Biochimica et Biophysica Acta 1791 (2009) 1023–1030

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

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

Subcellular localization of ceramide kinase and ceramide kinase-like protein requiresinterplay of their Pleckstrin Homology domain-containing N-terminal regionstogether with C-terminal domains

Philipp Rovina 1, Andrea Schanzer, Christine Graf 2, Diana Mechtcheriakova 3,Markus Jaritz 4, Frédéric Bornancin ⁎Novartis Institutes for BioMedical Research, Brunnerstrasse 59, A-1235 Vienna, Austria

⁎ Corresponding author. Novartis Pharma AG, ForumBasel, Switzerland. Tel.: +41 61 32 43136; fax: +41 61 3

E-mail address: frederic.bornancin@novartis.com (F.1 Present address: Roche Diagnostics GmbH, Engelhorn2 Present address: AFFiRiS AG, Viehmarktgasse 2A, A-3 Present address: Department of Pathophysiology,

A-1090 Vienna, Austria.4 Present address: Research Institute of Molecular Path

1388-1981/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.bbalip.2009.05.009

a b s t r a c t

a r t i c l e i n f o

Article history:Received 1 April 2009Received in revised form 15 May 2009Accepted 29 May 2009Available online 6 June 2009

Keywords:CeramideKinaseNucleocytoplasmicNucleolarPleckstrin homologyRetinal degeneration

Ceramide kinase (CERK) and the ceramide kinase-like protein (CERKL), two related members of thediacylglycerol kinase family, are ill-defined at the molecular level. In particular, what determines theirdistinctive subcellular localization is not well understood. Here we show that the Pleckstrin Homology (PH)domain of CERK, which is required for Golgi complex localization, can substitute for the N-terminal region ofCERKL and allow for wild-type CERKL localization, which is typified by nucleolar accumulation. Thisdemonstrates that determinants for localization of these two enzymes do not lie solely in their PH domain-containing N-terminal regions. Moreover, we present evidence for a previously unrecognized participation ofCERK distal sequences in structural stability, localization and activity of the full-length protein. Progressivedeletion of CERK and CERKL from the C-terminus revealed similar sequential organization in both proteins,with nuclear import signals in their N-terminal part, and nuclear export signals in their C-terminal part.Furthermore, mutagenesis of individual cysteine residues of a CERK-specific CXXXCXXC motif severelycompromised both exportation of CERK from the nucleus and its association with the Golgi complex.Altogether, this work identifies conserved domains in CERK and CERKL as well as new determinants for theirsubcellular localization. It further suggests a nucleocytoplasmic shuttling mechanism for both proteins thatmay be defective in CERKL mutant proteins responsible for retinal degenerative diseases.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Ceramide kinase (CERK) phosphorylates ceramide to yield cera-mide-1-phosphate (C1P). This lipid product is the focus of increasedattention because of its signaling properties [1–3]. Particularly welldescribed are the effects of C1P on cell proliferation and eicosanoidsynthesis, reviewed in [4,5]. The CERK protein has been only partiallycharacterized at the biochemical level. We previously analyzed itsPleckstrin Homology (PH) domain and identified a critical loop thatstabilizes the active protein conformation [6,7]. However, the precisefunction of the PH domain in CERK is incompletely elucidated. Inparticular,whether this domainprovides the soledeterminants for local-ization or may act as an independent module has not been studied yet.

The ceramide kinase-like protein (CERKL) is homologous to CERK[8–10]; however it does not phosphorylate ceramide and in fact still

1-Novartis Campus, CH-40562 43540.Bornancin).gasse 3, A-1211 Vienna, Austria.1030 Vienna, Austria.Medical University of Vienna,

ology, A-1030 Vienna, Austria.

ll rights reserved.

stands as an orphan kinase [10,11]. Interest in CERKL has been drivenby linkage between the CERKL gene and retinal degenerativepathologies [9,12–14]. The domain organization of the CERKLsequence has not been addressed in detail; in particular it is notclear if this protein has a PH domain.

CERK and CERKL have distinct subcellular localizations. CERKaccumulates at the Golgi complex and plasma membrane whereasCERKL can be found in various cell compartments among which,most notably, nucleoli [10]. Here we used this distingo as aparadigm to search for determinants that underlie subcellularlocalization of both proteins. Identification of novel conserveddomains in CERK and CERKL as well as testing of chimeric anddeletion mutant proteins has provided insight into the molecularorganization of both proteins with consequent functionalimplications.

