Tài liệu Báo cáo khoa học: Structural bases for recognition of Anp32⁄LANP proteins doc

As a prerequisite, we solved the structure in solution of the evolutionarily conserved N-terminal leucine-rich repeat LRR domain and modeled its interactions with other proteins, taking

Cesira de Chiara, Rajesh P Menon and Annalisa Pastore

National Institute for Medical Research, The Ridgeway, London, UK

The leucine-rich repeat acidic nuclear protein (Anp32a⁄

LANP) is a member of the Anp32 family of acidic

nuclear evolutionarily-conserved phosphoproteins,

which present a broad range of activities [1] They are

characterized by the presence of a highly conserved

N-terminal domain containing leucine-rich repeats

(LRRs), motifs known to mediate protein–protein

inter-actions, and of a C-terminal low-complexity region,

mainly composed of polyglutamates

Since their first description in the neoplastic

B-lym-phoblastoid cell line and their reported association

with proliferation [2], several Anp32a homologs, all

derived from a common ancestor gene by subsequent

duplication events, have been isolated in different

tis-sues and differently named [1] Members of the Anp32

family are widely recognized as nucleo-cytoplasmic

shuttling phosphoproteins that are implicated in

differ-ent signaling pathways and in a number of important

cellular processes, which include cell proliferation, dif-ferentiation, caspase-dependent and caspase-indepen-dent apoptosis, tumor suppression, regulation of mRNA trafficking and stability, histone acetyltransfer-ase inhibition, and regulation of microtubule-bacetyltransfer-ased functions [1,3]

The diverse activities of Anp32 proteins are achieved through an articulated network of interactions with several cellular partners Among them, two proteins are of particular relevance from the clinical point of view Anp32 proteins are powerful inhibitors of phos-phatase 2A (PP2A), a major serine⁄ threonine phospha-tase involved in many essential aspects of cellular function [4–8] PP2A, which is considered to be the principal guardian against cancerogenic transforma-tion, is a dynamic, structurally diverse molecule found

in several different complexes and able to react to a plethora of signals [9–12] The N-terminal LRR

Keywords

ataxin 1; leucine-rich repeats; NMR; PP2A

inhibitor; structure

Correspondence

A Pastore, National Institute for Medical

Research, The Ridgeway, London NW7

1AA, UK

Fax: +44 208 906 4477

Tel: +44 208 959 3666

E-mail: apastor@nimr.mrc.ac.uk

(Received 13 January 2008, revised 3 March

2008, accepted 14 March 2008)

doi:10.1111/j.1742-4658.2008.06403.x

The leucine-rich repeat acidic nuclear protein (Anp32a⁄ LANP) belongs to

a family of evolutionarily-conserved phosphoproteins involved in a com-plex network of protein–protein interactions In an effort to understand the cellular role, we have investigated the mode of interaction of Anp32a with its partners As a prerequisite, we solved the structure in solution of the evolutionarily conserved N-terminal leucine-rich repeat (LRR) domain and modeled its interactions with other proteins, taking PP2A as a paradig-matic example The interaction between the Anp32a LRR domain and the AXH domain of ataxin-1 was probed experimentally The two isolated and unmodified domains bind with very weak (millimolar) affinity, thus sug-gesting the necessity either for an additional partner (e.g other regions of either or both proteins or a third molecule) or for a post-translational modification Finally, we identified by two-hybrid screening a new partner

of the LRR domain, i.e the microtubule plus-end tracking protein Clip 170⁄ Restin, known to regulate the dynamic properties of microtubules and to be associated with severe human pathologies

Abbreviations

Anp32a ⁄ LANP, leucine-rich repeat acidic nuclear protein; Atx1, ataxin-1; AXH, ataxin-1 homology; Gal-X, 5-bromo-4-chloroindol-3-yl

b- D -galactoside; GST, glutathione S-transferase; HSQC, heteronuclear single quantum coherence; LRR, leucine-rich repeat; MAP,

microtubule-associated protein; PP2A, phosphatase 2A; RDC, residual dipolar coupling; SCA1, spinocerebellar ataxia 1; TCEP,

Tris(2-carboxyethyl)phosphine; +TIP, plus-end tracking proteins.

domain of Anp32 binds and strongly inhibits the

enzyme catalytic subunit PP2A-C [3–7], whose

struc-ture in a heterotrimeric complex with the scaffolding

A subunit and the regulatory B¢ ⁄ B56 ⁄ PR61 subunit

was solved recently [13,14] Although the role of

phos-phorylation in recognition remains debatable,

interac-tion between Anp32a and PP2A-C has been

independently confirmed by high-throughput yeast

two-hybrid screening [15]

