Human TRPV4 Channel Splice Variants Revealed a Key Role of Ankyrin Domains in Multimerization and Trafficking*

The TRPV4 cation channel exhibits a topology consisting of six predicted transmembrane domains (TM) with a putative pore loop between TM5 and TM6 and intracellular N- and C-tails, the former containing at least three ankyrin domains. Functional transient receptor potential (TRP) channels are supposed to result following the assembly of four subunits. However, the rules governing subunit assembly and protein domains implied in this process are only starting to emerge. The ankyrin, TM, and the C-tail domains have been identified as important determinants of the oligomerization process. We now describe the maturation and oligomerization of five splice variants of the TRPV4 channel. The already known TRPV4-A and TRPV4-B (Δ384–444) variants and the new TRPV4-C (Δ237–284), TRPV4-D (Δ27–61), and TRPV4-E (Δ237–284 and Δ384–444) variants. All alternative spliced variants involved deletions in the cytoplasmic N-terminal region, affecting (except for TRPV4-D) the ankyrin domains. Subcellular localization, fluorescence resonance energy transfer, co-immunoprecipitation, glycosylation profile, and functional analysis of these variants permitted us to group them into two classes: group I (TRPV4-A and TRPV4-D) and group II (TRPV4-B, TRPV4-C, and TRPV4-E). Group I, unlike group II variants, were correctly processed, homo- and heteromultimerized in the endoplasmic reticulum, and were targeted to the plasma membrane where they responded to typical TRPV4 stimuli. Our results suggest that: 1) TRPV4 biogenesis involves core glycosylation and oligomerization in the endoplasmic reticulum followed by transfer to the Golgi apparatus for subsequent maturation; 2) ankyrin domains are necessary for oligomerization of TRPV4; and 3) lack of TRPV4 oligomerization determines its accumulation in the endoplasmic reticulum.

The general topology of a TRP subunit consists of six predicted transmembrane domains (TM) with a putative pore loop between TM5 and TM6 and intracellular N-and C-terminal regions of variable length, the former containing multiple ankyrin (ANK) repeats in the TRPC, TRPA, TRPN, and TRPV subfamilies (1). ANK repeats are modular protein interaction domains, each composed by Ϸ33 amino acids with a highly conserved secondary structure of helix turn helix that determines its interaction properties (20). Functional TRP channels are supposed to result following the assembly of four TRP subunits. Although multimerization has been already demonstrated for various members of the TRP family (including TRPV4) (21)(22)(23)(24)(25), the tetrameric structure has only been proposed for a few TRP channels (26 -28).
The rules governing subunit assembly and the protein domains implied in this oligomerization process are just starting to emerge (29,30). Several studies suggest the participation of the cytosolic N-terminal region in the assembly of homo-or heteromultimeric TRP channels (23,31,32), and two recent papers narrowed down the role of the ANK domains in the oligomerization process of TRPV5 and TRPV6 channels (24,33). On the other hand, transmembrane domains (25) and the cytoplasmic C terminus (34) have also been identified as important determinants of the oligomerization process of TRPV channels.
In this study we characterized the oligomerization, maturation, localization, and ion channel function of five splice variants of the TRPV4 channel. Variants lacking part of the ANK domains did not oligomerize, were retained in the ER, and did not produce functional channels at the plasma membrane.
Cell Culture and Transient Transfection-CFT1-LCFSN human tracheal epithelial cells, HEK-293, and HeLa cells were grown as previously described (5,11). Fluorescent resonance energy transfer (FRET) was carried out on HEK-293 cells transfected with 0.4 g of CFP-tagged TRPV4 channel isoform and 1.6 g of YFP-tagged subunit using FuGENE 6 transfection reagent (Roche). For coimmunoprecipitation experiments HEK-293 cells were grown in 60-mm dishes and transfected with 2 g of FLAG-and 2 g of YFP-tagged plasmids. Immunofluorescence, glycosylation, and functional studies were carried out on HeLa cells transfected with ExGen500 (Fermentas MBI). 3 g of the TRPV4 constructs and 7 equivalents of polyethylenimine, together with the transfection reporter pEGFP-N1 (for functional studies) at a 10:1 ratio, were added to each 35-mm dish.
