Báo cáo khoa học: Functional association of human Ki-1⁄57 with pre-mRNA splicing events docx

Functionally, most of these identified RNA-binding proteins are involved in pre-mRNA splicing regulation, pointing to a role of Ki-1⁄ 57 in pre-mRNA splicing.. Seeksplic-ing for a possibl

splicing events

Gustavo C Bressan1,2, Alexandre J C Quaresma1, Eduardo C Moraes1,2, Adriana O Manfiolli3, Dario O Passos1, Marcelo D Gomes3and Jo¨rg Kobarg1,2

1 Laborato´rio Nacional de Luz Sı´ncrotron, Campinas, SP, Brasil

2 Instituto de Biologia, Universidade Estadual de Campinas, Campinas, SP, Brasil

3 Departamento de Bioquı´mica e Imunologia, Faculdade de Medicina de Ribeira˜o Preto da Universidade de Sa˜o Paulo, Ribeira˜o Preto, Brasil

Keywords

Arg methylation; nuclear bodies;

protein–protein interaction; RNA binding;

splicing

Correspondence

J Kobarg, Laborato´rio Nacional de Luz

Sı´ncrotron, Centro de Biologia Molecular

Estrutural, Rua Giuseppe Ma´ximo Scolfaro

10.000, C.P 6192, 13084-971 Campinas,

SP, Brasil

Fax: +55 19 3512 1006

Tel: +55 19 3512 1125

E-mail: jkobarg@lnls.br

(Received 12 April 2009, revised 8 May

2009, accepted 13 May 2009)

doi:10.1111/j.1742-4658.2009.07092.x

The cytoplasmic and nuclear protein Ki-1⁄ 57 was first identified in malig-nant cells from Hodgkin’s lymphoma Despite studies showing its phos-phorylation, arginine methylation, and interaction with several regulatory proteins, the functional role of Ki-1⁄ 57 in human cells remains to be determined Here, we investigated the relationship of Ki-1⁄ 57 with RNA functions Through immunoprecipitation assays, we verified the associa-tion of Ki-1⁄ 57 with the endogenous splicing proteins hnRNPQ and SFRS9 in HeLa cell extracts We also found that recombinant Ki-1⁄ 57 was able to bind to a poly-U RNA probe in electrophoretic mobility shift assays In a classic splicing test, we showed that Ki-1⁄ 57 can modify the splicing site selection of the adenoviral E1A minigene in a dose-dependent manner Further confocal and fluorescence microscopy analysis revealed the localization of enhanced green fluorescent protein–Ki-1⁄ 57 to nuclear bodies involved in RNA processing and or small nuclear ribonucleo-protein assembly, depending on the cellular methylation status and its N-terminal region In summary, our findings suggest that Ki-1⁄ 57 is probably involved in cellular events related to RNA functions, such

as pre-mRNA splicing

Structured digital abstract

l MINT-7041074: Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with SF2P32 (uni-protkb:Q07021) by two hybrid (MI:0018)

l MINT-7041232: Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with SFRS9 (uni-protkb:Q13242) by pull down (MI:0096)

l MINT-7041203: P80-Coilin (uniprotkb:P38432) and Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) colocalize (MI:0403) by fluorescence microscopy (MI:0416)

l MINT-7041217: SMN (uniprotkb:Q16637) and Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) colocalize (MI:0403) by fluorescence microscopy (MI:0416)

l MINT-7041189: SC-35 (uniprotkb:Q01130) and Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) colocalize (MI:0403) by fluorescence microscopy (MI:0416)

l MINT-7041169: NPM (uniprotkb:P06748) and Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) colocalize (MI:0403) by fluorescence microscopy (MI:0416)

l MINT-7041249: Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with SFRS9 (uni-protkb:O60506) by pull down (MI:0096)

Abbreviations

Adox, adenosine-2¢,3¢-dialdehyde; EGFP, enhanced green fluorescent protein; EMSA, electrophoretic mobility shift assay; GEMS, Gemini of coiled bodies; GST, glutathione S-transferase; hnRNP, heterogeneous nuclear ribonucleoprotein; SMN, survival of motor neurons; snRNP, small nuclear ribonucleoprotein; SR protein, Ser ⁄ Arg protein.

