|Home | About | Journals | Submit | Contact Us | Français|
The frequency sensitivity of auditory hair cells in the inner ear varies with their longitudinal position in the sensory epithelium. Among the factors that determine the differential cellular response to sound is the resonance of a hair cell's transmembrane electrical potential, whose frequency correlates with the kinetic properties of the high-conductance Ca2+-activated K+ (BK) channels encoded by a Slo (kcnma1) gene. It has been proposed that the inclusion of specific alternative axons in the Slo transcripts along the cochlea underlies the gradient of BK-channel kinetics. By analyzing the complete sequences of chicken Slo gene (cSlo) cDNAs from the chicken's cochlea, we show that most transcripts lack alternative exons. Transcripts with more than one alternative exon constitute only 10% of the total. Although the fraction of transcripts containing alternative exons increases from the cochlear base to the apex, the combination of alternative exons is not regulated. There is also a clear increase in the expression of BK transcripts with long carboxyl termini toward the apex. When long and short BK transcripts are expressed in HEK-293 cells, the kinetics of single-channel currents differ only slightly, but they are substantially slowed when the channels are coexpressed with the auxiliary β subunit that occurs more widely at the apex. These results argue that the tonotopic gradient is not established by the selective inclusion of highly specific cSlo exons. Instead, a gradient in the expression of β subunits slows BK channels toward the low-frequency apex of the cochlea.
The auditory system maps continuous sensory variables, such as the frequency of a sound or the spatial location of its source, onto cellular detector arrays whose individual elements are narrowly tuned. How sensory maps of any kind are established and maintained is an open question. In only a single case, the tonotopic frequency map in the inner ear, is the physiological mechanism responsible for the individual cell's differential tuning understood at the molecular level.
In the process of electrical resonance, a combination of voltage- and ion-dependent conductances operates in conjunction with the passive electrical properties of a hair cell's membrane to accentuate cellular responsiveness to a particular range of frequencies while attenuating that to frequencies outside this range. A dynamic interaction between two types of ion channels mediates this band-pass behavior. Voltage-sensitive Ca2+ channels activated by the current through mechanotransduction channels depolarize the hair cell's membrane and increase the intracellular concentration of Ca2+. With some delay, these ions activate the Ca2+-sensitive high-conductance Ca2+-activated K+ (BK) channels that repolarize the membrane, thereby closing the Ca2+ channels, decreasing the intracellular concentration of Ca2+, and deactivating the K+ channels. How rapidly this dynamic system progresses through the cycle depends on several factors, most notably on the kinetic properties of the BK channels (25).
The BK channel is a high-conductance K+ channel formed by tetramers of the α subunit. Membrane depolarization and intracellular Ca2+ activate the channel, and in many tissues auxiliary β subunits modulate its kinetics (26, 31, 38). The BK α subunit is encoded in vertebrates by the kcnma1 gene, the homologue of the slowpoke (Slo) gene in Drosophila.
The chicken Slo gene (cSlo) transcripts expressed in the cochlea contain several alternative exons that could potentially form more than 500 different combinations of alternative exons spliced into the reference sequence of the gene (1, 14, 30, 37). Some of these alternative exons have been cloned and inserted artificially into the reference Slo sequence, giving rise to channels that differ in physiological properties such as the single-channel conductance, kinetics, and Ca2+ sensitivity (8, 17, 36). The presence of different alternative exons of the α subunit, together with the graded expression of the β subunit of the BK channel, has been proposed to be the principal mechanism by which the cochlear tonotopic gradient is generated (34-36). In the chicken, this gradient extends from approximately 100 Hz at the apex to 5 kHz at the base with a continuous gradient between.
We hypothesized that the inclusion of alternative exons along the cochlea must be tightly regulated to generate the expression of channels with electrophysiological properties compatible with the characteristic frequency of a given region in the tonotopic gradient. Although the presence of alternative Slo exons has been documented in the cochlea of the chicken (30, 37), turtle (17), rat (4), and mouse (3), an analysis of the combination and usage of exons in cSlo transcripts is lacking. In the present study we quantified the expression of cSlo alternative exons and analyzed the expression of individual cSlo transcripts along the cochlea. We also extended the analysis to the expression of BK channels and identified splicing factors expressed in the cochlea.
We used chickens (Gallus gallus) of the White Leghorn strain at 2 to 3 weeks of age. Animals were housed and euthanized in accordance with the guidelines and regulations approved by The Rockefeller University Animal Care and Use Committee. After cochleae had been dissected by a medial approach, the auditory nerve and tegmentum vasculosum were removed, and the cochlea was cut into segments along its length. The tissue was moistened during the procedure with artificial perilymph (145 mM NaCl, 2 mM KCl, 1.5 mM CaCl2, 1 mM MgCl2, 0.1 mM Na2HPO4, 0.1 mM d-glucose, and 3 mM HEPES at pH 7.26). Cerebella were dissected from some animals to provide control tissue.
For the design of optimal primers to amplify the full cSlo transcripts, we performed rapid amplification of cDNA ends (RACE) to confirm downstream and upstream sequences. RNA was isolated from flash-frozen basal or apical halves of the cochlea and extracted with an RNeasy minikit (Qiagen, Valencia, CA). A sample of 2 μg of total RNA was used for 5′ and 3′ RACE performed with a GeneRacer kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The primer sequences are provided in the supplemental material. The 700-bp product of the 3′ RACE and the 600-bp product of the 5′ RACE were purified with the provided columns, cloned into the TOPO TA vector for sequencing (Invitrogen), and transformed into TOPten chemically competent bacteria. Colonies were cultured for minipreps (QIAprep spin miniprep kit; Qiagen, Valencia, CA), and the cDNA inserts were sequenced and assembled by using VectorNTI (Invitrogen).
