Phylogeny and effects of anoxia on hyperpolarization-activated cyclic nucleotide-gated channel gene expression in the heart of a primitive chordate, the Pacific hagfish (Eptatretus stoutii)

The regular vertebrate heartbeat requires no external input and is initiated by specialized pacemaker cells located in the sinoatrial node (DiFrancesco, 1993). These pacemaker cells have a Ca²⁺-mediated upstroke to the action potential rather than the Na⁺-mediated one that characterizes contractile cardiac myocytes. In addition, pacemaker cells have a slowly depolarizing membrane potential, known as the pacemaker potential, rather than the stable membrane potential present between action potentials characteristic of the working myocardium (Brown and DiFrancesco, 1980; Doerr et al., 1989; Baker et al., 1997; DiFrancesco, 2010). In mammals, the pacemaker potential is a result of inward cation currents. These comprise a mixed K⁺ and Na⁺ current (I_f) flowing through hyperpolarization-activated cyclic nucleotide-gated (HCN) protein channels in the cell membrane, and an inward sodium/calcium exchange current (I_NCX) activated by rhythmic cycling of Ca²⁺ through the sarcoplasmic reticulum (DiFrancesco and Ojeda, 1980; Yanagihara and Irisawa, 1980; Dobrzynski et al., 2007; Maltsev and Lakatta, 2008; Monfredi et al., 2013). The magnitude of these inward currents directly affects the slope of the pacemaker potential, which, in turn, sets heart rate by determining the time taken to reach the threshold voltage needed to open the Ca²⁺ channel (DiFrancesco and Tortora, 1991; Wainger et al., 2001).

The ancestral hagfishes are believed to have diverged from the vertebrate lineage over 500 million years ago and prior to the genome duplications postulated for the vertebrate lineage (Ota and Kuratani, 2007). Despite their evolutionary importance, knowledge on cardiac pacemaker mechanisms in hagfishes is very limited (Farrell, 2007; Wilson and Farrell, 2013). The hagfish heart is unique by being the only vertebrate-type, myogenic heart that is anatomically aneural. However, despite aneural control, heart rate can vary fourfold during recovery from prolonged anoxia (Greene, 1902; Augustinsson et al., 1956; Jensen, 1961; Hirsch et al., 1964; Jensen, 1965; Farrell, 2007; Cox et al., 2010). Recent work suggests that the sarcoplasmic reticulum plays a very minor, if any, role in hagfish cardiac pacemaking, whereas the central importance of HCN channels in setting the intrinsic heart rate is implied by the finding that the HCN channel antagonist zatebradine almost stops the heartbeat (Wilson and Farrell, 2013).

HCN channels are closely related to the ether-à-go-go and cyclic nucleotide-gated K⁺ channels. It is thought that HCN channels consist of four protein subunits, each with six transmembrane segments, a voltage sensor, an ion-conducting pore and an evolutionarily conserved cyclic nucleotide-binding domain (CNBD) (Doyle et al., 1998; Ishii et al., 1999; Jiang et al., 2003). CNBD inhibition of I_f is inversely related to the concentration of cAMP (Wang et al., 2001; Biel et al., 2002). Neural and humoral stimulation can modulate adenylate cyclase activity, changing cytosolic cAMP concentration, which then alters the inhibition imposed by the CNBD (Brown et al., 1979a; Brown et al., 1979b; Brown and DiFrancesco, 1980; DiFrancesco, 1985; DiFrancesco, 2010). Furthermore, in mammals, the relative expression of the four HCN isoforms (HCNs 1–4), each of which has different electrophysiological properties, can vary among tissues. For example, expression of HCN4 is dominant in adult mammalian atria, whereas HCN2 is dominant in the ventricle (Moosmang et al., 2001; Qu et al., 2002; Shi et al., 1999).

The genes encoding the four mammalian HCN isoforms are highly conserved among other vertebrate hearts. A large-scale phylogenetic study (Jackson et al., 2007) suggested that the four HCN isoforms arose from duplications of a single HCN ancestral gene most similar to HCN3, with some teleost fishes having extra isoforms that likely arose from lineage-specific gene or genome duplications. Outside the vertebrates, in Ciona, lineage-specific duplications of the ancestral gene have resulted in three Ciona-specific HCN isoforms (Jackson et al., 2007). Surprisingly, nothing is known about the quantity or regulation of HCN in hagfishes, which occupy the basal position in vertebrate evolution and possess a vertebrate-type, myogenic heart with valved chambers, unlike Ciona. Here, we postulate that modulation of HCN channels is central to the control of hagfish heart rate, in that the decrease in heart rate seen by Cox et al. (Cox et al., 2010) in anoxia results from a decrease in HCN channel expression, and the following increase in heart rate upon reoxygenation is coupled with an increase in HCN channel expression.

