Genomics 90 (2007) 59–71 www.elsevier.com/locate/ygeno

Molecular pathology of expanded polyalanine tract mutations in the Aristaless-related homeobox ⁎ Cheryl Shoubridge a, ,1, Desiree Cloosterman a,b,1, Emma Parkinson–Lawerence a,b, Douglas Brooks a,c, Jozef Gécz a,b

a Department of Genetic Medicine, Women’s and Children’s Hospital, Adelaide 5006, Australia b Department of Paediatrics, University of Adelaide, Adelaide 5006, Australia c Sansom Institute, University of South Australia, Adelaide 5006, Australia Received 21 November 2006; accepted 14 March 2007 Available online 9 May 2007

Abstract

The Aristaless-related homeobox gene (ARX) is one of the major causing X-linked mental retardation. We have been interested in the pathogenic mechanism of expanded polyalanine tract mutations in ARX. We showed that the c.304ins(GCG)7 mutation causing an increase from 16 to 23 alanines increased the propensity of ARX aggregation and a shift from nuclear to cytoplasmic localization. We proposed that mislocalization of ARX via cytoplasmic aggregation and subsequent degradation leads to a partial loss of function, contributing to the pathogenesis. We identified 13 (IPO13), a mediator of nuclear import for a variety of , as a novel ARX interacting protein. We predicted that the transport of ARX by IPO13 from the cytoplasm to the nucleus might be disrupted by expanded polyalanine tract mutations, but our data showed that in both yeast and mammalian cells these mutant ARX proteins were still able to interact with IPO13. We established the nuclear localization regions of the ARX homeodomain that were required for the interaction with IPO13 and correct localization of the full-length ARX transcription factor to the nucleus. Crown Copyright © 2007 Published by Elsevier Inc. All rights reserved.

Keywords: ARX; X-linked mental retardation; Mutation; Polyalanine tract; Yeast-2 hybrid; Coimmunoprecipitation; Nuclear localization sequence

X-linked mental retardation (XLMR) is frequent, affecting belongs to a subset of Aristaless-related Paired-class (Prd-class) an estimated 1:600 males [1,2]. Although to date, Fragile X homeodomain proteins [9]. Conserved orthologs in mouse, syndrome is the most common form of XLMR, mutations in zebrafish [9], Xenopus [10], and Caenorhabditis elegans [11] numerous other genes leading to both syndromic and have been isolated. ARX is expressed predominantly in the nonsyndromic forms of XLMR are emerging [3]. One of human fetal and adult brain and skeletal muscle [6,12,13].A these genes is the Aristaless-related homeobox gene (ARX) [4]. role for Arx in regulating specific cell fate decisions is indicated Our laboratory originally characterized ARX and, together with in a subset of GABAergic interneurons in the developing and others, implicated it in XLMR phenotypes with intellectual adult brain [14,15] and is crucial for mediating the correct disability with or without additional features including epilepsy, endocrine cell fate in the pancreas [16,17]. infantile spasms, dystonia, lissencephaly, autism, and dysarthia Expanded polyalanine mutations due to either a c.304ins [5–7]. As the prevalence of ARX mutations in XLMR continues (GCG)7 (16 increasing to 23 alanines) or a c.428_451dup (12 to emerge, it has been suggested that ARX may rival Fragile X increasing to 20 alanines) account for a considerable portion as a cause of mental retardation and epilepsy in males [8]. ARX (∼45%) of ARX mutations [4–6,12,18]. Expansion of the first alanine tract has been identified in three families with X-linked infantile spasms syndrome, also known as West syndrome (WS) ⁎ Corresponding author. Fax: +61 8 8161 7342. E-mail address: [email protected] (C. Shoubridge). (MIM 308350) with infantile seizures, hypsarrhythmia, and MR 1 The authors wish it to be known that, in their opinion, the first two authors [6,18,19]. The c.428_451dup mutation expanding the second should be regarded as joint first authors. tract has been found mainly in families with nonsyndromic MR

0888-7543/$ - see front matter Crown Copyright © 2007 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.ygeno.2007.03.005 60 C. Shoubridge et al. / Genomics 90 (2007) 59–71

[4,6,18,20,21]. This mutation has also been found in cases of were interested to examine the effect of the polyalanine tract WS [6], sporadic MR [12,22], and families with Partington mutations on the binding of ARX to importin 13 (IPO13) and syndrome (MIM 309510), whose symptoms include moderate subsequent transport to the nucleus. We identified IPO13, a to severe MR, dystonic movement of the hands, and dysarthria member of the importin β superfamily that mediates nuclear [6,21,23]. import of a variety of proteins [34,35], as a novel interacting There is growing experimental support for expanded protein partner of ARX using yeast-2 hybrid screens and polyalanine tract mutations contributing to disease via forma- coimmunoprecipitation studies. In yeast and mammalian cells tion of mutant protein aggregates. Nine hereditary diseases are both mutant ARX proteins were able to interact with IPO13, caused by expansions of polyalanine tracts [24,25], eight of indicating that the cytoplasmic aggregation of the c.304ins which occur in transcription factors. Protein aggregation was (GCG)7 mutant protein does not arise from a lack of binding to first noted in the expanded polyalanine tract mutations of the IPO13. We propose that mislocalization of ARX via cytoplas- PABPN1 gene [26]. This gene is implicated in oculopharyngeal mic aggregation and subsequent degradation leads to a partial muscular dystrophy (MIM 164300) and is unique among the loss of function, contributing to the pathogenesis. other disease-causing expanded polyalanine-tract-containing proteins as it is not a transcription factor but a polyadenylate- Results binding protein [27,28]. More recently, studies examining pathogenic mechanisms of expanded polyalanine tracts in c.304ins(GCG)7 ARX mutation forms cytoplasmic aggregates diseases involving FOXL2 (MIM 110100) [29], HOXD13 in SH-SY5Y cells and PC12 cells (MIM 186000), SOX3 (MIM 300123), RUNX2 (MIM 119600), and HOXA13 (MIM 140000) [30], and PHOX2B Before measuring aggregate formation with expanded (MIM 603851) [31] showed that the mutant proteins have the polyalanine tract mutations of ARX we established the rate propensity to form aggregates. Compared to the wild-type for and location of aggregate formation of the wild-type ARX each protein, smaller expansions of alanine tracts lead to an protein in SH-SY5Y cells and PC12 cells. The results in both increased percentage of transfected cells with nuclear aggre- cell types were almost identical and the data have been pooled. gates [29–31]. Larger alanine repeats, with a threshold above Transient overexpression of green fluorescent protein (GFP)- 18–22 alanines, resulted in a high proportion of transfected cells tagged wild-type ARX for 48 h resulted in aggregation in less with mislocalization of the expressed protein to aggregates than 7% of all transfected PC12 and SH-SY5Y cells (Fig. 1). within the cytoplasm [29–31]. The aggregates generally appeared as bright spots in the We have been particularly interested in the pathogenic nucleus. The rate of aggregate formation was consistently low at mechanism of the expanded polyalanine tract mutations due to 24 and 72 h posttransfection (data not shown). Transfection of either a c.304ins(GCG)7 or a c.428_451dup in ARX. The extent the c.428_451dup mutant construct into either cell type resulted of aggregate formation of expanded polyalanine tract mutations in similar low levels of abnormal localization or aggregate of ARX and their involvement in pathogenesis remains unclear. formation as seen in the wild-type ARX transfected cells (Fig. 1). Initial reports from an in vitro study indicated that there was no In contrast, overexpression of the c.304ins(GCG)7 ARX mutant aggregate formation with overexpression of c.304ins(GCG)7 protein for 48 h resulted in abnormal localization or aggregate expanded polyalanine tract mutant protein [14]. In contrast, a formation in 40% of cells (Fig. 1), an approx sixfold increase more recent study modeling the human c.304ins(GCG)7 compared to wild-type ARX. mutation in mouse ortholog of ARX demonstrated accumula- Subcellular localization of ARX expression (48 h post- tion and formation of nuclear aggregates in approximately 30% transfection) in SH-SY5Y and PC12 cells was categorized as of cells transfected with the mouse mutant construct [32]. either (i) normal diffuse nuclear staining with little or no signal Support for nuclear aggregation of the c.304ins(GCG)7 in the nucleolus or surrounding cytoplasm or (ii) cells with mutation was provided in a recent review of ARX in cortical bright spot(s) in the nucleus (nuclear inclusions) or (iii) diffuse development [33]. However, the more frequent 24-bp duplica- signal with aggregates across the nucleus and cytoplasm tion mutation was not tested. Despite these data, we would (cytoplasmic aggregates) (Figs. 2B–2D). Aggregates of the expect cytoplasmic aggregates with the longer 23-alanine tract normal and c.428_451dup mutant protein were predominantly arising from the human c.304ins(GCG)7 mutation in light of the nuclear inclusions (Fig. 2A). In contrast, the c.304ins(GCG)7 studies in other disease-causing expanded polyalanine tracts. ARX mutant protein formed aggregates with a predominant The aim of this current study was to characterize aggregate perinuclear or cytoplasmic localization. In 80% of cells with formation using the full-length human expanded alanine tract aggregation due to the c.304ins(GCG)7 ARX construct, the mutations of ARX. Data from our study have shown that an mutant protein localized to large aggregates in the cytoplasm, increase from 16 to 23 alanines in the c.304ins(GCG)7 mutation often to one side of the nucleus in addition to low levels of was associated with an increased propensity of ARX protein diffuse nuclear staining (Figs. 2A and 2D). This pattern of aggregation and a shift from nuclear to cytoplasmic localization. abnormal localization was also noted at 72 h posttransfection However, there was no significant increase in aggregate (data not shown). Both the ARX aggregates inside the nucleus formation either in the nucleus or in the cytoplasm with the (Fig. 2Ea) and those in the cytoplasm (Fig. 2Eb) costained with more frequent c.428_451dup mutation, indicating that aggre- an antibody against the heat shock protein Hsp70, regardless of gate formation is not a common pathogenic mechanism. We the construct tested. C. Shoubridge et al. / Genomics 90 (2007) 59–71 61

