The Dual-Pathway Model of Amygdala Remodeling: CO2 Tolerance as the Unified Upstream Variable Governing Both Direct Chemosensory Stimulation and Sleep-Dependent Structural Repair

Locke Draft Working Model, April 2026

Abstract

This document proposes a unified mechanistic model for amygdala dysfunction and remodeling following chronic stress. The standard model attributes therapeutic amygdala changes to the direct effects of cognitive behavioral therapy (CBT) or mindfulness-based interventions on neural circuitry through top-down prefrontal inhibition. This model proposes an alternative: that chronic amygdala hyperactivation and associated structural changes (dendritic hypertrophy) are maintained by two concurrent pathways, both governed by a single upstream variable, CO₂ tolerance, and that normalization occurs when both pathways are addressed simultaneously. Pathway One is a direct chemosensory stimulation pathway in which pH transients produced by ventilatory instability activate acid-sensing ion channels (ASIC1a) on amygdala neurons, providing a continuous activation signal independent of psychological content. Pathway Two is a sleep-dependent restoration pathway in which the same ventilatory instability degrades the quality of overnight tissue-level repair processes (glymphatic clearance, neurotrophic support, oxygen delivery) that would otherwise enable dendritic pruning. The model addresses both functional hyperactivation (ASIC1a-mediated reactivity) and structural correlates (dendritic changes and volume), while noting that volume changes alone are a non-specific marker and that ASIC1a-dependent functional activation may be the more mechanistically precise outcome variable. The model explains the partial efficacy of CBT, the dose-response plateau observed in therapy, the failure of passive recovery in animal models, and the differential response to CPAP between the prefrontal cortex and amygdala. It generates specific, testable predictions and identifies CO₂ tolerance training as a direct intervention targeting the upstream variable that current therapies address only indirectly and partially.

The Problem with the Standard Model

The prevailing model of structural brain recovery following chronic stress holds that cognitive and behavioral therapies (CBT, MBSR, exposure therapy) drive amygdala changes through top-down mechanisms: prefrontal inhibition of amygdala circuits, extinction learning, and cognitive reappraisal. Under this model, thinking certain thoughts in certain ways directly causes physical tissue to remodel.

This model has two fundamental unanswered questions. First, what is the specific molecular mechanism by which a cognitive event (a thought, a reappraisal, an inhibitory signal from the prefrontal cortex) causes a physical event (dendritic retraction in the basolateral amygdala)? The literature describes activity-dependent plasticity and Hebbian learning as the framework but does not specify the cellular pathway by which prefrontal inhibition of amygdala output leads to dendritic pruning within the amygdala itself. The mechanism is assumed rather than demonstrated.

Second, if the amygdala is capable of structural remodeling (which partial recovery demonstrates it is), why does recovery plateau at partial rather than proceeding to completion? The literature describes this plateau as a biological ceiling, invoking concepts such as 'social-neural homeostasis' and epigenetic scarring. But these are descriptions of the observation, not explanations for it. No mechanism has been identified that would prevent a dendrite capable of partial retraction from fully retracting given sufficient conditions. The plateau is documented but not explained.

The model proposed here offers a candidate mechanism for both problems through a unified physiological pathway that specifies the physical processes at every step in the causal chain.

This paper distinguishes established components from the proposed synthesis. The established components are: the amygdala can detect CO₂/acidosis through ASIC1a-dependent mechanisms (demonstrated in animal models with supporting human genetic evidence); chronic stress can produce persistent BLA dendritic hypertrophy; CO₂ strongly regulates cerebral blood flow; CSF bicarbonate participates in central pH regulation; and sleep-dependent neurovascular dynamics contribute to brain restoration. The novel claim is that these mechanisms may form an integrated maintenance loop in which respiratory instability both directly stimulates amygdala chemosensory channels and degrades the sleep-dependent repair environment required for structural reversal. It should be noted that amygdala volume changes, while frequently used as an outcome measure in imaging studies, are non-specific and do not necessarily map linearly to functional activation or clinical symptom severity. Indeed, amygdala lesioning can paradoxically increase fear and anxiety responses. The model therefore emphasizes ASIC1a-dependent functional activation as the primary mechanistic target, with structural and volumetric changes treated as one possible downstream correlate rather than the central outcome variable.

