Alternative Splicing of a Protein Domain Indispensable for Function of Transient Receptor Potential Melastatin 3 (TRPM3) Ion Channels

Background: TRPM3 proteins form Ca²⁺ permeable ion channels involved in insulin secretion and pain perception.

Results: A domain indispensable for TRPM3 channel function (ICF) is subject to alternative splicing.

Conclusion: This domain contributes essentially to the formation of TRPM channels and removing it by splicing modulates TRPM3-mediated Ca²⁺ signaling.

Significance: Alternative splicing of the ICF domain regulates biological functions attributed to TRPM3.

Abstract

TRPM3 channels form ionotropic steroid receptors in the plasma membrane of pancreatic β and dorsal root ganglion cells and link steroid hormone signaling to insulin release and pain perception, respectively. We identified and compared the function of a number of TRPM3 splice variants present in mouse, rat and human tissues. We found that variants lacking a region of 18 amino acid residues display neither Ca²⁺ entry nor ionic currents when expressed alone. Hence, splicing removes a region that is indispensable for channel function, which is called the ICF region. TRPM3 variants devoid of this region (TRPM3ΔICF), are ubiquitously present in different tissues and cell types where their transcripts constitute up to 15% of the TRPM3 isoforms. The ICF region is conserved throughout the TRPM family, and its presence in TRPM8 proteins is also necessary for function. Within the ICF region, 10 amino acid residues form a domain essential for the formation of operative TRPM3 channels. TRPM3ΔICF variants showed reduced interaction with other TRPM3 isoforms, and their occurrence at the cell membrane was diminished. Correspondingly, coexpression of ΔICF proteins with functional TRPM3 subunits not only reduced the number of channels but also impaired TRPM3-mediated Ca²⁺ entry. We conclude that TRPM3ΔICF variants are regulatory channel subunits fine-tuning TRPM3 channel activity.

Introduction

Transient receptor potential melastatin 3 (TRPM3)³ belongs to the large superfamily of TRP ion channels (1). TRP proteins exhibit common structural features such as six transmembrane domains with a pore loop between transmembrane domain 5 and transmembrane domain 6 and amino and carboxytermini located in the cytosol. TRP proteins assemble as tetramers, and members of TRP subfamilies were proposed to form heterooligomers (for review, see Refs. 2 and 3). Likewise, Förster resonance energy transfer (FRET) between fluorescent TRPM3 fusion proteins and investigation of dominant negative effects of a TRPM3 pore mutant onto Ca²⁺ entry in oligodendrocytes suggested direct molecular interactions of TRPM3 proteins (4). Furthermore, the closely related proteins TRPM3 and TRPM1 are able to form heterooligomeric channel complexes, as shown by co-immunoprecipitation, FRET, and by analyzing the biophysical properties of the currents (5).

TRPM3 channels can be directly activated by the neurosteroid pregnenolone sulfate from the extracellular side (6). They are expressed in pancreatic β cells, where they contribute to Ca²⁺ entry followed by enhancement of insulin release (6). TRPM3 proteins are also expressed in neurons of dorsal root ganglia, where they serve as thermosensitive channels implicated in the detection of noxious heat (7). Finally, TRPM3 channels have been suggested to be coupled to vascular smooth muscle contraction and interleukin-6 secretion from proliferating muscle cells (8) as well as to oligodendrocyte differentiation and neuronal myelination (4).

The TRPM3 gene encodes a vast number of variants (9–11). Most of these arise by alternative splicing of their primary transcripts. Splicing within exon 24 determines the ion selectivity of TRPM3 channels, as we showed in a previous study (10). The variant TRPM3α1 preferentially conducts monovalent cations. In contrast, the variant TRPM3α2 that lacks 12 amino acid residues within the pore region of the channel is highly permeable for Ca²⁺ and other divalent cations. This variant shows a unique biophysical and pharmacological signature such as inhibition by extracellular Na⁺ and activation by the Ca²⁺ channel blocker nifedipine (10, 12).

In the present study, we characterized additional splice variants and identified another important protein domain of 10 amino acid residues that we called ICF. Removal of the ICF domain by splicing resulted in a complete loss of channel activity. In the brain, ΔICF transcripts make up to 15% of TRPM3 mRNA. ΔICF proteins contribute significantly to TRPM3 channel formation and impair TRPM3-mediated Ca²⁺ entry. Our data indicate that splicing regulates the number of operative TRPM3 channels by direct interference of ΔICF proteins with functional channel subunits.

EXPERIMENTAL PROCEDURES

Cell Culture, Transfection, Generation of Stable Cell Lines, and Fluorescence-activated Cell Sorting (FACS)

Ins1 cells were cultured in RPMI 1640 as described (13). This medium contains 11 mm glucose. For experiments depicted in Fig. 2C Ins1 cells were washed twice and transferred into serum-free DMEM (Sigma D5030) containing either 2 mm or 13 mm glucose or 13 mm glucose/50 μm pregnenolone sulfate. For transfection of Ins1 cells, we used Lipofectamine 2000 (Invitrogen). Nonpigmented epithelial cells from the ciliary body were cultured as described (14). HEK293 or modified human embryonic kidney cells (HEKtsA201) were transfected with vectors allowing bicistronic expression of target proteins together with the enhanced green fluorescent proteins (EGFP, (15)) using FuGENE 6 (Roche Applied Science). Unless otherwise stated, cells were analyzed 48–72 h post transfection. Sometimes, cells were passaged to reduce their density 24 h before measurement. A clonal HEK293 cell line stably expressing Myc-tagged TRPM3α2 proteins (HEKα2) was generated essentially as described (10). In brief, the cDNA of the Myc epitope was introduced in-frame after the start codon of the TRPM3α2 cDNA. The whole sequence was ligated into pcDNA3 and transfected into HEK293 cells. Cells were cultured in selection medium containing 500 μg/ml G418 for 4 weeks. Single cells were separated by FACS on a MoFlo cell sorter (Beckmann Coulter) and expanded. Clones were tested for their expression of TRPM3α2 in Western blots using monoclonal anti-TRPM3 and anti-Myc antibodies.

