The Evolutionary Origin of C₄ Photosynthesis in the Grass Subtribe Neurachninae^,

In multiple Neurachne species, organelles and glycine decarboxylase show varying enhancement in the mestome sheath, suggesting that C₂ photosynthesis gradually evolved from ancestral C₃ plants.

Abstract

The Australian grass subtribe Neurachninae contains closely related species that use C₃, C₄, and C₂ photosynthesis. To gain insight into the evolution of C₄ photosynthesis in grasses, we examined leaf gas exchange, anatomy and ultrastructure, and tissue localization of Gly decarboxylase subunit P (GLDP) in nine Neurachninae species. We identified previously unrecognized variation in leaf structure and physiology within Neurachne that represents varying degrees of C₃–C₄ intermediacy in the Neurachninae. These include inverse correlations between the apparent photosynthetic carbon dioxide (CO₂) compensation point in the absence of day respiration (C_*) and chloroplast and mitochondrial investment in the mestome sheath (MS), where CO₂ is concentrated in C₂ and C₄ Neurachne species; width of the MS cells; frequency of plasmodesmata in the MS cell walls adjoining the parenchymatous bundle sheath; and the proportion of leaf GLDP invested in the MS tissue. Less than 12% of the leaf GLDP was allocated to the MS of completely C₃ Neurachninae species with C_* values of 56–61 μmol mol⁻¹, whereas two-thirds of leaf GLDP was in the MS of Neurachne lanigera, which exhibits a newly-identified, partial C₂ phenotype with C_* of 44 μmol mol⁻¹. Increased investment of GLDP in MS tissue of the C₂ species was attributed to more MS mitochondria and less GLDP in mesophyll mitochondria. These results are consistent with a model where C₄ evolution in Neurachninae initially occurred via an increase in organelle and GLDP content in MS cells, which generated a sink for photorespired CO₂ in MS tissues.

C₄ photosynthesis is a complex trait resulting from the evolutionary reorganization of C₃ leaf anatomy and biochemistry to form an efficient carbon dioxide (CO₂) concentrating mechanism. How the C₄ pathway evolved has been a central issue in plant biology since its discovery over 50 years ago (Downton, 1971). In the past 15 years, phylogenetic studies of C₃ and C₄ species enabled testing of hypotheses concerning when, where, and how the C₄ pathway originated (for example, McKown et al., 2005; Christin et al., 2010, 2011a,b; Kadereit et al., 2012, 2014). Together with physiological and biochemical investigations, these studies support the hypothesis of Monson et al. (1984) that the C₄ pathway arose via a series of innovations that initially enabled plants to refix a large fraction of photorespired CO₂ via the shuttling of Gly from the mesophyll (M) to a sheath of cells surrounding the vascular tissue, where the Gly is decarboxylated (for review, see Sage et al., 2014). The sheath layer is most commonly the bundle sheath (BS) cells, although in some grasses the mestome sheath (MS) cells is where decarboxylation of Gly occurs (Khoshravesh et al., 2016). Associated with the shuttling of Gly is a shift in the expression of the mitochondrial enzyme Gly decarboxylase (GDC) from M cells to the vascular sheath tissue. This requires that the Gly produced in photorespiration diffuses from M to sheath cells for metabolism. Gly shuttling, also known as C₂ photosynthesis, elevates CO₂ levels in the BS two to three times that of M cells, thereby improving Rubisco efficiency (von Caemmerer, 2000; Keerberg et al., 2014). Photosynthetic enhancements in C₂ species are greatest at CO₂ levels below the current atmospheric value of 400 µmol mol⁻¹, supporting a hypothesis that C₂ photosynthesis is a low-CO₂ adaptation that bridges the transition from C₃ to C₄ photosynthesis (Hattersley et al., 1986; Bauwe, 2011; Vogan and Sage, 2012). Consistent with this hypothesis, the C₂ condition is present in more than 40 intermediate species that branch between C₃ and C₄ species on the phylogenetic trees of more than a dozen distinct C₄ lineages (Sage et al., 2014; Lundgren and Christin, 2017; Schüssler et al., 2017).

An important early step proposed for C₂ evolution is the physiological activation of the BS of C₃ species (Gowik and Westhoff, 2011; Sage et al., 2014). Activation of the BS is facilitated by increases in the number and size of mitochondria and chloroplasts and repositioning of mitochondria plus a few chloroplasts to the centripetal pole of the BS, possibly to facilitate refixation of some photorespired CO₂ produced in the leaf (Muhaidat et al., 2011; Sage et al., 2013; Voznesenskaya et al., 2013; Khoshravesh et al., 2016). Together, organelle enrichment and repositioning have been termed the “proto-Kranz” phase of early C₄ evolution, and is recognized in close relatives of C₃ and C₂ species in the eudicot genera Flaveria, Heliotropium, and Salsola, and the monocot genus Steinchisma (Muhaidat et al., 2011; Sage et al., 2013; Khoshravesh et al., 2016; Schüssler et al., 2017). These observations support a hypothesis that proto-Kranz is a key early step in C₂ evolution. Transition to the C₂ condition occurs with further increases in mitochondrial size and numbers in the BS, increased chloroplast numbers in the BS, and a repositioning of most chloroplasts to the centripetal wall of BS cells (Brown and Hattersley, 1989; Muhaidat et al., 2011; Sage et al., 2013; Khoshravesh et al., 2016). Increases in vein density and BS cross-sectional area can occur in concert with organelle and GDC enrichment in sheath cells of proto-Kranz and C₂ species, and the size and numbers of M cells often decline as well (McKown and Dengler 2007; Muhaidat et al., 2011; Sage et al., 2013; Khoshravesh et al., 2016). Because metabolite flux between M and sheath cells increases as the C₂ and C₄ pathways evolve, increased intercellular transport capacity between these compartments should also increase. C₄ plants are known to have greater frequencies of plasmodesmata (PD) between M and BS cells than that in related C₃ species (Danila et al., 2016, 2018), but it is not known where along the C₃ to C₄ transition the PD frequency increases.

Although research on C₃–C₄ intermediates provides many insights into C₄ evolution, it is largely dominated by results from a handful of species. Over 80% of known intermediates are eudicot, with Flaveria species (Asteraceae) in particular dominating research on C₄ origins (Monson and Rawsthorne, 2000; Schulze et al., 2013; Sage et al., 2014). By contrast, only a handful of C₃–C₄ intermediate monocots are known, which is surprising given the predominance of monocot taxa among the C₄ flora. Despite accounting for 78% of all C₄ species, only four of 30 C₄ monocot lineages are known to have C₃–C₄ intermediate species (Hattersley and Stone, 1986; Khoshravesh et al., 2016; Lundgren et al., 2016; Sage, 2016). Of these, the grass genera Alloteropsis and Neurachne also contain C₄ species, but have only one recognized intermediate species each; Steinchisma has five or six species exhibiting C₂ photosynthesis, but it is not close to any C₄ clade (Brown et al., 1983; Aliscioni et al., 2003; Grass Phylogeny Working Group II, 2012). This relatively low number of C₃–C₄ intermediates in the monocots limits investigations into C₄ pathway evolution in grasses and sedges, and constrains development of theoretical models of C₄ evolution. For example, the evolutionary model of Williams et al. (2013) predicts C₄ evolved differently in grasses than eudicots; however, the low number of C₃–C₄ intermediates in the grasses limited parameterization of the monocot algorithms in their model.

The above considerations indicate that identification of additional C₃–C₄ intermediates in grasses and sedges would promote understanding of C₄ evolution in the monocots. To this end, C₃–C₄ intermediates in the grass genera Alloteropsis and Homolepis have been recently identified (Khoshravesh et al., 2016; Lundgren et al., 2016). Another approach is to re-examine the grass genus Neurachne and its allies in the subtribe Neurachninae, where work in the 1980s identified one C₂ species (Neurachne minor), two C₄ species (Neurachne munroi, and Neurachne muelleri—formerly Paraneurachne muelleri), and four putative C₃ species (Neurachne alopecuroidea, Neurachne lanigera, Neurachne queenslandica, Neurachne tenuifolia; Hattersley et al., 1982, 1986; Hattersley and Roksandic, 1983; Hattersley and Stone, 1986). Subsequent work discovered an additional C₃ species, Neurachne annularis (Macfarlane, 2007), whereas Thyridolepis, a sister genus in the Neurachninae, contains three putative C₃ species (Prendergast and Hattersley, 1985; Hattersley et al., 1986). This research indicates the potential of the Neurachninae to provide novel insights into C₄ evolution, particularly in the initial transition from a C₃ character state. Work with C₃ relatives of C₄ species describe certain anatomical shifts, such as increases in vein density, that may predispose lineages to evolve C₄ (Muhaidat et al., 2011; Christin et al., 2013; Griffiths et al., 2013; Sage et al., 2013). Similar studies in putative C₃ species of Neurachne could also identify anatomical traits that facilitated C₄ evolution in this lineage (Hattersley et al., 1986; Brown and Hattersley, 1989). The Neurachninae also promises unique insights not available in other C₄ lineages, in that the C₄ Neurachne species exhibit a distinct form of Kranz anatomy termed “Neurachneoid,” where a suberized MS is the site of CO₂ concentration (Dengler and Nelson, 1999; Edwards and Voznesenskaya, 2011). These cells are in turn surrounded by an outer sheath of parenchyma cells considered to have low photosynthetic activity (Hattersley et al., 1982, 1986). The combined resistance of the suberized MS wall and the parenchymatous sheath (PS) could have profound implications for how C₄ evolved in Neurachne, and other grass lineages where CO₂ is concentrated in the MS, such as Alloteropsis (Lundgren et al., 2019). In addition, it is now possible to evaluate evolutionary hypotheses in the Neurachninae with the publication of a species-level molecular phylogeny (Christin et al., 2012). The Neurachninae clade is young (<10 million years old), increasing the likelihood that unknown intermediate traits persist (Christin et al., 2012). Both C₄ Neurachne species use the NADP-malic enzyme subtype of C₄ photosynthesis (Hattersley and Stone, 1986; Moore and Edwards, 1989), which the molecular phylogeny indicates were independently acquired (Christin et al., 2012).

