Glucose-responsive AMPK-TET2-5hmc oscillation links metabolic stress to circadian transcription and Parkinson’s disease relevance

Abstract

Circadian disruption and metabolic stress are implicated in Parkinson’s disease (PD), but the epigenetic mechanisms linking nutrient fluctuations to circadian regulation remain poorly understood. Here, we combined genome-wide 5-hydroxymethylcytosine (5hmC) profiling, CRISPR-based genetic perturbations, and circadian models to investigate how glucose availability shapes epigenetic dynamics and circadian transcription. We found that low glucose activates AMP-activated protein kinase (AMPK), upregulating TET2 transcription and increasing global 5hmC remodeling. These effects are reversible upon glucose restoration and are abolished by AMPK deficiency, establishing AMPK as a key mediator of metabolic control over TET2. Notably, Tet2 exhibits intrinsic circadian rhythmicity and is temporally coupled with clock-associated gene expression. Disruption of TET2 impairs glucose-responsive 5hmC dynamics and alters circadian transcriptional outputs. Together, our findings define a glucose-responsive AMPK–TET2–5hmC axis that links metabolic state to epigenetic and circadian regulation. This work provides a mechanistic framework for understanding how metabolic dysfunction may contribute to circadian abnormalities in Parkinson’s disease.

Introduction

The circadian clock is a critical regulatory system that synchronizes biological processes to the 24-h day-night cycle, governing not only sleep‑wake behavior but also metabolic homeostasis, including glucose regulation, hormone secretion, and cellular metabolism^1,2. Disruption of circadian rhythms is increasingly recognized as a contributing factor to metabolic diseases^3–6 and, more recently, to neurodegeneration^7–10. In Parkinson’s disease (PD), circadian dysfunction—manifesting as fragmented sleep-wake cycles, altered hormone rhythms, and daily fluctuations in motor symptoms—is a common but underappreciated feature^11–13. Moreover, metabolic stress, such as region-specific brain hypometabolism and impaired glucose utilization, occurs in PD and may exacerbate neurodegeneration^14–17. Moreover, accumulating evidence suggests that metabolic dysfunction is not simply a secondary consequence of neurodegeneration but may actively contribute to PD pathogenesis. In our recent study, we demonstrated that glucose intolerance exacerbates dopaminergic neurodegeneration through mitochondrial dysfunction and ferroptosis, highlighting a direct pathogenic role of metabolic stress in PD progression¹⁷. Nevertheless, how metabolic stress reshapes circadian transcriptional programs and epigenetic regulation in PD remains poorly understood. Genetic studies have linked clock gene variants, particularly Per3, to PD risk^18,19, suggesting a connection between circadian gene regulation and disease pathogenesis. However, the molecular mechanisms that couple metabolic stress to circadian gene expression in the context of PD remain largely unknown, especially at the epigenetic level.

AMP-activated protein kinase (AMPK), a key metabolic sensor, has been implicated in the regulation of circadian rhythms by linking cellular energy states to the biological clock²⁰. AMPK activation occurs in response to energy stress, such as glucose deprivation, and regulates cellular metabolism by modulating key signaling pathways^21–23. It has been shown that AMPK interacts with circadian clock components, influencing the expression of clock genes and their downstream targets^24,25. However, the precise molecular mechanisms by which AMPK mediates circadian regulation remain elusive. Recent evidence suggests that epigenetic modifications, particularly DNA methylation, play a significant role in regulating circadian gene expression^26,27. Among the enzymes involved in DNA demethylation, Ten-eleven translocation (TET) proteins are critical for the conversion of 5-methylcytosine to 5-hydroxymethylcytosine (5hmC), a modification associated with active gene transcription^28,29. Despite the known role of TET proteins in gene regulation, their contribution to circadian rhythm modulation, particularly under conditions of metabolic stress, has not been fully explored.

In recent years, studies have focused on how fluctuations in glucose availability can affect circadian gene expression and the associated epigenetic modifications. Emerging evidence suggests that glucose deprivation can modulate the expression of circadian genes through AMPK-mediated pathways^30,31. However, the involvement of TET enzymes in this process has not been extensively investigated. TET proteins have been implicated in various biological processes, including development, differentiation, and metabolism^32,33, but their role in glucose-induced circadian rhythm regulation remains underexplored. It is hypothesized that AMPK activation during glucose deficiency could upregulate Tet1/2 expression, leading to changes in DNA methylation patterns that affect circadian gene expression, but direct evidence linking these pathways is lacking. In PD, Tet2 expression and 5hmC levels are altered in vulnerable brain regions^34,35, yet whether TET proteins mediate metabolic stress‑induced circadian changes has not been explored.

Here, we investigate how glucose fluctuations regulate circadian transcription through the AMPK–TET–5hmC axis. Using genome‑wide 5hmC mapping, CRISPR‑mediated knockouts, and circadian synchronization models, we show that low glucose activates AMPK, upregulates TET2 transcription, and drives dynamic 5hmC remodeling, thereby selectively controlling the rhythmic expression of Per3 independently of core CLOCK/BMAL1 machinery. While our mechanistic dissection uses mouse embryonic fibroblasts and HEK293T cells, the identified pathway is likely operational in PD‑relevant cell types, including dopaminergic neurons and microglia, which exhibit both metabolic vulnerability and circadian gene expression (Scheme 1). These findings establish a glucose‑responsive AMPK–TET2–5hmC axis that couples metabolic stress to circadian transcription, providing a mechanistic entry point for understanding circadian dysfunction in Parkinson’s disease.

Scheme 1. Schematic summary of the glucose-responsive AMPK-TET2-5hmC pathway regulating circadian transcription.

Glucose fluctuations activate AMPK, which induces TET2 expression and promotes 5hmC remodeling at clock-associated loci, exemplified by Per3, thereby shaping rhythmic transcriptional outputs. These epigenetic changes oscillate in synchrony with circadian gene expression and are reversible upon glucose restoration. CRISPR-mediated AMPK knockout and rescue experiments validate AMPK as an essential upstream regulator of TET2-dependent 5hmC dynamics. This pathway operates largely independently of the core CLOCK/BMAL1 oscillator and provides a mechanistic link between metabolic stress and circadian dysregulation in Parkinson’s disease.

