You're so vein: bundle sheath physiology, phylogeny and evolution in C3 and C4 plants



Bundle sheath (BS) anatomy is found in most C4 lineages, associated with low inter-veinal distances (IVD) and high BS:mesophyll ratio (BS:MC). The origins, function and selective advantages of the BS in C3 lineages are relevant for understanding the environmental, molecular and phylogenetic determinants of C4 evolution. Suggested functions for BS have included structural support, hydraulic isolation, storage for water, ions, and carbohydrates, and photorespiratory carbon metabolism; we propose a central role for cavitation repair, consistent with the BS as a control centre on regulating stem and leaf hydraulic continuity. An analysis of BS traits in the phylogenetic lineages giving rise to C4 grasses (the ‘PACMAD’ clade) shows an initial enhancement in BS:MC ratio in C3 lineages, although IVD is similar to the Pooideae sister group. Using a global database, a well-developed BS in the C3 PACMAD lineages was associated with higher precipitation and temperatures in the habitat of origin on an annual basis, with the C3 to C4 progression defined by the aridity index (AI). Maintaining leaf hydraulic conductance and cavitation repair are consistent with increased evaporative demand and more seasonal precipitation as drivers, first for the C3 BS, and then C4 diversification, under declining CO2 concentrations in the Palaeogene and Neogene.


The C4 pathway is known as a derived condition, which improves the operating efficiency of photosynthesis and water use (Sage & Monson 1998; Sage 2004; Hibberd & Covshoff 2010) and has arisen independently over 60 times from the basal C3 system (Christin et al. 2011; Sage, Christin & Edwards 2011). For all higher plants, there is a trade-off between demand for CO2, as regulated by stomata and transfer to the sites of carboxylation (stomatal and mesophyll conductances, respectively), and water availability delivered via root, stem and leaf hydraulic conductances. The Calvin–Benson–Bassham cycle is central to CO2 assimilation in C3 and C4 plants, although the C4 pathway improves both radiation and water use efficiencies. C4 carboxylation pathways are spatially separated within Kranz anatomy, coupling outer mesophyll cells (MC) and an inner ring of bundle sheath (BS), or occasionally mestome sheath, cells (Hattersley 1984; Dengler et al. 1994; Ogle 2003). An initial 4-carbon acid, produced in the MC via Phosphoenol pyruvate carboxylase (PEPc), diffuses to the BS cells and is used to generate elevated carbon dioxide concentrations around the primary carboxylase, Rubisco, also indirectly reducing water demand via lowered stomatal conductances (Hibberd & Covshoff 2010; Langdale 2011).

C4 plants are currently the focus of considerable research interest in both public and private sectors, for the development of highly productive crops such as maize and sugar cane, or bioenergy crops such as Miscanthus or switchgrass (Somerville et al. 2010) with the highest productivity recorded that of the C4 grass Echinochloa (Morison et al. 2000) in the Amazon basin. The prospect of introducing C4 traits into staple crops such as rice or wheat is being driven forward by promising research programmes (Hibberd, Sheehy & Langdale 2008; Hibberd & Covshoff 2010; Such research is identifying the key components of the C4 biochemical concentrating mechanism, and the molecular mechanisms controlling their cell-specific targeting mechanisms between mesophyll and BS (Leegood 2008; Brown et al. 2011; Langdale 2011).

There is some controversy as to the driving forces for C4 plant evolution, and likely climatic conditions which provided a selective advantage for the biochemical carbon concentrating mechanism (CCM) during the Miocene (Ehleringer, Cerling & Helliker 1997; Sage 2004; Edwards & Smith 2010). These conditions include the reduction in ambient CO2 concentration and changing precipitation patterns, while C4 advantages include quantum yield, stomatal conductance and hydraulic continuity (Sage 2004; Osborne & Freckleton 2009; Osborne & Sack 2012), as discussed below.

The aim of this review is to explore the original function of the BS in C3 lineages, providing an insight for selection pressures leading to the derived C4 pathway. Kranz anatomy is normally associated with reduced inter-veinal distance (IVD), and increased BS:MC ratio (Hattersley 1984; Dengler et al. 1994; Ogle 2003). The hypothesis we set out to test was that there is an evolutionary advantage conferred by the BS in terms of tolerance to moderate water deficits, perhaps through cavitation repair, and overall as a control centre for water supply and distribution. We also compared the extent of BS development, derived from published measurements of IVD and cross-sectional BS:MC area in PACMAD and Pooideae groups, augmented by new anatomical data. A secondary hypothesis was that both BS and IVD characteristics would be more pronounced in C3 members of the PACMAD clade. A phylogenetic comparison of BS proportions and IVD in grass clades was translated into global distribution patterns and climatic preferences using Global Biodiversity Information Facility (GBIF), for contrasting BS size-classes. Here, the aim was to investigate whether the extent of BS development as a function of annual precipitation or mean temperature, was also consistent with the C3 to C4 progression.

