Despite phylogenetic effects, C₃–C₄ lineages bridge the ecological gap to C₄ photosynthesis

Abstract

C₄ photosynthesis is a physiological innovation involving several anatomical and biochemical components that emerged recurrently in flowering plants. This complex trait evolved via a series of physiological intermediates, broadly termed ‘C₃–C₄’, which have been widely studied to understand C₄ origins. While this research program has focused on biochemistry, physiology, and anatomy, the ecology of these intermediates remains largely unexplored. Here, we use global occurrence data and local habitat descriptions to characterize the niches of multiple C₃–C₄ lineages, as well as their close C₃ and C₄ relatives. While C₃–C₄ taxa tend to occur in warm climates, their abiotic niches are spread along other dimensions, making it impossible to define a universal C₃–C₄ niche. Phylogeny-based comparisons suggest that, despite shifts associated with photosynthetic types, the precipitation component of the C₃–C₄ niche is particularly lineage specific, being highly correlated with that of closely related C₃ and C₄ taxa. Our large-scale analyses suggest that C₃–C₄ lineages converged toward warm habitats, which may have facilitated the transition to C₄ photosynthesis, effectively bridging the ecological gap between C₃ and C₄ plants. The intermediates retained some precipitation aspects of their C₃ ancestors’ habitat, and likely transmitted them to their C₄ descendants, contributing to the diversity among C₄ lineages seen today.

Biomes, C₃–C₄ intermediate, C₄ photosynthesis, ecology, evolution, phylogeny

Introduction

The C₄ photosynthetic pathway relies on a coordinated system of anatomical and biochemical traits that function to concentrate CO₂ around Rubisco, which in most C₄ plants is localized to the bundle sheath cells (Hatch, 1987). The enhanced CO₂ concentration substantially suppresses O₂ fixation and subsequent photorespiration, compared with the ancestral C₃ photosynthetic pathway, making C₄ photosynthesis advantageous in conditions that increase photorespiration (Chollet and Ogren, 1975; Hatch and Osmond, 1976). C₄ photosynthesis is consequently prevalent in the open biomes of warm regions where it boosts growth (Sage et al., 1999; Osborne and Freckleton, 2009; Atkinson et al., 2016), to ultimately shape entire ecosystems, such as the emblematic savannas (Sage and Stata, 2015).

It has been widely reported that some plants possess only a subset of the anatomical and/or biochemical components of the C₄ pump. These plants tend to be physiologically somewhere in between typical C₃ and C₄ plants and, as such, are termed C₃–C₄ intermediates (Kennedy and Laetsch, 1974; Monson and Moore, 1989; Sage, 2004; Schlüter and Weber, 2016). These physiologically intermediate plants use a photorespiratory CO₂ pump, or glycine shuttle, to rescue CO₂ released from mesophyll photorespiratory activity and transport it into the bundle sheath for re-use in the Calvin cycle located there (Hylton et al., 1988). Thus, the C₃–C₄ system establishes a CO₂ recycling mechanism based on the spatial segregation of metabolic reactions, the migration of the Calvin cycle to the bundle sheath, and the dual-compartment coordination that are characteristic of the C₄ pathway. These modifications improve the physiological performance of C₃–C₄ plants over the C₃ system in conditions that promote photorespiration, as they lessen the total carbon lost via photorespiration to improve net carbon assimilation (Vogan and Sage, 2011; Way et al., 2014). In addition to the glycine shuttle, a number of C₃–C₄ plants engage a weak C₄ cycle (Ku et al., 1983), which further reduces photorespiration and is predicted to increase biomass accumulation (Mallmann et al., 2014). Thus, this variation in C₄-associated traits forms a continuum between the C₃ condition and a diversity of C₄ phenotypes (Bauwe, 1984; McKown and Dengler, 2007; Lundgren et al., 2014; Bräutigam and Gowik, 2016).

Because C₃–C₄ plants share many anatomical, biochemical, and physiological traits with C₄ plants, they are often assumed to represent an evolutionary step facilitating C₄ evolution (Hylton et al., 1988; Sage, 2004; Sage et al., 2012; Bräutigam and Gowik, 2016), a hypothesis confirmed by the close relationships between C₃–C₄ and C₄ taxa in some groups (McKown et al., 2005; Christin et al., 2011b; Khoshravesh et al., 2012; Sage et al., 2012; Fisher et al., 2015). They are consequently widely studied and incorporated into models of C₄ evolution, which show that C₃–C₄ phenotypes can bridge the gap between C₃ and C₄ states by providing a series of stages that are advantageous over the preceding ones (Heckmann et al., 2013; Williams et al., 2013; Mallmann et al., 2014; Bräutigam and Gowik, 2016). This research program has been extremely successful in tracking the changes in leaf anatomy, organelles, metabolism, genes, and enzymes that likely took place during C₄ evolution, particularly in the eudicot genus Flaveria (e.g. Bauwe and Chollet, 1986; Svensson et al., 2003; McKown and Dengler, 2007, 2009; Sage et al., 2013). However, previous research failed to address the ecological consequences of these intermediate stages. Indeed, while models that predict the carbon gains of the intermediate stages exist (Heckmann et al., 2013; Mallmann et al., 2014), studies of natural distributions of extant C₃–C₄ taxa are nearly non-existent (but see Sudderth et al., 2009).

