SUPPLEMENT
McFarquhar M, Elliott R, McKie S et al. TOMM40 rs2075650 may represent a new candidate gene for vulnerability to major depressive disorder.
This supplementary material has been provided by the authors to give readers additional information about their work.
Contents:
1. Further recruitment details and the study design
2. Measuring Lifetime Depression in Level-1
3. Neuropsychological and fMRI task descriptions
4. Further genetic information
5. Further statistical information
6. Additional results
7. Power calculations for genetic association studies
1
1. Further recruitment details and the study design
Detailed description of the recruitment methodology has been published previously (Juhasz et al, 2009; Juhasz et al, 2011).
Level-1
For the Level-1 phase of the NewMood study (http://www.newmood.co.uk) more than 2,000 subjects completed and returned the postal questionnaire. From these individuals 1520 provided DNA by returning the buccal swab kit and a signed consent form (Juhasz et al, 2009). For the genotyping of TOMM40 rs2075650 we only included those DNA samples that have more than 90% call rate in our previous studies. Participants who reported history of manic or hypomanic episodes, or psychotic symptoms were excluded. Any participants with a self-reported history of anxiety disorder were not excluded due to high co-morbidity with depression and a substantial genetic overlap (Hettema, 2008). The only exception was obsessive-compulsive disorder (OCD), which we excluded, due to the evidence that high co-morbidity between OCD and major depressive disorder (MDD) is more likely to occur secondarily, as a result of OCD (Bartz and Hollander, 2006; Fineberg et al, 2007). Thus our final sample for this study consisted of n = 1220 subjects. We used the Brief Symptom Inventory (BSI) as our measure of current depression as the questionnaire is not too involved, but has demonstrated construct validity and convergence validity with the more comprehensive Symptom Checklist-90-R (SCL-90-R®) (Derogatis and Melisaratos, 1983). Similarly, we used the Big Five Inventory for its relative short-length (44-items), and because the BFI-44 has demonstrated high reliability, discriminant validity, and a high level of convergence with more comprehensive measures such as the NEO-PI-R (Soto and John, 2009).
Level-2
For Level 2 we invited eligible (of Caucasian origin, who provided DNA, with no self-reported history of manic or hypomanic episodes, psychotic symptoms, or obsessive-compulsive disorder) Level 1 participants who were either ‘non-depressed controls’ with no personal or family history of depression (based on Background Questionnaire) and a low score on the Brief Symptom Inventory (BSI) depression subscale (i.e. around or below the normal population mean of 0.21 for males and 0.36 for females); or were ‘remitted depressed’ who stated that they have been depressed (based on Background Questionnaire) but had a low BSI depression score (i.e. around or below the normal population mean of 0.21 for males and 0.36 for females); or were ‘currently depressed’ who have a history of depression (based on Background Questionnaire) and a high BSI depression score (i.e. around or greater than the out-patient mean of 1.65 for males and 1.90 for females). In addition, further subjects were recruited as required to obtain the targeted sample numbers. Altogether 264 face-to-face interview were carried out for the Level-2 phase of NewMood (Juhasz et al, 2011). 4 subjects were excluded for non-MDD primary psychiatric diagnosis (bipolar disorder, personality disorder, drug or alcohol abuse/dependence), 7 for not consenting to give DNA and 10 for not providing good quality DNA. Thus the final sample consisted of 243 subjects.
Based on SCID-I/NP (First et al, 2002) 102 participants fulfilled the criteria of controls, 97 of remitted depressed, 12 of partially remitted MDD, and 32 of current MDD. Partially remitted subjects (PRD) did not meet the DSM-IV criteria for current MDD but had a number of residual symptoms of a past MDD. A multivariate analysis of variance indicated no significant differences between the PRD and MDD groups in terms of age, or number of depressive episodes (all p >.05). As expected, significance was found between the PRD and MDD group on MADRS score (F(1,42)=6.524, p=0.014). An additional chi-square test found no significant differences in gender frequency between the two groups (χ2(1)=0.328, p=0.567). In the interests of power these two groups were collapsed.
15.8 % of participants in Level-2 were currently taking antidepressant medication; the most commonly reported being venlafaxine (3.3%), citalopram (2.5%), duloxetine (1.6%), clomipramine (1.2%), and fluoxetine (1.2%). Because the effect of medication was not significant on the outcome parameters of our tasks subjects with antidepressant medication were not excluded from the analysis in the interests of maintaining statistical power. Instead, any residual error in the model pertaining to the medication was reduced by using a dichotomised medication indicator variable in the model as an additional covariate.
