Two novel types of hexokinases in the moss Physcomitrella patens
© Nilsson et al; licensee BioMed Central Ltd. 2011
Received: 19 October 2010
Accepted: 14 February 2011
Published: 14 February 2011
Skip to main content
© Nilsson et al; licensee BioMed Central Ltd. 2011
Received: 19 October 2010
Accepted: 14 February 2011
Published: 14 February 2011
Hexokinase catalyzes the phosphorylation of glucose and fructose, but it is also involved in sugar sensing in both fungi and plants. We have previously described two types of hexokinases in the moss Physcomitrella. Type A, exemplified by PpHxk1, the major hexokinase in Physcomitrella, is a soluble protein that localizes to the chloroplast stroma. Type B, exemplified by PpHxk2, has an N-terminal membrane anchor. Both types are found also in vascular plants, and localize to the chloroplast stroma and mitochondrial membranes, respectively.
We have now characterized all 11 hexokinase encoding genes in Physcomitrella. Based on their N-terminal sequences and intracellular localizations, three of the encoded proteins are type A hexokinases and four are type B hexokinases. One of the type B hexokinases has a splice variant without a membrane anchor, that localizes to the cytosol and the nucleus. However, we also found two new types of hexokinases with no obvious orthologs in vascular plants. Type C, encoded by a single gene, has neither transit peptide nor membrane anchor, and is found in the cytosol and in the nucleus. Type D hexokinases, encoded by three genes, have membrane anchors and localize to mitochondrial membranes, but their sequences differ from those of the type B hexokinases. Interestingly, all moss hexokinases are more similar to each other in overall sequence than to hexokinases from other plants, even though characteristic sequence motifs such as the membrane anchor of the type B hexokinases are highly conserved between moss and vascular plants, indicating a common origin for hexokinases of the same type.
We conclude that the hexokinase gene family is more diverse in Physcomitrella, encoding two additional types of hexokinases that are absent in vascular plants. In particular, the presence of a cytosolic and nuclear hexokinase (type C) sets Physcomitrella apart from vascular plants, and instead resembles yeast, where all hexokinases localize to the cytosol. The fact that all moss hexokinases are more similar to each other than to hexokinases from vascular plants, even though both type A and type B hexokinases are present in all plants, further suggests that the hexokinase gene family in Physcomitrella has undergone concerted evolution.
Hexokinases catalyze the first step in hexose metabolism, the phosphorylation of glucose and fructose. Hexokinases that show a higher specificity for glucose than for fructose are sometimes called glucokinases. The yeast Saccharomyces thus has a glucokinase, ScGlk1, and two dual specificity hexokinases, ScHxk1 and ScHxk2. The eukaryotic hexokinases are all related to each other, but are unrelated to prokaryotic glucokinases and hexokinases. Plants also have a fructokinase which is unrelated to the hexokinases [1–3].
Hexokinases are found in several different intracellular locations. The three yeast hexokinases are cytosolic, but ScHxk2 can also enter the nucleus . Animal type I and II hexokinases have hydrophobic N-termini that target them to the outer mitochondrial membrane, whereas type III and IV hexokinases are cytosolic, but the latter can also enter the nucleus . We have previously described two types of plant hexokinases . Type A is exemplified by the Physcomitrella hexokinase PpHxk1, a soluble protein with a transit peptide  that localizes to the chloroplast stroma. Type B hexokinases exemplified by PpHxk2, have N-terminal membrane anchors. Both types are present also in vascular plants, where they localize to the chloroplast stroma and to the outer mitochondrial membrane, respectively [7–14].
In addition to their metabolic roles, eukaryotic hexokinases have also been implicated in signal transduction. Mitochondria-associated hexokinases have thus been shown to negatively affect programmed cell death in both animals and plants, by preventing the release of cytochrome c from mitochondria [14–17]. It should be noted, however, that this does not prove that a signal is transmitted by hexokinase, which could have a constitutive inhibitory effect on cytochrome c release. A more direct role for hexokinases in signal transduction is suggested by studies of the response to glucose in several organisms. Thus, early work in yeast showed that ScHxk2 is required for glucose repression [18, 19], but the molecular mechanism has resisted analysis for more than 30 years [20–23].