Evolution of the C4 pathway: current focus of palaeoclimatic considerations

Given the tremendous diversity in C4 plants, with over 60 independent origins (Sage 2004; Christin et al. 2011), there has been considerable debate about the environmental determinants of this derived pathway. From a palaeohistorical perspective, it has now been shown that both monocot and dicot C4 lineages show molecular divergence around 20–30 mya (Christin et al. 2011), in contrast to the later vegetative dominance of C4 grasses in savannah biomes (5–8 mya: Cerling et al. 1997; Ehleringer et al. 1997). Under declining atmospheric CO2 concentrations in the Miocene, the combined energetic cost of photorespiration and compensatory water loss (Ehleringer et al. 1997), could be ameliorated by the lower stomatal conductances and CCM inherent to C4 plant physiology (Osborne & Freckleton 2009; Osborne & Sack 2012). The reduction of ambient CO2 concentrations to 280 ppmv, perhaps 2 mya, is too late to have driven C4 diversification, and so the role of CO2 concentration as a sole driving force has been questioned (Sage 2004).

Recent phylogenetic comparisons of C3 and C4 traits suggest that the C4 pathway generally seems to have originated in seasonally dry, warm habitats (Osborne & Freckleton 2009; Edwards & Smith 2010) where a major driver was leaf hydraulic conductance (Osborne & Sack 2012). The higher evaporative demand in open, savannah habitats, and likely decrease in subtropical forest cover (Edwards & Smith 2010) are consistent with a need to cope with higher evaporative demand. Additionally, estimates of atmospheric CO2 decline (Beerling & Royer 2011) are consistent with recent modelling studies on ambient CO2 and comparative hydraulic limitations for C3 and C4 systems (Osborne & Sack 2012). Given the higher stomatal conductances in a C3 leaf, CO2 concentrations as high as 800–1500 ppmv would result in 50% of C3 veins suffering hydraulic failure under increasing temperatures and evaporative demand. Under equivalent conditions, C4 physiology would require much lower CO2 concentrations (200–500 ppmv), to suffer such extensive hydraulic failure (Osborne & Sack 2012).

There are complications: C4 plants need to maintain a high hydraulic conductance (Osborne & Sack 2012) and are often found physiologically to be less drought-tolerant than sympatric, and closely related, C3 plants (Taylor et al. 2011). Additionally, the diversification of some C4 lineages in wetlands, whether tropical or temperate (Long 1983; Morison et al. 2000) seems to represent a convergent use of C4 productivity potential, analogous to the co-evolution of Crassulacean acid metabolism (CAM) in semi-arid succulents, tropical forest epiphytes and temperate lake isoetids (Griffiths 1989). In general, as grasses diversified in savannah habitats, a major selective pressure was water demand under more exposed conditions, as seasonal precipitation patterns developed in more subtropical biomes (Edwards & Smith 2010; Osborne & Sack 2012).

Environmental determinants of BS development in C3 plants

In palaeohistorical terms, increasing xylem complexity and vein density is part of a phylogenetic progression as plant life has conquered and diversified on land (Raven 1993; Edwards, Kerp & Hass 1998; Brodribb & Feild, 2010; Brodribb, Feild & Sack, 2010). We now explore the divergence of more recent anatomical traits in the C3 lineages leading to C4 diversification in PACMAD clades, leading to reduced IVD and increased BS:MC ratio (Dengler et al. 1994; Sage 2001; Ogle 2003; Sage 2004; Christin et al. 2008; Hibberd & Covshoff 2010). The aim is to identify the physiological basis for the selective advantage of each trait, and their retention in the progression from C3 to C4 anatomy.

The scenario of decreasing ambient CO2, increased seasonality for precipitation inputs and evaporative demand in more exposed habitats suggests that the concept of ‘desiccation or starvation’ would have been a likely problem for C3 ecosystems in the Miocene (Ward et al. 2005; McDowell et al. 2008). Here, low ambient CO2 (Ca) causes higher rates of photorespiratory losses and reduced diffusive supply (‘starvation’); stomatal opening, to compensate for the low internal CO2 (Ci), leads to desiccation. One factor coupling water supply, and demand, would be partitioning of hydraulic conductance (Kh) between stem and leaves (Sack & Holbrook, 2006; Brodribb et al., 2010; Johnson et al. 2011). Recently there has been much interest in whether this is regulated symplastically or apoplastically in leaves (Sack & Holbrook, 2006; Nardini, Salleo & Jansen 2011). Furthermore, the correlation between reduced leaf size and increased vein density, likely to have been part of the progression towards C4 (Sage 2001; Sage 2004), has recently been suggested to provide a direct hydraulic benefit to leaves (Brodribb et al., 2010; Scoffoni et al. 2011).

Drought stress can have an impact on xylem conductivity via the direct effect of xylem collapse or indirect effect of cavitation (for review, see Scoffoni et al. 2011). With calculations by Osborne & Sack (2012) showing how low CO2 concentrations would disadvantage C3 systems (relative to any C4 species with the overall same Kh), then we might expect that the decreased IVD associated with the evolutionary progression within C3-C4 intermediates (Hattersley 1984; Dengler et al. 1994; Ogle 2003; McKown & Dengler 2009; Muhaidat et al. 2011; Sage et al. 2011) was initially set in course by water deficits. Thus, IVD might be expected to decrease in grasses when adapting to the more open savannah canopies, with increases in light, temperature and vapour pressure deficit as well as more seasonal precipitation inputs (as discussed above). Such changes in the vascular system would be concomitant with the progression in stomatal pore size and density, also observed for C4 grasses (Taylor et al. 2012).

