Neuro-PASC: symptoms, challenges, and solutions for viral persistence in the brain

The classification of Neuro-post-acute sequelae of SARS-CoV-2 infection (PASC) describes a wide range of neurological manifestations observed in individuals with PASC, or “long COVID” [1]. Cognitive impairment including brain fog and frontal network dysfunctions involving deficits in attention, memory, and executive function, can persist beyond one year in recovered individuals regardless of infection severity or hospitalization status [2–11]. Optical clearing and imaging studies observed the persistent presence and accumulation of the SARS-CoV-2 spike (S) and nucleocapsid (N) proteins at the skull-meninges-brain axis of postmortem skulls from patients who died from COVID-19. Whereas the S protein from SARS-CoV-2 was detected in 10 out of 34 postmortem skull samples obtained from patients who did not die of COVID-19 causes. The fact that skull marrow cells, meningeal cells, and brain tissue cells all tested positive for the S protein at rates of 2.3%, 2.8%, and 1%, respectively, implies that viral persistence is maintained in many individuals long after the clearance and resolution of infection. Similarly, healthy mice aerosol-inoculated with SARS-CoV-2 (Omicron B, XBB sub-lineage) virus exhibited clear persistence of the S protein in the head at 28 days post-infection after full recovery from acute symptoms and viral clearance from the skull marrow and brain cortex [12].

A prospective longitudinal cohort study of 475 adults (mean age 58.26 years) hospitalized with a clinical diagnosis of COVID-19 in the UK reported that participants experienced not only the emergence of new psychiatric and cognitive symptoms over the first 2–3 years but also the continued deterioration of existing symptoms present at 6 months post-hospitalization [13]. In 2023, an ambispective cohort study evaluated post-hospitalization Neuro-PASC (PNP, n = 50) and non-hospitalized Neuro-PASC (NNP, n = 50) patients in Medellin, Columbia diagnosed with COVID-19 between 2020 and 2023. The analysis of the data collected revealed that 64% of PNP, compared to 42% of NNP presented abnormal neurological symptoms that included brain fog (60%) and myalgia (42%). While numbness/tingling and abnormal body sensations were more prevalent in PNP (58% and 42%, respectively) than in NNP (24% and 12%, respectively). Conversely, brain fog was more common in NNP (64%) than PNP (56%). Both groups presented identical levels of fatigue (74%) and sleep dysregulation (46%) during examinations [14]. In November 2024, a large cross-sectional study involving 200 PNP and 1100 NNP patients reported that younger (18–44 years) and middle-aged (45–64 years) subjects, regardless of infection severity, were disproportionately affected by neurological symptoms compared to older (65+ years) subjects [15].

The potential mechanisms responsible for neurological and even structural alterations of the brain by the SARS-CoV-2 virus [16–21], remain to be fully elucidated. Nonetheless, an exhaustive list of reviews, hypotheses, and experimental work examining mechanisms including, but not limited to, neuroinflammation [22–25], mitochondrial quality control [26, 27], mitochondrial DNA content [28, 29], and endoplasmic reticulum stress [30] have been presented as causes responsible for the activation of Neuro-PASC [31–37]. Overwhelming evidence supports the existence of a strong, positive correlation between SARS-CoV-2 infection and neurological dysfunction [38–43]. A systematic analysis of resting-state electroencephalogram (rsEEG) in Long COVID patients revealed striking abnormalities resembling those exhibited in early developmental stages of neurodegenerative diseases (NDDs) including Alzheimer’s disease (AD) and related dementia (ADRD). The shift to lower frequencies in the rsEEG (reduced alpha rhythm, increased slow waves) and epileptiform-like EEG signals imply potential overlapping pathophysiology in both Neuro-PASC and ADRD [44]. Although a systematic review comprising 18 studies involving 412957 COVID-19 patients aged ≥ 65 years reported the identification of new-onset cognitive impairment in 64% of the patients [45], viral infection-induced cognitive impairment may not be unique to the SARS-CoV-2 virus, as recent advances link viral amyloidogenesis to neurodegeneration [46].

A dominant association exists between viral infections and NDDs

Many viral infections are associated with elevated risks in the development of NDD. A thorough data analysis of cohorts (> 800000) in the FinnGen and the UK Biobank was conducted to identify longitudinal and cross-sectional associations between viral exposures and common NDDs including AD, amyotrophic lateral sclerosis (ALS), dementia, Parkinson’s disease (PD), and multiple sclerosis (MS). The study reported 45 viruses, including viral encephalitis, viral hepatitis, influenza, pneumonia, herpes simplex, and varicella-zoster virus are associated with an elevated risk of NDD up to 15 years post-infection [47]. An analysis of transcriptomic and interactomic profiles of brain samples from the Gene Expression Omnibus repository of AD patients who died of COVID-19, AD without COVID-19, COVID-19 without AD, and control individuals found 21 interactome key nodes are dysregulated in AD patients who died of COVID-19, with the implication that SARS-CoV-2 infection exacerbated the AD condition due to the presence of higher levels of amyloid-beta (Aβ), inflammation, and oxidative stress [48]. Whether the SARS-CoV-2 virus features characteristics distinct from other viruses and exacerbates NDD is unclear and requires elucidation.

The development of Neuro-PASC is not specific to age or infection severity

Although age-related hippocampal atrophy and neuroinflammation can augment cognitive impairment [49], the development of Neuro-PASC is observed across all ages and health groups. A systematic review and meta-analysis of 11 studies involving > 6.8 million controls and 0.94 million infected subjects reported a higher risk in the development of new-onset dementia in adults aged > 60 years at different intervals post COVID-19 diagnosis [50]. Whereas, a longitudinal multimodal MRI study in Lombardy Italy, investigating the neural activities in mildly-infected, healthy young adults (mean age 24 years) 12 months post-infection, discovered persistent structural and functional changes in specific brain areas resulting in deficits in olfaction, visual/verbal domain of short-term memory, and reduction in spatial working memory compared to uninfected controls [51]. Imaging evidence from ultra-high field 7 T MRI of COVID-19 survivors (median time of 6.5 months post-hospitalization), showed magnetic resonance susceptibility abnormalities at multiple regions of the brainstem, including the medulla oblongata, pons, and midbrain [52].

A registered clinical trial (NCT04865237), conducted in the United Kingdom between March 2021 and July 2022, examined the changes in cognitive functions in 34 healthy, seronegative young adults aged 18–30 years, inoculated with Wildtype SARS-CoV-2. The study reported that subjects with mild infections exhibited statistically lower cognitive scores that persist for one year or longer, compared to uninfected test subjects [53]. SARS-CoV-2 intranasally inoculated ferrets demonstrated subclinical to mild respiratory disease. Even though there is no evidence of brain infection detected via in situ hybridization or immunohistochemistry, the presence of viral RNA in the brain, 7 days post infection, induced neurovirulence associated with microglial activation, astrocytic deactivation, and Alzheimer type II astrocytes [54]. Additionally, the discovery of substantial abnormal neuroimaging findings in a significant number of young children with neurological symptoms post COVID-19 infection [55] all point to the potential involvement of multifaceted, universal molecular mechanisms associated with phase separation and amyloidogenesis in the development of Neuro-PASC that may affect infected/exposed subjects of all ages and health conditions.

Melatonin as a cost-effective, time-tested solution for the potential treatment of Neuro-PASC

Melatonin (N-acetyl-5-methoxytryptamine) is an evolutionarily conserved molecule [56] that can effectively counter the development of Neuro-PASC by rebalancing the complex energetic scales of amyloidogenesis promoted by the SARS-CoV-2 virus. The impressive, extensive array of functions and features associated with melatonin since its isolation and identification in bovine pineal tissues in 1958 [57] provide convincing rationale for the ubiquitous presence of this ancient molecule, synthesized in all tested eukarya, bacteria, and archaea [58–60]. Nevertheless, proposals presenting melatonin as an effective regulator of phase separation under various contexts including neurodegenerative disorders [61], cancer multidrug resistance [62], viral phase separation [63], dementia [64], cataractogenesis [65], and mitochondrial function [66], offer a novel perspective in the investigation of the underexplored, expansive functions of the indoleamine for more than 2 billion years in Archaea and Bacteria prior to the advent of Eukarya [66].