The differing geographical and environmental distributions of C₃ and C₄ species have been widely studied (Teeri and Stowe, 1976; Rundel, 1980; Williams et al., 1995; Ehleringer et al., 1997, Epstein et al., 1997; Edwards and Still, 2008), with later incorporation of phylogenetic data providing estimates of the ecological shifts that happened before, during, or after photosynthetic transitions (Osborne and Freckleton, 2009; Edwards and Smith, 2010; Edwards and Ogburn, 2012; Kadereit et al., 2012; Lundgren et al., 2015). However, these efforts focused on comparisons between C₃ and C₄ plants, which are much more frequent and abundant than C₃–C₄ taxa. Previous discussions of C₃–C₄ ecology characterized their distributions in hot, sandy, and disturbed habitats with little competition (Powell 1978; Hedge and Patil, 1980; Prendergast and Hattersley, 1985; Vogan et al., 2007; Feodorova et al., 2010; Christin et al., 2011b; Sage et al., 2011, 2012). However, other groups with C₃–C₄ intermediates thrive in apparently very different habitats, with C₃–C₄Flaveria inhabiting a broad range of environments from open fields and scrublands (F. angustifolia) to pine forests (F. anomala), wetlands (F. floridana), and warm mineral springs (F. sonorensis; Powell 1978), yet field data failed to identify differences in the distributions of different photosynthetic types in Flaveria (Sudderth et al., 2009). The monocot C₃–C₄ intermediates of Eleocharis and Steinchisma thrive in wetland habitats (USDA/NRCS, 2016), C₃–C₄Alloteropsis grow in shady, deciduous forests of tropical Africa (Lundgren et al., 2015), and the recently identified intermediates in Homolepis (Khoshravesh et al., 2016) grow at the margins of South American rainforests. These disparate characterizations urge a careful, data based evaluation of the C₃–C₄ niche, its variation among evolutionary lineages, and its relation to that of C₃ and C₄ relatives.

In this study, we use available global occurrence data and local habitat descriptions to characterize the niche of C₃–C₄ lineages, along with their close C₃ and C₄ relatives. The ecological data are used to (i) quantitatively and objectively describe the abiotic habits of C₃–C₄ taxa and determine whether they inhabit uniform conditions, (ii) test whether phylogenetic effects partially explain the ecological sorting of C₃–C₄ lineages and whether their sorting explains the diversity in the ecology of C₄ relatives, and (iii) test whether, when controlling for phylogenetic effects, the C₃–C₄ physiology affects the niche, potentially bringing the plants closer to the C₄ niche. Our large-scale analyses, which consider all described C₃–C₄ lineages and their relatives, show that C₃–C₄ plants inhabit a large array of habitats, and that physiology closely interacts with evolutionary history to shape the niches of C₃–C₄, but also C₄, taxa.

Methods

Ecological distribution of individual C₃–C₄ species

A list of 56 C₃–C₄ intermediate taxa was assembled from the literature, and included 11 eudicot and two monocot families (Table 1). Occurrence data for each taxon were downloaded from the Global Biodiversity Information Facility (GBIF, http://www.gbif.org) using the RGBIF package in R (Chamberlain et al., 2016; data accessed 1 and 2 July 2016). Occurrence data for the Zambezian C₃–C₄ within Alloteropsis semialata were taken from Lundgren et al. (2015, 2016). All occurrence data were cleaned by removing any anomalous latitude or longitude points, points falling outside of a landmass, and any points close to GBIF headquarters in Copenhagen, Denmark, which may result from erroneous geolocation. To avoid repeated occurrences, latitude and longitude decimal degree values were rounded to two decimal places, and any duplicates at this resolution were removed. These filters are commonly applied to data extracted from GBIF (Zanne et al., 2014).

Table 1.