To test this hypothesis, we partially cloned hagfish cardiac HCN isoforms from atria and ventricles and compared them with known vertebrate isoforms to establish where the hagfish fits within the chordate phylogeny. In addition, real-time reverse transcriptase PCR (RT-PCR) was utilized to examine the mRNA expression of HCN isoforms in the hagfish heart before, during and after an anoxic challenge to test for plasticity in the expression of HCN. Both cardiac chambers were examined because pacemaker activity has been recorded from multiple areas of the atrium and ventricle in hagfishes, and both chambers continue beating following an atrioventricular ligature, albeit with the atrium outpacing the ventricle (Jensen, 1965; Wilson and Farrell, 2013).

All experiments were carried out in accordance with animal care policies of the University of British Columbia, the Bamfield Marine Sciences Centre, and the Department of Fisheries and Oceans Canada. Pacific hagfish [Eptatretus stoutii (Lockington 1878)] were caught near Bamfield Marine Sciences Centre, Bamfield, BC, Canada, using baited traps. Hagfish (100±1 g; mean ± s.e.m.) were transported to the Department of Fisheries and Oceans Canada, West Vancouver Laboratory, BC, Canada, where they were housed in a 4000 l fibreglass tank supplied with flow-through seawater (10±1°C) and fed squid weekly. Hagfish were fasted a minimum of 48 h prior to any experimental treatment.

Individual hagfish were placed in darkened 2.5 l chambers that were continuously flushed with aerated seawater (10±1°C) at a flow rate of 0.5 l min⁻¹ that was maintained throughout the experiments. Hagfish were left undisturbed for 24 h to habituate to the chamber, during which time they readopted their typical curled posture, as observed in the holding aquarium (see Cox et al., 2010). Anoxia was achieved in the same manner as in Cox et al. (Cox et al., 2010) by supplying the chambers with N₂-saturated seawater from two gas-exchange columns placed in series. The transition period to anoxia lasted less than 1 h (water oxygen concentration was measured continuously with an OxyGuard Handy MkIII, Birkerød, Denmark; accuracy ±0.1 mg l⁻¹). During the transition to anoxia, hagfish became slightly agitated before assuming a straightened position, as previously reported (Cox et al., 2010). Hagfish were maintained in flow-through anoxic seawater for 24 h, before quickly (under 30 min) restoring normoxic conditions for up to 36 h. Hagfish were sampled at the end of the 24-h anoxic period (no normoxic recovery, termed 0 h group), 2 h into normoxic recovery and 36 h into normoxic recovery (2 h and 36 h groups, respectively) (N=10 for each group). Control animals (N=10) were sampled after 84 h in the chamber in normoxic water. At each sample time, hagfish were rapidly removed from their chamber and killed by decapitation, and the branchial heart was dissected, freeing the atrium (7.7±1.2 mg) and ventricle (59.2±4.4 mg), prior to freeze-clamping each separately and storing them in liquid N₂. Tissues were stored at −80°C. The precise location of the hagfish sinoatrial node is unknown; therefore, it could not be reliably isolated.

