CGI55 interacts with nuclear proteins and co-localizes to p80-coilin positive-coiled bodies in the...

ORIGINAL ARTICLE

© Copyright 2006 by Humana Press Inc.All rights of any nature whatsoever reserved.1085-9195/(Online)1559-0283/06/44:463–474/$30.00

Cell Biochemistry and Biophysics 463 Volume 44, 2006

INTRODUCTION

The protein CGI-55 (PAI-RBP1) has an amino acidsequence similarity of 67.4% and an identity of 40.7%with the protein Ki-1/57 (1), and a possible ortholog isalso found in Caenorhabditis elegans (2). Recently, wehave shown that both CGI-55 and Ki-1/57 interact withthe chromatin-remodeling factor Chromo-helicase-DNA-binding domain protein (CHD)-3 (2). An alterna-tive splice variant of CGI-55 has been termed PAI-RBP1and was identified as a plasminogen activator inhibitor(PAI)-1-mRNA binding protein (3). This protein is iden-tical with CGI-55, with the exception of a 6-amino aciddeletion near residue 202 (3). The amino acid sequenceof CGI-55 contains a predicted nuclear localization sig-

nal (NLS), two putative coiled-coil motifs, and five can-didate lysine residues that might be modified by smallubiquitin modifier (SUMO)-1. Here, we used CGI-55 as“bait” in a yeast two-hybrid screen to identify its inter-acting protein patterns and obtain possible clues aboutits functional cellular context. In addition to CHD-3,which we have described previously (2), we identifiedeight other proteins that interacted with CGI-55: Daxx,Topoisomerase I binding RS (Topors), HPC2, UBA1,TDG, and protein inhibitor of activated STAT (signaltransducer and activator of transcription) (PIAS)-1, -3,and -y. Several of these proteins have been describedpreviously as permanent or transient components ofpromyelocytic leukemia nuclear bodies (PML-NBs).Our immunolocalization data demonstrate that CGI-55localizes to the nucleus, the nucleolus, and principallyto nuclear Cajal bodies. In summary, our data suggestthat CGI-55 engages in protein interactions that might

CGI-55 Interacts With Nuclear Proteins and Co-Localizes to p80-Coilin Positive-Coiled Bodies in the Nucleus

Taíla A. Lemos1,2 and Jörg Kobarg1,2,*1Centro de Biologia Molecular Estrutural, Laboratório Nacional de Luz Síncrotron, Campinas, SP, Brazil; and

2Departamento de Genética e Evolução, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, SP, Brazil

Abstract

The human protein CGI-55 has been described as a chromo-helicase-DNA-binding domain protein (CHD)-3interacting protein and was also found to interact with the 3′-region of the plasminogen activator inhibitor (PAI)-1 mRNA. Here, we used CGI-55 as a “bait” in a yeast two-hybrid screen and identified eight interacting proteins:Daxx, Topoisomerase I binding RS (Topors), HPC2, UBA2, TDG, and protein inhibitor of activated STAT (signaltransducer and activator of transcription) (PIAS)-1, -3, and -y. These proteins are either structurally or function-ally associated with promyelocytic leukemia nuclear bodies (PML-NBs), protein sumoylation, or the regulationof transcription. The interactions of CGI-55 with Daxx, Topors, PIASy, and UBA2 were confirmed by in vivo co-localization experiments in HeLa cells, by using green (GFP) and red fluorescence fusion proteins. A mappingstudy of the CGI-55 binding site for these proteins revealed three distinct patterns of interaction. The fact thatCGI-55-GFP has been localized in cytoplasm and nucleus in a dotted manner, and its interaction with proteinsassociated with PML-NBs, suggested that CGI-55 might be associated with nuclear bodies. Although Daxx andTopors co-localized with promyelocytic leukemia protein (PML), CGI-55 itself as well as PIASy and UBA2showed only little co-localization with PML. However, we observed that CGI-55 localizes to the nucleolus andco-localizes with p80-coilin positive nuclear-coiled bodies.

Index Entries: PML-NBs; coiled-bodies; Cajal bodies; protein–protein interaction; two-hybrid; domain map-ping; immunolocalization; p80-coilin.

* Author to whom all correspondence and reprint requestsshould be addressed. E-mail: [email protected]

be important to mediate its likely nuclear functions. Itsco-localization with Cajal bodies in the nucleus of HeLacells suggest that it might be involved in the processingof RNA or the regulation of transcription.

The C-terminal of CGI-55 contains, like that of itsparalog Ki-1/57 (2), several arginine- and lysine-richstretches that can represent putative hyaluronan bind-ing motifs (4). Recent studies, demonstrated, however,that Ki-1/57 but not CGI-55 interacts with the adap-tor/signaling molecules RACK1 and PKC (5). Thisinteraction might be relevant for the regulation of itsnuclear function, because it has been shown previouslythat RACK1 can bind to the p53-related transcriptionfactor p73 and thereby inhibit its transcription activa-tion function (6).

The nuclei of higher eukaryotic nuclei can containnumerous distinct substructures that are referred to asnuclear bodies (NBs) (7). The most predominant NBsare the coiled bodies, also termed “Cajal bodies” (CB)and the PML-NBs. CB typically occur in zero to sixcopies per nucleus, with the most frequently foundnumber being two (8). The exact function of the CB isstill unclear, but because they are highly enriched inseveral classes of small ribonucleoproteins, cell cyclecontrol proteins, and basal transcription factors, theymight be involved in the processing of RNA or the reg-ulation of transcription (8,9). The protein p80-coilin isan unambiguous molecular marker component of theCB (7,8).

PML-NBs are also called nuclear body 10 (ND10),Kremer bodies, or PML oncogenic domains (PODS) (7).Electromicroscopy analysis revealed that the PML-NB isa macromolecular structure of 0.2–1 µm. Cells typicallycontain 10–30 of these macromolecular structures,although their number and size may vary depending oncell type, cell cycle, and response to external stimuli.The major proteic component and marker protein of thePML-NB is the protein PML. Although the majority ofPML forms NBs, some protein is located in cytoplasmicbodies and some remains soluble in the nucleus. Severaladditional proteic components of the PML-NB havebeen identified, including Daxx, Topors, SUMO-1, p53,Sp100, sp140, CBP, BLM, and pBR (10,11). Involvementin various biological functions has been attributed toPML-NBs, including: tumor suppression (12), cell cycleregulation (13), transcription regulation (14), and DNAreplication and repair (15).

