This article has a correction. Please see:


In dogs and in humans, potassium channels formed by ether-a-go-go-related gene 1 protein ERG1 (KCNH2) and KCNQ1 α-subunits, in association with KCNE β-subunits, play a role in normal repolarization and may contribute to abnormal repolarization associated with long QT syndrome (LQTS). The molecular basis of repolarization in horse heart is unknown, although horses exhibit common cardiac arrhythmias and may receive drugs that induce LQTS. In horse heart, we have used immunoblotting and immunostaining to demonstrate the expression of ERG1, KCNQ1, KCNE1, and KCNE3 proteins and RT-PCR to detect KCNE2 message. Peptide N-glycosidase F-sensitive forms of horse ERG1 (145 kDa) and KCNQ1 (75 kDa) were detected. Both ERG1 and KCNQ1 coimmunoprecipitated with KCNE1. Cardiac action potential duration was prolonged by antagonists of either ERG1 (MK-499, cisapride) or KCNQ1/KCNE1 (chromanol 293B). Patch-clamp analysis confirmed the presence of a slow delayed rectifier current. These data suggest that repolarizing currents in horses are similar to those of other species, and that horses are therefore at risk for acquired LQTS. The data also provide unique evidence for coassociation between ERG1 and KCNE1 in cardiac tissue.

  • KCNH2
  • KvLQT1
  • minK
  • IsK
  • MiRP
  • repolarization
  • equine cardiology

the configuration of the cardiac action potential is modulated by the time- and voltage-dependent activation of a variety of potassium currents, the relative contributions of which vary significantly between species. The cardiac action potential is distinguished by a sustained plateau phase that is critical for adequate diastolic filling and systolic contraction. In most large mammals including humans, the duration of the plateau phase is strongly influenced by the activation of two distinct delayed rectifier potassium currents (34, 43): the fast delayed rectifier (I Kr) and the slow delayed rectifier (I Ks). Mutations in multiple human genes that encode the subunits of the ion channels that conductI Kr and I Ks have been associated with the congenital long QT syndrome (LQTS), a disease characterized by prolongation of ventricular repolarization, lengthening of the QT interval on the electrocardiogram, and enhanced potential for syncopal episodes, ventricular arrhythmias, and sudden death (1, 11, 47, 54).

Mutations linked to reduction of repolarizing of the delayed rectifier potassium current and the LQTS may occur in either the pore-forming (α-) or accessory (β-) subunits of the fast and slow delayed rectifier channels. Mutations in the α-(KCNQ1 or KvLQT1) subunit of the channels associated with I Ks have been linked to LQTS1, which is the most common inherited form of LQTS (54). Mutations in the α-subunit of the channels that conduct I Kr (ERG1 or KCNH2) have also been associated with inherited LQTS (LQTS2) (11). Furthermore, antagonism of I Kr by commonly prescribed cardiac and noncardiac drugs is the predominant cause of acquired LQTS, which is a related and much more common syndrome (20).

Mutations in the potassium channel β-subunits KCNE1 (minK or IsK) and KCNE2 (minK-related peptide 1, MiRP1) have also been associated with inherited and acquired LQTS (1, 47). Although it is clear that channels formed by coassembly of KCNQ1 and KCNE1 are required to recapitulate I Ks (4, 42), additional roles for KCNE1 and KCNE2 in native cardiac tissue are uncertain. In heterologous expression systems, KCNE2 can modulate KCNQ1 channel properties (51); in addition, both KCNE1 and KCNE2 can interact with ERG1 (1, 9, 10, 33, 36, 46, 50, 56).

Expression of the currents I Ks andI Kr and the putative molecular correlates KCNQ1, ERG1, KCNE1, and KCNE2 have been documented in a variety of species but not in the horse (6, 17, 29, 34, 38, 43, 55). Cardiac arrhythmias associated with abnormal repolarizations including quinidine-induced torsades de pointes have been documented in horses (40), and horses are routinely treated with other drugs that are known to prolong the QT interval in humans (12). Here, as necessary first steps in substantiating the potential for drug-induced LQTS in horses, we have evaluated the expression, post-translational modification, and coassociation of KCNQ1, ERG1, and KCNE1 proteins in equine heart. We demonstrate for the first time the coassociation of KCNE1 not only with KCNQ1 but also with ERG1 in native cardiac tissue. In addition, we provide evidence thatI Kr and I Ks contribute to action potential repolarization and we document the presence ofI Ks in equine cardiac myocytes.



Horses were donated to the Colleges of Veterinary Medicine at Kansas State University and Cornell University and had either life-threatening or career-ending disorders that were unrelated to the cardiovascular system. Horses underwent euthanasia with an overdose of pentobarbital sodium (85–110 mg/kg body wt) and then their hearts were rapidly excised.


Chemicals were obtained from Sigma (St. Louis, MO) unless stated otherwise. A polyclonal rabbit anti-KCNQ1 antibody directed against amino acids 522–540 of human KCNQ1 isoform 1 was generated by Spring Valley Laboratories (Woodbine, MD). A polyclonal rabbit antibody directed against amino acids 65–79 of KCNE1 was a gift from R. Dumaine (Masonic Medical Research Laboratory; Utica, NY). MK-499 was obtained from Merck Research Laboratories (Westpoint, PA). Human embryonic kidney (HEK)-293 cells stably transfected with the humanether-a-go-go-related gene were provided by Zhengfeng Zhou and Craig January (University of Wisconsin, Madison). HERG cDNA used for transient transfection of Chinese hamster ovary (CHO) cells was obtained from Robert Kass (Columbia University, New York, NY). CHO (CCL-61) and T84 (CCL-248) cell lines were obtained from the American Type Culture Collection (Manassas, VA). Primers for nonnested PCR of KCNE2 were a gift from Randy Wymore (University of Tulsa, OK).

Isolation of equine cardiac myocytes.