FIGURE 2.

Coexistence of ΔICF variants with functional TRPM3 channel subunits. A, oligonucleotide primers (arrows) flanking the ICF encoding region were used to amplify TRPM3 transcripts with (ICF) or without ICF encoding region (ΔICF) in different tissues and cells from different species. B and C, quantification of the abundance of TRPM3ΔICF mRNA relative to the total amount of TRPM3 transcripts using real-time RT-qPCR in different tissues (B) or in Ins1 cells after incubation in serum-free medium for the indicated time intervals (C) in the presence of 2 mm glucose (light gray bars), 13 mm glucose (black bars) and 13 mm glucose + 50 μm pregnenolone sulfate (PregS; dark gray bars). The percentage of ΔICF isoforms of the total amount of TRPM3 transcripts is shown (mean ± S.E.). The number of independent experiments is indicated in brackets. Analysis of variance and subsequent Bonferroni t test demonstrated, after 15 min of treatment, significant differences at the 95% level between 13% glucose and the other test conditions each, but not between 2% glucose and 13% glucose + 50 μm pregnenolone sulfate. DRG, dorsal root ganglion; NPE, nonpigmented epithelial cells.

Fluorescent Ca²⁺ Measurements

Cells were plated on poly-l-lysine-coated coverslips and measured the next day using an imaging system (TILL Photonics, Martinsried, Germany). Cells were loaded with 5 μm acetoxymethyl ester of fura-2 (fura-2-AM; Invitrogen) for 30–60 min at 37 °C, washed in modified Ringer solution containing 135 mm NaCl, 5.4 mm KCl, 1 or 2 mm CaCl₂, 2 mm MgCl₂, 20 mm glucose, 10 mm HEPES, (pH 7.25 adjusted with NaOH, 315 mosmol/kg) and analyzed in a perfusion chamber (Warner Instruments, Hamden, CT) with continuous perfusion of ∼1 ml per minute. For stimulation, pregnenolone sulfate was dissolved in Ringer solution. For stimulation with high Ca²⁺ concentrations in the absence of monovalent cations (monovalent-free), cells were perfused with buffer containing 10 mm CaCl₂, 260 mm glucose, 10 mm HEPES (pH 7.25 adjusted with N-methyl-d-glucamine (NMDG), 320 mosmol/kg). In some cases, a Ca²⁺-free Ringer solution was applied containing 2 mm EGTA instead of Ca²⁺. 1 m menthol stock solutions were prepared in dimethyl sulfoxide and dissolved in Ringer solution. Ratio images were obtained at excitation wavelengths of 340 and 380 nm every 3 s using a 20× Fluar objective (Zeiss). The fluorescence emissions at λ > 440 nm were captured with a CCD camera, digitized, and analyzed for individual cells after background subtraction using Till Vision software. Results are given as mean ± S.E. Unpaired two-tailed Student's t test was used to calculate p values as appropriate. In the figures, one asterisk indicates p < 0.05, two asterisks indicate p < 0.01, and three asterisks indicate p < 0.001.

Electrophysiology

Standard whole-cell patch clamping was performed with an EPC 10 amplifier under control of the Patchmaster software (HEKA, Lambrecht, Germany), using its automated capacity and series resistance compensation, essentially as described (12). The intracellular (pipette) solution contained 80–90 mm CsAsp, 45 mm CsCl, 4 mm Na₂ATP, 10 mm BAPTA, 5 mm EDTA, and 10 mm HEPES. The pH was adjusted to 7.2 with CsOH (adding ∼60 mm Cs⁺ to the solution), and the osmolality to values within the range of 305–320 mosmol/kg. The standard bath solution contained 145 mm NaCl, 3 mm KCl, 10 mm CsCl, 2 mm MgCl₂, 2 mm CaCl₂, and 10 mm HEPES. NaOH (2–5 mm) was used to adjust the pH to 7.2. Osmolality was brought to 320–330 mosmol/kg with d-glucose. The liquid junction potential (approx. −15 mV) between pipette and bath solutions was taken into account. Pregnenolone sulfate (35 μm and 150 μm) and nifedipine (20 μm and 50 μm) were added to standard bath solution from a stock solution (50 mm dimethyl sulfoxide). Solution exchange was accomplished with a custom-built gravity-driven local perfusion system. The holding potential was usually −15 mV, from which we applied fast (1 mV/ms) voltage ramps from −115 to +85 mV at a rate of ∼1 s⁻¹. From these ramps, the amplitudes of inward (at −80 mV) and outward (at +80 mV) currents were obtained off-line.

Detection and Quantification of ΔICF Transcripts

In accordance with the guidelines for minimum information for publication of quantitative real-time PCR experiment (MIQE (16)), total RNA was isolated with peqGOLD RNAPure reagent (peqLab) and controlled for its quantity, purity, and integrity by spectrophotometry (NanoDrop; Thermo Scientific) and microfluidic analysis using a Bioanalyzer 2100 (Agilent Technologies). Oligonucleotide primers that flank the ΔICF encoding region (5′-TCG CTC GCA GCC AGA TCT TTA T TT A (sense) and 5′-GGT ACA ATG TAT TTG AGG GCC CAT GTC (antisense)) and matching rat, mouse, and human sequences were used to amplify and visualize ICF and ΔICF encoding transcripts simultaneously (see Fig. 2A). To quantify the relative amount of ΔICF transcripts by real time reverse transcriptase-quantitative polymerase chain reaction (real time RT-qPCR) a pair of primers that specifically amplify ΔICF encoding transcripts (O_ΔICF, mouse, rat, 5′-TGG AAC AGA GTT GAC ATC GCT CG (sense); mouse, 5′-TGA GGG CCC ATG TCT TCC ATT TTC (antisense), rat, 5′-TGA GGG CCC ATG TCT TCC GTT TTC (antisense)) as well as control primers located apart from the splice site (O_con, mouse, rat, 5′-AGC CTG GAA CAG A GT TGA CAT CGC (sense); mouse, rat, 5′-TCT GTC CA G GAC TAG GGC ATC CAG (antisense)) were used. Real time RT-qPCR reactions using 30 ng of RNA each were performed in triplicate at a RotorGene 6000 real time analyzer (Qiagen) using the SensiMix one-step kit (Quantace) and the following cycling profile: 30 min/45 °C; 10 min/95 °C; 35–40 cycles of 95 °C/15 s, 68 °C/12 s, and 72 °C/15 s. Product identity was ensured by melting curve analysis and polyacrylamide gel electrophoresis. Quantification cycle (Cq) and amplification efficiency values (E) of each reaction were obtained from the comparative quantitation tool of the RotorGene 6000 analysis software. Using an adapted efficiency corrected quantification model (17), the relative expression of ΔICF transcripts was calculated with the following equation,