To develop a more comprehensive understanding of the tissue and cellular changes during C₄ grass evolution, this study re-examined the physiological and structural diversity of nine species of the Neurachninae subtribe using leaf gas exchange, leaf anatomical and ultrastructural studies, and immunolocalization of the P-subunit of GDC (GLDP) and Rubisco large subunit (LSU). Structural assessments include quantitative analyses of MS versus M cell dimensions, organelle size and numbers, interveinal distance (IVD), and the relative contribution of MS, PS, and M tissue to the leaf volume. PD frequencies between MS and PS cells and between PS and M cells were also evaluated to assess potential changes in intercellular transport capacities along the evolutionary transition. Gas exchange analysis allowed us to relate modifications in leaf structure to physiological performance, with the parameter C_* (the apparent CO₂ compensation point in the absence of day respiration) being our main index of photorespiratory CO₂ refixation after Gly shuttling (Sage et al., 2013). We used the recent phylogeny of Neurachninae (Christin et al., 2012) to interpret observed trait shifts in an evolutionary context.

RESULTS

Leaf Gas Exchange

At 32°C, the response of net CO₂ assimilation rate (A) to intercellular CO₂ partial pressure (C_i) for N. muelleri and N. munroi was typical of C₄ plants, showing a CO₂ compensation point (Γ) near 0 µmol mol⁻¹, a steep initial slope, and a CO₂ saturation point below a C_i of 400 µmol mol⁻¹ (Fig. 1; Table 1). N. alopecuroidea, N. queenslandica, and Thyridolepis mitchelliana exhibited A/C_i responses typical of C₃ photosynthesis, with Γ > 60 µmol mol⁻¹, a low initial slope relative to the C₄ species, and a CO₂ saturation point exceeding 1,000 μmol m⁻² s⁻¹ (Fig. 1; Table 1; A/C_i curves not shown for T. mitchelliana). We consider these three species “completely” C₃ plants to distinguish them from N. annularis that is functionally C₃ although it exhibits some intermediate structural characteristics. In N. annularis, Γ was C₃-like at 65 μmol mol⁻¹ at 32°C. The C₂ species N. minor exhibited responses to C_i that were intermediate between the C₃ and C₄ responses, with a Γ of 16 μmol mol⁻¹ at 32°C. In N. lanigera, Γ was 56 μmol mol⁻¹ at 32°C, and 58 μmol mol⁻¹ at 35°C; this latter Γ value was significantly less than in the Neurachne species that exhibi C₃-like Γ values at 35°C. Carboxylation efficiency was statistically the same in all non-C₄ species of Neurachninae that were examined at 32°C (Table 1).

Figure 1.

The response of net CO₂ assimilation rate to variation in intercellular CO₂ concentration in seven species of Neurachne at 32°C. Each curve is a single representation of three to five measurements per species. The highest CO₂ assimilation rate for N. alopecuroidea also presents ± se to show a typical se range.

Table 1. Summary of leaf gas exchange results for Neurachninae species.

Letters indicate common statistical groups at P < 0.05 (one-way ANOVA followed by a Tukey’s posthoc test). Mean ± se, n = 3–9 except for N. munroi where n = 2 at 35°C. ND, not determined.

Species	C*, μmol µmol mol−1	Γ at Saturating Light, µmol mol−1		Carboxylation Efficiency μmol m−2 s−1
	35°C	32°C	35°C	32°C	35°C
T. mitchelliana	61.1 ± 0.5 a	ND	62.4 ± 0.5 c,d	ND	0.22 ± 0.02 b
N. alopecuroidea	60.9 ± 2.0 a	71.1 ± 2.3 a	78.5 ± 2.0 a	0.05 ± 0.00 b	0.05 ± 0.01 d
N. queenslandica	55.9 ± 0.7 a	62.8 ± 1.0 a,b	69.2 ± 1.0 b	0.10 ± 0.02 b	0.08 ± 0.00 c,d
N. annularis	54.4 ± 0.7 a	65.0 ± 2.7 a,b	66.9 ± 1.0 b,c	0.09 ± 0.01 b	0.13 ± 0.00 c
N. lanigera	44.1 ± 0.9 b	55.6 ± 0.9 b	58.1 ± 1.7 d	0.13 ± 0.01 b	0.12 ± 0.03 c,d
N. minor	16.7 ± 0.9 c	15.6 ± 4.6 c	14.3 ± 1.2 e	0.11 ± 0.01 b	0.09 ± 0.02 c,d
N. munroi	ND	2.0 ± 0.4d	1.7 ± 0.3 f	0.50 ± 0.13 a	0.51 ± 0.03 a
N. muelleri	ND	0.3 ± 0.0d	ND	0.27 ± 0.02 b	ND

C_* was measured at 35°C to enhance photorespiration and the ability to resolve differences between C₃, proto-Kranz, and C₂ species (Fig. 2; Table 1). In C₄ plants, C_* cannot be measured due to C₄ metabolism, so Γ at high light was measured instead (for convenience in comparing species, C₄ Γ is treated as equivalent to C_* in the discussion of results). As shown in Figure 2, there is good convergence of the initial slopes of the A/C_i responses upon a common intercept at all light intensities (N. alopecuroidea, N. annularis, and N. lanigera), or light intensities at <500 µmol m⁻² s⁻¹ (N. minor). For N. alopecuroidea, N. queenslandica, N. annularis, and N. lanigera, all plots of the y-intercept versus the initial slope of the A/C_i response closely fit a linear response, with the slope of the response being similar for each species (Supplemental Fig. S1). For N. minor, by contrast, only the values <500 μmol m⁻² s⁻¹ fit a linear response; the value at high light did not correspond to the linear response due to a shift in the high-light A/C_i response to lower C_i (Supplemental Fig. S1). Mean C_* values at 35°C for T. mitchelliana, N. alopecuroidea, and N. queenslandica were C₃-like, being 56 μmol mol⁻¹ to 61 μmol mol⁻¹ (Table 1). For N. annularis, C_* was 54 μmol mol⁻¹, which is statistically equivalent to the C₃ group; in N. lanigera, C_* was 44 μmol mol⁻¹, significantly less than the values of C₃ Neurachninae species (Table 1). C_* for N. minor was 17 µmol mol⁻¹, a typical value for C₂ species (Khoshravesh et al., 2016). Because C_* reflects the magnitude of photorespiratory CO₂ refixation (Busch et al., 2013; Sage et al., 2013), the observed C_* values can be used to evaluate relationships between photosynthetic physiology and the underlying leaf structure.

Figure 2.

The response of net CO₂ assimilation rate to variation in intercellular CO₂ concentration (C_i) at <100 μmol CO₂ mol⁻¹ air in four species of Neurachne at 35°C and four to five light intensities. The C_i where the curves intercept is an estimate of C_*, the apparent CO₂ compensation point in the absence of day respiration (arrows). Curves are representative of three to eight A/C_i plots per species. Note the y axis range varies between the graphic representations. PPFD, photosynthetic photon flux density in μmol photons m⁻² s⁻¹.

Leaf and Cellular Architecture

The leaf architecture of Neurachninae species varies dramatically between the C₃ and C₄ phenotypes, in a manner that is consistent with differences between these photosynthetic pathways in many monocot and eudicot clades where the C₄ pathway evolved (Hattersley and Watson, 1975; Hattersley, 1984; Dengler and Nelson, 1999; Christin et al., 2013). Specifically, IVD, the ratio of M to MS tissue area in a cross section, and the M cell number between PSs of minor veins decreases from C₃ to C₄ species (Fig. 3, A–C; Supplemental Table S1). In species with C_* values > 40 μmol mol⁻¹, MS cells have a smaller cross-sectional area (<45 μm⁻²) in comparison to the larger (>100 μm⁻²) MS cells in species where C_* < 20 μmol mol⁻¹ (Fig. 3A; Supplemental Table S1). This increase in cross-sectional area of the MS cells in Neurachne as C_* declines mirrors changes in BS cell area and width (W) in planar cross sections from many clades where C₄ photosynthesis evolved, leading to the general view that C₄ species have more voluminous BS or MS cells relative to their C₃ relatives (Dengler and Nelson, 1999; Christin et al., 2013; Williams et al., 2013). To further evaluate changes in MS size and shape during C₄ evolution, we estimated MS cell volume by multiplying the cross-sectional area times the cell length (L) in a paradermal section, which is possible because MS and BS cells in grasses are cylindrical in shape (Leegood, 2008). Our estimated volume of MS cells does not support the common view that MS cells become larger during C₄ evolution. No increase in MS volume was observed in C₄ relative to C₃ or intermediate Neurachninae species and there is no relationship between MS cell volume and C_* (Fig. 3B; Supplemental Table S1). This is because MS cells decrease in L (the dimension along the proximo-distal axis parallel to the veins in grass leaves) and increase in W (the dimension along the medio-lateral axis perpendicular to grass veins) as C_* declines (Fig. 3, C–F). The MS cells in the C₃ species such as N. alopecuroidea are long and relatively thin (high L:W), whereas in C₄ species such as N. munroi, they are short and wide as shown by changes in L:W that decline from near 15 in C₃ species to <5 in C₄ species (Fig. 3, D–F; Supplemental Table S1). N. lanigera and N. minor have intermediate L:W, reflecting a transition from the C₃- to C₄-type MS cell. A second-order regression predicted 98% of the variation in this relationship (Fig. 3C). These morphological changes result in an increase in MS cell number along the length of the vascular strands (Fig. 3, D–F).

