Homocysteine metabolites inhibit autophagy, elevate amyloid beta, and induce neuropathy by impairing Phf8/H4K20me1‐dependent epigenetic regulation of mTOR in cystathionine β‐synthase‐deficient mice

The loss of cystathionine β‐synthase (CBS), an important homocysteine (Hcy)‐metabolizing enzyme or the loss of PHF8, an important histone demethylase participating in epigenetic regulation, causes severe intellectual disability in humans. Similar neuropathies were also observed in Cbs−/− and Phf8−/− mice. How CBS or PHF8 depletion can cause neuropathy was unknown. To answer this question, we examined a possible interaction between PHF8 and CBS using Cbs−/− mouse and neuroblastoma cell models. We quantified gene expression by RT‐qPCR and western blotting, mTOR‐bound H4K20me1 by chromatin immunoprecipitation (CHIP) assay, and amyloid β (Aβ) by confocal fluorescence microscopy using anti‐Aβ antibody. We found significantly reduced expression of Phf8, increased H4K20me1, increased mTOR expression and phosphorylation, and increased App, both on protein and mRNA levels in brains of Cbs−/− mice versus Cbs+/− sibling controls. Autophagy‐related Becn1, Atg5, and Atg7 were downregulated while p62, Nfl, and Gfap were upregulated on protein and mRNA levels, suggesting reduced autophagy and increased neurodegeneration in Cbs−/− brains. In mouse neuroblastoma N2a or N2a‐APPswe cells, treatments with Hcy‐thiolactone, N‐Hcy‐protein or Hcy, or Cbs gene silencing by RNA interference significantly reduced Phf8 expression and increased total H4K20me1 as well as mTOR promoter‐bound H4K20me1. This led to transcriptional mTOR upregulation, autophagy downregulation, and significantly increased APP and Aβ levels. The Phf8 gene silencing increased Aβ, but not APP, levels. Taken together, our findings identify Phf8 as a regulator of Aβ synthesis and suggest that neuropathy of Cbs deficiency is mediated by Hcy metabolites, which transcriptionally dysregulate the Phf8 → H4K20me1 → mTOR → autophagy pathway thereby increasing Aβ accumulation.


| INTRODUCTION
The sulfur-containing amino acids methionine (Met), homocysteine (Hcy), and cysteine (Cys) are metabolically related and play important roles in cellular physiology. 1 Met is an essential amino acid that participates in the biosynthesis of proteins (>20 000) and S-adenosylmethionine (AdoMet).AdoMet provides methyl groups for the biological methylation reactions (>300) and propyl groups for the polyamine spermidine and spermine biosynthesis.The methylation reactions generate S-adenosylhomocysteine (AdoHcy).Enzymatic hydrolysis of AdoHcy is the only known metabolic source of Hcy.Re-methylation of Hcy, which regenerates Met, is essential for the folate and one-carbon metabolism, which provides one-carbon units for nucleotides required for DNA and RNA biosynthesis.Transsulfuration of Hcy generates Cys, a semi-essential amino acid, participating in the biosynthesis of proteins, glutathione, hydrogen sulfide, and taurine.Hcy is also metabolized by methionyl-tRNA synthetase to Hcy-thiolactone, [2][3][4][5] which modifies proteins in a nonenzymatic N-homocysteinylation reaction that generates N-Hcy-protein. 5,6levated levels of Hcy due to nutritional or genetic deficiencies are associated with neuropathology affecting the central nervous system (CNS).Cystathionine β-synthase (CBS) deficiency due to mutations in the CBS gene is the most prevalent inborn error in the sulfur amino acid metabolism that results in severe hyperohmocysteinemia (HHcy) characterized by elevated levels of Hcy and its metabolites such as Hcy-thiolactone and N-Hcy-protein. 1 CBS deficiency affects the CNS and causes severe learning and intellectual disability, reduced IQ, 7 psychosis, obsessive-compulsive and behavior/personality disorders. 8ccelerated brain atrophy related to elevated plasma Hcy has been reported in patients with Alzheimer's disease, 9 in patients with alcoholism, 10 and in healthy elderly individuals. 11These phenotypes were replicated in Cbs À/À mice on the C57BL/6J background, which show cognitive impairment manifested as reduced problem-solving abilities, learning disability, and short-term and long-term memory in the puzzle box test. 12Cbs À/À mice on the C3H/HeJ background show cerebellar, although not cerebral, malformation and a transient learning deficit on day 2 in the passive avoidance step-through test. 13he molecular bases of neurological impairments in CBS deficiency are not fully understood.Although Hcy, Hcy-thiolactone, and N-Hcy-proteins accumulate in CBS-deficient patients and mice, [14][15][16][17] it is unclear whether each of these metabolites can contribute to neuropathy associated with CBS deficiency.Individuals with elevated plasma total Hcy show accelerated brain atrophy, impaired cognition, and are at higher risk and developing Alzheimer's disease 18 which is also associated with upregulated brain mTOR signaling. 19lant homeodomain finger protein 8 (PHF8) has been identified as one of the X chromosome genes linked to the intellectual disability syndrome, autism spectrum disorder, attention deficit hyperactivity disorder, 20 and severe intellectual disability. 21PHF8 is a histone demethylase that can demethylate H4K20me1, H3K9me2/me1, and H3K27me2.Demethylation of H4K20me1 by PHF8 is important for maintaining homeostasis of mTOR signaling.The human PHF8 deficiency phenotype has been replicated in Phf8 À/À mice, which show impaired hippocampal long-term potentiation and behavioral deficits in learning and memory. 22ow CBS or PHF8 depletion can cause neuropathy is not fully understood.The present work was undertaken to test a hypothesis that CBS depletion reduces the expression of PHF8, increases H4K20me1, mTOR expression and phosphorylation, inhibits autophagy, and increases amyloid β (Aβ) accumulation.Towards this end, we studied how Cbs gene deletion affects Phf8 expression and H4K20me1 levels, mTOR signaling/ autophagy, and neurodegeneration markers (Nfl, Gfap) expression in the mouse brain.We also used mouse neuroblastoma N2a and Aβ-overproducing N2a-APPswe cells to study how individual Hcy metabolites or depletion of Cbs or Phf8 by RNA interference affect Phf8 expression and its downstream effects on the expression of mTOR and autophagy-related proteins, as well as App expression and Aβ accumulation.