Equine myocardium was harvested into oxygenated Ca2+-free Tyrode solution consisting of (in mM) 140 NaCl, 5.4 KCl, 1.2 MgCl2, 5 HEPES, 5 glucose, and 30 2,3-butanedione monoxime (BDM). Small cubes (1–3 mm3) of cardiac tissue were incubated in oxygenated Ca2+-free Tyrode solution containing 1 μg/μl collagenase type II (GIBCO-BRL; Grand Island, NY) and 0.5 μg/μl protease type XIV at 37°C for 45 min before the supernatant was discarded and replaced with Ca2+-free buffer containing 1 μg/μl collagenase type II. After 15 min, myocytes were collected in the supernatant, pelleted by centrifugation (300 rpm for 3 min), then resuspended in incubation buffer consisting of (in mM) 118 NaCl, 4.8 KCl, 1.2 MgSO4, 1.2 KH2PO4, 0.68 glutamine, 11 glucose, 20 NaHCO3, 5.0 HEPES, 5.0 pyruvate, 10 taurine, 0.5 CaCl2, and 30 BDM; with 2% bovine serum albumin. The process of digestion and centrifugation of the cardiac tissues was repeated six additional times or until adequate cell numbers were obtained. Isolated cells were used for either immunocytochemistry or patch-clamp analysis.

Preparation of membranes and lysates.

For Western blot analyses including deglycosylation experiments, crude membrane fractions were isolated from equine atrium, ventricular endocardium, and ventricular epicardium using a previously described method that employs KI to minimize aggregation and nonspecific binding of proteins (5). Membrane protein fractions used for immunoprecipitation and coimmunoprecipitation experiments were isolated similarly but in the absence of KI to avoid inadvertent disruption of specific protein-protein interactions. Whole cell lysates were prepared from monolayer cultures of HEK-293, CHO, T84, and porcine granulosa cells in a buffer consisting of PBS with 1% Nonidet P-40 (NP-40), 0.5% sodium deoxycholate, 0.1% SDS, and a 1:500 dilution of P-8340 protease inhibitor cocktail (Sigma).

Glycosidase digestion.

Peptide N-glycosidase F (PNGase F, New England BioLabs; Beverly, MA) was used to cleave N-actelyglucosamine from asparagine residues. Membrane proteins (120 μg) were denatured by heat at 100°C for 5 min in 0.5% SDS and 1% β-mercaptoethanol. After the samples were cooled, 50 mM sodium phosphate buffer (pH 7.5) and 1% NP-40 were added at a 1:10 dilution. PNGase F (3,000 U) was also added, and the resultant reaction was incubated at 37°C for 2 h. Each experiment included negative controls consisting of identically treated proteins incubated in the absence of enzyme. Products of control incubations and glycosidase digestions were subsequently analyzed by SDS-PAGE and Western blotting.


Precipitation of equine ERG was performed using the Seize X protein G immunoprecipitation kit (Pierce) according to the manufacturer's instructions.

Precipitations of equine KCNE1 and Kv1.1 were performed using ImmunoPure Plus immobilized protein G (Pierce). To limit nonspecific binding, a preclearing step was performed: membrane protein (∼2,000 μg) from equine ventricle was incubated with 30 μl of protein G-agarose beads for 2 h at 4°C and pelleted by centrifugation (10,000 g for 3 min). A 1:200 dilution of antibody (Alomone Laboratories; Jerusalem, Israel) was added to the supernatant and incubated overnight at 4°C. Subsequently, the sample was incubated at 4°C for 2 h with protein G-agarose beads (40 μl). The beads were pelleted by centrifugation (10,000 g for 3 min) and washed three times before the proteins were eluted from the beads with SDS sample buffer that contained 5% β-mercaptoethanol.

Western blot analysis.

Precipitates or crude membrane proteins (∼65 μg) were separated using SDS-PAGE before being transferred to nitrocellulose, blocked for 1 h at room temperature in 5% nonfat milk in Tris-buffered saline with 0.1% Tween 20, and then incubated overnight at 4°C in primary antibody in blocking buffer. Primary antibody dilutions were anti-HERG1 (1:500, Alomone Laboratories), anti-KCNQ1 (1:1,000), anti-KCNE3 (1:400, Santa Cruz Biotechnology; Santa Cruz, CA), and anti-KCNE1 (1:5,000, R. Dumaine). Washed membranes were incubated for 1 h in horseradish peroxidase (HRP)-conjugated secondary antibody (Santa Cruz Biotechnology or Amersham Pharmacia Biotech; Piscataway, NJ). Bound antibodies were visualized using either the enhanced chemiluminescence detection system (Amersham Pharmacia) for blots probed with anti-HERG1, anti-KCNQ1, and anti-KCNE3 antibodies or the SuperSignal West Femto maximum sensitivity substrate (Pierce) for blots probed with anti-KCNE1 antibody. Results were recorded on radiographic film. Equal loading was confirmed by either immunoblotting for α-actin or Coomassie blue staining of gel lanes loaded and subjected to SDS-PAGE in an identical manner. In experiments where a single membrane was probed under more than one experimental condition, primary and secondary antibodies were removed using Restore Western blot stripping buffer (Pierce).


Myocytes were fixed in 2% paraformaldehyde in incubation buffer for 15 min and 4% paraformaldehyde alone for 15 min, washed three times with PBS, and then incubated in 50 mM ammonium chloride for 15 min to quench free aldehydes. After three additional washes, cell membranes were permeabilized by a 15-min incubation in 0.1% NP-40 in blocking buffer (1% goat serum and 1% nonfat dried milk in PBS). Cells were then washed three more times with PBS and incubated in blocking buffer overnight at 4°C. Subsequently, the myocytes were incubated with anti-HERG1 antibody in blocking buffer (1:200 dilution) for 1 h at room temperature and washed three times with PBS and then incubated in the dark in FITC-conjugated polyclonal goat anti-rabbit antibody (Molecular Probes; Eugene, OR) for 1 h at room temperature. After three additional washes, cells were placed under coverslips on slides in the presence of an antifade reagent (Slowfade, Molecular Probes). Slides were stored in the dark at 4°C until a Zeiss laser scanning microscope (LSM 410) was used to visualize the cells.


mRNA was isolated from equine atrium and ventricle samples using a Fast Track kit (Invitrogen; Carlsbad, CA). Reverse transcription was performed using random hexamers and enhanced avian RT (Sigma) under the manufacturer's recommended conditions. PCR detection of KCNE2 was performed using gene-specific oligonucleotide primers designed to amplify a 228-base-pair fragment (sense, 5′-CGCNCCNADGTTCTCATGGATGGTGGC; antisense, 5′-GTNGATGCNGAGAACTTCTACTAYG). PCR detection of KCNQ1 isoforms 1 and 2 was performed using primers (LQT201, LQT301, ET30A) described previously (24). The PCR reaction mixtures (25 μl) contained 0.4 μM each primer, 10 mM Tris · HCl (pH 8.3), 50 mM KCl, 0.2 mM each dNTP, 1.5 mM MgCl2, and 1 U REDTaq DNA polymerase. The PCR reaction was amplified for 30 cycles consisting of denaturation for 1.5 min at 94°C, annealing at 65°C for 1.5 min, and extension at 72°C for 1.5 min. The amplification product was extracted from the agarose gel using a Qiaex II gel extraction kit (Qiagen; Valencia, CA). The extracted DNA was reamplified using the described protocol. The subsequent PCR product was cloned into DH5α-competent cells using an Original TA Cloning kit (Invitrogen) and was sequenced using T7 and M13 primers.

