Panic disorder (PD), a complex anxiety disorder characterized by recurrent panic attacks, represents a poorly understood psychiatric condition which is associated with significant morbidity and an increased risk of suicide attempts and completed suicide. Recently however, neuroimaging and panic provocation challenge studies have provided insights into the pathoetiology of panic phenomena and have begun to elucidate potential neural mechanisms that may underlie panic attacks. In this regard, accumulating evidence suggests that acidosis may be a contributing factor in induction of panic. Challenge studies in patients with PD reveal that panic attacks may be reliably provoked by agents that lead to acid–base dysbalance such as CO2 inhalation and sodium lactate infusion. Chemosensory mechanisms that translate pH into panic-relevant fear, autonomic, and respiratory responses are therefore of high relevance to the understanding of panic pathophysiology. Herein, we provide a current update on clinical and preclinical studies supporting how acid–base imbalance and diverse chemosensory mechanisms may be associated with PD and discuss future implications of these findings.
Panic disorder (PD) is characterized by spontaneous and recurrent panic attacks that consist of incapacitating periods of acute-onset respiratory, cardiovascular, gastrointestinal, autonomic and cognitive symptoms. PD—which occurs in 6% of Americans1—typically begins in the second decade of life2 and exhibits a peak prevalence in the third and fourth decades of life.3 Thus, this condition is second only to major depressive disorder in terms of associated debility among psychiatric conditions in the United States.4 Importantly, PD also represents an independent risk factor for suicidality in diagnostically and demographically heterogeneous clinical populations5 and increases the risk of developing other anxiety disorders and secondary mood disorders.2 Yet, many patients suffering from PD are not clinically identified and frequently, do not receive even minimally effective treatment.6 Even still, available psychopharmacologic treatments (for example, selective serotonin reuptake inhibitors, benzodiazepines) and psychotherapies (for example, cognitive behavioral therapy, prolonged exposure therapy, psychodynamic psychotherapy) or the combination of psychotherapy+pharmacotherapy are often only modestly efficacious (for example, Cohen’s d=0.4–0.6)7, 8 and in some cases (for example, benzodiazepines) may be associated with treatment-specific side effects or risks such as sedation or the risk of dependence or tolerance.
Studies elucidating the pathoetiologic mechanisms of PD are urgently needed to reduce morbidity and mortality. Despite the prevalence, as well as associated morbidity and mortality of PD, relatively little regarding the neuropathophysiology of this condition is known. PD is highly heterogenous with variable symptom profile and intensity in panic episodes experienced by the same individual and across patients. According to the DSM-5,9 recurrent panic attacks in PD are categorized as being either spontaneous (unexpected) or cued (expected). Collective evidence from challenge studies in the laboratory, neuroimaging, symptomology, treatment responses and translational animal models have led to an increased understanding of PD.10, 11, 12, 13, 14 Accumulating evidence suggests that expected panic attacks are triggered by exteroceptive threats (that is, a panic attack context or other unrelated stressors) while spontaneous panic attacks may be provoked by interoceptive sensory triggers caused by fluctuations in the internal homeostatic milieu. An important internal homeostatic trigger for the genesis of panic attacks, supported by an emerging body of work, is acid–base imbalance and associated pH chemosensory mechanisms. Largely founded on panic provocation studies with agents promoting homeostatic pH imbalance and related to the false suffocation alarm theory, the role of acid–base and chemosensory systems in panic provides strong scientific insights on the genesis of uncued panic attacks which may sensitize fear-arousal-stress regulatory circuits to other triggers leading to full-blown PD (Figure 1, cycle of panic). Given the high relevance of interoceptive mechanisms in PD, this review provides an update on our current knowledge and understanding of the role of pH imbalance and chemosensory targets in PD. Although excellent reviews on this topic have appeared previously,15, 16, 17 here we focus on (1) current status on pH homeostasis, clinical studies of acid–base physiology and pathophysiology in patients with PD (2) preclinical rodent models of PD, especially those focusing on interoceptive pH imbalance and acid-chemosensory systems recruited in panic-like behaviors, and last (3) synthesize these findings to develop a working neurobiological model of PD that involves dysregulation of central acid sensing and associated circuitry, and finally, (4) translational relevance of these data, gaps in understanding and future implications with a discussion on neuropharmacologic interventions in patients with PD.