In initial experiments, cochlear segments were flash frozen in liquid nitrogen after dissection in artificial perilymph. After discovering that placing each cochlea in a drop of RNAlater (Qiagen) immediately after removal from the head gave a better RNA yield, we adopted that procedure. The dissection was then finished in RNAlater and tissue was stored at −80°C. Segments from 100 to 120 cochleae were pooled, and total RNA was extracted by using an RNeasy minikit (Qiagen) or, in later experiments, the Nucleo Spin RNA II kit (BD Biosciences, Mountain View, CA).
Genomic DNA was digested with DNase I. Starting with 200 to 500 ng of total RNA, we synthesized cDNA after downstream priming with the Superscript III first-strand synthesis system for reverse transcription-PCR (RT-PCR; Invitrogen). This step was followed by PCR amplification of half of the cDNA using Advantage cDNA mix (BD Biosciences) in a total volume of 100 μl. Primers for short or long cSlo included a 5′G to ensure equal TOPO cloning efficiency (5). The primer sequences are provided in the supplemental material.
The PCR products were mixed with loading buffer (30% glycerol and 100 mg of crystal violet/ml) and loaded into a gel containing 0.8% agarose and 1.5 mg of crystal violet/ml in TAE (50 mM Tris-acetate and 1 mM EDTA at pH 8). The bands were excised, and the cDNA was isolated with a QIAquick gel extraction kit (Qiagen) eluted in 30 ml of TE buffer (10 mM Tris-HCl and 1 mM EDTA at pH 7.5) and quantified by determing the absorbance at 260 nm. Equal amounts of clean cDNA were ligated into the TOPO XL vector (Invitrogen), and 1 to 2 μl of the product was electroporated into electrocompetent E. cloni cells (Lucigen, Middleton, WI) grown according to the manufacturer's instructions.
All new chicken KCNMA1 transcript sequences were scanned against the reference sequence (NM_204224) and submitted to GenBank. The accession numbers are GU223604 through GU223623.
The cSlo-containing TOPO XL plasmids behave as low-copy-number plasmids; it was therefore necessary to use midi or maxi preps to obtain sufficient material for complete sequencing of the interest. Because of the low percentage of alternative exon inclusion, we initially screened clones by PCR in a 96-well format using 0.65 μM concentrations of each primer for sites 1 through 6 (37) and PCR Super Mix High Fidelity (Invitrogen). The PCR products were cleaned with ExoSAP-IT (USB Corp., Cleveland, OH) and sequenced with the same primers.
To screen a large number of clones efficiently, we devised the following method. TOPO XL-containing clones from four different groups of plates, from short or long cSlo from the base or apex, were individually picked and each group was grown in three 96-well plates. Plates covered with Breathe-Easy gas-permeable membranes (BST Scientific Pte, Ltd., Singapore) were maintained for 14 h at 300 rpm and at 37°C. The live bacterial cultures from each well were transferred with a manual 96-pin replicator (VP scientific, San Diego, CA) to a precut nylon membrane (Nunc, Rochester, NY) forming 96 three-by-three arrays containing three replicates for each original well. There were a total of 864 dots from 288 different clones per membrane. For short cSlo clones, eight copies of each membrane were made, and for long cSlo, 10 copies were made. Each membrane was hybridized with a different probe; the probe sequences are provided in the supplemental material. A probe complementary to a constitutive rigion of cSlo was used as a positive control to eliminate clones not containing a cSlo insert. After blotting, the plates were sealed, frozen in dry ice, and stored at −80°C. For subsequent retrieval of a particular clone, the frozen surface of the well was scraped, and a fresh agar plate was streaked and grown overnight. Blotted membranes were denatured for 5 min with Whatman number 3 paper soaked in 1.5 M NaCl and 0.5 M NaOH, neutralized for 5 min with Whatman number 3 paper soaked in 1.5 M NaCl and 0.5 M Tris-HCl at pH 7.4, washed with agitation for 5 min in 2× SSC (300 mM NaCl and 30 mM sodium citrate), and fixed at 70°C for 2 h.
Portions (10 pmol) of each oligonucleotide probe were labeled with [γ-32P]ATP using Ready-to-Go T4 polynucleotide kinase (Amersham, Piscataway, NJ) according to the manufacturer's instructions. Each set of membranes to be hybridized with the same oligonucleotide was separated by a nylon mesh, placed in a hybridization bottle, and covered with ~15 ml of Rapid-Hyb solution (Amersham) for at least 30 min of prehybridization at 42°C. Labeled oligonucleotide was added to the solution and swirled immediately to mix before a 2-h hybridization at 42°C. The radioactive solution was carefully removed, and the membranes were rinsed once and washed once for 30 min with 5× SSC and 0.1% sodium dodecyl sulfate (SDS), twice for 20 min with 1× SSC and 0.1% SDS, and twice for 10 min with 2× SSC at 42°C. All clones with alternative exons in our first three screening experiments were confirmed by sequencing, and all clones containing exon 23 were sequenced to test the presence of exon 22, which we were unable to detect with a probe. Several clones in which the hybridization signal was unclear were also sequenced.
