Alternative splicing switches the divalent cation selectivity of TRPM3 channels.

TRPM3 is a poorly understood member of the large family of transient receptor potential (TRP) ion channels. Here we describe five novel splice variants of TRPM3, TRPM3alpha1-5. These variants are characterized by a previously unknown amino terminus of 61 residues. The differences between the five variants arise through splice events at three different sites. One of these splice sites might be located in the pore region of the channel as indicated by sequence alignment with other, better-characterized TRP channels. We selected two splice variants, TRPM3alpha1 and TRPM3alpha2, that differ only in this presumed pore region and analyzed their biophysical characteristics after heterologous expression in human embryonic kidney 293 cells. TRPM3alpha1 as well as TRPM3alpha2 induced a novel, outwardly rectifying cationic conductance that was tightly regulated by intracellular Mg(2+). However, these two variants are highly different in their ionic selectivity. Whereas TRPM3alpha1-encoded channels are poorly permeable for divalent cations, TRPM3alpha2-encoded channels are well permeated by Ca(2+) and Mg(2+). Additionally, we found that currents through TRPM3alpha2 are blocked by extracellular monovalent cations, whereas currents through TRPM3alpha1 are not. These differences unambiguously show that TRPM3 proteins constitute a pore-forming channel subunit and localize the position of the ion-conducting pore within the TRPM3 protein. Although the ionic selectivity of ion channels has traditionally been regarded as rather constant for a given channel-encoding gene, our results show that alternative splicing can be a mechanism to produce channels with very different selectivity profiles.

TRPM3 is a poorly understood member of the large family of transient receptor potential (TRP) ion channels. Here we describe five novel splice variants of TRPM3, TRPM3␣1-5. These variants are characterized by a previously unknown amino terminus of 61 residues. The differences between the five variants arise through splice events at three different sites. One of these splice sites might be located in the pore region of the channel as indicated by sequence alignment with other, bettercharacterized TRP channels. We selected two splice variants, TRPM3␣1 and TRPM3␣2, that differ only in this presumed pore region and analyzed their biophysical characteristics after heterologous expression in human embryonic kidney 293 cells. TRPM3␣1 as well as TRPM3␣2 induced a novel, outwardly rectifying cationic conductance that was tightly regulated by intracellular Mg 2؉ . However, these two variants are highly different in their ionic selectivity. Whereas TRPM3␣1encoded channels are poorly permeable for divalent cations, TRPM3␣2-encoded channels are well permeated by Ca 2؉ and Mg 2؉ . Additionally, we found that currents through TRPM3␣2 are blocked by extracellular monovalent cations, whereas currents through TRPM3␣1 are not. These differences unambiguously show that TRPM3 proteins constitute a pore-forming channel subunit and localize the position of the ionconducting pore within the TRPM3 protein. Although the ionic selectivity of ion channels has traditionally been regarded as rather constant for a given channelencoding gene, our results show that alternative splicing can be a mechanism to produce channels with very different selectivity profiles.
The transient receptor potential (TRP) 1 gene family comprises at least 28 mammalian genes divided into seven subfamilies (1,2). Most of the encoded proteins exhibit common structural features such as six predicted transmembrane (TM) domains with a putative pore loop between TM5 and TM6 and the so-called TRP box after TM6 (1,2). Although all members of this group have been reported to form cationic channels, their mechanisms of activation, their regulation, and their biological functions are remarkably diverse. They also display a large variety of different cation selectivities (1,2). For example, TRPM4 and TRPM5 have been described as impermeable for divalent cations (3)(4)(5), whereas TRPV5 and TRPV6 appear to be exclusively permeable for Ca 2ϩ (6,7). The diversity of TRP channels is further increased by the fact that most members of the TRP gene family can give rise to several different transcripts due to alternative splicing (8). In a few cases, the functional consequences of these alternative splice events are now beginning to emerge. For example, missplicing of TRPM6 transcripts is associated with a hereditary disorder called hypomagnesemia with secondary hypocalcemia (9,10), and an amino-terminal-truncated variant of TRPM4 appears to modulate Ca 2ϩ oscillations after receptor stimulation in T lymphocytes (11).
