Cardiac L-type calcium channel beta-subunits expressed in human heart have differential effects on single channel characteristics.

l-Type calcium channels are multiprotein complexes composed of pore-forming (CaV1.2) and modulatory auxiliary alpha2delta- and beta-subunits. We demonstrate expression of two different isoforms for the beta2-subunit (beta2a, beta2b) and the beta3-subunit (beta3a, beta3trunc) in human non-failing and failing ischemic myocardium. Quantitatively, in the left ventricle expression of beta2b transcripts prevails in the order of > beta3 >> beta2a. The expressed cardiac full-length beta3-subunit is identical to the beta3a-isoform, and beta3trunc results from deletion of exon 6 (20 nn) entailing a reading frameshift and translation stop at nucleotide position 495. In failing ischemic myocardium beta3trunc expression increases whereas overall beta3 expression remains unchanged. Heterologous coexpression studies demonstrated that beta2 induced larger currents through rabbit and human cardiac CaV1.2 pore subunits than beta3 isoforms. All beta-subunits increased channel availability at single channel level, but beta2 exerted an additional, marked stimulation of rapid gating (open and closed times, first latency), leading to higher peak current values. We conclude that cardiac beta-subunit isoforms differentially modulate calcium inward currents because of regulatory effects within the channel protein complex. Moreover, differences in the various beta-subunit gene products present in human heart might account for altered single channel behavior found in human heart failure.

fibrillation or hypertrophic obstructive cardiomyopathy Ca v 1.2subunit expression is altered, but in human diastolic or end stage heart failure most studies have demonstrated no change (see Refs. [1][2][3][4]. Our studies, however, revealed alterations of auxiliary ␤-subunit expression, which is reduced in diastolic heart failure of cardiac allografts (3) and possibly in end stage heart failure (5). Down-regulation of ␤-subunit mRNA expression is a phenomenon not restricted to cardiac tissue, because similar observations were made in pancreatic islets of diabetic rats (6). Although an altered stoichiometry of ␤-subunits relative to the Ca v 1.2-subunits theoretically suggests variation in calcium inward current, we and others have detected no change in whole-cell calcium inward currents in cardiomyocytes isolated from human failing heart. Instead, calcium channel inward current was even paradoxically increased at the single channel level and could not be further augmented by cAMPdependent phosphorylation (5). Because products from all four ␤-subunit genes modulate the ion conducting pore in a specific manner (7), we decided to study cardiac ␤-subunit gene expression in the human heart as a plausible reason for increased single calcium channel activity in heart failure.
In human myocardium expression of three different ␤-subunits genes (␤ 1 -␤ 3 ) (8 -10) was demonstrated at the mRNA and the protein level (3,10). In human non-failing (NF) 1 left ventricular (LV) mRNA our RT-PCR experiments with degenerated primers complementary to coding sequences with high similarity between ␤ 2 -and ␤ 3 -genes generated three amplification products of 650, 503, and 483 bp. The 650-and 503-bp amplification products are sequence-identical to ␤ 2 and ␤ 3 coding sequences (␤ 2 : GenBank TM accession number AF423189, nn 97-746; ␤ 3 : GenBank TM accession number X76555, nn 135-637). The 483-bp amplification product contained a deletion of 20 nucleotides matching exon 6 of the ␤ 3 -gene (11). At least two different isoforms of the human ␤ 2 -subunit gene are expressed in the human heart (␤ 2a , ␤ 2b ) (12,13) that differ with respect to their short amino termini. Our Northern blot experiments revealed transcripts for ␤ 2a -, ␤ 2b -, and ␤ 3 -subunits in human heart tissue; therefore, we assessed quantitative expression of these ␤-subunits by real-time PCR, which demonstrated substantially different expression levels with ␤ 2b Ͼ ␤ 3 Ͼ Ͼ ␤ 2a . To investigate the physiological effects of these ␤-subunits, we first cloned both the full-length coding sequence of the ␤ 3subunit (␤ 3a ) and the isoform containing the deletion of exon 6 (20 nn) (␤ 3trunc ) from human heart. Deletion of exon 6 results in truncation of the protein, and, interestingly, ␤ trunc is up-regulated in heart failure. Because of the high homology of the rabbit ␤ 2a -subunit to the human ␤ 2b we used this subunit together with the two cloned human ␤ 3 -subunit isoforms to study the functional impact of these ␤-subunits onto the calcium inward current at the single channel level. Electrophysiological studies were designed to answer the following questions. 1) Do human cardiac ␤ 3 -subunits functionally coexpress with pore subunits from rabbit heart (Ca v 1.2a), rabbit smooth muscle (Ca v 1.2b), and human heart (Ca v 1.2)? Is there a difference in quality and extent of their modulation when compared with a ␤ 2 -subunit? 2) Does alternative splicing of the ␤ 3 -subunits in heart failure explain the increase in single channel activity? These questions were addressed by transient coexpression in cell lines stably expressing Ca V 1.2 isoforms.