Deglycosylation with PNGase F and Endo H-Total protein was extracted from HeLa cells 24 h after transfection. Cells were lysed in a buffer containing (mM): 150 NaCl, 5 EDTA, 1% Nonidet P-40, 1 sodium orthovanadate, 1 phenylmethylsulfonyl fluoride, 0.05% aprotinin, and 1 dithiothreitol (1 h at 4°C). The lysis buffer was supplemented with a Complete Mini protease inhibitor mixture (1:7, v/v; Roche). The nuclear fraction was pelleted by centrifugation at 12,000 ϫ g for 15 min. 20 g of total protein were digested with 2 l of Endo H or PNGase F (New England Biolabs) for 1.5 h at 37°C following the manufacturers instructions. Proteins were subjected to SDS-PAGE (8%) and subsequently electroblotted onto nitrocellulose membranes. Incubation with anti-human TRPV4 antibody (1:500) for 1 h at room temperature was followed by incubation for 1 h at room temperature with the horseradish peroxidase-conjugated donkey anti-rabbit IgG (Amersham Biosciences) at a dilution of 1:2000. Detection was done with a SuperSignal West chemiluminescent substrate (Pierce).
Cytosolic Ca 2ϩ Measurements-Changes in the intracellular Ca 2ϩ concentration were measured in cells loaded with 5 M fura 2-AM (Molecular Probes) incubated for 30 min at 37°C followed by de-esterification for 15 min at room temperature. The ratio of emitted fluores-cence (510 nm) following alternating excitation at 340 and 380 nm was obtained as previously described (15). The isotonic bathing solution contained (mM): 2.5 KCl, 140 NaCl, 1.2 CaCl 2 , 0.5 MgCl 2 , 5 glucose, 10 HEPES (310 mosmol/liter, pH 7.35). The hypotonic bathing solution (220 mosmol/liter, pH 7.35) was prepared by omitting 50 mM NaCl from the isotonic Hanks' solution and osmolality was adjusted with D-mannitol when necessary. Results are expressed as mean Ϯ S.E. of n observations.

Cloning of TRPV4 Variants from Human Airway Epithelial Cells-A
reverse transcriptase-PCR-based cloning process identified 5 variants of the TRPV4 channel in human tracheal epithelial cells. Two of the cloned cDNAs corresponded to the already described TRPV4 isoform A (fulllength cDNA; GenBank number NM_021625) and TRPV4 isoform B (lacking exon number 7, number NM_147204; ⌬384 -444 amino acids). Variant B was originally described as an isoform lacking exon 6 (number AB073669), although subsequent analysis of the genomic structure of the TRPV4 gene identified a further 5Ј non-coding exon 1, therefore renumbering the TRPV4 exons. We also identified three new splice variants, named TRPV4-C, TRPV4-D, and TRPV4-E variants (Fig. 1A), affecting the cytoplasmic N-terminal region. TRPV4-C lacks exon 5 (number DQ59644; ⌬237-284 amino acids), TRPV4-D presents a short deletion inside exon 2 (number DQ59645; ⌬27-61 amino acids), and TRPV4-E (number DQ59646; ⌬237-284 and ⌬384 -444 amino acids) is produced by a double alternative splicing lacking exons 5 and 7. Splicing sites for the TRPV4-D variant are given by two canonical dinucleotides, GT and AG for donor and acceptor sites, respectively. All variants cloned contained the synonymous single nucleotide polymorphism in position 761 (dbSNP reference number rs3825394). Transcripts for these five variants have been also amplified from the human bronchial epithelial cell line HBE (results not shown).
Functional Measurements of the Human TRPV4 Variants-The TRPV4-A channel responds to a wide variety of stimuli (13). To functionally evaluate TRPV4 variants we transiently transfected HeLa cells and measured intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) in response to three different well known activators of TRPV4 (Fig. 1). As a negative control, HeLa cells were transfected with the backbone plasmid ( Fig. 1, B-D). 30% Hypotonicity, 1 M 4␣-PDD, and 10 M arachidonic acid induced increases in [Ca 2ϩ ] i for TRPV4-A (f) and TRPV4-D channels (light grey square) (Fig. 1, E-G), whereas no response was observed for TRPV4-B (OE), TRPV4-C (light grey triangle), and TRPV4-E (dark grey triangle) transfected cells (Fig. 1, H-J). No apparent differences in the Ca 2ϩ response were seen between TRPV4-A and TRPV4-D variants. JANUARY 20, 2006 • VOLUME 281 • NUMBER 3
Expression and Subcellular Localization of TRPV4 Channel Variants-Immunolocalization of TRPV4 variants was carried out on transiently transfected HeLa cells with an antibody raised against TRPV4 (5, 11). Fig. 3A shows that TRPV4-A and TRPV4-D proteins displayed a robust cell surface expression, whereas TRPV4-B, TRPV4-C, and TRPV4-E proteins showed a reticular intracellular immunostaining, indicative of retention in the ER.