| Brain protein extraction
Mice were euthanized by CO 2 inhalation, the brains collected and frozen on dry ice.Frozen brains were pulverized with dry ice using a mortar and pestle and stored at À80 C. Proteins were extracted from the pulverized brains (50 ± 5 mg) using RIPA buffer (4 v/w, containing protease and phosphatase inhibitors) with sonication (Bandelin SONOPLUS HD 2070) on wet ice (three sets of five 1-s strokes with 1 min cooling interval between strokes).Brain extracts were clarified by centrifugation (15 000 g, 15 min, 4 C) and clear supernatants containing 8-12 mg protein/mL were collected.Protein concentrations were measured with BCA kit (Thermo Scientific).
After reaching 70%-80% confluency, cell monolayers were washed 2-times with PBS and overlaid with DMEM medium without Met (Thermo Scientific), supplemented with 5% dialyzed FBS (MilliporeSigma) and non-essential amino acids.D,L-Hcy, L-Hcy-thiolactone (MilliporeSigma), or N-Hcy-protein (prepared by modification of FBS proteins with Hcy-thiolactone as described in ref. (26)) were added (at concentrations indicated in figure legends) and the cultures were incubated at 37 C in 5% CO 2 atmosphere for 24 h.
For gene silencing, siRNAs specific for the Cbs (Cat.# 100821 and s63474) or Phf8 gene (Cat.# S115808, and S115809) (Thermo Scientific) were transfected into cells maintained in Opti-MEM medium by 24-h Lipofectamine RNAiMax (Thermo Scientific) treatments.Cellular RNA for RT-q PCR analysis was isolated as described in section 2.5 below.For protein extraction, RIPA buffer (MilliporeSigma) was used according to manufacturer's protocol.

| RNA isolation, cDNA synthesis, RT-qPCR analysis
Total RNA was isolated using Trizol reagent (MilliporeSigma).cDNA synthesis was conducted using Revert Aid First cDNA Synthesis Kit (Thermo Fisher Scientific) according to manufacturer's protocol.Nucleic acid concentration was measured using NanoDrop (Thermo Fisher Scientific).RT-qPCR was performed with SYBR Green Mix and CFX96 thermocycler (Bio-Rad).The 2 (ÀΔΔCt) method was used to calculate the relative expression levels. 27Data analysis was performed with the CFX Manager™ Software, Microsoft Excel, and Statistica.RT-qPCR primer sequences are listed in Table S1.

| Chromatin immunoprecipitation assay
For CHIP assays we used CUT&RUN Assay Kit #86652 (Cell Signaling Technology, Danvers, MA, USA) following the manufacturer's protocol.Each ChIP assay was repeated three times.Briefly, for each reaction we used 100 000 cells.Cells were trypsinized and harvested, washed 3Â in ice-cold PBS, bound to concanavalin A-coated magnetic beads for 5 min, RT.Cells were then incubated (4 h, 4 C) with 2.5 μg of anti-PHF8 antibody (Abcam, ab36068) or anti-H4K20me1 antibody (Abcam, ab177188) in the antibody-binding buffer plus digitonin that permeabilizes cells.Next, cells are treated with pAG-MNase (1 h, 4 C), washed, and treated with CaCl 2 to activate DNA digestion (0.5 h, 4 C).Cells were then treated with the stop buffer and spike-in DNA was added for each reaction for signal normalization, and incubated (10-30 min, 37 C).Released DNA fragments were purified using DNA Purification Buffers and Spin Columns (CS #14209) and quantified by RT-qPCR using primers targeting the promoter, upstream, and downstream regions of the mTOR gene (Table S1).Rabbit (DA1E) mAb IgG XP ® Isotype control included in the CUT&RUN kit did not afford any signals in RT-qPCR assays targeting mTOR.

