HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Supplemental Figures All Versions of this Article: 29/3/231    most recent 00142.2006v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Jackson, H. A. Articles by Accili, E. A. Search for Related Content PubMed PubMed Citation Articles by Jackson, H. A. Articles by Accili, E. A.

Call For Papers: Comparative Genomics

Evolution and structural diversification of hyperpolarization-activated cyclic nucleotide-gated channel genes

¹ Department of Cellular and Physiological Sciences, Faculty of Medicine, University of British Columbia, Vancouver
² Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, British Columbia, Canada

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 
Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels are members of the voltage-gated channel superfamily and play a critical role in cellular pace-making. Overall sequence conservation is high throughout the family, and channel functions are similar but not identical. Phylogenetic analyses are imperative to understand how these genes have evolved and to make informed comparisons of HCN structure and function. These have been previously limited, however, by the small number of available sequences, from a minimal number of species unevenly distributed over evolutionary time. We have now identified and annotated 31 novel genes from invertebrates, urochordates, fish, amphibians, birds, and mammals. With increased sequence numbers and a broader species representation, a more precise sequence comparison was performed and an evolutionary history for these genes was constructed. Our data confirm the existence of at least four vertebrate paralogs and suggest that these arose via three duplication and diversification events from a single ancestral gene. Additional lineage-specific duplications appear to have occurred in urochordate and fish genomes. Based on exon boundary conservation and phylogenetic analyses, we hypothesize that mammalian gene structure was established, and duplication events occurred, after the divergence of urochordates and before the divergence of fish from the tetrapod lineage. In addition, we identified highly conserved sequence regions that are likely important for general HCN functions, as well as regions with differences conserved among each of the individual paralogs. The latter may underlie more subtle isoform-specific properties that are otherwise masked by the high identity among mammalian orthologs and/or inaccurate alignments between paralogs.

pacemaker channel; phylogeny; sequence analysis; molecular evolution

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 
HYPERPOLARIZATION-ACTIVATED CURRENTS, commonly referred to as I_h or I_f, have been identified in species distributed across the metazoan lineage and contribute to membrane potential in excitable and nonexcitable cells. In the mammalian heart and nervous system they contribute to spontaneous beating in pacemaker cells (11, 52). In nonpacing cells, I_h facilitates the regulation of resting membrane properties by limiting the fluctuation of membrane potentials (52). The proteins responsible for I_h have been cloned from both invertebrate and vertebrate species and are referred to as hyperpolarization-activated cyclic nucleotide-gated (HCN) channels (18, 26, 35, 57). When expressed in heterologous systems, HCN subunits form cation-selective channels that are directly modulated by cyclic nucleotides and exhibit biophysical properties that are similar to those of I_h observed in native tissue.

HCN channels are members of the voltage-gated cation channel superfamily and, based on sequence homology, are most closely related to the cyclic nucleotide-gated (CNG) channel and ether-a-go-go (EAG) potassium channel families. Individual subunits (Fig. 1) are predicted to have six transmembrane segments (S1–S6), with a voltage sensor domain in the S4 segment and an ion conducting pore between S5 and S6. Based on the crystal structures of related potassium channels (14, 27, 32) it is proposed that four HCN subunits come together to form a tetramer around a central pore. The distal termini of each subunit are cytoplasmic and from the crystal structure of the COOH terminus (78), it is now known that an -helical linker region joins the transmembrane region to an evolutionarily conserved cyclic nucleotide-binding domain (CNBD).

View larger version (33K): [in this window] [in a new window]   Fig. 1. Schematic representation of the overall topology of hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. Included are the 6 putative transmembrane segments (S1–S6), including an S4 voltage sensor (light gray) and the reentrant P-loop region between S5 and S6, all structural characteristics common to the entire voltage-gated potassium channel superfamily, including ERG, CNG, and HCN channels. HCN and ERG channels also share the potassium selectivity filter GY/FG. Based on the crystal structure of mHCN2 (78), the C-linker region in HCN channels contains 6 alpha-helices (A'–E') and connects the cytoplasmic end of S6 to the start of the cyclic nucleotide binding domain (CNBD). The CNBD consists of 2 alpha-helices (A and B) that flank an 8-stranded beta-roll followed by a 3rd alpha-helix at its COOH-terminal end (C-helix). The cyclic nucleotides bind to a pocket that is formed by the beta-roll and the C-helix. Thick black lines delineate the region that is included in the phylogenetic analyses. In HCN channels, these lines encompass vertebrate exons 2 through 7. Alignments of HCN vertebrate isoform-specific NH2-terminal (exon 1) and COOH-terminal (exon 8) regions include those residues that fall outside of this region.

In addition to the four mammalian isoforms, only a few HCN genes have been cloned from lower vertebrate and invertebrate species. A single HCN1 ortholog has been cloned from the rainbow trout (9), whereas one and two HCN homologs have been cloned from arthropods (20, 21, 30, 43) and the sea urchin (16, 18), respectively. These sequences demonstrate considerable identity with the mammalian homologs, as well as some intriguing differences that provide preliminary clues about the evolutionary relationships among HCN genes. But due to the lack of sequence representation from a diverse sampling of species distributed throughout evolutionary history, previous phylogenetic analyses of the HCN family and its relatives have been limited and have yielded inconsistent patterns of evolution (10, 16). Furthermore, the current sampling of the four vertebrate isoforms is limited primarily to closely related mammalian sequences. The high residue conservation among these sequences, combined with the lack of sequences from more distantly related vertebrates, renders comparative analyses of the orthologs from the individual vertebrate isoforms ineffective.

In this study, we report the first thorough phylogenetic analysis and sequence comparison of the HCN gene family. By performing an extensive search of the currently available protein and whole-genome databases we derived a comprehensive list of known and putative full-length sequences for HCN homologs from a wide variety of species including urochordates and lower vertebrates. The increased number of sequences and broadened species representation provide information about HCN gene structure at critical periods during its evolutionary history. We identified sequences that are conserved and likely important for general HCN functions, as well as regions that may underlie more subtle differences in function among the different isoforms. Exon structure and genomic organization were also determined. These analyses provide insight into the molecular evolution of this protein within different taxa and support the hypothesis that both lineage-specific and ancestral duplication and divergence events of the HCN genes have occurred throughout its history.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 
Sequence data.
Currently available HCN protein sequences were identified through BLASTP (1) searches of the National Center for Biotechnology Information (NCBI) and UniProt/SwissProt nonredundant protein databases. Human HCN2 (GenBank GI no. 21359848), trout HCN1 (GenBank GI no. 33312350) or drosophila HCN (GenBank GI no. 5326833) amino acid sequences were used as queries and default parameters were applied. To prevent the inclusion of incorrect sequences generated by computer prediction programs, all computer-derived annotations were temporarily removed until confirmed by sequence alignment or database annotation described below. Splice variants, short fragments, and other duplicates were also removed.