Meanwhile, it seems that the conditions leading to the C4 pathway in the PACMAD clade were more consistent with exposed, seasonally water-limited, subtropical habitats. The difference between mesophytic conditions which maintain forest cover, or promote savannah development can be slight (respectively, between 1500 and 1200 mm annual precipitation; Edwards & Smith 2010), and this precipitation crossover varies with latitude (Lehmann et al. 2011; Ratnam et al. 2011). The disadvantages of desiccation in a low CO2 world, and increased seasonality, are also consistent with the grass basal meristems, mobilization and storage of nutrients below-ground, and resistance to fire in subtropical habitats (Taylor et al. 2011; Osborne & Sack 2012).

The interplay between reduced IVD and enhanced BS development (as higher BS:MC ratio) is explored below for the PACMAD clade and closely related groups (Pooideae and Ehrharteae), and in relation to their global distribution. The Pooideae diversified into more temperate, cooler conditions (Edwards & Smith 2010). Here, low temperatures may also have caused problems of hydraulic continuity and cavitation, particularly as most temperate C3 grasses complete vegetative growth and flowering early in the growing season. If selection pressures for IVD and BS:MC for grasses evolving under both subtropical and cool-winter conditions were similar, it would indicate functional equivalence for maintenance of leaf hydraulic conductance in a wide range of C3 grassland habitats, and explain the selection pressure for enhanced BS development in both regions.

Form and function of IVD and BS in C3 progenitors

There are over 60 independent origins of the C4 pathway, and also 22 variants in BS morphology (Sage et al. 2011). In grasses there are nine identifiable BS variants, five of which have an inner mestome sheath lacking chloroplasts, while four others have the parenchyma BS cells in direct contact with the vasculature (Dengler & Nelson 1999). In order to address the physiological functions originally provided by the BS and mestome sheath in C3 leaves, Fig. 1 is a schematic which depicts a simplified Kranz anatomy and summarizes many key functions for a BS system in modern-day C3 plants (Fig. 1: BS cells), focusing on water supply rather than phloem redistribution. Adjacent panels in the figure contain associated functional attributes, and a more detailed description and rationale for inclusion of each nominal BS function is given in the sections that follow.

Figure 1.

Bundle sheath (BS) schematic depicting likely physiological role for BS and associated functions.

Hydraulic isolation and ‘smart pipe’

Recent research has focused on the role of the leaf vasculature in providing some degree of control over leaf hydraulic conductance (for review: Sack & Holbrook, 2006). Rather than acting as a bottleneck (Ache et al. 2010), evidence from Arabidopsis suggests the BS to have a more specific regulatory function in terms of maximizing leaf hydraulic conductance as a ‘stress-regulated valve’ (Shatil-Cohen, Attia & Moshelion 2011). Aquaporins in BS membranes respond to below-ground ABA signals, and the reduced water flow to the mesophyll would be one way to regulate the balance between water deficit and stomatal responses. The implications of this study are that hydropassive control of stomata (from changes in tissue or epidermal turgor), or hydroactive control (via metabolism, or ABA signals from leaf or root), could be mediated by the BS.

Additional functions as a ‘smart pipe’ were identified by the way that sclerified BS cells could prevent flooding of the mesophyll tissue during root pressure (Shatil-Cohen et al. 2011). This is another mechanism which would link the BS to hydraulic repair under moderately water-stressed mesophytic conditions, particularly in view of the high proportion of root and stem conducting tissues which cavitate on a daily basis in forbs and grasses, with up to 70% loss of conduction in maize xylem (McCully, Huang & Ling 1998). In conclusion, the BS could provide a key role in controlling hydraulic fluxes and offer a selective advantage for C3 plants under mesophytic subtropical and temperate conditions.

Mechanical support

The likelihood that water deficits were a driver of increased vein density have been discussed above, and reviewed by Brodribb & Feild (2010), and Brodribb et al. (2010). Increased vein density could provide a mechanical advantage in C4 plant lineages under dry, windy conditions (Sage 2004), where the cost of higher lignin content would be repaid rapidly under high light conditions (Sack & Frole, 2006). A comparative study on tropical tree leaf hydraulic conductivity showed a positive relationship between leaf density and resistance to cavitation, but not with specific leaf area (Markesteijn et al. 2011). Coordinated hydraulic traits have been suggested to have promoted angiosperm evolution, with the cost of investment in xylem repaid by acclimation to match water supply with demand (Brodribb & Feild, 2010). More detailed ultrastructural studies are needed if we are to understand whether the sclerified BS walls add to such a mechanical advantage, as well as considering the longitudinal extent of such BS cells, their role around the vasculature, and the partitioning caused by BS extensions, relative to those found in C4 plants today.

Capacitance – storage of water and increased surface area for regulated supply to mesophyll

One function of BS cells which seems to underpin many of the physiological roles incorporated and summarized in this section, is that of water storage. Suggested to act as a reservoir of water, which could cope with a sudden ‘evaporative surge’ under windy conditions (Sage 2001), many of the alternative BS functions identified in Fig. 1 could be supported by this capacitance function. By maximizing BS surface area, the surrounding MC can be replenished efficiently, allowing both liquid phase and gaseous diffusion path lengths to be optimized in relation to stomatal control (Sage 2004; Sack & Holbrook, 2006; Osborne & Sack 2012). Our subsequent understanding of the regulation of aquaporins to mediate symplastic transfer is also consistent with this role in water storage and supply (for review: Sack & Holbrook, 2006; Nardini et al. 2011; Shatil-Cohen et al. 2011).

