epigenetic clock (PedBE).
Notes: Correlations were made with either Pearson or Spearman. Bold
indicates p < 0.05.
In addition, we observed that children and adolescent patients with
psychosis were associated with accelerated epigenetic age (Levine
clock, β = 5.07, CI 95 = (2.74, 7.40), p = 0.001), Furthermore,
clinical characteristics appeared to influence this accelerated
epigenetic age. Our findings show that a reduction in schooling (Levine
clock, β = −5.01, CI 95 = (−7.55, −2.48), p = 0.001), a higher number
of comorbidities (BLUP clock, β = 0.49, CI 95 = (0.02, 0.97), p =
0.041) and more admissions (Wu clock, β = 0.81, CI 95 = (0.03, 1.60), p
= 0.042) were associated with higher epigenetic age in Mexican patients
with EOP ([95]Table 3).
Table 3.
Summary of the stepwise regression between epigenetic clocks and
sociodemographic and clinical characteristics in psychiatric patients.
Epiclock Age (Years) Sex Schooling (Years) Comorbidity Admissions
β, SE
(CI 95) p β, SE
(CI 95) p β, SE
(CI 95) p β, SE
(CI 95) p β, SE
(CI 95) p
BLUP 1.78, 0.55
(0.62, 2.94) 0.004 - - −1.82, 0.54
(−2.96, −0.67) 0.003 0.49, 0.22
(0.02, 0.97) 0.041 - -
DNAmTL −0.07, 0.02
(−0.11, −0.02) 0.006 - - 0.08, 0.02
(0.03, 0.13) 0.001 - - −0.13, 0.05
(−0.24, −0.02) 0.017
EN 1.36, 0.58
(0.12, 2.60) 0.032 - - −1.47, 0.58
(−2.69, −0.25) 0.020 - - - -
Horvath-1 1.75, 0.70
(0.26, 3.24) 0.023 - - - - - - - -
Horvath-2 0.60, 0.26
(0.04, 1.16) 0.034 - - −0.90, 0.26
(−1.45, −0.35) 0.002 0.31, 0.10
(0.08, 2.38) 0.010 - -
Levine 5.07, 1.10
(2.74, 7.40) 0.001 8.41, 2.37
(3.4, 13.40) 0.002 −5.01, 1.20
(−7.55, −2.48) 0.001 - - - -
PedBE 0.35, 0.10
(0.13, 0.58) 0.003 - - −0.31, 0.10
(−0.53, −0.09) 0.008 - - - -
Wu - - - - - - - - 0.81, 0.37
(0.03, 1.60) 0.042
Zhang 1.93, 0.66
(0.56, 3.31) 0.008 - - −1.81, 0.69
(−3.25, −0.36) 0.016 - - - -
[96]Open in a new tab
Abbreviations: β = estimated coefficient; SE = standard error; T =
t-value; BLUP = Best Linear Unbiased Prediction clock; DNAmTL = DNA
methylation-based telomere length; EN = Elastic Net clock; Horvath-1 =
Horvath clock (multi-tissue); Horvath-2 = Horvath clock (skin and
blood); PedBE = Pediatric Buccal Epigenetic clock. Bold values denote
statistical significance at the p < 0.05 level. Comorbidity values were
represented as the number of psychiatric disorders; sex was denoted as
a factor (0 = male, 1 = female); admissions were coded as a factor (0 =
no, 1 = yes).