Level-3
From Level-2 67 (control and remitted depressed) subjects agreed to take part at the fMRI scanning session for Level-3 phase of NewMood (Thomas et al, 2011a). In the present study 4 subjects (2 controls, 2 remitted) were excluded from the analysis due to technical problems (movement in the scanner exceeding 1 voxel (3.5 mm) or excessive drop-out in the functional images). An additional 5 subjects had missing genotype information. Thus data from 58 subjects were analysed.
All participants were right handed and had normal or corrected-to-normal vision. Individuals with other major medical disorder, history of neurological disorder or head trauma were not invited for Level-3. Based on the Level-2 interviews control subjects had no personal or immediate family history of psychiatric disorders. Remitted depressed subjects had at least one past MDD episodes, no history of bipolar disorder, psychosis or OCD, and were in remission for at least 2 months (21 were more than a year, 4 were less than a year). 3 of these individuals were on antidepressant medication (fluoxetine, paroxetine, and St. John’s Wort).
eFigure 1. Summary figure of the study design and selected intermediate phenotypes
2. Measuring lifetime Depression in Level 1
The Background Questionnaire for the NewMood study (http://www.newmood.co.uk) was adapted from the questionnaire used by the Epidemiology Unit of the University of Manchester (Hunt et al, 1999) and was reported in detail previously (Juhasz et al, 2011). In short, information about age, sex, ethnicity, socio-economic background, personal medical history and family psychiatric history was collected by 22 multiple-choice, self-rating questions. Reported lifetime depression was derived from 6 questions relating to personal psychiatric disorder history:
1. Have you ever had emotional or psychiatric problem?
2. If yes, was it depression?
3. Have you ever had professional help for it?
4. Have you sought or needed help for depression in the last year?
5. If you have ever suffered from depression, how many times has this happened in your life?
6. Have you ever had any medication/treatment because of depression?
The derived Lifetime Depression category was further validated in the Level 2 interview phase. By comparing the Background Questionnaire Lifetime Depression data with the diagnosis based on the Structured Clinical Interview for DSM-IV (SCID) (8) our questionnaire has 91.7% sensitivity to identify lifetime major depression with 89.8% specificity (false positive cases: 3.4% for depression reported in Level 1 but not meeting with SCID major depression criteria; false negative cases: 4.2% for absence of depression in Level 1 but meeting SCID major depression criteria; 2.2% for other psychiatric disorder).
3. Neuropsychological and fMRI task descriptions
N-back
The n-back paradigm is a well-reported measure of working memory function (Owen et al, 2005). Working memory deficits have been demonstrated in both AD and depression using the n-back task (Harvey et al, 2005; Waltz et al, 2004). Participants were shown a sequence of random numbers on a computer screen and were asked to report either the number currently shown (0-back), the number seen on the previous screen (1-back), the number seen 2 screens back (2-back), or the number seen 3 screens back (3-back). All conditions were randomised; with the number of correct responses from each of the conditions used as an index of working memory performance.
Stockings of Cambridge (SoC)
The SoC is a well reported task that has previously demonstrated executive deficits in depression (Beats et al, 1996; Elliott et al, 1996). Executive dysfunction is also found in AD, and has been reported even in early stages of the disease (Perry and Hodges, 1999; Waltz et al, 2004). The SOC task was presented using the Cambridge Neuropsychological Test Automated Battery software (CANTAB; http://www.camcog.com/) on a dedicated touch-screen computer. The task is a computerised variation of the Tower of London test (Shallice, 1982) in which individuals have to move coloured balls to match a target configuration using set rules (Owen et al, 1995). The task begins with versions of the puzzle that require 2 moves to complete with difficulty levels increasing to puzzles that require 5 moves to complete. The 3 main outcome measures included the initial thinking time (ITT), the subsequent thinking time (STT), and the average number of moves used for each of the 4 conditions. These measures were all taken as indices of the participant’s planning and executive function ability.
Emotional word memory
This task was chosen in order to investigate the interaction between the TOMM40 risk allele and affective memory bias. The emotional word memory task was based on a paradigm developed by Harmer (Harmer et al, 2009) and was divided into a learning session, an immediate recall session, and a delayed recall session. For the learning session participants were presented with a randomised list of 30 words (10 positive, 10 negative, 10 neutral, all matched on length and frequency), and were given 5 minutes to commit the words to memory. Participants were then distracted for approximately 2 minutes before beginning immediate recall. During immediate recall participants were asked to write down all the words they could remember within 5 minutes. This procedure was repeated after 30 minutes to constitute delayed recall. For both immediate and delayed recall the 2 main outcome measures included the total number of items correctly recalled for each valence, and the total number of intrusions for each valence, taken as any item recalled that was not on the original list.