EXPERIMENTAL PROCEDURES
Tissue Specimens-After local ethics committee approval, human myocardium specimens were obtained from explanted NF hearts not transplanted for technical reasons or from patients receiving orthotopic heart transplantation because of end-stage heart failure due to ischemic cardiomyopathy (ICM). All specimens were frozen directly after explantation.
Cell Culture and Transfection-Cell culture and transient cotransfection were done as described (15,16). In brief, Chinese hamster ovary (CHO) cells were stably transfected with the Ca v 1.2a-subunit cloned from rabbit heart (17) or with the Ca v 1.2b-subunit from rabbit lung (18). Human embryonic kidney cells (HEK 293) were stably transfected with cDNA encoding the human cardiac Ca v 1.2-subunit (NM_000719) (19). For experiments with pore subunits alone, cells were seeded in polystyrene Petri dishes (9.6 cm 2 ; Falcon, Heidelberg, Germany) at a density between 10 4 and 2 ϫ 10 4 cells cm Ϫ2 and used within 48 -96 h after plating.
Electrophysiological Recordings-Whole-cell and single channel currents through L-type Ca 2ϩ channels were measured at room temperature (19 -23°C). Whole-cell experiments were performed in external solution containing (mmol/liter) NaCl 120, BaCl 2 10.8, MgCl 2 1, CsCl 5.4, dextrose 10, HEPES 10 (pH 7.4). Pipettes (borosilicate glass, 5-7 megohms) were filled with (mmol/liter) CsCl 120, MgCl 2 3, MgATP 5, EGTA 10, HEPES 5 (pH 7.4). Ba 2ϩ currents were elicited at 0.2 Hz by depolarizing voltage steps from a holding potential of Ϫ80mV to various test potentials as indicated. Currents were sampled at 10 kHz and filtered (Ϫ3 dB) at 2kHz (List EPC-9; HEKA, Lambrecht, Germany). Leak and capacitive currents were subtracted by using a P/N pulse protocol. Peak currents were determined using the average of a 5-ms time window. Time-dependent inactivation was analyzed as I 100 , the current remaining at the end of a 100-ms test pulse, relative to peak current. The software PULSE/PULSE-FIT (version 9.12; HEKA) was used for data acquisition and analysis.
