Multiple mechanisms for enhanced plasmodesmata density in disparate subtypes of C4 grasses

High plasmodesmata density in C4 grasses is a result of larger pit fields and/or more abundant plasmodesmata per pit field area.


Introduction
Most plants obtain sugars by fixing atmospheric CO 2 using the enzyme Rubisco (ribulose bis-phosphate carboxylase oxygenase). This process is inherently inefficient as O 2 competes with CO 2 at the enzyme's active site, resulting in formation of compounds that cost energy to recycle in a process known as photorespiration. The first product of photosynthetic CO 2 fixation by Rubisco is a three-carbon sugar, hence this process is known as C 3 photosynthesis. Many tropical and sub-tropical plant lineages have independently evolved a more efficient photosynthetic biochemistry, termed C 4 photosynthesis (Hatch, 1987;Sage et al., 2011). Here, CO 2 is first captured in mesophyll (M) cells as C 4 acids, which then diffuse into bundle sheath (BS) cells, where Rubisco is located, and decarboxylated, resulting in greatly elevated local CO 2 concentrations (Furbank and Hatch, 1987). This CO 2 -concentrating mechanism reduces photorespiration and enables Rubisco to operate close to its catalytic maximum (von Caemmerer and Furbank, 2003;Sage et al., 2012). It is thought that a reduction in atmospheric CO 2 concentration ~35 million years ago may have driven the evolution of this CO 2 -concentrating mechanism (Sage, 2004).
Although many plants conduct C 4 photosynthesis, a range of anatomical and biochemical specialisations distinguish different C 4 lineages. For example, subcategories of C 4 plants are defined by the enzymes that catalyse the decarboxylation of the C 4 acid: NADP malic enzyme (NADP-ME) type, NAD malic enzyme (NAD-ME) type, and phosphoenolpyruvate carboxykinase (PCK) type (Hatch, 1987;Furbank, 2011). Particularly in grasses, these biochemical differences are further elaborated by anatomical specialisations that include the presence of the mestome sheath between the BS and the vasculature in NAD-ME and PCK types but not in the NADP-ME type (Hattersley and Watson, 1976); suberisation of the BS cells in NADP-ME and PCK types but not in the NAD-ME type (Hattersley and Browning, 1981); and oval chloroplasts positioned centrifugally with mitochondria in BS cells of NADP-ME and most PCK types but elongated chloroplasts positioned centripetally with mitochondria in BS cells of the NAD-ME type (Hattersley and Watson, 1976;Hattersley and Browning, 1981;Dengler et al., 1994;McKown and Dengler, 2007).
Over the last 35 million years, these evolutionary changes in anatomy and biochemistry arose independently at least 66 times (Sage et al., 2012). In grasses, 22-24 distinct C 4 lineages are found within the PACMAD (Panicoideae, Aristidoideae, Chloridoideae, Micrairoideae, Arundinoideae, and Danthonioideae) clade, specifically in the subfamilies Panicoideae, Aristidoideae, Chloridoideae, and Micrairoideae (GPWGII, 2012). These subfamilies comprise many highly productive crop species such as sugar cane, millets, and maize. On the other hand, the subfamilies Bambusoideae, Ehrhartoideae, and Pooideae, known as the BEP clade (GPWGII, 2012), contain no C 4 species. These subfamilies include staple food grains such as rice, wheat, and barley. Demand for food crops is predicted to increase by at least 50% in the next 35 years (Hibberd et al., 2008), and yield increases through traditional breeding of these C 3 species will not meet this requirement. Recent breakthroughs in biotechnology may provide the opportunity to engineer the C 4 photosynthetic pathway into C 3 crops, which could potentially meet required improvements to feed the growing human population (Hibberd et al., 2008).
The biochemistry of C 3 and C 4 photosynthesis has been well studied, with a strong focus on either down-regulation/ knockout (von Caemmerer and Furbank, 2016) or overexpression (Kajala et al., 2011) of one or more of the known key C 4 photosynthetic enzymes in various plant systems to understand their function. Previous work has shown how leaf anatomical traits can be used to gain insight into C 4 evolution (Hattersley and Watson, 1976;Dengler et al., 1994;McKown and Dengler, 2007). These studies mainly investigated traits related to the specialised vascular anatomy of C 4 plants known as Kranz anatomy: a wreath-like arrangement of BS and M cell layers enclosing the vascular bundles. This anatomical arrangement separates the biochemical CO 2 'pump' in the M from Rubisco in the BS, and provides a barrier to CO 2 diffusion out of the BS compartment (von Caemmerer and Furbank, 2003;Sage et al., 2012).
The requirement for metabolites to move at high rates between specialised C 4 cell types has long been recognised to be important for C 4 photosynthetic function (Osmond, 1971;Hatch and Osmond, 1976). Estimates of flux of photosynthetic metabolites between cell types in C 4 leaves assume that the M-BS cell wall is impermeable due to the secondary thickening of cell walls between these cells (Danila et al., 2016). Hence, these metabolites must move between cell types via plasmodesmata (PD), the symplastic nanochannels that span cell walls and provide both a cytoplasmic and an endoplasmic continuum for metabolite transport (Osmond and Smith, 1976;Robards, 1976;Overall and Blackman, 1996). In leaves, PD are distributed in groups called pit fields. The available data on PD distribution between leaf cells in C 3 and C 4 species show that C 4 plants have a greater density of PD than C 3 plants (Botha, 1992;Danila et al., 2016). A major barrier to quantitatively examining these symplastic connections across diverse species has been the difficulty of the microscopy required to acquire statistically robust data (e.g. Botha, 1992). Recently, a high-throughput technique has been developed to assess PD density (Danila et al., 2016). This technique combines high-resolution scanning electron microscopy (SEM), which allows analysis of individual PD within pit fields, and three-dimensional (3-D) immunolocalisation confocal microscopy for relatively rapid quantification of pit field distribution in a larger surface area across many cells within a leaf. Thus, it is now possible to quantify the PD connecting leaf cells of different C 4 species and determine whether increased PD density at the M-BS interface is a conserved trait of C 4 species and whether this density varies between different C 4 subtypes. In this study, we quantify PD density between leaf cells in a selection of C 3 and C 4 grass species. These species encompass both BEP and PACMAD clades and include four origins of C 4 photosynthesis comprising all C 4 subtypes.

