Spatial expression patterns of genes encoding sugar sensors in leaves of C4 and C3 grasses

Abstract Background and Aims The mechanisms of sugar sensing in grasses remain elusive, especially those using C4 photosynthesis even though a large proportion of the world's agricultural crops utilize this pathway. We addressed this gap by comparing the expression of genes encoding components of sugar sensors in C3 and C4 grasses, with a focus on source tissues of C4 grasses. Given C4 plants evolved into a two-cell carbon fixation system, it was hypothesized this may have also changed how sugars were sensed. Methods For six C3 and eight C4 grasses, putative sugar sensor genes were identified for target of rapamycin (TOR), SNF1-related kinase 1 (SnRK1), hexokinase (HXK) and those involved in the metabolism of the sugar sensing metabolite trehalose-6-phosphate (T6P) using publicly available RNA deep sequencing data. For several of these grasses, expression was compared in three ways: source (leaf) versus sink (seed), along the gradient of the leaf, and bundle sheath versus mesophyll cells. Key Results No positive selection of codons associated with the evolution of C4 photosynthesis was identified in sugar sensor proteins here. Expressions of genes encoding sugar sensors were relatively ubiquitous between source and sink tissues as well as along the leaf gradient of both C4 and C3 grasses. Across C4 grasses, SnRK1β1 and TPS1 were preferentially expressed in the mesophyll and bundle sheath cells, respectively. Species-specific differences of gene expression between the two cell types were also apparent. Conclusions This comprehensive transcriptomic study provides an initial foundation for elucidating sugar-sensing genes within major C4 and C3 crops. This study provides some evidence that C4 and C3 grasses do not differ in how sugars are sensed. While sugar sensor gene expression has a degree of stability along the leaf, there are some contrasts between the mesophyll and bundle sheath cells.


