FEBS Letters 580 (2006) 191–198

Mitochondrial targeting of growth suppressor protein DLC2through the START domain

David Chi-Heng Nga, Shing-Fai Chana, Kin Hang Koka, Judy Wai Ping Yama,b,Yick-Pang Chinga,b, Irene Oi-lin Ngb, Dong-Yan Jina,*

a Department of Biochemistry, Faculty of Medicine, The University of Hong Kong, Pokfulam, Hong Kongb Department of Pathology, Faculty of Medicine, The University of Hong Kong, Pokfulam, Hong Kong

Received 15 October 2005; accepted 27 November 2005

Available online 9 December 2005

Edited by Veli-Pekka Lehto

Abstract Deleted in liver cancer 2 (DLC2) is a candidate tumorsuppressor frequently found to be deleted in hepatocellular carci-noma. In this study, we determined the subcellular localization ofDLC2. Co-localization and biochemical fractionation studies re-vealed that DLC2 localized to mitochondria. In addition, theDLC2-containing cytoplasmic speckles were in proximity to li-pid droplets. A DLC2 mutant containing the steroidogenic acuteregulatory protein-related lipid transfer (START) domain onlyshowed a localization pattern identical to that of DLC2. Takentogether, we have provided the first evidence for mitochondriallocalization of DLC2 through the START domain. These find-ings might have implications in liver physiology and carcinogen-esis.� 2005 Federation of European Biochemical Societies. Publishedby Elsevier B.V. All rights reserved.

Keywords: Deleted in liver cancer 2; Steroidogenic acuteregulatory protein-related lipid transfer domain;Mitochondria; Lipid droplets; Growth suppressor protein;Hepatocellular carcinoma

1. Introduction

Hepatocellular carcinoma (HCC) is one leading cause of

cancer death in Asia and in the World [1]. Hepatocarcinogen-

esis is a multistage process, in which chromosomal aberrations

frequently occur on chromosome 1p, 8p, 13q, 16p and 17p,

leading to alteration of genes that govern cell growth and tu-

mor suppression [2]. Well-known tumor suppressor genes in-

volved in HCC include p53, c-myc, p16INKINK4 and b-catenin[3]. However, HCC is genetically heterogeneous and additional

tumor suppressors critical in HCC remain to be characterized.

Deleted in liver cancer 2 (DLC2), also known as steroido-

genic acute regulatory protein-related lipid transfer (START)

domain containing protein 13 (STARD13), is a novel growth

suppressor protein we have identified [4]. DLC2 is a paralog

of DLC1, a known tumor suppressor gene at chromosome

Abbreviations: DLC2, deleted in liver cancer 2; GAP, GTPase activa-ting protein; HCC, hepatocellular carcinoma; PBS, phosphate bufferedsaline; SAM, sterile a-motif; SRE, serum response element; StAR,steroidogenic acute regulatory protein; START, steroidogenic acuteregulatory protein-related lipid transfer; TRITC, tetramethylrhod-amine isothiocyanate

*Corresponding author. Fax: +852 2855 1254.E-mail address: dyjin@hkucc.hku.hk (D.-Y. Jin).

0014-5793/$32.00 � 2005 Federation of European Biochemical Societies. Pu

doi:10.1016/j.febslet.2005.11.073

8p22-p21.3 commonly deleted in HCC [5–7]. DLC1 and

DLC2 proteins share 51% identity and 64% similarity at the le-

vel of their amino acid sequences. Interestingly, DLC2 local-

izes to chromosome 13q12.3, which is also frequently found

to be deleted in HCC. Existing evidence from analysis of

DLC2 DNA and mRNA suggests that DLC2 is underex-

pressed in a significant number of HCC cases. Sequence anal-

ysis indicates that DLC2 is a multidomain protein containing

sterile a-motif (SAM), GTPase-activating protein (GAP) and

START domains [4]. All three domains are also highly con-

served in DLC1, since 76%, 74% and 60%, respectively, of their

amino acid residues are identical.

We have previously shown that DLC2 protein has GAP

activity specific for small GTPases RhoA and Cdc42 [4]. Thus,

DLC2 should stimulate the hydrolysis of GTP in those GTP-

ases, turning them into inactive GDP-bound form and shutting

down the signal transduction [8,9]. Consistent with this, the

introduction of human DLC2 into mouse fibroblasts sup-

presses Ras signaling and Ras-induced cellular transformation

in a GAP-dependent manner. Those findings suggest a role for

DLC2 in growth suppression and carcinogenesis [4].