An important question is where the enzyme exerts its signaling function. Early work in yeast focused on the cytosol, since the yeast hexokinases are cytosolic. However, further studies have shown that ScHxk2 also can translocate into the nucleus, where it forms a complex with the Mig1 repressor [4, 24]. Similarly, evidence from Arabidopsis  and rice [26, 27] suggest that plant type B hexokinases may enter the nucleus and participate in gene regulation.
The moss Physcomitrella patens is unique among plants in that gene targeting by homologous recombination works in it with frequencies comparable to yeast . This has made Physcomitrella a powerful model system for studies of plant gene function [29, 30]. The recent sequencing of the Physcomitrella genome has further strengthened it as a model plant . We have previously characterized the Physcomitrella hexokinase PpHxk1, which by gene targeting was shown to account for 80% of the glucose phosphorylating activity in protonemal tissue . Further studies of a PpHxk1 knockout mutant revealed a number of interesting phenotypes, but no conclusive evidence was obtained as to the possible role of this hexokinase in signaling . Part of the problem is that Physcomitrella like other plants possesses several hexokinases, which makes it difficult to draw conclusions about gene function from the knockout of a single gene.
We here report the characterization of all eleven genes encoding putative hexokinase proteins in the Physcomitrella genome. Seven of the genes predict proteins that clearly belong to the previously described types A and B . However, the remaining four genes encode two novel types of hexokinases, which we call C and D. The type C hexokinase PpHxk4 is a soluble protein which lacks both organelle targeting peptide and membrane anchor. The three type D hexokinases PpHxk9, PpHxk10 and PpHxk11 resemble the type B hexokinases in that they possess hydrophobic membrane anchors, but differ in sequence from the latter. The type D hexokinases also have a similar localization as the type B hexokinases, being found in the outer mitochondrial membrane, and to some extent in the chloroplast envelope.
The growth conditions used were growth at 25°C under constant light in a Sanyo MLR-350 light chamber with irradiation from the sides. Light was supplied from fluorescent tubes (FL40SS W/37, Toshiba) at 30 μmol m-2 s-1. Subculturing of Physcomitrella patens protonemal tissue was done on cellophane overlaid 0.8% agar plates containing BCD media (1 mM MgSO4, 1.85 mM KH2PO4, 10 mM KNO3, 45 μM FeSO4, 1 mM CaCl2, and trace elements ), supplemented with 5 mM ammonium tartrate.
In the same degenerative polymerase chain reaction (PCR) where we isolated PpHXK1 we also found several other hexokinase encoding sequences . From these, we could design primers to amplify full length cDNAs and genes of PpHXK2 and PpHXK3 (Additional files 1 and 2: Tables S1 and S2). The sequences of PpHXK1 and PpHXK2 were then used to search the PHYSCObase EST data base  for more hexokinase sequences. Based on the sequences found, primers were designed to amplify the PpHXK4 gene and cDNA and the PpHXK5 gene. Several of the partially sequenced EST clones identified in the PHYSCObase were also ordered from the RIKEN bioresource center and fully sequenced (Additional file 1: Table S1). When the sequence of the Physcomitrella patens genome became available , we searched it for additional hexokinase encoding sequences. This revealed six more putative hexokinase genes: PpHXK6-PpHXK11. Primers were designed to amplify genes and cDNAs of these hexokinases (Additional files 1 and 2: Tables S1 and S2).
In our localization studies we used the Green Fluorescent Protein (GFP) from the vector psmRS-GFP, a pUC118 based plasmid with the 35S promoter in front of a soluble modified red shifted GFP followed by the NOS1 terminator . Primers ending with BamHI or BglII sites were designed to facilitate sticky end ligation of PCR products into the BamHI site between the 35S promoter and the rsGFP coding region (Additional file 2: Table S2). GFP fusions were made for all eleven Physcomitrella hexokinases. For PpHXK2, 3, 4, and 7 the full length cDNAs were fused in frame to GFP, but for PpHXK5, 8, 9, 10 and 11 partial cDNAs were used since no full length cDNAs were available. No cDNA was available for PpHXK6, so the first exon amplified from the genomic DNA was used to construct a GFP fusion in that case. For all hexokinases two different versions of the hexokinase-GFP fusions were made: one containing the N-terminal membrane anchor or chloroplast transit peptide and one in which the membrane anchor or chloroplast transit peptide had been deleted (Additional file 3: Table S3). For PpHxk10 a hexokinase-GFP fusion was also made where the membrane anchor of PpHxk10 was fused directly to GFP.