2.3. Epigenome-Wide Association Study
Differential methylation analysis was performed to associate epigenetic
risk markers between EOP and non-EOP groups in Mexican patients. We
identified eight differentially methylated CpG sites associated with
EOP at nominal significance, CpG sites are distributed across six
chromosomes ([97]Figure 2). These sites were mapped to seven genes:
ADGRV1 (Adhesion G Protein-Coupled Receptor V1), HIST1H2BB (Histone
Cluster 1 H2B Family Member B), CEP164 (Centrosomal Protein 164),
IRF2BP1 (Interferon Regulatory Factor 2 Binding Protein 1), MAP1B
(Microtubule-Associated Protein 1B), NAALAD2 (N-Acetylated α-Linked
Acidic Dipeptidase 2), and SULT1C4 (Sulfotransferase Family 1C Member
4). The annotation indicated that four sites were situated in gene
bodies, two sites in exons, and one site in a transcription start site
(TSS200). In terms of their location relative to CpG islands, 25% of
the sites were within the island, 25% on the island shores, and 50% in
the open sea. Additionally, six CpG sites had lower methylation values,
while two CpG sites showed higher values in the EOP group compared to
the non-EOP group ([98]Table 4). In deep research of CpG methylation
sites, there are reports of cg06583549, cg08523325, cg20150189, and
cg26028573 in the EWAS catalog. Similarly, we conducted an enrichment
analysis using the Enrichr tool and found associations between the
seven annotated genes and pathways related to aspartate and asparagine
metabolism, the cytosolic sulfonation of small molecules, and Schwann
cell myelination ([99]Supplementary Sheet S1).
Figure 2.
[100]Figure 2
[101]Open in a new tab
Manhattan plot of EWAS analysis in EOP. Manhattan plot of CpG sites and
p-values. The X-axis represents the chromosome position, and the Y-axis
represents −log[10](p). The red horizontal line parallel to the X-axis
denotes nominal significance (p < 1 × 10^−5). All loci with a p-value <
1 × 10^−5 are annotated to genes according to the human genome assembly
(hg19). The association models are adjusted for sex, age, and five
surrogate variables. Adhesion G Protein-Coupled Receptor V1 (ADGRV1);
Histone Cluster 1 H2B Family Member B (HIST1H2BB); Centrosomal Protein
164 (CEP164); Interferon Regulatory Factor 2 Binding Protein 1
(IRF2BP1); Microtubule-Associated Protein 1B (MAP1B); N-Acetylated
α-Linked Acidic Dipeptidase 2 (NAALAD2); Sulfotransferase Family 1C
Member 4 (SULT1C4). λ = 0.9783.
Table 4.
Top differentially methylated positions in EOP.
CpG Gene Annotation Chr Position Relation to Island β SE T p
cg24772138 MAP1B (Body) 5 71,405,539 S Shore −0.4483 0.0508 −8.8233
4.30 × 10^−7
cg26028573 HIST1H2BB (Exon) 6 26,043,587 N Shore 0.2340 0.0264 8.8374
4.22 × 10^−7
cg05100917 CEP164 (TSS200) 11 117,198,460 Island −0.5790 0.0627 −9.2343
2.48 × 10^−7
cg20150189 SULT1C4 (Body) 2 108,999,219 Open Sea −0.3525 0.0498 −7.0718
5.58 × 10^−6
cg27181762 ADGRV1 (Body) 5 90,195,345 Open Sea −0.5829 0.0860 −6.7701
9.02 × 10^−6
cg13883911 - 8 33,430,120 Open Sea 0.3104 0.0449 6.9051 7.26 × 10^−6
cg08523325 NAALAD2 (Body) 11 89,901,450 Open Sea −0.3064 0.0448 −6.8337
8.14 × 10^−6
cg06583549 IRF2BP1 (Exon) 19 46,387,962 Island −0.2669 0.0350 −7.6212
2.40 × 10^−6
[102]Open in a new tab
Abbreviations: Chr = chromosome; β = effect size; TSS = transcription
start site; T = t-statistic; SE = standard error. Adhesion G
Protein-Coupled receptor V1 (ADGRV1); Histone Cluster 1 H2B Family
Member B (HIST1H2BB); Centrosomal Protein 164 (CEP164); Interferon
Regulatory Factor 2 Binding Protein 1 (IRF2BP1); Microtubule-Associated
Protein 1B (MAP1B); N-Acetylated α-Linked Acidic Dipeptidase 2
(NAALAD2); Sulfotransferase Family 1C Member 4 (SULT1C4). CpG sites
ordered by statistical p-value from EWAS analysis in early-onset
psychosis.