where Cq_con represents the quantification cycle obtained with O_con primer pair; E_con indicates the efficiency obtained with O_con primer pair; Cq_ΔICF indicates the quantification cycle obtained with O_ΔICF; and E_ΔICF indicates the efficiency obtained with O_ΔICF primer pair.

Western Blot Analysis

Western blotting was performed similar as described (18). Cells were washed 48 h after transfection with phosphate-buffered saline, pH 7.4, harvested from plates by trypsinization, and centrifuged at 200 × g and 4 °C for 5 min. Cell pellets were washed with phosphate-buffered saline, pH 7.4, and resuspended in ice-cold radioimmune precipitation assay buffer containing 150 mm NaCl₂, 50 mm Tris/HCl (pH 8.0), 1% Nonidet P40, 0.5% sodium desoxycholate, 0.1% sodium dodecylsulfate (SDS, w/v), 5 mm EDTA with added protease inhibitors (1 μg/ml leupeptine, 1 mm phenantroline, 1 μg/ml antipain, 0.1 mm phenylmethylsulfonyl fluoride, 1 μm benzamidine, 1 μm pepstatin A, 0.9 μm iodoacetamide, 0.3 μm aprotinine). Lysates were sheared by pipetting through a 27-gauge needle, and cellular debris was removed by centrifugation for 15 min at 12,000 × g and 4 °C. Proteins were dissolved in 2× protein loading buffer (120 mm Tris-HCl, pH 6.8, 8% (SDS), 20% glycerin, 0.01 bromphenol blue, 10% (v/v) β-mercaptoethanol) and separated by SDS-PAGE. Gels were transferred to nitrocellulose membranes (Hybond C extra; Amersham Biosciences) by tankblotting (PEQLAB Biotechnologie, Erlangen, Germany) in the presence of 20% methanol (350 mA, 1.5 h, 4 °C). Blots were stained with Ponceau red, incubated for 1 h at room temperature in 5% nonfat dry milk/Tris-buffered saline before the primary antibody was added for 12 h at 4 °C. The following primary antibodies were used: anti-TRPM3 (6), anti-TRPM8 (19), anti-Myc (clone 9E10, Roche Applied Science), anti-HA (clone 3F10, Roche Applied Science), anti-calnexin (Abcam), anti-α1 sodium potassium ATPase (AB7671, Abcam), and anti-EGFP (clones 7.1 and 13.1, Roche Applied Science). Blots were washed three times with Tris-buffered saline/0.1% Tween and incubated for 1 h with a horseradish peroxidase-conjugated secondary antibody. After three washes with Tris-buffered saline/0.1% Tween and one wash with Tris-buffered saline, signals were detected using the Western Lightning Chemiluminescence Reagent Plus Kit (PerkinElmer Life Sciences) and a cooled CCD system (LAS-3000, Fujifilm, Düsseldorf). Non-saturated signals were quantified densitometrically using AIDA software (Raytest).

Cell Surface Biotinylation

HEKtsA201 cells confluently grown in culture flasks (75 cm²) or FACS-sorted HEKα2 cells were washed 48 h after transfection and treated with EZ-Link-Sulfo-NHS-LC-biotin (final concentration 0.5 mg/ml; Pierce) as described previously (20). Subsequently, cells were lysed in ice-cold phosphate-buffered saline (pH 7.4) containing 1% Triton, 1 mm EDTA, and the same mixture of protease inhibitors as described for Western blot analysis. Lysates were sheared by pipetting through a 27-gauge needle, and agitated at 4 °C for 15 min, and cellular debris was removed by centrifugation for 5 min at 1000 × g and 4 °C. 900 μg of proteins (input) were added to 200 μl of avidin-agarose beads (Pierce) pre-equilibrated in the lysis buffer and agitated for 2 h at 4 °C. The biotin-avidin-agarose complexes were then harvested by centrifugation (400 × g, 2 min) and washed four times with the lysis buffer supplemented with 0.25 m NaCl. The beads were then resuspended in 120 μl of 2× protein loading buffer and incubated at 37 °C for 30 min prior to SDS-PAGE and Western blot analysis. Controls monitoring the input of TRPM3 protein were included.

Immunoprecipitation

Co-immunoprecipitation experiments were performed similar as described (21). Dishes with confluent HEK293 cells were co-transfected with Myc- and HA-tagged TRPM3 isoforms or TRPM4 as control. 48 h after transfection, cells were harvested, lysed, and separated from cell debris as described for Western blot analysis. For pre-clearing, protein samples were added to protein-G Sepharose beads (4 Fast Flow, GE Healthcare) pre-equilibrated in radioimmune precipitation assay buffer, agitated for 1 h at 4 °C, and centrifuged for 2 min at 400 × g. Samples (400 μl) were mixed with anti-Myc antibodies (6 μg; clone 9E10; Roche Applied Science) or anti-HA antibodies (6 μg; clone 12CA5; Roche Applied Science) and agitated for 1 h at 4 °C. Antigen-antibody complexes were then precipitated by adding 100 μl of protein G-Sepharose beads pre-equilibrated in radioimmune precipitation assay buffer. The immunoprecipitates were eluted by heating the samples in 50-μl 2× protein loading buffer for 5 min at 95 °C, subjected to SDS-PAGE and analyzed in Western blots that were either probed with anti-Myc or anti-HA antibodies.