While the RhoGAP activity of DLC2 has been characterized

[4], the function of its START domain remains to be under-

stood. The START domain has been found in a large variety

of proteins with different functions [10,11]. It is a well-con-

served lipid-binding domain widely found among lower pro-

karyotes, archaea and multicellular eukaryotes [12]. Among

known START containing proteins in mammals, steroidogenic

acute regulatory protein (StAR), metastatic lymph node-64

(MLN64) and phosphatidylcholine transfer protein (PC-TP)

are better characterized members, with their protein structure

resolved using X-ray crystallography and their lipid ligands

identified [12]. StAR and MLN64 specifically bind cholesterol;

for PC-TP, the lipid it binds is phosphatidylcholine [13]. StAR

is also known to stimulate the intake of cholesterol from cyto-

sol into mitochondria which initiates the production of sterol

in steroidogenesis [14]. In this regard, characterization of the

START domain in DLC2 will provide new avenues to func-

tional analysis of DLC2, DLC1 and related proteins. In partic-

ular, it will be of great interest to see whether the START

domain would have an impact on the subcellular localization

of DLC2 through interaction with lipids presented in the bio-

logical membranes.

In this study, we sought to determine the subcellular locali-

zation patterns of DLC2 in cultured hepatoma cells using

confocal immunofluorescence microscopy. Because other

blished by Elsevier B.V. All rights reserved.

192 D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198

well-studied START domain proteins such as StAR play a

critical role in lipid transport to mitochondria [14], we asked

particularly whether DLC2 might associate with mitochon-

dria. We also examined the role of the START domain in this

function. Additionally, we also investigated the localization of

DLC2 in relation to lipid droplets.

2. Materials and methods

2.1. PlasmidsReporter plasmid pSRE-Luc, in which the expression of firefly lucif-

erase is under the control of multiple copies of serum response ele-ments (SREs; AGGATGTCCATATTAGGACATCT, the consensussequences are bold), was purchased from Stratagene. Expression vec-tor for dominant active Ras mutant RasV12 has been described [4].Plasmids pCMV-DLC2, pCMV-DSAM, pCMV-START, pCMV-

DSTART and pCMV-dGAP are based on expression vector pCMV-tag3C (Stratagene). They contained, respectively, the full-lengthDLC2, truncated DLC2 mutant DLC2-DSAM (119–1113 amino acidresidues) without the SAM domain, truncated mutant DLC2-STARTcarrying only the START domain (906–1113 residues) of DLC2, trun-cated mutant DLC2-DSTART lacking the START domain (908–1113residues) only, and R699A point mutant DLC2-dGAP with a defectiveRhoGAP domain [4]. Notably, a Myc epitope has added to the N-ter-minal of DLC2 and DLC2-DSAM, while the other mutants contain anN-terminal V5 tag.

2.2. Cell culture and transfectionHuman hepatoma cell line Huh-7 [15] was obtained from Japanese

Collection of Research Bioresources. Huh-7 cells were maintained inDulbecco’s Modified Eagle’s Medium supplemented with 10% fetal bo-vine serum, 1.5 g/L sodium bicarbonate and antibiotics. Cultured cellswere transfected using GeneJuice reagents (Novagen).

2.3. Luciferase assayDual-luciferase assay was performed with extracts of transiently

transfected Huh-7 cells as previously described [4,16]. For this assay,both firefly and Renilla (sea pansy) luciferase reporter enzymes wereexpressed simultaneously in the cell. The activity of the co-transfectedRenilla reporter gene provides an internal control that can be used fornormalization of the firefly reporter activity recovered, thereby elimi-nating the differences in cell viability and transfection efficiency.

2.4. Western blottingWestern blotting was carried out as previously described [17].

Briefly, protein extracts prepared from transfected and untransfectedHuh-7 cells were solubilized with SDS gel loading buffer (60 mM Tris,2% SDS, 10% glycerol, 5% 2-mercaptoethanol, 0.1% bromophenolblue), separated by SDS–PAGE, and then electroblotted onto Immo-bilon-P PVDF-type membranes (Millipore) using a semi-dry blottingapparatus (Hoefer SemiPhor). Blots were blocked with 5% skim milk,followed by incubation with a monoclonal antibody against Myc tag(Amersham) or against V5 tag (Invitrogen). Blots were then incubatedwith goat anti-mouse secondary antibody conjugated to horseradishperoxidase (Amersham) and visualized by enhanced chemilumines-cence (ECL, Amersham).Mitochondria of Huh-7 hepatoma cells were prepared by using

mitochondria isolation kit (Sigma). Rabbit polyclonal antibodiesagainst human peroxiredoxin-I or against human peroxiredoxin-IIIwere purchased from Lab Frontier (Seoul, Korea). Secondary anti-body was goat anti-rabbit antibody conjugated to horseradish peroxi-dase (Amersham).