2.4. Methylation Risk Score
Epigenetic biomarkers could serve as predicting tools for psychosis
[[103]29], and currently there is no epigenetic biomarker for EOP. We
constructed the EOP-MRS to associate the biomarker with
sociodemographic and clinical characteristic in children and adolescent
Mexican patients with EOP. Our findings show that panic disorder,
auditory hallucinations, and visual hallucinations were associated with
a higher MRS using five different p-value thresholds: MRS[1×10^−1] (β =
2.10, CI 95 = 0.53–4.70; β = 2.26, CI 95 = 0.63–4.93; β = 1.93, CI 95 =
0.47–4.31; respectively); MRS[1×10^−2] (β = 2.08, CI 95 = 0.55–4.64; β
= 2.21, CI 95 = 0.63–4.85; β = 1.90, CI 95 = 0.47–4.23; respectively);
MRS[1×10^−3] (β = 1.92, CI 95 = 0.50–4.23; β = 1.96, CI 95 = 0.54–4.31;
β = 1.78, CI 95 = 0.44–3.94; respectively); MRS[1×10^−4] (β = 1.91, CI
95 = 0.52–4.16; β = 1.73, CI 95 = 0.45–3.78; β = 1.54, CI 95 =
0.33–3.39; respectively); and MRS[1×10^−5] (β = 1.36, CI 95 =
0.28–2.95; β = 1.28, CI 95 = 0.23–2.82; β = 1.22, CI 95 = 0.19–2.69;
respectively). Furthermore, a higher MRS was associated with tactile
hallucinations (MRS[1×10^−5], β = 1.59, CI 95 = 0.25–3.48), and
negative symptoms (MRS[1×10^−1], β = 1.99, CI 95 = 0.48–4.46;
MRS[1×10^−2], β = 1.71, CI 95 = 0.37–3.87) ([104]Figure 3). Three out
of six thresholds (MRS[1×10^−5], MRS[1×10^−2], and MRS[1×10^−1])
predicted three clinical manifestations of psychosis and one
psychiatric comorbidity. Our results shows that our EOP-MRS may be an
indicator for visual and auditive hallucinations in Mexican children
and adolescent patients.
Figure 3.
[105]Figure 3
[106]Open in a new tab
Association between clinical characteristics and early-onset psychosis
methylation risk score (EOP-MRS) in psychiatric patients. The β-values
represent the estimated coefficient (β) from logistic regression
analysis with a 95% confidence interval (CI 95). Significant
associations between clinical characteristics and MRS are displayed as
solid points (p < 0.05), while non-significant results are displayed as
hollow points. Each MRS calculated corresponds to a different cutoff
p-value from the EWAS. The forest plot shows only significant
associations among the sixty-eight clinical characteristics.
3. Discussion
This study performed the first characterization of genome-wide
epigenetic biomarkers in Mexican children and adolescents with EOP and
further explored the correlations and associations between epigenetic
age (11 epigenetic clocks), MRS and sociodemographic and clinical
characteristics in Mexican children and adolescents with EOP.
The current study demonstrated that lower levels of functioning,
admissions, and anxiety–stress disorder were associated with EOP. Our
findings are consistent with a study of European patients with FEP
(first episode of psychosis), who had lower GAF scores compared to
other psychiatric groups [[107]30]. These results support the
hypothesis that a lower global functioning may be influenced by
psychiatric comorbidities and admissions, leading to a poor prognosis
[[108]31] or premature death [[109]32].