Site-directed Mutagenesis

For mutagenesis, we used the QuikChange site-directed mutagenesis kit (Stratagene) and the following PCR conditions: 30 s at 95 °C, 12–18 cycles (95 °C, 30 s; 55 °C, 1 min; 72 °C, 1 min/kb of plasmid length). Mutations were verified by sequencing of both strands. A detailed list of oligonucleotide primers is shown in supplemental Table 3.

RESULTS

Splicing within Exon 13 Prevents TRPM3 Channel Activity

Comparison of TRPM3 protein sequences available in public databases revealed alternative splicing within the pore encoding exon 24 (10) and within exons 8, 13, 15, and 17 (Fig. 1A). Whereas exons 8, 15, and 17 are subject of exon skipping, the differences in exons 13 and 24 arise by alternative exclusion of a part of the exon by the use of an alternative 5′-splice site. Furthermore, the proteins differ in their amino termini encoded by exon 1 or exon 2 probably originating from the usage of alternative promoters (10). To identify functional differences related to splicing of exons 8, 13, 15, or 17, we introduced the cDNA of different TRPM3 variants TRPM3α2 to TRPM3α7 into HEK cells (Fig. 1A), which do not express TRPM3 endogenously (10). We analyzed the expression of TRPM3 proteins (Fig. 1B) and their Ca²⁺ signals (Fig. 1, C and D). Using monoclonal antibodies raised against TRPM3α2 amino acid residues Met¹²³⁰ to Ala¹³²⁰ (6), we found all variants expressed with expression levels ranging from 63–86% of TRPM3α2 (Fig. 1B). TRPM3α2 expressing cells responded strongly to stimulation with the agonist pregnenolone sulfate (Fig. 1, C and D) (6). We also applied an extracellular solution free of monovalent ions but containing 10 mm Ca²⁺ to boost the constitutive activity of TRPM3α2 and to release the channels from Na⁺-dependent inhibition (Fig. 1, C and D) (10). Even in the absence of receptor-specific stimuli, TRPM3α2 cells responded with a considerable increase of intracellular Ca²⁺.

FIGURE 1.

Identification of a protein region indispensable for TRPM3 channel function. A, scheme of TRPM3 isoforms functionally tested are shown as black bars and compared with the complete TRPM3 protein shown as yellow bars with spliced parts highlighted in orange, transmembrane regions are in gray, and alternative amino termini are in green (exon numbering according to Ref. 10). The protein region encoded by exons 13 and 14 of TRPM3α2 to TRPM3α6 is compared with that of TRPM3α7. Eighteen residues shown in orange are subject to alternative splicing. B, Western blot of HEKtsA201 cells transfected with different TRPM3 splice variants. Anti-TRPM3 antibodies (6) were used to detect recombinant TRPM3 proteins. Anti-EGFP antibodies served as loading controls. Signal intensities [%] relative to TRPM3α2 transfections each normalized to EGFP signals are indicated. C, fura-2 imaging experiments of HEK293 cells transfected with TRPM3α2 (red) or TRPM3α7 (blue). Transfected cells identified by their green fluorescence were compared with non-transfected neighboring control cells (black). Representative measurements of fluorescence ratios are shown in the presence of Ringer solution containing 2 mm Ca²⁺, during application of 30 μm pregnenolone sulfate (30 PregS), in solutions free of monovalents but containing 10 mm Ca²⁺ (10Ca/MVF) and in Ca²⁺-free solutions (EGTA). Each trace represents the average (±S.E.) of the number of cells indicated in brackets. D, maximal fluorescence ratios of experiments as shown in C of 64–289 transfected cells (gray bars) were compared with 87–333 non-transfected neighboring control cells (black bars). Control measurements of control cells from each experiment were not significantly different. At least four independent experiments were performed. E, current traces of whole-cell patch clamp recordings measured at −80 and +80 mV of HEK293 cells expressing TRPM3α2 (red) and TRPM3α7 (blue). Cells were exposed to 150 μm pregnenolone sulfate (150 PregS) or 50 μm nifedipine (50 Nif).

Then, we tested TRPM3 variants, which differed by the presence of protein regions encoded by exons 8, 13, 15, and 17 (Fig. 1A). TRPM3α3, TRPM3α4, TRPM3α5, and TRPM3α6 responded similarly as TRPM3α2 to removal of extracellular monovalents or to pregnenolone sulfate (Fig. 1D). In contrast, cells expressing the isoform TRPM3α7 did not show any increase in cytosolic Ca²⁺ after application of pregnenolone sulfate or relief of Na⁺-dependent channel inhibition in monovalent free solution (Fig. 1, C and D). Compared with TRPM3α2, this variant lacks 18 amino acid residues Val⁵¹² to Thr⁵²⁹ encoded by exon 13 (Fig. 1A).

In whole-cell patch clamp recordings, TRPM3α7 expressing cells displayed no currents in response to saturating concentrations of 150 μm pregnenolone sulfate or 50 μm nifedipine as a chemically different agonist of TRPM3 (Fig. 1E) (6). Apparently, 18 amino acid residues encoded by exon 13 are indispensable for TRPM3 proteins to function as ion channels. Therefore, this region is referred to as the ICF region (indispensable for channel function).

Splicing within Exon 13 Is Common in a Variety of Cell Types and Tissues

We next asked whether splicing within exon 13 is frequent and widespread in TRPM3-expressing tissues. So far, antibodies that specifically recognize the ICF region could not be obtained. Therefore, RT-PCR experiments were performed using primers flanking the TRPM3 ICF region and RNA from tissues and cells known to express TRPM3 (Fig. 2A). ICF containing and ICF lacking transcripts were detectable in all TRPM3 expressing tissues and cell types from mouse, rat, and human. By real-time RT-qPCR, we found substantial amounts of ΔICF transcripts in mouse and rat brain, where they contributed up to ∼15% of the total amount of TRPM3 transcripts (Fig. 2B), followed by up to ∼7% in the choroid plexus and insulinoma cells (Ins1) and 2–5% in eye and dorsal root ganglia. When we analyzed 144 independent full-length TRPM3 variants of the mouse choroid plexus in detail, we found 6.3% of the transcripts spliced within exon 13 (supplemental Fig. 1 and supplemental Tables 1 and 2). These data show that TRPM3ΔICF variants are significantly expressed and correspond to ∼2.5 up to ∼15% of the TRPM3 variants depending on tissue and cell type.

