Nocturnal Melatonin Amplitude Collapse Is Associated with Age-Independent Convergence of Microbiome and Glymphatic Biomarkers

1. Introduction

The mammalian circadian system coordinates physiology across multiple temporal scales through central pacemaker activity in the suprachiasmatic nucleus (SCN) and peripheral oscillators driven by transcription–translation feedback loops involving CLOCK, BMAL1, PER, and CRY proteins [1,2,3]. Modern environments frequently impose circadian challenges—shift work, nocturnal light exposure, irregular sleep–wake timing—that are linked epidemiologically to cardiometabolic disease, neurodegeneration, and cancer risk [4,5,6,7].

Melatonin (N-acetyl-5-methoxytryptamine) is the principal endocrine output of the circadian system. Synthesized primarily in the pineal gland under darkness via the tryptophan–serotonin pathway (TPH→AADC→SNAT→HIOMT), melatonin exerts its chronobiotic effects predominantly through two high-affinity G-protein-coupled receptors: MT1 (MTNR1A, Gi-coupled, expressed in SCN, pituitary, retina, and peripheral tissues) and MT2 (MTNR1B, Gi/Gq-coupled, expressed in retina, SCN, and vasculature). MT1 activation mediates the primary sleep-promoting and phase-shifting effects, while MT2 is critical for phase-advance capacity and vascular tone regulation [8,9]. Beyond receptor-mediated effects, melatonin acts on additional intracellular targets including nuclear receptors RORα and RZRβ, calmodulin, quinone reductase-2, cytochrome c, and the NLRP3 inflammasome, conferring broad antioxidant, anti-inflammatory, mitochondrial-protective, and oncostatic activity [10,11,12]. A substantial fraction of total body melatonin is synthesized in enterochromaffin cells of the gastrointestinal tract, linking circadian signaling directly to the gut ecosystem [13].

In parallel, the glymphatic system provides brain-wide clearance of interstitial waste via AQP4-dependent perivascular routes [14,15,16]. Glymphatic transport is strongly sleep- and circadian-dependent; melatonin specifically facilitates glymphatic function by promoting slow-wave sleep and directly modulating AQP4 channel polarization [17,18]. Impaired glymphatic clearance is increasingly recognized as a convergent pathophysiological mechanism in neurodegeneration, with white matter hyperintensity (WMH) accumulation and enlarged perivascular spaces serving as tractable neuroimaging surrogates [19,20]. The DTI-ALPS index—derived from diffusion tensor imaging along perivascular spaces—provides a non-invasive proxy for perivascular fluid transport and has demonstrated clinical validity in aging and neurodegeneration [21].

The gut microbiome oscillates diurnally [22], and its composition is sensitive to circadian disruption. Notably, urolithin A—a microbial metabolite of dietary ellagitannins produced specifically by Gordonibacter urolithinfaciens and Ellagibacter isourolithinifaciens—induces mitophagy, enhances mitochondrial biogenesis, and modulates neuroinflammation [23,24]. Because urolithin production depends on taxonomically restricted bacterial producers, circadian-associated dysbiosis could plausibly impair urolithin generation, creating a mechanistic bridge between disrupted rhythmicity and mitochondrial vulnerability. This pathway has not been evaluated in the context of melatonin amplitude collapse.

Critically, while individual components of the circadian–microbiome–brain axis have been studied separately, no clinical study has comprehensively characterized the co-disruption of melatonin rhythmicity (via fractionated urinary aMT6s and standardized dim-light plasma sampling), microbiome composition with urolithin metabolomics, and quantitative glymphatic imaging biomarkers within a single phenotyped cohort. We propose the microbiome–glymphatic–melatonin (MGM) axis as a conceptually novel integrated framework that explains how circadian desynchronization propagates simultaneously across endocrine, microbial, and central nervous system clearance pathways to produce a convergent multi-system pathological phenotype. Here, we test this hypothesis in a cohort of 179 patients with symptoms of circadian dysregulation and 107 healthy controls, and we further characterize its oncological relevance through translational chronotherapy data. The integrated MGM-axis framework examined in the present study is summarized in Figure 4.

2. Materials and Methods

2.1. Study Design and Ethics

This was a prospective, single-centre, observational cohort study conducted at the Institute for Personalized Medicine, Tbilisi, Georgia, between January 2019 and December 2023. The study was approved by the Institutional Review Board of Tbilisi State Medical University (Protocol no. 2019-IPM-01, renewed annually). All participants provided written informed consent prior to enrolment. The study was conducted in accordance with the Declaration of Helsinki.

