The compound MPP⁺ (1-methyl-4-phenylpyridinium) is a frequently used reference tool for eliciting experimental complex I deficiency (Beal, 2001; Richardson et al., 2005; Langston, 2017). To ensure selectivity and avoid toxicity of this drug, differentiated human LUHMES cells were treated with 10 µM MPP⁺ for 48 h, following closely related protocols (Krug et al., 2014; Smirnova et al., 2016). No overt signs of cytotoxicity were induced by this treatment regimen as reported (Krug et al., 2014; Baeken, 2020; Baeken et al., 2021); however, a moderate degree of microtubular reorganization was visible using immunocytochemistry (Supplementary Figure S1). Transcriptomic analysis of the MPP⁺-treated cultures indicated that the majority of RCC subunits were transcriptionally induced by complex I inhibition, consistent with a functional, compensatory response (Figure 1A; Supplementary Table S1) affecting many distant genomic loci (Supplementary Figure S2). Specifically, 28 out of 37 complex I genes were significantly upregulated (Figure 1D), as were 9 out of 10 complex III genes (Figure 1F), 11 out of 11 complex IV genes (Figure 1G), and 14 out of 16 complex V genes (Figure 1H). On average, global transcription of complex I genes was induced by 61%, complex III genes by 106%, complex IV genes by 123%, and complex V genes by 76%.

In inhibiting complex I electron flow from the aqueous NADH oxidation site to the ubiquinone binding site, MPP⁺ evokes two major biochemical effects: the loss of complex I as a proton pump contributing to ATP generation, and the production of superoxide radicals and other reactive oxygen species (ROS) (Fallon et al., 1997; Beal, 2001; Richardson et al., 2005). To distinguish which of these effects caused the transcriptional changes, we applied the mitochondrial antioxidant phenothiazine (PHT) to MPP⁺-treated cells. PHT is a nanomolar-activity antioxidant compound that permeates mitochondria and has shown high efficacy against mitochondrial ROS even in models where classic phenolic antioxidants generally fail (Hajieva et al., 2009; Mocko et al., 2010; Tapias et al., 2019). Concomitantly, PHT is not a two-electron reductant, meaning that it does not shuttle electrons from inhibited complex I to complex IV such as methylene blue (Atamna et al., 2008; Ohlow and Moosmann, 2011). Thus, it cannot ameliorate bioenergetic deficits, but merely acts as an ROS scavenger.

PHT treatment caused a significant, apparently uniform attenuation of the transcriptional effects of MPP⁺ (Figure 1B; Supplementary Table S1). More specifically, 26 out of 37 complex I genes were significantly downregulated compared to MPP⁺ only-treated cells (Figure 1D), as were 8 out of 10 complex III genes (Figure 1F), 11 out of 11 complex IV genes (Figure 1G), and 15 out of 16 complex V genes (Figure 1H). Across all subunits, transcription of complex I genes was reduced by 40%, complex III genes by 44%, complex IV genes by 51%, and complex V genes by 50%. Hence, approximately half of the regulatory effect of complex I inhibition was prevented by a low nanomolar dose of PHT. This may be considered a rather high fraction since kinetic scavenging systems can never intercept with all radicals if the radicals’ targets are present at the biologically typical, µM to mM concentrations (Moosmann and Hajieva, 2022).

Notably, mitochondrially encoded transcripts were severely reduced following MPP⁺ treatment (Figure 1I), consistent with earlier reports (Krug et al., 2014). Potentially, the engagement of the cells in preparatory mitochondrial DNA replication, which is known to be incompatible with mitochondrial transcription (Agaronyan et al., 2015), accounts for this effect. The mitochondrial transcripts were also unresponsive to PHT treatment like most of the nuclear encoded transcripts, indicating their disparate, redox-independent regulation. Interestingly, the three most prominently suppressed nuclear encoded RCC-related genes were either supernumerary, regulatory subunits involved in RCC assembly, like ATP5S (Belogrudov, 2009) and NDUFC2 (Alahmad et al., 2020), or they regulate supercomplex formation and have additional functions in the cytosol, like NDUFS1 (Elkholi et al., 2019). Evidently, assembly factors are less required as long as mitochondrial transcription and translation do not proceed.

