3.1.1. Potassium Channels
Potassium channels are proteins that facilitate the transport of K+
across membranes. They contain at least two TMs and a conserved sequence motif that acts as a filter for selecting K+
]. Potassium channels are the most abundant viral membrane transport proteins and they are detected in very different viruses and phages. All of them contain the highly conserved signature sequence G(Y/F)GD and at least two predicted TM domains. In addition to the already known K+
channels from the chloroviruses, prasinoviruses, the phaeovirus Ectocarpus siliculosus virus-1 (EsV-1), and two other phycodnaviruses (see Table S1
), four more phycodnavirus genes resembling K+
channels were found. One channel gene was detected in the recently described chlorovirus OSy-NE5 [33
] (94 amino acids (aa), 2 predicted TMs, accession #YP_009325597.1) and two genes were found in viruses from environmental samples, Dishui Lake phycodnavirus, and Yellowstone Lake phycodnavirus (116 aa, two predicted TMs, accession #AUT19143.1, and 96 aa, two predicted TMs, NCBI accession #YP_009174599.1). An additional gene was found in the newly discovered Tetraselmis virus-1 [34
] (TeV-1; 81 aa, two predicted TMs, accession #AUF82121.1). The alignment of the K+
channel consensus domain (Figure 1
, full alignment in Figure S1
) and the corresponding phylogenetic tree in Figure 2
show that the similarity to homologous K+
channel sequences was low for all four additional proteins. All channels contain the typical G(Y/F)G motive in the selectivity filter and a subsequent Asp, which is common in this position [35
]. Outside of this domain the channels exhibit some diversity from the canonical K+
channel consensus sequence although a Thr/Ser [35
] or a Thr/Val substitution has previously been reported for some other two pore domain K+
channels. Variability from canonical K+
channels is also apparent with respect to the pair of aromatic amino acids, which are generally present upstream of the selectivity filter in K+
]. They are not present in all sequences, however, and their absence does not prevent channel function in the case of the protein from virus MT325 [18
The blastp E values for the most similar hits were 4e−6 (OSy-NE5), 8e−8 (Dishui Lake phycodnavirus), 2e−7 (Yellowstone Lake phycodnavirus) and 2e−7 (TeV-1). The diversity of all phydcodnavirus K+
channels is high as indicated by the alignment in Figure S1
and the corresponding phylogenetic tree (Figure 2
). The proteins not only vary in size ranging from 79 to 156 amino acids in length, but also in overall amino acid composition.
Potassium ion channel encoding genes were also discovered in phage genomes. At least 1600 genomes of mycobacteriophages have been completely sequenced [37
] and in two of them we detected a K+
channel encoding gene: one in Mycobacterium phage Myrna (119 aa, two predicted TMs, accession #ACH62227.1) and one in Mycobacterium phage Phabba (119 aa, two predicted TMs, accession #ASZ74807.1). The two genes are very similar to each other (70% aa identity). The most similar non-viral protein is a protein, which is annotated as an “ion transport protein” from Acidobacteria bacterium
OLB7 (blastp E value 4e−19, 38% aa identity, accession #KXK00635.1).
At least nine Vibrio phages carry genes coding for a putative K+
channel protein with the consensus sequence of K+
channels (Figure 3
). The similarity in this domain and the agreement with the canonical K+
channel consensus sequence is quite high in this domain. Additionally, the typical aromatic amino acids upstream of the filter sequence [36
] are present. Diversity is only apparent in the first position of the consensus sequence, which is frequently, but not necessarily, a Thr in K+
channels. The phage proteins vary in size from 149 to 228 amino acids, but exhibit the overall architecture of K+
channel proteins (Table S3
and Figure S2A
). Interesting to note is that the putative channel protein of Vibrio phage phi-ST2 lacks the first TM domain. This does not exclude not per se a K+
channel function for these proteins. The conventional algorithms are optimized for predictions of structural domains in eukaryotes and prokaryotes and may fail to detect existing TM domains in viral proteins [39
]. We also found in the case of the chlorovirus AR158 Kcv that the protein was much shorter than its homologs and that the outer TM domain was lacking. We do not know how this finding fits into the general picture in which the viral channels seem to be essential for replication. At this point we speculate that the function of the truncated channel proteins might be restored by a separately, but unknown, virus or host encoded protein.