Involvement of Anp32 in spinocerebellar ataxia

type 1 (SCA1) pathogenesis was also suggested, on the

basis of the observation of an interaction with the

SCA1 gene product ataxin-1 (Atx1) [16] This protein

belongs to a family involved in neurodegenerative

dis-eases caused by anomalous expansion of polyQ tracts

[17] In SCA1, expanded Atx1 forms nuclear inclusions

that are associated with cell death

Immunofluores-cence studies demonstrated that Anp32a and Atx1

colocalize in nuclear matrix-associated subnuclear

structures The interaction was mapped onto the LRR

and AXH domain of Anp32 and Atx1 respectively,

and was shown to be stronger for expanded Atx1

[16,18] The temporal and cell-specific expression

pattern of Anp32 in cerebellar Purkinje cells, the

primary site of pathology in SCA1, as well as its

enhanced interaction with mutant Atx1, have

sug-gested a role for Anp32a in SCA1 pathogenesis

Despite the importance of molecular interactions

for the presumed cellular functions of Anp32a, very

little is known about their structural bases The

struc-ture of the Anp32 LRR domain was first predicted

by homology [1] and more recently solved by X-ray

crystallography [19] Interestingly, although both

reports described the domain as being formed by

tan-dem LRRs, the structures differed in the number of

repeats This information is not just academic, as

these details would allow accurate definition of the

domain boundaries and help our understanding of

how interactions could take place in different regions

of the molecule

As part of a long-term function-oriented structural

effort aimed at understanding the molecular bases of

polyQ disease proteins, we report here a study of the

structural determinants for the interactions of Anp32a

with other partners As a prerequisite for binding

stud-ies, we first solved the structure in solution of Anp32

by NMR spectroscopy This technique, which does not

need crystallization, also provides a powerful and

flexi-ble method for mapping binding interfaces Our

struc-ture, as described in the following sections, reveals the

presence of two extra N-terminal LRR motifs not

observed in the crystal, and allows accurate definition

of the C-terminal domain boundary Experimental

determination of the dynamic features of the domain

in solution, together with a comparison with the struc-ture of the spliceosomal U2A¢ in complex with U2B¢¢ [20], suggests new insights into the mechanism of Anp32 LRR–protein recognition By a combination of chemical shift perturbation techniques, molecular docking and two-hybrid screening, we also probed the interaction with Atx1 and PP2A, and identified a new partner of the Anp32a LRR domain

Results

Description of the Anp32a LRR domain structure

in solution The construct used for structure determination covers residues 1–164 of the mouse Anp32a sequence [21] These boundaries were chosen to include the sequence

up to the beginning of the acidic repeats, where sequence conservation breaks down (data not shown) The resulting sample was stable and well behaved, pro-viding NMR spectra typical of a folded monodisperse globular domain The final representative family of the

10 lowest-energy structures after water refinement could be superimposed on the average structure with overall rmsd values of 0.71 ± 0.15 A˚ and 1.16 ± 0.16 A˚, for backbone and heavy atoms respec-tively, in region 3–154 (Fig 1) The structure was solved at high precision and has an excellent whatif score (Table 1)

The domain topology (h1h2b1b2b3h3b4h4b5h5h6b6

b7b8) shows the secondary structure elements spatially arranged in the typical right-handed solenoid, which forms a curved horse-shoe fold A canonical parallel b-sheet is present on the concave side, whereas the convex surface contains both well-defined but irregular secondary structure elements (in the first and second repeats) and helical regions Among these, h1 and h6 are regular a-helices whereas h2, h3, h4 and h5 share features of 310-helices A search for tertiary structure similarity performed by dali [22] indicates that the Anp32a LRR domain structure belongs to the SDS22-like LRR subfamily [23]

Comparison with other Anp32 structures The Anp32a LRR domain is composed of five com-plete LRRs flanked by an a-helix at the N-terminus and by the C-terminal flanking motif termed LRRcap (SMART accession number SM00446), so far identi-fied in several ‘SDS22-like and typical’ LRR-contain-ing proteins, such as U2A¢, TAP, RabGGT, and dynein LC1 [23] The Anp32a LRRcap motif spans