| Confocal microscopy, Aβ staining in N2a-APPswe cells
Mouse neuroblastoma N2a-APPswe cells were cultured in Millicell EZ SLIDE 8-well glass (Merck).After treatments cells were washed 3Â with PBS for 10 min.Cells were fixed with 4% PFA (MilliporeSigma) (37 C, 15 min).After fixation, cells were again washed three times with PBS buffer and permeabilized in 0.1% Triton X-100 solution (RT, 20 min), blocked with 0.1% BSA (RT, 1 h), and incubated with anti-Aβ antibody (CS #8243; 4 C, 16 h).Cells were then washed three times with PBS and stained with secondary antibody Goat Anti-Rabbit IgG H&L (Alexa Fluor ® 488) (Abcam, ab150077; RT, 1 h) to detect Aβ.DAPI (Vector Laboratories) was used to visualize nuclei.Fluorescence signals were detected by using a Zeiss LSM 880 confocal microscope with a 488 nm filter for the Alexa Fluor ® 488 (Aβ) and 420-480 nm filter for DAPI, taking a z stack of 20-30 sections with an interval of 0.66 μm and a range of 15 μm.Zeiss Plan-Apochromat X40/1.2Oil differential interference contrast objective were used for imaging.Images were quantified with the ImageJ Fiji software (NIH).

| Statistical analysis
The results were calculated as mean ± standard deviation.Values for each experimental/treatment group were normalized to controls.Data were analyzed using oneway analysis of variance (ANOVA) with Tukey's multiple comparisons post-test using GraphPad Prism7 software (GraphPad Holdings LLC, San Diego, CA, USA, https:// www.graphpad.com).
We found similar Cbs genotype-dependent changes in autophagy-related mRNAs.Specifically, Becn1, Atg5, and Atg7 mRNAs (Figure 1O,P,R, respectively) were significantly downregulated, while p62 mRNA (Figure 1S), were significantly upregulated in Tg-I278T Cbs À/À compared with Tg-I278T Cbs +/À mice, reflecting changes in the corresponding protein levels.These findings indicate that Cbs gene exerts transcriptional control over the expression of mTOR, Becn1, Atg5, Atg7, and p62 in the mouse brain.

| Cbs deficiency upregulates App expression in the mouse brain
We have previously found that Aβ was elevated in brains of 1-year-old mice Cbs À/À mice compared to Cbs +/À controls. 28To determine whether upregulated App expression might be responsible for elevated Aβ, we quantified App protein and mRNA in Tg-I278T Cbs À/À and Tg-I278T Cbs +/À mice by western blotting (Figure 1M) and RT-qPCR (Figure 1T), respectively.
We also found that App protein significantly increased with age (Figure 1M), irrespective of Cbs genotype: old/young Tg-I278T Cbs À/À mice 1.80-fold, P < 0.0001; old/young Tg-I278T Cbs +/À mice 1.85-fold, P < 0.0001.However, age did not affect App mRNA levels (Figure 1T).These findings suggest that age exerts translational control over App expression independent of Cbs genotype or that age exerts posttranslational control by reducing the turnover of App protein.
3.4 | Cbs gene silencing downregulates the histone demethylase Phf8, upregulates H4K20me1, mTOR, pmTOR, APP, and inhibits autophagy in mouse neuroblastoma N2A-APPswe cells To elucidate the mechanism by which Cbs deficiency impacts Phf8 and its downstream effects on mTOR, autophagy, and App we first examined whether the findings in Cbs À/À mice can be recapitulated in cultured mouse neuroblastoma N2A-APPswe cells.We silenced the Cbs gene by transfecting corresponding siRNAs into N2A-APPswe cells and studied how the silencing impacts the level of Phf8 and its downstream effects.Changes in individual mRNA and protein levels in Cbs-silenced cells were analyzed by RT-qPCR and western blotting, respectively, using Gapdh mRNA and protein as references.
The western blot and RT-qPCR results show that the changes in Phf8, H4K20m31, mTOR signaling, autophagy, and APP induced by Cbs gene silencing in the mouse neuroblastoma cells (Figures S1 and S2) recapitulate the in vivo findings in the Cbs À/À mouse brain (Figure 1).Because Hcy-thiolactone and N-Hcy-protein, [14][15][16][17] in addition to Hcy, 29 are elevated in Cbs À/À mice, it is difficult to assign observed phenotypes to a specific metabolite in these mice.However, this limitation can be overcome in cultured cells by treatments with an excess of a specific metabolite.To determine how each metabolite affects the expression of Phf8 and its effects on downstream targets, we treated N2a cells with Hcy-thiolactone, N-Hcy-protein, and Hcy.

