Epigenetic control of adaptive or homeostatic splicing during interval-training activities

Abstract Interval-training activities induce adaptive cellular changes without altering their fundamental identity, but the precise underlying molecular mechanisms are not fully understood. In this study, we demonstrate that interval-training depolarization (ITD) of pituitary cells triggers distinct adaptive or homeostatic splicing responses of alternative exons. This occurs while preserving the steady-state expression of the Prolactin and other hormone genes. The nature of these splicing responses depends on the exon's DNA methylation status, the methyl-C-binding protein MeCP2 and its associated CA-rich motif-binding hnRNP L. Interestingly, the steady expression of the Prolactin gene is also reliant on MeCP2, whose disruption leads to exacerbated multi-exon aberrant splicing and overexpression of the hormone gene transcripts upon ITD, similar to the observed hyperprolactinemia or activity-dependent aberrant splicing in Rett Syndrome. Therefore, epigenetic control is crucial for both adaptive and homeostatic splicing and particularly the steady expression of the Prolactin hormone gene during ITD. Disruption in this regulation may have significant implications for the development of progressive diseases.


Introduction
Interval-training activities confer beneficial effects on cardiac, muscular and endocrine functions that cannot be attained through a single bout of activity (1)(2)(3)(4).Despite this, the underlying molecular and cellular mechanisms for the adaptive as well as homeostatic responses of different genes remain to be fully elucidated.The regulation of gene expression is critical in facilitating adjustments essential for functions such as hor-mone synthesis ( 2 , 5 , 6 ).However, the regulation of alternative pre-mRNA splicing ( 7 ), a fundamental mechanism driving the diversification of metazoan transcriptomes and proteomes to support intricate cellular functions ( 8 ,9 ), remains enigmatic in the adaptation process.
We have explored how cellular activities, particularly through membrane depolarization, regulate alternative splicing in pituitary cells (10)(11)(12)(13), where we predicted that cells adapt their splicing patterns after repeated stimulation compared to initial treatments ( 12 ).This regulation has been shown to be crucial for the alternative splicing of various genes in response to chronic changes in membrane potentials, significantly affecting neuronal electrical homeostasis or synaptic formation (14)(15)(16).The regulatory mechanisms involve Ca ++ / calmodulin-dependent protein kinase IV (CaMKIV) and downstream splicing factors such as hnRNP L / LL ( 10-12 ,17 ), Sam68 ( 15 ) or Nova-2 ( 14 ), depending on the target exon.Additionally, histone modifications and the methyl-DNA-binding protein MeCP2 play a key role in activity-dependent regulation, suggesting epigenetic influences (18)(19)(20).However, the role of DNA methylation in this process remains unclear.
DNA methylation is pivotal in adaptation ( 1 ,21 ), and generally correlates with exon inclusion in the genome / transcriptome ( 21 ), though not in all cases ( 22 ).The correlation aligns with the regulatory effects of MeCP2 and the methyl-free DNA-binding CTCF on splicing (23)(24)(25).MeCP2, which binds to specific nucleotide sequences mCAC and mCG ( 26 ), influences gene transcription and splicing (25)(26)(27).Notably, MECP2 mutations are identified in up to 96% of typical Rett syndrome cases (28)(29)(30), a severe neurodevelopmental disorder with autistic features and often exacerbated by abnormal brain activities like epilepsy ( 31 ).In Mecp2 -null mouse models of Rett syndrome, long genes' expression and synaptic exon splicing in the hippocampus, particularly after calcium signal-activating kainic acid treatment, are significantly impacted ( 19 ,32 ).Interestingly, both DNA methyltransferase DNMT3a and MeCP2 are regulated by calcium signaling: DNMT3a is recruited by CaMKIVregulated CREM α in T lymphocytes ( 33 ,34 ), and MeCP2 is phosphorylated by CaMKII in hippocampal neurons ( 35 ).MeCP2 likely modulates alternative splicing through its interaction with methylated DNA and splicing factors like YB-1 ( 25 ,36 ).Together, DNA methylation / MeCP2 dysfunction likely plays important roles in the development of neurological diseases as many other epigenetic changes ( 37 ).However, bioinformatics analyses suggest minimal global effects of DNMTs and MeCP2 on alternative splicing ( 38 ).These inconsistencies highlight the need for further research using comprehensive methylation and splicing analyses, including direct methylation of exon DNA in splicing assays, to clarify the effects of DNA methylation and MeCP2 on splicing.
Here, we show the distinct effects of interval-training depolarization (ITD) compared to a single round of treatment on the adaptive splicing of exons in the prolactin-and growth hormone-producing pituitary cells ( 39 ).We identified a critical role of epigenetic control in both adaptive or homeostatic splicing and Prolactin gene expression, a dual function in adaptation while preserving cell identity.