residues Leu128 to Asp146, and includes h6, which

belongs to the fifth LRR, and the short strand b6,

which runs parallel to b5 and is antiparallel to b7 The

solution and the crystal structures of the Anp32a LRR

domain superimpose with a 1.1 A˚ rmsd over the

back-bone atoms of the overlapping region 1–149

(Fig 2A,B) Despite the structural similarity, only four

repeats (44–65, 66–89, 90–114 and 115–138 in our

structure) were identified in the crystal structure [19],

whereas the first repeat (residues 19–43 in our

struc-ture) was considered by these authors to be an

N-CAP The presence of the first N-terminal LRR had

also not been predicted [1], probably because of the

low sequence conservation in this region In our

opin-ion, this region constitutes instead a bona fide full

repeat

Residues 147–149, which are truncated in the crystal

structure, form a b-hairpin (b7) with the strand 143–

145 (b6) This region shares a remarkable similarity

with U2A¢, the two protein with the highest structural

homology: the two proteins can be superimposed with

2 A˚ rmsd as calculated over 140 residues and have a

dali z score of 17 (Fig 2A,C) Although mainly

unstructured, a short additional strand C-terminal to

the hairpin (b8) is present in some of the NMR

struc-tures in region 149–164 (residues 151–153) This region,

which constitutes the linker between the LRR domain

and the acidic repeats, is thought to be involved in

interactions with the INHAT complex and the

phos-phorylation-dependent tumor suppressive⁄ proapoptotic

activity, which have been mapped to residues 150–180

and 150–174, respectively [15,24,25] Interestingly, this

tail contains one of the two CK2 phosphorylation

motifs (158–161) that have been proved to be natively

phosphorylated [26], out of the four putative sites

predicted by prosite [27] As expected, the structure of this region is flexible and is likely to be involved in the regulation of phosphorylation-dependent functions of Anp32a [1,3]

Probing the dynamics in solution of the Anp32a LRR domain

1H–15N relaxation studies were carried out to assess the dynamic properties of the Anp32a LRR domain (Fig 3) The correlation time, as estimated from the T1 and T2 relaxation data using the model-free approach [28], is 12.5 ± 0.1 ns at 27C, a value within the range expected for a single monomeric species of this size in solution [29] The flat profile of the relaxation parame-ters along the sequence indicates that, with the excep-tion of the first two N-terminal amino acids and of the C-terminal tail (Ala155–Val164), the structure is rigid and compact, which is in good agreement with what is observed from the local rmsd values and the residual dipolar coupling (RDC) values Of the seven resi-dues whose amide connectivities are missing in the

1H–15N heteronuclear single quantum coherence (HSQC) spectrum at pH 7 (Met3, Asp4, Ile30, Glu31, Ile34, Glu35 and Val52) the last five belong to regions without a regular secondary structure All seven resi-dues, including Met3 and Asp4 at the N-terminus of

h1, cluster together in the structure, suggesting that they experience chemical or conformational exchange The C-terminus is unstructured and highly mobile approximately from residue 154 onwards

It is also interesting to note a clear correlation between T1 and RDC values and the secondary struc-ture: the concave b-sheet is characterized by shorter

T1 and positive RDC values, the latter indicating that

A

B

Fig 1 Solution structure of the LRR domain of murine Anp32a (A) NMR bundle

of the 10 best structures in terms of energy (B) Average structure as obtained

by the WHEATSHEAF algorithm [62] Two orthogonal views are shown.

the corresponding residues are oriented parallel to the

external magnetic field B0 when in the anisotropic

medium [30] Conversely, the NH vectors in the long

helices (h1, h3, h4, and h6) running approximately

parallel to each other are mainly perpendicular to the

b-sheet vectors and, therefore, to B0 in the aligned

medium

These results suggest that interactions with other

molecules involving the LRR domain are not

medi-ated by an induced fit mechanism but by semirigid

docking of the partners onto the surface of the

Anp32a LRR domain Interactions with the INHAT

complex, mapped to the flexible C-terminus, may

induce structuring and stiffening of this region

Modeling the interaction of Anp32a with other proteins on the basis of the U2A¢–U2B¢¢ structure The structural similarity with U2A¢, whose structure is known in a complex with its target U2B¢¢ [20], may provide valuable hints on how Anp32 interacts with its partners We therefore analyzed this complex and compared its features with those of our structure Rec-ognition of the two molecules occurs by fitting a helix

of U2B¢¢ (residues 25–35) into the concave surface of the U2A¢ LRR (Fig 4) The size complementarity is almost perfect The nearby N-terminal b-hairpin of

Table 1 Structural statistics for the calculations of the Anp32a

LRR domain.

Final NMR restraints

Medium (residue i to i + j, j = 1–4) 663

Long-range (residue i to i + j, j > 4) 1392

Dihedral angle restraints b

1

Deviation from idealized geometry

Restraint violations

Distance restraint violation > 0.5 A ˚ 0

Dihedral restraint violation > 5 0

Coordinate precision (A ˚ ) with respect to the mean structure

Backbone of structured regions c 0.71 ± 0.15

Heavy atoms of structured regions c 1.16 ± 0.16

WHATIF quality check d

First-generation packing quality 0.01

Second-generation packing quality )2.44

Procheck Ramachandran statistics (%)

a Calculated for the 10 lowest-energy structures after water

refine-ment.bDerived from3J(HN, Ha) coupling constants and TALOS [48].

c Calculated for residues 3–154 The more positive the score, the

better it is Problematic structures typically have scores around )3.