3.1. Years of Schooling Was Associated with Epigenetic Age in EOP
The present study shows accelerated epigenetic age in psychiatric
patients with EOP. Consistent with our results, previous work
identified that monozygotic twins with psychiatric disorders had an
accelerated epigenetic age during early adolescence measured with the
Wu clock [[110]19], and in Mexican adults, the presence of a mental
disorder accelerated Horvath’s epigenetic age in discordant monozygotic
twins [[111]33]. In this sense, we suggest that Wu’s clock could be
used to estimate epigenetic age in the adolescent Mexican population.
Furthermore, the accelerated epigenetic age was associated with fewer
years of schooling in EOP, suggesting that more years of schooling
could be better to reduce biological age. This association may be
influenced by exposure to psychosocial stressors, such as increased
academic demands, and unhealthy lifestyles. Additionally, factors
characteristic of the Mexican population with low socioeconomic
status—such as limited access to health resources and lower self-care
literacy—may contribute to allostatic load, potentially activating the
hypothalamic–pituitary–adrenal (HPA) axis, altering cortisol release,
and modulating dopamine response [[112]9]. This biological response
could promote accelerated biological aging through DNA methylation. In
accordance with our results, previous work in the Mexican population
showed that schooling influences epigenetic age [[113]34]. We consider
that schooling could influence biological age in Mexicans with EOP.
Notably, the lack of concordance between different epigenetic clocks
could reflect how each clock is calibrated to specific sets of CpGs or
tissues. Nevertheless, further research is still needed to explore the
effect of other stressors and social determinants of health on the
relationship between EOP and epigenetic age [[114]35].
3.2. EWAS Suggested Potential Novel Associations with EOP
Our EWAS analysis identified eight CpG sites associated with EOP at a
nominal level. Nonetheless, a previous meta-analysis reported 95
differentially methylated positions associated with psychosis in adult
patients [[115]14]. The MAP1B gene regulates axon growth and synaptic
plasticity [[116]36], and its dysregulation leads to disruptions in
neurogenesis and synaptic plasticity [[117]37]. The HIST1H2BB gene is
involved in cell motility [[118]38] and neurodevelopment [[119]39]. In
addition, methylation changes in the CEP164 gene are associated with
brainstem malformation [[120]40], and the IRF2BP1 gene is involved in
immune suppression [[121]41]. There is evidence that SULT1C4, NAALAD2,
and ADGRV1 participate in neurotransmitter regulation [[122]42],
glutamate dysregulation in schizophrenia [[123]43], and epilepsy and
audiovisual disorders [[124]44,[125]45]. All these findings indicate
that DNA methylation is implicated in neurodevelopmental disorders,
immune modulatory pathways, and neurotransmitter regulation in EOP
patients. This suggestion is based on the following findings: (i) a
reduced level of methylation is observed at cg24772138, cg05100917,
cg06583549, cg20150189, cg08523325, and cg27181762; (ii) a higher level
of methylation is observed at cg26028573 and cg13883911. This result
should be interpretated with caution, as the nominal threshold
represents exploratory work. Further studies are needed to replicate
our findings.
We identified four CpG sites previously reported in EWASs
[[126]46,[127]47,[128]48,[129]49]. Islam et al. (2019) suggest that
cg06583549, cg08523325, cg20150189, and cg26028573 present DNAm
concordance between buccal epithelial cells and peripheral blood cells
in pediatric patients [[130]46]. Furthermore, these sites are
associated with children’s neurodevelopment [[131]49] and HIV infection
in adults [[132]48]. In addition, cg26028573 was associated with
alcohol consumption [[133]47]. Notably, this association has been
reported in psychotic patients and individuals with developmental
disorders [[134]50]. Our findings support evidence that DNAm signatures
overlap across different pediatric tissues in psychotic patients
[[135]51]. Moreover, these sites could reflect shifts across the
lifespan [[136]52] and neurodevelopmental disorders [[137]53],
suggesting that our results could provide insight into children’s
neurodevelopment within our population.