Figure 3.

The relationship between C_* and dimensions of MS cells. A, MS cell area in planar cross section. B, MS cell volume. C, MS cell L:W in a paradermal section in eight species of Neurachnineae. Values were determined from planar cross sections of TEM images (A and B). Values were determined from light microscopic images of planar paradermal sections (C). Mean ± se; n = 3 plants per species; 10 images per plant. Significant regressions at P ≤ 0.05 are shown. N. alopecuroidea (◑), N. annularis (▽), N. lanigera (◇), N. minor (○), N. munroi (⬜), N. muelleri (☆), N. queenslandica (△), and T. mitchelliana (◪). See Supplemental Table S1 for values and statistical tests. D to F, Paradermal view of vascular sheath cells. Nann, N. annularis; Nmin, N. minor; Nmun, N. munroi. Single MS cells are outlined (dashed line). Scale bars = 50 μm; *, MS, M, and PS.

Increased MS cell width, as well as reductions in interveinal M cell number, contribute to modifications in the percent of leaf tissue occupied by MS and M tissue in C₃ compared to C₂ and in turn C₄ species (Fig. 4, Supplemental Table S1; Supplemental Fig. S2). Whereas <10% of the cross-sectional area of photosynthetic tissue is occupied by MS in the C₃ species with C_* > 50 μmol mol⁻¹, MS tissue accounts for >20% of the leaf cross-sectional area in species with C_* < 20 μmol mol⁻¹ (Fig. 4A). The reduction in C_* from C₃ to C₂ to C₄ species was also associated with a decline in the percent of M tissue in a cross section, from >80% of the cross-sectional image area at a C_* >60 µmol mol⁻¹ to ∼60% at a C_* <20 µmol mol⁻¹ (Fig. 4C). As M:MS, IVD, and interveinal M cell number decline, so does C_* in a linear manner, with each regression having a significant slope (Fig. 4, D–F). No relationship was observed between percentage of PS tissue area and C_* (Fig. 4B).

Figure 4.

The relationship between leaf structural parameters and C_* in eight Neurachninae species. Values determined from light microscopic images of planar cross sections. Percent tissue values were calculated as the area of a given tissue in a leaf image divided by the sum of the areas of MS, PS, and M tissues in the same image, times 100% (Supplemental Fig. S2). A, Percent of image area composed of MS tissue. B, Percent of image area composed of PS tissue. C, Percent of image area comprised of M tissue. D, The ratio of M tissue area to MS tissue area in planar cross section. E, Mean distance between veins. F, The number of M cells between veins. Mean ± se; n = 3 plants per species; 10 images per plant. See Supplemental Table S1 for mean values and statistical tests. Significant linear regressions at P ≤ 0.05 are shown. N. alopecuroidea (◑), N. annularis (▽), N. lanigera (◇), N. minor (○), N. munroi (⬜), N. muelleri (☆), N. queenslandica (△), and T. mitchelliana (◪).

The allocation of chloroplasts to MS tissue also increases as C_* declines from C₃ to the C₄ values in Neurachninae (Figs. 5 and 6; Supplemental Table S2; Supplemental Fig. S3). In the C₃ species T. mitchelliana, N. queenslandica, N. alopecuroidea, and N. annularis, <5% of the chloroplast cross-sectional area within leaf photosynthetic tissue occurs in the MS (Figs. 5, A and B, and 6A). This fraction rises to >9% and 20% in N. lanigera and N. minor, respectively, and to near 50% in the C₄ species (Figs. 5, C and D, and 6A; Supplemental Table S2). Increases in chloroplast size and number within the MS protoplast contribute to greater chloroplast investment in MS tissue of intermediate and C₄ species (Figs. 5, E and F, and 6A; Table 2). Enhanced mitochondrial numbers and size contribute to greater mitochondrial investment in MS tissue as C_* declines from C₃ to C₂ species (Figs. 5, A–D, and 6B; Table 2; Supplemental Table S2; Supplemental Fig. S3). In N. lanigera and N. minor, >40% and 60%, respectively, of the mitochondrial cross-sectional area within leaf photosynthetic cells occurs in the MS (Figs. 5, C and 5D, and 6B; Supplemental Table S2). In the MS of the C₄ species, mitochondrial investment is lower than in C₂ species, due to a reduction in mitochondrial numbers (Figs. 5, E and F, and 6B; Table 2; Supplemental Table S2).

Figure 5.

TEM images of MS cells of six Neurachne species. A to F, Low magnification images illustrating ultrastructure. G to L, Immunogold labeling of GLDP in MS mitochondria. Immunogold labeling of GLDP is indicated by black dots in mitochondria. Nalo, N. alopecuroidea; Nann, N. annularis; Nlan, N. lanigera; Nmin, N. minor; Nmun, N. munroi; Nmue,N. muelleri; C, chloroplast; *, mitochondrion. Scale bars = 2 μm (A–F); 500 nm (G–L).

Figure 6.

The allocation of organelles and GLDP among leaf tissue types. A, Chloroplast. B, Mitochondria. C, GLDP among M (magenta, left bar), PS (blue, middle bar), and MS (orange, right bar) tissues in eight species of Neurachnineae. Values in parentheses represent mean C_* values for each species. n = 3 plants per species; 10 images per plant. See Supplemental Table S2 for mean values, statistical tests, and quantification procedure.

Table 2. Organelle parameters and gold density labeling of GLDP in tissues of Neurachninae species.

Values are from planar cross sections of TEM images. Mean ± se; n = 3 plants per species; 10 images per plant. Values followed by the same letter in each row are not significantly different (P ≥ 0.05) using a Tukey’s post hoc test for normally distributed data (*) and a Games–Howell test for all other data. See Supplemental Table S3 for peroxisome area per protoplast area. PK, Proto-Kranz phenotype.