Single channel measurements and analysis were done as reported (5). Cells were superfused with bath solution containing (mmol/liter) potassium glutamate 120, KCl 25, MgCl 2 2, HEPES 10, EGTA 2, CaCl 2 1, Na 2 -ATP 1, dextrose 10 (pH 7.4 with NaOH, 21-23°C). Pipettes (7-10 megohms) were filled with (mmol/liter) BaCl 2 110, HEPES 10 (pH 7.4 with tetraethylammonium hydroxide). Single calcium channels were recorded in the cell-attached configuration (depolarizing test pulses of 150-ms duration at 1.67 Hz, holding potential Ϫ100 mV). An Axopatch 1D amplifier and PClamp 5.5 or 6.0 software (both Axon Instruments, Foster City, CA) were used for pulse generation, data acquisition (10 kHz), and filtering (2 kHz, Ϫ3 dB, 4-pole Bessel filter). Experiments were analyzed whenever the channel activity persisted for at least 72 s (120 sweeps, using 180 sweeps in most cases). Linear leak and capacity currents were subtracted digitally. Openings and closures were identified by the half-height criterion. The availability (fraction of sweeps containing at least one channel opening), the open probability (p open , defined as the relative occupancy of the open state during active sweeps), and the peak ensemble average current (I peak , obtained visually) were analyzed from single channel and multichannel patches. In the latter case, they were corrected for n, the number of channels in the patch. n was defined as the maximum current amplitude observed, divided by the unitary current. Peak current was corrected by division through n. The availability was corrected by the square root method: (1 Ϫ availability corrected ) is the nth root of (1 Ϫ availability uncorrected ). Data Analysis-Effects of the ␤-subunits on single channel parameters were statistically examined by one-way ANOVA over the whole series (i.e. untransfected controls of a given pore subunit and cells cotransfected with ␣ 2 ␦, pGFP, and pcDNA3.1 or a given ␤-subunit subcloned into pcDNA3.1). Significant ANOVAs were followed by posttests against pcDNA3.1 and among ␤Ϫsubunits, applying Bonferroni correction for multiple comparisons. p Ͻ 0.05 was considered significant. All values are given as mean Ϯ S.E.

RESULTS
Specific antisense ␤ 2 -and ␤ 3 -RNA probes hybridized to distinct transcripts in Northern blots with human LV NF and ICM mRNA. The ␤ 3 -specific probe hybridized to two major transcripts in RA, RV, and LV in both human NF and ICM. The sizes of the large transcripts vary with respect to disease state and myocardial localization, 3.1 kb in NF RA, 3.1 kb in NF RV, 3.5 kb in NF LV, 3.5 kb in ICM RA, 2.9 kb in ICM RV, and 3.5 kb in ICM LV whereas the smaller transcript size was 1.1 kb in all heart tissues examined but 0.8 kb in the ICM RV (Fig. 1A). Because no major alternative splicing within the coding sequence was observed with cloning of ␤ 3 by RT-PCR, variable transcript sizes must be because of sequence differences in either the 5Ј-or 3Ј-untranslated regions. Major transcripts detected by the ␤ 2a -specific probe were of 9.5, 3.8, and 2.7 kb in NF/ICM LV/RV mRNA ( Fig. 2A) whereas the ␤ 2b -specific probe hybridized almost exclusively to 9.5-kb transcripts in the tissues examined (Fig. 2A). Because ␤ 2a and ␤ 2b coding sequences differ only with respect to their short N-terminal sequences, isoform-specific antisense probes contained major parts of isoform-specific 5Ј-untranslated sequences. Therefore, we proved the identity of the transcripts detected using a ␤ 2common -probe directed against C-terminal coding sequence common to human ␤ 2a and ␤ 2b . This probe marked transcripts of similar sizes and also another 11.2-kb transcript in NF LV that is barely detectable in ICM LV ( Fig. 2A). The specific probe for the housekeeping gene cardiac calsequestrin hybridized to major transcripts of 2.6 kb in NF and ICM LV mRNA (Fig. 2B) as described before (3,5). Generally, Northern blot intensity of all transcripts detected in RV tissues was less strong compared with LV tissues despite similar total mRNA loading reflecting left ventricular myocyte to matrix dominance.