INTRODUCTION
Given that C 4 species fix carbon and synthesize carbohydrates using a two-cell system compared with C 3 species, which use a single cell, it remains unclear if the sugars produced are sensed differently between them.C 4 photosynthesis evolved ~35 million years ago in response to a period of low atmospheric CO 2 , evolving in 62 independent lineages (Sage, 2004(Sage, , 2017;;Sage et al., 2011).Many agronomically important cereals, such as Zea mays (maize), Sorghum bicolor (sorghum), Panicum virgatum (switchgrass) and millets (such as Setaria italica), utilize C 4 photosynthesis.This evolution has led to major changes in gene expression, leaf morphology, biochemistry and the compartmentalization of photosynthetic reactions (Dengler and Nelson, 1999;Von Caemmerer and Furbank, 2003;McKown and Dengler, 2007;Muhaidat et al., 2007;Emms et al., 2016;Furbank and Kelly, 2021).
This compartmentalization of C 4 photosynthesis is enabled by a specialized leaf anatomy, known as Kranz anatomy, whereby mesophyll cells are arranged in a concentric layer around the bundle sheath cells (Haberlandt, 1904;Hattersley, 1984).In the mesophyll of C 4 leaves, CO 2 is hydrated into bicarbonate and is initially fixed by phosphoenolpyruvate (PEP) carboxylase (PEPC), using PEP as a CO 2 acceptor (Hatch and Slack, 1966;Furbank and Hatch, 1987).Oxaloacetate (OAA) is then produced and rapidly converted to two possible C 4 acids, malate or aspartate.These acids diffuse to the bundle sheath via the abundant plasmodesmatal connections, where they are decarboxylated, releasing CO 2 to be refixed by ribulose-1,5bisphosphate carboxylase-oxygenase (Rubisco) (Danila et al., 2016(Danila et al., , 2018)).The compartmentalization of the photosynthetic enzymes, the high PEPC/Rubisco activity ratio and the low permeability of the bundle sheath cell wall elevate CO 2 concentration around Rubisco, leading to near CO 2 saturation and reduced photorespiration (Hatch, 1987;Ghannoum et al., 2000;Von Caemmerer and Furbank, 2003;Danila et al., 2021).
During the evolution of C 4 photosynthesis, the expression of numerous genes was adjusted to enable distinct spatial separation, or altered regulation, relative to expression patterns seen in species that utilize C 3 photosynthesis (Hibberd and Covshoff, 2010;Westhoff and Gowik, 2010;Christin and Osborne, 2014).Some examples include the targeted expression of PEPC and the confinement of carbonic anhydrase to the mesophyll cell in C 4 plants as well as differences in Rubisco catalytic efficiencies between the two photosynthetic types (Gowik et al., 2004;Tetu et al., 2007;Tanz et al., 2009;Whitney et al., 2011;Ludwig, 2016).The evolution of photosynthesis into a two-cell process in C 4 plants has also resulted in the spatial partitioning of carbohydrate production.One of the main products of photosynthesis, triose phosphate, is used in the synthesis of soluble sugars such as glucose and sucrose, substrates that can then be synthesized into the storage carbohydrate starch.Generally, in C 4 plants sucrose biosynthesis occurs in the mesophyll, while starch synthesis occurs predominantly in bundle sheath chloroplasts (Lunn andFurbank, 1997, 1999;Lunn, 2007;Furbank and Kelly, 2021).In leaves of C 3 species these processes occur almost exclusively in the mesophyll.Carbohydrates are moved from the photosynthetic source leaves to the heterotrophic sink tissues such as seeds, stems, roots and young leaves for growth and development.
Photosynthesis and sink demand are tightly coordinated through metabolic feedback and signalling mechanisms (Blechschmidt-Schneider et al., 1989;Sheen, 1990).Sugar signalling integrates sugar production with plant development and environmental cues (Rolland et al., 2006).To date, there is a limited understanding of the molecular mechanisms underlying these feedback regulations in C 4 plants.C 4 species evolved in arid and warmer climates, conditions that may have also imposed specific selective pressures on aspects of sugar sensing.There is also evidence showing that the high photosynthetic activity in C 4 leaves can lead to the accumulation of higher levels of sugars, relative to C 3 species (Henry et al., 2020).Carbohydrate synthesis, metabolism and export differ in several ways between C 4 and C 3 photosynthetic species.As mentioned, sucrose and starch synthesis is compartmentalized in leaves of C 4 grasses (Lunn and Furbank, 1999).In addition, large metabolite pools are required to maintain a high concentration gradient across the mesophyll-bundle sheath interface with higher plasmodesmatal connections in C 4 grasses, allowing fast metabolite exchange and efficient carbon concentration around Rubisco (Leegood, 2002;Danila et al., 2016).Part of the 3-phosphoglycerate (PGA) produced by the Calvin cycle in bundle sheath cells is reduced in mesophyll cells due to lower photosystem II activity in the bundle sheath.Furthermore, there has been recent evidence that suggests that C 4 grasses have evolved sugar transporters, using a different strategy compared with C 3 grasses (Emms et al., 2016;Bezrutczyk et al., 2018;Chen et al., 2022;Hua et al., 2022).These factors suggest that sugar sensing may differ between C 4 and C 3 plants, and between the mesophyll and bundle sheath cells.
Three putative sugar sensor kinase proteins are known: target of rapamycin (TOR), SNF1-related kinase 1 (SnRK1), hexokinase (HXK) and the sugar-sensing metabolite trehalose-6-phosphate (T6P).TOR functions as a protein kinase and is a part of the TOR complex (TORC), which also includes RAPTOR (regulatory-associated protein of TOR 1) and LST8 (lethal with sec thirteen 8).These additional proteins can act as regulatory components of the TORC (Xiong and Sheen, 2014).In Arabidopsis thaliana, it has been established that the glucose-TOR signalling network can regulate numerous essential processes (Xiong and Sheen, 2012;Xiong et al., 2013).Hexokinase was one of the first proteins for which a direct link between sugar sensing and photosynthesis was established (Moore et al., 2003).Several homologues, such as AtHXK1 in Arabidopsis and OsHXK5 and OsHXK6 in rice (Oryza sativa), have been established as sugar sensor proteins (Moore et al., 2003;Cho et al., 2009a).There has also been evidence for SnRK1 as a sugar-sensing protein in plants (Jossier et al., 2009).The SnRK1 complex (SnRK1C) is made up of four subunits -the catalytic subunit (α), two regulatory subunits (β,γ) and a hybrid plant-specific subunit (βγ) -and can be involved in plant-pathogen interactions (Bouly et al., 1999;Lumbreras et al., 2001;Gissot et al., 2005Gissot et al., , 2006)).SnRK1 is thought to be upregulated when conditions are unfavourable for the plant (Baena-González et al., 2007;Zhang et al., 2009).There has been some evidence that SnRK1 is involved in the regulation of photosynthesis genes as the overexpression of KIN10 (the gene encoding the SnRK catalytic subunit in Arabidopsis) causes a downregulation of photosynthetic genes.Furthermore, SnRK1 is inhibited by T6P (the precursor to the disaccharide trehalose), but not by other sugars.In plants, T6P can only be made when sufficient levels of sucrose are present (Lunn et al., 2006), and therefore acts as a signalling molecule for sucrose and correlates with active growth (Schluepmann et al., 2003;Martínez-Barajas et al., 2011;Lunn et al., 2006Lunn et al., , 2014)).Trehalose phosphate synthase (TPS) is responsible for the synthesis of T6P, and T6P can subsequently be converted to trehalose via trehalose phosphate phosphatase (TPP) (Ponnu et al., 2011;Paul et al., 2020).Trehalose can then be broken back down into its glucose units by trehalase (TRE).While trehalose is found at relatively low levels in plants, it is thought that trehalose metabolism plays an important regulatory role (Goddijn and Smeekens, 1998).
Due to innate differences and the complexity of the signalling network, it is plausible to hypothesize that photosynthetic types may sense sugars differently.Sugar sensors may have evolved to accommodate the two-cell compartmentalization of C 4 photosynthesis.In this study, publicly available transcriptome data from C 3 and C 4 grasses were used to investigate the expression of putative genes encoding components of each sugar sensor.The overall aim was to determine if there were differences in expression patterns between C 3 and C 4 grasses that might alter how sugar is perceived.Data were used to (1) determine if there were C 4 -specific residues in the sugar-sensing genes associated with the evolution of C 4 photosynthesis, and (2) compare the transcript abundance (a) between the leaf (source) and seed (sink) in C 4 and C 3 grasses, (b) along the leaf gradient of C 4 and C 3 grasses, where a single leaf undergoes a sink (base)-source (tip) transition during development (Jones and Eagles, 1962;Turgeon and Webb, 1976;Harn et al., 1993;Kölling et al., 2013;Wang et al., 2014;Chen et al., 2022), and (c) between the bundle sheath and mesophyll cells of C 4 grasses, to investigate whether there was preferential expression to one photosynthetic cell type.This approach can shed light on those sugar sensors that might be linked with photosynthesis (source tissue).