Fig. 1. Percentage of abnormal subcellular localization of ARX expanded alanine tract mutations. (A) Schematic of the human ARX protein. Human ARX domains and regions are indicated above the schematic and their locations are indicated below. Known functional domains are highlighted, octapeptide (OP) as horizontally hatched rectangle, nuclear localization sequences (NLS) as three black rectangles, polyalanine tracts (PA) as four white rectangles, acidic domain as vertically hatched rectangle, homeodomain crosshatched, and Aristaless domain (OAR) hatched. The open reading frame is indicated in boldface. The expanded alanine tract mutations are shown in respect to the wild-type. (B) Full-length GFP-tagged ARX Wt (open bars), full-length c.428_451dup (gray bar), and full-length c.304ins(GCG)7 (black bar) constructs were transfected into SH-SY5Y or PC12 cells for 48 h. The percentage of transfected cells displaying abnormal localization as nuclear inclusions or as cytoplasmic aggregates was determined from between 800 and 4000 cells per construct from at least three separate transfection reactions in both cell types. An all-groups comparison of the percentage of abnormal cell localization of ARX expression in transfected cells was achieved by ANOVA.* p<0.001 vs ARX Wt and c.428_451dup.

To ensure that the aggregation was not due to the large GFP transferase-mediated dUTP nick end labeling (TUNEL) assay. tag used in these studies, we also used Myc-tagged and In nontransfected cells and lipofectamine-only-transfected cells, nontagged ARX normal and mutant constructs (data not the rate of TUNEL remained low at both 24 and 48 h shown) in both SH-SY5Y and PC12 cells. Our analysis showed posttransfection (data not shown). In ARX transfection experi- that the rate and subcellular localization of aggregate formation ments, only cells positive for ARX expression were then scored of Myc-tagged or nontagged ARX constructs were not different for TUNEL signal. The levels of ARX transfected cells positive from those of the GFP-tagged ARX constructs. Tagged and for TUNEL were low at 24 h posttransfection with increasing nontagged ARX were detected using an anti-ARX antibody that levels for all constructs at 48 h posttransfection (Fig. 4). At both we have generated. The specificity of this antibody was times tested the rate of cell death in cells transfected with confirmed by colocalization with GFP-tagged ARX by immu- c.304ins(GCG)7 ARX was markedly higher than that in either of noflorescence (IF) (Fig. 3A) and detection of the same size band the other two constructs, correlating with increased aggregation as a commercially available anti-Myc antibody in a Myc–ARX of the mutant protein in the cytoplasm. sample by Western immunoblot (WB) (Fig. 3B). Interaction of ARX and IPO13 Increased cell death in SH-SY5Y cells transfected with c.304ins(GCG)7 ARX To identify proteins that interact with ARX we performed GAL4-based yeast-2 hybrid screening. In this paper we report To ascertain whether the formation of aggregates correlated only on our findings of an interacting protein against the with programmed cell death, the rate of apoptosis was measured homeodmain of ARX. Bait proteins were produced by fusing in SH-SY5Y cells transfected with ARX normal, c.428_451dup, the human ARX homeodomain (amino acids (aa) 303–431) to the and c.304ins(GCG)7 ARX by terminal deoxynucleotidyl- GAL4 DNA binding domain (GAL4-DBD). The bait protein did 62 C. Shoubridge et al. / Genomics 90 (2007) 59–71

Fig. 2. Aggregate formation of expanded alanine tract mutations in ARX. (A) Percentage of transfected SH-SY5Y and PC12 cells displaying nuclear inclusions or cytoplasmic aggregates of GFP-tagged wild-type ARX (open bars), c.428_451dup (gray bars), and c.304ins(GCG)7 (black bars) protein 48 h after transfection. All- groups comparison of the percentage of abnormal cell localization of ARX expression as either nuclear inclusions or cytoplasmic aggregates in transfected cells was achieved by ANOVA. * p<0.001 vs ARX Wt and c.428_451dup transfected cells with cytoplasmic aggregate localization and to all ARX construct transfected cells with nuclear inclusions. Subcellular localization of (B) wild-type ARX, (C) c.428_451dup, and (D) c.304ins(GCG)7 show normal nonhomogenous staining in the nucleus (top) and abnormal localization as nuclear inclusions (middle) for all constructs and aggregates within the cytoplasm (bottom) for both polyalanine mutant constructs. Cell nuclei were stained with DAPI (left) and GFP-tagged ARX (center) and the two images were merged (right). (E) Costaining of nuclear inclusions (a) and cytoplasmic aggregates (b) of mutant ARX with an anti Hsp70 antibody. C. Shoubridge et al. / Genomics 90 (2007) 59–71 63