Phylogenetic tree construction
To construct a phylogenetic tree for this analysis the predicted protein sequences for each of the 18 species were subject to orthogroup inference using OrthoFinder (Emms and Kelly, 2015) and a set of 60 single-copy orthogroups containing sequences from at least 16 of the 18 grass species were identified (Supplementary Dataset S1 at JXB online). These protein sequences were aligned using MergeAlign (Collingridge and Kelly, 2012), edited to remove all gap-containing columns, concatenated, and subjected to 1000 replicates of a non-parametric bootstrapped maximum-likelihood phylogenetic inference using FastTree (Price et al., 2010). The full-length concatenated alignment was also used for Bayesian phylogenetic tree inference using mrBayes v 3.2.6 (Huelsenbeck and Ronquist, 2001). The amino acid model was set to JTT and the covarion was turned on. Two runs, each of four chains, were initiated and allowed to run for 100 000 generations sampling every 100 generations. Convergence was assessed through visual inspection of log-likelihood traces and through analysis of the standard deviation of split frequencies (σ 2 <0.00001).

Leaf anatomical sample preparation
All leaf tissue preparations for light microscopy, SEM, and 3-D immunolocalisation confocal microscopy were as described by Danila et al. (2016). The middle portion of the youngest fully expanded leaf from three individual 9-d-old seedlings per species were collected and pooled. From this sample pool, leaf tissues were fixed and processed accordingly. For 3-D immunolocalisation confocal microscopy, leaf tissue was cleared using PEA-CLARITY (Palmer et al., 2015), hybridised with β-1,3-glucan (callose) antibody, followed by Alexa488-tagged secondary antibody, and poststained with calcofluor white to visualise cell walls (Danila et al., 2016).

Microscopy
Transverse sections of all grass leaves were imaged for light microscopy under 10× and 40× objectives using a Nikon Eclipse 50i upright microscope (Nikon Instruments). SEM was performed using a Zeiss Ultra Plus field emission scanning electron microscope at 3 kV. To quantify pit field distribution, two z-stacks from two leaf tissues per species were obtained using a Leica SP8 multiphoton confocal microscope (Leica Microsystems). Details can be found in Danila et al. (2016).

