Biochimica et Biophysica Acta 1813 (2011) 1050–1058

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r.com/ locate /bbamcr

Sumoylation regulates nuclear localization of repressor DREAM☆

Malgorzata Palczewska a,b,1, Iñigo Casafont c,b, Kedar Ghimire a,b, Ana M. Rojas a, Alfonso Valencia a,2,Miguel Lafarga c,b, Britt Mellström a,b, Jose R. Naranjo a,b,⁎a Centro Nacional de Biotecnología, CSIC, Madrid, Spainb Centro de Investigación Biomedica en Red sobre Enfermedades Neurodegenerativas (CIBERNED), Spainc Departamento Anatomia y Biología Celular, Universidad de Cantabria and IFIMAV, Santander, Spain

Abbreviations: DREAM, downstream regulatory elKChIP, potassium channel interacting protein; SUMO, smGFP, green fluorescent protein; HA, hemaglutinin☆ This article is part of a Special Issue entitled: 11th Eur⁎ Corresponding author. Centro Nacional de Biotecn

Madrid, Spain. Tel.: +34 915854682; fax: +34 9158545E-mail address: naranjo@cnb.csic.es (J.R. Naranjo).

1 Present address: Instituto de Tecnologia Química e Bi2 Present address: Centro Nacional Investigaciones O

0167-4889/$ – see front matter © 2010 Elsevier B.V. Adoi:10.1016/j.bbamcr.2010.11.001

a b s t r a c t

a r t i c l e i n f o

Article history:Received 7 September 2010Received in revised form 29 October 2010Accepted 2 November 2010Available online 9 November 2010

Keywords:DREAMKChIPsNuclear localizationSUMOUbc9Trigeminal neuron

DREAM is a Ca2+-binding protein with specific functions in different cell compartments. In the nucleus,DREAM acts as a transcriptional repressor, although the mechanism that controls its nuclear localization isunknown. Yeast two-hybrid assay revealed the interaction between DREAM and the SUMO-conjugatingenzyme Ubc9 and bioinformatic analysis identified four sumoylation-susceptible sites in the DREAMsequence. Single K-to-R mutations at positions K26 and K90 prevented in vitro sumoylation of recombinantDREAM. DREAM sumoylation mutants retained the ability to bind to the DRE sequence but showed reducednuclear localization and failed to regulate DRE-dependent transcription. In PC12 cells, sumoylated DREAM ispresent exclusively in the nucleus and neuronal differentiation induced nuclear accumulation of sumoylatedDREAM. In fully differentiated trigeminal neurons, DREAM and SUMO-1 colocalized in nuclear domainsassociated with transcription. Our results show that sumoylation regulates the nuclear localization of DREAMin differentiated neurons. This article is part of a Special Issue entitled: 11th European Symposium on Calcium.

ement antagonist modulator;all ubiquitin-related modifier;

opean Symposium on Calcium.ología, CSIC, Darwin 3, 2804906.

ológica (ITQB), Oeiras, Portugal.ncológicas, Madrid, Spain.

ll rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Sumoylation is a post-translational modification with criticaleffects on protein function through the control of protein interactions,subcellular localization, and stability/degradation [1]. Sumoylation hasan important role in the nuclear import of several transcription factors,e.g., CREB and STAT [2,3] or in the nuclear-cytosolic shuttling of Elk-1[4], regulating their function in gene expression. The sumoylationprocess involves the sequential action of a small ubiquitin-likemodifier (SUMO), an E1-activating enzyme (SAE1/SAE2), an E2-conjugating enzyme (Ubc9), and usually an E3 SUMO protein ligase.Ubc9 recognizes substrate proteins directly and, often in conjunctionwith an E3 ligase, catalyzes the transfer of SUMO to the sumoylationmotif in the target protein [5]. Sumoylation sites having the strictconsensus sequence YKxE (where Y is a bulky, hydrophobic residue,typically Ile, Leu, or Val [2,6]) occur in many target proteins, althoughnon-consensus sequences may as well be sumoylated [2,7]. Thesimultaneous occurrence of another post-translational modification,e.g., phosphorylation, can alsomodify the sumoylation process [8]. The

structural context of the sumoylation process is not yet fullyunderstood, but it is safe to assume that Lys residues that are notsurface-exposed will not qualify for sumoylation [1]. Thus, a sumoyla-tion site is necessary, but not sufficient, for sumoylation to take place.

DREAM, alsoknownas calsenilin orKChIP3, is amemberof theKChIPfamily of structurally and functionally relatedproteins found indifferentcell compartments, with specific activities in each. In the nucleus,DREAM is a Ca2+-sensitive transcriptional repressor that binds to DNAand interacts with other nucleoproteins to regulate expression ofseveral genes, including prodynorphin [9–11]. Outside the nucleus,DREAM interacts with presenilins to regulate APP processing [12] andwith Kv4 potassium channels to regulate their membrane expressionand gating [13,14]. DREAM localization in the cytosol and the nucleushas been described in basal conditions for primary cultured neurons andmost cell lines [15,16]. Nuclear translocalization of DREAM is triggeredby serum deprivation in human H4 neuroglioma cells, whereas nuclearexport is induced by Ca2+ ionophore [17]. The molecular mechanismsthat mediate nuclear import of DREAM are presently unknown, and norecognizable nuclear localization signals have been found in theDREAMprotein (LOCtree http://www.psort.org/).

2. Materials and methods

2.1. Animals

Experiments involving the use of animals were conducted inaccordance with the standard ethical guidelines (European Communities

1051M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

Directive86/609EEC;National Institutes ofHealth1995) andapprovedbythe local ethical committees (CNB and UC).