Details of C₃–C₄ taxa used in this study and their local habitats

Comparison	Species	n	Habitat	Referencea
Acanthaceae
Blepharis	B. diversispina	42	Deciduous woodland, grasslands, soil sandy and gravelly, disturbed	Fisher et al., 2015; Hyde et al., 2016a,b; USDA/NRCS, 2016
	B. gigantea	6	Sandy to stony soils
	B. natalensis	6	Rocky slopes
	B. noli-me-tangere	2	Sandy soil, dry watercourses
	B. pruinosa	19	Sandy to stony soils
	B. sinuata	4	Bushland
	B. espinosa	5	Deciduous woodland, disturbed, various habitats
Amaranthaceae
Alternanthera	A. ficoidea	268	Uplands	Rajendrudu et al., 1986
	A. tenella	446
Salsola	S. divaricata	32	Semi-arid rocky zones near coastal areas; salt tolerant	Voznesenskaya et al., 2013
Sedobassia	S. sedoides	3	Ruderal, sandy, saline habitats	Eliáš and Dítě, 2014; Freitag and Kadereit, 2014
Asteraceae
Flaveria	F. pubescens	8	Wetlands, alkaline and saline soils, fine textured soils	Powell, 1978; Edwards and Ku, 1987; USDA/NRCS, 2016
	F. oppositifolia	36
	F. angustifolia	16	Pastures, fields, roadsides, disturbed
	F. anomala	44
	F. chloraefolia	16	Wetlands, saline and gypseous soils, disturbed
	F. floridana	3	Wetlands, woodlands, sandy, saline, disturbed
	F. linearis	77	Wetlands, woodlands, sandy, disturbed
	F. ramosissima	6
	F. sonorensis	3	Disturbed, semiarid soils
Parthenium	P. hysterophorus	11	Disturbed, mainly dry or saline soils	Hedge and Patil, 1980; Moore et al., 1987
Boraginaceae
Heliotropium	H. convolvulaceum	164	Sand dune specialist	Frohlich, 1978; Vogan et al., 2007
	H. lagoense	5
	H. greggii	49	Open site, lay, gravel soils
Brassicaceae
Diplotaxis	D. erucoides	2328	Disturbed	Apel et al., 1997; USDA/NRCS, 2016
	D. muralis	4828	Grazed grasslands, disturbed
	D. tenuifolia	7206	Wetlands, wet woods, mountain slopes, sandy, disturbed
Moricandia	M. nitens	285		Holaday and Chollet, 1984; USDA/NRCS, 2016
	M. sinaica	14
	M. spinosa	1
	M. suffruticosa	32
	M. arvensis	821	Grainfields, orchards, disturbed
Cleomaceae
Cleome	C. paradoxa	7	Arid, rocky soils	Voznesenskaya et al., 2007; Feodorova et al., 2010
Euphorbiaceae
Euphorbia	E. acuta	7	Dry limestone uplands, semi-arid scrublands, disturbed	Sage et al., 2011
	E. johnstonii	1	Dry limestone uplands, semi-arid scrublands; calcareous soils, caliche outcrops
	E. lata	52	Dry limestone uplands, semi-arid scrublands; calcareous soils, sandy plains
Molluginaceae
Hypertelis	Hypertelis spergulacea	16		Edwards and Ku, 1987; Christin et al., 2011b ; USDA/NRCS, 2016
	Paramollugo nudicaulis	203	Ruderal habitats lacking competition
Mollugo	M. verticillata	1686	Fields, gardens, disturbed, moist to dry soils; lacking competition
Portulacaceae
Portulaca	P. cryptopetala	35	Moist, warm habitats	Voznesenskaya et al., 2010
Scrophulariaceae
Anticharis	A. ebracteata	5	Quartz gravel	Khoshravesh et al., 2012
	A. juncea	7	Farm, granite rocks
Cyperaceae
Eleocharis	E. atropurpurea	355	Wetlands, disturbed	Roalson et al., 2010; USDA/NRCS, 2016
	E. brainii	6
	E. flavescens	182	Wetlands
	E. nigrescens	53	Wetlands, woodlands, sandy and peaty soils
	E. subfoliata	6
Poaceae
Alloteropsis	Zambezian A. semialata	13	Shady, miombo woodlands	Lundgren et al., 2015; 2016
Homolepis	H. aturensis	411	Rainforest	Khoshravesh et al., 2016
Neurachne	N. minor	69	Arid soils, often shallow	Prendergast and Hattersley, 1985; Hattersley et al., 1986
Steinchisma	S. cuprea	8		Edwards and Ku, 1987; USDA/NRCS, 2016
	S. decipiens	130
	S. hians	285	Wetlands
	S. spathellosum	57
	S. stenophyllum	6	Wetlands

^a References describe local habit. Those characterising C₃–C₄ intermediate status are italicized.

Environmental parameters that have been predicted to potentially explain the sorting of C₃, C₃–C₄, and C₄ photosynthetic types were selected (Christin and Osborne, 2014; Supplementary Table S1). Geographic distributions are characterized with latitudinal and altitudinal ranges, and broad climatic distributions are characterized via mean annual precipitation (MAP) and mean annual temperature (MAT) variables. The growing season temperature (i.e. temperature of the wettest quarter), minimum temperature (i.e. minimum temperature of the coldest month), number of annual frost days, minimum precipitation (i.e. precipitation of the driest month), number of annual wet days, percentage of maximum possible sunshine, rainfall seasonality, and fire return interval (FRI) were also used to characterize the environment. The rainfall seasonality data, which come from Lehmann et al. (2011), are based on an index that indicates how evenly dispersed rainfall is throughout a year, with zero indicating equal rain in all months and a value of 100 indicating that all annual rain fell within a single month (see Supplementary Table S1). The FRI data, which come from Archibald et al. (2013), are based on an index that indicates the growth time available to plants between fires, with greater FRI values indicating less frequent fire regimes and longer regrowth periods (Supplementary Table S1). Climate and soil fertility data were obtained by overlaying the occurrence coordinates onto high-resolution raster layers obtained from WorldClim (http://www.worldclim.org; Hijmans et al., 2005), Climatic Research Unit (New et al., 2002; http://www.ipcc-data.org), and the Harmonized World Soils Database (HWSD; FAO/IIASA/ISRIC/ISSCAS/JRC, 2012; http://webarchive.iiasa.ac.at; Supplementary Table S1).