The Ensembl Genome Browser (http://www.ensembl.org/) (6, 24) was used to determine the genomic position and distribution of the established HCN genes and either to examine new HCN genes identified by the computer gene-prediction programs and annotation process or to identify novel genes. Sequences classified as HCN genes by Ensembl were examined, and those that spanned the entire length of known sequences were downloaded. Protein annotations that resembled HCN but either lacked regions of the predicted sequence or showed signs of additional exons or exon fragments were not included. However, the genomic DNA underlying these protein predictions were used as further reference to help in the manual reannotation of the protein sequence. TBLASTN (1) searches of the available genome databases using either the low-sensitivity default parameters (optimized for near-exact matches) or medium-sensitivity default parameters (optimized to allow for local mismatch) were conducted to identify any further genes or genomic regions that showed significant sequence identity to the peptide query sequence used. In total, 13 genomes were analyzed, including: zebrafish (Danio rerio) (v. 35.5b), Japanese pufferfish (Fugu rubripes) (v. 29.2e), green pufferfish (Tetraodon) (v. 31.1c), opossum (Monodelphis domestica) (v35.2), dog (Canis familiaris) (v. 35.1d), cow (Bos taurus) (v.36.2), chimpanzee (Pan troglodytes) (v.31.2a), clawed frog (Xenopus tropicalis) (v. 31.1a), chicken (Gallus gallus) (v.35.1k), sea squirt (Ciona intestinalis) (v. 35.195b), pacific sea squirt (Ciona savignyi) (CSAV2.0), mosquito (Anopheles gambiae) (v. 23.2b.1), and worm (Caenohabditis elegans) (v 29.130). Human HCN2, trout HCN1, and drosophila HCN sequences were used as query sequences for vertebrate, fish, and insect genomes, respectively. For the urochordate genomes, sea urchin HCN (GenBank GI no. 74136757), trout HCN1, and human HCN2 sequences were used. In general, similar TBLASTN results were found regardless of the query sequence used, reflecting the high degree of sequence identity found within the core region throughout the HCN family. Full-length putative protein sequences were constructed from the conceptual translation of genomic DNA as previously described (42) with the proposed starting methionine in vertebrates supported by the presence of a consensus start sequence (29). Genome position and intron-exon structure were examined and recorded. A list of the sequences used in the analysis is shown in Table 1.

Phylogenetic analyses.
Sequences were trimmed to produce a core alignment spanning from the start of the transmembrane segment S1 to the end of the CNBD (see Fig. 1). This region exhibits high sequence identity within the HCN family and among the other related sequences that were used. As there is no known bacterial HCN channel, and based on a previous phylogenetic analysis of HCN and CNG channels (10), KAT1 (GenBank GI no. 44888080), human ERG1 (GenBank GI no. 7531135), human CNGA1 and CNGA3 (GenBank GI no. 2506302 and 13959682) sequences were included in the analyses to serve as an out-group for the rooted phylogenetic trees of the HCN family. Six sequences that were missing exons due to gaps in the genome assembly were removed from the alignment prior to running the programs. Neighbor-joining (NJ) trees were generated using ClustalX, followed by tree evaluation with bootstrap resampling of 1,000 times. Additional NJ, maximum parsimony (MP), and maximum likelihood (ML) trees were created using the Seqboot, Protdist, Neighbor, Protpars, Proml, and Consense programs from the PHYLIP package (version 3.65), bootstrapping with 100 replicates with randomized input order and 10 jumbles (15). The TreeView program (version 1.6.6) (51) was used to examine and display all trees.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 
To date, 21 unique full-length HCN coding sequences have been cloned and deposited in GenBank: 4 from arthropods, 2 from sea urchin, 1 from fish, and 14 from mammals, specifically human, rabbit, mouse and rat. Previous phylogenetic analyses of the HCN family and its relatives have been limited and inconsistent, in part due to the lack of available sequences from a diverse sampling of species. For example, the product of the first duplication event has been shown to be either HCN1 (10) or HCN3 (16). Furthermore, the recently cloned spHCN2 (16) sequence was shown to group with the HCN family but was situated at the base of the phylogenetic tree. This is not consistent with species evolution, which places the sea urchin genes with the other deuterostome species, after the arthropod clade division. Therefore, spHCN2 may instead represent either a new family of channels not yet identified in any other species or an HCN homolog that was the product of a duplication event followed by rapid sequence divergence. These inconsistencies underscore the need for more comprehensive analyses. Accordingly, the aim of this study was to expand the breadth of available HCN sequences through the manual annotation of available genomes and in turn, to provide the first comprehensive sequence comparison and phylogenetic analysis of this protein family.

HCN genes are present in multiple copies across a wide spectrum of species.
Using a BLASTP search at NCBI or by searching the genome databases available at the Ensembl website, we identified 58 nonredundant HCN sequences. Twenty-seven of these sequences were previously identified by cloning or by computer annotation and confirmed by genomic data. The remaining 31 are novel and were completed by data mining the genomes available at Ensembl or the manual reannotation of the predicted protein translations. Several of the Ensembl predictions were inconsistent with known sequences, or the predicted gene did not span the estimated length of the transcript. The inherent problems in gene prediction and computer protein annotation methods have been previously described (Ref. 42 and references therein). We resolved these problems by multispecies comparisons and manual reannotation on a gene-by-gene basis. A complete list of the sequences included in this analysis is shown in Table 1 and their protein sequences can be found in Supplementary Fig. S1. (The online version of this article contains supplementary material.) Included in the final list are: 23 mammalian sequences; including 2 and 3 new sequences from the chimpanzee and opossum genomes, respectively; 22 lower vertebrate sequences; 6 urochordate sequences, including 3 new sequences from both C. intestinalis and C. savignyi genomes; and 7 invertebrate sequences, with a new annotation of the single HCN gene found in the mosquito genome. Of the lower vertebrates, 2 new sequences were from chicken, 3 from frog, 6 new sequences from both the green pufferfish and the Japanese pufferfish genomes, and 2 new sequences from zebrafish. No HCN sequences were identified in C. elegans. With this substantial increase in the number of full-length sequences representing a wide range of species, we reconstructed the evolutionary history of this protein family and probed the relationship of channel structure and function in greater detail than was previously possible.