| Cbs gene silencing increased H4K20me1 biding to mTOR promoter in N2a cells
To determine whether increased levels of the histone H4K20me1 mark can promote mTOR gene expression by binding to its promoter, we carried out ChIP experiments using anti-H4K20me1 antibody (Figure 3).The Cbs gene was silenced by transfecting N2a using two different siR-NAs.The cells were permeabilized, treated with anti-H4K20me1 antibody and a recombinant micrococcal nuclease-protein A/G.DNA fragments released form N2a-APPswe cells we quantified by RT-qPCR using primers targeting the transcription start site (TSS) of the mTOR gene as well as upstream (UP) and downstream (DOWN) regions from the TSS.We found that in siRNA Cbs-silenced N2a-APPswe cells the binding of H4K20me1 was significantly increased at the mTOR TSS (2.7-fold, P = 0.002), UP (1.7-fold, P = 0.013), and DOWN (2.2-fold, P = 3EÀ4) sites (Figure 3A).Importantly, there were significantly more DNA fragments from the TSS site (2.62 ± 0.31 and 2.68 ± 0.19 for siRNA Cbs #1 and #2, respectively) than from the UP (1.67 ± 0.14 to 1.68 ± 0.41 for siRNA Cbs #1 and #2, respectively; P = 1EÀ4) and DOWN (2.22 ± 0.06 to 2.27 ± 0.19 for siRNA Cbs #1 and #2, respectively; P = 5EÀ4) sites (Figure 3A).Control experiments show that the binding of H3K4me3 to RPL30 intron was not affected by Cbs gene silencing (Figure 3B).These findings indicate that the binding of H4K20me1 was significantly higher at TSS than at UP and DOWN sites in siRNA Cbs-silenced cells.Similar results were obtained with N2a-APPswe cells (not shown).
CHIP experiments with anti-Phf8 antibody showed that Cbs silencing did not affect binding of Phf8 to the mTOR gene (not shown).

| Hcy-thiolactone, N-Hcy-protein, and Hcy increase H4K20me1 binding to mTOR promoter in N2a cells
Hcy-thiolactone and N-Hcy-protein [14][15][16][17] are elevated in Cbs À/À mice, as is Hcy. 29Each of these metabolites can affect mTOR expression by promoting H4K20me1 binding at the mTOR promoter.To examine this possibility, we treated N2a-APPswe cells with Hcy-thiolactone, N-Hcy-protein, or Hcy for 24 h and analyzed how these treatments affect levels of H4K20me1 bound to the F I G U R E 3 Cbs gene silencing or treatments with Hcy-thiolactone, N-Hcy-protein, and Hcy increase H4K20me1 binding at the mTOR promoter in mouse neuroblastoma N2a cells.(A) CHIP assays with anti-H4K20me1 antibody show the specific binding of H4K20me1 at the transcription start site (TSS) of the mTOR gene as well as downstream and upstream sites in Cbs siRNA silenced N2a cells.Bar graphs show the relative H4K20me1 binding at indicated regions of the mTOR gene in N2a cells transfected with two different siRNAs targeting the Cbs gene (siRNA Cbs #1 and #2).Transfections without siRNA (control-siRNA) or with scrambled siRNA (siRNAscr) were used as controls.(B) Control CHIP experiment with anti-H3K4me3 antibody shows that Cbs gene-silencing did not affect the binding of H3K4me3 at the Rpl30 intron.RT-qPCR was carried out on the input and precipitated DNA fragments.P values vs. 'control-siRNA' plus 'siRNAscr' were calculated by one-way ANOVA with Tukey's multiple comparisons test.(C) N2a cells were treated with shown concentrations of N-Hcyprotein, Hcy-thiolactone (HTL), or Hcy for 24 h at 37 C. Untreated cells were used as controls.CHIP assays with anti-H4K20me1 antibody show the binding of H4K20me1 at the transcription start site (TSS) of the mTOR gene as well as downstream and upstream sites.Bar graphs show the relative H4K20me1 binding at the indicated regions of the mTOR gene.(D) Control CHIP assay with anti-H3K4me3 antibody shows that Hcy-thiolactone, N-Hcy-protein, and Hcy did not affect the binding of H3K4me3 at the Rpl30 intron.RT-qPCR was carried out on the input and precipitated DNA fragments.P values were calculated by one-way ANOVA with Tukey's multiple comparisons test.*Significant difference vs. control, P < 0.05.mTOR gene using ChIP experiments with anti-H4K20me1 antibody.We found that Hcy-thiolactone at 20 μM increased binding of H4K20me1 at the mTOR TSS (2.4-fold, P = 0.004), UP (2.0-fold, P = 0.003), and DOWN (2.1-fold, P = 0.011) sites (Figure 3C).Similar results were obtained with 200 μM Hcy-thiolactone.
Neither Hcy, Hcy-thiolactone, nor N-Hcy-protein affected binding of Phf8 at the mTOR TSS, UP and down sites of the mTOR gene (not shown).
Control experiments show that H3K4me3 binding to RPL30 intron was not affected by Hcy, Hcy-thiolactone nor N-Hcy-protein (Figure 3D).These findings indicate that Hcy-thiolactone significantly increased binding of H4K20me1 at the mTOR gene TSS, UP, and DOWN sites while Hcy significantly increased binding of H4K20me1 at the mTOR TSS.