The protein PML is covalently modified by thesmall ubiquitin-like protein SUMO-1 at three lysineresidues. Only sumoylated PML is associated with thePML-NBs, whereas unmodified PML remains in thesoluble nucleoplasmic fraction. This segregation sug-gests that PML needs to be sumoylated to localize tothe NB (10). Other components of PML-NBs such as

Daxx and Sp100 must also be sumoylated to co-local-ize to nuclear bodies (16,17).

MATERIALS AND METHODS

Plasmid ConstructionsTo perform the two-hybrid screen, human CGI-55

(GenBank accession no. AL080119.1) was polymerasechain reaction (PCR) amplified from a full-length cDNAclone (DKFZp564M2423Q3) and subcloned in frame tothe LexA DNA-binding domain by using the EcoRI andBamHI sites of vector pBTM116 (18), as described previ-ously (2). The CGI-55 cDNA clone had been kindly pro-vided by the Resource Center/Primary Database (Berlin,Germany). This clone has been isolated from a humanfetal brain cDNA library (DKFZhfb2) created by StefanWiemann (DKFZ, Heidelberg, Germany). For the map-ping of the CGI-55 interaction site, seven deletion con-structs were created (Fig. 1A), which have been describedpreviously (2). In a similar manner, CGI-55 was cloned inframe into pEGFP-N1 (Clontech, Mountain View, CA) toexpress the CGI-55-green fluorescent protein (GFP) fusionprotein (2). The cDNAs encoding the hemagglutinin(HA)-tag fusions HA-Topors (amino acids [aa] 360-1045),HA-DAXX (aa 368-740), and HA-PIASy (aa 255-510 andHA-UBA2 (aa 368-640) were amplified from pACT2 andsubcloned in frame into pdSRedC1 (Clontech) to expressthe RED-HA-DAXX, RED-HA-PIASy, RED-HA-Topors,and RED-HA-UBA2 fusion proteins, respectively.

Yeast Two-Hybrid System ScreenCGI-55 fused to the LexA DNA-binding domain (in

plasmid pBTM-116) and a human fetal brain cDNAlibrary (Clontech) cloned in frame to the Gal4 activationdomain (vector pACT2) were used for the screen. Thetwo-hybrid screen was performed by a sequential trans-formation of bait and then library plasmids inSaccharomyces cerevisiae strain L40, carrying the genomi-cally integrated reporters LexA-HIS3 and LexA-LacZ.After transformation, yeast cells were plated on selectiveminimal medium (MM, -W-L,-H) (19) and incubated at30°C until transformants with interacting phenotypeappeared. Transformants were restreaked in duplicate inselective medium (MM, -W-L-H) and tested for β-galac-tosidase expression. The plasmid DNA of blue cloneswas isolated and amplified in Escherichia coli HB101.Plasmid DNA of positive clones was sequenced and ana-lyzed by similarity searches in databases (BLAST).

Mapping the Interaction Site of CGI-55The seven N- and C-terminal deletion constructs of

CGI-55 fused to LexA DNA-binding domain werecotransformed in Saccharomyces cerevisiae strain L40 with

464 Lemos and Kobarg

Cell Biochemistry and Biophysics Volume 44, 2006

the “bait”-plasmid DNAs isolated from the two-hybridscreening. Complete CGI-55 was used as a positive con-trol and an unrelated pACT-AUF1 construct (20) wasused as a negative control. After transformation, yeastclones were streaked out on MM,-W,-L,-H for testingtheir growth capacity under interaction-selective condi-tions. The resulting growth (interaction, +) or lack ofgrowth (no interaction, [minus]) is represented in Fig. 1B.The presence of both types of plasmids was controlled bygrowth on plates with MM, -W,-L (19) (not shown).

Cell Culture, Transfection, and Fluorescence Microscopy

HeLa cells were cultured at 37°C in Dulbecco’s mod-ified Eagle’s medium (DMEM) supplemented with 10%

fetal calf serum, 2 mM L-glutamine, penicillin (100U/mL), and streptomycin (100 µg/mL). For transienttranfection, cells were cultured on glass coverslips for24 h and then transfected with recombinant vector CGI-55-pEGFP-N1 using LipofectAMINE (Invitrogen,Carlsbad, CA). For permanent transfection, cells weregrown to approx 60% confluence on 24-well plates andthen transfected for 48 h with recombinant vector HA-Daxx-pDSRed-C1, HA-Topors-pDSRed-C1, HA-UBA2-pDSRed-C1, and HA-PIASy-pDSRed-C1 by using thelipid LipofectAMINE as above. After 48 h, transformedcells were incubated with DMEM containing 1 mg/mLG418 (Geneticin) for 1 wk. Positive clones were identi-fied by fluorescence microscopy, isolated, and main-tained in DMEM with 1 mg/mL G418. Cells were

Interacting Proteins and Localization of Novel Protein CGI-55 465


Fig. 1. Mapping of the regions of CGI-55 that are involved in the interaction with the proteins identified in the yeast two-hybrid screen. (A) Various N- and C-terminal truncations of human CGI-55 were fused in frame to the DNA-binding domainof LexA in plasmid pBTM116 and transformed into yeast strain L40 together with recombinant pACT2 plasmids encoding theinteracting proteins indicated in the first column of the table in panel B in fusion with the Gal4AD. (B) Table summarizing theobserved interactions between the CGI-55 deletions and the indicated interacting proteins. Interaction was determined by theability of the cotransformed cells to grow on MM,-W,-L,-H. +, growth; interactions; and –, no growth; no interaction. Presenceof “bait” and “prey” plasmids in the cotransformed cells was controlled by growth on MM-W,-L (not shown).

washed with phosphate-buffered saline (PBS) andmounted in 80% glycerol/10 mM Tris-HCl, pH 7.5, oncover glasses and analyzed with a fluorescence micro-scope (Eclipse E600; Nikon, Tokyo, Japan). Cell nucleiwere counterstained with 4,6-diamidino-2-phenylindole(DAPI) (2.5 µg/mL).