2.2. Plasmids

Eukaryotic and prokaryotic expression vectors for DREAMhave beendescribed [9]. DREAM was subcloned in the Myc-tagged pCS2+Mvector. The Ca2+-insensitive DREAM mutant (EFmDREAM), wild-typeDREAM, and antisense DREAM (AS-DREAM) were subcloned in pAS2-1[14]. Reporter plasmid pGL2-NCX3 [16]was used in the luciferase assay.The EGFP-SUMO and EGFP-SUMOΔ6 plasmids (gift from C. Santoro,Novara, Italy) and the Ubc9 plasmid (gift from R. Hay, St Andrews,Scotland) were previously described [5,18]. SUMO and Ubc9 weresubcloned in HA-tagged pcDNA3 and Ubc9 in pACT2. The DREAMmutants K26R, K90R, K101R, K241, and K26,90R were prepared in theeukaryotic and prokaryotic expression constructs using the Quick-Change method (Stratagene). Wild-type DREAM and sumoylationmutants were subcloned in pEGFP-C1 (Clontech).

2.3. Direct yeast two-hybrid assay

The yeast strain Saccharomyces cerevisiae AH109 (Clontech) wasused. Yeast were cotransformed with the bait (pAS2-1-EFmDREAM orpAS2-1-DREAM) and target (pACT2-Ubc9) plasmids. Cotransformantswere selected by plating on selective media SDDDO and SDQDO

(Clontech Matchmaker protocol handbook). Cotransformation of thebait pAS2-1-antisense-DREAM with target vector pACT2-Ubc9 wasused as negative control for the interaction.

2.4. Cell culture, transfection, and transduction with lentiviral vectors

HEK 293 and HeLa cells were cultured in Dulbecco's modifiedEagle's medium with 10% fetal bovine serum. Cells were transfectedusing Jet/PEI reagent (Poly+ Transfection), according to the manufac-turer's instructions. Primary cortico-hippocampal neurons fromDREAM−/− mouse E14 embryos were cultured and 2 days laterexposed to lentiviral particles (5 μg of p24/106 cells) as described [16].The medium was changed 8 h after addition of the virus, and cellswere harvested after 7 days. Infection efficiency was typically 90% asassessed using viral delivery of GFP. PC12 cells and PC12 subline UR61were grown in Dulbecco's modified Eagle's medium with 10% fetalbovine serum, 5% horse serum. Neuronal differentiation was inducedby exposure to dexamethasone as described [19] or by addition of NGF(25 ng/ml) to the culture medium.

2.5. Immunoprecipitation and Western blot

Whole cell extracts were prepared by incubation on ice for 30 minwith lysis buffer [20 mM Hepes, pH7.5; 100 mM KCl, 1 mM EDTA; 1%NP40] supplemented with protease inhibitor cocktail (Roche) and25 mM NEN (Sigma) for prevention of desumoylation. Cytosolic andnuclear extracts were prepared as described [20] and in the presenceof 25 mMNEN. Coimmunoprecipitation of DREAM and Ubc9was donewith whole cell extracts (300 μg) from HEK293 cells overexpressingMyc- and HA-tagged proteins, respectively. Antibodies used wereanti-Myc (9E10, St. Cruz and ab9106, Abcam) and anti-HA (715500,Zymed and F-7, St. Cruz). For coimmunoprecipitation of DREAM andSUMO-1, cytosolic or nuclear extracts (450 μg) from PC12 cells wereincubated with 1 μg of a DREAM-specific affinity-purified rabbitpolyclonal antibody (Ab1014) [21]. Western blot was with mousemonoclonal anti-SUMO-1 (21C7, Zymed). Proteins were visualizedwith HRP-conjugated secondary antibody (Jackson) followed by ECL(West Dura, Pierce). Blots were quantified using the Quantity Onesoftware (BioRad).

2.6. Sumoylation assay

Histidin-tagged proteins were expressed and purified according tostandard protocols. The purified proteins were dialyzed againstsumoylation buffer (20 mM Hepes, pH 7.5, 5 mM MgCl2). For the invitro sumoylation assay, the K007 kit (LEA Biotech International) wasused according to the manufacturer's protocol. Protein sumoylationwas analyzed by SDS-PAGE and Western blotting using anti-DREAM(Ab1014) [21]and anti-SUMO-1 (21 C7, Zymed).

2.7. Luciferase reporter and electrophoretic mobility shift assays

The luciferase reporter assay was performed as described [16]using a luciferase reporter driven by the NCX3 promoter. EMSAexperiments were performed as described [9], using a labeled double-stranded oligonucleotide corresponding to the human prodynorphinDRE site and with 5 μg protein extracts from HEK 293 cells transfectedwith DREAM, wild-type, or sumoylation mutants.

2.8. Immunofluorescence and confocal microscopy

HeLa cells were transfected with wild-type DREAM or DREAMsumoylation mutants and EGFP-SUMO. After 24 h, cells were fixedwith 4% formaldehyde for 30 min, washed 3 times with PBS, andpermeabilized with 2% Triton X-100. Blocking was performed byincubation in 10% FBS in PBS for 60 min, after which the cells wereincubated with anti-DREAM antibody (Ab1014) [21] and visualizedwith anti-rabbit secondary antibody Alexa 594 (1:500, Invitrogen).Microscopy was performed with a Radiance 2100 (Bio-Rad) LaserScanning System on a Zeiss Axiovert 200 microscope. The imageswere sequentially taken using LaserSharp v5.0 software (Bio-Rad) andanalyzed using LaserPix v.4 image software (Bio-Rad). The laser linesemployed were the appropriate for the different fluorophoreexcitation. Individual channels were overlaid using Image J software,and colocalization was determined using the Image J RG2B colocaliza-tion plug-in.

For preparation of freshly dissociated sensory neurons, rats wereperfused with 3.7% paraformaldehyde. Sensory neurons were disso-ciated from tissue fragments of trigeminal ganglia and processed forimmunofluorescence as previously described [22]. The followingprimary antibodies were used: rabbit polyclonal anti-DREAM(Ab1014) [21], mouse monoclonal anti-panHistone (Roche), mousemonoclonal anti-U2B″ (4G3, Euro Diagnostica) that recognizes thespliceosomal U2B″ small nuclear ribonucleoprotein (snRNP), andmouse monoclonal anti-SUMO-1 (21C7, Zymed).