Wrong structures have scores lower than )3 d

Calculated for resi-dues 1–154.

A

B

C

Fig 2 Comparison between the NMR (A) and the X-ray (B) struc-tures of the Anp32a LRR domain, and U2A¢ (C) [20] The coordi-nates were first superimposed using the DALI server, and then displaced.

U2A¢ (residues 13–26) provides further interactions by wrapping around the other molecule on one side There is also a good charge complementarity, as the concave surface of the U2A¢ LRR is negatively charged, whereas the U2B¢¢ helix, which is neutral overall, contains at least one positively charged residue (Arg28), which protrudes out into the solvent and is

Fig 3 Relaxation parameters and RDC values along the sequence

of the Anp32a LRR domain The data were recorded at 27 C and

800 MHz.

A

B

Fig 4 Modeling the interactions of the Anp32a LRR domain (A) Structure of the Anp32a LRR domain in a complex with the C sub-unit of PP2A as modeled by comparison with the U2A¢–U2B¢¢ complex The other two subunits shield most of the surface of PP2A–C (B) Structure of the U2A¢–U2B¢¢ complex [20].

able to form a salt bridge with Glu92 of U2A¢ Other

contacts will contribute with hydrophobic interactions

or intermolecular hydrogen bonds, which are likely to

be responsible for the specificity of recognition, which

seems to be tuned to the specific system Accordingly,

the U1A protein, which is closely related to U2B¢¢,

does not form a stable complex with the U2A¢ LRR,

whereas replacement of Asp24 and Lys28 with the

homologous Glu and Arg of U2B¢¢ [20] is sufficient to

re-establish formation of the complex [31]

We modeled, as a paradigmatic and particularly

inter-esting example, a complex between PP2A and the

Anp32a LRR The structure of PP2A has recently been

determined [13,14] It consists of a heterotrimeric

com-plex formed by the scaffolding subunit A, the regulatory

subunit B¢ ⁄ B56 ⁄ PR61, and the catalytic domain C

Interaction with Anp32 has been shown to involve the

catalytic subunit [15,32] and to inhibit its catalytic

activity, both in the absence and in the presence of the

scaffold subunit A and the regulatory subunit B, with

apparent Kiin the low nanomolar range [4] This implies

that the interaction involves an exposed region of

PP2A-C, without appreciable contributions from the

other two subunits Anp32 is also known to inhibit

PP2A in a noncompetitive manner, i.e without binding

to the active site of the enzyme [4] Finally, antibodies

recognizing the fourth LRR of Anp32e⁄ Cpd1

(resi-dues 87–101) are known to block the inhibitory PP2A

activity of Anp32e in protein extracts [7] Taken

together, these findings limit the region of interaction to

the only exposed surface of PP2A-C that contains a

semiexposed helix (residues 222–232)

The model of an Anp32 LRR–PP2A complex, built

using complex U2A¢–U2B¢¢ as a template, shows that,

by analogy with this structure, helix 222-232 of

PP2A-C protrudes out enough to fit well into the groove

formed by the concave surface of the LRR domain

Stabilizing interactions could form between His230 of

PP2A-C and Asp119 and Asn94 of Anp32a A salt

bridge could form between Glu226 of PP2A-C and

Lys67, Lys68 and Lys91

Testing the interaction with Atx1 experimentally

Interaction between the Anp32a LRR domain and the

Atx1 AXH domain was tested experimentally by

NMR chemical shift perturbation, with the aim of

mapping the surface of interaction between the two

proteins This method, which relies on the effect that

binding of a molecule has on the electron distribution

of another, causing a perturbation of its NMR

spec-trum, is routinely used to detect interactions and map

them on the structures of the individual components

We titrated the LRR domain with the AXH domain since this region had been proposed to be essential for the interaction on the basis of deletion mutants [18] When the effects were mapped onto the Anp32 surface (Fig 5), they all clustered around the concave surface However, even at high Atx1 AXH⁄ Anp32a LRR ratios (3.5 : 1 and low ionic strength), we observed only min-imal perturbations of the Anp32 LRR domain spec-trum (i.e.: <0.05 ppm in the proton dimension), which were absent in spectra recorded at a higher ionic strength (150 mm NaCl) Likewise, when we titrated the Atx1 AXH domain with the Anp32a LRR domain,

we observed only two very small effects

The interaction was independently probed by fluo-rescence spectroscopy, exploiting the intrinsic emission

at 327 nm of the only Trp residue present in Atx1 (Trp658; Anp32a does not contain Trp residues) after sample excitation at 295 nm During titration, fluores-cence quenching was observed along with a 4 nm blue shift of the kmax of emission (from 327 to 323 nm), suggesting a decrease in the Trp solvent exposure con-sequent to interaction (data not shown) However, the decrease in fluorescence intensity was far from reach-ing a plateau even at the highest Anp32a⁄ Atx1 ratio tested (60 : 1)