Immunofluorescence and AntibodiesThe cells grown on glass cover slips were washed

with PBS and fixed in 100% methanol for 3 min at –20(C.After 30-min incubation in PBS with 3% bovine serumalbumin (BSA), cells were incubated overnight with pri-mary antibodies: monoclonal mouse anti-CGI-55 (2),polyclonal rabbit anti-HA (Y-11), monoclonal mouseanti-HA (F-7), or polyclonal goat anti-PML (A-20) (SantaCruz Biotechnology, Santa Cruz, CA), as indicated in thefigure legends. The antibodies were diluted 1:200 in PBSwith 1% BSA, except for anti-CGI-55, which was used ashybridoma culture supernatant. After washing in PBS,1% BSA, 0.2% Tween-20, the cells were incubated for 1 hwith the secondary antibodies fluorescein isothiocyanate(FITC)-conjugated anti-mouse, rhodamine-conjugatedanti-rabbit, or rhodamine-conjugated anti-goat (SantaCruz Biotechnology), diluted 1:1000 in PBS with 1% BSA.After additional washes, cells were analyzed by fluores-cence microscopy as in the previous section of Materialsand Methods. Rhodamine was detected using a rho-damine filter set G-2E/C TRIC (excitation and emissionat 528–553 and greater than 565 nm, respectively),whereas FITC fluorescence was detected using fluores-cein filter set B-2E/C FITC (excitation and emission at465–485 nm and greater than 505 nm, respectively).

For co-localization of CGI-55 to coiled bodies, the tran-siently transfected CGI-55-GFP-expressing HeLa cellswere fixed and incubated for 2 h at 25°C with primarypolyclonal antibody rabbit anti-human p80-coilin R288 (8)diluted 1:100 in PBS with 1% BSA. As secondary antibody,we used rhodamine-conjugated anti-rabbit antibody.

RESULTS

Identification of Proteins That Interact With CGI-55

CGI-55 is a human protein of unknown function. Weperformed a yeast two-hybrid screen to identify pro-teins interacting with CGI-55 and thereby get a first hinton the possible functional cellular context of the protein.Tranformants (5 × 106) were tested, of which 582 grew inMM without histidine, 125 were blue in the β-galactosi-dase assays, and DNA sequences of greater than 20clones were retrieved. Forty percent of clones repre-sented the chromatin remodeling factor CHD-3. Thisinteraction has been published previously (2).

Human polycomb homolog 2 (HPC2), which is associ-ated with the modulation of the chromatin structure like

the protein CHD-3, was also identified to interact withCGI-55. The other proteins that were identified as CGI-55 interacting partners are listed in Table 1. Most ofthese proteins have been previously described to local-ize to PML-NBs or to be involved in the SUMO-1 mod-ification process.

The CGI-55-interacting proteins involved in SUMO-1modification are three members of the PIAS family ofproteins. PIAS1 and PIASy are inhibitors of activatedSTAT1, and PIAS3 is an inhibitor of activated STAT3. Inaddition to this STAT inhibition activity, PIAS family pro-teins were recently described as SUMO-modification cat-alyzing enzymes or SUMO-ligases (21,22). Furthermore,we identified the SUMO-1 activating enzyme subunit 2(UBA2) as a CGI-55 interacting protein.

The second group of proteins that interact with CGI-55 is associated with PML-NBs and include Daxx andTopors, which is also known as p53BP3 (p53-bindingprotein 3). There is a relationship between the latter twogroups of proteins, because most proteins found inPML-NB, including PML itself are sumoylated (22,23).

Three Patterns of Interaction of CGI-55 With Its Interacting Proteins

To delineate the regions of CGI-55 that engage ininteractions with the proteins identified in the yeasttwo-hybrid assay, various deletion constructs of CGI-55were generated (Fig. 1A). We were able to identify threebasic patterns of interaction (Fig. 1B): proteins thatinteract just with full-length CGI-55 (group A); the pro-tein UBA2, which interacts only with the C-terminalregion of CGI-55; and finally, proteins that interact withboth the C- and N-terminal regions of CGI-55 but fail tointeract with the deletion mutant 4, which representsthe central region of CGI-55 (group C).

Because the N-terminal region of CGI-55 contains aputative coiled-coil motif, we analyzed the interactingproteins for the presence of predicted coiled-coil regions.The interacting proteins can be separated in threegroups: (1) proteins with several predicted coiled-coilregions throughout the whole protein sequence (Daxx),(2) proteins with a single predicted coiled-coil at or nearthe C-terminal (Topors, PIASy, UBA2, and TDG), and (3)proteins without any predicted coiled-coil (HPC2,PIAS1, and PIAS3). Prediction for coiled-coil structuresfor CGI-55 and for the CGI-55-interacting proteins wasperformed by using the COILS-software available atwww.ch.embnet.org/software/COILS_form.html of theSwiss Institute for Experimental Cancer Research,using windows of 14 and 21 amino acids. In sum-mary, these data suggest that there is no strict neces-sity for the CGI-55-interacting proteins identified bythe two-hybrid screen to contain predicted coiled-coilstructures in their sequences.





Tabl

e 1

Cha

ract

eris

tics

of

the

CG

I-55

Int

erac

ting

Pro

tein

s as

Id

enti

fied

in th

e Ye

ast T

wo-

Hyb

rid

Scr

een

Prot

eins

that

inte

ract

w

ith

CG

I-55

C

oded

pro

tein

res

idue

s (s

ynon

yms)

(ret

riev

ed/

full-

size

)D

omai

n co

mpo

siti

onFu

ncti

onR

efer

ence

s

DA

XX

60-7

40/

740

Glu

tam

ic a

cid

ric

h re

gion

, A

popt

osis

, tra

nscr

ipti

on r

egul

atio

n29

368-

740/

740

JNK

act

ivat

or m

otif

Topo

rs (

=p5

3 bi

ndin

g 36

0-10

45/

1045

DN

A-b

ind

ing

dom

ain,

R

NA

poly

mer

ase

II-m

edia

ted

pr

otei

n, =

LU

N)

