DIFFERENTIALLY SPLICED ISOFORMS OF FAT1 ARE ASYMMETRICALLY DISTRIBUTED WITHIN MIGRATING CELLS

Cadherin FAT1 is localized along the leading edge of mammalian cells and is necessary for polarization and directed migration. It is essential for maintenance of the complex cytoarchitecture of the glomerular filtration barrier within the kidney. In this study, three novel splice isoforms of FAT1 with important functional differences in comparison with wild-type FAT1, FAT1(WT), were identified. The novel variants contained additional short peptide sequences at a specific site of the cytoplasmic domain (+12 or +32 or +8 amino acids, the latter resulting in a premature stop codon). FAT1(+12) was expressed in all peripheral tissues together with FAT1(WT), whereas FAT1(+32) and -(+8TR) were brain-specific. At the subcellular level, exclusively FAT1(WT) was localized along the cellular leading edge, whereas spliced FAT1 isoforms were confined to intercellular junctions. A shift of FAT1(WT) expression toward a predominance of FAT1(+12) was observed in migratory versus quiescent cells. A similar shift was observed in vivo when glomeruli from healthy individuals were compared with those from patients affected by glomerulonephritis. At the molecular level, the differential subcellular localization of FAT1 isoforms was mediated by a novel region harboring a phosphotyrosine-binding-like motif (DN_XYH), which was disrupted by the peptide inserts in the alternative splice variants. Overexpression of FAT1(WT) or specific knockdown of spliced FAT1 isoforms resulted in formation of cellular protrusions or increased wound healing, respectively. In summary, FAT1(WT) is the only FAT1 isoform located along the cellular leading edge. Only FAT1(WT) is up-regulated in migration, induces cellular process formation when overexpressed, and is necessary for efficient wound healing.

Cadherin FAT1 is localized along the leading edge of mammalian cells and is necessary for polarization and directed migration. It is essential for maintenance of the complex cytoarchitecture of the glomerular filtration barrier within the kidney. In this study, three novel splice isoforms of FAT1 with important functional differences in comparison to wildtype FAT1, FAT1(WT), were identified. The novel variants contained additional short peptide sequences at a specific site of the cytoplasmic domain (+12 aa or +32 aa or +8 aa, the latter resulting in a premature stop codon). FAT1(+12) was expressed in all peripheral tissues together with FAT1(WT), whereas FAT1(+32) and (+8TR) were brainspecific. At the subcellular level, exclusively FAT1(WT) was localized along the cellular leading edge, whereas spliced FAT1 isoforms were confined to intercellular junctions. A shift of FAT1(WT) expression towards a predominance of FAT1(+12) was observed in migratory versus quiescent cells. A similar shift was observed in vivo when glomeruli from healthy individuals were compared to those from patients affected by glomerulonephritis. At the molecular level, the differential subcellular localization of FAT1 isoforms was mediated by a novel region harboring a PTB-like motif (DN_xYH) which was disrupted by the peptide inserts in the alternative splice variants. Overexpression of INTRODUCTION FAT cadherins are large ∼500 kD transmembrane proteins with an extracellular domain containing 34 cadherin motifs and a cytoplasmic domain that harbors several protein interaction motifs (1). So far, Ena/VASP, Homer 3 and beta-catenin has been reported to interact with the FAT1 cytoplasmic domain (2)(3)(4)(5). In drosophila, the orthologs of mammalian FAT1, dfat (fl) and dfatlike (ftl), have been implicated in the establishment of planar cell polarity and formation of tracheal epithelia, respectively (6,7). We and others have shown that mammalian FAT1 is localized along the leading edge, filopodial protrusions, and intercellular junctions (2,4). As demonstrated by knockdown experiments, mammalian FAT1 is essential for cellular polarization and directed cell migration (2,4,5). FAT1 regulates actin dynamics at least in part via a direct interaction with Ena/VASP proteins (2). Consistently, genetic inactivation of FAT1 in mice leads to loss of podocyte foot processes in the kidney, which are a highly dynamic actin-based structures forming a specialized inter-cellular junction, termed slitdiaphragm (8)(9)(10)(11). In the present study, several novel splice isoforms of FAT1 were identified and characterized.