2.9. Electron microscopy immunocytochemistry

For immunoelectron microscopy, rats were fixed with 4%paraformaldehyde in 0.12 M phosphate buffer, pH 7.4. Small frag-ments of trigeminal ganglia were dehydrated in methanol andembedded in Lowicryl K4M at −20 °C. Ultrathin sections weremounted on nickel grids and sequentially incubated with 0.1 Mglycine in PBS for 15 min, with normal goat serum diluted 1:100 inPBS for 3 min and with rabbit polyclonal anti-DREAM antibody(Ab1014) [21], diluted in PBS containing 0.1 M glycine and 1% BSA, for1 h at room temperature. After washing, the sections were incubatedwith goat anti-rabbit secondary antibody conjugated to 10-nm gold-particles (BioCell, UK) diluted 1:25 in 1% BSA in PBS for 45 min atroom temperature. Sections were stainedwith uranyl acetate and leadcitrate and examined with a Philips EM-208 electron microscope. Ascontrols, ultrathin sections were treated as described above omittingthe primary antibody.

1052 M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

2.10. Quantitative real-time RT-PCR

RNA from primary neurons was prepared as described [16].Quantitative real-time PCR for prodynorphin and c-fos was performedusing specific kits from Applied Biosystems. The results werenormalized by quantification of HPRT mRNA using the specificprimers; forward 5′-TTGGATACAGGCCAGACTTTGTT-3′ and reverse5′-CTGAAGTACTCAT TATAGTCAAGGGCATA-3′, and the probe FAM-5′-TTGAAATTCCAGACAAGTTT-3′-MGB.

Fig. 1. DREAM interacts with Ubc9. (A) Direct yeast two-hybrid assay using EFmDREAM

2.11. Bioinformatics analysis

To detect suitable templates for structural modeling, the DREAMprotein was used as a query for the BLAST tool [23] against thestructural databases. As structural templates were found, a careful,manually curated pairwise alignment was generated for eachtemplate. We further conducted homology modeling using the bestalignments in terms of coverage and identity using the WHATIF [24]and SWISS-MODEL [25] servers. The models were evaluated using theWHATIF server and further visualized with pymol (http://pymol.sourceforge.net/). Secondary structure predictions were conducted inregions not covered by the template using JPRED (http://www.compbio.dundee.ac.uk/~www-jpred/). Accessibility values that reportresidues that are solvent-exposed were calculated on the model andthe template using the DSSP suite [26].

or wild-type DREAM as baits and Ubc9 as the target. DDO is nonselective and QDO isselective plating medium. Cotransformation of antisense DREAM (AS-DREAM) withUbc9 was used as negative control. (B) Coimmunoprecipitation of DREAM-Myc andUbc9-HA after overexpression in HEK 293 cells. Empty vectors or expression plasmidsused are indicated below. Incubation without antibody was used as negative controland is shown in the last lane (−Ab).

3. Results

3.1. DREAM interacts with the SUMO-conjugating enzyme Ubc9

To identify new protein–protein interactions that could revealDREAM functions in the cell, we previously screened a pre-transformedhuman brain library using the yeast two-hybrid assay and a Ca2+-insensitiveDREAMmutant (EFmDREAM)as thebait [14].One interactioncorresponding to a clone encoding Ubc9, a key enzyme for sumoylation[5], was further studied. Direct yeast two-hybrid assay confirmed theinteraction and showed that wild-type DREAM also interacts with Ubc9(Fig. 1A), suggesting that the interaction is not Ca2+-sensitive. To confirmthe interaction in mammalian cells, we coimmunoprecipitated Ubc9-HAand DREAM-Myc after overexpression in HEK 293 cells, a cell line thatdoes not express endogenous DREAM and shows good transfectionefficiency.Western blots showed a 37- and a23-kDaband correspondingto DREAM-Myc and Ubc9-HA, respectively (Fig. 1B), confirming DREAMas a potential sumoylation target.

Table 1Putative sumoylation sites in the DREAM sequence.

A

No. Positiona Group Score

1 K101 SLYRG FKNE CPTGL 0.85b

2 K241 FLEAC QKDE NIMSS 0.503 K221 VERFF EKMD RNQDG 0.504 K27 HTPLS KKEG IKWQR 0.31

B

Positionc Flanking peptide GPS score Cutoff Site type

26d LGHTKF TKKE GIKWQ 2.618 2.260 Non-consensus90 QAQTKF TKKE LQSLY 3.647 2.260 Non-consensus101 QSLYRG FKNE CPTGL 0.635 0.100 Ψ–K–X–E184 NKDGYI TKEE MLAIM 2.985 2.260 Non-consensus241 EFLEAC QKDE NIMSS 2.618 2.260 Non-consensus

a SUMOplot (http://www.abgent.com/tool/sumoplot/).b Sites found using SUMOplot (A) are listed in decreasing probability order (score 1

equals perfect site).c SUMOsp 1.0 (http://bioinformatics.lcd-ustc.org/sumosp/).d SUMOsp sites (B) are listed by position.

3.2. DREAM has six potential sumoylation sites

Since DREAM is a potential target for sumoylation, we searched forputative sumoylation sites within DREAM. SUMOplot (http://www.abgent.com/tool/sumoplot) and SUMOsp 2.0 (http://bioinformatics.lcd-ustc.org/sumosp/ [7]) predicted four sites at K26/K27, K101, K184,and K241, whereas sites at K221 and K90were detected only either bySUMOplot or by SUMOsp, respectively (Table 1). K101 falls within astrict consensus YKxE sumoylation sequence (FKNE), while the otherfive lysines (K26/27, K90, K184, K221, and K241) are within weakconsensus KxE/D sequences (Fig. 2). Homologymodeling based on thecrystal structure of KChIP-1 [27] and predictions from the NMRstructure of DREAM [28] indicated that K90, K101, and K241 aresurface-exposed, while the sites centered at K184 and K221 are lessexposed (Table 2). Surface exposure of K26/27 could not be judgedbecause the N-terminal region is highly variable among the KChIPfamily members and is not resolved in any of the structures availablefor these proteins.