Fig 5 Probing the interaction between the Anp32a LRR domain and the AXH domain of Atx1 by chemical shift perturbation Super-imposition of the HSQC spectra of a 0.2 m M solution of 15 N-labeled Anp32a LRR domain in 20 m M Tris (pH 7.0) and 2 m M TCEP, recorded at 600 MHz and 27 C in the absence (blue) and in the presence (red) of a three-fold excess of unlabeled Atx1 AXH domain.

This evidence indicates that interaction between the

two domains is very weak, i.e with binding constants

in the millimolar range Although such binding is

defi-nitely too weak to be significant, it is certainly possible

that, in vivo, the interaction is enhanced either by other

regions of the two molecules or by post-translational

modifications that are absent in our assays

Identification of new potential partners of the

Anp32 LRR domain

To identify new partners specific for the Anp32a LRR

domain, we used a construct spanning the same region

studied by structural techniques (residues 1–164) as a

bait in a two-hybrid screening assay This is at

vari-ance with previous studies, which were all carried out

on the full-length protein, thus inferring the role of the

LRR domain only indirectly By screening of a human

brain cDNA library (Clonetech, Mountain View, CA,

USA) for a total of approximately 5 million clones, we

found about 600 potential positives [i.e hits that were

positive both for quadruple-dropout media and for

5-bromo-4-chloroindol-3-yl b-d-galactoside (Gal-X)

overlay assays] Nearly 200 of these positives were

sequenced Among these, we identified 29 clones of the

C-terminus of the microtubule-associated protein

(MAP) Clip 170⁄ Restin, a microtubule plus-end

track-ing protein (+TIP), which associates with and

regu-lates the dynamic properties of microtubules and of

other MAPs [33] (Fig 6)

We tested the interaction further by expressing the

full-length proteins in mammalian cells In transfected

COS cells, Anp32a was predominantly nuclear, with a

limited number of cells showing extranuclear staining

(Fig 7A) In contrast, and as expected, Clip 170 was

excluded from the nucleus and localized to the

micro-tubule network Partial colocalization of Clip 170 and Anp32a was observed in the microtubules of cells coexpressing these proteins and showing extranuclear staining of Anp32a (Fig 7A, merged image)

To further validate the interaction, we carried out coimmunoprecipitation experiments to test the ability

of the endogenous proteins to associate HeLa cell lysates were immunoprecipitated with antibodies to Anp32a or with antibodies to histone H3 as a negative control The proteins from immunoprecipitation com-plexes were subjected to western blot analysis using antibodies to Clip 170 Clip 170 was associated with the complex pulled down by antibodies to Anp32a but not with the one pulled down by antibodies to histone H3 (Fig 7B,C)

Discussion

Here, we have explored the interaction properties of the LRR domain of Anp32, a family of LRR proteins potentially implicated in several important cellular pathways Two particularly interesting interactions have been described, with the PP2A phosphatase and with Atx1, two proteins of high medical importance

We first determined the domain boundaries of the domain by solving the solution structure at high reso-lution of a fragment spanning the whole conserved region up to some highly acidic repeats containing EA-EEE motifs We show that the domain contains a compact and rigid fold with five LRRs and a C-capping motif The structural information was used

to model the interaction with PP2A, which is known

to be mainly mediated by the PP2A-C subunit We suggest that, by analogy with the mode of recognition

of U2B¢¢ by U2A¢, which has the highest structural simi-larity with the Anp32a LRR domain, the interaction

-A

B

BD-Lanp.NT

BD-Lanp.NT

AD-Clip.CT

AD-Clip.CT AD-Clip.NT

Bait Prey Growth on QD plates X-Gal overlay

EMKKRESKFIKDADEEKASLQKSISITSALLTEKDAELEKLRNEVTVLRGENASAKSLHSVVQTLESDK

VKLELKVKNLELQLKENKRQLSSSSGNTDTQADEDERAQESQIDFLNSVIVDLQRKNQDLKMKVEM

MSEAALNGNGDDLNNYDSDDQEKQSKKKPRLFCDICDCFDLHDTEDCPTQAQMSEDPPHSTHHGS

RGEERPYCEICEMFGHWATNCNDDETF

Fig 6 Interaction of Clip 170 with Anp32a

in a yeast two-hybrid system (A) The N-ter-minus of Anp32a fused to the Gal4 DNA-binding domain (BD-Lanp.NT) interacts with the C-terminus of Clip 170 (AD-Clip.CT) fused to the Gal4 DNA activation domain as indicated by growth on quadruple-dropout (QD) plates and Gal-X overlay assays There was no growth on QD plates when either the N-terminus of Clip 170 (residues 1–