Results

Systemic glucose fluctuations drive TET2-dependent 5hmC rhythmicity and disrupted in PD

Circadian disruption and metabolic imbalance are increasingly recognized as key features of Parkinson’s disease (PD), yet how systemic metabolic cues are coupled to epigenetic regulation in the brain remains unclear. To address this, we first established a subacute PD mouse model using MPTP administration. Successful model induction was confirmed by behavioral deficits, including impaired motor coordination and reduced locomotor activity, as well as significant loss of tyrosine hydroxylase (TH)-positive dopaminergic neurons in the substantia nigra and striatum (Fig. S1A–F).

Based on this validated model, we performed time-resolved analyses of 5-hydroxymethylcytosine (5hmC) and TET family expression in metabolically relevant brain regions of wild-type (WT) and PD mice across a 20-h circadian cycle (2-h intervals; Fig. 1A, K). Under basal ad libitum feeding conditions, the medial basal hypothalamus (MBH) exhibited significantly higher global 5hmC levels than the prefrontal cortex (PFC) (Fig. 1B), indicating region-specific epigenetic activity. Time-series analysis revealed robust circadian oscillations of 5hmC in both regions (Fig. 1C), with stronger rhythmicity in MBH (JTK_Cycle: p < 0.001) than in PFC (p = 0.007), suggesting enhanced epigenetic plasticity in metabolically active hypothalamic nuclei.

Fig. 1. Systemic glucose fluctuations drive TET2-dependent 5hmC rhythmicity and are disrupted in Parkinson’s disease.

A Schematic of circadian sampling in wild-type (WT) mice maintained on a 12 h light/12 h dark cycle (ZT0, lights on). Medial basal hypothalamus (MBH) and prefrontal cortex (PFC) were collected every 2 h from ZT0 to ZT20 (n = 3). B, C Under ad libitum feeding, global 5hmC levels were assessed by dot blot and their circadian oscillations analyzed in MBH and PFC. Data in (C) were normalized to the mean across time points; rhythmicity was determined by JTK_Cycle (PFC, p = 0.007; MBH, p < 0.001). Uncropped blots are shown in Fig. S3. D Diurnal blood glucose levels in WT mice (n = 3). E qRT-PCR analysis of AMPK mRNA in MBH and PFC showed modest but non-rhythmic variation. F–H Circadian expression of Tet1, Tet2, and Tet3 in WT mice (n = 3). Only Tet2 was rhythmic (JTK_Cycle: PFC, p < 0.001; MBH, p = 0.002) and its pattern paralleled 5hmC oscillation. I, J qRT-PCR in WT mouse embryonic fibroblasts and HEK293T cells cultured under high-glucose (HG, 4.5 g/L) or low-glucose (LG, 1.0 g/L) conditions (n = 3), showing selective glucose responsiveness of TET2. K Schematic of circadian sampling in MPTP-induced PD mice under the same conditions as in (A) (n = 3). L, M Basal global 5hmC levels and circadian 5hmC oscillations in MBH and PFC of PD mice; uncropped blots are shown in Fig. S3. N Diurnal blood glucose in PD mice (n = 3 per time point) showed reduced-amplitude rhythmicity. O AMPK mRNA in PD mice remained non-rhythmic. P–R Tet2 oscillation was dampened in MBH and nearly lost in PFC, whereas Tet1 and Tet3 were unchanged. S, T qRT-PCR in SH-SY5Y cells and LUHMES dopaminergic neurons cultured under HG or LG conditions (n = 3) showed elevated basal TET2 and reduced glucose responsiveness. Data are mean ± SEM. Statistical significance was determined by two-way ANOVA with Tukey’s post hoc test unless otherwise indicated. ^*p < 0.05, ^**p < 0.01,^***p < 0.001.

We next examined systemic metabolic dynamics. Blood glucose levels displayed a clear diurnal pattern, peaking during the active phase and reaching a nadir during the resting phase (Fig. 1D). However, correlation analysis at zero lag revealed only a weak association between glucose and 5hmC dynamics, suggesting that their relationship is not temporally synchronized. To further resolve this, lagged correlation analysis was performed, which revealed that glucose fluctuations preceded 5hmC oscillations in the MBH with an estimated delay of approximately 8–10 h, indicating a temporally offset metabolic–epigenetic coupling. Given the central role of energy sensing pathways, we next assessed AMPK expression. qRT–PCR analysis showed that AMPK mRNA levels exhibited only modest fluctuations without a robust circadian rhythm in either MBH or PFC (Fig. 1E), consistent with its predominant regulation at the post-translational level.

We then examined the expression of TET family DNA demethylases. Among Tet1, Tet2, and Tet3, only Tet2 displayed significant circadian oscillations in both MBH and PFC (Fig. 1F–H), whereas Tet1 and Tet3 remained non-rhythmic. Notably, the temporal profile of Tet2 expression closely paralleled that of 5hmC, supporting a role for TET2 as a key mediator of rhythmic epigenetic remodeling. To determine whether glucose availability directly regulates TET expression, we performed in vitro experiments. In WT-MEFs, low-glucose conditions selectively increased Tet2 expression, whereas Tet1 and Tet3 remained unchanged (Fig. 1I). Similarly, in HEK293T cells, only TET2—but not TET1—responded to glucose fluctuations (Fig. 1J), confirming the selective glucose sensitivity of TET2.

We next investigated whether this metabolic–epigenetic coupling is disrupted in PD. In MPTP-treated mice (Fig. 1K), basal 5hmC levels remained higher in MBH than in PFC (Fig. 1L); however, circadian oscillations were markedly attenuated in both regions (Fig. 1M), indicating impaired epigenetic rhythmicity. Consistently, PD mice exhibited a blunted diurnal glucose rhythm, characterized by reduced amplitude and a flattened temporal profile (Fig. 1N), reflecting systemic metabolic dysregulation. AMPK mRNA remained arrhythmic in PD mice (Fig. 1O), similar to WT controls. In contrast, Tet2 oscillation was significantly dampened in the MBH and nearly abolished in the PFC, whereas Tet1 and Tet3 remained unchanged (Fig. 1P–R), indicating selective disruption of TET2 rhythmicity.

To further validate these findings in neuronal systems, we examined glucose-dependent TET expression in SH-SY5Y and LUHMES cells. In contrast to MEFs, PD-like neuronal cells exhibited elevated basal TET2 levels but a markedly attenuated response to glucose fluctuations, while TET1 and TET3 remained unresponsive (Fig. 1S, T).