www.fccc.edu/research/labs/gelemis/InteractionTrapInWork. html) of known false positives. The interaction between the ARX homeodomain and the isolated IPO13 protein was confirmed by yeast-based recombinational cloning. The interaction between the IPO13 and the entire ARX protein was initially confirmed in yeast and then in mammalian cells. In yeast, the HIS3, URA3, and LacZ reporter genes were not auto- activated with ARX normal protein fused to GAL4-DBD alone but were activated when both full-length GAL4-DBD-ARX normal protein and GAL4-AD-IPO13 were transformed into MaV203 yeast (data not shown). In mammalian cell studies, to increase the cellular amount of IPO13 for coimmunoprecipitation, the last 747 aa of IPO13 tagged to V5 was transiently transfected into HEK 293T cells along with Myc-tagged normal ARX. This truncated version of IPO13 lacks the RanGTP binding site and Fig. 3. Anti-ARX antibody detects tagged ARX proteins. The monoclonal anti- maximizes the likelihood of detecting the otherwise transient ARX antibody (ARX-2-1A3_B5) detects GFP-tagged wild-type ARX protein in interaction between ARX and IPO13. A monoclonal anti-V5 transfected SH-SY5Y cells. (A) The green GFP tag is visualized using a fluoroscein isothiocyanate filter. The antibody against ARX is detected using a antibody was used to immunoprecipitate V5-IPO13 from whole- CY3-conjugated secondary antibody and visualized by a tetramethylrhodamine cell extracts. Coimmunoprecipitated Myc-tagged ARX was isothiocyanate filter. The signal detected via the ARX2-1A3_B5 antibody detected by an anti-Myc antibody conjugated to horseradish overlaps the GFP tag expressed with the ARX protein. (B) Cell lysates of Myc– peroxidase (HRP) (Fig. 5A), indicating that IPO13 and full-length – ARX transfected cells were subjected to SDS PAGE and immunoblotted with ARX associate in HEK 293Tcells. The interaction between Myc- antibodies against ARX or Myc tag. There were no bands in the nontransfected 293T cell lysate (lane 1) but there was a band of ∼62 kDa (ARX 58 kDa+ Myc tagged ARX and IPO13 was confirmed in cotransfected cells by 4 kDa) in lysate from 293T cells transfected with Myc-ARX (lane 2) using either immunoflorescence. Endogenous IPO13 (Fig. 5Ba) and V5- our ARX-2 1A3_B5 antibody or the commercially available anti-Myc antibody. IPO13 (Fig. 5Bb) were detected as diffuse signals across the cell (For interpretation of the references to colour in this figure legend, the reader is cytoplasm and nucleus in untransfected cells. Endogenous IPO13 referred to the web version of this article.) signal was not affected by transfection of ARX (Fig. 5Ca), indicating the transient interaction of these proteins. However, the not autoactivate the HIS3 or URA3 reporter genes upon RanGTP-binding-site-deficient V5-IPO13 colocalized with transformation of MaV203 yeast but did autoactivate the lacZ cotransfected ARX (Fig. 5Cb). To ascertain whether the expanded reporter gene (data not shown). Upon screening a human fetal polyalanine mutations in ARX disrupt the normal interaction with brain cDNA library with the bait protein, 1 of 33 double IPO13, these procedures were repeated using c.304ins(GCG)7 transformants activated both the HIS3 and the URA3 reporter ARX or c.428_451dup ARX in place of normal ARX. Coimmu- genes (data not shown). Analysis of the single positive clone noprecipitation (Fig. 5A) and immunoflorescence confirmed that transformed with the homeodomain identified the cDNA insert both expanded polyalanine tract mutant proteins interact with from the interacting plasmid as the last 747 aa of IPO13 (importin IPO13 whether ARX was located diffusely to the nucleus or as 13, NM_014652). This cDNA was not present in the list (http:// aggregates in the nucleus (Fig. 5Cc) or the cytoplasm (Fig. 5Cd).

Fig. 4. Percentage of ARX transfected cells positive for apoptosis was measured by TUNEL. SH-SY5Y cells were transfected with GFP-tagged wild-type ARX (open bars), c.428_451dup (gray bars), and c.304ins(GCG)7 (black bars) constructs. Only cells positive for ARX transfection were scored for TUNEL signal at 24 and 48 h posttransfection. Over 250 ARX transfected cells were counted per construct for each time point, from at least three separate experiments. All-groups comparison of the percentage of TUNEL-positive ARX transfected cells was achieved by ANOVA. * p<0.001 vs ARX Wt and c.428_451dup within each time point. 64 C. Shoubridge et al. / Genomics 90 (2007) 59–71

Fig. 5. ARX interacts with IPO13 in mammalian cells. (A) HEK 293 T cells transfected with Myc-ARX (wt) or c.304ins(GCG)7 (in) or c.428_451dup (du) and V5- IPO13 were lysed and proteins were immunoprecipitated (IP) with rabbit anti-V5 antibody. Samples were loaded on 4–12% SDS–PAGE gels and analyzed for the presence of Myc-ARX bound to V5-IPO13 by immunoblotting with mouse anti-Myc HRP-conjugated antibody (top). The presence of Myc-ARX and V5-IPO13 present in the immunoprecipitate are shown in the middle and bottom, respectively. Cells transfected with Myc-ARX alone or V5-IPO13 alone were used as negative controls. (B) Endogenous IPO13 (a) and transfected V5-IPO13 (b) staining using a polyclonal anti-IPO13 antibody was seen as bright signal in the cytoplasm and across the nucleus. (C) In cells transfected with ARX alone, endogenous IPO13 was strongest in the cytoplasm and did not colocalize with the ARX signal in the nucleus (a), reflecting the transient interaction of these proteins. When Myc-ARX or c.304ins(GCG)7 or c.428_451dup ARX and RanGTPase-binding-site-deficient V5-IPO13 plasmids were cotransfected into cells, the IPO13 localization reflected ARX localization in all cells that expressed both proteins, regardless of whether ARX was localized to the nucleus (b) or in aggregates in either the nucleus (c) or the cytoplasm (d).

Interaction of IPO13 with nuclear localization sequences IPO13 was due to deletion of NLS3 residues, we mutated the (NLS) in the homeodomain of ARX individual residues in NLS3 (shown in boldface in Fig. 6A) by site-directed mutagenesis. Mutation of NLS3 residues also There are four naturally occurring mutations (R332P, resulted in no interaction between the mutant ARX home- R332H, R332C, and T333N) in the homeodomain NLS2 odomain and the IPO13 in yeast (Fig. 6C), suggesting that these region. The mechanism of pathogenesis of these mutations is residues in NLS3 are vital for this interaction. not known. The four point mutations were introduced into an otherwise normal homeodomain construct (Fig. 6A) to establish Interaction of IPO13 with the C-terminal nuclear localization whether these point mutations disrupt the interaction of ARX to sequences in the homeodomain of ARX is required for correct IPO13 in yeast. Constructs containing a point mutation in NLS2 nuclear localization activated the HIS3 (Fig. 6B) and URA3 reporter when GAL4AD-IPO13 was present (data not shown), suggest- We wanted to confirm the function of IPO13 binding to NLS ing that none of these point mutations by themselves were regions in the ARX homeodomain in the subsequent localization able to significantly affect the binding of IPO13 to ARX of this transcription factor to the nucleus in mammalian cells. homeodomain in yeast. However, deletion of NLS2 (ARX- Key basic residues in NLS regions of the homeodomain in full- HDΔNLS2) resulted in patchy activation of the reporter genes length Myc∼tagged ARX constructs were mutated by site- in only a few colonies (8/25 colonies in repeated experiments) directed mutagenesis (Fig. 7A) and then transiently transfected and deletion of NLS3 (ARX-HDΔNLS3) completely (0/25 into 293T cells with and without cotransfection of V5-IPO13. colonies) abolished reporter gene activation (Fig. 6C). To The putative NLS1 region was not targeted for mutation analysis ensure that this lack of binding between ARX-HDΔNLS3 and as subcellular localization studies using partial ARX constructs C. Shoubridge et al. / Genomics 90 (2007) 59–71 65