GFP fusion constructs were transiently expressed in wild type protoplasts after PEG-mediated transformation . The transformed protoplasts were analyzed after one to two days of incubation in the dark in a Zeiss Axioskop 2 mot fluorescence light microscope equipped with either a HR or MRm AxioCam camera from Zeiss. The GFP signal was detected using a FITC filter (excitation 480 nm, emission 535 nm, dichronic beamsplitter 505 nm) while chloroplast autofluorescence was detected using a TRITC filter (excitation 535 nm, emission 620 nm, dichronic beamsplitter 565 nm). The mitochondria specific dye MitoTracker® Orange was detected with Zeiss filter set number 20 (excitation 546/12 nm, emission 575-640 nm, dichronic beamsplitter 560 nm). The nucleic acid stain 4',6-diamidino-2-phenylindole dihydrochloride (DAPI) was used to visualize the nucleus and detected using a DAPI/Hoechst filter (excitation 360 nm, emission 460 nm, dichronic beamsplitter 400 nm).
A yeast strains with triple knockouts of the HXK1, HXK2 and GLK1 genes in the W303-1A background  was kindly provided by Stefan Hohmann . Hexokinase-encoding cDNA sequences from Physcomitrella were cloned into the high copy number 2 μm URA3 shuttle vector pFL61 , which expresses inserts in yeast from the constitutive PGK promoter (Additional files 2 and 3: Tables S2 and S3). Transformants were selected on synthetic media lacking uracil, with 2% galactose as carbon source in order to permit hexokinase deficient strains to grow. Colonies were picked to synthetic galactose plates lacking uracil, and the resulting grids were replicated to synthetic media lacking uracil and containing different carbon sources. Growth was scored after 6 days at 30°C.
The Vector NTI software package with ContigExpress (Invitrogen) was used for sequence editing, sequence analysis and building of contigs. The sequence of PpHxk1 differs in one position (leucine-55) from the published sequence  due to a sequence error that has now been corrected in GenBank. For the tree-building, we used the Neighbour-Joining method  as previously described .
The well-conserved protein sequences and the presence of cDNAs for most of the genes among our PCR products and in public EST data bases [34, 42] suggest that they encode functional products which are expressed in protonemal tissue. The only possible exception is PpHXK6, for which no transcript has been found. However, for four of the genes, PpHXK5, PpHXK9, PpHXK10 and PpHXK11, only aberrantly spliced transcripts causing premature termination have been sequenced. It should be noted that two other genes, PpHXK3 and PpHXK7, had both correctly and incorrectly spliced transcripts. This suggests that alternative splicing is common, and that correctly spliced products therefore could exist also for the four aberrantly spliced genes. A sequence analysis of the genes does not suggest that any of them is a pseudogene, since both predicted protein sequences and other important features such as consensus sites for splicing are well conserved. The only possible exception is PpHxk11 which has a few amino acid substitutions in positions suggested to be important for catalytic activity (see below).
We have previously classified plant hexokinases into two types  depending on their N-terminal sequences which contain either chloroplast transit peptides (type A) or hydrophobic membrane anchors (type B). Furthermore, the membrane anchors of the type B hexokinases are highly conserved between different plant species, suggesting a common evolutionary origin for this sequence . Most of the Physcomitrella hexokinases belong to the two previously described types. Thus, in addition to PpHxk1, two more type A hexokinases are encoded by PpHXK5 and PpHXK6. Based on the sequences, PpHxk6 appears to be more closely related to PpHxk1 than PpHxk5. Four of the predicted Physcomitrella hexokinases, PpHxk2, PpHxk3, PpHxk7 and PpHxk8 have N-terminal membrane anchors similar to the N-termini of type B hexokinases from other plants.
