Tangled history of a multigene family: The evolution of ISOPENTENYLTRANSFERASE genes

Abstract

ISOPENTENYLTRANSFERASE (IPT) genes play important roles in the initial steps of cytokinin synthesis, exist in plant and pathogenic bacteria, and form a multigene family in plants. Protein domain searches revealed that bacteria and plant IPT proteins were to assigned to different protein domains families in the Pfam database, namely Pfam IPT (IPT^Pfam) and Pfam IPPT (IPPT^Pfam) families, both are closely related in the P-loop NTPase clan. To understand the origin and evolution of the genes, a species matrix was assembled across the tree of life and intensively in plant lineages. The IPT^Pfam domain was only found in few bacteria lineages, whereas IPPT^Pfam is common except in Archaea and Mycoplasma bacteria. The bacterial IPPT^Pfam domain miaA genes were shown as ancestral of eukaryotic IPPT^Pfam domain genes. Plant IPTs diversified into class I, class II tRNA-IPTs, and Adenosine-phosphate IPTs; the class I tRNA-IPTs appeared to represent direct successors of miaA genes were found in all plant genomes, whereas class II tRNA-IPTs originated from eukaryotic genes, and were found in prasinophyte algae and in euphyllophytes. Adenosine-phosphate IPTs were only found in angiosperms. Gene duplications resulted in gene redundancies with ubiquitous expression or diversification in expression. In conclusion, it is shown that IPT genes have a complex history prior to the protein family split, and might have experienced losses or HGTs, and gene duplications that are to be likely correlated with the rise in morphological complexity involved in fine tuning cytokinin production.

Introduction

The evolution of gene families can be complex and may involve duplications within genomes or through polyploidization and loss or conversion events, these being the major forces enlarging gene families, with mutations accumulating over time further differentiating individual family members [1, 2]. ISOPENTENYLTRANSFERASE (IPT) enzymes regulate a rate limiting step in the biosynthesis pathway of cytokinin, an important hormone [3]. They also have other functions such as stabilizing codon recognition of tRNA through the modification of tRNA in yeast. In mammals they are linked to mitochondrial diseases [4, 5]. Cytokinins are not only found in plants, but also in plant pathogenic bacteria such as the crown-gall forming Agrobacterium tumefaciens (reviewed in [6]), the cyanobacterium Nostoc sp. PCC7120 [7], and the slime-mold Dictyostelium discoideum [8].

IPT genes were first identified in A. tumefaciens [9, 10] and only much later in Arabidopsis thaliana [11, 12], after the release of its genome sequence [13], and nine IPT genes were identified in the genome [14]. To date, IPTs have been studied in several angiosperms and mosses (e.g., Arabidopsis thaliana [14]; Oryza sativa [15]; Physcomitrella patens [16]; Solanum lycopersicum [17]), and were shown to belong to one multigene family [14, 15, 18, 19, 20, 21]. In A. thaliana, they are classified into two types depending on the substrates they use; Adenosine-phosphate IPTs (AP-IPTs) and tRNA-IPTs [14]. Agrobacterium tumefaciens also retaining AP-IPT, preferentially uses AMP whereas those in plants prefer ATP and ADP as substrates [22].

In previous studies, Frébort et al. [18] classified IPT genes into five groups: `bacterial adenylate IPTs`, `plant adenylate IPTs`, `eukaryotic origin plant tRNA IPTs`, `bacterial tRNA IPTs`, and `prokaryotic origin plant tRNA IPTs`, based on an unrooted gene tree reconstructed from full sequence lengths, where members of two plant families (A. thaliana; O. sativa) were included. Lindner et al. [19] carried out a more comprehensive analysis with 30 species across kingdoms including 12 plant families, in which they separated plant IPTs into `class I tRNA-IPTs`, `class II tRNA-IPTs`and `adenylate-IPTs`, and bacteria IPTs into `bacterial tRNA-IPTs`and `bacterial AMP-IPTs`, using a midpoint rooted Bayesian inference tree. The cytokinin synthesizing genes of the bacteria A. tumefaciens and the slime-mold D. discoideum were found to belong to the AMP-IPT clade and were separated from plant IPT clades in Lindner et al. [19]. The authors further showed that class I tRNA-IPTs are closely related to bacteria tRNA-IPTs, and class II tRNA-IPTs to adenylate-IPTs [19].

The two different classifications by Frébort et al. [18] and Lindner et al. [19] are not fully congruent, principally because they did not included the same groups of organisms (Table 1). Furthermore, the evolutionary history of IPTs was not fully explained in the two studies since the phylogenetic trees were unrooted, and the direction of evolution as well as the origin of the gene family unexplored. A further complication might have been that the full sequence and protein sequences between the different groups of IPTs are highly divergent and their alignment might have included ambiguous alignment information, obscuring the phylogenetic signal [23].

Table 1. Classification of ISOPENTENYLTRANSFERASE genes.

Gene classification in Frébort et al. [18]	Gene classification in Lindner et al. [19]	Gene classification in this study	Clade/Grade	Domain	Lineages found in clade
Bacterial adenylate IPTs	AMT-IPT	Outgroup		IPTPfam	Bacteria; Slime-mold
Bacterial tRNA IPTs	-	Bacteria miaA grade	A	IPPTPfam	Bacteria
Prokaryotic origin plant tRNA IPTs	Class I tRNA-IPT	Plant class I tRNA-IPT	B	IPPTPfam	Algae; Mosses; Lycophytes; Gymnosperms; Angiosperms
-	-	Unikont-SAR tRNA-IPT grade	C	IPPTPfam	Mammals; Insect; Fungi; Slime-mold; Zooplankton
-	Class II tRNA-IPT	Prasinophyte tRNA-IPT	D	IPPTPfam	Prasinophyte algae
Eukaryotic origin plant tRNA IPTs	Class II tRNA-IPT	Plant class II tRNA-IPT	E	IPPTPfam	Gymnosperms; Angiosperms
Plant adenylate IPTs	ADP/ATP-IPTs	Adenosine-phosphate IPT	F	IPPTPfam	Angiosperms

IPT^Pfam, IPPT^Pfam–referring to Pfam protein families IPT and IPPT.

Therefore, this study focused on the conserved protein domain of the IPTs, to infer the deep origin and evolution of this gene family. The conserved protein domain of IPT genes across kingdoms were assembled with a focus on plants and the matrix included 37 plants (of 21 families), three animals, two fungi, one amoeba, and one zooplankton species, selected across the evolutionary breadth of the tree of life [24, 25, 26]. The results of these domain based phylogenetic analyses are discussed in the light of the frequency and timing of duplication events, and linked to expression patterns of the gene copies and their intron positions as reported in previous studies. This is the first detailed analysis to illustrate the origin and pattern of diversification of IPT genes in plants in a phylogenetic context.