2.5. Confocal microscopyLaser-scanning confocal immunofluorescence microscopy was per-

formed as previously described [18,19]. Briefly, monolayer Huh-7 cellswere grown overnight on coverslips put in six-well plate. Transfectedcells were washed with phosphate buffered saline (PBS) and fixed withfreshly prepared 4% paraformaldehyde, buffered in PBS, for 10 min atroom temperature. Membrane on fixed cells was permeabilized with

0.1% Triton-X100/PBS for 10 min at room temperature. To minimizenon-specific staining, a 3% bovine serum albumin in PBS was used toincubate the treated cells for 30 min at room temperature. Incubationwith primary antibodies (rabbit polyclonal anti-Myc antibody fromSanta Cruz and mouse monoclonal anti-V5 from Invitrogen) was typ-ically for 8 h at 40 �C, and with secondary antibodies (fluorescein-con-jugated donkey polyclonal anti-rabbit IgG from Chemicon andfluorescein-conjugated rat polyclonal anti-mouse IgG from Zymax),it was typically for 2 h at room temperature. Mitochondria markerMitotracker Red CMXRos (Molecular Probes) was added to culturemedium 10 min prior to cell harvest to maximize staining and minimizebackground. Lipid stain Nile red (Sigma) was applied and incubated atroom temperature for 30 more minutes. The coverslips were washes forseveral times with PBS and then mounted on slides with Mowiolmounting medium (Mowiol 4-88 from Sigma, prepared in glyceroland Tris–Cl, pH 8.5). Antifade reagent 1,4-diazabicyclo-[2.2.2]octanecan be added to the prepared Mowiol solution to prevent photobleach-ing of the green signal. Double labeling was achieved by using differentfluorophores (e.g., fluorescein and tetramethylrhodamine isothiocya-nate (TRITC)). Images were mostly captured at 63· magnificationwith the help of the LaserSharp software. Some images were furtherenlarged using Adobe Photoshop.

3. Results

3.1. Expression of DLC2 in cultured hepatoma cells

We have previously demonstrated DLC2 to be a novel Rho-

GAP protein that has growth suppressor activity and is fre-

quently underexpressed in HCC tissues and cells [4]. In that

published work, we have not been able to express full-length

DLC2 in cultured HepG2 and Hep3B hepatoma cells, proba-

bly due to the cytotoxic effects induced by ectopic expression.

Further optimization of the conditions for transfection led to

successful expression of Myc-tagged DLC2 at low level in hu-

man Huh-7 hepatoma cells as detected by Western blotting

(Fig. 1A, lane 1 compared to lane 2). A V5-tagged DLC2 mu-

tant containing the START domain only (DLC2-START or

START) was also efficiently expressed in Huh-7, Hep3B and

other cell lines without inducing cytotoxicity (Fig. 1A, lane 3

compared to lane 4).

In order to obtain a quick answer on the functionality of

DLC2 and DLC2-START in cultured cells, we compared their

abilities to regulate cell signaling. If DLC2 is functional, it

should have an impact on Ras signaling, because DLC2 is a

GAP protein specific for RhoA and Cdc42 that are activated

by Ras [20,21]. Previously, we have shown that a truncated

version of DLC2 (i.e., DLC2-GAP) is capable of suppressing

Ras-induced activation of SRE and Ras-dependent transfor-

mation of NIH3T3 cells [4]. SRE is an important effector of

RhoA and Cdc42 [22] and can be activated potently by

RasV12, a constitutively active mutant of Ras [23]. In keeping

with this, we noted that RasV12 efficiently activated SRE-dri-

ven expression of luciferase reporter (Fig. 1B, compare column

2 to column 1). In this experimental setting, if DLC2 functions

as a RhoGAP protein intracellularly, it should inhibit the stim-

ulation of SRE by RasV12.

Indeed, we observed that DLC2, when expressed in Huh-7

cells, can significantly reduce the activation of SRE-depen-

dent luciferase reporter expression by RasV12 (Fig. 1B, com-

pare column 3 to column 2). In contrast, DLC2-START is

unable to inhibit SRE activity (Fig. 1B, compare column 4

to column 3). These results provided the first evidence that

DLC2 is functional in suppressing Ras signaling in mamma-

lian cells.

Fig. 1. Expression and activity of DLC2 in Huh-7 hepatoma cells. (A) Western blot analysis. Huh-7 cells were mock-transfected (lanes 2 and 4) ortransfected with an expression plasmid for Myc-tagged DLC2 (lane 1) or V5-tagged DLC2-START (lane 3). Equal amount (�10 lg) of protein wasloaded onto each lane. The blots were probed with anti-Myc (lanes 1 and 2) or anti-V5 (lanes 3 and 4) antibody. DLC2 of >120 kDa and DLC2-START of 27 kDa in size are indicated. Also shown are migration positions of molecular weight markers. (B) Suppression of Ras-induced activationof SRE by DLC2. Huh-7 cells were transfected with pSRE-Luc alone (column 1), pSRE-Luc + RasV12 (column 2), pSRELuc + RasV12 + DLC2(column 3) and pSRELuc + RasV12 + DLC2-START (column 4). Results are representative of three independent experiments and the error barsindicate S.E. Luc activity: firefly luciferase activity in arbitrary units normalized to Renilla luciferase activity.

D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198 193

3.2. Subcellular localization of DLC2

Subcellular localization of a protein often provides critical

information for its function. In agreement with previous find-

ings on rat DLC1/p122RhoGAP [24], the overexpression of

DLC2 by adding a higher concentration of expression plasmid

(1 lg/ml) in the transfection mixture caused substantial cell

death and detachment. This cytotoxicity is probably due to

the high RhoGAP activity of DLC2.