To experimentally ascertain the supposed non-interference of PHT with cellular bioenergetics under the employed conditions, LUHMES cells treated identically as before were surveyed for a series of functional metabolic readouts. As shown in Figure 2, MPP⁺ treatment had no significant effect on cellular ATP levels, with or without PHT coadministration (Figure 2A). NAD⁺ levels were also unchanged (Figure 2B), while NADH levels were increased by complex I inhibition as expected (Figure 2C), resulting in a significant drop in the NAD⁺/NADH ratio relevant for metabolic flux (Figure 2D). PHT had no modulatory effect on any of these parameters. Likewise, PHT did not alter the anticipated effects of MPP⁺ on lactate production (Figure 2E) and the acidification of the medium (Figure 2F), and it had an only minor modulatory effect on 2 out of 5 of the detected carbonic anhydrases (Supplementary Figure S3A). Hence, PHT did not prevent the expectable induction of anaerobic glycolysis by MPP⁺, which was further evidenced by an accelerated disappearance of glucose in the medium (Figure 2G) and an induction of the neuron-specific glucose transporter GLUT3 (Figures 2H,I). After all, the drop in membrane potential typical for complex I inhibition by MPP⁺ (Hajieva et al., 2009; Krug et al., 2014) was also only partially attenuated by PHT (Figures 2J,K). In contrast, analysis of cellular ROS levels by means of two indicator dyes, the untargeted sensor CellROX, and the mitochondrion-specific sensor MitoROX, clearly demonstrated the prooxidant effect of MPP⁺ and the antioxidant effect of PHT (Figures 2J,L,M), as widely reported (Beal, 2001; Hajieva et al., 2009; Mocko et al., 2010; Tapias et al., 2019). In summary, PHT was incapable of substantially modulating the bioenergetic effects of complex I inhibition, but selectively blunted the ensuing ROS effect. Besides, it had a small enhancing effect on the NAD⁺/NADH ratio at baseline (Figure 2D).

Considering the rather uniform patterns of up- and downregulation of numerous functionally related genes (Figure 1) scattered all over the genome (Supplementary Figure S2), we hypothesized that epigenetic mechanisms might be involved. Indeed, reduced levels of DNA methylation have been reported in different models of PD and in patient-derived tissue (Desplats et al., 2011; Yang et al., 2017), and several key enzymes of the epigenetic machinery are known to be modulated by ROS (Chen and Shen, 2025). Consistently, MPP⁺ caused a substantial reduction of global 5-methylcytosine levels in the nucleus of differentiated LUHMES cells within 48 h (Figures 3A,B). This effect was abrogated by PHT cotreatment. The control drug 6-thioguanine, an established DNA methylation suppressor (Agrawal et al., 2018), elicited comparable effects at the employed standard concentration of 1 µM (Mender et al., 2015).

DNA methylation in humans is established by three DNA methyltransferases. DNMT1 is considered a maintenance protein requiring a hemi-methylated template, whereas DNMT3A and DNMT3B are de novo methyltransferases (Hermann et al., 2004). Immunocytochemical analysis indicated that these proteins were variably localized to the nucleus in differentiated LUHMES cells (between 30% and 60%) (Figures 3A,B). Notably yet, all three proteins exhibited a significant shift away from the nucleus upon MPP⁺ exposure (Figures 3A,B), as also reported for DNMT1 in PD (Desplats et al., 2011). These effects were not rescued by PHT cotreatment except in the case of DNMT3B, which was therefore putatively assigned to be responsible for the observed, reversible changes in DNA methylation. The compound 6-thioguanine had a similar effect on the localization of DNMT1 and DNMT3A as MPP⁺, but did not influence DNMT3B. The former observation is consistent with the reported induction of proteasomal degradation of DNMT1 by 6-thioguanine (Yuan et al., 2011), which may also affect DNMT3A (Huang et al., 2022), but potentially spares DNMT3B. The reduction of nuclear DNMT expression by MPP⁺ may have been exacerbated by the pronounced transcriptional downregulation of genes controlling the production of the methylation cofactor, S-adenosyl methionine (Supplementary Figures S3B, C). These transcriptional effects were also relatively PHT-sensitive.