MATERIALS AND METHODS
C 4 and C 3 grass species utilized Sequences, transcript expression data or both were extracted from eight C 4 grasses (Panicum hallii, Panicum miliaceum, Panicum antidotale, Sorghum bicolor, Setaria italica, Setaria viridis, Saccharum spontaneum and Zea mays) and six C 3 grasses (Steinchisma laxum, Hymenachne amplexicaulis, Cyrtococcum patens, Panicum bisulcatum, Brachypodium distachyon and Oryza sativa).The species, transcriptomes and raw RNA sequencing data that were used for each aspect of this study are summarized in Supplementary Data Table S1.
De novo assembly of RNA-sequencing reads A de novo transcriptome assembly was built for those species that had no publicly available genomes at the time of analysis using published RNA-sequencing (RNA-Seq) data.All RNA-Seq data sets were obtained from https://www.ncbi.nlm.nih.gov/ or https://www.ebi.ac.uk/ using the project's associated accession number (Supplementary Data Table S1).Adapter sequences were first removed from all RNA-Seq reads using Trimmomatic (Bolger et al., 2014).The Trinity default pipeline was then implemented to create the de novo assemblies, each consisting of a set of contiguous sequences (contigs) for each species (Grabherr et al., 2011;Haas et al., 2013).For each de novo assembly, annotation of contigs was required to identify genes of interest.For this, de novo assemblies were loaded into Geneious Prime, 2022.2 (https://www.geneious.com;Kearse et al., 2012), and nucleotide databases were created using the inbuilt NCBI BLAST tool.The NCBI tool within Geneious Prime 2022.2 was then used to carry out BLASTN queries, using sequences of genes of interest from S. viridis (Bennetzen et al., 2012) and O. sativa (Ouyang et al., 2007) transcriptomes, for identification.

Estimation of transcript abundance
RNA-Seq reads obtained (Supplementary Data Table S1) were quantified using the quasi-align mode in Salmon (Patro et al., 2017).A mapping-based index was created for each transcriptome or de novo transcriptome.Trimmed reads were then mapped to the relevant index using the default settings of the quant command in mapping-based mode within Salmon.This produced normalized transcripts per million (TPM) values for each transcript or contig (Kearse et al., 2012).

Positive selection analysis using CodeML
To investigate for evidence of positive selection in genes of interest, CodeML was implemented to test residues for selection (Zhang et al., 2005).For this analysis, species were selected to ensure C 4 lineages were dispersed phylogenetically between C 3 species (Supplementary Data Table S1).Phylogenetic trees created using RaxML were processed as Newick files and annotated (using #1 to denote a foreground branch) to test the hypothesis of selection in foreground branches.Those branches labelled as foreground were those that contain a C 4 species, hence the following analyses test whether there is any positive selection associated with the evolution of C 4 photosynthesis (C 4 -specific selection).CodeML was then used to test the ratio of non-synonymous to synonymous substitutions (dN/ dS ratio; ω) under two scenarios: (1) a null model where all codons evolve under either purifying selection (ω < 1) or relaxed selection (ω = 1); and (2) sites evolve under purifying or neutral selection in the whole tree, except for foreground branches, where they evolve under positive selection (ω > 1).These scenarios were compared to estimate the posterior probability of each base evolving under positive selection using a Bayes empirical model.These scenarios were tested in pamlX, a package housing CodeML (Xu and Yang, 2013).

Source-to-sink expression data
Publicly available transcript expression data from the leaf and seed were downloaded for two C 4 grasses and two C 3 grasses (Supplementary Data Table S1).Except for data associated with P. miliaceum, all data were microarray data, where expression was represented as robust multichip average (RMA)normalized expression values.For P. miliaceum, the dataset was RNA-Seq, and therefore TPM values were extracted as described above.Using these data, the leaf expression values were divided by the seed expression values to obtain a sourceto-sink ratio.There were three biological replicates used for each study.Since there were only two species surveyed for each photosynthetic type, the biological replicates were used as the n value when representing the source-to-sink ratio for C 4 and C 3 grasses.Therefore n = 6 for the ratios in this experimental study.These values can be found in Supplementary Data File S2.

Leaf gradient expression data
Publicly available RNA-Seq data of leaf gradients were mined from four C 4 grasses and two C 3 grasses (Supplementary Data Table S1).The TPM values were extracted as described above.The expression profiles were represented as mean log 2 TPM values for at least three biological replicates (except for S. viridis, which had only one biological replicate).For each species, between 5 and 15 leaf sections were segmented and sampled for RNA-Seq analysis.These values can be found in Supplementary Data File S3.

Bundle sheath and mesophyll cell expression data
RNA-Seq data were obtained from the bundle sheath and mesophyll cells of several C 4 grasses (Supplementary Data Table S1).We extracted TPM values as described above.Each study had at least three biological replicates for both cell types.The expression of genes is represented as TPM compared between the bundle sheath and mesophyll cells for each species.The TPM values were also averaged across all the C 4 grasses that were surveyed for each cell type.These values can be found in Supplementary Data File S4.

Data analysis
Figures and statistical analyses were performed using GraphPad Prism v9.4.1.A paired Student's t-test was used to compare the leaf-to-seed expression ratios of each sugar sensor gene to determine if it there was preferential expression to the source or sink tissue (Supplementary Data Table S2).A paired Student's t-test was also carried out to compare the expression between bundle sheath and mesophyll cells of sugar sensor genes in C 4 grasses (Supplementary Data Table S3).