Ki-1 was the first monoclonal antibody used in the

specific detection of the malignant Hodgkin and

Sternberg–Reed cells in Hodgkin lymphoma [1] It

has been demonstrated that Ki-1 binds to the

120 kDa lymphocyte costimulatory molecule CD30

on the Hodgkin cell’s surface [1,2] However, it has

been noticed that there is a cross-reaction of Ki-1

with a functionally and structurally uncharacterized

intracellular phosphoprotein antigen of 57 kDa,

termed Ki-1⁄ 57 [3] Although its relationship with

Hodgkin disease has not been confirmed, initial

stud-ies revealed that Ki-1⁄ 57 is associated with Ser ⁄ Thr

protein kinase activity when isolated from tumor cells

[4] and localizes to both the cytoplasm and the

nucleus, where it could be found at the nuclear pores

and in several nuclear structures [2] Ki-1⁄ 57 was

found to associate with intracellular hyaluronic acid

and other negatively charged molecules in vitro, and

was therefore also named hyaluronic acid-binding

protein 4 [5]

Another human protein, CGI-55, shares 40.7%

iden-tity and 67.4% similarity with Ki-1⁄ 57, suggesting that

they could be paralogs and have similar or redundant

functions in human cells CGI-55 is also a

nucleus⁄ cytoplasmic shuttling protein [6] and, because

it was described as a protein able to bind to the

3¢-UTR region of the mRNA encoding the type 1

minogen activator inhibitor, it was also named

plas-minogen activator inhibitor RNA-binding protein 1

[7] We have recently found that Ki-1⁄ 57 and CGI-55

have overlapping interacting protein partners Among

them are the chromatin remodeling factor

chromo-heli-case DNA-binding domain protein 3 [8], DAXX, and

Topors [6,9] This suggests that the nuclear functions

of both proteins may be related to transcriptional

activity Despite the fact that these proteins share

reasonable sequence similarity, Ki-1⁄ 57, but not

CGI-55, interacts with the transcription factor MEF2C [10],

p53 [9], and the signaling⁄ scaffold receptor of activated

protein kinase C (RACK1) [11,12] Both Ki-1⁄ 57 and

CGI-55 mRNAs show ubiquitous expression in all

human tissues tested, and elevated expression in the heart, muscle, and liver [8] Ki-1⁄ 57 is also expressed

at higher levels in the brain [8]

Both Ki-1⁄ 57 and CGI-55 interact with and are methylated by the protein arginine methyltransferase PRMT1 [13] This enzyme is responsible for the meth-ylation of more than 85% of the cellular protein substrates [14], and targets the arginines embedded in typical Arg⁄ Gly-rich motifs (RG ⁄ RGG ⁄ RXR) [15] These are conserved motifs in many RNA-binding pro-teins, and have been reported to mediate RNA binding [16,17]

In previous yeast two-hybrid screenings, we found the interaction of Ki-1⁄ 57 with several RNA-binding proteins [9] (unpublished observations) Functionally, most of these identified RNA-binding proteins are involved in pre-mRNA splicing regulation, pointing to

a role of Ki-1⁄ 57 in pre-mRNA splicing Here, we show the first functional signatures for Ki-1⁄ 57 in human cells, mainly those concerning its possible involvement in mechanisms of splice regulation

Results

Protein–protein association analysis

In yeast two-hybrid analyses using Ki-1⁄ 57 and PRMT1 as baits, we found common RNA-binding proteins that are functionally associated with each other in the context of pre-mRNA splicing regulation (Fig 1) The splicing proteins SF2p32, YB-1 [9] and SFRS9 (unpublished observation) were found as posi-tive prey clones when we used the N-terminus of Ki-1⁄ 57 as bait in our screens (Fig 1) On the basis of the functional interconnections between the Ki-1⁄ 57-interacting and PRMT1-57-interacting proteins (Fig 1),

we reasoned that heterogeneous nuclear ribonucleo-protein hnRNPQ could also be functionally related to Ki-1⁄ 57 hnRNPQ has been reported to be associated with the regulation of pre-mRNA splicing [18], and has been previously found to be a novel interacting

l MINT-7041065: Ki-1⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with SFRS9 (uni-protkb:Q13242) by two hybrid (MI:0018)

l MINT-7041069: Ki-1 ⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with YB1 (uni-protkb:P67809) by two hybrid (MI:0018)

l MINT-7041079: Ki-1⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0915) with HNRPQ (uniprotkb:O60506) by two hybrid (MI:0018)

l MINT-7041087: Ki-1⁄ 57 (uniprotkb:Q5JVS0) physically interacts (MI:0218) with HNRPQ3 (uniprotkb:O60506-1), HNRPQ2 (uniprotkb:O60506-2) and HNRPQ-1 (uniprotkb:O60506-3)

by anti bait coimmunoprecipitation (MI:0006)

partner and target for Arg methylation by PRMT1

[13,19]

Aiming to confirm the endogenous association of

Ki-1⁄ 57 with proteins involved in splicing regulation,

we performed immunoprecipitation assays from HeLa

cell extracts We confirmed such an association

when we immunoprecipitated Ki-1⁄ 57 (and detected

SFRRS1⁄ 9 and hnRNPQ isoforms) (Fig 2A), and also

when we immunoprecipitated SFRS1⁄ 9 or hnRNPQ

isoforms (and detected Ki-1⁄ 57) (Fig 2B,C),

suggest-ing that these proteins might form complexes in vivo

The pan-antibody against hnRNPQ recognizes the

isoforms hnRNPQ1, hnRNPQ2, and hnRNPQ3, and

the antibody against SFRS1⁄ 9 recognizes the splicing

factors SFRS9 and SFRS1 The latter is also known as

SF2⁄ ASF, and its regulatory subunit, called SF2p32

[20], was also found to be a Ki-1⁄ 57-interacting

part-ner (Fig 1) [9]