DNMT3B expression in the Western blot showed two bands at ∼95 kDa (DNMT3BL) and ∼72 kDa (DNMT3BS) (Figures 3C,D). MPP⁺ caused a modest decline of both isoforms by approximately 30%, which was prevented by PHT treatment, but significantly only for DNMT3BS, the major form of DNMT3B in these cells. Exclusive PHT treatment also significantly increased DNMT3BS expression compared to the control group (Figures 3C,D) without rescuing DNMT3B at the mRNA level (Figure 3E). These data suggest that mitochondrial ROS lead to a loss of DNMT3B in the nucleus through a loss of protein expression, amplified by a shift of the remaining protein out of the nucleus. Like other epigenetic effector proteins (Park et al., 2016; Baeken et al., 2021; Baeken, 2024), DNMTs are now well established to be regulated by their controlled proteolytic degradation (Du et al., 2010; Yuan et al., 2011; Huang et al., 2022), which may also account for the current observations. Moreover, DNMT3A/3B have been shown in the past to be functionally affected by the thiol redox state of the cell (Chen et al., 2012). The PHT-indifferent baseline suppression of DNMT1 and DNMT3B transcription by MPP⁺ (Figure 3E) probably relates to another biochemical mechanism.

Histone lysine acetylation appears to be widely induced in idiopathic PD (Gebremedhin and Rademacher, 2016; Park et al., 2016; Toker et al., 2021) and in models of PD based on complex I inhibition (Park et al., 2016). Of the many sites that appear to be hyperacetylated in PD, including H2K15, H3K14, H3K18, H3K27 and H4K5, the site H3K14 may represent one of the most reproducible disease markers. No conflicting data as for H3K9 have been reported for H3K14 (Gebremedhin and Rademacher, 2016; Park et al., 2016), and it may also be more robustly induced than the related, more widely explored H3K9 site (Karmodiya et al., 2012). H3K14 has not been investigated after complex I inhibition.

Treatment of LUHMES cells with MPP⁺ as before evoked a significant increase in total lysine acetylation as well as H3K14 acetylation, which were entirely prevented by PHT cotreatment and thus redox-related (Figures 4A–D). The increase in H3K14 acetylation as per Western blot (∼600%) (Figure 4C) vastly exceeded the increase in total lysine acetylation (∼30%) (Figure 4D). Two minimally modified PHT derivatives that essentially lack antioxidant activity (Hajieva et al., 2009), namely, N-methylphenothiazine (MPHT) and N-acetylphenothiazine (APHT), were also tested in this assay because putatively involved histone deacetylases of the sirtuin (SIRT) family are highly sensitive to the levels of the somewhat related heteroaromatic molecule NAD⁺, containing nicotinamide (Anderson et al., 2017). MPHT and APHT were clearly less potent than PHT in their prevention of lysine acetylation (Figures 4C,D), corroborating that the PHT effect was caused by antioxidation.

To further define the cause of MPP⁺-induced hyperacetylation, we employed trichostatin A (TSA), an inhibitor of Zn²⁺-dependent histone deacetylases (HDACs), and selisistat (EX-527), a SIRT1 inhibitor with approximately 200-fold selectivity against SIRT2 and SIRT3 (Yoshida et al., 1990; Napper et al., 2005). Analysis of global lysine acetylation by immunocytochemistry (Figure 4B) and image quantification (Figure 4E) indicated that both agents caused an increase in lysine acetylation by approximately 50% and, thus, resembled MPP⁺ quantitatively. However, these increases were resistant to antioxidant PHT treatment as expected. Additive treatment with the inhibitors plus MPP⁺ gave a further (and PHT-reversible) increase only with TSA, but not with EX-527 (Figure 4E). Thus, the effect of MPP⁺ was concluded to be mediated by EX-527-inhibited deacetylases, but not by TSA-inhibited deacetylases.