Using random hexamers (Invitrogen) and Superscript III Reverse Transcriptase with RiboAmp (Arcturus/Molecular Devices, Sunnyvale, CA), we reverse transcribed 100 ng of total RNA or 0.5 to 1.4 mg of total amplified RNA from the base and the apex or from five cochlear segments. The cDNA was amplified using one-tenth of the total amount per 25 μl of iQ-SYBR green Supermix (Bio-Rad, Hercules, CA) in a 7900HT sequence detection system (Applied Biosystems, Foster City, CA). Primer pairs to detect exons 4, 11, 19, 23, 29, 33 (short cSlo), or 34 + 35 (long cSlo) were designed with Primer Express (Applied Biosystems) and tested for efficiency greater than 95% using cerebellar cDNA. For the detection of premature stop codons, we analyzed total RNA from the base and the apex as described above, except that up to 500 ng of total RNA was downstream-primed with short or long cSlo reverse primers and Platinum SYBR green qPCR SuperMix-UDG with ROX (Invitrogen) was used. Primers for the β-actin gene were used to standardize in all experiments; the sequences are provided in the supplemental material.
Cochleae were collected in artificial perilymph, flash-frozen in liquid nitrogen, and stored at −80°C. A total of 120 to 240 basal or apical cochlear segments was thawed over ice in 500 μl of a homogenization buffer solution containing 350 mM sucrose, 5 mM EDTA, 1× Halt protease inhibitor (Pierce, Rockford, IL), and 10 mM HEPES at pH 7.4 and then lysed by 10 strokes of a motor-driven Teflon-glass homogenizer at 350 rpm. The homogenized tissue was centrifuged at 2000 × g for 4 min at 4°C and the postnuclear supernatant was centrifuged for 1 h at 4°C at 100,000 × g in polycarbonate tubes in a TLA 100.3 rotor (Beckman Instruments, Palo Alto, CA). The pellets containing membrane-enriched cochlear proteins were dissolved in the homogenization buffer and sonicated for 4 to 8 s. Protein concentration was determined by BCA method (Pierce, Rockford, IL) to avoid interference by lipids and by detergents. Cell membrane samples (20 to 120 μg/lane) were electrophoresed in a 4 to 12% gradient NuPage NOVEX Bis-Tris gel (Invitrogen) and immunoblotted with a monoclonal anti-BK antibody (BD Biosciences, Ciudad, CA), a polyclonal anti-BK antiserum (Alomone, Jerusalem, Israel), a monoclonal α-tubulin antibody (Sigma, St. Louis, MO), or a custom-made rabbit polyclonal antiserum from Covance (Princeton, NJ). To detect short BK, we raised an antiserum against the peptide KYVQEDRL; to detect long BK, we used the peptide QEKKWFTDEPDNA; and to detect the β1 subunit, we raised an antiserum against the peptide EEIANNFKKYQT. The Rabbit IgG TrueBlot system (eBioscience, San Diego, CA) was used according to the manufacturer's instructions for secondary detection of the custom antibodies.
A chicken-genome array from Affymetrix (Santa Clara, CA) was used to compare the expression of genes between the basal and apical cochlear regions. Using ArrayScript reverse transcriptase and an oligo(dT) primer bearing a T7 promoter, we utilized 1 μg of total RNA isolated from 80 to 120 cochlear segments to synthesize the first strands of cDNA. Single-stranded cDNA was then converted into double-stranded DNA (dsDNA) by DNA polymerase I in the presence of Escherichia coli RNase H and DNA ligase. After column purification, dsDNA served as a template for in vitro transcription in a reaction containing biotin-labeled UTP, unlabeled NTPs, and T7 RNA polymerase. The amplified, biotin-labeled antisense RNA (aRNA) was purified, and its quality was assessed by using the Agilent 2100 Bioanalyzer and the RNA 6000 Nano kit.
A portion (20 μg) of labeled aRNA was fragmented, and 15 μg of the fragmented aRNA was hybridized to Affymetrix Chicken Genome Gene Chips for 16 h at 45°C as described in the Affymetrix technical analysis manual (Affymetrix, Santa Clara, CA). After hybridization, gene chips were labeled with streptavidin-phycoerythrin, followed by an antistreptavidin solution and a second streptavidin-phycoerythrin solution, with all liquid handling performed by a GeneChip Fluidics Station 450. Gene chips were then scanned with the Affymetrix GeneChip Scanner 3000. The raw intensity data of gene chips were normalized and further analyzed by using GeneSpring 7.2 (Agilent Technologies, Inc., Palo Alto, CA). The chicken-array annotations were completed by the use of the Affymetrix NetAffx analysis center to search human homologs and the links in the iHOP (http://www.ihop-net.org/) (10) and UniProt databases (http://www.uniprot.org/). Data from human homologues were used when available.
The cSlo clones selected for expression were first sequenced completely to confirm that the open reading frames were complete and then subcloned from the original TOPO-XL plasmid into pXOOM plasmids (13). Gels stained with crystal violet were used to reduce mutation rates. The correct orientation and sequence of subcloned variants was confirmed by restriction enzyme digestion and by sequencing the ends of the inserts or, in some cases, the complete cDNAs.
The β1 gene was amplified from total cerebellar cDNA with primers containing sites recognized by the restriction enzymes BamHI/EcoRI and directly cloned into pXOOM. The insert was sequenced and confirmed to match that cloned from the cochlea (accession number AF420468). For expression in HEK-293T cells, 150,000 cells per well were plated in 24-well plates and transfected on the next day with 1 μl per well of Lipofectamine 2000 (Invitrogen) at a concentration of 5 to 500 ng of plasmid per well. One day after the transfection, cells were treated with trypsin, diluted 5 to 20 times, and plated on polylysine-coated glass coverslips (BD Biosciences) for electrophysiological recordings. The expression of EGFP from a second promoter in the pXOOM plasmid was used as a marker of cSlo expression.