However, up to now, the largest number of different splice variants for any TRP family member has been described for TRPM3 (12,13). Lee et al. (12) reported six splice variants of human TRPM3, which they named TRPM3a-f. Their lengths range from 1544 to 1579 amino acid residues. Functional data were only reported for the TRPM3a splice variant. Heterologously expressed TRPM3a exhibits a constitutive, Ca 2ϩ concentration-dependent Ca 2ϩ entry that can be augmented by Ca 2ϩ store depletion or by stimulation of muscarinic receptors (12). In addition, Grimm et al. (13) reported one further human variant that has a total of only 1325 residues because of its considerably shorter carboxyl terminus (13). On its amino terminal end, however, this variant, which we will refer to as TRPM3 1325 throughout this report, possesses an additional 153 amino acids completely missing in the other reported TRPM3 variants. As is the case with TRPM3a, TRPM3 1325 generates constitutively active, Ca 2ϩ -permeable channels when heterologously expressed in human embryonic kidney 293 (HEK-293) cells. The activity of TRPM3 1325 channels can reportedly be enhanced by hypo-osmotically induced cell swelling (13) and by D-erythro-sphingosine (14).
In this report we describe five novel TRPM3 splice variants that we cloned as full-length cDNA constructs from mouse brain. These variants are 1699 -1721 amino acid residues long and are characterized by a novel amino terminal sequence of 61 residues not described so far. Interestingly, two of the novel variants, designated TRPM3␣1 and TRPM␣2, differ only in a region between the fifth and sixth transmembrane domains where the pore-forming region of the channel is assumed. This region is 12 amino acids longer in TRPM3␣1 compared with TRPM3␣2; additionally, a proline is substituted by an alanine. The principal finding of our study is that this alteration of the primary structure induces a large change in the ionic selectivity of the resulting ion channels. We thus identify alternative splicing as a mechanism to modify the selectivity of TRPM3 channels. A, 5Ј-GAG  AGC TGA GCG CAG GCT G-3Ј, and Z, 5Ј-TCC TGC AAC ACA CGG  TAA GCC-3Ј, were used to amplify Trpm3 transcripts after oligo(dT) 18 primed reverse transcription of mouse brain total RNA. For amplification we used the Long Expand PCR kit (Roche Applied Science) and the following conditions: addition of the enzymes at 62°C followed by 2 min at 94°C, 10 cycles (94°C, 10 s; 62°C, 30 s; 68°C, 5 min), 25 cycles (94°C, 10 s; 62°C, 30 s; 68°C, 5 min ϩ 20 s/cycle) and 7 min at 68°C. Sequencing both strands of subcloned fragments identified five independent clones encoding TRPM3␣1, three encoding TRPM3␣2, and three encoding TRPM3␣3. Regarding TRPM3␣4 and TRPM3␣5, one clone was found for each splice variant (Fig. 1C). We introduced the ribosome binding site ACC GCC ACC and a Myc tag in-frame immediately 5Ј to the start codon of Trpm3␣1 and Trpm3␣2 cDNAs, which were subsequently cloned into pCAGGS-IRES-GFP (15) for transient dicistronic expression of TRPM3 together with the green fluorescent protein (GFP). To generate stably transfected clones the TRPM3␣1 and TRPM3␣2 cDNAs were fused in-frame to the enhanced yellow fluorescent protein (EYFP) cDNA in the pEYFP-N1 vector (BD Biosciences Clontech) allowing expression of TRPM3-EYFP fusion proteins. The sequences of Trpm3␣1, Trpm3␣2, Trpm3␣3, Trpm3␣4, and Trpm3␣5 have been submitted to GenBank TM under accession numbers AJ544532, AJ544534, AJ544533 AJ544535, and AJ544536.

Cloning of TRPM3 Variants-Oligonucleotide primers
Cell Culture, Transfection, and Generation of Stable Cell Lines-HEK-293 cells were grown in minimal essential medium supplemented with 10% fetal calf serum in a humidified atmosphere containing 5% CO 2 . Cells were passaged 3 times a week until they reached passage 30. For transient transfection, cells grown to 70 -80% confluence were transfected with TRPM3␣1, Myc-tagged TRPM3␣1, Myc-tagged TRPM3␣2, or pCAGGS-IRES-GFP (mock) using PolyFect (Qiagen, Hilden, Germany). Cells were either used directly or passaged 24 h after transfection to reduce cell density. For generation of clones stably expressing Myc-tagged TRPM3␣1-EYFP or TRPM3␣2-EYFP fusion proteins, transfected cells were kept for 5 weeks in 500 g/ml G418. Single surviving cells were isolated by cell sorting using a MoFlo fluorescent-associated cell sorter (DakoCytomation, Hamburg, Germany) and expanded. Cell clones were tested for their EYFP fluorescence and for expression of TRPM3 in Western blots using monoclonal anti-Myc antibodies and anti-GFP antibodies (Roche Applied Science).
