Linking variability in brain chemistry and circuit function through multimodal human neuroimaging


  • P. M. Fisher,

    Corresponding author
    1. Center for Integrated Molecular Brain Imaging
    2. Neurobiology Research Unit, Copenhagen University Hospital Rigshospitalet, Copenhagen, Denmark
    Search for more papers by this author
  • A. R. Hariri

    1. Laboratory of NeuroGenetics, Department of Psychology & Neuroscience
    2. Institute for Genome Sciences & Policy, Duke University, Durham, North Carolina, USA
    Search for more papers by this author

Dr P. M. Fisher, PhD, Center for Integrated Molecular Brain Imaging, Neurobiology Research Unit, NRU 9201, Rigshospitalet, Blegdamsvej 9, Copenhagen O DK-2100, Denmark. E-mail:


Identifying neurobiological mechanisms mediating the emergence of individual differences in behavior is critical for advancing our understanding of relative risk for psychopathology. Neuroreceptor positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) can be used to assay in vivo regional brain chemistry and function, respectively. Typically, these neuroimaging modalities are implemented independently despite the capacity for integrated data sets to offer unique insight into molecular mechanisms associated with brain function. Through examples from the serotonin and dopamine system and its effects on threat- and reward-related brain function, we review evidence for how such a multimodal neuroimaging strategy can be successfully implemented. Furthermore, we discuss how multimodal PET-fMRI can be integrated with techniques such as imaging genetics, pharmacological challenge paradigms and gene–environment interaction models to more completely map biological pathways mediating individual differences in behavior and related risk for psychopathology and inform the development of novel therapeutic targets.

The capacity to assay various features of brain chemistry, structure and function using in vivo neuroimaging has provided a wealth of information about how neural circuitry shapes human behavior and related risk for psychopathology. Functional magnetic resonance imaging (fMRI) has provided detailed mappings between interindividual differences in the response of specific brain regions, distributed neural circuits, complex behaviors, personality traits and psychopathologic states. In parallel, neuroreceptor positron emission tomography (PET) has provided insight into specific molecular mechanisms associated with these same emergent phenomena. Despite their highly complementary information about brain function and brain chemistry, these methodologies and their idiosyncratic strengths have often been applied independently, limiting the degree to which either can inform neurobiological mechanisms of behavior and psychopathology. For example, fMRI can be used to assay interindividual variability in the response of the amygdala to threat within the environment or correlated activity between the amygdala and prefrontal cortex, offering a measure of distributed threat-related corticolimbic circuit function. Detailed molecular information such as the regional availability of serotonin (5-HT) transporters (5-HTT) within this circuit can be obtained through neuroreceptor PET.

Integrating these two neuroimaging techniques offers a unique opportunity to establish associations between such specific molecular mechanisms and aspects of behaviorally relevant brain function. When integrated with other methodological approaches such as imaging genetics and gene-by-environment (GxE) interaction studies, this multimodal neuroimaging strategy is poised to facilitate a more complete understanding of molecular and neurobiological pathways through which genetic and environmental mechanisms contribute to interindividual variability in behavior and psychopathology (Figure 1). Establishing these links furthers the degree to which genetic variation and assessment of environmental experience, which are less invasive and costly than neuroimaging, can be used to model underlying neurobiology and evaluate its impact on behavior and related risk for psychopathology (Caspi et al. 2010). Here, we discuss studies from our lab and others that have applied this multimodal neuroimaging approach to establish novel links between specific molecular mechanisms and aspects of brain function. We further highlight unique challenges of this research strategy and outline specific avenues for future work.

Figure 1.

Example model for evaluating molecular mechanisms underlying neural circuit function, complex behavioral phenotypes, and related risk for psychopathology utilizing a multimodal neuroimaging strategy. Serotonergic signaling pathways, which can be evaluated with neuroreceptor PET, modulate the function of a distributed threat-related corticolimbic circuit, which can be assayed with fMRI. Integration of this multimodal neuroimaging approach with complementary strategies such as imaging genetics or gene–environment interaction modeling can further show how the relationships between these neurobiological mechanisms, complex behavior and related risk for psychopathology may be moderated by common functional genetic polymorphisms, environmental factors and/or their interactions.