Next, we performed pull-down assays with

recom-binant proteins to test the interaction of Ki-1⁄ 57 with

SFRS9 and hnRNPQ in vitro We found that the

bacu-lovirus 6· His–SFRS9 was pulled down by the

bacterial glutathione S-transferase (GST)–Ki-1⁄ 57

(Fig 2D) Similarly, the bacterial GST–hnRNPQ

(1–443) was also pulled down by the bacterial 6· His– Ki-1⁄ 57 (Fig 2E) These results suggest that the inter-action of Ki-1⁄ 57 with these splicing proteins occurs directly and specifically, as no interaction was observed with GST alone

Moreover, we also detected the three hnRNPQ iso-forms when we immunoprecipitated SFRS1⁄ 9, although we did not observe direct in vitro binding activity between these proteins in our experimental conditions in pull-down assays (data not shown) This suggests that these proteins may form part of the same Ki-1⁄ 57-associated complex, although they might not all interact directly with each other

Yeast two-hybrid mapping assays Next, we were interested in knowing the regions of Ki-1⁄ 57 necessary for its interaction with splicing regulatory proteins Several N-terminal and C-terminal Ki-1⁄ 57 truncated forms fused to the LexA DNA-binding domain (Fig 2F) were cotransformed with constructs encoding the Ki-1⁄ 57-interacting proteins fused to a GAL4 activation domain, to test their abil-ity to interact with each other Only the full-length and N-terminal Ki-1⁄ 57 constructs were able to inter-act with the splicing proteins SFRS9, SF2p32, and YB-1 (Fig 2G, columns 1–3) This suggests that the interaction of Ki-1⁄ 57 with these molecules may occur predominantly through its N-terminal region This pattern was not verified for hnRNPQ (Fig 2G, col-umn 4), as we only observed its interaction with the full-length Ki-1⁄ 57 construct On the other hand, this finding may explain why we were not able to identify hnRNPQ in our yeast two-hybrid screens, where only the truncated forms of the N-terminus and C-terminus

of Ki-1⁄ 57 were used as ‘baits’

RNA-binding activity of Ki-1⁄ 57 in vitro Although Ki-1⁄ 57 does not have any classic RNA-bind-ing domains in its amino acid sequence, it has several Arg⁄ Gly-rich clusters (RGG-box) (Fig 3A) The RGG motif’s importance for the interaction of many RNA-binding proteins with RNA has already been reported [16,17] The two major Ki-1⁄ 57 RGG-boxes located at its C-terminal region are highly similar to those of its putative paralog, CGI-55 (Fig 3A) The exact role of CGI-55 in human cells also remains unknown, but it was found to bind to the 3¢-UTR of type 1 plasminogen activator inhibitor mRNA [7]

To test whether Ki-1⁄ 57 also binds RNA, we per-formed electrophoretic mobility shift assays (EMSAs)

We found that the recombinant GST–Ki-1⁄ 57 bound

Fig 1 Functional interconnections of Ki-1 ⁄ 57 with splicing

regula-tory proteins through direct physical interactions or participation in

common protein complexes Black bold lines: experiments

described in this article (see Results for details) Dotted lines:

previ-ously published findings; PRMT1 is found in the same complex

with SF2p32 [48], SFRS9 is associated with SF2p32 [20], and YB-1

and SFRS9 interact with each other [21] Thin dotted lines: YB-1

and hnRNPQ are functionally related, as both interact with

hnRNPD ⁄ AUF1 [42]; SFRS9 is highly similar in amino acid

sequence to SFRS1 (SF2 ⁄ ASF) Numbers in square brackets

indi-cate the respective references The database accession numbers

for the proteins shown are (in UniProt code): Ki-1 ⁄ 57, Q5JVS0;

SFRS9, Q13242; hnRNPQ, O60506; YB-1, P67809; SFRS1,

Q07955; SF2P32, Q07021; PRMT-1, Q99873.

to a U-rich RNA probe (Fig 3B, lanes 3–6) This was

relatively specific, as we also tested other RNA

homo-polymer probes (poly-A, poly-C, and poly-G), and

found no significant binding activity (data not shown)

These observations suggest that the binding of Ki-1⁄ 57

to its putative cellular RNA targets may involve U-rich regions, instead of A-rich regions as reported for CGI-55 [7]