However, some of the Physcomitrella hexokinases do not conform to the criteria that we used to define types A and B. One hexokinase, PpHxk4, has a truncated N-terminus without either a membrane anchor or an organelle import peptide. We will refer to this novel type as a type C hexokinase. Interestingly, no hexokinase with a truncated N-terminus is encoded by the Arabidopsis genome. The rice genome predicts two hexokinases with truncated N-termini, the OsHXK7 and OsHXK8 gene products , but their N-terminal sequences do not resemble PpHxk4. Instead, they look like truncated type B hexokinase membrane anchors, with most of the twelve first amino acid residues being alanines or valines.
Predicted intracellular locations and transmembrane helices of moss hexokinases
A number of conserved sequence motifs and structurally or functionally important amino acid residues have been identified by x-ray crystallography and comparisons of hexokinases from different organisms. Bork et al. [45, 46] described seven conserved regions in hexokinases which they named phosphate 1, sugar binding, connect 1, phosphate 2, helix, adenosine and connect 2, based on the known or suspected functions of these regions. Kuser et al. ( Table II) identified 20 amino acid residues that are highly conserved in 317 hexokinases. Mutational and structural studies have shown that the catalytic residue is an aspartic acid (D211 in the yeast hexokinase ScHxk2) whereas four other residues (S158, K176, E269 and E302 in ScHxk2) contribute to hexose binding .
First, we note that the catalytic aspartic acid is strictly conserved in all eleven Physcomitrella sequences, as are all but one hexose binding residue. The only exception is the K176 in ScHxk2, which is replaced by a glutamic acid in PpHxk11. As for the 20 most conserved residues , we note that 19 of them are strongly conserved in all plant hexokinases (the exception is C268 in ScHxk2). Interestingly, these 19 residues are strictly conserved in all Physcomitrella sequences except PpHxk11, which has four substitutions (Additional file 5: Figure S1). For comparison, we note that the highly divergent catalytically inactive AtHkl3 protein  has 12 substitutions in these 19 positions. This includes the catalytic aspartic acid, which is an asparagine in AtHkl3, and two of the hexose binding residues. The less divergent AtHkl1 and AtHkl2 proteins, also thought to be catalytically inactive, have two and three substitutions, respectively, in the 19 conserved residues, none of which involve the catalytic or hexose binding residues.
An inspection of the seven regions described by Bork et al.  shows that they all are well conserved in the Physcomitrella proteins (Additional file 5: Figure S1). There are however, some noteworthy exceptions. First, the type D hexokinases share several substitutions in the conserved regions which are not found in any other hexokinases. Thus, they have a cysteine followed by a leucine in the phosphate 1 motif where most other hexokinases have a valine followed by a glutamine. Furthermore, a phenylalanine in the sugar binding motif, which is strictly conserved in all other hexokinases, is replaced by a leucine in the three type D proteins. Finally, the latter also share a deletion of two residues at the end of the phosphate 2 motif which is not found in any other hexokinases. None of these changes involve residues shown to be critical for catalytic activity, but it is still possible that they could affect the activity and/or substrate specificity of the type D proteins. In addition to these changes, PpHxk11 has several more substitutions in the conserved regions, consistent with its generally more divergent sequence. Finally, we note that all Physcomitrella hexokinases have an insertion in the adenosine motif, which is found also in other plant hexokinases [7, 13].
In order to gain a better understanding of how the different hexokinases are related to each other, we used the predicted sequences of the Arabidopsis, rice and Physcomitrella hexokinases to construct an evolutionary tree. We limited the analysis to these three plant species since their genome sequences have been completed and since the rice and Arabidopsis hexokinases already have been fairly well studied [7, 13, 49]. The variable N-termini and C-termini were excluded from the analysis in order to avoid ambiguities in the sequence alignment, and to ensure that the result would be independent of the N-termini, thus making it possible to assess to what extent the latter have co-evolved with the rest of the proteins (Additional file 5: Figure S1).
The resulting tree is shown in Figure 3. Surprisingly, we found that all eleven Physcomitrella hexokinases are more closely related to each other than to other plant hexokinases, thus forming a single branch within the tree. This was unexpected since the Arabidopsis and rice sequences do not cluster in this way, but instead are interspersed (Figure 3). This is particularly evident in the case of the type A hexokinases, where the single proteins present in Arabidopsis (AtHxk3) and rice (OsHxk4) are more similar to each other than to the other Arabidopsis and rice hexokinases (Figure 3). In contrast, the three type A hexokinases in Physcomitrella, PpHxk1, PpHxk5 and PpHxk6, are more similar to the other Physcomitrella hexokinases than to their orthologs AtHxk3 and OsHxk4. We conclude from this that the Physcomitrella hexokinases show evidence of concerted evolution, unlike the Arabidopsis and rice proteins.