| Cbs gene silencing upregulates Aβ in N2a-APPswe cells
To determine whether Cbs depletion in neural cells could affect Aβ accumulation, we silenced the Cbs gene in N2a-APPswe cells and quantified Aβ by fluorescence confocal microscopy using anti-Aβ antibody.The cells were transfected with two different Cbs-targeting siRNAs, permeabilized, treated with anti-Aβ antibody, and Aβ was visualized with fluorescent secondary antibody (Figure 4A) and quantified (Figure 4B).We found that Cbs-silencing led to increased Aβ accumulation, manifested by significant increases in an average size of fluorescent Aβ puncta (from 0.32 ± 0.02 to 0.80 ± 0.02 and to 0.90 ± 0.06 μm 2 for siRNA Cbs #1 and #2, respectively; P = 1EÀ6) in Cbs siRNA-treated N2a-APPswe cells compared to siRNAscr-treated cells or control without siRNA (Figure 4B).Less significant increases were also observed in the area of fluorescent Aβ puncta (from 130 ± 7 to 189 ± 47 and 183 ± 13 μm 2 for siRNA Cbs #1 and #2, respectively; P = 0.055-0.079).
3.9 | Hcy-thiolactone, N-Hcy-protein, and Hcy upregulate Aβ accumulation in N2a-APPswe cells To ascertain how each Hcy metabolite can affect Aβ accumulation, we treated N2a-APPswe cells with Hcythiolactone, N-Hcy-protein, 30 or Hcy and quantified Aβ by fluorescence confocal microscopy using anti-Aβ antibody (Figure 4C,D).The metabolite-treated and untreated control cells were permeabilized, treated with anti-Aβ antibody, followed by fluorescent secondary antibody to visualize Aβ signals.We found that Aβ was significantly upregulated in N2a-APPswe cells treated with 20 μM Hcy, Hcy-thiolactone, or 10 μM N-Hcy-protein, manifested by significantly increased average size and signal intensity of the fluorescent Aβ puncta, compared to untreated cells (Figure 4D).Similar results were obtained with N2a-APPswe cells treated with 10-fold higher concentrations of Hcy, Hcy-thiolactone (200 μM), or 2-fold higher N-Hcy-protein (20 μM). 30owever, while treatments with Hcy or Hcy-thiolactone increased the size of the fluorescent Aβ puncta, treatments with N-Hcy-protein did not affect the size of Aβ signal (Figure 4D), suggesting different effects of these metabolites on the structure of Aβ aggregates.These findings indicate that Hcy and its downstream metabolites Hcy-thiolactone and N-Hcy-protein contribute to the accumulation of Aβ induced by the metabolic stress of HHcy.

| Phf8 gene silencing upregulates Aβ but not APP in N2a-APPswe cells
The findings that Phf8 expression was attenuated in brains of Cbs À/À mice and in Cbs-silenced or Hcy metabolite-treated mouse neuroblastoma N2a-APPswe cells raises a possibility that Phf8 loss by itself can affect biochemical pathways leading to Aβ accumulation.To examine this possibility, we silenced the Phf8 gene by transfection of N2a-APPswe cells with Phf8-targeting siR-NAs 31 and quantified by western blotting proteins that were affected by Cbs depletion in the mouse brain (Figure 1).We found significantly reduced Phf8 levels (by 80%, P < 0.0001; Figure 5A), significantly increased H4K20me1 (3-fold, P < 0.001; Figure 5B), mTOR (1.4-fold, P < 0.001; Figure 5C), and pmTOR (1.6-fold, P < 0.01; Figure 5D) levels in Phf8-silenced cells.Autophagy-related proteins Atg5 and Atg7 were significantly downregulated (by 20%-35%, P < 0.0001 and <0.01; Figure 5E,F) while Becn1 was not affected in Phf8-silenced cells (Figure 5G).Importantly, western blot analyses showed that the Phf8 gene silencing did not affect APP levels in N2a-APPswe cells (Figure 5H).In contrast, fluorescence confocal microscopy analyses showed that Aβ was significantly upregulated in Phf8-silenced cells, manifested by significantly increased average size (P < 0.001) and signal intensity (P < 0.001) of the fluorescent Aβ puncta compared to controls without siRNA or with siRNAscr (Figure 5J,K). 31Taken together, these findings clearly show that Aβ accumulation in the Phf8-silenced cells occurred independently of APP and suggest that most likely it was caused by impaired autophagy.

