The LONELY GUY gene family: from mosses to wheat, the key to the formation of active cytokinins in plants.

LONELY GUY (LOG) was first identified in a screen of rice mutants with defects in meristem maintenance. In plants, LOG codes for cytokinin riboside 5'-monophosphate phosphoribohydrolase, which converts inactive cytokinin nucleotides directly to the active free bases. Many enzymes with the PGGxGTxxE motif have been misannotated as lysine decarboxylases; conversely not all enzymes containing this motif are cytokinin-specific LOGs. As LOG mutants clearly impact yield in rice, we investigated the LOG gene family in bread wheat. By interrogating the wheat (Triticum aestivum) genome database, we show that wheat has multiple LOGs. The close alignment of TaLOG1, TaLOG2 and TaLOG6 with the X-ray structures of two functional Arabidopsis thaliana LOGs allows us to infer that the wheat LOGs 1-11 are functional LOGs. Using RNA-seq data sets, we assessed TaLOG expression across 70 tissue types, their responses to various stressors, the pattern of cis-regulatory elements (CREs) and intron/exon patterns. TaLOG gene family members are expressed variously across tissue types. When the TaLOG CREs are compared with those of the cytokinin dehydrogenases (CKX) and glucosyltransferases (CGT), there is close alignment of CREs between TaLOGs and TaCKXs reflecting the key role of CKX in maintaining cytokinin homeostasis. However, we suggest that the main homeostatic mechanism controlling cytokinin levels in response to biotic and abiotic challenge resides in the CGTs, rather than LOG or CKX. However, LOG transgenics and identified mutants in rice variously impact yield, providing interesting avenues for investigation in wheat.

Introduction

The cytokinins are involved in most aspects of plant growth and development, and there is significant interest in their role in enhancing yield (e.g. Jameson and Song, 2020; Liu et al., 2021; Nguyen et al., 2021a; Schwartz et al., 2020; Werner et al., 2021), including wheat (e.g. Beznec et al., 2021; Chen et al., 2020a, 2021a; Jablonski et al., 2020, 2021a, 2021b; Nguyen et al., 2020; Ogonowska et al., 2019). Historically, cytokinin biosynthesis in plants was considered a two‐step process: adenosine phosphate‐isopentenyl transferase (IPT) adding an isoprenoid side chain to an adenosine di‐ or triphosphate (ADP/ATP) to form inactive cytokinin nucleotides (Kakimoto, 2001; Takei et al., 2001), which were then predicted to be converted by a nucleotidase to form the ribosides (Chen and Kristopeit, 1981a), and then by a nucleosidase to form the bioactive free bases (Chen and Kristopeit, 1981b; Figure 1). However, while the reverse reactions (free base (nucleobase) to nucleotide via phosphoribosyl transferases; and riboside to nucleotide via adenine kinases) have been shown to occur via the purine salvage pathway (Chen et al., 1982; see reviews by Ashihara et al., 2018 and Witte and Herde, 2020), the genes/enzymes responsible for the two‐step forward reaction have not been identified in Arabidopsis thaliana (arabidopsis) or rice (Oryza sativa) (Kurakawa et al., 2007; Nguyen et al., 2021b), although a cytokinin riboside phosphorylase capable of interconverting cytokinin free bases and ribosides was identified in potato (Solanum tuberosum) stolons (Bromley et al., 2014). In a breakthrough paper, Kurakawa et al. (2007) showed that a one‐step reaction took place in rice converting the inactive cytokinin nucleotides directly to the active free base forms. This reaction was controlled by LONELY GUY (LOG) (Figure 1). LOG is named after the rice mutant that had only one stamen and no pistil (hence lonely guy) and was discovered in a screen for defects in meristem maintenance (Kurakawa et al., 2007). LOGs, which exhibit cytokinin riboside 5'‐monophosphate phosphoribohydrolase activity, have been misannotated as lysine decarboxylases (Jeon et al., 2006; Kurakawa et al., 2007; Naseem et al., 2018).

Figure 1

Outline of cytokinin biosynthetic and signal transduction pathway. Cytokinin riboside 5' monophosphates (cytokinin mononucleotides) are converted in one step by LONELY GUY (LOG) to the active cytokinin free bases (cytokinin nucleobases). In the putative two‐step process, ‘?’ refers to the paucity of support for cytokinin‐specific nucleotidases to convert the cytokinin nucleotides to ribosides, and ‘??’ refers similarly for cytokinin‐specific nucleosidases to convert cytokinin ribosides to nucleobases. Back conversion of nucleobases to nucleotides can occur via adenine phosphoribosyl transferase (APT1; Witte and Herde, 2020) and of ribosides to nucleotides by adenosine kinase (AK; Sakakibara, 2021). Cytokinin signal transduction is mediated by a two‐component system (TCS), consisting of histidine protein kinase receptors that sense the input and response regulators (RR) that mediate the output (Wang and Sheen, 2001). Cytokinin free bases are detected by histidine‐kinase receptors (HKs) on the ER (Romanov et al., 2018) or plasma membrane (Antoniadi et al., 2020; Kubiasova et al., 2020). Signal transduction is via a phospho relay with phosphate being transferred from the HKs by intermediary histidine phospho transfer proteins (HPs) leading to phosphorylation of type‐B response regulators (type‐B RRs) in the nucleus. The type‐B RRs are Myb‐type transcription factors that regulate transcription of primary response genes, leading to cytokinin responses. Negative regulator type‐A RRs are also transcribed (Zubo and Schaller, 2020). It should be noted that, while the essential TCS appears to be the same in monocots and dicots, rice also has a novel serine/threonine kinase receptor known as CHARK (Ito and Kurata, 2006; Halawa et al., 2021). Additionally, some rice RRs have acquired novel functions distinct from their roles in dicots (Worthen et al., 2019). A detailed figure of cytokinin biosynthesis and metabolism can be viewed in Chen et al. (2021a), and a model of the signal transduction pathway in Romanov et al. (2018).

In this review, using both sequence analysis and structural biology, we comment on the mega gene family to which the cytokinin nucleotide‐specific phosphoribohydrolases belong, consider the wider role of LOG in planta, and delve into the wheat genome to assess the potential utility of LOG in wheat breeding.

Evolutionary breadth of the LOG gene family

LOG gene family in rice

As mentioned above, the lonely guy rice mutant was identified in a screen for meristem mutants. LOG is expressed in the shoot apical meristem (SAM), in the panicle meristem at the primary branch initiation stage, and in the floral meristems of rice (Kurakawa et al., 2007). The effect of mutations in OsLOG resulted in a smaller vegetative meristem, although no abnormalities were apparent during vegetative growth. However, in log mutants there was a severe effect during the reproductive phase resulting in a reduction in panicle size, an abnormal branching pattern, a decrease in the number of floral organs and the abnormal floral feature (one stamen, no pistil) after which the gene is named, all of which were attributed to the reduction in proper meristem maintenance (Kurakawa et al., 2007; Table 1). Rescue of the log‐2 mutant occurred following the introduction of OsLOG cDNA driven by the LOG gene promoter. Functional testing showed that recombinant OsLOG converted cytokinin nucleotides isopentenyl riboside monophosphate (iPRMP) and trans‐zeatin riboside monophosphate (tZRMP) to the free base form with the release of a ribose 5'‐monophosphate. Further, LOG reacted only with cytokinin nucleoside 5'‐monophosphates, but not with the di‐ or triphosphates, nor with AMP, cytokinin ribosides and free bases; that is, rice LOG is a cytokinin nucleoside 5'‐monophosphate phosphoribohydrolase.