High sequence identity among four vertebrate HCN isoforms within the core region.
Due to the length and sequence variability that occurs in the NH₂ and COOH termini (see Supplementary Fig. S2), these regions were trimmed and a core alignment was produced corresponding to a region between S1 and the end of the CNBD and approximately to exon 2 through 7 of the mammalian isoforms. Sequence conservation of the four vertebrate isoforms within this region is high, at least 80–90% identity amongst the mammalian sequences and over 90% residue conservation between the newly identified fish sequences and their respective mammalian orthologs (Fig. 2). This indicates that this region has been slow to evolve during the 450 million years (MY) that separate the fish and human lineages. Also of interest is that HCN4 shares the highest sequence identity with all other vertebrate isoforms and that the mammalian HCN3 and fish HCN3 sequences are as similar to the other vertebrate isoforms as they are to each other. This suggests that HCN4 sequences have diverged the least from a common ancestral sequence, as is further evident from the branch lengths in the NJ and ML phylogenetic trees, and that HCN3 sequences have evolved independently within the fish and mammalian lineages to an equal degree but at different sites. Overall, the sequences of the invertebrates and urochordates display a lower conservation with the mammalian sequences. Arthropods share a general sequence identity of 60–65% with the mammalian homologs, whereas one of the sea urchin sequences, spHCN1 (also known as spIH or spHCN), and two of the Ciona homologs, here named HCNb and HCNa, are only 55% identical. The second sea urchin sequence, spHCN2, and the other sequence from the Ciona species (HCNc) are even more diverged.

View larger version (25K): [in this window] [in a new window]   Fig. 2. High sequence conservation observed between vertebrate isoforms. Sequence conservation between mammalian HCN1–4 (M1–M4) and fish HCN1–4 (F1–F4) isoforms indicates that mammalian HCN3 and fish HCN3 sequences are as similar to each other as they are to the other isoforms. This suggests that sequence divergence has continued independently in each lineage. Numbers indicate % sequence identity/sequence conservation generated by GeneDoc for the region between S1 and end of CNBD, where the total number of sites used was 457. For example, 85/94 correlates to 393 of 457 sites identical and 430 of 457 conserved. Human sequences represent mammalian sequences and HCNa_fugu represent fish sequences. Fish ‘b’ sequences show lower conservation.

To further support their validity, HCN expressed sequence tags (ESTs) were found in the database of the Kyoto University Ciona cDNA project (59) (http://ghost.zool.kyoto-u.ac.jp/indexr1.html). These ESTs span several exons and cover a large portion of the C. intestinalis sequences provided here, suggesting that the three C. intestinalis genes have been correctly identified and that their transcripts are expressed. A recent comprehensive analysis of the ascidian genome revealed the existence of these three putative HCN ion channels (HCN1 = HCNc, HCN2 = HCNa, and HCN3 = HCNb) (50); however, the sequences themselves were not reported. The sequences identified here were in turn used to correctly annotate those found in the C. savignyi genome.

Three different HCN duplication events occurred prior to the divergence of the fish lineage.
To examine the phylogeny of the HCN family, a multiple sequence alignment of the core region was created using 52 of the 58 identified sequences (see MATERIALS AND METHODS). Three different phylogenetic programs, two NJ methods and one MP method, produced similar results. Due to the high sequence conservation among mammalian sequences, ML methods were incapable of analyzing the complete list of sequences. In a subgroup of 41 sequences, ML methods produced similar results to the NJ and MP trees. Figure 3 is a rooted MP consensus tree produced by the PHYLIP program. The NJ rooted phylogram created by ClustalX and a single representative ML phylogram produced by PHYLIP are provided in Supplementary Fig. S3. Four sequences were included as an out-group: human CNGA1 and CNGA3, human ERG1 and KAT1. Tree topology did not change when any of these out-group sequences were removed.

View larger version (23K): [in this window] [in a new window]   Fig. 3. A maximum parsimony (MP) consensus tree of the HCN family. We included 52 HCN sequences in this analysis. The MP tree was constructed by resampling 100 datasets (bootstrap) with randomized input order of 10 jumbles using Seqboot, Protpars, and Consense programs in the PHYLIP software. To control for length differences between the different isoforms, only the region between the start of S1 and the end of the CNBD was used. The tree is rooted with an out-group of KAT1, hERG1, CNGA1, and CNGA3, 4 sequences that are related to the HCN family. Similar to the previous studies, HCN2_urchin branches before the invertebrate clade, which may be a consequence of the MP method's sensitivity to unequal rates of evolution. The six urochordate sequences from C. intestinalis and C. savignyi fall according to their evolutionary position, between the invertebrates and vertebrate species, which is supported by high bootstrap values. Because they group within their own species, they are not considered orthologous to any of the mammalian isoforms and are therefore named HCNa, HCNb, and HCNc. HCN3 is the product of the first duplication event. A split between the fish and mammalian sequences is observed but the partition is not well defined. HCN4 is the product of the second duplication event, followed by the emergence of HCN2 and HCN1. Low bootstrap values within isoform clades are most likely a consequence of the high sequence identity within the region used and in turn, the lack of informative sites required by the MP method. Duplicate copies of mammalian paralogs in the fish species are denoted with ‘a’ or ‘b’ in their sequence names. *Sequences annotated in this study. Numbers indicate bootstrap values and represent % support for the respective partitions.