Collectively, these findings demonstrate that systemic glucose fluctuations drive TET2-dependent 5hmC rhythmicity in hypothalamic regions and that this metabolic–epigenetic coupling is selectively disrupted in PD.

Glucose-dependent regulation of TET2 mediates 5hmC remodeling via metabolic signaling pathways

Building upon our in vivo findings that systemic glucose fluctuations are associated with TET2-dependent 5hmC rhythmicity, we next sought to dissect the underlying molecular mechanisms using a controlled in vitro system. Primary mouse embryonic fibroblasts (MEFs) harboring floxed Tet1/2/3 alleles were isolated and validated (Fig. 2A–C). providing a tractable model to investigate glucose-dependent epigenetic regulation independent of systemic influences.

Fig. 2. Glucose-dependent regulation of TET2 drives 5hmC dynamics via metabolic signaling pathways.

A Experimental workflow for the isolation and validation of primary mouse embryonic fibroblasts (MEFs) harboring floxed Tet1/2/3 alleles. B, C Lentiviral vector construction and PCR-based genotyping confirming the Tet1/2/3^fl/fl genotype in primary MEFs. D qRT–PCR analysis of Tet1, Tet2, and Tet3 mRNA expression in MEFs cultured under low-glucose (LG, 1.0 g/L) and high-glucose (HG, 4.5 g/L) conditions (n = 3). Low-glucose conditions selectively induce Tet2 expression, with only modest changes observed for Tet1 and no significant change for Tet3. E qRT–PCR analysis of Tet1 and Tet2 mRNA expression in HEK293T cells ectopically expressing Myc-TET1 or Flag-TET2 under HG and LG conditions (n = 3), confirming preferential glucose sensitivity of TET2. F qRT–PCR analysis of Tet1, Tet2, and AMPK mRNA levels during glucose transition (sustained HG, sustained LG, and LG → HG switch). Tet2 expression dynamically responds to glucose availability, whereas AMPK mRNA levels remain largely unchanged. G Immunoblot analysis of TET1, TET2, and AMPK phosphorylation (p-AMPK/AMPK) under HG, LG, and LG → HG conditions, showing glucose-dependent regulation of TET2 protein and AMPK activation. Uncropped original western blot images were seen in Fig. S3. H Densitometric quantification of protein levels normalized to β-actin or total AMPK (n = 3). I, J Correlation analysis between MAPK pathway activation (p-ERK/ERK) and TET1/2 protein levels, suggesting potential involvement of MAPK signaling in glucose-mediated TET2 regulation. K Dot blot analysis of global genomic 5hmC levels in MEFs under indicated glucose conditions (n = 3). Uncropped original dot blot images were seen in Fig. S3. L Quantification of normalized chemiluminescence signals demonstrating glucose-dependent changes in 5hmC abundance (n = 3). Data are presented as mean ± SEM. Statistical significance was determined by two-way ANOVA with Tukey’s post hoc test. ^*p < 0.05, ^**p < 0.01,^***p < 0.001.

We first examined whether glucose availability directly regulates TET expression at the transcriptional level. qRT–PCR analysis revealed that low-glucose (LG) conditions significantly increased Tet2 expression, whereas Tet1 exhibited only modest changes and Tet3 remained largely unchanged (Fig. 2D). This selective responsiveness of Tet2 is consistent with our in vivo observations and suggests that TET2 is a primary mediator of metabolic–epigenetic coupling.

To further validate this specificity, we assessed TET expression in HEK293T cells ectopically expressing Myc-TET1 or Flag-TET2. Under varying glucose conditions, TET2—but not TET1—displayed a robust transcriptional response to glucose availability (Fig. 2E), confirming the preferential glucose sensitivity of TET2.

We next investigated the dynamic response of TET expression during glucose transitions. Upon switching from sustained low-glucose to high-glucose conditions (LG → HG), Tet2 mRNA levels showed a rapid but partial reversal, whereas Tet1 remained relatively stable (Fig. 2F). Notably, AMPK mRNA levels remained largely unchanged across conditions, consistent with the notion that AMPK is primarily regulated at the post-translational level rather than through transcriptional control.

At the protein level, immunoblot analysis revealed that TET2 abundance was dynamically regulated by glucose availability and closely paralleled changes in AMPK activation, as indicated by increased p-AMPK/AMPK under low-glucose conditions (Fig. 2G). Glucose restoration attenuated both AMPK activation and TET2 expression. Quantitative analysis confirmed these trends (Fig. 2H), supporting a link between metabolic sensing and TET2 regulation.

Given that MAPK signaling has been implicated in cellular responses to metabolic stress, we next explored its potential involvement. Correlation analysis revealed a positive association between ERK phosphorylation (p-ERK/ERK) and TET2 protein levels (Fig. 2I, J), suggesting that MAPK signaling may contribute to glucose-dependent regulation of TET2.

Finally, we assessed the functional consequences of these molecular changes on epigenetic modification. Dot blot analysis demonstrated that global 5hmC levels were significantly increased under low-glucose conditions and decreased upon glucose restoration (Fig. 2K, L), mirroring the dynamic changes in TET2 expression.

Collectively, these results demonstrate that glucose availability regulates 5hmC dynamics through a TET2-dependent mechanism, coordinated by metabolic signaling pathways including AMPK activation and potentially MAPK signaling.

AMPK is required for glucose-dependent regulation of TET2 and 5hmC dynamics

To determine whether AMPK directly mediates glucose-dependent regulation of TET2, we generated AMPKα1/α2 double-knockout (AKO) MEFs using CRISPR/Cas9. Efficient ablation of AMPK signaling was confirmed by the loss of both AMPKα protein and Thr172 phosphorylation under high-glucose (HG), low-glucose (LG), and AICAR stimulation conditions (Fig. 3A, B).

Fig. 3. AMPK is required for glucose-dependent regulation of TET2.

