1. Neural activity in the amygdala is highlighted in panic disorder (PD) patients.

2. The hippocampus, left-brain stem, and cingulate cortices were found to have significantly higher activation in PD.

3. PD patients were found to have greater activation in the parahippocampal gyrus and thalamus compared with healthy controls.

4. Multiple prefrontal cortices were implicated in the neural activity of PD patients, including the ventromedial prefrontal cortex, ventrolateral prefrontal cortex, dorsomedial prefrontal cortex, and dorsolateral prefrontal cortex.

5. The dorsal midbrain, left brainstem, S1, and right caudate were found to be hyperactivated in PD patients.

Panic disorder (PD) is a severe anxiety disorder characterized by a high degree of distress that is often occupationally and socially disabling [1, 2]. PD is defined by spontaneous and recurrent panic attacks (PAs) [3], likely initiated by complex fear circuitry in the brain and which remains poorly understood [4].

The fear circuitry comprises the amygdala, thalamus, hippocampus, insula, and prefrontal cortex, and involves neurobiological fear responses including neurochemical, neuroendocrine, and behavioral responses adaptive to survival [5]. Several neuroanatomical models have been proposed to explain panic and to investigate the fear circuits involved in the brain [6, 7, 8].

The primary objective of this review is to identify which neuroanatomical areas are implicated in the pathophysiology and etiology of PD. The secondary objective is to identify sophisticated translational models to evaluate how animal research enhances our understanding of the neurobiological foundations and pathophysiology of PD in humans. This systematic review attempts to update and consolidate the knowledge of PD in neuroanatomical and neurocircuitry factors.

This systematic review was conducted according to PRISMA guidelines. This review draws on articles found via the PubMed database. Clinical studies and a randomized controlled trial, published in English between 2010 and 2020, were selected. The keyword search included panic disorder*(neur*/fear circuitry/fear network/serotonin/amygdala/noradrenalin/biomarker/hypothalamus/corticotropin releasing* OR CRF OR CRH/functional near infrared spectroscopy OR fNIRS/angiotensin II type 1 receptor OR AT1R).

In the first step of the process, titles and abstracts were manually screened against the inclusion/exclusion criteria. At this stage, retained articles were assessed against the following inclusion criteria: (1) an original research paper, (2) focused specifically on PD with/without comorbidity, (3) focused on panic/fear circuitry, and (4) adult participants/animal studies. Articles are excluded if they were: (1) a meta-analysis/systematic review/theoretical literature, (2) unrelated to PD, (3) focused on the therapy modalities/pharmacological intervention of PD, and (4) without an abstract.

Next, full-text articles were screened for their eligibility for qualitative synthesis. The article inclusion process was conducted independently by two reviewers, PK and CW, with a third reviewer, RCF, involved to resolve any inclusion disagreement before proceeding. Satisfactory articles were included in the synthesis and quality assessment used (see Fig. 1 for the PRISMA flow chart outlining the study identification and selection process, and see Table 1, Ref. [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28] and Table 2, Ref. [29, 30, 31] for quality assessments of human and animal studies, respectively). The Newcastle-Ottawa Scale [32] was used to assess the quality and risk of bias in human studies (see Table 3, Ref. [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]). For assessing animal studies, the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) [32] risk of bias tool was used (See Table 4, Ref. [16, 17, 18, 19, 20, 21, 22, 23, 24, 25]). In the systematic review, means and standard deviations were extracted from the primary data sources of included studies. This extraction process involved a detailed examination of the methodology and results sections of each study to identify the reported means and standard deviations pertaining to the relevant outcomes.

Fig. 1.

PRISMA flow chart outlining the study identification and selection process.

The keyword search generated a total of 2025 articles. Duplicates and irrelevant articles were removed, and 1063 articles were retained for abstract screening. The exclusion criteria removed 806 articles and 30 articles were included for the full-text assessment. In the full-text screening stage, 227 articles were excluded and 16 articles were included in the qualitative data synthesis. The 16 articles were analyzed, and findings were synthesized into one main category: neuroanatomical studies, with a primary focus on explaining PD etiology.

This paper reports solely on the neuroanatomical findings from this systematic review, which encompasses both human and animal studies.

3.1 Neuroanatomical Human Studies

The database search yielded 13 human studies meeting the inclusion criteria. Nine of the human studies compared the brain activity and/or connectivity between different brain regions in PD patients. The remaining studies investigated neurophysiological connections in the brain and the physiological effect of their intervention on PD. A summary of the neuroanatomical studies can be found in Table 3. Descriptive statistics for these studies include mean $\pm$ standard deviation (SD).