Detection of TRPM3 mRNA and Proteins-The 3450-and 1684-bp fragments, covering the entire coding sequence of the mouse Trpm3␣1 cDNA, were labeled by random priming with [␣ 32 P]dCTP and served as probes for Northern blots. Filters were exposed to x-ray films with intensifying screens at Ϫ80°C for 2 weeks before they were stripped and hybridized with a 239-bp cDNA fragment of the human glyceraldehyde-3-phosphate dehydrogenase as control.
In situ hybridizations were performed on 20-m cryosections of perfusion-fixed brain (16) and fresh brain from C57 Bl6/J ϫ 129 SVJ or NMRI mice with similar results. Sections from fresh tissue were air dried for 30 min, fixed for 20 min in 4% paraformaldehyde solution, washed three times in phosphate-buffered saline, and finally dehydrated and defatted in 70 and 100% ethanol. Oligonucleotide probes 5Ј-A GTG GGA GTG CAT GCT ATT GAG AAC GGT GA-3Ј and 5Ј-TT TGA GGG CCC ATG TCT CGT ATT GTA CAG TTC CTC TAG TC-3Ј labeled with [␣ 35 S]dATP using terminal deoxynucleotide transferase revealed the same distribution pattern of signals. Control reactions with a 10-fold excess of unlabeled probe resulted in background grain densities. Specimens were exposed to x-ray films for 5 days and then to Kodak NTB2 photoemulsion for 3 weeks.
Western blots were performed as described (17). Cells from four 3-cm dishes grown to 60 -70% confluence were lysed in 200 l of Laemmli buffer and analyzed using monoclonal anti-Myc antibodies and anti-GFP antibodies.
Electrophysiological Recordings-Standard tight-seal, whole cell patch clamp recordings were performed with an EPC9 amplifier controlled by Pulse software (HEKA, Lambrecht, Germany). Electrodes had resistances of 2.5-5 M⍀ measured in standard bath solution and were coated with Sigmacote (Sigma). TRPM3-expressing cells were identified by their green fluorescence on an inverted microscope (Axiovert 135; Zeiss) and used 24 -96 h after transfection. Cells were voltageclamped at Ϫ15 mV, and the instantaneous current-voltage relation-ship was probed at 1-s intervals with fast (1 mV/ms) voltage ramps ranging from Ϫ115 mV to ϩ85 mV. During recordings that lasted longer than 60 s, the interval between voltage ramps was often increased to 10 s. From the voltage ramps, the current at Ϫ80 mV and ϩ80 mV and the reversal potential were obtained by offline analysis.
Cells were maintained during the recordings at room temperature in Ringer solution. All extracellular solutions (Table I) were based on 10 mM HEPES; pH was adjusted to 7.2 (hydroxide of main cation and HCl) and osmolality to 315-335 mosmol/kg (glucose or H 2 0). Solutions were applied for short duration by a custom-made local perfusion system whose outlet was placed at Ͻ200 m from the recorded cell.
Fluorescent Ca 2ϩ Measurements-[Ca 2ϩ ] i measurements of HEK-293 cells using fura-2-AM were performed as described previously (18). In brief, HEK-293 cells and HEK cell clones stably expressing Myc-tagged TRPM3-EYFP fusion proteins were plated on poly-L-lysine-coated coverslips to ϳ50 -70% confluence and measured 48 -72 h later using an imaging system (T.I.L.L. Photonics, Martinsried, Germany). Cells were loaded with 5 M fura-2-AM for 30 min at room temperature, washed in Ringer solution containing (in mM) 138 NaCl, 5.4 KCl, 2 CaCl 2 , 2 MgCl 2 , 20 glucose, 10 HEPES, (pH 7.2 adjusted with NaOH, 315 mosmol/kg) for 20 min, and analyzed in a perfusion chamber (Warner Instruments, Hamden, CT). Ratio images were acquired every 3 s and analyzed online. Ratio values were converted to free Ca 2ϩ concentration after calibration of the setup using a calibration kit (Molecular Probes, Eugene, OR). Averages and S.E. values were calculated from traces obtained from individual cells and smoothed with a sliding window function (window size: 15 s) with the Origin data analysis software (Microcal, Northampton, MA). All reagents used in electrophysiological or Ca 2ϩ imaging experiments were from Sigma. Results are given as mean Ϯ S.E., and the two-tailed Student's t test was used to calculate p values.