1164, BD-Clip.NT) or the Gal4 DNA-binding domain was used as prey (B) Amino acid sequence of the region of Clip 170 interact-ing with Anp32a.

involves the helix-spanning residues 222–230 of

PP2A-C [13,14] This region is the only element of PP2A-PP2A-C

protruding out from the PP2A trimer, and its size and

shape mean that it could easily fit into the

complemen-tary concave surface of the Anp32a LRR domain

We tested binding to the AXH domain of Atx1

experimentally by chemical shift perturbation assays

We observed only very minor effects, which are

com-patible, at the very best, with millimolar affinities

The effects could be observed only at low ionic

strength, suggesting that the interaction is mainly of

an electrostatic nature and is nonspecific Would our

results shed doubts on an interaction originally

observed by two-hybrid screening? On the one hand,

it is interesting to note that none of the

high-through-put studies of the Atx1 interactome has reported any

evidence for this interaction [34,35] On the other

hand, however, very recent data provide the first

evi-dence of a functional link between Anp32a and Atx1,

showing that Atx1 relieves the transcriptional

repres-sion induced by Anp32a in complex with E4F [36]

As addition of exogenous Anp32a restores repression,

it was suggested that Atx1 sequesters Anp32a,

releas-ing its interaction with E4F Our evidence may

there-fore indicate that either a third component (which

could be another region of one or both proteins or another molecule) or a post-translational modification

is needed to give appreciable affinities The second possibility seems currently most likely: both Atx1 and Anp32a are known to be natively phosphorylated, and phosphorylation has been shown to modulate some of their functions [1,3,37] Atx1 contains two phosphorylation sites, both outside the AXH domain, one of which (Ser776) is located in the C-terminal region of the protein and is known to modulate the interaction with 14-3-3 [37,38] Two in vivo CK2 phosphorylation sites (Ser158 and Ser204) have also been identified in Anp32a [26] Phosphorylation of Ser158, which is immediately downstream of the LRR domain, could, for instance, induce a conforma-tional change of the adjacent region, which could be required for Atx1 binding

Finally, we used yeast two-hybrid screening to iden-tify new partners of Anp32a To our knowledge, ours

is the first study carried out using, for library screen-ing, the LRR domain only, i.e excluding the acidic C-terminus, which, being highly charged, could pro-duce false positives We observed an interaction between the Anp32a LRR domain and the microtubule +TIP Clip 170 This protein is known to associate

A

Fig 7 Anp32a and Clip 170 associate with each other in HeLa and transfected COS cells (A) Colocalization of Clip 170 and Anp32a in COS cells that were transfected with a plasmid vector carrying V5-tagged Clip 170 and c-Myc-tagged Anp32a Cells were analyzed by confocal microscopy Clip 170 was localized in the microtubule network (green), and Anp32a (red) was predominantly nuclear, with some cells show-ing localization in the microtubules The merged image shows colocalization of the proteins in the microtubules (B) Expression of endoge-nous proteins in HeLa cells HeLa cells were lysed in RIPA buffer, and input controls and immunoprecipitated samples were probed with the antibodies shown (C) Interaction of endogenous Clip 170 and Anp32a in HeLa cells HeLa cells were lysed in RIPA buffer and immuno-precipitated as above with antibodies to histone H3 or antibodies to Anp32a Proteins were transferred onto a poly(vinylidene difluoride) membrane and probed with antibodies to Clip 170.

with microtubules and with other MAPs, and to

regu-late the dynamic properties of microtubules [33]

Iden-tification of this new potential partner is particularly

interesting, because Anp32a has already been reported

to be involved in microtubule dynamics via its

interac-tion with several members of the family of MAPs, i.e

MAP1B, MAP2, and MAP4 [39–41] The interaction

with MAP1B was suggested to modulate the effects of

MAP1B in neurite extension [41] Microtubule +TIPs

have also been shown to be involved in modulating

neuronal growth cones, the motile tips of growing

axons [42,43] Interaction of Clip 170 with

micro-tubules has been suggested to be influenced by

phosphorylation, as phosphorylation by a

rampamy-cin-sensitive kinase (fluorescence recovery after

photo-bleaching; FRAP) increases the interaction of Clip 170

with microtubules [44] Interestingly, in our

coimmu-noprecipitation experiments, the Clip 170 band

appeared to be more intense when the cell lysates

incorporated a cocktail of phosphatase inhibitors,

sug-gesting that the association may be modulated by

phosphorylation events (not shown)