Total RNA was extracted from tissues using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Extractions followed a modified protocol outlined by Ellefsen et al. (Ellefsen et al., 2008). Briefly, each frozen tissue was weighed (atria 17.2±1.2 mg, ventricles 59.3±4.4 mg, means ± s.e.m.) and immersed in 60 μl TRIzol per milligram tissue. Then, 0.2 μl mg⁻¹ tissue of an external RNA control gene (mw2060) was added and the tissue/TRIzol/mw2060 mixture was homogenized for 1 min using a T-25 Basic Homogenizer (IKA Works, Wilmington, NC, USA). The use of an external RNA control has been argued to be the most accurate method for the normalization of real-time RT-PCR data (Huggett et al., 2005). mw2060, a 2060-bp-long mRNA species from the cyanobacterium Microcystis cf. wesenbergi that shows no sequence homology to known vertebrate mRNA species, was recently developed as a more reliable method for the normalization of real-time RT-PCR data from tissues of anoxia-tolerant species than other commonly employed normalization techniques (Ellefsen et al., 2008). Additionally, the methodology allows for inter-tissue comparisons of target gene expression (Stecyk et al., 2012). Samples were then chilled on ice, vortexed, incubated at room temperature for 15 min and vortexed again. The maximum amount of tissue/TRIzol homogenate (up to 1 ml) was transferred to an Eppendorf tube before adding chloroform (20% of the volume of tissue/TRIzol homogenate). The mixture was incubated at room temperature for 3 min, vortexed, centrifuged at 10,000 g for 15 min at 4°C and placed on ice. A volume corresponding to 40% of the tissue/TRIzol homogenate of the upper aqueous phase was then transferred to a new Eppendorf tube before adding an equal amount of ice-chilled isopropanol. The mixture was vortexed, incubated at −20°C overnight, incubated at room temperature for 10 min and centrifuged at 11,500 g for 10 min at 4°C. The supernatant was then discarded and the pellet was washed twice with ice-chilled 75% ethanol. Each ethanol wash was followed by centrifugation at 11,500 g for 10 min at 4°C. After the final ethanol wash, the pellet was air dried and eluted in 30 μl of nuclease-free water (Ambion, Austin, TX, USA). The mixture was then incubated at 65°C for 5 min to ensure the pellet was completely dissolved, and stored at −80°C.

Care was taken to avoid systematic errors introduced by sample processing during RNA extraction. All samples were handled without intermission and in a systematic, random order. Samples were processed in groups, wherein each group consisted of samples of a single tissue type and representing each of the treatment groups. The samples within each group were processed at random. Similar procedures were also employed for cDNA synthesis.

The concentration of total RNA in every sample was determined using a NanoDrop ND-1000 UV-Vis Spectrophotometer (NanoDrop Technologies, Rockland, DE, USA). One microgram of total RNA from each sample was then treated with DNase I (DNA-free; Invitrogen) and subsequently reverse transcribed using Random Hexamers (50 ng μl⁻¹) and Superscript III (both from Invitrogen) in reaction volumes of 20 μl and in accordance with the manufacturer's protocol. cDNA solutions were diluted 1:30 with nuclease-free water and stored at −20°C. Duplicate cDNA syntheses were performed on RNA samples when possible.

The HCN isoforms were partially cloned using PCR primers designed to recognize gene regions conserved among vertebrate species. The regions were located using GeneDoc (version 2.6.0.2, http://www.nrbsc.org/gfx/genedoc/) and ClustalX (version 1.81), while primers were designed using Primer3 (Nicholas et al., 1997; Thompson et al., 1997; Rozen and Skaletsky, 2000). Primers used for cloning are listed in Table 1.

PCR was performed on a mixture of 1:30 diluted cDNA from atrial and ventricular tissues and all treatment groups using Platinum Taq DNA Polymerase (Invitrogen; 94°C for 10 min, 94°C for 30 s, 48°C for 1 min, 72°C for 1 min, repeat steps two to four 44 times, 72°C for 10 min, hold 4°C). Resulting dsDNA fragments were ligated into the pGEM-T Easy Vector System I (Promega, Madison, WI, USA). Ligation reactions were transformed into CaCl₂-competent cells (TOP10 F′; Invitrogen) and positive colonies were checked for inserts of a correct size via agarose gel electrophoresis. PCR products from up to eight colonies were sequenced using T7 primers (ABI-lab, University of Oslo, Oslo, Norway). All procedures were carried out according to the manufacturer's protocol.

The partial sequences of cardiac HCN for hagfish were compared with known vertebrate HCN genes using ClustalX and NCBI BLAST (Altschul et al., 1990; Thompson et al., 1997). A phylogeny was created by running a neighbour-joining phylogenetic analysis with the translated amino acid products from the hagfish HCN isoforms, plus those of vertebrate species downloaded from Ensembl and GenBank. The closely related human ether-à-go-go-related gene (hERG) and Drosophila cyclic-nucleotide-gated gene (dCNG) were added to the sequence alignment as outgroups. A 37 amino acid segment was used because all partially cloned isoform protein segments overlapped in this area. One thousand bootstrap replications were conducted on the amino acid sequences and a consensus tree was produced. The phylogeny was completed using the Seqboot, Protdist, Neighbor, Consense and Drawgram programs of PHYLIP 3.65 (http://evolution.genetics.washington.edu/phylip.html), and viewed using TreeView 1.6.6 (http://taxonomy.zoology.gla.ac.uk/rod/treeview.html) and formatted using Adobe Illustrator (Page, 1996).