Table 1

Impacts of LOG mutants or manipulations of LOG

Species	Promoter/mutant	Environment	Characteristic	References
Cultivar/ecotype
Rice
Taichung65	log mutants		Normal vegetative development although smaller vegetative meristem Inflorescence and panicle branch meristems abort Reduction in panicle size and abnormal branching pattern Decrease in floral organs; often one stamen, no pistil Meristematic activity not properly maintained	Kurakawa et al. (2007)
	LOG::log2 complementation		Normal phenotype
O. rufipogon/O. sativa	LABA1 (OsLOGL6) laba1 allele		Long, barbed awns in wild rice Short, barbless awns; increased seed weight in cultivar	Hua et al. (2015)
	RNAi LABA1 LABA1::laba1 complementation		Awn length and barbs reduced Longer awns and barbs
O. rufipogon/O. sativa	An‐2 (OsLOG6)		Long, barbed awns in wild rice	Gu et al. (2015)
	An‐2 allele in awnless indica cv.		Awns elongated through increased cell division; decreased grain production through fewer grains/ panicle and fewer tillers/plant
	Complementation An‐2 oe		Awned progeny Awned progeny
Zhonghua	35S::LOGL5	Hydroponics	Shorter primary root Increased lateral root number	Wang et al. (2020)
		Field	Semi‐dwarf; narrow leaves Fewer tillers; fewer seeds/panicle; reduced 1000 grain weight
		Normal and low N Drought conditions	Decreased grain yield Decreased grain yield
	OsLOGL5: six knockouts at 3′ end	Hydroponics	Normal roots and shoots
		Field	Normal vegetative growth
	Edits A, B & F	Normal and low N	Increased grain yield
	Edits C, D & E	Low N Normal N	Reduced yield No effect
	Edits A‐C, E, F	Drought	Increased grain yield
	Edit D	Drought	Reduced tolerance (ns)
	Edits B, D, E & F	Drought	Increased seed setting rate, total grain number, full‐filled grain numbers/panicle and 1000 grain weight
Nipponbare	UNBRANCHED3	UB3‐oe	↓tillering; ↓panicle branching; ↓grains/panicle; ↓LOG1	Du et al. (2017)
BS208			Abnormal panicle branching pattern; ↓lateral grains on 2o branches	Li et al. (2021)
Teqing	REGULATOR OF GRAIN NUMBER1	rgn1 mutant	Absence of lateral grains on secondary branches; grain number decreased; grain size and weight increased; yield decreased LOG expression downregulated in NIL‐rgn1 compared to NIL‐RGN1
		RGN‐oe	LOG expression upregulated
		LOG‐oe in BS208	Partially rescued the absence of lateral grains in secondary branches
Taichung 65	OsWOX4	RNAi of WOX4	Severe defect in leaf development; vascular differentiation arrested; ↓OsLOG3 & 10 in vasculature	Yasui et al. (2018)
Arabidopsis thaliana
Columbia	AtLOG1‐9 Single, double and triple mutants of logs 3, 4 and 7		LOG6 & 9 non‐functional Single mutants: no visible phenotype Root growth assay: Single mutants not resistant to iPR; log2log7 and log3log4log7 mutants resistant to iPR log3log4log7 increased adventitious root formation; no change in primary root length; fewer flower buds and flowers	Kuroha et al. (2009)
	LOG7::log3log4log7		Rescued the reproductive stage phenotype
	35S::LOG2, 4, 5, 7, or 8		Increased cell division in embryos and leaf vascular tissues; reduced apical dominance; delayed leaf senescence; larger seeds; number of emerged lateral roots decreased
Columbia	Septuple log mutants		Severe retardation of shoot and root growth and development Reduction in size of root and shoot apical meristems Set flowers; Seeds larger than wild type	Tokunaga et al. (2012)
	Septuple mutant transformed with LOG7		Growth retardation rescued
Columbia	Log3,4,7 triple mutant		Abnormal root vasculature: vascular cell number reduced, vascular cell types eliminated (except for protoxylem)	Ohashi‐Ito et al. (2014)
	Triple mutant transformed with pTMO5::LOG3	xylem precursor‐cell‐specific promoter	Rescued triple mutant
Columbia	Log1,2,3,4,5,7,8		Reduced vascular cell file number; protoxylem only	De Rybel et al. (2014)
	Septuple mutant transformed with pTMO5‐LOG4 Septuple mutant transformed with pRPS5A‐LOG4	Constitutive promoter	Cell file number rescued; protoxylem and metaxylem present Vascular cell file number increased; metaxylem only
Columbia	LOG1,2,3,4,5,7,8	Grafted	Severe vegetative dwarfism due to smaller rosette leaves with reduced epidermal cell number, reduced SAM diameter and a delayed plastochron	Osugi et al. (2017)
	LOG1,2,3,4,5,7,8/WT	Grafted	WT root stock rescued leaf size and epidermal cell number, but not SAM size or plastochron number
	LOG1,2,3,4,5,7,8/abcg14	Grafted onto transporter mutant	Severe dwarfism when root‐to‐shoot cytokinin translocation impaired Recovered by tZ but not tZR application LOG function essential to meristem but not necessarily leaf growth
Columbia	LOG1,3,4,7 mutant		Vegetatively normal; smaller inflorescence meristems, fewer organs	Landrein et al. (2018)
	LOG4,7/Col‐0; LOG1,3,4,7/Col‐0	Grafted	Wild‐type root stock did not rescue LOG mutants
Columbia	AtLOG4	AtML1:LOG4 targeted to epidermal cells	Larger rosette and leaves; ↑cell number; early flowering; ↑vegetative meristem size and stem diameter; earlier transition to adult phase; lesser ↑inflorescence meristem size; ↑flower, gynoecia and silique size; ↑seed number/silique; ↑seed yield.	Werner et al. (2021)
Tomato	35S::TLOG1		Loss of apical dominance; de novo aerial mini‐tubers in tomato; extended aerial tuberisation in potato	Eviatar‐Ribak et al. (2013)
Medicago truncatula	GUS::MtLOG1 and 2		Showed expression in the upper dividing cells of the nodule primordia and, in more mature nodules, specifically in the meristem and early differentiating cells; CRE‐dependent	Mortier et al. (2014)
	GUS::MtLOG1		Expressed in the lateral root primordium; CRE‐independent
	MtLOG‐oe		Reduced nodule number, with loss of meristem in the nodules that did develop; Root thickening due to vascular tissue expansion and reduced primary root length
	Silencing of MtLOG1		Increased density of lateral roots but decreased the number of nodules formed
Lotus japonicus	LjLOG‐oe		Reduced nodule number	Reid et al. (2017)
	iPT‐Log‐Cyp735a‐oe		Spontaneous nodule formation occurred in Lotus japonicus
Cotton	GhLOG3‐oe		↑tolerance of Arabidopsis to NaCl	Wang et al. (2021)
	VIGS of GhLOG3_At		Enhanced sensitivity of cotton to salt stress
Kiwifruit	Gene edited SHY GIRL in male plant		LOG expression increased in ‘early flowers’; Enhanced feminisation	Varkonyi‐Gasic et al. (2021)
Chlorella variabilis	CvarLOG1‐oe	In vitro	“stay‐green” phenotype in culture	Nayar (2021)

Besides OsLOG, 10 other gene family members were recognized (Kurakawa et al., 2007), now referred to as OsLOGL1‐10. Using RT‐PCR, the expression of LOG gene family members was examined in eight tissues: OsLOG was expressed in all tissues, but most strongly in leaf blades; LOGL3 and LOGL7 were expressed in all tissues, with LOGL7 strongly expressed in the shoot apex and immature inflorescence tissues; LOGL2 and LOGL4 were not detected, and LOGL1 was barely detectable and only in the inflorescence meristem; LOGL5 was barely detectable; LOGL6 was not detected in stem or tiller buds, but was in other tissues; LOGL8 was specifically expressed in roots; and LOGL9 was expressed in root and leaf sheath whereas LOGL10 was expressed in all tissues except for root and leaf blade (Kurakawa et al., 2007).

LOG gene family in Arabidopsis thaliana

Nine arabidopsis LOG genes (AtLOG1‐9) were predicted as homologues of rice LOG (Kurakawa et al., 2007; Kuroha et al., 2009). However, the deduced amino acid sequence of AtLOG9 lacked the N‐terminal third of the rice LOG sequence, and the sequences of both AtLOG6 and 9 contained predicted premature stop codons (Kuroha et al., 2009). The remaining seven AtLOGs were tested for functionality. Purified recombinant proteins were shown to react with iPRMP, but not with the diphosphate or the triphosphate, or with the free base iP or its riboside. The recombinant proteins did not discriminate between the various isoprenoid nucleotide forms (iPRMP, tZRMP, cisZRMP and dihydroZRMP) but were generally less reactive towards the tested aromatic nucleotide (benzyl aminopurine, BARMP). In planta analysis, using an inducible promoter to drive overexpression of AtLOG2, showed a decrease in iPRPs, a slight increase in iP and iP‐9‐glucoside and a very strong increase in iP‐7‐glucoside (Kuroha et al., 2009), indicating overexpression of AtLOG2 resulted in the (subsequent) activation of cytokinin homeostatic mechanisms to deactivate excess cytokinin (Chen et al., 2021a).

Analysis of loss‐of‐function mutants showed that single Atlog mutants did not exhibit a visible phenotype. However, the double mutant Atlog2log7 and the triple mutant Atlog3log4log7 showed reduced sensitivity to iPR during the lateral root formation assay, but the double combinations of logs 3, 4 and 7 were not active. Complementation of the triple mutant by AtLOG7 rescued the reproductive stage, and overexpression of AtLOGs showed classic symptoms of excess cytokinin (Kuroha et al., 2009; Table 1).

Using GUS reporter gene constructs, detailed expression patterns of AtLOGs 1‐5, 7 and 8 were elegantly shown by Kuroha et al. (2009). AtLOGs 1, 4 and 8 were the most widely expressed across tissue types and AtLOG2 the least. The data indicated that AtLOGs were spatially and quantitatively differentiated, albeit with some overlaps between tissues, thus showing that active cytokinins could be synthesized in nearly all parts of the plant body via the direct pathway. Moreover, there are specific areas of overlap with the spatial expression patterns of the AtIPT GFMs shown by Miyawaki et al. (2004).