Our data also suggest that HCN4 is the product of the second duplication event followed by the emergence of HCN1 and HCN2. This proposed evolutionary pattern is evident in all three phylogenetic trees and is further supported by the sequence conservation pattern shown in Fig. 2. On the basis of phylogenetic results, the division between HCN1 and HCN2 remains unresolved. Nevertheless, functional data suggest a clear difference between these two isoforms. The discrepancy with the predicted order of species evolution within each mammalian clade is likely due to the high sequence conservation seen within this core region. This results in a limited number of informative sites and produces a low phylogenetic signal (17). The lack of sequence divergence in the mammalian clades may be due to insufficient evolutionary time to permit the accumulation of mutations and/or a strong selective pressure to retain the conserved sequence of this channel.

Fish lineage show evidence of duplicate HCN genes.
Genome data mining revealed multiple HCN genes in both the F. rubripes and Tetraodon species, in both cases exceeding the number found in the mammalian lineage. From the branching order in the phylogenetic tree shown in Fig. 3, it is clear that fish gene pairs group together within the clades of individual mammalian orthologs. This suggests that the common ancestor of teleost fish and tetrapods had four HCN genes and that these underwent further duplications independently in the fish lineage. These sequences have therefore been named according to their mammalian ortholog and subsequently designated ‘a’ or ‘b’. This pattern of duplication is clearly evident in the HCN2 and HCN4 clades, where the green puffer and fugu ‘a’ and ‘b’ sequences are grouped together and are separated from each other by high bootstrap values. Because the teleost fish are predicted to have undergone a complete genome duplication early in their own lineage (Ref. 23 and references therein), this distribution pattern is not unexpected. However, because some of the genes identified were not full length and some showed evidence of intron insertion, further analyses are required to determine whether these duplicates are expressed and functional or have become pseudogenes (73).

Ciona genes most likely arose through lineage-specific duplication events.
In contrast to the observed pattern of the multiple fish genes, the multiple copies of HCN sequences found within the urochordate species do not partition with any of the mammalian isoform clades. However, each gene did show a highly significant pairing between C. intestinalis and C. savignyi, indicating that these duplication events occurred before these two species diverged. The timing of this duplication is unknown, but with the presence of multiple HCN genes previously identified in the sea urchin, two different scenarios are possible. First, one duplication of an ancestral gene may have occurred prior to the divergence of the deuterostomes and given rise to the multiple HCN genes seen in the sea urchin and urochordate species. A lineage-specific duplication then occurred in the urochordates to produce the third Ciona HCN gene. Subsequently, the genes evolved rapidly and independently in the different lineages and thus no longer resemble each other or any specific HCN isoform. The loss of one ancestral gene and the duplication and diversification of the other would then have given rise to the four isoforms now common to all chordates. A second, and more parsimonious, pattern of evolution is that lineage-specific duplication events of a single HCN ancestor occurred and gave rise to the three HCN homologs within the urochordate lineage. Further ancestral duplication events occurred within the vertebrate lineage, after the divergence of the urochordates and prior to the fish lineage, and produced the four known mammalian isoforms. Lineage-specific gene duplication in Ciona has also been shown for the evolution of sodium channel gene family. Two sodium channel genes have been identified, but one possesses a sequence that has diverged considerably both from its paralogous pair and from the other known sodium channel gene sequences. The authors concluded that the duplication events occurred just prior to the fish lineage (33). Similarly, independent lineage-specific duplication was suggested for the ankyrin gene family based on their phylogenetic results and the differences in gene sequences between Ciona and the vertebrate homologs (7). The general branching order between the Ciona HCN homologs, in which HCNa and HCNc group together and HCNb is independent, is consistent among the different trees produced. Their position within the tree, however, is variable and is most likely a result of the different tree building methods used. It may also reflect the amount of lineage-specific sequence divergence that has occurred in these HCN genes, which has caused them to evolve independently of both the invertebrate and vertebrate clades and has blurred their phylogenetic position.

Predicted phylogenetic patterns are supported by exon boundary structure.
Our phylogenetic analyses indicate that the four vertebrate isoforms arose via three duplications of an ancestral HCN gene. This is supported by the exon structure of their coding sequences. The four human HCN genes are located on different chromosomes: 1q21.2 (HCN3), 5p12 (HCN1), 15q24–q25 (HCN4) (61), and 19p13 (HCN2) (72). Using the genome databases, we found this pattern of differential localization for all vertebrate HCN genes, from fish through mammals. Furthermore, the overall exon structure of the four isoforms has remained consistent since the duplication and divergence events (Fig. 4). It comprises eight exons, with highly conserved length and sequence in exons 2 through 7. An exception to this is observed in both of the fish HCN3b genes, which are predicted to have an intron of <75 bp inserted in the middle of exon 2. Exons 1 and 8, which directly correspond with the distal NH₂ and COOH termini, vary in both length and sequence for all vertebrate genes analyzed. In addition, exon boundary positions are highly conserved throughout the vertebrate lineage. This suggests that the extant vertebrate HCN sequences are derived from a single ancestral gene that had an exon structure similar to current mammalian HCN genes and that the duplication events occurred after the intron positions were fixed in the linear sequence.

View larger version (27K): [in this window] [in a new window]   Fig. 4. Conservation of exon structure reveals evolutionary patterns of HCN genes. Regions of one HCN subunit are delineated in the gray box and are not drawn to scale. These include: the NH2-terminus, 6 transmembrane segments (1–6), the region that forms the pore and selectivity filter in the tetrameric structure of the channel, C-linker, CNBD, and the distal COOH terminus. Below, the vertical lines represent exon boundary position of the coding sequence: black lines indicate an exon boundary position that is conserved with the human HCN isoforms, dark gray lines indicate conservation with invertebrate boundary positions, and light gray indicates boundary position specific to the individual protein. The exon number is shown above the horizontal black line, and the corresponding amino acid length is shown below. The exon structure of most vertebrate isoforms (HCN1–4) is conserved with that of human (exception is fish HCN3b genes, see text for details), consisting of 8 exons with 2 through 7 corresponding to the region between the start of S1 and the end of the CNBD. The sequence diverges in length between isoforms in the NH2 and COOH termini. A representative HCN human structure is shown at top with length differences of the four isoforms in exon 1 and 8 indicated by a range. The total lengths of the predicted exons and the exon boundary structure is variable among the three different C. intestinalis HCN coding sequences identified in this analysis. HCNb_c.intestinalis is composed of 15 exons, sharing all vertebrate HCN channels exon boundary positions, with an additional 6 unique boundaries and 1 boundary shared with the invertebrate HCN proteins of the fly and mosquito. HCNa_c.intestinalis is composed of 10 exons, but only 4 boundaries are conserved with vertebrates. HCNc_c.intestinalis is predicted to have 14 exons and shows no similarity in exon structure or boundary position to either of the other Ciona proteins or vertebrate HCN. In general, the mosquito and fly proteins have a higher number of exons compared with vertebrate HCN, sharing only a few boundary positions.