2. Materials and methods

2.1. Materials

C8-ceramide and cardiolipin were from Sigma. Octyl-D-β-gluco-pyranoside was from Fluka, [γ-32P]ATP (10 mCi/ml, 3000 Ci/mmol)

1024 P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

and [32P]orthophosphate (10 mCi/ml) were purchased from Amer-sham Biosciences. Complete™ protease inhibitors minitablets werefrom Roche Molecular Biochemicals. SDS-PAGE under reducingconditions was done on NuPAGE polyacrylamide gels (Invitrogen).Oligonucleotides were from VBC-Genomics. TOP10 competentEscherichia coli cells (Invitrogen) were used for plasmid expansion.DNA restriction enzymes and T4 DNA polymerase were from NewEngland Biolabs and from Fermentas. N-(5-(5,7-dimethyl BODIPY)-l-pentanoyl)-D-erythro-sphingosine (DMB-Cer) and N-((4-(4,4-difluoro-5-(2-thienyl)-4-bora-3a, 4a-diaza-s-indacene-3-yl)phe-noxy)acetyl)sphingosine (TRB-Cer) were from Molecular Probes.The anti-green fluorescent protein (GFP) antibody (ab290) wasfrom Abcam.

2.2. Plasmid constructs

hCERK and hCERKL cDNAs, corresponding to GenBank™ accessionnumbers AB079066 and AJ640141, were obtained and subcloned inGateway™ compatible entry vectors as described previously [10]. Sitedirected mutagenesis was performed with the QuickChange SiteDirectedMutagenesis system using primers designed according to theQuikChange® Primer Design Program (Stratagene). All constructswere then transferred to pcDNA-DEST53 (Invitrogen) in order toexpress N-terminally fused GFP.

Chimera 2, encoding a fusion protein of hCERK (1–123; Δ117–121)and hCERKL (160–532), was prepared by using CERK and CERKLconstructs previously cloned into pENTR TOPO vectors [10,15]. Toprepare the fragment derived from the CERK PH domain codingsequence, the CERK plasmid DNA was first cleaved by PstI. This wasfollowed by nucleotide fill-in in the presence of T4 DNA polymerase anddNTPs before further DNA cleavage by SacII. To release the 1–159 codingsequence of CERKL, the CERKL plasmid DNA was first cleaved by NdeI.This was followed by nucleotide fill-in in the presence of T4 DNApolymerase and dNTPs before further DNA cleavage by SacII. The twofragments were then ligated using the Fast-Link DNA ligation kit(Epicentre Biotechnologies). Chimera 1, encoding a fusion protein ofhCERK (1–123) andhCERKL (160–532),wasprepared fromchimera 2 bysite-directed mutagenesis using the following forward primer(insertion is underlined, encoding amino acids 117 to 121 of hCERK):ggagcagctgtgtcacttgtggctacagaccctgcgggagatgttggcaggctttccaaacagaccg.

2.3. COS cells, transfection and microscopy

COS-1 cells were obtained fromDSMZ (ref. ACC 63) and cultured inDMEM/10% FCS at 37 °C/5% CO2 in a humidified atmosphere. Cellswere seeded at 105 cells/well in six-well plates. After 24 h, cells weretransfected with 4-μg plasmid, using FuGENE 6 (Roche MolecularBiochemicals). Cells were incubated for 24 to 48 h followingtransfection. For harvest, cells were washed with ice-cold phos-phate-buffered saline (PBS), and scraped into an appropriate lysisbuffer (cf. below).

Fluorescence microscopy was performed as described previously[10] on an invertedmicroscope (Axiovert 200M, Zeiss) equipped witha high resolution microscopy camera (AxioCamMRc, Zeiss) as well asoil DIC objectives (Plan-Neofluar 40×/1.30 and Plan-Apochromat63×/1.40).

2.4. CERK assays

For in vitro kinase assays, cells were scraped into lysis buffer(10 mM MOPS pH 7.2; 2 mM EGTA, 150 mM KCl, 2% Triton X-100,1 mM DTT, and protease inhibitors). The suspension was homo-genized by 20 strokes in a Potter–Elvejhem homogenizer and usedimmediately. Kinase activity assays were performed exactly asdescribed in [7]. Cell-based CERK assays were performed as describedpreviously [3].

3. Results

3.1. Identification of novel conserved domains in CERK and CERKLproteins

CERK contains an N-terminal PH domain that is important formembrane association and subcellular localization of the full-lengthprotein [6]. The N-terminal region of CERKL starts with a stretch of 40amino acids among which more than 60% are either R, E, P or A.Therefore, in absence of a more detailed characterization of this N-terminal part of the protein, we have named this sequence the “REPA”box (Fig.1). Downstream of the REPA box CERKL displays a domain thathas homology with the PH domain of CERK. However, such a domaincould not be specifically assigned to CERKL previously [10]. To get moreinsight we now performed a fold recognition (threading) analysis usingthis CERKL domain and we found that many PH domain-containingproteins includingCERK rankhigh (data not shown). This suggested thatthis domain in fact probably adopts a PH domain fold. Remarkably, thecritical positively charged residues K90, R91, R96 and K98 previouslyidentified in theβ6–β7 loopof the PHdomain of CERK [7], are conservedin an similar position in CERKL: K127, K128, K132, and K134 (Fig. 1).Furthermore, the conserved tryptophan residue in the C-terminal α-helix of PH domains is also conserved in CERKL (W151). Altogether thissuggested that the domain spanning amino acids 40 to 160 in CERKL isindeed a PH domain (Fig. 1).

