PDLIM5 links kidney anion exchanger 1 (kAE1) to ILK and is required for membrane targeting of kAE1

Anion exchanger 1 (AE1) mediates Cl−/HCO3− exchange in erythrocytes and kidney intercalated cells where it functions to maintain normal bodily acid-base homeostasis. AE1’s C-terminal tail (AE1C) contains multiple potential membrane targeting/retention determinants, including a predicted PDZ binding motif, which are critical for its normal membrane residency. Here we identify PDLIM5 as a direct binding partner for AE1 in human kidney, via PDLIM5’s PDZ domain and the PDZ binding motif in AE1C. Kidney AE1 (kAE1), PDLIM5 and integrin-linked kinase (ILK) form a multiprotein complex in which PDLIM5 provides a bridge between ILK and AE1C. Depletion of PDLIM5 resulted in significant reduction in kAE1 at the cell membrane, whereas over-expression of kAE1 was accompanied by increased PDLIM5 levels, underscoring the functional importance of PDLIM5 for proper kAE1 membrane residency, as a crucial linker between kAE1 and actin cytoskeleton-associated proteins in polarized cells.

determinants 21 including a PDZ (Post-synaptic density protein, PSD-95; Drosophila disc large tumour suppressor, Dlg; Zonula occludens-1, ZO-1) binding motif formed by the last 4 residues (A 908 MPV 911 ) that has not been previously characterized. The M909T mutation or deletion of the motif results in trafficking defects of the mutant proteins in polarized kidney epithelial cells, implicating this motif in proper membrane residency of kAE1 8 .
Several proteins have so far been reported to associate with AE1C, and disruption of these interactions results in abnormal cellular location of kAE1 [21][22][23] . We report here a novel C-terminal binding partner, PDLIM5, also called Enigma homolog protein (ENH) and belonging to the Enigma subfamily of PDZ-LIM proteins. PDLIM5 contains one N-terminal PDZ and three C-terminal LIM domains. PDZ domains are a type of common structural domain of 80-90 amino acids and are known as organizers of protein complex assembly. Studies of PDZ proteins in various organisms have provided evidence of their involvement in cellular sorting and targeting (including basolateral membrane targeting) of their binding partners 24 .

Results
PDLIM5 is a potential binding partner for kAE1. Co-immunoprecipitation (IP) assays were performed using lysates from MDCK-Δ pMEP-GFP-kAE1 cells with anti-AE1 (Bric-170) as immunoprecipitating antibody and anti-CD63, which does not recognize dog CD63, as a negative control antibody. Mass Spectrometry identified 480 proteins which were classified into 26 functional enrichment categories (Fig. 1a, Supplementary Table 1). Among these potential binding candidates, PDZ domain-containing PDLIM5 was chosen for further investigation.
PDLIM5 expression in kidney. Lee et al. have recently reported gene expression patterns in each of 14 renal tubule segments in rat kidney using RNA-sequencing coupled with classic tubule microdissection 25 . There, PDLIM5 transcripts were found in almost every segment with enhancements in the medullary long descending limb of Henle's loop, connecting tubule, cortical collecting duct (CD) and inner medullary collecting duct. Dual-immunostaining for kAE1 and PDLIM5 in available sections of human kidney cortex demonstrated known kAE1 localization to the basolateral surface of α -IC in CD (Fig. 1b, panels b and e) while PDLIM5 was more extensively distributed, including proximal tubules and CD (panels a and d). Merged confocal optical sections suggested some co-localization of the two proteins in α -IC in the cortical collecting duct (panels c and f).
PCR products amplified from human kidney cDNA using specific primers displayed two clear bands of 1800 and 1400 bp (Fig. 2a) whose sequences correspond to isoforms a (ID: NP_006448.4) and b (ID: NP_001011513.3) in the NCBI database, respectively. PDLIM5a and PDLIM5b are splicing isoforms; both contain one N-terminal PDZ domain and three C-terminal LIM domains, but differ in the centre where PDLIM5b has 109 residues missing (panel b). In addition, both isoforms were identified in all kidney tissue/cell lysates examined (panel c) by Western blotting.
The PDZ binding motif in AE1C is important for direct PDLIM5 binding. To investigate direct protein interactions, we made GST-tagged AE1C-WT, AE1C-Δ 11 and AE1C-M909T proteins (Supplementary Figure 1a) as previously described 23 . Expression of tag-free full-length PDLIM5 caused degradation apart from a stable fragment of approximately 13 kD (Supplementary Figure 1b), which N-terminal sequencing showed to be the first 127 residues of PDLIM5 (designated PDLIM5-PDZ), which contains the entire PDZ domain of this protein. We were therefore able to use this domain for ELISA analysis. In addition, we attempted to express and purify a fragment containing the LIM domains of PDLIM5 but this also degraded rapidly during tag cleavage and purification steps.