Tokunaga et al. (2012) extended the work with AtLOG mutants, showing that the sextuple or lower‐order mutants showed little change in vegetative phenotype, but the septuple mutants showed severe retardation of both shoot and root growth, and severe impacts on the maintenance of the apical meristems (Table 1). However, some plants did flower and set seed, with those seeds being enlarged. The levels of cytokinin nucleotides were increased and those of the free bases and glucosides decreased in seedlings of the septuple mutant, indicating that the active cytokinins produced by the LOG‐dependent pathway were critical to maintenance of shoot and root meristems. Considering all the various mutant combinations, Tokunaga et al. (2012) concluded that AtLOG7 was the most important GFM for the maintenance of the SAM and for the provision of a basal level of cytokinin to support normal primary root growth. Moreover, the septuple mutant could be rescued by AtLOG7 alone (Table 1). Interestingly, in the spatial analyses shown by Kuroha et al. (2009), AtLOG7 was expressed in only a limited number of cell types, including strong expression in the elongation zone of roots, but was not shown to be expressed in the SAM. However, in situ hybridization by Yadav et al. (2009) showed AtLOG7 expression in the SAM.

Similar work with multiple mutants of LOG/LOGL GFMs has not been published for rice or other species. It also appears only OsLOG and OsLOGL6 have been functionally characterized (Hua et al., 2015; Kurakawa et  al., 2007, respectively).

LOG gene family in wheat

In order to identify LOG orthologues in wheat, we used LOG sequences from higher plants to do a genome‐wide screening of the wheat proteome sequences using a HMMER model. A total of 45 LOG‐like candidate proteins were identified. BLAST and motif analysis showed that five of the 45 LOG‐like candidates did not have what is now known as the conserved LOG motif PGGxGTxE (Dzurová et al., 2015; Jeon et al., 2006) and were excluded. The remaining 40 LOG‐like candidates, together with LOG sequences from other species, were then used to perform a phylogeny analysis (Figure S1).

The resulting phylogenetic tree comprised four clades (Figure S1 and Figure 2). Clade I includes those organisms whose LOG proteins are predicted to operate as tetramers (Mayaka et al., 2019) and hexamers (Seo and Kim, 2017), whereas LOGs in Clades II, III and IV likely function as dimers (Seo and Kim, 2017). Our initial phylogenetic tree, generated based on amino acid sequences, included archaea, bacteria, cyanobacteria and higher plants in Clade I (Figure S1). Clade II includes a mixture of pathogenic microorganisms, green algae, mosses and higher plants, whereas Clades III and IV contain only higher plants. Members of the wheat LOG gene family were initially found in all Clades and named TaLOG1‐12 (Figures S1 and S2). Clades II, III and IV contain functionally characterized AtLOGs. Each of TaLOGs1‐11 aligns with a rice LOG GFM. TaLOG11 aligns most closely with OsLOG, functionally characterized by Kurakawa et al. (2007); and TaLOG2 with OsLOG2 and OsLOGL6, functionally characterized by Hua et al. (2015). Their allocated numbers are detailed in Figure 3 and Table S1.

Figure 2

Phylogenetic tree of LOGs from selected species. The wheat LOGs were identified from the latest released version 2.1 of genome‐wide wheat peptides (Zhu et al., 2021) using the HMMER model and BLAST methods described in Chen et al. (2020b; 2021a). The HMMER model was developed based on published LOG data from higher plants (Immanen et al., 2013; Kurakawa et al., 2007; Kuroha et al., 2009; Mortier et al., 2014; Pecrix et al., 2018; Tokunaga et al., 2012), followed by genome‐wide screening of the wheat peptide genome (version 2.1). The candidate TaLOG sequences were validated by genome‐wide BLAST on the wheat omics database (http://202.194.139.32/blast/blast.html, Ma et al., 2021). A total of 45 LOG‐like candidate proteins were identified. BLAST and motif analysis showed that five of the 45 LOG‐like candidate proteins did not have the shorter conserved LOG motif PGGxGTxE and were excluded. Additionally, MEME analysis showed that the originally named TaLOG12 members do not have the longer conserved motif 1 sequence within the lysine decarboxylase domain (LDC) and were also excluded. The phylogenetic tree was constructed using MEGA X software via the neighbor‐joining algorithm (Kumar et al., 2018) and then beautified using the iTOL web tool (https://itol.embl.de/).

Figure 3

The phylogenetic tree of TaLOGs. The wheat LOG sequences were identified using HMMER model and BLAST methods, combining motif analysis and domain analysis as described for Figure 2. The tree was constructed using MEGA X software via the neighbor‐joining algorithm (Kumar et al., 2018), and then, beautification was performed using the iTOL web tool (https://itol.embl.de/).

Motif (Figure 4) combined with MEME analysis (Figure S3) showed that TaLOG1‐11 all coded for the shorter canonical plant LOG motif (PGGYGTxxE), and the extended LOG motifs (MHxRKx(2)Mx(5,20)FIxLxxGGYGTxxEE) (Dzurová et al., 2015) and [LR][AS]D[AG]F[VI]x[LM]PGGxGT(L)[DE]E(L)(F)Ex[LW]x(Q)[LT]xx[HVI] (Naseem et al., 2018) within motif 1. TaLOG1‐11 have the critical amino acids for phosphoribohydrolase activity (Dzurová et al., 2015; Seo and Kim, 2017).

Figure 4

The motif patterns of TaLOGs. The motif patterns were predicted using the MEME database (https://meme‐suite.org/meme/tools/meme) and were sorted according to the order in the LOG phylogenetic tree shown in Figure 3. The data were shown using TBtools software (Chen et al., 2020b) and merged with the phylogenetic tree in Photoshop software (CS 6).

The structures of two functionally characterized arabidopsis LOGs, AtLOG3 and AtLOG8, have been determined (Jeon et al., 2006). Based on these structures, we have developed homology models for the closest wheat homologues, TaLOG2 and TaLOG6, and TaLOG1, respectively, and compared these with the arabidopsis LOGs. Sequence identities on alignment were 71.2% for TaLOG1 and AtLOG8, 82.9% for AtLOG3 and TaLOG2, and 79.4% for AtLOG3 and TaLOG6 (Table S2). As wheat LOGs 1‐11 contain the key amino acids for substrate binding and catalysis, and structural homology to characterized AtLOGs, it is highly likely that the wheat LOGs1‐11 are functional cytokinin nucleoside 5'‐monophosphate phosphoribohydrolases.

An AtLOG has been shown to clade as a hexamer‐forming type IIb LOG (Frébortová and Frébort, 2021), and this AtLOG sequence was used in our initial phylogenetic analysis. In this analysis, both TaLOG12 and a Medicago truncatula LOG cluster with it (Figure S1). Indeed, the phylogenetic tree constructed by Seo and Kim (2017) shows two OsLOGs, AtLOGs, MtLOGs, and Prunus persica LOGs cladding as hexamer‐forming type IIb LOGs. However, MEME analysis showed TaLOG12 does not have the longer conserved motif 1 sequence within the lysine decarboxylase domain (LDC; Figures 4 and S4). Within the LDC domain of AtLOG‐type IIa and MtLOGL3 (Figure S5), motif and MEME analysis showed the AtLOG and the MtLOGL3 genes have a PGGIGTxxE (with an isoleucine replacing a tyrosine) within motif 11 (Figures S6 and S7). The ‘I’ is indicative of type IIb LOGs as proposed by Seo and Kim (2017). However, there is little consensus with the longer LOG motifs of Dzurová et al. (2015) and Naseem et al. (2018) within the LDC domain (Figure S7). Additionally, we show that TaLOG12 does not have close structural homology with other hexameric phosphoribohydrolases (Table S2). We consider that it is unlikely that wheat and other higher plants have hexamer‐forming cytokinin‐specific LOGs, as shown by their being omitted from the amended phylogenetic tree (Figure 2). TaLOG12 was omitted from this and from further analyses. What TaLOG12 codes for is unknown as the presence of the shorter consensus sequence indicates it is unlikely to code for LDCs (Naseem et al., 2018).

Origins of LOG

Tracking backwards though the evolution of Archaeplastida, it is clear that expansion of the LOG gene family took place within this lineage and, more specifically, within the mosses as exemplified in Physcomitrium (previously Physcomitrella) patens (Bryophyta sensu stricto; Kuroha et al., 2009; Lu et al., 2014; Nayar, 2021). Moss appears to have nine LOG‐like GFMs (Lu et al., 2014), whereas the lycophyte Selagenella moellendorffii possibly has four, and a red alga possibly two (Lu et al., 2014). PpLOG XP 001781548 is likely ancestral to all other moss and plant LOGs (Figure 2).

Only a single LOG gene has been detected in each green alga studied (Lu et al., 2014; Nayar, 2021). Recombinant Chlorella variabilis LOG showed cytokinin phosphoribohydrolase activity against iPRMP >> tZRMP, leading Nayar (2021) to suggest that the LOG enzymes are functionally conserved from unicellular algae to higher plants. This functional conservation aligns with the presence of components of the canonical two‐component signalling pathway for cytokinins (Rashotte, 2021; Figure 1), indicative of cytokinin‐regulated developmental processes. Critically, the expansion of the cytokinin LOG gene family within the mosses (Lu et al., 2014; Figure 2) is paralleled by the expansion of gene families associated with each step of the two‐component signalling pathway in moss, relative to liverwort and hornwort, and the charophyte green algae, as reviewed by Rashotte (2021).