All real-time RT-PCR primer pairs were designed using Primer3. The small differences between the paralogs of HCN2 and HCN3b prevented distinctions between paralogs using the tested primers; therefore, both paralogs of HCN3b and HCN2 are compared collectively as HCN3b and HCN2, respectively. Forward and reverse primers were targeted to either side of an exon–exon overlap as a further precaution against amplifying genomic DNA (i.e. in addition to the extraction of total RNA and DNase treatment). Amplification of the desired cDNA species by the primer pairs was verified by melting curve analyses (Lightcycler 480 software, Roche Diagnostics, Basel, Switzerland) and cloning and sequencing of all primer pair products (performed as described above for the cloning of HCN genes). Real-time RT-PCR primer sequences are summarized in Table 2. To find primers that worked well with the real-time RT-PCR protocol (primer concentration of 1 mmol l⁻¹ and annealing temperature of 60°C), a minimum of four primer pairs were tested for each gene. As an alternative to primer concentration/annealing temperature optimization, primer pairs that displayed distinct melting curves, the highest efficiency (calculated as described below for RT-PCR analyses) and the lowest C_q (quantification cycle) values were chosen. All procedures followed manufacturer's protocols.

Real-time RT-PCR for HCN isoforms 2, 3a, 3b and 3c were performed using Lightcycler 480. All real-time RT-PCR reactions were performed with a reaction volume of 10 μl that contained 5 μl of Lightcycler 480 SYBR Green I Master (Roche Diagnostics), 3 μl of 1:15 diluted cDNA as the template, 1 μl of 5 mmol l⁻¹ gene-specific forward primer and 1 μl of 5 mmol l⁻¹ gene-specific reverse primer (i.e. a final primer concentration of 1 mmol l⁻¹). The real-time RT-PCR program was: 95°C for 10 min, 95°C for 10 s and 60°C for 10 s, 72°C for 13 s, repeating steps two to four 42 times. Two real-time RT-PCR reactions were performed on each gene for each cDNA synthesis. The replicates were conducted on different plates and days. As a result, four real-time RT-PCR reactions were performed on each gene for each RNA sample.

For HCN4, real-time RT-PCR was conducted using an ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Each real-time RT-PCR reaction contained 10 μl SYBR Green Master Mix (Applied Biosystems), 4 μl of 1:15 diluted cDNA as the template, 0.8 μl of 5 mmol l⁻¹ gene-specific forward primer and 0.8 μl of 5 mmol l⁻¹ gene-specific reverse primer and 4.4 μl of double-processed tissue-culture water (Sigma-Aldrich, Ayrshire, UK), bringing the total reaction volume to 20 μl. The real-time RT-PCR program was: 50°C for 2 min, 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 1 min. Following the real-time RT-PCR, a melt curve analysis was performed. As for the other isoforms, two real-time RT-PCR reactions were performed on each cDNA synthesis, resulting in four real-time RT-PCR reactions for each mRNA sample.

For HCN2, HCN3a, HCN3b and HCN3c, C_q values were obtained for each reaction using the Lightcycler 480 software and were defined according to the second derivative maximum (Luu-The et al., 2005). Priming efficiencies were calculated for each real-time RT-PCR reaction using LinRegPCR software (Ruijter et al., 2009). For HCN4, each plate was run with a standard dilution of a mixture of randomly chosen cDNA samples from atrial and ventricular tissues. The primer efficiencies were then determined for each plate from the slope of the standard curve using Applied Biosystems 7000 System Sequence Detection Software 1.2.3. This software was also used to obtain the C_q values for each reaction.

Average priming efficiencies () were used for the final calculations, which were calculated separately for each tissue and primer pair from all real-time RT-PCR reactions (Čikoš et al., 2007). Then, to normalize HCN isoform gene expression to the expression of the external RNA control mw2060, was calculated for every reaction, as well as the ratio (R1) between and (where Tar is the target gene, E is priming efficiency and C_q is the quantification cycle). To compare the expression of each HCN isoform among the different treatments, R1 was referenced to the mean expression of the control, normoxic samples.