Abundance of the ␤ 2a -, ␤ 2b -, and ␤ 3 -subunit mRNA expression was determined by real-time RT-PCR in NF LV mRNA (n ϭ 3) and ICM LV mRNA (n ϭ 5; mean echocardiographic ejection fraction of examined ICM hearts, 21 Ϯ 2.23% (S.D.)). Expression data in each specimen were normalized to the respective expression of the housekeeping gene cardiac calsequestrin (3,21), which was also determined by real-time RT-PCR. Real-time RT-PCR for ␤ 2a /␤ 2b was designed with isoform-specific N-terminal sense primers but common C-terminal antisense primer and fluorescent probe. Because we detected no significant differences of copy numbers among ␤ 2a , ␤ 2b , and ␤ 3 when comparing NF LV and ICM LV, data were pooled. In the specimens examined the ␤ 2b copy number was 1.49 ϫ 10 5 Ϯ 6.18 ϫ 10 4 which is 35 times the ␤ 2a copy number of 4.11 ϫ 10 3 Ϯ 1.42 ϫ 10 3 (p ϭ 0.01) (Fig. 2C). Real-time RT-PCR for the ␤ 3 detected 4.83 ϫ 10 4 Ϯ 1.48 ϫ 10 4 copies in the eight specimens examined, which was significantly differ- FIG. 1. A, the radioactively labeled ␤ 3 -subunit antisense probe detects transcripts of marked sizes in mRNA isolated from LV, RV, and RA from NF and failing ICM. 10 g samples of mRNA were applied per lane. Exposure times were 96/108 h for NF/ICM. B, PCR with primers 1 and 2 amplifies the ␤ 3 -subunit region containing the alternatively spliced exon 6. C, 5% PAGE of RT-PCR amplification products with primers 1 and 2 as obtained from NF/ICM LV reverse-transcribed mRNA (each n ϭ 3). Co, control, reaction mixture without template DNA. Fragments obtained were 510 bp (␤ 3a ) and 490 bp (␤ 3trunc ). ent from either ␤ 2a (p ϭ 0.02) or ␤ 2b (p ϭ 0.03) (Fig. 2C). PCR efficiencies of ␤ 2a , ␤ 2b , ␤ 3 , and cardiac cardiac calsequestrin were very close to 100%, and correlation coefficients were Ͼ0.99, respectively (see "Experimental Procedures"), providing a solid basis for comparison. mRNA copy number was calculated on the basis of a 40% efficiency of reverse transcription (22). Gel electrophoresis of the real time PCR products demonstrated one amplification product of expected size per reaction.
The full-length ␤ 3a -sequence cloned from human NF LV mRNA by RT-PCR consists of 1455 bp and encodes 484 amino acids, and this ␤ 3 is identical to the ␤ 3a -subunit isoform cloned from human thyroidoma (GenBank TM accession number X76555). Except for exon 6 deletion (20 nn), which results in a premature stop of translation at nucleotide position 495 no other splicing product of the ␤ 3 -gene was observed in human heart. A similar truncation was also observed in mouse brain tissue (23). Because truncated proteins may play a role in cardiomyopathy, as demonstrated for troponin I (24), we further investigated the (patho-)physiological relevance of ␤ 3trunc by performing RT-PCR experiments using primers 1 ϩ 2 (Fig.  1B) in LV mRNA isolated from NF and ICM (Fig. 1C). In the specimens examined (LV NF/ICM, each n ϭ 7; mean echocardiographic LV ejection fraction in ICM patients, 23.75 Ϯ 2.75% (S.D.)) the ratio of ␤ 3a -/␤ 3trunc -subunit isoform expression changed from 83/17% in NF LV to 51/49% in ICM LV. CHO cells stably expressing the different splice variants of the Ca v 1.2 L-type calcium channel pore were transiently cotransfected with ␣ 2 ␦-subunits together with either rabbit ␤ 2aor the cloned human ␤ 3a -or ␤ 3trunc -subunits (Fig. 3). The existing rabbit ␤ 2a -clone was chosen for our coexpression experiments because of its high homology (96%) to the ␤ 2b , which in the human heart is predominantly expressed when compared with human ␤ 2a . Overall, cotransfection increased current density, shifted the current-voltage relationship toward more negative potentials, and increased the rate of inactivation, e.g. at the respective potential of peak current density (I 100 for non-cotransfected cells, 102 Ϯ 3%; ␤ 2a , 66 Ϯ 9%; ␤ 3a , 74 Ϯ 4%; and ␤ 3trunc , 68 Ϯ 13%; n as in Fig. 3), as reported (12,25). Although the qualitative effects with cotransfection of rabbit ␤ 2a -and human ␤ 3 -subunits were similar, the effects of ␤ 3 -subunits, and of ␤ 3trunc in particular, were less pronounced.   2. A, isoform-specific radioactively labeled antisense probes against N-terminal sequences (␤ 2a , ␤ 2b ) or the ␤ 2 -C-terminal sequence (␤ 2common ) detect transcripts of similar sizes in mRNA isolated from LV and RV from NF and failing ICM. 2 g were applied per lane. Exposure times were 2, 4, and 2 h for ␤ 2a , ␤ 2b , and ␤ 2common , respectively. B, the radioactively labeled antisense probe for cardiac calsequestrin hybridizes to 2.6-kb transcripts in LV from NF/ICM. C, real-time RT-PCR was performed with LV specimen from NF (n ϭ 3) and ICM (n ϭ 5). Data were pooled because of non-significant differences between NF/ICM. Expression levels for ␤ 2a , ␤ 2b , and ␤ 3 were normalized to the expression of cardiac calsequestrin in each sample.