Investigation of SIRT1 by Western blot (Figures 4F,G) indicated that this protein was significantly reduced upon MPP⁺-treatment in a PHT-reversible fashion; a corresponding result was obtained with a direct, fluorescent SIRT1 activity assay using cell lysates (Figure 4H). Here, the enzyme activity loss in the lysate was partially prevented by PHT, but unaltered by MPHT. In view of the unchanged transcription of SIRT1 (Figure 4I), these results demonstrate a reduction of the SIRT1 protein and of its enzyme activity due to a redox signal induced by mitochondrial complex I inhibition. The loss of this protein may be attributable to the redox-dependent induction of autophagic degradation of most sirtuins by MPP⁺ as recently reported (Baeken et al., 2021). Because SIRT1 enzyme activity is also known to be negatively affected by direct cysteine oxidation (Shao et al., 2014; Kalous et al., 2020), it is possible that both mechanisms operate in parallel.

Redox biological experiments in cell culture involve the general danger of returning exaggerated effects due to an unphysiologically oxidative environment (Kunath et al., 2020). Hence, we have tested the validity of some of the described molecular events in a mouse model of PD based on the same initiating event, namely, complex I inhibition by MPP⁺. Therefore, the pro-toxin MPTP (Beal, 2001; Langston, 2017) was administered intraperitoneally to 10-week-old male C57Bl/6J mice as sketched in the scheme in Figure 5. PHT was administered orally, including a roll-in period (for details, compare the Materials and Methods). Doses were chosen as to obtain an intermediate degree of toxicity only in the particularly vulnerable region, substantia nigra (SN), because widespread cell death could arguably give rise to strong secondary effects.

To verify the efficacy and selectivity of the employed MPTP dose and to probe any potential prevention by PHT, we analyzed the expression of tyrosine hydroxylase (TH), a commonly adopted marker of dopaminergic cell viability in vivo (Beal, 2001; Langston, 2017). Hence, midbrain slices were immunostained with antibodies against TH and counterstained with the chromatin dye Hoechst 33,258. The staining revealed a visible drop in the number of TH-positive cells in the SN, the primary region affected in PD (Figure 5A). The adjacent and larger ventral tegmental area (VTA) was essentially spared from toxicity as expected (Beal, 2001). Quantification by image analysis involving cell-cell border demarcation and counting of TH-positive cells yielded approximately 50% loss of SN neurons, with PHT treatment affording almost complete protection (Figure 5B), as already reported in related models (Mocko et al., 2010; Tapias et al., 2019). Motor performance experiments (“Rotarod”) done with the animals before sacrifice suggested a variable decline of capabilities from MPTP treatment including rescue by PHT, but statistical significance was not reached (Figure 5C).

Analysis of H3K14 acetylation in the SN demonstrated that MPTP treatment triggered this modification in vivo in a PHT-reversible fashion, with an increase of about 50% as per immunocytochemistry (Figures 6A,C) and about 150% as per Western blot (Figures 6B,D). Global lysine acetylation as per Western blot was raised by about 50% (Figure 6E). These results recapitulate the human cell culture outcome (Figure 4), but in an attenuated fashion. However, SIRT1 expression in the mouse in vivo was not decreased, but rather increased (Figure 6F). As PHT did not revert this effect, a redox-unrelated mechanism may have dominated here. Two additional SIRTs known to be rapidly degraded in vitro after MPP⁺ treatment (Baeken et al., 2021) were probed for control purposes, and indeed, SIRT3 and SIRT4 were suppressed by about 30% and 20%, respectively, and PHT-reversibly in vivo (Figures 6G,H). The origin of the differential behavior of SIRT1 is unclear at present. DNMT3B expression, in turn, was decreased in vivo by about 30% and rescued by PHT administration (Figure 6I), which recapitulates the in vitro situation. In summary, these results confirm the operability of a redox signal targeting epigenetic regulator proteins after complex I inhibition in vivo.

From the results presented so far, the hypothesis was derived that the redox-dependent effect of complex I inhibition towards the epigenetic effectors SIRT and DNMT3B may be part of a regulatory cycle to specifically maintain mitochondrial metabolic functioning. This hypothesis was tested by revisiting the transcriptomic changes that accompanied the induction of most RCC subunits as described in Figure 1. Out of 15,168 genes recovered in the RNA sequencing experiment, 3639 were upregulated, and 3531 were downregulated. A total of 826 genes coding for mitochondrially imported proteins (“MitoGenes”) as well as 78 RCC complex genes were selected for statistical analysis.