As we found the N-terminus to be an important region for the interaction of Ki-1⁄ 57 with its protein

A

B

C

E

H

G

F

D

Fig 2 Confirmation of the protein–protein interactions among Ki-1 ⁄ 57 and proteins involved in pre-mRNA splicing (A–C) Immunoprecipita-tion assays (IP) of endogenous proteins HeLa cell extracts were immunoprecipitated with G-Sepharose beads and the indicated antibodies The obtained protein complexes were analyzed by western blot (WB) as indicated in the figure panels Arrows indicate the positions of ana-lyzed proteins WL, whole cell lysate Immunoprecipitation with the indicated control antibodies is shown on the right side (D, E) In vitro pull-down assays Recombinant proteins from bacteria [GST, GST–Ki-1 ⁄ 57, GST–hnRNPQ(1–443), 6· His–Ki-1 ⁄ 57] or baculovirus (6· His–SFRS9) were loaded onto Ni2+–nitrilotriacetic acid (6· His-fusion) or glutathione–Sepharose beads (GST-fusion) and incubated with supernatants of cell lysates as indicated Arrows indicate the detected proteins Arrowheads point to the position of the control protein GST The additional bands observed correspond to proteolysis degradation products (F–H) Yeast two-hybrid mapping of Ki-1 ⁄ 57 regions (F) that interact with the indicated splicing proteins Black boxes in the diagrams represent the RGG-box motifs present in the sequence of Ki-1 ⁄ 57 (see also Fig 3A) L40 yeast cells were cotransformed with the plasmids encoding several Ki-1 ⁄ 57 truncated constructs fused to LexA and the plasmids encoding the prey proteins fused to the GAL4-activating domain (G) Protein–protein interactions were checked through analysis of reporter gene activation: b-galactosidase activity or capacity to grow on selective minimal medium (in the absence of the amino acids Trp, Leu, and His) (not shown) (H) Autoactivation control: inability of full-length Ki-1 ⁄ 57 to activate reporter genes in the absence of its interacting partners.

partners involved in splicing regulation (Fig 2G), we

further investigated which regions of Ki-1⁄ 57 were

involved in binding to the poly-U RNA We observed

that the two smaller RGG-box-containing constructs

6· His–Ki-1 ⁄ 57(151–260) and 6· His–Ki-1 ⁄ 57(261–

413) were able to bind, although weakly, to the

poly-U probe; however, only the larger C-terminal

Ki-1⁄ 57(122–413) construct could achieve binding as

strong as that of full-length GST–Ki-1⁄ 57 Hence, the

C-terminal region Ki-1⁄ 57(122–413) seems to be

necessary and sufficient for efficient interaction

with the RNA poly-U (Fig 3B, lanes 7–10) The observed ‘supershifted’ bands seemed to be stronger upon the incubation of poly-U with increasing quan-tities of Ki-1⁄ 57(1–413) and Ki-1⁄ 57(122–413) (Fig 3B, lanes 4–10) A 25-mer poly-U molecule is large enough to bind more than one molecule of Ki-1⁄ 57 Owing to its low expression yield and low stability in solution, an N-terminal construct, 6· His–Ki-1⁄ 57(1–222), could not be tested in these EMSA experiments

Influence of Ki-1⁄ 57 on E1A pre-mRNA splicing

in vivo The association of Ki-1⁄ 57 with splicing proteins pointed to a possible functional role in pre-mRNA splicing regulation Therefore, we investigated whether Ki-1⁄ 57 could modulate the splicing site selection of the adenoviral E1A test minigene, previously explored for the Ki-1⁄ 57-interacting proteins SFRS9 and YB-1 [21] Depending on the 5¢-splice site selection, the E1A pre-mRNA may generate five isoforms: 13S, 12S, 11S, 10S, and 9S (Fig 4A) [22] These isoforms can be monitored by RT-PCR followed by agarose gel analy-sis, where the intensity of each band in the gel directly correlates with the splicing site selection, which, in turn, reflects the positive or negative influence of regu-latory proteins [22,23]

We transiently cotransfected the encoding E1A mini-gene plasmid with increasing amounts of vectors expressing the recombinant enhanced green fluorescent protein (EGFP)–Ki-1⁄ 57 in COS7 cells We observed a significant effect of EGFP–Ki-1⁄ 57 in modifying the pattern of splicing of E1A mRNA in comparison with empty pEGFP vector (Fig 4B) Expression of EGFP– Ki-1⁄ 57 leads to formation of the 10S and 9S iso-forms, concomitantly with a reduction of 13S isoform formation, in a dose-dependent way (Fig 4B, lanes 2–4) This finding strongly suggests the functional involvement of Ki-1⁄ 57 in regulatory mechanisms of pre-mRNA splicing