It should further be noted that within the Physcomitrella sequences, the four above described hexokinase types form well-defined branches suggesting a distinct origin for each type. Thus, the three type D hexokinases are clearly more closely related to each other than to the four type B hexokinases, and vice versa. This suggests that each type of hexokinases arose from a single ancestral gene, which subsequently underwent duplications. This interpretation is further confirmed by the fact that the moss type B hexokinases have lost intron 2, which is present in the other moss hexokinases, including the type D hexokinases (Figure 1). Finally, we note that the sequence of the type C hexokinase, PpHxk4, is more distantly related to the other Physcomitrella hexokinases than they are to each other. This suggests that the type C hexokinase may represent an early branch on the tree, which has been lost in seed plants.
The three type D hexokinases PpHxk9, PpHxk10 and PpHxk11 also possess membrane anchors and show a similar, though more restricted localisation as the type B proteins. Thus, both PpHxk9 and PpHxk11 localize to the outer mitochondrial membrane, but only PpHxk11 is also associated with the chloroplast envelope, like the type B hexokinases (Figures 4, 5, 6). For PpHxk10, we were unable to clone a PpHXK10 full length transcript that was correctly spliced. We therefore made two incomplete PpHxk10-GFP fusions: one containing the entire region encoded by the first exon including the membrane anchor (Figure 4) and one containing the membrane anchor alone. Both fusions localized throughout the cytosol. It is, however, possible that these partial fusions are incorrectly folded due to the hydrophobic nature of the membrane anchor, and that the targeting signal is thus not functional. We cannot therefore rule out that a full-length fusion of PpHxk10 to GFP would localize to the outer mitochondrial membrane, similar to PpHxk9 and PpHxk11.
The PpHxk5-GFP and PpHxk6-GFP fusions, finally, had localizations resembling that of PpHxk1 . Thus, we found that they are imported into the chloroplast stroma (Figures 4 and 6). Truncated versions of PpHxk5-GFP and PpHxk6-GFP lacking the transit peptide were evenly distributed in the cytosol, similar to GFP expressed alone (Additional file 6: Figure S2). We conclude that chloroplast import of PpHxk5 and PpHxk6 is dependent of their N-terminal transit peptides, similar to PpHxk1 . Interestingly, a PpHxk5-GFP fusion with a shorter N-terminal truncation of amino acid residues 1-18 is still imported into the chloroplasts, so the targeting information is not immediately adjacent to the N-terminal end of PpHxk5 (Additional file 6: Figure S2).
During the sequencing of the cDNA and genomic clones we found a polymorphism in an AG microsatellite repeat in the 5'-untranslated region of the PpHXK3 gene. The two cDNAs that were sequenced differ by one AG (Additional file 7: Figure S3a), with the shorter variant being present in our genomic clone. We first considered the possibility that two duplicated genes might exist which differ only in this repeat. However, we saw no evidence of this, and only one PpHXK3 gene is found in the genome sequence . Interestingly, this gene has the longer variant, unlike our genomic clone. This made us consider the possibility that loss of one AG may have occurred recently in our moss line, which would still be heterogeneous for this mutation, thus explaining the two cDNAs. To test this we cloned two new PCR fragments from the 5'-untranslated region of the PpHXK3 gene. Significantly, we found that one has the extra AG and one does not, thus confirming the presence of a polymorphism in our genomic DNA. Two polymorphisms involving microsatellite repeats were also seen in PpHXK2, though we did not investigate these as carefully as the mutation in PpHXK3. We conclude that sequence evolution by acquisition or loss of microsatellite repeats seems to occur very rapidly in Physcomitrella. This could be a consequence of the high frequency of homologous recombination, since unequal sister chromatid exchange and gene conversion, both of which depend on homologous recombination, can generate this kind of polymorphisms.