A brief description of methodology

Before describing recent multimodal PET-fMRI studies that highlight this integrative approach, we briefly consider each methodology. For a more comprehensive review of these neuroimaging modalities, we direct the reader to other sources (Huettel et al. 2008; Price et al. 2010; Wernick & Aarsvold 2004).

fMRI refers to a class of noninvasive neuroimaging techniques using MRI technology to measure neural activity. The most commonly employed fMRI technique assays changes in the oxygenation of hemoglobin in blood associated with changes in neural activity. This blood oxygen level-dependent (BOLD) signal is an indirect but faithful index of neuronal activity, particularly local field potentials generated by neuronal ensembles supporting information processing (Lee et al. 2010; Logothetis et al. 2001). In a typical BOLD fMRI experiment, BOLD signal is measured during a task or active condition relative to BOLD signal during a baseline condition. Single-subject task-related BOLD signal is then evaluated across a cohort of participants to measure interindividual differences in task-related brain function. The general linear model is most commonly applied to evaluate associations between BOLD fMRI measured brain activity and variables of interest. Additional measures can be calculated from BOLD fMRI time series data, including functional connectivity, which is a measure of the correlation in BOLD signal between two brain regions over time, putatively reflecting neural circuit dynamics (Friston et al. 1997).

Neuroreceptor PET is a neuroimaging technique using an organic molecule called a ‘ligand’, which has both a high affinity and specificity for a specific target molecule (e.g. 5-HTT). A positron-emitting radioisotope is chemically attached to the ligand (often carbon-11 or fluorine-18), which is injected into the bloodstream and distributes throughout the body and brain, binding to its molecular target. Signal identified by a PET scanner is derived from the decay of the radioisotope and will be greater in areas where there are more molecular targets available for binding. Although a radioligand can have agonist or antagonistic properties at its molecular target the quantity of injected radioligand is so low as to have presumably negligible effects on targeted signaling pathways. When possible, quantified signal within a region- or voxel-of-interest is standardized with respect to a region presumed to be devoid of the target (reference region). For most radioligands, the primary outcome measure is binding potential (BP). Although there are variants in how BP is reported (e.g. BPF, BPP and BPND), all measures of BP are proportional to Bmax/KD, where 1/KD is the inverse of the dissociation constant and Bmax is the number of target molecules available for binding to the radioligand (Innis et al. 2007).

The collection of both fMRI and PET imaging data sets within a single cohort offers the unique opportunity to link task-related brain activity and underlying brain chemistry. Although this requires sequential neuroimaging scan sessions with different scanner systems (i.e. MRI and PET), it represents the best available method for linking brain chemistry and brain function in humans. Particular information offered by each scanning technique can be combined through this integrative multimodal neuroimaging strategy to obtain a better understanding of the neurobiological mechanisms underlying complex behavioral phenotypes and related psychopathology.

Threat sensitivity, corticolimbic circuitry and monoaminergic signaling

Individuals with high trait anxiety are at increased risk for mood and anxiety disorders (Kendler et al. 2006; Kotov et al. 2010). Identifying neurobiological mechanisms related to trait anxiety is critical for understanding how these individual differences emerge and for developing novel strategies for treating and preventing the related psychopathology (Hariri 2009). A corpus of research in humans and nonhuman animal models clearly implicates a corticolimbic circuitry comprising structural and functional interconnections between the amygdala and medial prefrontal cortex (mPFC) in generating and regulating both behavioral and physiologic responses to threatening stimuli (Hariri & Holmes 2006; Holmes 2008; Kim et al. 2011; Quirk & Mueller 2008). Similarly, a wealth of studies on both animal models and human studies implicate 5-HT and dopamine (DA) signaling in modulating sensitivity to threat through its effects on function of this corticolimbic circuitry (Bigos et al. 2008; Fakra et al. 2009; Forster et al. 2006; Hariri et al. 2002; Harmer et al. 2006; Holmes et al. 2003; Rosenkranz & Grace 2002).