A blastp search of the Vibrio phage phi-pp2 protein against non-viral organisms revealed a putative K+
channel protein from a Moraxellaceae bacterium as the most similar protein (E value 2e−26, 44% aa identity; accession #PCJ41331.1). The similarity between the viral and bacterial protein however is rather low and, interestingly, the viral protein is longer than the bacterial one (228 vs. 158 aa; Figure S2B
Additional K+ channel genes were found in Acinetobacter phage vB_AbaM_ME3 (124 aa, 2 TMs, accession #AND75308.1) and Pseudomonas phage ventosus (117 aa, 2 TMs, accession #ATW58311.1). Both proteins slightly resemble a hypothetical protein from Rheinheimera sp. F8 (accession #ALZ75942.1). The blastp E values and aa identities were 2e−28 and 41% (Acinetobacter phage) and 1e−33 and 44% (Pseudomonas phage), respectively. A K+ channel gene was also discovered in the genome of Streptomyces phage BillNye (123 aa, 2 TMs, accession #AVD99322.1) with some similarity to a hypothetical protein from Pararheinheimera texasensis (Gammaproteobacteria; blastp E value 1e−19, 34% aa identity, accession #WP_031568736.1). Streptomyces belongs to the Actinobacteria and is not related to Pararheinheimera. Hence, there is no direct evidence that a Steptomycete is a host for the channel carrying phage.
In addition, one K+ channel gene was found in Lactobacillus phage PLE (259 amino acids, four predicted TMs, accession #YP_009282368.1). The blastp search revealed that it is identical to a putative ion transporter from the Lactobacillus casei group, thus, a potential host of the phage (accession #WP_012491300.1).
The phylogenetic relationship of canonical K+
channels from viruses and phages (blue) and sequences from cellular organisms (red) that were most similar to the viral sequences are shown in Figure 2
. The figure shows that K+
channels from different viruses show no obvious relationship. The relationship between the viral channel and non-viral K+
channels exhibit two different extremes. In one case the viral protein is identical to that of its host while in the other cases there is no apparent relationship between genes from virus/phages and their hosts.
Finally, it is worth noting that several proteins from viruses belonging to Mimiviridae are annotated as K+
channels (Table S4
). They vary in size from 103 to 127 aa, contain two predicted TMs and show some similarity to bacterial K+
channels. However, they do not contain the crucial signature sequence of canonical K+
channels. Due to the functional importance of this domain for K+
channel function [40
] we did not include these proteins in the current analysis.
The main novelty from the present analysis in relation to viral K+ channels is that they are not only present in viruses from algae but also in bacterial phages. The large diversity between the viral K+ channels argues against a common origin of the genes. A comparison between the viral K+ channels genes with those of the host does not provide an answer on their origin because we find genes, which are very similar but also very different from the respective host genes.
3.1.2. Chloride Channels
Chloride channels consist of a larger and very heterogeneous family of proteins, which catalyze the passive diffusion of Cl−
across membranes. The proteins share little structural and functional features except that they all exhibit a rather low selectivity among anions [41
Some aquatic herpesviruses contain genes encoding two types of chloride channels: CLIC (chloride intracellular channel)-like chloride channels and MCLC (mid-1-related chloride channel)-like chloride channels. Anguillid herpes virus 1 (AngHV-1), which infects the eel Anguilla anguilla,
carries a gene for a CLIC-like protein (281 aa, two predicted TMs, accession #YP_003358251). The most similar homologous proteins are from fish species, with a protein from Pundamilia nyererei
being the most similar (blastp E value 7e−79, 67% aa identity; accession #XP_005746869.1) (Figure 4
). No homologous sequence was found in the host genome possibly due to incomplete data. The homologous sequences from other organisms are much longer than the viral sequence (>530 aa) and contain at least one more predicted TM (Figure S3
Abalone herpesvirus carries a gene for a MCLC-like protein (YP_006908742, 333 aa, three predicted TMs). Homologous sequences were found in the genomes of a variety of aquatic animals such as mussels, worms, fish and corals (Table S5
) although with low similarity. The homologous sequences are of similar size or longer (265–685 aa) than the viral protein. Again there was no homolog in the host genome. The blastp search also revealed two sequences from Abalone herpes virus from two different locations and years of sequencing (YP_006908742.1 and AET44204.1) that differ in one aa (in position 304, Q and H, respectively) plus a partial sequence (ADL16677.1, 134 aa, identical to YP_006908742.1). Further, two more sequences were found from herpesviruses that infect other mussels (Chlamys acute necrobiotic virus, ADD24788.1 and Ostreid herpesvirus 1, YP_024600.1; 316 aa each). They differ in one aa in position 312, K and R, respectively.
The resume from a search on viral Cl- channels is that there are different members of this family of anion channels in some DNA viruses. It is notable that they do not coexist with K+ channels suggesting that an efflux of KCl salt is not a functional feature.