Evolution of key residues in the voltage sensing domain and pore region.
HCN channels open in response to changes in membrane voltage and allow for the passage of Na⁺ and K⁺ ions across the plasma membrane. The transmembrane domains of the individual subunits, which form tetramers around a central pore, are primarily responsible for these functions. As might be expected from the natural constraints of the hydrophobic bilayer, sequence conservation is abundant in areas predicted to correspond to the six transmembrane segments. Similar to depolarization-activated K⁺ channels, the fourth transmembrane segment (S4) contains positively charged residues and is likely to sense changes in voltage across the cell membrane (Fig. 5A, ‘a’). In contrast to depolarization-activated K⁺ channels, however, HCN channels open instead of close in response to hyperpolarization. Interestingly, HCN channels possess an additional four or five charged residues at the NH₂-terminal end of S4 (Fig. 5A ‘b’). Together, in response to changes in voltage, these charges have been shown to move in the same direction as that in depolarization-activated K⁺ channels (3). Therefore, it has been suggested that the movement of the voltage-sensing domain is coupled to channel opening in the opposite way (40). Throughout the HCN family, there is high sequence identity in this region among the vertebrate isoforms and high conservation across all sequences. The more diverged sequences from the sea urchin and urochordate species, spHCN2, Ciona HCNa and Ciona HCNc, which are predicted to be the products of lineage-specific duplication, possess only three of the upstream positive charges. The effect of this loss of charge awaits functional characterization but could be due to relaxed evolutionary constraints in these duplicate genes.

View larger version (125K): [in this window] [in a new window]   Fig. 5. Sequence comparison of the voltage sensing domain and pore region. A: sequence of voltage sensing domain conserved throughout evolution. The fly, human, and fugu sequences were chosen to represent arthropod, mammal, and fish species, respectively. Sequences are organized into invertebrates, urochordates, and vertebrate groups. A high sequence identity is notable among all vertebrate sequences. Shading indicates 4 levels of sequence conservation (see MATERIALS AND METHODS), excluding the assembly gap regions. Black stars indicate regions of assembly gaps in genome database, and the arrow indicates the conserved exon boundary between mammalian exon 2 and 3. a, Region of charge common to all potassium channel voltage sensing domains; b, region of charge unique to HCN channels; c, region associated with normal channel closure. B: sequence comparison of the HCN pore region. The same representative sequences as in 5A were used, except HCN4_chicken. This was omitted due to a gap in the genome assembly across this region. The selectivity filter sequence CIGYG (a) is conserved throughout the HCN family, with the exception of 2 duplicate genes from sea urchin and urochordate species. The NXT/S glycosylation site (b) is located immediately prior to the pore helix and is conserved in urochordate and vertebrate sequences. Two conserved glycine residues (c and d) are located in S6, one of which was hypothesized to be the hinge that obstructs that pore. More recent evidence suggests that the hinge is located further downstream between the threonine and glutamine residues (e) at the COOH-terminal end of S6 in which a dipeptide is conserved throughout the entire HCN family. f, The residue identified as responsible for the difference in chloride sensitivity in HCN2 vs. HCN1.

Based on sequence homology to the crystal structure of a bacterial K⁺ channel (14), the region between S5 and the end of S6 is believed to form the ion conduction pore in HCN channels. It contains a pore helix and selectivity filter and is involved in both ion selectivity and transport. In K⁺ channels, the selectivity filter exhibits the K⁺ signature sequence (GY/FG), which is thought to confer their K⁺-selective nature (14). In the HCN family, this tripeptide has been conserved but the channels allow the passage of significant amounts of both Na⁺ and K⁺. In all but two HCN genes, the sequence motif that corresponds to the selectivity filter is CIGYG (5) (Fig. 5B ‘a’). The conservation of this sequence only differs from K⁺-selective channels at the cysteine residue, implicating this site's involvement in the reduced K⁺ permeability. The two exceptions are again found in gene duplicates from sea urchin and urochordate species. spHCN2 has a filter sequence of SIGFG (16), which makes it more similar to the filter sequences of channels in the EAG K⁺ channel family. Functional data do not exist for the spHCN2 channel, so whether this difference affects ion selectivity is not known. In Ciona HCNc genes, the selectivity filter motif is CIGYS. In mammalian HCN channels, substitution of the second glycine residue by serine (G404S in HCN2) reduced the slow activating current (39). Evidence of channel function is required to determine if this residue difference in the urochordate channel is involved in gene silencing. If Ciona HCNc is functional, a different mutational tolerance at this particular site seems likely and may reflect an adaptive process that has enabled these channels to fill a different functional niche specific to these species.

Recently, an N-linked glycosylation site (NXT/S) located in the outer turret between the end of S5 and the pore helix (Fig. 5B ‘b’) has been shown to play a role in membrane expression of mammalian isoforms (47). Similar to residues in the S4–S5 linker, this motif emerges with the urochordate sequences and is conserved in two of the three urochordate genes and throughout all vertebrate sequences. To understand the necessity of this motif and the role it plays in channel function throughout evolution, further studies are needed to determine if glycosylation or some other compensatory posttranslational modification that allows the channel to mature through the ER/Golgi process, occurs in these Ciona channels and other invertebrate sequences.