The brain imaging studies adopted a variety of interventions/exposure methods (i.e., fear/extinction conditioning, predictable and non-predictable cues [NPU paradigm], aversiveness task, emotional face observation task, and verbal fluency task [VFT]) as well as the absence of intervention (i.e., resting state functional connectivity [RSFC]). The non-brain imaging studies were observed to implement more CO₂ challenge interventions to induce panic.

The findings of the brain imaging studies are summarized in Table 4. Brain activation is noted in the Montreal Neurological Institute (MNI) coordinate system or the Talairach coordinates. Based on these coordinates, a map of brain activation was created to better illustrate the neurocircuitry of Parkinson’s disease patients undergoing various interventions (see Fig. 2).

Fig. 2.

Brain map illustration of the neural activation of the brain imaging study results. Note: The brain map illustration is mapped against a standardized model of a lateral brain view. The colored dots were mapped according to the MNI/Talairach coordinates stated in Table 2. There may be some degree of deviation in the location of the colored dots. vmPFC, ventromedial prefrontal cortex; vlPFC, ventrolateral prefrontal cortex; dmPFC, dorsomedial prefrontal cortex; dlPFC, dorsolateral prefrontal cortex; PCC posterior cingulate cortex; MCC, midcingulate cortex; ACC, anterior cingulate cortex; dACC, dorsal anterior cingulate cortex; rACC, rostral anterior cingulate cortex; IPS, intraparietal sulcus.

3.1.1 Amygdala

Neural activity in the amygdala in PD patients was highlighted in six studies. In this systematic review, only one neuroanatomical study within the perimeter of the database search adopted the resting state functional connectivity approach. Pannekoek et al. [16] studied the aberrant limbic and salience network resting-state functional connectivity in PD patients without comorbidity. The study implemented a seed-based correlation approach to investigate the RSFC in PD patients. Three seed regions using the MNI coordinates were observed that included the amygdala, dorsal anterior cingulate cortex (dACC), and posterior cingulate cortex (PCC). Based on this seed region, Pannekoek et al. [16] found that compared with healthy controls (HC), the amygdala has negative connectivity at coordinates with the following regions: bilateral precentral and postcentral gyrus, right supplementary motor cortex, and the rostral anterior cingulate cortex (rACC).

Meanwhile, four studies investigated the neural activity of PD patients using either panic, disorder-related, or fear-related scripts/scenes/contexts and demonstrated similar findings. Four studies noted PD patients experience significant hyperactivation in the amygdala region compared with their HC counterparts [17, 18, 19, 20]. Each study seems to show consistent results, indicating that PD patients exhibit increased activation in the amygdala in fear/panic/disorder-related scripts or contexts. Hyperactivation is noted in both the right amygdala [17, 18, 19, 21] and the left amygdala [17, 20].

3.1.2 Hippocampus

A similar part of the hippocampus was noted for its higher activation in two studies. Killgore et al. [20], who studied brain activity in a fear vs happy context, identified the hippocampus as one of the regions with significantly higher activation in PD compared with patients with social phobia (SP) and HC in a fear vs happy contrast context. Hippocampus activity was also observed in a fear conditioning study by Marin et al. [22], which found that HC had higher hippocampus activity compared with PD in late conditioning.

3.1.3 Parahippocampal Gyrus

Killgore et al. [20] found significantly greater activation in the parahippocampal gyrus in PD patients compared with HC (p 0.001) for a fear vs neutral contrast context. Fonzo and colleagues [18] assessed for brain activity in an emotional face task context and found significant positive relationships between trait anxiety and brain activation (p = 0.001). Both studies implemented conditioning intervention and activated a brain area that is similar and comparable.

3.1.4 Thalamus

Activities in the thalamus were mentioned in two studies. Feldker et al. [17] indicated that hyperactivation was distinct in PD patients for panic-related scenes compared with neutral scenes. This hyperactivation is significantly higher in PD patients compared with HC (p 0.05). The activation was noted in the left hemisphere. Balderston et al. [23] noted hyperactivation in the left hemisphere in PD for predictable threat compared with unpredictable and neutral contexts in the fear network (p 0.001).

3.1.5 Prefrontal Cortices

Multiple researchers highlighted the involvement of regions of the prefrontal cortex, which included the ventromedial prefrontal cortex (vmPFC), ventrolateral prefrontal cortex (vlPFC), dorsomedial prefrontal cortex (dmPFC), and dorsolateral prefrontal cortex (dlPFC), in the neural activity of PD patients under a variety of stimulus conditions.