Materials and methods

Genome resources

IPT genes were retrieved from publicly accessible genome or transcriptome databases. The list of species analysed and the databases used in this study are listed in S1 Table. The gene accession numbers are listed in S2 Table.

Domain searches

Domain searches were carried out using deduced amino acid sequences in Pfam v.31.0 [27]. Since IPTs are mostly single-domain proteins and retain either IPT^Pfam (Pfam family IPT) or IPPT^Pfam (Pfam family IPPT) domains, these were searched across kingdoms, including Archaea, bacteria, plants, yeast, animals, and slime-molds (S2 Table). Proteins possessing the IPT^Pfam domain are described as isopentenyl transferases or dimethylallyl transferases and synthesise cytokinin, while those possessing the IPPT^Pfam domain are IPP transferases/tRNA delta(2)-isopentenylpyrophosphate transferases and modify tRNA to stabilize codon recognition in a wide range of lineages (e.g., bacteria, fungi, mammals). They use AMP/ADP/ATP as substrates and contribute to cytokinin synthesis in plants [3]. The genome and transcriptome databases were BLAST searched (cut-off E < 0.1) using IPPT^Pfam and IPT^Pfam domains from A. thaliana and A. tumefaciens. Sequence matches were re-evaluated in Pfam searches, and only gene sequences clearly showing IPPT^Pfam and IPT^Pfam domain sequences were used for this study (S2 Table).

Assessing relationships among domain families

The protein families in the clan P-loop NTPase (CL0023), including IPT^Pfam (PF01745) and IPPT^Pfam (PF01715) protein domain families, were analysed. This clan included 217 protein domain families in Pfam v.31.0, and their Hidden Markov Model (HMM) profiles were downloaded from the Pfam website. HMM profiles estimate the true frequency of protein residues from the observed frequency by a Markov process with hidden status [28]. The HMM profile relationships were analysed and a distance matrix of HMM profiles and its unrooted Neighbor Joining tree generated using pHMM-tree [29].

Following the topology of the pHMM-tree, IPT^Pfam (PF01715) and IPPT^Pfam (PF01745) domain sequences were analysed using VirE^Pfam (PF05272) domain sequences as outgroup to focus on the phylogenetic relationship between IPT^Pfam and IPPT^Pfam. Sequences in the seed alignments of the three families were combined into a matrix. The seed alignment of IPT^Pfam contains seven, and that of VirE^Pfam six sequences and all were used in the analyses. The IPPT^Pfam seed alignment is large and contains 1247 sequences, and only representative sequences were selected for the analyses: to select sequences, preliminary phylogenetic analyses were carried out on the IPPT^Pfam seed alignment using all sequences. Hypervariable regions of the original IPPT^Pfam seed alignment were trimmed with BMGE v.1.12 [30], and a phylogenetic tree reconstructed with FastTree [31], and 162 topology-representative sequences selected. Finally, the reduced IPPT^Pfam seed alignment (162 sequences), IPT^Pfam seed alignment (7 sequences), and VirE^Pfam seed alignment (6 sequences), were combined with the MAFFT-merge subprogram in MAFFT v.7 [32], and the matrix was trimmed with BMGE v.1.12 [30]. An ML tree estimated with PhyML v.3.0 [33] with Smart Model Selection (SMS) [34] with the tree rooted on VirE^Pfam sequences. For branch support, values of an approximate likelihood ratio test with non-parametric branch support based on a Shimodaira-Hasegawa-like procedure (αLRT SH-like support) were estimated using PhyML. Additionally, an ultrafast bootstrap (UFBT) analysis of 1,000 replicates was carried out in W-IQ-TREE [35].

Building IPPT^Pfam HMM alignments with extended N-terminus region

The IPPT^Pfam original seed alignment with 1247 sequences was reduced to 103 representative sequences as described above. To confirm the similarity between the original (1247 sequences) and the representative sequences, HMM profiles were built for the 1247 and 103 sequences respectively, with hmmerbuild in HMMER v.3.0 [28], and HMM logos were generated with Skylign [36] and the logos compared. After confirming their similarity, full-lengths of the 103 representative sequences were retrieved from the database and the N-terminus region aligned manually. 101 out of the 103 sequences were found to have retained the approximately 40 AA long conserved region located in front of the starting point of the original IPPT^Pfam HMM (Fig 1). A new HMM profile was built that included those 40 AA sequences with hmmerbuild, its HMM logo generated, and the profile named IPPT^Pfam_N40.hmm. To annotate and check the protein alignment, the protein structures of IPT^Pfam and IPPT^Pfam domains were retrieved from the PDBsum-EMBL-EBI database (http://www.ebi.ac.uk/pdbsum), for IPT^Pfam from Agrobacterium tumefaciens (PDBsum accession number: 2ze5) and for IPPT^Pfam from Escherichia coli (3foz) as references.

Fig 1. Domain sequences used for HMM profiles.

Original IPPT^Pfam domain sequences were shorter than the IPT^Pfam domain by approximately 40 AA at the N-terminus. The expanded profile was retrieved from full sequences and used for the new HMM profile (IPPT^Pfam_N40.hmm). Arrowhead indicates the starting position of the original IPPT^Pfam.hmm. Box marked with an asterisk indicate the IPPT^Pfam region missing in the original IPPT^Pfam.hmm. Predicted positions of α-helixes are indicated by ‘H’, and those of β-sheet by ‘S’.

Assessing plant IPT^Pfam domain in Pfam database

Fragmental IPT^Pfam domains were found in species in a few plant families in the Pfam database (e.g., Musa acuminata, Solanum lycopersicum). Those plant IPT^Pfam domain genes registered in Pfam were retrieved and assessed with hmmersearch in HMMER v.3.0, which compares the protein sequences with IPPT^Pfam.hmm and IPT^Pfam.hmm from Pfam, and IPPT^Pfam_N40.hmm built in this study, to examine the similarities between the domain sequences and the HMM profiles.

In addition, a phylogenetic analysis was carried out with plant genes registered under IPT^Pfam domains in the database. The matrix was assembled with plant IPT^Pfam domain genes together with the bacterial IPT^Pfam domain genes, the bacterial IPPT^Pfam genes (miaA), and IPPT^Pfam genes from P. patens, A. thaliana, O. sativa, S. lycopersicum, S. tuberosum, M. acuminata. The IPPT^Pfam and IPT^Pfam domain sequences were first aligned separately using the hmmeralign in the HMMER v.3.0 with IPT^Pfam.hmm or IPPT^Pfam_N40.hmm. The two alignments were merged using MAFFT merge v.7 [32] and trimmed using BMGE v.1.12 [30]. The WAG model was selected under the AIC criterion [37] using Prottest v.3.0 [38], and an ML tree and αLRT SH-like support values were estimated with PhyML v.3.0 [33].