Results are expressed as means ± s.e.m. Statistical analyses were performed using SigmaPlot for Windows 11.0 (Systat Software, Chicago, IL, USA). Differences in HCN isoform expression within a tissue were assessed using a one-way ANOVA followed by a Tukey's post hoc test. A two-way ANOVA, followed by a Tukey's post hoc test, was used to determine statistically significant effects of anoxia and recovery on the expression of HCN gene isoforms. Differences in atrial and ventricular HCN isoform expression among each sampling time were assessed using a Student's t-test. Significance was accepted at P<0.05.

A total of six isoforms of HCN mRNA were found in hagfish cardiac tissue, with all six isoforms being represented at varying levels in both atrial and ventricular chambers and one isoform existing with two paralogs, for a total of seven partially cloned genes. The amino acid sequences aligned with mouse HCN3 are shown in Fig. 1 with the lengths of the partially cloned sequences shown in Table 1. Two isoforms of a hagfish representative of vertebrate HCN2, termed HCN2a and HCN2b, differed by five amino acid substitutions (P-T at 16, H-N at 56, P-H at 111, S-P at 253, S-G at 261) and one deletion of a G at location 280 in HCN2a. Three isoforms of HCN3 were found, HCN3a, HCN3b and HCN3c, with HCN3b existing as two paralogs. These two paralogs, termed HCN3bi and HCN3bii, differed by two substitutions (G-E) at amino acid locations 148 and 262. One hagfish representative of HCN4 was also found.

Neighbour-joining phylogenetic analysis (Fig. 2) was conducted on a 37 amino acid sequence from the CNBD of multiple vertebrate and urochordate HCN genes where six of the seven hagfish sequences overlapped, as shown in Fig. 3 (HCN3bi and HCN3bii paralogs are included as HCN3b on the tree because the G-E substitutions of HCN3bi and HCN3bii occurred outside of this segment).

To summarise the phylogenetic tree, HCN3b and HCN3c cluster together and, along with HCN3a, form a cluster with the HCN3 of teleosts. The HCN3 of mammals cluster together on the other side of the outgroup, which contains both the HCN-related human ERG and Drosophila CNG proteins, as well as the Ciona HCN isoforms. The HCN4 isoforms, including the hagfish sequence, all cluster together apart from the pufferfish HCN4a isoforms. The hagfish HCN2 sequences cluster with the mammalian HCN2 and pufferfish HCN2b isoforms, while the other teleost HCN2 sequences cluster in a second group either side of the HCN1 cluster. The splitting of HCN isoforms in the tree may be a consequence of using short sequences for the analysis. Despite the lack of a hagfish HCN1 sequence in the phylogeny, it is noted that the two teleost HCN1 isoforms cluster as a separate branch of the HCN1 cluster.

Of the five measured hagfish HCN isoforms, HCN3a had the highest expression in both the atrium and ventricle in normoxia. HCN3a was expressed over 17 times more than HCN4 in the atrium and 25 times more than HCN4 in the ventricle (Fig. 4). Expression of the other isoforms followed the order of HCN4, HCN3b, HCN2 and HCN3c for both the atrium and the ventricle. Because of the similarity of the two HCN2 and HCN3b mRNA segments, it was not possible to distinguish between them using the real-time RT-PCR primers, and both segments would have been duplicated together by their respective primers.

The effects of 24 h of anoxia and subsequent recovery on HCN isoform gene expression are shown in Fig. 5. In the atrium, the only statistically significant change was an increase in HCN3a at 2 h of normoxic recovery from anoxia, which returned to control expression levels by 36 h of recovery. In contrast, anoxia induced a significant decrease of HCN3a expression in the ventricle, which again progressively recovered back to the control level within the 36 h recovery period. The trends in HCN3b, HCN3c and HCN2 expression during anoxia, while not reaching statistical significance (Fig. 5), did return to the control values with normoxic recovery. Expressing HCN gene expression in the atrium and ventricle as ratio further illustrates the clear differences in the responses of the two chambers to anoxia (Table 3). For example, the ratio of atrial to ventricular HCN3b expression increased from <1 to 7 following 24 h of anoxia before returning back to <1 following 2 h of normoxic recovery.

The present study documents for the first time HCN expression in a representative of the basal chordates and elucidates differential changes in HCN expression in the atrium and ventricle of anoxic hagfish. Specifically, we: (1) described six isoforms of HCN from seven partial HCN amino acid sequences in the cardiac tissues of Pacific hagfish, (2) performed a phylogenetic analysis on six of these sequences, (3) made a quantitative comparison of the mRNA expression of five HCN isoforms for the atrium and ventricle and (4) discovered changes in HCN isoform expression in response to chronic anoxia that were reversed following normoxic recovery.