Downstream of the PH domain, both CERK and CERKL contain adiacylglycerol kinase (DAGK) domain that has homology with otherenzymes of the same family, e.g. sphingosine kinases (domains C1 toC3, [16]). Previous analysis of sequences further downstream did notdetect significant homology between CERK and CERKL [10]. Uponcomprehensive analysis based on Probcons1.1 alignments [17], wehave now identified three conserved domains in CERK and CERKLnamed CC1 to CC3 (Fig.1). The conservation between CERK and CERKLis high in these domains: 35% in CC1, 41% in CC2 and 32% in CC3 (Fig.1). CC1 encompasses the C4 domain of sphingosine kinases [16] towhich it has distant homology. CC2 is a domain that has no equivalentin sphingosine kinases and CC3 shows only distant homology to thesphingosine kinase C5 domain [16]. The CC1–CC2 inter-domain regionis twice as long in CERK and contains a CXXXCXXC motif that is criticalfor CERK activity [18]; this motif is absent in CERKL. The calcium/calmodulin regulation site previously identified in CERK [19] iscompromised in CERKL by the presence of a proline residue (P440).The CC2–CC3 inter-domain region is longer for CERK. In both proteinsit displays a conserved stretch of acidic residues that we coined “DE”box (Fig. 1).

3.2. Typical and contrasting subcellular localizations of CERK and CERKL

Despite their similarities at the primary sequence level, CERK andCERKL display distinct subcellular localization patterns. Whenexpressed in COS cells, CERK accumulates at the Golgi complex [6]and in microtubule-driven vesicles that traffic between the Golgicomplex and the plasma membrane [7]. This localization can beshifted towards the plasma membrane under conditions of osmoticshock [6]. The PH domain of CERK plays an essential role: when it isdeleted or compromised, subcellular localization is lost and mutantproteins appear dispersed in the cytoplasm [6,7].

CERKL, when expressed in COS cells, appears more diverselycompartmentalized than CERK. It is found in the cytoplasm, atperinuclear and plasma membrane levels, and in the nucleus where itis enriched in nucleoli [10]. Of note, a variety of point mutations weintroduced and tested (in particular G234A in the DAGK domain),abolished nucleolar staining (Table 1). Therefore, nucleolar localiza-tion of CERKL appears to be a hallmark of WT CERKL.

The typical localization of WT CERK and CERKL expressed in COScells is recapitulated in Fig. 2. We used TRB-Cer, a fluorescent short

Fig. 1. Primary sequence alignment showing conserved regions in hCERK and hCERKL. The alignment was made using ProbCons 1.1 [17]. Conserved regions are highlighted with colorcode. Amino acid conservation is shown between the two sequences in grey. “+” denotes closely related amino acids. The stretch of positive charges within the β6–β7 loop of the PHdomain in CERK is underlined [7] as well as corresponding amino acids in CERKL. The CXXXCXXCmotif [18] present in CERK between the CC1 and CC2 conserved regions is underlined.So is a type 1-8-14B CaM binding motif in CERK [19], present in the CC2 domain.

1025P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

chain ceramide that labels the ER and is slowly transported to theGolgi complex [3,20], in order to show the specific accumulation ofCERK in the later compartment. Hoechst 33342-staining was used toevidence the specific accumulation of CERKL at the nucleolar level.

Table 1Mutations in CERKL that abolish nucleolar localization.

CERKL protein Domain Localization Nucleolar

WT Full length CP+N YesChimera 1 CERK/CERKL CP+N YesChimera 2 CERK/CERKL CP NoG234D DAGK CP+N NoR379X CC1–CC2 NNCP NoR257X DAGK NNNCP NoE53X PH NNNCP NoΔ1–205 REPA-DAGK CP NoΔ2−9 (NLS1mut) REPA box CP No104-GGG-106 (NLS2mut) PH CP NoL101A, L103A PH CP NoI124A, L126A PH CP NoI141A, L143A PH CP No

The table lists the mutations introduced in the CERKL protein with reference to theprotein domains (cf. Fig. 1) as well as the resulting impact on cellular localizationincluding the incidence on nucleolar localization. “CP” stands for cytoplasmic(altogether plasma membrane, perinuclear and cytosolic localizations); “N” stands fornuclear localization.

3.3. Swapping N-terminal regions in CERK and CERKL identifies astructural role for the PH domain

We prepared a hybrid protein by exchanging the N-terminal regionof CERKL (REPA box+PH domain) for the N-terminal region of CERK(PH domain), resulting in chimera 1 (Fig. 3A). Unexpectedly, chimera1 localized identically to WT CERKL, i.e. it displayed altogethercytoplasmic, nuclear and very notably, nucleolar localization (Fig. 3B).We then made a closely related hybrid protein with a truncated PHdomain of CERK (chimera 2) (Fig. 3A). In chimera 2, the PH domain ofCERK is 5 amino acids shorter at the C-terminus and therefore bears acompromised α-helix that cannot stabilize the PH domain fold.Chimera 2 lost the ability to co-localize with nucleoli and mostlyaccumulated in the cytoplasm (Fig. 3).