RESULTS
Identification of genes encoding sugar sensor components in C 4 and C 3 grasses Identification of putative sugar sensors across selected C 4 and C 3 grasses was first carried out using BLAST searches of the respective genomes using known sequences from either Z. mays or Arabidopsis.Extracted sequences were translated and phylogenetic trees built to visualize homology to each other (Supplementary Data Figs S1-S6).Sequences from Z. mays, S. spontaneum, S. bicolor, S. viridis, S. italica, P. miliaceum, P. hallii, O. sativa and B. distachyon were extracted from their respective genomes where present.
Table 1 summarizes the sugar sensor genes that were present or absent in each grass species.The B. distachyon genome did not contain a copy of RAPTOR2 and some grasses did not have a copy of RAPTOR3.Several grasses, such as Z. mays, P. miliaceum and S. spontaneum, also contained a second hybrid subunit within their genomes (Supplementary Data Fig.S4).The number of hexokinase homologues varied between six and nine across the grasses analysed (Supplementary Data Fig.S5).Each genome contained a copy of HXK5 and a HXK6, the putative sugar sensors.Each species also contained a copy of TPS1, TPP and TRE (Supplementary Data Fig.S6).Analysis of C 4 and C 3 grasses did not appear to show gene duplication during evolution.
Positive selection analyses were carried out for each putative sugar sensor using the extracted protein sequences to determine if there was any detectable C 4 -dependent evolution within the set of genes.Sequences used were from four C 4 grasses, P. antidotale, S. bicolor, S. viridis and Z. mays, and five C 3 grasses, H. amplexicaulis, P. bisulcatum, S. laxum, O. sativa and C. patens.Nicotinamide-adenine dinucleotide phosphatemalic enzyme (NADP-ME) was used as a control for these analyses as it is known that several residues of this gene are under C 4 -specific selection.As expected, several residues were identified as being under C 4 -specific selection in NADP-ME (Supplementary Data Fig.S7).However, no residues were identified as being under C 4 -specific selection within any of the sugar sensor proteins tested here (TOR, LST8-1, RAPTOR1, RAPTOR2, SnRK1α, β, γ and βγ subunits, HXK5, HXK6, TPS1, TPP1 and TRE).
Sugar sensor genes are expressed in both the source and sink tissues of C 4 and C 3 grasses Source (leaf) and sink (seed) transcriptomic data were scrutinized to determine whether differences at the gene expression level are associated with the evolution of C 4 photosynthesis (Fig. 1, Supplementary Data File S1).Using publicly available leaf and seed transcriptomic data, gene expressions of sugar sensor genes were extracted for two C 4 grasses (Z.mays and P. miliaceum) and two C 3 grasses (O.sativa and B. distachyon) (Jain et al., 2007;Sekhon et al., 2011;Yue et al., 2016;Sibout et al., 2017).Data were combined for each photosynthetic type for analysis.Within these datasets, TOR was not in either of the datasets for the C 4 grasses, and RAPTOR3 expression was not detected in the C 3 grass B. distachyon.Genes encoding the TORC subunits exhibited heightened transcript expression in the leaves (source tissue) of C 3 grasses, when compared with the ratios exhibited by C 4 grasses.The source-to-sink ratios for the C 4 grasses were close to 1 for many of the sugar-sensing genes, suggesting they are expressed equally between leaf and seed tissues, at least for these grasses.Only the gene encoding the regulatory subunit of SnRK1C (SnRK1βγ1) and the TPP1 gene (which encodes the trehalose phosphate phosphatase enzyme) had significantly higher (P ≤ 0.05) expression in the leaf compared with the seed for the C 3 grasses.There was no significant expression of any gene for the C 4 species or towards the seed (i.e.<1 and significantly different from 1).It can also be noted that HXK5 source-to-sink ratio was ~5.6-fold higher for the C 3 grasses than the C 4 grasses, and this change was even more apparent for HXK6 gene expression, at an ~58.3-fold difference.However, these differences were not identified as significant.
There were more significant changes (P ≤ 0.05) within individual species that indicate some sugar sensor genes may be preferentially expressed in the source or sink tissues (Supplementary Data Fig.S8).Notably, many genes identified as significantly different in Z. mays had a ratio <1, indicating higher amounts of transcripts were identified in the sink tissues.Preferential expression to either the leaf or seed is more prominent within a species rather than collectively as a photosynthesis type (Supplementary Data Fig.S1).
Sugar sensor genes are largely stably expressed along the leaf gradient of C 3 and C 4 grasses Publicly available RNA-Seq data were mined for their expression along the leaf gradient of various C 3 and C 4 grasses and represented as heat maps (Li et al., 2010;Wang et al., 2014;Ding et al., 2015;Hu et al., 2018) (Figs 2-5).This analysis was carried out to determine whether there were possible changes in sugar sensing along the leaf and/or between C 4 and C 3 grasses during the sink-to-source transition from the base to the tip.The expression profiles for the genes encoding TORC subunits are displayed for TOR, LST8-1, RAPTOR1 and RAPTOR2 (Fig. 2A-D).RAPTOR3 was excluded since this subunit was only found in some species.Notably, B. distachyon orthologues were expressed at high levels compared with the other grasses.Further, LST8-1 and RAPTOR2 transcripts were also highly abundant along the leaf gradient in S. spontaneum and S. bicolor.
Similarly to genes that encode TORC subunits, the expression profiles of genes that encode the SnRK1C subunits were also examined along the leaf gradient of these grasses (Fig. 3).SnRK1γ2 and SnRK1βγ2 were excluded from the heat maps because they were absent in multiple genomes.The α subunits were found to be expressed in all species examined to varying  Asterisks represent a significant difference from 1 and predominating in the leaf.There were no genes expressed with a ratio <1 that were significantly different.levels (Fig. 3A-C).Expression was largely stable across the leaf in each species; however, some patterns were apparent.
The genes encoding the α subunits of S. bicolor generally had higher expression towards the tip of the leaf.A similar pattern was observed for SvSnRK1α1.Conversely, SsSnRK1β2 and ZmSnRK1β3 had higher expression at the base of the leaf.In the middle sections of the leaf, SvSnRK1β2 and SsSnRK1γ1 were expressed at higher levels.
The expressions of the genes encoding putative sugar sensors, HXK5 and HXK6, were examined to investigate whether glucose sensing may differ across the leaf gradient in C 4 and C 3 grasses (Fig. 4).HXK5 was expressed at high levels for all species except for Z. mays (Fig. 4A).Notably, for several C 4 grasses (S. spontaneum, S. bicolor and S. viridis) expression tended to be higher towards the base of the leaf for HXK5.Despite high homology between HXK5 and HXK6, HXK6 was largely expressed at low levels (Fig. 4B).Only S. viridis and B. distachyon exhibited high abundance of HXK6 transcripts (Fig. 4B).
Finally, genes encoding enzymes associated with trehalose metabolism were interrogated along the leaf gradient.Similarly to many other genes encoding sugar sensing components, TPS1, TPP and TRE transcripts were expressed at varying degrees across the leaf and between species (Fig. 5).For SbTPS1, leaf sections eight to ten (around the middle) exhibited the highest transcript expression compared with the rest of the leaf (Fig. 5A).SsTPP was expressed relatively ubiquitously along the leaf at high levels, while SvTPP and ZmTPP were preferentially expressed at the base of the leaf (Fig. 5B).SbTRE was highly expressed when compared with the other grasses, especially towards the tip of the leaf (Fig. 5C).

