2024-12-09
109_ Uncovering the epigenetics of suicidal behavior
Research Question and Aims
BACKGROUND
Suicidal behavior (SB) is a serious public health concern involving approximately one million annual deaths worldwide. According to
the WHO, suicide is the fourth leading cause of death worldwide and the second leading cause of death in Europe among young
individuals ranging from 15-29 years old. SB is highly prevalent in the course of severe mental disorders, with over 90% of suicides
and SA being related to psychiatric disorders (1).
Current evidence indicates that SB results from a complex and not yet understood interplay between premorbid clinical conditions,
environmental factors, and biological factors (2). New evidence points to the important role that epigenetic factors play in SB,
particularly DNA methylation (DNAm). Methylation can regulate gene expression/silencing by altering gene transcription without
changing the underlying DNA sequence. This makes DNA methylation one of the mechanisms responsible for phenotypic variability
and a key player in the interaction between genes and the environment.
Recent epigenome-wide association studies (EWAS) have investigated changes in the methylome of brain tissues indicating that
suicide is associated with widespread changes in DNAm patterns of neurotrophic and neuroprotective factors in the hippocampus
and prefrontal cortex (3–5). Additionally, our group investigated epigenome-wide differences in patients with bipolar disorder (BD)
with and without a previous history of suicide attempts, identifying 18 differentially methylated positions and 2 regions between
suicide attempters (BD/SA) and non-attempters /BD/non-SA) (in review at JAD).
Moreover, evidence showed that serious mental disorders, such as BD, are associated with an increased risk of medical illnesses and
premature mortality from natural causes, with lifespans up to 25 years shorter than the general population. This has raised the
possibility that severe mental disorders are associated with accelerated biological aging. Whereas chronological age is measured by
the passage of time, biological age is defined physiologically and functionally and is more closely associated with disease processes
and mortality (6). In this regard, age estimators based on methylation status are considered epigenetic clocks that can help us
determine how biological aging is progressing (7). In this context, evidence has shown accelerated epigenetic age (EA) in individuals
with suicide attempts (SA) (8–10). Our group assessed differences in epigenetic age acceleration (AgeAccel) by estimating several
epigenetic clocks, and we also found accelerated AgeAccel for the GrimAge clocks in BD/SA compared to BD/non-SA in our sample of patients with BD (currently under review at JAD). Conversely, studies on suicidal ideation (SI) are more limited, and thus far no
significant accelerated EA has been observed in individuals with SI compared to those without SI (11,12).
To conclude, though more studies are needed given that EWAS studies focused on SB are still scarce, this project aims to enhance
the existing knowledge regarding differential methylation patterns in these phenotypes. In addition, this study will be one of the first
to explore the aging pace by estimating the most novel pace predictor, DunedinPACE and will assess biological age acceleration
using six different epigenetic clocks.
OBJECTIVES
1. To identify genome-wide DNAm patterns and epigenetic age acceleration associated with SA and SI in BD. This will be achieved by
analyzing blood samples from 96 BD patients from the PsyCourse cohort, who have been assessed for SB and profiled using the
Infinium MethylationEPIC BeadChip v1.0.
2. To detect genome-wide DNAm patterns and epigenetic age acceleration associated with SA and SI in different psychiatric
disorders including BD, SCZ, and MDD. This will be done by analyzing blood samples from 192 cross-diagnostic patients from the
PsyCourse cohort, assessed for SB, and profiled using the Infinium MethylationEPIC BeadChip v2.0.
3. To meta-analyze the results obtained from objectives 1 and 2.
Analytic Plan
Participants
For objective 1 we will use 96 BD-I and II patients from the PsyCourse cohort, who have been assessed for SB and profiled using the
Infinium MethylationEPIC BeadChip v1.0.
For objective 2 we will use 192 patients with different psychiatric disorders, including BD-I, BD-II, schizophrenia (SCZ),
schizoaffective, and recurrent depression, from the PsyCourse cohort. These patients have been assessed for SB and profiled using
the Infinium MethylationEPIC BeadChip v2.0.