Like Anp32, which is linked to the SCA1 pathology

[16], Clip 170 is also known to be associated with

human disease The protein is overexpressed in

Hodg-kin’s disease and anaplastic large cell lymphoma

[45,46] Clip 170 has also been shown to interact with

the Lis1 protein, whose mutation causes type I

lissen-cephaly, a severe brain developmental disease [47]

Therefore, our results point out to an important role

of Anp32 proteins in human pathologies and

encour-age further studies to clarify the complete interactome

of this protein

Experimental procedures

Protein sample preparation

The LRR domain of Anp32a from Mus musculus

(resi-dues 1–164) was produced using an ampicillin-resistant

glu-tathione S-transferase (GST)-3C expression vector with a

human rhinovirus 3C protease recognition site This

con-struct resulted in the addition of five non-native residues

(GPLGS) at the N-terminus of the protein Isotopically

15

N-labeled and 13C⁄15

N-labeled samples were overexpres-sed in the Escherichia coli host strain BL21 (DE3) grown

on a minimal medium containing [15N]ammonium sulfate

and [13C]glucose as the sole sources of nitrogen and carbon

respectively The cells were grown at 37C until an

attenu-ance (D) at 600 nm of 0.5 was reached, and then cooled

to 18C, induced with isopropyl thio-b-d-galactoside

(0.5 mm), and harvested after overnight expression A

stan-dard purification protocol was performed, using Pharmacia

GST–Sepharose resin (GE Healthcare) Cleavage of the GST tag was achieved overnight at room temperature using the PreScission protease (GE Healthcare) The protein was further purified by HPLC size exclusion chromatography, using a prepacked HiLoad 16⁄ 60 Superdex 75 prep grade column (Pharmacia) The concentration of the NMR sam-ple used for structural studies was typically in the range 0.3–0.7 mm, in a buffer containing 10 mm Tris⁄ HCl and

2 mm Tris(2-carboxyethyl)phosphine (TCEP) at pH 7.0 in 90% H2O⁄ 10% D2O All the NMR experiments were per-formed at 27C on Bruker Advance and Varian Inova spectrometers, both equipped with cryoprobes and operat-ing at 14.1 and 18.8 T, respectively, and on a Varian Inova spectrometer operating at 14.1 T Samples of the Atx1 AXH domain (residues 567–689 and 567–694) were pro-duced as previously described [18]

Experimental restraints Resonance assignment of the LRR domain was performed

as previously described [21] Interproton distance restraints were derived from NOESY 15N HSQC and NOESY

13C HSQC spectra acquired at 27C with mixing times of

100 ms on a Varian Inova spectrometer operating at

800 MHz 1H frequency A set of 89 backbone / and u dihedral angles was obtained using the backbone torsion angle prediction package talos [48] Amide protection was inferred from deuterium exchange measurements per-formed at 27C on a freeze-dried 15N-labeled sample redissolved in a Tris⁄ HCl-buffered (pH 7.0) D2O solution and started immediately after redissolving the protein The intensity decay of the NH signals extracted from a series

of 40 1H–15N HSQCs of 35 min each allowed calculation

of the exchange rates Twenty slowly exchanging protons were identified as having an exchange time longer than

3 h Among these, a hydrogen bond restraint was added if

a hydrogen bond was consistently observed in at least 50% of the structures inspected at an advanced stage of the refinement 1DNH RDCs were measured at 27C, aligning the protein in 5% n-dodecyl-penta(ethylene gly-col)⁄ n-hexanol (r = 0.92) using a buffer composed of

20 mm Tris⁄ HCl, 2 mm TCEP and 0.02% NaN3 at

pH 7.0 The liquid crystalline medium gave a stable quad-rupolar splitting of the D2O signal of 21 Hz The final concentration of the protein in this medium was

 0.37 mm 92 1JNH splittings were obtained from a J-modulated 15N–1H HSQC spectrum [49] for NH vectors with a heteronuclear NOE value higher than 0.75 and used for the purpose of structure validation using the program module [50] The rmsd in hertz from RDC restraints (observed – calculated from structure generated without using RDCs) is 0.620 ± 0.035