Antibodies FAT1(+12)-specific polyclonal antiserum was raised in two rats and one guinea pig against a s y n t h e t i c p e p t i d e " C N K E E S L A A PDLSKPRGYH" N'-terminally fused to the antigenic protein KLH. Antiserum was centrifuged and used in dilutions of 1/50 for IF and of 1/500 for WB. Control preadsorption was done overnight at 4 C° (WB: 8 µg peptide/µl pure serum, IF: 800 µg peptide/µl pure serum). Other primary antibodies were used at indicated dilutions: previously described mouse α -FAT1(cyto) (2) (IF: 1/300 -1/3000, WB: 1/5000), mouse α-Flag (M2, Sigma, 1/2000-3000). Secondary antibodies conjugated to fluorescence dyes Cy2 and Cy3 were from J a c k s o n ImmunoResearch Laboratories, phalloidin-TRITC from Sigma and Mitotracker red CMXRos from Molecular Probes.
Cell culture and IF Cell lines were purchased from American Type Culture Collection: NRK-52E (#CRL-1571), RAT2 (#CRL-1764), COS-7 (#CRL-1651), HEK293T (#CRL-11268) and cultured in DMEM supplemented with 10% fetal calf serum and 200 U/ml of penicillin and streptomycin. A previously described immortalized podocyte cell line was cultured as described (15). Transfection was performed using Lipofectamin 2000 (Invitrogen) for NRK-52E cells and using F u g e n e -6  (Roche) or standard calcium phosphate method for COS-7 cells. IF studies were performed according to standard protocols. Briefly, cells grown on glass coverslips (collagen-coated for COS-7 cells) were fixed in 4% paraformaldehyde for 15 min and permeabilized with 0.2% Triton X-100 for 2 min (permeabilization omitted for experiments shown in Fig. 7B, C) or in methanol (-20° C) for 15 min and subsequently blocked with 10% donkey serum. Nuclei were stained with Hoechst 33342 DNA-dye (1/10 000, 10 min.). Coverslips were mounted on glass slides with Mowiol® (Calbiochem/Merck). Images were acquired with a Leica DM-IRBE inverted microscope connected to a digital camera and Openlab (Improvision) software or with a Leica DM-IRBE-TSP inverted confocal laser-scanning microscope connected to Leica confocal imaging software. Preparation for publication was done using Image J 1.33 U (NIH, USA) and Photoshop (Adobe) software.

Western blot
For Western blot of recombinant FAT1-mito, COS-7 cell lysates were processed by 10 % SDS-PAGE, (10 µg/lane), transferred on nitrocellulose membranes and bands were visualized using standard ECL (Western Lightning Plus, Cat.# NEL105, Perkin Elmer). For detection of endogenous FAT1, NRK-52E cells were grown on dishes of 15 cm diameter and processed by 5 % SDS-PAGE, (50-100 µg lysate/lane; in equal amounts for each group in an experimental series). Experiments were performed at least three times and representative immunoblots are shown.
Time lapse microscopy COS-7 cells grown on round coverslips were transfected with FAT1(WT)-mito for 36 h, stained with Mitotracker® red (1/500, 30 min), mounted on an air tight inverted chamber containing preconditioned medium and filmed at 37° C.
Mapping of the PTB-like motif FAT1-mito constructs were transfected into mammalian cells and assessed for presence of mitochondrial redistribution into the cell periphery. For each construct at least 100 cells from three independent experiments were examined and classified.

Statistical analysis
Quantitative data from standard RT-PCR band analysis, real-time-PCR quantification, and morphometric cell evaluation were statistically a n a l y z e d u s i n g P r i s m s o f t w a r e (www.GraphPad.com).

