The high-affinity K⁺ (HAK) transporter gene family – also called KT or KUP transporter family – is an ancient large family with members in Bacteria, Archaea, Fungi, Amoebozoa, and probably also in some species of Animalia (Grabov, 2007; Benito et al., 2011). Initially HAK genes have been deduced from plants by their similarity to K⁺ uptake permeases (KUP) from E. coli (Schleyer and Bakker, 1993) and high-affinity K⁺ transporters (HAK) from fungi (Banuelos et al., 1995; Quintero and Blatt, 1997; Santa-Maria et al., 1997; Fu and Luan, 1998; Kim et al., 1998). Several members of this family were shown to function as K⁺ uptake transporters in plants especially when the external potassium concentration was in the low μM range (Gierth et al., 2005; Aleman et al., 2011) indicating that HAK transporters are involved in high-affinity K⁺ uptake. Interestingly, all plant genomes analyzed so far contain genes encoding HAK transporters, while in Bacteria, Archaea, and Fungi they were found only in a subset of species (Grabov, 2007; Benito et al., 2011). The HAK family is the largest family of potential K⁺ transporters in plants and members of this family are expressed in nearly all tested plant tissues suggesting that HAK transporters have a general function in K⁺ supply (Banuelos et al., 2002).

To date, the topology of HAK transporters has neither been determined experimentally nor by in silico predictions; nevertheless, hydropathy profiles of these proteins suggest about 12 putative transmembrane segments and a long hydrophilic COOH-terminal region. In genome-wide screenings, HAK transporter proteins can be pinpointed by the presence of several conserved consensus motifs (see Materials and Methods). Our screenings identified 13 HAKs in Arabidopsis, 27 in rice, 22 in poplar, 18 in P. patens, and 11 in S. moellendorffii (Table 1; see Yang et al., 2009, for comparison). It is likely that the genome of P. trichocarpa contains more genes coding for HAK transporters, because especially in the screening of poplar we discarded partial sequences resulting from pre-mature gene annotation. Phylogenetic analyses allowed subdividing them into six independent groups (Figure 1A; see Rubio et al., 2000, for initial grouping-into Groups I–IV) and revealed that the last recent common ancestor of all embryophytes had two HAK transporters (Figure 1B). One of these diverged into current Group II, whereas the other was duplicated at least three times before the origin of tracheophytes. Two early duplication events got lost in the lineage leading to tracheophytes and led to P. patens-specific gene family amplifications (Groups V and VI). HAK transporters in S. moellendorffii spread over the other clades (Groups I–IV).

Functional information on HAK transporters is unfortunately still scarce. Most data are available for HAK transporters belonging to Group I. Ara-tha-HAK5, Ory-sat-HAK1, Ory-sat-HAK5, and Phy-pat-HAK1 were characterized as high-affinity K⁺ transporters (Rubio et al., 2000; Banuelos et al., 2002; Gierth et al., 2005; Garciadeblas et al., 2007; Qi et al., 2008; Horie et al., 2011b). It might thus be an educative guess to propose high-affinity K⁺-uptake properties also for the S. moellendorffii orthologs in the same clade, i.e., Sel-moe-HAK5, Sel-moe-HAK8, and Sel-moe-HAK9.

HAK transporters may not only mediate transport across the plasma membrane. Transient expression of the Ory-sat-HAK10::GFP fusion protein in living onion epidermal cells targeted this protein (from Group II) to the tonoplast (Banuelos et al., 2002); and Ara-tha-KUP/HAK/KT12 (from Group I) was found in the chloroplast proteome (Kleffmann et al., 2004; Peltier et al., 2004). Additionally, HAK transporters may not exclusively transport K⁺. Phy-pat-HAK1 and Ara-tha-HAK5, for instance, were reported to be permeable also for Cs⁺ (Garciadeblas et al., 2007; Qi et al., 2008); and Phy-pat-HAK13 belonging to Group IV was recently characterized as a high-affinity Na⁺ uptake transporter (Benito et al., 2012). We therefore propose for the closely related Sel-moe-HAK1 and Sel-moe-HAK11 from S. moellendorffii similar sodium-transport features. This phylogenetic divergence may indicate a – so far underexplored – diversity of HAK transporters in fine-tuned function of K⁺ uptake and re-distribution, cellular expression, and/or sub-cellular targeting.

