Introduction

Obsessive Compulsive Disorder (OCD) is a prevalent psychiatric disorder, with an incidence of between 1.5–3% across all ethnic and cultural groups studied (Kessler et al., 2005). The average age of onset is in adolescence or early adulthood with a separate peak in childhood, contributing to significant lifetime disease burden (Kessler et al., 2005). In addition, OCD can be disabling, with a significant impact on quality of life [1,2,3,4]. Despite this severity and prevalence, effective treatments for OCD are still limited. Only 10–15% of people experience remission with serotonin reuptake inhibitors, the only proven pharmacological monotherapy for OCD. While exposure-therapy with response prevention (ERP) shows significant efficacy, it can be challenging for patients to complete even if they are able to find experienced treatment providers. A greater understanding of the neural substrates of OCD is therefore necessary to develop improved treatments. Traditional circuitry models of OCD have focused on frontal cortical-striatal pathways, with significant emphasis on the role of the basal ganglia in the generation of compulsive behavior. However, accumulated evidence, described in detail throughout this review, supports an important role for particular prefrontal cortical networks in both the pathophysiology and treatment of this severe mental illness. Here we review current theories regarding how dynamic prefrontal cortical networks capable of highly distributed parallel processing and their connected subcortical circuits may contribute to OCD symptoms, and highlight areas for future research.

Clinical characteristics of OCD

Compared to other neuropsychiatric disorders, OCD has relatively clearly defined diagnostic criteria. According to DSM-5, OCD consists of obsessions, which are recurrent, intrusive thoughts, images, or impulses, and/or compulsions, which are repetitive behaviors or mental rituals that patients tend to perform in order to reduce the often severe distress associated with the compulsions [5]. Importantly, although it is not necessary to have both obsessions and compulsions to be diagnosed with OCD, most patients do have temporal linking of both of these broad categories of symptoms—i.e., obsession/ compulsion pairs are often triggered together by an external or internal stimulus.

Models of OCD symptoms highlight importance of prefrontal cortical networks

Though there is spirited and active debate in the field as to whether obsessions drive compulsions or vice versa [6,7,8], there is little question that these symptoms are related to each other, with support for the idea that they are linked by anxiety and avoidance behavior [9]. In turn, this linkage provides testable hypotheses regarding possible underlying neural substrates. Specifically, the temporal association of obsessions and compulsions points toward a neurobiological model in which obsessions may be generated cortically, and may mediate subcortical selection of compulsive behaviors in the context of anxiety. For example, one potential model of OCD suggests that excessive anticipation of dreaded consequences coupled with an erroneous Bayesian determination of action outcomes leads to symptoms [10]. This pathological process is proposed to stem from abnormalities in prefrontal cortical networks, which play an important role in anticipating probabilistic outcomes and formulating goal-directed behavior to deal with contextual contingencies. Importantly, work described in more detail below suggests that specific prefrontal cortical networks may be particularly responsible for the generation of obsessive thoughts in OCD. In addition, several important features and/or endophenotypes of OCD have been localized to prefrontal cortical areas (see below for details), suggesting a prominent role in the pathophysiology of OCD. It is therefore imperative that we discover the potential roles of abnormal prefrontal cortex (PFC) structure and function in the generation of OCD pathology and associated endophenotypes, to facilitate development of new treatments. Note, it is clear that prefrontal cortical networks do not act in isolation; rather, they are connected to various cortical and subcortical brain regions and nuclei that in concert mediate key functions such as action selection and execution. In addition, techniques that permit identification of individual cell types in both animal model systems and humans have demonstrated significant heterogeneity within the PFC that is reflected within complex microcircuit functions. However, here we will focus primarily on roles of PFC regions in OCD that can be delineated through human imaging studies.

Neuroimaging literature highlights prefrontal cortex hyperactivity associated with OCD symptoms

Over the past 30 years, convergent findings from human neuroimaging studies have highlighted abnormal structure and function of cortical-basal ganglia-thalamic loops in people with OCD. Though much translational work has focused on abnormalities in the striatum for reasons discussed below, findings in prefrontal cortical networks—particularly within the orbitofrontal cortex (OFC)—have been especially convergent across studies. In pioneering work, early positron emission tomography (PET) studies highlighted hyperactivity in medial and lateral OFC both at baseline and after symptoms were provoked either using real sensory stimulation with physical objects [11,12,13,14] [15] or evoked with imaginal/word exposures [16,17,18,19,20,21], suggesting that excessive OFC activity may generate obsessions and/or compulsions [meta-analyses of these results in [16, 22, 23]. Consistent with this theory, effective OCD treatment with either serotonin reuptake inhibitors, ERP, or deep brain stimulation (DBS) consistently normalizes OFC hyperactivity observed with fMRI [20, 21, 24,25,26,27,28,29,30]. Though it is not possible to directly test causality in people due to ethical concerns, together these results provide strong support for the idea that abnormal activity patterns in OFC networks can cause OCD symptoms. Note, however, that contrasting findings have been observed. For example, in [31], VMPFC hypoactivation and pre-SMA hyperactivation was observed in an fMRI study designed to provoke symptoms using autobiographical stimuli in OCD patients.

Though less thoroughly studied, similar findings have been seen in other prefrontal cortical brain regions. In particular, significant evidence points to a role for anterior cingulate cortex (ACC) in the generation of OCD symptoms. ACC hyperactivity is seen at baseline [32] and after symptom provocation [11,12,13, 24, 25, 33]. In addition, similar to findings in OFC, successful treatment of OCD symptoms is associated with normalization of this hyperactivity [32]. Moreover, anterior cingulotomy, a neurosurgical procedure involving destruction of ACC gray matter and adjacent white matter tracts, has been associated with reduction of OCD symptoms in treatment refractory cases [34,35,36]. Thus, while there has been less attention to these findings in the translational literature, ACC remains an important area of focus for pathophysiology and treatment studies.

Hyperactivity during symptom provocation has also been observed in fMRI studies of ventromedial prefrontal cortex (VMPFC) [37]. In addition, VMPFC hyperactivity has been observed in response to threatening shock during a Pavlovian fear task [38]. Hyperactivity in VLPFC has likewise been observed in OCD patients with high sensitivity to disgust when exposed to pictures of disgusted facial expressions [39]. Finally, although it is not formally considered to be part of the PFC, hyperactivity in supplementary motor area (SMA) has been associated with abnormal repetitive behaviors in Tourette Syndrome, which is highly comorbid with OCD [40,41,42]. PET studies in people with Tourette’s show increased tic-associated activity in SMA [43]. In addition, fMRI demonstrates that SMA is activated prior to tic onset [44, 45], and has different activity patterns during tics vs. normal intentional movements [44]. Together, these data indicate that hyperactivity in many different prefrontal cortical regions is associated with OCD symptoms, raising several important questions. First, are any of these findings directly responsible for generation of obsessions and compulsions? If so, which PFC abnormalities most contribute to symptoms? Alternatively, is this broad hyperactivity seen across many different PFC regions an epiphenomenon reflective of an entirely different upstream pathologic process? If so, compared to other cortical regions, what makes prefrontal cortical regions as a group inherently and potentially uniquely vulnerable in the context of OCD? Answering these and other questions will help develop new avenues for improved treatments.