We found at least one cDNA for ten of the eleven hexokinases in Physcomitrella, the only exception being PpHXK6. When the cDNA clones were sequenced and compared to other plant hexokinases we found several unexpected splice variants (Additional file 8: Table S5). Thus, we found both intron retention and exon skipping but the most frequent mode of alternative splicing was the use of alternative donor and/or acceptor sites. Most of these aberrantly spliced cDNA sequences would not encode functional hexokinases due to premature termination. The most interesting exception is the PpHXK7 cDNA clone pdp03464 that was obtained from the RIKEN bioresource center . PpHxk7 is a type B hexokinase with an N-terminal membrane anchor, but the anchor is not encoded by the alternatively spliced pdp03464 clone (Additional file 7: Figure S3b). In the resulting transcript, the predicted protein instead starts with the methionine codon at position 64. This truncated protein is likely to be functional since the deletion does not affect the phosphate, sugar or adenosine binding domains. Interestingly, we also cloned a normally spliced cDNA from PpHXK7 (Additional file 9: Table S6) which encodes a protein with an N-terminal membrane anchor (Additional file 7: Figure S3b). It thus appears that alternative splicing produces two PpHxk7 proteins, one with a membrane anchor and one without it.
Significantly, we found that the splice variant without a membrane anchor, PpHxk7a, localizes to the cytosol and in particular to the nucleus (Figure 7), whereas PpHxk7b localizes to mitochondrial membranes (Figure 7), consistent with the presence of a membrane anchor in that splice variant. It is therefore possible that the PpHxk7a splice variant could be involved in gene regulation. In this context, it should be noted that an artificial deletion of the membrane anchor in the two rice type B hexokinases OsHxk5 and OsHxk6 changed their localization to the nucleus, due to the presence of a cryptic nuclear localization sequence in these proteins . No obvious nuclear localization signal was found in PpHxk7a, but its nuclear localization could be the result of passive diffusion, as is seen also for GFP alone . Our finding suggests the interesting possibility that similar splice variants may exist for type B hexokinases in other plants, and that alternative splicing could provide a general mechanism by which type B hexokinases may enter the nucleus and affect gene expression.
We have previously reported that the major hexokinase in Physcomitrella, PpHxk1, which accounts for 80% of the glucose phosphorylating activity, is a novel type of plant hexokinase that is targeted to the chloroplast stroma . We have now extended our study of the hexokinase gene family in Physcomitrella by the cloning and characterization of ten new putative hexokinase-encoding genes (Figure 1). An inspection of the encoded protein sequences, in particular the N-termini (Figure 2), suggests that they represent several different types of hexokinases which are targeted to different intracellular compartments.
Four of the moss hexokinases, PpHxk2, PpHxk3, PpHxk7 and PpHxk8, are clearly type B hexokinases since they have N-terminal membrane anchors that are similar in sequence to those found in other type B hexokinases . We therefore expected that GFP fusions to these proteins would localize to mitochondria, as do several type B hexokinases in seed plants [7, 8, 10, 13]. We found that they indeed localize to the outer mitochondrial membrane, but surprisingly, also to the chloroplast envelope (Figs. 5 and 6). A similar dual localization was also seen for one type D hexokinase (PpHxk11). We note that subcellular fractionation data suggested that the spinach type B hexokinase SoHxK1 localizes to the chloroplast envelope , but more recent results with GFP fusions suggested that this is not the case . Dual targeting of proteins to mitochondria and chloroplasts has been described in Physcomitrella and several other plants, but only for proteins that are targeted to the interior of the organelles [64–66].
PpHxk5 and PpHxk6 are type A hexokinases as evidenced both from the sequence of their N-termini, which resemble organelle import peptides and from the fact that they are closely related to PpHxk1 (Figure 3). Consistent with this, we found that both the PpHxk5-GFP and PpHxk6-GFP fusions localize to the chloroplast stroma (Figure 6). This was surprising in view of our previous finding that a knockout of the PpHXK1 gene eliminates all glucose phosphorylating activity in the chloroplast stromal fraction . One possible explanation could be that PpHxk5 and PpHxk6 are minor hexokinases, which contribute only a small part of the total activity, or that they preferentially phosphorylate fructose, an activity which was not abolished in the absence of PpHXK1 . An alternative explanation could be that neither PpHxk5 nor PpHxk6 is expressed in the young protonemal tissue used for the subcellular fractionation experiments . Yet another possibility could be that PpHxk5 and PpHxk6 lack hexokinase activity and instead have some other function, as has been suggested to be the case for the Hkl1-3 proteins in Arabidopsis [13, 67]. Finally, it is possible that PpHXK5 and PpHXK6 are pseudogenes, since we have not been able to clone any cDNA from PpHXK6 and the two cDNAs from PpHXK5 that we sequenced were not correctly spliced. However, we think this is unlikely, since PpHXK5 and PpHXK6 show no other signs of being pseudogenes.