The first report of intracellular amyloids detected in the domain of Archaea stratifies an additional layer of significance in the synergy between melatonin and adenosine triphosphate (ATP) in the regulation of phase separation and amyloidogenesis that began close to 4 billion years ago (Ga). This review will apply current knowledge to explore potential molecular mechanisms employed by melatonin in synergy with ATP for the regulation of phase separation that can modulate pathological amyloidogenesis promoted by SARS-CoV-2 in Neuro-PASC (Figure 1).

Display full size

Figure 1. 

Schematic overview of the synergistic regulation of amyloid solubilization and clearance during slow wave sleep in Neuro-post-acute sequelae of SARS-CoV-2 infection (PASC). (A) During viral infection/persistence, exogenous melatonin enhances the production of adenosine triphosphate (ATP) and synthesis of mitochondrial melatonin, preventing SARS-CoV-2 inhibition of mitochondrial bioenergetics and dynamics; (B) melatonin augments the adenosine moiety effect that promotes the solubilizing effect of ATP to regulate protein phase transitions during aggregation. The hydrating effect of melatonin—facilitated by the indole π clouds and hydrogen bonding with water molecules—combined with ATP, prevents hydrophobic collapse and delays the transitioning into β-sheets during protein folding; (C) the SARS-CoV-2 virus and its structural, non-structural, and accessory proteins facilitate and accelerate the nucleation of physiological amyloid fibrils that have additional antimicrobial protective function into pathological amyloid aggregates. The SARS-CoV-2 virus further exacerbates the accumulation of insoluble amyloid aggregates by diverting nutrients away from mitochondrial oxidative phosphorylation (OXPHOS) ATP production into anabolic glycolytic pathways that support viral replication and maintenance. The suppression of fusion and fission creates dysfunctional hollow mitochondria devoid of the ATP synthase and elongated mitochondria. Unable to differentiate into functional subpopulations under nutrient deficiency, mitochondrial dysfunction further elevates imbalances between energy production and the synthesis of macromolecular precursors; (D) the optimal clearance of soluble amyloids during slow wave sleep by the glymphatic system must be supported by the adequate presence of both melatonin and ATP. The SARS-CoV-2 virus during infection and persistence prevents the effective brain clearance of excess insoluble amyloid aggregates due to deficient ATP and melatonin as a result of mitochondrial dysfunction. Created in BioRender. Loh, D. (2025) https://BioRender.com/b31s634

The ~4 Ga relationship between amyloids, phase separation, and viruses

Since its first proposal in 1991 [67], the amyloid hypothesis remains controversial. Notwithstanding the copious experimental evidence correlating the formation of Aβ with cytotoxic effects that may advance the development of NDDs [68–71], the detection of Aβ in cognitively normal individuals presents unresolved questions on the nature and function of Aβ oligomeric assemblies that inevitably increase with the aging process [72–77]. Recent advances in the antimicrobial functions of Aβ that may protect the brain against viral infections begin to reveal the multilayered complexity of amyloidogenesis [78]. Bacteria including Escherichia coli produce protective biofilms strengthened by amyloid fibrils known as curli [79]. Yet curli formation triggers the enhanced production of cytotoxic host Aβ oligomers that suppressed curli synthesis and inhibited E. coli biofilm density and elevated susceptibility to gentamicin antibiotic treatment [80]. The results of this explicit experiment open an unexpected door revealing the ancient, complex roles of amyloids in all three major domains of life where the neurodegenerative effects of aberrant Aβ aggregation may be the result of protective antimicrobial activity [81]. Furthermore, amyloids confer advantageous adaptive fitness via the inheritance of proteinaceous epigenetic memory [82].

The discovery of amyloid-forming, prion-like domains in Archaea consolidates their functional, physiological existence not only in all three major domains of life but also in the last universal common ancestor (LUCA) [83, 84]. Accordingly, amyloids may have played a major role in the origin of life, increasing the stability of RNA in extreme, fluctuating environments by acting as templates and catalysis for chemical reactions that favor replication [85]. While the exact origin of viruses remains enigmatic [86], and the debate on whether giant viruses constitute the fourth domain of life continues [87], the early cellular origin of viruses sharing a common evolutionary path with Archaea and Bacteria since their inception is, nonetheless, supported by phylogenomic data [88] as well as direct and indirect inferences from paleovirology [89]. All evidence examined to date appears to suggest that viruses are ancient and may even predate LUCA [90–92]. Upon invasion by viruses, ancient unicellular living organisms, especially archaea, inevitably formed cytotoxic amyloids as their only course of effective protection [46, 93–95].

Due to the nature of viral replication that is dependent upon phase separation, viral invasions of early organisms can readily transform functional proteins in their soluble monomeric states into insoluble, crystalline, Aβ-sheet assemblies [96]. Viruses facilitate the heterogeneous nucleation and phase transition of amyloids via direct physicochemical mechanisms that act as catalysts, promoting and accelerating the assembly of Aβ-sheets [97, 98]. Despite being potentially protective in nature, the cytotoxicity of amyloid oligomers is a conundrum requiring reconciliation and clarification [99–102].

Phase-separated biomolecular condensates promote amyloidogenesis

Phase separation is an intrinsic feature of nature used by all tested living organisms in the three major domains of life to organize and sustain fundamental biological processes. Spontaneous, nonequilibrium thermodynamic conditions governing the competition between entropy and enthalpy, drive intracellular phase separation in the assembly of membraneless organelles, also known as biomolecular condensates (BCs) [103–105]. BCs are dynamic, micron-scale, fluid compartments that can rapidly organize/reorganize cellular biochemistry to accelerate, slow, promote, or inhibit cellular functions and reactions, by either increasing, decreasing, including, or excluding reactants, enzymes, and substrates according to the fluctuating requirements of the cell [106]. Recent advances link the formation of amyloid fibrils to phase separation of condensates [107], where the interface of condensates promotes the accelerated formation of amyloid fibrils [108]. Viruses employ molecular sorting driven by phase separation and the exploitation of the host phase separation machinery to enable viral replication and persistence [63, 109–113]. Fluctuations in the energy landscape of aggregation pathways occurring at BC interfaces can subsequently become active sites for the heterogeneous nucleation of pathological amyloid fibrils promoted by viral phase separation [114–116].

BCs have been observed to both promote and inhibit the self-assembly of fibrillar proteins [117–120], which is a process commonly associated with NDDs including AD, PD, as well as prion diseases. The self-assembly of functional amyloids, dependent upon intrinsically disordered proteins (IDPs), may have played decisive roles in the origin of life and the evolution of species [121–123]. The formation of amyloid fibrils rich in insoluble β-sheets networks, stabilized by hydrogen (H) bonds acting as chemical adhesives [124], is intrinsically a phase transition of proteins from an amorphous liquid state into a solid, insoluble state via phase separation, nucleation, and growth processes [125–127]. The addition of phase-separated BCs near natively folded protein assemblies can trigger the abrupt acceleration of protein aggregation in the self-assembly of amyloid aggregates [117, 128]. Nevertheless, protein folding—whether self-assembling into stable, highly-ordered, β-sheets or physiological, well-defined, functional structures—is a common feature exhibited by, potentially, all polypeptide sequences, regardless of their native states or localization [129–133]. Perturbation in the energy landscape can directly affect the outcome of protein folding and aggregation. The transitioning into β-sheets during protein folding involves hydrophobic collapse, formation of H-bonds, and intermolecular electrostatic interactions that lower the Gibbs free energy but increase solvent entropy [134, 135].