A detailed comparison, and separation of possible effects of cotransfected ␣ 2 ␦-subunits, was carried out using the single channel technique.
Typical traces from single channel experiments using the recombinant Ca v 1.2a pore-forming subunit are depicted in Fig.  4. Single channel activity was sparse in patches from noncotransfected as well from cells cotransfected with ␣ 2 ␦-subunits (plus pcDNA3.1 and pGFP) only. ␤ 3a and ␤ 3trunc increased the probability of finding openings within a test pulse (availability). Open probability within such active sweeps appeared moderately enhanced. In contrast, rabbit ␤ 2a cotransfection led to marked increases in both parameters and to a higher increase in the peak current of the ensemble average. These visual impressions were statistically confirmed ( Table I). Analysis of fast gating parameters in one-channel patches revealed that the mean closed time is the parameter most prominently affected by rabbit ␤ 2a whereas human ␤ 3a or ␤ 3trunc have no effect on closed times. Histogram analysis showed that the reduction in mean closed times is because of a shortening of the time constant of the slow component closed,slow . Open time histograms revealed an increase in the time constant of the open state (Table I) but only in the case of rabbit ␤ 2a .
To examine whether the more marked modulation by rabbit ␤ 2a -subunits is restricted to the cardiac type Ca v 1.2a pore subunit, we examined the effect of ␤ 2a and ␤ 3a on the smooth muscle splice variant, Ca v 1.2b. As indicated in Fig. 5 and Table  II, very similar results were found as with Ca v 1.2a. Again, both ␤-subunits increased availability similarly. Ensemble average currents were more markedly affected by ␤ 2a because of a significant additional effect on rapid gating and open probability. Similar to Ca v 1.2a, closed,slow was reduced, which may explain this effect. Together, these findings confirm a role of the ␤-subunit structure on pore properties. As with Ca v 1.2a, non-transfected cells and cells cotransfected only with pcDNA3.1, ␣ 2 ␦-subunits, and GFP displayed a quite sparse pattern of activity of Ca v 1.2b (Fig. 5), similar to our previous observations (26).
To account for the functional differences found among the multiple splice variants of human cardiac-type Ca v 1.2 (27,28), we next examined the Ca v 1.2 pore-forming subunit usually expressed in human myocardium (19). We expressed cDNAs in HEK 293 cells for these studies, which enabled us to perform a more extended analysis, because of better seal stability. The inherent properties of pore subunits, as well as their differential modulation by rabbit ␤ 2a and human ␤ 3 isoforms (Table  III), were comparable with those of the CHO system. Another common feature with the CHO cell experiments was recognized: ␤ 2a but not the ␤ 3 -subunits increased open probability and open time duration. Closed time duration appeared to be shortened by ␤ 2a , mainly because of a change of its slow component. As expected, the first latency was shortened by ␤ 2a coexpression. ␤ 3trunc had no effects on fast gating, and ␤ 3a took a numerically intermediate position regarding these parameters. The effect of ␤ 3trunc on availability and peak current fell visibly short of the effect of ␤ 3a in this expression system. Finally, we examined a relevant window of test potentials (0 to ϩ20 mV) where voltage-dependent activation of gating parameters rises steeply. This reveals the biophysical nature of the differential modulation of the pore subunit by ␤ 2a versus ␤ 3a and ␤ 3trunc (Fig. 6). With ␤ 2a , the voltage dependence of open probability was shifted to the left. This was not as prominent with ␤ 3a and even less with ␤ 3trunc . Unfortunately, more positive potentials could not be examined because of bandwidth limitation (29). However, it becomes clear that the amount of voltage shift is larger with ␤ 2a . Considering availability, modulation by ␤ 2a and ␤ 3a again were quite similar (not shown).