LOG belongs to a superfamily of genes, and LOG‐like genes have been shown to be present in all domains of life (Frébortová and Frébort, 2021; Mayaka et al., 2019; Naseem et al., 2018; Samanovic et al., 2015). Burroughs et al. (2015) suggest that this superfamily is likely to encompass dual functions, with certain versions serving as (oligo)nucleotide sensors and others as enzymes that operate on nucleotides. Carlsson et al. (2018) suggest that the original function of the LOG family of proteins was to degrade non‐canonical nucleotides and that their role in cytokinin production was a later development in some organisms. Witte and Herde (2020) suggest that, in addition to recycling nitrogen, purine nucleotide catabolism could dampen stress responses through the production of catabolic intermediates. We suggest ‘LOG’ should be retained for those enzymes definitively acting as cytokinin riboside 5'‐monophosphate phosphoribohydrolases to produce free base cytokinins and which then trigger the canonical two‐component signal transduction pathway (Figure 1). We suggest that those enzymes that are not cytokinin riboside 5'‐monophosphate‐specific phosphoribohydrolases are simply referred to as phosphoribohydrolases (PRH) and not as LOG. We will expand our reasoning in a forthcoming article. Suffice to say that the presence of the PGGxGTxxE motif does not necessarily confer cytokinin riboside 5'‐monophosphate‐specific phosphoribohydrolase activity, contrary to the inference relating to E. coli in Nguyen et al. (2021b).

Chromosome location of wheat LOGs

There are 11 TaLOG GFMs associated with the A subgenome, 14 associated with the B sub‐genome and 12 associated with the D sub‐genome. There are instances of multiplication and loss of specific TaLOG GFMs on the sub‐genomes. Only the D sub‐genome has a copy of at least one of each of the 11 GFMs (Figure 5), whereas the B sub‐genome lacks LOG6, and the A sub‐genome lacks both LOG6 and LOG8. This may relate to the dominance of the D sub‐genome in bread wheat (El Baidouri et al., 2017).

Figure 5

Chromosome location of TaLOGs. The position file was extracted via the Linux shell programs from the newly released gff3 file for wheat genome (iwgsc_refseqv2.1_annotation_200916_HC.gff3). The gene positions were then mapped onto the chromosomes using RIdeogram package in R language environment. The images of the ABD sub‐genomes were merged and edited in Photoshop CS6 software. In order to shorten the gene names, we omitted the first two letters (Ta) and used ‘h’ instead of ‘00’ at the end of each name. The corresponding information is listed in Table S2.

Multiplication and loss of LOG GFMs has been noted in Populus trichocarpa and Prunus persica, respectively (Immanen et al., 2013). Medicago truncatula also appears to have lost LOG GFMs (Mortier et al., 2014). Recently, Wang et al. (2021) concluded that there were 28 LOG GFMs in Gossypium hirsutum (cotton). Cotton is tetraploid and, like wheat, the LOG genes are unevenly distributed between the sub‐genomes.

Cellular localization of LOGs

It might be anticipated that LOGs will be located close to the production of cytokinin nucleotides by IPT and/or in the vicinity of cytokinin receptors. Kasahara et al. (2004) showed, using IPT:GFP fusion proteins, that AtIPT1, AtIPT3, AtIPT5 and AtIPT8 localize in plastids and experiments with stable isotope‐labelled substrates indicated that plastids are a major subcellular compartment for the initial step in cytokinin biosynthesis in arabidopsis seedlings. Additionally, AtIPT7 was targeted to mitochondria, and AtIPT4 was localized in the cytosol. The tRNA‐IPT fusion protein, AtIPT2‐GFP, was also localized in the cytosol (Kasahara et al., 2004). Subsequently, Galichet et al. (2008) showed that post‐translational farnesylation of AtIPT3 regulated its subcellular localization and directed most of the protein to the nucleus.

Transient and stable expression of rice, arabidopsis and Chlorella variabilis LOG‐reporter gene fusions showed localization to both cytosol and nuclei (Gu et al., 2015; Hua et al., 2015; Kurakawa et al., 2007; Kuroha et al., 2009; Nayar, 2021). In contrast, Wang et al. (2021) reported that cotton LOG3_At transiently expressed in tobacco leaves ‘localized in the cell membrane’ – although the authors above have described similarly localized signals as being cytosolic.

Why cytokinin might be produced in the nucleus is an interesting question. Cytokinins themselves are not transcription factors. Cytokinins are recognized by membrane‐localized histidine‐kinase receptors and are transduced through a His‐Asp phosphorelay to activate the type‐B response regulators, which are a family of transcription factors in the nucleus. While there are identified differences between arabidopsis and rice (Vaughan‐Hirsch et al., 2021; Worthen et al., 2019), the essential features of signal transduction are conserved (Figure 1). Cytokinin receptors and the initial signal of the signal transduction chain are external to the nucleus (Kieber and Schaller, 2018; Romanov and Schmülling, 2021; Figure 1). As the function of LOG is to release active cytokinins, activity of both an IPT and LOG in the nucleus itself is hard to explain. Cytokinin controls cell division through its control of the cell cycle. As described by Yang et al. (2021), the dual regulatory modes of cytokinin in CYCD3 transcription (Riou‐Khamlichi et al., 1999) and MYB3R4 nucleocytoplasmic shuttling (Yang et al., 2021) ensure precise control of cell cycle transitions in response to different levels of cytokinin input. Yang et al. (2021) state that the variation in cytokinin levels is most likely perceived by intracellular cytokinin receptors. Indeed, cytokinin receptors have been identified in the membranes close to the perinuclear pore (Romanov and Schmülling, 2021). A role for cytokinin actually within the nucleus remains to be identified. Potentially, an alternative explanation for the location of LOG in organelles where DNA is synthesized could reside in what Carlsson et al. (2018) suggest is the original role of LOG: to degrade non‐canonical nucleotides thus preventing their incorporation into DNA.

Interestingly, in wheat, our in silico analyses showed that none of the TaLOGs were predicted to localize to the nucleus irrespective of the prediction method (Table S3), even with Cell‐PLoc 2.0 and ProtComp producing highly contrasting results. The most popular prediction method, WoLF PSORT (Horton et al., 2007; cited 2648 times), showed that more than half of the TaLOGs localize to the cytosol. Additionally, both chloroplast‐ and mitochondria‐localized TaLOGs account for 20%, respectively. Notably, all of the TaLOG3 and TaLOG10 proteins, which are co‐located on the phylogenetic tree and appear to be more recently evolved members of the TaLOG gene family (Figure 2), are predicted to localize to the mitochondria.

It is likely that TaIPTs and AtIPTs will be similarly located, so it is not unexpected to find TaLOGs located in the chloroplast and mitochondria, in which the free base could be active in situ or from where the free base could diffuse into the cytosol. TaLOGs localized in the cytosol could convert cytokinin nucleotides produced in the cytosol, or act on cytokinins arriving in the transpiration stream including cytokinin nucleotides and ribosides (back converted to nucleotides by adenosine kinases (Sakakibara, 2021)).

Clearly, the discrepancy between in vivo and in silico analyses requires resolving through the careful use of reporter gene constructs, as GFP itself is reported to translocate to the nucleus (Seibel et al., 2007). Additionally, different attachment positions can lead to differences in protein subcellular location (Marion et al., 2008; Weill et al., 2019), although Gu et al. (2015) attached GFP to both LOG termini, with no apparent difference in location of LOG.

Role of LOGs in planta

Rice LOG was originally detected in meristematic mutants resulting in reduced yield (Kurakawa et al., 2007) and, as a consequence, there has been significant focus on meristematic tissues. Cytokinin biosynthetic, catabolic, receptor and signalling mutants have implicated cytokinin in SAM and root apical meristem (RAM) organization in arabidopsis (Bartrina et al., 2011; Gordon et al., 2009; Landrein et al., 2018; Osugi et al., 2017), as well as in vascular development (Miyashima et al., 2013; Ohashi‐Ito et al., 2014). Cytokinins are intimately involved in the CLAVATA/WUSCHEL‐negative feedback loop of meristem maintenance in eudicots (e.g. Bartrina et al., 2011; Chickarmane et al., 2012; Gordon et al., 2009; Landrein et al., 2018; Sablowski, 2009), with AtLOG4 expression restricted to the L1 (epidermal) layer of the SAM and floral meristem, and AtLOG7 expressed in the developing primordia (Chickarmane et al., 2012; Yadav et al., 2009).

The grafting experiments of Osugi et al. (2017) and Landrein et al. (2018) using multiple AtLOG and other mutants provide convincing evidence that stem cell homeostasis in the SAM is controlled by a systemic signal (tZR) produced external to the meristem, which is converted by LOG to the active free base form (tZ) in the SAM, to control the size of the SAM and plastochron number (Table 1).