Stably transfected cell lines were also generated by clonal dilution and selection with Geneticin (Invitrogen). When a stable cell line expressing the β1 subunit was transfected with α subunit of cSlo, a dsRed-expressing plasmid was cotransfected to track the cells that were expressing these additional plasmids. The cell line expressing the short cSlo variant without alternative exons lost expression of the plasmid after several passages; there appeared to be selective pressure against cells expressing the gene.
HEK-293 cells were visualized on an upright microscope (Olympus, Center Valley, PA) fitted with a ×40 water-immersion objective lens of numerical aperture 0.8 and equipped with differential-interference-contrast optics and a fluorescence illuminator. Patch pipettes were fabricated from borosilicate glass capillaries (Sutter Instruments, Novato, CA) using a horizontal puller (P-80/PC; Sutter Instruments) and coated with either beeswax or nail polish to decrease their capacitance. Capacitative transients were electronically compensated at the beginning of each recording session, and the compensation was adjusted regularly. The series resistance was usually between 7 and 15 MΩ.
Single-channel currents were recorded using the cell-attached and the inside-out configurations with an Axopatch 200B amplifier (Molecular Devices, Union City, CA) whose headstage was mounted on a motorized manipulator (LN Junior; Luigs & Neumann, Ratingen, Germany). Current signals were filtered at 5 or 10 kHz with a four-pole analog Bessel filter and sampled at 10-μs intervals. A custom program written in LabVIEW (National Instruments, Austin, TX) was used for stimulus generation and data acquisition. Recordings were performed in symmetrical solutions containing 142 mM K+, 1.1 mM Ca2+, 0.5 mM Mg2+, 145 mM Cl−, 2 mM EGTA, and 5 mM HEPES at pH 7.3 The calculated free Ca2+ concentration was 70 nM. Experiments were performed at a room temperature of 20 to 25°C.
Data were digitally low-pass filtered with a Gaussian filter with a half-power frequency of 3 kHz. Transitions between closed and open states were detected by using the single-channel search routine of pClamp8 (Axon Instruments, Foster City, CA). All statistics were calculated in Matlab (MathWorks, Natick, MA). The statistics of the mean open time were estimated by using the maximum-likelihood method with an exponential distribution as the prior and also, for comparison, using a nonparametric bootstrap resampling method. The probability-density functions were calculated with a Markov chain Monte Carlo method using the Metropolis-Hastings search algorithm with a double-exponential distribution as the proposal (or target) density function. The probabilities in each marginalized distribution summed to unity. The two-sample Kolmogorov-Smirnov test was used to assess the probability that different data sets were samples of the same continuous distribution.
The results are expressed as means ± the standard errors except where otherwise indicated. An unpaired Student t test was used for the statistical comparison of means. A difference was considered statistically significant at a P value of <0.05.
The Slo-encoded BK channel has seven potential sites at which alternative splicing can generate sequence diversity (Fig. (Fig.1A).1A). By comparing cSlo sequences cloned from the cochlea and cerebellum to the reference genomic sequences in both the NCBI chicken genome version 2.1 (accession number NC_006093) and the UCSC genomic browser, we were able to relate our results to previously published data (3, 30, 37). In an effort to standardize the nomenclature, we used the numbering system for Slo alternative-splice sites and exons proposed while this work was in progress (3).
Performing RACE, we found only one amino terminus of the cSlo gene. A reported alternative amino terminus (37) corresponds to the carboxyl terminus of the paralemmin gene (21) and may therefore be a cloning artifact. In accordance with previous findings, we observed two alternative carboxyl termini (site 7) coding for KYVQEDRL (short cSlo) and REVEDEC (long cSlo; Fig. Fig.1B).1B). A third, previously reported terminus encoding KEMVYR (30) was found to end, like REVEDEC cSlo, in exon 35, but to lack exon 34 and therefore to have an earlier stop codon. We shall refer to all transcripts containing alternative exon 35 as long cSlo variants owing to the lengths of the transcripts rather than of the proteins.
We also identified two of the three alternative usages of splice acceptor sites in exon 35 that were previously found by analysis of different databases (3). Both long cSlo transcripts were found to share a 3′ untranslated region (3′UTR) distinct from that of the short cSlo 3′UTR. No insertion of alternative exons was observed in the chicken's cochlea or cerebellum at site 5, in accordance with the appearance of splicing at that site in the mammalian line that includes primates, canids, and bovids (3). Usage of the complete exon 19 was extremely rare, occurring in only 2 of 451 clones from the cerebellum. Nevertheless, the observed differences in length (37) and in sequence (3) suggest the possibility of additional acceptor or donor sites for exon 19.
To amplify the complete sequence of all possible cSlo transcripts, we designed two sets of primers: one for short cSlo and the other for long cSlo. We were surprised to discover that short cSlo transcripts exhibit a constant expression throughout the cochlea, whereas long cSlo transcripts are expressed less at the base (Fig. (Fig.2A).2A). These results disagree with the electrical-resonance model (12), which predicts more channels at the high-frequency basal region. In order to quantify the difference in transcript concentration between regions, we performed a quantitative PCR (qPCR) analysis on RNA isolated from five consecutive cochlear segments (Fig. (Fig.2E).2E). This study showed clearly that, whereas the level of short cSlo transcripts remains constant (Fig. (Fig.2B),2B), long cSlo transcripts are expressed in a continuous and increasing gradient from the base to the apex (Fig. (Fig.2C).2C). The expression patterns of alternative exons 11, 19, and 23 follow the same pattern as that for long cSlo (Fig. (Fig.2D).2D). Exons 4 and 29 were detected only in the cerebellum (not shown).