Sugar sensor genes exhibit species-specific preferential expression in either bundle sheath or mesophyll cells
Transcript expression associated with sugar-sensing genes was analysed within bundle sheath and mesophyll cells of several C 4 grasses from publicly available RNA-Seq datasets (John et al., 2014;Döring et al., 2016;Denton et al., 2017;Washburn et al., 2021).These results are presented for Z. mays, S. bicolor, S. viridis, S. italica and P. hallii (Fig. 6, Supplementary Data File S4, Supplementary Data Table S3).Datasets from C 3 grasses were not examined in this study since they were sparse or had poor mapping of reads to their respective genomes, likely due to the difficulty of isolating and separating these cells in C 3 species.In Z. mays, sugar sensor genes were generally not preferentially expressed in one photosynthetic cell type (Fig. 6A).ZmSnRK1β3 showed the highest overall expression out of all the genes surveyed, while ZmTPP expression was significantly higher (P ≤ 0.05) in the mesophyll cells.Unlike Z. mays, in S. bicolor and P. hallii there were numerous significant differences (P ≤ 0.05) in transcript expression of sugar sensor genes, with almost all those identified as significant being elevated in bundle sheath cells (Fig. 6B, C).All genes encoding the TORC subunits (except PhLST8-1) and SnRK1α were shown to be significantly preferentially expressed in the bundle sheath cells of S. bicolor and P. hallii.Where the two species differed was the transcript expression of SnRK1β3, which was expressed preferentially in mesophyll cells of P. hallii, but not S. bicolor.The largest fold changes were for TPS1, for which there was a 27-and 15-fold increase in bundle sheath cells in S. bicolor and P. hallii, respectively.Setaria viridis and S. italica are two close relatives within the millets.Both species exhibited less significant differences in sugar sensor gene expression between the two photosynthetic cell types when compared with S. bicolor (Fig. 6D, E).In addition, unlike S. bicolor, numerous genes had significantly higher (P ≤ 0.05) expression in mesophyll cells of both or one of the species.Transcript expression of LST8-1 was significantly elevated in mesophyll cells of both S. viridis and S. italica.Several genes encoding the SnRK1C subunits showed significant differences in transcript expression between the two cells.SvSnRK1α3 and SvSnRK1β2 showed preference for the bundle sheath and mesophyll, respectively.SiSnRK1α1 exhibited significantly elevated expressed in the mesophyll cells, although this was reversed for the other two SiSnRK1α subunit genes.Similar to S. bicolor and P. halli, there were also significant ~29.7-and ~47.3-fold increases in expression of TPS1 in bundle sheath cells of S. viridis and S. italica, respectively.
The values for all sugar sensor genes examined were averaged and the resulting log 2 TPM values were visualized for each cell type (Fig. 6F).Within this analysis, only SnRK1β1 and TPS1 transcript expressions were significantly different between the bundle sheath and mesophyll cells (Supplementary Data Table S3).The average SnRK1β1 expression was higher in mesophyll cells for the C 4 grasses examined; however, the fold differences were small.Although not significant, generally there was higher transcript abundance within the bundle sheath cell for SnRK1α2, SnRK1α3, SnRK1γ1, HXK5 and HXK6.Like with the source-to-sink expression comparison, significant changes within a species were more common than significant changes associated with photosynthetic type.