Data from the dominant soil type of the topsoil layer were extracted from the HWSD raster layers. Specifically, four soil parameters were used to characterize soil fertility and are described below as per the HWSD classifications (FAO/IIASA/ISRIC/ISSCAS/JRC, 2012). First, the percentage of organic carbon (OC) in the topsoil is a particularly good indicator of soil health, with moderate to high OC present in fertile, well-structured soils. Soils with less than 0.2% or 0.6% OC are considered very poor and poor, respectively, and soils with greater than 2% OC are considered fertile. Total exchangeable bases (TEB) is the sum of exchangeable cations of sodium, calcium, magnesium, and potassium in the topsoil and, as such, soils with more TEB have better fertility. The cation exchange capacity (CEC) of the topsoil indicates the total nutrient fixing capacity of the soil, with low CEC soils, such as sandy soils with CEC less than 4 cmol kg^–1, having little resilience and low nutrient stores, while soils with greater than 10 cmol kg^–1 have high nutrient fixing capacity and are suitable for crops. The pH of the topsoil indicates the acidity and alkalinity of the soil, with pH values less than 4.5, as found in mangrove soils or acid sulfate soils, being extremely acid and poorly draining, pH values of 5.5–7.2 are considered neutral, and those above 8.5 are alkaline and consequently may inhibit the bio-availability of nutrients in the soils.

The variation among environmental variables at individual plant occurrence points was summarized using a principal component analysis (PCA), as implemented in the FactoMineR package in R (Lê et al., 2008). A first PCA was conducted on climate variables, as described in Lundgren et al. (2015), and a second PCA was completed on the four soil fertility variables.

Testing for phylogenetic effects on the ecological sorting of C₃–C₄ lineages

To determine whether the ecological sorting of C₃–C₄ taxa is partially determined by the phylogenetic lineage to which they belong, we tested for an effect of the abiotic environment of the closest C₃ relatives on the sorting of C₃–C₄ lineages, and for an effect of the C₃–C₄ habitat on the sorting of the C₄ relatives. For this purpose, we identified sets of C₃–C₄ species and their C₃ and C₄ sister groups. An angiosperm-wide phylogeny including all of the C₃–C₄ groups and their relatives was unavailable, and thus groups were defined based on phylogenetic trees published for the different clades (see Supplementary Table S2). This endeavor was complicated by taxa with unknown photosynthetic types. In addition, while some small groups have well resolved phylogenetic trees with clearly identified photosynthetic types (e.g. Flaveria; McKown et al., 2005), many other systems have only been partially sampled or phenotyped. Nodes separating clearly identified C₃ and C₃–C₄, or C₃–C₄ and C₄ groups were selected, ignoring any groups with unknown photosynthetic types. For some C₃–C₄ lineages, either the C₃ or the C₄ sister group could not be identified. For example, Portulaca cryptopetala is nested in a group of C₄ species and the related species are potentially CAM (Ocampo and Columbus, 2010; Arakaki et al., 2011), and several C₃–C₄ intermediates lack close C₄ relatives (Supplementary Table S2). In cases where C₃–C₄ taxa were mixed with species of unknown type, the C₃–C₄ taxa were grouped and compared with a more distant clade with clearly established C₃ taxa (e.g. Eleocharis; Roalson et al., 2010; Paramollugo; Christin et al., 2011b), and C₃–C₄ groups forming paraphyletic clades with respect to C₄ species were merged (e.g. Flaveria; McKown et al., 2005; Lyu et al., 2015). However, C₃–C₄ belonging to the same family, but with distinct C₃ and C₄ relatives were considered separately (Supplementary Table S2). In other cases, where the phylogeny or photosynthetic categorization for a genus was incomplete, only taxa with clearly assigned photosynthetic types were considered and grouped based on the photosynthetic type independently of the phylogenetic relationships (e.g. Heliotropium; Supplementary Table S2). This approach decreases the number of contrasts, as closely related, yet independent C₃–C₄ lineages might have been merged. However, it ensures that no erroneous comparisons are included, for example when available plastid phylogenies do not perfectly match genome-wide relationships (e.g. Lyu et al., 2015). Indeed, our analyses only compare photosynthetic types within groups that are monophyletic, even if these are incompletely sampled. In conclusion, while the incomplete phylogenetic knowledge probably decreases our analytical power, our approach is statistically conservative.

The abiotic environment of C₃ and C₄ relatives of C₃–C₄ lineages was assessed as described for C₃–C₄ taxa. For each species and each variable, the median was used to avoid extreme values, which could be misidentifications or erroneously reported occurrence points. To obtain one value per group, the average of the medians was calculated for each C₃–C₄ lineage, its C₃ sister group, and its C₄ sister group. A phylogenetic effect on the sorting of C₃–C₄ taxa was evaluated with correlation tests between the climatic environment of the C₃ group and the environment inhabited by its closely related C₃–C₄ group. In the absence of phylogenetic effects, the values for C₃–C₄ taxa should be independent from those observed in the closely related C₃ group. These analyses were repeated by testing for a correlation between the environment of the C₃–C₄ lineage and that of the closely related C₄ group. Because many variables failed the Shapiro–Wilk test for normality, correlations were tested using the non-parametric Kendall rank correlation, which does not assume normality and is unbiased by small sample sizes. These tests were performed on the primary axis of the climate and soil PCAs, as well as on four climate variables (i.e. growing season temperature, minimum temperature, minimum precipitation, and rainfall seasonality) and two soil fertility variables (i.e. topsoil organic matter and TEB). These variables were selected to capture both temperature and precipitation patterns, which have classically been linked to photosynthetic types (reviewed in Christin and Osborne, 2014), and the two soil variables were selected to characterize the overall soil fertility. P-values of all tests were compared with a threshold corrected for eight comparisons (two PCA primary axes and six independent environmental variables; 0.00625).