Next, we investigated whether alternative splicing of the ICF region is a regulated process depending on alterations of the environmental conditions. We used Ins1 cells that express functional TRPM3 channels (6) to analyze changes in the amount of ΔICF variants (Fig. 2C). These cells show enhanced insulin secretion as well as elevated cytosolic Ca²⁺ in the presence of high glucose concentrations (13). When the cells were transferred into a medium containing low (2 mm) glucose, the percentage of ΔICF transcripts decreased significantly compared with cells that were maintained at high (13 mm) glucose concentrations. Furthermore, in high glucose, we found the fraction of ΔICF transcripts reduced when we applied pregnenolone sulfate. These findings clearly demonstrate that alternative splicing of the ICF region in Ins1 cells is tightly regulated by extracellular glucose and the TRPM3 agonist pregnenolone sulfate.

Characterization of the ΔICF Isoform TRPM3α7

To analyze interactions of TRPM3 isoforms, immunoprecipitation experiments were performed. HA-tagged TRPM3α2, or TRPM3α7 proteins were co-expressed with Myc-tagged variants in HEK293 cells after transfection of their cDNAs in a 1:1 ratio. Precipitations were accomplished with either anti-Myc or anti-HA antibodies (Fig. 3A). HA-tagged TRPM3α2 co-precipitated each of the Myc-tagged TRPM3 isoforms but not TRPM4, which served as a negative control (right lanes in the upper panels). Consistently, each of the Myc-tagged variants TRPM3α2 to TRPM3α7 co-precipitated HA-tagged TRPM3α2 (left lanes in the lower panels). Similar to TRPM3α2, TRPM3α7 showed homomeric interaction (rightmost panel pairs). These data provide strong evidence for direct interactions of the different TRPM3 isoforms and support a model of homo- and heteromeric TRPM3 complexes.

FIGURE 3.

Protein-protein interaction and cell surface occurrence of ΔICF proteins. A, co-immunoprecipitation of Myc-tagged TRPM3 variants and TRPM4 (as control) with HA-tagged TRPM3α2 (seven panel pairs) or TRPM3α7 (rightmost panel pair). Antibodies used for immunoprecipitation (IP) are indicated below the panels. For detection (Det.), anti-Myc antibodies (upper row) were used. Filters were stripped and reprobed with anti-HA antibodies (lower row). B, signal intensities [%] of co-precipitated proteins relative to the total amount of proteins [100%] precipitable with anti-Myc (A, left lanes, upper row) or anti-HA (A, right lanes, lower row) are displayed graphically as indicated as numbers below the blots in A. A dashed line indicates the maximal number of proteins that can coprecipitate in heterotetrameric channel complexes formed of two different subunits. C, surface biotinylation of Myc-tagged TRPM3α2 and TRPM3α7 proteins. 30 μg of proteins introduced into avidin-agarose pulldown (In) and one-half (left panels) or one-fourth (right panels) of the avidin-bound fractions (Av) were analyzed in two independent experiments with monoclonal anti-TRPM3 or anti-Myc antibodies. The ratios of the signals of the avidin-bound fractions normalized to their input signals are indicated. As control for intracellular proteins present in the avidin-bound fractions, the stripped filters were incubated with anti-calnexin antibodies.

For all variants except TRPM3α7, the amount of co-precipitated isoforms was higher than 50% (Fig. 3, A and B), and in many cases, it was even close to 87.5%, which is the theoretical maximum assuming that heterotetrameric TRPM3 channels are formed of two different types of subunits. For TRPM3α7 we found a reduced amount of co-precipitating proteins (≤40%). Apparently, the ICF region stabilizes the interaction of TRPM3 subunits.

Conceivably, therefore, also the cell surface expression of ΔICF isoforms might be diminished. To examine this issue, we performed surface biotinylation assays, in which we labeled all proteins accessible from the extracellular side with sulfo-NHS-LC-biotin. Labeled proteins were precipitated with avidin-agarose and analyzed in Western blots using either anti-TRPM3 or anti-Myc antibodies (Fig. 3C). Compared with TRPM3α2-expressing cells, significantly less TRPM3α7 proteins (∼25%) were biotinylated, indicating that less ΔICF isoforms reached the cell surface of HEK293 cells. However, despite the lack of any channel activity in cells overexpressing TRPM3α7 (Fig. 1), we still detected substantial amounts of biotinylated TRPM3α7 proteins.

The ICF Region Is Conserved throughout the TRPM Family

The TRPM3 ICF sequence is well conserved in other members of the TRPM subfamily (Fig. 4A), and the corresponding ICF-like sequences may be important for the functions of these other TRPM channels, too. To test this, we deleted Leu⁴⁸² to Asn⁴⁹⁹ of TRPM8 (TRPM8ΔICF) and investigated the presence of TRPM8 proteins at the cell surface as well as TRPM8-mediated Ca²⁺ signals. Similar to TRPM3ΔICF, the presence of TRPM8ΔICF proteins at the cell surface was reduced to ∼23.2% (Fig. 4B). TRPM8-expressing cells displayed strongly increased Ca²⁺ levels after addition of 100 μm menthol (Fig. 4C). However, cells expressing the cDNA of TRPM8ΔICF did not show signals different from control cells. Similar to TRPM3, TRPM8ΔICF proteins are expressed in significant amount at the cell surface but do not constitute functional channels themselves.

FIGURE 4.