Regulation of serotonin release and threat-related amygdala reactivity

The inhibitory 5-HT1A receptor, expressed within the dorsal raphe nucleus (DRN) as a somatodendritic autoreceptor, and 5-HTT provide direct negative feedback on 5-HT release and signaling at downstream targets (Barnes & Sharp 1999; Blakely et al. 1994; Blier et al. 1998; Riad et al. 2000). Preclinical studies on animal models strongly implicate 5-HT1A autoreceptor function in anxiety-related behavioral phenotypes and sensitivity to antidepressant treatments, many of which have their pharmacological effect via blockade of the 5-HTT (Richardson-Jones et al. 2010, 2011). Studies on animal models suggest that augmentation of 5-HT levels potentiates the response of the amygdala to threat (Burghardt et al. 2007; Christianson et al. 2010; Forster et al. 2006; Li et al. 2012). Using a PET-fMRI multimodal neuroimaging approach, two studies evaluated the association between individual variability in 5-HT1A autoreceptor binding or local amygdala 5-HTT binding, assessed with [11C]WAY100635 and [11C]DASB PET, respectively, and threat-related amygdala reactivity, assessed with fMRI in independent cohorts of healthy humans (Fisher et al. 2006; Rhodes et al. 2007). This novel multimodal neuroimaging strategy offered the opportunity to identify whether a specific molecular mechanism (i.e. capacity to negatively regulate 5-HT release) was associated with interindividual variability in threat-related amygdala reactivity in healthy humans.

Within a study from our lab, we found that DRN 5-HT1A autoreceptor binding was significantly inversely correlated with threat-related amygdala reactivity (Fisher et al. 2006). Rhodes et al. (2007) reported that local amygdala 5-HTT binding was significantly inversely correlated with threat-related amygdala reactivity. Each of these findings is consistent with a model wherein a greater capacity to regulate serotonin release (via negative autoreceptor feedback or local transporter-mediated extracellular clearance) is associated with reduced amygdala reactivity to threat-related stimuli (see Figure 2 from Hariri & Holmes 2006). These findings provide evidence for molecular mechanisms through which the regulation of 5-HT signaling modulates threat-related amygdala reactivity. Notably, these findings link molecular mechanisms (i.e. reduced 5-HT1A autoreceptor and 5-HTT) with a specific aspect of brain function (increased amygdala reactivity to threat), which have been implicated in the pathophysiology of mood and anxiety disorders (Drevets et al. 2007; Parsey et al. 2010; Smith & Jakobsen 2009; Stuhrmann et al. 2011). This multimodal neuroimaging finding extends these studies and provides in vivo evidence in humans linking the capacity to regulate 5-HT signaling and response to threat.

Figure 2.

Examples of various modeling techniques for assessing effects of molecular mechanisms underlying neural circuit function. One or more neuroreceptor PET data sets can be used to evaluate direct or interactive effects of serotonergic signaling on threat-related circuit function. This multimodal neuroimaging technique can be integrated with common genetic variants to evaluate moderated or mediated effects of genetic variation on serotonergic signaling, neural circuit function, threat-related behaviors and related risk for psychopathology. Although they do not imply causation, associations identified through multimodal neuroimaging implicates specific molecular mechanisms in modulating underlying neural function related to complex behavioral phenotypes.

Postsynaptic serotonin 1A and 2A receptors associated with amygdala reactivity, habituation and functional connectivity with prefrontal cortex