DISCUSSION
No co-optional evolution for C 4 sugar sensors but some speciesspecific preferential expression in bundle sheath or mesophyll cells In this study it was hypothesized that sugar sensors have evolved to accommodate the two-celled compartmentation  (Li et al., 2010;Wang et al., 2014;Ding et al., 2015;Hu et al., 2018).Scale bar to the right of each heat map represent log 2 TPM.
of C 4 photosynthesis.To determine whether sugar sensors diverged from their C 3 counterparts during the evolutionary transition from C 3 to C 4 photosynthesis, transcript sequences from C 4 species (P.antidotale, S. bicolor, S. viridis and Z. mays) and C 3 species (H.amplexicaulis, P. bisulcatum, S. laxum, O. sativa and C. patens) were utilized.No evidence for the positive selection of C 4 sugar sensors during C 4 evolution was identified within this study.This result was unexpected, as the evolution of C 4 photosynthesis has resulted in major changes, involving C 4 -specific residue changes in numerous key genes (Christin et al., 2009;Watson-Lazowski et al., 2018).However, it is plausible that selection pressures associated with C 4 photosynthesis have not influenced the sugar sensors in this specific way.For example, it has been well established that gene duplications have occurred during the evolution of C 4 photosynthesis but it was not observed for genes in this study (Marshall et al., 1996;Monson, 1999Monson, , 2003)).Moreover, changes to cis-regulatory elements in single-copy genes have contributed to the altered expression patterns that facilitate C 4 photosynthesis (Rosche and Westhoff, 1995).
Aspects such as these may still facilitate C 4 -specific expression patterns of sugar-sensing genes.
When examining the general expression of the genes that encode the proteins that make up TORC there was little change between the bundle sheath and mesophyll cells of the C 4 grasses (Fig. 6).The expression of TOR varied between species and was not significantly expressed in one cell type over another when collectively examining the C 4 grasses, which suggests that the TOR protein has a signalling role in both cells.Studies in the alga Chlamydomonas reinhardtii have shown that CO 2 fixation promotes TOR activity but has no effect on TOR or LST8 protein abundance (Mallen-Ponce et al., 2022).Furthermore, it was observed that photosynthesis inhibition decreases TOR activity.The variation in gene expression for the subunits that encode TORC could be also related to the role it has in the circadian rhythm (Xiong and Sheen, 2014;Dong et al., 2017).Therefore, tissue harvest time across studies could influence transcript abundance of the genes encoding TORC subunits.Moreover, its role as a master regulator across different tissues and processes could also account for the lack of  (Li et al., 2010;Wang et al., 2014;Ding et al., 2015;Hu et al., 2018).Where expression within a leaf section is represented as white there were no detectable reads.The scale bar to the right of each heat map represent log 2 TPM.(John et al., 2014;Döring et al., 2016;Denton et al., 2017;Washburn et al., 2021).(F) Heat map comparison of log 2 TPM means of bundle sheath and mesophyll expression in C 4 grasses.*P < 0.05, significantly different expression between bundle sheath and mesophyll cells (Student's t-test).

Zea mays
differences in gene expression between the two photosynthetic cell types when collectively analysing the C 4 grasses in this study (Pacheco et al., 2021).Sucrose and starch synthesis occurs in the mesophyll and bundle sheath cells of C 4 species, respectively.Although TORC is regulated by sugars, the complex also regulates starch accumulation.These observations could account for the presence of genes relating to this complex in both cell types.Like TORC, SnRK1C is thought to regulate many processes, and is usually upregulated under stress conditions when sucrose availability is low (Baena-González et al., 2007).Although the data were generated from plants grown in normal conditions, the overall transcript abundance of genes encoding SnRK1C subunits was high in comparison with the other sugar sensors (Fig. 6).On average, within the C 4 grasses examined, SnRK1β1 was preferentially expressed in mesophyll cells (Fig. 6F, Supplementary Data Table S3).SnRK1β1 encodes a regulatory component of the complex and β subunits can be expressed at varying levels depending on the tissue, developmental stage and environmental cues (Polge et al., 2008).Therefore, it is a possibility that SnRK1β1 regulates the interaction of the kinase with its targets within the mesophyll cells of C 4 grasses.When averaged across the C 4 species surveyed, SnRK1α genes (which encode the catalytic subunit of the complex) were not significantly different between cells.Nevertheless, it must be noted that many of the grasses had higher expression within the bundle sheath cells of the SnRK1α catalytic subunit genes (Fig. 6B-E).There is a possibility that SnRK1α subunits are important for sensing and/or signalling during sucrose translocation, in which photoassimilates pass through the bundle sheath cells for phloem loading to occur (Bezrutczyk et al., 2018;Chen et al., 2022).During this process, genes encoding regulatory subunits may be expressed according to translocation needs and photosynthetic activity.Alternatively, expression in the bundle sheath cells could be linked to a role in regulating genes associated with starch synthesis, which has been evidenced in the seed (Zhang et al., 2001;Tiessen et al., 2003).
As mentioned previously, T6P signalling has been closely linked with SnRK1C activity (Baena-González and Lunn, 2020).TPS1 expression was significantly higher in the bundle sheath cells when averaged across the C 4 grasses (Fig. 6F).TPS is involved in the synthesis of T6P, and its presence indicates elevated sucrose levels (Grennan, 2007).Therefore, it was surprising that TPS1 was higher in the bundle sheath cells, since sucrose biosynthesis occurs predominantly in the mesophyll cells of C 4 grasses (Lunn and Furbank, 1999;Furbank and Kelly, 2021).As suggested with SnRK1C, T6P signalling may play an important role in the phloem loading process (Emms et al., 2016;Bezrutczyk et al., 2018Bezrutczyk et al., , 2021;;Chen et al., 2022).Interestingly, trehalose increases the expression of ApL3, which encodes an ADP-glucose pyrophosphorylase that subsequently increases starch synthesis (Wingler et al., 2000).Therefore, the trehalose biosynthesis pathway may be important for starch production within the bundle sheath cells of C 4 grasses.The additional genes encoding enzymes associated with T6P metabolism were expressed at similar levels between the bundle sheath and mesophyll cells for all species analysed.This could suggest that the synthesis, breakdown and signalling of trehalose is important in both cells, or, since trehalose is a non-reducing disaccharide, it could also play a role in buffering sucrose loading into the phloem.
There were also differences in expression of the putative HXK sugar sensors, HXK5 and HXK6, between the two photosynthetic cells for several C 4 species.For example, SbHXK5, SbHXK6, PhHXK5 and PhHXK6 were all expressed at higher levels in the bundle sheath compared with mesophyll cells (Fig. 6B, C).This could indicate that the phosphorylation of glucose is more prevalent in the bundle sheath cells of C 4 species, or that glucose sensing predominates there.Research on the effect of HXK sugar sensing in C 4 species has been sparse, but seminal studies using Z. mays protoplasts have shown that glucose, the substrate for HXK, can repress photosynthesis genes (Sheen, 1990;Jang and Sheen, 1994).It must be noted that this was only examined in a single-cell system and did not examine the whole leaf or how the plant that might affect how sugar sensing occurs, given photosynthesis takes place in a two-cell system in Z. mays.