Note, when synthesizing these findings, it is important to recognize limitations of neuroimaging research that can impact data interpretation. First, significant heterogeneity is present within patient samples, particular with respect to comorbidities, and some studies do not assess for the presence of obsessive compulsive personality disorder (OCPD). Thus, it can be difficult to definitively assign observed behavioral or neural activity differences to OCD vs. highly comorbid disorders such as OCPD [46], anxiety disorders, and major depressive disorder [47]. Second, it is challenging to separate the contributions of state vs. trait in both neuroimaging studies and neurocognitive testing. For example, performance on neurocognitive tasks can be impacted by anxiety levels and interference from ongoing obsessions and compulsions, which may lead to heterogeneity within the clinical sample that can either mask or create differences between people with OCD and unaffected comparison subjects. Finally, fMRI is inherently limited in its ability to distinguish the cell types that are responsible for BOLD activation. Thus, a finding of BOLD signal in a particular PFC region could be accounted for by activation of excitatory and/or inhibitory neuron populations. This distinction has important implications for determining appropriate clinical targets. Preclinical studies in animal models can therefore be a useful bridge to dissect contributions of particular cell types to OCD-relevant behaviors.

How might disruptions in classic PFC-dependent functions lead to symptoms of OCD?

Though delineating the diverse functions of the PFC is beyond the scope of this review, here we will discuss executive functions classically thought to be PFC-regulated whose disruption may lead to the generation of obsessions and compulsions. Such a perspective is consistent with NIMH’s Research Domain Criteria (RDOC) framework, which aims to identify dimensions that cut across current diagnostic categories in order to facilitate translational investigations and identification of new treatment targets [48]. In particular, current theories suggest that abnormalities in the following domains may be key contributors to dysfunction in OCD: cognitive flexibility, threat processing/harm avoidance, habit formation/goal-directed planning, error monitoring, and response inhibition. However, whether and how abnormalities in these constructs actually generate the symptoms of OCD is still a matter of active investigation. Below, we will delineate the clinical and psychological evidence supporting these theories.

Cognitive flexibility

Based both on clinical observations and epidemiological studies demonstrating significant comorbidity of OCD and OCPD, OCD has been classically associated with impaired cognitive flexibility. These impairments are thought to be central to OCD pathology, in part due to the clinical perspective of obsessions and compulsions as fixed and rigid thoughts and behavioral patterns that are resistant to change. Many different tasks and variants within these tasks fall under the overarching rubric of cognitive flexibility, including reversal learning, set shifting, and cued task switching, leading to some difficulties in standardization across studies. In addition, it has at times been challenging to identify consistent abnormalities in tasks measuring cognitive flexibility vs. rigidity due to patient heterogeneity. However, impaired behavioral performance and/or altered neural activity patterns are generally seen on many neurocognitive tasks designed to test cognitive flexibility, with effect sizes typically in the medium range as reported in a relatively recent meta-analysis [49].

Several studies have identified abnormalities in reversal learning, task switching, and extra-dimensional set shifting [50, 51] in people with OCD, though it can be difficult to compare findings across studies because of differences in tasks. In some cases, differences in behavioral performance [52, 53] and reaction times [54,55,56] are noted between groups [though see [57,58,59]], but often neural activity differences are seen despite equivalent performance, suggesting potential compensatory processes (Chamberlain et al., 2007; Chamberlain et al., 2006; Gottwald et al., 2018). Specifically, evidence from neuroimaging studies conducted while OCD patients are performing reversal learning (Chamberlain et al., 2008; Remijnse et al., 2006; Remijnse et al., 2009; Remijnse et al., 2013) or a task switching paradigm (Gu et al., 2008) shows impaired recruitment of OFC, potentially due to excessive baseline activity. One study has additionally demonstrated that this OFC hypoactivity is seen in unaffected first-degree relatives of OCD patients. Notably, the abnormal activity during reversal learning [53, 54, 60, 61] occurs in the same orbitofrontal-striatal circuits that show abnormal activity during symptom provocation [11,12,13]. However, it is unknown whether the same abnormalities in OFC activity underlie both impaired reversal learning and compulsive behaviors. Dorsolateral PFC (DLPFC) also shows robust hypoactivity in OCD patients during reversal learning and task switching [53, 61,62,63,64], and reduced resting-state DLPFC-putamen functional connectivity was recently associated with impaired extra-dimensional set-shifting in people with OCD [65]. Finally, reduced ACC and VMPFC activity are also seen in people with OCD (Gu et al., 2008) during task switching. Together, these findings suggest a pattern in which people with OCD display less recruitment of multiple PFC structures compared to unaffected controls when faced with changing task conditions. It remains to be determined whether these decreases in activity represent primary pathologic changes, or are compensating for alterations in activity in other structures.

Threat processing/harm avoidance

Anxiety is often a prominent feature of OCD, despite its classification apart from other anxiety disorders in DSM-5 due to differences in underlying circuit abnormalities [66]. Whether anxiety is a primary pathologic driver in OCD or is simply a stressor that can exacerbate pre-existing obsessions and compulsions is still a matter of active debate. However, abnormalities in threat processing systems and aversive learning processes [38, 67, 68] have been documented in OCD. Impaired VMPFC recruitment in a subset of OCD patients with deficits in extinction recall is seen both during fear extinction training and recall deficits [68], consistent with its importance for flexible updating of stimuli that are no longer threatening [69, 70]. In addition, as described briefly above, OCD patients show impaired ability to compute safety signals in the VMPFC when they are required to update task contingencies with new information [38]. Specifically, using a threat reversal learning paradigm paired with fMRI, fear responses (as measured by skin conductance) were compared between people with OCD and unaffected comparison subjects in response to either a threatening (paired with shock) or safe (not paired with shock) visual stimulus, followed by contingency reversal. Excessive VMPFC activity predicted impaired adaptation on the task after reversal of neutral and shock-paired cues [38], suggesting negative functional consequences due to impaired generation of safety signals during value updating of previously-threatening cues. This may have important implications for delivery of exposure therapy, as particular OCD subpopulations may have worse response rates due to inherent difficulties with threat vs. safety processing. Normalizing these threat-associated circuits may therefore help facilitate behavioral treatments for OCD, by decreasing the propensity of patients to engage in harm avoidance strategies.