Six HCN isoforms were identified – HCN2a, HCN2b, HCN3a, HCN3b, HCN3c and HCN4 – a finding that differs from that in the adult mammalian heart, where four isoforms, HCN1–4, are expressed. Contrary to HCN4 dominance in mammalian pacemaker tissue (Shi et al., 1999; Marionneau et al., 2005), HCN3a was the dominantly expressed isoform in the hagfish atrium and ventricle. In addition, hagfish cardiac HCN3b existed as two paralogs, designated here as HCN3bi and HCN3bii. These findings have important implications for the phylogenetic appearance of pacemaker channels in the chordate lineage, as summarised in Fig. 6. Moreover, given what is known of the biophysical properties of the mammalian HCN isoforms, and assuming that mRNA expression reflects protein presence, the findings lend new but preliminary insights into the electrophysiological properties of the hagfish heart and possibly how heart rate is modulated during and following prolonged anoxia.

Based on the neighbour-joining phylogenetic analysis, the vertebrate HCN2, HCN3 and HCN4 isoforms had already appeared in cardiac tissues of the chordate lineage prior to the hagfish/gnathostome divergence. This conclusion is in accordance with previous work that compared HCN of Ciona with that of vertebrates, including more modern teleost species (Jackson et al., 2007). Jackson et al. proposed that HCN3 is ancestral, followed by the appearance of HCN4 and then HCN2 (Jackson et al., 2007). As shown here, all HCN3 isoforms cluster together and therefore appear ancestral. The two isoforms of HCN2 (2a and 2b) cluster together with the mammalian and avian forms of HCN2. Clearly, limitations do exist when using partial genes to create a phylogeny. Potentially, minor mutations of the CNBD during the long period between the appearance of hagfish and vertebrates may make ancestral genes appear more derived, something that will not be revealed with partial cloning and sequencing. However, a more robust phylogeny will require full hagfish HCN sequences. Nevertheless, the CNBD has been highly conserved throughout evolution (Jackson et al., 2007), likely because of its important regulatory function. Indeed, changing a single amino acid in this region is known to appreciably alter the effect of cAMP binding to the CNBD (Tibbs et al., 1998). Also, HCN2 and HCN4 are strongly gated by cAMP, whereas HCN1 is not (Wang et al., 2001). Moreover, because three lineage-specific HCN isoforms have been reported previously in urochordates (Jackson et al., 2007), it was previously hypothesised that, prior to the divergence of the deuterostomes, duplication of the ancestral HCN gene gave the sea urchin two HCN isoforms and, because of a further duplication, three HCN isoforms were found in the urochordates (Gauss et al., 1998; Galindo et al., 2005; Jackson et al., 2007). A further possibility is that the sea urchin also had three HCN isoforms, but subsequently lost one. It has been suggested that HCNa, HCNb and HCNc emerged early in Ciona, before C. intestinalis and C. savignyi had diverged; therefore, these three HCN isoforms may have been present in the urochordate ancestor (Jackson et al., 2007). Because the three hagfish HCN3 isoforms do not align with the Ciona isoforms, it is unlikely that the hagfish HCN3a, HCN3b and HCN3c isoforms represent an ancestral condition prior to the urochordate/chordate divergence, and more likely represent HCN3 duplication events following the hagfish/vertebrate split.

What is apparent from the available information is that lineage-specific HCN gene duplications are relatively regular evolutionary events, being previously identified in the urochordates and some teleost fishes (Gauss et al., 1998; Galindo et al., 2005; Jackson et al., 2007), and now in hagfishes. Having many isoforms of these genes could provide flexibility and fine-tuning of pacemaker activity, perhaps providing an evolutionary advantage, which we speculate on below.