Microelectrode recordings.

Thin sheets (∼2 mm) of equine epicardium were obtained using a 3-in Brown air dermatome (Zimmer; Dover, OH) placed in oxygenated incubation buffer and air shipped (US Airways, PDQ) the same day from Kansas State University to Cornell University where microelectrode studies were performed. These tissues were in transit for ∼8 h. Tissues were also obtained from horses euthanized at Cornell University.

Epicardium was further dissected into 2 × 1-cm sections, mounted in a Plexiglas chamber, and superfused at a rate of 15 ml/min with Tyrode solution, which included (in mM) 0.5 MgCl2, 0.9 NaH2PO4, 2.0 CaCl2, 137 NaCl, 24 NaHCO3, 4.0 KCl, and 5.5 glucose aerated with 95% O2-5% CO2 (Po 2 = 400–600 mmHg; pH = 7.35 ± 0.05; temperature = 37.0 ± 0.5°C). Transmembrane action potentials were recorded using standard microelectrode techniques (23).

Whole cell patch-clamp analysis.

Recording conditions, solutions, and voltage protocols used to measureI Ks have been described previously in detail (18, 19, 32). Briefly, the extracellular solution was a modified HEPES-buffered (pH 7.3) Tyrode solution containing 0.1 mM potassium, 500 nM nisoldipine, and 5 μM E-4031.I Ks was measured at room temperature (22°C) as time-dependent current at the end of 4-s depolarizing test pulses (−20 to +60 mV) and deactivating tail current after return to the −40 mV holding potential. Conductance-voltage (G-V) relationships were constructed using the transformG(V) =I(V)/(VE K), where I(V) is the measured current at each test potential (V) andE K is the calculated reversal potential for potassium; experimental data were fit using a Boltzmann function (Origin 6.1, OriginLab; Northampton, MA).

Horse ventricular myocyte membrane capacitance, calculated as the time integral of the capacitive response to a 5-mV hyperpolarizing step from a −40 mV holding potential, was 226 ± 11.5 pF (n= 13). Series resistance, estimated by dividing the time constant of the decay phase of the uncompensated capacitive transient by the calculated membrane capacitance, was 7.5 ± 0.7 MΩ, and was electrically compensated to minimize the duration of the capacity transient (>50%). Voltage errors associated with uncompensated series resistance did not exceed 4 mV.

Statistical analysis.

The apparent molecular mass of protein bands detected on Western blots was determined using the AlphaImager system (Alpha Innotech; San Leandro, CA). Immunoblotting results shown in Figs. 1-7 are representative of at least three independent experiments. Concentration-dependent effects of cisapride (Janssen Pharmaceuticals; Titusville, NJ) were assessed using repeated-measures ANOVA and Tukey's method for multiple comparisons (Statistix7, Analytical Software; Tallahassee, FL). Statistical significance was defined byP < 0.05.

Fig. 1.

A: Western blot analysis ofether-a-go-go-related gene protein 1 (ERG1) expression in equine atrium (Atria), equine ventricular endocardium (Endo), and equine ventricular epicardium (Epi). Antibody was preincubated for 1 h before use either without (−) or with (+) a fourfold excess of epitope-specific peptide. Arrows indicate molecular mass (kDa).B: whole cell lysates obtained from HEK-293 cells stably transfected with HERG1 (HEK-HERG) and membranes isolated from porcine granulosa cells (PGC) were used as positive and negative controls, respectively. C: Western blot analysis of ERG1 in membranes from a different horse than in A (5-min exposure) shown with associated positive control (1-min exposure).

Fig. 2.

Immunolocalization of equine ERG1 in myocytes isolated from equine atrium (A) and ventricle (B). Negative staining controls included atrial myocytes fixed with 4% paraformaldehyde that were then incubated with only secondary antibody (C) and ventricular myocytes fixed with 4% paraformaldehyde (D). Brightly staining clusters in A–D probably result from mitochondrial autofluorescence.

Fig. 3.

N-linked glycosylation of equine ERG1. PeptideN-glycosidase F (PNGase F) treatment of ERG1 from equine heart: membrane proteins from atria, ventricular endocardium, and ventricular epicardium and lysates of HEK-293 cells stably transfected with HERG (HEK-HERG). Cells were incubated at 37°C for 2 h in the presence or absence of PNGase F before Western blot analysis with anti-HERG1 antibody. Exposure time of HEK-HERG was approximately one-tenth that of cardiac membranes.

Fig. 4.

Expression and N-linked glycosylation of KCNQ1. A: crude membrane fractions from equine atrium (At), ventricular endocardium (En), and epicardium (Ep) were subjected to Western blot analysis using antibody against KCNQ1 (rabbit polyclonal). Cardiac membranes from guinea pig heart (GP) were included as a positive control. B: anti-KCNQ1 (goat polyclonal) also detected bands at 60 and 75 kDa in the fraction of equine atrial membranes precipitated by antibody directed against KCNE1 in the absence but not the presence of blocking peptide. C: RT-PCR for amplification of KCNQ1 isoform 1 (348 bp; lanes 1, 4), KCNQ1 isoform 2 (306 bp; lanes 2, 7–9, 13), and GAPDH (450 bp; lanes 3, 10–12) from equine cardiac mRNA (lanes 1–3, 7, 10), equine midmyocardial mRNA (lanes 8, 11), human cardiac cDNA (lanes 9, 12), and negative controls with no RT (lanes 4, 13). Lanes 5 and 6 contain 100-bp DNA ladder. D: before Western blot analysis with anti-KCNQ1 (rabbit polyclonal), equine cardiac membranes were treated with PNGAse F as described for Fig. 3. Apparent molecular mass of the 75-kDa band was shifted to ∼65 kDa. A nonspecific band is apparent in each lane at ∼40 kDa.