3.3. DREAM is sumoylated in vitro at multiple sites

To validate the in silico predictions, we analyzed DREAM sumoyla-tion in vitro. Purified recombinant DREAM showed two bands inWestern blot, a major 29-kDamonomeric form and a minor oligomericformmigrating at 60 kDa (Fig. 3A). After in vitro sumoylation, we foundan additional DREAM-immunoreactive doublet migrating at 40 kDa(Fig. 3A, lane 6). Of note, this doublet was detected only when allcomponents required for in vitro sumoylation were present in thereaction (Fig. 3A, lane 6) indicating that the 40-kDa band corresponds tosumoylated monomeric DREAM.

Fig. 2. Sequence alignment of DREAM/KChIP family members. Potential sumoylation sites are labeled in blue. Serine residues at position +5 from the key lysine residue insumoylation sites are labeled in red. The human sequences for DREAM (Acc. Q9Y2W7), KChIP 1 (Q9NZI2), KChIP 2 (Q9NS61), and KChIP 4 (Q6PIL6) were aligned using the T-Coffeeprogram [40]. Asterisk (*) denotes identical residue; colon (:), conservative change; period (.), related substitution.

1053M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

To identify the specific lysine residues in DREAM that are modifiedby SUMO, K-to-R mutations were screened by in vitro sumoylationassay. Directed by the in silico analysis, we prepared four single DREAMmutants, K26R, K90R, K101R, and K241R. Of these, K101R showed thesame in vitro sumoylation pattern aswild-typeDREAM(Fig. 3B). For theK26R and K90R mutants, the upper band of the 40-kDa doublet wasabsent in both cases,whereas for K241R, this bandwasweaker (Fig. 3B).The double DREAMmutant K26,90R confirmed the pattern observed forthe singlemutants (Fig. 3B). These results suggest that residues K26 andK90 are modified by SUMO in vitro, whereas residue K101 is not.Mutationof K241 altered the sumoylation to someextent. Persistence ofthe lower band in the doublet after in vitro sumoylation of K26R, K90R,and the double K26,90R mutants suggests additional sumoylationevents within the DREAM protein that are not affected by any of thesemutations.

3.4. DREAM sumoylation mutants do not translocate to the nucleus

To determinewhether sumoylation of DREAM regulates its nucleartranslocation/retention, we cotransfected HeLa cells with DREAM andGFP-SUMO or GFP-SUMOΔ6 and visualized the proteins by confocal

Table 2Potential sumoylation sites in the DREAM sequence.

Amino acid Homology model NMR structure

K90 103a 135K101 87 100K184 101 79K241 165 149E92 55 70E103 123 79E186 20 28E223 19 20E243 152 166

a Degree of surface exposure of K and E/D residues for the potential sumoylation sitesare shown as solvent accessibility values, representing the water-exposed residuesurface in Angstroms. Values N50 are generally accepted as exposed, whereas lowervalues are considered buried.

microscopy after immunostaining of DREAM. HeLa cells were used inthese experiments because they do not express endogenous DREAM.GFP-SUMOΔ6 is a sumoylation-deficient mutant that is unable to

Fig. 3. In vitro sumoylation of DREAM and mutation analysis. (A) Western blot withanti-DREAM showing in vitro sumoylation of recombinant DREAM by reconstitutedSUMO complex (lane 6). In vitro sumoylated DREAM migrates as a 40-kDa doublet(open arrowhead). Absence of E1, E2, SUMO, or all three preclude sumoylation ofDREAM (lanes 1–5). Nonsumoylated recombinant DREAM migrates as a 29-kDamonomer and a 60-kDa dimer (filled arrowheads). (B) Western blot with anti-DREAMshowing reduced in vitro sumoylation of K26R, K90R, K26,90R, and K241R DREAMmutants. Mutation of lysine 101 shows the same sumoylation pattern as wild-typeDREAM. Quantification of the upper band in the doublet is shown below each lane asratio of band intensity relative to wild-type DREAM. Arrowheads as in panel A.

1054 M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

interact with and to covalently modify target proteins. Whencotransfected with GFP-SUMO, DREAM immunoreactivity localizedboth in the cytosol and in the nucleus. In the nucleus, DREAMcolocalizedwith the green fluorescent signal for the GFP-SUMO fusionprotein (Fig. 4, upper row). In contrast, after cotransfection with GFP-SUMOΔ6, DREAM was largely excluded from the nucleus (Fig. 4,bottom row). These results suggest that sumoylation is a prerequisitefor the presence of DREAM in the nucleus of HeLa cells. Cotransfectionof DREAM K26R and K90R sumoylation mutants with GFP-SUMO

Fig. 4. Sumoylated DREAM is located in the nucleus. Confocal microscopy of HeLa cells cotranand DREAM mutants were visualized after immunostaining with anti-DREAM (red fluoresproteins (green fluorescent signal, right column). Merge is shown in the right column. Scal

resulted in predominantly cytosolic localization (Fig. 4, middle rows),confirming the in vitro sumoylation results. The residual and weaknuclear signal still visible after overexpression of DREAM K26R andK90R sumoylation mutants could be related to the overexpressionitself or could reflect the existence of additional weak sumoylationsites or alternative mechanisms for DREAM nuclear shuffling.Furthermore, overexpression of DREAM K241R mutant showed bothcytosolic and nuclear localization similar to wild type (Fig. 4, middlerows, bottom). This suggest that sumoylation at K241 either does not

sfected with wild-type (wt) DREAM and GFP-SUMO and the indicated mutants. DREAMcent signal, left column). GFP-SUMO and GFP-SUMOΔ6 were visualized as GFP-fusione bar corresponds to 5 μm.

1055M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

occur in vivo or is not involved in the nuclear translocation of DREAM.These results suggest that K26 and K90 are important residues forDREAM sumoylation and strengthen the idea that sumoylationregulates the nuclear localization of DREAM.

