A Double Dissociation in the Roles of Serotonin and Mood in Healthy Subjects Oliver J

BRIEF REPORTS A Double Dissociation in the Roles of Serotonin and Mood in Healthy Subjects Oliver J. Robinson and Barbara J. Sahakian Background: Affective disorders are associated with altered cognitive performance. However, the precise interaction between affect and cognition is unclear. The manipulation of serotonin (5-HT), a neurotransmitter implicated in affect, influences performance on “hot” cognitive tasks that require the processing of affective stimuli, but manipulation of affect via mood induction influences performance on “cold” cognitive tasks that do not involve affective stimuli. We attempted to disentangle the influence of affect on cognition by examining the effect of manipulating both serotonin (via acute tryptophan depletion [ATD]) and mood on established hot and cold cognitive tasks. Methods: In a double blind, placebo-controlled crossover design, 33 healthy mood-induced (positive, negative, or neutral) subjects completed the (hot) affective go/no-go (AGNG) and (cold) one touch tower (OTT) following both placebo and ATD. Results: Mood influenced performance on the OTT but not AGNG; ATD influenced performance on the AGNG but not OTT. Conclusions: A double dissociation was demonstrated between the influence of ATD and mood on cognition, indicating that serotonin and mood are not closely linked. We hypothesize that this is due to the differences between emotions and moods and that aberrant cognition in affective disorders may be provoked through both bottom-up and top-down mechanisms.

Key Words: Affect, cognition, depression, mania, mood, serotonin of affective stimuli and may therefore influence “hot” tasks in a bottom-up fashion (5,6,12). If mood induction influences pro- ffect and cognition are dissociable, but interacting, mental cessing within prefrontal regions (such as those recruited by the processes (1,2). Correspondingly, affective disorders, Tower of London task [13,14]), it may influence performance on A such as depression and mania, are associated with “cold” (but executive) cognitive tasks via a more top-down changes in cognitive performance that can help to maintain the mechanism (9). disease state (3). Understanding the interaction between affect We therefore tested this hypothesis by manipulating 5-HT (via and cognition is therefore crucial to a full understanding of acute tryptophan depletion [ATD]) in three groups of individuals affective disorders. undergoing positive, negative, or neutral mood induction. All Cognitive processes can be divided into “cold” processes that subjects completed the “hot” affective go/no-go (AGNG), which are purportedly independent of affect and “hot” processes that is influenced by ATD (6), and the “cold” one touch tower (OTT), require the processing of affective information (4). Conse- which is not influenced by ATD (6). We predicted that ATD quently, experimental manipulations of neurotransmitter systems would influence the AGNG (via a bottom-up mechanism) and implicated in affective processing, such as serotonin (5-HT) (5), that mood would influence the OTT (via a top-down mecha- alter performance on “hot” cognitive tasks (6,7), while leaving nism). performance on “cold” cognitive tasks intact (6). However, manipulation of affect via other methods, such as mood induction, can influence performance on the same “cold” tasks (1,8). Methods and Materials Affect therefore influences cognition beyond simply the pro- Experimental Procedure cessing of affective information. One reason for this is that affect Procedures were approved by the Norfolk Research Ethical is a multidimensional process that can be broken down into (at Committee (06/Q0101/5). Thirty-four subjects (18 female sub- least) emotions, which are typically short-lived affective states jects) were screened for psychiatric and neurological disorders triggered by specific stimuli and associated with specific auto- (Table 1, Supplement 1). One subject did not complete either nomic changes; and moods, which are long-term background session, two subjects completed a single session, and one subject affective states with cumulative or unclear causes (1). Emotions experienced technical difficulties with the AGNG. Subjects were often originate in midbrain and lower corticolimbic regions (9), assigned, double-blind, to the ATD-first group (n ϭ 15) or whereas mood states are often subserved by prefrontal cortical nutritionally balanced (BAL)-first group (n ϭ 19). Subjects were regions (1,10,11). assigned to negative (n ϭ 12, 4 female subjects), neutral (n ϭ 10, This distinction may therefore help to clarify the role of 5-HT 6 female subjects), and positive mood groups (n ϭ 11, 7 female and mood in cognition. If 5-HT primarily alters emotional subjects) and tested on two sessions separated by at least 1 week. processing, it will influence tasks that require the fast processing They were asked to consume only water from midnight prior to