RESULTS
Alternative splicing of the FAT1 cytoplasmic domain A comparison of the published mouse FAT1 mRNA sequences (Genbank acc.# AY256848 versus AJ250768) indicated the presence of a putative alternative splice site located at cytoplasmic nucleotide (nt) 529/530 (Fig. 1C, grey shaded sequences). Using RT-PCR, three splice inserts from mouse tissues or rat cell lines were cloned and sequenced. The corresponding exon-intron properties were established by comparative genomic sequence analysis (NC_000074.2). The mouse gene locus, mouse mRNA structure, and mouse mRNA/predicted protein sequences of the splice variants are shown in Fig. 1A-C, respectively. Three exons (cy1-3) encoded the previously described wildtype FAT1 cytoplasmic domain, FAT1(WT) (AY256848). Three alternatively spliced exons (A-C) were identified (Fig. 1A), which were invariantly directly flanked by introns carrying the splice-donor-acceptor consensus sequence 5' ag… …gt 3' (not shown). These exons were also present in the rat genome with 100 % nucleotide conservation (NC_005115) (not shown). showing 100 % nucleotide insert identity with the mouse and rat sequences, and of a truncated form similar to mouse FAT1(+8TR) (AY598440). The zebrafish genome contained a sequence corresponding to mouse FAT1 exon B that was also flanked by the intron splice consensus sequence 5' ag… …gt 3' (data not shown).
FAT1 isoform expression was investigated in various mouse tissues (Fig. 1D). FAT1(WT) mRNA was detected in all tissues. FAT1(+32) and (+8TR) were expressed specifically within the central nervous system where they predominated over FAT1(WT). FAT1(+12) was detected in highest amounts in lung, and in trace amounts in all other tissues except brain (Fig.  1D', brighter exposure of D).
Differential subcellular localization of endogenous FAT1 isoforms To investigate the potential functional characteristics of FAT1 isoforms, immunofluorescent double-stainings of various cell lines were performed. Immunofluorescent staining of early confluent native NRK-52E cells with α-FAT1(+12) revealed a signal at intercellular junctions and the nucleus (Fig. 2F) which could be abolished by antigen-preadsorption (Fig. 2F').
A striking differential subcellular distribution of FAT1 isoforms was observed in subconfluent migrating NRK-52E cells: α -F A T 1 (cyto) staining was observed along cellular leading edges as well as along inter-cellular junctions, and in a proportion of nuclei ( Fig. 3A; green fluorescence), as described previously (2,4). By contrast, α-FAT1(+12) staining was completely absent at leading edges. α-FAT1(+12) staining was only detected along intercellular contacts and within the nuclei (Fig. 3A'; red fluorescence). A similar subcellular distribution of FAT1 isoforms could be observed in RAT2 and COS-7 cells (not shown). It was verified by RT-PCR that cultured NRK-52E cells expressed both FAT1(WT) and (+12) mRNAs (Fig. 3B). Because FAT1 is essential for glomerular epithelial cells of the kidney in-vivo (8), the expression of FAT1(WT) and (+12) mRNAs was evaluated in a murine conditionally immortalized podocyte cell line (15) (Fig. 3C-E''). The subcellular localization of FAT1 isoforms in this podocyte cell line and in NRK-52E cells was identical (Fig. 3D-E''). These data indicate that FAT1(WT) was the only isoform localized along the leading edge of lamellipodial protrusions, while spliced FAT1(+12) was confined to intercellular junctions.
Splicing of endogenous FAT1 is regulated Since alternative splicing occurred in a tissuespecific fashion in the mouse (Fig. 1D, D'), it was investigated if expression of FAT1 isoforms was differentially regulated under different culture conditions in NRK-52E cells. RT-PCR of proliferating subconfluent NRK-52E cells expressed predominantly FAT1(WT) mRNA, while quiescent cultures grown to confluence for 72 h predominantly expressed FAT1(+12) (Fig.  4A). Quantitative analysis of digitally analyzed band intensities revealed a statistically significant shift by a factor of 2.2 from a relative predominance of FAT1(WT) to a relative predominance of FAT1(+12) between proliferating and quiescent cells (p<0.002; Fig.  4A). These data were corroborated at the protein level by immunoblotting of cellular lysates derived from subconfluent, early confluent or quiescent NRK-52E cells. Again, a significant shift toward the expression FAT1(+12) was observed (n=3; Fig. 4B). In addition, down-regulation of total FAT1 protein in confluent cells was observed as reported previously (4). Subcellular FAT1 isoform expression was analyzed by immunofluorescence. Consistent with the above described data, early confluent cultures expressed both FAT1(WT) and (+12) at intercellular junctions (Fig. 4C-C''), while in quiescent cells a down-regulation of FAT1(WT) was observed. After prolonged culture, both isoforms were eventually down-regulated (Fig.  4D-D'').