HKTs in plants belong to a family of monovalent cation transporters comprising also the fungal TRKs (K⁺ transporters) and bacterial KtrABs (Na⁺-dependent K⁺ transporter), for instance (Corratge-Faillie et al., 2010). Proteins of this family share a common structure of four TM-P-TM motifs (every two transmembrane α-helices are connected by ∼30 aa-long pore-forming P segments), which might have evolved from an ancestor related to the bacterial KscA K⁺ channel of Streptomyces lividans (Durell and Guy, 1999; Figure 2A). The plant HKT family comprises transporters that mediate Na⁺ uptake in roots or in other plant organs. They accumulate Na⁺ from the soil and recirculate it throughout the plant. There are two types of plant HKT transporters that can be distinguished by the amino acid sequence of the selectivity filter (the narrowest part of the permeation pathways that selects one ion species over others) of the first TM-P-TM motif: (i) S-S-M and (ii) [T,S,I]-G-L. Two HKTs from O. sativa and A. thaliana belonging to the first type have been well characterized in planta as Na⁺ uptake transporters (Uozumi et al., 2000; Rus et al., 2001; Maser et al., 2002a,c; Berthomieu et al., 2003; Garciadeblas et al., 2003; Sunarpi et al., 2005; Horie et al., 2007; Xue et al., 2011). The function of the second type has not been studied in plants. Nonetheless, two members of this group, from barley and wheat, mediate K⁺ or Na⁺uniport or Na⁺-K⁺symport – depending on the protein expression level – when heterologously expressed in yeast cells (Haro et al., 2005; Banuelos et al., 2008). Functional expression of this type of transporters in Xenopus oocytes produced similar results with slight variations regarding K⁺ versus Na⁺ permeability, symport activity, and permeability to divalent cations (Rubio et al., 1995; Gassman et al., 1996; Jabnoune et al., 2009; Lan et al., 2010; Horie et al., 2011a; Oomen et al., 2012).

In genome-wide screenings, proteins of the HKT type can be pinpointed by the presence of several conserved consensus motifs (see Materials and Methods). Our screenings identified one HKT-coding gene in Arabidopsis, seven in rice, one in poplar, one in P. patens, and six in S. moellendorffii (Table 2). Phylogenetic analyses grouped all of them into a single group of orthologs (Figure 2B) indicating that the most recent common ancestor of all embryophytes comprised a single protein of the HKT type. P. patens has a single extant representative (Phy-pat-HKT1), whereas in tracheophytes several duplication events occurred in different lineages (Figure 2C). The obviously independent multiplication of HKT-coding genes in rice and S. moellendorffii may be correlated with the affinity of these vascular plants to moisture environments. Probably, a larger variety of Na⁺/K⁺ transporters provides some advantage for better adaptation. Initially, the HKT family has been partitioned into the two subfamilies one and two, and transporter nomenclature was adjusted accordingly of the type “species HKT subfamily; No” (Platten et al., 2006). Subfamily one gathers transporters with the S-S-M signature in the selectivity filter of the first TM-P-TM motif. Our analysis now reveals that this subfamily division emerged in land plants only after the separation of Lycopodiophyta. Thus, the proposed unified nomenclature cannot be applied to all plant HKT genes. The rules fail, for instance, for HKTs from S. moellendorffii and P. patens (see also Haro et al., 2010).

At the functional level, the six HKTs of S. moellendorffii very likely share properties of the orthologs from other species. They may thus be implicated in K⁺/Na⁺ recirculation in this vascular plant and could contribute not only to K⁺ transport but in first line to desalination and Na⁺ detoxification.

Potassium channels play important roles in many physiological aspects of higher plants such as osmoregulation, turgor-driven movements, and ion uptake. It is estimated that K⁺ channels can contribute to more than 50% of the nutritional K⁺ uptake under most field conditions (Spalding et al., 1999; Amtmann and Blatt, 2009). In angiosperms there are two large groups of K⁺ channels: voltage-gated channels, the activity of which is regulated by the transmembrane voltage (Dreyer and Blatt, 2009), and non-voltage-gated K⁺ channels. Non-voltage-gated K⁺ channels form the class of tandem-pore K⁺ (TPK) channels. Functional TPK channels are proposed to form dimers consisting of two identical subunits (Maitrejean et al., 2011). Each subunit is characterized by a structure with four transmembrane domains and two pore-forming loops between the first and second and the third and forth membrane-spanning domain, respectively (Figure 3A; Voelker et al., 2010).