Potential pathogenesis of uncued and cued panic attacks in panic disorder: Initial unexpected attacks may result from an acid/base imbalance or from altered chemosensory mechanisms that represent a ‘threat to homeostasis’. Although the exact origin of pH disturbance is unknown, it may arise due to genetic predisposition, respiratory abnormalities and other factors. This may produce a state of alarm and subsequent activation of threat response systems leading to elevated fear, cardiovascular and respiratory symptoms which, phenomenologically, constitute a panic attack. Further, experiencing uncued panic attacks may sensitize threat response systems to exteroceptive triggers such as stress, panic context and associated phobic cues leading to cued panic attacks. Persistence of uncued and cued panic attacks results in full-blown panic disorder.
Many clinical and preclinical studies have focused on dysregulation of central fear circuitry that includes components of the limbic network, which involves connections between the amygdala and the anterior cingulate cortex (Broadman’s area 25, 24/32) as well as midbrain regions including the periaqueductal gray matter in panic.10, 14 These studies have provided useful information relevant to cued panic attacks and phobias in PD subjects; however, the genesis of unexpected panic attacks still remains elusive to panic researchers. James18 first proposed that feelings and emotion can derive from interoceptive sensing of our body states. These internal triggers and interoceptive chemosensory pathways are of particular relevance to PD as initial attacks come ‘out of no-where’. Moreover, accumulating evidence supports a principal role of pH homeostasis in panic physiology and suggests that acidosis may be an interoceptive trigger for panic attacks. Consistent with this, a recent study with ambulatory monitoring, a valid approach for studying spontaneous panic, reported pH disturbances and altered respiratory rhythms in subjects during the final minutes before the onset of a panic attack.19 Well-characterized and clinically relevant methods of inducing panic, such as CO2 inhalation or sodium lactate infusion, cause acid–base disturbances.16 Importantly, CO2 inhalation and sodium lactate infusion stimulate respiration, which is itself tightly regulated by pH.20 Thus, it is intriguing that despite their disparate effects, respiratory acidosis and metabolic alkalosis, both CO2 inhalation and sodium lactate administration lead to extracellular and intracellular acidosis in the brain.16, 21 In addition, neuroimaging studies also raise the possibility of dysregulated acid–base buffering and increased plasma and brain lactate responses to metabolic challenges in PD.22, 23, 24 There is also a high prevalence of hyperventilation and other respiratory abnormalities among patients with PD.21, 25, 26 The link between panic attacks and pH disturbances also forms the basis of the false suffocation alarm theory of spontaneous panic, where CO2 hypersensitivity may exist due to a malfunctioning suffocation alarm monitor.27 Below we discuss specific areas of investigation that support the panic-pH link.
Brain pH in patients with PD: evidence from neuroimaging
Neuroimaging studies on PD patients support a role of homeostatic pH disturbances in panic physiology (Table 1). The majority of these studies have focused on lactate responses to homeostatic or activity-dependent challenges. Given the close relationship between lactate and pH in the brain, the findings are consistent with a model of brain metabolic and pH dysregulation associated with altered function of acid-sensitive fear circuits as a trait vulnerability factor in PD. Exaggerated activity-dependent brain lactate responses are observed in PD patients, even remitted patients, as compared with healthy controls, suggestive of underlying pH abnormalities.23, 28 Further, activity-dependent changes in glutamate–glutamine were smaller in PD patients.28 Others have shown increased (and more prolonged) brain lactate levels compared with healthy comparison subjects during sodium lactate infusion, although resolution was limited.29 In a subsequent study, the greater and prolonged brain lactate rises in the insula in patients with PD during and following lactate infusion30 were observed—a finding of great importance given the central role of the insula in interoceptive pathways.31 Interestingly, differential brain metabolic responses fail to normalize following treatment with the selective serotonin reuptake inhibitor fluoxetine suggesting that abnormal metabolic lactate responses represent a trait feature of PD. Accentuated increases in brain lactate levels have also been reported in PD subjects during hyperventilation, that promotes pH imbalance, as compared with comparison subjects,22 suggesting that increased lactate possibly reflects altered cerebral blood flow in response to hypocapnia evoked by increased breathing. Using phosphorous (31P) spectroscopy, pH dynamics during hyperventilation were assessed by Friedman et al.,32 and reported an abnormal pH-buffering capacity in PD. Very recently, CO2-evoked activation in the brain stem area was reported in panic patients, controls and divers (representing a group with reduced CO2 sensitivity).33 The authors reported significant increases in activation within brain stem and insular areas in panic patients as compared with controls and diver groups supporting the role of these areas in interoceptive sensing and processing in PD. Recently, new magnetic resonance imaging techniques such as the T1 relaxation in the rotating frame (T1ρ) have been developed and have facilitated greater pH sensitivity and improved temporal and spatial resolution over 31P spectroscopy.34 Increased T1ρ was observed in PD patients in the occipital cortex as compared with healthy controls35 consistent with abnormal pH regulation in PD. Interestingly, the magnitude of the T1ρ response correlates with the severity of anxiety symptoms in patients with PD (but not in healthy subjects). Thus, it is increasingly apparent that this new imaging technique will allow for greater understanding of pH-dependent changes in PD. Finally, although studies of panic symptom provocation are still needed, evidence from functional magnetic resonance imaging and functional magnetic resonance imaging supports differences in metabolic activity or other factors such as cerebral blood flow may lead to altered pH homeostasis in PD.