However, it is the location of receptor and type‐B response regulators (RRs) that define the region of cytokinin action below the epidermal layer (Chickarmane et al., 2012), with Meng et al. (2017) suggesting type‐B ARRs bind to the WUS promoter. Recently, Yang et al. (2021) confirmed the intimate role played by cytokinin in driving mitosis and cytokinesis (i.e. controlling the cell cycle) in the arabidopsis SAM with cytokinin acting by changing the nuclear concentration of a transcription factor, thereby regulating cell numbers in the shoot stem cell niche.

Although CLV and WUS‐like genes have been identified in wheat, their detailed roles are only beginning to be examined (Li et al., 2019, 2020). However, Ohmori et al. (2013) suggest that there are two types of meristems in rice – vegetative (not affected by OsLOG mutants) and reproductive, which are affected by OsLOG mutants through a failure to maintain the floral meristem. As described earlier for rice, in situ hybridization shows OsLOG GFMs to be clearly located in meristematic regions. Cytokinin‐inducible RRs 1 and 5 were also expressed in the floral meristems of wild‐type rice, but not in the meristems of a log mutant, suggesting that active cytokinin was being produced in the wild‐type meristem. In terms of meristem maintenance in rice, OsLOG could control cell‐specific activation of cytokinin (Kurakawa et al., 2007), but its activity is likely spatially defined by receptor and/or signalling components.

Interesting examples of development implicating LOG include ectopic tuber formation on tomato (Eviatar‐Ribak et al., 2013), nodule formation (Azarakhsh et al., 2020; Mortier et al., 2014; Reid et al., 2017) and feminization of male kiwifruit plants (Varkonyi‐Gasic et al., 2021; Table 1).

Wheat LOGs express in multiple organs

Using the RNA‐seq data obtained by Ramírez‐González et al. (2018), we extracted all of the expression data sets for the TaLOGs from 70 tissues/organs. The data are displayed in a heatmap (Figure 6) and in detailed graphs (Figure S8) showing expression in roots, shoots, leaves and spikes across development. It is evident that TaLOG GFMs are expressed in multiple tissues, both vegetative and reproductive. TaLOG1 is constitutively and strongly expressed in all tissues and organs analyzed across development, including in the spike. TaLOG10 and LOG11 are also constitutively expressed, but much less strongly. There does not seem to be a strong correlation in terms of expression analysis between the TaLOGs and the aligned OsLOGs. For example, OsLOGL1 was barely expressed in rice (Kurakawa et al., 2007), but TaLOG1 is constitutively expressed in wheat; OsLOG10 was widely expressed including the leaf sheath but not leaf blade or roots, whereas TaLOG10 is constitutively expressed. However, the relative transcript abundances reported by Tsai et al. (2012) in rice in shoots and roots [OsLOG, LOGL5, LOGL7, LOGL9 and LOGL10 were greater in shoots relative to roots, whereas the transcripts of LOGL1, LOGL6 and LOGL8 had higher expression in roots] are somewhat replicated in wheat.

Figure 6

The developmental expression patterns of the TaLOGs. The expression data for each TaLOG gene family member were extracted from the published data (expVIP: http://www.wheat‐expression.com/, Ramırez‐Gonzalez et al., 2018), which leveraged 850 RNA‐seq samples. The data set was grouped into four tissue types using the R program and is shown on the top of the figure. The order of the genes was sorted using Linux commands and corresponds to the LOG phylogenetic tree shown in Figure 3. The expression data is shown as a heatmap using the image function in the R program. Detailed expression graphs are shown in Figure S8.

RNA‐seq data lack the cellular precision of promoter:reporter constructs but, interestingly, AtLOG8, which is the most closely associated GFM with TaLOG1 in Clade II, is the most widely expressed arabidopsis GFM, expressing in vascular tissue and other cells in roots and shoots (Kuroha et al., 2009). Clade IV AtLOG4 is also strongly expressed in vascular tissue, and, between AtLOG8 and 4, there is LOG expression in both root and shoot meristems. The constitutively expressed TaLOGs10 and 11 are also in Clade IV. Not surprisingly then, an RNA‐seq of extracted whole organs containing vascular tissue revealed expression. As observed by Kuroha et al. (2009) for arabidopsis and is also apparent in strawberry (Mi et al., 2017) and cotton (Wang et al., 2021), cytokinin can be synthesized in most parts of a plant via the direct LOG pathway.

Reproductive Tissues

Clearly there are TaLOGs expressing in both vegetative and reproductive shoot apical meristems (Figures 6 and S8). This was expected as Harrop et al. (2016), using laser microdissection, showed OsLOG to be expressed in the four meristem types (rachis meristem, primary branch meristem, elongating primary branch including axillary meristems and the spikelet meristem), whereas OsLOGL1 and LOGL6 (LABA1) were expressed more highly in the latter two meristem types (Harrop et al., 2016).

More recently, and again using NanoString technology, Yamburenko et al. (2017) monitored, among other cytokinin‐related genes, expression of all the rice LOG/LOGL GFMs, across five stages of panicle development in relation to the shoot apical meristem. Several LOGs exhibited greater expression in the SAM and during primary branch initiation (OsLOGL 2, 3, 4, 5, 7, 8) than at later stages. Peak expression of OsLOG occurred at the time when most branch and spikelet meristems had formed, and expression had decreased at the stage when floret meristems emerge from the spikelets, a stage when LOGL1, LOGL6, LOGL9 and LOGL10 were most strongly expressed, although these also expressed earlier in the SAM (Yamburenko et al., 2017).

In contrast to the transcript profiling by Feng et al. (2017) during wheat inflorescence development, which did not reveal expression of TaLOG GFMs, most wheat LOGs, with the exception of TaLOG6 and TaLOG2, expressed during spike and spikelet emergence. The differential expression shown by Yamburenko et al. (2017) between rice LOG GFMs during panicle development is not apparent in the wheat data during spike emergence due to the limited sampling of early stages of spike development. However, TaLOG9‐5A0834700 [a homologue of OsLOGL9 that strongly expressed at stage 5 when floret meristems emerge from the spikelets (Yamburenko et al., 2017)] showed strong activity in the spikes and spikelets, and in anthers at anthesis, with little expression in other organs. The expression of both TaLOG9‐5A0833800 and TaLOG9‐5B0873000, although relatively lower, was restricted to the spikes and spikelets. Earlier developmental stages, when meristematic tissues are present, were not included in the RNA‐seq samples. This is unfortunate, as elevated levels of TaIPT8 and the high expression levels of a β‐glucosidase GFM (TaGLU1‐1) and TaRR9 and 1 (Song et al., 2012), all indicate that this is a stage when cytokinin is likely to be critical.

However, wheat spikes are also known to be particularly susceptible to stress at booting, with low temperature stress impacting spikelet, floret and awn development (e.g. Zhang et al., 2019), and heat stress causing significant losses to floret fertility (grain set) and hence yield (e.g. Erena et al., 2021). All three TaLOG9 GFMs are, therefore, candidate targets for breeding, as these stresses may decrease cytokinin levels (Cortleven et al., 2019a). Moreover, the lack of expression of the TaLOG9 GFMs in roots and the shoot apical meristem at tillering could be useful in this context, as increased cytokinin levels generally impact root growth negatively (Werner et al., 2003). An increased tiller number in wheat is considered undesirable as multiple small tillers redirect resources away from the main tillers and ultimately decrease yield (Hendriks et al., 2016; Kebrom et al., 2012).

To enhance yield, one would anticipate increasing the cytokinin content of relevant tissues (Jameson and Song, 2016). While it will be challenging to find upregulated mutants of TaLOG9, mutations in an upstream negative regulator could be targeted. For example, UNBRANCHED3, a member of the SPL gene family, binds to the promoter and negatively regulates OsLOG and positively regulates OsCKX, among other genes (Du et al., 2017). Strong overexpression of UB3 reduced active cytokinin levels and negatively regulated tillering, panicle branching and grains per panicle in rice (Table 1). In the ub3::mum maize mutant line, which has additional kernel rows, Du et al. (2017) showed that genes associated with cytokinin biosynthesis were upregulated, and those associated with degradation downregulated. Du et al. (2017) concluded that UB3 regulates vegetative and reproductive branching by modulating cytokinin biosynthesis and signalling in both maize and rice. While multiple components interact to control yield in cereals, mutants of UB3 in wheat would be interesting to investigate.

In contrast to UB3, Li et al. (2021) recently identified a positive regulator of LOG. A japonica variety of rice, BS208, showed a highly abnormal panicle branching pattern, with reduced numbers of lateral grains on secondary branches. To determine the genetic mechanism underlying the abnormal phenotype, BS208 was crossed with Indica variety Teqing (TQ). Map‐based cloning using a segregating population with different grain numbers per panicle led to the identification of an R2R3 MYB transcription factor, REGULATOR OF GRAIN NUMBER1 (RGN1). Near isogenic lines (NIL) were developed: plants with the loss‐of‐function mutant allele, rgn1, displayed highly decreased grain number. OsLOG expression was downregulated in rgn1‐1 and NIL‐rgn1 compared with japonica cultivar Nipponbare and NIL‐RGN1 (Table 1), whereas the expression of LOG in RGN1‐OE3 plants was upregulated. Moreover, yeast one‐hybrid assays showed that RGN1 bound directly to the promoter region of OsLOG, and overexpression of OsLOG from Nipponbare in BS208 partially rescued the absence of lateral grains in secondary branches (Table 1), leading Li et al. (2021) to the conclusion that RGN1 affects lateral grain number and panicle architecture in rice by regulation of OsLOG expression.