| Cbs deficiency accelerates agedependent neurodegeneration in the mouse brain
Glial fibrillary acidic protein (GFAP) 31 and neurofilament light chain (NfL) 32 increase with age and have been associated with Alzheimer's disease.GFAP, an astrocytic cytoskeletal protein, is a marker of astrogliosis, abnormal activation and proliferation of astrocytes due to neuronal damage. 31NfL is a marker of neuro-axonal damage, resulting because of neurologic diseases. 33To find whether Cbs deficiency accelerates neuronal damage we examined the expression of Nfl and Gfap in brains of Tg-I278T Cbs À/À mice and Tg-I278T Cbs +/À sibling controls by using western blotting and RT-qPCR.We found that the Nfl and Gfap protein levels were significantly increased in brains of older (1-year-old) Tg-I278T Cbs À/À mice vs. Tg-I278T Cbs +/À sibling controls (Nfl: 2.48 ± 0.39 vs. 1.74 ± 0.38, P < 0.0001, Figure 6A; Gfap: 2.14 ± 0.24 vs. 1.63 ± 0.21, P < 0.0001, Figure 6B).In contrast, in younger (9-week-old) mice levels of Nfl and Gfap proteins were not affected by Cbs deficiency (Figure 6A,B).
Although Nfl and Gfap proteins and mRNAs increased significantly with age in Cbs À/À mice and Cbs +/À , the increases were faster in Cbs À/À animals (Figure 6A-D).These findings show that Cbs À/À genotype transcriptionally upregulates Nfl and Gfap genes in old but not in young mice and that Cbs protects mice from age-dependent neurodegeneration.