Identification and Cloning of TRPM3 Variants from Mouse
Brain-Screening public databases we identified two partial human TRPM3 cDNA clones, AL136545 and XM_0136545 (Fig.  1C). Both clones derive from a single gene locus on human chromosome 9. The corresponding mouse gene shows a highly similar organization with 28 exons and spans more than 850 kb on mouse chromosome 19b (NCBI gene number 226025, Fig. 1A). Primers A and Z ( Fig. 1A) were deduced from the mouse genomic sequence. Their sequences are located 5Ј and 3Ј to stop codons flanking an open reading frame with a predicted translation start that is also present in the human cDNA of clone AL136545. Using this primer combination, ϳ5.2-kbp cDNA fragments from mouse brain total RNA were amplified by reverse transcription PCR and subcloned (Fig. 1B). Similar products were amplified from mouse eye (data not shown). Each clone contained the complete coding sequence of mouse Trpm3. We identified five different splice variants, designated TRPM3␣1, TRPM3␣2, TRPM3␣3, TRPM3␣4, and TRPM3␣5 (Fig. 1C). Their amino acid sequences comprise 1699 up to 1721 amino acids, which are encoded by exons 1, 3-7, and 9 -28. The predicted exons 2 and 8 (NCBI gene number 226025) are not present in the cDNA of these TRPM3 variants from mouse, but 18 amino acid residues encoded by the corresponding exon 8 have been detected in the recently described human hTRPM3f variant (12). Exon 2 encodes a 59-amino acid sequence that corresponds to the amino terminal part of the human TRPM3 1325 variant (13) but is absent in the a-f variants of hTRPM3. Exon 1 codes for a completely novel amino terminal sequence that is not present in any of the already described human TRPM3 variants. Among 13 independent mouse cDNA clones, we could not find a single one that contained exon 1 together with 2 ( Fig. 1C), suggesting that these exons are expressed in a mutually exclusive fashion. This could be explained either by tightly regulated posttranscriptional processing or by expression of these variants from alternative promoters.
Starting from residue 156, however, mouse TRPM3 proteins show ϳ97% amino acid sequence identity to the human variants. Therein, mouse TRPM3␣1 corresponds to human TRPM3c, mTRPM3␣2 to hTRPM3a, mTRPM3␣3 to hTRPM3b, mTRPM3␣4 to hTRPM3e and mTRPM␣5 to hTRPM3d (Fig.  1C). The close correspondence between the respective mouse and human transcripts indicates that not only the amino acid sequence but also the splice events are highly conserved between mouse and human. Such strong conservation is highly suggestive of functional importance of TRPM3 proteins generally but also of their regulation and modification by alternative splicing.
Expression Pattern of TRPM3-We subsequently analyzed the expression pattern of the mouse Trpm3 gene by Northern blot and by in situ hybridization (Fig. 2). We found Trpm3 expression in brain and eye with transcripts of ϳ1.8, ϳ2.6, ϳ3.7, ϳ5.8, ϳ7.6, ϳ9.4, Ͼ12, and Ͼ15 kb. (Fig. 2A). In agreement with previous results (13), we could not detect Trpm3 transcripts in mouse kidney by Northern blot (data not shown). In situ hybridization experiments showed Trpm3 transcripts in several regions of the mouse brain such as the dentate gyrus, the intermediate lateral septal nuclei, the indusium griseum, and the tenia tecta (Fig. 2, C and D). Strongest Trpm3 expression was found in the epithelial cells of choroid plexus (Fig. 2, B-D) where transcripts could readily be detected in 20 g of total RNA ( Fig. 2A).
TRPM3␣1 and TRPM3␣2 Form Cation Channels Regulated by Intracellular Mg 2ϩ -Alternative splicing within exon 24 of both the mouse and the human TRPM3 gene leads to the presence of 12 additional amino acid residues and the addi-tional replacement of an alanine by a proline residue (Fig. 1D) (12). Interestingly, this domain, which is present in mTRPM3␣1 and hTRPM3c (12) but absent in all other variants, lies between the presumed fifth and sixth transmembrane domains (Fig. 1C). In analogy to the topologically similar members of the TRPV subfamily (19), it is likely that this part of the protein contributes to the pore-forming region of the channel (Fig. 1D). The variants TRPM3␣1 and TRPM3␣2 differ only in this region (Fig. 1, C and D). We set out to test whether this change in the primary sequence altered the biophysical characteristics of the resulting ion channels. To this end, we expressed TRPM3 variants in HEK-293 cells, which do not express TRPM3 endogenously ( Fig. 2A).