each session. At T0, a blood sample was taken and a nutritionally From the Department of Psychiatry and Behavioural and Clinical Neuro- balanced (BAL) or a tryptophan free (ATD) amino acid drink was science Institute, University of Cambridge, Cambridge, United Kingdom. ingested. After approximately 5 hours, a second blood sample Address reprint requests to Oliver J. Robinson, B.A., University of Cambridge, was taken (T1). Department of Psychiatry, Box 189, Level E4, Addenbrooke’s Hospital, Subjects then completed mood induction procedures (MIP), Hills Road, Cambridge CB2 2QQ, United Kingdom; E-mail: ojr23@cam. followed by the affective go/no-go and then one touch tower ac.uk. tasks. Visual analogue scales (VAS) were completed to determine Received August 4, 2008; revised September 15, 2008; accepted October 1, self-reported mood state. Further details are included in Supple- 2008. ment 1.

Data Analysis All data were analyzed via repeated-measures analysis of variance (ANOVA) in SPSS 10 (SPSS Inc, Chicago, Illinois). Error rates were square-root transformed. Simple effects were calcu- lated from the estimated marginal means.

Results

Blood Sample Analysis There was a significant two-way drink ϫ time interaction for the critical tryptophan (TRP)/total long neutral amino acids (ΑLNAA) ratio [drink ϫ time: F(1,23) ϭ 71.2, p Ͻ .0001]. Simple effects analysis revealed that this was due to a 84.7% decrease in Α the TRP/ LNAA ratio between T0 and T1 in the ATD group [simple effect of time: t (23) ϭ 12.2, p Ͻ .0001] but a 21.7% Α increase in the TRP/ NAA ratio between T0 and T1 in the BAL group [simple effect of time: t (23) ϭϪ2.9, p ϭ .007].