Quantitative leaf anatomical measurement
Different from the conventional use of resin-embedded leaf tissue, BS cell area was measured from 25 to 50 individual cells of minor veins using virtual z-sections through entire confocal z-stacks for each species. BS cell volume was calculated by multiplying the BS cell area by BS cell length, which was measured using cell images (n=30 to 160) obtained from the paradermally orientated confocal micrographs of the same leaf z-stacks (Turrell, 1936). Vein diameter and interveinal distance (IVD) were measured using 10 to 25 individual minor veins from light micrographs of transverse leaf sections (see Supplementary Fig. S1). The bundle sheath surface area per unit leaf area, S b , was calculated using the equation described in Pengelly et al. (2010). To determine cell-to-cell PD connectivity among different subtypes of C 4 photosynthesis, the frequency of PD within pit fields and density of pit fields per cell interface were analysed using SEM and confocal microscopy, respectively, in PCK (n=3 species), NAD-ME (n=4), and NADP-ME (n=6) grasses. For reference, representative grasses from C 3 BEP (n=4) and C 3 PACMAD (n=1) were also measured. Quantitation of PD per µm 2 pit field (n=18 or more whole pit fields obtained from SEM), percent pit field per cell interface area (n=5 or more maximum intensity projection images generated from two confocal z-stacks), and PD per µm 2 cell interface were carried out as described in Danila et al. (2016). PD quantification values used for O. sativa, T. aestivum, Z. mays, and S. viridis were as reported in Danila et al. (2016) (see Supplementary Table S1 for specific details). The cross-sectional area of at least 40 individual PD enclosed by the wall collar (Faulkner et al., 2008) (termed as PD area here) located in the M-BS cell interface was measured from SEM images. PD area per M-BS interface area and PD area per unit leaf area were calculated as follows: All anatomical measurements were performed using ImageJ software (https://imagej.nih.gov/ij/).

Statistical analysis
Statistical analyses were carried out using one-way (photosynthetic type and species) ANOVA (OriginPro 9.1, OriginLab Corporation).
Means were grouped using a post hoc Tukey test.

C 4 origin and lineage representation
A phylogenetic tree of the 18 species was adapted from GPWGII (2012). It is currently thought that this set of species encompass four independent origins of C 4 photosynthesis (GPWGII, 2012). The independent evolutionary origins of C 4 are indicated in Fig. 1 where species are colour-coded according to their photosynthetic type (Table 1); this coding and species order are retained throughout the paper.

Plasmodesmata in C 4 grasses
Analysis of pit field size (Fig. 2, Supplementary Table S1) and patterns of pit field distribution (Fig. 3) revealed that NAD-ME species had the largest and most abundant pit fields (in terms of area coverage) on the M-BS cell interface. Both NADP-ME and PCK species had smaller and less abundant pit fields. C 3 species, from both the BEP ( Table S1). Indeed, NAD-ME species had fewer PD per pit field area on the M-BS cell interface compared to NADP-ME and most PCK species (Fig. 4A, Supplementary Table S1). This was offset by the greater percent pit field area per M-BS cell interface area in NAD-ME species compared to NADP-ME and PCK species (Fig. 4B, Supplementary Table S1). The resulting PD density per M-BS cell interface was greater in NAD-ME compared to NADP-ME species, with large variation observed amongst the PCK species (Fig. 4C, Supplementary Table S1). C 4 species also had greater PD density on the M-M cell interface relative to the C 3 species but there was no substantial variation among the C 4 subtypes (Figs. 4D-F, Supplementary Table S1). Estimates of the cross-sectional area of individual PD revealed no significant differences between the two photosynthetic pathways and among decarboxylation types (Fig. 4G, Supplementary Table  S1). The proportion of the M-BS cell interface populated by PD (Fig. 4H, Supplementary Table S1) and the M-BS PD area per unit leaf area (Fig. 4I, Supplementary Table S1) were greater in C 4 species than C 3 species, and followed the pattern of PD density for the C 4 decarboxylation types.