Table 1 Studies of pH or lactate-related brain changes in patients with panic disorder
Full size table
Symptom provocation challenge studies with agents producing acid–base imbalance
Panic attacks can be induced in PD individuals by a variety of agents such as CO2,36, 37 sodium lactate,38 doxapram,39 isoproterenol,40 caffeine,41 yohimbine,42 cholecystokinin (CCK-4 and agonists),43 the benzodiazepine receptor antagonist, flumazenil,44 serotonin receptor agonists45 and opioid receptor antagonists.46 It is important to note, however, that homeostatic perturbations such as CO2, lactate and doxapram that cause direct or compensatory pH shifts reliably evoke panic attacks that closely resemble spontaneous panic attacks. Furthermore, as opposed to other challenges, these agents induce panic attacks specifically in PD subjects versus other disorders such as depression, premenstrual dysphoric disorder, generalized anxiety disorder and posttraumatic stress disorder.47, 48 Panic provocation agents linked to homeostatic pH imbalance are discussed below.
Inhalation of CO2—a commonly studied interoceptive stimulus—produces intense fear, autonomic and respiratory responses that can evoke panic attacks in individuals with PD. For this reason, CO2 is frequently used as a biological challenge and pathological marker of PD.11, 36 First described in 1951 by Cohen and White,37 CO2 inhalation is established as a reliable panicogen in patients with PD.11, 36, 49, 50
The partial pressure of CO2 in the blood increases following CO2 inhalation challenge. In addition, and of direct relevance to central nervous system (CNS) physiology, CO2 readily crosses the blood–brain barrier and is sensed by H+ and CO2 chemoreceptors in the CNS and periphery.51 In the extracellular fluid, CO2 is hydrolyzed to carbonic acid (H2CO3) by carbonic anhydrase which readily dissociates into bicarbonate (HCO3-) and H+.51 The resulting acidosis is thought to be the trigger for the panic symptoms caused by this challenge including hyperventilation and increased blood pressure.52 Klein puts forth in his false suffocation theory that hyperventilation may have a protective role to combat attacks caused by increases in CO2 (acidosis).27 However, this is a faulty response because the respiratory alkalosis caused by hyperventilation is always associated with a compensatory metabolic acidosis produced by tissue buffering systems that release H+ ions.21 Panic challenge studies with acetazolamide shed light on the role of protons as effector molecules for generating panic responses; acetazolamide, a carbonic anhydrase inhibitor, blocks the facilitated conversion of CO2 to bicarbonate and H+, leading to increases in CO2 concentrations. Interestingly, administration of intravenous acetazolamide fails to induce panic attacks in patients with PD53, 54 suggesting that H+ ions, rather than CO2 per se may facilitate panicogenesis.
Currently, two CO2 inhalation techniques are used in panic challenge studies. In the first, steady-state inhalation, a low concentration of CO2 (5–7.5%) is inhaled for approximately 1–20 min or until a panic attack occurs. In the second approach, individuals inhale a high concentration of CO2 (35%).36 The advantage of modeling CO2-induced panic is that these CO2-induced panic attacks closely resemble spontaneous panic attacks and the attacks resolve quickly.11 Interestingly, although PD is twice as likely to occur in women,3 sex differences in CO2-reactivity are less clear. Although there is some evidence that women report greater fear and anxiety following a CO2 challenge,55, 56, 57 not all studies have observed gender effects.50, 58, 59
CO2 inhalation has also been useful for exposure-based treatments in patients with PD60, 61 and has been utilized for validation of current treatments such as selective serotonin reuptake inhibitors: paroxetine, sertraline, fluvoxamine62 and benzodiazapine alprazolam.63 In addition, screening of potential anti-panic medications such as CRF1 receptor antagonist, R317573,64 GABA agonist: zolpidem63 and neurokinin-1 receptor antagonist: vestipitant65 has also been conducted using this challenge. Thus, CO2 inhalation appears to have utility for testing the efficacy of pharmacotherapeutic agents and for identifying vulnerability to PD.