Investigation of RGN1 sequence variations in wild and cultivated rice revealed two haplotypes, with accessions with the RGN1C allele exhibiting longer panicles and greater expression of LOG. They suggested that this elite haplotype RGN1C for panicle length identified in natural germplasm has potential to improve the yield of rice, but had not been selected during domestication, potentially due to linkage drag. This may have been associated with the elimination of undesirable allele(s) controlling awn formation preventing selection of the RGN1C allele during domestication (Li et al., 2021).

These data provide evidence of the critical role of OsLOG in controlling determinants of yield. We located three putative wheat RGN1 sequences in the NCBI database, most likely associated with the A, B and D sub‐genomes. This sequence is highly conserved in Aegilops tauschii (the progenitor of the D genome of bread wheat; Li et al., 2021) and is conserved in the first half of the RGN1 sequence in wheat (Table S4). Additionally, promoter motif scanning analysis showed that all 37 TaLOG GFMs have the highly conserved motif, ACCAAA (located at the end of the 18th motif of Figure S9), which was shown by Li et al. (2021) to be bound by OsRGN1. This indicates that TaLOG may be a conserved downstream target of RGN1. Identification of sequence variants, similar to that found in rice, should be of interest to wheat breeders.

Leaf tissues

TaLOGs, particularly TaLOG5 and TaLOG1, are expressed in leaf tissues at all stages of development, including in senescing flag leaves (Figures 6 and S8). The recent model suggested by Sakakibara (2021) shows tZR (and some tZ) moving in the xylem stream, with the tZR being converted initially to tZRMP, by adenosine kinase, prior to activation by LOG either in the SAM or in the leaf. Following activation of the two‐component signalling system, the model shows cytokinin positively impacting the leaf emergence rate and leaf maintenance. As several TaLOGs are strongly expressed in wheat leaves, and others in the SAM, targeting of the cytokinin produced is likely via positioning of receptors and/or response regulators (Werner et al., 2021).

Interestingly, Werner et al. (2021) recently showed that elevated AtLOG4 expression targeted to the epidermis significantly increased the size of the vegetative meristem but less so the inflorescence meristem. Leaf size and rosette size were increased as a result of increased cell number, with the authors suggesting cytokinin was acting on the epidermal cells in a partially layer‐autonomous fashion delaying their differentiation. Seed yield was also enhanced due to larger siliques with more seeds per silique, likely as a consequence of increased cytokinin synthesis in the outer layer of reproductive tissues and in the placenta, enhancing ovule formation (Table 1). They suggest that the increased seed yield through the expression of AtML1:LOG4 as a single dominant gene is of potential biotechnological interest (Werner et al., 2021).

Leaf development in monocots, with their long, narrow leaves with parallel veins, is quite different from that of the eudicots. However, not surprisingly, LOG is again implicated. In rice, OsWOX4 plays important roles in early leaf development through its expression in leaf primordia (Yasui et al., 2018). OsWOX4 functions upstream of LOG as in situ analysis showed that knockdown of OsWOX4 reduced the expression of OsLOGL3 and OsLOGL10 in developing vascular bundles. Consistent with this, cytokinin levels were downregulated by OsWOX4 knockdown and cell cycle progression was reduced (Yasui et al., 2018; Table 1). OsWOX4 is also involved in the maintenance of the vegetative meristem in rice (Ohmori et al., 2013; Yasui et al., 2018), whereas AtWOX4 is specifically associated only with the maintenance of vascular stem cells in arabidopsis, indicating diversification of function of the two orthologous genes between rice and arabidopsis.

Root tissues

Multiple TaLOGs are expressed in root tissues at various stages of development (Figures 6 and S8). Cytokinins have generally been considered to be negative regulators of root elongation and branching, and reduction in cytokinin levels through overexpression of CKX leads to enhanced root growth in multiple species (e.g. Ramireddy et al., 2021; and reviewed in Chen et al., 2020a). However, interestingly, decreased expression of CKX led to greater root mass in seedling roots of wheat and barley (Gasparis et al., 2019; Jablonski et al., 2020, 2021a; Zalewski et al., 2010). Cytokinins are critical to cambial activity in roots (Matsumoto‐Kitano et al., 2008), and the septuple AtLOG mutant showed severe retardation of root growth and defects in the RAM (Tokunaga et al., 2012; Table 1). Furthermore, Ohashi‐Ito et al. (2014) showed an arabidopsis bHLH transcription factor complex targets AtLOG3 and 4 in xylem precursor cells in the RAM, with the produced cytokinin acting as an initial signalling event that induces cell proliferation in adjacent procambial cells (Table 1). De Rybel et al. (2014) also showed that AtLOG4 is a direct target gene of the TMO5/LHW transcription complex and plays a critical role in establishing vascular patterning (Table 1). Moreover, cytokinin, through the inhibitory activity of AtIPT and AtLOG4, is intimately involved in determining the spacing of lateral roots (Bielach et al., 2012; Chang et al., 2015). Consequently, it is not surprising to find TaLOG GFMs expressing in wheat roots.

OsLOG mutants unexpectedly show increased yield

OsLOGL genes have been the target of selection during the domestication of rice (Gu et al., 2015; Hua et al., 2015). The wild progenitor of rice, O. rufipogon, has many long, barbed awns, fewer tillers per plant and fewer grains per panicle in contrast to cultivated rice. An‐1 and An‐2 are two quantitative trait loci (QTLs) associated with awn formation and awn elongation, respectively, and both loci are considered to have been targets of artificial selection during rice domestication (Gu et al., 2015; Hua et al., 2015; Luo et al., 2013). An‐1 positively regulates the formation of awn primordia and negatively regulates grain number (Gu et al., 2015; Luo et al., 2013). Gu et al. (2015) fine mapped An‐2 to the LOGL6 gene.

Gu et al. (2015) showed that An‐2 exhibited a pleiotropic effect during rice development and morphology: it promoted awn elongation by enhancing cytokinin levels and cell division. Additionally, grain production was decreased through fewer grains per panicle and fewer tillers per plant; root length was also decreased (Table 1). An‐2 was mainly expressed in young panicles, branching crowns and roots, and specifically in the epidermal cells and vascular bundles of awns. Complementation and overexpression experiments confirmed that a LOG homologous gene was An‐2. Further, this LOG encoded a functional OsLOGL6 in wild rice. The an‐2 allele has a C‐nucleotide deletion in exon 1 that resulted in a frameshift and a premature non‐functional protein that had lost nearly all the conserved sequences. They concluded that the An‐2 locus was probably a target of artificial selection during rice domestication, for suppression of awn elongation and/or promotion of yield (Gu et al., 2015). Additionally, they concluded that the functions of homologous LOG enzymes have differentiated, as the expression pattern of An‐2 in panicles was more dispersed compared with LOG. Moreover, the log mutant of rice had fewer grains per panicle and abnormal spikelets. Gu et al. (2015) suggested this differentiation is controlled by regulatory regions rather than protein functional variations.

In a parallel study, Hua et al. (2015) fine mapped the QTL LONG AND BARBED AWN1 (LABA1). Complementation and RNAi interference of the putative gene led Hua et al. (2015) to the conclusion that the LABA1 gene ‘conditions’ both awn length and the presence of barbs (Table 1). LABA1 was preferentially expressed in developing panicles, 1–2 cm in length, notably in young epidermal cells of awn primordia, with lesser expression in vegetative tissue and the awn itself after heading. The laba1 mutation (a frameshift deletion) was associated with short, barbless awns, with increased unit weight and was selected during rice domestication (Hua et al., 2015). Functional analyses showed LABA1 encodes cytokinin nucleoside 5'‐monophosphate phosphoribohydrolase activity, with sequencing indicating it to be a rice LOG homologue. Hua et al. (2015) concluded that LABA1 encodes the cytokinin‐activating enzyme OsLOGL6 that increases cytokinin content and cell division in epidermal cells of the awn primordia in wild rice.

Wang et al. (2020) selected a drought‐tolerant rice activation tagged line with 100% amino acid match to OsLOGL5. Focusing on the fact that there are diverse N‐terminal and C‐terminal ends among the LOG GFMs, they developed a series of gene edits targeting the final 25 amino acids of the C‐terminal end of the protein. Tagged lines overexpressing OsLOGL5 were semi‐dwarfed and showed reductions in primary root growth, fertile panicles, seeds per panicle and seed weight, with yield significantly reduced under drought conditions. Gene‐edited lines with mutations at the 3′‐end of the coding sequence resulted in normal vegetative growth, and with increased yield in the field under normal conditions as well as under drought and low N (Table 1). Under drought, this was a consequence of increases in grain setting rate, total grain number and 1000 grain weight. Of the six gene‐edited lines, two showed an increase in tiller number, while two others had fewer tillers. Wang et al. (2020) concluded that OsLOGL5 was likely to be involved in both seed development and grain filling and that the C‐terminal end of the protein was important in the response of OsLOGL5 to abiotic stress conditions.