Interval training depolarization by KCl treatments and RNA / DNA extraction
For the ITD KCl treatment group, GH 3 cultures were treated with KCl (50 mM) for 6 h, then washed and supplied with complete fresh medium, followed by incubation for 18 h, completing the 1 st round (day) of the KCl treatment, which was repeated up to the 6 th time (6 th KCl).For the single KCl treatment group (1st KCl), cells went through the same medium change process except that KCl (50 mM) was added on the 6 th day.Where applicable, DMSO or 5-aza-Cytidine (50 μM) was added to fresh culture medium 18 h before the 1st or 6th KCl-treatment.Cell density was maintained throughout the experiment by splitting them into extra dishes.
For samples for RNA-Seq only, we extracted total RNA with the GenElute™ Mammalian Total RNA Miniprep Kit (#R TN350-1KT, Sigma Aldrich, US A).For both RNA-Seq and whol-genome bisulfite sequencing (WGBS), we extracted cytoplasmic RNA for RNA-Seq and the corresponding nuclear DNA for WGBS, using our previous nucleo-cytoplasmic fractionation protocol ( 40 ,41 ).For RT-PCR of the non-RNA-Seq samples, cytoplasmic RNA was used.

RNA-Seq and WGBS analyses
RNA-Seq analyses were performed the same as our previous procedures ( 40 ), except that the Illumina HiSeq4000 pairedend 100-bp sequencing was used for the total RNA of nontreated (NT), 1 st KCl or 6th KCl samples, and the Illumina NovaSeq 6000 S2 paired-end 100-bp sequencing was used for the cytoplasmic RNA of the 6 th KCl samples with or without 24 h pre-treatment by 5-azaC (50 μM).Alternative exons, alternative transcription starts and alternative polyadenylation were identified by DEXSeq ( 40 ,42 ); alternative splice junctions by MATS ( 43 ); differential gene expression by edgeR ( 44 ).
For WGBS analyses, approximately 1 μg of gDNA each sample was subject to bisulfite conversion for shotgun library construction (NEB Ultra II) and Illumina HiSeqX PE150 sequencing, yielding 150-bp paired-end reads.DNA quality control, library preparation, Illumina library quality control and Illumina HiSeqX PE150 sequencing were conducted at the McGill University Génome Québec Innovation Centre (Montréal, Québec, Canada).We obtained an average of 66 ± 4 million of paired-end reads.The sequence quality was verified using FastQC ( 45 ), with the high-quality reads mapped to the rat genome assembly Rnor_6.0.84 (GH 3 samples) or mouse assembly GRCm38 (mm10, hippocampus tissue samples), using BSMAP ( 17 ,46 ).The DNA methylation status of individual cytosines of each exon was obtained by filtering the BSMAP output list with the genomic coordinates within the DEXSeq list of changed exons.Total DNA methylation level of an exon (mCpG or mCpH) was calculated by multiplying the average methylation ratio of CpG or CpH cytosines with the total number of mCpG or mCpH sites, respectively, in the sense strand of each exon.
For functional enrichment analysis and functional annotation of genes, we used the Database for Annotation, Visualization and Integrated Discovery (DAVID, developed at the U.S. National Institute of Allergy and Infectious Diseases, https://david.ncifcrf.gov) ( 47 ).For sequence motif analysis, the MEME was used with default parameters ( 48 ,49 ).The motifs presented have the highest scores for tested data set.

Semi-quantitative RT-PCR
RT-PCR was performed based on our previous procedures ( 10 ,40 ).Briefly, for reverse transcription, 300 ng of cytoplasmic RNA was used in a 10 μl-reaction and incubated at 45