Within prefrontal cortex, the excitatory 5-HT2A receptor is predominantly localized on glutamatergic neurons at proximal portions of the apical dendrite (de Almeida & Mengod 2007; Jakab & Goldman-Rakic 1998). Studies on animal models and humans implicate 5-HT signaling and the 5-HT2A receptor in modulating prefrontal function in the context of fear- or anxiety-related behaviors and depression (Bhagwagar et al. 2006; Forster et al. 2006; Frokjaer et al. 2008; Weisstaub et al. 2006). We evaluated the association between mPFC 5-HT2A binding, assessed with [18F]altanserin PET, and threat-related corticolimbic circuit function in a cohort of healthy individuals (Fisher et al. 2009). Consistent with its localization and how that suggests 5-HT2A signaling would modulate corticolimbic circuit function, we found that mPFC 5-HT2A binding was significantly inversely correlated with threat-related amygdala reactivity. Additionally, mPFC 5-HT2A binding was positively correlated with habituation of the amygdala response over time and with functional connectivity between the amygdala and mPFC. Notably, our finding that mPFC 5-HT2A binding is associated with multiple facets of distributed corticolimbic circuit function provides compelling evidence for its role in mediating the effects of 5-HT signaling on this circuitry in a manner that is consistent with its localization in this reciprocal neural circuitry.

Next, we considered whether this effect was moderated by the postsynaptic 5-HT1A receptor, which is highly colocalized with the 5-HT2A receptor in mPFC, situated proximal to the cell body and thus a potential mediator of 5-HT signaling on glutamatergic neuronal excitability (Amargos-Bosch et al. 2004; de Almeida & Mengod 2008). On the basis of this colocalization, the inhibitory 5-HT1A receptor is situated to negatively moderate or ‘gate’ the capacity for the excitatory 5-HT2A receptor to facilitate regulation of threat-related amygdala reactivity. Consistent with this model, we found that mPFC 5-HT1A binding significantly moderated the negative association between 5-HT2A binding and threat-related amygdala reactivity such that mPFC 5-HT2A binding was significantly inversely correlated with amygdala reactivity, but only when mPFC 5-HT1A binding was relatively low (see Figure 4 from Fisher et al. 2011). These findings indicate that the effects of mPFC 5-HT1A and 5-HT2A receptors interact in mediating the effects of 5-HT signaling on threat-related corticolimbic circuit function, consistent with their colocalization within prefrontal cortex and their opposing effects on excitability of glutamatergic neurons (Puig & Gulledge 2011).

These four studies highlight how a PET-fMRI multimodal neuroimaging strategy can link specific molecular mechanisms and distributed neural circuit function, which is associated with individual variability in behavior and related risk for psychopathology. Multiple studies on healthy individuals have reported associations between these neuroimaging measures and behavioral indices of sensitivity to threat (Frokjaer et al. 2008; Kim et al. 2011; Neumeister et al. 2004; Pezawas et al. 2005). Similarly, previous studies on clinical cohorts have reported alterations in threat-related corticolimbic circuit function (Stuhrmann et al. 2011) as well as 5-HT1A, 5-HT2A and 5-HTT binding (Drevets et al. 2007; Smith & Jakobsen 2009). This multimodal neuroimaging approach provides a framework for linking molecular mechanisms with brain function in a manner that cannot be accomplished by either neuroimaging technique independently. The finding that prefrontal 5-HT1A binding moderates the effect of 5-HT2A binding on amygdala reactivity is consistent with the myriad known 5-HT receptors whose colocalization suggests they interact to mediate the effects of 5-HT signaling on threat-related brain function (see Figure 3 from Holmes 2008). Modeling these interactions in future studies through multitracer protocols may help further elucidate how 5-HT signaling contributes to threat-related behavioral phenotypes, related risk for psychopathology and, within clinical cohorts, sensitivity to specific therapeutic strategies through its effects on distributed neural circuit function.

DA signaling and sensitivity to threat

In addition to 5-HT, human neuroimaging and animal studies implicate DA signaling in the modulation of amygdala reactivity to threat (Hariri et al. 2002; Rosenkranz & Grace 2002). A recent PET-fMRI study evaluated the association between DA storage capacity, assessed with 6-[18F]fluoro-l-3,4-dihydroxyphenylalanine (l-DOPA) PET, and amygdala reactivity to aversive stimuli, assessed with fMRI (Kienast et al. 2008). The authors utilized a novel approach for quantifying the PET data to calculate a measure of vesicular storage of DA, thus reflecting the ‘state of readiness' for local DA release. Consistent with DA signaling within the amygdala potentiating its response to aversive stimuli, Kienast and colleagues reported a significant positive correlation between DA storage capacity and amygdala reactivity to aversive stimuli as well as reactivity of the anterior cingulate cortex to aversive stimuli.