Although we also observed a significant modification

of the E1A mRNA splicing pattern by Ki-1⁄ 57(1–222) and Ki-1⁄ 57(122–413), respectively, it only occurred at the highest plasmid concentrations used (Fig 4C,D) Moreover, the effects seemed to be isoform specific for each Ki-1⁄ 57 region, as the formation of 10S mRNA was only increased by the C-terminal region of Ki-1⁄ 57 although with a lower efficiency in compari-son with the full length protein (Fig 4C–E) The influ-ence of the C-terminal construct Ki-1⁄ 57(122–413) on the generation of the 9S mRNA was also more pronounced (Fig 4F)

A

B

Fig 3 GST–Ki-1 ⁄ 57 binds poly-U RNA in vitro (A) Schematic view

of the RGG-box motif localization in the sequence of Ki-1 ⁄ 57 and

its paralog CGI-55 (database accession number: Q8NC51) (B, C)

EMSA results (B) Interaction of Ki-1 ⁄ 57 with poly-U RNA On top

of the panel: the three different truncated versions of recombinant

Ki-1 ⁄ 57 used in the mapping experiments Increasing

concentra-tions (0.5, 1, 2 and 4 m M , respectively) of the recombinant protein

GST–Ki-1 ⁄ 57 and the three 6· His-fused truncated versions of

Ki-1 ⁄ 57 (122–413, 151–260, and 261–413) were incubated with the

32 P-labeled poly-U probe (25-mer) and subjected to native

polyacryl-amide gel (10% gel, with a 29:1 acrylpolyacryl-amide:bis-acrylpolyacryl-amide ratio)

electrophoresis The arrow indicates shifted bands and the dashed

arrowheads indicate supershifted bands.

B

F D

Fig 4 Ki-1 ⁄ 57 influences the splicing pattern of the E1A pre-mRNA (A) Diagram showing the splicing events that generate the 13S, 12S, 10S and 9S mRNAS of the E1A reporter gene [22,23] (B–D, G) In vivo splicing assays COS7 cells were transiently cotransfected with an E1A minigene encoding plasmid, an empty pEGFP vector and increasing amounts (1·, 5 lg; 2·, 10 lg; 3·, 15 lg) of pEGFP-Ki-1 ⁄ 57 (full length) or pEGFP–Ki-1⁄ 57(1–222) and pEGFP–Ki-1 ⁄ 57(122–413) vectors The empty pEGFP vector was used to keep constant the DNA con-centration in each transfection Splicing activity quantization was performed as described in Experimental procedures The displayed figures are representative of at least three independent experiments Vertical bars in the graphs indicate ± standard deviation Wherever it exists, the significance of the difference relative to the control (empty pEGFP vector alone; line 1) is indicated by *P < 0.05 (B–D) Influence of the overexpression of full-length Ki-1 ⁄ 57 (B) and its N-terminal (C) or C-terminal (D) constructs on E1A splice site selection Essentially the same results were obtained in HEK293 cells and when we used a flag-tagged construct of Ki-1 ⁄ 57 (data not shown) (E, F) Treatment ⁄ control band intensity ratios – comparison of the splicing site selection efficiency of Ki-1 ⁄ 57 and its N-terminal or C-terminal truncated forms The average

of band intensity values obtained for the isoforms 10S (E) or 9S (F) in (B), (C) and (D) in comparison to the average of the intensities in the control samples were plotted in the graphs, and represent the fold induction of each isoform in relation to the control We achieved approxi-mately 60% transfection efficiency in all experiments performed Open circles, unspliced pre-mRNA; M, marker.

Effect of SFRS9 on E1A pre-mRNA splicing in the

presence of Ki-1⁄ 57

SFRS9 and many other Ser⁄ Arg proteins (SR proteins)