3.1.3. Ligand-Gated Ion Channels
Ligand-gated ion channels are activated by various molecules such as glycine and glutamate. They vary in size and structure. For example, human glutamate receptors are trimeric and are activated by glutamate, while glycine receptors are pentamers and are activated by glycine and other similar small aas [42
Chlorovirus PBCV-1 (Phycodnaviridae
) and the closely related OSy-NE5 virus carry a gene for a putative glutamate receptor (411 aa, three TMs, NP_048510 and 406 aa, three TMs, YP_009325583.1, respectively; Figure S4A,B
). No other viruses have a similar gene. The similarity to other glutamate receptors is small, however. The most similar protein is one annotated as ‘glutamate (NMDA) receptor subunit epsilon-2′ from Cricetulus griseus
(E value 0.011, aa identity 22%; accession #ERE 66598.1). The viral protein is much smaller (411 aa vs. 1121 aa) and lacks a conserved n
-aspartate receptor 2B3 C-terminus domain, which is present in the mammalian homolog (Figure S4A,B
]). This large hydrophilic extra-cellular domain is typical of NMDA receptor proteins and presumably involved in receptor function. Apart from this domain the general architecture of the pore region in the viral protein slightly resembles a mammalian glutamate receptor (Figure S4B
). Attempts to activate the channel after expression in Xenopus
oocytes with glutamate failed. However when we measured channel activity in HEK293 cells expressing the protein we recorded single channel fluctuations with a conductance of ca. 20 pS. These fluctuations, examples are shown in Figure 5
, were only seen in recordings when the pipette solution contained 1 mM of glycine (n
= 5). In the absence of glycine (n
= 5) or with 1 mM glutamate in the pipette (n
= 5) these channel openings were not seen. This suggests that they originate from a glycine activation of the viral protein. An activation of this type of ligand activated channels is not completely surprising considering that channels with a similar architecture from plants are activated by glycine, but not by glutamate [43
Several chlorovirus genomes contain a gene that is annotated as a ligand-gated ion channel, e.g., ORF A163L from PBCV-1 (433 aa, three predicted TMs, accession #NP_048511). The proteins are weakly similar to a hypothetical protein from the plant Vigna angularis (blastp E value 0.017, 24% aa identity, accession #KOM43022.1) and a “glutamate receptor 2.8-like” protein from the plant Vigna radiata var. radiata (blastp E value 0.025, 25% aa identity, accession #XP_014499575.1). There is no indication that these two plants are hosts for these viruses.
The interesting conclusion, which can be drawn from a screening of viral genomes for ligand gated channels, is that distantly related members of this family of channels are present in a few viral genomes. The genes products seem to be functional and indeed gated by ligands. The physiological function of the viral channels remains unknown. However, their low abundance in viral genomes suggests that they may not be as essential as K+ channels.
3.1.5. Mechanosensitive Channels
Mechanosensitive channels are predominantly present in bacteria but they also exist in eukaryotes where they facilitate a rather nonselective flux of ions. The name refers to their type of gating; they are activated by mechanical changes to the membrane [45
Aureococcus anophagefferens virus has a gene coding for a 261-aa long putative mechanosensitive channel that contains three predicted TMs (accession #YP_009052121.1). It is most similar to a “mechanosensitive ion channel protein MscS” from Nesiotobacter exalbescens. However, the similarity is low (blastp E value 8e−10, 25% aa identity, accession number WP_028481655.1) and the bacterial protein is much larger (843 aa vs. 261 aa).
An even smaller putative mechanosensitive channel was detected in Tetraselmis virus 1. The 101 aa long protein contains one to two predicted TMs (accession #AUF82136.1) and is most similar to a ‘large conductance mechanosensitive channel protein’ from a bacterium in the Parcubacteria group. Again, the similarity is rather low (blastp E value 4e−5, 29% aa identity, accession #KKP92070.1), but the proteins are similar in size (127 aa vs. 101 aa).
A protein of the same size was also detected in the genome of the Cafeteria roenbergensis virus BV-PW-1 (accession #YP_003969926.1). It has two predicted TMs and has low similarity to a putative large-conductance mechanosensitive channel from an Arc I group achaeon (blastp E value 1e−4, 27% aa identity, accession #KYC46208.1). Interestingly, the three viral proteins are not similar to each other (Figure S5
). This suggests at least three independent transfers of mechanosensitive ion channel genes into a viral genome; however, evidence for a function for these proteins is still missing. In addition, the low similarities to proteins from cellular organisms do not provide much information on a potential origin of the genes.
Our databank screening confirms that different types of mechanosensitive channels are present in viral genomes. This includes the large mimiviruses but also the much smaller Tetraselmis virus, which has an alga as host. The question as to whether the putative channels are active and whether they play a role in viral replication remains unanswered.