Fig. 6. Yeast screening of IPO13 interaction with normal and mutant ARX homeodomain constructs. (A) Schematic of the partial ARX protein, with the amino acid location indicated above, containing the homeodomain is depicted with the NLS2 and NLS3 regions as black boxes. The sequences in boldface above the black boxes for NLS2 and NLS3 depict basic residues that are part of NLS-like motifs. Missense mutations identified in NLS2 are shown above the sequence. (B) ARX homeodomain constructs are indicated on the left, with the yeast-2 hybrid screening results on the right. All MaV203 yeast colonies transformed with both the GAL4- DBD-ARX construct and the GAL4-AD-IPO13 are able to grow on the -L-W master plate. Colony growth on the HIS3 (30 mM 3AT) reporter plate is achieved only if the two constructs interact to allow reporter gene expression. Five Yeast Control MaV103 Strains (Control A–E) included with the ProQuest Yeast-2 Hybrid System contain plasmids expressing fusion proteins which have a spectrum of interacting strengths from no interaction in Strain A to very strong interaction in Strain E. (C) ARX homeodomain constructs with either the NLS2 or the NLS3 region deleted or individual residues in NLS3 mutated (left). Two rows of 10 MaV203 yeast colonies per construct transformed with both the GAL4-DBD-ARX and the GAL4-AD-IPO13 constructs all grow on the -L-W master plate (middle) with protein interaction determined by growth on the HIS3 (20 mM 3AT) reporter plate (right). in pCMV-Myc expression vectors indicated that this region was localization of the full-length ARX protein in transfected not crucial for ARX nuclear localization. ARX-Poly(A) partial cells. Mutations in NLS2 resulted in large aggregates either in or construct (aa 61–174) contained the NLS1 region (aa 82–89) over the nucleus (Fig. 7Cb) in 32% of transfected cells (Fig. and the homeodomain construct (aa 303–431) contained both 7D), suggesting that NLS3 can function in ∼70% of cells to the NLS2 and the NLS3 regions (Fig. 7B). These partial correctly localize ARX to the nucleus. In contrast, when constructs were transiently transfected into 293T cells and residues in NLS3 are mutated, over 60% of all transfected cells expressed protein was visualized using an anti-myc antibody. (Fig. 7D) display the mutant protein localized to large The homeodomain clearly locates to the nucleus (Fig. 7Bb) aggregates in the cytoplasm (Fig. 7Cc). These data clearly while the construct containing the NLS1 region is localized suggest that a functioning NLS2 alone is not enough to correctly diffusely in the cytoplasm (Fig. 7Ba). localize the ARX transcription factor to the nucleus. Cotrans- Disruption of residues in the NLS2 region or the NLS3 fection of V5-IPO13 with the Myc-tagged ARX wt (Fig. 7Eb) region (Fig. 7Cc) both resulted in abnormal subcellular and ARX-NLS2 mutant (Fig. 7Ec) constructs shows that the 66 C. Shoubridge et al. / Genomics 90 (2007) 59–71

Fig. 7. Changes in nuclear localization of full-length ARX in 293T cells when the NLS regions are disrupted. (A) Schematic of the full-length ARX protein (domains as illustrated in Fig. 1A) is depicted with the NLS regions as three black boxes. The sequences in boldface above the black boxes for NLS2 and NLS3 depict basic residues that are part of NLS-like motifs, with the mutations introduced by site-directed mutagenesis listed above individual residues. (B) Partial ARX constructs containing the NLS1 region (ARX-Poly(A)) and the ARX homeodomain (ARX-HD) region are depicted, with the amino acid location indicated above. Partial constructs transfected into 293Tcells and expression detected via an anti-Myc antibody and a TRITC-labeled secondary antibody (left) with the merged image showing DAPI staining of cell nuclei (right). The NLS1-containing construct did not localize to the nucleus of transfected 293T cells although partial constructs with the NLS2 and NLS3 regions did localize exclusively in the nucleus. (C) The subcellular localization of Myc-ARX wild-type (a) changes when ARX constructs with mutations in NLS2 (b) or NLS3 (c) are transiently transfected into 293T cells, detected using our anti-ARX-21A3_B5 antibody and a FITC-labeled secondary antibody (left) and merged with image of DAPI-stained cell nuclei (right). (D) Percentage of transfected cells with abnormal localization increased following expression of ARX-NLS2 mutant (gray bar), with even higher levels with ARX-NLS3 mutant (black bar) as compared to ARX Wt (open bar). At least 2000 transfected cells were counted for each construct, from three separate experiments. (E) V5-IPO13 transfected into HEK 293T cells localizes diffusely across the cell cytoplasm. All-groups comparison of the percentage of abnormal cell localization of ARX expression in transfected cells was achieved by ANOVA. * p<0.001 vs ARX Wt and ARX-NLS2 mutant, with ϒ P<0.001 vs ARX Wt and ARX-NLS3 mutant. (E) (a) Cotransfection of V5-IPO13 with the Myc-tagged ARX Wt (b) and ARX-NLS2 mutant (c) constructs shows that the IPO13 changes its localization pattern to mirror the nuclear localization of ARX, indicating that these proteins are interacting. Cotransfection of V5-IPO13 with ARX-NLS3 mutant constructs results in cells with large cytoplasmic aggregates that are positive for both ARX and IPO13 (d) but are not localized to the nucleus. C. Shoubridge et al. / Genomics 90 (2007) 59–71 67

