Phosphorylation of Neuronal Lysine‐Specific Demethylase 1LSD1

JOURNAL OF NEUROCHEMISTRY | 2014 | 128 | 603–616 doi: 10.1111/jnc.12457

, , , , ,

*Department of Medical Biotechnology and Translational Medicine, University of Milan, Milan, Italy †Department of Biology and Biotechnology, University of Pavia, Pavia, Italy ‡Consiglio Nazionale delle Ricerche, Institute of Neuroscience, Milan, Italy §Neuromuscular Diseases and Neuroimmunology, Neurological Institute Foundation Carlo Besta, Milan, Italy ¶Department of Veterinary Science and Public Health (DIVET), University of Milan, Milan, Italy **Filarete Foundation, Milan, Italy

Abstract structure analysis using a phospho-mimetic mutant Epigenetic mechanisms play important roles in brain develop- (Thr369bAsp), indicate that phosphorylation affects the residues ment, orchestrating proliferation, differentiation, and morpho- surrounding the exon E8a-coded amino acids, causing a local genesis. Lysine-Speciﬁc Demethylase 1 (LSD1 also known as conformational change. We suggest that phosphorylation, KDM1A and AOF2) is a histone modiﬁer involved in transcrip- without affecting demethylase activity, causes in neurons tional repression, forming a stable core complex with the CoREST and HDAC1/2 corepressors detachment from LSD1- corepressors corepressor of REST (CoREST) and histone 8a and impairs neuronal LSD1-8a repressive activity. In deacetylases (HDAC1/2). Importantly, in the mammalian CNS, neurons, Thr369b phosphorylation is required for morphogenic neuronal LSD1-8a, an alternative splicing isoform of LSD1 activity, converting neuronal LSD1-8a in a dominant-negative including the mini-exon E8a, sets alongside LSD1 and is isoform, challenging LSD1-mediated transcriptional repression capable of enhancing neurite growth and morphogenesis. Here, on target genes. we describe that the morphogenic properties of neuronal LSD1- Keywords: corepressor, epigenetics, KDM1A/LSD1, neuronal 8a require switching off repressive activity and this negative maturation, transcription. modulation is mediated in vivo by phosphorylation of the J. Neurochem. (2014) 128, 603–616. Thr369b residue coded by exon E8a. Three-dimensional crystal

Neurons acquire a proper morphology during late embryonic Received August 06, 2013; revised manuscript received September 10, fi 2013; accepted September 16, 2013. development and early post-natal life. Speci c morphoge- Address correspondence and reprint requests to Elena Battaglioli, nesis together with the possibility to remodel neurite Department of Medical Biotechnology and Translational Medicine, arborization during the adult life, provide neurons with University of Milan, Via Viotti 3/5, 20133 Milano MI, Italy. E-mail: unique plasticity essential to complex cognitive functions (de [email protected] Coordinate deposition la Torre-Ubieta and Bonni 2011). Transcriptional repression : Atomic coordinates and structure factors have been deposited with the Protein Data Bank (code 4bay). through epigenetic mechanisms is emerging as a key process Abbreviations used: LSD1, lysine-specific demethylase 1; PTM, post- to control developmental programs and crucial to modulate translational modifications.

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 603 604 E. Toffolo et al. neuronal morphology required for adaptive plasticity in the corepressors from LSD1-8a. We show that phosphorylation CNS (Perissi et al. 2010). In this context, an important of Thr369b is necessary for LSD1-8a to drive neurite corepressor is represented by the epigenetic enzyme Lysine- outgrowth and branching. We propose that phosphorylation Specific Demethylase 1 (LSD1, also known as KDM1A). converts LSD1-8a into a dominant negative enzyme isoform, Both in non-neuronal tissues and in the CNS, LSD1 forms a essentially unable to repress transcription of genes that are stable complex with Corepressor of REST (CoREST), associated with neuronal growth and morphogenesis, reveal- histone deacetylases (HDAC) HDAC1, and HDAC2, which ing a fascinating molecular mechanism in which alternative enables a functional crosstalk between the demethylase and splicing and phosphorylation finely regulate the activity of a deacetylase enzymes (Lee et al. 2006a; Forneris et al. 2009). chromatin-modifying complex in the mammalian CNS. Indeed, histone deacetylase activity and CoREST are required for LSD1 histone H3-Lys4 demethylation both in Methods vivo and in vitro (Lee et al. 2005; (Forneris et al. 2005a). The recruitment of LSD1/CoREST/HDAC1/2 core complex Plasmids by a number of specific transcription factors settles its Gal4-LSD1 and Gal4-LSD1-8a containing the full length biological roles in development and differentiation both in cDNAs were generated by PCR and cloned into pSG424- normal and pathological conditions (Godmann et al. 2007; vector. The HA-LSD1, HA-LSD1-8a, and the E. coli Saleque et al. 2007; Yokoyama et al. 2008; Fuentes et al. expression plasmids have been described (Zibetti et al. 2012) and regulates chromatin structure and gene activity 2010). Thr369bAla, Thr369bAsp, and Lys661Ala mutants (Chosed and Dent 2007; Godmann et al. 2007; Saleque where obtained by site-specific mutagenesis using the et al. 2007; Yokoyama et al. 2008; Wang et al. 2009). The QuikChange II Site-Directed Mutagenesis Kits (Stratagene, role of LSD1 in neuronal differentiation was originally La Jolla, CA, USA). For lentiviral transduction, HA-fused inferred through its function as corepressor of the transcrip- LSD1 and LSD1-8aThr369bAsp cDNAs were generated by tional silencer repressor element-1 silencing transcription PCR using pCGN-vectors as templates and cloned into factor/neuro-restrictive silencer factor REST/NRSF), which lentiviral transfer vector C-FUW in the AscI and HpaI sites. targets several genes involved in neuronal differentiation and All plasmids were sequenced. neurite outgrowth (Ballas et al. 2005). More recently, LSD1 has been implicated in neural stem cells proliferation through Animals the association with the nuclear receptor homologue of the Animals were obtained from Charles River, Calco, Italy. All Drosophila tailless gene TLX affording repression of p21 and the animal treatments followed the guidelines established by PTEN genes and maintenance of neural stemness in mouse the Italian Council on Animal Care and were approved by the hippocampus (Yokoyama et al. 2008; Sun et al. 2010). Italian Government decree No. 27/2010 being in compliance Along these lines, a fascinating finding was that a LSD1- with the ARRIVE guidelines. All efforts were made to specific inhibitor can block memory consolidation in mice minimize the number of subjects used and their suffering. (Neelamegam et al. 2012). In a previous work, our group identified a neurospecific isoform of LSD1, neuroLSD1 Immunopurification of LSD1 from brain extract and in situ (LSD1-8a) which is able to enhance neuronal maturation digestion and mass spectrometry when over-expressed in cortical neurons, representing one of Total protein extracts were obtained from post-natal day 1 rat the few examples of chromatin modifying enzyme devoted to brains (Charles River) lysed in immunoprecipitation (IP) neurons (Wu et al. 2007; Abidi et al. 2008). This isoform is buffer (10% glycerol, 0.5 mM EDTA, 0.5% Triton-X100, generated by alternative splicing through inclusion of the 1 mM phenylmethylsulfonyl fluoride), and 19 Protease mini-exon E8a, coding for the tetra-peptide Asp-Thr-Val-Lys Inhibitors Cocktail, (Sigma-Aldrich, St. Louis, MO, USA). (Zibetti et al. 2010). Neuronal LSD1-8a retains substrate Two milligram of cellular extract were reacted with 2.5 lgof specificity and catalytic properties as well as the ability to rabbit polyclonal anti-LSD1 antibody (AbCam, Cambridge, interact with CoREST and HDAC1/2. The alternative UK) overnight rocking at 4°C. The immunoprecipitates were splicing mechanisms regulating inclusion of exon E8a and collected with Protein G Agarose (Invitrogen, Carlsbad, CA, generation of LSD1-8a are dynamically regulated during USA). After incubation, the beads were washed four times mammalian brain development and synaptic maturation. with the IP buffer. The immunoprecipitates were then eluted Here, we provide evidences about the molecular mecha- with 1x sodium dodecyl sulfate Sample buffer and separated nism at the basis of the neurospecific function of LSD1-8a. on 1D polyacrylamide. Following mass compatible silver We discovered that the second residue coded by the staining, the protein band at 110 KDa band corresponding to neurospecific exon E8a, Thr369b, can be phosphorylated in the LSD1 molecular weight excised, destained with 50% the brain. Using a phospho-mimetic mutant, we propose that acetonitrile in ammonium bicarbonate 0.1 M (40 min at phosphorylation, without affecting in vitro demethylase 25°C), dried in a Speed Vac, soaked with ammonium activity, causes detachment of CoREST and HDAC1/2 bicarbonate 0.1 M, reduced and derivatized by iodoacetamide,