Parameter by Tissue Type	Completely C3 Species			C3 PK	Modest C2a	C2 Speciesb	C4 Species
	T. mitchelliana	N. queenslandica	N. alopecuroidea	N. annularis	N. lanigera	N. minor	N. munroi	N. muelleri
Area per chloroplast, μm2
Mestome sheath	0.6 ± 0.2 a	4.1 ± 0.6 a	3.4 ± 0.4 a	3 ± 0.3 a	3.5 ± 0.3 a	8.3 ± 0.8 b	19.2 ± 1 b	8.6 ± 0.5 b
Mesophyll	5.4 ± 0.4 a	8.9 ± 0.7 b	6.4 ± 0.4 a	7.5 ± 0.4 a	6.8 ± 0.4 a	6.7 ± 0.3 a	10.6 ± 0.9 b	6.6 ± 0.4 a
Parenchymatous sheath	3.5 ± 0.5 a,b	4.8 ± 1 b	2.8 ± 0.3 a	2.9 ± 0.2 a	2.3 ± 0.3 a	3.7 ± 0.3 a,b	7.7 ± 0.4 c	3.4 ± 0.2 a,b
Area per mitochondrion, μm2
Mestome sheath	0.08 ± 0.02 a	0.21 ± 0.04 a	0.22 ± 0.02 a	0.3 ± 0.03 b	0.3 ± 0.02 b	0.36 ± 0.02 b	0.42 ± 0.02 c	0.32 ± 0.02 b
Mesophyll	0.24 ± 0.06 a,b	0.31 ± 0.03 b	0.21 ± 0.03 a	0.28 ± 0.02 a,b	0.25 ± 0.02 a,b	0.21 ± 0.02 a	0.37 ± 0.03 c	0.25 ± 0.02 a,b
Parenchymatous sheath	0.15 ± 0.02 a	0.21 ± 0.02 a	0.15 ± 0.01 a	0.16 ± 0.01 a	0.18 ± 0.01 a	0.21 ± 0.03 a	0.37 ± 0.05 b	0.21 ± 0.03 a
Chloroplast area/protoplast area, %
Mestome sheath	1.4 ± 0.4 a	12.9 ± 2.7 b	12.2 ± 2.9 b	16.4 ± 1.9 b	27.5 ± 2.8 c	46.5 ± 3.3 d	74.2 ± 2.4 e	38.7 ± 1.7 d
Mesophyll	15.7 ± 1.2 a,b	21.6 ± 1.9 a,b	38.2 ± 2.4 d	33.4 ± 2.2 c,d	24.4 ± 1.9 b,c	21.4 ± 1.8 a,b	18.8 ± 1.9 a,b	11.4 ± 1.0 a
Parenchymatous sheath	4.7 ± 0.8 a,b	16.3 ± 3.8 c	8.0 ± 1.0 c	4.2 ± 0.5 a	4.7 ± 0.6 a,b	6.5 ± 0.7 a,b,c	11.0 ± 1.0 c	7.3 ± 0.8 b,c
Chloroplast #/protoplast area, μm−2 × 10−3
Mestome sheath	20.0 ± 4.7 a	32.4 ± 7.2 a,b	34.8 ± 7.5 a,b	56.3 ± 5.9 b,c	84.0 ± 8.7 c	62.6 ± 4.9 b,c	41.9 ± 2.6 b	47.8 ± 2.3 b
Mesophyll*	29.9 ± 2.2 c	26.8 ± 1.8 c,b	58.4 ± 3.3 a	45.3 ± 2.6 b	36.7 ± 3.2 c	30.7 ± 1.9 c	19.0 ± 1.6 d,e	17.5 ± 1.1 e
Parenchymatous sheath	13.1 ± 1.6 a	33.3 ± 5.5 c	27.5 ± 1.7 c	14.9 ± 1.4 a,b	20.3 ± 2.0 a,b,c	18.6 ± 1.6 a,b,c	15.7 ± 1.8 a,b	21.4 ± 2.0 b,c
Mitochondrion area/protoplast area, %
Mestome sheath	0.5 ± 0.1 a	1.4 ± 0.4 a,b	3.4 ± 0.7 b	8.2 ± 0.8 c	10.6 ± 1.1 c	9.1 ± 0.8 c	2.2 ± 0.2 b	1.6 ± 0.1 b
Mesophyll	0.4 ± 0. 1 a	0.8 ± 0.1 b,c	1.6 ± 0.2 d,e	1.8 ± 0.2 e	1.2 ± 0.2 c,d	0.6 ± 0.1 a,b,c	1.0 ± 0.1 c,d	0.5 ± 0.1 a,b
Parenchymatous sheath	0.2 ± 0.0 a	1.0 ± 0.2 b,c,d	0.9 ± 0.1 b,c	0.5 ± 0.1 b	1.0 ± 0.1 c,d	0.7 ± 0.1 b	1.4 ± 0.2 d	0.8 ± 0.1 b,c
Mitochondrion #/protoplast area, μm−2 × 10−3
Mestome sheath	47 ± 24 a	100 ± 25 b,c	164 ± 24 c,d	313 ± 23 e,f	389 ± 24 f	246 ± 24 d,e	53 ± 24 a,b	53 ± 24 a,b
Mesophyll	1.8 ± 0.2 a	2.9 ± 0.3 a,b	7.0 ± 0.7 c	6.8 ± 0.6 c	5.2 ± 0.8 b,c	2.4 ± 0.3 a	2.8 ± 0.3 a,b	2.2 ± 0.2 a
Parenchymatous sheath	1.2 ± 0.2 a	4.9 ± 0.8 b,c	5.6 ± 0.6 b,c	3.4 ± 0.3 c	6.0 ± 0.6 b	3.5 ± 0.4 c	4.3 ± 0.5 b,c	3.5 ± 0.5 c
Gold particles/mitochondrion area, µm−2
Mestome sheath	7.0 ± 3.5 a	47.3 ± 3.7 b,c	28.7 ± 3.5 b	37.4 ± 3.4 b,c	42.3 ± 3.5 bc	49.8 ± 3.5 c	23.8 ± 3.5 b	28.1 ± 3.5 b
Mesophyll	86.4 ± 5.8 a	53.1 ± 3.5 a	44.1 ± 2.1 a,b	46.1 ± 2.4 a	20.3 ± 1.7 b	2.2 ± 0.3 c	1.5 ± 0.3 c	2.0 ± 0.3 c
Parenchymatous sheath	53.9 ± 4.7 a	25.4 ± 2.4 a	16.2 ± 1.7 a	18.5 ± 1.3 a	4.0 ± 0.6 b	2.5 ± 0.4 b	1.3 ± 0.3 b	3.3 ± 0.7 b
Gold particles/protoplast area, μm−2
Mestome sheath	0.07 ± 0.28 a	0.84 ± 0.29 b,c	1.45 ± 0.28 c	2.92 ± 0.27 d	4.42 ± 0.28 d	4.39 ± 0.27 d	0.50 ± 0.27 b	0.44 ± 0.28 b
Mesophyll	0.32 ± 0.04 a,b	0.42 ± 0.06 a,b	0.76 ± 0.12 a,b	0.83 ± 0.09 a	0.29 ± 0.06 b	0.01 ± 0.00c	0.01 ± 0.00 c	0.01 ± 0.00 c
Parenchymatous sheath	0.13 ± 0.02a	0.19 ± 0.03 a	0.15 ± 0.02 a	0.10 ± 0.01 a	0.04 ± 0.01 b	0.02 ± 0.00 b	0.02 ± 0.00 b	0.02 ± 0.00 b

^aIncomplete C₂ phenotype.

^bComplete C₂ phenotype, where all GDC is in the MS compartment.

Immunodetection of GLDP and Rubisco LSU

To quantify GLDP in tissues of the Neurachninae species, a polyclonal antiserum was generated against a highly conserved 16-amino acid-peptide present in eudicot and monocot GLDP isoforms. The specificity of the antiserum was determined by immunoblot analysis of Neurachne leaf extracts after denaturing PAGE (Supplemental Fig. S4A). A single polypeptide band of ∼98 kD corresponding to the molecular mass of GLDP was detected in all Neurachne leaf extracts (Supplemental Fig. S4B). The labeling shows lower content of GLDP protein, on a total protein basis, in leaves of the C₄ species N. munroi and N. muelleri relative to that in leaves of N. annularis, N. queenslandica, N. lanigera, and N. minor (Supplemental Fig. S4B). Less intense GLDP labeling in N. alopecuroidea extracts is due to protein degradation, which is inherently high in this species compared with other Neurachninae even in the presence of protease inhibitors, and is evident in the Coomassie-stained gel (Supplemental Fig. S4A). The relative abundance of transcripts encoding GLDP was generally greater in the C₃ and C₂ species than in the C₄ species (Supplemental Fig. S4D), reflecting the relative patterns of GLDP protein in the leaves of the various species (Supplemental Fig. S4B).

When the anti-GLDP serum was used to label sections of Neurachninae leaves, the density of GLDP labeling per protoplast planar area increased in MS cells as C_* declined from near 60 μmol mol⁻¹ in completely C₃ species to the 44 μmol mol⁻¹ observed in N. lanigera (Table 2). The density of GLDP labeling per protoplast area in the MS cells of N. annularis was double that of N. alopecuroidea, which had the highest density of GLDP gold labeling per protoplast area in the MS cells of the three completely C₃ species (2.92 versus 1.45 gold particles μm⁻²; Table 2). The density of GLDP labeling per protoplast area in the MS cells was almost three times higher in N. lanigera and N. minor than the highest completely C₃ mean (4.4 versus 1.45 particles μm⁻²; Table 2). These differences largely reflected the increase in mitochondrial area in the protoplast of MS cells of N. annularis, N. lanigera, and N. minor relative to the completely C₃ species, because the amount of GLDP labeling per mitochondrion area did not consistently differ between the Neurachninae photosynthetic types (Fig. 5; Table 2). GLDP labeling per planar area of MS protoplasts was low in the C₄ species relative to N. lanigera and N. minor, which reflects both low mitochondrial area in the C₄ MS and reduced GLDP labeling per mitochondrion (Fig. 5, C–F and I–L; Table 2). Low mitochondrial area in the MS or BS cells is commonly observed in NADP-ME species, which include N. muelleri and N. munroi (Hattersley et al., 1986; Sage et al., 2014). T. mitchelliana exhibited very low amounts of GLDP labeling in the mitochondria of MS cells, which reflects low mitochondrial cross-sectional area and low GLDP content per mitochondrion (Table 2). Along with low values of chloroplast area in the MS cells, these results indicate the MS cells of C₃ T. mitchelliana are photosynthetically inactive compared to those of the other completely C₃ Neurachne species, which have higher organelle content and GLDP levels. In contrast to the patterns observed in the MS cells of the completely C₃ species, the density of GLDP labeling per protoplast and mitochondrial planar area declined in M cells as C_* declined, such that GLDP labeling of the M cells is negligible in the C₄ species and N. minor (Table 2; Supplemental Fig. S5). The near-absence of GLDP labeling in M cells of N. minor corresponds to its robust C₂ cycle as indicated by the low C_* of 17 μmol mol⁻¹ at 35°C. GLDP content per protoplast planar area also declined with C_* in the PS cells (Fig. 6; Table 2). The increase in GLDP labeling in the MS cells coupled with its reduction in the M and PS cells results in an increase in the fraction of leaf GLDP labeling present in the MS tissue, and this correlates with a reduction in C_* in the Neurachninae (Fig. 6C; Supplemental Table S2). The quantification of immunogold particles indicated 100% of the leaf GLDP was present in the MS tissue of N. minor, whereas 67% and 22% of leaf GLDP was present in the MS tissue in N. lanigera and N. annularis, respectively. In the MS tissue of the completely C₃ species N. alopecuroidea, N. queenslandica, and T. mitchelliana, the fraction of GLDP labeling was <22%. In the two C₄ species, all of the leaf GLDP labeling detected was in MS tissue (Fig. 6C; Supplemental Table S2).