A Immunoblot analysis of total AMPK and phosphorylated AMPK (p-AMPK, Thr172) in wild-type (WT) and AMPKα1/α2 double-knockout (AKO) MEFs under high-glucose (HG), low-glucose (LG), and AICAR treatment conditions. Uncropped original western blot images were seen in Fig. S3. B Densitometric quantification of p-AMPK/AMPK ratios normalized to β-actin (n = 3). Time-course qRT–PCR analysis of Tet1 (C) and Tet2 (D) mRNA expression in WT and AKO MEFs under sustained HG, sustained LG, and glucose-switch conditions (LG → HG). qRT–PCR analysis of Tet1 (E) and Tet2 (F) mRNA expression following pharmacological modulation of AMPK activity, including AICAR-mediated activation and Compound C-mediated inhibition (n = 3). Data are presented as mean ± SEM. Statistical significance was determined by two-way or three-way ANOVA with Tukey’s multiple comparisons test.

In wild-type (WT) MEFs, LG exposure robustly increased AMPK activation, as indicated by elevated p-AMPK/AMPK levels, which was recapitulated by pharmacological activation using AICAR (Fig. 3A, B). In contrast, this response was completely abolished in AKO cells, confirming successful disruption of AMPK-dependent metabolic sensing.

We next examined whether AMPK is required for glucose-dependent regulation of TET expression. Time-course qRT–PCR analysis revealed that LG treatment induced Tet2 mRNA expression in WT cells, whereas this induction was largely absent in AKO cells (Fig. 3C, D). Consistent with our previous observations, Tet1 showed only modest changes, further supporting a predominant role of TET2 in metabolic responsiveness.

Pharmacological activation of AMPK using AICAR similarly increased Tet2 expression in WT cells but failed to do so in AKO cells (Fig. 3E, F), indicating that AMPK activation is both necessary and sufficient for TET2 induction under metabolic stress.

To further assess the dynamic nature of this regulation, we performed glucose-switch experiments. Restoration from LG to HG conditions led to a rapid attenuation of Tet2 expression in WT cells, whereas AKO cells remained unresponsive to glucose fluctuations (Fig. 3C, D), demonstrating that AMPK is essential for the reversibility of this process. Consistent with these findings, pharmacological inhibition of AMPK using Compound C effectively blocked LG-induced Tet2 upregulation (Fig. 3E, F), further establishing a causal role of AMPK in mediating glucose-dependent TET2 regulation.

Together, these results demonstrate that AMPK functions as a critical metabolic sensor that links extracellular glucose availability to TET2-dependent epigenetic remodeling.

AMPKα coordinates glucose-dependent hydroxymethylation dynamics

Quantitative dot blot analysis of genomic 5hmC levels—performed at 25 ng and 50 ng DNA inputs to assess dose-dependent signal linearity—revealed no baseline differences between AMPKα1/α2 knockout (AKO) and wild-type (WT) MEFs under high glucose (HG: 1.02 ± 0.11 vs. AKO: 0.98 ± 0.09, p = 0.327), confirming AMPKα is dispensable for maintaining steady-state hydroxymethylation (Fig. 4A).

Fig. 4. Genetic disruptions of AMPK signaling impairs glucose-mediated 5hmC regulation.

Dot blooting quantification of global 5hmC levels in: A WT MEFs and B AMPKα-KO (AKO) MEFs under three metabolic regimes: High-glucose (HG, 25 mM) baseline, Low-glucose (LG, 5.5 mM) challenge (24 h) HG with 2 mM AICAR co-treatment (24 h). Uncropped original dot blot images were seen in Fig. S4. C, D Normalized 5hmC chemiluminescence signals showing glucose transition effects (LG → HG switch). Data represent mean ± SEM (n = 3). ^**p < 0.01, ^***p < 0.001 by three-way ANOVA with Tukey’s multiple comparisons.

WT MEFs exhibited dynamic glucose-sensitive 5hmC remodeling. Low glucose (LG) treatment for 24 h increased 5hmC signals by 1.8 ± 0.2-fold (p = 0.005) and 2.1 ± 0.3-fold (p = 0.001) at 25 ng and 50 ng DNA loads, respectively. AMPK agonist AICAR (12-h treatment) partially recapitulated this effect, inducing 1.4 ± 0.1-fold (p = 0.042) and 1.7 ± 0.2-fold (p = 0.007) increases at corresponding DNA inputs (Fig. 4B, C). These dual-load measurements confirmed signal robustness across detection ranges. AKO MEFs lacked metabolic stress responsiveness: LG caused no significant 5hmC changes (25 ng: 1.1 ± 0.2-fold, p = 0.215; 50 ng: 1.0 ± 0.1-fold, p = 0.886), and AICAR failed to induce hydroxymethylation (25 ng: 0.9 ± 0.1-fold; 50 ng: 1.0 ± 0.2-fold, p > 0.05) (Fig. 4D), mirroring their TET expression defects.

TET1/2 deficiency disrupts rhythmic oscillation of core clock components

To investigate glucose fluctuation-dependent modulation of TET-mediated epigenetic-circadian crosstalk, we first established a synchronized circadian system in vitro. Cultured cells were subjected to serum shock, a well-characterized protocol that transiently resets cellular circadian oscillators by synchronizing individual cellular clocks through high-concentration serum exposure, thereby enabling robust detection of intrinsic circadian rhythmicity in gene expression³⁶. Using this approach, Tet1/2/3^fl/fl MEFs were synchronized (Fig. 5A). Time-course qPCR revealed robust circadian oscillations of Tet2 mRNA (p < 0.01) in WT-MEFs, whereas Tet1 exhibited minimal rhythmicity (p = 0.213). Cre-mediated triple knockout reduced baseline Tet1/2/3 expression by 87–92% (p < 0.001) and abolished Tet2 transcriptional rhythms (Fig. 5B, C), indicating that Tet2 rhythmicity is independent of its basal expression level.

Fig. 5. Genetic deletions of Tet genes impair serum shock-induced circadian rhythmicity in MEFs.

A Serum shock protocol: MEFs synchronized with 50% horse serum (2 h) followed by circadian sampling at 4-h intervals over 24 h. ZT0 marks serum withdrawal; gray shading indicates subjective night. Circadian profiling of Tet1 (B) and Tet2 (C) mRNA in Tet-KO (Cre-recombined) vs. control (GFP) MEFs (n = 3). qRT-PCR analysis of core clock genes Per3 (D) and Bmal1 (E) (n = 3). qRT-PCR analysis of core clock genes: Per1 (F), Per2 (G), Cry1 (H), Clock (I). Data represent mean ± SEM. ^*p < 0.05 by JTK_CYCLE analysis with Benjamini-Hochberg correction.