Parallel ELISA analyses were then performed using AE1C-WT, AE1C-Δ 11 or AE1C-M909T GST fusion proteins incubated with GST-PDLIM5a or PDLIM5-PDZ recombinant proteins. AE1C-WT and AE1C-M909T, but not AE1C-Δ 11, were able to bind to both tagged PDLIM5a (Fig. 2, panel d) and untagged PDLIM5-PDZ (panel e). This indicates firstly that A 908 MPV and A 908 TPV at the far C-terminus of AE1 (wild-type and mutant respectively) are both PDZ-binding motifs and secondly that the PDZ domain of PDLIM5 is important for binding. PDLIM5 is required for membrane residency of kAE1 in kidney cells. Since loss of the last 11 residues of AE1 causes not just dRTA/kAE1 mistargeting but also abolishes PDLIM5 binding, we next assessed a possible role for PDLIM5 in kAE1 membrane targeting/retention in kidney cells. We first used siRNA oligonucleotides directed against the 3 rd exon of PDLIM5 (which encodes part of the PDZ domain in both long and short forms) to target endogenous PDLIM5 mRNA in HEK-Δ pMEP-GFP-kAE1 cells. This achieved approximately 70% reduction of endogenous PDLIM5a expression (Fig. 3a), while levels of PDLIM5b were better maintained at about 80%. PDLIM5 depletion resulted in a marked decrease (at least 70%) of kAE1 levels at the cell surface as assessed by biotinylation with accompanying but less marked reduction (~20%) in kAE1 levels in total cell lysates. In concert, confocal microscopy (Fig. 3b) demonstrated a similar reduction of membrane presence in knock down cells, with most kAE1 protein retained intracellularly, suggesting impaired membrane trafficking. As a non-radioactive alternative to pulse-chase analysis, we performed a time-course series at 2, 4, 6 and 8 hours after induction of kAE1 expression in HEK-Δ pMEP-eGFP-kAE1 cells depleted of PDLIM5 (Fig. 3c,d). Surface kAE1 was low in knockdown cells at all time-points, suggesting a defect of membrane targeting rather than achievement of membrane residency followed by internalisation.
A slightly lower level (~50%) of PDLIM5a depletion was achievable in MDCK kAE1-expressing cells, with no reduction in the level of PDLIM5b (Supplementary Figure 2). Interestingly, kAE1 levels were unaffected in the MDCK knock down samples, implying possible compensation by the PDLIM5b isoform for the loss of PDLIM5a in this system. Second, we examined induction of kAE1 membrane expression in HEK-Δ pMEP-eGFP-kAE1 cells. Steady state was reached at 16 hours (Supplementary Figure 3) and was accompanied by significantly increased total levels of PDLIM5 protein (Fig. 3e), of which the majority was the a-isoform. To differentiate between new PDLIM5 protein synthesis or its stabilization, we examined additional earlier timepoints by real-time quantitative PCR (RTqPCR) and Western blot (Fig. 3f). We found no significant change in mRNA levels for PDLIM5a (right panel), whereas PDLIM5 protein levels rose sequentially, suggesting that PDLIM5 is required to stabilise kAE1's membrane residency. PDLIM5, kAE1 and ILK are found in the same complex in human kidney. Pull downs using immobilized GST_AE1C-WT fusion protein and incubation with human kidney tissue lysates yielded both a and b isoforms of PDLIM5 (Fig. 4a, representative of three replicate experiments), confirming association of PDLIM5 with the AE1C domain. As integrin linked kinase (ILK) is reported to associate with kAE1 in HEK293 cells and also glomeruli (where a low level of kAE1 has been reported) 26,27 , we re-probed the same blot for ILK, with positive results (Fig. 4a). This suggests a possible multiprotein complex formed of kAE1, through its C-terminal domain, with PDLIM5 and ILK.