Many primers were used without success to target conserved regions of the vertebrate HCN1 isoform in the hagfish heart. We propose three possible explanations for the absence of HCN1 in the hagfish heart. First, HCN1 is commonly associated with central neural tissue in addition to the heart (Franz et al., 2000; Nolan et al., 2003). Therefore, HCN1 expression in hagfish may be limited to non-cardiac tissues. Another explanation could be related to a very low expression of HCN1 that was below the detection level. Finally, HCN1 may not have appeared in the vertebrate lineage until after the hagfish/vertebrate split. Failure to find HCN1 in hagfish would suggest that duplication of the ancestral HCN2-like gene likely produced HCN1 following the hagfish/vertebrate divergence. Indeed, in another phylogenetic study, HCN1 was suggested as the most recent HCN gene, although the exact timing of its appearance is unknown (Jackson et al., 2007). Data on HCN expression in elasmobranchs and basal teleosts would be particularly informative in this regard. All the same, conclusive knowledge on the exact number and type of HCN isoforms will not be possible until the entire hagfish genome has been sequenced and searched.

In mammalian studies, HCN4 is by far the most dominant HCN isoform in the heart, with expression in the sinoatrial node dwarfing that of other HCN channels (Shi et al., 1999; Marionneau et al., 2005). Contrary to this, we discovered that in hagfish, HCN3a mRNA expression in both the atrium and ventricle dominates HCN4 mRNA expression by at least an order of magnitude (Fig. 4). In this regard, the hagfish heart bears some resemblance to mammalian embryonic stem cells, where, in early stage mouse embryonic stem cells, the dominant HCN isoform is HCN3 (Qu et al., 2008). As mouse development progresses, expression of HCN3 decreases and HCN4 becomes dominant in association with the onset of cardiac contractions (Qu et al., 2008; Schweizer et al., 2009). While we are not prescribing Lamarkian recapitulation of phylogeny, the similarity points to a functional role of variations in cardiac HCN isoform expression. Nevertheless, care must be taken when speculating on the physiological implications of differential HCN isoform expression in the hagfish because the electrophysiological properties of these new isoforms have yet to be examined. Therefore, the electrophysiological ramifications of the present results that are proposed below are based on the rich knowledge of mammalian HCN electrophysiology.

In lieu of electrophysiological evidence on the activation dynamics of the hagfish HCN isoforms, it is not possible to know exactly what the effect of HCN3 dominance over HCN4 might be on hagfish cardiac excitation–contraction coupling. However, Moosmang et al. (Moosmang et al., 2001) showed that HCN3 activates at a similar voltage to HCN4, but has activation kinetics more similar to the faster HCN2 isoform, allowing a larger current to flow at lower membrane potentials. Therefore, a large proportion of HCN3 may allow the hagfish heart to become easily activated and then allow the spread of excitation throughout the myocardium to occur rapidly. Indeed, evidence exists for the presence of pacemaker potentials throughout the hagfish myocardium, which may be a requirement to allow enough current to flow in order to activate contraction in a heart known to have a low diastolic membrane potential (Jensen, 1965).

A higher HCN mRNA expression in the hagfish atrium compared with the ventricle for all isoforms of HCN (apart from the similar expression of HCN3b) was to be expected for synchronous cardiac contraction, as it would cause the atrial pacemaker potential to be shorter than that of the ventricle. Indeed, HCN gene expression per unit mass in rabbit atria and especially the sinoatrial node are much higher than that in the ventricles (Shi et al., 1999; Moosmang et al., 2001; Marionneau et al., 2005). In turn, a larger sarcolemmal ion channel density allows a larger current to pass when the membrane channels are open (Mukherjee et al., 1998). Thus, a higher atrial HCN channel expression leads to a larger I_f and a faster depolarisation to the threshold potential between action potentials, which ensures that atrial contraction precedes ventricular contraction. Although expression of four out of five isoforms was roughly twice as high in the atrium compared with the ventricle, we have no information on sarcolemmal protein expression. However, it seems that relative excitation, and therefore synchrony of contraction, of hagfish cardiac chambers is controlled primarily by HCN3a channel density rather than upregulation and downregulation of specific isoforms in the chambers. Indeed, it is possible that regulation of HCN channel density throughout the myocardium may be a precursor of the cardiac conduction system seen in other vertebrate hearts.