Phenotype definition
For both samples, patients will be categorized into two groups based on the presence or absence of a prior history of SA at visit 1. In
the same way, SI will be defined based on the presence or absence of a prior history of SI in these individuals at visit 1.
QC methods
First, quality–-control of all samples will be done by filtering: (i) probes with low detection p-value (p>0.01), (ii) probes with <3 beads
in at least 5% per probe, (iii) non-CpGs probes contained in the dataset, (iv) SNP-related probes, (v) cross-hybridizing probes, and (vi)
probes located in chromosomes X and Y. Additional filtering steps for the Infinium HumanMethylationEPIC v2.0 will be applied,
including Illumina’s flagged probes, inaccurately mapped probes, and probes located on chromosome 0.
In addition to the filtering of probes and samples, normalization using preprocess Quantile and Beta Mixture Quantile Normalization
(BMIQ) will be performed to adjust the β-values of type-II probes into a type-I probe statistical distribution (13). Finally, Combat will
be used to correct for batch effects.
All of these steps will be performed by following the publicly available pipeline by Nathan Yusupov and Alexandra Halberstam:
“https://github.molgen.mpg.de/mpip/EPIC_Preprocessing_Pipeline“, which can currently be implemented to fit Infinium
HumanMethylationEPIC v2.0.
Plans to address population stratification/other confounders
Ancestry principal components (PCs) will be calculated using PLINK v.1.9 (Chang et al., 2015) using raw genomic data. PCs will be
added as covariates in our differential methylation analysis to correct for population ancestry/stratification.
Moreover, blood cell type proportions and the smoking score will be estimated for all individuals utilizing EpiDISH and EpiSmokEr R
packages, respectively. These variables are known to correlate with DNA methylation changes, so it is crucial to account for them
when analyzing methylation differences between groups (14,15).
Statistical power
Statistical power for both PsyCourse samples will be calculated employing pwrEWAS, an R package that provides a reliable estimate
of the statistical power of the study (16).
Differential Methylation Analysis
Differentially methylated positions (DMPs) and regions (DMRs) between groups will be calculated using Limma and DMRcate,
respectively (17,18). Blood cell counts, sex, age, and smoking status (smoking score) will be added as covariates. For both, DMPs and
DMRs, Benjamini-Hochberg multiple-testing correction will be used to correct the false discovery rate (FDR), and a p-value of 0.05
will be considered for significance.
Blood-brain DNA methylation correlation
The IMAGE-CpG tool (https://shinozaki-lab-image-cpg.web.app/), based on CpG sites measured on the Infinium
HumanMethylationEPIC v1.0 BeadChip (19), will be employed to infer the blood-brain correlation of identified DMPs to strengthen
the interpretation of findings.
Number of suicide attempts and DNA methylation correlation
If possible, correlations between the number of suicide attempts and the residuals of the beta-values (DNA methylation) after
adjusting for age, sex, blood cell counts, and smoking score will be analyzed. This will assess whether DNAm has a dose-response
relationship with suicide attempt burden.
Epigenetic aging
EA in each patient will be estimated using Horvath’s online calculator (https://dnamage.clockfoundation.org/) based on different
epigenetic clocks: Horvath (7), Skin and Blood (20), Hannum (21), PhenoAge (22), GrimAge (23) and the second version of GrimAge
(24). Different EA measures will be estimated for all patients including AgeAccel (age acceleration) for all clocks. Additionally,
DunedinPACE, which is a novel DNAm measure of the pace of biological aging (25), will be calculated using the Dunedin R package
and pipeline (https://github.com/danbelsky/DunedinPACE). Correlation coefficients between chronological age and epigenetic age
estimates will be calculated. Moreover, differences in AgeAccel corrected by age, sex, blood cell counts, and smoking score will be
compared between groups based on the presence of SA/SI. FDR multiple-testing correction will be applied and a p-value<0.05 after correction will be considered for significance. All statistical analyses will be performed using IBM SPSS 27.0 (https://www.ibm.com/es-es/spss).
Meta-analysis
Finally, a meta-analysis combining the EWAS data obtained from the BD (n=96) sample and the sample with several psychiatric
diagnoses (n= 192) will be performed by employing the R package metafor (26). Analyzing these data will uncover more robust
associations of the DNA methylation patterns in SB.