Self-Report Mood Figure 1. Performance on the “hot” AGNG is influenced by serotonin but ϭ ϭ not mood. Female subjects make significantly more errors in response to There was no treatment by time [F(1,29) 1.4, p .25] happy distracters than sad distracters (valence difference ϭ happy - sad) interaction between T0 and T1 on the happy - sad VAS, but there under placebo (BAL). This bias is abolished by acute tryptophan deple- was a time by MIP interaction between T1 and T2 before and after tion (ATD). * p Ͻ .05. AGNG, affective go/no-go; ATD, acute tryptophan the MIP [F(2,25) ϭ 3.8, p ϭ .035]. This demonstrates that the depletion; BAL, nutritionally balanced; N.S., not significant. mood induction, but not ATD, successfully altered subjects’ moods. Simple effects analysis is presented in Supplement 1. F(5,24) ϭ .49, p ϭ .78] or by a treatment ϫ mood interaction [treatment ϫ mood ϫ difficulty: F(10,50) ϭ .67, p ϭ .75]. Simple Affective Go/No-Go effects analysis revealed that this was due to a significant increase There was no interaction between mood state, treatment, and in the number of attempts required to complete the harder word valance in the distracter (no-go) errors [n ϭ 32, F(2,28) ϭ problems in subjects under negative [main effect of difficulty: 1.3, p ϭ .30] or between mood state and word valence [F(2,28) ϭ n ϭ 12, F (5,24) ϭ 3.62, p ϭ .014] or positive mood [n ϭ 11, .62, p ϭ .55]. However, there was a significant interaction F (5,24) ϭ 3.68, p ϭ .013] but not neutral mood [n ϭ 10, between gender, treatment, and word valence [n ϭ 32, F(1,29) ϭ F (5,24) ϭ 1.71, p ϭ .17] (Figure 2, Supplement 1). 8.8, p ϭ .006]. Simple effects revealed an interaction between treatment and word valence in female [n ϭ 16, F(1,14) ϭ 7.2, p ϭ Discussion .018] but not male subjects [n ϭ 16, F(1,15) ϭ 1.7, p ϭ .21]. This female specific effect was driven by a significant increase in We demonstrate a double dissociation in the influence of errors in response to happy distracter words (i.e, a positive bias) serotonin and mood on cognition. The affective go/no-go task on placebo [main effect of valence: F(1,14) ϭ 8.3, p ϭ .009] but was mediated by serotonin but not mood manipulation, whereas not following ATD [F(1,15) ϭ 2.4, p ϭ .14]. (Figure 1, Table 2, performance on the one touch tower task was mediated by mood Supplement 1). but not serotonin manipulation. This finding is, to the best of our knowledge, the first One Touch Tower of London experimental evidence that serotonin and mood are not closely There was a significant interaction between mood state and linked. A recent meta-analysis (15) found no effect of serotonin difficulty on the number of attempts required to complete manipulation on the mood state of healthy individuals. However, problems [MIP ϫ difficulty: n ϭ 33, F(10,50) ϭ 2.3, p ϭ .017], the reviewed studies largely relied on self-report of mood which was not confounded by treatment [treatment ϫ difficulty: following 5-HT manipulation. Here, by manipulating both and by demonstrating a double dissociation in their influence over Table 1. Group Demographic and Trait Characteristics cognition, we indicate that 5-HT and mood state cannot be closely linked. This is of clear importance to our understanding Negative Positive Neutral of serotonergic function and its role in affective disorders. Measure (SD) (SD) (SD) Fp Table 2. Affective Go/No-Go Age 26.1 (6.5) 23.7 (5.0) 22.5 (3.0) 1.41 .26 BDI 3.5 (3.0) 4.4 (2.5) 6.7 (4.4) 2.68 .08 Happy Words Sad Words BIS 19.5 (3.3) 19.9 (2.7) 19.3 (3.8) .11 .90 BAS 39.0 (4.4) 37.2 (7.4) 38.5 (5.5) .28 .76 Female IVE-Implusiveness 7.0 (4.9) 6.9 (3.1) 8.8 (4.4) .72 .49 BAL 4.40 (.81) 2.80 (.88) IVE-Venturesomeness 9.7 (4.7) 10.0 (4.4) 10.5 (4.3) .08 .93 ATD 2.60 (.74) 3.67 (.73) IVE-Empathy 11.1 (4.3) 13.7 (2.3) 13.0 (3.0) 1.87 .17 Male Barrat Impulsiveness BAL 3.19 (.78) 3.94 (.85) Scale 63.8 (10.6) 60.6 (10.7) 64.5 (13.3) .35 .70 ATD 3.63 (.72) 3.25 (.71) ANOVA reveals the groups to be matched (F/p). Distracter errors on the affective go/no-go within each condition (acute ANOVA, analysis of variance; BDI, Beck Depression Inventory-II; BIS, be- tryptophan depletion [ATD] vs. placebo [BAL]/happy vs. sad words) for both havioral inhibition system score; BAS, behavioral activation system score; male and female subjects (SEM). IVE, Impulsiveness Venturesomeness Empathy questionnaire. ATD, acute tryptophan depletion; BAL, nutritionally balanced placebo. www.sobp.org/journal O.J. Robinson and B.J. Sahakian BIOL PSYCHIATRY 2009;65:89–92 91

that mood and 5-HT are not closely linked in healthy individuals. While both manipulations influence affect, it may be that ATD mediates emotions and influences “hot” cognition via a bottom-up mechanism, whereas mood influences “cold” cognition via a top-down mechanism. This framework may help us to understand the influence of affect on cognition and hence the changes in cognition seen in affective disorders.