The study of these two hybrid proteins indicates that the N-terminal region of CERK, if it bears a structurally intact PH domain, cansubstitute for the N-terminal region of CERKL that contains the REPA

box and the PH domain. Therefore, the results suggested that the PHdomain of CERK does not bear the signature used for WT CERKlocalization (plasma membrane and Golgi complex). Reciprocally, the

Fig. 2. Subcellular localization of CERK and CERKL analyzed with markers. COS-1 cells transiently expressing GFP-CERK or GFP-CERKL were co-stained with Hoechst-33342 and TRB-Cer to label the nucleus and endoplasmic reticulum/Golgi complex respectively. The cells were then observed with fluorescence microscopy.

Fig. 3. Subcellular localization of CERK–CERKL chimeras. (A) Chimera 1 results from thefusion of the PH domain of CERK (grey) to a CERKL deletion construct (black) lackingthe REPA box (white) and the PH domain of CERKL. Chimera 2 is a shorter version ofchimera 1, with a CERK PH domain truncated in the terminal α-helix (amino acids 117to 121). (B) Subcellular localization of GFP-chimera 1 and GFP-chimera 2 observed byfluorescence microscopy and compared to GFP-CERK and GFP-CERKL used as controls.

1026 P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

results also suggested that the N-terminal region of CERKL does notbear the signature for nucleolar localization.

3.4. The C-terminal region including the CC3 domain is required for CERKlocalization and activity

The results above implied that some of the information forsubcellular localization of CERK (and CERKL) may be encoded by theC-terminal region of these proteins. To address this possibility, wemade progressive C-terminal deletions in CERK and studied theireffect on subcellular localization and activity.

The C-terminal CC3 domain of CERK (Fig. 1) is very conserved invarious species (Fig. 4A). Beyond the CC3 domain sequences divergeboth in amino acid content and length (Fig. 4A). Premature truncationof the C-terminus of CERK did affect neither CERK localization noractivity as long as it did not reach the CC3 domain (e.g. P533X, E528X)(Fig. 4B). However, as soon as the CC3 domainwas truncated (A524X),both localization and cellular activity were lost in the resultingmutantproteins (Fig. 4B). Remarkably, the protein variants localized as ΔPH–CERK, a mutated CERK protein lacking the entire PH domain [6] (Fig.4B). Although localization and activity were preserved in the E528Xmutant protein when assayed at the cellular level with exogenouslyadded substrate (Fig. 4B), activity was lost after cell lysis in vitro (Fig.4C). This suggested that the E528X mutant protein has an unstableconformation (Fig. 4C).

3.5. Nucleocytoplasmic shuttling motifs map to similar regions in CERKand CERKL

Further deletion from the C-terminus of CERK was undertaken. ACERK mutant protein lacking a third of the CC3 domain (L519X) wasmostly cytosolic but was also detected in the nucleus (Fig. 5A). Adeletion that removed the last 198 amino acids (G339X) resulted inmore significant accumulation into the nucleus at the expense ofcytosolic localization (Fig. 5A). Deletion of the complete sequencebeyond the PH domain (E124X) resulted in almost completeaccumulation in the nucleus, as shown previously [6]. The effects ofthese progressive deletions suggest the existence of nuclear exportsignals (NES) in the C-terminal part of CERK. A search for NES [21] inCERK indeed identified a traditional NES 511-IEVRVHCQLVRL-522 inthe CC3 domain and a class 2-NES 347-CRAGCFVC-354 between theCC1 and the CC2 domains. The location of these NES in the sequence of

CERK is consistent with the degree of nuclear localization obtainedwith the V513X and G339X CERK mutant proteins and with anothermutant protein we made which is deleted between amino acids 220and 495 (Fig. 5A). Remarkably the class 2-NES sequence coincideswith a cysteine motif (CXXXCXXC) that is important for catalytic

Fig. 4. CERK distal sequences are required for localization and activity. (A) Alignments of the C-terminus of CERK fromvarious species, featuring the CC3 domain, made using CLUSTALW [29]. The ⁎ symbols refer to sites wheremutations have been introduced and studied. (B) Fluorescence microscopy (left panel), Anti-GFPWestern Blot (middle panel) and cellularactivity (right panel) of GFP-CERKWT, P533X, E528X and A524X mutant proteins. Activity was measured by incubation with 5 μMDMB-Cer for 2 h followed by lipid extraction, thinlayer chromatography and quantification of the formed DMB-C1P; Mean±SD of triplicate determinations. (C) In vitro specific kinase activity (Mean±SD of triplicatedeterminations) of GFP-CERK P533X, E528X and A524X mutant proteins compared to GFP-WT CERK.