Similar results were obtained using HEK-Δ pMEP-GFP-kAE1 cell lysates (panel b). Finally, a blot overlay assay showed a direct interaction between PDLIM5 and ILK (panel c), but neither ELISA (panel d) nor blot overlay assays (panels e/f) yielded detectable ILK binding to AE1C or the isolated PDZ domain of PDLIM5. Together these data indicate that PDLIM5 forms a bridge between kAE1 and ILK. These results are further supported by bioinformatic analyses (as described in Methods) that suggest that the LIM domains in PDLIM5 are capable of interaction with the ankyrin repeats in ILK. and three LIM domains (grey) at N and C-terminus, respectively, but with a deletion of 109 residues in the middle of PDLIM5b. (c) Western blot analysis using anti-PDLIM5-ct detected both PDLIM5a and PDLIM5b isoforms in human kidney cytosol (HKC), HEK-Δ pMEP-eGFP-kAE1 (HEK) and MDCK-Δ pMEP-eGFP-kAE1 (MDCK) cell lysates. (d,e) ELISA plate coated with GST wild-type (GST_AE1C-WT) or mutant (GST_AE1C-M909T and GST_AE1C-Δ 11) fusion protein or GST alone incubated with either GST_PDLIM5 (d) or tag-free PDZ domain of PDLIM5 (e) followed by detection with anti-PDLIM5-nt antibody. Signals were expressed relative to WT (100%) ± SEM. Specific binding of AE1C-WT or AE1C-M909T to PDLIM5 or PDLIM5-PDZ is shown, demonstrating the importance of the PDZ binding motif in AE1 binding to PDLIM5's PDZ domain. Signals from GST alone were significantly low (P < 0.0001 analyzed using ANOVA) confirming specificity.

Discussion
We have here identified PDLIM5 as a binding partner of the terminal 11 residues of kAE1 and demonstrated a direct interaction through PDLIM5's PDZ domain. Our results suggest that the PDLIM5/kAE1 interaction plays a role in kAE1's basolateral membrane targeting/retention. The role of the last 11 residues of AE1 for PDLIM5 binding is of particular interest, the presence of basolateral targeting motifs/determinants within this region of AE1 having previously been identified through expression studies of the truncating mutant 5,6,18,28 . In those studies, the mutation led to intracellular retention in HEK293 cells and mis-targeting apically in fully polarized MDCK cells.
We have previously also demonstrated the importance of the far C-terminus for AE1's membrane residency, firstly by removing just the last 4 residues resulting in intracellular retention, and secondly showing non-polarized targeting with some intracellular retention of kAE1-M909T, a mutation causing dRTA 8 . In that study, we introduced the idea that M909T creates a Type I PDZ binding motif (X(S/T)XΦ , where X and Φ are 'any' and 'hydrophobic' amino acids, respectively) into the C-terminus. The last four residues (A 908 MPV) in wild-type AE1 are likely a Class II (XΦ XΦ ) motif 24 . Thus, the preservation of PDLIM5 binding by AE1-M909T or AE1-WT suggests that the PDZ domain in PDLIM5 can function as either Class I or II.
SiRNA-induced depletion of endogenous PDLIM5 led to overall reduction of kAE1 with a major fall in kAE1 levels on the plasma membrane and increased intracellular retention. Furthermore, in the PDLIM5-depleted time-course assay, surface kAE1 was low in knockdown cells at all time-points. The overall reduction is likely the result of degradation of non-delivered kAE1 by a lysosomal pathway as described by Almomani et al. 29 . Conversely, when membrane overexpression of kAE1 was induced in these cells, significantly increased levels of endogenous PDLIM5 protein, but not mRNA, were observed, suggesting increased protein stability due to PDLIM5's association with kAE1. Since plasma membrane expression of eGFP-kAE1 reaches steady state at around 14-16 hours after induction of its expression and PDLIM5 levels rise in parallel, we believe that PDLIM5 is involved in both kAE1 translocation to the membrane and in its retention. These results together indicate a requirement of PDLIM5 for the proper membrane residency of kAE1.
The nature of kAE1's linkage to the underlying actin cytoskeleton has been uncertain. In 2007, Keskanokwong et al. reported an association between the kAE1's N-terminus and ILK via a yeast two-hybrid screen 26 . Further characterization employed co-IPs using various fragments of the two proteins in HEK293 cells. Overexpression of ILK increased kAE1's presence at the membrane, with a parallel increase in ion transport; pulse-chase assay showed that the two proteins associated early in biosynthesis and travelled together from endoplasmic reticulum to plasma membrane. It was therefore proposed that ILK acts as a bridging molecule between the N-terminus of kAE1 and actin, via paxillin and actopaxin. However, a year later Williamson et al. showed that deletion of the majority of kAE1's proposed ILK binding region failed to alter membrane residency of kAE1 30 , weakening the hypothesis and calling the proposed direct interaction into question. Our GST pulldown results using both human kidney and HEK-Δ pMEP-GFP-kAE1 lysates support Keskanowong's findings but importantly, implicate the C-terminus of AE1 and not the N-terminus. Our ELISA and/or blot overlay assays, which showed no detectable direct interaction between AE1C and ILK, but a clear interaction between PDLIM5 and ILK that link AE1's C-terminus to ILK, would account for Williamson's results.