A PTB-like motif targets FAT1(WT) towards the leading edge and can be inactivated by alternative splicing The differential subcellular distribution of FAT1 splice isoforms indicated that this effect was mediated by a specific protein-protein interaction that was regulated by alternative splicing. Similar to endogenous FAT1(WT), ectopic expression of the FAT1(WT) cytoplasmic domain on mitochondrial outer leaflets (schematic in Fig.  5A) resulted in polarized mitochondrial redistribution towards the cellular leading edge (Fig. 2B-B''; Fig. 5B-B'', H), as described previously (2). Mitochondria are normally distributed in a perinuclear fashion close to the Golgi apparatus, as shown in untransfected cells in Fig. 5B', cell 1. Mitochondrial targeting of control proteins (i.e. the cytoplasmic domain of protocadherin Dachsous, Fig. 5C, or two repeats of the FK506-binding protein, Fig. 5D) did not result in mitochondrial redistribution, indicating that mitochondrial redistribution was a characteristic intrinsic to the wild-type FAT1 cytoplasmic domain. Thus, mitochondrial redistribution was a useful assay to map the domain that mediates the unique subcellular distribution of FAT1(WT) towards the leading edge. Various mutants of FAT1-mito including all splice variants (+12), (+32), and (+8TR) were tested ( Fig. 6A; splice site indicated by vertical bar). A quantitative grading system for the extent of distribution towards the leading edge was established and transfected cells were scored into one of three classes of either perinuclear (0 %), incomplete (<50 %) or complete (>50 %) mitochondrial redistribution towards the leading edge, respectively (exemplary cells are shown in Fig. 5E-L). Equal protein expression of each mutant was verified by immunoblotting (Fig.  5M). The statistical analysis for each mutant is shown in Fig. 6B (n=100 cells; 3 independent experiments each). FAT1(WT)mito mediated complete mitochondrial redistribution towards the leading edge in 90% of the cells, while mitochondrial distribution remained unchanged in cells transfected with any of the splice variants (+12) and (+32) similar to controls (Dhs and 2xFKBP). As predicted, FAT1(+8TR)mito was not targeted to mitochondria owing to the premature stop codon, resulting in its diffuse cytoplasmic distribution (Fig. 5K). These data show that polarized redistribution of FAT1 was abolished in all spliced FAT1 isoforms. To map the essential amino acids mediating redistribution of FAT1(WT), C-terminal and Nterminal deletion constructs of FAT1(WT)mito were analyzed (Fig. 6A Fig. 6C. Alignment for consensus sequence homology of mouse, chicken and zebrafish FAT1 together with mammalian FAT3 identified a conserved domain at aa 158-204, as indicated by a box in Fig. 6A and a grey bar in Fig. 6C. Of note, the identified motif was the only motif of vertebrate FAT1 that was conserved in its drosophila homologue ftl, suggesting an important evolutionary conserved function (Fig. 6D).
The alternative splice site was located immediately within the identified minimal motif separating 4 highly conserved amino acids 175 DNxYH 179 similar to a phosphotyrosinebinding (PTB) motif ( Fig. 6C; black vertical line indicating the splice site). Indeed, when Y178 was mutated to glutamic acid (Y178E) mitochondrial redistribution was totally abolished (Fig. 6A, B). Substitution of Y178 to phenylalanine (F), a residue resembling a tyrosine residue that cannot be phosphorylated, had no effect (Y178F, Fig. 6A, B). In agreement with this, no evidence for tyrosinphosphorylation of the wild-type FAT1 cytoplasmic domain was observed (not shown). Mutation of a second conserved tyrosine residue Y200 (to both E and F) had no effect (not shown). These findings support the notion that 175 DNxYH 179 represents a minimal proteinprotein interaction core motif that is not regulated by tyrosin-phosphorylation. Indeed, the identified motif shared characteristics with the Disabled1 (Dab1)-PTB-binding motif of C. e l e g a n s protein kinase C 3 (PKC3) (xxDNxxFHxx) which does not require phoshphorylation of the essential tyrosine residue (19,20). To investigate if the motif is inactivated by insertion of an irrelevant sequence or by insertion of a specific inactivating sequence, three additional FAT1-mito constructs were generated. Insertion of only 3 irrelevant amino acids (AAS) at the splice site (FAT1(+3)mito) significantly reduced mitochondrial redistribution while insertion of longer irrelevant fragments e.g. of 98 aa (FAT1(+98)mito) totally abolished redistribution (Fig. 6A, B). Furthermore, the functionality of the splice isoform FAT1(+12)mito could be rescued by insertion of the last three aa (DDN) derived from the N'terminal half of the FAT1 cytoplasmic domain (see Fig. 6C) next to the essential tyrosine Y178 reconstituting the essential core sequence of the P T B -l i k e m o t i f ( s e q u e n c e : .. 174 DDNeslaapdlskprDDNGYHWD.., Fig. 6A,  B). These experiments show that the splice site was located at a functionally critical position. It was concluded that splice inserts (+12) and (+32) sterically disrupted a PTB-like motif.