To further substantiate the relationship between sumoylation andnuclear translocation of DREAM, we assayed the sumoylation mutantsfor their capability to repress transcription. For this, first we used aluciferase reporter assay in HEK 293 cells devoid of endogenousDREAM. Single K26R, K90R, and double K26,90R DREAM sumoylationmutants showed reduced repressor activity on the NCX3 reportercompared to wild-type DREAM or the K241R and K101R DREAMmutants (Fig. 5A). Second, we used primary cultures of cortico-hippocampal neurons from DREAM null embryos and quantified theeffect of wild-type or sumoylation mutant DREAM overexpression onendogenous levels of prodynorphin and c-fos, two established DREAMtarget genes in neurons [9]. Overexpression of wild-type DREAM, butnot the double K26,90R DREAM sumoylation mutant, significantlyreduced basal expression of these two target genes (Fig. 5B). Thelimited reduction (25%) of prodynorphin and c-fos mRNA levels afterDREAM over-expression in DREAM null cultured neurons is related tothe endogenous expression of other KChIP family members (Fig. 5B).These results suggest that in both HEK 293 cells and primary neurons,DREAM sumoylation is associated with the nuclear location neededfor transcriptional repressor activity.

To ensure that the lack of repressor activity of DREAM sumoylationmutants is not due to reduced interaction with DNA, we performedelectrophoreticmobility shift assays usinga labeledDREoligonucleotide

Fig. 5. DREAM sumoylation mutants are not repressors in vivo but bind to DNA in vitro.(A) Luciferase reporter assay with pGL2-NCX3 and DREAM or the indicated sumoylationmutants after transient transfection inHEK 293 cells. The results are shown as themean±SD of three independent experiments in triplicate. The K26R, K90R and K26,90R mutantsshow impaired repressor activity compared to wild-type (wt) DREAM (one-way ANOVA,Pb0.05 with Dunnett's post test). (B) Quantitative RT-PCR analysis of endogenous c-fosand prodynorphin expression in primary cortico-hippocampal cultures of DREAM nullneurons after infection with lentiviral vectors overexpressing wild-type (wt) DREAM orthe double K26,90R mutant. Infection with a lentiviral vector encoding GFP was used ascontrol. The results are shown as the mean ± SD of three independent experiments intriplicate. The K26,90R mutant does not modify endogenous expression of c-fos orprodynorphin, while the effect of wild-type DREAM is statistically significant compared tocontrol (*Pb0.05, two-tailed, unpaired t-test) or to doubleK26,90Rmutant (#Pb0.05, two-tailed, unpaired t-test). (C) Electrophoreticmobility shift assaywithwild-type andmutantDREAM proteins expressed in HEK 293 cells and a labeled DRE probe. Wild-type andmutant proteins showed the same mobility shift pattern. Competition with excessunlabeled Sp1 (cSp1) or DRE (cDRE) is shown.

derived from the prodynorphin gene. The result showed that all K-to-Rmutants conserved the in vitro DNA-binding property of wild-typeDREAM (Fig. 5C).

3.5. DREAM is sumoylated in vivo and accumulates in the nucleus duringPC12 neuronal differentiation

Undifferentiated PC12 cells, which express large amounts of DREAM[29], can be induced to attain a neuron-like phenotype by exposure toNGF and activation of Ras signaling, by chronic depolarization, or byincrease in cAMP levels [30]. Therefore, we choose the PC12 cell systemto investigate DREAM in vivo sumoylation in basal conditions andduring neuronal differentiation. In the UR61 subline of PC12 cellsexposure to dexamethasone induces N-ras expression and neuronaldifferentiation [19,31]. Differentiation of PC12 UR61 was characterizedby an increase in cell size, neurite outgrowth, and increased nuclearaccumulation of endogenous DREAM immunoreactivity (Fig. 6A).Coimmunoprecipitation of DREAM and SUMO-1 showed that sumoy-lated DREAM is present exclusively in the nucleus of undifferentiatedPC12 cells and that the nuclear levels are increased after NGF-induceddifferentiation (Fig. 6B andC). SumoylatedDREAMmigrates as a bandofapproximately 80 kDa, corresponding to a dimer of sumoylated DREAM(Fig. 6B). Thus, nuclear sumoylated DREAM in PC12 cells corresponds tomultimeric DREAM, the formwith highest affinity for specific binding toDRE sites in the DNA [32]. NonsumoylatedDREAM is present in both thecytosol and nucleus as a monomer (Fig. 6B). Following NGF-induceddifferentiation, monomeric DREAM also accumulates in the nuclearcompartment, while it is less abundant in the cytosol (Fig. 6B and C).These results suggest that nuclear translocation/retention of DREAM isenhanced after neuronal differentiation in PC12 cells and thatsumoylated DREAM is present exclusively in the nucleus.

3.6. DREAM and SUMO-1 colocalize in the nucleus in trigeminal neurons

Neurons involved in sensory processing express high levels ofDREAM [33]. In sensory neurons from the trigeminal ganglia, we foundthat DREAM immunoreactivity is diffuse in the cytoplasm andconcentrated in the nucleus, excluding the nucleolus (Fig. 7A). Doubleimmunolabeling of DREAM and pan-histone, a chromatin marker,demonstrated colocalization in chromatin domains (Fig. 7A–C), whilecostaining with spliceosomal snRNP showed that DREAM did notcolocalize with splicing factors in nuclear speckles or Cajal bodies(Fig. 7D–F, arrowhead), nuclear compartments in which snRNP isconcentrated and where no transcription takes place [34,35]. Impor-tantly, coimmunolabeling for DREAM and SUMO-1 demonstrated thenuclear colocalization of both molecular components (Fig. 7G–I andinset). The slight increase of signal intensity of SUMO-1 at the nuclearperiphery (Fig. 7I) is consistent with the sumoylation of RanGAP1 at thenuclear pore complex, a modification associated with nucleo-cytoplas-mic traffic regulation [36]. The nuclear distribution of DREAM wasfurther analyzed using immunogold electronmicroscopy. Gold particlesof DREAM immunoreactivity appeared in domains of decondensedchromatin (euchromatin) and were conspicuously absent in bothnucleolus and Cajal bodies (Fig. 8A and B). Interestingly, tangentialsectionsof thenuclear envelope illustrated thepresence of goldparticlesdirectly associated with the rings of the nuclear pore complex (Fig. 8C)supporting the nuclear traffic of DREAM and the results from thecolocalization analysis (see Fig. 7G–I).