In thinking about kAE1's tethering to the actin cytoskeleton, our bioinformatic analysis supports earlier studies reporting that ILK interacts with actopaxin and paxillin 26,31,32 . Therefore, our working model (Fig. 5a) is one in which PDLIM5 physically tethers kAE1 to ILK, which would be required for correct movement of kAE1 towards its final basolateral membrane destination. Our previous studies of wild-type kAE1 in kidney epithelial cells also demonstrated that it is stabilized on the membrane by Na + , K + -ATPase through interaction of kAE1 with the pump's β 1 subunit 21 . Our current GST pulldown data (Fig. 5b) indicate that levels of β 1 in the AE1C-M909T mutant sample were significantly lower (P < 0.0001 by ANOVA) than for wild-type, indicating loss of sodium pump binding by AE1-M909T. This difference may explain why, despite preservation of the PDLIM5 interaction in this mutant and therefore basolaterally directed travel, less appears basolaterally and more is intracellularly retained 8 .
Finally, a number of other basolateral membrane determinants involving residues D 902 EYDEV 907 are reported within the last 11 residues of AE1 23,29,30 , most recently including adaptor protein subunit 1B 33 , but how all these determinants collaborate to regulate kAE1 membrane residency is not yet clear. In summary, we have identified a direct interaction between kAE1 and PDLIM5, and our data indicate that PDLIM5 is not only a novel chaperone for kAE1, but also provides a bridge between kAE1 and the underlying actin cytoskeleton. In addition, combining our data with previous reports, a molecular model is emerging of kAE1's polarized cellular behaviour.
Scientific RepoRts | 7:39701 | DOI: 10.1038/srep39701 cDNA encoding PDLIM5 was amplified from human kidney cDNA pool by high fidelity PCR using primers PDLIM5_F CCGGAGCTCATGAGCAACTACAGTGTGTCA and PDLIM5_R CCGGCTCGA-GTCAAAAATTCACAGAATGAGCATG to introduce SacI and XhoI restriction sites, respectively. These sites were used to clone the PCR product into pSUMO3 vector (LifeSensors, Inc.) which contains His 6 -SUMO3 double tags upstream of a SacI site. All constructs were sequence-verified prior to use. To express intact kAE1 in mammalian cells, full-length cDNA was cloned into inducible vector Δ pMEP4 to create N-terminal eGFP-tagged kAE1 (Δ pMEP-eGFP-kAE1) as previously described 8,21 .
cDNA encoding PDLIM5 within pSUMO3 vector was expressed in E. coli BL21 cells to produce an N-terminal His 6 -SUMO3 tagged PDLIM5 fusion protein, which was purified using Ni-NTA agarose resin (QIAGEN). Eluate containing 200 mM imidazole (pH 7.4) from the Ni-NTA resin was digested with SUMO proteinase 2 (LifeSensors, Inc.) to remove the His 6 -SUMO3 tag. Digests were further purified by size-exclusion chromatography using HiLoad 26/60 Superdex 75 (GE life science) column equilibrated with PBS. Fractions collected were initially analyzed by SDS-PAGE coomassie staining, then Western blotting and N-terminal sequencing (Department of Biochemistry, University of Cambridge).
Plasmid or siRNA transfection. MDCKII and HEK293 express endogenous PDLIM5, but not AE1. To overexpress kAE1, both lines were transfected with Δ pMEP-GFP-kAE1 followed by stable clone selection, and maintenance as described 21 .

Protein expression in mammalian cells. Stable MDCK-Δ pMEP-GFP-kAE1 cells were seeded on
Corning ® Transwell ® polycarbonate membrane cell culture inserts (Transwell filters, Corning Life Sciences) and grown for 4 days to form polarized monolayers. HEK-Δ pMEP-GFP-kAE1 cells were grown either on glass coverslips or in normal tissue culture plates/flasks. GFP-tagged kAE1 protein expression was induced for 8-12 h as previously described 8,21 .