Association of cellular process-formation and FAT1(WT) As shown in Fig. 5B and H, the identified PTBlike motif mediated redistribution of recombinant FAT1(WT) in a polarized fashion towards the cellular leading edge. To verify this finding, movements of FAT1(WT)mito decorated mitochondria were observed for 12 hours using time-lapse video microscopy in transiently transfected COS-7 cells (supplemental movie S1; green cell, transfected with FAT1(WT)mito; red cell, non-transfected control). FAT1(WT)mitodecorated mitochondria were rearranged in a highly dynamic fashion into cellular processes towards the leading edge of migrating cells. The majority of FAT1-decorated mitochondria were redistributed into the direction of the overall cellular movement. Redistribution of FAT1(WT)mito-decorated mitochondria was closely associated with polarized process formation. To test if FAT1 was also quantitatively associated with the formation of cellular processes, a partial full-length clone of wild-type FAT1 lacking the N-terminal 34 cadherin repeats was expressed within the cellular plasma membrane (FAT1(WT)∆Cadh) (Fig. 7A). As a control, a clone lacking the identified PTB-like motif was generated (FAT1(∆160-188)∆Cadh; see Fig. 6A, C for sequence). Both constructs efficiently localized to the plasma membrane as shown by indirect immunofluorescence staining of transiently transfected COS-7 cells omitting the permeabilization step: Staining for the extracellular FLAG-epitope yielded a signal (green), while co-staining for the intracellular FAT1-cytoplasmic domain by α-FAT1(cyto) (red) was negative (Fig. 7B, C). Importantly, FAT1(WT)∆Cadh was localized towards the cellular leading edge and into cellular processes similar to FAT1(WT)mito (Fig. 7B). This polarized redistribution was clearly abolished in the mutated clone, FAT1(∆160-188)∆Cadh (Fig.  7C) indicating that FAT1(WT)∆Cadh was not enriched as a result of cellular artifacts.
The FAT1(WT) isoform is sufficient for directed cell migration The increased number of cellular protrusions in FAT1(WT)∆Cadh transfected cells suggested a differential role of the FAT1 splice isoforms in cell migration. To test this hypothesis, the expression of FAT1 splice isoforms was selectively silenced using RNA interference (16). NRK-52E cells were transduced with a lentivirus (KD(+12)) expressing a shRNA sequence specific for the splice inserts common only to the FAT1(+12) and (+32) isoforms. Expression of FAT1 isoforms was selectively silenced as demonstrated by immunoblotting of cellular lysates and by RT-PCR of total RNA extracts (Fig. 8A,B). Splice specific knockdown did not affect expression of wild-type FAT1. Knockdown of all FAT1 isoforms was achieved using a previously described lentivirus (KD(FAT1)) (2). These results also confirmed the specificity of the α-FAT1(+12) antiserum. NRK-52E cell monolayers were transduced with control virus (KD control), or virus specific for FAT1 splice isoforms (KD(+12)) or for total FAT1 (KD(FAT1)) and subjected to a wound assay. Lentiviral transduction efficiency was adjusted to 90% transduced cells, verified by coexpression of a viral eGFP reporter gene (2). Ten hours after wounding, would closure was significantly impaired in FAT1 deficient cells (KD(FAT1)) compared to controls (KD control), as described previously (2,4). In contrast, wound closure was improved in cells that only expressed wild-type FAT1 (KD(+12)) compared to controls by a factor of 1.7 (mean remaining wound diameter of 106 µm vs. 180 µm). These results confirm that only the wild-type FAT1 isoform at the cellular leading edge is essential for directed cell migration.