Detecting the presence of IPT^Pfam domain genes in bacteria and slime-mold and their phylogenetic relationship to IPPT^Pfam domain genes

To show the presence or absence of IPPT^Pfam and IPT^Pfam domain genes in bacteria and slime-mold, a species tree based on Battistuzzi et al. [39] and Tomitani et al. [40] was generated and annotated with the presence and absence of the domain genes. Yeast was added as outgroup. A Newick file was generated manually in a text editor and the tree modified in TreeView v.1.6.6 [41] and FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).

The phylogenetic tree of IPPT^Pfam and IPT^Pfam domain sequences of bacteria, slime-mold and yeast was build alongside the species tree generated above. IPPT^Pfam and IPT^Pfam domain sequences were aligned separately using hmmeralign with IPPT^Pfam_N40.hmm or IPT^Pfam.hmm. The two alignments were merged using MAFFT merge v.7 and trimmed using Gblocks [42]. The LG model was selected under the AIC criterion using Prottest v.3.0, and an ML tree estimated with PhyML v.3.0. αLRT SH-like support values were estimated using PhyML and an UFBT analysis of 10,000 replicates carried out in W-IQ-TREE [35].

Comprehensive phylogeny of IPPT^Pfam domain genes across kingdoms

To build the comprehensive phylogenetic IPPT^Pfam domain gene tree, IPPT^Pfam domain genes were retrieved from genome databases from algae to angiosperms. bacteria, animals, yeast, slime-molds, and zooplankton genes were also included in the analyses (S2 Table). The matrix was generated as described above for the bacterial IPPT^Pfam and IPT^Pfam phylogeny and trimmed using BMGE v.1.12. The LG model was selected under the AIC criterion using Prottest v.3.0, and an IPT^Pfam rooted ML tree with branch support value estimated as above with PhyML and W-IQ-TREE.

Estimating the timing of duplication of ISOPENTENYLTRANSFERASE in plants

To estimate the number and timings of duplications of IPTs specifically among plants, gene duplication and losses (DL) analyses were carried. Gene subtrees containing plant IPTs were reconciled and rearranged with plant species trees separately for class I tRNA IPTs and class II tRNA-IPT/AP-IPTs (S9–S14 Figs) in DL mode, considering duplications and losses, with NOTUNG v.2.9 using default settings [43]. To allow the topological support existing within the IPT clades to optimize the duplication-loss events, the αLRT SH-like branch support values of the ML analysis were transferred to the gene subtree in the NOTUNG analyses.

To place the history of IPTs in a phylogenetic timeframe, divergence times for major lineages and species were referred from key published analyses (angiosperms-liverworts [44], charophytes-red algae [45], eukaryotic lineages [46], prokaryotic lineages [39]), and a metric summary tree of life phylogeny constructed and the transfer and duplication event placed in that tree.

Intron distribution

Intron-exon structures were also examined, by interrogating databases and comparing genome and transcribed sequences. The number of nucleotides in exons and introns were determined and schematic illustrations based on their number, size and position drawn with GSDS v.2.0 [47].

Diversification of expression patterns

Literature searches were carried out to obtain an overview of gene expression patterns in relation to the duplication history of IPT genes. For interspecific comparisons, the expression data were categorised into root, leaf, flower, and fruit. For mosses, the protonema, mature gametophytic stage, and sporophytic stage were reported and these categories were used here. The literature used in this study regarding gene expression patterns are summarized in S3 and S4 Tables.

Accession numbers

The accession numbers of the sequences used in this study are listed in S2 and S5 Tables.

Datasets used in this study

The matrixes and tree files used in this study are deposited in TreeBASE (study accession http://purl.org/phylo/treebase/phylows/study/TB2:S22409). The files include a FastTree inferred approximately-ML tree of the IPPT^Pfam domain seed alignment with 1247 sequences (M46567), a IPPT^Pfam_N40.hmm new seed alignment with 103 sequences (M46568), the IPPT^Pfam/IPT^Pfam/VirE^Pfam merged matrix (with 175 sequences) and tree shown in S3 Fig (M46562, Tr112785), a plant IPT^Pfam domain matrix (with 101 sequences) and tree file shown in S6 Fig (M46563, Tr112786), the bacterial IPT^Pfam/IPPT^Pfam domain matrix (with 64 sequences) and tree shown in Fig 2 and S7 Fig (M46565, Tr112787), and a IPT^Pfam/IPPT^Pfam domain matrix (with 215 sequences) and tree across kingdoms shown in Fig 3, S8 Fig (M46566, Tr112788).

Fig 2. ML tree of the bacterial IPT^Pfam and IPPT^Pfam domain genes.

The αLRT (left) and UFBT values (right) are shown along major branches. An asterisk indicates support values < 0.5 and < 50%. Thickened branches indicate support values > 0.9 and > 90%, medium-thick branches indicate > 0.7 and > 70%. A tree with all support values is shown in S7 Fig. The species with both IPT^Pfam and IPPT^Pfam domain genes are highlighted blue and yellow. The classification of the species is indicated by two characters at the end of the gene names; Ac: Actinobacteria, Al: α-Proteobacteria, Am: Amoebozoa, Aq: Aquficae, As: Ascomycota, Be: β-Proteobacteria, Ch: Chlamydiae, Cy: Cyanobacteria, Ep: ε-Proteobacteria, Fi: Firmicutes, Fu: Fusobacteria, Ga: γ-Proteobacteria, Sp: Spirochaetes, Th: Thermotogae.

Fig 3. ML tree of IPPT^Pfam domain genes across kingdoms.