Immunofluorescence microscopy. Ethically-approved and formally patient-consented samples of normal human kidney were obtained from the Addenbrooke's Hospital Tissue Bank (Cambridge Research Ethics Committee approval 03/279). All studies were carried out in accordance with relevant guidelines and regulations. 4% formaldehyde-fixed paraffin wax-embedded 5 μ m-thick kidney sections were dual-immunostained for kAE1 and PDLIM5 based on previously described methods 35 . Following citrate buffer antigen retrieval, sections were blocked with 10% FBS in PBS containing 0.01% Tween 20 (blocking buffer-1), then incubated with primary antibodies (Bric-170 and anti-PDLIM5-mid), at 1:100 dilution at 4 °C overnight. Fluorochrome-conjugated secondary antibodies were used at 1:500 dilution for 1 h at room temperature. Sections were mounted in Vectashield Mounting Medium (Vector Laboratories) and examined with a Confocal Laser Scanning Microscope (LSM880). As controls, primary antibodies were replaced by either non-immune serum or isotype-specific antisera; all steps were followed unchanged.

Co-IP-coupled Mass Spectrometry and GST pull-down assays.
For co-IP assays using MDCK cells stably expressing full-length kAE1 8,21,23 , cell lysates were pre-cleared and transferred to protein G-agarose beads preloaded with mouse monoclonal antibodies Bric-170 or anti-CD63 (which does not cross-react with dog CD63) for overnight incubation at 4 °C. Beads were then thoroughly washed with co-IP lysis buffer containing reduced NP-40 (0.1%). Proteins co-immunoprecipitated were separated by SDS-PAGE. Proteomic analyses were carried out at the Cambridge Centre for Proteomics. All gel fragments generated were excised and subjected to to Liquid Chromatography tandem Mass Spectrometry (LC-MS/MS) after in-gel trypsin digestion. LC-MS/MS was performed using an Eksigent NanoLC-1D Plus (Eksigent Technologies) HPLC system and an LTQ Orbitrap Mass Spectrometer (Thermo Fisher Scientific), as described in detail elsewhere 36 . Digests from gel segments were run with dynamic exclusion. MS data were processed using the SEQUEST Bioworks Browser (version 3.3.1 SP1; Thermo Fisher Scientific) to generate MS/MS peak lists. Combined peak list files were submitted to the MASCOT search algorithm (version 2.2.1; Matrix Science) and searched against the IPI-Human Database, version 4.3. All ambiguous peptides were excluded unless matched only to products of a single gene. Protein identification required at least two unique peptides, with a false discovery rate of < 0.1. In addition, the DAVID bioinformatics tool was used to generate functional enrichment categories with a false discovery rate of < 0.01. GST-pull down assays using cytosol or membrane fractions of human kidney cortex lysates prepared as described 37 , or lysates from HEK-Δ pMEP-GFP-kAE1 cells were carried out as described 38 . Bound proteins were probed on blots with anti-PDLIM5-ct and anti-ILK antibodies, and band intensity measured using ImageJ software (National Institutes of Health). The same blot was re-probed with anti-β 1, instead of anti-GST antibody, as loading control (Supplementary Figure 4). ELISA and blot overlay analysis. ELISA was performed 23 by immobilizing 100 μ l of 30 μ M of GST_ AE1C-WT, GST_AE1C-Δ 11, GST_AE1C-M909T or GST alone onto a 96-well plate and incubating with 100 μ l of 2 μ g/ml of GST_PDLIM5 (AbNova, H00010611-P01) or tag-free PDLIM5-PDZ or GST_ILK (AbNova, H00003611-P01) recombinant protein. Detection was with anti-PDLIM5-nt or anti-ILK and HRP conjugated antibodies, and signals were visualized with ABTS.
Blot overlay assay was carried out 39 using equal moles (at pmole level, see Fig. 4 legend for details) of one of GST-AE1C, GST-PDLIM5, PDLIM5-PDZ or GST alone spotted onto gridded 0.45 μ m Nitrocellulose Transfer Membrane (Whatman International Ltd) and air dried for 3 h. Following blocking in PBS containing 2% skimmed milk and 0.1% Tween 20 (blocking buffer-2), the membrane was first overlaid with GST-ILK recombinant protein (50-65 pmoles), then washed and detected with anti-ILK and HRP conjugated antibodies. Blocking buffer-2 was used for all washes and for dilutions of the ligand and antibodies.
Bioinformatic analysis. String-DB [http://www.ncbi.nlm.nih.gov/pubmed/25352553] was searched using Homo sapiens gene names for PDLIM5, ILK, paxillin and actopaxin. DOMINE 40 and InterPro databases were used to seek evidence of interaction between the domains of every possible pair of proteins among them, domains of each protein having been collected from the Interpro 41 database.