In order to achieve better expression of DLC2 in cultured

Huh-7 cells, we adjusted the conditions for transient transfec-

tion. This optimization, which involved a 10-fold reduction of

plasmid concentration to 0.1 lg/ml in the transfection mixture,

led to significant reduction of cell death or detachment. Under

this condition, about 15% of the cells were normally transfec-

ted. All viable transfected cells showed expression of DLC2 in

the cytoplasm (Fig. 2, panels A–C). In particular, the DLC2-

specific signal was punctuate, suggesting that DLC2 was con-

centrated in cytoplasmic speckles (Fig. 2, panels A–C).

Notably, the localization patterns of full-length DLC2 and

DLC2 mutant DLC2-DSAM without the N-terminal SAM do-

main and DLC2-START were very similar (Fig. 2, compare

panels D–F to panels A–C). Thus, the N-terminal SAM do-

main is dispensable for DLC2 localization.

In addition to the catalytic GAP domain, DLC2 and DLC2-

DSAM also share a START domain that might have regula-

tory function. Hence, we next addressed the question as to

whether the START domain is essential for the subcellular

localization of DLC2. The START domain is located at the

very end of DLC2, near the C terminal. Analysis of the amino

acid sequence has revealed no noticeable localization signal on

this portion of DLC2. However, the START domain is

thought to be lipophilic [14]. A mutant of DLC2 containing

the START domain alone, named DLC2-START, was con-

structed and expressed as a V5-tagged protein in Huh-7 cells

(Fig. 1A, lane 3). Under the confocal microscope, DLC2-

START was found to be in dotted form throughout the cyto-

sol, although some cells had DLC2-START protein aggregates

in the perinuclear region (Fig. 2, panels G–I). This localization

pattern of DLC2-START resembles that observed in cells

transfected with DLC2 and DLC2-DSAM (Fig. 2, compare

panels G–I to panels A–F). In another word, the START do-

main can sufficiently target DLC2 to a particular intracellular

compartment. In further support of this role of DLC2-

START, a DLC2 mutant lacking the START domain only

(DLC2-DSTART) was found to distribute homogenously in

the cytoplasm (Fig. 2, panels J–L). In contrast, the R699A

point mutant of DLC2 with a defective RhoGAP domain

(DLC2-dGAP) and an intact START domain, which could

be expressed abundantly in various cell lines without inducing

cytotoxicity, localized to perinuclear speckles. The localization

Fig. 2. Subcellular localization of DLC2 and its mutants in Huh-7cells. Huh-7 cells were transfected with 0.1 lg/ml of plasmid expressingMyc- or V5-tagged DLC2 (panels A–C), DLC2-DSAM (D–F), DLC2-START (G–I), DLC2-DSTART (J–L) or DLC2-dGAP (M–O). Aschematic diagram of the domain structure of DLC2 and its mutants isalso shown (P). Cells were stained for DLC2 with mouse monoclonalanti-Myc (panels B and E) or anti-V5 (panels H, K and N). Nuclearmorphology was visualized by counterstaining with propidium iodide(PI; panels A, D, G, J and M). The green (representing DLC2) and red(representing PI) fluorescent signals were merged by computerassistance (panels C, F and I). The same fields of cells are shown inpanels A–C, D–F, G–I, J–L and M–O.

194 D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198

patterns of DLC2, DLC2-DSAM, DLC2-START and DLC2-

dGAP were very similar.

Collectively, our data demonstrated for the first time an

intracellular function of the START domain in DLC2.

3.3. Mitochondrial localization of DLC2

As the subcellular localization of DLC2 holds the clues to its

function, it will be of great interest to define the identity of

these DLC2-containing speckles. In light of the facts that mito-

chondria are important in lipid metabolism [25] and that other

START domain proteins such as StAR play a role in lipid

transport into mitochondria [14], this organelle is one possible

target of DLC2. To study this possibility, a commercially

available chemical stain for mitochondria, known as Mito-

Tracker-TRITC [26], was used to co-stain Huh-7 cells transfec-

ted with either Myc-tagged DLC2 (Fig. 3) or V5-tagged

DLC2-START (data not shown).

Interestingly, a significant portion of DLC2 co-localized

with mitochondria, as shown in Fig. 3 (see panel C for yel-

low patches indicating co-localization). Computer-assisted

analysis showed that the DLC2 staining overlapped that of

mitochondria perfectly. In fact, over 75% of the fields cap-

tured showed such co-localization of DLC2 and mitochon-

dria. Noteworthily, almost all DLC2 signals overlapped

those of mitochondria, but not all mitochondria were tar-

geted by DLC2.

The co-localization of DLC2 with mitochondria (Fig. 3,

panels A–C) and the similar localization patterns of DLC2

and DLC2-START (Fig. 2) strongly suggest that DLC2-

START would also localize to mitochondria. Indeed,

concentrated DLC2-START speckles co-stained with Mito-

Tracker-TRITC both in both Huh-7 and HeLa cells (Fig. 3,

panels D–I), indicating that the START domain targets

DLC2 to mitochondria.