Strong labeling of an ∼56-kD polypeptide, corresponding to the LSU of Rubisco, was detected in all Neurachninae species when blots of total leaf protein were labeled with an anti-Rubisco LSU antiserum (Supplemental Fig. S4C). Immunolabeling of the Rubisco LSU was detected in chloroplasts of MS and PS cells in all species, and in M cells of all species except the C₄ N. munroi and N. muelleri (Supplemental Figs. S6–S8). Rubisco LSU immunolabeling in the PS cells was very low in N. annularis, N. lanigera, N. minor, and the two C₄ species. The absence of Rubisco LSU immunolabeling in M cells of N. munroi and N. muelleri reflects their use of the C₄ pathway.

PD Frequency

Because an increase in the frequency of PD between M and BS cells is characteristic of C₄ relative to C₃ species (Danila et al., 2016, 2018), we examined PD frequency in the MS, PS, and M walls of the different Neurachninae species to evaluate the hypothesis that PD frequency increases in C₂ species to accommodate the increased flux of photorespiratory metabolites among M, PS, and MS cells. The PD frequency at the MS–PS cell wall boundary significantly increases in Neurachninae species when C_* declines from >60 μmol mol⁻¹ to <15 μmol mol⁻¹ (Fig. 7A; Supplemental Table S3). The regression between PD frequency in the MS–PS wall and C_* predicts 98% of the variation in the relationship. A significant inverse relationship was also present between PD frequency in the PS–M wall and C_*, with an R² of 0.74 (Fig. 7B; Supplemental Table S3). When species means were pooled by photosynthetic type, the C₄ mean for PD frequency in the MS–MS cell wall was greater than the C₃ mean (P ≤ 0.05).

Figure 7.

The relationship between C_* and PD frequencies in walls between leaf tissue types. A, The frequency of PD between MS and PS cells. B, The frequency of PD between PS and M cells. Mean ± se; n = 3 plants per species; 10 images per plant. Significant linear regressions at P ≤ 0.05 are shown. N. alopecuroidea (◑), N. annularis (▽), N. lanigera (◇), N. minor (○), N. munroi (⬜), N. muelleri (☆), N. queenslandica (△), and T. mitchelliana (◪). See Supplemental Table S3 for mean values and statistical tests.

Phylogenetic Principal Component Analysis

Of the many anatomical, cellular, and GLDP traits examined (Table 2; Supplemental Tables S1–S3), 13 are significantly correlated with C_* (Supplemental Table S4). These 13 traits include IVD, MS L:W, the allocation of mitochondria, chloroplasts, and GLDP labeling in MS versus M tissue, as well as PD frequency between MS–PS, MS–MS, and PS–M cells. The pair comparison of phylogenetic independent contrasts between these 13 variables exhibited similar significance of correlation (Supplemental Fig. S9). To correct for phylogenetic signal, a phylogenetic principal component analysis (pPCA) was performed using the traits that exhibited high correlation in species values and the corresponding phylogenetic contrasts for C_* values (Fig. 8; Revell, 2009, 2012). Five of seven principal components identified by pPCA had eigenvalues > 0.1 (Supplemental Table S5), and pPC1 and pPC2 explained 79% and 12% of variance, respectively (Fig. 8; Supplemental Table S5). In the pPCA plot, the values for N. annularis place it near the completely C₃ species on the right-hand side of the pPC1 axis, but displaced toward the C₄ species along pPC1. C₄ species cluster on the left side of the pPC1 axis, whereas values for N. lanigera and N. minor are spaced between the C₃ and C₄ values along the pPC1 but shifted upwards along pPC2 axis relative to the C₃ and C₄ species (Fig. 8). pPC1 is negatively correlated with parameters associated with the MS tissue, and positively correlated with M tissue-related parameters. Species with C₃, incomplete C₂, complete C₂, and C₄ photosynthesis spread from right to left, respectively, along pPC1 in response to reductions in chloroplast, mitochondria, and GLDP investment in M tissue, and to increases in chloroplast, mitochondria, and GLDP in MS tissue (Fig. 8; Supplemental Table S6). pPC2 is negatively correlated with M tissue mitochondrial number and positively correlated with MS tissue mitochondrial number. The correlation of pPC2 with the remaining variables is negligible (<0.3). A linear model with C_* as the response of the five first pPCs indicates that pPC1 significantly predicts the C_* values (adjusted R² = 0.99, P = 0.004). This is consistent with significant regressions between individual variables and C_* in Figs. 3, 4, and 7.

Figure 8.

pPCA for eight species of Neurachnineae and correlation of 13 quantified variables with the first and second principal component. Data points indicate mean values for each species. Capital letters indicate cell or tissue type: M, PS, and MS. Small letters indicate measured variables as follows: chl, pecent chloroplast allocated to the specified tissue type; mit, percent mitochondria allocated to the specified tissue type; ivd, interveinal distance; g, GLDP density. See Supplemental Tables S1–S3 for mean values and allocation calculations.

Trait Reconstruction

The Neurachninae is nested within a C₃ clade in the Paniceae tribe (Grass Phylogeny Working Group II, 2012), and trait reconstructions support a hypothesis that the C₃ state represents the ancestral condition in the Neurachninae, with two origins of C₄ arising from an intermediate state (Fig. 9; Supplemental Table S7; Christin et al., 2012; Grass Phylogeny Working Group II, 2012). Ancestral state reconstruction of physiological (C_{* or} Γ), cellular (percent GDC in MS), and anatomical (MS L:W) traits using the Neurachninae phylogeny of Christin et al. (2012) consistently show intermediate trait values along the internal branches of the phylogeny, as indicated in Figure 9 by darker (for C₄) versus lighter (for C₃) shading on the branches (see also Supplemental Table S7). The 95% confidence interval (CI) for C_* and numerous other traits overlaps with the CI observed for C₃ species at nodes 1, 7, or 8 only, supporting a hypothesis that the ancestral state of the internal nodes 2–6 was C₂-like or proto-Kranz (Fig. 9; Supplemental Table S7). Although the trait reconstruction indicates the backbone of the Neurachne phylogeny corresponds to a C₂-like condition. Alternatively, the backbone of the tree could be predominantly C₃, with multiple origins of C₂ and C₄ species. Resolution of these alternative models may require discovery of additional populations and/or species of Neurachne, a possibility suggested by the relatively recent discovery of N. annularis by Macfarlane (2007).

Figure 9.

The molecular phylogeny of the Neurachninae (from Christin et al., 2012) with three traits mapped onto the tree. A, Values for C_*. B, % of GDC in the MS tissue. C, L:W of MS cells in paradermal section. No data were available for N. tenuifolia, as indicated by the dotted branch. Nlan, N. lanigera; Nalo, N. alopecuroidea; Nque, N. queenslandica; Nmue, N. muelleri; Nmin, N. minor; Nten, N. tenuifolia; Nmun, N. munroi; Nnann, N. annularis; Tmit, T. mitchelliana. Ninety-five percent confidence intervals for C_* and 12 other traits at ancestral nodes are presented in Supplemental Table S7.

DISCUSSION

In this work, we present a comprehensive analysis of changes in structure and enzyme distribution associated with the C₃ to C₄ transition in the grass subtribe Neurachninae. We demonstrate correlated shifts in multiple structural traits of leaves, and in the density and location of GLDP and Rubisco LSU in the observed C₃, C₂, and C₄ species of the Neurachninae. These variations are correlated with C_* and Γ, implying that simultaneous evolutionary modifications occurred to first assemble a C₂ pathway and then upregulate the C₄ pathway during the evolution of the two known C₄ clades in the Neurachninae. A key question is whether the phenotypes in the extant species represent the evolutionary phases that occurred in the transition from C₃ to C₄ photosynthesis in Neurachne, as opposed to being distinct nonintermediate states, for example, as may arise by reversion or hybridization events (Kadereit et al., 2017). In other clades of C₄ origin, phylogenetic data supports the hypothesis that intermediate phenotypes represent bona fide evolutionary states, as indicated by correspondence between phylogenetic and phenotypic gradients (McKown et al., 2005; Christin et al., 2011b; Yang and Berry, 2011; Sage et al., 2013; Fisher et al., 2015; Lundgren et al., 2016; Schüssler et al., 2017; Dunning et al., 2019). The Neurachninae phylogeny does not clearly follow these patterns, however. Hypotheses supported by the Neurachninae phylogeny of Christin et al. (2012; see also Fig. 9) include the following: (1) C₃ is the ancestral condition in the Neurachninae as indicated by the phylogenetic sister position of C₃ Thyridolepis and the nesting of the Neurachninae in a C₃ clade by the Grass Phylogeny Working Group (GPWGII, 2012). (2) In Neurachne itself, N. annularis branches at the base of the genus, allowing us to infer the proto-Kranz state described here is the ancestral condition within Neurachne. (3) N. minor branches in a sister position to the two C₄ species, supporting a hypothesis that it shares an intermediate common ancestor with the C₄ species. (4) A phylogenetic reconstruction within Neurachne indicates the ancestral state along the backbone of the genus tree is an intermediate condition that could have given rise to C₃ revertants, fully developed C₂ species, and multiple origins of C₄ species (Fig. 9).