SARS-CoV-2 escalates the self-assembly of crystalline Aβ-sheets

Viruses can accelerate the phase transition of amorphous amyloids into stable β-sheet aggregates. Acting as chemical catalysts, viruses modulate hydrophobic collapse and the nucleation/condensation of proteins, potentially by altering the activation barrier (ΔG#) for protein folding via a wide array of molecular mechanisms—including phase separation—that will be addressed, as allowed by scope, in this review [97, 98, 134]. Accordingly, abundant in vitro studies report that different structural, non-structural, and accessory proteins of the SAR-CoV-2 virus from wildtype and prominent variants readily form amyloid aggregates that are commonly associated with NDDs. Even SARS-CoV-2 nucleic acids that have been deactivated by UV radiation are able to induce amyloid aggregation in vitro, by directly acting as physicochemical mechanistic catalysts that promote amyloid aggregation via heterogeneous nucleation and phase transition that do not involve templating. The aggregation of disease-associated proteins is often seeded by BCs containing proteins with unfavorable molecular interactions [117]. The enhanced cytotoxicity of ground-state amyloid crystals assembled from SARS-CoV-2 accessory open reading frame 6 (ORF6) protein can increase the level of neuronal apoptosis compared to higher free energy, non-crystalline amyloid assemblies [136–153]. Please see Table 1 for individual study descriptions and corresponding observed effects. The formation of β-sheet aggregates as a result of viral infection is possibly an evolutionarily conserved response in eukaryotes and archaea. As such, it is not unreasonable to expect living organisms to be able to effectively manage the cytotoxic effects via the maintenance of proteostasis.

It is almost implausible that an intrinsic, fundamental flaw in the design of the molecular chaperone disaggregase system is responsible for the increase in neurological dysfunctions associated with viral infections in humans, especially when considering the alarmingly high prevalence of Neuro-PASC in infected/exposed individuals. While the emergence of amyloids may have begun as early as LUCA [84], the existence of a proteostasis-defending heat shock response system that arose around the same time implies that this relationship is not only evolutionarily conserved but should also be robust and effective [154]. Yet this dynamic relationship is dependent upon an intricate balance between proteostasis and energy homeostasis [155]. Effective induction of stress-induced chaperones must be supported by the reduction of other housekeeping proteins to manage proteotoxic stress. Aberrant phase separation prevents the requisite assembly of molecular condensates that exert gain-of-function to amplify chaperone synthesis, and loss-of-function to repress expendable cellular functions. Thus, the ineffective shut-down of housekeeping processes during chaperone induction due to macromolecular assembly defects often results in less-than-optimal performance [156]. The successful assembly of BCs is ultimately determined by thermodynamic principles that govern phase separation, and the presence or absence of ATP.

The glymphatic system and the clearance of soluble amyloids in Neuro-PASC

In neurodegenerative disorders, proteostasis failure can be a vicious cycle where reduced protein degradation may be directly responsible for the ineffective clearance of amyloid aggregates [176, 183, 184]. Deficient proteostasis allows the transition of amyloid fibrils from soluble, amorphous states into insoluble, crystalline β-sheet aggregates [185], via breaching solubility limits coupled with a breakdown of supersaturation—promoted by defective clearance—where the solute concentration overwhelms its thermodynamic solubility [186]. Hence, maintaining the solubility of proteins could be one of the most important deciding factors that can ensure the effective clearance and efflux of amyloids from the brain via the glymphatic system to prevent amyloid crystallization [187]. Higher levels of soluble Aβ42 in cerebrospinal fluid (CSF) is associated with better neuropsychological function and increased volume in the hippocampus [188].

The efflux of soluble amyloid fibrils in the brain is facilitated by various clearance mechanisms including the blood brain barrier, blood-CSF barrier, the glymphatic drainage system, and intramural periarterial drainage [189]. Since the twin discoveries of meningeal lymphatics and the glial-lymphatic (glymphatic) system, increased understanding of how the movement of CSF is responsible not only for the delivery of nutrients but also for the clearance of brain waste [190, 191] begins to emerge. The dysfunction of the glymphatic system and meningeal lymphatics [192] is associated with defective clearance of metabolic waste in the CSF and the subsequent deposition of insoluble amyloid crystals resulting in the progression of NDDs [183, 193–196]. The discovery of the extension of the CSF flowing contiguously and continuously from the central nervous system to distal ends of peripheral nerves expands the potential reach and effect of the glymphatic system in the clearance of waste and the delivery of nutrients in health and disease [197]. However, the extent to which glymphatic-lymphatic brain clearance contributes to total brain clearance is currently unknown, and whether yet unidentified direct/indirect drainage routes exist, awaits elucidation [198, 199].

Sleep deprivation impairs glymphatic clearance of Aβ biomarkers

A novel, cross-sectional and longitudinal study, employing the analysis along the perivascular space (ALPS) index, investigated the relationship between whole-brain glymphatic activity and clinical/pathological features in participants with AD dementia (n = 47), mild cognitive impairment (n = 137), and healthy controls (n = 235). Although the ALPS index reflects only global glymphatic activity that may not identify regional glymphatic dysfunction, the results of the study revealed that glymphatic dysfunction precedes amyloid pathology including increased Aβ PET burden [200]. While Aβ42 in the CSF breaching the positive threshold often predicts brain atrophy and cognitive decline [201]. A multi-site randomized crossover clinical trial measured the changes in the plasma levels of Aβ42/Aβ40 and tau peptides from evening to morning in 39 healthy participants (between 50–65 years old) under in-laboratory conditions of both normal and deprived sleep conditions. Results of the study showed sleep-active elements—including higher EEG Delta power and lower brain parenchymal resistance—enhanced the overnight glymphatic clearance of amyloid aggregates; whereas sleep deprivation associated elements reduced the ability of the glymphatic system to clear Aβ biomarkers to plasma [202].

While the glymphatic system is only capable of removing soluble amyloid fibrils [203], the SARS-CoV-2 virus enhances the production of insoluble amyloid crystals, reducing soluble proteins [152] to severely compromise the cerebral efflux of Aβ via the glymphatic system. Consequently, dysfunction of the glymphatic system is often reported in recovered COVID-19 patients with only mild infections [204], but are associated with neurological impairments [205]. It is perhaps not a coincidence that the higher EEG Delta power that is correlated with higher overnight glymphatic clearance of amyloids is associated with deep, restorative, non-REM stage 3 slow wave sleep [206]. During slow wave sleep, melatonin is released from the pineal gland into the circulatory system, where maximum slow wave activity has been reported to coincide with melatonin secretion rhythm, with frequency band crests occurring before the rise of plasma melatonin concentrations [207]. Unexpectedly, fragmentation of stage 3 slow wave sleep in healthy male subjects was found to elevate morning melatonin in saliva by 1.8 times compared to controls [208]. Hypothetically, the increase in saliva melatonin may be associated with the activation of the serotonergic system during slow wave sleep suppression [209]. Alternatively, if melatonin is utilized for the glymphatic clearance of amyloids via solubilizing and preventing crystallization of amyloids, the observed 50% reduction of stage 3 slow wave sleep could severely inhibit glymphatic clearance during active sleep, resulting in higher levels of morning melatonin detected in the saliva as a result of unused melatonin during slow wave sleep glymphatic clearance. The subsequent increase in urinary melatonin excretion in cataract surgery subjects may be another example where the reduced utilization of melatonin for solubilizing and clearing cataractous amyloid aggregates in the crystalline lens via the ocular glymphatic system results in elevated circulatory/excretion melatonin levels [65, 210, 211].