IPO13 changes localization pattern from diffuse cytoplasmic support a loss of transcriptional activity due to expanded staining (Fig. 7Ea) to mirror the nuclear localization of polyalanine tract mutations in ARX [37]. Arx has been ARX. This indicates that, although the NLS2 region of the established as a transcriptional repressor [17] influencing cell homeodomain is mutated, the NLS3 region is adequate to bind fate decisions [14,16,17]. Studies in our laboratory have IPO13 and in the majority of cells can enable correct transport confirmed, in cell-based assays, that ARX acts as a strong of ARX to the nucleus. The IPO13 signal in cells cotransfected transcriptional repressor and have established that the majority with the ARX-NLS3 mutation construct show an overlapping of the repression resides in the fourth, highly conserved subcellular signal with ARX-NLS3 mutant protein (Fig. 7Ed), polyalanine tract [37]. The repression activity of ARX was indicating that NLS2 alone can at least bind IPO13. However, unaffected by the c.428_451dup mutation and was increased by the subcellular localization of these proteins was in large the c.304ins(GCG)7 ARX mutant construct. Hence, we can aggregates outside of the nucleus in the cytoplasm (Fig. 7Ed). speculate that one cause of ARX dysfunction due to expanded Hence, the interaction of IPO13 with the NLS3 region of the polyalanine tracts may arise in part from increased repression ARX homeodomain is vital for the correct transport of this activity of mutant ARX homeobox protein. Disturbance to the homeobox transcription factor protein into the nucleus. normal transcriptional repression during early brain develop- ment could lead to inappropriate gene expression and Discussion subsequent disturbance of normal cell development and tissue patterning. The severe “loss of function” phenotypes of ARX do In this study we have shown that overexpression of the indeed result in a disturbed patterning of the cortical tissue in the c.304ins(GCG)7 mutant caused a significant increase in protein brain, thought to arise from incorrect migration of neurons aggregation compared to either the normal wild-type ARX during early brain development [13]. Hence, changes in the sequence or the c.428_451dup mutant. This increased propen- transcriptional activity of ARX due to these expanded sity of the c.304ins(GCG)7 mutant alanine tract to aggregate polyalanine tract mutations may have subtle effects on neuronal was in agreement with the data previously reported [32]. organization and may contribute to pathogenesis of the affected However, this previous report indicated that aggregates of individuals. mouse Arx sequence, with insertion of additional alanines to In our study we did not observe any significant increase in mimic the human situation, were observed as nuclear the propensity of the c.428_451dup mutant protein to aggregate inclusions. In contrast, we have shown that the location of or change in localization compared to the wild-type protein. aggregates shifted from predominantly nuclear inclusions with These data indicate that very likely soluble protein with altered normal and duplication mutation protein expression to a function is the key molecular pathogenesis of ARX polyalanine predominantly perinuclear or cytoplasmic location of aggrega- tract expansions. In contrast, as the aggregates of the c.304ins tion with the 23 alanine tract mutant protein. Experimental data (GCG)7 mutant protein mislocalized to the cytoplasm, we can presented for FOXL2, HOXA13, HOXD13, PHOX2B, predict that the lack of the transcription factor being located to RUNX2, and SOX3 also show that mutations leading to longer the nucleus may lead to partial loss of function, even though the polyalanine tracts result in cytoplasmic aggregation of the mutation itself does not abolish the transcriptional activity of mutant protein [29–31]. These experimental data are well ARX [37]. Despite experimental evidence of longer expanded aligned with the theory of expanded alanine protein aggregation polyalanine tract mutations forming aggregates, the clinical in an alanine-dose-dependant manner, with cellular environ- relevance of these aggregates remains to be established for ment influencing the threshold or propensity to aggregate. many of the genes implicated. For example, PABPN1, a poly(A) ARX belongs to a growing number of disease-causing genes binding protein, is the only disease-causing gene due to alanine with mutations that lead to expanded polyalanine tracts. These expansion mutations that is not a transcription factor and in expanded alanine tracts are stably transmitted across multiple which intranuclear inclusions are reported as a pathological generations with many giving rise to congenital defects [25,36]. hallmark of OPMD (MIM 164300) [28,38,39]. However, the The lengths of normal alanine tracts are similar in all filaments of PABPN1 have also been observed in neurons as transcription factors (12–20 alanines) and so is the amount of dynamic structures whose appearance depends upon changes to increased length of these tracts that cause disease (18–29 cellular activity [40]. Hence, the contribution of aggregates of alanines). The pathogenic mechanism of expanded polyalanine expanded alanine mutant PABPN1 protein to the pathogenesis tract mutations is not clear. In the case of ARX, a gain of of OPMD remains speculative. function hypothesis is unlikely as these polyalanine expansion Another possible pathogenic mechanism of these expanded mutations are inherited in an X-linked recessive fashion with alanine tract mutations would be a disruption of the binding of carrier females largely unaffected. The clinical similarity of the mutant protein to its normal DNA targets or interacting polyalanine expansion mutations and some point mutations proteins. In the case of ZIC2, the phenotype arising from loss of [6,13] indicates a partial loss of function of this homeobox function mutations is not noticeably different from that arising transcription factor as a likely pathogenic mechanism. This from the naturally occurring mutation increasing the 15 alanine could occur by a number of different mechanisms. Aggregation tract to 25 alanines [41]. Functional studies indicate that the of the mutant protein could result in a loss of transcriptional expanded alanine mutant protein of ZIC2 influences the activity or it could disrupt normal cellular processes due to the strength of DNA binding, resulting in a near complete loss of aggregates themselves. Recent work in our laboratory does not the normal levels of transcriptional activity [41]. Similarly, 68 C. Shoubridge et al. / Genomics 90 (2007) 59–71 expanded polyalanine mutations of SOX3 lead to a reduced distribute the protein once inside the nucleus. Subtle differences activation of luciferase reporter gene downstream of the in the amounts of ARX imported to the correct location within proximal promoter of Hesx1, a purported binding site for the nucleus during development may affect important transcrip- SOX proteins [42]. This reduced activity is thought to be tional regulatory events during GABAergic interneuron pro- primarily due to an inability of the mutant protein to bind to the liferation, differentiation, and migration and may contribute to DNA binding site correctly. A complete “loss of function” of XLAG or Proud syndrome. Alternatively, these point mutations ARX due to expanded polyalanine mutations appears unlikely may make no significant difference to the import of ARX into given frameshift mutations that lead to premature termination of the nucleus but may affect the binding of ARX to DNA or other ARX protein, resulting in a very severe phenotype of X-linked transcriptional factors or may influence other posttranslational lissencephaly with abnormal genitalia (XLAG) (MIM 300215) modifications of this homeobox transcription factor. [13], and this phenotype has never been reported for either expanded alanine tract mutation of ARX. Nor does it appear that Materials and methods these mutations abolish the binding of ARX to the novel interacting protein partner, IPO13. We have established an Construction of ARX-containing vectors interaction between ARX and transport protein IPO13 by yeast- Gateway cloning and expression vectors (Invitrogen) were used unless 2 hybrid studies and coimmunoprecipitation. Although we otherwise stated. All clones were sequenced to exclude introduced errors and the predicted that the formation of cytoplasmic aggregates with the concentration was determined spectrophotometrically. c.304ins(GCG)7 mutant protein may arise from a lack of Full-length human ARX construct (ARX Wt) was produced by a series of adequate delivery of the mutant protein to the nucleus, we PCRs from genomic DNA purified from blood and the PCR products were demonstrated that IPO13 continues to bind to ARX despite ligated together via compatible restriction sites. The construct was cloned into a range of expression vectors, including phrGFP-N1 (Stratagene) and the Gateway expansion of the polyalanine tracts. However, we cannot vectors including pcDNA3.2/V5-DEST and Gateway-compatible pCMV-Myc exclude the possibilities that the binding of these expanded (BD-Biosciences) expression vectors for cell transfections and pDEST32 vector polyalanine tract mutant ARX proteins to IPO13 may be less for yeast-2 hybrid screening. The expanded polyalanine mutations were efficient than that of the normal ARX protein and may lead to amplified from genomic DNA of affected individuals and inserted into full- compromised nuclear import, contributing in part to the length constructs (c.304ins(GCG)7 ARX and c.428_451dup ARX). The residues in the homeodomain NLS regions of the full-length pCMV-Myc-ARX molecular pathogenesis of these disorders. constructs predicted to be important in binding to IPO13 (underlined) were IPO13 has been shown to mediate the nuclear import of the mutated by site directed mutagenesis using the QuikChange site-directed proteins PAX6, PAX3, and CRX by binding to NLS sequences mutagenesis kit (Stratagene) following manufacturer’s instructions. ARX-NLS2 associated with Paired-type homeodomains [35]. These same normal Lys-Arg-Lys-Gln-Arg-Arg-Tyr-Arg → mutant Lys-Ser-Ile-Gln-Gly- → highly conserved NLS sequences have also been shown to be Gly-Tyr-Gly, ARX-NLS3 normal Arg-Arg-Ala-Lys-Trp-Arg-Lys-Arg mutant Gly-Gly-Ala-Lys-Trp-Ser-Lys-Arg. IPO13 (aa 217–747) were cloned important for the nuclear localization of the Paired-class into pcDNA3.1/nV5-DEST for mammalian cell culture studies. homeodomain proteins CART1, VSX-1, and OTX1 [43–45]. For yeast-2 hybrid screening, standard PCR and subcloning techniques were Four naturally occurring