PpHxk4 represents a novel type of hexokinase, distinct from both types A and B, which we call type C. We base this distinction on two facts. First, the sequence of PpHxk4 shows that it is more distantly related to the type A and B hexokinases in Physcomitrella than the latter are to each other. PpHxk4 is thus clearly a separate type of hexokinase. In addition, the truncated N-terminus of PpHxk4 differs both from the membrane anchors found in type B hexokinases and from the organelle targeting peptides found in type A hexokinases. The predicted location of PpHxk4 is in the cytosol, since this is where a protein will end up in the absence of a targeting peptide. Consistent with this, we found that a PpHxk4-GFP fusion shows a diffuse fluorescence throughout the cell, indicating a primarily cytosolic, but also nuclear localization (Figure 7). We further note that PpHxk4 occupies a basal position among the Physcomitrella hexokinases, being the most divergent member of the protein family (Figure 3). This suggests that it may represent an ancient type of hexokinase that has been lost in seed plants. Interestingly, while no orthologue of PpHxk4 appears to be present in seed plants, the two rice hexokinases OsHxk7 and OsHxk8 also have truncated N-termini, and OsHxk7 was shown to have a cytosolic localisation . However, the N-termini of OsHxk7 and OsHxk8 look more like truncated versions of the type B hexokinase membrane anchor, with most of the twelve first amino acid residues being hydrophobic. Furthermore, OsHxk7 and OsHxk8 also group together with the type B hexokinases in the phylogenetic tree (Figure 3). Thus, they seem to be divergent type B hexokinases rather than orthologs of the Physcomitrella type C hexokinase. Still, it is conceivable that OsHxk7 and OsHxk8 may have a function in rice which is analogous to that of PpHxk4 in moss.
Our finding of a new type of cytosolic hexokinase in Physcomitrella is interesting in view of the proposed role of plant hexokinases in glucose sensing and signaling [21, 61, 68–70]. These discussions have so far focused on membrane integrated type B hexokinases such as AtHxk1 and AtHxk2, which was the only type of plant hexokinase that had been studied prior to the discovery of the type A hexokinases . It has recently been shown that some type B hexokinases can translocate into the nucleus and affect gene expression [25, 27]. However, it is not clear how these membrane anchored hexokinases are released from their membrane association and translocated into the nucleus. In contrast, the type C hexokinase PpHxk4, which lacks membrane anchor and is a soluble protein, could more easily move into the nucleus.
In this context, we note that we found evidence that moss type B hexokinases also may translocate to the nucleus. Thus, PpHXK7 encodes two differently spliced cDNAs, one of which is missing the membrane anchor (Additional file 7: Figure S3b). The intracellular localization of these two proteins is also very different as seen from the expression of the translational fusions with GFP. In protoplasts expressing the PpHxk7a splice variant that lacks the membrane anchor, the fluorescence is thus localized throughout the cytosol but is also associated with the nucleus (Figure 7). This suggests that alternative splicing could be a molecular mechanism whereby membrane bound type B hexokinases, perhaps also in other plants, may become soluble and thus exert a function inside the nucleus.
The other new type of plant hexokinase, type D, appears to have a similar localization as the type B hexokinases, i.e. in the outer mitochondrial membrane and also to some extent in the chloroplast envelope. However, they differ from the type B hexokinases in the sequences of their membrane anchors (Figure 2), and form a distinct clade in the evolutionary tree (Figure 3). Furthermore, they do not share the fusion of exons 2 and 3 that is found in all moss type B hexokinases (Figure 1). Still, the overlapping localizations, and the fact that the type D hexokinases also have membrane anchors, suggests that the type B and D hexokinases may have similar functions. In this context it should be noted that the type B hexokinases is a large and diverse group in seed plants, and that some members have N-termini that are less well conserved (labelled BX in Figure 3). Thus, while no obvious orthologues of the type D hexokinases exist in seed plants, it is conceivable that the more divergent members of the type B group may perform an analogous function as the type D hexokinases do in moss.