3.3. Association Between Clinical Characteristics Associated with Psychosis
MRS
Our study showed clinical characteristics associated with EOP-MRS. We
constructed an EOP-MRS using six different thresholds to obtain varying
weighted sums from distinct CpG sites. The MRS[1×10^−5] includes the
eight CpG sites that were differentially methylated and nominally
associated with early-onset psychosis. Our findings demonstrated that
increases in this MRS predict three clinical manifestations of
psychosis and panic disorder associated with early-onset psychosis.
Previous associations of MRS with clinical characteristics have been
identified in other populations. For example, in an adult Australian
population diagnosed with schizophrenia, the MRS was associated with
clozapine administration [[138]54].
The potential utility of our EOP-MRS lies in its ability to predict
psychotic symptoms directly measured in the adolescent population. In
addition, this score may be determined in individuals who do not
exhibit psychosis. Furthermore, to establish it as a biomarker for
psychosis, the MRS must be validated in different tissues and evaluated
in relation to the stage of the disease. Our study was conducted during
the prodromal phase, suggesting its potential utility in identifying
risk factors. However, it could be also explored as a biomarker for
treatment response, disease severity, or progression in clinical
applications. Our EOP-MRS (the most predictive ones: MRS[1×10^−5],
MRS[1×10^−2], and MRS[1×10^−1]) could represent a predictive tool to
identify panic disorder and hallucinations in Mexican children and the
adolescent population. We suggest that EOP-MRS reflects the cumulative
effect of multiple epigenetic markers associated with EOP and
recapitulates its clinical manifestations.
3.4. Limitations
We considered several limitations: This is a hypothesis-generating
study rather than a conclusive investigation of the biological
mechanisms involved in early-onset psychosis. A small sample of
psychiatric patients with EOP was included in our research, which
limits our statistical power to detect associations and transform this
study into an exploratory one. Additionally, epigenetic studies are not
ideal for identifying causal risk factors, as DNA methylation (DNAm)
may be influenced by environmental confounders. Moreover, EPIC arrays
cover only 3% of CpG sites in the genome, and polymorphisms and
mutations were not accounted for in psychosis risk calculations.
However, this could be addressed using different technologies, such as
third-generation sequencing for quantifying methylation levels.
Furthermore, DNAm profiles focus on peripheral blood as a surrogate for
brain tissue, and variations associated with early-onset psychosis may
not represent specific brain tissue biomarkers in the early stages of
the disease. Additionally, epigenetic age, EWAS, and psychosis MRS
associated with EOP could be biased due to ancestry, pubertal stage,
childhood adversity, or environmental exposures not included in the
analysis. Likewise, the associations between the EOP-MRS and clinical
manifestations could be lost after adjusting the MRS model, possibly
due to the use of different thresholds and the small sample size. We
did not have a replication cohort given the difficulty in recruiting
those with EOP. In addition, we should consider pruning, as variability
in epigenetic signals may differ across individuals, populations, and
ages, particularly during early stages of life. Finally, future
research should address these issues including larger samples, DNAm
profiles from other tissues, and social determinants of health to
replicate our results.
4. Materials and Methods
4.1. Sample Population
Thirty-three patients with previously diagnosed psychiatric disorders
were recruited from the Dr. Juan N. Navarro Children’s Psychiatric
Hospital in Mexico City. A psychiatrist assessed all participants
according to the DSM-5 (Diagnostic and Statistical Manual of Mental
Disorders, 5th Edition). We included patients aged 10 to 18 years of
Mexican descent, with the onset of psychotic symptoms before the age of
18. Exclusion criteria included patients with psychosis secondary to a
brain infection, neurodegenerative disease, or severe
neurodevelopmental disorder.
4.2. Study Design
This was a cross-sectional study with a convenience sample design.
Recruitment took place from January 2022 to July 2023. The sample was
divided into two groups: the EOP group (patients with psychotic
symptoms) and the non-EOP group (patients without psychotic symptoms).