1027P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

activity of CERK [18]. Mutation of any of the cysteines in this motifincreased nuclear retention of the resulting mutant proteins ascompared to WT CERK (Fig. 6). Moreover, mutation of either C347or C351 severely impaired localization at the Golgi complex, resultingin cytosolic accumulation instead (Fig. 6). Altogether, this suggestedthat the CXXXCXXC motif in CERK contains a functional NES and mayalso play a role in addressing the protein to the Golgi complex.

In CERKL, a similar progressive nuclear retention that parallelsincreasing C-terminal deletion was observed (Fig. 5B). Consistently,two NES were identified in the protein sequence of CERKL, 503-LMEVASEVHIRL-514 in the CC3 domain and 259-LHIIMGHVQL-303, inthe terminal part of the DAGK domain.

Accumulation of C-terminally truncated CERK and CERKL proteinsin the nucleus suggests the existence of nuclear localization signals(NLS) in the N-terminal part of both proteins. In CERK, neither theNucPred [22] nor the PredictNLS [23] software could predict a NLS.However, use of the PSORTII [24] server pointed to the 93-RRHR-96sequence as a putative NLS. A search with this peptidic sequenceagainst PredictNLS [23] recovered 82 proteins in addition to CERK,among which 78% were identified as nuclear proteins (data notshown). Remarkably, the RRHR sequence in CERK is located in theβ6–β7 loop of the PH domain, which we previously showed to containkey basic amino acids [7]. Therefore, in addition to stabilizing thestructure, charges in the β6–β7 loop of the PH domain may alsoregulate nuclear accumulation of CERK.

In CERKL, we previously suggested the existence of two NLS in theN-terminal part of the protein, a first one at 2-PWRRRRNR-9 (NLS1),and a second one at 102-KLKRR-106 (NLS2) [10]. A recent report byIgarashi et al. showed that NLS2 is functional [25]. To investigate thepossible contribution of NLS1 we prepared the CERKL E53X mutantprotein (Fig. 5B), which lacks NLS2, and found that it is stronglyretained in the nucleus. Altogether, this suggests that both NLS1 andNLS2 are functional in CERKL.

3.6. Mutations in CERKL, responsible for retinal degeneration, preventnucleocytoplasmic trafficking and nucleolar localization

While activity of the CERKL protein still remains unknown, severalreports have shown that mutations in the CERKL gene can lead toretinal degenerative diseases, e.g. Retinitis pigmentosa [9,12,13]. Thefirst mutant identified [9] encodes a protein that is prematurelytruncated in the DAGK domain, after R257. Our previous analysis [10]revealed that the R257X CERKL mutant protein is strongly retained inthe nucleus, leading us to propose that nuclear accumulation ofmutated CERKL might correlate with disease. We now tested inparallel a CERKL mutant protein with a compromised NLS2 sequence(NLS2mut CERKL), similar to the CERKL protein product (R106S CERKL)reported in a second study of CERKL associated diseases [13]. NLS2mut

CERKL was essentially cytoplasmic, as expected for a mutation in afunctional NLS (Fig. 7). The third CERKL mutant reported so far refers

Fig. 5. C-terminal deletion analysis of CERK and CERKL proteins. Progressive C-terminal deletions were obtained in CERK (A) and CERKL (B). The resulting impact on subcellularlocalization was addressed using fluorescence microscopy as described in the legend of Fig. 3. Quantification of cytosolic and nuclear retention is done with a 4-circle scale: 4 emptycircles (0%) means no localization to this compartment, 4 filled-circles (100%) means complete associationwith that compartment. Localization to the Golgi complex or to nucleoli isall or none and therefore is evaluated using one circle only. Pictures of selected mutant proteins are displayed on the right of the figure.

1028 P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

to two types of aberrantly-spliced products, one bearing the entireintron 1 and the other one only part of this intron [12]. Although theexact effect of this splicing mutation on CERKL transcripts in humanretina is not known, the expected outcome is incorrect splicing,leading to an abnormal protein product. Retention of intron 1 isexpected to lead to premature translation termination, after insertion

Fig. 6.Mutation of the CXXXCXXCmotif in CERK impairs export from the nucleus and localizatindividual cysteines of the C347XXXC351XXC354 motif was analyzed by fluorescence microscopThe arrowheads point to nuclear accumulation; the arrows show residual Golgi complex lo

of 115 incorrect amino acids, starting at position 80 of the CERKLprotein [12]. Based on the study performed in Fig. 5, this would resultin accumulation of the CERKL truncatedmutant protein in the nucleus.In case the sequence might proceed at translational level, that wouldresult in a long extension of the β3–β4 loop of the PH domain [7],likely disruptive for the PH domain fold. The expected consequence

ion to the Golgi complex. Subcellular localization of GFP-tagged CERKwith mutagenizedy upon transient expression in COS-1 cells, and compared to GFP-CERK used as a control.calization of the C354A CERK mutant protein.