2.2. Patient Inclusion and Exclusion Criteria

Inclusion criteria (patients): (1) age 18–90 years; (2) presence of ≥3 of the following symptoms lasting ≥6 months and not explained by a diagnosed primary psychiatric or neurological disorder: persistent fatigue unrelated to exertion, non-restorative sleep confirmed by Pittsburgh Sleep Quality Index score ≥6, cognitive complaints (concentration/memory) confirmed by MoCA ≤26, unexplained metabolic disturbance (BMI ≥25 or HbA1c ≥5.7%), diurnal mood variability, or gastrointestinal complaints consistent with dysbiosis; (3) willingness to undergo comprehensive sampling protocol; (4) stable medication regimen for ≥3 months.

Exclusion criteria (patients): (1) current use of exogenous melatonin, β-blockers, fluvoxamine, or other agents known to substantially alter melatonin secretion within 30 days of sampling; (2) active neoplastic disease under chemotherapy or radiotherapy; (3) primary sleep disorders (confirmed obstructive sleep apnoea, narcolepsy, REM behaviour disorder); (4) severe hepatic or renal insufficiency (eGFR <30 mL/min/1.73m²); (5) active autoimmune or inflammatory disease; (6) pregnancy or breastfeeding; (7) inability to complete the 24-hour urine collection protocol.

Inclusion criteria (controls): (1) age 18–80 years; (2) no chronic systemic disease; (3) PSQI ≤5; (4) MoCA ≥27; (5) BMI 18.5–24.9 kg/m²; (6) no use of melatonin-altering medications; (7) regular sleep schedule (bedtime 22:00–24:00, wake time 06:00–08:00) confirmed for ≥3 months.

2.3. Melatonin Sampling Protocol

Urinary sampling: All participants collected a continuous 24-hour urine sample beginning at 08:00. The collection was additionally fractionated into two 12-hour intervals: daytime (06:00–18:00) and nocturnal (18:00–06:00). Participants were instructed to avoid light exposure >10 lux after 20:00 during the collection period, and all sampling was performed using standardized low-light melatonin collection precautions. No awakening for forced nocturnal voids was required; participants voided upon natural awakening only. Samples were stored at −20°C within 30 minutes of collection and analysed within 14 days. Urinary 6-sulfatoxymelatonin (aMT6s) was quantified using a validated commercial enzyme immunoassay (IBL International GmbH, Hamburg, Germany; intra-assay CV <8%, inter-assay CV <12%, sensitivity 0.5 ng/mL). Results were expressed as ng per mL of urine and normalized to creatinine for sensitivity analysis (data in Supplementary Table S1).

Plasma sampling: Plasma melatonin was measured at two standardised time windows: daytime (14:00–16:00) and nocturnal (02:00–04:00). Nocturnal sampling was performed under dim-light conditions (<10 lux red light). Blood was collected into EDTA tubes, immediately centrifuged at 4°C (2000×g, 10 min), and plasma stored at −80°C. Plasma melatonin was quantified by ELISA (Bühlmann Laboratories AG, Switzerland; sensitivity 0.7 pg/mL; intra-assay CV <10%). Plasma cortisol was measured at the same time windows using a chemiluminescent immunoassay (Roche Cobas e411; CV <5%). Urinary cortisol was measured in the fractionated 12-hour samples by ELISA (IBL International; CV <10%). Table 2 reports the standardized single-visit cross-sectional case–control comparison, whereas the supplementary urinary workbook provides repeated weekly measurements (Weeks 1–7) for the patient cohort only and is intended for longitudinal rhythm tracking and quality-control documentation rather than direct reproduction of Table 2 medians.

2.4. Microbiome Analysis

Faecal samples were collected in OMNIgene•GUT stabilisation tubes (DNA Genotek) and processed within 72 hours. DNA was extracted using the QIAamp PowerFecal Pro DNA Kit (Qiagen). 16S rRNA gene amplicon sequencing targeted the V3–V4 hypervariable region using primers 341F/806R on an Illumina MiSeq platform (2×300 bp paired-end reads; minimum 50,000 reads/sample). Bioinformatics processing used QIIME2 (v2023.2) with DADA2 denoising; taxonomic classification employed the SILVA 138 reference database. Alpha-diversity was summarised by the Shannon index. Relative abundances of Gordonibacter spp. and Ellagibacter spp. were determined at species level. Targeted plasma metabolomics for urolithin A quantification used LC-MS/MS (Waters Acquity UPLC coupled to Xevo TQ-S; internal standard: urolithin A-d4; limit of detection 0.1 ng/mL; CV <8%).