Resources needed
v1_adv
v1_Antidepressants
v1_Antipsychotics
v1_asrm_sum
v1_bdi2_sum
v1_bmi
con_psy_snc_lst
v1_cur_psy_trm
v1_cur_work_restr
v1_curr_paid_empl
v1_disabl_pens
gsa_id
v1_lith
v1_lith_prd
v1_liv_aln
med_pst_sx_mths
med_pst_wk
v1_medchange
v1_Mood_stabilizers
v1_mss_sum
v1_no_cig
v1_Other_psychiatric
v1_panss_g1
v1_panss_g10
v1_panss_g11
v1_panss_g12
v1_panss_g13
v1_panss_g14
v1_panss_g15
v1_panss_g16
v1_panss_g2
v1_panss_g3
v1_panss_g4
v1_panss_g5
v1_panss_g6
v1_panss_g7
v1_panss_g8
v1_panss_g9
v1_panss_n1
v1_panss_n2
v1_panss_n3
v1_panss_n4
v1_panss_n5
v1_panss_n6
v1_panss_n7
v1_panss_p1
v1_panss_p2
v1_panss_p3
v1_panss_p4
v1_panss_p5
v1_panss_p6
v1_panss_p7
v1_panss_sum_gen
v1_panss_sum_neg
v1_panss_sum_pos
panss_sum_tot
v1_pst6_ill_drg
psyc_id
v1_spec_emp
v1_Tranquilizers
v1_1st_ep
v1_age
v1_age_1st_inpat_trm
v1_age_1st_out_trm
v1_age_f_birth
v1_age_fst_suic_att
v1_age_m_birth
v1_age_sec_suic_att
v1_age_smk
v1_age_thr_suic_att
v1_alc_pst12_mths
v1_brothers
v1_can_cat_evr
v1_cat_daypat_outpat_trm
v1_center
v1_cntr_brth
v1_cntr_brth_f
v1_cntr_brth_m
v1_daypat_inpat_trm
v1_dur_illness
v1_ed_status
v1_epic_id
v1_epic2_id
v1_ever_smkd
v1_evr_hvy_usr
v1_evr_ill_drg
v1_fam_hist
v1_hal_cat_evr
v1_hlf_brthrs
v1_hlf_sstrs
v1_id
v1_inh_cat_evr
v1_kok_cat_evr
v1_leq_A_1A
v1_leq_A_1B
v1_leq_A_2A
v1_leq_A_2B
v1_leq_A_3A
v1_leq_A_3B
v1_leq_A_4A
v1_leq_A_4B
v1_leq_A_5A
v1_leq_A_5B
v1_leq_A_6A
v1_leq_A_6B
v1_leq_A_7A
v1_leq_A_7B
v1_leq_A_8A
v1_leq_A_8B
v1_leq_A_9A
v1_leq_A_9B
v1_leq_B_10A
v1_leq_B_10B
v1_leq_B_11A
v1_leq_B_11B
v1_leq_B_12A
v1_leq_B_12B
v1_leq_B_13A
v1_leq_B_13B
v1_leq_B_14A
v1_leq_B_14B
v1_leq_B_15A
v1_leq_B_15B
v1_leq_B_16A
v1_leq_B_16B
v1_leq_B_17A
v1_leq_B_17B
v1_leq_B_18A
v1_leq_B_18B
v1_leq_B_19A
v1_leq_B_19B
v1_leq_C_20A
v1_leq_C_20B
v1_leq_C_21A
v1_leq_C_21B
v1_leq_C_22A
v1_leq_C_22B
v1_leq_C_23A
v1_leq_C_23B
v1_leq_D_24A
v1_leq_D_24B
v1_leq_D_25A
v1_leq_D_25B
v1_leq_D_26A
v1_leq_D_26B
v1_leq_D_27A
v1_leq_D_27B
v1_leq_E_28A
v1_leq_E_28B
v1_leq_E_29A
v1_leq_E_29B
v1_leq_E_30A
v1_leq_E_30B