are well known as regulators of E1A pre-mRNA

splic-ing [21,24] Seeksplic-ing for a possible functional influence

of Ki-1⁄ 57 on SFRS9 splicing activity, we performed

splicing assays in which both proteins were

coex-pressed in COS-7 cells

When we cotransfected the construct EGFP–SFRS9

alone with the pMTE1A vector, we observed a strong

inhibitory effect on the formation of the 12S and 10S

mRNAs (Fig 5, lanes 2 and 3), but, similarly to what

was found for EGFP–Ki-1⁄ 57, we also observed

stimu-latory activity in generating the 9S isoform (Fig 5,

lanes 2 and 3) This finding may suggest that although

both proteins may act together in selecting the most

distal splice site region that generates the 9S isoform,

they can also be involved in different regulatory

splic-ing mechanisms, as EGFP–Ki-1⁄ 57 has an opposite

stimulatory activity in generating the 10S isoform in

comparison with SFRS9, which is inhibitory

Interest-ingly, upon adding increasing amounts of EGFP–

Ki-1⁄ 57 we consistently observed that the inhibitory

effect of EGFP–SFRS9 in selecting the 10S isoform

can be partially reversed

Colocalization analysis of Ki-1⁄ 57 to nucleoli and splicing speckles

We have previously shown that, upon treatment with the inhibitor of methylation adenosine-2¢,3¢-dialdehyde (Adox), the endogenous Ki-1⁄ 57, instead of showing a uniform nuclear⁄ cytoplasmic distribution, relocalizes predominantly to the nucleus, where it appears as nuclear dots [13] As the results that we present here pointed to involvement of Ki-1⁄ 57 with RNA-binding proteins related to RNA⁄ mRNA processing, and as most of the so far characterized nuclear subdomains are sites for RNA maturation and processing [25], we decided to investigate, through confocal microscopy analysis, the identity of the nuclear substructures where Ki-1⁄ 57 is present in Adox-treated cells

We tested two lineages of adherent cells, COS7 and HEK293, and found, in both of them, that upon Adox treatment the recombinant EGFP–Ki-1⁄ 57 displayed similar nuclear relocalization, at several dots, as displayed by the endogenous Ki-1⁄ 57 in HeLa cells [13] (data not shown) We then decided to use the recombinant EGFP-fused form of Ki-1⁄ 57 in our con-focal analysis, mainly because of the insufficient qual-ity of the images obtained by labeling the endogenous Ki-1⁄ 57 with monoclonal antibodies We noticed that

Fig 5 Effect of Ki-1 ⁄ 57 on SFRS9 activity COS7 cells were transiently cotransfected with an E1A minigene-encoding plasmid [21,23], an empty EGFP vector, and increasing amounts of pEGFP vectors encoding full-lengths constructs for SFRS9 or Ki-1 ⁄ 57 (1·, 4 lg; 2·, 8 lg; 3·,

12 lg) The empty pEGFP vector was used to keep constant the DNA concentration in each transfection Splicing activity quantization was performed as described in Experimental procedures The displayed figures are representative of at least three independent experiments Ver-tical bars in the graphs indicate ± standard deviation Wherever it exists, the significance of the difference relative to the control (empty pEGFP vector alone; line 1) is indicated by *P < 0.05 Lanes 2 and 3 display the activity of SFRS9 alone, whereas lines 4–6 (darker gray bars) show the effect of the increasing amounts of Ki-1 ⁄ 57 The white triangle indicates that the value plotted for the 10S isoform in line 6 is dif-ferent (P < 0.05) to that in line 4 We achieved approximately 60% transfection efficiency in all experiments performed Open circles, unsp-liced pre-mRNA; M, marker The expression of cotransfected Ki-1 ⁄ 57 and SFRS9 was controlled by RT-PCR and is shown in Fig S1.

the most evident dot-forming Ki-1⁄ 57 in the nuclei of

Adox-treated cells seemed to be related to nucleoli,

mainly because of the well-known large area that this

structure occupies in the cell nucleus

Although EGFP–Ki-1⁄ 57 shows a diffuse

distribu-tion throughout the nucleus, it showed a stronger

signal that colocalizes with the staining of the nucleoli

marker nucleophosmin (Fig 6Aiii) in Adox-treated

cells This suggests that the methylation status of

Ki-1⁄ 57 is important for its colocalization to this

nuclear subcompartment

Besides the larger, nucleolar-associated bodies, we

also observed, in the Adox-treated cell nuclei, several

small dot-forming Ki-1⁄ 57 domains Owing to the

interaction of Ki-1⁄ 57 with proteins associated with

pre-mRNA splicing, a plausible hypothesis would be

that these regions corresponded to nuclear speckles,

which are nuclear substructures known to be enriched

in small nuclear ribonucleoprotein (snRNP) and many

other transcription-related and pre-mRNA

splicing-related proteins [25,26] To investigate this possibility,

we studied Ki-1⁄ 57’s colocalization with the SR

pro-tein SC-35, a marker propro-tein for splicing speckles [27]

Despite the diffuse distribution of Ki-1⁄ 57 in the

nucleus, we saw partial colocalization with the SC-35

dots in nontreated cells (Fig 6Bjjj) In Adox-treated

cells, we noticed, however, a juxtaposition of the

EGFP–Ki-1⁄ 57 and SC-35 nuclear substructures

(Fig 6Biii) This is an indication that the cellular

methylation status has specific effects on the

localiza-tion of EGFP–Ki-1⁄ 57 among different subnuclear compartments

Colocalization of EGFP–Ki-1⁄ 57 with Cajal and Gemini of coiled bodies (GEMS) nuclear bodies The partial colocalization of EGFP–Ki-1⁄ 57 with splicing speckles in untreated control cells has led us

to test antibodies against molecular marker proteins for Cajal bodies and GEMS (Gemini of coiled bodies), both of which are considered to be nuclear compart-ments involved in snRNP storage and⁄ or in the assem-bly of pre-mRNA splicing complexes [25]