It has been proposed that some hexokinases may lack catalytic activity but still have other functions, based on data in fungi , flies  and plants [13, 67]. In particular, three of the six predicted hexokinases in Arabidopsis, AtHkl1-AtHkl3, appear to lack glucose phosphorylating activity but are conserved between A. thaliana and A. lyrata, which suggests that they still are under selection . On the other hand, all ten hexokinases in rice could complement a hexokinase-deficient yeast strain, indicating that they are catalytically active . This raises the question whether non-enzymatic hexokinases are peculiar to Arabidopsis or more widespread in plants. The knockout phenotypes and enzymatic activities of PpHxk2-PpHxk11 remain to be determined, but we note that PpHxk2-PpHxk10 are as strongly conserved as PpHxk1, which is an active enzyme , suggesting that they also may be active. Consistent with this, we found that PpHxk3 can complement a hexokinase deficient yeast strain (Figure 7). In contrast, PpHxk11 has several substitutions which could affect its activity. With four substitutions in the 19 most conserved residues it is not as divergent as AtHkl3 (12/19), but instead resembles AtHkl1 (3/19) and AtHkl2 (2/19). This is also evident from the tree in Figure 3 where AtHkl3 has a very long branch, whereas AtHkl1, AtHkl2 and PpHxk11 have much shorter branches. It is therefore conceivable that PpHxk11 could have a non-enzymatic function, perhaps in regulation or signaling, as has been suggested for the AtHkl proteins .
Surprisingly, we found that the eleven Physcomitrella hexokinases are more closely related to each other than to any other plant hexokinase, despite the fact that they represent different types of hexokinases, some of which are found also in seed plants (Figure 3). This is in contrast to the situation in seed plants, where hexokinases of the same type from different plants typically are more closely related to each other than hexokinases of different types from the same plant (ref.  and Figure 3). There are at least two possible explanations for this. One is that the different hexokinases in Physcomitrella originated by gene duplications after the separation of mosses from seed plants. This is the most straightforward interpretation of the tree in Figure 3, but we do not think that this is a likely explanation since the sequence of the membrane anchor in the type B hexokinases is highly conserved between seed plants and Physcomitrella (Figure 2). This suggests a common origin for the latter, since it is unlikely that this unique sequence would have been created twice in evolution just by chance.
A more likely explanation is therefore that several genes encoding different types of hexokinases were present already in the common ancestor of mosses and seed plants, and that these genes co-evolved in Physcomitrella by gene conversion , making them appear to be more closely related to each other than they really are. It has already been noted that tandemly arrayed genes in Physcomitrella are highly similar in sequence, which suggests that they may undergo concerted evolution by gene conversion . The hexokinase genes are not tandemly arrayed, in fact they are all located on different scaffolds in the draft sequence of the Physcomitrella genome , and those scaffolds that could be linked to the genetic map  were all in different linkage groups. However, work in yeast has shown that gene conversion also can be ectopic, i. e. take place between related genes on different chromosomes . Such ectopic gene conversion could have provided a mechanism by which the moss hexokinases co-evolved. A testable prediction of this hypothesis is that other dispersed gene families also will show evidence of co-evolution in Physcomitrella.
We have characterized all 11 hexokinase encoding genes in the moss Physcomitrella and classified them into different types based on sequence motifs and intracellular localization. We found that the hexokinase gene family is more diverse in Physcomitrella than in other plants studied so far, encoding two novel types of hexokinases, types C and D. The presence of a cytoplasmic and nuclear hexokinase (type C) sets Physcomitrella apart from vascular plants, and instead resembles yeast, where all hexokinases localize to the cytosol. The fact that all moss hexokinases are more similar to each other than to hexokinases from vascular plants, even though both type A and type B hexokinases are present in all plants, further suggests that the hexokinases in Physcomitrella have undergone concerted evolution.
We thank Stefan Hohmann for providing us with the hexokinase deficient yeast strain. This work was supported by grants from Formas, the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning, and from SSF, the Swedish Strategic Research Foundation.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.