IPT^Pfam domain genes were used as outgroup. The αLRT (left) and UFBT support values (right), are shown along the major branches. An asterisk indicates support values < 0.5 and < 50%. Thickened branches indicate support values > 0.9 and > 90%, medium-thick branches indicate > 0.7 and > 70%. Trees with all support values are shown in S8 Fig. A. Bacterial IPPT^Pfam genes, miaA. B. Plant class I tRNA-IPTs. Two IPPT^Pfam domain genes from Dictyostelium discoideum are nested in this clade (red arrow). The Mosses IPPT^Pfam clade included multiple copies of tRNA-IPTs from Sphagnum fallax and Physcomitrella patens. C. Unikont-SAR IPTs. IPPT^Pfam domain genes of zooplankton, yeast, animals arranged in grades. One copy of the IPPT^Pfam gene of D. discoideum appeared as sister grade to the animal clade (red arrow). D. Prasinophyte algae tRNA-IPTs. Prasinophyte clade indicated by asterisk and pink box in B and D. E. Plant class II tRNA-IPTs. F. AP-IPTs. Two clades (F¹, F²) were observed and the basal angiosperm Amborella trichopoda retained two copies, one belonging to each clade (black arrows). Derived angiosperms retained diverged copies within F² (F²a,b). The multiple copies of Arabidopsis thaliana (grey arrows) and Oryza sativa (green arrows) are indicated. Arrowheads indicate gene duplication events inferred from NOTUNG analyses (see also Fig 5). Red arrowhead indicates gene duplication event prior to class II tRNA-IPT and AP-IPT splits, and blue and green arrowhead indicates events within plant AP-IPTs.

Results

Protein domains of cytokinin biosynthesis genes

Both A. thaliana and A. tumefaciens IPTs (AtIPTs for A. thaliana, Tzs and Ipt for A. tumefaciens) are single domain proteins of about 250–460 amino acid (AA) length (S1 Fig). We found that the cytokinin biosynthesis IPT genes in A. thaliana, AtIPT, and Oryza sativa, OsIPTs, possess an IPPT^Pfam domain, while these genes in A. tumefaciens, Ipt and Tzs, and Nostoc sp. PCC7120 NoIPT1, have an IPT^Pfam domain (S6 Table). Thus, cytokinin biosynthesis IPTs in plants and bacteria retain different domains.

IPPT^Pfam has a 228 AA long Hidden Markov Model (HMM) profile, and AP-IPT in A. thaliana has a truncated IPPT^Pfam domain lacking ca. 75–140 AA of the IPPT^Pfam HMM profile, while those of tRNA-IPT have almost the full length of the IPPT^Pfam HMM profile (S1 Fig). IPT^Pfam has a 233 AA HMM profile and Tzs or Ipt of A. tumefaciens possess almost the entire region of the IPT^Pfam HMM profile. Both, the IPPT^Pfam and IPT^Pfam domain, belong to the P-loop NTPase clan (CL0023) in the Pfam database v.31.0 [27]. This clan contains 217 families and these often perform chaperone-like functions [48, 49]. The pHMM-tree analyses of the P-loop NTPase clan suggested that IPPT^Pfam and IPT^Pfam HMM profiles are closely related and appear as sisters in the Neighbor Joining tree (S2 Fig).

When adding VirE^Pfam sequences as outgroups to IPPT^Pfam and IPT^Pfam domain sequences, the ML phylogenetic analyses performed using the protein domain sequence alignment showed that IPPT^Pfam and IPT^Pfam domain sequences formed individual clades with high branch support each (IPT^Pfam: αLRT SH-like = 0.89, UFBT = 99; IPPT^Pfam: αLRT SH-like = 0.85, UFBT = 82) and were highly supported sister to each other (αLRT SH-like = 0.97, UFBT = 100), suggesting that the origin of IPPT^Pfam and IPT^Pfam proteins could be traced back to before the emergence of the protein families (S3 Fig).

IPPT^Pfam and IPT^Pfam domain proteins in plants and bacteria

The presence of IPPT^Pfam and IPT^Pfam domains assigned to IPTs was investigated across kingdoms including Archaea, bacteria, slime-mold, yeast, plants and animals. Intriguingly, IPT^Pfam domain genes were only found in the genomes of bacteria and the slime-mold D. discoideum (S1 Table), and in very few plant species: P. patens, S. lycopersicum, S. tuberosum, Musa acuminata, and Oryza barthii and Oryza brachyntha (S5 Table). On the other hand, IPPT^Pfam domain genes were found in most other organisms examined, except in Archaea and the Mycoplasma lineage in Firmicutes of bacteria (S1 Table, S4 Fig).

The bacterial IPT^Pfam domain genes (e.g., Tzs and Ipt in A. tumefaciens; S1 Fig) are well characterized, whereas those in plants only exist in a few species, many of those are located as very fragmented proteins shorter than 100 AA. These sequences matched only positions 2 to 112 of the 288 AA IPT^Pfam.hmm, which indicated that they only retain the N-terminus region of IPT^Pfam.hmm (S5 Table). In the seed alignment the IPT^Pfam domain was found to be about 40 AA longer than those of the IPPT^Pfam domain towards the N-terminus (Fig 1). Evaluation of the sequences in the IPPT^Pfam seed alignment showed that the IPPT^Pfam HMM profile can be extend towards the N-terminus to match the length of the IPT^Pfam.hmm (Fig 1, S5 Fig). Thus, a new HMM alignment was built that included an additional 40 AA (IPPT^Pfam_N40.hmm; S5 Fig). HMM searches revealed that plant IPT^Pfam domain gene sequences had a higher or equivalent similarity to IPPT^Pfam_N40.hmm compared to IPT^Pfam.hmm (S5 Table). The ML tree also showed that the plant IPT^Pfam domain genes grouped together in the IPPT^Pfam domain gene clade with a high support value (αLRT = 1), and not in the IPT^Pfam domain gene clade (S6 Fig). Therefore, the plant IPT^Pfam domains might be mis-assigned in the IPT^Pfam domains in the Pfam database since the original IPPT^Pfam.hmm lacks the N-terminus region where IPPT^Pfam and IPT^Pfam domains have high similarities. However, our analyses indicated that those mis-assigned plant IPT^Pfam domains were more similar to IPPT^Pfam domain genes. Since those plant IPT^Pfam domains lack a functional annotation and are fragmental, these were excluded from further analyses.

Across bacteria, D. discoideum, and yeast, the phylogenetic analyses of IPPT^Pfam and IPT^Pfam domain genes showed that each clustered separately with maximal clade support (αLRT = 1; UFBT = 100) (Fig 2, S7 Fig). The bacterial IPPT^Pfam domain genes, termed miaA, clustered predominantly following species tree relationships (S4 Fig; see [39, 50]), except for those in ɛ-Proteobacteria (Ep) and Borrelia burgdorferi (Spirochaetes, Sp). IPT^Pfam domain genes were only found in a few species: in Proteobacteria (α-Proteobacteria: Al, ß-Proteobacteria: Be, γ-Proteobacteria: Ga) they formed a clade (αLRT = 0.99; UFBT = 100), and with Cyanobacteria (Cy) and Actinobacteria (Ac) in sister grades (Fig 2, S4 Fig). One gene of D. discoideum (amoeba: Am) was also assigned to the IPT^Pfam domain clade.