A subsequent multimodal neuroimaging study provides further support for an impact of DA signaling within the amygdala on its response to threat in humans. Studies on animal models evaluating molecular mechanisms mediating the effects of DA signaling within the amygdala suggest the D1 and D2 receptors positively modulate excitability of amygdala projection neurons via disinhibitory mechanisms (Kroner et al. 2005; Rosenkranz & Grace 2002). In light of these findings, Takahashi et al. (2010) evaluated whether amygdala D1 or D2 receptor binding, assessed with [11C]SCH23390 and [11C]FLB457 PET, respectively, was associated with amygdala reactivity to novel and fearful stimuli, assessed with fMRI. Within the amygdala, both the D1 and D2 receptors are localized both pre- and postsynaptically and appear to be expressed on both excitatory and inhibitory neurons (Muly et al. 2009; Pinto & Sesack 2008). The authors reported that amygdala D1 binding, but not D2 binding, was significantly positively associated with amygdala reactivity to novel and fearful stimuli. It is noteworthy that an association with amygdala D1 binding emerges considering the complex effects its distribution pattern would suggest. However, this finding is consistent with the predominantly anxiogenic effects on behavior following D1 agonism in animal models (de La Mora et al. 2010).

The above-mentioned studies extend previous human neuroimaging findings implicate both pre- and postsynaptic DA signaling in the potentiation of amygdala reactivity to salient and negative stimuli. Considering the collective findings from studies of the independent effects of 5-HT and DA signaling, future work evaluating the interactions of 5-HT and DA signaling within the same cohort will advance our understanding of how monoaminergic systems contribute to threat-related corticolimbic circuit function and related behavior.

Reward anticipation & impulsivity, ventral striatum reactivity and DA release

The ventral striatum (VS) plays a key role in sensitivity to reward-related stimuli and its function is modulated by dopaminergic inputs derived from the substania nigra (SN) and ventral tegmental area (VTA; Sesack & Grace 2010). This reciprocal mesolimbic circuitry plays a critical role in reward-related behaviors such as impulsivity, which are associated with risk for psychopathic and substance abuse disorders (Everitt et al. 2008; Kalivas & Volkow 2005). Recent PET-fMRI studies have provided additional insight into molecular mechanisms mediating the effects of DA signaling on reward-related VS activity within humans.

Schott et al. (2008) evaluated the association between reward-related DA release and reward anticipation during a monetary incentive delay task (MIDT). Reward-related DA release was determined as the difference in regional D2 binding assessed during two separate [11C]raclopride PET scans: one scan while the participant completed the MIDT with many rewarding trials and another scan while the participant completed the same task, but with fewer rewarding trials. [11C]raclopride and other D2 radioligands are particularly interesting because their binding is sensitive to acute release of DA. Thus, challenges that elicit increases in DA release will result in reduced D2 binding, which can be quantified with neuroreceptor PET. Brain activation in response to reward anticipation was determined using the same reward-related paradigm in an independent fMRI scan session.

Schott and colleagues found that the magnitude of DA release within VS was positively associated with BOLD response to reward anticipation within brainstem-SN/VTA and VS. DA release within VS was also associated with BOLD response to reward anticipation in other brain regions, including hippocampus and amygdala; however, effects of reward on [11C]raclopride binding within these regions was not reported, possibly due to low signal within these regions for this radioligand. A second study using the MIDT attempted to follow-up on this result, using a similar protocol of PET and fMRI scans (Urban et al. 2011). Although Urban and colleagues observed significant reward-anticipation-related BOLD activation within VS, they did not observe significant displacement of [11C]raclopride within VS following completion of the MIDT. Methodological differences including the ratio of reward/neutral trials and task completion relative to time of PET scan may have contributed to the little change in [11C]raclopride binding being observed. These findings provide a compelling link between reactivity to reward of the DA system and reactivity to reward of the VS, which has been linked to individual variability in reward-related behaviors such as impulsive tendencies (Hariri et al. 2006).