Under 12/12 h light-dark cycles, WT-MEFs maintained stable Per3 phase synchronization the cycles. In contrast, TET-KO cells displayed selective circadian disruption: Per3 oscillations exhibited severe amplitude attenuation (62 ± 8%, p = 0.004), phase delay (4.0 ± 0.7 h, p = 0.012), and reduced peak expression (2.1 ± 0.3-fold, p = 0.007) (Fig. 5D-E). Other core clock genes (Per1/2, Cry1, Clock, Bmal1) maintained stable oscillations, with amplitude changes <15% (p > 0.1) and phase shifts <1 h (p > 0.2) (Fig. 5F–I). These data establish TET proteins as critical regulators of Per3 phase stability within the circadian network.

TET1/2 deficiency disrupts glucose-coupled circadian oscillations

To dissect TET-dependent circadian regulation uncoupled from serum-mediated signaling, we employed a medium-change synchronization protocols to couple metabolic signals with circadian oscillations³⁷. Serum-shocked MEFs were maintained in low glucose for 12 h prior to rhythm induction. High-resolution time-series monitoring (2-h intervals over 36 h) revealed glucose-dependent circadian dynamics (Fig. 6A). Acute LG exposure (0–12 h) increased Tet2 oscillation amplitude by 1.8 ± 0.3-fold (p = 0.004) with a phase advance of 3.0 ± 0.5-h (p < 0.05). During the metabolic memory phase (12–24 h), Tet2 amplitude remained elevated (1.5 ± 0.2-fold, p < 0.05) despite HG restoration, accompanied by persistent phase shifts (≥8 h, p < 0.05). Rhythmic parameters normalized by 24–36 h (amplitude ≤1.1-fold, phase shift ≤1 h, p > 0.1) (Fig. 6B, C). phase synchronization index (PSI)

Fig. 6. Metabolic reprogramming reshapes circadian dynamics through TET-dependent mechanisms.

A Experimental paradigm for circadian synchronization under glucose modulation: baseline synchronization in HG (25 mM), metabolic challenge (LG, 5.5 mM) initiated at ZT0, and longitudinal sampling every 4 h for 36 h. gray shading denotes subjective night. Circadian transcriptional rhythms of Tet1 (B) and Tet2 (C) in glucose-modulated Tet-KO (Cre-recombined) vs. control MEFs (JTK_Cycle analysis, n = 3). Phase shift analysis of core clock genes Per3 (D) and Bmal1 (E) under sustained LG (red) vs. HG (blue), calculated by continuous wavelet transform (CWT) (n = 3; p < 0.01 for LG-induced phase delay). Data represent mean ± SEM. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001 by mixed-effects model with Šidák’s multiple comparisons test.

Bivariate wavelet coherence analysis revealed glucose-Tet2-Per3 coupling. In WT cells, LG synchronized Tet2 and Per3 rhythms (cross-wavelet coherence: 0.87 ± 0.05), enhancing Per3 amplitude (2.1 ± 0.3-fold, p < 0.05) and phase synchronization (phase synchronization index [PSI]: 0.82 ± 0.05 to 0.94 ± 0.03, p < 0.01). In contrast, Tet-knockout (KO) cells exhibited reduced basal Per3 expression (53 ± 6%, p < 0.001) and abolished rhythmic plasticity (amplitude ≤1.1-fold, PSI = 0.79 ± 0.06, p > 0.1) (Fig. 6D, E). Core clock analysis demonstrated pathway specificity: Bmal1 maintained stability (amplitude change <15%, p = 0.215; phase shift <1 h, p = 0.307), while Per3 responsiveness to LG strictly depended on Tet2 (Δexpression ≤8%, p = 0.528) (Fig. S2A–D).

Collectively, these findings designate Tet2 as a glucose-sensitive rheostat for Per3 node dynamics, selectively reshaping clock network hierarchy without global circadian integrity.

Discussions

In this study, we identify a metabolic–epigenetic regulatory axis in which glucose availability modulates 5hmC dynamics through AMPK-dependent regulation of TET2. Our findings extend the current understanding of AMPK beyond its canonical role as a metabolic sensor by demonstrating its involvement in coordinating epigenetic remodeling with metabolic fluctuations. Specifically, we show that glucose deprivation activates AMPK, which in turn selectively upregulates TET2 expression and drives global 5hmC changes, thereby linking extracellular nutrient status to epigenetic regulation.

AMPK, a highly conserved metabolic sensor, plays pivotal roles in maintaining cellular energy homeostasis by regulating metabolic pathways, cell proliferation, and stress responses^25,38. Our study reveals that glucose deprivation activates AMPK to transcriptionally upregulate Tet1 and Tet2 expression, establishing a direct link between extracellular glucose levels and dynamic 5hmC remodeling. This finding aligns with recent reports demonstrating glucose concentration-dependent regulation of Tet2 expression in colorectal cancer cells³⁹, where hyperglycemic conditions (9.0 g/L glucose, 6-h exposure) suppressed Tet2 mRNA levels. The observed discrepancies in Tet1 regulation across studies likely stem from methodological variations, including differences in glucose concentrations (4.5 g/L vs. 9.0 g/L), treatment durations (24 h vs. 6 h), and cell type-specific regulatory mechanisms (MEFs vs. cancer cells).

Importantly, integrating our in vivo data, we demonstrate that TET2 exhibits circadian oscillations in metabolically active brain regions, closely paralleling 5hmC dynamics. While DNA methylation was traditionally considered relatively stable, accumulating evidence indicates that epigenetic modifications, including DNA methylation and hydroxymethylation, undergo circadian regulation⁴⁰. However, rhythmic expression of TET enzymes has been primarily reported in non-mammalian systems⁴¹. Our observation of TET2 rhythmicity in mammalian tissues therefore, expands the conceptual framework of epigenetic–circadian coupling and suggests that TET2 may act as a convergence node integrating metabolic and temporal signals. These findings are particularly relevant in the context of Parkinson’s disease, where metabolic dysfunction and circadian disruption are increasingly recognized as key pathological features. PD patients exhibit impaired cerebral glucose metabolism and altered energy-sensing pathways, including dysregulated AMPK signaling^42,43. In parallel, emerging studies have implicated epigenetic dysregulation, including altered DNA methylation and hydroxymethylation, in dopaminergic neuron vulnerability and disease progression^44–46. Based on our results, we propose that metabolic disturbances in PD may perturb the AMPK–TET2–5hmC axis, thereby contributing to aberrant epigenetic remodeling and circadian dysfunction. Although our mechanistic studies were performed in non-neuronal systems, previous evidence indicates that both AMPK and TET2 play critical roles in neuronal function and survival, supporting the potential relevance of this pathway in neurodegenerative contexts^47–49.