Origin and diversification of ISOPENTENYLTRANSFERASEs

The cytokinin synthesizing IPTs in the plant species examined here all retained IPPT^Pfam domains (S1 and S6 Tables). In the phylogenetic tree rooted on IPT^Pfam domain genes (αLRT = 1; UFBT = 100), the IPPT^Pfam domain genes formed a maximally supported clade (αLRT = 1; UFBT = 100) and could be divided into two grades and four clades with mostly high branch support (Fig 3, Table 1, S8 Fig). The bacterial miaA genes formed grades at the base of the IPPT^Pfam clade and each of the two IPPT^Pfam subclades, one leading to plant class I tRNA-IPTs (Fig 3 clade B, S1 Table), the other to Unikont-SAR tRNA-IPTs including animal, fungi, zooplankton, and some copies from slime-mold (Fig 3 grade C). The prasinophyte tRNA-IPTs followed in the next grade (Fig 3 clade D), to which euphyllophyte IPTs were sister (Fig 3 clades E + F). Class II tRNA-IPTs (Fig 3 clade E) included genes from euphyllophytes, i.e. monilophytes, gymnosperms, and angiosperms. The clade and grade structures shown in Fig 3 is summarized along the tree of life in Fig 4.

Fig 4. Schematic dated tree of life with absence and presence of IPT^Pfam and IPPT^Pfam domain genes and IPPT^Pfam gene clades/grades shown in Fig 3.

Grey-shaded or open squares indicate IPT^Pfam and IPPT^Pfam domain presence or absence respectively. Presence (coloured squares) or absence (open squares) of class I tRNA-IPTs, class II tRNA-IPTs, and Adenosine-phosphate IPTs (AP-IPTs) for plants indicated by blue, green, or orange respectively. Class I and class II tRNA-IPTs and AP-IPTs are IPPT^Pfam domain genes. Shaded-circles indicate the presence of the possible direct ancestral IPPT^Pfam domain genes of plant class I and class II tRNA-IPTs. LECA: the last eukaryotic common ancestor, CK: point of cytokinin signal establishment [51]. Organism phylogeny is based on the Tree of Life Web Project [52], Qiu et al. [53] and Hug et al. [54], Popper et al. [55], Derelle et al. [56]. Dates are transferred from Magallón et al. [44], Heron et al. [45], Parfrey et al. [46], Battistuzzi et al. [39].

Duplications of ISOPENTENYLTRANSFERASEs within plant clades

The high copy number of IPPT^Pfam genes found in mosses and angiosperms had different patterns of distribution: the mosses Sphagnum fallax and Physcomitrella patens possessed five and eight IPPT^Pfam genes respectively, all of which belonged to the class I tRNA-IPT clade (‘Mosses’ in Fig 3). Most angiosperms in this clade, on the other hand, had only single copies, except for Brassica rapa and Sorghum bicolor which had two copies. Angiosperms, however, possessed additional IPPT^Pfam genes across the class II tRNA-IPTs, and a high-copy number in the AP-IPTs clade (Fig 3, S1 and S2 Tables). The basal angiosperm Amborella trichopoda possessed two copies of AP-IPTs and each was assigned to a different clade (black arrows in Fig 3 clade F), where otherwise extensive gene duplications had occurred. For instance, A. thaliana possessed four genes in clade F¹ and three in clade F², and Oryza sativa three in F¹ and five in F² respectively. Within the clades the gene trees roughly followed the species tree with some discrepancies, but many of these branches were not highly supported or unsupported (S8 Fig). The NOTUNG analyses provided some context for the interpretation of these discrepancies.

The reconciled NOTUNG tree for plant class I tRNA-IPT genes had a DL score (duplications and losses event score) of 48, and suggested 18 duplications and 21 losses. Rearranging the tree topology around poorly supported branches resulted in a greatly reduced DL score of 23.5, with 13 duplications and 4 losses (S11 Fig). Most duplications were inferred in the moss lineage, with two of the nine occurring at the time of diversification of Physcomitrella patens and Sphagnum fallax and alone five in P. patens after its diversification from S. fallax (Fig 5). Isolated duplications of class I tRNA-IPT inferred to have occurred once in Marchantia polymorpha. In angiosperms, class I tRNA-IPT duplications were rarely inferred, once prior or at the time of diversification of Poaceae, once within Poaceae at or prior to the split between Zea mays and Sorghum bicolor, and once in Brassica rapa after its split from A. thaliana (Fig 5).

Fig 5. Duplications and major losses in IPT genes inferred in NOTUNG analyses on the tree of life for plants.

Gene duplication resulting in class II tRNA-IPT and AP-IPT (red arrowhead), was followed by AP-IPT losses in ferns and gymnosperms (`L`in black). AP-IPTs duplications were inferred before or at angiosperm diversification (blue and green arrowheads). D: gene duplications, L: gene losses, `D`in black: duplication leading to class II tRNA-IPT and AP-IPT, `D`in blue: duplications within AP-IPT-1, `D`in grey: duplications within AP-IPT-2, `D`in red: duplication within class II tRNA-IPT, `D`in brown: duplications within class I tRNA-IPT.

For class II tRNA-IPTs/AP-IPTs, the reconciled NOTUNG tree prior rearrangement had a DL score of 220.5, involving 61 duplications and 129 losses. After rearrangement, the DL score was reduced to 88.5 with 39 duplications and 30 losses (S14 Fig). One early duplication of class II tRNA-IPT and AP-IPT was inferred to have occurred after the acquisition of IPT genes by euphyllophytes perhaps coinciding with the diversification of the lineage (S14 Fig, Figs 4 and 5, red arrowhead), with the monilophytes and gymnosperms appear to have consecutively lost their AP-IPT copies. Two successive duplications were inferred for angiosperms prior or at the time of their first diversification, the first giving rise to AP-IPT-1 (Fig 3F¹) and AP-IPT-2 (Fig 3F²), the following one resulting in AP-IPT-2a (Fig 3F²a) and AP-IPT-2b (Fig 3F²b). Some lineages such as Amborella trichopoda and monocots were inferred to have lost their AP-IPT-2b copy (S14 Fig). More local duplications are scattered across the angiosperms. The monocot lineage Poaceae and Brassicaceae showed a high clustering of duplications, with the former having six duplication events prior or at the time of diversification and five such events were inferred for the lineage of Brassica rapa (Fig 5).