The Upstream Variable: CO₂ Tolerance and Bicarbonate Buffering

Carbon dioxide dissolved in blood and cerebrospinal fluid (CSF) reacts with water to form carbonic acid, which dissociates into hydrogen ions and bicarbonate. Higher CO₂ levels produce more hydrogen ions and lower pH. Lower CO₂ levels produce fewer hydrogen ions and higher pH. The body maintains blood pH at approximately 7.4 regardless of baseline CO₂ levels through renal regulation of bicarbonate: the kidneys retain or excrete bicarbonate to compensate for the prevailing CO₂ baseline. CSF bicarbonate is actively regulated through choroid plexus mechanisms in a parallel process.

This compensation can substantially normalize steady-state pH, but it does so by changing bicarbonate availability. A person with a baseline PaCO₂ of 35 mmHg and a person with a baseline of 42 mmHg both have a resting blood pH of approximately 7.4. However, the compensation carries a critical structural consequence: the total quantity of bicarbonate available as a buffer is proportional to the baseline CO₂ level. A chronically low PaCO₂ from habitual overbreathing leads to chronic bicarbonate depletion as the kidneys excrete bicarbonate to normalize pH at the lower CO₂ setpoint.

This is particularly consequential in the cerebrospinal fluid, which lacks the robust protein buffers (hemoglobin, albumin) present in systemic blood. In CSF, bicarbonate is the dominant physiologically relevant buffer, because protein buffering is far less robust than in blood. The model proposes that when bicarbonate is depleted from the CSF through chronic hyperventilation, the fluid surrounding the brain becomes more sensitive to transient CO₂ fluctuations, increasing the likelihood of pH excursions that would be absorbed without consequence in a fully buffered system.

The size of the bicarbonate buffer is therefore partly determined by the prevailing CO₂ baseline and the compensatory acid-base state associated with it. Higher CO₂ tolerance would generally be expected to track with a more stable PaCO₂ baseline, greater CO₂ reserve, and more robust bicarbonate buffering, producing more stable pH in the face of transient CO₂ fluctuations. Lower tolerance would be expected to produce the reverse: a thinner buffer and larger pH swings from the same magnitude of CO₂ fluctuation.

Pathway One: Direct Chemosensory Stimulation of the Amygdala

The amygdala expresses acid-sensing ion channels, specifically the ASIC1a subtype, on the surface of its neurons. These are proton-gated ion channels whose activation and desensitization kinetics make them particularly responsive to sufficiently rapid extracellular acidification. In a system with repeated sharp pH drops, ASIC1a-mediated signaling would be engaged more frequently than in a system where CO₂/pH changes occur slowly and remain well buffered. A slow, smooth pH drift produces minimal ASIC1a activation because the channels desensitize before significant signaling accumulates. A rapid pH drop can produce strong activation because the change outpaces desensitization.

ASIC1a activation in the amygdala has been shown, at least in animal models, to produce fear behavior relevant to panic physiology. CO₂ inhalation can evoke fear behavior through ASIC1a-dependent amygdala mechanisms, and ASIC1a knockout animals show impaired fear conditioning; restoring ASIC1a expression selectively in the basolateral amygdala rescues the fear response. In human populations, genetic association studies provide corroboration: polymorphisms in the ACCN2 gene (the human ortholog encoding ASIC1a), particularly the rs10875995 variant, are significantly associated with panic disorder, and risk-allele carriers showed increased baseline amygdala volume and heightened amygdala reactivity on functional MRI. The amygdala is not merely a processor of cognitively mediated threat signals. It is a direct chemosensor that detects the pH signature of respiratory disturbance independently of any cognitive content.

The Stimulation Pattern Under Low CO₂ Tolerance

In a person with chronically low CO₂ tolerance (measurable through end-tidal CO₂, hypercapnic ventilatory response, or behavioral proxies such as voluntary breath-hold duration), two features converge to produce frequent pH transients in the CSF. First, the bicarbonate buffer is depleted, meaning CO₂ fluctuations translate into disproportionately large pH swings. Second, the chemoreceptors are hypersensitive and loop gain is elevated, meaning breathing is unstable and oscillatory. The respiratory control system overcorrects to small perturbations, producing repeated cycles of hyperventilation (CO₂ crash, pH spike upward) and hypoventilation or apnea (CO₂ surge, pH spike downward).