Searching for proteins with the characteristic pore-forming region, in the genome of S. moellendorffii four genes coding for TPK channel subunits could be identified (Table 3). Together with the six TPKs from Arabidopsis (Ara-tha-TPK1–5 and Ara-tha-KCO3), three from rice, ten from poplar, and three from P. patens they could be classed into two groups of orthologs (Figure 3B). This implies that the ancestor of land plants had already two of these genes. A deeper phylogenetic analysis revealed several duplication events in the two groups, both species-specific and at higher levels (Figure 3C). A remarkable example in this context is KCO3 from A. thaliana. This subunit lacks the first of the two pore loops and was originally considered as founder of a separate channel family with structural features (TM-P-TM; two transmembrane α-helices, and pore-forming P segment) similar to the simplest class of K⁺ channels from bacteria and animals. It became evident, however, that Ara-tha-KCO3 developed through a very recent evolutionary event involving gene duplication of the Ara-tha-TPK2 gene followed by partial deletion (Marcel et al., 2010; Voelker et al., 2010). And indeed, in line with this concept, neither the genome of S. moellendorffii nor that of P. patens appears to contain genes coding for K⁺ channels of the TM-P-TM type.

TPK channels in plants were reported to be targeted to the vacuolar membrane (Czempinski et al., 2002; Voelker et al., 2006; Latz et al., 2007; Dunkel et al., 2008; Isayenkov et al., 2011a,b). The exception is Ara-tha-TPK4 which has been reported to be targeted also to the plasma membrane (Becker et al., 2004). However, orthologs of Ara-tha-TPK4 were only found in the genus Arabidopsis so far (i.e., A. thaliana and A. lyrata; Voelker et al., 2010) but not in other plant species indicating a rather recent evolutionary event in channel specialization. Therefore, we have justified reasons to hypothesize that TPKs in bryophytes and lycophytes are vacuolar K⁺ channels. Their physiological role, however, remains as speculative as that of TPKs in other plants (Voelker et al., 2010).

Voltage-gated K⁺ channels are tetrameric proteins built of four α-subunits. One subunit shows usually a structure with six transmembrane domains and one pore loop (5TM-P-TM). The first four transmembrane domains fold into the voltage-sensor module, and the pore loop together with the fifth and sixth transmembrane domains establishes the permeation pathway module (Figure 4A; Dreyer and Blatt, 2009).

Voltage-gated potassium channels in angiosperms are targeted to the plasma membrane and could normally be grouped into the class of Shaker-like K⁺ channels that subdivides into four functional subgroups: (a) Inward-rectifying (K_in) channels open at membrane hyperpolarization and are responsible for K⁺ uptake. (b) Silent (K_silent) channel subunits assemble with K_in subunits and modulate K⁺ uptake channel properties. (c) Weakly rectifying (K_weak) channels are specialized K_in channels that show a bi-modal gating behavior. They appear to play a special role in the energy household of vascular tissues. (d) Outward-rectifying (K_out) channel subunits open at depolarizing voltages and mediate K⁺ release, e.g., during xylem loading or stomata closure (see Dreyer and Uozumi, 2011, for a contemporary review). Our screening strategy based on the characteristic pore-forming region identified nine – already known – genes coding for Shaker-like channels in Arabidopsis, eleven in rice, and eleven in poplar. Genes coding for Shaker-like K⁺ channels were also identified in the moss P. patens and in the fern ally S. moellendorffii (Table 4). However, whereas four K_in-like channels were identified in P. patens, despite a very careful screening strategy no such gene could be found in S. moellendorffii. Instead there, a gene coding for a K_out channel subunit was discovered as the only Shaker-like channel (Figure 4). This channel comprises all the essential structural features that were shown in the Ara-tha-SKOR K_out channel to be responsible for a unique K⁺ sensing property (Johansson et al., 2006). K_out channels open upon depolarization but additionally adjust their gating to the prevailing concentration of K⁺ outside. As a consequence, they open only at voltages positive of the K⁺ equilibrium voltage, E_K, when the electrochemical driving force is directed outward and so ensure K⁺ efflux regardless of the extracellular K⁺ concentration. This ability to adapt channel gating to the cation concentration outside guarantees an efficient K⁺ release during xylem loading and stomatal closure, for instance, even under varying external K⁺ (from 10 nM to 100 mM; Blatt, 1988; Schroeder, 1988; Wegner and de Boer, 1997; Gaymard et al., 1998; Ache et al., 2000). From analogy we may postulate that the presence of a K_out channel in the vascular plant S. moellendorffii and its absence in the non-vascular plant P. patens is correlated with the important evolutionary step of vascularization. In contrast, it is rather difficult to find an explanation for the loss of the K_in/K_weak/K_silent channel branch in S. moellendorffii.