The exon-intron structure showed that class I and class II tRNA-IPTs possessed multiple introns, but in Poaceae intron losses occurred in class I tRNA-IPTs (S15 Fig, S2 Table). Unlike tRNA-IPTs, AP-IPTs in general rarely possessed introns (S15 Fig, S2 Table). To understand the differentiation and similarities of function of the multiple copies of IPT, published results for gene expression patterns in moss, gymnosperm, and angiosperms were summarised alongside the phylogenetic IPT^Pfam/IPPT^Pfam tree (S15 Fig).

Discussion

IPPT^Pfam and IPT^Pfam domains

The Pfam database v.31.0 (released on 8 March 2017) contains 16,712 protein families and 604 clans. Each family is based on the manually curated seed-alignment of protein domains and thus each has a unique Hidden Markov Model (HMM) profile. A Pfam clan is a structural unit of families that share a related structure, function, and significantly matching HMM profile, suggesting that they have a single evolutionary origin [57, 58]. The two protein families, IPPT^Pfam and IPT^Pfam, assigned for cytokinin biosynthesis IPT genes are both in the P-loop NTPase clan and closely related, suggesting that genes in the IPPT^Pfam and IPT^Pfam families share a common ancestor before the two protein families diverged, and followed independent evolutionary trajectories. This has been confirmed here in our analysis including the VirE^Pfam family (S3 Fig).

IPT^Pfam domain genes are only found in a few bacteria, whereas IPPT^Pfam domain genes are found in most organisms except the Archaea and Mycoplasma lineages. It is unclear whether IPPT^Pfam is lost in Archaea or gained in bacteria since the relationships between the two groups are still unclear (e.g. [52]). It appears, however, to more likely represent a gain in bacteria that spread into the eukaryote lineages (see e.g. [49]). The Firmicute Mycoplasma is known to have a very small genome that is missing many genes, which might be a reason for the absence of IPPT^Pfam domain genes here [59].

The IPT^Pfam domain genes are phylogenetically scattered and found only in some members of Actinobacteria, Cyanobacteria, ɑ-Proteobacteria, β-Proteobacteria, and γ-Proteobacteria and in the eukaryote D. discoideum. The IPT^Pfam domain clade showed long branches and its topology was mostly congruent with the species tree. One could hypothesize that they were present in the ancestor of bacteria, and as a result of a strong evolutionary selection only the plant pathogenic lineages retained the IPT^Pfam domain genes, perhaps because of the importance of cytokinins in plant pathogenicity (e.g. [60]). However, this would require multiple losses of IPT^Pfam domain genes in the other bacteria lineages. Overall, a more parsimonious scenario would be HGTs that caused the scattered distribution of IPT^Pfam domain genes in bacteria, perhaps events that occurred in the more distant past that allowed some phylogenetic patterns to be retained among the IPT^Pfam domain genes. In support of this scenario, D. discoideum could be cited where HGT events are widely observed in its genome and this might explain the presence of IPT^Pfam in this organism [61].

One might expect that cytokinin synthesising genes in bacteria and plants are closely related. However, bacteria and slime-mold cytokinin synthesising IPTs appear to be only distantly related to plant IPTs. Plant IPPT^Pfam domain IPTs were found indeed closer related to bacteria IPPT^Pfam domain miaA genes that however, do not synthesise cytokinins (Fig 3). Thus we infer that the cytokinin synthesis pathways in plants and bacteria have evolved or have been acquired twice independently.

Origins and early evolution of ISOPENTENYLTRANSFERASEs

The present study has shown that plants IPTs have two different evolutionary sources, class I tRNA-IPTs originating from bacterial miaA genes, and class II tRNA-IPTs and AP-IPTs linked to the Unikont-SAR IPT grade (Fig 3C) through prasinophyte algae tRNA-IPTs (Fig 3D). The class I tRNA-IPT clade included all plant lineages examined in this study, ranging from red algae to angiosperms. The basal relationships of the tree of life around the last eukaryotic common ancestor (LECA) are still unresolved which somewhat hampers the clarification of the origin of IPT genes as well as the limited sampling in non-plant lineages in this study. However, based on the distribution of the genes among lineages (Figs 3–5), several hypotheses can be proposed (Fig 6): It is possible that plants have acquired class I tRNA-IPT genes from bacteria through their LECA early on in time 1,900 MYA and then following the tree of life with subsequent losses in the lineages leading to animals/fungi (Unikonts) and SAR (Fig 6A). Alternatively, plants could have acquired class I tRNA-IPTs via HGT from bacteria, perhaps before the diversification of plantae 1,600 MYA (Fig 6B). In this case, the brown algae and slime mold lineage would have acquired the genes independently, perhaps through further HGT events.

Fig 6. Schematic summary of hypotheses for ISOPENTENYLTRANSFERASE gene evolution inferred in this study.

Lines indicate possible evolutionary pathways from bacterial or eukaryotic ancestral IPPT^Pfam domain genes to plant IPT genes with IPPT^Pfam domain. Open boxes: gene loss, shaded boxes: gene gain, LECA: the last eukaryotic common ancestor. A, B. Hypotheses for class I tRNA-IPT evolution. A. class I tRNA-IPTs in plants directly descended from LECA gene and loss in Unikont and SAR independently. B. class I tRNA-IPTs in plants acquired via HGT from bacteria and secondary transfer to brown algae and slime molds. C, D. Hypotheses for class II tRNA-IPT/AP-IPT evolution. C. class II tRNA-IPTs in euphyllophytes were obtained via HGT from eukaryotic organisms, using prasinophytes as stepping stone. D. class II tRNA-IPTs evolved directly from LECA, but loss in brown algae, red algae and in several basal lineages of green plants independently.

Also for the origin of plant class II tRNA-IPT/AP-IPT, two hypotheses for can be postulated (Fig 6C and 6D): In one hypothesis, a common ancestor of the red algae and green plants (green lineage) lost the original eukaryotic tRNA-IPT of the LECA, and around 411 MYA, euphyllophytes secondarily acquired class II tRNA-IPTs by two HGT events from Unikont-SAR tRNA-IPT using the prasinophyte algae as stepping stone (Fig 6C). This hypothesis is supported by the unique genome structure of prasinophyte algae. It harbours large viral DNA in addition to their own genome [62, 63], and HGT events are commonly observed between eukaryote genomes and viral DNAs [62, 63, 64]. In an alternative hypothesis, class II tRNA-IPT/AP-IPT could have originated by descent of the original eukaryotic tRNA-IPT from the LECA, following the tree of life to the green lineages, but was later lost in several plant lineages (Fig 6D). However, this would require seven independent losses of the genes, in red algae, core chlorophytes and charophytes in green algae, liverworts, mosses, hornworts, and lycophytes (Fig 6D). The fact that the publicly available 14 genomes of the seven lineages investigated here all lack class II tRNA-IPT genes might suggest that the stepping-stone hypothesis is more likely because it requires fewer events to explain the scenario. There is some controversy surrounding the paraphyly of bryophytes, with the latest work suggesting various scenarios [65, 66]. Even if they were monophyletic, this would reduce the number of losses of class II tRNA-IPT genes by only two. Overall, a better understanding of the deep origin of tRNA-IPT genes can only be gleaned once the number of available genomes increases in the future and a better resolution of the eukaryote origin is achieved.