T1, T2 and heteronuclear NOE measurements were per-formed at 27C and 800 MHz, using adapted standard

pulse sequences The T1⁄ T2ratios of residues not

undergo-ing large amplitude motions or exchange were used to

esti-mate the correlation time (sc), assuming the model-free

approach [28] Residues with T1and T2values that differed

by more than one standard deviation from the mean were

excluded from the sccalculation

Structure calculation for Anp32a

Structure calculations were performed using the aria

pro-gram (version 1.2) [51] A typical run consisted of nine

iter-ations At each iteration, 20 structures were calculated by

simulated annealing using the standard cns protocol [52]

with numbers of steps equal to 15 000 and 12 000 in the

first and second cooling stages of the annealing,

respec-tively Floating assignment for prochiral groups and

correc-tion for spin diffusion during iterative NOE assignment

were applied as previously described [53,54] At the end of

each iteration, the best seven structures in terms of lowest

global energy were selected and used for assignment of

additional NOEs during the following iteration In the final

ariarun, the number of structures generated in iteration 8

was increased to 100, and after refinement by molecular

dynamics simulation in water of the 50 lowest-energy

struc-tures [55], the 10 lowest-energy strucstruc-tures were selected as

representative of the Anp32a LRR domain structure and

used for statistical analysis In the final iteration, 3774

unambiguous and 1377 ambiguous NOEs were assigned

Among the 5151 total NOEs, 2021 were intraresidue, 1075

sequential, 663 medium range, and 1392 long range

Struc-ture quality was evaluated using the programs procheck

[56] and whatif [57] The coordinates are deposited with

the Protein Data Bank (accession code 2jqd)

Comparative modeling

The structure of an Anp32a–PP2A complex was modeled

on the U2A¢–U2B¢¢ coordinates (1a9n) [20] The available

information strongly indicates that the interaction is

domi-nated by the C subunit of PP2A Of this, the main region

that protrudes out into solution and is not protected by

interactions with the other two subunits comprises

helix 222–232 Assuming a similar modality of interaction,

we superimposed this region on helix 1 of U2B¢¢

(resi-dues 24–34) The resulting complex did not involve major

steric clashes except with the flexible C-terminus of Anp32a

The structure was energy minimized by the gromacs

pack-age [58] using the gromos96 force field [59] to relieve

possi-ble structural strain

Atx1 interactions

Interaction of Anp32a with the Atx1 AXH domain was

probed both by NMR spectroscopy and by fluorescence

spectroscopy Two different constructs of the Atx1 AXH domain with different C-terminal boundaries (residues 567–

689 or residues 567–694) were used Typically, 0.2–0.3 mm solutions of the 15N-labeled Anp32a LRR domain in

20 mm Tris (pH 7.0) and 2 mm TCEP were used for the NMR experiments They were titrated with stepwise addi-tions of concentrated stock soluaddi-tions of unlabeled Atx1 AXH domain up to a two to threefold excess of this The inverse titration using labeled Atx1 AXH domain and unla-beled LRR domain was also probed The experiments were carried out at 27C, both at low ionic strength to enhance even weak electrostatic interactions, and at physiological ionic strength (150 mm NaCl)

Fluorescence measurements were performed on a SPEX Fluoromaxspectrometer, by exciting at 295 nm (slit width 0.4 nm) a 10 lm sample of Atx1 AXH domain in 20 mm Tris (pH 7.0) and recording the emission intensity from 300

to 450 nm (slit width 1.5 nm) Titration was carried out by stepwise additions of a 0.87 mm stock solution of Anp32a LRR domain up to a 60 : 1 ratio The data were evaluated using the origin program package (Micro-Cal Software, Bletchley, UK)

Yeast two-hybrid analysis The DNA fragment encoding the murine Anp32a N-termi-nus (1–164 amino acids) was cloned into the pGBKT7 vec-tor (Clontech, Mountain View, CA, USA) for expression as

a Gal4 DNA-binding domain fusion protein This bait was transformed into an AH 109 yeast strain and used to screen

a human brain two-hybrid cDNA library from Clonetech

as previously described [60] DNAs recovered from clones selected by growth in quadruple-dropout media and Gal-X overlay assays were sequenced and compared with known sequences

Confocal microscopy cDNAs encoding full-length Anp32a and Clip 170 (Gene-Service, IMAGE 3592614) were cloned into the pBudCE4.1 vector (Invitrogen, Paisley, UK) The immu-nofluorescence assay was carried out essentially as described previously [61] Briefly, COS cells were grown overnight in chamber slides and transfected with pBudCE4.1 vector expressing V5-tagged Clip 170 and c-Myc-tagged Anp32a Forty-eight hours after transfection, cells were fixed using 4.0% paraformaldehyde, permeabi-lized with 0.2% Triton X-100⁄ NaCl ⁄ Pi and probed with fluorescein isothiocyanate-conjugated antibodies to V5 (Invitrogen) and Cy3-conjugated antibodies to c-Myc (Sigma, Poole, UK) for 1 h at room temperature After being washed with NaCl⁄ Pi, slides were mounted using Citifluor (Agar Scientific) before analysis by confocal microscopy Cells were visualized under a Leica laser