Fig. 5.

Expression of KCNE family members in equine heart.A: Western blot analysis of KCNE1 expression in membrane fractions from equine atrium (A) and ventricle (V) detected three bands with molecular mass between 18 and 25 kDa that correspond closely to the three bands present in the guinea pig heart positive control (GP), which is presumed to represent differentially glycosylated forms of KCNE1. B: Western blot analysis of KCNE3 expression in equine atrium, ventricle, and the positive control (T84 cells) revealed a single band with an apparent molecular mass of 27 kDa in all three lanes (top). When the blot was stripped and reprobed with anti-KCNE3 antibody preincubated with blocking peptide, the band was no longer apparent (bottom). C: RT-PCR was used to amplify a 228-bp fragment of KCNE2 from equine ventricular (V), cardiac (C, atrial + ventricular), and atrial (A) mRNA.

Fig. 6.

Specific immunoprecipitation of equine ERG1.A: cardiac membrane fractions were immunoprecipitated using the Seize X protein G kit (left) or immobilized protein G (right) and immunoblotted with the same anti-HERG antibody. A porcine granulosa cell lysate (PGC) and protein immunoprecipitated from Chinese hamster ovary (CHO) cells transiently transfected with HERG (CHO-HERG) are also shown (right). Arrows indicate molecular mass (kDa). B: membrane fractions from equine (EQ) ventricle were precipitated with anti-HERG antibody as described, and the resultant immunoprecipitate (IP) and supernatant (S) were immunoblotted with antibody against KCNQ1. KCNQ1 was also present in membranes from EQ ventricle and guinea pig heart (GPH) not subjected to immunoprecipitation (−) but was absent from lysate of HEK-293 cells stably transfected with HERG (HEK-HERG). IgG bands are apparent at 50 kDa.

Fig. 7.

Coimmunprecipitation of ERG1 and KCNE1 but not KCNE3. A: membranes from equine atrium and ventricle were precipitated using anti-HERG1 and immunoblotted with anti-KCNE1 antibody. Four bands were identified between 30 and 18 kDa in immunoprecipitates; three were also present in the supernatants. Guinea pig heart membranes not subjected to HERG immunoprecipitation (−) were included as a positive control for KCNE1 (arrowheads). Ventricular membranes precipitated using anti-Kv1.1 (α-Kv1.1 immunoprecipitate) were the negative control. A single ∼30-kDa band consistent with IgG light chain was present in the anti-Kv1.1 and anti-HERG precipitates (double-headed arrow). B: equine cardiac membranes were precipitated using anti-KCNE1 and immunoblotted with anti-HERG. Three ERG1 bands with apparent molecular masses of 100, 145, and 250 kDa were apparent in the anti-KCNE1 immunoprecipitate (left), and the anti-HERG1 immunoprecipitate (right) was performed concurrently as a positive control. C: when anti-HERG1 precipitates and supernatants were immunoblotted with anti-KCNE3, a 27-kDa band was detected only in the supernatants.


Expression of ERG1 in horse heart.

An equine homolog of ERG1 was identified by immunoblot with an antibody directed against a COOH-terminal epitope. Two bands that correspond to reported molecular masses of ERG proteins were observed on immunoblots of equine atrium, ventricular endocardium, and ventricular epicardium at 145 and 100 kDa (Fig. 1). Similar results were obtained using membranes isolated from hearts of five horses; however, the presence and relative intensity of the 100-kDa band varied between preparations obtained from different animals (Fig.1, A vs. C). Protein comparable to the 145-kDa protein detected in the horse heart was also present in the HEK-293 cells that served as a positive control (59), whereas no protein was detected in the porcine granulosa cells that served as a negative control (Fig. 1 B; Ref. 32). The bands detected at 100 and 145 kDa on blots probed with anti-HERG1 antibody were diminished visibly when the antibody was incubated with an excess (8 μg) of peptide corresponding to the anti-HERG1 epitope (Fig.1 A). These data demonstrate specificity of the anti-HERG1 antibody for equine ERG1 and suggest that two forms of ERG1 are expressed throughout the equine myocardium.

Protein detected by the ERG1 antibody at molecular mass >200 kDa in the equine heart and transfected HEK-293 cells (Fig. 1) is likely to represent aggregated protein associated with boiling samples before SDS-PAGE. Similar material was present on immunoblots of boiled HERG-transfected HEK-293 membranes (58, 59) but was absent from blots of similar samples heated to 70°C (37).

Immunocytochemistry was performed to confirm the expression of ERG in the equine heart and to determine the subcellular distribution of the protein in atrial and ventricular myocytes. As described previously for the rat heart (38), equine ERG1 has a generalized distribution pattern throughout atrial and ventricular myocytes and does not appear to be clustered at the intercalated discs (Fig.2, A and B). Furthermore, the linear fluorescence pattern suggests that ERG1 is associated with the T tubules of both atrial and ventricular myocytes. Similar fluorescence patterns were not observed in the negative controls (Fig. 2, C and D).

Glycosylation of ERG1 in horse heart.

To determine whether equine ERG is N-linked glycosylated, crude cardiac membrane fractions were incubated in the presence and absence of PNGase F before Western blot analysis. PNGase F cleaves oligosaccharide moieties from asparagine. Therefore, changes in molecular mass demonstrated by immunoblotting may reflect glycosylation of mature proteins as well as immature proteins within the endoplasmic reticulum and golgi. PNGase F treatment increased the electrophoretic mobility of bands detected by immunoblotting with anti-HERG1 (Fig.3, left). Close inspection of the data presented in Fig. 3 reveals that the antibody detects a doublet centered at 145 kDa in membranes from horse heart as well as HEK-293 cells transfected with HERG1. As reported previously (59), PNGase F treatment decreased the apparent molecular mass of the upper bands by ∼10 kDa and the lower bands by ≤5 kDa. These data suggest that in equine heart the 145-kDa form but not the 100-kDa form is N-linked glycosylated. The immunoreactive protein with apparent molecular mass >200 kDa, which is assumed to be aggregated protein, was insensitive to PNGase F (Fig. 3) and thus cannot represent extensively glycosylated ERG1 that is comparable to the 205-kDa ERG1 described in rodent heart (38).

Expression and glycosylation of KCNQ1 in horse heart.