Interestingly, we found, through confocal analyses, that EGFP–Ki-1⁄ 57 was again localized in a diffusive fashion throughout the nucleoplasm, but showed stronger spotted staining that colocalized with the Cajal body protein marker p80-coilin in the nucleus of HEK293 cells treated with Adox (Fig 7Ai–iii) This finding may, in addition, strengthen the hypothesis of the involvement of Ki-1⁄ 57 in pre-mRNA processing events

The GEMS are regions enriched with the survival of motor neurons (SMN) protein complexes and are con-sidered to be Cajal body-like domains [25] Although they may be found as distinct structures, they can also

be found colocalized This suggests that they are functionally related [28,29] However, an interesting particularity of the SMN protein is its demand for the presence of Arg⁄ Gly-rich regions in most of its

inter-Fig 6 Localization of Ki-1 ⁄ 57 to nucleoli

and splicing speckles HEK293 cells were

transfected with EGFP–Ki-1 ⁄ 57 and treated

or not treated with the methylation inhibitor

Adox After fixation, the cells were

immuno-stained with the antibodies against the

nuclear proteins nucleophosmin (NPM;

marker protein of nucleoli) or SC-35 (marker

protein of speckles), and analyzed by

laser-scanning confocal microscopy (A) Partial

colocalization of EGFP–Ki-1 ⁄ 57 to nucleoli

(nucleophosmin) in Adox-treated cells (Aiii),

but not in the control cells (Ajjj) (B) Partial

colocalization of EGFP–Ki-1 ⁄ 57 to speckles

(SC-35) only in the control cells (Bjjj) The

Adox treatment seems to cause only a

close juxtaposition between the speckles

and the Ki-1 ⁄ 57-associated substructure

(Biii, inset) Bars: 5 lm Figure insets

emphasize colocalizations or close

juxtaposi-tion of structures.

acting partners [30], suggesting the possible requirement

of Arg methylation for these protein–protein

associa-tions Through confocal analyses, we observed the

partial colocalization of Ki-1⁄ 57 with SMN ⁄ GEMS

in Adox-treated cells (Fig 7Biii) Despite the diffuse

nuclear distribution of EGFP–Ki-1⁄ 57, the majority

of the red GEMS spots coincide with spots of

brighter EGFP–Ki-1⁄ 57 staining (Fig 7Bi–ii) This

finding unveils a novel nuclear body with which

Ki-1⁄ 57 is associated in HEK293 cells treated with

Adox We, like others before us, observed also that

Cajal bodies colocalize with nucleoli in some cells [31]

(not shown)

Subcellular localization of Ki-1⁄ 57 truncated

forms

In order to obtain further clues on the functions of

different regions of the Ki-1⁄ 57 amino acid sequence

in human cells, we fused to EGFP the various

trun-cated forms used in the mapping studies described

before (Figs 2G, 3B and 4C–D) Through fluorescence

microscopy analysis of transfected HEK293 cells, we

observed that all tested C-terminal constructs showed

similar nuclear and cytoplasmic localization as

observed for full-length Ki-1⁄ 57 (Fig 8D–F) In turn,

the N-terminal construct displayed an exclusively

nuclear localization that, after careful analysis, could

be found in a few regions consisting of nuclear bodies

(Fig 8C) This may suggest that the targeting of

Ki-1⁄ 57 to nuclear subdomains requires its N-terminal

region On the other hand, when we treated the HEK293 cells with Adox, we observed a small but sig-nificant change in the localization of the C-terminal construct In the majority of analyzed cells Ki-1⁄ 57(122–413) was seen more predominantly in the nuclear compartment, in contrast to the diffusely nuclear⁄ cytoplasmic distribution observed in control cells (compare panels I and D in Fig 8) It is interest-ing to observe that, upon Adox treatment, the N-ter-minal construct showed pronounced relocalization from the nucleoplasm to several well-defined nuclear bodies (Fig 8H), as observed for the full-length Ki-1⁄ 57 construct (Fig 8G) More than 90% and 98%

of the cells transfected with full-length EGFP–Ki-1⁄ 57 (Figs 6 and 7) and with EGFP–Ki-1⁄ 57(1–222) (not shown), respectively, were found in the nucleus at nuclear substructures upon Adox treatment (Fig 8L) This suggests that the C-terminus of Ki-1⁄ 57 is not a required region for its association with these nuclear subdomains, and therefore suggests that some signal for localization control may exist at the N-terminus This localization may occur via protein–protein inter-actions, as suggested by our mapping results with the Ki-1⁄ 57-interacting proteins involved in pre-mRNA splicing (Fig 2G)

Discussion

We here describe Ki-1⁄ 57 as a novel human protein that is functionally related to regulatory events of pre-mRNA processing From a wider point of view, the