Melatonin facilitates brain clearance of soluble amyloids in Neuro-PASC

Pineal melatonin is released directly into the CSF via the pineal recess of the third ventricle. Ventricular CSF melatonin measured in animals is at least one order of magnitude higher compared to levels detected in the blood of these animals. Whereas the ratio of pineal melatonin released into the CSF and the blood in humans still awaits quantification [212–215]. The indisputable correlation between sleep and the clearance of brain waste [216–218] suggests a definitive, protective role of melatonin in the brain during sleep [212, 219–221]. Postmortem ventricular CSF melatonin was conclusively demonstrated to be negatively correlated with the severity of AD pathology. Aged individuals with early neuropathological changes, including an accumulation of amyloid aggregates, exhibited a significant ~3.5- to ~7-fold decline in their CSF melatonin levels compared to individuals without any amyloid aggregates [222]. Similarly, a staggering 5-fold decline in ventricular postmortem CSF melatonin was identified in 85 AD patients compared to 82 age-matched controls without AD [223].

The adequate release of melatonin from the pineal gland into the CSF during slow wave sleep becomes an important consideration for the clearance of brain waste. An analysis of postmortem human brain material and melatonin levels in postmortem pineal glands obtained from AD subjects and healthy controls revealed that alterations in CSF melatonin levels closely mirrored the changes in the pineal melatonin content [224]. Young adult Sprague-Dawley (SD) rats, when subjected to both pinealectomy and intracerebroventricular (icv) Aβ1–42 infusion, experienced increased cognitive impairment and increased accumulation of Aβ1–42 and γ-secretase compared to pinealectomy-only and icvAβ1–42-only groups. However, 50 mg/kg melatonin supplementation via IP injection for 40 days improved all tested parameters in all groups to closely match control levels [225]. Accordingly, pineal gland dysfunction is believed to be responsible for many NDD-associated symptoms, and that the use of melatonin supplementation is an efficacious remedy for these conditions [226–229]. Importantly, pinealectomy experiments on animals revealed that the reduction in hippocampal melatonin levels is not only exaggerated by treatment with Aβ, but also induced learning and memory deficits [230, 231].

Accordingly, the dysregulation of melatonin from the fragmentation of NREM slow wave sleep resulting in increased amyloid aggregation and related symptoms are increasingly associated with proteostasis failure in neurodegenerative disorders [175, 232]. The insoluble amyloid crystalline assemblies induced by the SARS-CoV-2 virus not only present huge obstacles against clearance by the glymphatic system, but also accelerate and promote the self-assembly of additional insoluble Aβ-sheet-rich, inert, stable structures that encourage sequestration of proteins, promoting loss-of-function toxicity. The lower free energies and reinforced stability of these amyloid crystalline assemblies may be responsible for their increased cytotoxicity as a result of delayed/defective clearance, which is exacerbated by elevated depletion of soluble proteins by SARS-CoV-2 nucleic acids in human CSF compared to uninfected controls [149, 152, 233]. The Omicron variant was associated with sleep deterioration (new onset insomnia or exacerbation of chronic insomnia) together with altered brain structures in infected patients compared to healthy never-infected controls [234]. While asymmetric bilateral glymphatic dysfunction was reported in mild COVID-19 patients, where cognitive abnormalities were more evident in older patients and subjects identified with right-sided glymphatic dysfunction [204, 205].

Nonetheless, at the concentration of ~1.54 mM, melatonin administered for four weeks to chronic unpredictable mild stress (CUMS) mouse model of depression was able to restore glymphatic system function and improve sleep structure [235]. For close to three decades, a wealth of experimental in vivo and in vitro work studied the ability of melatonin to prevent amyloid fibrillization and disassemble amyloid aggregates in neurodegenerative disorders [225, 236–251]. Table 2 is a representative—but not comprehensive—collection, presented in chronological, descending order, of in vivo, ex vivo, and in vitro studies that examined the effects of melatonin on Aβ and phase-separated amyloidogenic aggregates under various contexts.

The conundrum of resolving contradictory results from in vivo and in vitro experiments on melatonin and amyloid disaggregation

The results obtained from in vivo studies delineated a distinct dose- and time-dependent relationship between melatonin supplementation and the prevention of amyloid deposition-induced neurodegenerative symptoms in healthy and diseased animals tested. The highest melatonin supplementation recorded in the experiments in Table 2 was 6 mg/day in drinking water administered to transgenic Tg2576 AD mice for 11.5 months, starting at 4 months of age [242]. Compared to the same Tg2576 mice given ~65 µg melatonin per mouse in drinking water for only 10 weeks [247], the differences were stark. Transgenic AD mice supplemented with high-dose melatonin (HDM) exhibited lower mortality rate similar to non-transgenic mice. The brains of HDM-treated animals showed impressive declines in oligomeric Aβ40 and Aβ42 compared to same-age untreated controls. It is noteworthy that HDM treatment significantly elevated lymphatic clearance of soluble monomeric Aβ40 and Aβ42 peptides in transgenic AD mice compared to non-treated wildtype controls [242]. Tg7256 AD mice supplemented with a 92-fold reduction in melatonin in drinking water, starting 10 months later, for a total duration of only 10 weeks, fared poorly. These mice showed disappointing results where melatonin treatment failed to reduce neither brain Aβ levels nor oxidative damage [247].

Recent advances begin to reveal that the wide range of molecular mechanisms employed by melatonin to improve cognitive function and reverse amyloid-induced cognitive impairment are dependent upon the regulation of phase separation. By modulating BC dynamics, melatonin effectively adjusts the behavior of immune system mediators such as the NOD-like receptor family, pyrin domain-containing 3 (NLRP3) inflammasome and transcription factor EB (TFEB) [238, 252, 253]. Melatonin regulation of phase separation processes that drive the fibrillization of amyloids reflects an ancient synergistic relationship dependent upon ATP and other molecular interactions. Consequently, the variation and fluctuation in these conditions may explain the discrepancies and inconsistent results observed in both in vitro and in vivo studies examining the ability of melatonin to prevent and/or disassemble amyloid aggregates.

The removal of water molecules from protein hydration shells drives phase separation and nucleation

Overcrowding, hydrophobic, and electrostatic interactions are understood to drive phase separation of Aβ peptides, where solid-liquid interfaces serve as sites for heterogeneous primary and accelerated secondary nucleation events [115, 255]. In particular, dehydration—the removal of water molecules from protein hydration shells into bulk water—is the central player in the entropy tug-of-war in phase separation thermodynamics during condensate formation, where released water gains entropy and confined water molecules in the dense phase are penalized entropically [258–260]. In dehydration, the resultant positive change in water entropy promotes nucleation where amyloid aggregation is accomplished via the release of water molecules surrounding hydrophobic, monomeric amyloid peptides [261, 262]. Accordingly, the hydrophobic collapse [134, 263] of amyloid protofilaments as a result of the release of water molecules leads to amyloid aggregation, and the increase in ion concentration that augments the salting-out effect also produces similar aggregation effects [264, 265]. The phase separation of tau droplets into gel-like states before transitioning into cytotoxic fibrils that are associated with NDDs is also fueled by the dehydration of hydrophobic protein residues [266, 267].

Therefore, the presence of melatonin in the critical stage where water is expelled from hydrophobic residues may be a deciding factor that can prevent further aggregation and nucleation of amyloids. As such, the ability of melatonin to disassemble preformed, post-nucleation fibrils may reflect an important step that is present in in vivo conditions, but not necessarily in experimental setups using preformed fibrils. Unexpectedly, when preformed α-syn fibrils [268] were treated with 25 μM melatonin for 6 h, there was an obvious destabilization and reduction in fibril aggregates and the formation of shorter fibrils. At compound:peptide ratios of 5:14 and 2:14, 25 μM and 10 μM melatonin completed blocked α-syn oligomerization, respectively. Whereas, a mere 2.5 μM concentration of melatonin at compound:peptide ratio of 1:28 was able to exert an inhibitory effect on α-syn oligomerization [244]. The effectiveness of melatonin in the disaggregation of preformed α-syn and the blocking of further oligomerization underscores a unique synergy between melatonin and ATP in the regulation of phase separation by maintaining the hydration of proteins to prevent aberrant aggregation.