Mitogenes were found to be significantly more highly expressed after MPP⁺ treatment than average genes. Of these mitogenes, RCC complex genes were significantly more highly expressed than the other genes (Figure 7A). Hence, complex I inhibition caused a selective increase in the transcription of genes counteracting any potential malfunction of complex I and the respiratory chain. The selective targeting of mitogenes and RCC complex genes by MPP⁺ was also selectively affected by PHT treatment: coadministration of PHT lowered mitogenes significantly more than average genes, and RCC complex genes significantly more than other mitogenes (Figure 7B). Notably, the effect of PHT was strong enough to entirely abrogate the differential targeting of mitogenes and RCC complex genes by complex I inhibition (Figure 7C), which suggests that this differential targeting (Figure 7A) was essentially mediated by redox signaling, and not by other mito-nuclear signaling mechanisms (Quiros et al., 2016; Walker and Moraes, 2022).

The above results were compared with an arbitrary selection of other, individually assembled metabolic gene clusters from the Reactome database, as Gene Ontology (GO) term analysis had been inconclusive, failing to return any metabolic clusters (Supplementary Figure S4). Of nine investigated clusters, two were significantly induced by MPP⁺ treatment, namely, glycolysis and cholesterol biosynthesis (Figure 8A). PHT treatment had a significant suppressive effect on both MPP⁺-induced clusters and, in addition, on several other clusters such as the amino acid catabolism cluster (Figure 8B). In the direct comparison of the control transcriptome with the PHT/MPP⁺ transcriptome, none of the clusters attained significance (Figure 8C) except glycolysis and cholesterol biosynthesis, being modestly induced. These analyses demonstrate that redox signaling was responsible not only for the transcriptional targeting towards the mitochondrion, but also partially responsible for the targeting of glycolysis, a functionally connected metabolic pathway. The relationship between complex I inhibition and induced cholesterol biosynthesis is elusive, even if a substantial literature exists as regards the potential role of cholesterol in PD (Garcia-Sanz et al., 2021; Al-Kuraishy et al., 2024).

All cell culture media and supplements including Advanced Dulbecco’s Modified Eagle’s Medium (DMEM)/Ham’s F-12 (F12), Phosphate-Buffered Saline (PBS), N2 supplement, Hoechst 33,258, and 6-diamidino-2-phenylindole (DAPI) were obtained from Invitrogen. General laboratory chemicals and biochemicals, neurotoxins, phenothiazine and its derivatives were purchased from Sigma-Aldrich at the highest available purity unless otherwise specified.

LUHMES cells were kindly provided by Dr. Jürgen Winkler (Division of Molecular Neurology, University of Erlangen, Germany). The cells were grown and differentiated as described (Krug et al., 2014; Baeken et al., 2020). At the fourth day post differentiation, the medium was exchanged, and the cells were treated with different compounds at the following concentrations: MPP⁺: 10 μM; PHT: 20 nM; MPHT: 20 nM; APHT: 20 nM; 6-thioguanine: 1 μM; EX-257: 100 nM; TSA: 50 nM.

In vivo experiments in male C57Bl/6J mice (from Charles River) were performed by QPS Austria GmbH (Grambach, Austria). QPS Austria is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Animal care, housing and experimentation were approved by the Institutional Animal Care and Welfare Committee and complied to the animal welfare legislation of the Ministry of Science of the Austrian government.

Thirty mice aged 10 ± 2 weeks were allocated to three different treatment groups: a “vehicle group”, an “MPTP group”, and an “MPTP plus phenothiazine” group, involving repeated applications or injections of the two agents phenothiazine (10 mg/kg, ten doses in toto, vehicle DMSO:corn oil 1:50) and MPTP (20 mg/kg, four doses in toto, vehicle saline). The adopted treatment regimen is detailed in Figure 5 and has been described before (Baeken et al, 2020).

Animals from all groups were subjected to the Rotarod motor performance test on day 6 (1 day after the final PHT treatment). Prior to the first test session, the mice were habituated to the testing system until they were able to stay on the rotating rod at a constant speed of 2 rpm for approximately 1 minute. During testing, a single animal was exposed to the apparatus for three 180 s trials. The initial speed was increased from 2 rpm to 20 rpm during these 180 s. When the mouse fell, the session was over. The mean latency to fall in every single testing was determined.