Lentivirus transduction and gene overexpression
The lentiviruses were purchased from GeneChem.The target sequence of MeCP2 shRNA is 5 -C AGC ATCTGC AAAGAGGAGAA-3 , hnRNP L shRNA is 5 -GCT A TGGTGGAGTTTGA TTCT-3 , and the nontargeting control is 5 -TTCTCCGAA CGTGTCA CGT-3 .These sequences were cloned into the lentiviral vector GV493 containing sequences of hU6-MCS-CBh-gcGFP-IRESpuromycin.GH 3 cells were transduced with supernatants containing virus carrying the shMeCP2 or shhnRNP L construct using the Lipofectamine 3000 reagent (Invitrogen).After 48 h, the transduction effect was verified with fluorescent microscopy, and the infected cells were then selected with 10 mg / ml puromycin for 75 days.
For MeCP2 and hnRNP L co-expression, plasmid GV366 (CMV-MCS-HA-SV40-Neomycin) was used to clone full length of hnRNP L (NM_001134760) using XhoI and BamHI restriction enzymes.The plasmid GV657 (CMV-MCS-3flag-polyA-EF1A-zsGreen-sv40-puromycin) was used to clone full length MeCP2 (NM_022673) using BamHI and KpnI restriction enzymes.The control plasmid is CON237.All constructs generated were confirmed by sequencing.HEK293T cells were transiently transduced with plasmid constructs for the overexpression of MeCP2-Flag and hnRNP L-HA using the Lipofectamine 3000 reagent (Invitrogen).
Co-immunoprecipitation for the MeCP2-Flag and hnRNP L-HA interaction was performed using a Pierce™ Co-IP Kit (Thermo Scientific, 26149) following manufacturer's protocol.Briefly, anti-FLAG (Sigma, F1804) or anti-HA (CST, 3724) antibodies were immobilized onto the AminoLink Plus Coupling Resin.The nuclear lysate was pre-cleared with control agarose beads before the antibody-resin addition and overnight incubation at 4 • C. The resin-protein complex was then washed twice in IP Wash Buffer (25 mM Tris, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 5% glycerol; pH 7.4), followed by a final wash in 1 × Conditioning Buffer (pH 7.2), then incubated with 40 μl of Elution Buffer (pH 2.8) at room temperature for 5 min.The resulting elute was mixed with 10 μl of loading buffer (300 mM Tris •HCl, 5% SDS, 50% glycerol, pH 6.8) and heated at 100 • C for 5 min for western blot analysis.

Genome / transcriptome analysis of the datasets from wild type or Rett syndrome mice
We analyzed the DNA methylation or alternative splicing of the hippocampal tissue samples of wild type or Rett syndrome mice using the raw reads from two published datasets ( 19 ,53 ).Briefly, we analyzed the raw reads of RNA-Seq sequences from the total RNA of the hippocampi of male littermate mice (wild type and Mecp2 -null mice) at 7 weeks of age upon KA treatment by intraperitoneal injection ( 19 ), and the raw reads of WGBS from gDNA of the hippocampal dentate gyri of 8-10-week-old male mice (C57BL / 6, same as the Mecp2 -null background) ( 53 ).The reads were quality-controlled by FASTQC, trimmed and mapped to the mouse assembly GRCm38 (mm10) for DEXSeq or BSMAP analyses.

RNA samples from Rett syndrome patients
Usage of the patient samples in the study was under the approval of the Health Research Ethics Board of the University of Manitoba.Total RNA samples were isolated from human hippocampus and cerebellar tissues using TRIzole (Life Technologies), as we reported ( 54 ,55 ).Briefly, 0.5 ml of TRIzol was added to the frozen brain powders of about 50 mg in each tube, then homogenized and incubated for 5 min at room temperature.We then added 0.1 ml chloroform, incubated it for another 3 min, and centrifuged for 15 min (12 000 ×g, 4 • C).We collected the aqueous phase and added 5 μg RNase-free glycogen and 0.25 ml isopropanol, incubated it for 10 min at room temperature, and centrifuged for 10 min (12 000 ×g, 4 • C).We washed the pellet with 0.5 ml of 75% ethanol and centrifuged for 5 min, 12 000 ×g, 4 • C. RNA pellets were airdried and re-suspended in 30 μl of RNase-free water, quantified by NanoDrop 2000 micro-volume spectrophotometer, and stored at -80

Statistical tests
We used two-tailed Student's t -test, except for the built-in tests in DAVID (modified Fisher's exact test).The DEXSeq uses Fisher's test ( 42 ).