The clinical evaluation of patients was performed using the Kiddie
Schedule for Affective Disorders and Schizophrenia Present and Lifetime
version DSM-5 (K-SADS PL-5), Spanish edition, validated in children and
the adolescent Mexican population [[139]55]. Each individual diagnosis
was categorized into one of five groups as follows: (1) mood disorders
(disruptive mood dysregulation disorder, dysthymia, major depressive
disorder, bipolar disorder); (2) anxiety and stress disorders
(separation anxiety disorder, agoraphobia, panic disorder, social
phobia, generalized anxiety disorder, post-traumatic stress disorder);
(3) conduct disorders (enuresis, encopresis, oppositional defiant
disorder, conduct disorder, tic disorder, obsessive-compulsive
disorder); (4) neurodevelopmental disorders (attention
deficit/hyperactivity disorder, autism spectrum disorder); (5) eating
disorders (bulimia nervosa, other eating disorders).
Clinicians collected sociodemographic information of age, gender,
schooling (completed years of education), and clinical data, including
weight, height, body mass index (BMI), number of psychiatric
hospitalizations (admissions), number of medications (psychiatric
treatment), and number of psychiatric comorbidities. BMI values were
converted into z-scores with standard deviations. The global assessment
of functioning (GAF) was measured on a 100-point scale, where lower
scores indicated severe symptoms or impairment [[140]56]. GAF is
validated in Spanish and adolescent populations [[141]57]. At the end
of the interview, we collected a peripheral venous blood sample (4 mL)
from each patient.
4.3. DNA Extraction
Genomic DNA was obtained from whole blood samples using the DNA kit
(Qiagen, Germantown, MD, USA), according to the manufacturer’s
instructions. Quality and integrity were evaluated using a NanoDrop
spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Sample
processing for the methylation study protocol was performed with a
genomic DNA concentration of 80 ng/mL. The genomic DNA was
bisulfite-converted using the EZ DNA Methylation Kit (Zymo Research
Corporation, Irvine, CA, USA), according to the manufacturer’s
protocol, “Infinium Assay for Methylation Protocol” (Illumina Inc., San
Diego, CA, USA), specific for the “Infinium Methylation EPIC BeadChip
Kit.” The Microarray Unit at the National Institute of Genomic Medicine
(Mexico City, Mexico) processed the microarray.
4.4. Genomic-Wide Quantification of DNA Methylation
Fluorescence intensities were transformed into IDAT files using
GenomeStudio software version 2.0 (Illumina, USA). The IDAT data were
imported into the R environment. β-normalized methylation levels were
obtained following the ENmix (1.22.0) pipeline. Preprocessing included
background correction, RELIC dye bias correction, and RCP probe-type
bias adjustment. We removed CpG sites with low variability (12,730 CpG
sites) and sites that coincided with SNP loci from the analysis.
Additionally, 10 samples were eliminated due to sex mismatch, poor
quality, and being outliers. Five surrogate variables were estimated to
correct for batch effects using ‘csva()’; and cell proportions
(neutrophils, monocytes, natural killer cells, CD8+ T-lymphocytes, CD4+
T-lymphocytes, and B-lymphocytes) were imputed using
‘estimateCellProp()’ (ENmix R library). Finally, 848,643 CpG sites from
23 samples were analyzed to obtain the methylation β (β)- and M-values.
4.5. Epigenetic Clocks
Epigenetic age was estimated using β-values with two libraries. The
methylClock and methylCIPHER packages were chosen for their broad
availability of epigenetic calculators. Eleven epigenetic clocks were
selected from all three generations of age estimators. Each epigenetic
clock was trained on different tissues and different CpG sites to
capture aging from diverse sources using different methodological
algorithms. The Wu and Pediatric Buccal Epigenetic (PedBE) clocks were
included because they are designed to predict age in children and
adolescents. The Dunedin Pace of Aging (DunedinPoAm38) and Levine
(PhenoAge) clocks were selected for their ability to calculate
biological age and predict mortality. DNA methylation-based telomere
length (DNAmTL), Hannum, Horvath-1 (multi-tissue), and Zhang clocks
were included because they predict chronological age in blood samples.