To verify the mitochondrial localization of DLC2, biochem-

ical fractionation experiments were carried out. Mitochondria

were isolated from the lysate of Huh-7 cells transfected with

Myc-DLC2 and probed with an anti-Myc antibody (Fig. 4).

The Myc-tagged DLC2 protein was found both in the total cell

lysate and in the mitochondrial fraction (Fig. 4A, lanes 1 and

2). As a control, we also examined human peroxiredoxin-I and

peroxiredoxin-III, which are known to localize primarily to the

cytoplasm and mitochondria, respectively [18,27,28]. While

mitochondrial peroxiredoxin-III was found both in the total

cell lysate and in the mitochondria (Fig. 4C), cytoplasmic per-

oxiredoxin I was detected only in the total cell lysate, but not

the mitochondrial fraction (Fig. 4B, compare lane 1 to lane 2).

That is to say, the mitochondria prepared in our experiment

were not contaminated with the cytoplasmic fraction. Thus,

our results from biochemical fractionation confirmed the mito-

chondrial localization of DLC2.

3.4. Localization of DLC2 in relation to lipid droplets

The START domain is thought to have lipid-binding activ-

ity [14]. To study the interaction of DLC2 with intracellular

lipids, we asked whether the DLC2-specific fluorescent signals

would overlap with lipid-reactive dyes. It is well known that

intracellular lipids can be stained with various dyes such as

Nile red and Sudan III [29]. Hence, we co-stained lipid-rich

Huh-7 hepatoma cells for both DLC2-START and lipids using

an anti-Myc antibody and Nile red, respectively (Fig. 5).

DLC2-START was used in this experiment because the local-

ization patterns of DLC2 and DLC2-START are very similar

(Fig. 2).

As shown in Fig. 5, the speckles containing DLC2-START

localized around the lipid droplets stained by Nile red. Inter-

estingly, the two signals did not overlap substantially with each

Fig. 3. Mitochondrial localization of DLC2 and DLC2-START. Huh-7 cells were transfected with 0.1 lg/mL pCMV-DLC2 plasmid expressingDLC2 (panels A–C). Huh-7 (panels D–F) or HeLa (panels G–I) cells were transfected with 0.5 lg/mL pCMV-START plasmid expressing DLC2-START. Cells were then stained for DLC2 or DLC2-START with mouse monoclonal anti-Myc (panel B) or anti-V5 (panels E and H), and formitochondria with MitoTracker CMXRos conjugated to TRITC (Molecular Probes; panels A, D and G). The green (representing DLC2 or DLC2-START) and red (representing mitochondria) fluorescent signals were merged by computer assistance (panels C, F and I). Co-localization is in yellow(highlighted with arrows). The same fields of cells are shown in panels A–C, D–F and G–I. Similar results were also obtained with DLC2-START inNIH3T3 cells (data not shown).

D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198 195

other, but a major portion of DLC2-START was found to be

in proximity to the droplets of lipid (Fig. 5, panels C and F).

Thus, consistent with its putative lipid-binding or lipid-transfer

activity, the START domain of DLC2 likely serves a lipid-re-

lated function in the cell and it targets DLC2 to areas proximal

to the lipid droplets.

4. Discussion

In this study, we determined the subcellular localization of

growth suppressor protein DLC2 in human Huh-7 hepatoma

cells using confocal immuno-fluorescence microscopy. First

we performed Western blotting to verify the transient expres-

sion and functionality of full-length DLC2 in Huh-7 hepatoma

cells (Fig. 1). Then we defined a novel pattern for DLC2 local-

ization and demonstrated the role of the START domain in

intracellular targeting of DLC2 (Fig. 2). We established that

DLC2 localized to mitochondria via the START domain (Figs.

3 and 4). Finally, we also documented that the DLC2-contain-

ing cytoplasmic speckles were in proximity to lipid droplets

(Fig. 5).

4.1. Function of START domain in DLC2

Our findings establish one function of the START domain in

targeting DLC2 to subcellular compartments overlapping with

mitochondria and proximal to the lipid droplets. This function

is consistent with the predicted lipid-binding or lipid-transfer

activity of the START domain [14,30]. Actually the START

domain in some other proteins might also play a similar role

in protein targeting by tethering to membrane or to other

START containing proteins [31]. Further investigation on

DLC2-START will provide additional insight into the cellular

function of DLC2 and related proteins. In this context, the

identification of the lipid ligand of DLC2-START remains

an important task for the next stage of our investigation. In

addition, we will also investigate the relevance of this lipid-

binding activity to the predicted interaction with and regula-

tory role on phospholipase Cd1, which has been demonstrated

for rat DLC1 [32].