Fig. 7. CERKL mutant proteins involved in retinal degeneration cannot traffic between cytoplasm and nucleus. The subcellular localization of GFP-tagged CERKL proteins, bearingmutations that correspond to reportedmutants associated with retinal diseases, was analyzed by fluorescencemicroscopy upon transient expression in COS-1 cells, and compared toGFP-CERKL used as a control. The NLS2mut protein is described in Table 1. The arrows point to nucleoli; “cytoplasmic” stands altogether for plasma membrane, perinuclear andcytosolic localizations.

1029P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

would be cytoplasmic accumulation of the CERKL mutant protein.Therefore, both scenarios would impact on trafficking of CERKL,leading to its accumulation either in the nucleus or in the cytoplasm.

Remarkably, none from all these known CERKLmutant proteins didlocalize (R257X, NLS2mut) or would be expected to localize (aber-rantly-spliced products) at the nucleolar level (Fig. 7 and cf. Table 1).

4. Discussion

A closer analysis of CERK and CERKL primary sequences hasrevealed a number of previously unrecognized conserved features.First, the N-terminal region of CERKL contains a bona fide PH domain(Fig. 1) which, similar to CERK, is necessary but not sufficient forappropriate subcellular localization of the full-length protein. Second,beyond the DAGK domain, three conserved stretches of sequence(CC1, CC2 and CC3, Fig. 1) as well as an acidic box, (DE box, Fig. 1)could be identified in CERK and CERKL. Third, a clear and similarpattern of nucleocytoplasmic signals was evidenced in both CERK andCERKLwhereby the N-terminal region controls nuclear uptake and theC-terminal one is responsible, at least in part, for export into thecytoplasm (Figs. 5, 6). Treatment with leptomycin B, an inhibitor ofthe CRM-1 exportin did not have an effect on either CERK or CERKLnuclear export (data not shown), thus suggesting alternativemechanisms. For CERKL, the “frozen” nuclear or cytoplasmic localiza-tion of mutant proteins associated with retinal degeneration, inparticular those that still bear an intact catalytic domain [12], suggeststhat trafficking of CERKL between cytoplasm and nucleus is probablyrelevant to function. Impairment in the shuttling process more thanaccumulation in a particular compartment (the nucleus for R257X orthe cytoplasm for NLS2mut, Fig. 7) may have pathological relevance.

Experiments on the PH domain of CERK previously showed thatthis domain is crucial for stabilization of the full-length protein.

Truncation of the PH domain either from the N-terminus or C-terminus resulted in disruption of the PH domain fold and loss of WTCERK localization [7]. Positive charges in the β6–β7 loop of the PHdomain were also identified as critical for the stability of this domainand, as a result, for maintenance of the full-length protein structure[7]. The chimeric approach used in the present work (Fig. 3) hasprovided direct evidence for a structural role of the PH domain inCERK. In addition, this study has identified that the C-terminal regionof CERK is equally important (Fig. 4). Remarkably, the E528X CERKmutant protein was correctly localized and active in the cell, but lostactivity when subsequently assayed in vitro (Fig. 4), thus demonstrat-ing that the C-terminal part of CERK, like the PH domain, is required tomaintain the structure of CERK. Therefore, the structural organizationof CERK appears to require interplay of both the PH domain and C-terminal regions of the protein.

What determines the typical enrichment of CERK at the Golgicomplex and/or plasmamembrane (Fig. 2) has been difficult to assignbecause modifications of the protein sequence that result in loss oflocalization also result in loss of protein stability. This is the case ofmutations in the PH domain [7] and at the C-terminus of CERK (Fig. 4).However the present work has shed some further light. First,experiments with chimera 1 showed that the PH domain of CERK,when fused to N-terminally deleted CERKL does not bring theresulting protein to the Golgi complex but instead allows the chimerato localize as if it were full-length CERKL (Fig. 3). Therefore, althoughthe isolated PH domain of CERK can bind to lipid vesicles in vitro [6],its sequence does not appear to bear a motif that, per se, mediateslocalization to the Golgi complex. This is in contrast to PH domains ofproteins such as Four-Phosphate-Adaptor-Proteins (FAPPs) that canfunction as isolated modules and bind to the Golgi complex. In thecase of FAPPs, binding occurs via specific recognition of thephosphoinositide PIP4 and through protein–protein interaction [26].

1030 P. Rovina et al. / Biochimica et Biophysica Acta 1791 (2009) 1023–1030

The PH domain of CERK, instead, has a promiscuous phosphoinositidebinding profile [7,27] and specific binding partners have not beenreported. Second, Fig. 6 shows that the CXXXCXXC motif is importantnot only for migration out of the nucleus, but also for localization atthe Golgi complex. Therefore, the CXXXCXXC motif may represent anovel determinant for subcellular addressing of CERK. Because anequivalent motif in the radical S-adenosylmethionine enzymes [28] isregulated by [4Fe.4S] complexes, studying how this motif regulatesCERK is now warranted.