Testing for differences among photosynthetic types, while controlling for phylogeny

Phylogenetic effects and photosynthetic types can both potentially contribute to the ecological sorting of plants. We consequently tested for differences among photosynthetic types, while controlling for phylogenetic effects. A sister group approach was adopted to compare C₃, C₃–C₄, and C₄ photosynthetic types within each lineage (see Supplementary Table S2), an approach that removes phylogenetic effects in a similar manner to phylogenetic independent contrasts (Garland et al., 1992). Indeed, a directional shift consistently associated with a given photosynthetic type within each group is strongly indicative of non-random processes (Vamosi and Vamosi, 2005; Edwards and Still, 2008; Edwards and Smith, 2010; Spriggs et al., 2014). The age of the different groups varies (Christin et al., 2011a, 2014), which means that the amount of divergence between the photosynthetic types is not necessarily constant among groups. However, our analyses are based on rank or sign tests and are therefore unaffected by variation in the magnitude of differences between photosynthetic types within each group. Consistent shifts between photosynthetic types were evaluated as the number of clades where the mean of the medians of the type of interest (either C₃–C₄ or C₄) was larger than the mean of the medians of the comparison (C₃ and C₃–C₄, respectively). The probability of observing such a shift with a random process (i.e. a probability of success of 0.5) was calculated based on a binomial distribution, in a two-tailed sign test. These tests were performed on the same eight variables used to assess the phylogenetic effects on C₃–C₄ sorting, and using the same corrections for multiple testing.

Results

Geographic distribution of C₃–C₄ intermediates

As a whole, C₃–C₄ intermediates are broadly distributed across Australia, Asia, Europe, Africa, and the Americas (Fig. 1). While the sampling is clearly biased toward western Europe, Central America, and specific countries (e.g. Israel), the data clearly indicate that intermediates can occur in most tropical and temperate regions. The C₃–C₄ occurrences span a latitudinal belt between 50°S and 65°N (Fig. 1, Table 2, and Supplementary Dataset S1), with Diplotaxis intermediates reaching from northern Europe to the south of Australia, Africa, and South America (see Supplementary Fig. S1). Eleocharis and Mollugo C₃–C₄ plants are similarly widespread, spreading across the Americas, Europe, Africa, Asia, and Australia (Table 2, Supplementary Figs S1 and S2, and Supplementary Dataset S1). Other intermediate lineages, such as Alloteropsis, Neurachne, Blepharis, and Sebodassia, have smaller geographic ranges, according to the available occurrence data (Table 2, Supplementary Figs S1 and S2, and Supplementary Dataset S1). Many intermediates occur well below sea level, along the Dead Sea (e.g. Diplotaxis erucoides, Moricandia sinaica, and Parthenium hysterophorus), in The Netherlands (e.g. Diplotaxis tenuifolia and Diplotaxis muralis), or along the Gulf of Mexico (e.g. Flaveria linearis, Eleocharis atropurpurea; Table 2, Supplementary Figs S1 and S2, and Supplementary Dataset S1). C₃–C₄ intermediates also occur at high elevations, along the Andes mountains (e.g. Steinchisma decipiens, Steinchisma hians, Mollugo verticillata, Diplotaxis muralis), in Lesotho (e.g. Diplotaxis muralis, Blepharis espinosa), and in the highlands of Mexico (e.g. Mollugo verticillata, Berkheya spinosissma; Table 2; Supplementary Figs S1 and S2; Supplementary Dataset S1).

Fig. 1.

Global distribution of C₃–C₄ taxa. Each dot represents an occurrence point for a single C₃–C₄ intermediate plant.

Table 2.

Ranges of geography, climate, and soil characteristics of C₃–C₄ taxa within each lineage group