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 Splicing and phosphorylation modulate LSD1/KDM1A 605 and digested overnight with trypsin sequencing grade reaction was read using a microplate reader Wallac 1420 (Roche, Monza, Italy) at 37°C (Pastori et al. 2010). The in VICTOR3 V (Perkin Elmer, Waltham, MA, USA). gel tryptic digest was extracted with 50% acetonitrile in 0.1% triﬂuoroacetic acid. Digested aliquots were removed and Lentiviral infections subjected to a desalting/concentration step on a lZipTipC18 HA-LSD1 and Ha-LSD1-8a-T369bD were subcloned into

(Millipore, Bedford, MA, USA) using 40% CH3CN in 0.1% the lentiviral transfer vector FUW which contains the trifluoroacetic acid as eluent before MALDI analysis using ubiquitin promoter (Lois et al. 2002). The preparation of an Autoflex III instrument (Bruker Daltonics, Bremen, the lentivirus vectors has been previously described (Naldini Germany). For protein identification, the Biotools software et al. 1996; Lois et al. 2002). Neurons were infected after (Bruker Daltonics, Milano, Italy) was used to search the 4 days in vitro (DIV4) and analyzed on DIV8. NCBI protein Data Bank by setting carbamidomethylation as fixed modification and by setting oxidation (M), phosphor- Total RNA extraction and RT-PCR analysis ylation (Ser, Thr and Tyr), and deamidation (Asn and Gln) as Total RNA was isolated using the Trizol reagent (Sigma- variable modifications. Two missed cleavages were allowed. Aldrich), and the purified RNA was treated with RNase-free Peptide quality scores were derived by processing against DNase set (Qiagen, Valencia, CA, USA) to remove any decoy shuffled databases. residual DNA. Quantitative RT-PCR analysis was performed on an iQ5 Real-Time PCR Detection System (Biorad, Stable cell lines Hercules, CA, USA) using the iScriptTM two-Step RT-PCR HA-tagged human LSD1 or LSD1-8a cDNAs (Zibetti et al. Kit with SYBRâ Green (Biorad). The relative expression of 2010) were transfected into HeLa cells together with pPURO the investigated genes was quantified after normalization (Clontech Takara-Bio, Mountain View, CA, USA). Stable against ribosomal protein SA (RPSA) and glyceraldehyde-3- clones were selected against Puromycine (0.5 lg/mL). phosphate dehydrogenase (GAPDH). RT-PCR was performed with BioTaq DNA Polymerase (Boline, London, UK). Immunoprecipitation and HDAC activity assay Primers for semiquantitative RT-PCR included actin Immunoprecipitation experiments were performed as previ- forward primer (5′-ACC TGG CCG TCA GGC AGC TC- ously reported (Battaglioli et al. 2002). HeLa cells or cortical 3′) and actin reverse primer (5′-CCG AGC GTG GCT ACA neurons were harvested in IP buffer (10% glycerol NaCl GCT TC-3′). Primers for real-time RT-PCR included RPSA 150 mM, imidazole 10 mM, 0.5 mM EDTA 0.5% Triton- forward primer (5′-ACCCAGAGGAGATTGAGAAGG-3′) X100, dithiothreitol 0.5 mM) supplemented with 1 mM and RPSA reverse (5′-TGGGGAACTGCTGAATGGGC-3′), phenylmethylsulfonyl fluoride and 1x Protease Inhibitors GAPDH forward primer (5′-GGAAACCCATCAC- Cocktail and 1x Phosphatase Inhibitor Cocktail (Roche). CATCTTCC-3′) and GAPDH reverse (5′-GAAGGGGCGG- 0.5 mg of cellular extract were reacted and collected with AGATGATGACC-3′), CK5R1 forward primer (5′- HA-conjugated Agarose beads (Santa Cruz, Santa Cruz, CA, AGCCCTTCCTGGTGGAGAG -3′) and CDK5R1 reverse USA). After incubation, the beads were washed four times (5′-AAGTCAGAGAACACTTGTGTG-3′), EGR1 forward with IP buffer. The immunoprecipitates were then eluted primer (5′-TTCAATCCTCAAGGGGAGC-3′) and EGR1 with 1x sodium dodecyl sulfate Sample buffer and analyzed reverse (5′- AACCGGGTAGTTTGGCTGGGA-3′), c-FOS using western Blot. Quantification was performed using forward primer (5′-CTGCAGCCAAGTGCCGGAAT-3′) ImageJ software (ImageJ 1.43u National Institute of Health, and c-FOS reverse (5′-TTGGCAATCTCGGTCTGCAAC- Washington, DC, USA). Immunoprecipitates with 5 lgof 3′), PCTAIRE forward primer (5′-TCGTGTTCCAGTCT- the monoclonal anti-HA antibody (Santa Cruz) obtained GATCTCC-3′) and PCTAIRE reverse (5′-tcgtgttccagtct- from 2 mg of total cell extracts were tested for their histone gatctcc-3′), GRIN1 forward primer (5′- deacetylate activity (Active Motif, Carlsbad, CA, USA). GGTGGCTGTGATGCTGTAC-3′) and GRIN1 reverse (5′- Immunoprecipitates collected using Dynabeads (Invitrogen) TCCTCCTCCTCACTGTTCAC-3′), PSD95 forward primer were incubated for 60 min with a short peptide substrate (5′-CAAGATCCTGGCGGTCAAC 3′) and PSD95 reverse containing an acetylated lysine residue that can be deacet- (5′-CGTCATATGTGTTCTTCAGGG-3′). ylated by Class I, II, and IV HDAC enzymes. Once the substrate is deacetylated, the lysine reacts with the develop- Cortical neuron cultures and immunostaining ing solution and releases a fluorescent product measured Cortical neuron cultures were prepared from embryonic day using a fluorescent plate reader with an excitation wave- 18 (E18–E19) rat brain (Charles River) as previously length of 340–60 nm and emission wavelength of 440– described (Romorini et al. 2004), plated on 18 mm diameter 465 nm. We performed developing reaction for 10 min. cover slips, and grown on 12-well plastic tissue culture plates Histone deacetylase activity was normalized for immuno- (Iwaki; Bibby Sterilin, Staffordshire, UK). The neurons were precipitation efficiency on western blot using an anti-HA transfected using calcium phosphate precipitation (transfec- antibody and expressed as fold over the mock condition. The tion efficiency 1%). Cells were fixed with a phosphate-