point mutations within these highly used to generate a partial ARX construct which contains the homeodomain – conserved residues of NLS2 in ARX have been shown to cause (ARX-HD; aa 303 431) from human fetal brain cDNA (Clontech) fused in X-linked lissencephaly with ambiguous genitalia (XLAG) frame with the GAL4 DNA binding domain in the Gateway pDEST32 expression vector. IPO13 (aa 217–747) was also cloned into the vector (R332P, R332H, and R332C) [7,13,46] and Proud syndrome pDEST22 where it would then be fused in frame with the GAL4 activation (T333N) [13]. We therefore examined the interaction between domain. Missense mutations introduced into the ARX-HD construct included IPO13 and ARX homeodomain with mutations in NLS2 by c.994C>T leading to R332C (ARX-HD-R332C), c.995G>C leading to R332P yeast-2 hybrid analysis. Interestingly, we found that the four (ARX-HD-R332P), c.995G>A leading to R332H (ARX-HD-R332H), and point mutations had no effects on the interaction between ARX c.998C>A leading to T333N (ARX-HD-T333N). NLS regions within the ARX homeodomain were deleted to examine the effects of each on binding to IPO13 homeodomain and IPO13. However, since the binding of IPO13 in yeast. ARX-HDΔNLS2 contains aa 337–431, ARX-HDΔNLS3 contains aa to PAX6 was found to be stronger with the NLS at the C- 303–376, and point mutations in NLS2 without the NLS3 sequence included terminal end of the homeodomain [35], we speculated that the ARX-HD-R332CΔNLS3 and ARX-HD-T333NΔNLS3. interaction between IPO13 and NLS3 of ARX was sufficiently strong to activate yeast reporter gene expression. Our study Cell culture confirmed that the interaction of IPO13 was abolished when the ’ NLS3 region of the ARX homeodomain was removed or Rat pheochromocytoma cells (PC12) were maintained in Dulbecco s modified Eagle’s medium (DMEM) supplemented with 5% (v/v) fetal calf mutated. Indeed, disruption of key residues within the NLS3 serum (FCS) and 10% (v/v) horse sera. Human neuroblastoma SH-SY5Y cells region of the full-length protein resulted in the protein were cultured in a 1:1 mixture of DMEM and Ham’s F12 medium with 10% inappropriately localizing outside the nucleus in the majority FCS. HEK 293T were maintained in DMEM supplemented with 10% (v/v) FCS. of transfected cells. When key residues of the NLS2 region were All cells were cultured in the presence of 100 U/ml sodium penicillin and μ disrupted, however, a smaller proportion of cells with abnormal 100 g/ml of streptomycin sulfate in 5% CO2 at 37 °C. ARX localization was observed. Therefore, although point mutations in NLS2 of ARX may still enable the protein to Antibody production interact with IPO13 through its binding to the intact NLS3, We have produced a monoclonal antibody to a small peptide region of the these mutations may have subtle affects on the ability of IPO13 mature ARX protein (KISQAPQVSISRSKSYREN; (aa 161–17) with 100% to import ARX into the nucleus efficiently or to correctly amino acid homology across human mouse and zebrafish species. A custom- C. Shoubridge et al. / Genomics 90 (2007) 59–71 69 made synthetic peptide (Mimotopes) was conjugated to diptheria toxoid linked citrate used as the permeabilization buffer. After the secondary antibody was to the peptide via maleimidocapryl-N-hydoxysuccinimide. BALBc mice were washed off the sections were rinsed twice with PBS and then incubated with TdT immunized with 50 μg protein per mouse in 200-μl bolus administered by enzyme and UTP-TMR red at 37 °C for 1 h (Roche). Transfected cells were subcutaneous injection. A priming injection of peptide in Freud’s complete scored for presence of TUNEL staining. The rate of TUNEL staining in adjuvant was followed with a second and third boost of peptide in Freud’s nontransfected cells was determined by counting all TUNEL-positive cells in incomplete adjuvant and a fourth boost of peptide in phosphate-buffered saline one pass of the middle of each coverslip from top to bottom. There were no (PBS), with each treatment at fortnightly intervals. The mice were culled 3 days differences in the rates of TUNEL staining in nontransfected cells at each time after the final boost to harvest the spleen. The dissociated spleen cells were fused point in the three ARX constructs (data not shown). to myeloma cells to allow the formation of antibody-producing hybridomas under the selection of HAT media. Conditioned media from the resulting Yeast-2 hybrid screening and interaction testing hybridomas were screened for reactivity against the peptide used to inoculate the mice. Clones that gave a positive enzyme-linked immunosorbent assay result Plasmids encoding the ARX domains were fused to the GAL4 DNA binding against the specific peptide and little or no cross-reactivity to either diptheria domain and the library proteins were fused to the GAL4 activation domain. The toxoid or bovine serum albumen (BSA) were tested against the mature protein MaV203 yeast strain was cotransformed with the ARX-HD construct and the by immunoflorescence and Western immunoblot. A successful clone, ARX-2 human fetal brain cDNA library (ProQuest, Invitrogen). Colonies were selected 1A3, was haplotyped as an IgG antibody and, after subcloning the B5 daughter on media lacking leucine, tryptophan, and histidine and positive clones were clone, was selected for use. analyzed for the three reporter genes HIS3, URA3, and lacZ which were measured by β-galactosidase activity. Library inserts of positive clones that Antibodies activated more than one reporter gene were amplified by PCR and sequenced to determine the identity of the library clone. Interaction between the ARX The following antibodies were used for immunoflorescence and/or Western homeodomain and the isolated clones was confirmed by yeast-based immunoblot analyses: mouse anti-ARX (2 μg/ml final) (IF and WB), rabbit anti- recombinational cloning as described previously [47]. HSP70 (1:200) (IF) (Chemicon Int.); mouse anti-Myc HRP-conjugated antibody (1:5000) (WB) (Invitrogen); rabbit anti-V5 antibody (1:5000) (WB) (Bethyl Coimmunoprecipitation Laboratories); and rabbit anti-IPO13 (1:1000) (IF) kindly provided by Professor Dirk Görlich, (ZMBH, Heidelberg, Germany). The secondary antibodies were To verify the interactions between Myc-ARX and V5-IPO13, cotransfected goat anti-mouse antibody conjugated to CY3 (IF) and goat anti-rabbit antibody HEK 293T cells were lysed and immunoprecipitated with mouse anti-Myc conjugated to FITC or CY3 secondary (1:1000) (IF) (Jackson Laboratories) and antibody (Santa Cruz Biotechnology) and then coupled to protein-A Sepharose. peroxidase goat anti-mouse IgG (1:1000) (WB) (DakoCytomation). Immunoprecipitated proteins were analyzed for the presence of Myc-ARX by Western immunoblotting using mouse anti-Myc HRP-conjugated antibody and Transfection studies V5-IPO13 was performed with rabbit anti-V5 antibody followed by goat anti- rabbit HRP-conjugated secondary antibody. Cell lysates were prepared using Cells were plated onto sterile glass coverslips at 1×106 cells per well in a lysis buffer (120 mM NaCl, 50 mM Tris–HCl (pH 8.0), 0.5% Nonidet-P40 (NP- 5 six-well plate for HEK 293T cells and at 2×10 cells per well for SH-SY5Y and 40), 1X protease inhibitor cocktail (Sigma), 1 mM Na3VO4, 1 mM NaF, 1 mM PC12 cells. Routinely, cells were transfected with 1–2 μg of plasmid DNA using phenylmethylsulfonyl flouride), clarified by centrifugation (15 min, 13,000 g), lipofectamine 2000 (Invitrogen) following manufacturers instructions. Cells and precleared with protein-A Sepharose. Aliquots of extracts were immuno- were fixed at 24 and 48 or 72 h after transfection with 3.7% formaldehyde–PBS precipitated overnight at 4 °C. The lysates were then incubated with 40 μlof before mounting onto slides with Dapi-anti-fade mounting media visualized by protein-A Sepharose for 1 h at 4 °C before being washed four times with 0.4 ml standard fluorescence microscopy (Olympus Australia) or stored at 4 °C until of wash buffer (500 mM NaCl, 20 mM Tris–HCl (pH 8.0), 1 mM EDTA, 0.5% probed with antibodies (see below). NP-40) and eluted in SDS loading buffer (62.5 mM Tris–HCl (pH 6.8), 2% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.001% bromophenol blue). Western immunoblotting Lysates from transfected HEK 293T cells producing either Myc-ARX alone or V5-IPO13 alone were used as controls. 293T cells transfected with 1 μg Myc∼ARX constructs were harvested and cell lysates were made for coimmunoprecipitation. Lysates were then run in Statistical analysis replicates (3 μl per lane), subjected to SDS–PAGE, transferred to nitrocellulose, and analyzed by immunoblotting. Signal was detected via chemiluminesence. All data are reported as mean±SE for the percentage of transfected cells with abnormal subceullular localization or TUNEL. These data were determined Immunoflorescence and microscopy from a minimum of three separate transfection reactions, with the number of transfected cells counted per construct indicated in each figure legend. Fixed cells were permeabilized in 0.2% (v/v) Triton PBS for 5 min. Differences in the percentage of transfected cells with abnormal subceullular Blocking of nonspecific binding of the secondary antibody was achieved localization or TUNEL within each experiment were analyzed by one-way routinely by addition of 5% goat sera in 1% BSA in Tris-buffered saline and ANOVA using least-significant difference post hoc test for all pairwise 0.5% Tween (TBS-T) before incubation of the primary and secondary comparisons when significance was reached. p <0.05 was considered antibodies diluted in 1% BSA-TBS-T. Removal of excess block and antibody significant. was achieved with multiple washes of TBS-T. Nuclei were counterstained with DAPI (Molecular Probes, Invitrogen). Over 1000 transfected cells were counted for each construct from at least three different transfection reactions. The Acknowledgments formation of aggregates was assessed using standard florescence microscopy. The percentage of transfected cells with aggregates was determined as the The authors thank May Tan for assistance in the production number of cells containing aggregates divided by the total number of ARX- of mutagenic ARX homeodomain constructs and Kate positive cells. Dowling of the Faculty of Health Sciences, University of Adelaide for assistance with statistical analysis of the data. Cell death assay Desiree Cloosterman was supported by M.S. McLeod Post- SH-SY5Y cells transfected for 24 and 48 h were fixed and stained with anti- graduate Scholarship. This work was supported by a NHMRC ARX antibody as described above, with the exception of 0.1% (v/v) Triton in program grant. 70 C. Shoubridge et al. / Genomics 90 (2007) 59–71