Among plants, only prasinophyte algae, monilophytes, gymnosperms and angiosperms possessed additional tRNA-IPTs besides class I tRNA-IPTs. In a previous study, these were classified together as class II tRNA-IPTs [19]. The present study showed that prasinophyte algae tRNA-IPTs formed a grade between Unikont-SAR tRNA-IPTs, and a clade with plant class II tRNA-IPTs and AP-IPTs. None of the other algae lineages (i.e. red algae, core chlorophytes, charophytes), bryophytes, and lycophytes retained class II tRNA-IPTs and AP-IPTs (Fig 4, S1 Table). A study on the evolution of cytokinin receptor genes suggested that the cytokinin signal transduction pathway established later towards the evolution of land plants in charophytes. Since prasinophytes algae lack the complete set of genes responsible for cytokinin signal transductions [51, 67], the additional copies of tRNA-IPTs in prasinophyte algae might not possess the function for cytokinin production but have their own as yet unknown roles. Therefore, in this study prasinophyte algae tRNA-IPTs (Fig 3D) were placed in their own class, `prasinophyte tRNA-IPTs`, separate from plant class II tRNA-IPTs (Table 1).

Duplication and redundancy of plant ISOPENTENYLTRANSFERASEs

The evolutionary history of IPTs in plants is marked by multiple gene duplication and major loss events that strikingly differed between plant lineages (Figs 3 and 5). It was noticeable that class I tRNA-IPTs showed many duplications in mosses, and very few in angiosperms, while the reverse was the case for class II tRNA-IPT/AP-IPT genes. This might be linked to functional redundancies (see below). The time of acquisition of a second set of tRNA-IPTs in euphyllophytes was estimated to around 411 MYA, sometime after the emergence of land plants [68], and coincided with a gene duplication event that gave rise to class II tRNA-IPT and AP-IPT. The latter was apparently lost in monilophytes and gymnosperms (Figs 4 and 5, S14 Fig), or not yet found at least in gymnosperms where only two genomes of one family, Pinaceae, were available at present. Two further duplications among AP-IPTs led to a further increase in copy numbers around the time of first divergence of angiosperms 194 MYA. Further duplications occurred, often in parallel in AP-IPT-1 and AP-IPT-2 throughout the diversification of angiosperms (Figs 3–5). Some of the earlier events might be linked to whole genome duplications that have been indicated in the evolution of seed plants and angiosperms (e.g. [69]]. The strong clustering of duplication events in Brassicaceae and Poaceae may stem from the much denser genome data available for these lineages that included model plants such as A. thaliana or O. sativa.

Overall, the pattern of IPT gene duplications across plants showed a tendency of an increased rate towards derived clades and increased morphological complexity with a peak in the AP-IPT clade with some plants possessing more than 10 copies (Figs 3 and 5, S1 Table). Comparing the function of these copies indicated that some IPTs show ubiquitous expression, while others show tissue specific patterns, and great redundancies among copies (S15 Fig; [14, 20]). A tendency was observed in that copies with specific roles occur in the most derived class of IPT genes in each species. For mosses it was the class I tRNA-IPTs, for gymnosperm class II tRNA-IPTs, and for angiosperm AP-IPTs; e.g. suppression of PpIPT4 expression in the moss sporophytic stage (S9 Fig; [21]), differential expression of PatIPT_IIa and PatIPT_IIb in female cones (S9 Fig; [70]), and in angiosperms, AP-IPTs showed differential expression patterns in different organs and differential response to external cytokinin treatments (S15 Fig, S4 Table). This might be a typical pattern for gene duplications from a ubiquitously expressed copy that allowed the acquisition of redundant copies to have a specific roles [71]. Thus, multiple but specific plant IPT copies may be important in fine-tuning the cytokinin concentration locally.

Introns are rarely found in AP-IPTs in contrast to class II tRNA-IPTs (S15 Fig). Considering the more likely stepping stone origin for class II tRNA-IPTs through prasinophytes, the lack of introns in prasinophytes might indicate that intron-gain in plant class II-IPTs is more likely (Figs 2 and 5, S9 Fig) rather than the intron-loss in AP-IPTs. The expression of AP-IPTs with few or no introns might be regulated by specific promoters reacting in the temporal-spatial manner at different plant growth stages (e.g. [20]). Considering the effects of presence and absence of introns, it was shown that rapidly transcribed genes retained lower numbers of introns [72]. It can be speculated that intron-less AP-IPT genes might result in more rapid transcription during different developmental stages when a finely tuned rapid cytokinin production is required, for example during flower development or when responding rapidly to environmental changes (e.g. [73]). A unique case was found in the Poales clade showing an absence of introns in class I tRNA-IPTs, whereas other lineages retained introns. While AP-IPTs produce trans-zeatin or isopentenyladenine type cytokinins, which have been considered as major cytokinins in angiosperms, tRNA-IPTs are thought to produce cis-zeatin type cytokinin, which is supposed to have minor or no function as cytokinin [3]. However, cis-zeatins are abundant in Poales [52, 74] and even retain their biological functions as cytokinins [75]. It might just be that intron loss in Poales class I tRNA-IPTs affect the regulation of cis-zeatin type cytokinin production in plants, an aspect that would be worthwhile testing in the future.

Conclusions

The roles and functions of ISOPENTENYLTRANSFERASEs, key genes for the production of cytokinins, have been studied intensively over the last two decades. The accumulating genome knowledge of model and non-model plants and an accompanying advancement in statistical analytical methodology applied here allowed us to reveal the phylogenetic origin and evolution of these genes across the tree of life. This study revealed that plant IPTs are closely related to bacteria miaA genes (IPPT^Pfam) and not to bacteria IPT genes (IPT^Pfam). Further, plants possess two independent IPTs, class I tRNA-IPTs and class II tRNA-IPT/AP-IPTs. Their exact deep origin could not be fully resolved due to uncertain relationships in basal eukaryotes. However, class II tRNA-IPTs and AP-IPTs are the consequence of a gene duplication event at the onset of euphyllophyte diversification. Further gene duplication events in the plant lineage were inferred with increasing frequency towards angiosperms, coinciding with emerging increased specialisation of functions. This study is an example for the elucidation of the deep history of cytokinin synthesis genes that involved an interplay of possible horizontal gene transfers, gene duplications, losses and diversification in function in the evolution of a multigene family.