For CERKL, the mechanism responsible for the typical nucleolarlocalization (Fig. 2) remains elusive. On the one hand, the N-terminalregion is clearly necessary for nuclear localization. It contains 2nuclear localization signals NLS1 and NLS2 [10]. NLS2 is functional andmay also be acting as a nucleolar localization signal [25]. Because theE53X mutant protein was highly retained in the nucleus (Fig. 5) thepresent work indicates that NLS1 is also functional in CERKL. On theother hand, we show here that determinants for nucleolar localizationmust exist downstream of the PH domain because the N-terminal partof CERK could effectively replace the N-terminal part of CERKL inchimera 1, without compromising nucleolar localization (Fig. 3).Defining these determinants further appears challenging because avariety of mutations we introduced in CERKL invariably led to loss ofnucleolar accumulation (Table 1). Nevertheless, nucleolar associationremains an intriguing feature of CERKL that is lost in CERKL proteinmutants associated with retinal degeneration (Fig. 7).

In summary, our work underscores similarities between CERK andCERKL in their protein sequence organization. Both proteins have a PHdomain that is necessary but not sufficient for subcellular localizationof the full-length protein. Because the PH domain of CERK cansubstitute for the N-terminal region of CERKL, this identifies astructural rather than assigning role for the PH domain. The CXXXCXXCmotif which is specific to CERK is critical for localization at the Golgicomplex. For CERKL the actual determinants remain to be identified.Both CERK and CERKL rely on their N-terminal region for nuclearuptake and on their C-terminal region for export out of the nucleus.Shuttling of these proteins between cytoplasm and nucleus mayrepresent a novel functional mechanism.

Acknowledgements

Weare grateful toDr. TamarBen-Yosef (Israel Institute of Technology,Haifa, Israel) for helpful comments on the manuscript. We thank ClaireMalinverni (Novartis Institutes for BioMedical Research, Basle, Switzer-land) for help with the experiments performed at revision stage.

References

[1] B.J. Pettus, A. Bielawska, S. Spiegel, P. Roddy, Y.A. Hannun, C.E. Chalfant, Ceramidekinase mediates cytokine- and calcium ionophore-induced arachidonic acidrelease, J. Biol. Chem. 278 (2003) 38206–38213.

[2] A. Gomez-Munoz, J.Y. Kong, B. Salh, U.P. Steinbrecher, Ceramide-1-phosphateblocks apoptosis through inhibition of acid sphingomyelinase in macrophages, J.Lipid Res. 45 (2004) 99–105.

[3] A. Boath, C. Graf, E. Lidome, T. Ullrich, P. Nussbaumer, F. Bornancin, Regulation andtraffic of ceramide 1-phosphate produced by ceramide kinase: comparative analysisto glucosylceramide and sphingomyelin, J. Biol. Chem. 283 (2008) 8517–8526.

[4] A. Gomez-Munoz, Ceramide 1-phosphate/ceramide, a switch between life anddeath, Biochim. Biophys. Acta 1758 (2006) 2049–2056.

[5] N.F. Lamour, C.E. Chalfant, Ceramide kinase and the ceramide-1-phosphate/cPLA2alpha interactions as a therapeutic target, Current Drug Targets 9 (2008)674–682.

[6] A. Carre, C. Graf, S. Stora, D. Mechtcheriakova, R. Csonga, N. Urtz, A. Billich, F.Bornancin, Ceramide kinase targeting and activity determined by its N-terminalpleckstrin homology domain, Biochem. Biophys. Res. Commun. 324 (2004)1215–1219.

[7] P. Rovina, M. Jaritz, S. Hofinger, C. Graf, P. Devay, A. Billich, T. Baumruker, A criticalbeta6–beta7 loop in the pleckstrin homology domain of ceramide kinase, Biochem. J.400 (2006) 255–265.

[8] M. Sugiura, K. Kono, H. Liu, T. Shimizugawa, H. Minekura, S. Spiegel, T. Kohama,Ceramide kinase, a novel lipid kinase: molecular cloning and functional character-ization, J. Biol. Chem. 277 (2002) 23294–23300.

[9] M. Tuson, G. Marfany, R. Gonzalez-Duarte, Mutation of CERKL, a novel humanceramide kinase gene, causes autosomal recessive retinitis pigmentosa (RP26), Am. J.Hum. Genet. 74 (2004) 128–138.

[10] F. Bornancin, D. Mechtcheriakova, S. Stora, C. Graf, A. Wlachos, P. Devay, UrtzN. , T.Baumruker, A. Billich, Characterization of a ceramide kinase-like protein, Biochim.Biophys. Acta 1687 (2005) 31–43.

[11] C. Graf, S. Niwa, M. Muller, B. Kinzel, F. Bornancin, Wild-type levels of ceramideand ceramide-1-phosphate in the retina of ceramide kinase-like-deficient mice,Biochem. Biophys. Res. Commun. 373 (2008) 159–163.