C3–C4 group	n	Latitude	Altitude (m)	MAT (°C)	MAP (mm)	OC (% weight)	TEB (cmol/kg)	CEC (cmol/kg)	pH (–log (H+))
Eudicots
Alternanthera	714	35°S–51°N	0–2873	8–29	363–4523	0.1–16	0.2–76	1–76	3.3–8.4
Anticharis	12	29°S–22°S	289–1831	18–23	27–442	0.3–0.7	1.5–16	2–16	5.5–8.6
Blepharis	84	33°S–12°S	182–2555	10–23	100–1228	0.1–1.6	0–41	0–41	4.9–9.8
Cleome	7	11°N–16°N	23–777	25–29	38–503	0.3–0.7	6.8–17	6–17	6.5–8.1
Diplotaxisa	14362	50°S–65°N	–409 to 3959	–2 to 26	33–2990	0.1–39.4	0.8–68.2	1–87	4.1–8.8
Euphorbia	60	25°N–38°N	59–1913	11–23	245–736	0.4–1.8	4.4–31.1	5–23	6.0–8.4
Flaveria	209	17°N–35°N	–1 to 3116	10–27	214–1581	0.3–14	1.7–83	4–83	4.5–8.4
Heliotropium	218	15°S–40°N	0–2543	9–26	63–2183	0.1–14	1.1–44	2–44	4.7–8.4
Hypertelis	16	29°S–28°S	68–1086	16–23	41–98	0.4–0.7	4.0–16	4–16	6.5–8.5
Mollugoa	1889	38°S–53°N	–5 to 4209	0–30	1–4048	0.1–35.3	0.2–83	2–85	3.3–10.2
Moricandiaa	1153	35°S–60°N	–251 to 2701	6–25	10–1328	0.2–2.7	2.0–46.6	3–43	4.4–8.7
Partheniuma	11	22°S–33°N	–228 to 904	18–23	325–1685	0.4–1.6	1.7–45.2	6–44	4.9–8.1
Portulaca	35	34°S–17°S	2–1948	15–26	308–1749	0.4–2.5	0.6–43.4	2–43	4.9–9.0
Salsola	32	28°N–40°N	5–1066	14–21	97–545	0.5–1.4	4.5–24.3	5–16	6.4–8.0
Sedobassia	3	43°N–48°N	64–97	10–12	527–540	1.1–1.7	38.0–40.9	23–43	6.9–7.8
Monocots
Eleocharis	604	35°S–51°N	–1 to 3805	–1 to 29	163–4614	0.1–35.3	0.2–76	2–84	3.3–8.9
Alloteropsis	13	13°S–6°S	958–2264	18–24	812–1439	0.7–2.5	0.8–12	5–20	4.6–6.5
Homolepis	411	18°S–20°N	0–3548	8–28	671–7731	0.1–28	0.2–83	1–85	3.3–8.3
Neurachne	69	34°S–23°S	205–637	14–24	166–1128	0.3–2.1	2.1–18.1	2–15	4.5–8.3
Steinchismaa	486	35°S–37°N	2–4524	3–27	229–3104	0.2–5.3	0.2–45.2	2–46	3.5–9

CEC, topsoil cation exchange capacity; MAP, mean annual precipitation; MAT, mean annual temperature; OC, topsoil organic matter content; TEB, topsoil total exchangeable bases.

^a C₃–C₄ lineages lacking close C₄ relatives.

Environmental distribution of C₃–C₄ intermediates

As a whole, C₃–C₄ taxa are broadly distributed across environments, inhabiting a variety of warm biomes, from tropical rainforests to deserts (Fig. 2C, D and Tables 1 and 2). In particular, C₃–C₄ eudicots are distributed within tropical seasonal forests, savannas, the woodland/grassland/shrubland habitats, temperate forests, and deserts (Fig. 2C, G). C₃–C₄ monocots are primarily distributed within tropical seasonal forests and savannas (Fig. 2D, H). Unlike C₃–C₄ eudicots, they are largely excluded from deserts and are present in tropical rainforests. They also have a smaller presence in the woodland/grassland/shrubland habitats than eudicot intermediates (Fig. 2D, H).

Fig. 2.

Comparative C₃–C₄ distributions across biomes. The median ± 10th and 90th quantiles for mean annual temperature (MAT) and precipitation (MAP) are plotted for eudicot (left) and monocot (right) C₃ sister (blue; A, B), C₃–C₄ (green, C, D), and C₄ sister (red, E, F) taxa. The bottom row overlaps the three distributions for eudicots (left, G) and monocots (right, H). All panels contain biome classifications (see Ricklefs, 2008) for tropical rainforest (TrRF), temperate rainforest (TeRF), temperate forest (TF), tropical seasonal forest (TSF), woodland/grassland/shrubland (WGS), savanna (S), desert (D), taiga (Ta), and tundra (Tu).

While the exact conditions in which the plants grow are not captured by average climatic variables, especially for annuals, annual precipitation may be virtually absent (e.g. Mollugo verticillata in the warm coastal deserts of Peru) or over 7700 mm (e.g. Homolepis aturensis in the tropical rainforests of Colombia) in habitats supporting C₃–C₄ intermediates (Table 2 and Supplementary Dataset S1). C₃–C₄ plants can inhabit regions with mean annual temperatures just below zero (e.g. Diplotaxis muralis, Diplotaxis tenuifolia, Eleocharis flavescens), but also as high as 30°C (e.g. Paramollugo nudicaulis; Table 2 and Supplementary Dataset S1). They exist in areas with winter temperatures down to –25°C (e.g. Diplotaxis muralis and Mollugo verticillata in Ontario and Saskatchewan, Canada) and 285 days of frost per year (e.g. Mollugo verticillata in the Rocky Mountains of Colorado and Eleocharis flavescens in the Andes of Chile) and growing season temperatures as low as –10°C (e.g. Eleocharis flavescens in Wyoming) but also above 32°C (e.g. Paramollugo nudicaulis in Pakistan, Heliotropium convolvulacea in California, Eleocharis atropurpurea in Western Australia, and Cleome paradoxa in Ethiopia; Supplementary Dataset S1). These broad climatic variables do not encapsulate the micro-environment of each species. Of the plants that inhabit the coldest climates, Mollugo verticillata and Diplotaxis muralis are annuals, and the perennial Eleocharis flavescens occurs in aquatic environments connected to warm thermal water (Simpson and Simpson, 2015). However, these regional climatic variables do highlight the broad-scale variation among C₃–C₄ taxa. The broad ecological distribution of C₃–C₄ taxa found in the global raster datasets is supported by species-specific habitat descriptions from the literature (Table 1). These descriptions report C₃–C₄ plants from deciduous woodlands, grasslands, wetlands, scrublands, and mountainous slopes, as well as from a variety of soil types (e.g. from fine-textured, to sandy, gravelly, and rocky soils; Table 1).