Regarding the vmPFC, Marin et al. [22] found greater activation in the left vmPFC in HC compared with anxiety-disordered patients (PD included) in the early conditioning paradigm (p 0.009) and less activation in the left vmPFC in extinction recall (p 0.02). Burkhardt et al. [19] identified significantly decreased activation in the right vmPFC in PD patients during the imagination phase of the disorder-related vs neutral script (p 0.05). Balderston et al. [23] found significantly lower bilateral activation of the vmPFC toward predictable and unpredictable cues compared with neutral cues in PD patients (p 0.001).

vlPFC neural deactivation was noted only by one study, Burkhardt et al. [19]. They found vlPFC deactivation in both the left and right vlPFC. The hypoactivation is more relevant for the phase of the related disorder than the neutral script imagination.

Similar hypoactivation in relation to the related disorder and neutral script imagination was noted in the left dmPFC [18]. Interestingly, PD patients experienced significantly higher activation of the dmPFC during a predictable cue compared with an unpredictable and neutral cue only in the fear network (p 0.001) [23].

The same two studies, Burkhardt et al. [19] and Balderston et al. [23], highlighted similar neural responses of the dlPFC in PD patients. Burkhardt et al. [19] discovered that the right dlPFC showed decreased activation during disorder-related imagination compared with neutral script (p 0.05). The left dlPFC is significantly less active for predictable cues vs neutral cues in the default mode network (DMN).

3.1.6 Insula

Six studies identified activity in the insula at 11 coordinate points. As seen in Table 2, activity in the insula was observed in both the left and right hemispheres.

In the study by Marin et al. [22], anxiety disorder patients (PD patients included) showed higher activation in the insula during the extinction recall paradigm. Hyperactivation of the insula was also observed in a study by Feldker et al. [17] that utilized panic-related scene intervention where hyperactivation was noted at three points in the left insula. The three points of hyperactivation that were significantly activated for panic-related scenes vs neutral scenes compared with HC were noted. Interestingly, Killgore et al.’s study [20] that implemented a happy vs neutral context found that PD showed greater happy vs neutral contrast compared with HC.

In the NPU paradigm studies, Lieberman et al. [21] observed significant activation of the right insula for all participants in an unpredictable threat condition. More comprehensive results were found by Gorka et al. (2014) [24], whereby PD with Major Depressive Disorder (MDD) comorbidity patients showed greater left and right insula activation during unpredictable threat compared with MDD patients and HC. In addition, Balderston et al. [23] found that significantly higher insula activation was observed for predictable threat conditions in the fear network only in PD patients. Significant activities were noted in both hemispheres. The findings of these studies appear to strongly support that the insula has greater activation during an adverse related stimulus, i.e., unpredictable threat, predictable threat, and panic-related scenes.

3.1.7 Cingulate Cortices

Amongst the other brain regions, five studies observed notably one or more activities in the regions of the cingulate cortices, which include: midcingulate cortex (MCC), anterior cingulate cortex (ACC), dorsal anterior cingulate cortex (dACC), and rostral anterior cingulate cortex (rACC).

Feldker et al. [17] found significant bilateral hyperactivation in the MCC region at two peak coordinates. The two points demonstrated hyperactivation, which was higher in panic-related scenes compared with neutral scenes of the intervention for PD patients than in HC.

Feldker et al. [17] also found that the ACC experienced significant hyperactivation for the same intervention phase as the MCC. Two coordinates in the left ACC demonstrated hyperactivation in PD patients during panic-related scenes as compared with neutral scenes.

Notable activity in the dACC was observed in two studies. Pannekoek et al. [16] discovered that at resting state, PD patients have decreased connectivity between the left dACC and the bilateral frontal pole and superior/medial frontal gyrus compared with HC. However, increased connectivity between the left dACC and the bilateral pre-central and post-central gyrus was noted. Lieberman et al. [21] found that greater activation in the dACC is associated with increased panic symptoms as measured with IDAS-II during U-threat.

For the rACC, two studies suggest similar findings that PD and AD experience less or lower activation in the rACC compared with HC. Marin et al. [22] identified that AD (including PD patients) show less activation in extinction recall. Meanwhile, Burkhardt et al. [19], found that HC has higher activation in the left rACC during imagination of disorder-related script compared with PD patients. Furthermore, the reduced activation in PD was noted particularly on the right rACC.