Habit formation/goal-directed planning

The conceptualization of OCD as a shift from normal goal-directed behavior to pathologic habits [described in [71] reflected the emerging recognition that striatal pathology plays an important role in OCD. Since this theory was first proposed [11, 33, 72], several investigations have tested this idea using a variety of paradigms designed to probe different features of these constructs. Using a novel outcome devaluation-based task (The Fabulous Fruit Game) to test whether actions are habitual or goal-directed, Gillan et al. [73] found that people with OCD were less likely than controls to respond appropriately to devaluation by changing their performance, suggesting an overreliance on habits. In addition, when explicit knowledge of contingencies was tested in a separate section of the task, OCD patients showed decreased understanding of the associations between actions and outcomes compared to controls. Increased habit formation in OCD has also been inferred using contingency degradation paradigms, with OCD patients demonstrating greater levels of responding for reward than controls when the causal link between a particular action and the outcome is diminished [74]. Note, however, that the OCD subjects in this study did explicitly report accurate assessments of the contingencies, raising questions about whether these behavioral patterns could instead be construed as an atypical version of goal-directed behavior. In addition, the constructs of model-based and model-free learning, which can be loosely mapped onto goal-directed and habitual strategies, have been examined in OCD using a two-step task [see [75] for detailed task description]. Here again OCD patients showed a propensity for habitual behavior, as manifested in a bias toward model-free learning, which was associated with a decrease in mOFC and caudate gray matter volume [76].

Extending this work to investigate devaluation of negative outcomes using an avoidance-based task, [67] also found that people with OCD continued to avoid a negative stimulus (i.e., shock), even when this outcome had been clearly devalued–i.e., the shocker was unplugged and therefore acknowledged as non-functional by the patients when their explicit knowledge was tested. This impairment in adjusting behavior to an updated negative contingency was found to be associated with hyperactivation of the medial OFC during the learning of the avoidance behavior [77]. Further supporting these results, a more recent study used a novel task with a milder aversive outcome– needing to prevent simulated “machines” from breaking as opposed to avoiding shock delivery [78]. Though no overall group differences were seen between patients with OCD and controls on this task, there was an association between unsuccessful devaluation of cues and reduced activity in pre-motor cortex and left inferior frontal gyrus, areas which have been linked to cognitive and response inhibition (Aron et al., 2007; Hung et al., 2018; Picton et al., 2007). Reduced mOFC activity in response to valued cues was also observed in OCD patients, consistent with [77]. However, during presentation of valued cues, unsuccessful devaluers within the group of OCD patients had greater mOFC activity than successful devaluers, suggesting a complex relationship in which either hyperactivity or hypoactivity in mOFC can lead to impaired devaluation. Together, these findings may yield insight into why OCD patients are prone to incorporating avoidance rituals into their compulsions, and thus inform the genesis of avoidance-related compulsive behaviors. However, it is important to note that the extent of avoidance in [77] did not correlate with compulsive behaviors as measured by the Yale Brown Obsessive Compulsive Scale (YBOCS), raising important questions about whether there is a direct relationship between abnormal performance on neurocognitive tasks and OCD symptoms.

Goal-directed planning has also been assessed independently of the investigation of habits in OCD. Using a self-paced version of the visuo-spatial planning Tower of London task, van den Heuvel et al. [79] found significant planning impairments in people with OCD that were associated with decreased activation of the DLPFC and caudate. Similarly, Den Braber et al. [80] found hypoactivation of DLPFC during planning in a sample of people at high risk for OCD, though behavioral deficits were not observed in this less severe clinical sample. Consistent with these findings, Vaghi et al. (2017) found that both OCD patients and their asymptomatic relatives showed both slower reaction times and hypoactivation of the DLPFC during goal-directed planning on the Tower of London task, suggesting this may be an endophenotype of the disorder [64].

Impairments in goal-directed planning in OCD are also supported by the fact that patient self-report in studies assessing habit formation yields accurate assessments about action-outcome contingencies. These findings suggest that patients are performing the “habitual” behavior intentionally, despite explicit knowledge that this is not necessary to achieve the desired outcome. Similarly, in an “illusion of control” task, people with OCD had more accurate assessments of their amount of control in the task [67], in keeping with the idea that they have the ability to correctly estimate the value of performing particular actions, but may have other competing value assignments that weight performance of compulsive behaviors more heavily. Thus, performance of a compulsive behavior may actually be goal-directed, but lead to pathology since the compulsions rank higher in the value hierarchy than other goal-directed behaviors necessary for health and well-being. This is consistent with metacognitive theories suggesting that compulsive behaviors are goal-directed [81]—i.e., performed consciously to achieve a valued outcome, such as anxiety reduction, neutralization of negative thoughts, or confirmation of uncertain priors, potentially based on overestimation of credibility of obsessive thoughts. For example, an attractive possibility consistent with clinical experience is that performance of compulsions can in some cases provide temporary anxiety relief via harm avoidance, which may be assigned a high value.

A similar pattern was seen in a predictive inference task, in which people with OCD displayed excessive actions that only took recent evidence into account, despite reporting similar confidence as controls in the fact that performing these actions was not the ideal strategy [82]. While other studies have also described decreased confidence in correctness of decisions in people with OCD [83] or non-diagnosed people with high levels of compulsive traits [84], these findings suggest that complexity beyond the dichotomy of goal-directed vs. habitual behavior must be integrated into theories of decision-making in OCD. Recent theories integrate these ideas by suggesting that OCD symptoms result from dysfunction in the process of arbitration between goal-directed and hierarchical systems, with a specific suggestion that the selection of goal-directed actions is hierarchically organized under a goal-directed controller—i.e., once a particular sequence of well-practiced actions (i.e., a “chunk” that could correspond to a compulsive behavior) is selected, the individual actions are insensitive to outcome, completed in an automated manner, and not individually monitored by the goal-directed system. Hyperactivity in mOFC could therefore drive repeated retrieval of unobservable but anxiety-provoking outcomes, as suggested by the findings in [85, 86] and proposed in [87], leading to repeated performance of chunked sequences of behavior like handwashing.

Together, these studies have found evidence consistent with the theory that people with OCD have impairments in the balance between goal-directed behavior and habits [87, 88]. However, depending on the task used, distinct abnormalities are seen that point to either excessive habit formation or maladaptive goal-directed behavior. Future studies testing these paradigms longitudinally within the same subjects will help to dissociate which set of impairments predominates in OCD.