360-

820/

1045

topo

isom

aras

e I

bind

ing

tran

scri

ptio

n, tr

ansc

ript

iona

l 10

0-60

0/10

45d

omai

n, p

53 b

ind

ing

dom

ain,

pr

oces

sing

26Z

f-C

3HC

4 m

otif

PIA

S136

0-65

5/65

5SA

Pd

omai

n, Z

f-M

IZ m

otif

Prot

ein

inhi

bito

r of

act

ivat

ed S

TAT

1,

30,4

0E

3 SU

MO

-1 li

gase

PIA

S320

4-61

9/61

9Z

f-M

IZ m

otif

Prot

ein

inhi

bito

r of

act

ivat

ed S

TAT

3,

40E

3 SU

MO

liga

se

PIA

Sy25

5-51

0/51

0SA

Pd

omai

n, Z

f-M

IZ m

otif

Prot

ein

inhi

bito

r of

act

ivat

ed S

TAT

1,

40,5

1E

3 SU

MO

liga

se

HPC

2, H

uman

PC

263

-420

/55

0C

hrom

o-d

omai

nD

e-re

pres

sion

of

prot

o-on

coge

nes,

38

regu

lati

on o

f tr

ansc

ript

ion,

ce

llula

r tr

ansf

orm

atio

n

TD

G18

0-41

0/41

0D

NA

glyc

osyl

ase

dom

ain

DN

Are

pair

46

UB

A2

368-

640/

640

Thi

f d

omai

n, U

BA

CT

dom

ain

SUM

O-1

act

ivat

ion

enzy

me

E1

44

CH

D3,

15

51-2

000/

2000

2 PH

D d

omai

ns, 2

chr

omo

Chr

omat

in-r

emod

elin

g,

1807

-200

0/20

00

dom

ains

, 1 h

elic

ase

dom

ain,

tr

ansc

ript

ion

regu

lati

on2

(in

tota

l sev

en d

iffe

rent

1

DN

A-b

ind

ing

dom

ain

clon

es w

ere

retr

ieve

d)

Cellular Localization of CGI-55We observed that CGI-55 occurs in a dotted pattern

in the cytoplasm, perinuclear region, and nucleus ofHeLa cells (Fig. 2A). Such a spotted pattern has beenobserved previously for other proteins, including sev-eral proteins that we found to interact with CGI-55:PIAS1, -3, and -y (24,25); Topors (11,26–28); and Daxx(17,29). When HeLa cells were permanently transfectedwith CGI-55-GFP (Fig. 2A), the expressed CGI-55-GFPfusion showed a dotted pattern in cytoplasm andnucleus (Fig. 2A, bottom row).

The endogenous CGI-55, which was detected byimmunofluorescence, using the anti-CGI-55 monoclonalantibody 10.5.6 (2), showed a similar staining pattern(Fig. 2A, top row). The pattern was also dotted, but ithad a more diffuse overall staining and a marked label-ing of the perinuclear region, the nucleoli, and othersmall and distinct nuclear spots.

CGI-55-GFP Co-Localizes In Vivo With Its Interacting Proteins Identified in the Two-Hybrid Screen

HeLa cells were permanently transfected with plasmidconstructs expressing Red-HA-Daxx, Red-HA-Topors,Red-HA-PIASy, and Red-HA-UBA2 and transientlytransfected with a CGI-55-GFP and analyzed as indicatedin the legend of Fig. 2B.

In HeLa cells transiently transfected with CGI-55-GFP and permanently transfected with RED-HA-Daxx,both proteins showed co-localization (Fig. 2B), asdemonstrated by the merge of the DAXX and CGI-55images (Fig. 2B, fourth row).

CGI-55-GFP also co-localizes with RED-HA-Topors(Fig. 2B, first row) and with RED-HA-PIASy (Fig. 2B,third row). The co-localization occurred in the cyto-plasm and perinuclear. RED-HA-UBA2 also co-localizeswith CGI-55-GFP (Fig. 2B, last row) in the cytoplasmand perinuclear region.

Co-Localization of CGI-55 and Its Interacting Proteins With PML

Based on our observation that some of the proteinsthat interacted with CGI-55 in the yeast two-hybrid sys-tem (Topors and Daxx) have previously been reportedto be physically localized to PML bodies (11,17) and onthe finding that several of the CGI-55-interacting pro-tein partners show a punctuated nuclear localizationpattern (17,29,30), we set out to test whether CGI-55 co-localizes with PML or CBs in human cells.

HeLa cells were cotransfected with the CGI-55-GFPand RED-HA-Daxx, RED-HA-Topors, RED-HA-PIAS,and RED-HA-UBA2 expression vector constructs andimmunostained with anti-PML (red fluorescence) anti-

body and either anti-CGI-55 antibody 10.5.6 or anti-HAantibody (green fluorescence).

HA-Daxx occurs as dots, but also diffusely in thenucleus and cytoplasm and at the nucleoli (Fig. 3A, firstrow), in a similar pattern as seen with endogenous CGI-55 (Fig. 2A, top row). Most of the Daxx dots coincidewith the PML spots, confirming the published findingthat Daxx has been described as associated with PML(17,29,30). Other data from the literature show thatDaxx, when associated with Ask1, is found in the cyto-plasm (31), and when interacting with protein MSP58, itis localized to the nucleoli (32).

For protein Topors, several dots coincide with PML(Fig. 3A, second row). PML, however, does not co-local-ize as much with Topors as the protein Daxx. Rasheed etal. (11) already have described that in some cells, GFP-Topors dots do not have corresponding PML dots, andthey suggested that GFP-Topors may also localize tonuclear bodies distinct from PML-NBs. Interestingly,Topors was found additionally localized to the nucleoli,identically to what was observed.

The HA-PIASy nuclear dots co-localize only partiallywith those of the PML-NB spots (Fig. 3A, third row).The homologous proteins PIAS1 and PIAS3 have beendescribed as SUMO-1 ligases and co-localize withSUMO-1 protein at some nuclear dots (22). Other pro-teins such as Daxx are located at PML-NBs only afterSUMO-1 modification (30). The co-localization of PIASyand PML may suggest that PIAS could function as aSUMO-1 ligase and might cause the translocation ofsumoylated proteins such as PML and DAXX to thePML-NBs.

We further found a partial co-localization of PMLwith HA-UBA2 (Fig. 3A, fourth row). PML shows a dot-ted pattern in both cytoplasm and nucleus, whereasUBA2 has a more diffuse staining predominantly in thecytoplasm. The homolog of UBA2 from Drosophilamelanogaster had been found in the nucleus, perinuclearregion, and also in the deeper cytoplasm, depending onthe cell cycle period (33).

CGI-55-GFP showed a dotted pattern both in the cyto-plasm and in the nucleus, with staining of the nucleolusin the latter (Fig. 3A, last row), just like seen for endoge-nous CGI-55 detected by monoclonal antibody 10.5.6(Fig. 1A, first row). Only some of the CGI-55-GFP dotsco-localized with PML dots in the nucleus (Fig. 3A, lastrow). It seems that the CGI-55 spots also co-localize withsome additional non-PML nuclear spots.