| DISCUSSION
The loss of CBS, an important Hcy-metabolizing enzyme, 12,13 or the loss of PHF8, 21,22 an important histone demethylase participating in epigenetic regulation, causes severe intellectual disability in humans.Similar neuropathies are also observed in Cbs À/À and Phf8 À/À mice.Because PHF8 participates in epigenetic regulation, we surmised that dysregulation of epigenetic mechanisms involving PHF8 could underlie neuropathy associated with CBS deficiency.The present study substantiated this supposition by showing that the expression of Phf8 mRNA and protein was downregulated while H4K20me1 epigenetic mark was upregulated in brains of Cbs À/À mice (Figure 1) and in Cbs-depleted mouse neuroblastoma cells (Figures S1,S2).
We also found that metabolites that are elevated in CBS-deficient humans and mice-Hcy-thiolactone, N- Hcy-protein, and Hcy-downregulated Phf8 (Figure 2A) and upregulated histone H4K20me1 (Figure 2B) in cultured mouse neuroblastoma cells.This suggests that Hcythiolactone, N-Hcy-protein, and Hcy are responsible for Phf8 downregulation (Figure 1A) and H4K20me1 upregulation (Figure 1B) found in vivo in brains of Cbs À/À mice.We also showed that treatments with siRNA targeting Cbs gene or with Hcy-thiolactone or Hcy, which metabolically downregulated Phf8 expression (Figure S1B, Figure 2A), promoted the accumulation of both APP (Figure 2H) and Aβ (Figure 4) in mouse neuroblastoma cells.In contrast, attenuation of Phf8 expression by treatments with siRNAs targeting Phf8 gene promoted Aβ accumulation (Figure 5J,K), but did not affect APP levels (Figure 5H).These disparate effects of metabolic and genetic Phf8 depletion on APP expression suggest that Aβ accumulation can occur by two different mechanisms in Cbs-depleted brain and neural cells: upregulated APP expression caused by Hcy-thiolactone, N-Hcy-protein, and Hcy, and impaired Aβ clearance caused by downregulation of autophagy (Figure 7A).In Phf8-depleted cells, only one mechanism, downregulated autophagy, leads to accumulation of Aβ.
The findings that upregulation of APP (Figure 1M,T; Figures S1F and S2G) was associated with downregulation of Becn1 in Cbs À/À mice (Figure 1I,O) and Cbs-depleted mouse neural cells (Figures S1G and S2D) suggest a third pathway that can contribute to Aβ accumulation.In this pathway Becn1 is a negative regulator of APP expression and processing (Figure 7B).This pathway is consistent with the findings showing that Becn1, a protein with a key role in the initiation of autophagy, is decreased in human AD brains and that genetic depletion of Becn1 in mice that overexpress APP (APP + Becn +/À mice) increased Aβ accumulation in neurons. 34Becn1 can also regulate APP processing and turnover.For example, depletion of Bcln1 by RNA interference in rat neuroblastoma cells expressing human APP transgene (B103/hAPPwt cells) increased APP, Lc3, and Aβ, while Bcln1 overexpression downregulated APP. 35The involvement of autophagy in Cbs depletion-induced APP and Aβ accumulation needs to be confirmed in future experiments by boosting autophagy (e.g., by treatment with TAT-Beclin1), which should rescue APP and Aβ accumulation in Cbs-depleted cells.
Our findings suggest that Phf8 regulates Aβ accumulation through its effects on mTOR and autophagy.Specifically, we found that treatments with Hcy-thiolactone, N-Hcy-protein, or Hcy, which downregulated Phf8 expression (Figure 2A) and upregulated the histone mark H4K20me1 (Figure 2B), also increased H4K20me1 binding to the mTOR promoter in mouse neuroblastoma cells (Figure 3C).These findings provide direct mechanistic evidence linking each of the Hcy metabolites with dysregulated mTOR signaling and its downstream consequences.Concomitant inhibition of autophagy (Figure 2E-G) and upregulation of Aβ (Figure 4C,D) by Hcy-thiolactone, N-Hcy-protein, and Hcy identify a likely mechanism (Figure 7) that accelerates neuronal damage in CBS deficiency, manifested by elevated Nfl and Gfap levels (Figure 6).This mechanism can also account for the association of HHcy with Alzheimer's disease. 18revious studies showed that Phf8 bound at the TSS of Kras, Camk2d, and Rps6ka1 genes involved in mTOR signaling and that depletion of Phf8 increased binding of H4K20me1 at the TSS in Kras, Camk2d, and Rps6ka1 genes in mice 22 ; H4K20me1 binding at the mTOR gene was not examined in that study.In the present study we found that Phf8 depletion in mouse neuroblastoma N2a cells by treatments with Hcy-thiolactone, N-Hcy-protein, or Hcy (Figure 2A), or by silencing Cbs gene with siRNAs (Figure S2B), increased H4K20me1 binding at the TSS as well as down and up sites of the mTOR gene (Figure 3A,C).Our findings add mTOR to the list of genes regulated by H4K20me1.However, in contrast to TSS of the Kras, Camk2d, and Rps6ka1 genes, which bound Phf8, 22 we could not detect any binding of Phf8 to TSS of the mTOR gene.
We found that the effects of Cbs deficiency were mimicked by Hcy, Hcy-thiolactone, and N-Hcy proteins.This may not be surprising given that Hcy is metabolized to Hcy-thiolactone which then modifies proteins generating N-Hcy-protein, as originally demonstrated in 1997 5 and confirmed in a variety of cell types and in animal and human studies. 1 Similarities in pathophysiological effects of each of the three metabolites also include their cytotoxicity (reviewed in 1 ), demonstrated for N-Hcyprotein (e.g., 36,37 ), Hcy-thiolactone (e.g., 38 ), and Hcy (e.g., 39 ).However, how each of these metabolites exerts its effects needs to be examined in future studies.
Previous studies have found that Cbs À/À mice have elevated Hcy-thiolactone and N-Hcy-protein, [14][15][16][17] in addition to Hcy, 29 which was accompanied by increased brain Aβ accumulation 28 and cognitive impairments. 12,13In the present study we have found significantly depleted Phf8 in brains of Cbs À/À mice (Figure 1A).Although depletion of Phf8 is linked to intellectual disability, autism spectrum disorder, attention deficit hyperactivity disorder, 20 and mental retardation, 21 it was not known to be associated with Alzheimer's disease.Our present findings that Phf8 depletion in the mouse neuroblastoma cells, induced by Phf8 siRNA interference (Figure 5A), or by supplementation with Hcy-thiolactone, N-Hcy-protein, or Hcy (Figure 2A), significantly increased Aβ accumulation (Figure 5J,K), suggest that Phf8 depletion can contribute to the association of Alzheimer's disease with HHcy. 18n conclusion, our findings find the histone demethylase Phf8 as a regulator of Aβ accumulation and suggest that neuropathy of Cbs deficiency is mediated by Hcy metabolites-dependent depletion of Phf8, which upregulates mTOR through increased H4K20me1 binding at the mTOR promoter, which in turn inhibits autophagy, leads to Aβ accumulation, and neuronal damage, manifested by elevation of brain Nfl and Gfap levels.We have also shown that Cbs-depletion and Hcy metabolites, independently of their inhibitory effect on Phf8 expression, upregulated APP expression, which can also contribute to Aβ accumulation.

AUTHOR CONTRIBUTIONS
Łukasz Witucki performed and analyzed the experiments; Hieronim Jakubowski conceived the idea for the project, designed the study, bred the mice, collected mouse tissue samples, analyzed data, and drafted the paper.