Despite these insights, several limitations should be acknowledged. First, our mechanistic investigations were primarily conducted in fibroblast-based models, and the relevance of the AMPK–TET2 axis in neurons, particularly dopaminergic neurons, remains to be validated. Future studies should extend these findings to neuronal systems to establish disease-specific relevance. Second, while we demonstrate global changes in 5hmC levels, the locus-specific genomic targets of TET2 remain undefined. High-resolution epigenomic approaches, such as hMeDIP-seq or CUT&Tag, will be necessary to identify specific regulatory regions and target genes involved in circadian and metabolic regulation. Third, although we observed disrupted rhythmicity and epigenetic alterations in PD mouse models, the causal contribution of the AMPK–TET2 pathway to PD pathogenesis has not yet been directly established. In vivo genetic or pharmacological modulation of this pathway will be essential to determine its functional role in disease progression.

Collectively, our study identifies a metabolic-epigenetic regulatory axis wherein glucose deprivation activates AMPK to selectively induce TET2 expression and drive dynamic 5hmC remodeling, thereby linking metabolic state to circadian epigenetic regulation through the AMPK–TET2–5hmC axis. These findings uncover a previously unappreciated mechanism of metabolic–circadian crosstalk and provide a conceptual framework for understanding how metabolic stress may drive epigenetic dysregulation in neurodegenerative diseases, highlighting TET2 as a key mediator through which metabolic dysfunction contributes to circadian disruption in Parkinson’s disease.

Methods

Animals and Circadian sampling

All animal procedures were approved by the Institutional Animal Care and Use Committees of Wenzhou Medical University and carried out according to the “Guidelines and Regulations for the Use and Care of Animals of the Review Board of Wenzhou Medical Laboratory Animal Center (SYXK(Zhe)-2021-0020). Male C57BL/6J mice (8–10 weeks old, Jackson Laboratory) were housed under a 12 h light:12 h dark cycle (LD) with ad libitum access to food and water. For circadian 5hmC profiling, mice (n = 2 per timepoint) were sacrificed at 2-h intervals over a 20-h cycle (Zeitgeber Time [ZT] 0–20), with ZT0 defined as lights-on. Brains were rapidly dissected, and hypothalamic preoptic area (PFC) and medial basal hypothalamus (MBH) were micro-punched in ice-cold PBS, flash-frozen, and stored at −80 °C.

MPTP-induced subacute Parkinson’s disease model

Mice received intraperitoneal injections of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP, 30 mg/kg, Sigma-Aldrich) once daily for 7 consecutive days. Control mice received an equal volume of saline. Behavioral tests were performed at day 7 after the last MPTP injection.

Behavioral tests

All behavioral tests were conducted during the light phase (9:00 AM–2:00 PM) by an investigator blinded to group allocation. Mice were habituated to the testing room for at least 1 h before each test.

Rotarod test

Mice were placed on a rotating rod (4–40 rpm over 5 min, Ugo Basile). The latency to fall was recorded. Each mouse underwent three trials, and the average latency was calculated.

Open field test

Mice were placed in a square arena (40 × 40 × 40 cm) and allowed to explore freely for 10 min. Total distance traveled was recorded and analyzed using EthoVision XT (Noldus).

Pole test

A rough-surfaced vertical pole (50 cm height, 1 cm diameter) was placed in the home cage. Mice were placed head-up near the top, and the time to turn completely downward (turn time) and to descend to the base (descent time) were recorded.

Detection of genomic 5-hydroxymethylcytosine (5hmC)

Global 5hmC levels were quantified by dot blot hybridization. Genomic DNA (1 μg) was denatured at 95 °C for 10 min, chilled on ice for 5 min, and mixed with an equal volume of 2 mol/L ammonium acetate to prepare 50 ng/μL and 25 ng/μL dilution series. Samples (1 μL) were spotted onto a 0.45 μm nylon membrane, air-dried for 5 min, and crosslinked under UV light (120 mJ/cm², 2 min). Membranes were blocked with 5% non-fat milk in TBST for 30 min, then incubated with 5hmC primary antibody (rabbit monoclonal, 1:1000 dilution) overnight at 4°C, followed by HRP-conjugated secondary antibody (goat anti-rabbit, 1:5000) for 1 h at room temperature. After three 10-min TBST washes, signals were detected by ECL chemiluminescence using an ImageQuant LAS 4000 system and quantified via ImageJ grayscale analysis.

Experimental models and cell culture

The human embryonic kidney 293T (HEK293T) cell line (ATCC CRL-3216) was utilized for lentiviral packaging and validation of exogenous protein expression. Primary mouse embryonic fibroblasts (MEFs) derived from C57BL/6J mice served as the core experimental models: wild-type (WT-MEF), AMPKα1/α2 double-knockout (AKO-MEF), and Tet1/2/3 conditional knockout (Tet-TKO-MEF) cells. AKO-MEF and WT-MEF were provided by Professor Shengcai Lin’s laboratory at the School of Life Sciences, Xiamen University. Tet-TKO-MEF were isolated from Tet1/2/3^f1/f1 embryos via primary culture.

To extend findings to neuronal systems, the human neuroblastoma cell line SH-SY5Y (ATCC CRL-2266) and the human dopaminergic neuronal precursor cell line LUHMES were also included. SH-SY5Y cells were maintained in DMEM/F12 medium supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin–streptomycin. LUHMES cells were cultured in advanced DMEM/F12 supplemented with N2 supplement, 1 mM dibutyryl cAMP, 2 ng/mL GDNF, and 1 μg/mL tetracycline to induce neuronal differentiation, as previously described.