Other START domain proteins such as MLN64 are criti-

cally involved in carcinogenesis or metastasis [33]. The func-

tion of the START domain in protein targeting raises

another interesting question as to whether this domain is re-

quired for the growth suppressive activity of DLC2. It also

remains to be understood whether the START domain of

DLC2 might serve as a molecular switch to regulate Rho-

GAP activity upon binding with lipids. In this regard, Rho-

GAP protein p190 has previously been shown to be

regulated by phospholipids [34]. On the other hand, it is un-

clear whether the RhoGAP activity of DLC2 might have an

impact on the function of the START domain. All these

questions concerning the START domain of DLC2 warrant

further study.

Fig. 4. Detection of DLC2 in the mitochondrial fraction. Huh-7 cellswere transfected with an expression plasmid for Myc-tagged DLC2.Biochemical fractionation was performed to isolate mitochondria.Equal amounts (�8 lg) of protein from total cell lysate (total; lane 1)or from mitochondrial fraction (mito; lane 2) were loaded. The blotswere probed with anti-Myc (A), anti-Prx-I (B) or anti-Prx-IIIantibody.

196 D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198

In general, our findings on DLC2-START are in keeping

with a recent study which has suggested that the START do-

main in DLC1 is functional [35]. While the START domains

in DLC2 and DLC1 are highly conserved, the subcellular

localization of DLC1 has not been extensively studied. Thus,

it will be of great interest to see whether the START domain

might also target DLC1 to mitochondria.

Fig. 5. Subcellular localization of DLC2-START in relation to lipid dropletV5-tagged DLC2-START. Cells were then stained for DLC2 with mouse mo(panels A and D). The green (representing DLC2) and red (representing li(panels C and F). Co-localization should be in yellow. The same fields of ce

4.2. Mitochondrial localization of DLC2 in relation to lipid

droplets

Our findings that DLC2 localizes to cytoplasmic speckles

overlapping with mitochondria and proximal to lipid droplets

are consistent with recent results from proteomic analysis of

adipocytes, which reveal a close association of lipid droplets

with mitochondria [36].

Lipid droplets are a reservoir of lipids for the synthesis and

maintenance of membranes [36]. The mechanisms by which

lipids are transported into and out of lipid droplets are poorly

understood. Yet, various lipid metabolic enzymes have been

found in or around the lipid droplets. The appearance of

DLC2 in cytoplasmic speckles surrounding the lipid droplets

suggests that DLC2 might serve a lipid-related function.

Further characterization of the cellular function of DLC2 in

relation to lipid droplets requires the verification of the lipid-

binding activity of DLC2 and the identification of its lipid

ligand. Prototypic START domain proteins StAR and

MLN64 play a pivotal role in lipid transport into mitochon-

dria [14,37]. Particularly, StAR localizes to the mitochondria

and promotes the translocation of cholesterol from outer to in-

ner mitochondrial membranes [31]. The localization of DLC2

to mitochondria in tight association with lipid droplets impli-

cates a role of DLC2 in mitochondrial lipid transport. This no-

tion is further strengthened by a recent report on a ‘‘RhoGAP

encoded on chromosome 13q12’’, which turns out to be DLC2

[38]. In that study, the entire coding sequence of DLC2 was

used as bait in a yeast-2-hybrid screen and several interesting

proteins that potentially bind to DLC2 were identified. Among

those candidates, HMG-CoA reductase is more interesting

than others as it is involved in the energy pathway found in

mitochondria. HMG-CoA is the rate limiting enzyme in cho-

lesterol biosynthesis [39]. Since the START domain of DLC2

is actively targeted to the mitochondria, it is reasonable to

hypothesize that this START domain will indeed tether with

HMG-CoA in vivo and play a pivotal role in the lipid biosyn-

thesis pathway.

s. Hepatoma Huh-7 cells were transfected with expression plasmid fornoclonal anti-V5 (panels B and E) and for lipid droplets with Nile redpid droplets) fluorescent signals were merged by computer assistancells are shown in panels A–C and D–F.

D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198 197

DLC2 is thought to be a tumor suppressor protein in HCC

[4]. The mitochondrial targeting of DLC2 raises the interesting

possibility that DLC2 might fulfill its growth suppressive func-

tion by regulating mitochondrial membrane permeability and

the mitochondrial pathway of apoptosis [40]. Thus, it will be

of interest to investigate whether DLC2 modulates cell survival

and how this modulation contributes to its effects on cell pro-

liferation. In connection to this, the relationship between mito-

chondrial targeting of DLC2 and the induction of apoptosis

merits further investigations.

Acknowledgments:D.-Y.J is a Leukemia and Lymphoma Society Scho-lar. This work was supported in part by a grant to I.O.-l.N. and D.-Y.J. from the Hong Kong Research Grants Council (Project HKU7281/01M).

References

[1] Bosch, F.X., Ribes, J., Dıaz, M. and Cleries, R. (2004) Primaryliver cancer: worldwide incidence and trends. Gastroenterology127, S5–S16.

[2] Nagai, H., Pineau, P., Tiollais, P., Buendia, M.A. and Dejean, A.(1997) Comprehensive allelotyping of human hepatocellularcarcinoma. Oncogene 14, 2927–2933.