The ability of melatonin to disaggregate preformed amyloids is dependent upon mitochondrial ATP production

Importantly, the differences in neuronal cell lines used and the amount of fetal bovine serum (FBS) incubated with the cells may dramatically alter the amount of ATP that may be available to interact with melatonin synergistically to solubilize and disassemble preformed fibrils. Commercially available fetal bovine sera from four different sources (Gibco Thermo Fisher, Euroclone, Microgem, and Sigma-Aldrich) were found to contain a wide range of ATP from as high as 1300 pmol/L (Gibco) to as low as 250 pmol/L (Sigma) [269]. Thus, the amount of FBS added to the testing medium may affect total ATP available. In addition, the synthesis of ATP by different cell lines including primary neurons and Neuro2A (N2a) cells under different culture conditions may also be different, leading to different experimental outcomes reported in literature [270]. Mixed neuronal cultures of mesencephalon and neostriatum were used to study the effect of melatonin on preformed α-syn fibrils. These cells were incubated in 10% FBS for 48 h and switched to 5% calf serum for the assay. No glucose was used [244]. Most commercially available FBS is collected from bovine fetuses from pregnant cows during slaughter; whereas fetal calf serum (FCS) is collected from calves under 12 months old, and may contain higher protein than FBS. The level of ATP in commercially available FCS has not been determined. Conversely, neuroblastoma N2a cells were used to study the effect of melatonin on preformed full-length tau, using 0.5% FBS and without any glucose added during experiments [240].

ATP: the biphasic, universal cosolute of BCs

In 1952, ATP was first reported to function as a classic hydrotrope that can dissolve insoluble aggregates [275]. This role was independently verified in different experiments decades later [276, 277]. Recent advances reveal the unique molecular structure of ATP—comprising a high-energy, hydrophilic, negatively charge triphosphate moiety, and a hydrophobic adenine nucleobase attached to a hydrophobic ribosyl group—is the primary reason why ATP is selected as the preferred energy currency used by all living organisms to regulate phase separation [278, 279]. The emergence of life required an efficient energy transfer system. Nature selected ATP as the ultimate energy transfer system for kinetically stable and thermodynamically active phosphates despite the availability of other viable phosphate carriers [280]. In the context of phase separation, phosphate carriers release high-energy phosphates to phosphorylate peptides to form and stabilize membraneless condensates. The high negative charge of phosphates can suppress charge-charge repulsion to facilitate and initiate phase separation [281].

Both ATP and high energy phosphate carriers can engage proteins via nonspecific weak interactions that promote phase separation by displacing water H-bonds, favoring the folded state by substituting water H-bonds with triphosphate H-bonds. Nevertheless, the highly negatively charged triphosphates continue to interact with hydrophobic protein patches to cause severe aggregation [282, 283]. The triphosphate moiety of ATP has been demonstrated to enhance the fibrillization of tau K18 fibrils in vitro [284]. Whereas, the stoichiometric incorporation of ATP (0.45 mM) with oligo-lysine segments containing amyloidogenic residues drives aberrant phase separation towards the self-assembly of fibrils [285]. Yet, the attachment of the adenosine moiety distinguishes ATP from other high-energy phosphate carriers. The hydrating, aromatic adenosine moiety can shield the aggregating effects of the triphosphate moiety, solubilizing solvent-exposed hydrophobic patches via π-π, π-cation and van der Waals interactions [279, 283, 286].

Consequently, a biphasic effect—where low levels of ATP promote phase separation and aggregation, but high levels have the opposite effect—has been extensively observed and reported [278, 287]. The addition of 5 mM ATP promoted phase separation of IDP-rich HNRNPG protein via charge neutralization and conformational compaction. Whereas at levels above 24 mM, ATP reduced intermolecular contacts to dissolve condensates [288]. While the addition of merely 0.25 mM ATP to wild-type α-syn monomers was adequate to begin the induction of fibril aggregation in a dose-dependent manner. However, the continued increase of ATP results in the inhibition of α-syn aggregation. In the presence of 10 mM ATP, a significant loss in late-stage fibril aggregation was clearly indicated by a reduction of Thioflavine T (ThT) fluorescence, confirming the biphasic effect of ATP in α-syn aggregation [289].

The solubilizing effect of ATP is dependent upon the adenosine moiety

The recognition that ATP may facilitate the transition of a droplet into a fibril and back into a droplet [285, 290] supports previous computational and experimental studies that showed that at physiological levels between 0.5 mM to 10 mM, ATP prevents Aβ misfolding via stabilization through the triphosphate moiety and the solubilizing effect of the adenosine moiety [291]. Even though ATP can displace water H-bonds, altering protein hydration, leading to protein folding, the solubilizing effect of the adenosine moiety is highly effective in preventing or delaying further transitioning into insoluble aggregates. Despite the fact that hydrophobic adenosine easily binds to accessible hydrophobic sites in unfolded proteins, its solubility in water is surprisingly low [292]. However, when adenosine is combined with the hydrophilic triphosphate moiety that is surrounded by three–four layers of hypermobile water [293], its solubility is greatly enhanced. This ingenious combination results in, at minimum, an order of magnitude increase in anti-fibrillation performance compared to classic hydrotropes [292, 294].

Nucleation changes the course of fibril aggregation and the subsequent dissolution requirement for ATP

Nevertheless, experiments employing all-atom molecular dynamics for different force fields revealed that ATP, at 150 mM concentration, can destabilize and inhibit the aggregation of preformed Aβ16–22 peptides via π-π stacking, NH-π interactions, and dose-dependent increases in H bonding [296]. Although primary neurons are capable of producing 150 mM ATP under glucose replete conditions [271], the tested concentration of ATP in 136 organ, tissue, and cell sources obtained from the three major domains of life averaged only ~4.41 mM [297]. Consequently, the use of ATP at concentrations between 100–500 mM to dose-dependently inhibit the aggregation and dissolve preformed Aβ40 fibrils in complementary wet-lab experiment and molecular simulation may require justification [298]. It is unclear why a wide range of ATP, from 10 mM to 500 mM, is required to inhibit and dissolve amyloid aggregates. The fact that wild-type α-syn monomers require only 10 mM ATP, but Aβ16–22 prefibrils and Aβ40 preformed aggregates require 150 mM and 500 mM, respectively, to dissolve aggregates reveals the important step in nucleation that changes the course of fibril aggregation. Viruses including SARS-CoV-2 are convenient catalysts that promote aggregation of amyloids via heterogeneous nucleation [152]. Consequently, the ancient synergistic relationship between melatonin and ATP becomes even more critical in the prevention and inhibition of amyloid aggregation as the result of viral infections and viral persistence.

Melatonin compensates for low ATP availability

The potential requirement for exceptionally high concentrations of ATP to inhibit and dissolve aggregates post nucleation events could present living organisms, especially early unicellular organisms without mitochondria, with an energetic cost challenge that is detrimental to survival. The efficient management of proteostatic energetic cost confers significant survival advantages during evolution [157]. This advantage is further enhanced when smaller proteins with low production and maintenance cost, including melatonin, are utilized. It is not a coincidence that the molecular mass of the penultimate, rate-limiting enzyme in the melatonin biosynthetic pathway—serotonin N-acetyltransferase (SNAT)—is only ~15-, ~17-, and 23-kDa in archaea, cyanobacteria, and humans [arylalkylamine N-acetyltransferase (AANAT)], respectively [60, 299, 300]. In comparison, the molecular mass of the ATPase enzymes responsible for the production of ATP is ~730-, ~545-, and ~592-kDa in archaea, bacteria (E. coli), and humans, respectively [301–303].