Capacitance – storage of ions

A role for osmolyte storage has also been proposed for BS cells. X-ray imaging has shown that depending on the type of BS cells, specific ions are preferentially accumulated in barley, such as K+ and Mg2+ (Williams et al. 1993). Additional evidence has shown that sodium is excluded from the leaf mesophyll by the BS in banana leaves, with the exception of leaf margins, where the BS is incomplete (Shapira et al. 2009). Other work demonstrating ion-specific selectivity and the ‘smart-pipe’ concept (Shatil-Cohen et al. 2011) relates to boron, which is largely excreted in guttation fluid rather than passing into the leaf mesophyll (Sutton et al. 2007). The function of these osmolytes may also have direct role in cavitation repair, as we consider below.

Capacitance – storage of sugars

Williams, Farrar & Pollock (1989) studied carbohydrate accumulation in BS cells of barley, and defined three cell categories as well as structural cells, found at the top of the xylem arc. Such structural differentiation, often with an incomplete ring of cells, may well indicate how additional functions, such as the C4 pathway, have been co-opted by various types of BS and mestome sheath configurations in the PACMAD lineages (Dengler & Nelson 1999). Williams et al. (1989) showed contrasting degrees of soluble and insoluble carbohydrate accumulation between mesophyll and BS cells. Subsequent studies showed that the fructan:sucrose ratio was higher in BS cells, perhaps associated with a storage role for BS cells, either associated with control of carbohydrate export, or, as we explore below, repair of cavitation.

Loss of xylem conductivity and cavitation repair

Cavitation and loss of hydraulic conductivity is a real and present danger in all transpiring systems: detectable as audible clicks, when emboli occur at high transpiration rates (Jackson, Irvine & Grace 1995; Johnson et al. 2011), cavitation is an issue for trees and herbaceous plants (McCully et al. 1998; McCully 1999; Sack & Holbrook, 2006; Zwieniecki & Holbrook 2009). Maize is likely to suffer from 50 to 70% loss of xylem conductivity on a daily basis (McCully et al. 1998), with repair occurring dynamically in root xylem even during the day (McCully 1999). Under well-watered conditions, root pressure is normally thought to repair cavitation at night, with a positive sap pressure derived from active transport of solutes from the root stele symplast into the xylem apoplast. Root pressure can be visualized as guttation, whereby water is channelled via the BS to hydathodes in the leaf margin (Shatil-Cohen et al. 2011). However, as soil water deficits develop, it would be interesting to determine the thermodynamic threshold for root pressure at night, and the frequency of repair to embolized vessels dynamically during the day, for roots, stems and leaves within herbaceous plants (McCully 1999). Factors which maintain ‘safe’ hydraulic conductances as modelled by Osborne & Sack (2012), are in addition to decreasing IVD and enhanced BS cells, the extent that repairs to cavitated vessels could be made in actively transpiring stems and leaves (McCully 1999; Zwieniecki & Holbrook 2009).

A solution has recently been suggested by Secchi and Zwieniecki (2011) and by Zwieniecki & Holbrook (2009). An earlier explanation had described how active solute transport could be used to reload cavitated vessels, with air progressively re-dissolving as the vessel lumen filled, until a critical point when hydraulic continuity and transpiration flow would be re-established by the expediency of refilling vessel pits synchronously (Holbrook & Zwieniecki 1999; Zwieniecki & Holbrook 2009). However, a theoretical analysis of water storage in xylem parenchyma, and the volume and flux needed to refill cavitated vessels, has now led to the suggestion that water vapour diffuses across from adjacent, actively transpiring vessels (Zwieniecki & Holbrook 2009; Secchi & Zwieniecki, 2011).

While Sage (2001) discussed whether BS cells might help to prevent cavitation, we hypothesize that cavitation, and active repair, could have been a significant driver for increased BS development and reduced IVD, by maintaining Kh and helping to resist desiccation and carbon starvation in C3 grasses and forbs in the Miocene. While the BS cells would be needed to act as a reservoir for water, solutes and carbohydrates (as described above), their requirement to drive active transport, and cavitation repair, has yet to be tested experimentally.

Refixation of (photo)respiratory CO2

Photosynthetic activity in BS cells has long been associated with the origins of the C4 pathway. Organic acids supplied in the transpiration stream from roots were decarboxylated and the CO2 refixed photosynthetically within cells adjacent to the vascular bundle (Hibberd & Quick 2002; Brown et al. 2010). Refixation of photorespiratory CO2 within the BS of C3-C4 intermediates was suggested to have initiated the progression to C4, because of the localization of glycine decarboxylase only in BS mitochondria (Hylton et al. 1988). Subsequently, this ‘C2’ pathway is recognized as a common step in C4 development (Griffiths 1989; Gowik & Westhoff, 2011; Muhaidat et al. 2011; Nelson 2011), which also increases the demand for re-assimilation of ammonia generated during the glycine decarboxylase reaction (Berry, personal communication).

However, other evidence suggests that 14C fixation within barley BS cells contributes to carbon storage and partitioning (Koroleva et al. 2000; Pollock et al. 2003). The photosynthetic metabolism associated with vascular tissues (Hibberd & Quick 2002), and distribution of chloroplasts within BS cells to recapture such (photo)respiratory sources (Hylton et al. 1988; Williams et al. 1989) would contribute to both carbon and nitrogen balance, and perhaps the considerable energetic demand for cavitation repair (Zwieniecki & Holbrook 2009; Secchi & Zwieniecki, 2011).