After finishing the behavioral testing on day 6, the mice were deeply anesthetized by pentobarbital injection (600 mg/kg), transcardially perfused with saline, and dissected. The brains were hemisected, and the left hemisphere was subdivided into striatal tissue, midbrain and residual brain. These samples were rapidly frozen and kept at −80°C until further analysis. The right hemispheres were immersion-fixed in freshly prepared 4% paraformaldehyde in PBS for 1 h at RT. Thereafter, they were transferred to 15% sucrose in PBS until they sank down, indicating sufficient cryoprotection. The hemispheres were cryo-embedded in O.C.T. medium with dry ice-cooled isopentane and afterwards stored at −80°C.

O.C.T.-embedded brain hemispheres were cut into 10 µm slices with a cryostat. After washing with PBS, the sections were incubated with 3% BSA/0.1% Triton X-100 in PBS for 1 h at 4°C for blocking and permeabilization. After several washes with PBS, the slices were incubated overnight with primary antibodies at 4°C, repeatedly washed again, and secondary antibodies were applied for 2 h at RT. Following three more washing steps with PBS, 50 ng/mL Hoechst 33,258 was added for 15 min at RT before another PBS wash. Finally, the sections were mounted in anti-fading solution (polyvinyl alcohol/p-phenylendiamine) and stored at −80°C until microscopic evaluation.

The following antibodies were employed: anti-tyrosine hydroxylase (ab112, Abcam, 1:1000 in PBS containing 1% BSA), anti-H3K14ac (7627, Cell Signaling, 1:100), anti-SLC6A3 (DAT) (NBP2-22164, Novus Biologicals, 1:100). Secondary antibodies Cy3-anti-rabbit and Cy2-anti-mouse (Jackson Immunoresearch) were used at 1:1000 dilution for tyrosine hydroxylase, and at 1:200 dilution for SLC6A3 and H3K14ac. All sections were visualized using standard fluorescence microscopy.

LUHMES cells were plated at 7 × 10⁴ cells/cm² on glass cover slips with standard coating. The experiments were performed on day 6 post differentiation following published protocols (Baeken et al., 2020; Baeken et al., 2021). Cells were incubated with primary antibodies overnight at 4°C and thereupon incubated with secondary antibodies for 2 h at RT. DAPI (1 μg/mL) was used as nuclear counterstain.

The following primary antibodies were used: anti-acetylated lysine (9441, Cell Signaling, 1:200), anti-DNMT1 (ab13537, Abcam, 1:100), anti-DNMT3A (2160, Cell Signaling, 1:100), anti-DNMT3B (67,259, Cell Signaling, 1:100), anti-5-methylcytosine (28,692, Cell Signaling, 1:200), anti-TUBB (MAB1637, Millipore, 1:500). Cy3-anti-rabbit/Cy2-anti-mouse (Jackson Immunoresearch), 1:400 were used as secondary antibodies.

Slides generated through immunohistochemistry were recorded with an Axiovert 200 fluorescence microscope from Zeiss using blue, green and red filters and objectives for ×4 and ×10 magnifications. Immunocytochemistry slides were photographed with a laser scanning microscope (LSM) TCS SP5 from Leica (higher magnifications).

Immunohistochemistry pictures were evaluated using the open-source software ImageJ (imagej.net). Tyrosine hydroxylase (TH) staining was measured after the signal was „watershed“: this algorithm calculates signal maxima and thereby allocates cell borders. The cells were quantified using the “analyse particles” function from ImageJ that counts each continuous signal as one particle. Through this approach, cells can be evaluated regardless of the size of their cell body. For H3K14ac staining, total nuclei on the slide were quantified with DAPI, again using the “analyse particle” function of ImageJ. The same process was repeated for the H3K14ac staining, and the quotient H3K14ac/nucleus was calculated. Total acetyl-lysine and 5-methylcytosine levels were also quantified by dividing the intensity measured with ImageJ by the number of cells (nuclei). Stainings of DNMT1, DNMT3A and DNMT3B were evaluated by integration over all z-stacks of the signal intensity that colocalized with DAPI (assigned as “nuclear”) or did not colocalize with DAPI (assigned as “cytosolic”). The added signal intensity of both compartments was defined as 100%.