Results
Adaptive splicing induced by ITD of GH 3 pituitary cells and its disruption by the DNA methylation inhibitor 5-azacytidine The depolarization effect on splicing is reversible by washing off / adding back depolarizing concentrations of KCl (50 mM) ( 41 ).We thus mimicked interval-training activities of cells by treating GH 3 pituitary cells with interval-training depolarization (ITD), in comparison to the single round of treatment established in these cells in our previous studies ( 11 ,12 ).The cells were treated once or six times with depolarizing concentrations of KCl for 6h, followed by 18h wash-off intervals (Figure 1 A, see also S_Figure 1 A and B for pre-tests of the STREX exon ( 10 )).Our treatment did not alter the growth curve of the cells (S_Figure 1 C), and showed strong homeostatic expression for the majority of exons (99%, S_Figure 1 D) and the signature Prolactin and Growth hormone genes of the pituitary cells ( 56 ) (see below).
A subset of exons (1878 exons of 1204 genes) displayed substantial changes from the 1 st KCl treatment (Figure 1 B), in contrast to the continued exon repression upon sustained KCl treatment up to 24 h without wash-off ( 41 ) (see also S_Figure 1 A).Around 81% of them were alternatively spliced exons and the remainder were produced from alternative transcription start or polyadenylation sites.These genes mainly clustered for functions at the synapse or for RNA recognition (S_Figure 1 E).Their alternative exons exhibited three primary response patterns upon ITD: homeostatic, desensitized, or hypersensitive compared to their responses to the single (1 st ) KCl treatment (S_Figure 1 F-H).Validation by semiquantitative reverse transcription-polymerase chain reaction (RT-PCR) confirmed adaptive splicing in 78.3% of exons ( n = 23) examined by their altered net percent changes of exon inclusion upon ITD over the 1 st treatment ( Supplementary Figure S1 G and H, desensitized or hypersensitive, P < 0.05 by %).Therefore, ITD induces adaptive splicing of a group of exons in genes associated with key cellular functions amidst the homeostatic response of most exons.
Specifically, prior addition of 5-azaC to the cells before the 6 th KCl treatment led to global alterations in exonic DNA methylation (EDM) and exon usage, which inversely correlated with each other overall (S_Figure 3 A and B).Consistent with these effects, KCl depolarization itself indeed also caused EDM changes in across 346 exons detected along the course of the treatments compared to that of the nontreated (Figure 1 C).Importantly, 41 of the top 52 adaptive exons ( ∼79%) were disrupted by at least 1.5 folds when treated with 5-azaC (Figure 1 D).RT-PCR analysis confirmed that the adaptive splicing of these exons was mostly abolished to homeostatic responses by 5-azaC (Figure 1 E and Supplementary Figure S3 C), and two of their response patterns reversed (Dlg and Phldb1 exons).More interestingly, we also noticed that even some homeostatically responsive exons became adaptive upon the 5-azaC treatment (e.g.STREX in Supplementary Figure S1 G and Mapt E7a in Supplementary Figure S3 C).The 5-azaC effect on the patterns of splicing responses to ITD was accompanied by EDM changes (Figure 1 F, G, and S_Figure 3 D).Together, these data suggest that EDM probably play a role in the ITD-induced patterns of adaptive or homeostatic splicing in an exon-specific manner.

EDM level-dependent adaptive or homeostatic splicing of in vitro methylated exons
To directly explore the influence of EDM on adaptive or homeostatic splicing, we created synthetic reporter exons harboring either 3 or 7 copies of CpG dinucleotides that can be specifically methylated in vitro (Figure 2 A, from exon 14 of the synaptic m Baiap2 gene ( 60)).The methylation efficiency reached at least 90% (Figure 2 B) by the M. SssI CpG methyltransferase ( 61 ).The reporters with various EDM levels were tested with co-expressed CaMKIV, a key mediator of depolarization-induced splicing ( 10 , 12 , 13 , 15 , 41 ).
Upon co-transfection into HEK293T cells with the constitutively active Flag-CaMKIV-dCT or its kinase-dead mutant ( 12 , 13 , 17 ), we observed that the (CpG) ×3 exon was repressed by CaMKIV by approximately 20% reduction, regardless of its methylation status, consistent with a homeostatic response (Figure 2 C, lanes 1-4).However, increasing the CpG count to 7, resembling the ITD-induced EDM changes (Figure 1 C, F-G), augmented the CaMKIV repression from 34% to 51% reduction upon hypermethylation, consistent with a hypersensitive response (lanes 5-8).In contrast, mutating the additional 4 copies of CpG dinucleotides eliminated this enhancement effect, reversing it back to homeostatic response (lanes [9][10][11][12].This indicates that the EDM level determines the exon's homeostatic (lanes 1-4, 9-12) or adaptive (lanes 5-8) response to the same stimulation by the co-expressed CaMKIV, and the EDM change is both necessary and sufficient for an altered / adaptive splicing response to CaMKIV.Taken together with the ITD-induction of EDM changes and their correlation with the splicing changes (Figure 1 C, G and S_Figure 3 A-B, D), these results support an essential role of EDM level and its changes in ITD / CaMKIV-induced adaptive or homeostatic splicing.