The position of N. lanigera in the tree is intriguing because it is nested within a C₃ clade that is sister to the C₄ N. muelleri. This raises a number of possibilities. First, N. lanigera, N. alopecuroidea, and N. queenslandica may have been derived via reversion from an N. muelleri-like C₄ ancestor. This is unlikely because unequivocal evidence of reversion from a C₄ state has yet to be presented, and revertants may have poor fitness because they may not readily reacquire a functional C₃ pathway in the M tissue (Christin et al., 2010; Oakley et al., 2014; Bräutigam and Gowik, 2016). In addition, Christin et al. (2012) present evidence indicating N. muelleri obtained a C₄-adapted PEPC gene by lateral transfer from an ancestor of N. munroi. If N. lanigera, N. alopecuroidea, or N. queenslandica were revertants, they would be expected to share this introgressed gene, which they do not (Christin et al., 2012). It is also possible that the partial C₂ physiology of N. lanigera is the result of C₃ × C₄ hybridization. Evidence of C₃ × C₄ hybridization in Neurachne was noted by Christin et al. (2012) through incongruence of chloroplast and nuclear genomes in polyploid N. queenslandica. The accession of N. lanigera used in this study is diploid (Christin et al., 2012), implying that it is not a hybrid. The available evidence supports a third interpretation, that N. lanigera ultimately arose de novo from an ancestral C₃ state. Resolution of this issue will require additional populations or species of Neurachne and/or detailed analysis of genomic data, which are underway. For now, the most parsimonious interpretation is that N. lanigera is derived from the C₃ condition by vertical selection, and thus may represent a genuine intermediate state in the evolution of full C₂ photosynthesis. If it is assumed the N. annularis and N. minor are also derived vertically from C₃ ancestors, and that C_* reflects the degree of intermediacy, then a hypothetical model of C₂ evolution can be generated for Neurachne.

The novel physiological and structural phenotypes shown here demonstrate the Neurachninae is more dynamic with respect to C₃–C₄ intermediacy than previously understood. Instead of containing just one C₃–C₄ intermediate species as described in Hattersley et al. (1986), the Neurachninae contains three intermediate species of differing character states. If the assumption that C_* reflects degrees of evolutionary intermediacy is valid, then the results are consistent with a progressive model of C₄ evolution as proposed for eudicot lineages such as Flaveria (Monson and Rawsthorne, 2000; Sage et al., 2013; Schulze et al., 2013; Bräutigam and Gowik, 2016). The structural and immunolocalization data provide some key details of how the C₂ pathway may have been assembled in Neurachne, and in the process highlight similarities with Flaveria, but also differences associated with the co-option of the MS versus BS tissue as the compartment of CO₂ concentration. To summarize, N. annularis has a higher organelle investment within the MS compared to the average of completely C₃ species, whereas N. lanigera has greater organelle investment that is intermediate between N. annularis and N. minor. N. annularis shows higher labeling of GLDP per MS protoplast, nearly double that of completely C₃ species, and approaching the high values observed in N. lanigera and N minor; however, N. annularis maintains a C₃-like pattern of GLDP expression in the M tissue. GLDP labeling in the M tissue of N. lanigera is half that of N. annularis, and is negligible in N. minor, typical of other well-developed C₂ species (Khoshravesh et al., 2016). N. annularis also exhibits initial characteristics of Kranz anatomy, as indicated by a 20% to 30% reduction in L:W of MS cells compared to the completely C₃ species. The L:W of the MS cells is even lower in N. lanigera, with values that are 50% of the completely C₃ species values, and in N. minor where L:W average 25% of the completely C₃ values. These changes in physiological and cellular characteristics are associated with lower C_* values: C_* in N. annularis is at the low end of the C₃ range, whereas C_* in N. lanigera falls between the C₃-like value of N. annularis and the completely C₂ value of N. minor. From these results, we conclude N. annularis is largely C₃ in function but shows evidence for an incipient C₂ cycle in the MS tissue, similar to what has been described in BS tissue of proto-Kranz species of Flaveria, Heliotropium, Salsola, and Steinchisma (Muhaidat et al., 2011; Sage et al., 2013; Khoshravesh et al., 2016, Schlüssler et al., 2017). N. annularis differs from proto-Kranz eudicots by not exhibiting a centripetal arrangement of mitochondria, but the thick suberized wall, coupled with the PS layer, indicates the accumulation of photorespiratory CO₂ released by MS mitochondria, thereby improving Rubisco efficiency. N. lanigera shows evidence for a stronger C₂ cycle, but this is not fully optimized for C₂ photosynthesis as it is in N. minor, largely due to the continued presence of GDC in the M cells. We conclude N. lanigera is an incomplete or partial C₂ species (C₂-like) that relies upon the MS mitochondria to metabolize some of the photorespiratory Gly from M tissue. In doing so, C_* is lowered to 44 µmol mol⁻¹ at 35°C. The near complete loss of GLDP expression in the M tissue of N. minor results in the vast majority of photorespiratory Gly being metabolized in the MS tissue. Because of an abundance of mitochondria in the MS tissue, this species has sufficient capacity to metabolize all the photorespiratory Gly coming from the M tissue, and consequently, to reduce C_* to <20 μmol mol⁻¹ as is often observed in completely C₂ species (Monson and Rawsthorne, 2000). The pronounced anatomical shifts in N. minor relative to C₃ species, including a lower MS cell L:W, greater MS volume, and higher PD frequency, demonstrate extensive cellular elaboration and developmental modification to support the C₂ pathway. N. minor can thus be considered to express a form of Kranz anatomy that is highly specialized for efficient C₂ function.

PD frequency through BS–M cell walls is known to be greater in C₄ than C₃ species (Hattersley et al., 1986; Danila et al., 2016, 2018). Data from this study likewise show more frequent PD through MS–PS cell walls, as well as through the PS–M and MS–MS cell walls of the C₄ Neurachninae. Moreover, results demonstrated there is a significant relationship between C_* and PD frequency through the MS–PS cell walls, and C_* and PD frequency through PS–M walls. Assuming C_* reflects the degree of C₃–C₄ intermediacy, these results are consistent with a model whereby the increase in PD is progressive across the C₃–C₄ transition in the Neurachninae. Because enhancement of PD frequency also occurs in proportion to greater organelle and GLDP labeling in the MS tissue, we conclude that a greater flux of photorespiratory metabolites is linked to formation of more PD. PD enhancement in the Neurachninae could be due to increased levels of GOLDEN2-LIKE (GLK) transcription factors, which have been hypothesized to promote activation of BS cells during C₄ evolution (Waters et al., 2009). In rice (Oryza sativa), constitutive expression of GLK genes resulted in greater chloroplast and mitochondrial volume in BS and MS cells that corresponded with more Rubisco and GLDP content as well as higher PD frequency between M and BS cells (Wang et al., 2017), similar to what is observed in N. annularis and N. lanigera.

Declines in M cell number and size contribute to reductions in IVD in C₄ relative to C₃ leaves and are considered essential for optimal C₄ function because fewer M cells reduce the mean diffusion distance for C₄ cycle metabolites (Hattersley, 1984; Dengler and Nelson, 1999; Ogle, 2003; Griffiths et al., 2013). The reduction in IVD in Neurachninae species occurs in proportion to a decline in C_*, and is due to progressively fewer M cells between veins. In the C₄ N. munroi and N. muelleri, these changes converge upon approximately two M cells between the veins, a commonly observed number in most C₄ lineages (Hattersley and Watson, 1975; Dengler and Nelson, 1999; Christin et al., 2013). Notably, reductions in IVD are apparent in N. annularis and N. lanigera compared to the completely C₃ N. alopecuroidea and T. mitchelliana. In N. minor, two M cells are already apparent between veins and IVD approaches C₄ values, indicating that the rapid flux of photorespiratory metabolites between M and MS tissues in this C₂ species requires short diffusion distances that are similar to C₄ values. If so, meeting the rapid transport imperative in C₂ photosynthesis establishes the low transport distance required for efficient C₄ function.

Enhanced BS cell size is commonly thought to be essential for C₄ function by allowing for greater organelle volume, and creating a large vacuole in species lacking suberized lamellae (Dengler and Nelson, 1999; Christin et al., 2013). Large vacuoles can serve as a resistive barrier that facilitates CO₂ trapping in sheath tissues (von Caemmerer and Furbank, 2003). Because it is uncertain where along the C₃–C₄ transition that the size of the Kranz sheath cell increases, we examined changes in MS cell morphology in three dimensions among the various photosynthetic types. In contrast to the common view, we observed no correlation between C_* and MS volume. MS cells of C₄ Neurachne species are not larger (that is, more voluminous) than their C₃ counterparts, although they are wider and shorter. The MS cell dimensions changed such that the L:W was ∼80% lower in the C₄ than in the C₃ species. Danila et al. (2018) also noted C₃ and C₄ grasses have similar BS cell volume. Based on results from the Neurachninae and Danila et al. (2018), we conclude that Kranz anatomy in grasses does not consist of larger BS or MS cells, but only an increased width of BS or MS cells. As indicated by the proportional decline in C_* and the L:W of MS cells, the evolutionary reorientation of the MS cells may have occurred progressively in Neurachne, with N. annularis representing an early stage in this process. The change in MS tissue shape arises from a reorientation in the polarity of cell expansion from a pattern emphasizing proximo-distal (lengthwise, or base to tip in a grass leaf) to medio-lateral (widthwise) expansion, with an increase in cell division along the proximo-distal axis. In doing so, the MS tissue volume (rather than cell volume) increases, creating additional space for organelles and possibly increasing the diffusive distance for CO₂ efflux. The increase in MS cell division along the proximo-distal leaf axis is in contrast to declines in M cell division along the medio-lateral leaf axis, highlighting the contrasting shifts in tissue development from C₃ to C₂ to C₄ phenotypes.