Sodium lactate infusion
In addition to CO2, sodium lactate is a reliable panicogen38 frequently used in challenge paradigms. A masked intraveneous infusion of a 0.5 M sodium lactate (10 ml kg−1) produces panic attacks in vulnerable individuals.38, 66 Lactate-induced panic attacks, like CO2-induced panic attacks, phenomenologically mirror spontaneous panic attacks (that is, symptoms of dyspnea, generalized fear, a desire to flee and fear of losing control.67 Clinically, susceptibility to lactate-induced panic attacks are frequently used as treatment outcome measures for psychopharmacologic treatments.68, 69, 70
A byproduct of cellular metabolism, lactate serves as an energy source for neurons,71 and alters systemic acid–base balance. Pertinent to lactate infusions, lactate can cross the blood–brain barrier through monocarboxylate transporters and there is evidence that lactate becomes a significant fuel source in the brain when elevated in blood.72 When administered intravenously to lower primates, lactate decreases brain pH73 as H+ is co-transported with lactate via monocarboxylate transporters. Although lactate infusion may evoke acidosis, a direct role of pH in lactate-evoked panic has not been demonstrated. Interestingly, patients with PD show exaggerated lactic acid production in response to alkalosis evoked by sodium lactate infusion suggestive of increased compensatory drive and impaired acid–base buffering in these individuals.17 Other studies reported that a rapid overload of sodium and resultant acute hypernatremia may contribute to sodium lactate-evoked panic since hypertonic saline (3%) facilitated panic symptoms similar to 0.5 M sodium lactate.74 An interesting observation in the study was the induction of mild acidosis by hypertonic saline while sodium lactate-evoked hyperventilation and associated alkalosis, although specific parameters such as blood pCO2 were not measured. Lactate-evoked panic attacks do not recruit neuroendocrine responses as a dissociation between autonomic activation and cortisol has been reported in ‘panickers’ following sodium lactate.75 Potential downstream mechanisms for lactate sensitivity in PD are not clear. Involvement of GABAergic system has been suggested by effective blockade of lactate-evoked panic in subjects treated with gabapentin,76 while presynaptic, α2adrenergic agonist, clonidine had partial effects. Additionally, concentrations of endogenous neuroactive steroid modulators of the GABAA receptor, allopregnanolone and pregnanolone are decreased in patients with PD during lactate-evoked panic.77 Elegant preclinical studies by Shekhar and colleagues have highlighted the role of circumventricular organs (CVOs), hypothalamic GABA, angiotensin and orexin systems in sodium lactate-evoked panic responses (see section ‘Sodium lactate rodent model and hypothalamic GABAergic and acid-chemosensitive orexin targets’). Thus, the underlying mechanism(s) or a direct role of acidosis in sodium lactate-induced panic attacks has not been elucidated to date, and while lactate may contribute to decreased brain pH, the exact effector in lactate-evoked panic in PD subjects is still unclear.
Doxapram, a respiratory stimulant first synthesized in 1962 (Ward and Franko78) has been examined in the management of acute respiratory failure during the 1960s and 1970s and likely had a specific role in the treatment of individuals with chronic obstructive pulmonary disease. Specifically, administration of doxapram increases tidal volume and ventilation frequency.79 Clinically, doxapram use is primarily limited to post-anesthesia shivering prophylaxis and stimulation of respiratory drive in premature infants. However, its use is also associated with second-degree atrioventricular block and QTc prolongation. Regarding patients with PD, doxapram has been examined in several studies, given its tendency to cause hyperventilation. In this challenge, doxapram (0.5 mg kg−1, IV)39, 80, 81 produces panic attacks that phenomenologically mirror spontaneous panic attacks with associated hyperpnea, tachycardia, increased blood pressure and fear symptoms. Contribution of forebrain regions in doxapram-evoked panic is supported by recent studies where doxapram was administered to PD patients and healthy subjects before positron emission tomography with 18F-deoxyglucose.82 Cardiac responses were accentuated and patients with PD exhibited decreased prefrontal activity (relative to controls) and increased activity within the cingulate gyrus and amygdala, suggesting a failure to activate prefrontal inhibitory structures in patients with PD.