To determine whether B, C, and E variants were trapped in the ER, we analyzed their glycosylation profile and carried out colocalization studies with calnexin (an ER resident protein) in transfected HeLa cells. Total cell protein was treated with Endo H to remove N-linked high mannose glycosylations (indicative of ER forms) or PNGase F to remove both high mannose and complex glycosylation (indicative of post-ER forms). TRPV4-A was used as a control. As shown in Fig. 3B, three immunoreactive protein bands with apparent molecular masses of 96, 100, and 110 kDa were detected for TRPV4-A (and TRPV4-D, results not shown) in untreated samples. The 96-kDa band represents the nonglycosylated form of the TRPV4-A. The 100-kDa band, but not the 110-kDa band, was sensitive to Endo H and both bands were sensitive to PNGase F. Thus the 100-kDa band represents the high mannose and the 110-kDa band the complex glycosylated form. An identical glycosylation pattern was observed for the TRPV4-D isoform (results not shown). An identical glycosylation pattern and Endo H/PNGase F sen-sitivity was observed for the TRPV4-D isoform (results not shown). In contrast, the TRPC4-B, -C, and -E variants gave rise to two immunoreactive bands migrating at 91 and 96 kDa, 92 and 97 kDa, and 86 and 90 kDa, respectively. The upper band disappeared upon treatment with Endo H or PNGase F (Fig. 3B) indicating the presence of high-mannose glycosylation, but a lack of further maturation. O-Glycosylation of TRPV4-A can be excluded because PNGase F treatment produces a unique band with an apparent molecular weight corresponding to the predicted size of the non-glycosylated protein. Taken together, the data suggest that B, C, and E variants are retained in the ER, whereas TRPV4-A and TRPV4-D are fully processed and reach the plasma membrane where they produce functional channels. Confirmation of ER retention for B, C, and E variants was obtained by colocalization  studies. All three variants overlap with the ER resident chaperone calnexin (Fig. 3C). No colocalization of TRPV4-B variant with the ER/Golgi intermediate compartment (ERGIC) or the Golgi markers, ␣ERGIC-53 and GM130, respectively, was detected (Fig. 4A). Identical results were obtained with TRPV4-C and -E (not shown).
Homomerization of TRPV4 Variants-Functional TRP channel complexes require the tetrameric assembly of pore-forming subunits (30). Therefore, quality control mechanisms must be met to ensure the correct quaternary structure. Several compartments, including ER and the Golgi, have been identified as sites for oligomerization (37,38). Thus, we next checked the multimerization of the TRPV4 variants. For this purpose, we evaluated the proximity of CFP-and YFP-tagged TRPV4 variants, as an indication of subunit assembly by FRET technique (25). The relative CFP and YFP fluorescence intensities were determined for every single cell to ensure comparable levels of TRPV4 expression and to adjust the 1.5-3-fold molar ratio between acceptor-(YFP) and donor (CFP)-labeled subunits. Fig. 5 shows FRET efficiencies obtained in HEK-293 cells transiently co-transfected with expression plasmids encoding CFP-or YFP-fused TRPV4 isoforms. TRPV4-A and TRPV4-D variants displayed a higher FRET efficiency (17.5 Ϯ 0.94 and 18.32 Ϯ 1.07, respectively) compared with TRPV4-B, TRPV4-C, and TRPV4-D (around 5%). FRET results indicate that only TRPV4-A and TRPV4-D efficiently assemble into homomeric channel complexes.