To examine the usage of exon combinations along the cochlea, we cloned PCR products from short and long cSlo transcripts and sequenced the entire inserts. Because the majority of transcripts lacked alternative exons, we screened clones directly blotted onto nylon membranes with radioactive oligonucleotide probes specific for different alternative exons. In three independent experiments on 3,042 clones, 1701 corresponded to short cSlo variants and 1,341 corresponded to long cSlo products. The average of the three experiments showed that most transcripts, including 91% ± 1% of short cSlo and 67% ± 6% of long cSlo transcripts, have no alternative exons (Fig. 3A and B). A small number of transcripts possess two or three alternative exons. Only four long cSlo clones, or 0.13% ± 0.03% of the total, were found to contain all four of the exons 11, 19, 22, and 23. These clones also contained the constitutive exon 10, which is mutually exclusive of exon 11 and generates an early stop codon. The only difference along the tonotopic gradient was the slight but significant (P = 0.02) decrease in the expression at the apex of short cSlo with no alternative exons (Fig. (Fig.3A3A).
When clones from the base and apex with at least two alternative exons are considered together, 21% ± 6% correspond to long cSlo transcripts and 1.2% ± 0.7% to short cSlo transcripts (P < 0.05). Because equal starting amounts of PCR product were used for these experiments, these numbers must be corrected for the relative expression at the base and apex as assessed by qPCR. Short cSlo transcripts were found to be expressed evenly along the cochlea (Fig. (Fig.2B2B and and8C).8C). In contrast, we found 11-fold as many long cSlo transcripts at the apex as at the base (Fig. (Fig.2C2C and and8E).8E). When the number of short and long cSlo clones is corrected for regional expression, there is an apparent enrichment of transcripts with multiple alternative exons at the apex (Fig. 3C and D). This difference is not statistically significant, however, owing to the variability between experiments in the number of clones with specific numbers of alternative exons.
Qualitative analysis of cSlo clones showed that all alternative exons expressed in the cochlea occur both at the base and at the apex (Fig. (Fig.4A4A and and4B).4B). The relative expression of each alternative exon was not significantly different along the cochlea except for the preferential inclusion of exon 23 at the base (Fig. 4A and B). The exon preference between short (Fig. (Fig.4A)4A) and long cSlo transcripts (Fig. (Fig.4B)4B) differed only in the reduced inclusion of exon 23 (in the absence of exon 22) in the long transcripts. When these data are corrected for the differential expression of short (Fig. (Fig.4C)4C) and long (Fig. (Fig.4D)4D) cSlo variants along the cochlea, the results demonstrate a preferential inclusion of all exon 11-containing cSlo transcripts at the apex. As expected, there is also a general increase in the expression of all exons and in long cSlo compared to short cSlo transcripts at the apex. The exon preference is similar for the two long cSlo coding variants REVEDEC (Fig. (Fig.5A)5A) and KEMVYR (Fig. (Fig.5B).5B). Less than 10% of the total clones are KEMVYR variants, and these occur preferentially at the base (Fig. (Fig.5C5C).
The exon preference in the cochlea differs from that in the cerebellum; in particular, the relative expression of exons 11 and 22 is inverted between the two organs. The cerebellum also expresses exon 4, the complete exon 19 (19L), and exon 29, all of which are absent from the cochlea. Finally, there is only a small difference in the exon preference of short and long cSlo transcripts in the cerebellum.
From the analysis of all clones, we could test the hypothesis that particular exon combinations are used in different regions of the cochlea to achieve a functional tonotopic gradient. We compared the representation of each possible combination in the short cSlo variant (Fig. (Fig.6A).6A). More than 95% of transcripts had no alternative exons, especially at the base. The remainder expressed each alternative exon with equal probability at the base and apex. An exception was exon 11, which alone or in combination with exon 19 showed higher expression at the apex. The three combinations with more than one alternative exon were all expressed at the apex in accordance with the higher expression of alternative exons in this region (Fig. (Fig.2D).2D). There are several possible combinations that were not encountered at all, possibly because of their low abundance. In REVEDEC-encoding cSlo transcripts (Fig. (Fig.6B),6B), the variants with alternative exons constitute about 50% of the total. Whether expressed alone or in combination with exon 22/23, REVEDEC variants also show a preferential expression of exon 11 at the base. Otherwise, there was no preference for particular combinations of exons along the cochlea. With a few exceptions, all possible combinations were found at both the apex and the base. No long cSlo transcripts were determined to express both exons 19 and 23 but to lack exon 22 (Fig. (Fig.4B).4B). The rare KEMVYR variants (Fig. (Fig.6C)6C) expressed only one of the alternative exons in any given transcript, except for the combination of exons 11 and 19 that was found at the apex.
While analyzing complete sequences of the cSlo transcripts to be subcloned into expression vectors, we found many transcripts that included premature stop codons. In some cases the introduction or deletion of a single base probably reflected a PCR artifact; however, we also found intron retention in several clones. To quantify splicing fidelity we designed primers to detect three common events: (i) retention of exon 10 when the mutually exclusive exon 11 was spliced into the transcript, (ii) retention of a particular segment of an intron sequence after exon 19, and (iii) retention of a specific part of an intron sequence after exon 23 (Fig. (Fig.7).7). The qPCR analysis of cochlear RNA showed an increased retention of exon 10 at the apex for short cSlo (Fig. (Fig.7A)7A) and a decrease in intron retention after exon 19 at the apex for both short and long cSlo variants (Fig. (Fig.7B).7B). There is no difference between the base and apex in the fidelity during exon 23 splicing (Fig. (Fig.7C).7C). Although these results reveal premature stop codons in the population of cSlo transcripts owing to intron retention, they do not signal any difference between the regions of the cochlea. Unproductive splicing (24) of cSlo therefore does not regulate the gene's expression in a way that contributes to the establishment of the tonotopic gradient.