Sufficiently sharp downward pH excursions from CO₂ surges would be expected to activate ASIC1a channels on amygdala neurons. In a system with high loop gain and a thin buffer, these events would be expected to occur most often during sleep-stage transitions, unstable NREM breathing, apnea/hypopnea recovery cycles, and waking periods in which breathing slows after prior overventilation. The amygdala would receive frequent pH-triggered activation signals across the 24-hour cycle.

In many stress- and learning-related circuits, repeated activation can bias neurons toward activity-dependent structural remodeling, including dendritic growth or stabilization. The model proposes that repeated ASIC1a-mediated activation could contribute to maintenance of BLA dendritic hypertrophy, particularly when paired with impaired sleep-dependent pruning. The amygdala hypertrophy observed under chronic stress is therefore not solely a consequence of glucocorticoid toxicity or threat-related learning. It may be maintained, at least in part, by direct chemosensory stimulation from pH transients produced by respiratory instability in a buffer-depleted system. This stimulation requires no psychological content. It is a purely physical, chemical process acting on ion channels embedded in the neuronal membrane.

The Intracellular Mechanism: Calcium Influx and Structural Growth

A critical feature of the ASIC1a homomeric channel, distinguishing it from other ASIC subtypes, is its permeability to calcium (Ca²⁺) in addition to sodium. When a pH transient opens homomeric ASIC1a channels, calcium influx can engage intracellular signaling pathways relevant to synaptic plasticity, including CaMKII- and ERK-linked mechanisms. CaMKII regulates AMPA and NMDA receptor anchoring at the postsynaptic density while promoting actin filament polymerization, which can physically expand the dendritic spine. This provides a plausible molecular mechanism by which repeated ASIC1a activation could translate into measurable structural hypertrophy. Supporting evidence exists: global deletion of the Asic1a gene significantly alters dendritic spine morphology and density, establishing ASIC1a as a contributor to this form of structural plasticity.

The BDNF-ASIC1a Amplification Loop

The hypertrophic process described above may be further amplified by a positive feedback mechanism mediated by brain-derived neurotrophic factor (BDNF). BDNF signaling through TrkB receptors has been shown to increase ASIC1a surface expression via PI3K/Akt-dependent trafficking, specifically through phosphorylation of the ASIC1a protein at Serine-25, a modification that prevents channel endocytosis and increases the insertion of ASIC1a on the neuronal membrane. If repeated acidosis and ASIC1a activation also promote local neurotrophic signaling in amygdala circuits, this could create a positive feedback loop.

If this mechanism operates in the amygdala as the model proposes, chronic pH transient exposure would not only drive dendritic growth through the CaMKII/ERK pathway but would simultaneously populate the expanded dendritic surface with additional ASIC1a receptors. This would create a self-reinforcing loop: pH transients activate ASIC1a, ASIC1a activation drives BDNF release, BDNF upregulates ASIC1a surface expression, and the increased receptor density amplifies the response to subsequent pH transients. Such a mechanism would explain the remarkable persistence of amygdala hypertrophy in the Vyas paradigm and the resistance to passive recovery: the receptor population itself has been upregulated by the process that caused the growth, requiring active intervention to break the cycle rather than mere removal of the original stressor.

The Stimulation Pattern Under High CO₂ Tolerance

In a person with high CO₂ tolerance, bicarbonate buffering has normalized toward a higher-capacity state. The chemoreceptors tolerate normal CO₂ fluctuations without overcorrection. Loop gain is low and breathing during sleep is smooth and regular. The normal nocturnal rise in PaCO₂ of 3 to 8 mmHg occurs as a slow, gradual ramp that the large bicarbonate buffer absorbs without producing sharp pH transients. The ASIC1a channels on amygdala neurons remain largely quiescent because pH changes remain slow and well buffered.

During waking hours, CO₂ fluctuations from normal activity variations (sitting, standing, light exertion, relaxation) are similarly absorbed by the intact buffer. The amygdala is activated by genuine emotional and cognitive events through normal neural pathways, but it is not receiving the continuous background bombardment of pH-triggered activation that characterizes the low-tolerance state. The signal-to-noise ratio is improved: amygdala activation is less contaminated by recurrent internal chemosensory noise and more tightly coupled to ordinary affective, interoceptive, and environmental inputs.