In two recent studies, Buckholtz et al. (2010a,b) provided compelling evidence for a link between mesolimbic DA release, reward-anticipation-related striatal reactivity and the personality trait impulsivity. DA release was measured as the displacement of [18F]fallypride, a D2 antagonist with sensitivity to acute DA release like that of [11C]raclopride, following an amphetamine challenge. Reactivity of the VS to reward-anticipation was evaluated using the MIDT. Buckholtz and colleagues observed significant displacement of [18F]fallypride binding within VS following amphetamine challenge and significant reward-anticipation activity within VS, assessed with fMRI. Each of these measures was positively associated with the personality trait impulsivity and these measures were also significantly positively correlated with each other, suggesting that reward-related DA release may facilitate activation within VS in response to reward-anticipation, which in turn contributes to increased impulsivity (Buckholtz et al. 2010a). Buckholtz et al. (2010b) extended this PET finding to show that DA release within VS mediated the effect of DA release within SN/VTA on impulsivity. These findings provide a compelling link between dopaminergic signaling, its effects on reward processing and subsequent impact on individual variability in the behavioral phenotype impulsivity.

Summary of findings

Critical to understanding how neurobiological mechanisms contribute to individual variability in behavioral phenotypes and related risk for psychopathology is a thorough understanding of how molecular mechanisms affect the function of brain regions and distributed neural circuits. The findings described here highlight how a multimodal neuroimaging strategy incorporating fMRI and neuroreceptor PET can pair complimentary techniques to advance our understanding of how 5-HT and DA mechanisms contribute to interindividual variability in threat- and reward-related brain function. Although previous studies have implicated functionality of these mechanisms either within animal models or indirectly with paradigms in humans, these in vivo findings provide compelling evidence that can be extended toward evaluating how these associations in turn contribute to interindividual variability in aspects of personality and related risk for psychopathology. These associations could not have been evaluated in humans through the use of either neuroimaging modality on its own. Previous studies implicating these mechanisms in threat- and reward-related behaviors in animal models support the multimodal neuroimaging findings and provide a strong context for the interpretation of observed associations. Importantly, this multimodal neuroimaging approach provides a compelling in vivo approach in humans for furthering our understanding of neurobiological mechanisms that contribute to interindividual variability in sensitivity to threat, reward processing and related risk for psychopathology. Although we did not have the opportunity to discuss all PET-fMRI studies here, we point the reader toward additional studies within healthy cohorts (Braskie et al. 2011; Siessmeier et al. 2006) and clinical cohorts (Fusar-Poli et al. 2010; Heinz et al. 2004) that further highlight this approach.

Limitations of PET-fMRI multimodal neuroimaging

Although these neuroimaging techniques and their integration offer a unique opportunity to evaluate neurobiological mechanisms underlying aspects of personality and behavior, limitations of these techniques must be well understood in order that they are effectively applied. Neuroreceptor PET is not a direct measure of receptor function, but rather a measure related to quantity of receptors available for binding to the radioligand. The impact of radioligand binding to inactive or internalized receptors is difficult to quantify, in vivo, and likely varies between receptor systems and radioligands. Studies using PET tend to include smaller sample sizes compared with fMRI. This in part stems from the limitation that PET requires a broader infrastructure, including a radiopharmaceutical production unit. Exposure to radioactivity and the invasiveness of intravenous or arterial sampling creates additional limitations. The restriction of PET to adults unless medically necessary (e.g. cancer) further limits our understanding of how these systems develop in childhood and adolescence, periods of highest vulnerability for mood and anxiety disorders.