DISCUSSION
In this study, three alternative splice isoforms of FAT1 with important functional differences from wild-type FAT1 were identified. The data show that endogenous FAT1(WT) represents the isoform through which FAT1 mediates functions of directed cell migration at the leading edge that have been previously described (2). This notion is supported by the fact that only FAT1(WT) was associated with cellular process formation and was essential for directed cell migration in a wound healing assay. In addition, FAT1(WT) was the predominant isoform in migrating cells in vitro and in healthy glomeruli with intact podocyte foot processes in vivo. Furthermore, cells are able to modulate their FAT1(WT) to FAT1(+12) expression ratios in response to extracellular signals (quiescent versus migratory). The isoform-specific antisera described in this work will therefore be an important tool to study specific FAT1 functions.
The association of endogenous FAT1(WT) with cellular leading edges was mapped to a novel PTB-like motif ( 175 DNxYH 179 ). This motif was functionally inactivated in alternatively spliced FAT1 isoforms. As shown by insertion of irrelevant sequences into the splice site, this mechanism was sterical in nature rather than dependent on a specific additional sequence. Consequently, FAT1(+12) and (+32) are predicted to be identical in function, although it cannot be ruled out that the sequence exerts other functions not detected in our analysis. Of note, the PTB-like motif and its surrounding region was highly conserved from drosophila ftl to mammals, suggesting its importance for FAT1 function, and supporting the notion that ftl is the true ortholog of mammalian FAT1 (6). It remains to be determined whether the PTB-like motif directly mediates targeting of FAT1(WT) to the leading edge, e.g. by vesicle transport. Alternatively, the identified PTB-motif may be implicated in the transduction of cellular polarization signals via asymmetrical enrichment of wild-type FAT1 along the cellular leading edge. Indeed, asymmetrical enrichment in "cortical platforms" (21) as seen with FAT1(WT) is an established mode of transducing polarity signals in directed cell migration (22)(23)(24)(25)(26)(27). Major events regulating directed cell migration are actin reorganization and polarized capture of microtubuli (21). Recombinant FAT1 was redistributed into similar locations as the cytosolic microtubular bridging protein or kinesin-transported cargos suggesting that the PTB-like motif interacts at the leading edge with the microtubular system (25,(28)(29)(30)(31).
At the nucleotide level, a striking evolutionary sequence conservation of the splice inserts including their surrounding introns was observed. This suggests the presence of important regulatory sequences for the splicing machinery. In a recent study in human patients, the Cterminus of the FAT1 gene has been shown to harbor alleles linked to bipolar disease (32). In the present study, mRNAs encoding spliced isoforms FAT1(+32) and (+8TR) were expressed exclusively within the CNS. One of the SNPs associated with bipolar disease lies within the PTB-like motif at S167 but does not result in a change of the protein sequence (TCC to TCT; rs1298865) (32). However, such silent point mutations may influence mRNA processing (33). Susceptibility to bipolar disease may be conferred by non-coding mutations that change the ratio between the FAT1 splice isoforms. Within the renal glomerulus, upregulation of the FAT1(+12) splice isoform was associated with diseases that go along with heavy protein-loss into the urine, namely minimal change disease, membranous glomerulonephritis, and focalsegmental glomerulosclerosis. These glomerulopathies are associated with loss of the complicated structure of the actin cytoskeleton (termed foot process effacement) and formation of adherens junctions. No glomerular leukocyte infiltration or other glomerular cell proliferation occurs in minimal change disease and membranous glomerulonephritis excluding artifacts from other cell types. Our data therefore suggest that FAT1 isoform expression is also regulated in-vivo and that loss of podocyte foot processes is associated with a relative increase in expression of the junction-associated isoform FAT1(+12). These observations are of potential diagnostic or even therapeutic value.