To collect total RNA, LUHMES cells were harvested in 500 µL TRI-Reagent (Sigma-Aldrich) following the manufacturer’s instructions.

Residual DNA was removed by addition of 2 µL DNAse I (#18047019 from ThermoFisher) and incubation at 37°C for 1 h. Afterwards, 500 µL 75% ethanol were added, and the samples were washed through centrifugation at 7,500 g for 5 min at 4°C. The supernatant was discarded, and the cleaned RNA pellet was reconstituted in 50 µL DEPC-treated H₂O.

The cleaned RNA was analyzed with a 2100 Bioanalyzer from Agilent and quantified using the Qubit dsDNA HS Assay Kit using a Qubit 2.0 Fluorometer (Life Technologies). All nine samples were pooled in equimolar ratio and sequenced in a NextSeq 500 High Output Flowcell with unique adapter sequences, SR for 1 × 84 cycles, plus 7 cycles for the index read.

Sample demultiplexing and FastQ file generation was performed using Illumina bcl2fastq v2.19.1, and overall sequence quality was assessed with FastQC v.0.11.5. Sequence reads were aligned to the human reference genome GRCh38 with annotation from Gencode release 25 using STAR v.2.5.2b with parameters “--outFilterMismatchNmax 2 --outFilterMultimapNmax 10”. Secondary alignments were removed with SAMtools v.1.5, and data quality was examined using RSeQC v2.6.4 and dupRadar v.1.8.0. Read summarization on the gene level was performed using Subread featureCounts v.1.5.1 with stranded option “-s 2”. Pairwise differential expression comparisons between sample groups were performed using the Bioconductor package DESeq2 v.1.18.1 using a cutoff of 1% FDR.

Whole midbrain tissue and LUHMES cells were homogenized in lysis buffer (50 mM Tris-HCl, pH 6.8; 2% SDS; 10% sucrose; 0.5 mM EDTA; 0.5 mM EGTA plus protease and phosphatase inhibitor cocktails (Sigma-Aldrich)) and were briefly sonicated. Protein concentrations were determined through BCA kit (Pierce) following the manufacturer’s protocol. A total of 20 µg protein was loaded onto 12% SDS-PAGE gels and separated with a Mini protean III system (Bio-Rad). The proteins were transferred onto nitrocellulose membranes by electroblotting adopting standard protocols. Following 30 min incubation with 4% low-fat dry milk in PBST, the membranes were incubated with the following primary antibodies: anti-H3K14ac (7627, Cell Signaling, 1:1000), anti-acetylated lysine (9441, Cell Signaling, 1:1000), anti-DNMT3B (67,259, Cell Signaling, 1:1000), anti-GLUT3 (ab191071, Abcam, 1:1000), anti-SIRT1 (8469, Cell Signaling, 1:1000), anti-SIRT3 (2627, Cell Signaling, 1:1000), anti-SIRT4 (ab90485, Abcam, 1:1000). Anti-α-tubulin (T9026, 1:1000, Sigma-Aldrich) and anti-histone H3 (14,269, Cell Signaling, 1:1000) were used as controls for equal protein loading. Primary antibodies were generally detected with horseradish peroxidase-conjugated secondary antibodies (Jackson Immunoresearch). All antibodies used for Western blotting were diluted in PBST. Densitometric analysis of the immunoreactive bands was performed with ImageJ.

ATP, NAD⁺, NADH, glucose and lactate were measured using commercial test kits, following the instructions of the suppliers. Colorimetric or fluorimetric quantifications of the respective readouts were done in a multiwell plate reader (PerkinElmer Wallac Victor 3V) and conducted in white 96-well plates.

For the determination of cellular ATP levels, the medium was aspirated, and the cells were lysed in the supplied buffer. The employed ATP Detection Assay Kit (#700410 from Cayman Chemicals) uses a firefly luciferase system to translate ATP from cell and tissue lysates into luminescence.