Phylogenetic effects on the sorting of C₃–C₄ taxa and their C₄ relatives

The C₃ relatives of C₃–C₄ lineages occur in a variety of temperature regimes from dry habitats to moderately wet ones, a pattern that is similar in monocot and eudicot systems (Fig. 2A, B). The medians of the C₃–C₄ lineages are widely distributed along the primary PCA axis for climatic variables, which explains 50.23% of the variation, and these are not correlated to those of their close C₃ relatives (Fig. 3A, C, E and Table 3). However, the soil fertility conditions experienced by C₃–C₄ plants, extracted from the primary PCA axis for soil variables, which explains 55.58% of the variation, are correlated to those of their C₃ relatives, which might be driven by topsoil TEB (Fig. 3B, D, F and Tables 3 and 4). Similarly, variation in minimum precipitation experienced by C₃–C₄ lineages is correlated to that of closely related C₃ lineages (Fig. 4C and Table 3), indicating a strong phylogenetic effect.

Fig. 3.

Distribution of photosynthetic types in ecological space. The median ± 10th and 90th quantiles for the first two principal component axes (PC1 and PC2) of the climate (A) and soil fertility (B) PCAs for C₃ sister (blue), C₃–C₄ (green), and C₄ sister (red) taxa. The associated variable factor maps for the climate and soil fertility PCAs are shown in (C, D). Shifts in the primary axis of the climatic (E) and soil fertility (F) PCAs, as comparisons between C₃–C₄ taxa and their closely related C₃ (blue) and C₄ (red) sister taxa within each phylogenetic group. Comparisons of C₃–C₄ taxa and their C₃ relatives in groups that lack close C₄ relatives are presented as blue triangles. Black lines indicate the 1:1 relationship. Linear relationships are shown for correlations significant after correction for multiple testing (P<0.00625), in the relevant color (see Table 3).

Fig. 4.

Ecological shifts between photosynthetic types. Shifts in growing season temperature, minimum temperature of the coldest month, minimum precipitation of the driest month, rainfall seasonality, topsoil organic matter content, and topsoil total exchangeable bases (as labelled) between C₃–C₄ taxa and their C₃ (blue) and C₄ (red) close relatives were evaluated. Each point represents an average for all species within each comparison group (see Methods). Comparisons of C₃–C₄ taxa and their C₃ relatives in groups that lack close C₄ relatives are presented as blue triangles. Black lines indicate the 1:1 relationship. Linear relationships are shown for correlations significant after correction for multiple testing (P<0.00625), in the relevant color (see Table 3).

Table 3.

Kendall correlation tests for environmental medians among photosynthetic types across angiosperms

	C3–C4vs. C3		C4vs. C3–C4
Variable	P-value	tau	P-value	tau
Climate PCA axis 1	0.27	0.19	0.0059*	0.52
Soils PCA axis 1	0.0032*	0.50	0.02	0.46
Growth season temperature	0.14	0.25	0.03	0.42
Minimum temperature	0.78	–0.05	0.85	0.05
Minimum precipitation	0.0041*	0.48	0.0025*	0.59
Rainfall seasonality	0.07	0.31	0.0011*	0.63
Topsoil organic content	0.92	0.02	0.02	0.47
Total exchangeable bases	0.04	0.34	0.03	0.42

* Tests that were considered significant, using a threshold of 0.00625, which corresponds to a 0.05 threshold corrected for eight tests.

The close C₄ relatives of C₃–C₄ plants exist along a broad range of temperatures in eudicots, but are restricted to warmer areas in monocot species, resulting in less overlap between photosynthetic types in the latter than in the former (Fig. 2E–H). The variation among C₄ lineages on the first axis of the climate variable PCA is correlated with that of their C₃–C₄ relatives (Fig. 3A, E and Table 3), indicating an overall phylogenetic effect on the sorting of C₄ lineages. The soil fertility conditions experienced by C₄ plants, assessed with the PCA on soil variables, is weakly correlated to that of their C₃–C₄ relatives (Fig. 3B, F and Table 3). Among the individual variables, the minimum precipitation and rainfall seasonality experienced by C₄ lineages are correlated to that of their C₃–C₄ relatives (Fig. 4C, D and Table 3). Moreover, the growth season temperature and topsoil properties of C₄ lineages are also weakly correlated to those of their close C₃–C₄ relatives; however, these do not remain significant after correcting for multiple tests (Fig. 4A, E, F and Table 3). Thus, the precipitation, and possibly the temperature and soil fertility, preferences of C₄ lineages depend, to varying degrees, on phylogenetic effects.