Error monitoring

Impairments in error monitoring have also been observed in OCD, and suggested as the basis of a subtype of OCD obsessions often termed a “just not quite right” feeling. Across multiple studies, increased error-related negativity (ERN) [89] is observed as a negative deflection of the event-related potential on EEG with an ACC source in OCD patients compared to controls [90, 91] [also see [92, 93] for reviews]. Stability of these findings independent of treatment [94] and presence of increased ERN in unaffected first-degree relatives [95] suggests that impaired performance monitoring could be an OCD endophenotype. A recent quantitative meta-analysis indicates a strong effect size across 38 studies, specifically in tasks that create response conflicts [96]. Consistent with this, rostral ACC hyperactivity is seen in OCD patients during error commission on a cognitive task specifically designed to generate errors, and the extent of hyperactivity was positively correlated with symptom severity [97]. In addition, greater activation of the rostral ACC in the post-error trial period in OCD patients was found to positively correlate with severity of compulsions. Increased functional anisotropy was also seen in cingulum bundle underlying rACC, potentially serving as a substrate for an increase in the communication between anterior and posterior cingulate cortex that is necessary for error processing [98]. Combined fMRI and EEG studies have also observed a positive correlation between activity in the preSMA and ERN, which could reflect attempts to exert proactive control [99]. Further investigation of the relationship between ERN changes and other prefrontal cortical nodes of the error processing network (e.g., ventromedial PFC, DLPFC, anterior insula) is needed to gain greater understanding of mechanism [100]. Notably, an initial study has started to examine the therapeutic potential of attention bias modification in people with OCD [101], finding significantly reduced ERN in OCD subjects, but not in controls undergoing training. Future work is needed to determine if this reduction in ERN leads to symptom improvement.

Response inhibition

A prominent theory in the OCD literature suggests that response inhibition deficits may contribute to the symptomatology of OCD. This theory has been attractive because it makes some intuitive sense—i.e., compulsions are often colloquially considered as an inability to restrain oneself from performing particular actions. A variety of tasks have been used to formally test this idea, including Go/NoGo(Snyder et al., 2015), Stop Signal Reaction Time (SSRT) [60, 102, 103], and Stroop [104,105,106,107]. Though a meta-analysis has not demonstrated impairments on the Go/NoGo task (Snyder et al., 2015), OCD patients demonstrate moderate performance deficits on the Stroop task, with the strongest impact on interference scores and time on incongruent Stroop [108]. One study provides evidence that Stroop abnormalities in children and adolescents with OCD are correlated with decreased functional anisotropy in the cingulum bundle [105], but further work is needed to examine the role of PFC regions in mediating these phenotypic changes in OCD. Impairments in inhibition on the SSRT have been specifically associated with abnormalities in the PFC, including alterations in gray matter volume (decreased in OFC and inferior frontal; increased in anterior cingulate [109]). In addition, fMRI demonstrated functional differences in PFC during SSRT performance, including increased activity in preSMA during successful inhibition trials in OCD patients compared to controls, and decreased activity in the inferior frontal cortex [110], a region that has been consistently linked to response inhibition on this task [111]. However, note that though there is evidence that response inhibition is impaired in OCD and may even serve as an endophenotype (i.e., impairments are also observed in first-degree relatives [60]), a comprehensive meta-analysis has suggested a more complex picture [49], consistent with the fact that there is also significant evidence of normal performance on many of these tasks in OCD (Stop signal–[109, 112]; Go/No-go (Snyder et al., 2015); Stroop [104,105,106,107]. Although abnormal interference in the Stroop task was associated with a medium effect size across 23 studies, only a small effect size was seen for commission errors across 15 studies of multiple response inhibition tasks (Go/No-Go; continuous performance task; SSRT), with the upper limit of the confidence intervals approaching zero. Thus, there may be challenges in generalizing these findings due to heterogeneity of patient populations and/or differences in task parameters across testing sites [49, 113]). These findings highlight a potential need to identify subgroups of OCD patients in order to identify abnormalities in particular neurocognitive domains. In addition, performing longitudinal assessments of the same subjects on different response inhibition tasks while undergoing functional imaging would clarify whether particular tasks are more sensitive for detecting OCD-relevant response inhibition abnormalities in certain patient subgroups.

Emerging OCD postmortem literature also highlights molecular abnormalities in OFC that could be consistent with abnormal function

Although the etiology of OCD is unknown, both twin studies [114] and emerging genome-wide association studies (GWAS) support a genetic contribution. In particular, though OCD GWAS studies are not yet powered to detect significant common variants, preliminary evidence suggests that genes encoding proteins found at or around excitatory synapses may play an important role [115, 116]. More recent exome sequencing studies have identified rare damaging coding variants (CHD8, SCUBE1, and SLITRK5) that also provide support for synaptic abnormalities [117, 118]. Until relatively recently, however, there has been little direct investigation of molecular abnormalities in OCD.

Recent work from our group and others has begun to address this hole in the OCD literature, specifically focusing on examination of gene expression in the PFC and striatum. The first study to examine postmortem brain tissue from people with OCD did so in a trans-diagnostic way across several disorders characterized by obsessions and compulsions (eating disorders, OCD, OCPD, and tic disorder) (Jaffe et al., [119]. Using a microarray, 286 differentially-expressed genes were detected in the DLPFC of people with OCD/OCPD/tic disorder compared to unaffected comparison subjects. Gene set analyses indicated significant enrichment for genes associated with Alzheimer’s, nuclear phosphoproteins, and oxidative phosphorylation. However, this study did not examine gene expression in prefrontal cortical regions shown to be abnormal in OCD. To address this gap, we performed targeted qPCR of excitatory and inhibitory synaptic genes [120] and RNAseq [121] in the OFC. Strikingly, despite having only eight OCD subjects and eight comparison subjects, our initial qPCR findings showed a significant downregulation of a group of genes associated with glutamatergic synapse gene expression in OFC. In contrast, there was little change in expression of transcripts associated with GABAergic synapses. To follow up on this work, we performed unbiased RNAseq and gene set enrichment analysis to identify large scale changes in gene expression in the same subjects. A combined analysis of expression in OFC and striatal regions demonstrated downregulation of synaptic transmission gene sets, consistent with our previous findings, as well as upregulated gene sets associated with cardiovascular system development and transmembrane tyrosine kinase activity. In addition, analysis of cell-type fractions using a deconvolutional approach [121] showed an increase in the vascular cell-type fraction and decrease in the interneuron fraction in the OFC. Both of these findings could be consistent with the previous observations of hypermetabolism of the OFC in OCD patients [28, 33, 122, 123].