Pathway Two: Sleep-Dependent Tissue Restoration

The structural remodeling of neural tissue (dendritic pruning, synaptic reorganization, waste clearance) is a physical process requiring specific biochemical conditions that are established primarily during sleep. Three processes are directly relevant to amygdala remodeling:

Glymphatic clearance: sleep-dependent neurovascular dynamics, including synchronized changes in cerebral blood volume, vasomotion, and CSF movement, drive the clearance of metabolic waste products from brain tissue. The model proposes that unstable nocturnal CO₂ physiology could degrade these vascular-fluid dynamics and thereby reduce the efficiency of overnight waste clearance.

Neurotrophic and metabolic support: sleep-dependent structural remodeling likely depends on a constellation of conditions including BDNF-related plasticity pathways, adequate perfusion, oxygen availability, and waste clearance. The model does not require any single mediator; rather, unstable respiratory physiology may degrade the broader plasticity-supportive environment that these processes collectively provide.

Oxygen delivery: low CO₂ can impair brain oxygen availability through two converging mechanisms: cerebral vasoconstriction and a leftward shift in hemoglobin oxygen affinity under alkalotic conditions. Conversely, stable tolerated CO₂ during sleep may support tissue oxygen delivery by preserving cerebral perfusion and facilitating oxygen unloading.

Restoration Quality Is Governed by CO₂ Tolerance

All three restoration processes are modulated by the same respiratory variables that govern Pathway One. Cerebral blood flow is highly CO₂-sensitive over normal physiological ranges, with human studies estimating approximately 3 to 4 percent increase in flow per 1 mmHg increase in PaCO₂ during hypercapnia and roughly 2 to 3 percent decrease per 1 mmHg during hypocapnia. A person with a higher tolerated nocturnal PaCO₂ has wider cerebral vasodilation, stronger arterial pulsations, more effective glymphatic flow, and better tissue oxygenation throughout the night. A person with lower tolerance has narrower vessels, weaker pulsations, reduced glymphatic clearance, and impaired oxygenation.

Critically, these variables are not captured by standard polysomnography. Two individuals can have identical sleep staging, identical continuity, identical total sleep time, and dramatically different tissue-level restoration quality based on their CO₂ tolerance. Standard sleep measurement does not detect the sub-architectural quality variable that determines how much physical remodeling work each hour of sleep actually accomplishes.

The Competing Signal Problem

For the amygdala specifically, the restoration process faces a challenge that does not apply to the prefrontal cortex or hippocampus. During sleep in a low-tolerance system, the same ventilatory instability that degrades restoration quality is simultaneously activating ASIC1a channels on amygdala neurons. The overnight repair machinery is attempting to prune dendrites on neurons that are being actively stimulated by pH transients from the very breathing instability that impairs the repair process. The repair environment is degraded and the damage signal is active, simultaneously, during the window when repair should be occurring.

Although ASIC1a is expressed in multiple brain regions, the amygdala has a uniquely demonstrated functional role in converting CO₂/acidosis signals into fear and threat activation. The model therefore does not depend on

ASIC1a being absent from the PFC or hippocampus. It depends on the amygdala being especially positioned, functionally, to convert repeated internal pH perturbations into threat-related activation signals. PFC and hippocampal remodeling would therefore depend more heavily on the quality of the restoration environment (Pathway Two) without the same competing chemosensory stimulation signal (Pathway One). This is why these structures may recover more readily under partial sleep improvement: they need the repair environment to improve. The amygdala needs both the repair environment to improve and the competing stimulation signal to stop.

The Unified Mechanism

Both pathways converge on a single upstream variable: CO₂ tolerance, operationalized as the chemoreceptor setpoint and its downstream consequences for bicarbonate buffering and ventilatory stability.

Chronic amygdala hyperactivation and associated structural hypertrophy are maintained by the simultaneous operation of two processes. Pathway One: pH transients from ventilatory instability in a buffer-depleted system directly stimulate amygdala neurons through ASIC1a channels, providing a continuous activation and growth signal. Pathway Two: the same ventilatory instability degrades the overnight tissue-level restoration environment, impairing the physical processes that would otherwise enable dendritic pruning.