In the above cases, it is apparent that reduced expression of OsLOGL6 and OsLOGL5 leads to increased yield (Table 1).

Response to biotic and abiotic stressors

Cytokinins are generally considered to be negative regulators of stress (e.g. Nishiyama et al., 2011; Pavlů et al., 2018; Tran et al., 2010). Wang et al. (2021) showed that expression levels of GhLOG were modulated by drought, high salt and hormone stress. They suggested that overexpression of GhLOG3_At improved tolerance of cotton to NaCl potentially because root growth was less inhibited by salt relative to control roots, whereas reduced expression enhanced sensitivity (Wang et al., 2021), a result that appears contrary to cytokinins as negative regulators of stress.

In wheat, those LOG GFMs constitutively expressed in all organs (TaLOG1, TaLOG10 and TaLOG11; Figure 6) showed little response to biotic or abiotic stressors (Figure 7). TaLOG6 showed no response. TaLOG5 GFMs, two of which were strongly expressed in leaves (Figures 6 and S8), showed only a low level of expression in response to the biotic and abiotic stressors. Overall, there were no consistent patterns of up‐ or down‐regulation in response to various stressors; for example, in terms of drought, both upregulated and downregulated GFMs can be seen. In a similar fashion, the analyses conducted by Pavlů et al. (2018) showed little overlap in the expression patterns between AtLOG GFMs and a wide range of stress‐associated genes, in contrast to expression of IPT and signal transduction components.

Figure 7

Expression response pattern of the TaLOGs under biotic and abiotic stressors. The expression data were extracted from the expVIP database (http://www.wheat‐expression.com/) and shown as fold change. The expression values are shown in fold change (FC). FC is defined as: FC = Control/Treat. In order to avoid a meaningless calculation when the denominator was zero, 0.001 was added to both the numerator and denominator. The FC data were normalized using log2 transformation. The gene order corresponds to the LOG phylogenetic tree (Figure 3). The expression data are shown as a heatmap using the image function in R program.

LOGs are intron‐rich

Jeffares et al. (2008) suggest that there has been selection against introns in genes that require rapid expression changes. They showed that genes whose expression changed rapidly in response to stress contained significantly lower intron densities. Specifically relating to plants, Liu et al. (2021) speculate that loss of introns may be a factor in early land plant evolution allowing fast response to stress conditions. They also suggested loss of introns may have facilitated faster developmental adaptations to life on land.

Generally, the reported LOG GFMs appear to be relatively intron‐rich (having ≥4 introns; Liu et al., 2021). The LOG gene of rice has seven exons and six introns (Kurakawa et al., 2007), Strawberry (Mi et al., 2017) and Medicago truncatula LOGs have six to seven introns (Azarakhsh et al., 2020). While 10 of the 11 wheat LOGs are relatively intron‐rich (with between five and six introns), TaLOG5 on all sub‐genomes is intron‐poor (Figure 8). Notably, OsLOGL5 is also reported to have no introns (Wang et al., 2020), in contrast to OsLOG. The single TaLOG6 has six introns, with one being extremely long. The intron pattern is in general agreement with Wang et al. (2021) where most cotton LOGs had six introns, although one GFM had an extremely long intron.

Figure 8

The intron–exon gene structure pattern of TaLOGs. The gene intron–exon structure information was extracted from the gff3 file of wheat genome (Version 2.1, Zhu et al., 2021) using Linux shell commands. The intron–exon gene structure pattern was drawn in TBtools (Chen et al., 2020b).

The generally intron‐rich character of LOGs is in contrast to the intron‐poor character of the cytokinin glucosyl transferase (CGT) gene family (Chen et al., 2021a), which is considered to respond fast to stress (Vyroubalová et al., 2009) and to decrease active cytokinin levels (Chen et al., 2021a). Intron numbers for TaCKX are somewhat in between, ranging from 0 to 4, with the majority of GFMs having either 2 or 4 introns (Figure S10). Indeed, several CKX and CGT gene family members are regarded as primary response genes in arabidopsis (e.g. Brenner et al., 2012; Zubo and Schaller, 2020). However, TaLOG5 GFMs (0‐1 intron) showed little response to biotic or abiotic stress (Figure 7) and did not have more over‐represented stress‐associated cis‐regulatory elements (CREs) than other TaLOGs (Figure 9), and so potentially do not follow the general rule relating to intron‐poor genes responding rapidly to stress.

Figure 9

Cis‐regulatory element distribution patterns on 3 kb potential TaLOG promoter. The 3 kb potential promoter sequences were extracted from the wheat genome sequence (Zhu et al., 2021) via Linux shell commands. The cis‐regulatory elements (CRE) were predicted in plantCARE database (Chou and Shen, 2007; 2008; 2010). The heatmap was made based on a comparative analysis of CREs on the TaLOGs with a representative promoter CRE data set of 100 genes chosen randomly from the wheat genome as described in Chen et al. (2021a). We calculated a value of FC (Fold Change, TaLOG CRE number/average CRE number) to confirm over‐ (FC > 1) or under‐represented (FC < 1) CREs. The FC values were normalized by log2 calculation and developed into the heatmap using the R program.

Long introns have been considered to be costly in terms of transcription (Seoighe et al., 2005). The very long intron in TaLOG6D, and the loss of TaLOG6A and B (Figure 8), possibly reflects selection against this gene. TaLOG6D was not expressed during development nor responsive to stress (Figures 6 and 7). We subsequently discovered that TaLOG6 is likely truncated on the A and B genomes of Durum wheat and, as such, was not captured in our analysis of bread wheat. As TaLOG6D protein has high sequence similarity and possibly similar function to OsLOGL6/LABA1 in rice, the long intron might be the result of unconscious selection events towards the reduction and/or elimination of its negative effects on grain number as shown for OsLOGL6 in rice, leading to loss‐of‐function of the gene. Selection might have induced the complete elimination of the non‐functional sequence fragments of this gene loci on chromosomes A and B. We noted that TaLOG6D protein aligns closely to AtLOG6 which was shown to be non‐functional. Interestingly, no MtLOGs or PtLOGs aligned with AtLOG6, and FvLOG3 (homologue of AtLOG6) was not expressed in strawberry (Mi et al., 2017), so it is possible that this LOG is non‐functional across the angiosperms.

Cis‐regulatory element analysis

We conducted an analysis of the cis‐regulatory elements (CRE) on 3 kb of the putative promoters of the TaLOG GFMs, as this has been used as an indicator of responsiveness of genes to various signals. Here, we grouped the CREs according to responsiveness to abiotic, biotic, hormone and ‘internal control’. The heatmap shows whether the various elements are over‐ or under‐represented relative to 100 randomly selected wheat genes (Figure 9).

There are several consistent patterns among the CREs on the promoters of TaLOG GFMs: light‐responsive elements and anaerobic‐induction response elements were over‐represented, whereas CMA3, cell cycle and binding site of ATBP.1 were under‐represented (Figure 9).

As cytokinin is intimately involved in the response of the plant to the light environment (reviewed in Cortleven et al., 2019a; Ikeda et al., 2021; Pavlü et al., 2018), it is not surprising that light‐responsive elements are over‐represented. CMA3 (conserved DNA module array) is a binding site for 3AF3, which, along with other proteins, activates light transcription of the small subunit of RUBISCO (Sarokin and Chua, 1992). As this cis element is under‐represented, it implies expression of TaLOGs/cytokinin is not required for this particular process.

ATBP‐1 (AT‐rich DNA binding protein‐1), first identified by Tjaden and Coruzzi (1994), binds to AT‐rich promoter elements. These elements are considered to play a role in transcription of plant genes. As they are under‐represented on the TaLOG promoters, we infer that the expression of the TaLOG genes is under the regulation of specific regulators, rather than a common regulator.

An anaerobic environment is one totally lacking in free or bound oxygen. Anaerobic‐induction response elements are over‐represented on the TaLOG promoters. Interestingly, during a genome‐wide association study (GWAS), OsLOGL8 was highly associated with anaerobic germination in rice (Hsu and Tung, 2015). Maintenance of the RAM under such conditions is essential.

The MYBHv1‐binding site is generally underrepresented. MYBHv1 is a positive regulator of drought tolerance and is associated with mediating the action of abscisic acid (ABA; Alexander et al., 2019). Lee et al. (2007) showed the wheat orthologue (TaMYB1) to be upregulated by abiotic stress in wheat roots, and overexpression of the wheat paralogue TaMYB1D in tobacco enhanced drought and oxidative stress resistance (Wei et al., 2017). However, the MYB drought response element is generally over‐represented on the TaLOG promoters, although there is evidence to suggest that some of these elements may be negative regulators of stress responses (e.g. Baldoni et al., 2015; Gao et al., 2016). As cytokinin levels and signal transduction are generally reduced during drought (e.g. Nishiyama et al., 2011; Nguyen et al., 2016; reviewed in Cortleven et al., 2019a), negative regulation would appear likely.