Emotional faces fMRI paradigm
The emotional face processing paradigm was based on the work of Harmer et al. (2001) and has been reported previously (Thomas et al, 2011b). During the task participants were asked to identify the gender of faces displaying different emotions. The presented faces were based on the work of Ekman & Friesen (1976) and comprised 6 actors (3 male and 3 female) demonstrating either a neutral expression, happiness, sadness, or fear. The faces were displayed on the screen for 3000ms with an inter-stimulus interval (ISI) of 500ms in which the participants could respond. The task was designed in blocks containing 6 faces of a single emotion (one from each actor). The blocks were pseudo-randomised with a neutral block at the beginning of the task, the end of the task, and after each emotion. Participant responses were made via a fibre-optic button box held in the right hand.
4. Further genetic information
The ApoE genetic locus
eFigure 2. HapMap of the ApoE genetic locus
Figure was created with Haploview 4.2 software (Barrett et al, 2005) (http://www.broadinstitute.org/scientific-community/science/programs/medical-and-population-genetics/haploview/haploview) based on the HapMap CEU+TSI population (http://hapmap.ncbi.nlm.nih.gov/).
SNP selection
The most significant SNP for Alzheimer’s disease based on the NHGRI GWAS Catalog (Available at www.genome.gov/gwastudies. Accessed October 2010) was selected for our study. The genotype and allele frequency of rs2075650 in the HapMap CEU population can be seen in eTable 1.
eTable 1: The genotype and allele frequency of rs2075650 in the HapMap CEU populationGenotype / Allele / Hardy-Weinberg equilibrium
rs2075650 / AA / AG / GG / A / G / p-value
HapMap CEU / 72% / 25% / 3% / 84% / 16% / 0.439
Details of genotyping
Genotyping was performed using buccal mucosa cells taken with a cytology brush and 15ml tubes containing 2ml of buffer solution. DNA extraction was performed using a protocol suggested by Freeman et al. . The TOMM40 rs2075650 SNP was genotyped using the Sequenom® MassARRAY technology (Sequenom®, San Diego). The IplexTM assay was followed according to manufacturers instructions (http://www.sequenom.com) using 25ng of DNA. All laboratory work was performed under the ISO 9001:2000 quality management requirements and genotyping was blinded for the phenotype. The genotyping call rate for TOMM40 rs2075650 SNP was 99%.
5. Further statistical information
pFDR correction
The multiplicity correction applied individually to all hypothesis test in Level-1 and all hypothesis test in Level-2 was the positive FDR (pFDR) with its p-value analogue the q-value, as integrated into the Qvalue software. The pFDR method involves using the information in the uncorrected p-values to estimate the minimum FDR incurred for each test. We then reject or fail to reject based on q ≤ α, setting α at the standard 0.05. This method has been shown to have greater power over equivalent FDR methods, such as the Benjamini–Hochberg method (Storey, 2002) as the number of tests increase, and is therefore a more appropriate correction tool in the exploratory multiple hypothesis testing of the current study. For more details on the pFDR method see Storey (2002), Storey & Tibshirani (2003), and Storey, Taylor, & Siegmund (2004).
6. Additional results
Medication effects in Level-3
To investigate the possible influence of including medicated participants in our Level-3 imaging models we re-formulated our model whilst excluded the three medicated participants (all AA carriers). This resulted in a marginal loss of significance for the PCC (p(FWE) = 0.070), with a larger change in significance for the ACC (p(FWE) = 0.112). Because these shifts may have resulted from the loss of error degrees of freedom we explored the impact of exclusion in the section on power below.
7. Power calculations for genetic association studies
Quanto version 1.2.4 (http://hydra.usc.edu/gxe) was employed to calculate the power of the recruited populations. In our Level-1 cohort (n = 1220, with a minor allele frequency ≥ 10%) we have higher than 93% power to detect genetic effects, where explanations are associated with 1.5 odds ratio for a disease or 1% of variations in a continuous variable. We also have 99% power to detect a gene x depression status interaction that explains 2% of variations in a continuous variable.