Co-Localization of CGI-55 With p80-CoilinPositive Nuclear Coiled-Bodies

Because we did not observe a significant co-localiza-tion of the CGI-55 positive nuclear spots with PML-NBs,we speculated that CGI-55 might localize to other NBs,




Fig. 2. Cellular localization of CGI-55-GFP, endogenous CGI-55, and co-localization of CGI-55 with Red-HA-Daxx, Red-HA-Topors, Red-HA-PIASy, and Red-HA-UBA2. (A) HeLa cells were (bottom row) or were not (top row) transiently trans-fected with the expression vector CGI-55-pEGFP-N1. Cells were grown on glass cover slips and double stained with DAPI(middle column). Endogenous CGI-55 was detected by immunostaining via primary mouse monoclonal antibody 10.5.6and fluorescein-labeled secondary anti-mouse antibody (top row). The last column of images shows the superimpositionof the CGI-55 and DAPI stainings. Immunolabeled proteins or DAPI staining is indicated in the top right corners of the indi-vidual panels. (B) HeLa cells were permanently transfected with pDSRed1-C1-HA-Daxx, pDSRed1-C1-HA-Topors,pDSRed1-C1-HA-PIASy, and pDSRed1-C1-HA-UBA2 and transiently transfected with recombinant expression vector CGI-55-pEGFP-N1. Cells were grown on glass cover slips, fixed, and immunostained with primary mouse monoclonal antibodyanti-CGI-55 (10.5.6) and rabbit polyclonal anti-HA (Y-11). As secondary antibodies, fluorescein anti-mouse (first column)and rhodamine conjugated anti-rabbit (second column) were used. All cells also were stained with DAPI (third column).Immunolabeled proteins are indicated in the top right corners of the individual panels. Superimposition of the green andred fluorescence colors (fourth column) was visualized as yellow/orange. The fifth column shows the merged images ofCGI-55 and DAPI stainings. Color available in the online version only.

such as the coiled bodies. To test this hypothesis, we per-formed immune co-localization studies of CGI-55 withthe coiled-body protein marker p80-coilin (7,8). Weexpressed CGI-55-GFP by transiently transfecting HeLa

cells and detected it with monoclonal antibody 10.5.6,whereas we used antiserum R228 (8) to detect p80-coilin(Fig. 3B). We found that all p80-coilin positive NBs co-stained with anti-CGI-55 monoclonal antibody 10.5.6



(Fig. 3B, white arrows). As for the co-staining of PMLand CGI-55, we again observed additional nuclear spotsin the anti-CGI-55-labeled cells (Fig. 3B, first column).

DISCUSSION

We performed a yeast two-hybrid screen with thenovel protein CGI-55 and identified in addition toCHD-3 (2), eight proteins that interact with CGI-55:Daxx, Topors, HPC2, TDG and UBA2, and PIAS1, -3,and -y (Table 1). The identification of these new inter-acting proteins suggests that CGI-55 is a nuclear proteinthat is possibly involved in the regulation of transcrip-tion, like these proteins.

Three clones coding the protein Topors were identi-fied. Two of them encompass both the topoisomerase-binding and p53-binding domains at its C-terminalregion; the third clone encompasses the DNA-bindingand ZFC3HC4 domains from the N-terminal region.Topors has been identified independently in yeast two-hybrid screens as Topoisomerase I binding RS protein(26) and also as the p53BP3 (27). Topors amino acids51–375 contain the ring finger motif, leucine zipper, andcoiled-coil regions and binds to DNA (34). The presenceof RING finger and RS domains may suggest thatTopors is involved in RNA polymerase II-mediatedtranscription and mRNA processing. The localization ofGFP-Topors fusion protein in punctuated nuclear sub-domains is consistent with such a hypothesis (26).

Daxx was first identified in a yeast two-hybrid screenfor cDNAs encoding proteins capable of binding to thecytosolic domain of the Fas receptor, an apoptosis-inducing member of the tumor necrosis factor (TNF)receptor family (35). In our screen with CGI-55, twoDaxx clones were identified, one missing only 59 aa atits N-terminal and the other representing approxi-mately the C-terminal half of Daxx. Lin et al. (32)

showed that Daxx, when missing its C-terminal region(aa 501–740), failed to interact with the glucocorticoidreceptor (GR). They further showed that the Daxx C-ter-minal region not only interacts with GR but alsorepresses its transcriptional activity. Lin et al. (32) sug-gested that Daxx C-terminal contains a major dockingdomain for protein–protein interactions, because thisregion has been reported to interact also with transcrip-tion factor Pax5 (36), PML (17), and centrometric proteinCENP-C (37).

HPC2 is a member of the polycomb (PC) protein fam-ily and was identified in our two-hybrid screen as aCGI-55-interacting protein. HPC2 is a member of theDrosophila PC gene family that is involved in the inher-itance of gene activity by progeny cells (38). Usingimmunofluorescence, Satijn et al. (38) found that humanHPC2 co-localizes with another human PC homolog,CBX2, in interphase nuclei, suggesting that these pro-teins are part of a larger complex. Based on the resultsof studies with mutant proteins and overexpression ofwild-type protein, these authors suggested, that humanPC2 is a repressor of proto-oncogene activity and thatinterference with human PC2 function could lead toderepression of proto-oncogene transcription and possi-bly to cellular transformation.

Three members of the PIAS family of proteins wereidentified to interact with CGI-55: PIAS1, -3, and -y. Twoof them contain the complete zinc finger domain Zf-Miz, and the clone encoding PIAS1 contains the C-ter-minal half of this domain. PIAS1 was first identifiedusing a yeast two-hybrid screen with a portion of STAT1as bait (39). Functionally, PIAS1 inhibited STAT1-medi-ated gene activation in response to interferon and co-immunoprecipitated specifically with STAT1 in vivo(39). Furthermore, PIAS1, but not the other analyzedPIAS proteins, inhibited the DNA-binding activity ofSTAT1 in vitro.