To establish whether the gradients in transcript expression are reflected in gradients of protein expression, we raised polyclonal antisera against peptides specific to the KYVQEDRL (short cSlo) or REVEDEC (long cSlo) carboxyl terminus. In accordance with the pattern of transcript expression (Fig. (Fig.8A),8A), the expression of BK channels detected with a commercial monoclonal antibody was enhanced at the apex (Fig. (Fig.8B).8B). Also in accord with transcript expression (Fig. (Fig.8C),8C), the expression of KYVQEDRL channels was relatively constant along the cochlea (Fig. (Fig.8D).8D). The expression of REVEDEC channels was enhanced at the apex (Fig. (Fig.8F),8F), again in accordance with transcript expression (Fig. (Fig.8E)8E) and with immunohistochemical observations (data not shown). These results indicate that the gradient in cSlo transcripts gives rise to a gradient in BK channel expression with greater expression at the cochlear apex.
A systematic variation in the average time that a BK channel spends in the open state is a conspicuous feature of tonotopic organization in the cochleae of the chicken and the turtle; this trait is invariably taken into account in models of electrical tuning (2, 6, 15, 34, 43). It has been suggested that alternative splicing of the cSlo gene coding for the pore-forming α subunit of the BK channels is responsible for the observed variation of the channel's kinetics (37). We therefore estimated the open times of the two most abundant splice variants, KYVQEDRL and REVEDEC (Fig. (Fig.9A),9A), whose relative expression changes between the low- and the high-frequency regions of the chicken's cochlea. The mean open times calculated using the maximum-likelihood method were 1.69 ms (95% confidence interval, 1.67 to 1.72 ms) for REVEDEC channels and 1.22 ms (95% confidence interval, 1.19 to 1.24 ms) for KYVQEDRL channels (Fig. (Fig.9B).9B). Because this determination was parametric, based on a model with exponentially distributed open times, we also used a bootstrap resampling method that was free from assumptions about the shape of the underlying distribution. From 1,000 iterations of the bootstrap, the means of the open times were 1.69 and 1.21 ms for, respectively, the REVEDEC and the KYVQEDRL variants, essentially the same values as those obtained by the maximum-likelihood method (Fig. (Fig.9B,9B, inset). It is interesting, however, that the standard deviations, 1.83 ms for REVEDEC channels and 1.57 ms for KYVQEDRL variants, do not equal the means. The fact that the open-time distributions deviate slightly from exponential, for which the two parameters are equal, implies an additional source of variability. This could result from channel flicker, correlations between channels in the patch, or other factors affecting the rate of the transition from the open to the closed state.
The key result of the analysis was that the REVEDEC channel has a longer mean open time than the KYVQEDRL variant (Fig. (Fig.9B).9B). Indeed, the probability that the samples of the channels' mean open times belong to the same continuous distribution is 10−102 as determined by a two-sample Kolmogorov-Smirnov test.
It is instructive to compare the effect of splicing on the channels' open times with the effect exerted by the accessory β subunits that are known to delay the transition from the open to the closed state (34, 36). Although a previous report indicated that the β subunit is more highly expressed at the apex, the evidence was an in situ hybridization in the quail's cochlea (36). Because we had previously detected β2 and β4 subunits in cochlear samples (data not shown), we used a custom antiserum raised against a peptide unique to the β1 isoform. Even though the expression of the cognate transcript shows no significant difference along the cochlea (Fig. 10A) and even a tendency toward greater expression at the base, our immunoblotting (Fig. 10B) and immunohistochemical results (not shown) indicated that the β1 subunit is more extensively expressed at the apex. Taken together, these results suggest the importance of a posttranscriptional mechanism in the regulation of β1 expression.
In the presence of β1 subunits, single-channel currents for the REVEDEC variant are prolonged (Fig. 10C), a finding in accord with the well-known kinetic effect of the β1 subunit. In keeping with published data (34, 36), we also observed this effect with the KYVQEDRL variant as well as with putatively chimeric channels obtained by cotransfecting the cells expressing the β1 subunit with plasmids for both REVEDEC and KYVQEDRL splice variants (data not shown). The following strategy was used to estimate the parameters of the underlying open-time distribution. We supposed that the effect of β1 subunits is to create a second kinetic population of channels with exponentially distributed open times and a larger mean (15). We then used a Markov-chain Monte Carlo method to estimate the probability-density functions for the means of the two underlying distributions as well as the relative weights of the two. Based on the data from six cells, the two probability-density functions showed means of 1.65 ± 0.08 ms and 6.88 ± 0.40 ms (Fig. 10D). The estimated probability that a given channel opening belongs to the “slow” population is about one third. Note that the first mean is close to 1.69 ms, the mean open time of the REVEDEC channel without the β1 subunit (Fig. (Fig.9A).9A). It is probable that some channels in the recorded patches simply did not contain β1 subunits because the stoichiometry of the subunit interaction is not regulated in an expression system. Also, even when a β1 subunit is present in the channel assembly, some openings may be as short as in the absence of the accessory subunit. The second mean is over 4-fold as large as the first, which reflects the effect of the β subunit on the channels' open time. The presence of the β subunit therefore exerts a potent effect on the channel kinetics, prolonging the mean open time more than 4-fold.