[12] N. Auslender, D. Sharon, A.H. Abbasi, H.J. Garzozi, E. Banin, T. Ben-Yosef, A Commonfounder mutation of CERKL underlies autosomal recessive retinal degenerationwith early macular involvement among Yemenite Jews, Invest. Ophthalmol. Vis.Sci. 48 (2007) 5431–5438.

[13] M. Ali, V.L. Ramprasad, N. Soumittra, M.D. Mohamed, H. Jafri, Y. Rashid, M.Danciger, M. McKibbin, G. Kumaramanickavel, C.F. Inglehearn, A missensemutation in the nuclear localization signal sequence of CERKL (p.R106S) causesautosomal recessive retinal degeneration, Mol. Vis. (2008) 1960–1964.

[14] A. Avila-Fernandez, R. Riveiro-Alvarez, E. Vallespin, R. Wilke, I. Tapias, D.Cantalapiedra, J. Aguirre-Lamban, A. Gimenez, M.J. Trujillo-Tiebas, C. Ayuso,CERKL mutations and associated phenotypes in seven Spanish families withautosomal recessive retinitis pigmentosa, Invest. Ophthalmol. Vis. Sci. 49 (2008)2709–2713.

[15] A. Billich, F. Bornancin, P. Devay, D. Mechtcheriakova, N. Urtz, T. Baumruker,Phosphorylation of the immunomodulatory drug FTY720 by sphingosine kinases,J. Biol. Chem. 278 (2003) 47408–47415.

[16] H. Liu, M. Sugiura, E. Nava, L.C. Edsall, K. Kono, S. Poulton, S. Milstien, T. Kohama, S.Spiegel, Molecular cloning and functional characterization of a novel mammaliansphingosine kinase type 2 isoform, J. Biol. Chem. 275 (2000) 19513–19520.

[17] C.B. Do, M.S.P. Mahabhashyam, M. Brudno, S. Batzoglou, ProbCons: probabilisticconsistency-based multiple sequence alignment, Genome Res. 15 (2005)330–340.

[18] E. Lidome, C. Graf, M. Jaritz, A. Schanzer, P. Rovina, R. Nikolay, F. Bornancin, Aconserved cysteine motif essential for ceramide kinase function, Biochimie 90(2008) 1560–1565.

[19] S. Mitsutake, Y. Igarashi, Calmodulin is involved in the Ca2+-dependent activationof ceramide kinase as a calcium sensor, J. Biol. Chem. 280 (2005) 40436–40441.

[20] R.E. Pagano, O.C. Martin, H.C. Kang, R.P. Haugland, A novel fluorescent ceramideanalogue for studying membrane traffic in animal cells: accumulation at the Golgiapparatus results in altered spectral properties of the sphingolipid precursor, J.Cell Biol. 113 (1991) 1267–1279.

[21] S. Kosugi, M. Hasebe, M. Tomita, H. Yanagawa, Nuclear export signal consensussequences defined using a localization-based yeast selection system, Traffic 9(2008) 2053–2062.

[22] M. Brameier, A. Krings, R.M. MacCallum, NucPred Predicting nuclear localization ofproteins, Bioinformatics 23 (2007) 1159–1160.

[23] M. Cokol, R. Nair, B. Rost, Finding nuclear localization signals, EMBO Rep. 1 (2000)411–415.

[24] K. Nakai, P. Horton, PSORT: a program for detecting sorting signals in proteins andpredicting their subcellular localization, Trends Biochem. Sci. 24 (1999) 34–35.

[25] Y. Inagaki, S. Mitsutake, Y. Igarashi, Identification of a nuclear localization signal inthe retinitis pigmentosa-mutated RP26 protein, ceramide kinase-like protein,Biochem. Biophys. Res. Commun. 343 (2006) 982–987.

[26] A. Godi, A.D. Campli, A. Konstantakopoulos, G.D. Tullio, D.R. Alessi, G.S. Kular, T.Daniele, P. Marra, J.M. Lucocq, M.A.D. Matteis, FAPPs control Golgi-to-cell-surfacemembrane traffic by binding to ARF and PtdIns(4)P, Nat. Cell Biol. 6 (2004)393–404.

[27] T.-J. Kim, S. Mitsutake, Y. Igarashi, The interaction between the pleckstrinhomology domain of ceramide kinase and phosphatidylinositol 4,5-bisphosphateregulates the plasma membrane targeting and ceramide 1-phosphate levels,Biochem. Biophys. Res. Commun. 342 (2006) 611–617.

[28] S.C. Wang, P.A. Frey, S-adenosylmethionine as an oxidant: the radical SAMsuperfamily, Trends Biochem. Sci. 32 (2007) 101–110.

[29] J.D. Thompson, D.G. Higgins, T.J. Gibson, CLUSTAL W: improving the sensitivity ofprogressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice, Nucl. Acids Res. 22 (1994)4673–4680.