All cells were maintained at 37 °C under 5% CO₂ in high-glucose DMEM (4.5 g/L glucose) supplemented with 10% fetal bovine serum (FBS; Gibco) and 100 U/mL penicillin-streptomycin. For glucose deprivation experiments, cells were seeded in 6-well plates (5 × 10⁵ cells/well) and switched to low-glucose DMEM (1.0 g/L glucose) after 24 h of attachment. Rescue groups received high-glucose medium after 12 h of low-glucose treatment. AMPK activation assays used 12-well plates (3 × 10⁵ cells/well), with cells treated with 1 mM AICAR in high-glucose medium for specified durations. Cell viability (>95% by trypan blue exclusion) and mycoplasma contamination (monthly PCR screening) were routinely monitored.

Isolation and culture of primary mouse embryonic fibroblasts

Pregnant mice at embryonic day 14 were euthanized by cervical dislocation. Uterine tissues were aseptically dissected along the ventral midline and transferred to a 10 cm dish containing 10 mL PBS for rinsing. Under sterile conditions, embryos were extracted from amniotic sacs using ophthalmic scissors and forceps, then placed in fresh PBS. Cephalic regions, limbs, and visceral organs were meticulously excised, and remaining trunk tissues were minced into 1 mm³ fragments with a surgical scalpel.

Tissue fragments were digested with 1 mL 0.25% trypsin-EDTA solution (pre-warmed to 37 °C) for 30 min at 37 °C. Digestion was terminated by adding 1 mL DMEM supplemented with 10% fetal bovine serum (FBS). The cell suspension was filtered through a 200-μm mesh, pelleted by centrifugation (1000 × g, 5 min), and resuspended in complete medium (DMEM with 10% FBS, 1% penicillin/streptomycin). Cells were seeded onto 0.1% gelatin-coated 10 cm dishes and cultured at 37 °C under 5% CO₂. Primary cells were maintained with medium replacement 24 h post-plating, followed by daily monitoring of morphology and confluency.

Immortalization of embryonic fibroblasts

The simian vacuolating virus 40 large T antigen (SV40 LT) was utilized to immortalize primary mouse embryonic fibroblasts (MEFs) by disrupting cell cycle regulation through inactivation of tumor suppressor proteins, including p53 and Rb. Lentiviral particles encoding SV40 LT and a puromycin resistance gene were generated by co-transfecting HEK293T cells (80% confluent in 6-well plates) with 2 μg pBabe-SV40-LargeT-puro and 2 μg PCL-eco packaging plasmid using calcium phosphate. Viral supernatants were collected 48 h post-transfection, filtered through 0.45 μm membranes, and stored at −80 °C.

For transduction, passage 3 (P3) primary MEFs were seeded at 3 × 10⁵ cells/well in 6-well plates and incubated with lentiviral supernatant (600 μL) supplemented with 8 μg/mL Polybrene to enhance transduction efficiency. Following 48 h of incubation, the medium was replaced with fresh complete DMEM, and cells were cultured to 90% confluency. To select stably transduced populations, cells were treated with 5 μg/mL puromycin for 72 h, with daily medium replacement to eliminate non-transduced cells. Surviving clones were expanded through 5–6 passages at a 1:5 split ratio and validated for immortalization by assessing proliferative capacity (population doubling time <24 h), mycoplasma-free status (PCR-negative), and SV40 LT expression (western blot). Functional validation included consistent morphological stability across passages and retained responsiveness to metabolic stimuli, confirming the establishment of a robust immortalized MEF line suitable for downstream experiments.

Plasmid construction and gene cloning

The lentiviral expression vector pBOBI cs2.0, containing a multiple cloning site (MCS) downstream of the CMV promoter, was engineered to express N-terminal Myc/Flag-tagged Tet1 and Tet2. The coding sequences (CDS) of mouse Tet1 (GenBank ID: NM_001305093.1) and Tet2 (GenBank ID: NM_001347457.1) were inserted into the MCS via XbaI/AgeI and XbaI/XhoI restriction sites, respectively. Recombinant plasmids were validated by Sanger sequencing (Tsingke Biotechnology) and restriction digestion (1% agarose gel electrophoresis).

For primer design, Tet1/2 CDS sequences were retrieved from the NCBI RefSeq database. Gene-specific primers (Table S1) were designed using Primer Premier 6.0, incorporating XbaI (5′-TCTAGA-3′), AgeI (5′-ACCGGT-3′), or XhoI (5′-CTCGAG-3′) restriction sites with protective bases at the 5′ termini. PCR-amplified products were gel-purified (AxyPrep DNA Gel Extraction Kit), cloned into the pMD19-T vector (Takara) via TA cloning, and transformed into DH5α competent cells for colony screening.

Recombinant plasmid assembly

Tet1/2 CDS regions were amplified from mouse cDNA using PCR conditions specified in Tables S2 and S3. Amplified fragments were verified by 1% agarose gel electrophoresis, purified, and quantified (NanoDrop 2000; A_260/280 = 1.8–2.0). The pBOBI cs2.0 vector and purified inserts were digested with XbaI/AgeI (Tet1) or XbaI/XhoI (Tet2) (Table S4) at 37 °C for 3 h. Digestion efficiency was confirmed by gel electrophoresis. Purified vector and insert fragments were ligated at a 3:1 molar ratio using T4 DNA ligase (16 °C, 12 h). Ligation products were transformed into DH5α competent cells (30 min on ice, 45 s heat shock at 42 °C, 2 min on ice) and plated on LB agar containing 100 μg/mL ampicillin (37 °C, 16 h). Positive clones were screened by colony PCR and restriction digestion, followed by Sanger sequencing (Tsingke Biotechnology) to confirm insert integrity.

Genomic DNA isolation

Genomic DNA was isolated using the cell DNA Isolation Mini Kit (Vazyme). Cells from 6-well plates were trypsinized, pelleted by centrifugation, and resuspended in PBS. Suspensions were treated with 10 μL RNase A and 20 μL Proteinase K for 30 min at room temperature. Lysates were mixed with Buffer GB, incubated at 65 °C for 30 min, and precipitated with absolute ethanol. DNA was purified via silica membrane columns (12,000 × g centrifugation), washed sequentially with 500 μL Wash Buffer A and 650 μL Wash Buffer B, and eluted in 50 μL pre-warmed (70 °C) Elution Buffer. DNA concentration and purity were measured by NanoDrop (A_260/280 = 1.8–2.0), with integrity verified by 0.8% agarose gel electrophoresis (dominant band >20 kb).