Bundle sheath of C 4 grasses
Our 3-D approach to measure BS cell cross-sectional areas and volumes used confocal micrographs derived from z-stacks of the leaf (Fig. 5). Measurement of the BS cell cross-sectional areas revealed no significant difference between the C 3 and C 4 species examined (Fig. 6A, Supplementary Table  S2). Although shorter BS cell length in C 4 species compared to C 3 species was observed (Fig. 6B, Supplementary Table  S2), the calculated BS cell volumes were similar for the C 3 and C 4 species examined (Fig. 6C, Supplementary Table S2).
Measurements from light micrographs of transverse leaf sections showed NAD-ME and NADP-ME species had the largest and smallest vein diameter, respectively, with C 3 and PCK species being intermediate (Fig. 6D, Supplementary Table  S2). As expected, leaf interveinal distance (IVD) was larger in C 3 species than in C 4 species (Fig. 6E, Supplementary Table  S2), but among the C 4 species the IVDs were not significantly different (Fig. 6E, Supplementary Table S2). Conversely, BS  cell surface area per unit leaf area (S b ) of C 4 species was double that of C 3 species (Fig. 6F, Supplementary Table S2).

Discussion
Previous studies have shown that PD are more abundant at the M-BS cell interface in a C 4 leaf compared to a C 3 leaf (Botha, 1992;Danila et al., 2016), presumably to accommodate the higher demand for metabolite transport between cell types in the C 4 leaf that is required to support the C 4 photosynthetic mechanism (Hatch and Osmond, 1976;Weber and von Caemmerer, 2010). However, little is known about the variation of PD density at the M-BS interface amongst C 4 species and the different decarboxylation types.
Modelling of metabolite movement between M and BS cells has been hampered by the lack of quantitative data on PD density at this key cellular interface (Danila et al., 2016). To address this, we extended our PD density quantification to a larger subset of grasses representative of C 3 photosynthesis, in both BEP (n=4) and PACMAD (n=1) clades, and in all the C 4 decarboxylation types: NAD-ME (n=4), NADP-ME (n=6), and PCK (n=3) ( Table 1). We confirmed that not only did all the C 4 species examined have greater PD density than the C 3 species, but that there was also substantial variation in how the high M-BS PD density was achieved among different C 4 subtypes.

Increased PD density in C 4 grasses is a result of larger pit fields and/or more abundant PD per pit field area
Here we show that C 4 species have evolved greater PD density than C 3 species. Greater symplastic connectivity can be achieved by increasing the pit field area or by increasing the PD per pit field area. Interestingly we saw both solutions in our data, particularly in the M-BS interface. In NAD-ME species we saw an increase in pit field area without increasing PD per pit field area, while in PCK and NADP-ME species we saw an increase in both. We propose that the NAD-ME solution is due to their larger veins and thus their need for larger pit fields to facilitate transport. However, in PCK and NADP-ME types where veins are smaller it is sufficient just to increase PD per pit field area without much increase in pit field area to achieve the same effect. These results indicate that there is genetic plasticity in the way in which increased PD transport is achieved in C 4 grasses.

NAD-ME grasses have the greatest pit field area per M-BS interface among the C 4 decarboxylation types
High PD density at the M-BS interface in NAD-ME grasses was solely determined by increases in pit field area per M-BS area driven by larger pit fields (rather than more numerous small pit fields). This is interesting in light of the unique aspects of NAD-ME BS cell anatomy (Dengler et al., 1994). The distinct characteristics of BS cell chloroplasts and mitochondria in NAD-ME species compared to NADP-ME and PCK types (Dengler et al., 1994) are consistent with the different solution they used to achieve high M-BS PD density. This supports the suggestion that PD function and formation are strongly coordinated with the function of both chloroplasts and mitochondria (Brunkard et al., 2013;Wang et al., 2017). Both the centripetal arrangement of mitochondria and larger BS cell cross-sectional area provide a longer diffusion pathlength in the NAD-ME leaf (von Caemmerer et al., 2007). While this anatomical attribute minimises CO 2 leakage across the M-BS interface, it may limit the rate of the C 4 cycle activity (von Caemmerer et al., 2007). More PD connections between M and BS cells allows rapid metabolite shuttling for the C 4 cycle in the NAD-ME type (von Caemmerer et al., 2007), therefore sustaining the high C 4 photosynthetic rate (Henderson et al., 1992;Pinto et al., 2014). Correspondingly, the centrifugal arrangement of chloroplasts and mitochondria in the NADP-ME type presents a greater possibility of CO 2 leakage (von Caemmerer et al., 2007). However, fewer PD between M and BS cells in NADP-ME leaves, in combination with the suberin lamella surrounding the BS, minimises this possibility (von Caemmerer et al., 2007). It is interesting that there was almost as much diversity in the PD density at the M-BS interface among PCK species as there was across C 4 grasses as a whole. This wide range of PD densities in the PCK species examined could result from both PCK and NAD decarboxylation located in the BS (von Caemmerer and Furbank, 2003) together with their considerable variation in BS cell cross-sectional areas, BS chloroplast morphology and positioning, and abundance of mitochondria in the BS (Hattersley and Browning, 1981).