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 606 E. Toffolo et al. buffered saline solution containing 4% paraformaldehyde for Statistics 10 min. Cells were incubated anti-HA (1 : 100 sc80; Santa GraphPad Prism 4 (GraphPad Software, San Diego, CA, Cruz) for 3 h at 23°C in gelatin dilution buffer (30 mM USA) was used for statistical analysis of data. Error bars in phosphate buffer, pH 7.4, containing 0.2% gelatin, 0.5% graphs represent SEM. Non-parametric one-way ANOVA using Triton X-100, and 0.8 M NaCl), followed by FITC-conju- Kruskal–Wallis was applied followed by Dunn’s; for Gal4- gated secondary antibodies (Jackson Laboratories, Bar based transcriptional repressor assays we applied one-way Harbor, ME, USA) for 1 h. The images were acquired using ANOVA followed by Newman-Keuls’ post-test for multiple a Zeiss LSM5 510 laser-scanning confocal microscope comparisons. (Oberkochen, Germany, generously donated by Fondazione Monzino) at 63x magnification, and an Axioplan fluorescent Results microscope at 259 magnification. Exon E8a coded Thr369b can be phosphorylated in the brain Reporter gene assays Three-dimensional crystal structure analysis of recombinant Cortical neurons were cultured and transfected at DIV4 using LSD1/CoREST proteins indicated that the four additional calcium phosphate precipitation. 5xUAS-TK-LUC reporter amino acids of exon E8a form a loop protruding out of the plasmid (Chen et al. 1998) was used at the indicated molar protein surface on the rim of the active site cleft. Importantly, ratio relative to the expression plasmids pGal4-LSD1, pGal4- such a loop appeared to be fully accessible for possible post- LSD1-8a, and mutants. Control experiments were carried out translational modifications (PTM) (Zibetti et al. 2010). For by using equivalent molar amounts of pGal4 empty vectors. consistency with the previous publications and with num- DNA was kept constant by adding pBSIIKS (Stratagene) in bering scheme used for the coordinates deposited with the every experiment; pRL-TK-vector (Promega, Madison, MI, Protein Data Bank, we shall number the four residues of the USA) reporter vector was co-transfected to normalize for exon E8a, inserted between Ala369 and Val370, as Asp369a- transfection efficiency. The luciferase reporter activity was Thr369b-Val369c-Lys369d. We scanned the 856 amino acid- determined with the Dual-Luciferase reporter assay system long LSD1-8a variant using NetPhos v2.0 software (http:// (Promega) according to the manufacturer’s instructions. For www.cbs.dtu.dk/services/NetPhos) to identify putative phos- each construct, the values of Firefly luciferase were normal- phorylation sites. Thr369b coded by exon E8a was predicted ized over Renilla luciferase (both expressed as relative as a high-score potential site of phosphorylation (Fig. 1a). In luminescent units). The activity of each construct was order to verify this prediction, rat brain tissues collected at expressed as a percentage of the promoter-less plasmid P1, when neuronal LSD1-8a protein is highly expressed, pGal4-vector. were used to immuno-isolate LSD1 complexes with a pan- LSD1 antibody. Immunocomplexes were separated by Biochemical assays and structural analysis of the sodium dodecyl sulfate–polyacrylamide gel electrophoresis, Thr369bAsp phospho-mimetic mutant and the 110 KDa band corresponding to LSD1 was Recombinant LSD1-8a/CoREST proteins were expressed in processed for mass spectrometry. The analysis indicated E. coli and purified following the same protocols used for the the presence of LSD1 peptides containing phosphorylated expression and purification of wild-type LSD1/CoREST and non-phosphorylated exon E8a (Fig. 1b and c). More- (Zibetti et al. 2010). Likewise, the enzymatic activities were over, Tyr363 was also phosphorylated, in line with its very measured as described (Forneris et al. 2007). Crystals of high predicted phosphorylation score (Fig. 1a and b). Thr369bAsp LSD1/CoREST were grown at 20°C by hang- ing-drop vapor diffusion method by mixing equal volumes of Phosphorylation of Thr369b alters the local three- protein samples with reservoir solutions containing 1.2 M dimensional structure of LSD1-8a sodium/potassium tartrate and 100 mM N-(2-acetamido) To investigate the possible functional and structural effects of iminodiacetic acid, pH 6.5 (Forneris et al. 2007). Crystals Thr369b phosphorylation, we generated a phospho-mimetic were flash cooled in liquid nitrogen and data collections were mutant (Greif et al. 2004) carrying an Asp residue at position performed at the beam-lines of the European Synchrotron 369b (Thr369bAsp). Recombinant LSD1-8a-Thr369bAsp Radiation Facility and of the Swiss Light Source. Data was tandem-affinity purified with CoREST (amino acids processing and crystallographic refinements were performed 308–440) (Forneris et al. 2007) and its enzymatic properties using programs of the CCP4 package (Winn et al. 2011). were evaluated by measuring in vitro demethylase activity The crystallographic statistics are the following: resolution, using a 21 amino acid monomethylated H3-Lys4 peptide as 3.1 A; number of unique reflections, 44180; Rmerge, 0.102 substrate. We observed no large alterations in the enzymatic (0.601 in the highest resolution shell); multiplicity 4.5 (4.1); functional properties as gathered from the comparison of kcat

Rfactor and Rfree of the reﬁned model, 0.209 and 0.226; and Km values (Table 1). We also solved the crystal structure root-mean-square deviation for bond-lengths and bond- of the mutant protein complex, which indicated no signiﬁcant angles, 0.005 A and 0.91°. overall (including the active site) conformational changes

(a)

(b) (c)

(d)

Fig. 1 Lysine-Specific Demethylase 1 (LSD1)-8a Thr369b coded by Ser, Tyr) or a decrease of 18 units because of the loss of the exon E8a can be phosphorylated in the brain. (a) Probability plot of phosphate and a water molecule (Thr, Ser). The molecular weights are phosphorylation sites was computed on NetPhos v.2.0 software given as monoisotopic mass. Cys_CAM indicates the presence of a spanning through the 856 amino-acid long LSD1-8a compared to carbamidomethylated Cys residue because of the derivatization of the conventional LSD1. (b) Mass spectrometry analysis of LSD1 immuno- protein by iodoacetamide as described in Methods. (c) Spectrum of the purified from post-natal day 1 rat brain revealed peptides containing peptide 360–369 in phosphorylated (1725.10) and non-phosphorylated phosphorylated and non-phosphorylated exon E8a. The presence of (1566.04) state. (d) Thr369b position is indicated relatively to exons one phosphate group leads to an increase in mass of 80 units (Thr, and functional domains in hLSD1. with respect to wild-type LSD1-8a/CoREST (Fig. 2a and b). However, the mutation causes a significant alteration in the The room-mean-square deviation calculated from the super- area surrounding the exon site. In particular, the loop of the position of the two structures is 0.61 A for 667 Ca atoms. Thr369bAsp mutant is shifted up to 5.1 A (for Ca atom of

Gly367), remaining fully solvent-exposed. Furthermore, the and the phosphorylation site may inﬂuence in vivo the adjacent a-helix 721–735 also moves by 1.1 A. These local interaction of the phosphorylated protein with its corepres- conformational changes most likely reﬂect the altered sors CoREST, HDAC1, and HDAC2 or may be instrumental electrostatics associated with presence of the negatively to the generation of a properly structured site for recognition charged Asp side-chain in place of a Thr. They can be of one or more unknown protein partners. especially relevant in light of the hypothesis that exon E8a Phosphorylation of Thr369b induces CoREST, HDAC1, and HDAC2 disassembly Table 1 Kinetic parameters of the LSD1 and LSD1-8a mutants To determine whether the structural alteration induced by the phospho-mimetic mutation Thr369bAsp on LSD1-8a affects k À1 a K l a cat (min ) m ( M) its ability to recruit CoREST, HDAC1, and HDAC2, we

LSD1d+CoRESTc 7.35 Æ 0.28b 5.12 Æ 1.04b performed immunoprecipitation assays using HeLa cells LSD1-8ad + CoRESTc 5.19 Æ 0.48 4.55 Æ 1.65 expressing HA-tagged LSD1, LSD1-8a, or the phospho- LSD1-8a-T369bDd+CoRESTc 2.44 Æ 0.04 3.50 Æ 0.30 mimetic LSD1-8a-Thr369bAsp as well as phospho-defective LSD1-K661Ad +CoRESTc Inactive – LSD1-8a-Thr369bAla mutants. We chose HeLa cells mainly LSD1-8a-K661Ad +CoRESTc Inactive – because of the experimental manipulability of the cell type LSD1-8a- T369bA-K661Ad+CoRESTc Inactive – when performing biochemical studies. Using anti-HA conjugated agarose beads, we isolated LSD1 complexes in aSteady-state kinetic parameters were determined as previously described using a 21 amino acid monomethylated H3-Lys4 peptide protein extracts obtained from HeLa clones, each stably (Forneris et al. 2005b). expressing the described HA-LSD1 isoform and mutants. As bData taken from Forneris et al. 2007 (Forneris et al. 2007). mock condition, we used HeLa clones stably expressing the cCoREST from aa 308 to 440. HA-tag from the pCGN empty vector. In accordance with dLSD1 from aa 171–840. previously published data (Zibetti et al. 2010), LSD1 and