Circadian gene oscillation profiling

This study employed serum shock and medium synchronization protocols to induce cellular circadian rhythm synchronization. Time-course qPCR was utilized to monitor the dynamic expression of core clock genes. The methodological selection was based on the following rationale: The serum shock protocol enables rapid elimination of intercellular phase heterogeneity in circadian rhythms³⁶, while the medium synchronization approach mimics physiological stimuli to achieve low-interference rhythm induction⁵⁰. Combined application of these dual models allows cross-validation of the universality of metabolism-rhythm regulatory mechanisms.

Serum shock protocol

MEF cells (3 × 10⁵ cells per well) were seeded in 12-well plates. After 24 h (reaching 80% confluency), cells were stimulated for 2 h with high-glucose medium containing 50% horse serum (intensive synchronization phase), followed by replacement with serum-free medium to initiate circadian oscillations. The synchronization initiation timepoint was defined as Zeitgeber Time 0 (ZT0). Sequential sampling was performed at 0, 2, 4, 8, 12, 16, 20, and 24 h post-synchronization for transcriptional profiling of Bmal1, Per3, and other core clock genes.

Medium synchronization protocol

Cells were seeded identically. At 80% confluency, cultures were switched to serum-free, low-glucose medium for 12-h serum starvation (eliminating cell cycle interference), followed by reactivation with high-/low-glucose medium containing 10% FBS to simulate physiological metabolic fluctuations. Reactivation time was designated ZT0. Sampling occurred at 3-h intervals over 36 h (spanning 1.5 circadian cycles) to enhance the precision of rhythm parameter fitting.

Western blot analysis

Target protein expression was analyzed by standard SDS-PAGE. Cells were lysed in RIPA buffer containing protease and phosphatase inhibitors, followed by sonication (20% amplitude, 15 s) and centrifugation (16,000 × g, 15 min, 4 °C). Supernatants were mixed with 2× SDS loading buffer, denatured at 95 °C for 10 min, and loaded (20 μg total protein per lane) onto 8–12% gradient separating gels with 4% stacking gels. Electrophoresis was performed at 80 V until the bromophenol blue dye migrated into the separating gel, then continued at 120 V until target molecular weights were resolved.

Resolved proteins were transferred to methanol-activated PVDF membranes (0.45 μm) using a wet transfer system (25 mM Tris, 192 mM glycine, 20% methanol) at 250 mA for 90 min. Membranes were blocked with 5% non-fat milk in TBST for 30 min, incubated with primary antibodies (1:1000 dilution in TBST containing 5% BSA) overnight at 4 °C, washed three times with TBST (10 min each), and probed with HRP-conjugated secondary antibodies (1:5000 dilution) for 1 h at room temperature. After final washes, signals were detected using ECL substrate (ImageQuant LAS 4000).

RT-qPCR

Total RNA was extracted with TRIzol (Invitrogen) and reverse-transcribed using the High-Capacity cDNA Kit (Applied Biosystems). qPCR was performed on a QuantStudio 6 Flex system (Applied Biosystems) with PowerUp SYBR Green Master Mix. Primer sequences for Tet1, Tet2, Per3, Per1, Per2, Per3, Cry1, Bmal1, Clock, and Gapdh are listed in Supplementary Table S5. Data were normalized to Gapdh and analyzed via the ΔΔCt method.

Lagged correlation analysis

To evaluate the temporal relationship between blood glucose levels and 5hmC dynamics, Spearman’s correlation was calculated using time-series data at 2-h intervals. Mean values at each time point were used for analysis.

Lagged correlation analysis was performed by shifting the glucose time series relative to 5hmC levels (−12 h to +12 h, 2-h steps). Negative lags indicate that glucose changes precede 5hmC alterations. Correlations were calculated using overlapping time points only, and the lag with the highest correlation coefficient was defined as the optimal temporal offset.

Circadian rhythm profiling

Time-series data were collected at 2-h intervals over 36 h post-synchronization. Oscillation parameters (amplitude, phase) were derived using nonlinear least-squares sinusoidal fitting (y = A⋅sin(2πt/T + ϕ)+C; MATLAB R2022a, nlinfit). Phase shifts were calculated relative to baseline rhythms.

Wavelet coherence and phase synchronization analysis

Cross-wavelet coherence between Tet2 and Per3 oscillations was computed using Morlet continuous wavelet transforms (Python pywt). The phase synchronization index (PSI) was defined as:

$$\mathrm{PSI}=1/\mathrm{N}\sum N{e}^i\left[{\phantom{1}}_{\mathrm{Tet}2}^{\theta }{{\phantom{1}}^{\left(t\right)-\theta }}_{\mathrm{Per}3}{\phantom{1}}^{\left(t\right)}\right],$$

where θ_(t) denotes instantaneous phase angles derived from wavelet transforms. PSI significance was assessed via 1000 bootstrapped permutations.

Statistical analysis

Chemiluminescence signals from western and dot blots were quantified by grayscale intensity analysis using ImageJ software. Data are expressed as mean ± SEM. Between-group differences were evaluated using unpaired Student’s t-tests (two groups) or one-way ANOVA with Tukey’s post hoc test (multiple groups). Time-series data were analyzed using sinusoidal regression to quantify circadian amplitude and phase. Phase synchronization between Tet2 and Per3 oscillations was assessed via cross-wavelet coherence and the phase synchronization index (PSI)^51–53, calculated as the modulus of the mean phase coherence across time points. Statistical significance of PSI changes was determined by bootstrapping (1000 permutations). Multiple comparisons were adjusted using the Benjamini-Hochberg method. Nonlinear regression and circular statistics were performed in MATLAB or Python. Graphical visualization and final statistical tests were conducted in GraphPad Prism 9.0.0.

Author contributions

L.X.C.: Conceptualization, methodology, software, data curation, investigation, funding acquisition, and writing—review and editing. S.S.W., L.L.W. and F.Y.D.: Methodology, software, data curation, investigation, and writing—original. W.Y.W., J.J.M., Y.Q.L., W.J.H., R.H.T., C.Y.Z., C.G. and W.J.W.: Data curation, investigation. S.T., J.C. and Y.L.Z.: Conceptualization and writing—review and editing.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available as they consist of raw experimental data and original laboratory records that have not been deposited in a public repository, but are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41531-026-01447-z.