v1_leq_E_31A
v1_leq_E_31B
v1_leq_E_32A
v1_leq_E_32B
v1_leq_E_33A
v1_leq_E_33B
v1_leq_E_34A
v1_leq_E_34B
v1_leq_E_35A
v1_leq_E_35B
v1_leq_E_36A
v1_leq_E_36B
v1_leq_E_37A
v1_leq_E_37B
v1_leq_E_38A
v1_leq_E_38B
v1_leq_E_39A
v1_leq_E_39B
v1_leq_E_40A
v1_leq_E_40B
v1_leq_E_41A
v1_leq_E_41B
v1_leq_E_42A
v1_leq_E_42B
v1_leq_F_43A
v1_leq_F_43B
v1_leq_F_44A
v1_leq_F_44B
v1_leq_F_45A
v1_leq_F_45B
v1_leq_F_46A
v1_leq_F_46B
v1_leq_F_47A
v1_leq_F_47B
v1_leq_F_48A
v1_leq_F_48B
v1_leq_F_49A
v1_leq_F_49B
v1_leq_F_50A
v1_leq_F_50B
v1_leq_G_51A
v1_leq_G_51B
v1_leq_G_52A
v1_leq_G_52B
v1_leq_G_53A
v1_leq_G_53B
v1_leq_G_54A
v1_leq_G_54B
v1_leq_G_55A
v1_leq_G_55B
v1_leq_H_56A
v1_leq_H_56B
v1_leq_H_57A
v1_leq_H_57B
v1_leq_H_58A
v1_leq_H_58B
v1_leq_H_59A
v1_leq_H_59B
v1_leq_H_60A
v1_leq_H_60B
v1_leq_H_61A
v1_leq_H_61B
v1_leq_H_62A
v1_leq_H_62B
v1_leq_H_63A
v1_leq_H_63B
v1_leq_H_64A
v1_leq_H_64B
v1_leq_H_65A
v1_leq_H_65B
v1_leq_H_66A
v1_leq_H_66B
v1_leq_H_67A
v1_leq_H_67B
v1_leq_H_68A
v1_leq_H_68B
v1_leq_I_69A
v1_leq_I_69B
v1_leq_I_70A
v1_leq_I_70B
v1_leq_I_71A
v1_leq_I_71B
v1_leq_I_72A
v1_leq_I_72B
v1_leq_I_73A
v1_leq_I_73B
v1_leq_J_74A
v1_leq_J_74B
v1_leq_J_75A
v1_leq_J_75B
v1_leq_J_76A
v1_leq_J_76B
v1_leq_J_77A
v1_leq_J_77B
v1_leq_J_78A
v1_leq_J_78B
v1_leq_J_79A
v1_leq_J_79B
v1_lftm_alc_dep
v1_lith
v1_opi_cat_evr
v1_outpat_psy_trm
v1_prep_suic_attp_ord
v1_prof_dgr
v1_rel_act
v1_rel_chr
v1_rel_isl
v1_rel_oth
v1_school
v1_scid_age_fst_psyc
v1_scid_age_hypomania
v1_scid_age_mania
v1_scid_age_MDE
v1_scid_dsm_dx
v1_scid_dsm_dx_cat
v1_scid_ever_delus
v1_scid_ever_halls
v1_scid_ever_psyc
v1_scid_evr_suic_ide
v1_scid_no_hypomania
v1_scid_no_mania
v1_scid_no_MDE
v1_scid_no_suic_attmpt
v1_scid_suic_ide
v1_scid_suic_note_thgts
v1_scid_suic_thght_mth
v1_scid_yr_fst_psyc
v1_seas_birth
v1_sex
v1_sisters
v1_sti_cat_evr
v1_stp_brthrs
v1_stp_sstrs
v1_suic_attmpt
v1_suic_note_attmpt
v1_tbi
v1_tms_daypat_outpat_trm
v1_tra_cat_evr
v1_twin_fam
v1_twin_slf
v1_var_cat_evr
v1_wrk_abs_pst_5_yrs
v1_yob
v1_marital_stat
v1_partner
v1_no_bio_chld
v1_no_adpt_chld
v1_stp_chld