Large PD size between photosynthetic cells is found in leaves of all the grass species examined
Using TEM sections for PD studies presents advantages and disadvantages, highly dependent on the purpose of the study (i.e. ultrastructure versus quantification). For our purpose of analysing many samples, SEM analysis was a more rapid and rigorous way to quantify the cross-sectional area of individual PD in grasses. Random tearing of leaf tissue to reveal M-BS or M-M cell interfaces allowed us to expose the cross-sectional area of PD, almost always within the middle cavity because the tissue tended to separate along the middle lamella between tissue layers. In our previous paper (Danila et al., 2016), our PD area measurements from TEM micrographs used only PD cross-sections with distinct central desmotubules. Similar measurements taken from previously published TEM micrographs of grass leaf PD ( fig. 8 from Botha, 1992) generated values similar to both our TEM and SEM results. In fact, the similarity of PD area values we obtained when we compared our TEM and SEM measurements for O. sativa, Z. mays, and S. viridis encouraged us to use SEM in place of TEM for PD area measurement. We find it interesting that the PD areas we obtained in all cases were very large (about 0.006 μm 2 ), the diameter being in the range of 90 nm, while the majority of published values for land-plant PD, many of which were obtained from root PD measurements, have a diameter of 50 nm or smaller (Overall, 1999;Ehlers and Kollmann, 2001).

Anatomical enablers of C 4 photosynthesis
It has been proposed that enlargement of the BS cells in C 4 leaves compared to C 3 leaves and their 'functionalisation' by increases in chloroplast number was an early step in C 4 evolution (Sage, 2004). Clearly, in this study we showed that there is little evidence for BS cells in C 4 grasses being larger in volume than their C 3 counterparts. What we saw instead was a distinctively large BS surface area to leaf area ratio (S b ) in C 4 grass leaves compared to C 3 leaves, consistent with the report of Hattersley (1984). We therefore argue that increasing S b , but not BS cell size, is important in C 4 leaf physiology (von Caemmerer et al., 2007). This finding emphasises the importance of looking at the 3-D geometry of cells in addition to the 2-D view for a more global cell perspective and for improved accuracy in terms of reporting measured values (Théroux-Rancourt et al., 2017). Large IVD is also not always a clear indication of C 3 anatomy because it can be interpreted as either an increase in interveinal M cell number (as in C 3 grasses) or an increase in BS cell crosssectional area (as in most NAD-ME and PCK grasses). Our observations showed consistently greater PD density on the M-BS interface in all the C 4 species examined relative to the C 3 species. Another potentially useful diagnostic character is pit field density, as seen in the substantial difference between the NAD-ME types and other decarboxylation types. This could be used to distinguish not only C 3 from C 4 photosynthesis but also among C 4 biochemical types, at least in grasses.