On the basis of the K⁺ channel crystal structure (14), the S6 segments of HCN likely form the inner vestibule of the pore and the gate. They show high sequence conservation with other protein relatives, such as ERG and CNG, and are almost completely conserved throughout the HCN family. Not surprisingly, the S6 segments of Ciona HCNa and HCNc and of spHCN2, which are predicted to be the result of lineage-specific duplication, are the exceptions. Two glycine residues (Fig. 5B ‘c’ and ‘d’) are completely conserved among the HCN genes. Based on sequence homology with voltage-gated K⁺ channels (62), it has been suggested that one of these may form a glycine hinge involved in the opening of the channel gate in response to the movement of the S4–S5 linker. However, more recent experimental and homology modeling evidence (19, 55, 56) has shown that a threonine residue, positioned after the glycines in the linear sequence and on the intracellular side of the channel, is only accessible in the open state. Furthermore, a glutamine residue at the end of S6 is accessible in both the open and closed positions, suggesting that the putative hinge position is located between these two residues (Fig. 5B ‘e’).

Evolution of the cyclic nucleotide binding and modulatory domains.
Cyclic nucleotides bind directly to the CNBD (12) located in the intracellular COOH terminus and modify channel opening. The CNBD is an evolutionarily conserved domain that is found in several cyclic nucleotide binding proteins, including the bacterial catabolic activating peptide and the protein kinase A family (4). The crystal structure of the COOH terminus of mouse HCN2 has been solved (78). Despite a low overall sequence conservation, the tertiary structure of its CNBD is highly conserved with the crystal structures of other CNBDs (4). Furthermore, residues identified as being critical to structure and nucleotide binding are conserved throughout the HCN family (Fig. 6). This includes the phosphate binding cassette (PBC), the most conserved feature of the CNBD that makes direct contact with cAMP (13), and the hinge region, important for the capping of cAMP by the C-helix of the CNBD. The middle of the PBC has diverged in one of the sea urchin genes and all of the urochordate genes, but these residues are also variable in other cyclic nucleotide-binding proteins, suggesting that they may not be critical for nucleotide binding. Because the HCN family possesses these conserved key residues, it seems probable that all of their CNBDs have a stable tertiary structure and bind cyclic nucleotides. One exception to this is seen in urochordate HCNc genes. These two sequences are missing a key hydrophobic residue in the PBC (Fig. 6 ‘’) that forms a conserved interaction with cAMP (53). In these genes, this residue is threonine, a hydrophilic amino acid that could disrupt this cAMP interaction. This difference is consistent with our hypothesis that the Ciona HCNc genes have undergone diversification following a lineage-specific duplication. Whether this difference also corresponds to a functional change in cAMP binding is an interesting question that will require functional studies to confirm.

View larger version (72K): [in this window] [in a new window]   Fig. 6. Sequence comparison of the cyclic nucleotide binding domain. A sequence alignment of the CNBDs from representative sequences of the HCN family indicates that functionally important residues in the CNBD are conserved throughout evolution. Shading indicates 4 levels of sequence conservation (see MATERIALS AND METHODS). Evolutionarily conserved glycine residues important for maintaining the fold of the ß-roll found in all CNBDs (located between A and B) and those residues known to interact with the cyclic nucleotide are indicated by stars below the alignment. The phosphate binding cassette (PBC) and hinge region common to all CNBDs are identified by the black lines below the sequence. Cyclic nucleotides are known to bind to the second conserved arginine in the PBC. The arrow indicates the hydrophobic site where divergence has occurred in Ciona HCNc genes (as explained in text). Cylinders above the sequence represent helices, and arrows indicate b-strands that together form the B-roll.

Overall, the role of the C-linker in HCN channels is unusual compared with other regions in the channel. Its sequence and function show variation throughout the family, but it connects two domains that are themselves highly conserved in sequence throughout evolution and confer highly conserved, but distinct, functions in a cooperative manner (transmembrane = voltage-sensing, channel opening, and ion permeation; CNBD = cAMP binding and modulation of channel open by the transmembrane domain). Because of its position between these two domains, the sequence variability in the C-linker may be, in part, the result of an adaptive process that has enabled the diversification of cyclic nucleotide binding and the modulation of channel opening by the CNBD within the HCN family.

Sequence variability and functional divergence of vertebrate HCN paralogs.
From our data, we predict that the four vertebrate isoforms are paralogs of each other resulting from three duplication events that occurred before the divergence of the teleost and tetrapod lineages. At the time of their origin, the gene pairs produced by these events would have been functionally redundant. Because the vast majority of duplicate genes are silenced throughout the course of evolution (37), the retention of all four is probably due to the acquisition of unique functional characteristics and/or expression patterns that result from tolerated mutations specific to each paralog (65). Throughout the core region, there is high sequence identity among the four vertebrate HCN isoforms. Some of these invariant residues probably contribute to functional properties common among all vertebrate channels. In contrast, some of the residues that vary among paralogs within this conserved region probably contribute to the more subtle isoform-specific differences in function (31). The search has begun to identify which residues underlie differences in function among the four mammalian isoforms using direct sequence comparisons followed by single site mutagenesis and/or construction of chimeras, and functional assays. Using these approaches, three studies have identified residues or regions responsible for phenotypic variation among the four mammalian HCN channels. First, differences in the rates of channel opening and cyclic nucleotide modulation between the HCN2 and HCN4 isoforms have been localized to a variant residue in the S1 segment (68). Second, a single residue difference in the inner selectivity filter was shown to confer chloride sensitivity to the HCN2 channel (74) (see Fig. 5B). Lastly, differences in the C-linker have been shown to contribute to differences in cAMP efficacy between HCN1 and HCN2, although the specific residues involved were not identified (76). From these few examples, it is clear that the functional consequences of residue variation among the four vertebrate isoforms, which may encompass not only overt differences in channel opening and closing but also differences in permeation, cyclic nucleotide affinity, homo- and heterotetrameric assembly, glycosylation status, protein-channel interactions, and abilities to traffic to the cell surface, cannot be predicted simply by the location within the channel. Differences in function may involve interactions among multiple residues and domains located throughout the channel. Therefore, functional assays in combination with site-directed mutagenesis and/or the construction of chimeras between vertebrate isoforms may not be sufficient to determine all differences.