Essential role of MeCP2 and its partner hnRNP L in ITD-induced adaptive or homeostatic splicing of a group of exons
Based on the EDM analysis, the roles of the mCbinding MeCP2 and CaMKIV-target splicing factor hn-RNP L in activity-dependent splicing ( 10 ,19 ), and synaptic gene exons changed by both 5-azaC and hnRNP L ( 40 ) ( Supplementary Table S1 ), it's likely that MeCP2 and hn-RNP L are involved in the ITD-induced splicing patterns as well.Interestingly, these two proteins interact in reciprocal co-immunoprecipitation assays (Figure 3 A).We thus established GH 3 cell colonies stably knocked down of MeCP2 or hnRNP L using lentiviral shRNAs (Figure 3 B and C).Upon knockdown of either factor, all five adaptive exons of the synaptic genes tested lost their adaptive splicing patterns to homeostatic responses to ITD (Figure 3 D-E), in a similar way as the same Epb41l3 , Nrg1 and Mapt exons by 5-azaC in Supplementary Figure S3 C and Figure 1 E. The homeostatic STREX exon (see also Supplementary Figure S1 G), in contrast exhibited desensitized or hypersensitive responses to ITD upon knockdown of the two factors, respectively.
In further support of hnRNP L's role in the control of additional adaptive / homeostatic exons besides the reported STREX ( 10 , 17 , 40 ), we examined the sequences of the 119 exons regulated by both hnRNP L and 5-azaC (Figure 4 A and Supplementary Table S1 ).They share specifically an hnRNP L-preferred CA-rich consensus motif in or nearby the exons (Figure 4 B).Moreover, UV-crosslinking-immunoprecipitation and CA-to-CG mutation assays of the Mapt exon 6 motif at the 3 splice site supported hnRNP L direct binding to the site in a CA-dependent manner (Figure 4 C).In the WGBS analysis (Figure 1 C), the mC ratio around the 3 splice site of the Mapt exon 6 (4a) changed from 0.75 to 0.54 by the 1 st and 0.66 to 1.0 by the 6 th KCl treatment, respectively, and then reduced to 0.29 by 5-azaC.
Together, these findings support that both MeCP2 and its associated hnRNP L are required for either adaptive or homeostatic splicing responses of the synaptic alternative exons upon ITD.

Essential role of MeCP2 for the homeostatic splicing and expression of the Prolactin hormone gene upon ITD of the pituitary cells or for the proper splicing of the adaptive exons of synaptic genes in the hippocampus of Rett syndrome patients
In addition to disrupting ITD-dependent adaptive or homeostatic splicing, we also found that treatment with 5-azaC exacerbated aberrant splicing of a number of hormone or hormone-related genes including the Prolactin gene upon ITD, accompanied by EDM changes ( Supplementary Figures S4 -S6).This observation suggests disruption of the homeostatic splicing of even the constitutive exons during ITD upon epigenetic changes.We thus examined the Prl transcripts in the MeCP2-knockdown cells and found significantly increased Prl transcript level and exacerbated aberrant splicing, both in an ITD-dependent way (Figure 5 A).Particularly   the aberrant splicing includes the skipping of constitutive exons 2 and 3 coding for the conserved domain of the growth hormone-like superfamily ( 40 ,62 ), or the inclusion of a cryptic 93nt exon also found in hnRNP L-knockdown cells ( 40 ).In contrast, the growth hormone Gh1 gene expression and splicing were not affected in these cells.Thus, the homeostatic splicing and expression of the pituitary cell's signature gene Prolactin specifically requires sufficient MeCP2 in response to ITD.This finding of the ITD-aggravated effect not only aligns closely with but also goes beyond the MECP2 mutationaggravated aberrant splicing induced by chronic neuronal activities with sustained treatment ( 19 ), as well as with abnormal neuronal activities and synaptic plasticity associated with the progression of Rett syndrome ( 31 ,63-66 ).Moreover, the increased Prl mRNA transcripts upon MeCP2 knockdown may help explain the origin of the hyperprolactinemia in about 14% of the Rett syndrome patients ( 67 ,68 ).Together our findings support that MeCP2 is critical for the maintenance of homeostatic splicing and steady expression of the prolactin hormone transcript upon ITD in the pituitary cells.
To corroborate these in vitro findings from cultured cells, we evaluated the splicing of the ITD-induced adaptive ex-ons in MeCP2-defective mouse or human samples.In Mecp2null mice ( 19 ,53 ), the calcium signal-activator kainic acid (KA) generally exacerbated splicing changes ( Supplementary Figure S7 A), consistent with previous studies ( 19 ).Interestingly, these splicing alterations were also inversely correlated with EDM globally, including 8 of the top adaptive exons affected by 5-azaC (from Figure 1 D).Importantly, we identified several of the adaptive exons in Rett syndrome patients with MECP2 mutations ( 29 , 64 , 69 ).These exons, particularly the synaptic EPB41L3 exon 15, showed significant splicing changes in the hippocampi but not cerebella of patients (Figure 5 B, and Supplementary Figure S7 B-D), although with differences from the pituitary cells (Figures 1 and 3 C, and Supplementary Figure S3 ), a context-dependent effect ( 70 ).Nonetheless, this result indicates that at least a group of the ITD-induced adaptive exons also undergo hippocampusspecific aberrant splicing upon loss-of MECP2 function.
Taken together, the epigenetic control is likely essential for both adaptive and homeostatic splicing of alternative exons and the homeostatic splicing and expression of the signature Prolactin hormone gene upon ITD of pituitary cells, with implications for the aberrant splicing of such exons in the progressive genetic disease Rett syndrome.