To explore the possibility that a differentially expressed splicing activator or repressor underlies the increased exon inclusion along the cochlea, we performed a large-scale analysis of cochlear gene expression using a chicken-genome array. Five separate pairs of total RNA samples from the base and apex were analyzed, each isolated from 80 to 120 cochlear segments. We confirmed the robust expression of cochlear marker genes such as those encoding otoferlin and myosin VI. Although several genes showed at least a 4-fold difference in expression between the apex and base, no significant difference in the expression of the splicing factors was observed between the two regions. The data from the base and the apex were therefore combined and the four microarray experiments were pooled, after which the genes with reproducible expression (P < 0.05) in at least one of the replicates were identified. The splicing-factor genes that emerged were clustered by similar expression profiles and displayed with their significances, a measure of the reproducibility of the expression values across samples (Fig. (Fig.11).11). The fact that genes with lower expression have lower significance highlights the difficulty of measuring splicing factors, which generally have low expression levels.
The splicing factors that we identified are associated primarily with the major spliceosome. Half of them occur in the spliceosome C complex, the unstable intermediate of the splicing reaction that forms after the early spliceosome E has been assembled. The major spliceosome's core components, such as the small nuclear RNAs U1, U2, U4, U5, and U6, were also present but were filtered out mainly due to variations in concentration between samples. The presence in the list of the U11/12 25-kDa protein and splicing factor 3B subunit 3 suggests a role for the minor spliceosome (20) in the cochlea. One-tenth of the listed splicing factors have a function in RNA stability, and two factors have been implicated in nonsense-mediated decay, highlighting the importance of these pathways in the cochlea.
To validate the list, we used qPCR to confirm the presence of PRP8 and the absence of KHSRP in cochlear cDNA. We also tested the expression levels of splicing factors BAT1 and YT521B, which were filtered out in the analysis. Both were detected by qPCR, showing that there are additional cochlear splicing factors that fall below the detection threshold of microarray analysis. This list of expressed splicing factors, the first from cochlear RNA, is likely to represent genes of high functional relevance and provides a starting point for further study.
We have tested the hypothesis that the inclusion of alternative cSlo exons along the cochlea is tightly regulated to generate channels with variable kinetics matching the tonotopic gradient. We examined the sequences of more than 4,000 complete cDNA clones of cSlo channels expressed at the cochlear base and apex.
The most striking observation was that the actual variability of the cSlo transcripts is quite low. First, not all alternative splicing sites are used. Sites 1 and 6 are not used in the chicken's cochlea; site 5 seems entirely inactive in the chicken. Site 3 uses only one alternative sequence. Finally, there appear to be no alternative amino termini. Considering three possible carboxyl termini, there are 48 possible variants, far fewer than the 500 predicted if all of the splicing sites were active. The use of an alternative splice-acceptor site that introduces a single additional amino acid at the carboxyl terminus is unlikely to have an impact on the channel's behavior.
The second reduction in complexity is the low number of transcripts with alternative exons: fewer than 20% of all clones analyzed display alternative exons. This number is qualified, however, by the fact that we used equal amounts of PCR products for the amplification of short and long cSlo variants from the base and apex. Correcting the relative numbers with the quantitative data, we obtain an estimate that 30% of the transcripts possess alternative exons; of these, only 10% have more than one alternative exon. Finally, of the 48 possible variants, only 28 were found in the cochlea. This trend is probably conserved in tetrapods, for Slo clones isolated from a rat cochlear cDNA library contain no alternative exons and RT-PCR of partial Slo sequences discloses a low abundance of alternative exons (23). A possible interpretation of the low usage of alternative exons is that the inclusion of one variant of the α subunit in a channel tetramer suffices to induce a change in the kinetics, whereas the incorporation of additional variants disrupts the channel's function. There might thus be an advantage in maintaining low heterogeneity in channel tetramers.
The analysis of complete cSlo transcripts failed to reveal a clear difference in the combinations of exons expressed along the cochlea. Instead, it appears that all combinations of exons occur throughout. The small total number of variants with multiple alternative exons can explain why some possible combinations were not found. These results suggest that the pattern of the alternative exons is not relevant to the formation of the tonotopic gradient. This idea has already emerged from a study of the expression of six constructs containing different combinations of alternative exons from the turtle's Slo channel (16). It was found that longer inserts correlate with slower channels, regardless of the particular sequences included.
Our results suggest that there is no particular combination of exons preferred along the cochlea. It is important to note, however, that long cSlo transcripts containing exon 11 that are uncommon in the cerebellum are expressed in the cochlea, preferentially at its base, apparently as the result of a cochlea-specific splicing reaction. Cloned from a human cell line and expressed in HEK-293 cells, exon 11-containing Slo transcripts gave rise to channels that had a higher open probability than the exon 10-based variants (39). The former channels could therefore be appropriate for high-frequency hair cells, in which case the distribution of transcripts seems reasonable.
The most prominent splicing regulation found was the selection of exons encoding the carboxyl termini of BK channels. The observation that BK channels are more highly expressed at the low-frequency end of the cochlea runs counter to electrophysiological evidence (11, 12). It is possible that the excess channels at the apex, whether KYVQEDRL or REVEDEC variants, are not functional. Variations in BK sequence have been reported to change the subcellular distribution of the channels (22, 44). Furthermore, the REVEDEC BK isoform is largely retained in the endoplasmic reticulum (18, 27), requiring expression of the β1 subunit to be transported to the cell surface (19). The β1 subunit also increases the membrane localization of the human KYVQEDRL channel (40) by recognizing a carboxyl-terminal sequence (YLSIL) that occurs with a conservative substitution (YVSVL) in the chicken. In addition to its effects on the electrophysiological properties of the α subunit (29, 33), the β subunit might act as a chaperone, a function that has been ascribed to the β subunits of other ion channels (9).