(a) (b)

Fig. 2 Structure. (a) Overall crystal structure of LSD1-8a–CoREST outlined by a circle. (b) The superposition of the LSD1-8a/CoREST Thr369bAsp mutant solved at 3.1 A resolution (PDB entry 4bay). (colored in yellow; PDB entry 2XOL) and LSD1-8a-Thr369bAsp/ Mutant LSD1-8a (residues 171–840) is colored in blue, whereas CoREST (blue) three-dimensional structures highlights the local CoREST (residues 308–440) in red. The FAD cofactor is shown as conformational change in the residues surrounding the Thr369bAsp yellow ball-and-stick. The insertion site of E8a (residues Asp369a- mutation. Generated with CCP4 mg (Winn et al. 2011). Thr369b-Val369c-Lys369d) and the site of mutation Thr369bAsp is

(a)

(b)

(d)

(c)

Fig. 3 Phosphorylation of Thr369b induces CoREST, HDAC1 and immunocomplexes from rat cortical neurons virally transduced with HDAC2 detachment from LSD1-8a and reduces associated deacety- HA-LSD1 or HA-LSD1-8a-Thr369bAsp, probed with the indicated lase activity. (a) HA-immunocomplexes from HeLa clones stably antibodies. The asterisk indicates the 55 KDa band of HDAC2. The expressing HA-LSD1, HA-LSD1-8a, HA-LSD1-8a-Thr369bAsp, and slower migrating is a non-specific band. (d) Histone deacetylase HA-LSD1-8a-Thr369bAla, probed with the indicated antibodies. (b) activity associated to immunocomplexes from HeLa cells, normalized Quantification of co-immunoprecipitated factors normalized for immu- for immunoprecipitation efficiency, expressed as ratio of mock condi- noprecipitation efficiency, expressed as ratio of LSD1. Mean val- tion. Mean values Æ SEM. Statistical analysis revealed significant ues Æ SEM. Statistical analysis revealed significant differences differences (**p < 0.01 vs. Vector; #p < 0.05 vs. LSD1-8a). (b–d) (*p < 0.05, **p < 0.01 vs. LSD1; #p < 0.05 vs. LSD1-8a). (c) HA- Kruskal–Wallis one-way ANOVA followed by Dunn’s post hoc test.

LSD1-8a immunocomplexes were found to contain similar vs. 0.68 Æ 0.16; HDAC1 0.11 Æ 0.01 vs. 0.81 Æ 0.03; amounts of CoREST, HDAC1, and HDAC2 (Fig. 3a, b; HDAC2 0.25 Æ 0.1 vs. 1.08 Æ 0.04, p < 0.05; LSD1-8a- LSD1 vs. LSD1-8a: CoREST 1 Æ 0.23 vs. 0.68 Æ 0.16; Thr369bAsp vs. LSD1: CoREST 0.17 Æ 0.04 vs. 1 Æ 0.23, HDAC1 1 Æ 0.08 vs. 0.81 Æ 0.03; HDAC2 1 Æ 0.04 vs. p < 0.01; HDAC1 0.11 Æ 0.01 vs. 1 Æ 0.08, p < 0.05; 1.08 Æ 0.04). Remarkably, the immunocomplexes isolated HDAC2 0.25 Æ 0.1 vs. 1 Æ 0.04; p < 0.05). Conversely, by the phospho-mimetic LSD1-8a-Thr369bAsp mutant the phospho-defective mutant LSD1-8a-Thr369bAla did not showed almost complete lack of the three co-repressors show differential association with corepressors, compared to when compared to either LSD1-8a or LSD1 (Fig. 3a, b; LSD1-8a (Fig. 3a, b; LSD1-8a-Thr369bAla vs. LSD1-8a: LSD1-8a-Thr369bAsp vs. LSD1-8a: CoREST 0.17 Æ 0.04 CoREST 0.71 Æ 0.09 vs. 0.68 Æ 0.16; HDAC1

0.82 Æ 0.003 vs. 0.81 Æ 0.03; HDAC2 0.85 Æ 0.003 vs. (Active Motif, Inc.) as a substrate, we found that residual 1.08 Æ 0.04). These quantifications derived from at least enzymatic activity bound to HA-LSD1-8a-Thr369bAsp was three independent experiments with two independent HeLa comparable to the mock condition and reduced compared to clones for each isoform or mutant. To confirm that phospho- LSD1 and LSD1-8a (Fig. 3d; Mock vs. LSD1 1 vs. mimetic LSD1-8a-Thr369bAsp mutant is unable to assemble 15.3 Æ 3.5; p < 0.01; Mock vs. LSD1-8a 1 vs. 7.8 Æ 1.5; CoREST/HDAC1/2 corepressor complex, we used anti-HA Mock vs. LSD1-8a-Thr369bAsp: 1 vs. 1.9 Æ 0.5, p < 0.05). conjugated agarose beads to isolate LSD1 complexes from Note that the difference in HDAC activity associated to protein extracts obtained from primary cortical neurons, LSD1-8a showed a tendency towards reduction compared to virally transduced with HA-tagged LSD1, or the phospho- LSD1. We cannot exclude that, at least in part, over- mimetic LSD1-8a-Thr369bAsp mutant. As for HeLa cells, expressed LSD1-8a could be phosphorylated in HeLa cells we found that also neuronal LSD1-8a-Thr369bAsp immu- being responsible for partial loss of HDAC activity. nocomplexes contained largely reduced amounts of Co- These results strongly suggest that the presence of the REST, and HDAC2 (Fig. 3c). phospho-mimetic mutation renders LSD1-8a unable to Given the possibility that other deacetylases might still be recruit any deacetylase activity. In this regard, it should be present in the LSD1-8a-Thr369bAsp complex or might be de noticed that a short fragment of human CoREST (residues novo recruited, we measured residual deacetylase activity in 308–440) is able to bind in vitro the phospho-mimetic LSD1- HeLa cells. Using fluorescent acetylated histone peptides 8a mutant (Fig. 2) indicating either that full-length CoREST

(a)

(b)

(c)

Fig. 4 Repressive activity of LSD1s isoforms is modulated by phos- and mutant proteins used. (c) RNA transcripts from cortical neurons phorylation. (a) LSD1 isoforms and mutants fused to Gal4 were after viral transduction with mock, LSD1, LSD1-8a-Thr369bAsp, and assayed for their ability to repress the 5xUAS-TK-LUC reporter gene in LSD1-8a-Thr369bAla for indicated genes (Cdk5R1, EGR1, c-FOS, rat cortical neurons at different reporter:repressor molar ratios (1 : 1 PCTAIRE, Grin1, PSD95) were quantified by reverse-transcription and 1 : 2). The luciferase activity normalized on the co-transfected qPCR. Transcript levels were normalized to RPSA and expressed as renilla reporter is expressed as a percentage of the Gal4 empty vector. the fold expression relative to mock. Pooled data show mean Æ SEM. Mean values Æ SEM. Statistical analysis revealed significant differ- Statistical analysis revealed significant differences (**p < 0.01, § §§ ences (***p < 0.001 vs. Gal4-Vector; #p < 0.05, ##p < 0.01, ***p < 0.001 vs. LSD1; #p < 0.05 vs. Vector; p < 0.05, p < 0.01 ###p < 0.001 vs. LSD1-8a). One-way ANOVA followed by Newman– vs. LSD1-8a-Thr369bAsp). (a–c) Kruskal–Wallis one-way ANOVA fol- Keuls’ post hoc test. (b) Protein expression levels of LSD1 isoforms lowed by Dunn’s post hoc test.