Amygdala normalization (reduced ASIC1a-mediated functional hyperactivation and, potentially, associated structural changes including dendritic retraction) occurs when both pathways are addressed: the direct stimulation signal is reduced and the restoration environment is optimized. Because both pathways are downstream of the same upstream variable, a single intervention targeting CO₂ tolerance addresses both simultaneously. Recalibrating chemoreceptor sensitivity upward raises baseline PaCO₂, which rebuilds the bicarbonate buffer (reducing ASIC1a stimulation), widens the CO₂ reserve and lowers loop gain (stabilizing ventilation during sleep), and improves cerebral blood flow, glymphatic clearance, and tissue oxygenation (optimizing the repair environment).

Explanatory Applications

Why CBT Produces Partial Amygdala Recovery

CBT reduces rumination and chronic threat-vigilance, which lowers tonic sympathetic activation. The calmer autonomic state produces a modest reduction in habitual overbreathing and a modest rise in baseline PaCO₂. The kidneys and choroid plexus respond by retaining slightly more bicarbonate. The CSF buffer partially rebuilds. pH transients from CO₂ fluctuations are partially blunted. ASIC1a-mediated amygdala stimulation decreases. Simultaneously, the modest improvement in breathing stability produces modest improvement in overnight restoration quality. The combined partial reduction in Pathway One stimulation and partial improvement in Pathway Two restoration enables partial functional and structural normalization.

Under this model, the specific cognitive content of CBT is not the primary structural mediator. Cognitive interventions may still matter clinically, but their contribution to amygdala remodeling would be mediated largely through downstream autonomic, respiratory, and sleep-related variables. Any intervention producing equivalent autonomic calming (exercise, MBSR, breathing techniques, anxiolytic medication) would produce equivalent effects on CO₂ dynamics and equivalent structural outcomes, because all converge on the same downstream respiratory pathway. This prediction is consistent with the empirical finding that diverse therapeutic modalities with entirely different cognitive content produce comparable structural remodeling magnitudes.

Why the CBT Plateau Exists

CBT's effect on breathing patterns is indirect and modest. It can calm the system enough to reduce minute ventilation from crisis levels, but it cannot push baseline PaCO₂ beyond the point where a modestly calmer nervous system naturally settles. The chemoreceptors are never directly challenged. The bicarbonate buffer rebuilds to match the modestly improved PaCO₂ and stabilizes. Further therapy sessions produce no additional change in breathing dynamics, no further bicarbonate retention, no further reduction in pH transients, and no further improvement in overnight restoration quality. The amygdala stops remodeling because both pathways have stopped improving.

The plateau is not a biological ceiling on tissue recovery capacity. It is a ceiling on the intervention's ability to move the upstream variable. The tissue remains fully capable of further remodeling if conditions improve further. The intervention has simply exhausted its indirect leverage on the respiratory pathway it does not know it is targeting.

Why Passive Recovery Fails for the Amygdala (Vyas et al.)

In the Vyas passive recovery paradigm, the external stressor was removed and animals were housed in stress-free conditions for 21 days. Hippocampal dendritic atrophy reversed. Amygdala hypertrophy did not. Respiratory variables were not measured during the recovery period.

Under this model, the hippocampus recovered because removing the stressor reduced cortisol and improved the general restoration environment sufficiently to cross the hippocampus's lower threshold for passive remodeling. The hippocampus does not face the competing ASIC1a stimulation signal that the amygdala does.

Under this model, the failure of amygdala recovery would be explained if the animals maintained an altered respiratory/chemosensory baseline during the recovery period. Because respiratory variables were not measured, the Vyas paradigm leaves open the possibility that removal of the external stressor normalized some restorative variables while leaving a persistent interoceptive driver of BLA activation intact. Chronic stress recalibrates chemoreceptor sensitivity downward through the bicarbonate depletion mechanism. Removing the stressor removes the active cortisol and orexin driving the overbreathing, but would not necessarily reverse the chemoreceptor recalibration or restore the depleted bicarbonate buffer. The pH transients would continue. The ASIC1a stimulation would continue. The amygdala would continue to receive a direct chemical growth signal despite the absence of any external threat. Simultaneously, sub-optimal ventilatory stability would continue to degrade overnight restoration quality, impairing the repair processes that would have enabled dendritic pruning.