2.5. Neuroimaging and Cognitive Assessment

Brain MRI was performed on a 3T scanner (Siemens Magnetom Prisma) using a standardised protocol including FLAIR (TR/TE 9000/81 ms) for WMH volumetry (automated segmentation: FreeSurfer/LST toolbox), T2-weighted sequences for perivascular space scoring using the Potter et al. semiquantitative scale (0–3) [26], and multishell diffusion tensor imaging (b=0, 1000, 2000 s/mm²) for DTI-ALPS index calculation [21]. BNP was measured by electrochemiluminescence immunoassay (Roche Cobas; CV <5%). CD4 and CD8 lymphocyte counts were determined by flow cytometry (BD FACSCanto II). Cognitive assessment used the Montreal Cognitive Assessment (MoCA).

2.6. Statistical Analysis

Continuous variables are reported as median (interquartile range, IQR). Between-group comparisons used two-sided Mann–Whitney U tests. Effect sizes were computed as Cliff's δ. Associations between chronological age and biomarkers within the patient group were tested by Spearman rank correlation. Statistical significance was set at α=0.05 (two-tailed). Multiple testing correction (Benjamini–Hochberg FDR) was applied to Table 2 and Table 3 comparisons; all p-values reported remained significant after correction. All analyses were performed in Python 3.11 (scipy.stats, pingouin). Complete comparison data including effect sizes are provided in Supplementary Table S1.

3. Results

3.1. Participant Characteristics

Participant baseline characteristics are summarised in Table 1. Patients were older (median 57 vs 47 years; p=0.003) and had higher BMI (28.0 vs 23.2 kg/m²; p<0.0001). Sex distribution did not differ between groups (χ² p=0.737). MoCA scores were substantially lower in patients (20 vs 29; p<0.0001), consistent with neurocognitive involvement.

3.2. Circadian Endocrine Findings—Melatonin Amplitude Inversion

Results are presented in Table 2 and Figure 2. Patients displayed a profound reduction in total 24-hour melatonin output (aMT6s 19.2 vs 46.6 ng/mL; Cliff's δ=0.95; p<0.0001) with near-complete attenuation of the nocturnal peak (night aMT6s 10.2 vs 40.6 ng/mL; p<0.0001) despite relative daytime elevation (day aMT6s 8.4 vs 5.3 ng/mL; p<0.0001). This pattern produced a markedly elevated Day/Night aMT6s ratio (0.81 vs 0.14; p<0.0001), constituting a rhythmic amplitude inversion pattern—approaching phase equivalence across the 24-hour cycle. Plasma nocturnal melatonin confirmed this collapse (12.7 vs 54.4 pg/mL; p<0.0001). The cortisol axis showed parallel loss of rhythmicity: elevated nocturnal urinary cortisol (7.4 vs 4.0 µg/dL; p<0.0001) and elevated Night/Day ratio (0.83 vs 0.31; p<0.0001), consistent with HPA axis phase compression. Together, these findings indicate that patients are characterised by simultaneous melatonin and cortisol circadian amplitude collapse rather than a simple phase shift.

Table 2. Circadian endocrine and microbiome/metabolic markers.