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 Splicing and phosphorylation modulate LSD1/KDM1A 611 is required to be disassembled, or that in the cellular context, gave a clear indication about the opposite effect exerted by an additional factor is recruited by the phosphorylated loop phospho-defective and phospho-mimetic mutants. LSD1-8a- leading to CoREST displacement. Thr369bAla negatively modulated gene transcription, whereas LSD1-8a-Thr369bAsp over-expression induced Phosphorylation of Thr369b on LSD1-8a modulates up-regulation of the same genes, behaving as a dominant transcriptional repression in neurons negative isoform. These data are consistent with the reporter- The primary known function of LSD1 is to repress gene assays and provide physiological evidence of the transcription through an epigenetic mechanism based on phosphorylation role in regulating neuronal morphogenesis both histone H3-Lys4 demethylation and recruitment of through the modulation of morphogenesis-related genes. HDAC1/2 (Shi et al. 2004; Forneris et al. 2005a; Lee et al. 2006a). Consistently, both LSD1 and LSD1-8a are capable to Phosphorylation of Thr369b drives morphogenesis in repress transcription in neurons. We therefore tested whether cortical neurons phosphorylation, releasing the interaction with HDAC1/2 as We previously demonstrated that the presence of exon E8a well as with CoREST, could hamper the ability of LSD1-8a provides LSD1 with the ability to modulate neurite mor- to function as a corepressor in neurons where LSD1-8a is phology (Zibetti et al. 2010). On these bases, we asked physiologically expressed. We generated Gal4-LSD1-8a whether phosphorylation regulates maturation in rat cortical mutants (Thr369bAsp and Thr369bAla) mimicking a phospho- neurons. Based on immunofluorescence and quantitative rylated and unphosphorylated status, and evaluated their protein measurements, all enzyme variants over-expressed in repressive strength on the basis of the 5X-UAS-TK-Luc cortical neurons showed nuclear localization and similar reporter gene expression in rat cortical neurons. In each levels of expression (Figs. 5a, b). In line with previous experiment, the promoter activity was measured as luciferase findings (Zibetti et al. 2010), the neurospecific LSD1-8a but activity normalized over a co-transfected renilla reporter. not LSD1 had a significant effect on neuronal maturation as Repressive strength was inferred through luciferase reduction evaluated by cumulative neurite length (Fig. 5a and c; mock upon normalization over the promoter activity in the vs. LSD1-8a, 909 Æ 128 lm vs. 1740 Æ 64 lm, p < 0.001) presence of the Gal4 empty vector (Fig 4a, b). Thr369bAsp and the number of neurite branches (Fig. 5a and d; mock vs. phospho-mimetic mutation was found to cause loss of LSD1-8a, 6.6 Æ 1.1 vs. 14.9 Æ 1.2, p < 0.001). The repressive activity compared to LSD1-8a (Fig. 4a; LSD1- phospho-mimetic mutant (Thr369bAsp) showed a further 8a-Thr369bAsp vs. LSD1-8a at 1 : 1 molar ratio tendency to enhance morphogenic effect of LSD1-8a on 92.45 Æ 4.56 vs. 74.49 Æ 9.35, p < 0.05), becoming similar cumulative neurite length (Fig. 5a and c; LSD1-8a vs. LSD1- to the Gal4 empty vector. Conversely, Thr369bAla mutation 8a-Thr369bAsp, 1740 Æ 64 lm vs. 1935 Æ 50.7 lm) and caused a significant increase in repressive activity compared number of neurite branches (Fig. 5a and d; LSD1-8a vs. to LSD1-8a (Fig. 4a LSD1-8a-Thr369bAla vs. LSD1-8a at LSD1-8a-Thr369bAsp 14.9 Æ 0.7 vs. 17.8 Æ 1.2). On the 1 : 1 molar ratio 53.64 Æ 4.23 vs. 74.49 Æ 9.35, p < 0.001. contrary, the phospho-defective mutation Thr369bAla abol- These data suggest that phosphorylation of Thr369b reduces ished the pro-maturation effect played by LSD1-8a, as no repressive activity of neuronal LSD1-8a. statistical difference could be found when comparing mock To further verify this notion, we infected cortical neurons conditions with LSD1-8a-Thr369bAla (Fig. 5a and c, LSD1- with lentiviral vectors carrying HA-LSD1, HA-LSD1-8a- 8a-Thr369bAla vs. mock, 889 Æ 45 vs. 909 Æ 128 lm; Thr369bAsp, HA-LSD1-8a-Thr369bAla, or green fluores- Fig. 5a and d, LSD1-8a-Thr369bAla vs. mock 7.0 Æ 0.6 vs. cent protein (mock condition) at DIV4 and we analyzed the 6.6 Æ 1.1) providing further evidence for transfected HA- effect of isoforms over-expression on transcription of LSD1-8a phosphorylation. These data suggest that phos- selected known HDAC2 brain targets. Indeed, brain HDAC2 phorylation is indeed necessary to induce morphogenesis in plays a central role as negative regulator of complex cortical neurons. processes such as memory formation and neuronal maturation (Guan et al. 2009). Total RNA was analyzed at DIV8 by Demethylase inactivating mutation phenocopy LSD1-8a qRT-PCR. We probed genes involved in neuronal maturation phosphorylation and morphology (CDK5R1 and PCTAIRE), neuronal func- The data hereby presented suggest that Thr369b phospho- tion (GRIN1, PSD95) and two immediate early genes rylation impairs LSD1-8a mediated transcriptional repression involved in plasticity c-FOS and EGR1, which has already by displacing histone deacetylase activity from the corepres- been validated as LSD1 direct target outside the nervous sor complex and promoting neuronal maturation and mor- system (Lee et al. 2006b). Transcriptional activity is phogenesis. LSD1 transcriptional repression requires the described as fold gene expression normalized to the house- cooperative action of histone H3-Lys4 demethylation and keeping gene RPSA, relative to mock condition (Fig. 4c; HDAC1/2 activity (Forneris et al. 2005a; Lee et al. 2005, similar results were obtained normalizing gene expression to 2006a; Shi et al. 2005). Therefore, we reasoned that a GAPDH as housekeeping, not shown). These experiments demethylase inactivating mutation should be able to revert

(a)

(b)

Fig. 5 Phosphorylation at Thr369b modulates LSD1-8a morphogenetic properties in neurons. Rat cortical neurons were co-transfected with pCGN vector (mock), HA-LSD1, HA-LSD1-8a, the phospho-mimetic HA-LSD1-8a-Thr369bAsp mutant and the phospho-defective HA-LSD1-8a- Thr369bAla mutant together with pEGFP. (a) Morphological analysis at DIV8 for EGFP- and HA-positive neurons. Scale bars: 20 lm. (b) Protein expression level of wild types and mutant isoforms. (c, d) Quantification of cumulative neurite length and number of secondary branches is indicated Æ SEM. ***p < 0.001 vs. mock; §§§p < 0.001 vs. LSD1; ###p < 0.001 vs. HA- LSD1-8a-Thr369bAsp. Kruskal–Wallis test one- way ANOVA followed by Dunn’s post hoc test was applied to values. the effect of the phospho-defective mutation of LSD1-8a. To that demethylase inactivating mutation is able to phenocopy this aim, we introduced a demethylase inactivating mutation LSD1-8a phosphorylation, demonstrating that interfering (Lys661Ala) (Lee et al. 2005) in the phospho-defective with the ability of LSD1-8a to repress transcription, either mutant (LSD1-8a-Thr369bAla). First, we verified that by Thr369b phosphorylation or by a demethylase inactivat- recombinant LSD1-8a-Thr369bAla/Lys661Ala is a deme- ing mutation, promotes neuronal maturation and morpho- thylase dead mutant, evaluating its activity on histone H3 genesis through a dominant negative effect. peptide (Table 1). Next, we showed that the mutant is unable to repress transcription in neurons (Fig. 6a; LSD1-8a- Discussion Thr369bAla vs. LSD1-8a-Thr369bAla/Lys661Ala; 53.64 Æ 4.23 vs. 107.58 Æ 6.35, p < 0.001), whereas it can partially This work explores the biology underlying the isoform- revert the phenotype of the phospho-defective mutant specific properties enabling neuronal LSD1-8a to modulate promoting maturation as shown by cumulative neurite length transcriptional pathways regulating morphogenesis in neu- (Fig. 6c and d; LSD1-8a-Thr369bAla vs. LSD1-8a- rons. Purifying LSD1 directly from rat brain, we demonstrate Thr369bAla/Lys661Ala 889 Æ 45 lm vs. 1530 Æ 74 lm, by mass spectrometry that a threonine residue coded by exon p < 0.001) and number of neurite branches (Fig. 6c and e; E8a, Thr369b, can be phosphorylated in vivo. Maturation LSD1-8a-Thr369bAla vs. LSD1-8a-Thr369bAla/Lys661Ala experiments in cortical neurons provide a perspective on the 7.0 Æ 0.6 vs. 12.0 Æ 0.8, p < 0.001). This result showed physiological consequence of LSD1-8a phosphorylation