The underlying mechanism, specifically a role of pH in doxapram-evoked panic has not been established. Hyperventilation induces alkalosis, which has been reported to evoke a compensatory increase in lactic acid release; a response that is exaggerated in PD patients.17 It is unclear, however, whether this compensatory increase in acidosis is associated with doxapram-evoked panic attacks. There is evidence that the effects of doxapram may be related to the inhibition of TASK-1 and TASK-3 acid-sensitive potassium channels located in brain stem serotonergic neurons.83, 84 In this regard, inhibition of the TASK-1 and TASK-3 channels could increase the excitability of brain stem pH-sensitive neurons and may link the panicogenic action of CO2 inhalation and doxapram administration.85 In addition, increased respiratory drive by doxapram may aggravate a pre-existing respiratory abnormality in PD. As for lactate, a potential role of pH and acidosis in panic provocation by doxapram may be speculated, however, direct evidence for this link is currently lacking.
A strong contribution of genetics and family history in PD prevalence was first reported by Crowe and colleagues.86 In support of a genetic component in vulnerability to interoceptive triggers and PD, higher sensitivity to 35% CO2 was observed in first-degree relatives of patients with PD.87 CO2 hypersensitivity has been proposed as a genetic risk and disease-specific trait marker for PD88 also supported by twin studies.89, 90 Importantly, a distinction between genetic vulnerability to CO2 hypersensitivity versus trait anxiety experienced pre-CO2 inhalation was found suggesting that there are specific genetic factors associated with responsivity to stimulation via CO2 versus factors related to underlying trait anxiety.91
However, as PD does not develop in all individuals with CO2 hypersensitivity, it underscores the relevance of other risk factors for development of PD. A combination of genetic factors and early adversity such as childhood parental loss may determine hypersensitivity to CO2 and PD.92 An interesting preclinical study in cross-fostered mice pups revealed persistent expression of enhanced CO2-evoked respiratory responses in mice with a history of interference with dam–pup interactions suggestive of significant gene-by-environment effects on heightened CO2 sensitivity.93 In any case, hypersensitivity to elevated CO2 may help identify childhood groups at familial risk for subsequent development of PD94, 95 Interestingly, association of polymorphisms within the tryptophan hydroxylase-2 (TPH-2) gene and CO2 responses is observed suggestive of a role of the serotonergic (5-HT) system in the effects of CO2.96 Accumulating evidence strongly supports an association of polymorphisms in multiple markers of the 5-HT system, including polymorphisms in the gene locus and 3’ polyadenylation site of the serotonin transporter (5-HTT) with PD.97, 98 An association of 5-HT biosynthetic enzyme TPH-2 and 5-HT receptor subtypes R1 and R2 with PD has also been reported.99, 100, 101 The 5-HT system is of interest given its role in the regulation of panic-like behaviors.102 Importantly, evidence of chemosensory serotonergic neurons in the medullary raphe (see section ‘Acid chemosensory serotonergic neurons in the medullary raphe nucleus’) underscores the role of the 5-HT system in translation of pH fluctuations to panic-relevant ventilatory responses.
Lactate sensitivity on the other hand, did not show familial vulnerability.103 However, an association of polymorphism within the exon of the lactate dehydrogenase A gene was reported with CO2 sensitivity where the LDH polymorphism was a risk factor for increased CO2 responses.104 This is relevant, given the role LDH in lactate metabolism and its dependence on cell pH.
A recent study identified two single nucleotide polymorphisms within the acid sensing ion channel 1 (ASIC1) gene, ACCN2 in individuals with PD, which was associated with increased amygdala volume and hyperactivity in these subjects.105 This observation strongly supports an association of altered pH sensing within the amygdala with increased risk for PD. Collectively, all evidence support a strong genetic vulnerability component to interoceptive threats (represented by CO2 inhalation) and polymorphisms in chemosensory targets such as ASICs. However, other environmental factors may be required for the development of symptomology of PD.
Collectively, evidence from neuroimaging, challenge studies for panic provocation and genetics support that pH homeostasis and acid–base disturbances may contribute to PD at least in a large subset of patients with PD. In the sections below, we discuss pH chemosensory mechanisms and our understanding so far on their potential contributions to panic pathophysiology using preclinical models.