A Model of C₂ Evolution in the Neurachninae

Understanding of the evolutionary origin of C₄ photosynthesis is largely based on eudicot systems, notably Flaveria (Monson and Rawsthorne, 2000; Sage et al., 2013; Schulze et al., 2013; Covshoff et al., 2014). Evolution in monocots is less understood due to the low number of C₃–C₄ intermediates in grasses and sedges that have been identified and studied (Sage et al., 2011a; Lundgren and Christin, 2017). For monocots, this deficiency constrains development of evolutionary landscape models of C₄ origin. Williams et al. (2013) used a Bayesian approach to fill in gaps in the evolutionary sequence of monocots and coupled this to a Markov model to compare evolutionary landscapes in eudicots and monocots. Only two C₃–C₄ intermediate monocots (N. minor and Steinchisma hians) were available to parameterize the monocot responses and both exhibit well-developed C₂ photosynthesis (Williams et al., 2013; Khoshravesh et al., 2016). The model predicted significant divergence in the paths to C₄ photosynthesis, with monocots delaying reallocation of GDC to the sheath cells relative to eudicots. Monocots were also predicted to delay chloroplast accumulation and enhancement of sheath size in comparison with eudicots, while accelerating reductions in vein spacing. Our results do not support these interpretations of Williams et al. (2013). The Neurachne data, along with recent results from Flaveria (Sage et al., 2013; Schulze et al., 2013), Heliotropium (Muhaidat et al., 2011), and Steinchisma (Brown et al., 1983; Khoshravesh et al., 2016) show similar patterns in organelle enhancement and GDC reallocation in grasses and eudicots. Instead of multiple distinct evolutionary pathways between monocots and eudicots, as concluded by Williams et al. (2013), the pattern we interpret is one of similarities between monocots and eudicots in terms of first acquiring the C₂ pathway via organelle and GDC enrichment in the sheath cells, concurrent with reductions in IVD. This new information, along with recent data from Alloteropsis semialata intermediates (Lundgren et al., 2019), will be valuable in updating the Williams et al. (2013) model. Whereas significant differences in the general pattern of C₂ evolution are not obvious, subtle differences are observed between Flaveria and Neurachne, which are two lineages with enough variation between intermediates to evaluate potential patterns of evolutionary change. The nature of the ancestral sheath tissue likely accounts for these variations.

From studies of both eudicots and Neurachne, the initial phase of C₂ evolution appears to be the accumulation of organelles in the sheath tissue (BS in Flaveria, Heliotropium, and Euphorbia, MS in Neurachne), to strengthen, or “activate,” the photosynthetic potential of the sheath tissue (Muhaidat et al., 2011; Sage et al., 2011b, 2013; Schulze et al., 2013). In C₃ Flaveria, Euphorbia, and Heliotropium, this activation occurs in tandem with increased BS width. In C₃ Flaveria cronquistii, chloroplasts and mitochondria line the walls along the intercellular air spaces (IAS), allowing the BS cells to conduct diffusion-based C₃ photosynthesis as occurs in the M cells (Sage et al., 2013). CO₂ rapidly diffuses into the peripheral BS chloroplasts from the IAS while adjacent mitochondria quickly recycle photorespiratory metabolites (Bauwe et al., 2010). The shift to the proto-Kranz phase is observed in Flaveria pringlei and Flaveria robusta, which are branch sisters to F. cronquistii in the Flaveria phylogeny. In these two species, the mitochondria are repositioned along the centripetal wall of the BS cells, opposite most chloroplasts. Similar relocation of mitochondria occurs in the BS of proto-Kranz Heliotropium, Salsola, and Steinchisma species (Sage et al., 2014). This relocation leads Gly produced from glycolate made in the peripheral chloroplasts to be metabolized by centripetal mitochondria, with the result that photorespiratory CO₂ accumulates in the inner region, enhancing Rubisco efficiency in nearby chloroplasts. This arrangement is proposed to establish a single-cell Gly shuttle that can scavenge some photorespiratory CO₂ with slight benefits to the plant (Sage et al., 2013, 2014).

In Neurachne and Thyridolepis, MS cells have suberized outer walls and are surrounded by tightly packed PS cells, both of which can enhance resistance to diffusive efflux of CO₂ generated by decarboxylation of Gly or C₄ acids (Hattersley et al., 1986; von Caemmerer and Furbank, 2003; Alonso-Cantabrana et al., 2018). In the MS of C₃ species, however, suberized walls and the PS sheath would reduce diffusive CO₂ influx and thus starve chloroplasts of CO₂. A boost in chloroplast numbers in the MS of C₃ species would therefore be of little value unless CO₂ were imported from another source, such as photorespiratory Gly. Because N. annularis grows in warm-temperate, semiarid landscapes of western Australia with hot summers (Macfarlane, 2007), high rates of photorespiration would produce an abundance of Gly that could potentially deliver CO₂ to the MS cells. If elevated GDC activity reduces Gly concentration in the MS, a Gly diffusion gradient between M and MS cells would be established, thereby drawing in additional Gly. We hypothesize that the boost in organelles in the MS of N. annularis creates a sink for photorespiratory Gly and can thus increase photosynthetic potential in the MS cells in a manner that slightly enhances leaf photosynthetic efficiency. Because this enhancement results from the operation of a weak photorespiratory Gly shuttle, we propose N. annularis fits the criteria of the proto-Kranz species, even though there is no mitochondrial relocation within the MS cell as observed in eudicots. We hypothesize the suberized walls and PS in N. annularis slow CO₂ efflux, thereby eliminating the need to reposition mitochondria, whereas in Flaveria and other eudicots, trapping of photorespired CO₂ requires a wide vacuole between centripetal mitochondria and the outer periphery of the BS cells to form the barrier to CO₂ efflux.

As indicated by the phenotypes of the C₂-like species N. lanigera and Flaveria angustifolia, once numerous organelles are in place in the sheath tissue and a limited photorespiratory Gly shuttle is functioning, step-wise progression to an optimized C₂ pathway is envisioned where parallel changes occur in (1) increased organelle number in the sheath tissue, (2) greater GDC content in the sheath tissue, (3) reduction of GDC in M cells, (4) increasing PD connections between sheath and M tissue, and (5) increasing Kranz-like anatomy via fewer M cells between veins and declining MS L:W, resulting in increased proportion of BS or MS cell area in cross section (Sage et al., 2013). Because the establishment of a Gly sink in the sheath tissue is beneficial during high photorespiratory episodes, it may stabilize the proto-Kranz character state in hot climates and by doing so become a foundation that enables subsequent evolution that strengthens the photorespiratory Gly shuttle (Sage et al., 2012; Mallmann et al., 2014; Bräutigam and Gowik 2016). As has been proposed for later stages in C₄ evolution (from the C₂ to C₄ condition), this “facilitation” cascade could establish an evolutionary trajectory that resembles a Mount Fuji-type of fitness landscape (Heckmann et al., 2013).

CONCLUSION

The information presented here from the Neurachninae indicates how the structural requirements for C₄ photosynthesis are largely met by the evolution of the C₂ pathway, where MS tissue width, PD frequency, and IVD in N. minor equals or approaches C₄ values. This illustrates how the evolution of C₂ photosynthesis could establish the foundation of structure and metabolite transport upon which a C₄ metabolic cycle may be assembled. We also provide evidence that is consistent with a model whereby the assembly of the C₂ pathway occurs gradually, beginning with modifications in MS cell division and expansion, reduced M cell division, and greater organelle allocation to the MS cells. Assuming C_* reflects the degree of evolutionary intermediacy, the results support a hypothesis that organelle enrichment initiates a progressive reallocation of GDC from M to MS tissue, with loss of GDC expression in M cells occurring after mitochondrial enhancement increases GDC capacity in the MS tissue. Pre-existing anatomy in Neurachne likely eliminates the need for the repositioning of organelles observed in eudicots, because the PS and thick MS wall enable CO₂ trapping without a need for a large vacuole acting as a diffusive barrier. In light of these results, the Neurachninae, along with numerous eudicot groups such as Flaveria, indicates how evolutionary processes reconfigured the variable anatomy, biochemistry, and physiology of C₃ ancestors into a common, complex trait that can address challenges imposed by global environmental change.

MATERIALS AND METHODS

Plant Material

Seeds or rhizomes of Thyridolepis mitchelliana, Neurachne alopecuroidea, Neurachne annularis, Neurachne lanigera, Neurachne minor, Neurachne queenslandica, Neurachne tenuifolia, Neurachne munroi, and Neurachne muelleri were collected in their natural habitat (Supplemental Table S8) and used to generate plants that were grown in naturally illuminated glasshouses at the Plant Growth Facility, University of Western Australia, Perth, Western Australia (latitude 33°59’ south). N. muelleri was classified by Blake (1972) as Paraneurachne muelleri but was previously known as N. muelleri (Hackel, 1895; Lazarides, 1970). As a recent molecular study showed the species branches within the Neurachne phylogeny (Christin et al., 2012), the original name is used here. T. mitchelliana, N. alopecuroidea, N. annularis, and N. tenuifolia were grown at mean temperatures of 22°C/18°C (day/night), whereas N. lanigera, N. minor, N. munroi, and N. muelleri were grown at mean temperatures of 30°C/25°C (day/night). N. queenslandica was grown at day/night temperatures of 28°C/18°C in a separate, sealed greenhouse, as it is not native to Western Australia and thus by regulatory statute required complete isolation from the outside environment. Plants were grown in a mixture of a commercial potting mix and soil collected from west-central Australia, and fertilized with slow release fertilizer for Australian native plants (Osmocote Plus Trace elements: Native Gardens, Scotts Australia). Growth photon flux density routinely reached 1,600 μmol m⁻² s⁻¹. Recently fully expanded leaves of mature plants were used for all measurements. All species were examined for gas exchange, microscopic imaging, and immunolocalization analysis except for N. tenuifolia, which died after preliminary sampling and was unavailable for immunolocalization and gas exchange analyses.