The high expression of both the REVEDEC α subunit and the β1 subunit at the cochlear apex implies that the channel is not only correctly localized and functional but also exhibits slower kinetics. The increased expression of the β1 subunit at the apex could have evolved both to rescue the retained BK channels that would otherwise stress the endoplasmic reticulum (28) and to stabilize the channel's open conformation.
Our single-channel recordings show that REVEDEC channels have a greater mean open time than KYVQEDRL channels, which is consistent with the prevalence of the former variant in the low-frequency region. Although this difference has a high statistical significance, the physiological relevance of such a small kinetic effect is questionable. On the other hand, it is clear that the β1 subunit significantly delays the transition from the open to the closed state in the KYVQEDRL channel (34, 36). We confirmed this result and demonstrated a similar effect on the REVEDEC channels that occur abundantly with β1 subunits in the low-frequency region of the cochlea. We conclude that, in comparison to a very modest difference owing to alternative splicing, the presence of the auxiliary β1 subunit is more important in determining cochlear tonotopy. It nevertheless remains possible that these BK isoforms display a different kinetic behavior when expressed in the chicken's cochlea. To test this possibility is technically challenging and represents an open question for further investigation.
The activity at the resting potential of brain BK channels encoded by exon 11-containing Slo is not affected by the β4 subunit (39), the predominant isoform in the brain. The corresponding cochlear variant might also be unaffected by the expression of β1 subunits, which would constitute another reason for reducing the expression of exon 11 in the low-frequency region. The cSlo transcript found at the apex and containing exons 10, 11, 19, 22, and 23 was subcloned after the removal of exon 10, which generated an early stop codon, and expressed in HEK 293 cells. Our preliminary recordings indicate that this variant is also not affected by the β1 subunit (not shown), which is biophysically interesting but of arguable physiological significance. The rarity of this subunit, which represents only 0.13% ± 0.09% of transcripts, and the early stop codon suggest that the corresponding channel is not expressed significantly in the cochlea. Further testing is needed to characterize this channel and to determine whether the inclusion of exon 11 confers resistance to the β1 subunit.
The mechanism that regulates the gradient of exon inclusion along the cochlea remains to be determined. We used a microarray containing all available sequences from the complete chicken genome project, giving a total of 32,000 genes. Although we identified 48 splicing factors likely to be important in the normal functioning of hair and supporting cells, no gene encoding a splicing factor was differentially expressed between the base and the apex of the cochlea. The lack of differentially expressed splicing factors suggests that the putative splicing activator or repressor that generates the gradient is yet to be recognized. Alternatively, a posttranscriptional modification might change the activity along the cochlea of splicing factors present on the list.
cSlo expression might also be regulated by intron retention. In a process called regulated unproductive splicing, the production of premature stop codons provides an additional level of posttranscriptional regulation (24). We did not observe a general decrease in splicing fidelity in samples from the apex or base. Furthermore, the two splicing factors in our list that play a role in nonsense-mediated decay showed constant expression along the cochlea. This suggests that regulated RNA degradation is likely to be important in the cochlea, but not as a mechanism to increase splicing fidelity at the base. Our finding by RACE that short and long cSlo transcripts have different 3′UTRs suggests the possibility of differential regulation by microRNAs, small RNAs that can induce the degradation of messenger RNAs or inhibit their translation (41). In fact, miR-9 regulates the stability of Slo splice variants in the rat brain (32). From bioinformatic predictions (7) we determined that gga-miR-205a is a good candidate to bind the 3′UTR of long cSlo but not that of short cSlo. Our preliminary results suggest, however, that this microRNA is not expressed in the cochlea (not shown), a conclusion supported by the absence of miR-205 in the murine cochlea (42).
In conclusion, our results indicate that the inclusion of alternative exons in cSlo transcripts gradually wanes from the low-frequency apex to the high-frequency base of the chicken's cochlea. Variants of the short cSlo sequence that contain exons 34, 35, and a part of exon 33 are more common in the low-frequency apex and give rise to slower channels. A still stronger kinetic effect on single-channel currents is exerted by the auxiliary β1 subunits that are abundant at the apex and quadruple the channel's mean open time. The original hypothesis that places an emphasis on the tightly regulated inclusion of alternative cSlo exons along the cochlea as a means of establishing the tonotopic gradient should be revised.
We thank D. Andor-Ardó for assistance with Bayesian statistics; B. Fabella for LabVIEW programming; S. Pylawka, J. Villiers, and M. Vologodskaia for technical assistance; and A. Le Boeuf for comments on the manuscript. We are grateful to D. Betel (Memorial Sloan-Kettering Cancer Center) for help in predicting miRNA targets for the cSlo 3′UTR sequence and to M. Grunnet for discussions about unproductive splicing.
This research was supported by grant DC00241 from the National Institutes of Health and by a Sue and Frank Binswanger Grant in Auditory Science from the National Organization for Hearing Research Foundation. S.M.-R. was supported by an F. M. Kirby Postdoctoral Fellowship; A.J.H. is an Investigator of Howard Hughes Medical Institute.
Published ahead of print on 17 May 2010.
†Supplemental material for this article may be found at http://mcb.asm.org/.