Discussion
Traditionally, studies on gene expression and particularly alternative splicing have focused on the effects of single or continuous treatments.However, cells often experience repeated extracellular stimuli interrupted by periods of inactivity, as seen in neurons, muscle cells and hormone-producing pituitary cells in such activities as interval-training exercises or drug addiction ( 1 , 2 , 5 ).Our study extends the existing body of work by examining how cells adapt their alternative splicing pattern in response to interval-training depolarization (ITD) while maintaining their identity by homeostatic gene expres-sion and splicing responses.The ITD perhaps did not recapitulate the physiological stimuli precisely in terms of strength and duration but it does provide a proof-of-principle for interval training activity-dependent splicing regulation that is different from the sustained treatment on exons ( 19 ,41 ).The results here demonstrate clearly that exons may exhibit different responses to cell activities depending on how many times the cells are stimulated, likely due to altered epigenetic and splicing components (Figures 1 C,  (4a) as an example of the exon-dependent, diverse changes of methylation / splicing .W ith a single round of KCl treatment (I), the DNA methylation at the 3 splice site is at a comparatively low level (54%); therefore, binding of MeCP2 and its associated hnRNP L to the DNA and pre-mRNA is inefficient, and insufficient to o v ercome the effect of splicing activators ( A ) in cells.Upon the 6 th KCl-treatment (II, ITD), the DNA is hypermethylated (100%, Left panel) recruiting MeCP2 and its associated hnRNP L leading to e x on repression (re v ersed splicing response from the 1 s t KCl treatment).Right panel: with 5-azaC pre-treatment (Figure 1 E, F), the methylation is greatly reduced (29%) and the exon is inhibited by a depolarization non-responsive repressor (R) thereby homeostatic splicing response.*: In case of MeCP2-or hnRNP L-knockdown (Figure 3 D), the repressor R is likely replaced by a depolarization-non-responsiv e activ ator also causing homeostatic splicing though with higher basal le v el of the e x on.Similar decoupling of the methyl-DNA / MeCP2 and associated hnRNP L is likely most effective when the TRD and MDB domains of MECP2 are mutated ( Supplementary Figure S7 D).The mC changes and their disruption during the adaptive splicing are also consistent with the in vitro methylation / mut agenesis dat a of the reporter e x on in Figure 2  or stress in endocrine cells ( 2 , 3 , 5 , 6 ), and to electrical firing activities in neurons ( 11 , 12 , 14-16 ).The dysregulation of the adaptive / homeostatic splicing upon epigenetic disruption (Figure 5 ) likely contribute to the aberrant gene expression or splicing in the progression of neurological diseases.The diverse effects of the epigenetic control of different exons probably suggest an important role of the exon-dependent interplay between the epigenetic and splicing machineries (Figure 6 ) in adaptive and homeostatic cell physiology and progressive diseases.The interplay is worthy of further study for the exon-dependent effect and underlying molecular mechanisms that may involve more epigenetic / splicing factors in the process ( Supplementary Figure S8 and Supplementary Tables S2  and S3 ).

Figure 1 .
Figure 1.A daptiv e splicing of a group of e x ons upon I TD of GH 3 pituitary cells and effect of the DNA meth ylation inhibitor 5-azacytidine.( A ) Scheme of the single or ITD KCl (50mM) treatments of GH 3 pituitary cells.( B ) Heatmap of the fold changes (FC) of 720 exons between the 6 th and 1 st KCl-treated cells by DEXSeq analysis ( ≥1.1-fold, P < 0.01, average base mean > 20, exon base mean > 20).Additional exons were detected by MATS analysis.( C ) EDM le v els upon the 1 st or ITD KCl treatments versus their levels in NT samples by BSMAP analysis of WGBS data ( n = 346 exons with measurable mC in all samples).Grey, yellow, orange or red dot: 1 st , 5 th , 6 th and 6 th KCl plus 5-azaC treatment, respectively.*: 18 h after wash-off of the 5 th KCl media, before the 6 th KCl addition.NT, not treated.( D ) Heatmap of the usage fold changes of 41 top adaptive exons (FC > 1.5) that are prevented by 5-azaC (FC > 1.5).( E ) R epresentativ e agarose gels of RT-PCR products (Left) of 5-azaC effects on the adaptive splicing of exons in response to single round of depolarization or ITD treatments ( n ≥ 3).Arrowheads: PCR primers.The exon numbers are based on reference transcripts in the UCSC Genome Rat Jul.2014 (RGSC 6.0 / rn6) Assembly: Mapt exon 6, M84156, equivalent to human MAPT exon 4a (NM_001123066.3,GRCh38 / hg38); Mapt exon 10, M84156, equivalent to human MAPT exon 10 (NM_001123066.3,GRCh38 / hg38).The changes by depolarization ( %) are all significant except for the Mapt exon 6(4a) with 5-azaC treatment.( F ) The corresponding mC ratios of the exons in E upon the 6th KCl treatment with (red) or without (black) 5-azaC.( G ) An example of EDM changes ( Mapt exon 10, in 100% scale) by the 1 st and 6 th depolarization treatment, alongside its exon inclusion levels.