Unicellular dinoflagellates, including L. polyedra, have significantly reduced mitochondrial genomes, comprising only three protein-coding genes, while the majority of 45 other genes are permanently lost and the remainder transferred to the nucleus [304]. Under duress, L. polyedra may not be able to elevate ATP synthesis to regulate the phase separation of guanine crystals that control their bioluminescence. Instead, L. polyedra increases melatonin synthesis from ~1 μM to staggering levels of as much as ~20 mM (4.61 μg/mg protein) [66, 305]. The synthesis of extreme levels of melatonin by L. polyedra to regulate the phase separation of guanine crystals is highly reminiscent of in vivo studies that administered HDM (267 mg/kg b.w.) to transgenic AD mice with dysfunctional mitochondria [242]. Despite the potential effects of reduced ATP concentration in neurons of transgenic mice [306], HDM was able to suppress the formation of oligomeric Aβ40/Aβ42 and markedly increase the glymphatic clearance of soluble Aβ40 and Aβ42 compared to untreated mice at the same age [242].

Melatonin forms H-bonds with water and stacks with the adenine moiety

Melatonin is able to compensate for lower ATP concentration and reduced solubilizing capacity, by enhancing the adenosine moiety. Melatonin is an amphiphile with hydrophobic parts that change the conformation of melatonin in the presence of water molecules as a result of the hydrophobic effect [307]. Even though melatonin has low solubility in water, it can form stable H-bonds with water. Car-Parrinello molecular dynamics and ultraviolet/infrared spectroscopy studies indicate that melatonin possesses five distinct H-bond sites for water. Melatonin’s amide NH and indole NH groups are H-bond donors to water molecules. Conversely, the amide carbonyl, methoxy oxygen, and indole π cloud of melatonin accept H-bonds from water molecules. Helmholtz free energy studies reveal that water hydrogen molecules when coordinated with the O of the amide group can exhibit an infinite average residence time due to the high degree of stability [308, 309]. Just one single water molecule attached to melatonin changes its conformational preference, where strong H-bonds produce substantial electronic frequency shifts [309] that can affect how melatonin reacts with other molecules, especially free radicals [310].

The adenosine moiety of ATP has a high structural homology to the indole moiety of melatonin (Figure 2). The aromatic indole side chain of tryptophan has been clearly demonstrated to stack comfortably with the aromatic ring systems of adenine in experiments testing molecular interactions between ATP and homopeptides [292]. The dominant π-π-stacking interactions observed between the adenosine moiety and indoles [281] may explain the synergistic association between melatonin and ATP in the regulation of phase separation that began close to 4 Ga [66]. The homologous affinity permits effortless stacking of the indole ring in melatonin with the adenine nucleobase in ATP. Melatonin is not only hydrophobic, but is also capable of binding water molecules that can enhance the solubilizing effect of the hydrophobic adenosine moiety. The stacking of aromatic side chains between tryptophan and adenine in ATP can elevate the van der Waals attraction energy to 83 kJ/mol [292]. The delocalized π-systems of melatonin not only confer increased stability, but also greatly enhance the anti-fibrillation suppressive effects of the adenosine moiety of ATP (Figure 2) [292, 311].

Display full size

Figure 2. 

Diagram depicting the homologous molecular structures between melatonin [312] and adenosine triphosphate (ATP) [313] that augment the adenosine moiety effect of ATP in the regulation of protein folding and phase transitions during phase separation. The hydrophobic adenine significantly enhances the solubilizing effect of ATP to prevent aggregation of phase-separated proteins as a result of hydrophobic interactions with the triphosphate moiety. The aromatic indole ring of melatonin can stack comfortably with the adenosine moiety to further promote its hydrating effect via van de Waals interactions that effectively prevent the aggregation of solvent-exposed hydrophobic patches induced by the triphosphate moiety of ATP. Created in BioRender. Loh, D. (2025) https://BioRender.com/t35s942

Therefore, increased melatonin in the vicinity of ATP may significantly elevate the solubilizing effect without the need to increase total ATP concentration. Early organisms may have depended upon this synergy to manage proteostatic energy costs and enhance survival advantage. Conversely, an absence or reduction in melatonin would require extremely high levels of ATP to inhibit further progression of amyloid aggregates post nucleation. Untargeted metabolomic analyses measured both ketone bodies and metabolites derived from the glycolytic and pentose phosphate energy metabolism pathways collected from postmortem prefrontal cortex brain tissues from AD and subjects without clinical AD diagnosis. The results obtained revealed a significant deficit in global energy metabolism in the brains of AD patients [314]. The dramatic reduction in pyruvate essentially suppresses mitochondrial oxidative phosphorylation (OXPHOS). Without the chemical reduction of pyruvate, nicotinamide adenine dinucleotide (NAD⁺) cannot be regenerated to fuel continued glycolysis that supplies the reduced form of NAD⁺ (NADH)—the essential source of reducing equivalents—for the mitochondrial tricarboxylic cycle and the electron transport chain during OXPHOS [315, 316].

Reduced production of ATP suppresses slow wave sleep and impedes effective clearance of amyloid aggregates

The strong correlation between sleep and the clearance of brain waste has been clearly identified in literature [216–218]. The exact molecular mechanism responsible for the evidence reported remains elusive. In order to sustain synaptic transmission, neurons require approximately 40% of total ATP produced in the cortex [317]. Whereas during slow wave sleep, the surge of ATP from the reduction of neuronal activity [318] may be critical in facilitating the solubilization of amyloid crystals for removal by the CSF and associated waste-clearing systems. Since the adenosine moiety of ATP is largely responsible for the solubilizing effect of ATP as a hydrotrope, it is not surprising that increased adenosine levels in the CSF enhances both glymphatic flow and clearance of wastes in rodent models [319]. Conversely, fragmentation of slow wave sleep may lead to insufficient ATP in the CSF that not only prevents the effective solubilization and clearance of amyloid aggregates, but also the ability to utilize melatonin for the enhancement of the solubilizing effect of the adenosine moiety. Thus, the puzzling phenomenon of increased circulating melatonin associated with slow wave sleep fragmentation may merely reflect a deficiency in ATP during slow wave sleep [208].

Recent advances reveal both positive and negative feedback relationships between ATP and sleep. Adequate ATP is required for the induction of sleep, while slow wave sleep increases the amount of ATP to promote sleep. Whereas, sleep deprivation not only reduces the amount of ATP available for use by the CSF but also inhibits slow wave sleep. Experimental work studying the slow wave sleep-promoting neurons innervating the dorsal fan-shaped body (dFBN) from Drosophila brains reported that mitochondrial fusion and fission of dFBNs regulate sleep by altering ATP surplus and deficits, respectively. Increased ATP production in dFBNs induces sleep, whereas mitochondria fragmentation that reduces ATP production in dFBNs inhibits sleep induction [320, 321]. This distinct dualistic feedback cycle underscores the relevance of melatonin, which can enhance mitochondrial ATP production, and the SARS-CoV-2 virus, which can cause mitochondrial dysfunction. The multidirectional association between mitochondrial dysfunction, including reduced ATP synthesis, resulting in failure of glymphatic clearance of brain waste [322] emphasizes the relevance of the sleep-promoting effects of melatonin [323] and the dysregulation of sleep by SARS-CoV-2 [234]. Sleep dysregulation suppresses the effective, timely clearance of brain waste (Figure 1). While the SARS-CoV-2 is essentially a natural magnet for amyloid aggregation that can significantly elevate the risk for development and progression of Neuro-PASC, unexpected challenges complicate the targeting of amyloid aggregation in Neuro-PASC and other viral infections.