BS as control centre integrating xylem, mesophyll and stomatal conductances

Sack and Holbrook (2006) suggest a role for the BS as a control centre to match xylem hydraulic conductance with mesophyll hydraulic demand. Additional roles, as a hydraulic smart pipe, would include responding to changing stomatal conductances and evaporative fluxes and sensing ABA signalling, as well as regulating root pressure flows and buffering water supply to the mesophyll (Osborne & Sack 2012). Thus, the increased BS volume allows storage of water, ions and carbohydrates, as well as the capacity to refix respiratory CO2, and control of water flux to the mesophyll via aquaporins (Sack & Holbrook, 2006; Shatil-Cohen et al. 2011). Anatomically, the BS functions to reduce diffusion path lengths between veins and MC, with increased vein density helping to regulate water flow under increased evaporative demand and reduced ambient CO2 concentrations. Additionally, key reactive oxygen signalling processes, integrating light and ABA signalling, water, have also been demonstrated to be regulated within the BS cells (Galvez-Valdivieso et al. 2009). Empirical observations and modelling outputs suggest that a high hydraulic conductance is needed by the C4 pathway (Osborne & Sack 2012), and that relatively high CO2 thresholds led to selective advantages over sympatric C3 systems. These observations are consistent with the BS being needed to maintain hydraulic continuity, as well as then providing a spatial location for carboxylation pathways in developing C4 systems.

IVD and increased BS:MC ratio in PACMAD and sister lineages leading to C4 pathway

As a means to investigate the development of BS traits in the C3 lineages in the PACMAD clade, measurements of IVD and BS:MC ratio were compared with both C4 PACMAD lines and the sister groups, Pooideae and Ehrharteae. This builds on recent analyses of the phylogeny of C4 evolution, which focused on the origins and dispersal of C4 traits (Christin et al. 2008; Edwards & Smith 2010; Sage et al. 2011; Osborne & Sack 2012). Qualitative relationships between C3 and C4 lines within the PACMAD clade are shown in Fig. 2, which depicts the relative shifts in IVD and BS:MC ratio in C3 and C4 lineages, for which anatomical data were published (Hattersley 1984; Ogle 2003) or have been obtained experimentally (see Methodology below, and Supporting Information Table S1).

Figure 2.

Abbreviated phylogenetic tree of PACMAD grasses and closely related lineages, showing C3 and C4 origins and bundle sheath (BS) characteristics [inter-veinal distances (IVD), BS% area]. Figure is derived from PACMAD phylogeny (Christin et al. 2008): highlighted clades (enhanced lines) represent condensed C4 grass lineages, with C3 lineages as fine lines. Individual species were sorted into C3 or C4 lineages for measurements of IVD (Ogle 2003; Supporting Information Table S1) and percentage of BS area (Hattersley 1984; Supporting Information Table S1). Mean IVD and BS values for each lineage are indicated by a qualitative colour scale representing C4-like (red) to C3-like (green) values. [Correction added on 04 December 2012, after first online publication: In the ‘BS%’ column, the 13th data point has been corrected from 18% to 13%; 16th data point, from 29% to blank.]



In addition to data presented by Hattersley (1984) and Ogle (2003), leaf cross section slides were analysed from the Herbarium grass collection, Royal Botanic Gardens, Kew (Collector, sample origin, and corresponding herbarium sheet number are all listed in Supporting Information Table S1). Samples were viewed using a Leitz DMRB microscope, with Leica camera attachment [Leica Microsystems (UK) Ltd, Milton Keynes, UK], and analysed at 2.5× magnification using Olympus ‘analySIS’ software package (Olympus Imaging and Audio Ltd, Southend on Sea, UK). IVD was measured as the distance between the centre points of two adjacent major and minor vein vascular bundles, as a mean of 3–5 measurements for each leaf section, avoiding any central vein.

BS to mesophyll ratio (BS%)

Using the methodology of Hattersley (1984) the Herbarium slides were viewed normally under 20 × magnification, with three replicate images of one major and one minor vein analysed using the ‘closed polygon’ tool in analySIS, areas for total mesophyll (M), BS and vasculature; BS area ratio was calculated as a percentage from (BS Area/[BS Area + M Area]) for each species.

Data analysis

For C3 PACMAD species BS % area was divided into two groups, respectively, less than 16% BS (Low) or more than 19% BS (High). This limit was chosen based on a break in the data from c. 15% to 20% BS area, this also separated species into roughly equal numbers. Morphometric (IVD, BS%) data were compared using analysis of variance with post hoc Tukey's test, and shown as box plots with median (in bold), interquartile range (box) and whiskers (1.5× interquartile range), with outlying data indicated as circles. Occurrence data were sourced from GBIF data portal ( The GBIF portal was queried for georeferenced data (i.e. those species with latitude/longitude coordinates). Distribution maps were plotted globally using Q-GIS (2012 Quantum Geographic Information System; Open Source Geospatial Foundation Project, Phragmites australis (BS area of 27%) was excluded from the distribution maps as this species had a disproportionally large representation in occurrence records from temperate regions and is semiaquatic which differs from the other species examined that have terrestrial ecology.