Starting with TRPM3␣1, we noticed that transfected cells, but not control cells, exhibit a constitutively active, outwardly rectifying current. This current was already visible directly after establishing the whole cell configuration (Fig. 3, A and B). Replacing the extracellular cations by the impermeant cation N-methyl-D-glucamine (NMDG ϩ ) shifted the reversal potential from Ϫ9.5 Ϯ 0.5 to Ϫ66.2 Ϯ 0.8 mV (n ϭ 5), indicating that the current was predominately carried by cations (Fig. 3A). When we used pipette solutions without Mg 2ϩ , TRPM3␣1-induced currents increased strongly within the first minutes of whole cell recording (Fig. 3, B and C). Conversely, when we increased the free Mg 2؉ concentration in the recording pipette to 9 mM (Fig. 3, C and D), TRPM3␣1-induced currents vanished rapidly. Using 0.9 mM free Mg 2ϩ in the recording pipette, currents of intermediate size were observed (Fig. 3D). These data indicate that TRPM3␣1 is regulated by physiological concentrations of free intracellular Mg 2ϩ , which typically is found to be in the order of 0.5-1 mM (20).
We next turned to TRPM3␣2 and asked whether this variant also forms a functional channel. In extracellular Ringer solu-FIG. 1. Identification of TRPM3 variants from mouse brain. A, schematic diagram of the mouse Trpm3 gene, comprising 28 exons. Location of primers A and Z used to amplify the cDNA by reverse transcription PCR and stop codons in-frame to the translation start are indicated. B, reverse transcription PCR amplification of 5.2-kbp fragments using primers A and Z. C, schematic presentation of TRPM3 with transmembrane domains 1-6, coiled coil region (cc), and TRP homology domain (Trp). Novel mouse TRPM3 protein variants shown as thick black lines are compared with the human variants hTRPM3a-f (12) and hTRPM3 1325 (13). The numbers of amino acid residues of each variant are indicated in parentheses. D, putative pore regions of TRPM3␣1 and TRPM3␣2 compared with the corresponding mouse sequences of TRPM6 (accession number NP_700466), TRPM7 (accession number NP_067425), TRPV5 (accession number JC7795), and TRPV6 (accession number CAD62684). The 12 additional amino acid residues present in TRPM3␣1 are indicated. Identical residues are boxed in black, conserved in gray. An aspartate residue that determines Ca 2ϩ permeation of the TRPV5/TRPV6 pore is marked by an asterisk. Residues proposed to build the selectivity filter of TRPV6 are underlined (29). tion, TRPM3␣2 currents were small and increased only slightly during prolonged recording (Fig. 4A). However, when we again replaced all extracellular cations with NMDG ϩ , we observed large outward currents, which disappeared as soon as standard conditions were re-established (Fig. 4, A and B). Such currents were not observed in control cells measured under identical conditions. The absence of detectable inward currents under these conditions indicates that, similar to TRPM3␣1, the TRPM3␣2-induced conductance is mainly permeable to cations. We then tested whether intracellular Mg 2ϩ also regulated TRPM3␣2 channels. Very similar to our results with TRPM3␣1, we found large currents in the absence of intracellular Mg 2ϩ but no detectable currents with 7 mM free Mg 2ϩ in the pipette solution (Fig. 4C). Therefore, block by intracellular Mg 2ϩ appears to be a common property of the two splice variants.