Combined with data from our lab [120, 121] and others [124] indicating similar downregulation of excitatory synaptic genes in the striatum, these findings support a model of impaired excitatory synaptic transmission in two key circuit nodes implicated in OCD pathophysiology. It remains to be seen whether: (1) similar gene expression changes are observed in other prefrontal cortical regions, or are unique to the OFC; (2) if decreases in excitatory synaptic gene expression are also observed in non-prefrontal cortical areas; and (3) whether these gene expression changes correspond to downregulation of expression vs. synapse loss. However, out of the 904 genes that were differentially expressed in OFC and striatum in our RNAseq study [121], only 13 were identified as differentially expressed by Jaffe et al. [119], suggesting that OCD gene expression abnormalities in DLPFC may be distinct from those observed in the adjacent OFC. Answering these questions will refine our understanding of the molecular pathology associated with OCD by helping us determine if these abnormalities stem from widespread brain changes or pathology localized to specific circuits.

Theories to link these findings together

Together, the accumulated data described above provide significant support for the idea that abnormal PFC function plays an important role in the pathogenesis of OCD. However, because of the inherent difficulty of assessing causality in human studies, as well as the lack of cellular resolution in human imaging studies, this cannot be readily tested in humans. Despite this fact, synthesis of these data can lead to several alternative theories that can potentially be supported or refuted by evidence from human intervention studies and animal models (Fig. 1). Here, in order to ground the theories in the previously described findings from neuroimaging studies, we first focus on four broad alternatives describing how findings of either hyperactivity or hypoactivity in prefrontal cortical networks could lead to OCD symptoms through network dysfunction in the context of a signal processing framework. We believe this is a useful lens for considering novel treatment interventions, as even the most advanced currently available neuro-interventional treatments (i.e., focused TMS, DBS) are only able to broadly tune activity up or down within particular PFC regions, leading to impacts in their downstream connected networks. Within each of these Alternatives, we then highlight detailed theories of how dysfunction in particular PFC networks consistent with that broad scenario could produce obsessions and/or compulsions, generating testable hypotheses.

Fig. 1
figure 1

Alternative theories of how abnormal PFC function could lead to OCD symptoms. Treatments for each of these scenarios and testable predictions from animal models are provided for each theory.

Full size image

Alternative 1: Hyperactivity in prefrontal cortical regions directly causes OCD symptoms

Functional neuroimaging studies showing hyperactivity at baseline and after symptom provocation in prefrontal cortical regions including OFC, ACC, VMPFC, and DMPFC provide strong support for the idea that hyperactivity in these regions (potentially generated through a conserved synaptic abnormality that impacts these areas in concert; see section on postmortem findings) may directly lead to OCD symptoms. Though not a comprehensive list, possible theories that would be consistent with this Alternative include the following examples.

  • Anticipatory autonomic arousal upon viewing a subjectively threatening stimulus could drive hyperactivity in either OFC or ACC. In turn, this OFC/ACC hyperactivity could lead to overvaluation of actions that would generate decreases in autonomic arousal, such as leaving the room to avoid the perceived potential threat or performing a compulsive behavior that subjectively “neutralizes” the perceived threat (e.g., checking the stove, handwashing, etc.).

  • Hyperactivity in OFC could lead to impaired action-outcome monitoring, such that performance of a particular behavior does not lead to the expected sense of completion. Its repeated selection would therefore be continuously assigned a high value, making it difficult—and potentially even illogical given their priors and decision-making framework—for the patient to choose an alternative, less valued outcome.

  • Increased activity in ACC could lead to enhanced error-related signals, which has been observed in patients with OCD [91, 97, 125, 126], resulting in increased error aversion. In turn, this could cause a corresponding increase in harm avoidance strategies to avoid the perception of committing an error—e.g., performing compulsions despite consciously understanding that the perceived negative outcome is unlikely to happen, as seen in [67].

If Alternative 1 is correct, hypoactivity in PFC regions during neurocognitive tasks may result from impaired ability to recruit these structures due to baseline circuit hyperactivity. Thus, while PFC hypoactivity may lead to cognitive dysfunction, this would not be the primary driver of symptoms in OCD in this scenario. This concept is supported by the fact that a comprehensive meta-analysis suggests that the effect sizes observed for neurocognitive deficits in OCD are relatively moderate in size, and are therefore unlikely to yield a major contribution to the significant symptom burden experienced by people with OCD [49]. In this model, the observed neurocognitive abnormalities would be considered an epiphenomenon of the other pathologic factors that are more central to symptom production in OCD. Thus, while it might be inherently valuable to improve these neurocognitive deficits if they have an impact on patients’ quality of life, such a therapeutic strategy would not be expected to lead to resolution of obsessions and compulsions.

Alternative 2: Hypoactivity in prefrontal cortical regions directly causes OCD symptoms

Hypoactivity in PFC regions during performance of neurocognitive tasks is quite consistent across multiple tasks and PFC regions, providing support for the idea that this hypoactivity is a cause, and not a consequence, of OCD symptoms. Possible theories that would be consistent with this Alternative include the following examples.

  • Deficits in the PFC cognitive control network [including ACC/pre-SMA (ACC/pSMA), dorsolateral prefrontal cortex (DLPFC), inferior frontal junction, anterior insular cortex, dorsal pre-motor cortex, and posterior parietal cortex [127, 128]] could directly lead to compulsive behaviors. For example, humans and monkeys with prefrontal damage display stimulus-bound behaviors—that is, salient cues lead to reflexive actions due to strong associations between the sensory stimulus and a particular action. Though not formally tested (but suggested in [129]), this phenomenon is commonly seen in people with OCD, who anecdotally report reacting to particular trigger stimuli using harm avoidance strategies before they can consciously engage in response prevention strategies learned through EX/RP (exposure therapy with response prevention). Findings of PFC hypoactivity could thus be consistent with impairments in inhibitory control of prepotent responses seen in OCD, which may lead to more automatic engagement in compulsive behaviors, particularly in stressful circumstances.

  • Hypoactivity in ACC/OFC could also be consistent with findings that people with OCD have less capacity to engage in online integration of changing values and application of this information to behavior selection in complex environments or emotionally stressful situations, thus reverting to established strong stimulus-outcome relationships. This is consistent with the clinical phenomenon of exposures being easier to perform in the relatively sparse environment of the clinician’s office, but harder to achieve in more complex “real-world” environments with changing and unpredictable contingencies and emotional states [130].