F I G U R E 1
Deletion of the Cbs gene affects the expression of histone demethylase Phf8, histone H4K20me1 epigenetic mark, mTOR signaling, autophagy, and App in the mouse brain.Nine-week-old and 1-year-old Tg-I278T Cbs À/À mice (n = 7 and 14) and their Tg-I278T Cbs +/À sibling controls (n = 10 and 10) were used in experiments.Bar graphs illustrating quantification of the following brain proteins by western blotting are shown: Phf8 (A), H4K20me1 (C), mTOR (E), pmTOR (G), Becn1 (I), Atg5 (J), Atg7 (K), and p62 (L), and App (M).Pictures of western blots used for protein quantification are shown in panels (D), (H), and (N).Bar graphs illustrating quantification of the following brain mRNAs by RT-qPCR are also shown: Phf8 (B), mTOR (F), Becn1 (O), Atg5 (P), Atg7 (R), and p62 (S), and App (T).Gapdh protein and mRNA were used as references for normalization.P values were calculated by one-way ANOVA with Tukey's multiple comparisons test.

F
I G U R E 2 Hcy-thiolactone, N-Hcy-protein, and Hcy downregulate Phf8, upregulate H4K20me1 epigenetic mark, mTOR signaling, App, and impair autophagy in mouse neuroblastoma N2a cells.N2a cells were treated with indicated concentrations of N-Hcy-protein, Hcythiolactone (HTL), or Hcy for 24 h at 37 C as described in Materials and Methods.Bar graphs illustrating the quantification of Phf8 (A), H4K20me1 (B), mTOR (C), pmTOR (D), Becn1 (E), Atg5 (F), Atg7 (G), and App (H) based on western blot analyses with corresponding antibodies are shown.Gapdh was used as a reference protein.Data are averages of three independent experiments.P values were calculated by one-way ANOVA with Tukey's multiple comparisons test.

F I G U R E 4
Cbs depletion or treatments with Hcy-thiolactone, N-Hcy-protein, and Hcy promote Aβ accumulation in the mouse neuroblastoma N2a-APPswe cells.(A, B) The cells were transfected with siRNAs targeting the Cbs gene.Transfections without siRNA (control-siRNA) or with scrambled siRNA (siRNAscr) were used as controls.Aβ was detected and quantified by confocal immunofluorescence microscopy using anti-Aβ antibody.(A) Confocal microscopy images of Aβ signals from Cbs-silenced N2a-APPswe cells.(B) Bar graphs show quantification of Aβ signals in Cbs-silenced and control cells.(C, D) N-Hcy-protein, Hcy-thiolactone (HTL), or Hcy promote Aβ accumulation in mouse neuroblastoma N2a-APPswe cells.N2a-APPSwe cells were treated with indicated concentrations of N-Hcy-protein, Hcy-thiolactone (HTL), or Hcy for 24 h at 37 C. Untreated cells were used as controls.Confocal microscopy images (C) and quantification of Aβ signals (D) from treated and untreated N2a-APPswe cells.P values vs. 'control-siRNA' plus 'siRNAscr' were calculated by one-way ANOVA with Tukey's multiple comparisons test.Data for cells treated with Hcy-thiolactone and N-Hcy-protein in panels (C, D) were reproduced with permission from Ref. (30).

F
I G U R E 5 Silencing Phf8 gene promotes Aβ accumulation mediated by upregulation of mTOR signaling and inhibition of autophagy in the mouse neuroblastoma N2a-APPswe cells.The cells were transfected with siRNAs targeting the Phf8 gene (Phf8 siRNA #1 and #2).Transfections without siRNA (control-siRNA) or with scrambled siRNA (siRNAscr) were used as controls.Proteins were quantified by western blotting.Bar graphs illustrate levels of (A) Phf8, (B) H4K20me1, (C) mTOR, (D) pmTOR, (E) Atg5, (F) Atg7, (G) Becn1, and (H) APP.Aβ was detected and quantified by confocal immunofluorescence microscopy using anti-Aβ antibody.Representative western blot is shown in panel (I).Confocal microscopy images of Aβ signals from Phf8-silenced and control N2a-APPswe cells.(J) Bar graphs show quantification of Aβ signals.P values vs. 'control-siRNA' plus 'siRNAscr' were calculated by one-way ANOVA with Tukey's multiple comparisons test.NS, not significant.

F
I G U R E 6 Deletion of the Cbs gene upregulates Nfl and Gfap expression of neurodegeneration markers in the mouse brain.Nineweek-old and 1-year-old Tg-I278T Cbs À/À mice (n = 7 and 14) and their Tg-I278T Cbs +/À sibling controls (n = 10 and 10) were used in experiments.Bar plots show the quantification of brain Nfl (A) and Gfap (B) proteins by western blotting-representative images shown in panel (C)-and of brain Nfl (D) and Gfap mRNAs (E).Gapdh protein and mRNA were used as references for normalization.P values shown on the graphs were calculated by one-way ANOVA with Tukey's multiple comparisons test.

F
I G U R E 7 Mechanisms underlying neuropathy in the Cbs-deficient brain.Up and down arrows show the direction of changes in the indicated metabolites, proteins, and molecular processes.App, amyloid beta precursor protein; CBS, cystathionine β-synthase; Hcy, homocysteine; HTL, Hcy-thiolactone; mTOR, mammalian target of rapamycin; Phf8, plant homeodomain finger protein 8.