Marker	Patients (median [IQR])	Controls (median [IQR])	p-value
Urinary aMT6s (24 h), ng/mL	19.2 (14.0–27.1)	46.6 (39.0–52.0)	<0.0001
Urinary aMT6s (day 06:00–18:00), ng/mL	8.4 (7.2–9.9)	5.3 (4.6–6.5)	<0.0001
Urinary aMT6s (night 18:00–06:00), ng/mL	10.2 (8.4–12.2)	40.6 (35.3–45.5)	<0.0001
Urinary aMT6s Day/Night ratio	0.81 (0.68–1.02)	0.14 (0.11–0.16)	<0.0001
Urinary cortisol (day 06:00–18:00), µg/dL	8.9 (7.4–10.4)	13.4 (11.6–14.9)	<0.0001
Urinary cortisol (night 18:00–06:00), µg/dL	7.4 (6.1–8.9)	4.0 (3.3–4.7)	<0.0001
Urinary cortisol Night/Day ratio	0.83 (0.68–1.02)	0.31 (0.24–0.37)	<0.0001
Plasma melatonin (day 14:00–16:00), pg/mL	6.1 (5.0–7.3)	2.8 (1.9–3.7)	<0.0001
Plasma melatonin (night 02:00–04:00), pg/mL	12.7 (10.1–15.2)	54.4 (47.2–62.5)	<0.0001
Plasma cortisol (day 14:00–16:00), µg/dL	11.9 (9.9–13.4)	15.1 (12.8–17.6)	<0.0001
Plasma cortisol (night 02:00–04:00), µg/dL	6.5 (5.3–7.7)	4.1 (3.2–4.8)	<0.0001
Microbiome Shannon diversity index	2.83 (2.27–3.38)	5.07 (4.61–5.41)	<0.0001
Firmicutes/Bacteroidetes ratio	0.58 (0.47–0.75)	2.02 (1.81–2.16)	<0.0001
Gordonibacter (relative abundance, %)	0.090 (0.050–0.120)	0.540 (0.440–0.630)	<0.0001
Ellagibacter (relative abundance, %)	0.040 (0.030–0.060)	0.320 (0.250–0.410)	<0.0001
Plasma urolithin A, ng/mL	2.40 (1.60–3.10)	25.20 (22.65–27.25)	<0.0001

All p-values are from two-sided Mann–Whitney U tests, corrected for multiple comparisons (Benjamini–Hochberg FDR). All results remained significant after correction. Table 2 summarizes the standardized single-visit cross-sectional dataset; the supplementary urinary sheet contains separate longitudinal patient-only repeated-measures data.

Table 2. Circadian endocrine and microbiome/metabolic markers.

Marker	Patients (median [IQR])	Controls (median [IQR])	p-value
Urinary aMT6s (24 h), ng/mL	19.2 (14.0–27.1)	46.6 (39.0–52.0)	<0.0001
Urinary aMT6s (day 06:00–18:00), ng/mL	8.4 (7.2–9.9)	5.3 (4.6–6.5)	<0.0001
Urinary aMT6s (night 18:00–06:00), ng/mL	10.2 (8.4–12.2)	40.6 (35.3–45.5)	<0.0001
Urinary aMT6s Day/Night ratio	0.81 (0.68–1.02)	0.14 (0.11–0.16)	<0.0001
Urinary cortisol (day 06:00–18:00), µg/dL	8.9 (7.4–10.4)	13.4 (11.6–14.9)	<0.0001
Urinary cortisol (night 18:00–06:00), µg/dL	7.4 (6.1–8.9)	4.0 (3.3–4.7)	<0.0001
Urinary cortisol Night/Day ratio	0.83 (0.68–1.02)	0.31 (0.24–0.37)	<0.0001
Plasma melatonin (day 14:00–16:00), pg/mL	6.1 (5.0–7.3)	2.8 (1.9–3.7)	<0.0001
Plasma melatonin (night 02:00–04:00), pg/mL	12.7 (10.1–15.2)	54.4 (47.2–62.5)	<0.0001
Plasma cortisol (day 14:00–16:00), µg/dL	11.9 (9.9–13.4)	15.1 (12.8–17.6)	<0.0001
Plasma cortisol (night 02:00–04:00), µg/dL	6.5 (5.3–7.7)	4.1 (3.2–4.8)	<0.0001
Microbiome Shannon diversity index	2.83 (2.27–3.38)	5.07 (4.61–5.41)	<0.0001
Firmicutes/Bacteroidetes ratio	0.58 (0.47–0.75)	2.02 (1.81–2.16)	<0.0001
Gordonibacter (relative abundance, %)	0.090 (0.050–0.120)	0.540 (0.440–0.630)	<0.0001
Ellagibacter (relative abundance, %)	0.040 (0.030–0.060)	0.320 (0.250–0.410)	<0.0001
Plasma urolithin A, ng/mL	2.40 (1.60–3.10)	25.20 (22.65–27.25)	<0.0001

All p-values are from two-sided Mann–Whitney U tests, corrected for multiple comparisons (Benjamini–Hochberg FDR). All results remained significant after correction. Table 2 summarizes the standardized single-visit cross-sectional dataset; the supplementary urinary sheet contains separate longitudinal patient-only repeated-measures data.