Elements responsive to ABA, gibberellin and auxin are generally averagely to under‐represented on the TaLOG promoters (Figure 9). As this is in marked contrast to the over‐representation of these CREs on the CGTs (Chen et al., 2021a), we also analyzed the CREs on the TaCKX promoters (Figure S11). Interestingly, the patterns of CREs on the TaCKX promoters matched those on the TaLOG promoters, which reinforces suggestions in the literature that CKX plays a strong role in maintaining cytokinin homeostasis (refer Chen et al., 2020a). However, the well‐known antagonism between cytokinin and ABA (Zubo and Schaller, 2020) would appear to be better reflected in the over‐representation of ABA‐REs on the promoters of the TaCGTs (Chen et al., 2021a) but not on TaLOGs or TaCKXs. This may reflect the conservation (rather than destruction) of cytokinin during stressful conditions, and its availability for subsequent release.

Jasmonates (JA) are involved in many aspects of plant development and are viewed as part of the growth‐defence trade‐off (Howe et al., 2018). They are key factors in the plant’s response to wounding by chewing insects including in rice (Fu et al., 2021), resistance to necrotrophs (e.g. Chen et al., 2021b) and certain biotrophs (Figueiredo et al., 2017). Interactions between cytokinin and jasmonates are generally viewed as antagonistic (Pavlů et al., 2018; Zubo and Schaller, 2020), and a second obvious contrast in the CREs relates to the JA‐REs. These are over‐represented on the TaCGT promoters and under‐represented on both the TaLOG and TaCKX promoters, again indicating control of cytokinin levels during stress responses through glucosylation rather than through transcription of TaLOG or destruction of cytokinin by CKX. This conclusion is supported by the overall low level and inconsistent responses of the TaLOGs to biotic stressors (Figure 7).

In contrast to the above, the light‐REs were over‐represented on the promoters of the TaLOGs, TaCKXs and TaCGTs, potentially reflecting the critical involvement of cytokinin in light‐regulated responses (e.g. Cortleven et al., 2019a, 2019b; Pavlů et al., 2018).

Two steps or just the one?

In a recent paper, Nguyen et al. (2021b) endeavor to reactivate the two‐step pathway as well as suggesting that the cytokinin ribosides may be active compounds and have significant roles beyond being transport forms. Interestingly, they do not refer to Lomin et al. (2015) who used a plant‐based system to confirm that only the free bases and not the ribosides interacted with the cytokinin receptors, in contrast to previous receptor work where bacterial or yeast‐based systems were used (see references in Nguyen et al., 2021b). Indeed, they cite a yeast complementation assay that showed binding to receptors not only of free bases and ribosides but also of nucleotides and O‐glucosides (Daudu et al., 2017) and infer that Brassica napus receptors (Kuderová et al., 2015) also bind a wide range of cytokinin forms (see Nguyen et al., 2021b). This inference by Nguyen et al. (2021b) is in contrast to that by Hluska et al. (2021) who state that O‐glucosides or ribosides exert cytokinin activity in various bioassays, but O‐glucosides do not bind to receptors at all and ribosides very weakly. They are active because they can be metabolized to the active forms. Moreover, structural biology assessments of the ligand‐binding sensor modules of the cytokinin receptors indicate free bases are a ‘good fit’ but not so for ribosides (Hothorn et al., 2011; Lomin et al., 2015); binding of nucleotides and O‐glucosides to cytokinin receptors is, therefore, highly improbable.

More worryingly, Nguyen et al. (2021b) cite several papers in which cytokinin ribosides were applied to plants and suggest that the impacts on growth and/or development were due to the riboside itself. This ignores the metabolism papers from the Letham and Wareing labs that show that exogenously supplied cytokinins are rapidly converted to the nucleotides in various plants: for example 3H‐ZR/3H‐DZR/3H‐Z supplied to Lupinus angustifolius (e.g. Jameson et al., 1987; Parker et al., 1978; Zhang et al., 2002) and L. luteus (Knypl et al., 1985); 3H‐Z/3H‐ZR supplied to soybean (e.g. Singh et al., 1988); 3H‐Z supplied to Zea mays (e.g. Parker and Letham, 1974) and radish (Gordon et al., 1974). Early work from Wareing et al. (1977) indicated that 8‐14C zeatin supplied to roots and buds of Phaseolus  vulgaris was converted to zeatin nucleotide rather than to glucosides, and 14C‐dihydrozeatin supplied to stem segments of P. vulgaris was rapidly metabolized to dihydrozeatin nucleotide (Palmer et al., 1981). Indeed, in the seminal paper testing functionality of arabidopsis LOGs, iPR was supplied to arabidopsis on the expectation of conversion to nucleotide and subsequent direct conversion by LOG to active free bases (Kuroha et al., 2009). While endeavoring to reactivate the two‐step pathway, Nguyen et al. (2021b) do admit that strong and specific activity towards cytokinins of the enzymes of this pathway has yet to be confirmed (see references cited in Nguyen et al., 2021b; Figure 1). The most recent model presented by Sakakibara (2021) shows xylem‐translocated tZR converted by adenosine kinase to tZRMP and thence by LOG to tZ to control both leaf emergence rate and leaf expansion. The recent commentary by Romanov and Schmülling (2022) is a critique of Nguyen et al. (2021b) and concludes that the free bases and not the ribosides are the active forms of cytokinins.

Conclusion

Structural biology comparison of wheat LOGs with arabidopsis LOGs indicates that wheat LOGs 1‐11 likely form dimers. Sequence and expression analyses indicate that TaLOGs 1‐5 and 7‐11 are functional cytokinin riboside 5'‐monophosphate phosphoribohydrolases. We suggest that the main homeostatic mechanism controlling cytokinin levels resides with CKX but, in response to biotic and abiotic challenges, the activities of the CGTs, rather than CKX or LOG, become more important. The TaLOGs variously expressed in all tissues analyzed and those tissues constitutively expressing LOGs may provide the plant with basal levels of cytokinin (Hluska et al., 2021). Even though cytokinins are generally regarded as negatively impacting root growth, a basal level of cytokinin is required for maintenance of the RAM (Tokunaga et al., 2012), as well as cytokinin being implicated in lateral root spacing (Chang et al., 2015). Even senescing leaves need to maintain basic mitochondrial functionality and energy production well into senescence (Kim et al., 2018). The more subtle roles for cytokinin, such as implicated in the SAM, are more dependent on the precise positioning and activity of RRs and other components of the signal transduction pathway, than on active cytokinin forms per se.

However, results such as those from the overexpression of AtLOG4 in epidermal cells leading to enhanced yield components (Werner et al., 2021) provide avenues for future investigation. Interestingly, selection during domestication of a mutation in rice LOGL6 is implicated in increased yield (Gu et al., 2015; Hua et al., 2015), and overexpression of OsLOGL5 led not only to a reduction in primary root growth, as might be expected, but also to a reduction in yield. Further, gene editing the C‐terminal of OsLOG5 conferred enhanced stress tolerance and enhanced yield (Wang et al., 2020). Consequently, targeted mutations in wheat LOGs may provide useful material for breeders.

Conflicts of interest

The authors declare they have no conflicts of interest.

Author contributions

L.C. interrogated the wheat genome databases and prepared relevant figures. G.B.J. provided the structural biology analyses. Y.G. collected the raw seq data of LOG and the RNA‐seq data, and did some manual pre‐processing of the raw data sets. J.S. interrogated some phylogenetic data and prepared the relevant figures. P.E.J wrote the first draft of the manuscript with input subsequently from all authors.

Figure S1 The cluster pattern of the unrooted LOG phylogenetic tree clades.

Figure S2 The rectangular pattern of the unrooted LOG phylogenetic tree.

Figure S3 Detailed information of the TaLOG motif patterns.

Figure S4 Conserved domain analysis of TaLOGs.

Figure S5 Conserved domain analysis of TaLOGs, TaLOG12, AtLOG Type IIb and MtLOGL3.

Figure S6 Motif patterns of MtLOGs, AtLOGs, OsLOGs and MtbLOG.

Figure S7 The detailed information for the motif patterns of MtLOGs, AtLOGs, OsLOGs and MtbLOG.

Figure S8 RNA‐seq graphs for each of the TaLOG gene family members.

Figure S9 Detailed information of the TaLOG promoter motif patterns.

Figure S10 The intron‐exon gene structure pattern of TaCKX.

Figure S11 Cis‐regulatory element distribution patterns on 3 kb potential TaCKX promoter.

Table S1 TaLOG gene family members: gene ID, positions, symbols, and homologues.

Table S2 Percent identity of TaLOG2, 6, 1 and 12 with AtLOG3 and 8, and other sequences.

Table S3 Cellular localization prediction of TaLOGs.

Table S4 Rice, Aegilopes tauchii, and bread wheat RGN1 amino acid sequence alignment.

Click here for additional data file.