Expression of sugar sensor genes changes in source-sink developmental models
To investigate the expression patterns of genes encoding sugar sensor components in source and sink tissues, we interrogated leaf and seed tissues as well as developmental leaf gradients to determine if there was preferential expression in the source or sink tissue.Given that source-to-sink expression gradients have been observed with sugar transporters and other genes associated with sugar metabolism in C 4 grasses, it may be expected that similar expression patterns will be found for genes encoding the sugar sensor proteins (Bezrutczyk et al., 2018;Hu et al., 2018;Chen et al., 2022).
As previously established, TORC is a master regulator of many different processes in the plant.Therefore, genes encoding this complex would more likely be found in all tissue types.For the C 4 species examined, gene expression of the regulatory subunits of TORC were close to 1, whereas for the C 3 species the expression of numerous TORC genes trended towards source tissue (ratio of <1) (Fig. 1).The expression of these genes was also ubiquitous along the leaf gradient of the four C 4 grasses and the two C 3 grasses (Fig. 2).Interestingly, it has been shown that when TORC repression is initiated, S. viridis showed a milder phenotype and a smaller magnitude of changes relating to primary metabolites and global gene expression when compared with the C 3 Arabidopsis (da Silva et al., 2021).This might suggest that plant growth in C 4 species is less rigorously controlled by TORC, or is less sensitive to changes in carbon status.Previous work on the relationship between CO 2 fixation and TOR activity has suggested that CO 2 fixation status can influence TOR activity, but not necessarily change the protein abundance (Mallen-Ponce et al., 2022).Thus, this might also mean that the transcript abundance of the subunits of TORC may not change between source and sink tissues, but rather activity is modulated via other factors.
Like TORC, SnRK1C is thought to regulate numerous processes throughout the plant.The seed-to-leaf expression ratios of genes encoding the catalytic subunits of SnRK1C for both C 4 and C 3 species were close to 1, demonstrating that they are found in both the seed and leaf tissues of the analysed grasses (Fig. 1).This suggests that these genes play a similar role within the plant regardless of whether they are C 4 or C 3 species.When examining the expression of genes that encode subunits of SnRK1C over a leaf gradient, again there were no immediate trends that differentiate source to sink comparisons between the leaf and seed or within C 4 and C 3 leaves along the leaf gradient (Fig. 3).These data would suggests that SnRK1 is largely equally distributed across source and sink tissue of C 3 and C 4 grasses.However, SnRK1C is known to be activated in response to unfavourable conditions or during a starvation response (Baena-González et al., 2007).Therefore, the lack of differences between source and sink tissues might not be uncommon since these plants were grown in normal conditions.In addition, there is evidence that SnRK1α is regulated on the post-transcriptional level (Lu et al., 2007), which may also explain the limited differences identified.
While the link between photosynthesis and SnRK1C is not well documented, a direct link between HXK sugar sensing and modulating photosynthesis gene expression has been identified.This was first established using Z. mays protoplasts, as mentioned previously (Sheen, 1990;Jang and Sheen, 1994).This was later confirmed using gin2 mutants of Arabidopsis, showing AtHXK1 could sense glucose and in turn influence photosynthesis gene expression (Moore et al., 2003).Sugar sensors have also been established in rice (C 3 grass) via overexpression lines of OsHXK5 and OsHXK6, which exhibited heightened sensitivity to glucose (Cho et al., 2006;Cho et al., 2009b).These rice lines were generally smaller than wild-type and showed decreased expression of key photosynthesis genes, such as the Rubisco small subunit gene (rbcS).The homologues of HXK5 and HXK6 were expressed in both the leaf and seed tissues of the C 4 and C 3 species examined (Fig. 1).The source-to-sink ratio for C 4 grasses was close to 1, whereas for C 3 grasses it was well above 1.Although these differences were not significant within our dataset, the extent of the differences would suggest that HXKs predominate in the leaves of C 3 grasses.Expression of HXK5 and HXK6 homologues was apparent all along the leaves of both C 4 and C 3 grasses, and changes along the leaf were subtle (Fig. 4).This is unlike sugar transporters and starch and sugar metabolism genes, which exhibited a more prominent gradient as the tissue changes from sink to source from the base to the tip of the leaf (Chen et al., 2022).For HXK5, there was higher abundance at the base of the leaf, where it is more sinklike tissue, for the C 4 grasses S. spontaneum, S. bicolor and S. viridis (Fig. 4A).This may suggest a role in sensing incoming photoassimilates that break down to glucose for utilization as the tissue matures.However, like TOR and SnRK1α, HXK sugar sensors could also be post-transcriptionally regulated, and so changes in expression may be minor and not correlate with activity.
T6P abundance is thought to modulate SnRK1C activity, subsequently de-repressing anabolic processes (Baena-González et al., 2007;Lawlor and Paul, 2014).SnRK1C is known to be involved in starch synthesis during grain filling of grasses by regulating the expression of genes encoding proteins involved in this process, and there have also been suggestions that its activity is controlled by T6P levels (Laurie et al., 2003;Lu et al., 2007;Gazzarrini and Tsai, 2014).In this study, the TPP1 source-to-sink expression ratio was significantly above 1 for C 3 grasses, which may suggest that T6P (or trehalose itself) have a larger role in sugar sensing and signalling within the leaves (Fig. 1).The transgenic manipulation of TPP1 in Z. mays showed that T6P plays a large role in coordinating photoassimilate partitioning to the reproductive tissues by regulating photosynthesis (Oszvald et al., 2018).The authors showed that Sugars Will Eventually be Exported Transporters (SWEET) genes were upregulated in the transgenic lines, increasing the movement of photoassimilates to sink tissue, particularly under drought conditions.Other genes associated with T6P metabolism also had source-to-sink ratios above 1 in this study, although these differences were not significant.Further analysis on the expression of these genes along the leaf gradient of C 4 and C 3 grasses showed that they were expressed throughout, and many only showed small changes from the base to the tip (Fig. 5).