NAD⁺ and NADH measurements were done with the NAD/NADH-Glo Assay Kit (#G9071 from Promega). Cells were grown in 24-well plates, washed, and lysed in 50 µL PBS/50 µL 0.2 M NaOH supplemented with 1% dodecyltrimethylammonium bromide (DTAB). A volume of 50 µL of the lysate was directly heated to 60°C for 15 min (to eliminate NADH), another 50 µL of the lysate were heated to 60°C for 15 min after adding 25 µL 0.4 M HCl (to eliminate NAD⁺). To stabilize the pH, 25 µL 0.5 M Tris, or 50 µL 0.5 M Tris/HCl were subsequently added to the lysates, respectively. NAD⁺ and NADH levels were then quantified using the luciferin-coupled, NAD-cycling enzymatic test system provided with the kit.

Glucose and lactate in the medium were determined using the Glucose Colorimetric Assay Kit (#10009582 from Cayman Chemicals) and the Glycolysis Cell-Based Assay Kit (#600450 from Cayman Chemicals), respectively. Prior to analysis, pH values were documented with a calibrated mini-electrode. To avoid interference of phenol red in the medium, a baseline background measurement was performed for each sample. For the lactate measurements, 10 µL medium were used and diluted with water.

Mitochondrial membrane potential was assessed by MitoTracker Red (#M22425 from Invitrogen) staining, followed by microscopic image analysis as described (Hajieva et al., 2009).

ROS levels were determined by quantitative fluorescence microscopy. Cells grown on coverslips were treated as indicated and were incubated with 5 µM CellROX Deep Red (#C10422 from Invitrogen) for 30 min at 37°C under standard cell culture conditions. Following fixation with 4% PFA and counterstaining with 1 μg/mL DAPI, fluorescence microscopic images were taken from three separate fields of view with approximately 50 cells per image and analyzed with ImageJ.

Mitochondrial ROS levels were also evaluated with the superoxide-selective dye MitoROS 580 (#ABD-16052 from Biomol). Cells grown and pretreated in 96-well plates as indicated before were incubated with 2x MitoROS 580 working solution in Hank’s buffer supplemented with 20 mM HEPES for 30 min at 37°C according to the manufacturer’s protocol. Afterwards, the cells were gently washed, and fluorescence intensity was measured with a microplate reader (Tecan Infinite 200 Pro) using 540 nm/590 nm filters.

SIRT1 activity in cells lysates was determined with the FLUOR DE LYS SIRT1 fluorometric drug discovery assay kit from Enzo Lifesciences (BML-AK555-0001). In brief, LUHMES cells were harvested in 200 µL SIRT1 assay buffer, briefly sonicated and assessed for protein content with a NanoDrop 1000 photometer. The highest possible amount of protein (181 µg) was diluted in SIRT1 assay buffer to a final volume of 35 µL for all samples. Four assay wells per sample were loaded with protein lysate and kept on ice for the remaining procedure. 64 μM SIRT1 substrate, FLUOR DE LYS SIRT1, and 500 µM NAD⁺ were diluted in 15 µL assay buffer and added to three of the four sample wells. The fourth sample received only 15 µL assay buffer to allow evaluation of the lysate background, while one well only received 50 µL SIRT1 assay buffer, and one well received sample buffer with NAD⁺ and FLUOR DE LYS SIRT1 to allow quantification of the components’ background. The plate was incubated at 37°C for 1 h. The developer solution was prepared with SIRT1 assay buffer, 2 mM nicotinamide to stop additional reactions, and 1x FLUOR DE LYS Developer II. A volume of 50 µL of the developer solution was added to each well except for the background controls, which instead received only 50 µL SIRT1 assay buffer. Two additional wells were prepared, one that only received 50 µL SIRT1 assay buffer and 50 µL developer solution, and one that received 64 µM FLUOR DE LYS Deacetylated Standard (in 50 µL assay buffer) and 50 µL developer solution, to allow for the quantification of the developer solution’s background as well as the potential signal maximum. The plate was then incubated at 37°C for 45 min. Afterwards, all fluorescent signals were quantified with a multilabel counter (PerkinElmer Wallac Victor 3V).

All data are expressed as mean ± standard deviation (SD) of the indicated number of independent experiments. Statistically significant differences between the treatment groups were identified by either one-way or two-way ANOVA as indicated, followed by Benjamini-Hochberg multiple comparisons post hoc test. Significance levels are either provided numerically or, in case of the RNA sequencing experiments, coded as symbols denoting p ≤ 0.05.