Why CPAP May Recover Some Regions but Not the Amygdala

CPAP resolves obstructive airway events, eliminating acute hypoxic episodes and sympathetic surges associated with obstructive sleep apnea. A longitudinal VBM study by Canessa et al. (2011) examined 17 treatment-naive men with severe OSA before and after three months of compliant CPAP therapy. The study found robust, statistically significant gray matter volume increases in the prefrontal cortex, superior frontal gyrus, and multiple hippocampal subdivisions, with corresponding reversal of neuropsychological deficits.

However, the hippocampus-amygdala transition area (HATA), an amygdala-adjacent medial temporal region, showed no statistically significant volume recovery (left HATA p=0.13; right HATA p=0.11) following the same three months of therapy.

However, CPAP does not directly target chemoreceptor sensitivity, CO₂ tolerance, or loop gain, and in some patients may leave central ventilatory instability partly unresolved. Sub-threshold respiratory oscillations may persist. The bicarbonate buffer remains at whatever level the patient's baseline CO₂ tolerance dictates.

The model predicts that residual amygdala abnormalities, where present after CPAP, may reflect persistence of respiratory control instability or pH/ASIC1a signaling not fully addressed by airway stabilization alone. CPAP addresses Pathway Two (restoration quality) but may leave Pathway One (ASIC1a stimulation from pH transients) untouched. Patients with persistent amygdala abnormalities after CPAP would be expected to show residual nocturnal CO₂ variability, elevated loop gain, or abnormal CO₂ chemosensitivity.

Why Full Responders Differ from Partial Responders

Autonomic variables such as heart rate variability and vagal tone have been proposed and studied as predictors or mediators of treatment response in stress-related disorders. Under this model, such measures would be expected to outperform therapy-hour counts when the outcome of interest is amygdala functional normalization. HRV serves as a proxy for the state of the autonomic and respiratory system. Higher HRV indicates better parasympathetic function, which correlates with more stable breathing patterns, higher baseline CO₂ tolerance, more complete bicarbonate buffering, fewer pH transients, less ASIC1a stimulation, and better overnight restoration quality. Full responders are individuals whose autonomic and respiratory physiology reaches a threshold sufficient to address both pathways. Partial responders are individuals whose physiology improves but does not cross that threshold, leaving residual Pathway One stimulation and suboptimal Pathway Two restoration.

The Integrated Causal Chain

The complete causal sequence, from chronic stress to amygdala hyperactivation to recovery, is as follows:

Critical Bridge Claims

The model depends on several bridge claims that have not yet been directly tested as an integrated pathway. These are not assumed to be proven; they are the core empirical targets generated by the model:

Bridge Claim 1: Chronic stress or hyperarousal can produce persistent reductions in CO₂ tolerance and increased nocturnal CO₂ variability in susceptible individuals.

Bridge Claim 2: These respiratory changes are sufficient to alter CSF/extracellular pH dynamics in regions relevant to amygdala ASIC1a signaling.

Bridge Claim 3: Repeated pH transients can produce enough ASIC1a-mediated activation to maintain BLA dendritic hypertrophy or interfere with pruning.

Bridge Claim 4: Respiratory stabilization, through direct CO₂ tolerance training, can reduce this chemosensory signal while simultaneously improving the sleep-dependent neurovascular environment required for structural remodeling.

Bridge Claim 5: Repeated ASIC1a activation in amygdala neurons engages Ca²⁺/CaMKII/ERK and BDNF/TrkB/PI3K/Akt-related mechanisms strongly enough, in vivo, to stabilize or expand BLA dendritic structure under chronic stress conditions.

Testable Predictions

The model generates the following specific, falsifiable predictions:

Prediction 1: Individuals with higher baseline CO2 tolerance (measured by end-tidal CO2, hypercapnic ventilatory response, or behavioral proxies such as breath-hold duration) should show faster and more complete normalization of amygdala ASIC1a-dependent reactivity, and potentially associated volume changes, following any therapeutic intervention, independent of the type of therapy received.

Prediction 2: A CO2 tolerance training protocol added to standard CBT should produce functional amygdala normalization beyond the plateau observed with CBT alone, because the direct intervention on chemoreceptor sensitivity pushes the upstream variable past the ceiling that CBT's indirect effect cannot reach.