(a)

(b)

Fig. 6 Effect of demethylase inactivating mutation in Lysine-Specific Demethylase 1 (LSD1)-8a on (c) repressive activity and neuronal maturation in cortical neurons. (a) LSD1-8a-Thr369bAla and a double LSD1-8a-Thr369bAla-Lys661Ala demethylase null mutant, fused to Gal4, were assayed for their ability to repress the 5xUAS-TK- LUC reporter gene in rat cortical neurons. The luciferase activity normalized on the co-transfected renilla reporter is expressed as a percentage of the Gal4 empty vector. Mean values Æ SEM. Statistical analysis revealed significant differences (**p < 0.01 vs. Gal4-Vector; (d) (e) ###p < 0.001 vs. LSD1-8a-Thr369bAla). (b) Protein expression level of mutant isoforms. (c) Morphological analysis at DIV8 for EGFP- and HA- positive neurons. Neurons were co-transfected with pCGN vector (mock), the phospho-defective HA-LSD1-8a-Thr369bAla mutant and the phospho-defective/demethylase null HA-LSD1- 8a-Thr369bAla-Lys661Ala mutant together with pEGFP. Scale bars: 20 lm. (d, e) Quantification of cumulative neurite length and number of secondary branches is indicated Æ SEM (*p < 0.05 vs. mock; ###p < 0.001 vs. HA-LSD1- 8aThr369bAla). (c–e) Kruskal–Wallis one-way ANOVA followed by Dunn’s post hoc test was applied to values.

during neuronal morphogenesis, unraveling that the pro- functionally relevant. Indeed, LSD1-8a phospho-mimetic maturation effect exerted by LSD1-8a depends on phospho- mutant looses the ability to recruit corepressors CoREST, rylation of Thr369b of the transfected protein. Indeed, HDAC1, and HDAC2. We suggest that complex disassembly over-expression in cortical neurons of the unphosphorylable could be directly mediated by phosphorylation or that mutant (Thr369bAla) knocks out the LSD1-8a pro-maturation phosphorylation, being instrumental to the generation of a effect suggesting that part of the LSD1-8a over-expressed properly structured protein-binding site, allows recruitment protein must be phosphorylated in vivo. In line with partial of a new interactor, which could directly or indirectly compete phosphorylation of the transfected LSD1-8a protein, over- for the binding with corepressors. The functional conse- expression of the phospho-mimetic mutant (Thr369bAsp) quence of complex disassembly on repressive activity stems further enhances LSD1-8a pro-maturation properties. from the tight functional interdependence and synergistic The crystal structure analysis of LSD1-8a shows that interplay between LSD1-mediated histone demethylation and substitution of Thr369b with an Asp side chain leads to a histone deacetylation (Forneris et al. 2005a; Shi et al. 2005; local conformational change, which affects the residues Lee et al. 2006a) that provides the biochemical bases to surrounding the phosphorylation site without altering the explain how loss of deacetylase activity is per se sufﬁcient to catalytic pocket and its demethylase activity in vitro. In vivo, impair repression. Indeed, it has been demonstrated that using endogenous proteins, we demonstrate that the struc- histone deacetylation is required for proper H3-Lys4 deme- tural alteration caused by the phospho-mimetic mutation, is thylation and the enzymatic activity of LSD1 is required for

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 614 E. Toffolo et al. optimal activity of histone deacetylases. We suggest that the 2001). When Cdk5 kinase activity is reduced by over- sole histone deacetylase loss is sufficient to explain in vivo expressing dominant-negative Cdk5 mutants as well as lack of repression and pro-maturation effect. Further proof antisense oligonucleotides of Cdk5 and CDK5R1 in cultured that the synergistic interplay between demethylase and primary neurons, neuronal morphogenesis is inhibited (Nik- deacetylase activities is required to repress gene transcrip- olic et al. 1996). CDK5R1 protein has been described as a tion, is that pro-maturation effect can be achieved also by highly unstable protein with a very short half-life of about knocking out LSD1-8a repressive function interfering with 20–30 min (Patrick et al. 1998), in this frame importance of its demethylase activity. Indeed, we were able to phenocopy regulating CDK5R1 transcription is maximal (Dhavan and the pro-maturation effect of the phospho-mimetic mutation Tsai 2001). All these evidences are consistent with the with a paradigmatic demethylase-null mutant (Lys661Ala) hypothesis that morphogenic properties of the phospho- (Lee et al. 2005). In conclusion, thanks to the phospho- mutant LSD1-8a (Thr369bAsp) could be exerted by posi- mimetic and the demethylase-null mutants, we here demon- tively regulating Cdk5 activator CDK5R1. In vivo Pctaire1 is strate that LSD1-8a-mediated repression requires both enriched at the level of neurites growth cones of early histone H3-Lys4 demethylase and deacetylase activities to differentiating post-mitotic hippocampal neurons. Indeed, functionally regulate gene transcription and morphogenesis. during morphogenesis, Pctaire1 co-localizes with synaptic These findings also underline the dual function of LSD1-8a marker in dendrites (Fu et al. 2011). Expression of Pctaire as histone demethylase enzyme and as scaffold protein. increases in the perinatal window (Graeser et al. 2002) being Phosphorylation, regulating disassembly of LSD1-8a from its concomitant with the elevation of LSD1-8a levels in vivo corepressor partners, represents a powerful tool to regulate (Zibetti et al. 2010). Moreover, knock-down of Pctaire1 LSD1-8a repressive potential. Since over-expression of the abolishes dendrite development (Fu et al. 2011), pheno- phospho-mimetic mutant (Thr369bAsp) in cortical neurons copying morphological outcome of LSD1-8a down-regula- increases target genes transcription, we propose that tion on neurite growth and branching in cortical neurons phospho-LSD1-8a behaves as a transient dominant negative (Zibetti et al. 2010). Altogether, these data are consistent isoform of LSD1-8a, competing with the endogenous with the idea that LSD1-8a pro-maturation properties could unphosphorylated form for the binding to the chromatin be mediated, at least in part, by modulating Pctaire substrate. Dimeric nature of the LSD1/CoREST/HDAC1/2 transcription together with CDK5R1 levels in neurons. It complex has been unraveled (Humphrey et al. 2001) and we would be of tremendous relevance to better characterize demonstrated that this corepressor complex can encompass transcriptional regulatory pathways providing LSD1-8a with two identical or different LSD1 isoforms (Zibetti et al. the ability to regulate morphogenic genes and it will be of 2010). Therefore, the dominant negative action of phospho- particular impact to unravel transcription factors that endow LSD1-8a might indirectly affect also LSD1 repressive LSD1 with gene specificity and that are in turn coregulated activity. Indeed, in cortical neurons, genes whose transcrip- by LSD1 epigenetic activity. tion is increased upon phospho-mimetic mutant LSD1-8a Studies directed toward the identification of the kinase (Thr369bAsp) over-expression, are repressed by the over- acting on LSD1-8a will also be crucial. In this context, we expression of LSD1. notice that the bioinformatic tool NetPhosK (http://www.cbs. Signal-dependent displacement of HDAC2 from chromatin dtu.dk/services/NetPhosK/) strongly predicts protein kinase has already been described as fine strategy to regulate gene C as the likely candidate for Thr369b phosphorylation (Saito transcription in response to stimuli (Palacios et al. 2010). In & Shirai 2002, Schmalz et al. 1998, Leach et al. 1989), particular in neurons, NO production and S-nitrosylation of providing a first clue for future studies along this research HDAC2 triggers its release from target gene promoters, and direction. the activation of genes that are associated with neuronal The key conclusion of our studies is that exon E8a development (Nott et al. 2008). In this regard, we suggest alternative splicing and phosphorylation of Thr369b coded that phosphorylation of neuronal LSD1-8a, inducing disso- by the same exon, provide a fine dual mechanism to set the ciation of HDAC1/2 and CoREST, could account for an LSD1 repressive potential in the brain. During a specific additional strategy to get rid of HDAC1/2 from targets. developmental window, which correlates with neuronal On a biological point of view, LSD1-8a morphogenic maturation, alternative splicing of exon E8a is dynamically properties relies on the transcriptional regulation of genes controlled (Zibetti et al. 2010). During the adult life LSD1 regulating morphogenesis such as CDK5R1 (Dhavan and and neuronal LSD1-8a reach a specific equilibrium, therefore Tsai 2001) and PCTAIRE (Graeser et al. 2002; Mokalled LSD1-8a repressive potential could be modulated by external et al. 2010). CDK5R1 is the main cellular activator of Cdk5 signals promoting phosphorylation or de-phosphorylation. kinase. Cdk5 displays pleiotropic effect, which is evident However, we cannot exclude that upon specific stimuli, also given the variety of neuronal processes that it drives, exon E8a alternative splicing could respond, increasing or spanning from axon guidance to actin dynamics passing decreasing the total amount of neuronal LSD1-8a and through synapse shaping and plasticity (Dhavan and Tsai consequently LSD1. Indeed, several examples of activity or