Conversely, fMRI is only an indirect measure of brain function based on signal relative to a baseline or control task. Changes in the fMRI signal do not directly reflect neural activity and may perhaps more closely reflect changes in local field potential, which are thought to reflect incoming neural signaling (Logothetis et al. 2001). Functional connectivity and other measures of correlated fMRI signal across brain regions (e.g. effective connectivity) are limited in its capacity to establish causal effects. Importantly, correlations between neuroreceptor PET binding and fMRI brain function are similarly correlational in nature. As such, these measures should be evaluated cautiously and paired with strong evidence supporting circuit dynamics. Despite these shortcomings, these two methodologies represent the most effective methods currently available for assaying brain chemistry and brain function. The use of well-documented fMRI paradigms that have been applied across multiple cohorts and repeatedly linked to relevant aspects of personality and behavior and well-validated PET radioligands benefit the application of this technique because identified associations can be considered in the context of a broader literature. The importance of using suitably powerful paradigms is underscored by one described study, which was not able to observe significant DA release within the context of a behavioral paradigm (Urban et al. 2011). Age-related changes have been reported for many PET radioligands and fMRI paradigms. Collecting the imaging measures within close temporal proximity to one another is important for avoiding potential age-related confounds.

Future applications

Multimodal imaging genetics

Imaging genetics is a widely used strategy for identifying genetic sources of individual variability in neurobiology (Hariri 2009). Most commonly, imaging genetics has been applied toward identifying links between common genetic polymorphisms and task-related brain function (e.g. 5-HTTLPR S allele associated with increased threat-related amygdala reactivity). Multimodal neuroimaging genetics using PET and fMRI offers a unique opportunity to ‘fill in the blanks' and identify molecular mechanisms mediating the effect of genetic variation on underlying brain function, behavior and related risk for psychopathology. For example, the S allele of the 5-HTTLPR has been associated with increased neuroticism and risk for depression relative to the L allele at this locus (Caspi et al. 2003; Lesch et al. 1996). Previous studies have reported that the 5-HTTLPR S allele is associated with reduced 5-HT1A binding (David et al. 2005) and increased threat-related amygdala reactivity (Munafo et al. 2008). As discussed here, 5-HT1A autoreceptor binding has also been inversely associated with threat-related amygdala reactivity (Fisher et al. 2006). Evaluating whether the effect of the S allele on increased amygdala reactivity is mediated through its effects on 5-HT1A autoreceptor binding would provide a compelling link from a genetic variant to brain chemistry and function, which contributes to threat-related behavior and related risk for psychopathology (Figure 2).

Alternatively, genetic variants can be used as putative indicators of relative differences in 5-HT signaling and, against this genetic background, specific models of relative 5-HT receptor effects can be evaluated. For example, the 5-HTTLPR S allele can be used to model relatively increased 5-HT neurotransmission in comparison to the L allele. A finding that 5-HTTLPR status moderates the observed inverse correlation between mPFC 5-HT2A binding and threat-related amygdala reactivity would suggest the impact of 5-HT2A on brain function depends on individual variability in 5-HT neurotransmission (Figure 2). This approach has been successfully employed in a recent study using single-photon emission computer tomography to assess DA transporter (DAT) binding and fMRI to assess default-mode network (DMN) connectivity as a function of a genetic polymorphism associated with DA signaling (Sambataro et al. 2011). Specifically, Sambataro and colleagues reported that the correlation between DAT binding and DMN connectivity differed as a function of genotype status for a common variant in the DA D2 receptor gene.

Finally, a multimodal neuroimaging strategy integrated with GxE models could inform how these interactions contribute to personality through effects on molecular mechanisms and underlying brain function (Caspi et al. 2010). For example, a recent study reported that the impact of 5-HTTLPR genotype on 5-HTT binding was moderated by fluctuations in daylight minutes throughout the year (Kalbitzer et al. 2010). Specifically, S allele carriers showed a more significant fluctuation in 5-HTT binding throughout the year, with relatively high binding in winter months and low binding in summer months. Individuals homozygous for the L allele showed less fluctuation, which the authors speculate may be protective against susceptibility for seasonal affective disorder and related psychopathology. In the context of a multimodal neuroimaging strategy, such a model could be further evaluated to determine whether these fluctuations in 5-HTT binding are associated with similar fluctuations in threat-related amygdala reactivity providing a compelling link between molecular and neural pathways that emerge due to interactions between genetic and environmental factors. The application of sophisticated statistical modeling techniques such as path models, including mediation and ‘moderated-mediation’, offer a powerful quantitative method for leveraging rich multimodal neuroimaging data sets toward better understanding neurobiological mechanisms that shape individual differences in behavior and related risk for psychopathology (Hyde et al. 2011).