We postulated, and our data support the idea, that alterations in HCN channel expression could play a role in controlling heart rate during prolonged anoxia and subsequent recovery. Consistent with the 50% decrease in heart rate during anoxia, an increased heart rate during the initial phase of recovery and the return of heart rate to the control normoxic rate after 36 h of recovery (Cox et al., 2010), we discovered significant changes in HCN mRNA expression with anoxia and recovery. Ventricular HCN3a expression decreased significantly, with the HCN3a:HCN4 expression ratio decreasing by over threefold, from 25:1 to 7:1 during anoxia and returning to 20:1 during normoxic recovery (Table 4). In contrast, the HCN3a:HCN4 expression ratio increased almost twofold in the atrium, from 17:1 to 26:1 during anoxia, and returned to 12:1 during normoxic recovery (Table 4). In mammals, HCN4 allows a smaller current to flow compared with HCN3; therefore, these findings may represent a decrease in hagfish ventricular excitability during prolonged anoxia. The immediate response to restoration of normoxic conditions for the hagfish heart is a quadrupling of the anoxia-induced heart rate after 2 h (Cox et al., 2010), which correlates with the increase in atrial HCN3a and a restored level of ventricular HCN3a. An increase in HCN3a would increase I_f, leading to an increased pacemaker potential, and therefore an increase in heart rate because the action potential upstroke threshold would be reached earlier. HCN3a expression returned to control levels by the end of the 36 h recovery period as did the ratio of HCN3a:HCN4 in both cardiac chambers. Therefore, it is possible that regulation of HCN3a may play a role, amongst others, in controlling heart rate during prolonged anoxia and subsequent recovery; however, these hypotheses need to be tested using protein expression studies.

The above discussion considered the HCN channel as the primary vertebrate cardiac pacemaker, yet a recent alternative hypothesis suggests that pacemaker activity involves a calcium clock (Rubenstein and Lipsius, 1989; Li et al., 1997; Hüser et al., 2000; Ju and Allen, 2000; Dobrzynski et al., 2007; Maltsev and Lakatta, 2008; Monfredi et al., 2013). In brief, this hypothesis proposes that spontaneous Ca²⁺ sparks released from the sarcoplasmic reticulum via ryanodine receptors interact with local sodium-calcium exchanger proteins to cause an overall inward depolarizing current. The Ca²⁺ is then pumped back into the sarcoplasmic reticulum, thus resetting the clock. The most recent synthesis of experimental evidence in support of the calcium clock hypothesis includes HCN as a part, albeit a minor one, of the inter-related cycling of ions in pacemaker tissue (Monfredi et al., 2013). That said, this model does recognize that future work is still needed to completely explain how HCN blockers such as ZD7288 exert their effect within the calcium clock paradigm. Moreover, future calcium clock studies will have to resolve why another HCN blocker, zatebradine, which slows pacemaker activity in rainbow trout hearts (Gamperl et al., 2011), almost completely abolished pacemaker activity in both atrial and ventricular chambers isolated from hagfish (Wilson and Farrell, 2013), and how recent work in mice has shown that an increase in heart rate during pregnancy is coupled with an increase in HCN2 expression (El Khoury et al., 2013).

In summary, we have provided novel insights into cardiac pacemaker expression and their evolution using real-time RT-PCR of partially cloned HCN mRNA expression in hagfish hearts under normoxic and anoxic conditions. Of the six isoforms, HCN3a dominated in atrium and ventricle. The presence of HCN2, HCN3 and HCN4 in the hagfish heart suggests their presence prior to the hagfish/vertebrate divergence and an importance in cardiac pacemaker activity at the time of the emergence of the chambered myogenic heart. HCNa and HCNb appeared as either hagfish-specific duplications or remnants of the forms previously identified in ancestral deuterostomes and urochodates. Our postulate that altered HCN expression is involved in the control of heart rate in the aneural hagfish heart was supported by several observations, including a decrease in the ventricular HCN3 expression during anoxia and an increase in atrial HCN3 expression shortly after return to normoxia. Functional studies on the activation and cAMP gating of each isoform, and their respective roles in cardiac pacing, will require electrophysiological study, which, when coupled with HCN protein expression studies, will provide even further insight into regulation of heart rate in the aneural hagfish heart.

Gratitude is given to Dr P. Schulte and T. Healy at the University of British Columbia for expertise and lab space during HCN4 measurements.

 
• CNBD

  cyclic nucleotide-binding domain

 
• C_q

  quantification cycle

 
• dCNG

  Drosophila cyclic-nucleotide-gated gene

 
• E

  priming efficiency

 
• 

  mean priming efficiency

 
• HCN

  hyperpolarization-activated, cyclic nucleotide-gated channel

 
• hERG

  human ether-à-go-go-related gene

 
• I_f

  mixed K⁺ and Na²⁺ (funny) current

 
• I_NCX

  sodium/calcium exchange current

 
• RT-PCR

  reverse transcriptase polymerase chain reaction

 
• Tar

  target gene