DISCUSSION
In the present study, we describe expression of two ␤ 2 -subunit isoforms (␤ 2a , ␤ 2b ) and of two ␤ 3 -subunits (␤ 3a , ␤ 3trunc ) in the human heart. The ␤ 2a isoform-specific probe detected distinct transcripts of 9.5, 3.8 and 2.7 kb in left ventricular mRNA from human non-failing and failing ischemic cardiomyopathy whereas the ␤ 2b -specific probe hybridized to a major transcript of 9.5 kb in both specimens. Because of the short length of the isoform-specific N-terminal coding sequences, each antisense probe contained untranslated sequences to major parts that might anneal non-specifically to other non-coding sequences. To prove the specificity of the hybridization signals we therefore used a ␤ 2common antisense RNA probe complementary to the coding C-terminal sequence common to both ␤ 2a and ␤ 2b . This probe detected transcripts of similar sizes, thus, unspecific hybridization of the isoform-specific probes is unlikely. Our probes detected transcript sizes different from the sizes observed by Freise et al. (8) who demonstrated in human heart mRNA a major band of 3.5 kb and a faint band of 6.9 kb when using a radioactively labeled murine ␤ 2 145-bp cDNA probe. Differences in the transcript sizes detected may result either from experimental discrepancies or inherent properties of the various probes. However, in our Northern blots the detection of transcripts of similar sizes by two different antisense RNA probes that recognize diverse sequences of the same mRNA molecule was re-assuring. Our probes marked relatively large transcripts, although full-length coding sequences of the ␤ 2a and ␤ 2b are 1818 and 1821 bp, respectively. We can rule out genomic contamination of our mRNA preparations, because ␤ 2a and ␤ 2b real-time RT-PCR from the same mRNA preparations resulted in single amplification products despite amplification across boundaries of N-terminal and common C-terminal coding exons. Most likely, these long transcripts represent progenitor forms of the finally translated mRNA molecules reflecting the large dispersion of the ␤ 2 gene over chromosome 10. In contrast to earlier studies that did not detect ␤ 3 -subunit mRNA expression in human LV mRNA (30), we detected transcripts of 3.5 kb and smaller sizes in RA, RV, and LV from both human non-failing and failing ischemic myocardium. Transcripts of 3.5 kb were also reported for other human tissues (30). Both the expression of ␤ 3 transcripts in the whole heart and the finding of differential splicing of exon 6 suggest a physiological role of the ␤ 3 in the human heart, which is also supported by our real-time PCR results that revealed a higher ␤ 3 copy number when compared with ␤ 2a .
Real-time experiments revealed unreckoned prevailing ␤ 2b expression when compared with ␤ 2a , which suggests an important physiological relevance of the ␤ 2b , although it lacks the palmitoylation sequence contained in the ␤ 2a . Compared with both ␤ 2b and ␤ 3 , ␤ 2a expression level is much lower, which was surprising in view of its attributed important role for modulation of the L-type calcium inward current in the cardiomyocyte (13). However, gene expression levels of the same magnitude were also reported for transcripts of the ␤ 1b -subunit isoform as detected by the competitive RT-PCR technique (3). At the moment, the physiological relevance of the different magnitudes in the expression levels of ␤ 1b , ␤ 2a , ␤ 2b , and ␤ 3 remains unclear but certainly warrants further investigation.
As with previous gene expression studies in human myocardium (3,5), expression levels of ␤ 2a , ␤ 2b , and ␤ 3 were all normalized to cardiac calsequestrin expression in the same sample. This sarcoplasmic calcium-binding protein was chosen as "housekeeping gene," because different studies in human heart failure showed that its expression is consistently unchanged at the mRNA and the protein level when compared with non-failing myocardium (21). Furthermore, cardiac calsequestrin is almost exclusively expressed in the cardiomyocyte (31); thus, in human heart expression of the gene of interest can be related to the cardiomyocyte content in the specimen examined.