The mean annual climatic data were derived for each location as a single mean value for each species to limit bias from over-represented individuals, plotted for C3 PACMAD BS subgroups and compared with C4 distribution data. Mean annual precipitation (MAP: mm year −1), mean annual temperature (MAT: °C) were derived from Hijmans et al. (2005). Mean annual potential evaporative transpiration (PET: mm year −1) and aridity index expressed as percentage [aridity index (AI): MAP.PET−1] were derived from Trabucco & Zomer (2009). Data display as box plots (see above), and compared using Dunnetts modified Tukey–Kramer pairwise multiple comparison test.

Combined data presentation

In Fig. 2, higher values for IVD and low BS:MC ratios (represented as green by the qualitative colour scale) are generally found in C3 lineages, as compared with the lower IVD and high BS:MC (represented as red by the qualitative colour scale) found in C4 lineages (Fig. 2). The relationship between reduced IVD and enhanced BS:MC ratio has been long recognized for C4 plants (Hattersley 1984; Dengler & Nelson 1999; Sage 2004). The qualitative colour scale in Fig. 2 indicates IVD convergence in some instances between C4 and C3 lineages of the PACMAD clade, compared with the Pooideae and Ehrhateae outgroups. This is illustrated in the subfamily Micrairoideae where there is a progression in IVD from C3 to C4 (192, 170, 107 µm), with a corresponding increase in BS% (respectively, 6, 13 to 31%). From the C3 Danthonioideae to C4 subfamily Chloridoideae, the progression is similar (IVD: 263 to 129 and 124; BS%: 15 to 23 and 42%) In the C3 subfamily Arundinoideae, the BS:MC ratio was 20%, as compared with a very wide range in C3 members of the Panicoideae (13 to 38%) where in many lineages C3 BS development was more pronounced than in the C4 Panicoids (16 to 33%) (Fig. 2).

A quantitative analysis of IVD and BS:MC ratio between C3 and C4 taxa belonging to the PACMAD clade and sister subfamilies of Pooideae, Ehrharteae and Bambusoideae (BEP) is provided in Fig 3. Species belonging to C4 PACMAD lineages have significantly lower IVD than those belonging to C3 PACMAD or BEP, for which groups the interquartile ranges are all similar.

Figure 3.

Comparison of (a) inter-veinal distances (IVD), and (b) bundle sheath (BS) characteristics between C3 and C4 lineages. Box plots represent taxa from C4 PACMAD, C3 PACMAD and combined sister groups Pooideae and Erharteae and Bambusoideae (PEB), with data for (a) IVD (µm) (data from Ogle 2003; Supporting Information Table S1) and (b) Bundle Sheath Area Ratio, derived as %: BS = (Bundle Sheath Area/[Mesophyll Cell Area + Bundle Sheath Area]) from Hattersley, 1984 and Supporting Information Table S1. (Sample numbers for Figs 3a and 3b, respectively: PACMAD C4, n = 52, 44; C3, n = 24, 18; combined PEB Pooideae, n = 12,9; Erharteae, n = 4,4; Bambusoideae, n = 3,3). Significance at P < 0.05 indicated by letters (A, B, C) via analysis of variance with post hoc Tukey's test. [Correction added on 04 December 2012, after first online publication: IVD and BS boxplots, as well as the figures in the caption, have been amended. Figure 3a: IVD C4 PACMAD max, from 200 to 169; C3 PACMAD min, from 124 to 146; 1st quartile, from 176 to 184. Figure 3b: bundle sheath area ratio C3 PACMAD median, from 21 to 18.5; 3rd quartile, from 25.5 to 23.5; PEB median, from 11 to 11.5; 3rd quartile, from 13 to 14.]

In contrast, the BS:MC ratio shows a clear progression within the PACMAD clade, with convergence between the C3 and derived C4 lineages (Fig. 3b), and increased BS:MC ratio relative to BEP. Similar to the comparison of IVD, BEP lineages maintained significantly different BS:MC ratios compared with PACMAD C4 (Fig. 3b). As a preliminary conclusion, differences between IVD and BS:MC ratio suggest that BS proliferation has occurred prior to increased vein density, in contrast to the progression often seen in C3-C4 intermediates within a single genus (e.g. Cleome: Marshall et al. 2007; Flaveria: McKown & Dengler 2007; Heliotropium: Muhaidat et al. 2011).

Distribution and climatic preferences within PACMAD as a function of BS development

C3 or C4 groups within PACMAD clade were further subdivided in terms of their BS:MC area ratios. Global distribution maps of known occurrence data for species with higher or lower BS:MS ratios were overlaid to contrast their environmental and spatial preferences (see Methodology). Within the PACMAD C3 lineages, taxa with higher BS:MC ratios (20–30% BS:MC) occurred in subtropical, seasonally arid regions, in contrast to a more temperate spread of species with a lower BS:MC ratio (5–15% BS:MC) (Fig. 4).

Figure 4.

Global distribution of PACMAD C3 grasses for currently available datasets as a function of bundle sheath (BS) development, with high BS% area (red) compared with those with low BS% area (green), with overlapping distributions in both categories shown as brown. Data (Hattersley 1984; Supporting Information Table S1; see Methodology) were defined by BS % area, respectively, if less than 16% BS (Low) or more than 19% BS (High). Occurrence data were sourced from Global Biodiversity Information Facility data portal and plotted globally using Q-GIS (see Methodology).