In addition to the experiments focused on ERG1, the α-subunit of the channels that conduct I Kr, we performed Western blot analysis to evaluate expression of KCNQ1, the α-subunit ofI Ks channels. Immunoblotting with anti-KCNQ1 antibody revealed a prominent band with an apparent molecular mass of 75 kDa as well as more faint bands at 60 and 100 kDa in equine atrium, ventricular epicardium, and endocardium and in the guinea pig heart positive control (Fig.4 A).

The proteins detected at 60 and 75 kDa could represent either different isoforms of KCNQ1 or differentially glycosylated forms of a single KCNQ1 isoform (14, 22, 53). RT-PCR amplified product of the expected size from equine cardiac mRNA with a primer set specific for isoform 1 but not isoform 2 (Fig. 4 C, left). In contrast, product of the expected size was amplified from reactions containing human cardiac cDNA (Clontech) with isoform 2-specific primers (Fig. 4 C, right). The identity of the PCR products was confirmed by sequencing. The equine KCNQ1 isoform 1 gene fragment (GenBank accession no. AF482704) shared 83% identity with the human clone.

To determine whether the KCNQ1 bands at 75 and 60 kDa in equine heart represent differentially glycosylated protein, crude membrane fractions from atrium, ventricular endocardium, and epicardium were treated with PNGase F. Although the 60-kDa protein was unaffected, the apparent mass of the 75-kDa protein shifted to ∼65 kDa, which indicates that the latter form of KCNQ1 is N-linked glycosylated in equine heart (Fig. 4 D).

The mobility of the 100-kDa band was also affected by PNGase F treatment (Fig. 4 D). This band may correspond to a KCNQ1/KCNE1 complex that was not disrupted by heat and SDS. We have documented previously (27) in guinea pig hearts and pig granulosa cells the presence of bands >100 kDa that were detected by both anti-KCNE1 and anti-KCNQ1 and resolved into bands consistent with KCNQ1 and KCNE1 monomers when heated to temperatures >80°C. Although we did not investigate directly the nature of the PNGase F-sensitive 100-kDa band that reacted with anti-KCNQ1, experiments were performed to confirm expression of KCNE1 and coassociation of KCNQ1 and KCNE1 in equine heart.

To demonstrate directly the coassociation of KCNQ1 and KCNE1 in equine heart, we performed Western blot analysis with anti-KCNQ1 on anti-KCNE1 precipitates from equine atria and ventricle. A commercially available goat-polyclonal KCNQ1 antibody (Santa Cruz Biotechnology) was used in lieu of the anti-rabbit KCNQ1 antibody in this experiment to avoid detection of IgG heavy chain. The anti-KCNE1 antibody (Alomone Laboratories) precipitated KCNQ1 protein from both equine atrial (Fig.4 B) and ventricular (data not shown) membranes. Immunoprecipitation of KCNQ1 protein by this antibody has been shown previously in porcine granulosa cells (32). The specificity of this immunoprecipitation in horse heart is evident from Fig. 6 B, which shows KCNQ1 protein in the supernatant but not the precipitate after anti-HERG immunoprecipitation. We were unable to perform the reciprocal immunoprecipitation to confirm coassociation between KCNQ1 and KCNE1, because neither of the KCNQ1 antibodies immunoprecipitates the target protein.

Expression of KCNE accessory subunits.

KCNQ1 can associate not only with KCNE1, as shown in Fig.4 A, but also with KCNE2 (MiRP1) and KCNE3 (MiRP2) (4,42, 44, 51). Each of these three KCNE family β-subunits is also capable of coassociating with ERG1 and influencingI Kr (1, 33, 44). Here, Western blot analysis was used to confirm the expression of KCNE1 and examine the expression of KCNE3 protein in equine atrium and ventricle. In the absence of commercially available antibody to KCNE2, RT-PCR was performed to evaluate the expression of KCNE2 message in horse heart.

As expected from the immunoprecipitation data in Fig. 4 B, Western blot analysis using anti-KCNE1 antibody revealed three bands with apparent molecular masses between 18 and 26 kDa in equine atrium and ventricle as well as in the positive control, guinea pig heart (Fig. 5 A). The relative mobilities of the detected bands are consistent with those previously reported for fully glycosylated, partially glycosylated, and unglycosylated KCNE1 (4, 33, 49). Immunoblotting with anti-KCNE3 antibody revealed a single band with an apparent molecular mass of 27 kDa in equine atrium and ventricle (Fig. 5 B), which is consistent with that detected in the cell lysates prepared from T84 cells that served as the positive control (44). These bands were not apparent when immunoblotting was performed using anti-KCNE3 antibody preincubated with an excess of the epitope-specific (blocking) peptide (Fig. 5 B). These results suggest that both KCNE1 and KCNE3 are expressed in equine atrium and ventricle.

RT-PCR was used to assess KCNE2 expression in equine atrium and ventricle. Amplification products of the expected size were obtained from atrial, ventricular, and cardiac (atrial plus ventricular) mRNA (Fig. 5 C). Atrial and total cardiac products were cloned. Sequence alignment (BLAST) showed that the gene fragment obtained shared 94% identity with human KCNE2 (GenBank accession no. AF387764). These data provide evidence that the equine myocardium expresses message for KCNE2 protein.

Association of equine ERG with accessory subunits.

The experiments shown in Figs. 6 and7 were performed to determine whether either KCNE1 or KCNE3 coassociate with equine ERG1 in heart. First, sequential immunoprecipitation and immunoblotting were performed to determine whether the anti-HERG1 antibody could precipitate equine ERG1 (Fig. 6 A). The anti-HERG1 antibody detected protein at 30, 50, 100, and 145 kDa and >200 kDa on Western blots of equine cardiac precipitates, and at 130 and ∼250 kDa in the positive control precipitate from a lysate of CHO cells transiently transfected with HERG1. Immunoprecipitation of equine cardiac membranes with anti-HERG1 antibody yielded similar results regardless of the immunoprecipitation technique employed (Fig. 6 A). The 145- and 100-kDa bands in the horse heart were diminished when immunoblotting was performed with antibody preincubated with blocking peptide (data not shown) and were thus identified as the forms of equine ERG precipitated by anti-HERG1 antibody. The 130-kDa band was similarly identified as the form of HERG1 precipitated from the CHO lysate. The bands at 30 and 50 kDa are most likely the light and heavy chain regions of IgG that were either leached from the column (Fig. 6 A, left) or precipitated by immobilized protein G (Fig. 6 A,right). As expected from Fig. 1 B, the intensity of the immunoreactive material detected at >200 kDa and assumed to be aggregated protein was diminished in the presence of blocking peptide.