Pharmacological challenge, endogenous neurotransmitter release and combined MRI-PET

Pharmacological challenge paradigms can provide compelling evidence for a direct effect of receptor function on task-related brain function. This is of interest in the case of 5-HT and threat, for better understanding both how 5-HT modulates threat sensitivity and developing more effective and patient-centered treatment strategies. Coupled with a multimodal neuroimaging approach, effects of SSRI administration on threat-related brain function can be linked to changes in specific 5-HT receptor mechanisms, extending results from previous challenge studies reporting an effect of short-term administration of SSRIs on threat-related brain function (Bigos et al. 2008; Harmer et al. 2006). Similarly, evaluating whether baseline measures of binding are associated with observed changes in fMRI signal may provide insight into molecular mechanisms that shape sensitivity to pharmacological challenge and may inform antidepressant treatment strategies. Findings from Buckholtz and colleagues provide compelling evidence for using a challenge paradigm for evaluating DA release, reward-related brain function and impulsivity (Buckholtz et al., 2010a,b).

The capacity to measure endogenous DA release via displacement of receptor binding for many D2 radioligands has benefited this field of research. Despite great effort, there is no equivalent radioligand that can be used to measure endogenous 5-HT release (Paterson et al. 2010). Recent radioligands targeting the 5-HT1B receptor and agonist 5-HT receptor radioligands have emerged as potential candidates to fill this need with promising results; however, they are still undergoing validation in animals (Finnema et al. 2010). Multimodal neuroimaging paradigms incorporating a measure of endogenous 5-HT release would provide novel and exciting new insights into how 5-HT signaling help shape threat-related corticolimbic circuit function and related behaviors.

The value of measuring endogenous 5-HT or DA release will be further enhanced by the development of dual MRI-PET scanners. Obviating the need to perform sequential scan sessions in independent MRI and PET scanner systems, a dual MRI-PET system offers the opportunity to measure fMRI and PET signal, simultaneously. An MRI-PET system may reduce noise introduced by coregistration though acquisition of different scans with the participant in the same spatial position, which can benefit precision of these sensitive measures. In the context of measuring endogenous neurotransmitter release, this technology offers the exciting opportunity to measure simultaneous changes in receptor binding and neural activity providing direct associations between changes in brain chemistry and brain function.

Looking Forward

In vivo neuroimaging offers a powerful technique for outlining molecular mechanisms and neural pathways underlying specific aspects of brain function and cognition. The development of these techniques and their application has provided unprecedented insight into how the brain works. PET and fMRI have been invaluable tools for understanding how neurobiological mechanisms shape related aspects of personality and risk for psychopathology through effects on distributed neural circuits. Recent application of a multimodal neuroimaging strategy that integrates these two modalities highlights the capacity to extend this understanding beyond studies that implement either of these neuroimaging techniques alone. The current review has focused on the specific benefits of PET-fMRI multimodal neuroimaging; however, the general point is that complimentary neuroimaging modalities collected within a single cohort offers the opportunity to ask relevant research questions that cannot be approached through to the use of a single neuroimaging technique. Implementing multimodal neuroimaging with other methods will undoubtedly further our understanding of relevant how genetic variation and environmental experience modulate the development and function of neurobiological mechanisms that shape behavior, personality and risk for illness.


We would like to thank C. Becker, J.C. Price and L.W. Hyde for their insightful comments and suggestions in preparing this article. Parts of the review article have been adapted from Fisher PM & Hariri AR (under review) Identifying serotonergic mechanisms underlying the corticolimbic response to threat in humans, which is planned for inclusion in a special issue of Philosophical Transactions of the Royal Society B: Biological Sciences aimed for publication in Summer 2012. This special issue is in conjunction with the 33rd International Symposium of the Groupe de recherche sur le système nerveux central (GRSNC): The Neurobiology of Depression: Revisiting the Serotonin Hypothesis, which was held May 2011, in Montreal, Quebec, Canada. The authors declare no conflicts of interest.