SARS-CoV-2 alters the molecular grammar of host physiological condensates

The phase separation of FUS and TDP-43 forms condensates comprising a diverse range of material properties in both health and disease. The aberrant aggregation of FUS and TDP-43 changes the molecular grammar of these condensates that drive the progression of NDDs including ALS, and frontotemporal dementia (FTD) [354]. Consequently, the partitioning of N proteins into host condensates may alter the molecular grammar of functional condensates, turning them into pathological aggregates that are associated with disease-states. The ultimate irony is that host physiological condensates are hijacked by viruses to serve as ready scaffolds that promote the formation of RNA condensates that support and accelerate viral replication. Nonetheless, the double-edged sword of phase separation cuts both ways. Recent experiments reported that AMPs can partition into N protein condensates with a range of different results, from the reduction of condensate dynamics to the formation of fibrillar structures that impair the normal function of RNA transcription and virion assembly in N protein condensates [355].

The aberrant aggregation of α-syn transforms functional, physiological oligomers and protofibrils into pathogenic, fibrous filamentous Lewy body assemblies that are responsible for promoting a range of α-synucleinopathies including PD, dementia with Lewy bodies, and multiple system atrophy (MSA). The SARS-CoV-2 virus is believed to cause α-synucleinopathies associated with NDDs by intensifying the propensity for α-syn aberrant aggregation [356]. In vitro experiments using neuroblastoma SH-SY5Y cells reported observing direct interactions between the N protein, but not the S protein, of SAR-CoV-2 and α-syn. The interactions result in the formation of complexes where the N protein acts as a critical nucleus that can significantly speed up the formation and aggregation of α-syn amyloid fibrils [145]. Conversely, bioinformatics analysis that was experimentally verified in human embryonic kidney (HEK) 293 cells demonstrated high affinity binding of α-syn to both the S and the N protein of SARS-CoV-2. Furthermore, both the S and N protein induced upregulated expression of α-syn and accelerated its aggregation, causing the formation of Lewy-like bodies (Table 1) [143].

SARS-CoV2 exploits the biphasic features of ATP to enhance phase separation for replication and maintenance

Phase-separated condensates, regardless of origin, promote the generation of amyloid fibrils at BC interfaces [108]. Recombinant SARS-CoV-2 proteins—S protein 1, 2 (S1, S2), receptor binding domain (RBD), and N—can spontaneously self assemble into nanoparticles and nanofibrils to form amyloid-like structures in receptor specific and non-specific mammalian cell lines [142]. The low complexity domain (LCD) of the SARS-CoV-2 N protein phase separates with RNA. Although the formation of amyloid fibrils by LCD condensates is not dependent on RNA, the maturation and the stacking of stable, Aβ-sheets can be modulated by RNA interactions (Table 1) [144]. Notwithstanding, it is determined that phase separation of SARS-CoV-2 N protein is maximally optimized by the presence of oligonucleotides at a ratio of 1:1. Whereas, phase separation of the N protein can be dramatically terminated by just a slight increase in either N protein or nucleic acid [111].

The hydrophobic adenosine moiety that is attached to the ATP nucleotide comprises the nucleobase adenine. Therefore, the concentration of ATP, as a result of the solubilizing effect of adenosine, regulates the phase separation propensity of SARS-CoV-2 in a biphasic manner [357]. In vitro experiments reported that at a molar ratio of 1:500 (N protein:ATP), all phase-separated N protein droplets formed at a lower ratio (up to 1:200) were dissolved [287]. Unexpectedly, ATP was observed to displace nucleic acids such as arginine that bind to the N protein to induce phase separation. By replacing the displaced nucleic acid, ATP can then proceed to solubilize and dissolve the phase-separated droplets formed by N protein and other nucleic acids [358]. Therefore, as part of an effective hijacking strategy that is evolutionarily conserved, viruses including SAR-CoV-2 must be able to hijack and control mitochondria—capable of large-scale ATP production—that can derail and terminate viral replication and maintenance.

Diversion of carbon energy substrates to glycolysis suppresses endogenous melatonin production

The faster generation of a lower concentration of ATP via glycolysis favors the production of viral factories via phase separation due to the biphasic effect of ATP that stabilizes condensates at lower concentrations. Consequently, the dysregulation of metabolic flexibility [379] and the disturbance of the delicate balance between ATP production rate and yield that maintains energy homeostasis [380] may prevent the timely delivery and utilization of ATP. Reduced ATP impacts not only neurogenesis [381] but also causes neurological dysfunction [382] as a result of impaired, timely clearance of amyloid aggregates by CSF brain waste clearance systems during slow wave sleep that is dependent upon high concentrations of ATP and melatonin released from the pineal gland.

Fusion and fission regulate the proper differentiation of mitochondrial subpopulations

During nutrient deficiency, the disturbance of mitochondrial dynamics that suppress fusion or fission can derail the formation of mitochondria enriched in ATP synthase and cristae and others that are only enriched in the enzyme, pyrroline-5-carboxylate synthase (P5CS), responsible for reductive synthesis of macromolecular precursors. The inhibition of either mitofusin (Mfn, responsible for fusion) or dynamin-like-protein-1 (Drp1, responsible for fission), forces the selection between the preservation of either OXPHOS or reductive synthesis of macromolecular precursors (Figure 1) [390]. Knocking out Mfn1/2 in mouse embryonic fibroblasts (MEFs) rendered the cells incapable of separating into subpopulations with distinct functions, under both nutrient replete and deficient conditions. While mitochondria from Drp1^–/– MEFs cultured under nutrient-deficient conditions not only could not subdivide but also exhibited morphological changes including elongation and bulb-like enlargement [390].

SARS-CoV-2 dysregulation of mitochondrial dynamics produces dysfunctional, hollow, elongated mitochondria with high membrane potential

A transcriptomic and metabolomic data analysis of the changes in mitochondrial morphology and function induced by SARS-CoV-2 accessory proteins in A549 human lung carcinoma cells reported the detection of two distinct subpopulations of mitochondria with high and low activity. Hollow mitochondria devoid of cristae exhibited lower membrane potential and activity level. These effects were associated with the downregulation of optic atrophy 1 (OPA1) responsible for maintaining cristae morphology [359, 394]. Whereas, mitochondrial subpopulations with higher activity and membrane potential still exhibited diminished ATP production capacity [359], supporting the finds of Ryu et al. [390] (2024).

The high-curvature apex at cristae invaginations of the inner mitochondrial membranes are exclusive anchors for ATP synthase dimers. Accordingly, ATP output during OXPHOS can be dramatically impacted by dysregulation in cristae morphology, or worse, the complete absence of cristae invaginations [395, 396]. Intriguingly, the suppression of fission 1 (Fis1), the primary recruiter of Drp1, did not cause elongation of mitochondria by SARS-CoV-2 accessory proteins [359]. Conversely, an in vitro study employing HEK293 cells infected with the SAR-CoV-2 virus reported the suppression of Drp1 expression resulted in the aberrant elongation of mitochondria with high ΔΨm when exposed to N protein-RNA clusters, but not single RNAs from SARS-CoV-2 (Figure 1) [397].