[3] Thorgeirsson, S.S. and Grisham, J.W. (2002) Molecular patho-genesis of human hepatocellular carcinoma. Nat. Genet. 31, 339–346.

[4] Ching, Y.-P., Wong, C.-M., Chan, S.-F., Leung, T.H.-Y., Ng,D.C.-H., Jin, D.-Y. and Ng, I.O.-L. (2003) Deleted in liver cancer(DLC) 2 encodes a RhoGAP protein with growth suppressorfunction and is underexpressed in hepatocellular carcinoma. J.Biol. Chem. 278, 10824–10830.

[5] Yuan, B.-Z., Miller, M.J., Keck, C.L., Zimonjic, D.B., Thorge-irsson, S.S. and Popescu, N.C. (1998) Cloning, characterization,and chromosomal localization of a gene frequently deleted inhuman liver cancer (DLC-1) homologous to rat RhoGAP. CancerRes. 58, 2196–2199.

[6] Ng, I.O.L., Liang, Z.-D., Cao, L. and Lee, T.K. (2000) DLC-1 isdeleted in primary hepatocellular carcinoma and exerts inhibitoryeffects on the proliferation of hepatoma cell lines with deletedDLC-1. Cancer Res. 60, 6581–6584.

[7] Wong, C.M., Lee, J.M.F., Ching, Y.P., Jin, D.-Y. and Ng, I.O.L.(2003) Genetic and epigenetic alterations of DLC-1 gene inhepatocellular carcinoma. Cancer Res. 63, 7646–7651.

[8] Hall, A. (1998) Rho GTPases and the actin cytoskeleton. Science279, 509–514.

[9] Moon, S.Y. and Zheng, Y. (2003) Rho GTPase-activatingproteins in cell regulation. Trends Cell Biol. 13, 13–22.

[10] Masucci, J.D., Rerie, W.G., Foreman, D.R., Zhang, M., Galway,M.E., Marks, M.D. and Schiefelbein, J.W. (1996) The homeoboxgene GLABRA2 is required for position-dependent cell differen-tiation in the root epidermis of Arabidopsis thaliana. Development122, 1253–1260.

[11] Ponting, C.P., Schultz, J., Milpetz, F. and Bork, P. (1999)SMART: identification and annotation of domains from signal-ling and extracellular protein sequences. Nucleic Acids Res. 27,229–232.

[12] Iyer, L.M., Koonin, E.V. and Aravind, L. (2001) Adaptations ofthe helix-grip fold for ligand binding and catalysis in the STARTdomain superfamily. Proteins 43, 134–144.

[13] Soccio, R.E., Adams, R.M., Romanowski, M.J., Sehayek, E.,Burley, S.K. and Breslow, J.L. (2002) The cholesterol-regulatedStarD4 gene encodes a StAR-related lipid transfer protein withtwo closely related homologues, StarD5 and StarD6. Proc. Natl.Acad. Sci. USA 99, 6943–6948.

[14] Strauss III, J.F., Kishida, T., Christenson, L.K., Fujimoto, T. andHiroi, H. (2003) START domain proteins and the intracellulartrafficking of cholesterol in steroidogenic cells. Mol. Cell. Endo-crinol. 202, 59–65.

[15] Nakabayashi, H., Taketa, T., Miyano, K., Yamane, T. and Sato,J. (1982) Growth of human hepatoma cells lines with differenti-

ated functions in chemically defined medium. Cancer Res. 42,3858–3863.

[16] Huang, G.-J., Zhang, Z.-Q. and Jin, D.-Y. (2002) Stimulation ofIKK-c oligomerization by the human T-cell leukemia virusoncoprotein Tax. FEBS Lett. 531, 494–498.

[17] Chun, A.C.S. and Jin, D.-Y. (2003) Transcriptional regulation ofmitotic checkpoint gene MAD1 by p53. J. Biol. Chem. 278,37439–37450.

[18] Jin, D.-Y., Chae, H.Z., Rhee, S.G. and Jeang, K.-T. (1997)Regulatory role for a novel human thioredoxin peroxidase in NF-jB activation. J. Biol. Chem. 272, 30952–30961.

[19] Chin, K.-T., Zhou, H.-J., Wong, C.-M., Lee, J.M.-F., Chan, C.-P., Qiang, B.-Q., Yuan, J.-G., Ng, I.O.-L. and Jin, D.-Y. (2005)The liver-enriched transcription factor CREB-H is a growthsuppressor protein underexpressed in hepatocellular carcinoma.Nucleic Acids Res. 33, 1859–1873.

[20] Bar-Sagi, D. and Hall, A. (2000) Ras and Rho GTPases: a familyreunion. Cell 103, 227–238.

[21] Sahai, E. and Marshall, C.J. (2002) RHO-GTPases and cancer.Nat. Rev. Cancer 2, 133–142.