References the dup24 mutation in the ARX (Aristaless related homeobox) gene, BMC Med, Genet 6 (2005) 16–19. [21] G. Turner, M. Partington, B. Kerr, M. Mangelsdorf, J. Gécz, Variable [1] D.S Herbst, J.R Miller, Nonspecific X-linked mental retardation II: expression of mental retardation, autism, seizures, and dystonic hand the frequency in British Columbia, Am. J. Med. Genet. 7 (1980) movements in two families with an identical ARX gene mutation, Am. J. 461–469. Med. Genet. 112 (2002) 405–411. [2] B. Kerr, G. Turner, J.C. Mulley, A.K. Gedeon, M. Partington, Non- [22] K. Gronskov, H. Hjalgrim, I. Nielsen, K. Brundum-Nielsen, Screening of specific X linked mental retardation, J. Med. Genet. 28 (1991) the ARX gene in 682 retarded males, Eur. J. Med. Genet. 12 (2004) 378–382. 701–705. [3] H.H Ropers, X-Linked mental retardation: Many genes for a complex [23] S.G. Frints, G. Froyen, P. Marynen, D. Willekens, E. Legius, J.P. Fryns, disorder, Curr. Opin. Genet. Dev. 16 (2006) 1–10. Re-evaluation of MRX36 family after discovery of an ARX gene mutation [4] J. Gécz, D. Cloosterman, M. Partington, ARX: a gene for all seasons, Curr. reveals mild neurological features of Partington syndrome, Am. J. Med. Opin. Genet. Dev. 16 (2006) 308–316. Genet. 112 (2002) 427–428. [5] M. Partington, G. Turner, J. Boyle, J. Gécz, Three new families with [24] L. Brown, S. Brown, Alanine tracts: the expanding story of human illness X-linked mental retardation caused by the 428-451dup(24(bp) mutation in and trinucleotide repeats, Trends Genet. 20 (2004) 51–58. ARX, Clin. Genet. 66 (2004) 39–45. [25] A. Albrecht, S. Mundlos, The other trinucleotide repeat: polyalanine [6] P. Strømme, M.E. Mangelsdorf, M. Shaw, K. Lower, S.M. Lewis, H. expansion disorders, Curr. Opin. Genet. Dev. 15 (2005) 285–293. Bruyere, et al., Mutations in the human ortholog of Aristaless cause [26] V. Shanmugam, P. Dion, D. Rochefort, J. Laganie`re, B. Brais, G.A. X-linked mental retardation and epilepsy, Nat. Genet. 30 (2002) Rouleau, PABP2 polyalanine tract expansion causes intranuclear inclu- 441–445. sions in oculopharyngeal muscular dystrophy, Ann. Neurol. 48 (2000) [7] K. Kitamura, M. Yanazawa, N. Sugiyama, H. Miura, A. Iizuka-Kogo, M. 798–802. Kusaka, et al., Mutation of ARX causes abnormal development of [27] B.M. van der Sluijs, B.G.M. van Engelen, L.H. Hoefsloot, Oculophar- forebrain and testes in mice and X-linked lissencephaly with abnormal yngeal muscular dystrophy (OPMD) due to a small duplication in the genitalia in humans, Nat. Genet. 32 (2002) 359–369. PABPN1 gene, Hum. Mut. 21 (2003) 553–558. [8] E. Sherr, The ARX story (epilepsy, mental retardation, autism, and cerebral [28] E. Uyama, T. Tsukahara, K. Goto, Y. Kurano, M. Ogawa, Y.J. Kim, et malformations): one gene leads to many phenotypes, Curr. Opin. Ped. 15 al., Nuclear accumulation of expanded PABP2 gene product in (2003) 567–571. oculopharyngeal muscular dystrophy, Muscle Nerve 23 (2000) [9] H. Miura, M. Yanazawa, K. Kato, K. Kitamura, Expression of a novel 1549–1554. aristaless related homeobox gene 'Arx' in the vertebrate telencephalon, [29] S. Caburet, A. Demarez, L. Moumne', M. Fellous, E. De Baere, R.A. diencephalon and floor plate, Mech. Dev. 65 (1997) 99–109. Veitia, A recurrent polyalanine expansion in the transcription factor [10] H.M. El-Hodiri, X.L. Qi, D.W. Seufert, The Xenopus arx gene is expressed FOXL2 induces extensive nuclear and cytoplasmic protein aggregation, in the developing rostral forebrain, Dev. Genes. Evol. 212 (2003) Med. Genet. 41 (2004) 932–936. 608–612. [30] A. Albrecht, U. Kornak, A. Boddrich, K. Suring, P. Robinson, A. Stiege, A [11] T. Melkman, P. Sengupta, Regulation of chemosensory and GABAergic molecular pathogenesis for transcription factor associated poly-alanine motor neuron development by the C. elegans Aristaless/Arx homolog tract expansions, Hum. Mol. Genet. 13 (2004) 2351–2359. alr-1, Development 132 (2005) 1935–1949. [31] T. Bachetti, I. Matera, S. Borghini, M. Duca, R. Ravazzola, I. Ceccherini, [12] T. Bienvenu, K. Poirier, G. Friocourt, N. Bahi, D. Beaumont, F. Distinct pathogenetic mechanisms for PHOX2B associated polyalanine Fauchereau, et al., ARX, a novel Prd-class-homeobox gene highly expansions and frameshift mutations in congenital central hypoventilation expressed in the telencephalon, is mutated in X-linked mental retardation, syndrome, Hum. Mol. Genet. 14 (2005) 1815–1824. Hum. Mol. Genet. 11 (2002) 981–991. [32] I.M. Nasrallah, J.C. Minarcik, J.A. Golden, A polyalanine tract expansion [13] M. Kato, S. Das, K. Petras, K. Kitamura, K. Morohashi, D.N. Abuelo, in Arx forms intranuclear inclusions and results in increased cell death, et al., Mutations of ARX are associated with striking pleiotropy and J. Cell Biol. 167 (2004) 411–416. consistent genotype-phenotype correlation, Hum. Mut. 23 (2004) [33] G. Friocourt, K. Poirer, S. Rakic, J.G. Parnavelas, J. Chelly, A role 147–159. for ARX in cortical development, Eur. J. Neurosci. 23 (2006) [14] E. Colombo, R. Galli, G. Cossu, J. Gécz, V. Broccoli, Mouse orthologue of 869–876. arx, a gene mutated in several x-linked forms of mental retardation and [34] J.M. Mingot, R.K. Kostka, E. Hartmann, D. Gorlich, Importin 13: a epilepsy, is a marker of adult neural stem cells and forebrain gabaergic novel mediator of nuclear import and export, EMBO J. 21 (2001) neurons, Dev. Dyn. 231 (2004) 631–639. 3685–3694. [15] K. Poirer, H. Van Esch, G. Friocourt, Y. Saillour, N. Bahi, S. Backer, et al., [35] J.E. Ploski, M.K. Shamsher, A. Radu, Paired-type homeodomain Neuroanatomical distribution of ARX in brain and its localization in transcription factors are imported into the nucleus by Karyopherin 13, GABAergic neurons, Mol. Brain Res. 122 (2004) 35–46. Mol. Cell. Biol. 24 (2004) 4824–4834. [16] P. Collombat, J. Hecksher-Sørensen, V. Broccoli, J. Krull, I. Ponte, T. [36] B. Utsch, K. Becker, D. Brock, M.J. Lentze, F. Bidlingmaier, M. Ludwig, Mundinger, et al., The simultaneous loss of Arx and Pax4genes promotes A novel stable polyalanine [poly(A)] expansion in the HOXA13 gene asomatostatin-producing cell fate specification at the expense of the α-and associated with hand-foot-genital syndrome: proper function of poly(A)- β-cell lineages in the mouse endocrine pancreas, Development 132 (2005) harbouring transcription factors depends on a critical repeat length? Hum, 2969–2980. Genet 110 (2002) 488–494. [17] P. Collombat, A. Mansouri, J. Hecksher-Sørensen, P. Serup, J. Krull, G. [37] O. McKenzie, I. Ponte, M. Manglesdorf, M. Finnis, G. Colasante, C. Gradwohl, et al., Opposing actions of Arx and Pax4 in endocrine pancreas Shoubridge, S. Stifani, J. Gecz, V. Broccoli, Aristaless-related homeobox development, Genes Dev. 17 (2003) 2591–2603. gene, the gene responsible for West syndrome and related disorders, is a [18] P. Strømme, M.E. Mangelsdorf, I.E. Scheffer, J. Gécz, Infantile spasms, Groucho/transducin-like enhancer of split dependent transcriptional repres- dystonia, and other X-linked phenotypes caused by mutations in Aristaless sor, Neuroscience (in press), doi:10.1016/j.neuroscience.2007.01.038. related homeobox gene, ARX, Brain Dev. 24 (2002) 266–268. [38] A. Caldo, F.M.S. Tome, B. Brais, G.A. Rouleau, U. Kuhn, E. Wahle, et al., [19] G. Wohlrab, G. Uyanik, C. Gross, U. Hehr, J. Winkler, B. Schmitt, Nuclear inclusions in oculaopharyngeal muscular dystrophy consists of et al., Familial West syndrome and dystonia caused by an Aristaless poly(A) binding protein 2 aggregates which sequester poly(A) RNA, Hum. related homeobox gene mutation, Eur. J. Pediatr. 164 (2005) Mol. Genet. 9 (2000) 2321–2328. 326–328. [39] M.W. Becher, J.A. Kotzuk, A.H. Sharp, S.W. Davies, G.P. Bates, D.L. [20] M. Stepp, A.L. Cason, M. Finnis, M. Manglesdorf, E. Holinski-Feder, D. Price, et al., Intranuclear neuronal inclusions in Huntington’s disease and Macgregor, et al., XLMR in MRX families 29, 32, 33 and 38 results from dentatorubral and pallidoluysian atrophy: correlation between the density C. Shoubridge et al. / Genomics 90 (2007) 59–71 71