© 2013 International Society for Neurochemistry, J. Neurochem. (2014) 128, 603--616 Splicing and phosphorylation modulate LSD1/KDM1A 615 developmentally regulated neuronal splicing factors promote Benayoun B. A. and Veitia R. A. (2009) A post-translational the inclusion of alternative exons, many of which contain modification code for transcription factors: sorting through a sea 19 – regulatory phosphorylation sites (Boutz et al. 2007; Zhang of signals. Trends Cell Biol. , 189 197. Boutz P. L., Stoilov P., Li Q., Lin C. H., Chawla G., Ostrow K., Shiue et al. 2008; Lee et al. 2009). L., Ares M. and Black D. L. (2007) A post-transcriptional Whereas there are numerous examples on the role of regulatory switch in polypyrimidine tract-binding proteins PTMs in regulating transcription factor functions (Benayoun reprograms alternative splicing in developing neurons. Genes and Veitia 2009; Everett et al. 2009), our characterization of Dev. 21, 1636–1652. the post-transcriptional and post-translational mechanism Caretti G., Palacios D., Sartorelli V. and Puri P. L. (2011) Phosphoryl- EZH-ion. Cell Stem Cell 8, 262–265. regulating LSD1 function represents one of the very few Chen G., Nguyen P. H. and Courey A. J. (1998) A role for Groucho cases of epigenetic enzyme modulation by PTMs (Nott et al. tetramerization in transcriptional repression. Mol. Cell. Biol. 18, 2008; Guan et al. 2009; Caretti et al. 2011). The most 7259–7268. fascinating aspect of our findings is that, to the best of our Chosed R. and Dent S. Y. (2007) A two-way street: LSD1 regulates fi chromatin boundary formation in S. pombe and Drosophila. Mol. knowledge, neuronal LSD1-8a is the rst example of a 26 – fi Cell , 160 162. mammalian epigenetic enzyme speci cally devoted to neu- Dhavan R. and Tsai L. H. (2001) A decade of CDK5. Nat. Rev. Mol. Cell rons that is potentially capable of responding to intrinsic or Biol. 2, 749–759. extrinsic cell signals, enabling adaptive transcriptional Everett L., Vo A. and Hannenhalli S. (2009) PTM-Switchboard–a responses in neurons by modifying chromatin. The neuro- database of posttranslational modifications of transcription factors, 37 specific alternative splicing of exon E8a therefore creates a the mediating enzymes and target genes. Nucleic Acids Res. , D66–D71. clade-restricted, developmentally regulated PTM modulatory Forneris F., Binda C., Vanoni M. A., Battaglioli E. and Mattevi A. site that may contribute to the increased complexity of (2005a) Human histone demethylase LSD1 reads the histone code. mammalian brain (Barbosa-Morais et al. 2012). The nature J. Biol. Chem. 280, 41360–41365. of the extracellular signals that modulate the epigenetic Forneris F., Binda C., Vanoni M. A., Mattevi A. and Battaglioli E. landscape in the nervous system during development, (2005b) Histone demethylation catalysed by LSD1 is a flavin-dependent oxidative process. FEBS Lett. 579, 2203– learning and memory has just begun to be characterized 2207. (Nott et al. 2008; Riccio 2010) and given its specificity, Forneris F., Binda C., Adamo A., Battaglioli E. and Mattevi A. (2007) neuronal LSD1-8a phosphorylation represents a model to Structural basis of LSD1-CoREST selectivity in histone H3 decipher these pathways. recognition. J. Biol. Chem. 282, 20070–20074. Forneris F., Battaglioli E., Mattevi A. and Binda C. (2009) New roles of flavoproteins in molecular cell biology: histone demethylase LSD1 Acknowledgements and chromatin. FEBS J. 276, 4304–4312. Fu W. Y., Cheng K., Fu A. K. and Ip N. Y. (2011) Cyclin-dependent This study was supported by the Epigenomics Flagship Project kinase 5-dependent phosphorylation of Pctaire1 regulates dendrite EPIGEN (MIUR-CNR) (E.B. and A.M.), Fondazione Cariplo development. Neuroscience 180, 353–359. Projects 2009-2640 (E.B.) and 2010-778 (A.M.), and Associazione Fuentes P., Canovas J., Berndt F. A., Noctor S. C. and Kukuljan M. Italiana Ricerca sul Cancro (IG11342, A.M.); ‘Dote ricerca’: FSE, (2012) CoREST/LSD1 control the development of pyramidal cortical neurons. Cereb. Cortex 22, 1431–1441. Regione Lombardia (F.R.). We are very grateful to Maria Estela Godmann M., Auger V., Ferraroni-Aguiar V., Di Sauro A., Sette C., Andres, Beatrice Bodega and Emma Paola Sturani for their Behr R. and Kimmins S. (2007) Dynamic regulation of histone H3 fl stimulating discussions. The authors have no con icts of interest methylation at lysine 4 in mammalian spermatogenesis. Biol. to declare. Reprod. 77, 754–764. Graeser R., Gannon J., Poon R. Y., Dubois T., Aitken A. and Hunt T. (2002) Regulation of the CDK-related protein kinase PCTAIRE-1 References and its possible role in neurite outgrowth in Neuro-2A cells. J. Cell Sci. 115, 3479–3490. Abidi F. E., Holloway L., Moore C. A., Weaver D. D., Simensen R. J., Greif D. M., Sacks D. B. and Michel T. (2004) Calmodulin Stevenson R. E., Rogers R. C. and Schwartz C. E. (2008) phosphorylation and modulation of endothelial nitric oxide Mutations in JARID1C are associated with X-linked mental synthase catalysis. Proc. Natl Acad. Sci. USA 101, 1165–1170. retardation, short stature and hyperreflexia. J. Med. Genet. 45, Guan J. S., Haggarty S. J., Giacometti E. et al. (2009) HDAC2 787–793. negatively regulates memory formation and synaptic plasticity. Ballas N., Grunseich C., Lu D. D., Speh J. C. and Mandel G. (2005) Nature 459,55–60. REST and its corepressors mediate plasticity of neuronal gene Humphrey G. W., Wang Y., Russanova V. R., Hirai T., Qin J., chromatin throughout neurogenesis. Cell 121, 645–657. Nakatani Y. and Howard B. H. (2001) Stable histone Barbosa-Morais N. L., Irimia M., Pan Q. et al. (2012) The evolutionary deacetylase complexes distinguished by the presence of SANT landscape of alternative splicing in vertebrate species. Science 338, domain proteins CoREST/kiaa0071 and Mta-L1. J. Biol. Chem. 1587–1593. 276, 6817–6824. Battaglioli E., Andres M. E., Rose D. W., Chenoweth J. G., Rosenfeld Leach K. L., Powers E. A., Ruff V. A., Jaken S. and Kaufmann S. (1989) M. G., Anderson M. E. and Mandel G. (2002) REST repression of Type 3 protein kinase C localization to the nuclear envelope of neuronal genes requires components of the hSWI.SNF complex. J. phorbol ester-treated NIH 3T3 cells. J. Cell. Biol. 109, 685–695. Biol. Chem. 277, 41038–41045.