  • Because ACC plays an important role in error detection, evaluation of the degree of the error, and selection of the subsequent most appropriate actions [131, 132], ACC hypofunction may lead to impairments in updating values and altering decision-making strategies in response to errors, leading instead to persistent selection of maladaptive behaviors.

  • Hypofunction of prefrontal cortical networks including OFC and ACC could contribute to impairments in learning of new associations between cues, actions, and outcomes (i.e., reward vs. punishment), which could be detrimental for efficacy of exposure therapy.

If Alternative 2 is correct, hyperactivity in PFC regions at baseline and during symptom provocation may be compensatory responses from connected PFC regions and intraregional microcircuits that are attempting to boost performance of critical PFC executive control functions, including decision making and goal-directed planning.

Alternative 3: Either hyperactivity or hypoactivity in prefrontal cortical regions can lead to similar dysfunctional states

As described above, both hyperactivity and hypoactivity of PFC regions are commonly observed in OCD, though typically in different task states. If Alternative 3 is correct, both PFC hyperactivity or hypoactivity could have a similar negative impact on circuit dynamics in the PFC, thus effectively “jamming the system” and impacting downstream output in a similar way. For example, it has been proposed that OCD is associated with impairments in value-based decision making—specifically, with assignment of excessive value to unlikely negative outcomes associated with obsessions [133, 134], or inability to imagine positive outcomes through impaired metacognition [85, 87, 135]. Either OFC hyperactivity or hypoactivity could therefore lead to impairments in this key executive function.

Alternative 4: Altered prefrontal cortical activity is an epiphenomenon resulting from other primary pathologic sources

Although PFC abnormalities are prominent in OCD, leading theories suggest that symptoms result from abnormal activity within parallel vs. interconnected cortico-basal ganglia-thalamic loops [71, 136, 137]. Thus, pathologic disruptions at any node in the circuit could lead to downstream abnormalities in any connected node. Though disruptions in these extended networks connected to PFC regions are not the focus of this review, one example is that thalamic hyperactivity is also very consistently observed in OCD [12, 138,139,140,141,142], and could therefore be a primary pathologic feature. In turn, hyperactivity in medial thalamic nuclei could result in the observed pattern of PFC hyperactivity/ hypoactivity in OCD through long-range projections acting either directly on PFC glutamatergic pyramidal neurons or indirectly on inhibitory interneurons in a state-specific or task-specific way. Another example is that dysfunction in dopaminergic projections to PFC could cause abnormal representation of reward-related information via either positive or negative prediction error [143,144,145], leading to either overvaluation of the impact a compulsive behavior will have on decreasing the likelihood of a negative outcome, or undervaluation of the value of refraining from performing the compulsion ([146,147,148,149]. In addition, dopamine release in PFC can directly affect capacity for plasticity through its effects on NMDA-receptor mediated glutamatergic transmission [150] or facilitation of activity-dependent plasticity [151], thus impacting capacity to form new associations based on updated action values. Thus, it is possible that signal integration and readout within PFC regions is working correctly, but they are receiving faulty information from their inputs. In the case of Alternative 4, although therapeutic interventions can potentially be made at any point in the circuit, understanding the location of the primary pathology may help to develop new, more targeted treatments that directly attack the source of the problem—e.g., restoring normal thalamic rhythmicity or tuning function of dopaminergic neurons.

Approaches for testing these alternative theories

It is important to develop approaches to distinguish between these Alternatives, as different treatments may be required depending on regional localization of pathology, directionality of abnormal activity (e.g., hyper- vs. hypoactivity), and cell type origins of pathologic activity patterns. Specifically, determining the causal relationship between activity in PFC regions and clinically-important behaviors may help to direct novel, more specific pharmacologic, behavioral, and neurostimulation-based interventions.

One path forward toward this goal is to test these alternative theories in preclinical models using a combination of sophisticated behavioral paradigms and optogenetic and chemogenetic techniques for precise manipulation of activity in particular neural circuits. However, it is also increasingly clear that recent advances now allow for greater levels of precision and causality testing in humans than previously achievable. For example, Alternative 1 is consistent with our recent study in humans, in which we randomized individuals with compulsive behaviors (a majority of whom had OCD) to either intermittent TBS, which is expected to potentiate the OFC, or continuous TBS (cTBS), which is expected to depotentiate the OFC. In both conditions, TBS was paired with a “habit override” computer task, in which participants were trained to override an overlearned shock avoidance behavior. We found that cTBS paired with the habit override task led to a decrease in both urges and compulsive behaviors that persisted for at least a week after the intervention [152], consistent with findings from another recent study [153]. These findings are now being followed up with a sham-control randomized clinical trial.

It is attractive to try to develop a unified theory of the pathologic origins of OCD that will aid the field in the development of new treatments. However, the examples described above demonstrate that many different theories could be consistent with current neuroimaging findings, particularly given the inherent limitations of neuroimaging that prevent us from assigning BOLD activation to excitatory neurons, inhibitory neurons, or nonneuronal subtypes (e.g., astrocytes, microglia). In addition, it is important to recognize that there are likely many different pathological paths that can lead to similar circuit and symptom outputs, highlighting the need for patient stratification into subgroups based on differences in underlying biological mechanisms that may require different treatment approaches. For example, [154] recently compared the efficacy of two different DBS sites—ventral capsule/ventral striatum (VC/VS) and anteromedial STN—in the same patients with treatment refractory OCD, coupled with neurocognitive testing and tractography. They found that stimulation at either site within an individual patient led to similar improvements in OCD symptoms. However, amSTN DBS also led to significant improvements in cognitive flexibility, while VC/VS DBS had a greater impact on mood symptoms; these findings were associated with connectivity to the lateral OFC and medial OFC, respectively. Thus, a patient’s individual symptoms and network function may dictate the most effective treatment targets.

Interventions in PFC circuits can be effective OCD treatments

Consistent with the accumulated findings from the studies described above, there is evidence that interventions in prefrontal cortical circuits can lead to therapeutic effects in people with OCD. Some of these surgical interventions predated modern neuroimaging technology highlighting functional and structural abnormalities in the targeted regions. Other more recent interventions have taken advantage of neuroimaging findings to guide treatment approaches. Importantly, effective interventions, be they exposure therapy, pharmacotherapy, or DBS, appear to share the feature of ultimate normalization of hyperactivity in OFC, ACC, and mPFC [20, 26, 155,156,157,158]. Though therapeutic approaches for OCD will be covered in more detail in another article in this volume (see Rasmussen and Goodman, 2021), here we will briefly describe several different therapeutic PFC interventions in OCD that may help us to test the four alternative theories described above. Note that although the efficacy of interventions in prefrontal cortical circuits implies that the targeted PFC regions are involved in pathology, we acknowledge that this is not necessarily the case, because compensatory brain changes are likely to occur during the intersection of the disease process with normal development as OCD evolves. However, this does indicate that interventions in PFC circuits can improve functional outcomes, implying that restoration of normal PFC functions can override, if not correct, abnormal pathology.