Immunoblotting with anti-KCNQ1 antibody was done to assess the specificity of the immunoprecipitation technique. Bands corresponding to the molecular mass of KCNQ1 were not observed in the lane with the anti-HERG precipitated proteins but were present in the associated supernatant as well as the membranes prepared from equine and guinea pig heart that served as positive controls (Fig. 6 B). These data not only show that equine ERG is specifically precipitated by the anti-HERG1 antibody, but also confirm that as expected, KCNQ1 does not associate with equine ERG1 in cardiac membranes.

Sequential immunoprecipitation and immunoblotting were used to determine whether ERG1 is associated with either KCNE1 (Fig. 7,A and B) or KCNE3 (Fig. 7 C) in equine atrial and ventricular membrane preparations. Anti-KCNE1 antibody recognized four bands with apparent molecular masses of between 18 and 30 kDa in anti-HERG1 precipitates from equine atrium and ventricle (Fig. 7 A). Consistent with the apparent molecular masses of glycosylated and unglycosylated KCNE1 (18–26 kDa as shown in Fig.5 A), three of these four bands were also detected in the associated supernatants and the guinea pig heart membranes used as a positive control (Fig. 7 A). In contrast, only a single band with an apparent molecular mass of ∼30 kDa, which is consistent with IgG light chain, was detected in the anti-Kv1.1 precipitate (Fig.7 A). These data suggest that ERG1 and KCNE1 associate in horse heart, which is a conclusion that is supported by the results of the reciprocal immunoprecipitation. Anti-HERG1 antibody detected protein consistent with ERG1 in anti-KCNE1 precipitates from equine atrial and ventricular membranes (Fig. 7 B).

As expected, the KCNE1 protein remaining in the supernatants from the anti-HERG1 precipitations (Fig. 7 A) was associated with KCNQ1. When a second immunoprecipitation using anti-KCNE1 antibody was performed on these supernatant fractions, KCNQ1 protein was detected in the KCNE1 immunoprecipitate as well as the resultant supernatant (data not shown).

Anti-HERG1 precipitates were probed with anti-KCNE3 to evaluate association between equine ERG and KCNE3. KCNE3 protein was absent from the anti-HERG1 precipitates but was present in the supernatants (Fig.7 C). These data suggest that KCNE3 and ERG1 do not associate in equine cardiac membranes. We were unable either to confirm the lack of KCNE3/ERG1 coassociation with the reciprocal experiment or to test for association between KCNE3 and KCNQ1, because the anti-KCNE3 and anti-KCNQ1 antibodies available do not immunoprecipitate their respective target proteins.

Effect of IK antagonists on equine cardiac action potential duration.

The slow component of the cardiac I Ks is specifically antagonized by chromanol compounds including 293B (7) and a series of benzodiazepine compounds (45); however, only chromanol 293B has been shown to antagonize I Ks effectively in superfused multicellular preparations (48). Here, chromanol 293B was used to establish a functional correlation between cardiac expression of KCNQ1/KCNE1 and action potential repolarization in the horse heart. Figure 8 A shows that chromanol 293B, at concentrations that specifically antagonizeI Ks, increases equine epicardial action potential duration (APD). Similar results were obtained in two additional experiments. These data suggest thatI Ks is a repolarizing potassium current in equine ventricle. Consistent with the predicted presence ofI Ks, slowly activating, noninactivating outward potassium current could be recorded from horse ventricular myocytes during long (4 s) depolarizing test pulses to positive test potentials (Fig. 8 B) in the presence of low extracellular potassium and E-4031 (Wako Pure Chemicals Industries; Osaka, Japan). Equine cardiacI Ks activated with a half-activation voltage and slope factor of 20.7 ± 7.1 and 12.6 ± 1.0 mV, respectively (n = 3) under these recording conditions (Fig.8 C).

Fig. 8.

Slow delayed rectifier current (I Ks) contributes to ventricular repolarization. A: action potentials recorded from equine ventricular epicardium in the absence (Con) and presence of 10 μM chromanol 293B. Basic cycle length (BCL) = 1 s. B: I Ksrecorded from an equine ventricular myocyte during a +50 mV test pulse applied from a −40 mV holding potential; extracellular potassium concentration = 0 mM. C: conductance-voltage relationship for time-dependent I Ks elicited using 4-s depolarizing pulses applied from a −40 mV holding potential (n = 3); extracellular potassium concentration = 0.1 mM.

The rapid component of I Kr is specifically antagonized by methanesulfonanilide class III antiarrhythmic agents and a variety of noncardiac drugs including cisapride (15, 20, 30,35, 39, 43). Here, MK-499 and cisapride were used to establish a functional correlation between cardiac expression of ERG1 and action potential repolarization in the horse heart. Both the methanesulfonanilide (Fig. 9 A) and cisapride (Fig. 9 B) prolonged equine ventricular APD. The effects of cisapride were concentration dependent and statistically significant at 0.5 and 1.0 μM (Fig. 9 C). Moreover, early afterdepolarizations (EAD) occurred in epicardial muscle preparations paced at a basic cycle length of 2 s and exposed to 1.0 μM cisapride (Fig. 9 D).

Fig. 9.

Antagonists of fast delayed rectifier current (I Kr) prolong action potential duration (APD) and induce early afterdepolarizations in equine ventricular muscle. Epicardial action potentials recorded in the absence and presence of MK-499 (5 μM; A) or cisapride (1 μM; B); BCL = 1 s. C: effects of cisapride on APD were concentration dependent and statistically significant at 0.5 and 1 μM (P < 0.05; n = 6). Solid symbols, individual preparations; □, means. D: early afterdepolarizations occurred in epicardial muscle preparations paced at a BCL of 2,000 ms and exposed to 1 μM cisapride for 15 min.


Horses exhibit resting heart rates of 26–50 beats/min but have the capacity to rapidly increase heart rate up to 204–241 beats/min during exercise (16). The wide range of heart rates is accompanied by significant changes in the QT interval, which indicates marked rate-related adaptation of repolarization (21). In addition, a number of different cardiac arrhythmias have been documented in horses, including ventricular tachycardias associated with electrolyte disturbances and administration of drugs that prolong the QT interval (26, 31,40). On this basis, we hypothesized thatI Kr and I Ks would contribute to action potential repolarization in the horse heart, and that the proteins that form the associated potassium channels would be expressed in equine myocardium. Our data, discussed here in detail, support the initial hypothesis.