In this study, we expanded the list of sequences for each of the four HCN paralogs by the addition of sequences from lower vertebrates. Broadening the evolutionary representation of the four vertebrate isoforms improves the sequence signal-to-noise ratio and enhances the identification of residues that are conserved among orthologs but differ among paralogs. However, genes from lower vertebrates (e.g., fish) and mammals have continued to evolve under different selective pressures, since their divergence from a common ancestor, several MYA. Therefore, information derived from this expanded list of sequences is complex. We identified three groups of residues based on similarities among vertebrate orthologs. First are residues conserved among orthologs and could contribute to a function specific to each vertebrate isoform. Second are residues conserved in the mammalian and fish orthologs but differ between these two groups. These residues may contribute to functional differences between the mammalian and fish channels of a particular isoform. Finally are residues conserved in orthologs of mammals or fish, but not both. These could be involved in species-specific or paralog-specific functions within each species. Alternatively, they may be the result of relaxed evolutionary constraints. By analyzing these different sets of conserved sites together with functional characterization of the various channels belonging to each of the vertebrate isoforms, we can more easily identify sites that may contribute to paralog-specific differences in channel function.

An informative subset of sites identified when comparing HCN genes from lower vertebrates and mammals are those that are conserved with a paralog rather than its own ortholog. If we assume that the probability of identical sites mutating to the same residues independently following duplication and species divergence is low, then these probably represent sites that have been retained from ancestral genes. Differential retention of sites from ancestral genes among orthologs suggests that channel phenotypes may not be completely conserved among them. In conjunction with the identification of conserved and nonconserved sites, the functional analysis of HCN channels from lower vertebrates and comparisons of their properties with those of mammalian channels will help to identify residues that contribute to different phenotypes and will also have the potential to shed light on the sequences and functions of ancestral HCN channels.

Vertebrate isoform-specific alignments of sequences spanning 450 MY reveal conserved motifs in the NH₂ and COOH termini.
The large number of vertebrate HCN sequences assembled in this study has greatly increased the power to identify isoform-specific motifs in the NH₂ and COOH termini that may contribute to unique functions and specific patterns of expression within cells and tissues. The high sequence conservation among the four vertebrate isoforms extends beyond the core region analyzed above and includes regions just upstream of the S1 segment and downstream of the CNBD. The more distal NH₂ and COOH termini, however, vary in both length and sequence among paralogs and do not align well. The sequences of the mammalian orthologs identified prior to this study were too close in evolutionary time to reveal divergence in the distal NH₂ and COOH termini. On the other hand, the four paralogs from a single species were too diverged to align reliably. By expanding the taxa sampling to include species from different time periods of vertebrate evolution, the signal-to-noise ratio that is inherent in the sequence information was considerably improved (2). Alignments of the distal termini, using this expanded list of vertebrate sequences, revealed several isoform-specific blocks of conserved sequence interspersed with diverged regions of variable length (Fig. 7).

View larger version (91K): [in this window] [in a new window]   Fig. 7. Individual isoform alignments of the distal termini of HCN channels. A: alignment of the NH2-terminal region of vertebrate HCN1, 2, and 4 isoforms. Sequence alignments of available vertebrate sequences for the NH2-terminal region of 3 of the 4 individual isoforms were produced using ClustalX and edited and shaded using GenDoc. Shading indicates 4 levels of sequence conservation (see MATERIALS AND METHODS). a, Example of conserved motif in vertebrate HCN2 sequences that may play a role in channel surface expression; b, region conserved among all vertebrate HCN sequences that may be involved in tetrameric interactions in HCN2 and may play a similar role in HCN1, 3, and 4. Arrows between conserved blocks represent variable regions with the range of amino acid length listed above. Numbers at the start and end of the sequences correspond to their position in the full-length sequences provided in Supplementary Fig. S1. B: alignment of the COOH-terminal region of vertebrate HCN1, 2, and 4 isoforms. Sequence alignments of all available vertebrate sequences for the COOH-terminal region of 3 of the 4 individual isoforms were produced using ClustalX and edited/shaded using GenDoc. Shading indicates 4 levels of sequence conservation (see MATERIALS AND METHODS). a, Example of conserved motif in vertebrate HCN1 sequences has been shown to interact with Filamin A; b, region conserved among all vertebrate HCN sequences downstream of the CNBD; c, tripeptide sequence shown to be involved in protein-protein interactions with scaffolding proteins in vivo and to regulate cell surface expression. Dotted line indicates conserved region among 4 vertebrate isoforms previously identified but of unknown function. Arrows between conserved blocks represent variable regions with the range of amino acid length listed above. Numbers at the start and end of the sequences correspond to their position in the full-length sequences provided in Supplementary Fig. S1.

Finally, both the NH₂ and COOH termini possess regions of high sequence conservation among the four vertebrate isoforms, in addition to those of the core region used for the phylogenetic analyses. These regions probably confer properties that are not unique to any individual isoform but are important for all vertebrate HCN channels. In the NH₂ terminus, a region of 50 residues immediately upstream of the start of S1 is conserved in all four vertebrate isoforms (Fig. 7A ‘b’, HCN3 not shown). In mouse HCN2, this region interacts with itself and may be involved in intersubunit interactions of tetrameric assembly (70). Furthermore, the removal of this region, along with the rest of the NH₂ terminus, prevents the formation of functional channels. Based on the high level of sequence conservation among the four isoforms, it seems probable that this region provides similar interactions for HCN1, 3, and 4. If this is true, the few differences among paralogs within this region may modify intersubunit interactions and thus regulate homomeric and/or heteromeric assembly.

In the COOH terminus, a block of residues conserved among the vertebrate HCN1, 2, and 4 isoforms was identified immediately downstream of the CNBD, which corresponds to the start of the last exon (Fig. 7B ‘b’). Deletion of this region, together with the entire distal COOH terminus and the C-helix of the CNBD, does not affect HCN1, HCN2, or HCN4 channel cell surface expression or function in heterologous systems (54, 75). A function for this conserved block of residues is not known, but whatever this function may be, based on the lack of sequence conservation, it is not likely possessed by HCN3 channels. Finally, a motif found at the distal end of the COOH terminus, SNL, is conserved in HCN1, 2, and 4 (Fig. 7B ‘c’). This motif, which qualifies as a consensus PDZ-binding domain, interacts with PDZ-containing proteins in vitro (28), and also with the TRP8 protein in vitro and in vivo, where it regulates channel cell surface expression (58). The absence of this motif in the HCN3 genes suggests that either this isoform does not interact with PDZ-containing proteins or that this interaction takes place elsewhere within the channel.