Ablative surgical procedures

Several different types of ablation procedures have been used to treat refractory OCD, based on the theory that disruption of connections between hyperactive prefrontal cortical regions and their downstream basal ganglia circuits may lead to a decrease in transmission of abnormal neural signals [159]. These moderately efficacious procedures include anterior cingulotomy, capsulotomy, subcaudate tractotomy, and limbic leucotomy [34]. In anterior cingulotomy, the cingulum bundle is lesioned bilaterally to disrupt connections between cingulate cortex and subcortical regions. Similarly, white matter tracts in the anterior limb of the internal capsule (IC) are transected in capsulotomy, resulting in presumptive disconnection of the OFC and mediodorsal thalamus. Subcaudate tractotomy, a lesion below and anterior to the head of the caudate, is also thought to disrupt frontothalamic circuits, as is limbic leucotomy, which combines cingulotomy and subcaudate tractotomy.

Transcranial magnetic stimulation (TMS) in PFC

Though results from ablative neurosurgical procedures and DBS have been promising (see Rasmussen and Goodman, 2021 in this volume; [160]), nonsurgical interventions are preferable if they are equally efficacious. In this context, recent studies have highlighted the potential utility of transcranial magnetic stimulation (TMS) in OCD treatment. Early work suggested possible efficacy of low frequency rTMS of SMA [161, 162]. More recently, high-frequency (20 Hz) deep TMS of the mPFC-ACC has been found to yield clinical improvement on the YBOCS in a multi-center randomized placebo controlled trial, which correlated with increased ERN in the Stroop task in an initial study [163, 164]. Improvements in YBOCS were maintained at 1 month follow up, with 45.2% in the active TMS group and 17.8% in the sham group showing >30% reduction in YBOCS. This treatment has now been FDA approved [165] and investigations of predictors of response are ongoing, with some initial evidence that older adults and people with less severe illness may have greater responses [166]. In addition, as described above, work from our group [152] highlights the potential efficacy of deep TMS of OFC using continuous theta-burst stimulation (cTBS) for OCD treatment.

Investigation of prefrontal cortex in animal models for OCD-relevant behaviors and treatment

Studies in animal models can provide an important bridge to gain insight into pathophysiology. To date, mechanistic dissection of OCD-related behaviors in animal models has tended to focus on repetitive and/or compulsive behaviors, in part because it is not possible to assay obsessions in species that cannot verbally report their experiences. In addition, the designation of PFC subregions in rodents is quite controversial due to the absence of dysgranular cortex [167], highlighting the importance of including primate models in the study of psychiatric disorders with significant PFC pathology [168]. Thus, much of the OCD animal literature has focused on dissecting the contributions of the basal ganglia to abnormal repetitive and compulsive behaviors. With these caveats in mind, efforts have more recently been made to dissect contributions of PFC-homologous regions to OCD-relevant behaviors in rodents. Because we have recently performed comprehensive reviews of the OCD animal literature [169, 170], here we will focus on studies that may yield mechanistic insights into the proposed alternative theories described above.

Several studies in mice have recently focused on the role of both medial and lateral OFC in the generation and treatment of abnormal repetitive behaviors. In two studies using transgenic mice, constitutive knockout of synaptic proteins—SAPAP3 [171] or Slitrk5 [172]—led to compulsive grooming behavior, anxiety-like behavior, and OFC abnormalities. A corresponding increase in mOFC activity was inferred in Slitrk5-KOs by increased levels of the neural activity marker FosB in KOs compared to littermate controls [172]. In contrast, while SAPAP3-KOs had no evidence of baseline differences in lOFC activity using in vivo electrophysiology [173], our subsequent work using in vivo microscopy has shown that compulsive grooming is associated with an increased number of inhibited cells in the lOFC [174]. Consistent with these findings, prior work has shown that acute stimulation of lOFC-centromedial striatum projections normalized compulsive grooming in SAPAP3-KOs, potentially by boosting activity in striatal parvalbumin positive interneurons to decrease overall striatal activity [173]. In addition, we [175] and others [176,177,178] have identified reversal learning deficits in SAPAP3-KOs, with evidence of associated abnormal neural activity patterns in both mPFC [175] and lOFC (Manning et al., bioRxiv). Most recently, [178] found that impaired function of OFC GABAergic neurons tagged via expression driven by the non-specific Dlx GABAergic promoter is responsible for reversal learning abnormalities in SAPAP3-KOs, due to lack of inhibition of OFC pyramidal neurons during value updating following reversal. Furthermore, our new work indicates baseline deficits in the strength of encoding of both the correct lever press and reward cue in LOFC neurons from SAPAP3-KOs [174]. Thus, while the data from Slitrk5-KOs supports the idea that mOFC hyperactivity may play a role in the generation of compulsive grooming behavior (Alternative 1), the data from SAPAP3-KOs suggests that lOFC hypoactivity may be important in the generation of impaired reversal learning and potentially in compulsive grooming (Alternative 2). These seemingly disparate results could be synthesized either by Alternative 3, in which any disruption of signaling in a particular brain region interferes with normal processing and has the same impact on downstream neural signaling, or by the fact that these two constitutive transgenic models may have very different developmental compensatory effects, leading to distinct neural activity patterns associated with the same behavioral abnormalities. Two other sets of studies also highlight the potential role of OFC hyperactivity in the generation of abnormal repetitive behaviors. First, 5-HT1B receptor activation in ventrolateral OFC leads to perseverative locomotion and PPI disruption [179], which has translational relevance to OCD [180,181,182,183,184,185]. In addition, consistent with Alternative 1, repeated optogenetic stimulation of mOFC-striatal projections to simulate OFC hyperactivity in OCD patients led to a progressive increase in grooming accompanied by an increase in evoked firing rate [186]. Together, these findings are potentially consistent with the idea that a medial-lateral gradient exists, in which medial OFC and its connected circuits are responsible for “goal-directed” repetitive behaviors such as grooming, while lateral OFC and its downstream targets are more important for cognitive flexibility impairments seen in OCD. Though at first glance this idea seems inconsistent with our recent finding of recruitment of more grooming-associated inhibited cells in LOFC of SAPAP3-KOs (Manning et al., bioRxiv), decreased activity in the lOFC may lead to abnormal encoding of the stimulus→outcome associations necessary for Pavlovian instrumental transfer (PIT) or impaired contingency degradation, leading to grooming that appears “habit-like” due to its resistance to devaluation (Ostlund 2007, Balleine, Leung, Ostlund, 2011, Bradfield 2017. This view could be consistent with behavioral findings from a water-drop induced grooming test in [173].