Loss of function after deletion of the ICF region in TRPM8. A, alignment of ICF regions of TRPM proteins 1–8. Identical residues present in at least four members of the TRPM family are labeled in gray and are indicated in the consensus sequence. The ICF domain identified in TRPM3 (see Fig. 5) is highlighted in red. Genbank^TM accession numbers are indicated. B, surface biotinylation of TRPM8 and TRPM8ΔICF proteins. Proteins introduced into pulldown experiments (In) and the avidin-bound fraction (Av) were analyzed in with anti-TRPM8 antibodies. The ratios of the signals of the avidin-bound fractions normalized to their input signals are indicated. Filters were stripped and incubated with anti-calnexin antibodies to control the presence of intracellular proteins in the avidin-bound fractions. C, representative fura-2 measurements of TRPM8 expressing HEK293 cells compared with TRPM8ΔICF and non-transfected, neighboring control cells after application of 100 μm menthol. Each trace represents the average (±S.E.) of the number of cells indicated in brackets. Three independent transfections with two experiments of at least 23 cells each showed similar results.

Identification of an ICF Domain

Secondary structure prediction of the TRPM3α2 sequence indicated an α helical conformation of the ICF region (22–24), which may be essential for channel function. However, substitution of the ICF region in TRPM3α2 by an α helical peptide of 18 alanine residues (TRPM3ICF/A18) did not maintain channel activity (Fig. 5A). Alternatively, a single amino acid residue or a particular recognition pattern could be important within the ICF region. For example, a substrate motif for casein kinase 2 is present in the ICF region predicting phosphorylation of serine 521. We deleted each single amino acid residue of the ICF region within TRPM3α2 (Fig. 5, B and C) and analyzed Ca²⁺ signals after introduction of the mutants in HEK293 cells. Single deletions of residues Val⁵¹² to Phe⁵¹⁷, Asn⁵²⁸, and Thr⁵²⁹ did not affect pregnenolone sulfate induced channel activity. In contrast, deletions of each of the ten residues Leu⁵¹⁸ to Tyr⁵²⁷ abolished TRPM3-mediated Ca²⁺ increase almost completely and mimicked the phenotype of the ΔICF splice variant. We then replaced each of these 10 residues by alanine and found that all mutated proteins including the S521A mutant formed active TRPM3 channels (Fig. 5, D and E). The experiments demonstrate that not a single amino acid side chain but rather the complete motif of 10 residues form a domain that is indispensable for TRPM3 channel function. Three of these 10 residues are leucines and present in nearly all TRPM proteins (see Fig. 4A).

FIGURE 5.

Identification of the ICF domain in TRPM3. A–E, pregnenolone sulfate (PregS) induced Ca²⁺ responses of HEK293 cells expressing TRPM3α2 mutants with complete replacement of the ICF region by 18 alanine residues (TRPM3ICF/A₁₈; A), single amino acid deletions (B and C), or single exchanges to alanine (D and E). A, B, and D, representative fura-2 measurements after stimulation with 30 μm pregnenolone sulfate. Each trace represents the average (±S.E.) of the number of cells indicated in brackets. C and E, maximal changes of the fluorescence (±S.E.) before and after application of pregnenolone sulfate of experiments performed as in B (22–141 cells) and D (27–45 cells) were calculated and compared with non-transfected neighboring control cells (not shown).

Regulation of TRPM3 Channel Activity by TRPM3 ΔICF Variants

Because ΔICF proteins are electrically silent (Fig. 1) but interact with Ca²⁺-permeable TRPM3 subunits and reach the plasma membrane (Fig. 3), we wondered whether TRPM3α7 might interfere with TRPM3-mediated Ca²⁺ entry. First, we introduced the TRPM3α7 cDNA into pancreatic Ins1 cells, which express TRPM3 endogenously (6), and we examined pregnenolone sulfate induced Ca²⁺ signals at 48, 72, and 96 h post transfection (Fig. 6, B and E). TRPM3α2 transfections served as positive (Fig. 6A), mock (vector) transfections as negative controls (data not shown). Mock transfections had no influence upon pregnenolone sulfate-induced Ca²⁺ signals, but introduction of TRPM3α2 strongly increased cytosolic Ca²⁺ at all time points, indicating that recombinant TRPM3 proteins are expressed in Ins1 cells after 2 days, up to 4 days (Fig. 6, A and E). Transfection of TRPM3α7 cDNA reduced pregnenolone sulfate-induced Ca²⁺ signals up to ∼50% (Fig. 6, B and E). These data show that TRPM3α2 enhances and TRPM3α7 decreases TRPM3 activity in Ins1 cells. TRPM3α2 expression may add additional functional TRPM3 channels, whereas TRPM3α7 may interact with endogenous TRPM3 channels and inhibit their activity. Accordingly, increase of channel activity has been observed after 48 h, whereas significant inhibition has been detected after 72 and 96 h only (Fig. 6E). Presumably, the onset of inhibition depends on the turnover of the endogenous TRPM3 proteins.

FIGURE 6.

Interference of TRPM3-mediated Ca²⁺ signaling by ΔICF variants. A–D, fluorescence ratios measured in Ins1 cells (Ins1) during application of 30 μm pregnenolone sulfate (30 PregS) 96 h post transfection (A and B) or in HEKα2 cells 72 h post transfection (C and D) of TRPM3α2 (red traces) or TRPM3α7 (blue traces) in comparison with non-transfected neighboring control cells (black traces). Each trace represents the average (±S.E.) of the number of cells indicated in brackets. E and F, changes of fluorescence upon pregnenolone sulfate application to TRPM3α2- and TRPM3α7-transfected Ins1 cells (E) or HEKα2 cells (F) 48, 72, and 96 h post transfection in experiments performed as in A–D. Values were normalized for each time point to 84, 64, and 26 mock-transfected Ins1 cells or 186, 89, and 59 mock transfected HEKα2 cells, respectively. Numbers of cells analyzed on each day are indicated in brackets. G, biotinylation of Myc-tagged TRPM3α2 surface proteins after introduction of HA-tagged TRPM3α7. Proteins introduced into pulldown experiments (In) and the avidin-bound fraction (Av) were first analyzed with anti-Myc antibodies to test for the presence of TRPMα2 (upper part of the blot). Then, the blot was stripped and reprobed with anti-HA antibodies. As a loading control of membrane proteins, the lower part of the blot was probed initially with anti-ATP1A1 antibodies and after stripping with anti-calnexin antibodies. The relative signal intensities of the input and the avidin-bound fractions are indicated. H, fluorescence ratios measured during application of 100 μm pregnenolone sulfate (100 PregS) of HEK293 cells co-transfected with 1.8 μg TRPM3α2/0.2 μg TRPM3α7 (10% TRPM3α7) or 1.8 μg TRPM3α2/0.2 μg empty vector (control). Each trace represents the average (±S.E.) of the number of cells indicated in brackets. I, changes of fluorescence in experiments performed as shown in H 48, 72, and 96 h post transfection of 10% TRPM3α7 (blue) or 20% TRPM3α7 (green). Values were normalized for each time point to control cells. Numbers of sample (colored) and control (black) cells analyzed on each day are indicated in brackets. Two independent experiments gave similar results. rel. ratio, relative ratio; calnex., calnexin.