Lastly, the subgenual cingulate cortex was noted in one study. Tuescher et al. [25] identified that there is reduced activation in the subgenual cingulate cortex in response to threats and increased sensitivity of this region to safe conditions was reported in PD patients during an instructed fear-conditioning paradigm.

3.1.8 Dorsal Midbrain

Only one study noted the activity of the dorsal midbrain for PD patients. Tuescher et al. [25] found a relative increase in the interaction contrast in PD patients compared with post-traumatic disorder (PTSD) patients in the threat vs safe condition paradigm.

3.1.9 Brainstem

Finding for the brainstem region were like most findings in the cingulate cortices. The left brainstem was more hyperactivated in PD during panic-related than during neutral scenes compared with HC at two points.

3.1.10 Right Caudate

Tuescher et al. [25] also identified activity in the right caudate where there was a relative increase of interaction contrast for PD vs PTSD and threat vs safe conditions.

3.1.11 S1

Balderston et al. [23] found that S1 activity was significantly higher in response to predictable cues compared with unpredictable and neutral cues in PD patients, but this was observed only within the fear network.

3.1.12 Intraparietal Sulcus

Balderston et al. [23], also found that the left intraparietal sulcus (IPS) showed significant hypoactivation for predictable cues compared with unpredictable and neutral cues for the default mode network (DMN).

This systematic review yielded findings related to the neuroanatomical factors playing a role in the etiology and pathophysiology of PD. A qualitative systematic review is best suited to highlight the most significant findings.

Neuroanatomical brain imaging findings in humans highlighted several key areas involved in the pathophysiology of Parkinson’s disease, including the amygdala, hippocampus, parahippocampal gyrus, thalamus, brainstem, prefrontal cortex (PFC), insula, and cingulate cortices. The cingulate cortices are comprised of the midcingulate cortex (MCC), the anterior cingulate cortex (ACC), the dorsal anterior cingulate cortex (dACC), and the rostral anterior cingulate cortex (rACC). The dorsal midbrain, right caudate [35], and left brainstem [20] are also implicated in a relative increase in interaction contrast in PD patients. Hypersensitivity in the brainstem and the amygdala play a role in the pathogenesis of PD and in the activation of the fear network which involves sub-cortical and cortical regions.

The amygdala was highlighted in six studies in this review, with four studies indicating that PD patients have significant hyperactivation in the amygdala region compared with HC. Moreover, the coordinates of the amygdala activity across the four conditioning studies indicate a substantial degree of overlap in both left and right lateralization [17, 19, 20, 21]. Overall, PD patients appear to have either hyperreactive or hypersensitive amygdala when stimulated with a non-neutral stimulus (i.e., fear contrast stimulus, happy contrast stimulus, angry contrast stimulus, panic-related scenes, disorder-related scripts). According to Pannekoek et al.’s study [16] the connectivity between the amygdala and the bilateral pre-central and post-central gyrus, the right supplementary motor cortex, and the rACC appear to be reduced in PD. Therefore, the connectivity related to emotional processing between the amygdala and the abovementioned linked brain region may be impaired in PD patients. The morphometric measurements of the amygdala may point to the pathophysiological mechanisms underlying PD [25]. The resilience in anxiety states such as PD might be inhibited by altered neuronal integration and validation of anxiety-related emotional stimuli [36]. Abnormalities in regulating emotional processing have also been noted to contribute to the pathophysiology of PD [37, 38].

Several neurotransmitters that have lower receptor binding in the amygdala, including GABAA and serotonin, have been reported. Particularly, a study that used a 20% CO₂-panic provocation model in rats showed that orexin (ORX) neurons in the dorsomedial/perifornical regions are important for triggering coordinated panic reactions. Activation of ORX-synthesizing neurons induces a panic-prone state in the rat panic model [33]. ORX1 receptor antagonists reduce panic responses via neuronal networks involving the extended amygdala, periaqueductal gray, and medullary autonomic regions [33].

The hippocampus has also been implicated in fear circuitry, given its significant role in emotional regulation and contextualizing fear responses. Research has demonstrated that fear conditioning is compromised in patients with amygdala lesions; however, fear conditioning is not affected by hippocampal lesions [5, 39]. The hippocampus processes risk assessment, which is a fundamental aspect of emotional regulation aimed at appraising potential danger versus rewards [22]. Moreover, the role of the hippocampus in PD is in the expression of fear and anxiety elicited by learned fear contributing to the integration of defensive neural networks that make up the fear circuitry, comprising of the hippocampus, amygdala, nucleus accumbens, periaqueductal gray, ventromedial hypothalamus, thalamic nuclei, insular cortex, and several brain stem and prefrontal regions [22].