The derivation of environmental preferences compared in Fig. 5 (MAT, MAP, annual PET and the AI) is described above (Hijmans et al. 2005; Trabucco & Zomer 2009). The interquartile range and median MAT and MAP data for C4 PACMAD and high or low percentage BS groups in the C3 PACMAD lineages are shown. One drawback to this approach is that such annual climatic data are in practice delimited by growing season conditions, given the eventual advantage of C4 systems to survive dry seasons exposure and promote fire, as discussed above.

Figure 5.

(a) Mean annual precipitation (mm year −1) (MAP); (b) mean annual temperature (°C) (MAT) (Hijmans et al. 2005); (c) mean annual potential evaporative transpiration (mm year −1) (PET); and (d) aridity index (MAP PET −1) (AI) (Trabucco & Zomer, 2009), in habitats occupied by PACMAD C3 high bundle sheath (BS)% area, PACMAD C3 low BS% area and C4 taxa. Data selection for PACMAD C3 taxa: as in Fig. 4. Data for C4 taxa: Hattersley, 1984; Supporting Information Table S1. MAT, MAP, PET and AI data were averaged for each species and presented as box plots; C4 BS%, n = 44; C3 High BS%, n = 9; C3 low BS%, n = 9. Significance (P < 0.05) indicated by letters (A, B) using Dunnetts modified Tukey–Kramer pairwise multiple comparison test. [Correction added on 04 December 2012, after first online publication: Numerical figures in the legend have been amended.]

The higher BS:MC ratio, C3 PACMAD C3 taxa, are significantly distributed in warmer (median 22 °C) habitats than full C4 members of the PACMAD clade (median 18 °C), and C3 PACMAD taxa with lower BS:MC ratios intermediate (Fig. 5a). Both C3 groups have a significant preference for higher mean annual rainfall areas (Fig. 5b), as compared with the C4 lineage, where the median is close to that reported earlier from a similar analysis (1200 mm year −1: Edwards & Smith 2010). However, the high BS development C3 group are found across a wide range of annual mean precipitation, with the lower quartiles encompassing that of the C4 group (Fig. 5b).

When temperature and precipitation are combined to determine PET (Fig. 5c), the tolerance of significantly higher PET by the high BS C3 PACMAD group, relative to the C4 group, suggests PET was an environmental driver which led to enhanced BS development. Finally, the AI, which integrates precipitation and PET, reveals a striking correlation, whereby there is a significant difference between the low BS C3 group and the C4 group, and systematic progression from low BS development to the full C4 pathway (Fig. 5d). Granted the limitation of using annual climatic data, the evidence that low BS development is also important for more temperate species, suggests a dual role for the BS under both semi-arid and cooler conditions.

This preliminary investigation of environmental preferences for the higher BS:MC ratio C3 PACMAD taxa suggests that enhanced BS development provides advantages to the C3 PACMAD grasses in hot climates where there is a high evaporative demand. Environmental preferences of the higher BS:MC ratio C3 PACMAD lineages have been treated as synonymous with those environmental conditions which promoted the enhanced BS development. The advantages or functions of the BS could then be interpreted as the cause for the difference between the environmental preferences of higher and lower BS:MC ratio C3 PACMAD taxa.

It seems that BS development in C3 PACMAD grasses is consistent with tolerance of higher MAT and MAP than full C4 PACMAD lineages (Fig. 5a,b). The median values in the C3 and C4 box plots are entirely consistent with the BS initially providing a selective advantage under conditions of higher evaporative demand. These observations are consistent with the wide range of climatic conditions under which savannah habitats developed (Lehmann et al. 2011; Ratnam et al. 2011). C4 grasses need a relatively high annual rainfall (1200 mm year −1: Edwards & Smith 2010) to maintain a relatively high hydraulic conductance overall (Osborne & Sack 2012), and can be quite susceptible to drought (Taylor et al. 2011). However, the preferred distribution of C4 within low AI areas (Fig. 5d) is consistent with additional environmental tolerances being conferred by the full C4 pathway expression.

The C3 PACMAD group tolerance of high PET, in both high and low BS:MC ratio groups, supports the contention that the BS initially had an important role in water relations. The function as a hydraulic conductance control centre (Sack & Holbrook, 2006) could be one major driver for adoption and development of the BS. Similarly, the BS may provide some function in cold-induced cavitation repair, given the enhanced development in of the smaller BS size-class in more temperate provenances (Figs 4 & 5a) with low-temperature preferences (see complete correlation analysis, Supporting Information Fig. S1). Potentially, hydraulic advantages of the BS could optimize water delivery under mesophytic conditions (acting as a ‘smart pipe’: Shatil-Cohen et al. 2011), while assisting cavitation repair dynamically under more water-limited, or low-temperature, conditions (c.f.: Holbrook & Zwieniecki 1999; McCully 1999; Zwieniecki & Holbrook 2009).

Additional anatomical, experimental and modelling data are now required to prove these functions and advantages of the BS in C3 PACMAD grasses, which would confirm any tolerance to exposure and evaporative demand, or the extent of cavitation under low temperatures. Such studies would provide a rationale for spatial segregation in BS:MC anatomy, which provided a platform for development of the C4 condition.