3.3. Microbiome and Metabolic Markers

Microbiome alpha-diversity was substantially reduced in patients (Shannon 2.83 vs 5.07; p<0.0001; Cliff's δ=0.97). The Firmicutes/Bacteroidetes ratio was markedly lower in patients (0.58 vs 2.02; p<0.0001), indicating a dysbiotic compositional shift. Critically, urolithin-producing taxa were profoundly depleted: Gordonibacter relative abundance was 83% lower (0.090% vs 0.540%; p<0.0001) and Ellagibacter 88% lower (0.040% vs 0.320%; p<0.0001). Consistent with the loss of urolithin-producing capacity, plasma urolithin A was reduced by >90% in patients (2.40 vs 25.20 ng/mL; p<0.0001; Cliff's δ=0.98), representing the most extreme effect size in the dataset. These findings provide a mechanistic bridge between circadian-associated dysbiosis and compromised mitochondrial resilience signalling.

3.4. Systemic Stress, Immune, and Neuroimaging Markers

Results are presented in Table 3. BNP was markedly elevated in patients (294 vs 55 pg/mL; p<0.0001), indicating myocardial stress. CD4/CD8 ratio was significantly reduced (0.81 vs 1.77; p<0.0001), consistent with an immunosenescent phenotype. Neuroimaging demonstrated convergent findings: WMH volume was approximately 6.5-fold higher in patients (9.99 vs 1.54 mL; p<0.0001), perivascular space scores were markedly elevated (median 2 vs 0; p<0.0001), and DTI-ALPS index was substantially reduced (1.05 vs 1.70; p<0.0001), consistent with impaired perivascular fluid transport and glymphatic dysfunction. These neuroimaging markers co-segregated with neurocognitive findings (MoCA 20 vs 29; p<0.0001). Representative MRI schematics and the age–WMH relationship are shown in Figure 5.

Table 3. Systemic stress, immune, and neuroimaging markers.

Marker	Patients (median [IQR])	Controls (median [IQR])	p-value
BNP, pg/mL	294 (193–362)	55 (38–70)	<0.0001
CD4/CD8 ratio	0.81 (0.48–1.20)	1.77 (1.51–2.10)	<0.0001
Perivascular spaces score (0–3)	2 (2–3)	0 (0–1)	<0.0001
WMH volume, mL	9.99 (8.09–12.22)	1.54 (1.15–2.08)	<0.0001
DTI-ALPS index	1.05 (0.86–1.25)	1.70 (1.64–1.79)	<0.0001

All p-values are from two-sided Mann–Whitney U tests; corrected for multiple comparisons as above.

Table 3. Systemic stress, immune, and neuroimaging markers.

Marker	Patients (median [IQR])	Controls (median [IQR])	p-value
BNP, pg/mL	294 (193–362)	55 (38–70)	<0.0001
CD4/CD8 ratio	0.81 (0.48–1.20)	1.77 (1.51–2.10)	<0.0001
Perivascular spaces score (0–3)	2 (2–3)	0 (0–1)	<0.0001
WMH volume, mL	9.99 (8.09–12.22)	1.54 (1.15–2.08)	<0.0001
DTI-ALPS index	1.05 (0.86–1.25)	1.70 (1.64–1.79)	<0.0001

All p-values are from two-sided Mann–Whitney U tests; corrected for multiple comparisons as above.

3.5. Age-Independent Convergence of MGM-Axis Biomarkers

To test whether the MGM-axis phenotype represents an age-dependent acceleration of normal aging or an age-independent pathological state, we examined Spearman correlations between chronological age and key biomarkers within the patient group. Across all domains—circadian endocrine (24-h aMT6s: r=−0.04, p=0.57; Day/Night ratio: r=+0.01, p=0.92), microbiome/metabolic (urolithin A: r=−0.05, p=0.54; Shannon index: r=−0.10, p=0.18), immune (CD4/CD8: r=+0.09, p=0.24), and cognitive (MoCA: r=−0.01, p=0.87)—no significant age associations were identified (Figure 6). This age-independence implies that the MGM-axis phenotype does not simply reflect accelerated chronological aging; rather, circadian desynchronization appears to drive diverse individuals toward a shared pathological biochemical state independently of their age.