Lee M. G., Wynder C., Cooch N. and Shiekhattar R. (2005) An essential Romorini S., Piccoli G., Jiang M., Grossano P., Tonna N., Passafaro M., role for CoREST in nucleosomal histone 3 lysine 4 demethylation. Zhang M. and Sala C. (2004) A functional role of postsynaptic Nature 437, 432–435. density-95-guanylate kinase-associated protein complex in LeeM.G.,WynderC.,BocharD.A.,HakimiM.A.,CoochN.and regulating Shank assembly and stability to synapses. J. Neurosci. Shiekhattar R. (2006a) Functional interplay between histone 24, 9391–9404. demethylase and deacetylase enzymes. Mol. Cell. Biol. 26,6395–6402. Saito N. and Shirai Y. (2002) Protein kinase C gamma (PKC gamma): Lee M. G., Wynder C., Schmidt D. M., McCafferty D. G. and function of neuron specific isotype. J. Biochem. 132, 683–687. Shiekhattar R. (2006b) Histone H3 lysine 4 demethylation is a Saleque S., Kim J., Rooke H. M. and Orkin S. H. (2007) Epigenetic target of nonselective antidepressive medications. Chem. Biol. 13, regulation of hematopoietic differentiation by Gfi-1 and Gfi-1b is 563–567. mediated by the cofactors CoREST and LSD1. Mol. Cell 27, 562– Lee J. A., Tang Z. Z. and Black D. L. (2009) An inducible change in 572. Fox-1/A2BP1 splicing modulates the alternative splicing of Schmalz D., Hucho F. and Buchner K. (1998) Nuclear import of protein downstream neuronal target exons. Genes Dev. 23, 2284–2293. kinase C occurs by a mechanism distinct from the mechanism used Lois C., Hong E. J., Pease S., Brown E. J. and Baltimore D. (2002) by proteins with a classical nuclear localization signal. J. Cell. Sci. Germline transmission and tissue-specific expression of transgenes 111, 1823–1830. delivered by lentiviral vectors. Science 295, 868–872. Shi Y., Lan F., Matson C., Mulligan P., Whetstine J. R., Cole P. A. and Mokalled M. H., Johnson A., Kim Y., Oh J. and Olson E. N. (2010) Casero R. A. (2004) Histone demethylation mediated by the Myocardin-related transcription factors regulate the Cdk5/Pctaire1 nuclear amine oxidase homolog LSD1. Cell 119, 941–953. kinase cascade to control neurite outgrowth, neuronal migration Shi Y. J., Matson C., Lan F., Iwase S., Baba T. and Shi Y. (2005) and brain development. Development 137, 2365–2374. Regulation of LSD1 histone demethylase activity by its associated Naldini L., Blomer€ U., Gallay P., Ory D., Mulligan R., Gage F. H., factors. Mol. Cell 19, 857–864. Verma I. M. and Trono D. (1996) In vivo gene delivery and stable Sun G., Alzayady K., Stewart R., Ye P., Yang S., Li W. and Shi Y. transduction of nondividing cells by a lentiviral vector. Science (2010) Histone demethylase LSD1 regulates neural stem cell 272, 263–267. proliferation. Mol. Cell. Biol. 30, 1997–2005. Neelamegam R., Ricq E. L., Malvaez M. et al. (2012) Brain-penetrant de la Torre-Ubieta L. and Bonni A. (2011) Transcriptional regulation of LSD1 inhibitors can block memory consolidation. ACS Chem. neuronal polarity and morphogenesis in the mammalian brain. Neurosci. 3, 120–128. Neuron 72,22–40. Nikolic M., Dudek H., Kwon Y. T., Ramos Y. F. and Tsai L. H. (1996) Wang J., Hevi S., Kurash J. K. et al. (2009) The lysine demethylase The cdk5/p35 kinase is essential for neurite outgrowth during LSD1 (KDM1) is required for maintenance of global DNA neuronal differentiation. Genes Dev. 10, 816–825. methylation. Nat. Genet. 41, 125–129. Nott A., Watson P. M., Robinson J. D., Crepaldi L. and Riccio A. (2008) Winn M. D., Ballard C. C., Cowtan K. D. et al. (2011) Overview of the S-Nitrosylation of histone deacetylase 2 induces chromatin CCP4 suite and current developments. Acta Crystallogr. D Biol. remodelling in neurons. Nature 455, 411–415. Crystallogr. 67, 235–242. Palacios D., Mozzetta C., Consalvi S. et al. (2010) TNF/p38a/polycomb Wu J. I., Lessard J., Olave I. A., Qiu Z., Ghosh A., Graef I. A. and signaling to Pax7 locus in satellite cells links inflammation to the Crabtree G. R. (2007) Regulation of dendritic development by epigenetic control of muscle regeneration. Cell Stem Cell 7, 455– neuron-specific chromatin remodeling complexes. Neuron 56,94– 469. 108. Pastori V., Sangalli E., Coccetti P., Pozzi C., Nonnis S., Tedeschi G. and Yokoyama A., Takezawa S., Schule€ R., Kitagawa H. and Kato S. (2008) Fusi P. (2010) CK2 and GSK3 phosphorylation on S29 controls Transrepressive function of TLX requires the histone demethylase wild-type ATXN3 nuclear uptake. Biochim. Biophys. Acta 1802, LSD1. Mol. Cell. Biol. 28, 3995–4003. 583–592. Zhang C., Zhang Z., Castle J., Sun S., Johnson J., Krainer A. R. and Patrick G. N., Zhou P., Kwon Y. T., Howley P. M. and Tsai L. H. (1998) Zhang M. Q. (2008) Defining the regulatory network of the tissue- p35, the neuronal-specific activator of cyclin-dependent kinase 5 specific splicing factors Fox-1 and Fox-2. Genes Dev. 22, 2550– (Cdk5) is degraded by the ubiquitin-proteasome pathway. J. Biol. 2563. Chem. 273, 24057–24064. Zibetti C., Adamo A., Binda C. et al. (2010) Alternative splicing of the Perissi V., Jepsen K., Glass C. K. and Rosenfeld M. G. (2010) histone demethylase LSD1/KDM1 contributes to the modulation of Deconstructing repression: evolving models of co-repressor action. neurite morphogenesis in the mammalian nervous system. J. Nat. Rev. Genet. 11, 109–123. Neurosci. 30, 2521–2532. Riccio A. (2010) Dynamic epigenetic regulation in neurons: enzymes, stimuli and signaling pathways. Nat. Neurosci. 13, 1330–1337.