In independent experiments, we transfected the TRPM3α2 or TRPM3α7 cDNAs into HEK293 cells stably overexpressing the Myc-tagged TRPM3α2 variant (HEKα2, Fig. 6, C, D, and F). Ca²⁺ signals were significantly reduced to ∼73% at 48 and 72 h and to ∼86% at 96 h after introduction of TRPM3α7 (Fig. 6, D and F), supporting the results obtained in Ins1 cells. Introduction of TRPM3α2, however, did not increase Ca²⁺ entry significantly (Fig. 6, C and F), probably because these cells already express maximal amounts of recombinant TRPM3α2 channels.

Next, we tested whether the introduction of TRPM3α7 affects the abundance of TRPM3α2 within the cells or at the cell surface. HEKα2 cells were transfected with plasmids allowing bicistronic expression of the HA-tagged TRPM3α7 and the green fluorescent protein. Eight million transfected, green fluorescent cells and non-transfected, non-fluorescent cells (control) were each separated by fluorescence activated cell sorting (supplemental Fig. 2) and analyzed by cell surface biotinylation. We found very similar amounts of NaK-ATPase (ATP1A1) in the total protein fraction (input) as well as in the avidin-bound fraction of cell surface proteins in HA-TRPM3α7-expressing cells (105%, Fig. 6G) and control cells (100%, Fig. 6G), confirming that identical numbers of cells have been analyzed. The amount of Myc-tagged TRPM3α2 was reduced in TRPM3α7-expressing cells both in the avidin-bound fraction of cell surface proteins (84%) and in the total protein lysate (79%), indicating that the expression of TRPM3α7 reduced the total number of functional TRPM3α2 subunits as well as their occurrence at the plasma membrane.

Depending on the tissue, ΔICF transcripts represent ∼2.5 to 15% of the total number of variants (Fig. 2). To simulate this situation, we analyzed the impact of ΔICF variants on TRPM3 channel function under more defined conditions. Keeping the total amount of plasmids constant (2 μg), we cotransfected 1.8 μg of TRPM3α2 and 0.2 μg of TRPM3α7 (representing 10%) or 1.6 μg of TRPM3α2 and 0.4 μg of TRPM3α7 (representing 20%) into HEK293 cells and analyzed Ca²⁺ signals after addition of a nearly saturating concentration of pregnenolone sulfate (100 μm, Fig. 6, H and I). Co-transfections of TRPM3α2 plasmids and empty vectors replacing TRPM3α7 served as controls. Already 10% of TRPM3α7 cDNA cotransfected with TRPM3α2 significantly reduced peak calcium signals to 88.0% after 48 h. With 20% TRPM3α7 cDNA, we found an even more pronounced reduction to 75.4%. The experiments demonstrate that naturally occurring amounts of ΔICF transcripts have a strong impact upon TRPM3-dependent calcium entry.

DISCUSSION

In the present study, we compared TRPM3 splice variants and used them as a guiding principle to identify a protein domain of TRPM3 ion channels that is indispensable for channel function. We showed that splicing of a part of exon 13 led to non-functional splice variants without any channel activity.

Splicing within Exon 13 Is Physiologically Relevant

The non-functional splice variant Na_v1.5f of voltage-dependent sodium channels accounts for nearly 50% of the total Na_v1.5 transcripts in human brain, and its generation has been suggested to limit the number of undesired functional channels (25). Likewise, a non-functional variant of TRPV1 channels has been suggested to act as a dominant-negative regulator of responses of sensory neurons to noxious stimuli (26). A similar scenario might also be true for TRPM3. As we showed, removal of the ICF domain by splicing within exon 13 is common in many tissues, and ΔICF transcripts are abundantly present in the brain where they contribute ∼15% to the total number of TRPM3 transcripts. This means that after splicing, 15% of primary TRPM3 transcripts do not contribute to the formation of functionally active TRPM3 subunits but rather encode non-functional variants. These non-functional variants reduce TRPM3-mediated signaling additionally by direct interference with intact TRPM3 subunits and impairment of the TRPM3 channel function. The degree of this additional impairment in brain tissue can be estimated from our experiments assuming that the properties of native and recombinant ΔICF proteins are similar. Cotransfection of 10 and 20% TRPM3α7 cDNA with TRPM3α2 revealed a reduction of Ca²⁺ signals of ∼12 and ∼25%, respectively. Therefore, one can expect ∼18.5% functional subunits to be affected in the presence of 15% ΔICF transcripts. Taken together, one can estimate that in the brain, 15% (spliced to ΔICF isoforms) plus 18.5% (affected by ΔICF isoforms) of the TRPM3 pre-mRNA does not contribute to the formation of operative TRPM3 channels. Thus, splicing of 54 nucleotides within exon 13 is a very effective tool to limit TRPM3-mediated calcium entry, and it is reasonable to assume that removal of the ICF domain by splicing has significant impact on biological functions attributed to TRPM3 such as pain perception and insulin release.