3.6. Translational Context: Tumour Metabolism and Chronotherapy

Because MT1 and MT2 receptor signalling suppresses tumour cell proliferation via Gi-coupled inhibition of cAMP, regulation of E-cadherin, and modulation of oestrogen receptor sensitivity [12], melatonin amplitude collapse in the MGM-axis phenotype carries direct oncological implications. Figure 1 illustrates tryptophan metabolic reprogramming in triple-negative breast cancer (TNBC), in which IDO-mediated diversion to kynurenine suppresses anti-tumour immunity while depleting the serotonin–melatonin synthetic precursor pool. Among a contextual cohort (TNBC n=48 vs Luminal A n=24), daytime melatonin was profoundly reduced (0.1 vs 7.7 arbitrary units; p<0.001) and kynurenic acid elevated (117.5 vs 67.9), consistent with IDO pathway upregulation.

Chronotherapy data from our centre [25,27] further demonstrate that alignment of cytostatic administration to the nocturnal window (02:00–04:30) preserves melatonin rhythmicity across ABVD cycles in Hodgkin lymphoma (nocturnal melatonin at Cycle 3: 70 vs 30 pg/mL; 2.3-fold preservation; p=0.0001) and substantially improves treatment response in hepatocellular carcinoma (AFP reduction 36.5% vs 13.7%; tumour size reduction 27.1% vs 6.3%). These observations underscore that clinically measurable melatonin amplitude—the core deficit of the MGM-axis phenotype—is a pharmacologically tractable and clinically relevant endpoint.

Figure 3. Chronotherapy outcomes in oncology cohorts. (A) Nocturnal melatonin preservation across 6 ABVD cycles: chronotherapy-aligned administration maintains melatonin amplitude (2.3-fold difference at Cycle 3; p=0.0001). (B) Liver enzyme activity improvement with chronotherapy alignment (mean 55.5%; p<0.01). (C,D) Hepatocellular carcinoma: AFP and tumour size reductions are significantly greater with nocturnal cisplatin (02:00–04:30) vs standard daytime administration.

Figure 3. Chronotherapy outcomes in oncology cohorts. (A) Nocturnal melatonin preservation across 6 ABVD cycles: chronotherapy-aligned administration maintains melatonin amplitude (2.3-fold difference at Cycle 3; p=0.0001). (B) Liver enzyme activity improvement with chronotherapy alignment (mean 55.5%; p<0.01). (C,D) Hepatocellular carcinoma: AFP and tumour size reductions are significantly greater with nocturnal cisplatin (02:00–04:30) vs standard daytime administration.

Table 4. Oncology cohorts and chronotherapy-related outcomes (translational context).

Cohort / Study	Comparison (n)	Key outcomes
Breast cancer metabolic profile	TNBC (n=48) vs Luminal A (n=24)	Melatonin day: 0.1 vs 7.7; melatonin night: 3.9 vs 15.7; serotonin: 24.8 vs 53.7; kynurenic acid: 117.5 vs 67.9 (p<0.001 all).
Hodgkin lymphoma chronotherapy	Chronotherapy vs standard timing (n=38)	Nocturnal melatonin preserved across 6 ABVD cycles; Cycle 3: 70 vs 30 pg/mL (≈2.3×; p=0.0001).
Hodgkin lymphoma (enzyme activity)	Chronotherapy vs standard timing	Liver enzyme activity improvement vs standard: mean 55.5% (p<0.01).
Hepatocellular carcinoma	Cisplatin 02:00–04:30 vs 10:00–12:00	AFP reduction: 36.5% vs 13.7% (p<0.01); tumor size reduction: 27.1% vs 6.3% (p<0.001).

Data from cohorts illustrated in Figure 1 and Figure 3, provided for translational context. AFP, alpha-fetoprotein; ABVD, doxorubicin/bleomycin/vinblastine/dacarbazine.

Table 4. Oncology cohorts and chronotherapy-related outcomes (translational context).