Conclusions
In this study, transcriptomic data across various C 4 and C 3 grasses were analysed to determine gene expression patterns of the components of TORC, SnRK1C and HXK sugar sensors, as well as T6P metabolism.These analyses focused on the role of these sugar sensors in relation to photosynthesis, where sugars are produced, and whether sugars may be perceived differently between C 4 and C 3 grasses.Even though C 3 grasses perform photosynthetic and carbohydrate production reactions in one cell type, unlike C 4 grasses, which is compartmentalized, not many changes in sugar sensor gene expression were observed between the two types of plants.
There were few distinct gradient transitions of expression for sugar sensor genes, suggesting sugar sensing is important along the whole young leaf.Moreover, when expression was examined in the two photosynthetic cell types in C 4 grass leaves only, SnRK1β1 and TPS1 were preferentially expressed in the mesophyll and bundle sheath cells, respectively.However, it must be noted that within species there were more distinct changes in expression of each sugar sensor gene.
Future studies could be incorporated to analyse sugar sensors between C 4 and C 3 grasses by examining protein abundance and activity to determine if sugars are perceived differently.These studies can also be expanded into a larger variety of C 4 and C 3 species that include dicots and monocots and different subtypes of C 4 photosynthesis.Nevertheless, this study provides a foundation for which the role of sugar sensors can be scrutinized, especially in terms of how it may relate to C 4 and C 3 photosynthesis.SUPPLEMENTARY DATA Supplementary data are available online at https://academic.oup.com/aob and consist of the following.Figure S1: phylogenetic tree of monocot TOR complex subunits.Figure S2: phylogenetic tree of monocot SnRK1α subunits.Figure S3: phylogenetic tree of monocot SnRK1β subunits.Figure S4: phylogenetic tree of monocot SnRK1βγ and SnRK1γ subunits.
Figure S5: phylogenetic tree of monocot hexokinases.Figure S6: phylogenetic tree of monocot proteins related to T6P metabolism.Figure S7: C 4 -dependent evolution of NADP-ME. Figure S8: leaf-to-seed expression ratio of sugar sensor genes in C 4 and C 3 grasses.Table S1: summary of species used in this study, data accession numbers and references.Table S2: leaf-to-seed expression ratio of sugar sensor genes in C 4 and C 3 grasses.Table S3: bundle sheath and mesophyll cell sugar sensor gene expression in C 4 grasses.File S1: gene IDs.

Fig. 1 .
Fig.1.Leaf-to-seed expression ratio of sugar sensor gene comparisons between C 4 and C 3 grasses.RMA-normalized or TPM values from either microarray or RNA-Seq data of sugar sensor genes were used to calculate leaf-to-seed ratios from the C 4 grasses Z. mays and P. miliaceum and the C 3 grasses O. sativa and B. distachyon(Jain et al., 2007;Sekhon et al., 2011;Yue et al., 2016;Sibout et al., 2017).Each species consisted of three biological replicates for each tissue sampled.Data represent the mean leaf-to-seed ratios of genes from C 4 and C 3 grasses (n = 6).Error bars represent the standard error of the mean.Ratios <1 indicate expression of the gene predominating in the seed whereas ratios >1 indicate expression predominating in the leaf.The broken line indicates 1. Asterisks represent a significant difference from 1 and predominating in the leaf.There were no genes expressed with a ratio <1 that were significantly different.

Fig. 2 .Fig. 3 .
Fig.2.Expression of genes encoding TORC subunits along the leaf gradient of C 4 and C 3 grasses.Heat maps displaying log 2 TPM values of the genes TOR (A), LST8-1 (B), RAPTOR1 (C) and RAPTOR2 (D), encoding subunits that make up TORC.RAPTOR3 is omitted due to its absence in multiple genomes.RAPTOR2 was absent in the Brachypodium genome.The C 4 species examined were Z. mays (15 sections), Saccharum spontaneum (15 sections), Sorghum bicolor (13 sections) and Setaria viridis (10 sections), the C 3 species being O. sativa (11 sections) and B. distachyon (5 sections)(Li et al., 2010;Wang et al., 2014;Ding et al., 2015;Hu et al., 2018).Leaf sectioning is indicated to the left (A).Where expression within a leaf section is represented as white, there were no detectable reads.The scale bar to the right of each heat map represents log 2 TPM.

Fig. 5 .
Fig. 5. Expression of genes encoding proteins involved in T6P signalling along the leaf gradient of C 4 and C 3 grasses.Heat maps displaying log 2 TPM values of the genes TPS1 (A), TPP (B) and TRE (C).The C 4 species examined were Z. mays (15 sections), Saccharum spontaneum (15 sections), Sorghum bicolor (13 sections) and Setaria viridis (10 sections), the C 3 species being O. sativa (11 sections) and B. distachyon (5 sections)(Li et al., 2010;Wang et al., 2014;Ding et al., 2015;Hu et al., 2018).Where expression within a leaf section is represented as white there were no detectable reads.The scale bar to the right of each heat map represent log 2 TPM.
File S2: leaf-to-seed expression ratios of C 4 and C 3 grasses.File S3: leaf gradient expression of sugar sensors from C 4 and C 3 grasses.File S4: bundle sheath and mesophyll expression of sugar sensors from C 4 and C 3 grasses.FUNDING This work was funded by the Australian Research Council Centre of Excellence for Translational Photosynthesis (Grant Number CE140100015) and Australian Research Council Discovery Project (DP210102730) awarded to O.G. and R.T.F.