Whole-Leaf Gas Exchange

Leaf gas exchange measurements were performed on healthy, attached leaves using a model no. LI-6400XT gas exchange system (LICOR Biosciences). Responses of net CO₂ assimilation rate (A) to intercellular CO₂ concentration (C_i) were measured at a leaf temperature of 32°C and a light intensity of 1,500 μmol m⁻² s⁻¹. After a leaf was acclimated to these conditions inside the LI-6400XT leaf chamber, the A/C_i curves were measured with a sequence of reference CO₂ concentrations of 400, 300, 200, 150, 100, 75, 50, 400, 400, 600, 800, 1,000, 1250, 1,500, and 2,000 μmol mol⁻¹. Data were recorded in quick sequential order, with an equilibration time of between 90 and 180 s at each CO₂ concentration. The CO₂ compensation point in the absence of day respiration (C_*) was measured using the common-intercept method at 35°C (Brooks and Farquhar, 1985). For the C_* measurements, leaves were acclimated to a leaf temperature of 35°C, a light intensity of 1,000 μmol m⁻² s⁻¹, and a reference CO₂ concentration of 400 μmol mol⁻¹ until A reached steady state. CO₂ responses were then measured at CO₂ concentrations of 400, 120, 110, 100, 90, 80, 70, and 60 μmol mol⁻¹ at five different light intensities of 1,000, 400, 300, 200, and 150 μmol m⁻² s⁻¹. At each light intensity, leaves were given an additional 10 min to acclimate before proceeding with the next CO₂ step change. Typically, lower light intensity reduces the slope of the A versus C_i plot at low C_i, creating a distinct series of linear responses that can be used to estimate C_* by plotting the regression slope of these responses versus their corresponding y-intercept (Walker and Ort, 2015). C_* is then estimated as the slope of this relationship, assuming that M conductance is constant (Supplemental Fig. S1; Walker and Ort, 2015; Walker et al., 2016). Outlier points at high light were excluded from the regression analysis if they resulted from shifts in the A/C_i response to lower C_i, as can occur in C₂ species at >800 μmol photons m⁻² s⁻¹ (Sage et al., 2013). In C₄ plants, the common intercept method does not work for C_* as Γ varies little with declining photosynthetic photon flux density (Table 1); hence, only Γ values were determined in C₄ plants at saturating light.

Leaf Anatomy, Ultrastructure, and Immunolocalization

The internal anatomy of leaves was assessed on sections sampled from the middle of the most recent, fully expanded leaves (three plants per species; one leaf per plant; 10 images per leaf). Plants were sampled between 9 am and 12 am. The youngest cohort of fully expanded leaves was sampled in full sun for all procedures. Samples were fixed in 0.5% (v/v) glutaraldehyde, 4% (w/v) paraformaldehyde in sodium cacodylate buffer for light and transmission electron microscopy (TEM), and embedded in Spurr’s resin and London Resin White as described by Khoshravesh et al. (2017). Immunocytochemical detection of GLDP and Rubisco LSU followed Khoshravesh et al. (2016, 2017) with the same antibodies used for the immunoblots (Supplemental Fig. S4; see Supplemental Table S9 for more details). Light microscope images were obtained with an Axioplan (Zeiss) equipped with an image analysis system (model no. DP71, Olympus; Empix Imaging). A Philips 201 TEM equipped with an Advantage HR camera system (Advanced Microscopy Techniques) was used to capture leaf ultrastructure and immunolabeling results. To quantify all M, PS, and MS cellular features, and immunodetection events (gold particles) of GLDP and Rubisco LSU, TEM images were analyzed using a Cintiq graphics tablet (Wacom Technology) and ImageJ software (Schneider et al., 2012) as described in Sage et al. (2013) and Stata et al. (2014). To quantify protoplast area, we subtracted the cross-sectional area of the cell covered by cell walls from the cell area values. IVD was measured as the distance between PS tissue of minor veins, because the PS tissue is part of the vascular tissue complex. MS cell volume was estimated as the length along the proximo-distal axis parallel to the vein (in paradermal sections) times the area of the MS cell in planar cross section. Allocation coefficients representing the fraction of organelles and GLDP in specific tissues were calculated as: % organelle allocation = (planar area of an organelle within a tissue cross section divided by the sum of organelle area in all of the MS, M, and PS tissue within a sample image) × 100%; and percent GLDP allocation = (the number of immunogold particles within a tissue cross section divided by the total number of immunogold particles in the MS, M, and PS tissue within a sample image) × 100% (Wang et al., 2017). IAS were not included in tissue measurements. PD frequency was quantified as described by Wang et al. (2017); the presence and absence of PD in the walls between M, MS, and PS cells in planar cross sections of TEM images were recorded as absence = 0, presence = 1. Relative frequency of PD was then quantified as the number of cells that had PD connections with the adjacent cell, divided by the total number of observed cells for that cell-type, multiplied by 100%.

Data Analysis

For analysis of light microscopy and TEM images, samples were collected from one leaf of each of three distinct plants. Ten images from each leaf were analyzed with the values averaged to provide a mean for the plant. These individual plant values were the unit of replication for statistical analysis. For leaf gas exchange, three to eight measurements were conducted on two to seven plants per species, with each leaf serving as the unit of replication. Results were analyzed with the software SigmaPlot v12.5 (Systat Software) or SPSS v20.0. (IBM) using one-way ANOVA followed by a Tukey’s test of variance (for normally distributed data) or a Games–Howell test (when the variances were not equal) to evaluate statistical grouping of the species means, unless otherwise indicated.

To analyze whether the anatomical variables and GLDP labeling predict C_* values in the Neurachninae, linear regression analysis was performed. Predictors were selected based on their correlation (Pearson correlation coefficient [r] ≥ |0.65|) with C_*. To correct for phylogenetic signal, phylogenetic independent contrast on the mean value of selected traits was performed using the R package “ape” (Paradis et al., 2004), with evolutionary model = Brownian Motion. The list of predictors was then finalized by selecting traits exhibiting Pearson correlations coefficient (r) ≥ |0.65| with C_* in the contrasting analysis. To eliminate collinearity between selected variables, a phylogenetic pPCA was performed using the phylo.pca function from the R package phytools v0.6-44 (Revell 2009, 2012). Then a linear regression (principal component regression) was performed using pPCs with eigenvalues > 0.1 as the predictors.

To evaluate possible ancestral character states, C_*, percent GLDP in the MS, L:W of MS cell, plus other structural traits were mapped onto the Neurachninae phylogeny of Christin et al. (2012) using the program Mesquite (Maddison and Maddison, 2018) and maximum parsimony (Fig. 9). The 95% confidence intervals for physiological (C_* or for C₄ species, Γ), cellular, and anatomical traits were reconstructed using a maximum likelihood approach and Brownian Motion evolutionary models using R package “ape” (Paradis et al., 2004) and the Neurachninae phylogeny (Christin et al., 2012; Supplemental Table S7).

Supplemental Data

The following supplemental materials are available.

• Supplemental Appendix 1. Supplemental Tables S1-S9.

• Supplemental Table S1. Characteristics of cells and tissues from Neurachninae leaves.

• Supplemental Table S2. The allocation of organelles and GLDP in photosynthetic tissues of eight Neurachninae species.

• Supplemental Table S3. Peroxisome parameters and PD frequency in eight Neurachninae species.

• Supplemental Table S4. Pearson correlations coefficients between measured parameters in this study.

• Supplemental Table S5. Summary of pPCA.

• Supplemental Table S6. Pearson correlations coefficients between traits and phylogenetic principal components.

• Supplemental Table S7. Means and 95% confidence interval for reconstructed trait values at the phylogenetic nodes in Figure 9.

• Supplemental Table S8. Species examined, with voucher information and source location.

• Supplemental Table S9. Details of immunolocalization methods.

• Supplemental Appendix 2. Supplemental Figures S1–S9.

• Supplemental Figure S1. The relationship between the initial slope of the A/Ci response and its extrapolated y-intercept for five species of Neurachne.

• Supplemental Figure S2. Light micrographs of leaf cross sections from nine Neurachninae species.

• Supplemental Figure S3. TEM images of MS cells in N. queenslandica, N. tenuifolia, and T. mitchelliana.

• Supplemental Figure S4. Immunoblot and transcript analyses of Neurachne GLDP and the Rubisco LSU.

• Supplemental Figure S5. TEM images of GLDP immunolocalization in M cells of six Neurachne species.

• Supplemental Figure S6. Immunolocalization of Rubisco LSU in C₃ Neurachninae species.

• Supplemental Figure S7. Immunolocalization of Rubisco LSU in N. annularis, N. lanigera, and N. minor.

• Supplemental Figure S8. Immunolocalization of Rubisco LSU in C₄ Neurachne species.

• Supplemental Figure S9. Phylogenetic independent contrast plots of pair comparisons between 13 anatomical traits selected for pPCA and C_*.