ERG1 in equine heart.

ERG1 mRNA expression is abundant in the cardiac tissues of all species studied, including rabbits, ferrets, guinea pigs, dogs, humans, and rats (6, 34, 55), and expression of ERG protein has been verified in rats, mice, ferrets, and humans (6, 38). Species-dependent variation has been documented in the apparent molecular mass, extent of post-translational processing, and regional distribution of the ERG1 protein expressed in cardiac tissue (6,38). In human and rodent hearts, the molecular masses of the ERG1 proteins detected exceed the predicted molecular mass of full-length HERG1 (127 kDa) (38). Asparagine (N)-linked glycosylation of human ERG1 is required for surface-membrane expression in heterologous expression systems, and defective trafficking of mutant HERG proteins has been identified as a major mechanism for inherited LQTS (37, 58, 59). Species-dependent differences in I Kr current density and gating kinetics have also been described (52). Furthermore, species has been identified as a key determinant of the potential for drug-induced prolongation of APD and EAD in cardiac tissue (20, 29).

In the present investigation, we characterized ERG1 expression in the horse heart. ERG1 proteins with apparent molecular masses of 100 and 145 kDa were identified in equine atrium, ventricular endocardium, and ventricular epicardium. The 145-kDa band in horse heart is comparable to the 165-kDa ERG1 protein detected in rodent heart, the 145-kDa ERG1 protein detected in human heart, and the 155-kDa HERG1 protein overexpressed by the stably transfected HEK-293 cells used as a positive control (38, 59). Like that of the 145-kDa equine protein, the apparent molecular masses of 165-kDa rodent ERG1 and 155-kDa heterologously expressed HERG1 are shifted to ∼130 kDa by treatment with PNGase F (37, 38, 59). ThusN-linked glycosylation of ERG1 in horse heart is apparently less extensive than in rodent and is similar to human heart.

The protein detected at 100 kDa in equine cardiac membranes but absent from HERG1-transfected cells could represent either a truncated isoform of ERG1 or a cleavage product of 145-kDa ERG1. NH2-terminal alternatively spliced variants of ERG1 with similar molecular mass have been cloned from mouse (MERG1b) and human (HERG1b) hearts (25,28). When expressed heterologously, these NH2-terminal truncated forms of ERG1 can either form homomeric channels or coassemble with the full-length ERG1 proteins (25, 28). The functional significance of these abbreviated ERG1 forms in native tissue is uncertain, because immunoblotting failed to demonstrate expression of either MERG1b or HERG1b in cardiac membranes (38). If the 100-kDa protein detected in equine heart corresponds to an NH2-terminal truncated form of ERG1, its lack of PNGase F sensitivity is unexpected; the previously cloned NH2-terminal splice variants contain bothN-glycosylation sites present in the full-length ERG1 protein (25, 28). The 100-kDa protein detected in horse heart is unlikely to represent HERGUSO, the described COOH-terminal splice variant of HERG1, because the commercial anti-HERG1 antibody used to detect equine ERG identifies a COOH-terminal epitope not present in HERGUSO.

ERG1 expression has been reported to vary between epicardium and endocardium and between atria and ventricles in a species-dependent fashion (6, 38). Although the present investigation was not specifically designed to address this point, it is clear from our results that ERG1 expression does not vary dramatically between atrium, ventricular epicardium, and ventricular endocardium in any consistent fashion. The subcellular distribution of ERG1 in equine myocytes was similar to that of rat and ferret, in that the protein was distributed uniformly along the T tubular membranes rather than concentrated in the regions of the intercalated disks (6, 38).

KCNQ1 in horse heart.

Expression of the mRNA encoding KCNQ1 gene products has been studied extensively. KCNQ1 mRNA is abundant in many epithelial tissues as well as in cardiac muscle (8, 13, 42, 53, 57). In heart as well as other tissues, both the full-length (isoform 1) and NH2-terminally truncated (isoform 2) KCNQ1 isoforms are expressed (13, 14, 22, 24, 57). KCNQ1 isoform 2 exerts a strong dominant negative effect on the current associated with heterologous expression of isoform 1, and this effect is modulated by the presence of the KCNE1 accessory subunit (14, 22). On this basis, it has been hypothesized that both of the isoforms contribute to cardiac I Ks channels, and that the amplitude of cardiac I Ks in a particular species or region of the heart is dependent on the relative abundance of isoforms 1 and 2 and KCNE1 (14, 22). Both KCNQ1 isoforms contain a potential N-linked glycosylation site in the S5-P segment (4), and N-linked glycosylation of KCNQ1 isoform 1 has been shown to influence current amplitude in a heterologous expression system (19). Surprisingly, despite the demonstrated functional significance, neither the relative expression levels nor the distribution of alternatively spliced and/or differentially processed KCNQ1 proteins has been examined directly in mammalian myocardium.

Two KCNQ1 proteins with apparent molecular masses of 75 and 60 kDa were detected in horse heart using an antibody generated against a COOH-terminal sequence common to both KCNQ1 isoforms. These relative mobilities are not only consistent with the predicted molecular masses of KCNQ1 isoforms 1 (68–75 kDa) and 2 (61–62 kDa) but are also similar to those of the two protein bands detected when either isoform was expressed alone in COS-7 cells (14). Although the 60-kDa protein was unaffected, the molecular mass of the 75-kDa protein shifted to ∼65 kDa by PNGAse F treatment, which indicates that the latter form of KCNQ1 is N-linked glycosylated in equine heart. The failure of the upper band to converge completely with the lower 60-kDa band could reflect either the expression of two KCNQ1 isoforms or the additional post-translational modification of a single isoform by a process other than N-glycosylation. The predicted sequences of KCNQ1 isoforms 1 and 2 containN-myristolation sites as well as phosphorylation sites for a variety of protein kinases including protein kinases A, C, CaMII, and CKII (2, 35a). We were able to detect message for KCNQ1 isoform 1 but not KCNQ1 isoform 2 in horse heart. Additional experiments are required to confirm the absence of the latter splice variant.

KCNE proteins in horse heart.