Fig. 3. Immunocolocalization of CGI-55, HA-Daxx, -Topors, -PIASy, and -UBA2 with PML or p80-coilin. (A) HeLa cellspermanently transfected with pDSRed1-C1-Daxx, pDSRed1-C1-Topors, pDSRed1-C1-PIASy, and pDSRed1-C1-UBA2 andtransiently transfected with CGI-55-pEGFP-N1 were grown on glass cover slips, fixed, and immunostained with primarymouse monoclonal antibody anti-HA (F-7) or anti-CGI-55 monoclonal antibody 10.5.6 and goat polyclonal anti-PML (A-20).The secondary antibodies were fluorescein-labeled anti-mouse (first column) and rhodamine-conjugated anti-goat (secondcolumn). Cell nuclei were stained with DAPI (third column). Immunolabeled proteins are indicated in the top right cornersof the panels. Superimposition of the green and red fluorescence colors (fourth column) was visualized as yellow/orange.The fifth column shows the merged images of CGI-55 and DAPI stainings. (B) HeLa cells were transiently transfected withrecombinant expression vector CGI-55-pEGFP-N1, fixed, and immunostained with anti-CGI-55 monoclonal mouse anti-body 10.5.6 and rabbit polyclonal anti-R288 anti-p80-coilin antibody. The secondary antibodies were fluorescein-labeledanti-mouse (first column) and rhodamine-conjugated anti-rabbit (second column). Cell nuclei were stained with DAPI(third column). Immunolabeled proteins and DAPI staining are indicated in the bottom or top regions of the panels.Superimposition of the green and red fluorescence colors (fourth column) was visualized as yellow/orange. The fifth col-umn shows the merged images of CGI-55 and DAPI stainings. The white arrowheads indicate the nuclear bodies thatshowed a superimposition of the CGI-55 (green) and p80-coilin stains (red). Color is available in the online version only.

In contrast, Valdez et al. (24) identified PIAS1 as aGu/RNA helicase II-binding protein (GuBP) in yeasttwo-hybrid studies. Gu/RNA helicase II belongs to theDEAD box family of proteins (24). Using immunofluo-rescence, Valdez et al. (24) found that epitope-taggedPIAS1/GuBP localized to the nucleus in a dotted anddiffuse pattern. PIAS1 was also found associated withSUMO-1, p53, and UBC9 (21,22,40).

However, recent data suggest that PIAS proteins canfunction as SUMO ligases, or possibly as tightly boundregulators of sumoylation (21,22,40,41). PIAS1, for exam-ple, catalyzed the sumoylation of p53 both in U2OS cellsand in vitro in a domain-dependent manner (22).

Most interestingly, we also found the SUMO-1 acti-vating enzyme subunit 2 (UBA2) as a CGI-55-interactingprotein in our two-hybrid screen. The CGI-55-interactingUBA2 clone contains the UBACT domain. SUMO acti-vating enzyme is a heterodimeric enzyme that consistsof Aos1 (Sua1 and SAE1) and UBA2 (SAE2) (42–44).Aos1 and UBA2 form thioester bonds with SUMO-1, -2,and -3 (43) and are required for the SUMO-1 modifica-tion of IκBα (42), p53 (45), and RanGAP1 (44).

Finally, we identified thymine DNA-glycosylase(TDG) as a CGI-55-interacting protein in the yeast two-hybrid screen. TDG was first identified by Neddermannet al. (46) and mediates the repair of G/T and G/U mis-matches, which are commonly associated with CpGislands, by removing the thymine and uracil moieties.G/U and G/T mismatches are generated by either mis-incorporations during replication or by spontaneoushydrolytic deaminations of the cytosine or 5-methylcytosine bases (47). Recently, it has been reported thatTDG is modified by SUMO and that the sumoylationalters its enzymatic activity and substrate specificity(48). Takahashi et al. (49) demonstrated that TDG alsointeracts in a noncovalent form with SUMO, that thisinteraction is important for TDG covalent modificationwith SUMO, and that TDG localizes to PML-NBs.

Most of the CGI-55-interacting proteins are eitherstructurally (Daxx and Topors) or functionally (PIAS1,PIAS3, PIASy, and UBA2) associated with PML-NBs orwith the process of protein sumoylation (PIAS andUBA2). PML-NBs are important nuclear subdomainsthat are involved in transcriptional regulation, apopto-sis, and cell cycle control.

Our fluorescence microscopy analysis showed thatRedHA-Daxx, -Topors, -PIASy, and -UBA2 co-localizedwith CGI-55-GFP, suggesting that these proteins can alsointeract with CGI-55 in vivo in human cells. BecauseDaxx (29,30,33), Topors (11), and SUMO-1 (23,50) weredescribed to localize to PML-NBs, CGI-55 might localizeto these structures, too. Immunolocalization, however,showed that CGI-55-GFP co-localizes only partially withPML-NBs but clearly co-localizes with p80-coilin-con-

taining nuclear coiled-bodies. Interestingly, Ki-1/57 (theputative paralog of CGI-55), shows a dotted, corpuscu-lar, nuclear staining, too, and it also interacts with sev-eral CGI-55-interacting proteins in the yeast two-hybridsystem: CHD3 (2), Daxx, Topors, and PIAS3 (F. C. Neryand J. Kobarg, unpublished data).

The association of CGI-55 to PML-NBs or CBs maydepend on its SUMO modification. This has beendescribed for several other proteins, including PML(23). Zhong et al. (10) proposed a model where PMLneeds first to be sumoylated and then recruits otherscomponents, also sumoylated, to form PML-NBs.

We have found that CGI-55 interacts with the SUMOactivating enzyme subunit 2 UBA2 (42) and with theSUMO ligases PIAS1 and -3 (21,22,40,41), indicating thatCGI-55 might be a substrate for sumoylation (Fig. 4).

In summary, GCI-55 co-localizes with CBs and par-tially with PML-NBs and interacts and co-localizes withproteins that are either targets of sumoylation or itselfactively involved in the process of sumoylation of pro-teins. These findings suggest that CGI-55 could be a tar-get of sumoylation, and future studies will addresswhether CGI-55 is sumoylated and what might be pos-sible functional consequences of the described interac-tions in vivo.

ACKNOWLEDGMENTS

We thank the Resource Center/Primary Databaseand Stefan Wiemann for providing the CGI-55 cloneand Dr. Nilson I. T. Zanchin for maintenance of the flu-orescence microscope facility at Centro de BiologiaMolecular Estrutural/Laboratório Nacional de LuzSíncrotron (LNLS). We also thank Maria Eugenia R.Camargo for technical support and Dr. Carlos H. I.Ramos and Luciana R. Camillo for DNA sequencingsupport. We thank Dr. Luis E. C. Andrade (São Paulo,Brazil) for providing the p80-coilin antibody R288. J. K.would like to thank Drs. Alberto Spisni and Thelma A.Pertinhez for the hospitality and organization of a verysuccessful and scientifically excellent 1st LAPS meetingin Angra dos Reis. This work was supported by theLNLS, the Conselho Nacional de DesenvolvimentoCientífico e Tecnológico (CNPq), and by the Fundaçãode Amparo à Pesquisa do Estado São Paulo (FAPESP),through the CEPID and SMOLBNet projects and a per-sonal grant to J.K. (02/09592-3). T.A.L. received anFAPESP doctoral fellowship (00/02216-0).