The hippocampus has been found to have higher activation in PD compared with patients with social phobia (SP) and HC in a fear vs happy contrast context [20] and higher hippocampus activity has been found in HC in comparison with PD patients in late conditioning [20, 22]. Killgore et al. [20] reported significantly greater activation in the parahippocampal gyrus in PD patients when compared with HC. Moreover, the left intraparietal sulcus showed significant hypoactivation for predictable cues compared with the unpredictable and neutral cues in PD patients for DMN. Although the S1 area showed significantly higher activity for predictable cues compared with unpredictable and neutral cues in PD patients, it was only in the fear network [23].

The human neuroanatomical brain imaging findings regarding PD might be a consequence of neurochemical alterations in the PD central nervous system, resulting in neuroimage alteration. This review described some common findings regarding the neurochemical factors involved in PD. For instance, the effects of brain-derived neurotrophic factor (BDNF) and tyrosine receptor kinase B (TrkB) signaling in the PD model were demonstrated to be an important site for dPAG activation, as BDNF panicolytic-like effects occur via $\gamma$-aminobutyric acid type A (GABAA)-dependent mechanisms. BDNF exerts a modulatory effect on the serotonergic system in brain loci (periaqueductal grey and dorsal raphe nucleus) [40]. The periaqueductal grey (PAG) has high levels of TrkB and BDNF receptor messenger RNAs and proteins [41].

In animal models, electrical or chemical stimulation of the PAG induces escape responses and autonomic changes that are like those observed in aversive situations [41, 42]. Using electrical stimulation of the dPAG as a model of panic [42, 43], intra-dPAG injections of serotonin (5-HT) [44], or GABA-enhancing drugs reduce the escape response triggered by this stimulation, suggesting a panicolytic-like effect [31]. Insights from animal studies suggest that GABAergic neurons can exert a strong inhibitory effect on the dorsomedial and posterior hypothalamic nuclei, thereby reducing the excitability of neurons involved in the development and expression of panic-like responses [45]. A specific hypothalamic nucleus, the dorsalmedial hypothalamic nucleus (DMH), and a dysfunction in its regulatory mechanism may be relevant in the genesis/maintenance of panic disorder [46].

Regarding other neurotransmitters, most of the brain lactate and glutamate concentrations change in PD patients. A significant difference in visual cortex lactate/N-acetylaspartate was observed in PD patients, during and following the visual stimulation and recovery period. An important finding also suggests that glutamatergic baseline concentration mainly determines the degree of glutamate + glutamine/creatine. Brain lactate in PD is argued to be influenced by excessive cerebral vasoconstriction that leads to brain hypoxia and metabolic disturbance [47]. In addition, investigation of the relationship between PD and serotonin reuptake inhibitors (SRIs) on the coupling of cortical and cardiac activity has found that PD patients have higher N300H magnitudes compared with HC. This phenomenon has been labeled ‘N300H’ to indicate a negative association between EEG amplitude at 300 ms and the heart period (the acceleration at subsequent beats) following an external stimulus. Moreover, SRI treatment resulted in greater N300H activity spread in PD patients than in non-SRI-treated PD patients.

In this review, we have presented animal and human studies regarding the neuroanatomical areas that are salient in PD. These studies have identified patterns of altered expression in several biological systems, such as neurotransmission, the hypothalamic pituitary adrenal axis, and neuroplasticity, resulting in neuroanatomical modifications.

Complex emotional and cognitive processing in neuropsychiatric illnesses is associated with abnormal functioning of neural circuits, which incorporate several brain regions [22] that are responsible for varying types of defensive responses and fear circuitry. Therefore, an understanding of the brain regions involved and their functional connectivity may further inform our understanding of the neurobiological foundation of PD, further leading to the development of effective interventions [22].

This systematic review is registered under PROSPERO with registration number CRD42021247285. PROSPERO registration can be retrieved from https://www.crd.york.ac.uk/prospero/.

The PRISMA protocol was used for this systematic review and is described in the methods section of the article.

Data is available in the original research articles available via the PubMed database.

Conception–PK, CW, RCF, AEN; Design–PK, CW; Supervision–PK, RCF, AEN; Fundings–PK; Materials–PK, CW; Data Collection and/or Processing–PK, CW; Analysis and/or Interpretation–PK, CW, LQ; Literature Review–PK, CW; Writing–PK, CW, LQ; Critical Review–PK, CW, LQ, RCF, AEN. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Not applicable.

This research received no external funding.

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/AP38756.