This work was conducted within the Behavioural and Clini- cal Neuroscience Institute, which is co-funded by the Medical Research Council (MRC) and the Wellcome Trust. The research was funded by a programme grant from The Wellcome Trust (Grant Number 076274/Z04/Z to T.W. Robbins, B.J. Everitt, A.C. Roberts, and B.J. Sahakian). We are grateful to Stuart Fuller and the staff of the Wellcome Trust Clinical Research Facility, Addenbrooke’s Hospital, Cam- bridge. We thank Mike Franklin for analysis of plasma data. OJR Figure 2. The number of attempts required to complete the “cold” one touch tower planning task is mediated by mood state, but not the serotonin: holds an MRC Research Studentship. subjects in either a positive or negative mood make signiﬁcantly more Oliver Robinson reports no biomedical financial interests or mistakes on harder problems. NEG, negative mood; NEUT, neutral mood; potential conflicts of interest. Professor Barbara Sahakian con- POS, positive mood. sults for Cambridge Cognition. She holds shares in CeNeS. She has consulted for Novartis, Shire, GlaxoSmithKline, and Lilly. She The finding that 5-HT mediates performance on the AGNG has also received honoraria for Grand Rounds in Psychiatry at has been demonstrated before (6) and indicates that reduced Massachusetts General Hospital (continuing medical education 5-HT, rather than negative mood, causes the disruption of AGNG [CME] credits) (Boston, 27 April 2007) and for speaking at the performance found in depression (16). The restriction of this International Conference on Cognitive Dysfunction in Schizo- finding to female subjects is redolent of previous findings (15,17) phrenia and Mood Disorders: clinical aspects, mechanisms, and and suggests that women are more susceptible to the effects of therapy (Brescia, 17–19 January 2007). She is on the Medical 5-HT fluctuation than men. This, in turn, may underlie the Research Council Neurosciences and Mental Health Board and increased incidence (2:1) of depression in women. on the Science Coordination Team for the Foresight Project on The second finding, that mood state mediates performance on Mental Capital and Wellbeing (Office of Science, The Depart- the OTT, replicates findings from the original Tower of London ment of Innovation, Universities and Skills). As an Associate task (8,18) and demonstrates that both positive and negative Editor, she also receives an honorarium from the journal Psy- mood can impair planning ability. Mood state, rather than altered chological Medicine. serotonin, is therefore likely to cause the impairments in plan- Supplementary material cited in this article is available ning found in depression (19) and during the manic phase of online. mania (20), although additional factors may contribute to executive dysfunction during euthymia. 1. Mitchell RLC, Phillips LH (2007): The psychological, neurochemical and Integrating these findings, it may be that serotonin acts on functional neuroanatomical mediators of the effects of positive and emotion perception systems and influences cognition in a bot- negative mood on executive functions. Neuropsychologia 45:617–629. tom-up fashion (5,12), whereas mood disrupts more complex 2. Gray JR, Braver TS, Raichle ME (2002): Integration of emotion and cog- cognitive processes in higher cortical regions (such as lateral nition in the lateral prefrontal cortex. Proc Natl Acad SciUSA99:4115– prefrontal cortex [LPFC]) (2) through a more top-down route. 4120. 3. Tavares JVT, Drevets WC, Sahakian BJ (2003): Cognition in mania and This is broadly consistent with recent models of emotion pro- depression. Psychol Med 33:959–967. cessing that posit the presence of ventral and dorsal streams of 4. Roiser JP, Rubinsztein JS, Sahakian BJ (2003): Cognition in depression. affective processing (9) and merits further research. Resistance to Psychiatry 2:43–47. these pressures may, furthermore, contribute toward resilience to 5. Cools R, Roberts AC, Robbins TW (2008): Serotoninergic regulation of affective disorders. emotional and behavioural control processes. Trends Cogn Sci 12:31–40. As a caveat, these findings may be specific to the tasks 6. Murphy FC, Smith KA, Cowen PJ, Robbins TW, Sahakian BJ (2002): The effects of tryptophan depletion on cognitive and affective processing in studied. Cognitive processes that require the integration of healthy volunteers. Psychopharmacology (Berl) 163:42–53. executive processing with emotional processing may be influ- 7. Chamberlain SR, Muller U, Blackwell AD, Clark L, Robbins TW, Sahakian enced by both mood and 5-HT (which would explain recent BJ (2006): Neurochemical modulation of response inhibition and prob- findings in which mood state and 5-HT interact to bias cognition) abilistic learning in humans. Science 311:861–863. (17), and simple “cold” tasks, which do not require higher 8. Phillips LH, Smith L, Gilhooly KJ (2002): The effects of adult aging and prefrontal processing, may be unaffected by both mood and induced positive and negative mood on planning. Emotion 2:263–272. 9. Phillips ML, Drevets WC, Rauch SL, Lane R (2003): Neurobiology of emo- ATD. A further caveat is that we varied 5-HT within subjects but tion perception I: The neural basis of normal emotion perception. Biol mood between subjects. Future research should vary mood Psychiatry 54:504–514. induction within subjects to remove the potential confound of 10. Habel U, Klein M, Kellermann T, Shah NJ, Schneider F (2005): Same or within- versus between-subject effects on task performance. different? Neural correlates of happy and sad mood in healthy males. Neuroimage 26:206–214. 11. Mayberg HS, Liotti M, Brannan SK, McGinnis S, Mahurin RK, Jerabek PA, Conclusions et al. (1999): Reciprocal limbic-cortical function and negative mood: In summary, we demonstrate a double dissociation in the Converging PET ﬁndings in depression and normal sadness. Am J Psy- influence of 5-HT and mood on cognition and therefore suggest chiatry 156:675–682.