Block by intracellular Mg 2ϩ , however, is not a unique feature of TRPM3 channels but has also been described for the related ion channels TRPM6 (21) and TRPM7 (22). TRPM7 has been shown to be endogenously expressed in HEK-293 cells and is thought to be the molecular correlate of an endogenous current that has been named MagNuM (22) or MIC (23). In the experiments presented above we have taken several measures to ensure that the contribution of the endogenous MagNuM/MIC current does not compromise our data. 1) We verified that current densities in control cells examined under identical conditions were much smaller than current densities in transfected cells (Figs. 3B and 4, A and B). 2) Measurements with divalent free extracellular solutions (NMDG solution, Table I Despite these precautions, however, the formal possibility exists that the observed outwardly rectifying currents do not flow through TRPM3 proteins but result from up-regulation of endogenous TRPM7 channels. We found, however, that the currents in TRPM3␣1-expressing cells exhibited a time-dependent facilitation upon depolarization to values higher than ϩ60 mV (Fig. 3E, arrows), similar to those described for TRPM4, TRPM5, and TRPM8 (4,5,24). Channels encoded by TRPM7 or TRPM6 do not show such a voltage-dependent facilitation (21,25). Also, when we exposed TRPM3␣2-expressing cells to an extracellular solution containing only NMDG ϩ (to evoke the large outward currents), a voltage-dependent facilitation of the current could be observed (Fig. 4D, arrows). Therefore, TRPM3␣1 and TRPM3␣2 channels show a unique combination of biophysical properties that are not found in any other member of the TRPM channel family.
Splicing Changes the Ion Selectivity of TRPM3 Channels-To determine the ionic selectivity of the TRPM3␣1 and TRPM3␣2 splice variants, we measured TRPM3-dependent currents under bi-ionic conditions and analyzed the reversal potential (Fig. 5). Fig. 5B shows that the reversal potential is significantly more positive in cells expressing TRPM3␣2 compared with TRPM3␣1-expressing cells when Ca 2ϩ and NMDG ϩ are the only cations present in the extracellular solution. Similar results were obtained with Mg 2ϩ instead of Ca 2ϩ . An estimation of the relative permeability ratios utilizing the Goldman-Hodgkin-Katz formalism (26) showed that TRPM3␣2 channels are at least 10 times more permeable for Ca 2ϩ (Fig.  5C) and at least 100 times more permeable for Mg 2ϩ than TRPM3␣1 channels (Fig. 5D). This large difference in ion permeation properties establishes that the ion-conducting pore of TRPM3 channels is affected by the sequence differences between the two splice variants ␣1 and ␣2 (Fig. 1C). It therefore provides evidence that this region of the protein is part of the ion-conducting pore of TRPM3. Furthermore, this finding proves conclusively that the observed currents after overexpressing TRPM3 proteins are mediated by these proteins and are not the result of up-regulated, endogenous conductances.
Divalent Inward Currents through TRPM3␣2-The higher selectivity for divalent cations of the TRPM3␣2 variant does not necessarily imply that this variant is capable of conducting sizeable divalent inward currents. However, when we applied high extracellular concentrations of either Ca 2ϩ or Mg 2ϩ (Fig.  6), TRPM3␣2-expressing cells conducted inward currents of up to 1 nA (at Ϫ80 mV, Fig. 6). Such currents were seen neither in TRPM3␣1-expressing cells nor in control cells. These data directly demonstrate that, in sharp contrast to TRPM3␣1, TRPM3␣2-encoded channels not only have a rather high selectivity for divalent cations but are also capable of conducting sizeable divalent currents.
TRPM3 Channel Variants Are Differentially Regulated by Extracellular Cations-Comparison of TRPM3␣1 and TRPM3␣2 outward currents obtained in extracellular Ringer solution and a solution containing NMDG ϩ as the only cation implied that outward currents through TRPM3␣2 are inhibited by cations in the Ringer solution (Figs. 3A and 4B). We were intrigued by the fact that this block affected only outward currents through TRPM3␣2 but not through TRPM3␣1. We therefore investigated the nature of this inhibition using solutions that contained only one permeable cation in addition to NMDG ϩ . We found that Ca 2ϩ blocks outward currents through TRPM3␣2 as well as through TRPM3␣1 dose dependently (Figs. 5A and 7B). Similar results were obtained for Mg 2ϩ (Fig. 7B). On the other hand, the monovalent cations Na ϩ and K ϩ quite strongly reduced TRPM3␣2 outward currents, whereas neither of them affected currents through TRPM3␣1 (Fig. 7, A and B).
TRPM3␣2 thus is inhibited by all cations tested on the extracellular side. It is important to note, however, that at their respective physiological concentrations, none of the cations blocked outward currents through TRPM3␣2 completely. Nevertheless, these results imply that TRPM3 channel activity is tightly regulated by the concentration of extracellular cations. While TRPM3␣1 channels are only sensitive to divalent cations, TRPM3␣2 channels are sensitive to all extracellular cations.