To check whether TRPV4 channels oligomerize in the ER we carried out FRET analysis of HeLa cells transfected with CFP-and YFP-fused TRPV4-A variants and treated for 10 h after transfection with the intracellular transport inhibitor brefeldin A (0.5 g/liter), which prevents transport from the ER to the Golgi (39). FRET efficiencies (measured 14 h after transfection) for homomeric TRPV4-A were not statistically different between solvent-treated and brefeldin A-treated cells (24.1 Ϯ 1.3 and 19.4 Ϯ 3.3%, respectively; n ϭ 3; p Ͼ 0.05). This result suggests that assembly of TRV4 occurs primarily in the ER and that ER sequestration of TRPV4-B, -C, and -E variants is related to their inability to oligomerize in the ER.
One possible mechanism explaining the intracellular sequestration of channel subunits involves the participation of bi-arginine (RXR) retention motifs, which are hidden under the tetrameric structure but when exposed in non-assembled channel subunits determine their intracellular retention (40). TRPV4 protein contains 4 RXR motifs, two on the N-terminal ( 122 RWR and 269 RGR) and two on the C-terminal ( 816 RLR and 819 RDR) intracellular tails. Expression of TRPV4-B variants containing alanine mutations in the first arginine of either of the four RXR motifs did not modify the localization of TRPV4-B (data not shown). Likewise, a TRPV4-B construct in which all four RXR motifs were modified by alanine mutations was still retained in the ER and did not migrate to the ERGIC or Golgi (Fig. 4B).
Heteromerization of TRPV4 Variants-Heteromeric assembly of TRP channels can produce a change in the intracellular localization and/or modify the function of the new assembled protein (33,41,42). To test whether TRPV4-A could rescue TRPV4-B, TRPV4-C, and TRPV4-E isoforms retained in the ER, we co-expressed TRPV4-A-YFP and CFPfused variants in HEK-293 cells and visualized the subcellular distribution by confocal laser microscopy. As shown in Fig. 6A plasma membrane overlapping was only observed in cells expressing TRPV4-A and TRPV4-D (with almost identical localization), whereas partial overlap in intracellular compartments was detected for all combinations of TRPV4 variants. It is also worth noting the apparent lack of redistribution to the plasma membrane of the intracellular variants (B, C, and E) following coexpression with the YFP-tagged TRPV4-A isoform. TRPV4-B, TRPV4-C, and TRPV4-E proteins failed to reach the plasma membrane even if coexpressed with excess amounts of targeting competent TRPV4-A-YFP.
Heteromerization of TRPV4 variants was also tested by FRET and coimmunoprecipitation analysis. Using TRPV4-A-YFP as the FRETacceptor and CFP-fused TRPV4-B-E variants as donors, a high FRET efficency of 21.7 Ϯ 3.5 could only be demonstrated for the heteromeric channel formation between TRPV4-A and TRPV4-D subunits. By contrast, FRET efficiencies remained low for TRPV4-A coexpressed with TRPV4-B, TRPV4-C, or TRPV4-E (Fig. 6B). Heteromeric TRP proteins may modify the response of the resulting channel compared with the homomeric channel, a possibility we also evaluated. Measurements of [Ca 2ϩ ] i in cells expressing heteromultimeric TRPV4-A/TRPV4-D channels did not differ from the homomultimeric TRPV4-A and  TRPV4-D channels (Fig. 1, E-G) probed with 30% hypotonicity, 1 M 4␣-PDD, and 10 M arachidonic acid.
Coimmunoprecipitation experiments confirmed the FRET data. From all the combinations tested with the TRPV4-A-FLAG-tagged protein coexpressed with TRPV4-YFP isoforms in HEK-293 cells, we were able to coimmunoprecipitate only the TRPV4-D variant (Fig. 6C, top panel). Lysate was probed with anti-YFP antibody (Fig. 6C, middle panel), to check the expression of the transfected YFP variants, and the immunoprecipitates with anti-FLAG antibody (Fig. 6C, lower panel), proving the recovery of the FLAG protein, are also shown. No FRET or coimmunoprecipitation between TRPV4-B, -C, and -E variants was detected (results not shown).