The aberrant elongation of mitochondria may reflect the inability to differentiate into optimal subpopulations. Whereas the high ΔΨm observed in recent studies is controversial. Under physiological conditions of nutrient deficiency, a high ΔΨm in mitochondrial subpopulations can imply the presence of elevated reductive biosynthesis that correlates with a high NAD(P)H/NAD(P)⁺ ratio [where NAD(P)H and NAD(P)⁺ represent the reduced and oxidized forms of nicotinamide adenine dinucleotide phosphate, respectively] [390]. Whereas under pathological virus-induced nutrient deficiency conditions, a high ΔΨm observed with elongated mitochondrial morphologies may imply dysregulated mitochondrial dynamics that resulted in mitochondria capable of only ATP production, explaining the elevated ATP levels when compared to uninfected controls [397]. It would not be surprising that if the production of macromolecular precursors by elongated mitochondria were also measured, their levels may also exceed those of uninfected controls. Therefore, more studies may be necessary to elucidate how the SARS-CoV-2 virus modulates mitochondrial dynamics and their ability to differentiate into subpopulations that can respond to nutrient deficiency efficiently. Ultimately, the disturbance of mitochondrial dynamics leading to mitochondrial dysfunction and deficiency of ATP and melatonin is indirectly responsible for exacerbating amyloidogenesis in Neuro-PASC. Supplementation with melatonin is a time-tested solution that can overcome these challenges by protecting mitochondrial dynamics and elevating ATP synthesis [398].

Melatonin regulates fusion and fission to elevate ATP production

Melatonin is extensively reported in literature to protect mitochondrial dynamics, enhancing ATP production under different contexts by maintaining membrane potential and balancing mitochondrial fusion and fission [401, 402]. By enhancing the expression of OPA1, melatonin normalizes mitochondrial fusion and fission, elevating ATP production to protect against cardiac ischemia/reperfusion (I/R) injury in vivo and hypoxia reoxygenation injury in vitro [403]. Similarly, melatonin normalized mitochondrial function, structural integrity, and ATP production by enhancing the expression of Opa1 and Mfn1/2 in SD rat aorta vascular smooth muscle cells treated with beta-glycerophosphate to induce calcification [404]. Conversely, in red vastus lateralis muscles of Zücker diabetic fatty rats, melatonin restored ATP production to control levels by downregulating fusion proteins Opa1 and Mfn2 while upregulating fission proteins Drp1 and Fis1 (for additional details on study design and effects, please see Table 3) [405].

Whereas, in a streptozotocin (STZ)-induced diabetic mouse model, daily melatonin supplementation at 10 mg/kg b.w. rescued neural retinal dysfunction and prevented retinal microvascular permeability via Mfn2 upregulation that rebalanced fusion and fission in retinal mitochondria of diabetic mice compared to healthy controls [406]. Notwithstanding, 10 mg/kg was inadequate for normalizing hyperglycemia in STZ diabetic mice. While both 10 mg/kg and 30 mg/kg upregulated the expression of Opa1 and downregulated Drp1 in the hearts of rats subjected to non-lethal mechanical trauma (MT)-induced myocardial injury, only treatment with 30 mg/kg melatonin reduced apoptosis rates by > 50%, compared to < 10% in rats treated with 10 mg/kg melatonin [407]. While H9c2 cardiomyoblasts pretreated with 100 μM melatonin before incubation with 20% traumatic plasma for 12 h were able to restore ∆Ψm loss and maintained ATP production similar to control levels (for additional details on study design and effects, please see Table 3) [407].

Melatonin’s impressive array of additional ATP-elevating mechanisms

Since the early 1990s, melatonin has been demonstrated to utilize an extensive range of distinct mechanisms to increase mitochondrial ATP synthesis in in vivo, in vitro, and in situ experiments that studied different organs, tissues, and cell lines. In addition to modulating mitochondrial dynamics, these diverse mechanisms include elevating the expression of OXPHOS enzymes in skeletal muscles and bone marrow mesenchymal stem cells in a dose-dependent manner [408, 409], and reducing state 4 respiration proton leak and mitochondrial permeability transition pore activity in brown adipose tissue [410]. Melatonin increased the lifespans of both aged senescent and senescent-resistant mice by increasing ATP production and restoring optimal mitochondrial function [411]. Melatonin treatment in septic mice reversed damage to mitochondrial proteins that increased ATP production and normalized ATP/ADP ratios compared to controls [412]. In rat liver cold-storage models, both in vitro [413] and in situ [414] experiments verified that melatonin added to perfusates during reperfusion restores liver morphology and function, mitigating I/R injury by elevating ATP production (please refer to Table 3 for additional details on study designs and effects reported in the aforementioned studies).

Mitochondria from normal mouse liver cells treated with 1 nM to 1 mM melatonin exhibited different levels of activity enhancement in OXPHOS complexes in a dose-specific manner. Melatonin at 1 nM in complex I, 1 nM to 1 μM in complex II, 1 mM in complex III, and 1 mM in complex IV produced 56%, > 31%, 30%, and a staggering 609% increase in activity levels, respectively [415]. Although melatonin treatment during reperfusion in cold-storage livers was unable to completely restore the loss of ATP production to control levels [414], melatonin pretreatment in male Wistar rats subjected to ruthenium red-induced mitochondrial damage not only prevented the reduction in complex I and IV activity in a time-dependent manner, but also increased complex I and IV activities in the liver mitochondria by 33% and 233%, respectively, compared to controls. Male Wistar rats not subjected to ruthenium red toxicity (60 μg/kg, IP) but treated with 10 mg/kg (IP) all showed significant increases of up to 166% in complex I and IV activity levels in the liver and brain mitochondria in a time-dependent manner. Importantly, the additive effect of melatonin on brain and liver mitochondria OXPHOS activity completely vanishes starting at 120 min for complex I in the brain and complex IV in the liver mitochondria (please refer to Table 3 for additional details on study designs and effects reported in the aforementioned studies) [416].

Melatonin regulates sirtuins to elevate ATP synthesis in mitochondria

Sirtuins (SIRTs) are NAD⁺-dependent class III histone deacetylases that can tune the assembly and disassembly of BCs by acetylation and deacetylation of regions within IDPs [417–419]. A substantial amount of evidence in literature supports a definitive, dependent relationship between melatonin and SIRTs, especially SIRT1 and the mitochondrial SIRT3 [420], in enhancing mitochondrial function and energy homeostasis. In vivo melatonin treatment enhanced mitochondrial function in STZ-induced diabetic rats subjected to myocardial I/R (MI/R) injury. Melatonin reduced MI/R injury by elevating activity levels in complexes II, III, and IV to increase ATP production in an AMPK-PGC-1α-SIRT3 signaling-dependent manner [421]. Oocytes from aging female mice treated with 1 µM melatonin increased mitochondrial ATP production by 54% compared to controls. The addition of melatonin to the terminal culture medium elevated the expression of Sirt1 and Sirt3 to improve oocyte maturation, fertilization, and blastocyst formation [422]. In vitro porcine oocytes supplemented with 500 nM melatonin were able to abolish rotenone-induced mitochondrial dysfunction causing embryo development impairment. Melatonin increased SIRT1 expression that not only promoted mitochondrial biogenesis but also elevated the expression of mitochondrial OXPHOS enzymes complex I and V to return the 50% reduction of ATP production back to 75% of control levels (please refer to Table 3 for additional details on study designs and effects reported in the aforementioned studies) [423].

It is noteworthy that both cancer cells and viruses divert nutrients away from mitochondrial OXPHOS into the oxidative and non-oxidative branches of the PPP to produce metabolites that fuel reductive biosynthesis instead of producing ATP [368, 424]. Consequently, by redirecting carbon flux back to mitochondrial OXPHOS and increasing ATP production, melatonin has been shown to cause cancer cell apoptosis. Lung tumor tissue samples from patients who took melatonin (5 mg/day) for one month showed a 183% increase in SIRT3 expression. Whereas Lewis lung cancer (LLC) mice that were treated with 10 mg/kg b.w. (IP) also showed enhanced expression of SIRT3 that significantly elevated ATP production by more than 320%. In vivo melatonin treatment in LLC mice restored ΔΨm, and increased mitochondrial complex I and IV activity levels by 1.75- and > 14.5-fold, respectively. The redirection of glycolytic intermediates from anabolic pathways to mitochondrial OXPHOS resulted in an impressive 5-fold reduction in tumor size at 16 days of treatment (Table 3) [425].