PD density, metabolite flux, and CO 2 diffusion in the C 4 BS
PD density at the M-BS interface affects not only metabolite diffusion but also leakage of inorganic carbon and O 2 out of the BS compartment, an important determinant of the efficiency of C 4 photosynthesis (von Caemmerer and Furbank, 2003). While it is difficult to directly determine the CO 2 concentration in the BS of C 4 plants, this parameter can be modelled using certain assumptions concerning leaf cell anatomical dimensions, inorganic carbon equilibration, and diffusion properties of membranes and PD (Furbank and Hatch, 1987;Jenkins et al., 1989;von Caemmerer and Furbank, 2003). Based on permeability coefficients determined for metabolites moving into isolated BS cells through PD (Weiner et al., 1988) and for CO 2 in C 4 leaves and isolated BS cells Jenkins et al., 1989), Jenkins et al. (1989) calculated that approximately 40% of the CO 2 leakage from the BS occurs via an apoplastic route and 60% via the PD. For oxygen, which moves poorly through lipid bilayers and polymeric barriers such as suberin , the majority of diffusion will be through the aqueous route, i.e. via PD. Likewise, bicarbonate moves poorly through lipid membranes and will diffuse out of the BS mostly via PD . Given these modelled values for inorganic carbon and O 2 movement at the M-BS interface, the 'optimal' PD density for a C 4 leaf would be a compromise between high density for metabolite transport and dissipation of O 2 (in species with PSII in the BS chloroplast), and leakage of inorganic carbon from the BS compartment. The quantification of PD density made here is a first step towards enabling more accurate modelling of metabolite flux and leakage of inorganic carbon from the BS across a range of species and decarboxylation types (Wang et al., 2014). The key parameter required for predicting metabolite gradients and leaf cell metabolite concentrations required to support the C 4 pathway is the proportion of M-BS interface comprised of PD pores (Osmond, 1971;Hatch and Osmond, 1976). The current work provides these data for a range of species. The values provide an upper limit given that not all PD may be functional and the obstruction of the PD area by the desmotubule has not been taken into account. Nevertheless, these data facilitate modelling of inorganic carbon leakage rates from the BS without resorting to the use of permeability values obtained from isolated cells ). This will be particularly useful for interspecific comparisons given the diversity in suberisation of the M-BS interface between species and the diverse arrangements of organelles and cellular sites of C 4 acid decarboxylation across the three biochemical types.

Evolution of symplastic connections to the BS in C 4 leaves
It has been reported previously that the transition from C 3 to C 4 photosynthesis involved a series of genetic alterations, mostly gain of function, leading to numerous changes in plant anatomy and biochemistry (Sage et al., 2012;Bräutigam et al., 2014;Wang et al., 2014;Emms et al., 2016). Recently it was suggested that evolution of C 4 photosynthesis has involved a change in the apoplastic transport of sugars in the BS cells of C 4 leaves (Emms et al., 2016). A sugar effluxer (or SWEET protein) appears to have been recruited from a relatively minor role in the M cells of C 3 grasses to become one of the most highly expressed transcripts in the C 4 BS cells (Emms et al., 2016). While we have not as yet identified the genetic changes responsible, we show here that PD density was at least doubled in C 4 species compared to C 3 species, a clear indication of enhanced expression of PD developmental genes. However, to date, the genes underpinning PD development remain largely unknown (Brunkard and Zambryski, 2017). It is intriguing to consider whether the proposed evolutionary pressures to recruit a highly expressed sugar effluxer to the BS cells of C 4 plants were linked to the proliferation of PD at the M-BS interface in C 4 leaves. Recent data from genome-wide gene-tree/species-tree reconciliation in grasses has revealed multiple functional categories of genes, of which 10 genes were found to have plausible association with the symplastic transport function (Emms et al., 2016). This leads us to be optimistic that the discovery of candidate genes controlling PD development may be not too far away. Fig. 6. Distribution of bundle sheath-related trait values among photosynthetic types. The distribution of six variables is summarised by boxplots individually for the C 3 BEP and PACMAD (black, n=5), C 4 PCK (green, n=3), C 4 NAD-ME (blue, n=4), and C 4 NADP-ME (magenta, n=6). Boxes and whisker plots are as described in Fig. 4. Data for individual species are given in Supplementary Table S2. BS, bundle sheath; S b , bundle sheath surface area per unit leaf area.

Supplementary data
Supplementary data are available at JXB online. Dataset S1. Single-copy orthologous gene sequences used for phylogenetic tree construction (supplied as a BZIP2 file). Fig. S1. Light micrographs of transverse sections of C 3 and C 4 grass leaves. Table S1. Quantitative plasmodesmata traits of the 18 grass species examined. Table S2. Leaf anatomical traits quantified in the 18 grass species examined.