Cellular localization of the ␤ 2b -and ␤ 3 -subunit isoforms cannot be delineated from our gene expression studies, and appropriate experiments might also be crippled by their low abundance of expression. For this reason, we addressed this question at the functional level by expressing the ␤ 3a -and ␤ 3trunc -subunit in cells that stably express the Ca V 1.2a, Ca V 1.2b, and human Ca V 1.2. The calcium currents obtained by coexpression with the ␤ 3 isoforms were compared with currents induced by rabbit ␤ 2a because of its 96% homology to the human ␤ 2b . Indeed, the ␤ 3a , and to a lesser extent the ␤ 3trunc splice variant, elicited functional impact when coexpressed with the cardiac pore-forming subunit. Interestingly, the extent of modulation by ␤ 3a was substantially less than with ␤ 2a . This difference is not because of a species mismatch between the rabbit and human clones, because similar findings were obtained with both rabbit and human cardiac pore-forming subunits. We also tend to exclude confounding effects of a different dose of the subunit isoforms; all technical precautions (same expression vector, same amount of plasmid DNA) were appropriately implemented, and, more convincingly, the distinctive features of single channel modulation by ␤ 2a and ␤ 3a help to rule out this concern. Although ␤ 2a and ␤ 3a equivalently increased channel availability, only ␤ 2a showed a large and statistically robust stimulation of the fast gating properties. A more pronounced influence of the same kind of modulation, as with an increased gene dose, should have affected both parameters in parallel. One may argue that our single channel analysis could be biased by presence of a multitude of channels in the patch, even in cases where no multiple simultaneous openings were detected. However, this would even sharpen the distinction between ␤ 2a and ␤ 3a , because we can exclude multichannel patches with greater certainty in the case of ␤ 2a . The probability that the absence of stacked openings indicates the presence of truly one channel is Ͼ98% in a typical experiment with ␤ 2a but only ϳ92% in case of ␤ 3a (32).
The electrophysiological distinction between ␤-subunits is not without precedence. When rat ␤-subunit-GFP fusion pro-teins were overexpressed in native myocytes, ␤ 2 -subunits caused a more prominent increase of whole-cell currents than ␤ 3 -subunits (33). An abstract (34) and a recent paper (13) showed that ␤-subunit isoforms can be discriminated regarding closed time distributions at the single channel level, a finding in line with our data.
The observed functional consequence of differential splicing of ␤ 3 -subunits cannot explain the increase in channel activity as observed in human heart failure (5). ␤ 3trunc -subunits, which appear to modulate the pore in a qualitatively similar manner as ␤ 3a , have a less pronounced influence in almost every respect. Although channel activity is typically doubled by ␤ 3trunc , this effect does not reach significance when corrected for the multiple comparisons we were obliged to perform. Viewed in isolation, even the ␤ 3trunc isoform is able to modulate the cardiac pore subunits, albeit in a more subtle manner. This is not unexpected, because N-terminal interaction sequences remain despite the premature translational stop, which results in loss of the principal ␤-subunit interaction domain and, furthermore, of potential phosphorylation sites for protein kinase C and the casein kinase II (30). Also, such structural features may alter interaction with G-proteins (35,36). ␤ 3trunc expression may actually compete with other isoforms and thus reduce single channel activity in native heteromultimeric complexes, but channels behaving in such a manner were not detected in native failing human myocytes (5,29,37).
In summary, ␤-subunit gene products modulate cardiac calcium channel pores in an isoform-and splice variant-specific manner. Therefore, it is crucial to quantify not only the abso- lute amount of mRNA or protein of these subunits under pathological conditions but rather the relative contribution of the subtypes actually expressed. Future studies should dissect such ␤-subunit alterations. These could serve as a molecular basis of altered single channel behavior in heart failure, which in turn may critically compromise excitation-contraction coupling.