DISCUSSION
Our study of oligomerization, localization, and channel activity of human TRPV4 splice variants has allowed us to identify the N-terminal ANK repeats as key molecular determinants of subunit assembly and subsequent processing of the assembled channel. Using a combined approach of several independent methods we have described five TRPV4 variants (TRPV4-A-E) cloned from human airway epithelial cells, three of which were described for the first time (TRPV4-C-E). We have grouped them into two classes: group I (including TRPV4-A and TRPV4-D) and group II (including TRPV4-B, TRPV4-C, and TRPV4-E). Group I variants are correctly processed and targeted to the plasma membrane where they form functional channels with similar electrophysiological properties. Variants from group II, which are lacking parts of the ANK domains are unable to oligomerize and were retained intracellularly, in the ER.
Misfolded channel proteins and/or oligomerization-deficient subunits can be identified and intracellularly retained by quality control mechanisms in the ER (37,43). Therefore, it is particularly important to identify the structural domains involved in the oligomerization process. In the case of TRP channels, identification of the assembly pathway is starting to emerge (30). The N-tail of TRP channels, as it occurs for voltage-gated K ϩ channels (33), has also been identified as a possible multimerization domain (23)(24)(25)(31)(32)(33). In the case of TRPV5 and TRPV6 a key oligomerization domain has been narrowed down to the N-terminal ANK domains (24,33). C-tails and transmembrane domains have also been suggested to participate in the oligomerization and/or stabilization of assembled TRP channels (33,34), including TRPV4 (25).
Our study of TRPV4-B, -C, and -E splice variants (missing part of the ANK domains) by in vivo FRET and coimmunoprecipitation methods has allowed us to show for the first time three important traits of TRPV4 biogenesis. 1) Glycosylation of TRPV4 channel involves ER to Golgi transport with the corresponding change in the N-linked oligosaccharides from the high mannose type characteristic of the ER to the complex type characteristic of the Golgi apparatus, without apparent O-glycosylation. 2) TRPV4-A subunits oligomerize in the ER (no significant differences in FRET efficiency were seen in the absence or presence of brefeldin A, an inhibitor of ER to Golgi transport). 3) Impaired subunit assembly of type II variants is because of the lack of N-terminal ANK domains and causes protein retention in the ER. Although we propose that ER retention of TRPV-B, -C, and E variants is directly related to their inability to oligomerize we cannot unequivocally discard that such retention is because of protein misfolding.
Subunit misfolding is a cause of inefficient ion channel assembly, but folding events of ion channel subunits also occur after subunit assembly, i.e. subunits oligomerize even if their folding is not complete (37,44). A possible mechanism for ER retention of unoligomerized subunits involves exposure of a retention signal sequence (RXR), which otherwise is shielded in completely assembled channels (40,45). RXR motifs play a role in monitoring the assembly of multimeric proteins and TRPV4 presents four RXR motifs, two in each intracellular tails. However, mutation of all RXR motifs did not modify the intracellular distribution of type II variants, similarly to what have been reported for other membrane proteins (46). The implication of other ER-transport signals (47,48) in the retention of type II TRPV4 variants is a mechanism that remains to be addressed.
Intriguingly, an early study on truncated rat TRPV4 lacking the three ANK domains reported a delayed response to hypotonicity (3). At present, we cannot explain how these truncated proteins, which according to our results should not oligomerize, can produce functional channels. Certainly, the original observation by Liedtke and colleagues (3) should be re-evaluated in view of our results and those of others demonstrating impaired function of TRPV4 (49) and TRPV6 (24) channels with deleted ANK domains.
Ion channel functional diversity is greatly enlarged by both the presence of splice variants and heteromerization of different pore-forming and regulatory subunits. Alternative splicing is a major contributor to protein diversity (50). Within the TRP family of ion channels several splice variants have been identified, some of them resulting in lack of responses to typical stimuli, others modifying the pore properties, and those exerting dominant negative effects (42,(51)(52)(53)(54). Group II TRPV4 splice variants have been identified in two unrelated, human airway epithelial cell lines (CFT1-LCFSN and HBE). Considering the relevance of TRPV4 channels in epithelial physiology (5,11,55), it may be conceivable that a change in the expressed ratio of group I to group II variants, favoring the later, would modify normal epithelial functioning. Splicing can be regulated by several stressing stimuli (56) including pH, osmotic, and temperature shocks, all of them being also activating stimuli of the TRPV4. Therefore, it will be interesting to test whether such stimuli modify the expressed ratio of TRPV4 group I to group II variants.