REFERENCES

1. Kobarg, J., Schnittger, S., Fonatsch, C., et al. (1997)Characterization, mapping and partial cDNA sequenceof the 57-kDa intracellular Ki-1 antigen. Exp. Clin.Immunogenet. 14, 273–280.



2. Lemos, T. A., Passos, D. O., Nery, F. C., and Kobarg, J.(2003) Characterization of a new family of proteins thatinteract with the C-terminal region of the chromatin-remodeling factor CHD-3. FEBS Lett. 533, 14–20.

3. Heaton, J. H., Dlakic, W. M., Dlakic, M., and Gelehrter, T. D.(2001) Identification and cDNA cloning of a novel RNA-binding protein that interacts with the cyclic nucleotide-responsive sequence in the type-1 plasminogen activatorinhibitor mRNA. J. Biol. Chem. 276, 3341–3347.

4. Huang, L., Grammatikakis, N., Yoneda, M., Banerjee, S.D., and Toole, B. P. (2000) Molecular characterization of anovel intracellular hyaluronan-binding protein. J. Biol.Chem. 275, 29,829–29,839.

5. Nery, F. C., Passos, D. O., Garcia, V. S., and Kobarg, J.(2004) Ki-1/57 interacts with RACK1 and is a substrate forPMA activated PKC. J. Biol. Chem. 279, 11,444–11,455.

6. Ozaki, T., Watanabe, K.-I., Nakagawa, T., Miyazaki, K.,Takahashi, M., and Nakagawara, A. (2003) Function ofp73, not of p53, is inhibited by the physical interactionwith RACK1 and its inhibitory effect is counteracted bypRB. Oncogene 22, 3231–3242.

7. Matera, A. G. (1999) Nuclear bodies: multifaceted subdo-mains of the interchromatin space. Trends Cell Biol. 9,302–309.

8. Andrade, L. E. C., Tan, E. M., and Chan, E. K. L. (1993)Immunocytochemical analysis of the coiled body in the cellcycle and during cell proliferation. Proc. Natl. Acad. Sci.U.S.A. 99, 1947–1951.

9. Ogg, S. C., and Lamond, A. I. (2002) Cajal bodies andcoilin-moving towards function. J. Cell Biol. 14, 17–21.

10. Zhong, S., Salomoni, P., and Pandolfi, P. (2000) The tran-scriptional role of PML and the nuclear body. Nat. Cell Biol.2, E85–E90.

11. Rasheed, Z. A., Saleem, A., Ravee, Y., Pandolfi, P. P., andRubin, E. H. (2002) The topoisomerase I-binding RINGprotein, topors, is associated with promyelocytic leukemianuclear bodies. Exp. Cell Res. 277, 152–160.

12. Salomoni, P. and Pandolfi, P. P. (2002) The role of PML intumor suppression. Cell 108, 165–170.

13. Everett, R. D., Lomonte, P., Sternsdorf, T., van Driel, R.,and Orr, A. (1999) Cell cycle regulation of PML modifica-tion and ND10 composition. J. Cell Sci. 112, 4581–4588.

14. Zhong, S., Müller, S., Ronchetti, S., Freemont, P. S., Dejean,A., and Pandolfi, P. P. (2000) Role of SUMO-1-modifiedPML in nuclear body formation. Blood 95, 2748–2752.

15. Borden, K. L. (2002) Pondering the promyelocyticleukemia protein (PML) puzzle: possible functions fromPML nuclear bodies. Mol. Cell. Biol. 22, 5259–5269.

16. Sterndorf, T., Jensen, K., and Will, H. (1997) Evidence forcovalent modification of the nuclear dot-associated pro-teins PML and SP100 by PIC1/SUMO1. J. Cell Biol. 139,1621–1634.

17. Ishov, A. M., Sotnikov, A. G., Negorev, D., et al. (1999)PML is critical for ND10 formation and recruits the PML-interacting protein daxx to this nuclear structure whenmodified by SUMO-1. J. Cell Biol. 147, 221–234.

18. Fields, S. and Song, O. (1989) A novel genetic system todetect protein-protein interactions. Nature 340, 245–246.

19. Vojtek, A. B. and Hollenberg, S. M. (1995) Ras-Raf interac-tion: two-hybrid analysis. Methods Enzymol. 255, 331–342.

20. Moraes, K. C., Quaresma, A. J., Maehnss, K., and Kobarg,J. (2003) Identification and characterization of proteinsthat selectively interact with isoforms of the mRNA bind-ing protein AUF1 (hnRNP D). Biol. Chem. 384, 35–37.

21. Schmidt, D. and Müller, S. (2002) Members of the PIASfamily act as SUMO ligases for c-Jun and p53 and



Fig. 4. Schematic representation. Summary of the putative protein interactions of CGI-55 as identified by the yeast-twohybrid system and the localization of CGI-55 and its interacting proteins to cellular compartments and nuclear domains (seeDiscussion for details).

repress p53 activity. Proc. Natl. Acad. Sci. U.S.A. 99,2872–2877.

22. Kotaja, N., Karvonen, U., Janne, O. A., and Palvimo, J. J.(2002) PIAS proteins modulate transcription factors byfunctioning as SUMO-1 ligases. Mol. Cell Biol. 22,5222–5234.

23. Müller, S., Matunis, M. J., and Dejean, A. (1998)Conjugation with the ubiquitin-related modifier SUMO-1regulates the partitioning of PML within the nucleus.EMBO J. 17, 61–70.

24. Valdez, B. C., Henning, D., Perlaky, L., Busch, R. K., andBusch, H. (1997) Cloning and characterization of Gu/RH-II binding protein. Biochem. Biophys. Res. Commun. 234,335–340.

25. Miyauchi, Y., Yogosawa, S., Honda, R., Nishida, T., andYasuda, H. (2002) Sumoylation of Mdm2 by proteininhibitor of activated STAT (PIAS) and RanBP2 enzymes.J. Biol. Chem. 277, 50,131–50,136.