of inclusions and IT15 CAG triplet repeat length, Neurobiol. Dis. 4 (1998) protein Vsx-1 independent of SUMO-1 modification, Proc. Natl. Acad. 387–397. Sci. USA 98 (2001) 5602–5607. [40] M.T. Berciano, N.Y. Villagra, J.L. Ojeda, J. Navascues, A. Gomes, M. [44] K. Furukawa, T. Iioka, M. Morishita, A. Yamaguchi, H. Shindo, H. Lafarga, et al., Oculopharyngeal muscular dystrophy-like nuclear inclu- Namba, et al., Functional domains of paired-like homeoprotein Cart1 and sions are present in normal magnocellular neurosecretory neurons of the the relationship between dimerization and transcription activity, Genes hypothalamus, Hum. Mol. Genet. 13 (2004) 829–838. Cells 7 (2002) 1135–1147. [41] L. Brown, M. Paraso, R. Arkell, S. Brown, In vitro analysis of partial loss- [45] Y.A. Zhang, A. Okada, C.H. Lew, S.K. McConnell, Regulated nuclear of-function ZIC2 mutations in holoprosencephaly : alanine tract expansion trafficking of the homeodomain protein Otx1 in cortical neurons, Mol. modulates DNA binding and transactivation, Hum. Mol. Genet. 14 (2005) Cell. Neurol. 19 (2002) 430–446. 411–420. [46] G. Uyanik, L. Aigner, P. Martin, C. Gross, D. Neumann, H. Marschner- [42] K. Woods, M. Cundall, J. Turton, K. Rizotti, A. Mehta, R. Palmer, et al., Schafer, et al., ARX mutations in X-linked lissencephaly with abnormal Over-and underdosage of SOX3 is associated with Infundibular hypoplasia genitalia, Neurology 61 (2003) 232–235. and hypopituitarism, Am. J. Hum. Genet. 76 (2005) 833–849. [47] K.R. Oldenburg, K.T. Vo, S. Michaelis, C. Paddon, Recombination- [43] A.L. Kurtzman, N. Schechter, Ubc9 interacts with a nuclear localization mediated PCR-directed plasmid construction in vivo in yeast, Nucleic signal and mediates nuclear localization of the paired-like homeobox Acids Res. 25 (1997) 451–452.