Prediction 3: In the Vyas passive recovery paradigm, if animals were subjected to a CO2 tolerance protocol (controlled intermittent hypercapnic exposure) during the recovery period, the amygdala hypertrophy should reverse, because the direct chemosensory stimulation pathway would be addressed alongside the passive restoration improvement.

Prediction 4: CPAP patients with residual amygdala abnormalities who add CO2 tolerance training to their treatment should show amygdala functional and potentially structural normalization, because the intervention addresses Pathway One (ASIC1a stimulation) that CPAP leaves untouched.

Prediction 5: Continuous capnography during sleep should reveal that patients whose amygdalas fail to recover have greater PaCO2 variability (larger and more frequent CO2 fluctuations) than patients whose amygdalas do recover, independent of sleep staging or total sleep time.

Prediction 6: CSF bicarbonate concentration should be measurably lower in chronic stress patients with persistent amygdala hyperactivation compared to recovered patients, reflecting the buffer depletion that enables pH transients to reach ASIC1a channels.

Prediction 7: Pharmacological or experimental blockade of ASIC1a channels (such as amiloride or PcTx1) would be predicted to accelerate amygdala functional normalization in chronic stress patients even without improvement in sleep quality or CO2 tolerance, because Pathway One would be blocked pharmacologically while Pathway Two continues to operate at whatever level the patient's sleep provides.

Prediction 8: The rate of emotional reactivity reduction in recovering patients should correlate with measured CO2 tolerance progression rather than with hours of cognitive therapy received, because CO2 tolerance is the upstream variable governing both pathways while therapy is an indirect and partial input.

Relationship to Existing Therapeutic Interventions

This model does not claim that CBT, MBSR, exercise, or pharmacological interventions are ineffective. They demonstrably produce measurable improvement. The model reframes the mechanism of their efficacy: they work not through their specific cognitive or behavioral content but through their indirect, partial effect on the autonomic and respiratory variables that govern the two pathways maintaining amygdala hyperactivation and hypertrophy.

Under this reframing, these interventions are indirect, low-efficiency approaches to a physiological target that can be addressed directly. CO₂ tolerance training, through progressive breath-hold protocols and nasal breathing discipline, is proposed to recalibrate chemoreceptor responsiveness, which directly rebuilds the bicarbonate buffer, which directly reduces the ASIC1a stimulation signal, while simultaneously directly optimizing overnight restoration quality. Every step is a specified physical mechanism acting on identified biological substrates.

The practical implication is not that CBT should be abandoned. For many patients, the autonomic calming produced by CBT may be a necessary precondition for engaging with CO₂ tolerance training, particularly if acute anxiety makes breath-hold work intolerable. The implication is that CO₂ tolerance training should be recognized as a direct intervention targeting the upstream variable that determines recovery rate and completeness, and that its addition to existing therapeutic protocols would be expected to push recovery past the plateau that current interventions reach alone.

Limitations

This model is derived from the integration of established physiology across multiple domains (respiratory control, acid-base chemistry, ion channel biophysics, sleep neuroscience, structural neuroplasticity) with observational data from an N-of-1 recovery account. Direct measurement of ASIC1a activation in the human amygdala during sleep is not currently feasible with non-invasive methods. CSF bicarbonate concentration is not routinely measured in stress recovery studies. The causal chain proposed here is mechanistically specified at every step but has not been validated as an integrated pathway in a controlled experimental setting.

Additionally, amygdala volume, which features in several of the model's predictions as a measurable outcome, is recognized as a non-specific marker. Amygdala volume changes do not map linearly to functional activation or clinical symptom severity, and amygdala lesioning can paradoxically increase fear responses. Future testing of this model should prioritize functional outcome measures (such as ASIC1a-dependent amygdala reactivity on fMRI, or CO₂ challenge-evoked amygdala activation) alongside or in place of volumetric measures where feasible.

The model's primary strength is its explanatory power: it offers candidate mechanistic answers to questions that the standard model leaves unanswered (the mechanism of cognitive-to-structural causation, the reason for the recovery plateau, the Vyas passive recovery failure, the CPAP amygdala differential). Its primary limitation is the absence of direct experimental confirmation of the integrated dual-pathway mechanism operating as described. The testable predictions generated in section “Testable Predictions” provide a roadmap for such confirmation.

Sources and related reading

This manuscript does not contain a standalone numbered bibliography. Follow the evidence library and the related core manuscripts for source-linked background.