Expression of TRPM3 Variants Increases Resting Ca 2ϩ Levels-We compared intracellular Ca 2ϩ concentrations in untransfected control cells to HEK-293 cells stably expressing either TRPM3␣1 or TRPM3␣2 at similar levels (Fig. 8A). In Ringer solution containing 2 mM Ca 2ϩ , TRPM3␣2-expressing cells showed significantly (p Ͻ 0.001) increased steady state Ca 2ϩ levels of 145 Ϯ 2.6 nM compared with TRPM3␣1-expressing cells (114 Ϯ 2.3 nM) and control cells (84 Ϯ 1.6 nM) (Fig. 8B). When we exposed the cells to an extracellular solution devoid of Ca 2ϩ (containing 2 mM EGTA), intracellular Ca 2ϩ levels dropped to 70 -80 nM, irrespective of the expression of TRPM3 channels (Fig. 8B). Upon re-addition of 2 mM extracellular Ca 2ϩ , TRPM3␣2-expressing cells showed an overshooting rise of the intracellular Ca 2ϩ concentration (Fig. 8B). Such an overshoot was seen neither in TRPM3␣1-expressing cells nor in control cells (Fig. 8B). These data indicate that TRPM3␣2expressing cells possess an increased permeability for Ca 2ϩ ions under physiological extracellular ionic conditions that likely causes the larger basal intracellular Ca 2ϩ levels. On the other hand, TRPM3␣1-expressing cells seem to have a reduced permeability for Ca 2ϩ (compared with TRPM3␣2-expressing cells) as witnessed by the reduced Ca 2ϩ influx after extracellular Ca 2ϩ re-addition and the lower basal intracellular Ca 2ϩ concentration. These observations agree with our electrophysi-  Ringer  145  3  10  0  2  2  10  NMDG  0  0  0  145  0  0  40  1 Ca  0  0  0  145  1  0  40  ological data that TRPM3␣2 channels have a much larger divalent permeability compared with TRPM3␣1 channels (Figs. 5 and 6). They also correspond well with previous observations that the human TRPM3a and TRPM3 1325 variants, which both contain the same pore region as TRPM3␣2, mediate Ca 2ϩ entry under similar experimental conditions (12,13). DISCUSSION In this report, we have described five novel splice variants expressed in mouse brain and eye tissues that originate from the Trpm3 gene. Presently, four regions of modifications within the protein are known, and, additionally, variations of the length and sequence of the amino-and carboxyl-terminal regions have been described (Refs. 12 and 13 and this report). The Trpm3 gene therefore potentially encodes for a plethora of different proteins, 12 of which have up to now been verified experimentally. Here, we have focused on splice variants that differ only in a domain predicted to contribute to the ionconducting pore.
Alternative Splicing Switches the Ion Selectivity of TRPM3 Channels-The selectivity of ion channels is thought to be determined by the geometry and charge distribution of the selectivity filter, usually envisioned as the narrowest part of the channel pore (26). Typically, all members of an ion channel family, such as voltage-gated Na ϩ , K ϩ , or Ca 2ϩ channels, share common ionic selectivities. The TRP family of ion channels is already somewhat unusual in this respect as it encompasses members with quite diverging cationic selectivity profiles (reviewed in Ref. 1). The Trpm3 gene adds extra complexity to this picture, because two channels can be expressed from this gene with entirely different ionic selectivities. One channel, TRPM3␣1, preferentially conducts monovalent cation influx, whereas TRPM3␣2 strongly favors divalent entry. In vivo, such a change in ionic selectivity must be expected to have considerable consequences for the function of the channel and the  Table I) in order to block endogenous K ϩ -selective channels. physiology of the cell that expresses it.
Previously, it has been shown that subtle manipulations of key amino acid residues can substantially influence the divalent selectivity of ion channels. Voltage-gated Na ϩ channels that are transformed to conduct Ca 2ϩ ions by exchanging a lysine and/or an alanine to a glutamate residue are an extreme example (27). However, these changes in the primary amino acid sequence have been introduced artificially, whereas the variations in selectivity we describe for TRPM3 have evolved naturally and occur in vivo. Although alternative splicing is increasingly recognized as a potent mechanism to dynamically modify ion channel properties (28), it has not yet been implicated in directly modifying the selectivity of a channel. Interestingly, Trpm1, the closest relative of Trpm3, also encodes splice variants that differ in the pore-forming region (8) and might therefore represent a further example for such a mechanism.
Editing of RNA is another mechanism of posttranscriptional modification occurring in vivo. Deamidation of a selected adenosine in primary transcripts of non-NMDA ionotropic glutamate receptors leads to the substitution of a glutamine by an arginine within the channel pore. This has been shown to reduce the Ca 2ϩ selectivity of those channels with important consequences for their physiology (27).