4. Discussion

Transcriptional repressor activity of DREAM is regulated by its Ca2+-dependent interaction with DRE sites in the DNA. Importantly, thesumoylation-dependent nuclear translocalization of DREAM identifiedin this study represents a novel mechanism to further regulate therepressor activity of DREAM.

Fig. 6. DREAM sumoylation in PC12 cells. (A) Immunostaining of DREAM in UR61 PC12cells showing its preferred nuclear localization after neuronal differentiation withdexamethasone. Co-staining with propidium iodide is shown. Scale bar corresponds to5 μm. (B) Coimmunoprecipitation of endogenous DREAM and SUMO-1 in non-differentiated and NGF-differentiated PC12 cells. Sumoylated DREAM migrates as adimer (upper panel, arrow) and is present in the nuclear and absent in the cytosolicfraction. *Denotes nonspecific band. Nonsumoylated DREAM migrates as a monomer(lower panel, arrow) and is present both in the cytosolic and the nuclear fraction.(C) Quantification of the sumoylated and non-sumoylated DREAM bands shows theincreased accumulation of DREAM in the nuclear fraction in NGF-differentiated PC12cells.

1056 M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

In silico analysis of the DREAM primary sequence indicated fivenon-canonical KxE/D motifs and one canonical YKxE motif. Thisrepresents four times the statistically expected number based onprotein length suggesting that the original yeast two-hybrid obser-vation corresponds to a functionally relevant interaction betweenUbc9 and DREAM. Instead of analyzing all possible single/multiplesumoylation mutants, we discarded sumolyation sites consideredunfavorable based on low solvent accessibility. At difference from thecase of calmodulin, in which Ca2+ binding results in a largeconformational change [37], Ca2+-dependent long-range conforma-tional changes are not expected to occur in neuronal calcium sensors,including the KChIPs and recoverin [38], among others. Thus, ourassumptions that K184 and K221 are not surface exposed are likely tobe valid for a range of modified states of DREAM.

Results from the mutational analysis are consistent with that atleast two sites, K26 and K90, are important for sumoylation-dependent nuclear localization of DREAM, since conservative muta-tion of either K26 or K90 to arginine is sufficient to inhibit nucleartranslocation/retention. In vitro DREAM sumoylation at K241 is notdirectly related to nuclear localization but may nonetheless beimportant for modulation of other yet to be characterized DREAM

functions in the cytosol. The perfect sumoylation consensus site atK101 is not sumoylated in vitro, and its mutation does not alter thetranscriptional repressor activity of DREAM. Lack of modification atK101 could be due to an unfavorable position within the helixstructure, which would impede Ubc9–target interaction [1]. Sequenceanalysis of the four members of the KChIP family shows that thesumoylation consensus site at K101, in the DREAM sequence, is theonly site conserved in all four, whereas K26, located in the DREAM-specific N-terminal region, is not present in the other three members.The sumoylation site at K90 is conserved in KChIP 1 and 2 but not inKChIP4, and K241 is absent in KChIP1. Thus, sumoylation of otherKChIP family members may occur and remains to be furtherinvestigated.

It has been proposed that phosphorylation at serine residueslocated five amino acids downstream from the sumoylation site isassociated with increased sumoylation and modified transcriptionalactivity of several nucleoproteins, including HSF1, PPAR-γ and MEF2[8,39]. We have not determined, however, whether simultaneouspost-translational modifications are needed or can influence thesumoylation of DREAM. A serine residue at position 95, five aminoacids downstream from the K90 sumoylation site, is present inDREAM (Fig. 2), while the othermembers of the DREAM/KChIP family,but not DREAM, have a serine residue at position 106, five amino acidsdownstream from the K101 sumoylation site. GRK2-mediatedphosphorylation at Ser95 in DREAM regulates the intracellulartrafficking of Kv4 potassium channels. Ser-to-Ala mutation of thisresidue affects the interaction with the Kv4 potassium channel. Thismutant, however, retains its repressor activity [14], suggesting thatphosphorylation at Ser95 in DREAM is not required for sumoylationand nuclear translocation. Whether sumoylation also regulatesnuclear translocation of the other KChIP proteins and whetherphosphorylation at Ser106 has a functional role in their potentialsumoylation at K101 has not been characterized in this study.

The coexistence of sumoylated and nonsumoylated DREAM inPC12 cells suggests that the two forms are in equilibrium with (i)sumoylation and nuclear import controlled by NGF-induced neuronaldifferentiation and (ii) desumoylation and nuclear export of DREAMcontrolled by an active, presently unknown, mechanism. Theincreased presence of sumoylated DREAM in the nucleus followingneuronal differentiation supports the physiological significance of thesumoylation process in this cell system. Other important aspects thatremain to be analyzed are whether sumoylation alters the interactionof DREAM with other nucleoproteins and whether sumoylatedDREAM modifies the affinity or the specificity of the interactionwith DRE sites. Importantly, sumoylated DREAM in PC12 cells isexclusively nuclear and is present as a multimer, the form with thehighest affinity for DNA and thus with the highest capacity to repressDRE-dependent transcription.

Neurons involved in sensory processing express high levels ofDREAM [33], and we have found that in sensory neurons fromtrigeminal ganglia, the DREAM protein is mostly concentrated in thenucleus, where double immunohistochemical detection showed thatDREAM colocalizes with SUMO-1 immunoreactivity. Biochemicalanalysis, e.g., coimmunoprecipitation of sumoylated DREAM intrigeminal neurons is hampered by the low availability of biologicalsample and the accessibility to reagents for detection in theseconditions. The substantial presence of DREAM in nuclear euchroma-tin domains in trigeminal neurons indicates an important role forDREAM in transcriptional control of the processing of sensoryinformation and makes DREAM and maybe the inhibition of DREAMsumoylation new targets for the treatment of trigeminal pain.