Cohort / Study	Comparison (n)	Key outcomes
Breast cancer metabolic profile	TNBC (n=48) vs Luminal A (n=24)	Melatonin day: 0.1 vs 7.7; melatonin night: 3.9 vs 15.7; serotonin: 24.8 vs 53.7; kynurenic acid: 117.5 vs 67.9 (p<0.001 all).
Hodgkin lymphoma chronotherapy	Chronotherapy vs standard timing (n=38)	Nocturnal melatonin preserved across 6 ABVD cycles; Cycle 3: 70 vs 30 pg/mL (≈2.3×; p=0.0001).
Hodgkin lymphoma (enzyme activity)	Chronotherapy vs standard timing	Liver enzyme activity improvement vs standard: mean 55.5% (p<0.01).
Hepatocellular carcinoma	Cisplatin 02:00–04:30 vs 10:00–12:00	AFP reduction: 36.5% vs 13.7% (p<0.01); tumor size reduction: 27.1% vs 6.3% (p<0.001).

Data from cohorts illustrated in Figure 1 and Figure 3, provided for translational context. AFP, alpha-fetoprotein; ABVD, doxorubicin/bleomycin/vinblastine/dacarbazine.

4. Discussion

In a clinically phenotyped cohort evaluated with a comprehensive, multi-domain protocol, we report a coherent and highly statistically significant pattern linking circadian endocrine disruption to gut dysbiosis, urolithin A depletion, and quantitative brain imaging markers of impaired glymphatic clearance. The consistency of this pattern across diverse biomarker domains—endocrine, microbial, metabolic, immune, cardiovascular, and neuroimaging—supports the concept of an integrated MGM axis as a unifying pathophysiological framework for circadian desynchronization.

The novel conceptual contribution of this work is three-fold. First, we demonstrate that melatonin amplitude inversion (Day/Night ratio approaching 1.0) rather than simple melatonin deficiency is the cardinal endocrine signature—a qualitative change in circadian organisation, not merely a quantitative one. MT1-mediated phase-setting requires a sufficient melatonin signal amplitude above daytime baseline; when the Day/Night contrast is lost, MT1 signalling in the SCN and peripheral tissues is expected to be profoundly compromised [8]. Second, we show for the first time that this endocrine pattern co-segregates with the near-complete loss of urolithin A, a microbially-derived mitophagy inducer whose production depends on circadian-oscillating bacterial taxa. Third, glymphatic imaging abnormalities (DTI-ALPS reduction, WMH accumulation) directly co-segregate with melatonin collapse, mechanistically linking the loss of MT2-dependent AQP4 polarisation support to impaired perivascular clearance [18].

The absence of age correlation within the patient biomarker profiles is a particularly striking finding. Melatonin production, microbiome diversity, and glymphatic efficiency all decline with chronological age in healthy populations. The fact that patients across the 20–89-year age range showed uniformly pathological biomarker profiles—without the expected age gradient—argues strongly that the MGM-axis phenotype is driven by circadian disruption rather than aging per se. This age-independence has important clinical implications: it identifies young adults with circadian disruption as a population at risk equivalent to elderly individuals, and it suggests that circadian resynchronisation (rather than age-targeted interventions) is the appropriate therapeutic target.

Limitations: This observational cross-sectional study cannot establish causality among MGM-axis components. Patients were older on average than controls, which may confound some comparisons, though effect sizes remained very large and age-correlation analyses within the patient group were negative. We employed imaging proxies rather than direct glymphatic tracer studies. Melatonin sampling, while conducted with standardized fractionated urinary and dim-light nocturnal collection procedures, relied on clinic-based nocturnal collection rather than home ambulatory protocols. The cohort reflects a single Georgian centre, and generalizability across ethnic and environmental contexts requires validation. Longitudinal and interventional studies are required to establish whether MGM-axis resynchronisation normalises biomarkers and improves clinical outcomes.

5. Conclusions

Patients evaluated for chronic multisystem symptoms of circadian dysregulation demonstrate a coherent, age-independent MGM-axis phenotype: MT1/MT2-relevant melatonin amplitude inversion, parallel HPA axis phase compression, gut dysbiosis with near-complete loss of urolithin-producing taxa, profound plasma urolithin A depletion, systemic immune and cardiac stress, and quantitative neuroimaging evidence of impaired glymphatic clearance. This integrated multi-domain signature—characterised by large-to-extreme effect sizes across all domains—supports the MGM axis as a clinically measurable and mechanistically coherent framework for circadian medicine. The proposed clinical algorithm (Figure 7) provides an evidence-based workflow for fractionated melatonin assessment, MGM-axis biomarker screening, and targeted resynchronisation therapy including MT1/MT2-directed chronopharmacology.