Locating the Ion-conducting Pore in TRPM Channels-The switch of ionic selectivity in TRPM3 variants is brought about by removing a short stretch of 12 amino acid residues and exchanging 1 further residue within the linker domain between the presumed fifth and sixth transmembrane regions (Fig. 1B). The differences in ion selectivity seen for the TRPM3 splice variants strongly indicate that this linker domain constitutes the pore of TRPM3. Although this domain could already be suspected to be the ion-conducting pore, due to direct evidence obtained for TRPV1, TRPV4, TRPV5, and TRPV6 channels (19), this prediction has not been confirmed up to now for any member of the TRPM subfamily.
Compared with the presumed pore regions of other members of the TRP family, the pore loop of TRPM3 is considerably longer by 8 (TRPM3␣2) and 20 (TRPM3␣1) additional amino acid residues (Fig. 1C). The domains that build the proposed selectivity filter of the Ca 2ϩ -selective TRPV5/V6 channels (29) are conserved in TRPM3 proteins. The splicing within the TRPM3 channel pore introduces additional, positively charged amino acid residues into this domain. This might decrease the Ca 2ϩ permeability of TRPM3␣1 compared with TRPM3␣2, perhaps simply because of increased electrostatic repulsion. In line with this reasoning, in AMPA and kainate receptors the aforementioned replacement of a glutamine by a positively charged arginine residue by RNA editing reduces the Ca 2ϩ selectivity of those channels as well (30). Conversely, artificially introducing negatively charged glutamate residues in the pore of Na ϩ channels increases their divalent selectivity (27).
Block of TRPM3 Channels by Intra-and Extracellular Cations-Our data show that both TRPM3␣1 and TRPM3␣2 are regulated by physiological concentrations of intracellular Mg 2ϩ , similar to related members of the TRPM family such as TRPM6 and TRPM7 (21,22). Previously, the short human variant TRPM3 1325 has been reported to mediate increased calcium entry when the osmolality of the extracellular solution was shifted from 300 to 200 mosmol/kg (13), a stimulus that induces considerable swelling of the cells. Conceivably, the Mg 2ϩ dependence that we observed in whole cell patch clamp experiments offers a mechanistic explanation of the sensitivity to changes in osmolality. Because the intracellular cation concentration and the cell volume are mutually and inversely related, regulation of TRPM3 activity by intracellular Mg 2ϩ might explain why TRPM3 shows increased or reduced activity in hypotonic and hypertonic solutions, respectively.
Our experiments were performed in isotonic solutions. Under these conditions we found that TRPM3 currents are also tightly regulated by extracellular cations. However, only TRPM3␣2 channel activity was influenced by the extracellular concentration of monovalent cations, whereas both TRPM3 variants were inhibited by high extracellular divalent concentrations. Block by extracellular Na ϩ is a highly uncommon feature of ion channels, but not entirely unprecedented. Inward rectifier (31) and, especially, HERG (32) potassium channels have also been shown to be inhibited by extracellular Na ϩ . Although the block of TRPM3␣2 by extracellular Na ϩ concentrations in the physiological range seemed to be severe in electrophysiological recordings (Fig. 7), it does not appear to be complete. The intracellular Ca 2ϩ concentration in TRPM3␣2overexpressing cells was elevated (Fig. 8), indicating that, also under physiological extracellular conditions, Ca 2ϩ can enter the cell through TRPM3␣2 channels at a rate too low to be detectable in electrophysiological recordings but measurable in Ca 2ϩ imaging experiments.
Functional Role of TRPM3-We found that TRPM3 is strongly expressed in the choroid plexus (Fig. 2). This tissue in the ventricles of the brain is responsible for the formation and the regulation of cerebrospinal fluid. It has been demonstrated that the concentrations of ions such as Ca 2ϩ in cerebrospinal fluid are carefully regulated and are independent of variations in the plasma concentrations of these ions (33). The high expression in the choroid plexus might indicate involvement of TRPM3 proteins in the production of cerebrospinal fluid or the regulation of its ionic composition ( Fig. 2A). At present we do not know which of the various TRPM3 variants are present in the choroid plexus. However, the high selectivity for divalent cations of TRPM3␣2 channels makes this variant a good candidate to play a role in the regulation of divalent cation concentration in cerebrospinal fluid.