A molecular perspective on the BS

Considerable interest has been stimulated by the need to understand the molecular regulation of the C4 pathway, because of the need to increase crop productivity for the future (Hibberd et al. 2008). In particular, comparisons of the progression from C3 to C4, via C3-C4 intermediates, are particularly relevant and revealing (e.g. Moricandia: Hylton et al. 1988; Cleome: Marshall et al. 2007; Flaveria: Gowik et al. 2004; McKown & Dengler 2007, 2009; Heliotropium: Muhaidat et al. 2011). However, grasses seem largely absent from these detailed comparisons recently, despite the well-documented Panicoid progression (Brown & Hattersley 1989). If we are to manipulate rice, to introduce the C4 pathway (Hibberd et al. 2008; Hibberd & Covshoff 2010; then the focus could usefully include more studies on the C3 to C4 progression in equivalent grass lineages, exploring the origins and development of BS anatomy.

There have been significant advances in understanding how gene expression can be targeted between BS and mesophyll in C4 plants (Gowik et al. 2004; Brown et al. 2011, Gowik et al., 2011; Gowik & Westhoff, 2011; Kajala et al. 2011; Nelson 2011); what are the emerging lessons for vein and BS characteristics? In developmental terms, the proliferation of veins is generally thought to occur in advance of differentiation of surrounding ground (BS) tissue (for review: Nelson 2011). This is somewhat counter to the phylogenetic progression presented above, whereby in relative terms, BS proliferation was more advanced in C3 PACMAD lineages as compared with any reduction in IVD (Fig. 3). However, identifying the molecular drivers of BS expansion and differentiation, as well as those increasing vein density, offers an intriguing target for future research and manipulation. Perhaps the wealth of transcriptomic data which are emerging, usually framed by the developmental progression along a maturing C4 leaf, will contain elements of the genetic signals which increase BS size and differentiation (Gowik et al., 2011). Furthermore, it should be possible to use molecular phylogenies to investigate whether those C4 grasses and Cyperaceae associated with aquatic habitats are secondarily aquatic and derived, relative to C3 ancestors with low IVD.

From a developmental perspective, it has been recognized that veins act as organizing centres for C4 processes (Langdale 2011). A range of effectors for changing IVD and BS development have been identified, including cytokinins, brassinosteroids small RNAs, transcription factors (Nelson 2011). Other candidates for gene expression and development studies are probably the interaction between auxin and ABA, which are compelling for two reasons: firstly, the developmental similarity between root endodermis, vein proliferation and suberization (Nelson 2011); secondly, the likely interactions between vein density and BS in maintaining leaf hydraulic conductance under high evaporative demand, seasonality of water availability and declining CO2 concentrations in the Miocene (see above). In drought-stressed roots, ABA triggers primary root expansion by upregulating expansins in the zone of elongation (Sharp et al. 2004), while inhibiting the auxin stimulation of lateral root development (De Smet et al. 2006).

Also, we note that some of the genes noted as being upregulated in a comparison of the C3 to C4 progression found in Flaveria, include such genes as AVP1 (Gowik et al. 2004; Li et al. 2005; Gowik et al., 2011) which seems to promote root endodermis formation and possibly leaf vein proliferation, although the mechanism seems more complicated than the cell wall acidification, originally envisaged by Wabnik et al. (2011).

As recently summarized by Gowik et al. (2011) there are a number of unknown processes which lead to increased vein density and BS proliferation. Specifically, these interactions seem to involve auxin transport and polarity of the transporter PIN1 expression, whereby preprocambial strand formation and extension increase vein differentiation in a dicot leaf (Scarpella, Barkoulas & Tsiantis 2010). However, the control of BS proliferation is likely also to reflect polar plant growth regulator interactions, such as the auxin interactions seen in rice leaf-vein differentiation (Qi et al., 2008). Ultimately, understanding the control of BS development, as well as that of veins, should be considered a key desiderata to be investigated with the vigour currently being applied to understanding C4 cell-specific targeting processes.


Given the current interest in understanding the molecular regulation and control of C4 pathway components, and relevance for achieving future food security, it is timely to reconsider developmental control over BS anatomy. Firstly, the contrasting roles attributed to the BS from physiological perspective actually suggest that the development of this anatomical condition is rather more important for maintaining overall hydraulic conductance and optimal delivery of water to the mesophyll than had been traditionally realized. Secondly, while the hypothesis that the BS is important for the dynamic repair of cavitated vessels in stems and leaves is yet to be tested experimentally, there is compelling circumstantial evidence for the involvement of BS cells in such processes. Thirdly, the development of enhanced BS:MC ratio, which seems to precede the reduction in IVD for the limited anatomical data currently available, is an interesting observation which is rather at odds with the traditional assertions regarding both the developmental control of vein differentiation, and likely adaptations of vein density as an adaptation to enhanced evaporative demand. Finally, there are intriguing implications for the role of BS within both PACMAD and sister clades (Pooideae, Ehrharteae) for phylogeny and diversification of grasses in both subtropical and temperate areas. In conclusion, the hypothesis that we have set out is relevant for both crop development and manipulation, as well as understanding the origins and diversification of the C4 pathway in the Palaeocene and Neogene: experimental and molecular validation await!


We are grateful for helpful discussions with Drs. Colin Osborne (APS Sheffield), Julian Hibberd and Jill Harrison (Plant Sciences, Cambridge) as well as data handling advice from Dr. Francisco Rodriguez-Sanchez, Dr. Jessica Royles and Beccy Wilebore, and the assistance and support of Maddie Mitchell (Plant Sciences, Cambridge). Drs. David Simpson and Paula Rudall (RBG, Kew) kindly provided access to the Kew Herbarium specimens, and the analyses were undertaken by Lydia Toy as an MRes project, also supervised by Dr. Tim Barraclough, Imperial College.