Figure 2 .
Figure 2. Effect of the EDM status on the adaptive or homeostatic splicing in response to CaMKIV.( A ) Diagram of the in vitro EDM minigene splicing reporter assay.The synthesized reporter exons are based on the adaptive exon 14 (48nt, ENSMUST0 0 0 0 0 1 06233.1,GRCm38 / mm1 0) of the mouse Baiap2 gene (see below).Its partial flanking introns here do not harbor CpG sites.The vector backbone is the DUP175, derived from the constitutive beta-globin e x ons.Arro wheads: PCR primers.( B ) T he in vitro meth ylation efficiency of the reporter e x ons [(CpG) x3 , (CpG) x7 or (CpG) x3+4m ] b y M. SssI CpG methyltransferase, verified by using the CpG methylation-sensitive BstUI (restriction site: CG ↓ CG).The percentage of the full-length insert (minus the unclea v ed full-length, unmeth ylated insert in the preceding lane) out of all fragments in each lane w as tak en as the meth ylation efficiency.( C ) Agarose gels of semi-quantitative RT-PCR products of the splicing reporters transiently expressed in HEK293T cells with CaMKIV or its mutant (CaMKIVm).( D ) Bar graph of net percent changes of the reporter exons in (C) by CaMKIV ( n ≥ 3).Homeo: homeostatic; HS: hypersensitive; ns: not significant; **** P < 0.0 0 01.

Figure 3 .
Figure 3.Effect of knocking down MeCP2 or its partner hnRNP L on adaptive or homeostatic splicing.( A ) Reciprocal co-immunoprecipitation assay of MeCP2-Flag and hnRNP L-HA co-expressed in HEK293T cells.IgG: rabbit IgG, negative control.( B, C ) Western blot analyses of stable lentiviral shRNA-expressing GH 3 cell lines showing specific knockdown of MeCP2 (B) or hnRNP L (C) proteins compared to the non-targeting control shRNA (NC).Mettl3, hnRNP F / H and GAPDH are shRNA-negative and protein loading controls.( D ) R epresentativ e agarose gels of RT-PCR products of the major patterns of MeCP2 or hnRNP L knockdown effects on the adaptive or homeostatic response of exons that are also disrupted by 5-azaC ( n ≥ 3) in GH 3 cells.Arrowheads: PCR primers.( E ) Bar graphs of net percent changes of adaptive or homeostatic exons upon single (1 st ) or ITD (6 th ) KCl treatment, with / without MeCP2 or hnRNP L knockdown ( n ≥ 3).Homeo: homeostatic; DS: desensitized; HS: hypersensitized; R: reversed.ns: not significant, ** P < 0.01; *** P < 0.001; **** P < 0.0 0 01.

Figure 4 .
Figure 4. hnNRP L-binding motifs of e x ons changed in 5-azaC-treated and in hnRNP L-knockdown GH 3 cells.( A ) Diagram of differentially spliced exons in the GH 3 transcriptome upon 5-azaC treatment of the 6 th KCl samples (Figure 1 C) or upon lentiviral knockdown of hnRNP L protein by shL, with the L-like (LL) targets in comparison.( B ) Consensus motifs of shLL-or shL-changed e x ons that were also affected by 5-azaC, identified by MEME analysis.( C ) UV crosslinking of wild type (WT) CA repeat or CG mutant (Mut) RNA probes of the Mapt e x on E6 upstream 3´splice site in HeLa nuclear extracts.Upper: probe diagram.Red dots: corresponding DNA mCpAs, of which methylation levels were reduced by 5-azaC.Underlined: 3´AG.Lower: phosphorimages of proteins crosslinked to the probes and resolved in SDS-PAGE gels.Immunoprecipitating antibody is against hnRNP L (L), YB-1 (YB) or PTB (PTBP1).A sixth of the crosslinking mix for immunoprecipitation was loaded in lanes 3 and 4.