[22] Hill, C.S., Wynne, J. and Treisman, R. (1995) The Rho familyGTPases RhoA, Rac1, and CDC42Hs regulate transcriptionalactivation by SRF. Cell 81, 1159–1170.

[23] Joneson, T., White, M.A., Wigler, M.H. and Bar-Sagi, D. (1996)Stimulation of membrane ruffling and MAP kinase activation bydistinct effectors of RAS. Science 271, 810–812.

[24] Sekimata, M., Kabuyama, Y., Emori, Y. and Homma, Y. (1999)Morphological changes and detachment of adherent cells inducedby p122, a GTPase-activating protein for Rho. J. Biol. Chem. 274,17757–17762.

[25] Daum, G. (1985) Lipids of mitochondria. Biochim. Biophys. Acta822, 1–42.

[26] Simpson, P.B., Mehotra, S., Lange, G.D. and Russell, J.T. (1997)High density distribution of endoplasmic reticulum proteins andmitochondria at specialized Ca2+ release sites in oligodendrocyteprocesses. J. Biol. Chem. 272, 22654–22661.

[27] Gourlay, L.J., Bhella, D., Kelly, S.M., Price, N.C. andLindsay, J.G. (2003) Structure–function analysis of recombi-nant substrate protein 22 kDa (SP-22): a mitochondrial 2-CYSperoxiredoxin organized as a decameric toroid. J. Biol. Chem.278, 32631–32637.

[28] Zhou, Y., Kok, K.H., Chun, A.C.S., Wong, C.M., Wu, H.W.,Lin, M.C.M., Fung, P.C.W., Kung, H.-f. and Jin, D.-Y. (2000)Mouse peroxiredoxin V is a thioredoxin peroxidase that inhibitsp53-induced apoptosis. Biochem. Biophys. Res. Commun. 268,921–927.

[29] Fukumoto, S. and Fujimoto, T. (2002) Deformation of lipiddroplets in fixed samples. Histochem. Cell Biol. 118, 423–428.

[30] Alpy, F. and Tomasetto, C. (2005) Giving lipids a START: theStAR-related lipid transfer (START) domain in mammals. J. CellSci. 118, 2791–2801.

[31] Soccio, R.E. and Breslow, J.L. (2003) StAR-related lipid transfer(START) proteins: mediators of intracellular lipid metabolism. J.Biol. Chem. 278, 22183–22186.

[32] Homma, Y. and Emori, Y. (1995) A dual functional signalmediator showing RhoGAP and phospholipase C-delta stimulat-ing activities. EMBO J. 14, 286–291.

[33] Moog-Lutz, C., Tomasetto, C., Regnier, C.H., Wendling, C.,Lutz, Y., Muller, D., Chenard, M.-P., Basset, P. and Rio, M.-C.(1997) MLN64 exhibits homology with the steroidogenic acuteregulatory protein (STAR) and is over-expressed in human breastcarcinomas. Int. J. Cancer 71, 183–191.

[34] Ligeti, E., Dagher, M.C., Hernandez, S.E., Koleske, A.J. andSettleman, J. (2004) Phospholipids can switch the GTPasesubstrate preference of a GTPase-activating protein. J. Biol.Chem. 279, 5055–5058.

[35] Yamaga, M., Sekimata, M., Fujii, M., Kawai, K., Kamata, H.,Hirata, H., Homma, Y. and Yagisawa, H. (2004) A PLCd1-binding protein, p122/RhoGAP, is localized in caveolin-enrichedmembrane domains and regulates caveolin internalization. GenesCells 9, 25–37.

[36] Brasaemle, D.L., Dolios, G., Shapiro, L. and Wang, R. (2004)Proteomic analysis of proteins associated with lipid droplets ofbasal and lipolytically stimulated 3T3-L1 adipocytes. J. Biol.Chem. 279, 46835–46842.

198 D.C.-H. Ng et al. / FEBS Letters 580 (2006) 191–198

[37] Zhang, M., Liu, P., Dwyer, N.K., Christenson, L.K., Fujimoto,T., Martinez, F., Comly, M., Hanover, J.A., Blanchette-Mackie,E.J. and Strauss III, J.F. (2002) MLN64 mediates mobilization oflysosomal cholesterol to steroidogenic mitochondria. J. Biol.Chem. 277, 33300–33310.

[38] Nagaraja, G.M. and Kandpal, R.P. (2004) Chromosome 13q12encoded Rho GTPase activating protein suppresses growth ofbreast carcinoma cells, and yeast two-hybrid screen shows its

interaction with several proteins. Biochem. Biophys. Res. Com-mun. 313, 654–665.

[39] Danesh, F.R. and Kanwar, Y.S. (2004) Modulatory effects ofHMG-CoA reductase inhibitors in diabetic microangiopathy.FASEB J. 18, 805–815.

[40] Breckenridge, D.G. and Xue, D. (2004) Regulation of mitochon-drial membrane permeabilization by BCL-2 family proteins andcaspases. Curr. Opin. Cell Biol. 16, 647–652.