Summary and Perspectives
The availability of an increasing number of genome sequences has enabled us to generate a list of putative HCN coding sequences that has doubled the number of those previously known and covers a significantly greater portion of evolutionary time. With improved species representation, we were able to more accurately complete sequence comparisons, phylogenetic analyses and exon structure comparisons of the HCN gene family and put forward a model of its molecular evolution. We propose that the vertebrate isoforms evolved from a single ancestral sequence that had an exon structure similar to current mammalian HCN genes and that the duplication events occurred after the intron positions were fixed in the linear sequence. Duplications appear to have occurred independently in the sea urchin, urochordate, and fish lineages. Increasing the evolutionary distance between the sequences for each HCN isoform provided a good contrast and enabled us to unmask and identify regions putatively important for isoform-specific, as well as species-specific, functions. Together, our study provides a strong basis from which to refine the proposed model of evolution as more genomes become available and as the functional analysis of HCN genes progresses. Finally, our study provides a valuable tool to aid in the planning of experiments that probe the relationship between structure and function of HCN channels and the functional significance of sequence similarities and differences among them.

	GRANTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 
Supported by grants from the Natural Sciences and Engineering Research Council of Canada.

	ACKNOWLEDGMENTS

 
We gratefully acknowledge Dr. Fiona Brinkman (Molecular Biology and Biochemistry, Simon Fraser University) for advice, training, and feedback on the manuscript, as well as the Ouellette laboratory at the University of British Columbia Bioinformatics Centre (http://bioinformatics.ubc.ca) for resource advice that has made this work possible.

Current address of C. R. Marshall: The Centre for Applied Genomics and Program in Genetics and Genomic Biology, The Hospital for Sick Children, Department of Molecular and Medical Genetics, University of Toronto, Ontario, Canada.

	FOOTNOTES

 
Address for reprint requests and other correspondence: E. A. Accili, Univ. of BC, Vancouver, BC, Canada (e-mail: eaaccili{at}interchange.ubc.ca).

Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION GRANTS REFERENCES

 

1. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol 215: 403–410, 1990.[CrossRef][ISI][Medline]
2. Anderson PA, Greenberg RM. Phylogeny of ion channels: clues to structure and function. Comp Biochem Physiol B Biochem Mol Biol 129: 17–28, 2001.[CrossRef][Medline]
3. Bell DC, Yao H, Saenger RC, Riley JH, Siegelbaum SA. Changes in local S4 environment provide a voltage-sensing mechanism for mammalian hyperpolarization-activated HCN channels. J Gen Physiol 123: 5–19, 2004.[CrossRef][ISI][Medline]
4. Berman HM, Ten Eyck LF, Goodsell DS, Haste NM, Kornev A, Taylor SS. The cAMP binding domain: an ancient signaling module. Proc Natl Acad Sci USA 102: 45–50, 2005.[Abstract/Free Full Text]
5. Biel M, Schneider A, Wahl C. Cardiac HCN channels: structure, function, and modulation. Trends Cardiovasc Med 12: 206–212, 2002.[CrossRef][ISI][Medline]
6. Birney E, Andrews TD, Bevan P, Caccamo M, Chen Y, Clarke L, Coates G, Cuff J, Curwen V, Cutts T, Down T, Eyras E, Fernandez-Suarez XM, Gane P, Gibbins B, Gilbert J, Hammond M, Hotz HR, Iyer V, Jekosch K, Kahari A, Kasprzyk A, Keefe D, Keenan S, Lehvaslaiho H, McVicker G, Melsopp C, Meidl P, Mongin E, Pettett R, Potter S, Proctor G, Rae M, Searle S, Slater G, Smedley D, Smith J, Spooner W, Stabenau A, Stalker J, Storey R, Ureta-Vidal A, Woodwark KC, Cameron G, Durbin R, Cox A, Hubbard T, Clamp M. An overview of Ensembl. Genome Res 14: 925–928, 2004.[Abstract/Free Full Text]
7. Cai X, Zhang Y. Molecular evolution of the ankyrin gene family. Mol Biol Evol 23: 550–558, 2006.[Abstract/Free Full Text]
8. Chen J, Mitcheson JS, Tristani-Firouzi M, Lin M, Sanguinetti MC. The S4–S5 linker couples voltage sensing and activation of pacemaker channels. Proc Natl Acad Sci USA 98: 11277–11282, 2001.[Abstract/Free Full Text]
9. Cho WJ, Drescher MJ, Hatfield JS, Bessert DA, Skoff RP, Drescher DG. Hyperpolarization-activated, cyclic AMP-gated, HCN1-like cation channel: the primary, full-length HCN isoform expressed in a saccular hair-cell layer. Neuroscience 118: 525–534, 2003.[CrossRef][ISI][Medline]
10. Craven KB, Zagotta WN. CNG and HCN channels: two peas, one pod. Annu Rev Physiol 68: 375–401, 2006.[CrossRef][ISI][Medline]
11. DiFrancesco D. Pacemaker mechanisms in cardiac tissue. Annu Rev Physiol 55: 455–472, 1993.[CrossRef][ISI][Medline]
12. DiFrancesco D, Tortora P. Direct activation of cardiac pacemaker channels by intracellular cyclic AMP. Nature 351: 145–147, 1991.[CrossRef][Medline]
13. Diller TC, Madhusudan Xuong NH, Taylor SS. Molecular basis for regulatory subunit diversity in cAMP-dependent protein kinase: crystal structure of the type II beta regulatory subunit. Structure 9: 73–82, 2001.[Medline]
14. Doyle DA, Morais Cabral J, Pfuetzner RA, Kuo A, Gulbis JM, Cohen SL, Chait BT, MacKinnon R. The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280: 69–77, 1998.[Abstract/Free Full Text]
15. Felsenstein J. Inferring phylogenies from protein sequences by parsimony, distance, and likelihood methods. Methods Enzymol 266: 418–427, 1996.[ISI][Medline]
16. Galindo BE, Neill AT, Vacquier VD. A new hyperpolarization-activated, cyclic nucleotide-gated channel from sea urchin sperm flagella. Biochem Biophys Res Commun 334: 96–101, 2005.[CrossRef][ISI][Medline]
17. Gallin WJ. Evolution of the "classical" cadherin family of cell adhesion molecules in vertebrates. Mol Biol Evol 15: 1099–1107, 1998.