Figure 5 .
Figure 5. Exacerbated o v ere xpression and aberrant splicing of the Prolactin gene in MeCP2-knockdown GH 3 cells upon ITD and aberrant splicing of the adaptiv e e x ons of synaptic genes in R ett syndrome patients.( A ) R T-PCR products of the hormone P rolactin ( P rl ) gene upon MeCP2 knockdo wn and I TD treatment of GH 3 pituitary cells.Gh : growth hormone gene as a negative control of the hormone genes; Gapdh : RNA loading control.Arrowheads: aberrant splice variants with the deduced cryptic skipped / included exons to the right.+93nt: inclusion of a 93nt cryptic exon in the Prl intron 4 as reported by Lei, et al., MCB 18. RNA samples are as in Figure 3 B and D. The bar graph to the right shows the ITD-dependent increase of not only the le v el of the Prolactin mRNA transcript normalized to Gapdh (Top), but also its aberrant transcripts with exons skipped or the intron fragment included (Bottom).( B ) (Left) Agarose gel RT-PCR products of the adaptive EPB41L3 synaptic gene / exon in the hippocampal tissues (Hippo) of Rett syndrome patients and healthy controls.Cerebella of the same individuals are tissue controls.*: Product from primers without cDNA input.(Right) Bar graphs of the percentages of e x on inclusion (mean ± s.d., n = 5 Rett syndrome patients, and 4 healthy controls).ns: not significant; ** P < 0.01; *** P < 0.001.
3 and 5 A).The altered patterns of splicing responses would likely allow cells to fine-tune their adaption during such activities as exercises

Figure 6 .
Figure 6.Summary of the epigenetic components in the control of adaptive or homeostatic splicing upon ITD.Shown are effects on the Mapt exon 6 (4a) as an example of the exon-dependent, diverse changes of methylation / splicing .W ith a single round of KCl treatment (I), the DNA methylation at the 3 splice site is at a comparatively low level (54%); therefore, binding of MeCP2 and its associated hnRNP L to the DNA and pre-mRNA is inefficient, and insufficient to o v ercome the effect of splicing activators ( A ) in cells.Upon the 6 th KCl-treatment (II, ITD), the DNA is hypermethylated (100%, Left panel) recruiting MeCP2 and its associated hnRNP L leading to e x on repression (re v ersed splicing response from the 1 s t KCl treatment).Right panel: with 5-azaC pre-treatment (Figure1 E, F), the methylation is greatly reduced (29%) and the exon is inhibited by a depolarization non-responsive repressor (R) thereby homeostatic splicing response.*: In case of MeCP2-or hnRNP L-knockdown (Figure3 D), the repressor R is likely replaced by a depolarization-non-responsiv e activ ator also causing homeostatic splicing though with higher basal le v el of the e x on.Similar decoupling of the methyl-DNA / MeCP2 and associated hnRNP L is likely most effective when the TRD and MDB domains of MECP2 are mutated ( Supplementary FigureS7 D).The mC changes and their disruption during the adaptive splicing are also consistent with the in vitro methylation / mut agenesis dat a of the reporter e x on in Figure2.Star: p-Ser 513 of hnRNP L b y CaMKIV.R ed dots on DNA: meth yl-Cytosines.Lighter colors of the factors / spliceosomes represent their reduced effects.
Figure 6.Summary of the epigenetic components in the control of adaptive or homeostatic splicing upon ITD.Shown are effects on the Mapt exon 6 (4a) as an example of the exon-dependent, diverse changes of methylation / splicing .W ith a single round of KCl treatment (I), the DNA methylation at the 3 splice site is at a comparatively low level (54%); therefore, binding of MeCP2 and its associated hnRNP L to the DNA and pre-mRNA is inefficient, and insufficient to o v ercome the effect of splicing activators ( A ) in cells.Upon the 6 th KCl-treatment (II, ITD), the DNA is hypermethylated (100%, Left panel) recruiting MeCP2 and its associated hnRNP L leading to e x on repression (re v ersed splicing response from the 1 s t KCl treatment).Right panel: with 5-azaC pre-treatment (Figure1 E, F), the methylation is greatly reduced (29%) and the exon is inhibited by a depolarization non-responsive repressor (R) thereby homeostatic splicing response.*: In case of MeCP2-or hnRNP L-knockdown (Figure3 D), the repressor R is likely replaced by a depolarization-non-responsiv e activ ator also causing homeostatic splicing though with higher basal le v el of the e x on.Similar decoupling of the methyl-DNA / MeCP2 and associated hnRNP L is likely most effective when the TRD and MDB domains of MECP2 are mutated ( Supplementary FigureS7 D).The mC changes and their disruption during the adaptive splicing are also consistent with the in vitro methylation / mut agenesis dat a of the reporter e x on in Figure2.Star: p-Ser 513 of hnRNP L b y CaMKIV.R ed dots on DNA: meth yl-Cytosines.Lighter colors of the factors / spliceosomes represent their reduced effects.
• C before RT-PCR.The Rett Syndrome human brain tissues used in this study are: * Donated brain tissues to the Rastegar lab for research( 54 ,55 ).