Recent preclinical studies have also focused on the role of prefrontal cortical circuits in OCD treatment. One of the most effective treatments for OCD is EX/RP, but it can be quite challenging for patients to complete treatment due to associated distress and difficulties in achieving extinction. To investigate mechanisms underlying this process that might help guide development of improved treatments, [187] modified a previously developed platform avoidance task [188, 189]. Briefly, after rats were trained to press a lever to receive reward, they learned that a tone predicted an upcoming shock. After tone onset, they had 30 s in which they could either continue pressing the lever to receive reward, or escape to a plexiglass platform where they couldn’t be shocked. After 10 days, rats were exposed to a 3-day “EX-RP” phase, during which access to the avoidance platform was blocked. Rats that were resistant to EX-RP (i.e., displayed high freezing throughout extinction) also showed persistent avoidance on a probe test during which the platform was no longer blocked. Interestingly, while this persistent avoidance in EX-RP resistant rats was reduced after lOFC inactivation, persistent avoidance was actually increased after lOFC inactivation in rats that had successful EX-RP. This is reminiscent of differential effects of optogenetic stimulation of OFC circuits in wild-type [186] and mutant [173] mice, and highlights the importance of examining the mechanisms underlying individual variation in both healthy and pathological model systems. Interestingly, DBS of the dorsal portion of the ventral striatum (dVS) (thought to be comparable to the VC/VS target used in DBS for OCD [190,191,192]) also led to a decrease in persistent avoidance in this model. Based on previous findings, it is possible that these therapeutic effects result from plasticity in PFC regions including mOFC, lOFC, and mPFC (PL/IL) [193,194,195] through antidromic activation [196,197,198]. A recent study examining the mechanism of action of DBS in SAPAP3-KO mice also found that DBS of either the IC or the dVS reduced compulsive grooming, with IC stimulation being more effective but also causing increased locomotion [199]. Note, both IC and dVS DBS led to c-fos induction in PL, mOFC, and lOFC (only in IC), but it remains to be tested whether these changes in PFC activation were related to DBS mechanism of action. Finally, in our recent study, we found that both behavioral deficits (impaired reversal learning, compulsive grooming) and abnormal lOFC activity patterns were improved by treatment with chronic fluoxetine, a first-line OCD treatment (Manning et al., bioRxiv). Though causality remains to be determined, this suggests that increased serotonin in the lOFC may be an important treatment mechanism in this animal model.

Future directions and challenges for the field

A wealth of behavioral, neuroimaging, and treatment studies in humans highlight the importance of the PFC in OCD pathology and treatment, with most evidence pointing to important roles for the OFC and ACC, and more recent meta- and mega-analyses highlighting the emerging roles of pre-supplementary and SMAs. However, several key questions remain to be addressed that will facilitate transforming these findings into improved treatment approaches.

First, how do we make sense of the pattern of hyperactivity of PFC regions at rest and during symptomatic states that are normalized with successful treatment, and hypoactivity during performance of specific neuropsychological/cognitive tasks? One potential explanation for this apparent contradiction is that by virtue of technical limitations, neuroimaging studies in OCD have often treated subregions of the PFC as homogenous structures. However, as discussed above, distinct cell-types within cortical regions can be defined not only by their neurotransmitter composition (e.g., glutamatergic vs. GABAergic), but also by their unique patterns of projections and inputs (e.g., PT and IT neurons [200]. For example, based on work in animal models it is conceivable that lOFC neurons that project to dorsal/central striatum are more involved in generation of OCD symptoms [173], while lOFC neurons projecting to BLA are more involved in mediating reversal learning [201, 202]. Different activity patterns in these distinct projections could therefore lead to hyperactivity vs. hypoactivity in different behavioral states. Assessing the same subjects in both symptom provocation and neurocognitive tasks could potentially help test this idea. However, it will also be important to further isolate postmortem abnormalities to particular circuits and cell types, and then take advantage of the capacity of animal models to dissect the contributions of specific PFC cell-types and projections to OCD-relevant behaviors using techniques such as optogenetics and chemogenetics [8]. In turn, this information could be used to develop more targeted neurostimulation-based treatments as suggested by [203,204,205].

In addition, to date, subregions of the PFC in OCD human imaging studies have been treated as fairly unitary structures–e.g., mOFC, lOFC, and ACC. However, there is evidence from humans, primates, and rodents supporting functional distinctions within these regions. For example, there are indications from rat studies that anterior and posterior mOFC may play distinct roles, with anterior mOFC being responsible for retrieval of unobservable outcomes [85], while posterior mOFC is involved in delay discounting [206]. This is consistent with findings from a meta-analysis of human neuroimaging studies indicating that more abstract and complex reinforcers are represented in activity patterns in the anterior, and not posterior, mOFC [207]. Higher resolution 7T neuroimaging methods may allow further parcellation of PFC subregions to determine if differential functions across anterior-posterior gradients contribute to the distinct PFC activity patterns seen during symptom provocation vs. neurocognitive tasks. This delineation could help develop more localized TMS protocols for particular OCD symptoms.

A final important question for the refinement of OCD treatment is how symptom subtypes should be delineated. Though there have been many proposed subtype designations, with some suggestions that particular subtypes may be differentially responsive to treatment [e.g., hoarding is generally less responsive to treatment [208,209,210], while tic-related OCD may be more responsive to antipsychotic augmentation [211]], large scale application of current symptomatic designations based on the five-factor designation [212, 213] have not yet led to development of targeted treatments. An example from the field of depression may provide a model for success [214]. Here, fMRI was performed on a large multisite sample (n = 1188), which afforded the statistical power to subdivide patients with depression into four neurophysiological subtypes (‘biotypes’) defined by distinct patterns of dysfunctional connectivity in limbic and frontostriatal networks. Though these biotypes couldn’t be differentiated solely on the basis of clinical features, they were associated with differing clinical symptom profiles. Importantly, in a subset of the patients that were treated with repetitive TMS of DMPFC (n = 154), these biotypes predicted treatment responsiveness. Similar large-scale studies in OCD could therefore begin to develop treatment-defining biotypes for OCD. Concerted investment of resources in OCD research could therefore realistically lead to the development of new, more tailored and effective treatments for this severe psychiatric illness.