The Best Linear Unbiased Prediction (BLUP), Elastic Net (EN), and
Horvath-2 (skin and blood) clocks were chosen because they were trained
on blood, skin, and saliva samples using different arrays (450k and
EPIC arrays)
[[142]58,[143]59,[144]60,[145]61,[146]62,[147]63,[148]64,[149]65,[150]6
6].
4.6. Statistical Analysis
Clinical characteristics are presented as means and standard deviations
for continuous variables, and as frequencies and percentages for
categorical variables. Comparisons between groups were made using
Student’s t-test or the Mann–Whitney U test. The associations were
assessed with the chi-square test (χ^2). In psychiatric patients the
correlations between epigenetic age and clinical and sociodemographic
characteristics were calculated using corrplot R package. Multiple
linear regressions (stepwise regression) were performed to associate
sociodemographic and clinical characteristics with epigenetic age (the
response variable). Statistical analysis was conducted using R
software, version 4.3.3 ([151]https://CRAN.R-project.org, accessed 30
December 2024). The significance level was set at p < 0.05.
4.7. EWAS and Methylation Risk Score
Genome-wide CpG association analysis with EOP was performed using
normalized β-values. EOP was the response variable, and each individual
CpG methylation β-value served as the predictor of interest. A logistic
model was implemented using the ‘CpGassoc()’ R package [[152]67], with
PC1 (principal component 1), PC2, PC3, PC4, PC5, age, and sex included
as covariates. A variance inflation factor (VIF) was used to detect
multicollinearity among predictors and to select covariates. A QQ plot
was examined for evidence of genomic inflation (λ = 0.9783, [153]Figure
S1). We manually search for previous associations of the CpG sites
using the EWAS catalog [[154]68].
CpG sites were stratified by cut-off p-values from the EWAS results to
calculate the EOP-MRS (methylation risk score to psychosis):
MRS[1×10^−6] (3 CpG sites, p < 1 × 10^−6), MRS[1×10^−5] (8 CpG sites, p
< 1 × 10^−5), MRS[1×10^−4] (101 CpG sites, p < 1 × 10^−4), MRS[1×10^−3]
(997 CpG sites, p < 1 × 10^−3), MRS[1×10^−2] (9832 CpG sites, p < 1 ×
10^−2), MRS[1×10^−1] (88,916 CpG sites, p < 1 × 10^−1).
[MATH:
MRS=∑n=CpGM va
lue × Z value :MATH]
(1)
The MRS was calculated as the sum of the individual products of the
M-values and the Z-values (effect size/standard error) for each CpG
site (CpG1 + CpG2 + … + CpGn). The resulting score was then normalized
into z-scores.
[MATH:
MRS z score=MRS
score−mean MRS valuestand<
mi>ard dev<
/mi>iation
:MATH]
(2)
Finally, linear and logistic regressions were performed to assess
associations between EOP-MRS and sociodemographic and clinical
characteristics. Sociodemographic and clinical variables served as the
response variable, and the EOP-MRS was considered a predictor of
interest.
5. Conclusions
We found that patients with EOP have lower levels of global functioning
and accelerated epigenetic age. Furthermore, fewer years of education,
comorbidity, and psychiatric hospital admissions impact epigenetic age.
Additionally, our study suggests associations between DNAm and EOP,
specifically in genes involved in immune modulatory and
neurotransmission pathways. Finally, we developed an EOP-MRS that could
predict clinical manifestations of psychosis and psychiatric
comorbidity associated with EOP. This score could be used as a risk
biomarker for EOP in Mexican adolescents.
Acknowledgments