In addition to Shaker-like channels the genomes of both, S. moellendorffii and P. patens, contain members of another class of putatively voltage-gated potassium channels (Table 5). These channels show some similarity with large conductance Ca²⁺-activated K⁺ channels (“big K” = BK channels), a channel type that is widely present in animals (including humans) but absent in flowering plants, for instance. BK channels activate in response to membrane depolarization and binding of intracellular Ca²⁺ and Mg²⁺ (Latorre et al., 2010). These channels are built of α- and β-subunits, where – as in Shaker-like channels – four α-subunits form the per se functional permeation pathway-establishing unit and the β-subunits just modulate and fine-tune channel properties. In contrast to Shaker-like channels the BK channel protein consists of seven (instead of six) transmembrane domains (6TM-P-TM structure) that lead to an exoplasmic N-terminus. Also BK-like channel α-subunits from P. patens and S. moellendorffii show a larger N-terminal region compared to plant Shaker-like channels. It might thus be speculated that also these proteins fold into a 6TM-P-TM structure instead of the 5TM-P-TM Shaker-like topology.

The functional properties and the physiological roles of plant BK-like channels are unknown. In mammalian tissues, BK channels serve as a negative-feedback mechanism for excitatory events that lead to increases in calcium concentration or membrane depolarization. In this way, they play a key role, for instance, in regulating the contractile tone in vascular smooth muscle cells or help to terminate the action potential and thus modulate secretion in chromaffin cells. It might be speculated that – at least in S. moellendorffii – the two BK-like channels could compensate for the absent Shaker-like K_in channels in carrying out functions in K⁺ uptake and distribution.

To assess the evolutionary origin of the non-Shaker-like channels we screened the genomes of the green algae Chlamydomonas reinhardtii, Coccomyxa sp.C-169, Micromonas sp. RCC299, Ostreococcus tauri, and Volvox carteri for voltage-gated K⁺ channels. Despite the fact that this transporter class in algae exhibits a huge structural diversity comprising also homologs of plant Shaker-like channels (Figure 5), a clear trace leading to BK-like channels in S. moellendorffii or P. patens could not be identified. The two most similar channels from Volvox and Chlamydomonas share an identity of 16–19% over a stretch of ∼500 amino acids. In comparison, a BLAST search at NCBI¹ limited to a query coverage of >45% resulted as best hit outside Animalia, S. moellendorffii, or P. patens in a voltage-gated K⁺ channel from Phytophthora infestans (XM_002998337) with ∼28% identity over a stretch of ∼500 amino acids. Unfortunately, from all these results we cannot resolve unequivocally the origin of BK-like channels in S. moellendorffii and P. patens. Neither can we exclude the possibility that bryophytes and lycophytes may have acquired these K⁺ channel genes from Fungi or Protozoa. However, our data (Figure 5) clearly indicate that the diversity of voltage-gated K⁺ channels observable in Chlorophyta collapsed contemporaneously with the transition of green plants from an aqueous to a dry environment. In higher plants the subsequent functional diversification into K_in/K_out/K_weak/K_silent (Figure 4) took place on the basis of only one surviving channel class.