www.sobp.org/journal 92 BIOL PSYCHIATRY 2009;65:89–92 O.J. Robinson and B.J. Sahakian

12. Harmer CJ (2008): Serotonin and emotional processing: Does it help medicated patients with major depressive disorder. Am J Psychiatry explain antidepressant drug action [published online ahead of print 162:2171–2173. June 17]? Neuropharmacology. 17. Robinson OJ, Cools R, Crockett M, Sahakian BJ (in press): Mood State 13. Newman SD, Carpenter PA, Varma S, Just MA (2003): Frontal and parietal moderates the role of serotonin in cognitive biases. J Psychopharmacol. participation in problem solving in the Tower of London: fMRI and 18. Oaksford M, Grainger B, Morris F, Williams JMG (1996): Mood, reasoning, computational modeling of planning and high-level perception. Neuro- and central executive processes. J Exp Psychol Learn Mem Cogn 22:476– psychologia 41:1668–1682. 492. 14. Duncan J, Owen AM (2000): Common regions of the human frontal lobe 19. Elliott R, Sahakian BJ, Herrod JJ, Robbins TW, Paykel ES (1997): Ab- recruited by diverse cognitive demands. Trends Neurosci 23:475–483. normal response to negative feedback in unipolar depression: Evi- 15. Ruhe HG, Mason NS, Schene AH (2007): Mood is indirectly related to dence for a diagnosis speciﬁc impairment. J Neurol Neurosurg Psychi- serotonin, norepinephrine and dopamine levels in humans: A meta- atry 63:74–82. analysis of monoamine depletion studies. Mol Psychiatry 12:331–359. 20. Murphy FC, Sahakian BJ, Rubinsztein JS, Michael A, Rogers RD, Robbins 16. Erickson K, Drevets WC, Clark L, Cannon DM, Bain EE, Zarate CA, et al. TW, et al. (2000): Emotional bias and inhibitory control processes in (2005): Mood-congruent bias in affective go/no-go performance of un- mania and depression. Psychol Med 29:1307–1321.