Acknowledgements

We thank P. Gonzalez and S. Gutierrez-Erlansson for assistanceand P. Groves for discussions. Editorial help from C. Mark is

Fig. 7. Endogenous DREAM localizes in the nucleus of freshly dissociated trigeminal ganglia neurons. (A–C) Double immunolabeling for DREAM and pan-histone, a chromatinmarker, shows DREAM colocalization with nuclear domains of diffuse pan-histone staining (euchromatin), whereas the nucleolus appears free of labeling. (D–F) Co-immunostainingof DREAM and snRNP splicing factors (U2B″ antibody). DREAM does not colocalize with either nuclear speckles of splicing factors or the Cajal body (arrowhead), two nuclearcompartments enriched in snRNPs. (G–I) A neuron co-immunostained for DREAM and SUMO1. Themerged image (I) shows the nuclear colocalization of DREAM and SUMO1. (Inset)Enlargements of the dashed box shown in panel I. (Left panel) Detail of a nuclear colocalization (yellow) domain. (Right panel) The application of the colocalization filter (RG2BImage J), which replaces pixels containing signal in both the red and green channel with a grayscale pixel of similar intensity, confirms the colocalization (white) of DREAM andSUMO1 in the dashed box. Scale bar corresponds to 5 μm.

1057M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

acknowledged. This work was supported by grants from MinisterioCiencia e Innovacion (SAF2007-62449 to J.R.N., SAF2005-04682 andSAF2008-03469 to B.M., and BFU2008-00175 to M.L.); CIBERNED to B.M., I.C., M.L., and J.R.N. and from CAM; Fundación La Marató (2007-

Fig. 8. Immunogold electron microscopy of DREAM in trigeminal neurons. (A) Immunogold(No) and Cajal body (CB) appear free of DREAM immunolabeling. (B) Detail of a domain of e(C) Tangential section of the nuclear envelope showing several rings of the nuclear pore comseveral nuclear pores (arrows). Nucleus (N) and cytoplasm (C). Scale bars correspond to 15

062231); Fundación La Caixa (BM04-167-0); and the EU 6thFramework Program (NeuroNE) to J.R.N. M.P. and I.C. were supportedby a FEBS Long Term Fellowship and a CIBERNED contract,respectively.

particles are distributed on domains of euchromatin (arrows), whereas the nucleolusuchromatin illustrating the association of gold particles with chromatin fibrils (arrows).plexes. Note the direct association of gold particles of DREAM immunoreactivity with0 nm.

1058 M. Palczewska et al. / Biochimica et Biophysica Acta 1813 (2011) 1050–1058

References

[1] R. Geiss-Friedlander, F. Melchior, Concepts in sumoylation: a decade on, Nat. Rev.Mol. Cell Biol. 8 (2007) 947–956.

[2] E.S. Johnson, Protein modification by SUMO, Annu. Rev. Biochem. 73 (2004)355–382.

[3] F. Melchior, M. Schergaut, A. Pichler, SUMO: ligases, isopeptidases and nuclearpores, Trends Biochem. Sci. 28 (2003) 612–618.

[4] S. Salinas, A. Briancon-Marjollet, G. Bossis, M.A. Lopez, M. Piechaczyk, I. Jariel-Encontre, A. Debant, R.A. Hipskind, SUMOylation regulates nucleo-cytoplasmicshuttling of Elk-1, J. Cell Biol. 165 (2004) 767–773.

[5] J.M. Desterro, J. Thomson, R.T. Hay, Ubch9 conjugates SUMO but not ubiquitin,FEBS Lett. 417 (1997) 297–300.

[6] R.T. Hay, SUMO: a history of modification, Mol. Cell 18 (2005) 1–12.[7] Y. Xue, F. Zhou, C. Fu, Y. Xu, X. Yao, SUMOsp: a web server for sumoylation site

prediction, Nucleic Acids Res. 34 (2006) W254–257.[8] X.J. Yang, S. Gregoire, A recurrent phospho-sumoyl switch in transcriptional

repression and beyond, Mol. Cell 23 (2006) 779–786.[9] A.M. Carrion, W.A. Link, F. Ledo, B. Mellstrom, J.R. Naranjo, DREAM is a Ca2+-

regulated transcriptional repressor, Nature 398 (1999) 80–84.[10] H.Y. Cheng, G.M. Pitcher, S.R. Laviolette, I.Q. Whishaw, K.I. Tong, L.K. Kockeritz, T.

Wada, N.A. Joza, M. Crackower, J. Goncalves, I. Sarosi, J.R. Woodgett, A.J. Oliveira-dos-Santos, M. Ikura, D. van der Kooy, M.W. Salter, J.M. Penninger, DREAM is acritical transcriptional repressor for pain modulation, Cell 108 (2002) 31–43.

[11] MellstromB. , SavignacM. , Gomez-VillafuertesR. , NaranjoJ.R. , Ca2+-operatedtranscriptional networks: molecular mechanisms and in vivo models, Physiol Rev88 (2008) 421–449.

[12] J.D. Buxbaum, E.K. Choi, Y. Luo, C. Lilliehook, A.C. Crowley, D.E. Merriam, W.Wasco, Calsenilin: a calcium-binding protein that interacts with the presenilinsand regulates the levels of a presenilin fragment, Nat. Med. 4 (1998) 1177–1181.

[13] W.F. An, M.R. Bowlby, M. Betty, J. Cao, H.P. Ling, G. Mendoza, J.W. Hinson, K.I.Mattsson, B.W. Strassle, J.S. Trimmer, K.J. Rhodes, Modulation of A-type potassiumchannels by a family of calcium sensors, Nature 403 (2000) 553–556.

[14] A. Ruiz-Gomez, B. Mellstrom, D. Tornero, E. Morato, M. Savignac, H. Holguin, K.Aurrekoetxea, P. Gonzalez, C. Gonzalez-Garcia, V. Cena, F. Mayor Jr., J.R. Naranjo, Gprotein-coupled receptor kinase 2-mediated phosphorylation of downstreamregulatory element antagonist modulator regulates membrane trafficking ofKv4.2 potassium channel, J. Biol. Chem. 282 (2007) 1205–1215.

[15] J.W. Chang, H. Choi, H.J. Kim, D.G. Jo, Y.J. Jeon, J.Y. Noh, W.J. Park, Y.K. Jung,Neuronal vulnerability of CLN3 deletion to calcium-induced cytotoxicity ismediated by calsenilin, Hum. Mol. Genet. 16 (2007) 317–326.