Putative K⁺ transporters were identified using the conceptual proteomes of A. thaliana (TAIR10 Genome release), O. sativa spp. Indica, P. trichocarpa, P. patens, and S. moellendorffii (Phytozome v6.0) and the algae genomes Coccomyxa_C169, Micromonas RCC299, O. tauri (v2, v3, and v4 respectively²), V. carteri, and C. reinhardtii (Phytozome v8.0) by screening with different transporter class-specific protein motifs: three motifs for K⁺ channels ((1) [S,T]-x-x-T-x-G-[Y,F,L]-G-[D,E], (2) R-[L,F]-x-R-[L,V,I,A,G]-x-[R,C,K]-[V,A,L,M], (3) [A,V,S]-Y-[L,I]-[I,L]-G-[N,I]-[M,I]-T-[N,A]-L-[V,I]); two motifs for HKTs ((4) [S,T,A]-x-[F,Y,V,L,C]-x-[D,N,S]-G, (5) [G,A]-[Y,F]-[G,A]-x-[V,A,I]-G-[L,M,Y,F]-[S,T]); and five motifs for HAK transporters ((6) [A,G]-[D,S,G]-[V,L,I,M]-x-x-[S,A]-P-L-Y; (7) [A,G]-[N,D,H,S]-[D,N]-x-G-[E,Q,D,N]-[A,G]; (8) [A,G,S]-[D,N]-[G,S,A,C]-x-[L,I,V,F]-x-P-x-[V,I,L,M]-[A,S]; (9) G-[S,A,T,C]-E-[A,G]-x-[F,Y]-A-[D,N,E]-[L,I,V]-[G,C,S,A]-x-F; (10) [Y,F]-x-x-x-x-x-[H,F,Y]-G-Y-x-[E,D]) using the FUZZNUC program from EMBOSS (Rice et al., 2000). Additionally, results were checked against BLAST searches in the five genomes using known transporters of different classes from Arabidopsis and rice as templates. In order to eliminate false-positives the resulting raw-data were curated in a semi-automatic way. In a first step sequences with a length <70% of the average length between the outermost motifs in the corresponding Arabidopsis transporters were discarded. Subsequently, the remaining n protein sequences of each transporter type of each species were pairwise aligned using ClustalW2³. From the resulting n(n−1)/2 pairs those with a score of <20 and of 100 (identical sequences) were removed. The residual pairs fragmented the sequences into distinct groups. That group with the highest similarity to the corresponding Arabidopsis transporters was selected for further analyses.

To verify whether the screening for K⁺ channels in S. moellendorffii was exhaustive, its genome was screened in the six-frame translations using the program SIXPACK from EMBOSS (Rice et al., 2000) for the presence of the K⁺-selectivity filter motif G-Y-G in ORFs. Following a positive hit, the closer environment of the GYG was inspected manually for further characteristic sequence features allowing categorizing the peptide to be part of a K⁺ channel. As a result – besides the K⁺ channels obtained already in the first screening – only the K_out channel SmORK could be identified in addition.

Sequences from each family were aligned using MAFFT (Katoh and Toh, 2010), and alignments were filtered using GBlocks (Castresana, 2000) in order to eliminate regions of low quality. Briefly, the minimum number of sequences for a conserved position was half the number of sequences, the minimum number of sequences for a flanking position was half the number of sequences, the maximum length of contiguous non-conserved positions was 20, and the minimum length of a block was two, positions with gaps were not treated differently from other position. Evolutionary relationships were inferred by Maximum Likelihood using RAxML and 1000 bootstrap replicates (Stamatakis, 2006). The evolutionary model used for phylogenetic analyses was inferred using ProtTest (Darriba et al., 2011). For two pore channels and HAK transporters the model was LG + γ, for HKTs and Shaker-like channels it was JTT + γ. In order to root and resolve the gene trees we performed a gene tree-species tree reconciliation analysis using the species tree from Lang et al. (2010; TreeBase 10409). Reconciliation analysis was carried out in Notung 2.6 (Chen et al., 2000; Vernot et al., 2007). To get an idea of the phylogenetic structure of the other voltage-gated K⁺ channels displayed in Figure 5, the sequences were hierarchically clustered based on pairwise identities between two sequences using UPGMA (Unweighted Pair Group Method with Arithmetic Mean). UPGMA analyses were carried out in MAFFT⁴.