Fig 7 EGFP–Ki-1 ⁄ 57 is found in Cajal bodies and GEMS upon Adox treatment HEK293 cells transfected with EGFP– Ki-1 ⁄ 57 were treated or not treated with the inhibitor of methylation Adox The fixed cells were stained with antibodies against p80-coilin, which label Cajal bodies (A), or against SMN, a marker for GEMS (B), and analyzed by laser-scanning confocal microscopy Bars: 5 lm Figure insets focus on colocalizations or point out high-magnification images of the selected structures.

involvement of Ki-1⁄ 57 in modulating the splicing site

selection of the E1A pre-mRNA reported here may

unify its roles in the two main functional ‘worlds’ of

the nuclear context: RNA metabolism and

trans-criptional regulation It is well known that in

eukary-otes, transcription and pre-mRNA maturation events

(5¢-capping, splicing, 3¢-end processing and

polyadeny-lation) occur cotranscriptionally and that the

machin-eries responsible for these activities are functionally

and physically associated [31,32] Furthermore, it could

be speculated that not only the processing⁄ maturation,

but also the expression, of some subsets of mRNA

may be regulated by Ki-1⁄ 57 in a defined cellular

context Several of the identified Ki-1⁄ 57-interacting

proteins are involved in transcriptional control, such

as CHD3, RACK1, p53 and others p53-associated

proteins [8,9,11], thereby reinforcing a putative

tran-scriptional regulation role for Ki-1⁄ 57

Here, as well as in a previous study, Ki-1⁄ 57 has

been observed as dot-like structures in the cell nucleus

[2,13] Growing evidence points to important roles of

these nuclear subdomains, not only as storage spaces

but also as dynamic structures involved in RNA

tran-scription, processing, and maturation [25,27,33,34] We

further showed the importance of methylation for its localization at distinct nuclear ‘spots’, and showed that

in control cells, but not in Adox-treated cells, EGFP– Ki-1⁄ 57 partially localizes to nuclear speckles, whereas

in Adox-treated cells it partially localizes to GEMS, Cajal bodies, and nucleoli The nuclear speckles are known to be storage places for pre-mRNA splicing complexes and, in turn, Cajal bodies and GEMS are regions involved in snRNA modification, snRNP bio-genesis, and trafficking of small nucleolar RNPs (small nucleolar RNPs)⁄ snRNPs to nucleoli or speckles, respectively [25] Therefore, the methylation process may be an important step in the migration of Ki-1⁄ 57

in ‘assembled’ RNP complexes at Cajal⁄ GEMS bodies

to nuclear speckles, from where they could be recruited

to transcriptionally active sites, which are regions that are likely to involve synchronous splicing activity Interestingly, the methylation effects on snRNP assembly in both cytoplasmic and nuclear phases have been demonstrated in other studies Gonsalvez et al [35] have shown that in cells treated with the methyla-tion inhibitor 5¢-deoxy-5¢-(methylthio)adenosine, the process of snRNP assembly in the cytoplasmic compart-ment is disrupted Similarly, the Arg methylation of the

100

50

0

Nuclear - with dots

Nuclear/Cytopl without dots

n = 3

Ki-1/57 Ki-1/57(1–222)

Ki-1/57(1–413)

B

A

L

Ki-1/57(1–222) Ki-1/57(122–413) Ki-1/57(151–263) Ki-1/57(264–413)

Fig 8 The localization of Ki-1 ⁄ 57 to nuclear

bodies depends on its N-terminal region (A)

Schematic representations of the truncated

Ki-1 ⁄ 57 constructs fused to EGFP, used for

the localization assays in untreated or

Adox-treated HEK293 cells (B–F) UnAdox-treated cells.

The full-length EGFP–Ki-1 ⁄ 57 and its

C-ter-minal constructs (122–413), (151–263) and

(264–413) show a diffuse nuclear and

cyto-plasmic localization (B, D–F), whereas the

N-terminal construct (1–222) shows an

exclusively nuclear localization (C), at

dis-crete nuclear dots (white arrowheads) (G–L)

Adox-treated cells The full-length

EGFP–Ki-1 ⁄ 57 and its N-terminal construct (1–222)

predominantly show nuclear localization (G,

H), at several nuclear bodies (white

arrow-heads) A small but significant amount of

nuclear relocalization can be observed for

the C-terminal construct (122–413) (I) No

changes were observed for the smaller

C-terminal constructs (151–263) and (264–413)

(J, K) (L) Proportion of Adox-treated cells

containing nuclear bodies More than 90%

and 98% of the cells transfected with

full-length EGFP–Ki-1 ⁄ 57 and EGFP–Ki-1 ⁄ 57

(1–222), respectively, were found in the

nucleus at several dots Approximately 100

cells were analyzed in each of three

inde-pendent experiments.