Supporting information

S1 Fig. Domain structure of ISOPENTENYLTRANSFERASEs in Arabidopsis thaliana and Tzs, Ipt, and miaA genes in Agrobacterium tumefaciens.

Domains are shown as green boxes. Coordinates to the Pfam HMM profiles are shown below the boxes.

(PDF)

S2 Fig. Neighbor-joining tree of HMM profiles of P-loop NTPases in the Pfam database.

The tree was calculated by pHMM-tree. IPT^Pfam and IPPT^Pfam families appeared as sister clades (arrow).

(PDF)

S3 Fig. ML tree based on sequences of IPPT^Pfam, IPT^Pfam, and pVirE^Pfam seed alignments.

The tree is rooted on VirE^Pfam sequences. The αLRT SH-like values (left) and UFBT values (right) are shown on major branches leading to each protein family. Branches with above 70% support values are emphasized by a thick line.

(PDF)

S4 Fig. Species tree of bacteria.

Species retaining the IPPT^Pfam domain gene are shown in black, for species with both IPPT^Pfam and IPT^Pfam domain genes in orange, and for species lacking IPPT^Pfam and IPT^Pfam domain genes in grey.

(PDF)

S5 Fig. Comparison of HMM logos between the original HMM registered in Pfam v.31.0 and the expanded HMM build in this study.

IPT^Pfam.hmm and the original IPPT^Pfam.hmm (1247 seed seq) were retrieved from Pfam v.31.0. IPPT^Pfam.hmm (103 seed seq) built with 103 representative out of 1247 seed sequences. Logos build from 1247 sequences in the original seed alignment and the 103 representative sequences were very similar. New HMM profile with extended N-terminus (IPPT^Pfam_N40.hmm) built in this study. N40: additional N-terminus region in IPPT^Pfam_N40.hmm.

(PDF)

S6 Fig. ML tree calculated by PhyML including plant sequences registered in the IPT^Pfam family of Pfam database.

Plant sequences in IPT^Pfam family of Pfam indicated by red arrows, and those shown in the IPPT^Pfam domain clade but not in the IPT^Pfam clade (αLRT SH-like = 1).

(PDF)

S7 Fig. ML tree shown in Fig 2 with all support values.

The αLRT (left) and UFBT values (right) are shown along major branches. An asterisk indicates support values < 0.5 and < 50%. Thickened branches indicate support values > 0.9 and > 90%, medium-thick branches indicate > 0.7 and > 70%. The classification of the species is indicated by two characters at the end of the gene names; Ac: Actinobacteria, Al: α-Proteobacteria, Am: Amoebozoa, Aq: Aquficae, As: Ascomycota, Be: β-Proteobacteria, Ch: Chlamydiae, Cy: Cyanobacteria, Ep: ε-Proteobacteria, Fi: Firmicutes, Fu: Fusobacteria, Ga: γ-Proteobacteria, Sp: Spirochaetes, Th: Thermotogae.

(PDF)

S8 Fig. ML tree shown in Fig 3 with all support values.

IPT^Pfam domain genes were used as outgroup. The αLRT (left) and UFBT support values (right), are shown along the major branches. An asterisk indicates support values < 0.5 and < 50%. Thickened branches indicate support values > 0.9 and > 90%, medium-thick branches indicate > 0.7 and > 70%.

(PDF)

S9 Fig. Species tree used for NOTUNG analyses of plant class I tRNA-IPTs.

(PDF)

S10 Fig. Gene tree used for NOTUNG analyses of plant class I tRNA-IPTs.

(PDF)

S11 Fig. NOTUNG DL analyses of plant class I tRNA-IPTs.

Weak edges highlighted yellow. Gene duplications marked by red ‘D’.

(PDF)

S12 Fig. Species tree used for NOTUNG analyses of plant class II tRNA-IPTs/AP-IPTs.

(PDF)

S13 Fig. Gene tree used for NOTUNG analyses of plant class II tRNA-IPTs/AP-IPTs.

(PDF)

S14 Fig. NOTUNG DL analyses of plant class II tRNA-IPTs/AP-IPTs.

Weak edges highlighted yellow. Gene duplication marked by red ‘D’.

(PDF)

S15 Fig. Summary of intron positions, expression patterns, cytokinin interaction of plant ISOPENTENYLTRANSFERASEs alongside the phylogenetic tree.

The tree is a cladogram of the tree shown in Fig 3. Intron positions are shown as schematic illustrations. Asterisks indicate absence of introns in the gene. Genes without intron information are shown with ‘?’. Gene expressions are shown in square boxes: red indicates strong expression, orange indicates medium expression or expression without quantification, white indicates very weak or no expression in the tissues indicated (see also S3 Table). The response to external cytokinin treatments are indicated by upper or lower triangles: Upper triangles indicate the responses in the above ground parts of plants, and lower triangles indicate the responses in roots. Increase in gene expression is shown in yellow, no change in blue, and reduced expression in white (see also S4 Table).

(PDF)

S1 Table. List of species used in this study and their classification, with the numbers of IPT^Pfam and IPPT^Pfam domain genes.

(PDF)

S2 Table. Gene accession numbers used in this study.

Gene ID is the ID used in the large phylogeny in Fig 3. Asterisks indicate the gene name retrieved from Frébort et al. (2011) [18].

(PDF)

S3 Table. References used for the summary of gene expressions in S15 Fig.

(PDF)

S4 Table. References used for cytokinin interactions in S15 Fig.

(PDF)

S5 Table. List of plant IPT^Pfam domain genes in the Pfam database and results of the hmmsearch.

(PDF)

S6 Table. Domains assigned in ISOPENTENYLTRANSFERASE genes in model plants and cytokinin biosynthesizing bacteria.

(PDF)

Data Availability

The matrices and tree files shown in this study are available from TreeBASE and are accessible using the following link: http://purl.org/phylo/treebase/phylows/study/TB2:S22409.

Funding Statement

This work was supported by The Japan Society for the Promotion of Science (JSPS KAKENHI grant numbers 15K18593 and 18K06375 to K.N.; www.jsps.go.jp/j-grantsinaid/), the Sumitomo Foundation, Japan to K.N. (grant number 170204; http://www.sumitomo.or.jp/), the Sibbald Trust at Royal Botanic Garden Edinburgh, RBGE, UK (to K.N.; https://sibbaldtrust.wordpress.com/), and through travel funds from National Taiwan University, Taiwan (to Y.C). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.