In accordance with this assumption, we also found higher amounts of ΔICF variants in Ins1 insulinoma cells growing in high extracellular glucose that were largely reduced when the cells were incubated in either low concentrations of glucose or in the presence of high glucose and pregnenolone sulfate. As a result, under the latter conditions the number of functional TRPM3 channels might be increased. Thus, in high glucose, pregnenolone sulfate may not only lead to enhanced insulin secretion by direct stimulation of TRPM3-mediated Ca²⁺ entry (6). In addition, it may trigger an amplifying pathway of insulin secretion by enhanced formation of functional TRPM3 channels.

The ICF Region Contains a Domain Essential for Protein Folding

Splice variants of the thyroid-stimulating hormone receptor display a loss of function phenotype because they have lost their capability to bind their ligand (27). However, a similar scenario for TRPM3ΔICF variants comprising a selective elimination of the binding site for pregnenolone sulfate is improbable as the ICF region is predicted to be located in the cytosol, and we recently showed that pregnenolone sulfate acts on TRPM3 channels from the extracellular side (6). Furthermore, the absence of the ICF region also affected nifedipine stimulation of the channel as well as TRPM3-mediated Ca²⁺ signals induced by the removal of sodium from the bath solution.

Alternative splicing can also modify posttranslational modifications of a protein. For example, isoform-dependent phosphorylation has been shown to change Kv4.3 potassium channel currents (28). The ICF region of TRPM3 contains a putative casein kinase 2 phosphorylation site at Ser⁵²¹, but we could show that an exchange of this residue to alanine did not affect TRPM3 channel activity.

Likewise, protein-protein interactions with auxiliary subunits might be affected by alternative splicing through deletion or creation of binding domains. Splicing of eight amino acids in the 200-kDa protein agrin strongly affected the binding of agrin to nicotinic acetylcholine receptors and rendered the shortened isoform biologically inactive in clustering of the receptors (29). Hitherto, auxiliary subunits of TRPM3 are unknown, and we found that removal of the ICF region of TRPM8 inactivates this channel as well. Thus, a role of the ICF region as binding site for unknown TRPM specific channel subunits is also rather unlikely.

Another important property of proteins that is regulated by alternative splicing is their insertion into membranes (30). By deleting transmembrane or membrane association domains, non-membrane bound isoforms can be generated. These soluble isoforms might lose the function of the membrane bound form but may serve as modulators. Such a mechanism has been proposed for TRPM1, the closest relative of TRPM3 (31). It has been reported that a short, N-terminal isoform devoid of any transmembrane segment, reduce constitutive Ca²⁺ entry of the long TRPM1 variant when coexpressed in HEK293 cells, most likely by inhibiting translocation of the long variant to the plasma membrane. Correspondingly, our surface biotinylation assays indicated that a loss of the ICF region decreases the presence of TRPM3 in the plasma membrane. Therefore, the ICF region might play a role for correct integration of the channel into the plasma membrane and/or trafficking of the channel complex to the cell surface.

However, a trafficking defect alone might be insufficient to explain the total loss of function of ΔICF variants. A reduced amount of TRPM3ΔICF variants still reached the plasma membrane but showed neither any Ca²⁺ signal nor any ionic current. This clearly demonstrates that the ICF region is essential for protein folding of functional TRPM3 channels. A change of the ternary structure of ΔICF proteins also explains very well their reduced capability to interact with each other as well as with functional TRPM3 variants because it might not only interfere with the formation of an functional channel pore but also might change the conformation of domains needed for the assembly of channel subunits. The structural change in the end might lead to reduced stability and enhanced degradation of the protein. Correspondingly, we found that not only the plasma membrane fraction but also the total amount of TRPM3α2 proteins were reduced when they were co-expressed with TRPM3α7.

Biotinylation experiments showed that the cell surface expression of ΔICF proteins was reduced to ∼25%, and therefore, the fraction of ΔICF proteins in a cell might be significantly smaller than the fraction of their transcripts. It might be 2% in HEK293 cells after cotransfection of 10% TRPM3α7cDNA and TRPM3α2. This means that only one TRPM3α7 protein might coexist with 49 functional TRPM3α2 proteins that together might form ∼12 tetrameric channels complexes. However, we still observed a significant reduction of Ca²⁺ signals. Therefore, our data imply that the presence of only one ΔICF subunit in a tetramer of TRPM3 proteins might be sufficient to obliterate the formation of a functional channel.

Replacement of the α-helical ICF region by 18 alanine residues was not sufficient to recover functional channels indicating that the ICF region does not simply serve as a spacer to bring other protein regions into their correct positions but rather constitutes a domain that communicates with other domains of the protein. Consistently, we identified 10 essential amino acid residues, with each of them exchangeable to alanine but indispensable when deleted. Taken together, our data provide strong evidence for a central function of the ICF domain as an intramolecular module that is essential for protein folding of single subunits as well as formation of functional homo- and heterooligomeric channel complexes. The model is in line with our finding that removal of the ICF region in TRPM8 inactivates this channel, too. In addition, this finding strongly suggests a general importance of the ICF domain for the folding and function of all TRPM channels.

How the ICF domain fulfills this central function remains an open question. We noticed that most of the essential residues in the ICF domain are leucine residues, which are highly conserved throughout the TRPM family (Fig. 4A). Within a region spanning the residues 168 to 790 of TRPM3α2, the occurrence of leucine residues is increased and makes up 13.5%. Using an online analysis resource (32), we found, that the residues Asp⁴⁹⁴ to Glu⁵²⁵ in TRPM3α2 strongly resemble known leucine-rich repeats of the protein family PF07723. Leucine-rich repeats form protein interaction domains, and all known leucine-rich repeat structures form curved solenoids (33). The ICF domain in TRPM3α2 is formed by the residues Leu⁵¹⁸ to Tyr⁵²⁷ and might be an essential part of a similar structure. Its removal by alternative splicing may change the curvature of this structure and may interfere in this way with the folding and the function of TRPM3 ion channels.