26. Haluska, P., Jr., Saleem, A., Rasheed, Z., et al. (1999)Interaction between human topoisomerase I and a novelRING-finger/arginine-serine protein. Nucleic Acids Res.27, 2538–2544.

27. Zhou, R., Wen, H., and Ao, S. Z. (1999) Identification of anovel gene encoding a p53-associated protein. Gene 235,93–101.

28. Rechsteiner, M., Rogers, S. W. (1996) PEST sequences andregulation by proteolysis. Trends Biochem. Sci. 21, 267–271.

29. Torii, S., Egan, D. A., Evans, R. A., and Reed, J. C. (1999)Human Daxx regulates Fas-induced apoptosis fromnuclear PML oncogenic domains (PODs). EMBO J. 18,6037–6049.

30. Li, R., Pei, H., Watson, D. K., and Papas, T. S. (2000)EAP1/Daxx interacts with ETS1 and represses transcrip-tional activation of ETS1 target genes. Oncogene 19,745–753.

31. Ko, Y. G., Kang, Y. S., Park, H., et al. (2001) Apoptosis sig-nal-regulating kinase 1 controls the proapoptotic functionof death-associated protein (Daxx) in the cytoplasm. J.Biol. Chem. 276, 39,103–39,106.

32. Lin, D. Y., Lai, M. Z., Ann, D. K., and Shih, H. M. (2003)Promyelocytic leukemia protein (PML) functions as a glu-cocorticoid receptor co-activator by Sequestering Daxx tothe PML oncogenic domains (PODs) to enhance its trans-activation potential. J. Biol. Chem. 278, 15,958–15,965.

33. Shih, H. P., Hales, K. G., Pringle, J. R., and Peifer, M. (2002)Identification of septin-interacting proteins and character-ization of the Smt3/SUMO-conjugation system inDrosophila. J. Cell Sci. 115, 1259–1271.

34. Chu, D., Kakazu, N., Gorrin-Rivas, M. J., et al. (2001)Cloning and characterization of LUN, a novel ring fingerprotein that is highly expressed in lung and specificallybinds to a palindromic sequence. J. Biol. Chem. 276,14,004–14,013.

35. Yang, X., Khosravi-Far, R., Chang, H. Y., and Baltimore, D.(1997) Daxx, a novel Fas-binding protein that activatesJNK and apoptosis. Cell 89, 1067–1076.

36. Emelyanov, A. V., Kovac, C. R., Sepulveda, M. A., andBirshtein, B. K. (2002) The interaction of Pax5 (BSAP) with

Daxx can result in transcriptional activation in B cells. J.Biol. Chem. 277, 11,156–11,164.

37. Pluta, A. F., Earnshaw, W. C., and Goldberg, I. G. (1998)Interphase-specific association of intrinsic centromer proteinCENP-C with Daxx, a death domain-binding protein impli-cated in Fas-mediated cell death. J. Cell. Sci. 111, 2029–2041.

38. Satijn, D. P., Olson, D. J., van der Vlag, J., et al. (1997)Interference with the expression of a novel human poly-comb protein, hPc2, results in cellular transformation andapoptosis. Mol. Cell Biol. 17, 6076–6086.

39. Liu, B., Liao, J., Rao, X., et al. (1998) Inhibition of Stat1-mediated gene activation by PIAS1. Proc. Nat. Acad. Sci.U.S.A. 95, 10,626–10,631.

40. Kahyo, T., Nishida, T., and Yasuda, H. (2001) Involvementof PIAS1 in the sumoylation of tumor suppressor p53.Mol. Cell 8, 713–718.

41. Jackson, P. K. (2001) A new RING for SUMO: wrestlingtranscriptional responses into nuclear bodieswith PIASfamily E3 SUMO ligases. Genes Dev. 15, 3053–3058.

42. Desterro, J. M., Rodríguez, M. S., Kemp, G. D., and Hay, R.T. (1999) Identification of the enzyme required for activa-tion of the small ubiquitin-like protein SUMO-1. J. Biol.Chem. 274, 10,618–10,624.

43. Gong, L., Li, B., Millas, S., and Yeh, E. T. (1999) Molecularcloning and characterization of human AOS1 and UBA2,components of the sentrin-activating enzyme complex.FEBS Lett. 448, 185–189.

44. Okuma, T., Honda, R., Ichikawa, G., Tsumagari, N., andYasuda, H. (1999) In vitro SUMO-1 modification requirestwo enzymatic steps, E1 and E2. Biochem. Biophys. Res.Commun. 254, 693–698.

45. Rodriguez, M. S., Desterro, J. M., Lían, S., Midgley, C. A.,Lane, D. P., and Hay, R. T. (1999) SUMO-1 modificationactivates the transcriptional response of p53. EMBO J. 18,6455–6461.

46. Neddermann, P., Gallinari, P., Lettieri, T., et al. (1996)Cloning and expression of human G/T mismatch-specificthymine-DNA glycosylase. J. Biol. Chem. 271, 12,767–12,774.

47. Lindahl, T. (1982) DNA repair enzymes. Annu. Rev.Biochem. 51, 61–87.

48. Hardeland, U., Steinacher, R., Jiricny, J., and Schär, P.(2002) Modification of the human tymine-DNA-glycosy-lase by ubiquitin-like proteins facilitates enzymaticturnover. EMBO J. 21, 1456–1464.

49. Takahashi, H., Hatakeyama, S., Saitoh, H., and Nakayama,K. I. (2005) Noncovalent SUMO-1 binding of thymineDNA glycosylase (TDG) is required for its SUMO-1 mod-ification and colocalization with the promyelocyticleukemia protein (PML). J. Biol. Chem. 280, 5611–5621.

50. Boddy, M. N., Howe, K., Etkin L. D., Solomon, E., andFreemont, P. S. (1996) PIC 1, a novel ubiquitin-like proteinwhich interacts with the PML component of a multipro-tein complex that is disrupted in acute promyelocyticleukaemia. Oncogene 13, 971–982.

51. Long, J., Matsura, I., He, D., Wang, G., Shuai, K., and Liu,F. (2003) Repression of SMAD transcriptional activity byPIASy, an inhibitor of activated STAT. Proc. Natl. Acad. SciU.S.A. 100, 9791–9796.



CGI55 interacts with nuclear proteins and co-localizes to p80-coilin positive-coiled bodies in the...

Documents

Transcript of CGI55 interacts with nuclear proteins and co-localizes to p80-coilin positive-coiled bodies in the...