[16] R. Gomez-Villafuertes, B. Torres, J. Barrio, M. Savignac, N. Gabellini, F. Rizzato, B.Pintado, A. Gutierrez-Adan, B. Mellstrom, E. Carafoli, J.R. Naranjo, Downstreamregulatory element antagonist modulator regulates Ca2+ homeostasis andviability in cerebellar neurons, J. Neurosci. 25 (2005) 10822–10830.

[17] N.F. Zaidi, E.E. Thomson, E.K. Choi, J.D. Buxbaum, W. Wasco, Intracellular calciummodulates the nuclear translocation of calsenilin, J. Neurochem. 89 (2004)593–601.

[18] L. Collavin, M. Gostissa, F. Avolio, P. Secco, A. Ronchi, C. Santoro, G. Del Sal,Modification of the erythroid transcription factor GATA-1 by SUMO-1, Proc. NatlAcad. Sci. USA 101 (2004) 8870–8875.

[19] I. Guerrero, A. Pellicer, D.E. Burstein, Dissociation of c-fos from ODC expressionand neuronal differentiation in a PC12 subline stably transfected with aninducible N-ras oncogene, Biochem. Biophys. Res. Commun. 150 (1988)1185–1192.

[20] W. Fischle, In nucleo enzymatic assays for the identification and characterizationof histone modifying activities, Methods 36 (2005) 362–367.

[21] M. Savignac, B. Pintado, A. Gutierrez-Adan, M. Palczewska, B. Mellstrom, J.R.Naranjo, Transcriptional repressor DREAM regulates T-lymphocyte proliferationand cytokine gene expression, EMBO J. 24 (2005) 3555–3564.

[22] E. Pena, M.T. Berciano, R. Fernandez, J.L. Ojeda, M. Lafarga, Neuronal body sizecorrelates with the number of nucleoli and Cajal bodies, andwith the organizationof the splicing machinery in rat trigeminal ganglion neurons, J. Comp. Neurol. 430(2001) 250–263.

[23] S.F. Altschul, W. Gish, W. Miller, E.W. Myers, D.J. Lipman, Basic local alignmentsearch tool, J. Mol. Biol. 215 (1990) 403–410.

[24] VriendG. , WHAT IF: a molecular modeling and drug design program, J Mol Graph8 (1990) 52–568 29.

[25] T. Schwede, J. Kopp, N. Guex, M.C. Peitsch, SWISS-MODEL: an automated proteinhomology-modeling server, Nucleic Acids Res. 31 (2003) 3381–3385.

[26] W. Kabsch, C. Sander, Dictionary of protein secondary structure: patternrecognition of hydrogen-bonded and geometrical features, Biopolymers 22(1983) 2577–2637.

[27] W. Zhou, Y. Qian, K. Kunjilwar, P.J. Pfaffinger, S. Choe, Structural insights into thefunctional interaction of KChIP1 with Shal-type K(+) channels, Neuron 41 (2004)573–586.

[28] J.D. Lusin, M. Vanarotti, C. Li, A. Valiveti, J.B. Ames, NMR structure of DREAM:implications for Ca(2+)-dependent DNA binding and protein dimerization,Biochemistry 47 (2008) 2252–2264.

[29] F. Ledo, L. Kremer, B. Mellstrom, J.R. Naranjo, Ca2+-dependent block of CREB-CBPtranscription by repressor DREAM, EMBO J. 21 (2002) 4583–4592.

[30] D. Vaudry, P.J. Stork, P. Lazarovici, L.E. Eiden, Signaling pathways for PC12 celldifferentiation: making the right connections, Science 296 (2002) 1648–1649.

[31] J. Navascues, M.T. Berciano, K.E. Tucker, M. Lafarga, A.G. Matera, Targeting SMN toCajal bodies and nuclear gems during neuritogenesis, Chromosoma 112 (2004)398–409.

[32] M. Osawa, K.I. Tong, C. Lilliehook, W. Wasco, J.D. Buxbaum, H.Y. Cheng, J.M.Penninger, M. Ikura, J.B. Ames, Calcium-regulated DNA binding and oligomeriza-tion of the neuronal calcium-sensing protein, calsenilin/DREAM/KChIP3, J. Biol.Chem. 276 (2001) 41005–41013.

[33] P.I. Hammond, T.A. Craig, R. Kumar, S. Brimijoin, Regional and cellular distributionof DREAM in adult rat brain consistent with multiple sensory processing roles,Brain Res. Mol. Brain Res. 111 (2003) 104–110.

[34] I. Casafont, J. Navascues, E. Pena, M. Lafarga, M.T. Berciano, Nuclear organizationand dynamics of transcription sites in rat sensory ganglia neurons detected byincorporation of 5′-fluorouridine into nascent RNA, Neuroscience 140 (2006)453–462.

[35] A.I. Lamond, D.L. Spector, Nuclear speckles: a model for nuclear organelles, Nat.Rev. Mol. Cell Biol. 4 (2003) 605–612.

[36] M.J. Matunis, E. Coutavas, G. Blobel, A novel ubiquitin-like modificationmodulatesthe partitioning of the Ran-GTPase-activating protein RanGAP1 between thecytosol and the nuclear pore complex, J. Cell Biol. 135 (1996) 1457–1470.

[37] K.L. Yap, J.B. Ames, M.B. Swindells, M. Ikura, Diversity of conformational states andchanges within the EF-hand protein superfamily, Proteins 37 (1999) 499–507.

[38] J.B. Ames, M. Ikura, Structure and membrane-targeting mechanism of retinal Ca2+-binding proteins, recoverin and GCAP-2, Adv. Exp. Med. Biol. 514 (2002)333–348.

[39] J. Anckar, L. Sistonen, SUMO: getting it on, Biochem. Soc. Trans. 35 (2007)1409–1413.

[40] C. Notredame, D.G. Higgins, J. Heringa, T-Coffee: a novel method for fast andaccurate multiple sequence alignment, J. Mol. Biol. 302 (2000) 205–217.