The KCNE family β-subunits are small, single transmembrane domain proteins that form stable complexes with potassium channel α-subunits and thereby influence the cell surface expression, gating properties, and pharmacology of the resultant channels (41). In heterologous expression systems, KCNE1 (minK or IsK), KCNE2 (MiRP1), and KCNE3 (MiRP2) can combine with either HERG1 or KCNQ1 (1, 4,9, 10, 33, 42, 44, 46, 50, 51). The expression of either protein or message for KCNE1, KCNE2, and KCNE3 is well documented in cardiac tissue from species other than the horse (1, 17, 18, 44); however, all possible subunit combinations are not believed to be functionally significant in heart.

It is well accepted that KCNQ1 and KCNE1 associate in native cardiac tissue to form the channels that conduct I Ks, because 1) both proteins are expressed in cardiac tissue,2) coexpression of these subunits is required to recapitulate cardiac I Ks, and 3) mutations in both KCNQ1 and KCNE1 have been associated with dysfunctional I Ks channels and enhanced arrhythmia susceptibility (4, 42, 47, 54, 57). Analogous criteria have been satisfied for HERG and KCNE2, and it is generally accepted that these two subunits coassemble to form the channels that conduct cardiac I Kr (1, 9-11, 46,50). Similar reasoning has also been used to hypothesize that KCNE3 interacts with KCNQ1 to form potassium channels in colonic epithelia, but does not associate with either HERG1 or KCNQ1 in heart (44).

Whether HERG1 also interacts with KCNE1 in heart is less certain. Both HERG1 and KCNE1 are expressed in cardiac tissue (11, 17, 18,38). Furthermore, it has been established that HERG1 and KCNE1 can coassemble in heterologous expression systems, where KCNE1 has been shown to modulate the expression, gating, and pharmacology of theI Kr current (10, 33, 36). Indirect evidence from antisense experiments conducted using cardiac myocytes and atrial tumor cells strongly suggests that KCNE1 associates with HERG1 as well as KCNQ1 (36, 56), but to date there has been no biochemical evidence for such an association in native tissue (52).

Here, we confirm expression of KCNE1, KCNE2, and KCNE3 in equine heart. We also demonstrate that KCNE3 does not associate with ERG1, but KCNE1 associates with both KCNQ1 and ERG1. We were unable to investigate coassociation between the α-subunits and KCNE2 because of the lack of a specific antibody to KCNE2.

The demonstrated coassociation between KCNE1 and KCNQ1 and the lack of KCNE3 coassociation with ERG1 were expected for the reasons we have discussed. To our knowledge, this is the first report to demonstrate coassociation of ERG1 and KCNE1 in native cardiac membranes. Our coimmunoprecipitation data indicate that ERG1 interacts with both the glycosylated and unglycosylated forms of KCNE1, which is consistent with the proposal that KCNE β-subunits form stable complexes with α-subunits early in protein processing (33). These data are inconsistent with the hypothesis that preferential association of KCNE2 with ERG1 prohibits association of KCNE1 and ERG1 in cardiac myocytes (1).

IK antagonists and repolarization.

At concentrations that specifically antagonizeI Ks and I Kr, chromanol 293B and methanesulfonanilide MK-499, respectively, prolonged APD in equine ventricular epicardium. These data suggest that both the rapid and slow components of the cardiac delayed rectifier contribute to action potential repolarization in the horse heart and provide a functional correlate to the documented expression of ERG1, KCNQ1, and KCNE1 channel proteins in equine cardiac membranes. The presence ofI Ks was confirmed using patch-clamp analysis.

In humans and large mammals other than horses, whereI Kr is important for limiting cardiac APD, a large number of cardiac and noncardiac drugs have been associated not only with induction of early EAD in isolated tissue but also with prolongation of the QT interval and induction of potentially lethal arrhythmias in the whole animal (20). Our data indicate that I Kr antagonists may produce similar effects in the horse. Cisapride, a prokinetic agent with an established propensity to block I Kr and induce LQTS (15, 35, 39), prolonged equine APD and induced EAD in equine ventricular muscle. The effects of cisapride were concentration dependent and occurred at concentrations similar to those measured in horses given therapeutic doses of the drug (12). These data have significant clinical implications for veterinary practice, where horses are frequently treated with drugs labeled for human use. Drugs known to block I Kr in other species should be used with caution in horses to minimize the risk of QT prolongation.

In conclusion, the potassium channel α-subunits ERG1 and KCNQ1 along with the β-subunits KCNE1, KCNE2, and KCNE3 are expressed in equine heart, and the apparent molecular masses and posttranslational modifications are similar to human heart. In equine cardiac membranes, the KCNE1 accessory subunit associates not only with KCNQ1, the α-subunit of I Ks channels, but also with ERG1, the α-subunit of I Kr channels. Both of these outward potassium currents contribute to repolarization of the equine cardiac action potential, and antagonism of either one prolongs APD. Together, these data indicate that the molecular basis for cardiac repolarization in horses resembles that in humans. Thus horses should be considered susceptible to drug-induced arrhythmias and acquired LQTS.


The authors thank Dr. Thomas Divers of Cornell University for obtaining horse hearts. Dr. Randy Wymore and the members of Dr. Wymore's laboratory assisted with PCR optimization.


  • These studies were supported by grants from the Kansas State University College of Veterinary Medicine Dean's Fund (to L. C. Freeman and J. Lillich) and the Harry M. Zweig Memorial Fund for Equine Research (to R. F. Gilmour, L. C. Freeman, and J. Lillich) and Grant HD-36002 from the National Institutes of Heath (to L. C. Freeman).

  • Address for reprint requests and other correspondence: L. C. Freeman, Dept. of Anatomy and Physiology, College of Veterinary Medicine, Kansas State Univ., 228 Coles Hall, Manhattan, KS 66506-5802 (E-mail: Freeman{at}

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

  • First published March 14, 2002;10.1152/ajpheart.00622.2001


  1. 1.
  2. 2.
  3. 4.
  4. 5.
  5. 6.
  6. 7.
  7. 8.
  8. 9.
  9. 10.
  10. 11.
  11. 12.
  12. 13.
  13. 14.
  14. 15.
  15. 16.
  16. 17.
  17. 18.
  18. 19.
  19. 20.
  20. 21.
  21. 22.
  22. 23.
  23. 24.
  24. 25.
  25. 26.
  26. 27.
  27. 28.
  28. 29.
  29. 30.
  30. 31.
  31. 32.
  32. 33.
  33. 34.
  34. 35.
  35. 35a.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
View Abstract