Effects of photosynthetic types after correcting for phylogenetic signals

The five C₃–C₄ lineages without close C₄ relatives do not behave in the same manner as the lineages that did evolve C₄ photosynthesis. With the exception of Eleocharis, which contains aquatic plants that grow in warm waters within cold climates, four of these five lineages are those that occupy the coldest environments experienced by intermediate plants (Table 2) and are primarily in habitats with higher minimum precipitation than their C₃ relatives (Fig. 4C). All five of these C₃–C₄ lineages inhabit areas with more organic soils than their close C₃ relatives (Fig. 4E). These lineages without C₄ relatives are also among the most widely distributed of all intermediates groups (i.e. Diplotaxis, Mollugo verticillata; Supplementary Figs S1 and S2), which likely reflects an ability to tolerate diverse ecological conditions.

Considering the C₃–C₄ lineages with close C₄ relatives, their distributions are significantly shifted toward positive values of the primary axis of the climate variable PCA, which corresponds to drier and warmer environments, compared with their paired C₃ relatives (Fig. 3A, C, E and Table 4). This shift is reflected within the individual variables, with C₃–C₄ lineages occupying regions with warmer growing season temperatures, higher minimum winter temperatures, and more seasonal rainfall patterns than their C₃ relatives (Table 4). Therefore C₃–C₄ intermediates tend to inhabit relatively warm regions, regardless of the habitat in which their C₃ relatives occur, while their preference for habitat aridity does depend on the minimum precipitation experienced by their C₃ relatives (Fig. 4A–C and Table 4).

Table 4.

Tests for environmental shifts among photosynthetic types across angiosperms

	C3–C4vs. C3 (all lineages)		C3–C4vs. C3 (only lineages with close C4 relatives)		C4vs. C3–C4
Variable	Observeda	P-value	Observeda	P-value	Observeda	P-value
Climate PCA axis 1	14/19	0.019	12/14	0.0018*	8/15	0.61
Soil fertility PCA axis 1	10/19	0.65	8/14	0.42	6/15	0.61
Growth season temperature	14/19	0.019	13/14	0.00012*	5/15	0.30
Minimum temperature	13/19	0.064	12/14	0.0018*	8/15	0.61
Minimum precipitation	7/19	0.36	3/14	0.057	7/15	1
Rainfall seasonality	14/19	0.019	12/14	0.0018*	6/15	0.61
Topsoil organic content	11/19	0.36	6/14	0.79	5/15	0.30
Total exchangeable bases	8/19	0.65	5/14	0.42	6/15	0.61

^a The number of points higher in the focal group is indicated.

* Tests that were considered significant, using a threshold of 0.00625, which corresponds to a 0.05 threshold corrected for eight tests.

None of the studied environmental parameters, including both of the composite PCA variables and the six individual environmental variables, show a significant shift between close C₃–C₄ and C₄ relatives (Table 4). Therefore, with the data available here, the C₄ physiology is not linked to consistent ecological shifts when controlling for phylogenetic effects.

Discussion

A uniform C₃–C₄ niche does not exist

C₃–C₄ taxa are remarkably widespread across geographical and environmental space, maintaining the ability to exist in both typical C₃ and C₄ niches (Figs 1–3 and Supplementary Figs S1 and S2). It should be noted that the GBIF occurrence data, if anything, represent a subset of the total geographic range for each species and the realized geographical and environmental ranges of these taxa may be larger than presented here, especially for groups distributed in poorly sampled areas, such as Africa and southeast Asia. However, since related taxa tend to occur in similar regions, a sampling bias would likely affect the different photosynthetic types within a lineage to a similar degree, and the dataset is therefore still representative of the relative distribution of each type. Furthermore, several of the C₃–C₄ groups likely include more intermediate species than we present here, as we considered only those taxa for which the photosynthetic type has been assessed with confidence. For instance, the photosynthetic type of only one species within the genus Homolepis has been determined (Khoshravesh et al., 2016), while the remaining five congeners have not yet been characterized. The same is true of Eleocharis, where several species have been characterized as only possible intermediates (Roalson et al., 2010) and, as such, were not included in the study. Finally, it is unknown whether the various occurrences for each taxon are using the same photosynthetic type, or whether these vary intraspecifically across space or environments, as has been observed in the grass Alloteropsis semialata (Lundgren et al., 2015, 2016), and suggested for other taxa (e.g. Khoshravesh et al., 2012). When this variation had been reported but not clarified, the taxon was ignored, but in most cases, only a limited number of plants have been screened per species. With these caveats in mind, it is clear that the physiology of C₃–C₄ plants does not strongly restrict the migration of species geographically or into new environments.

Evolutionary history influences the realized ecology

While differences between sister groups can result from shifts in either group, they do allow for comparisons among character states independent of phylogeny. Interestingly, these analyses clearly show that the precipitation niches of C₃–C₄ taxa are statistically correlated to those of their close C₃ relatives, specifically with respect to minimum precipitation. This suggests that C₃–C₄ plants can occur in arid habitats if their C₃ relatives are already adapted to do so, and not specifically as a result of the C₃–C₄ physiology. Similarly, statistical evidence indicates that soil preferences of C₃–C₄ are correlated to those of their close C₃ relatives. C₃–C₄ physiology is only part of the attributes that a plant can use to tolerate environmental conditions, which tend to be similar among relatives (Christin and Osborne, 2014). These attributes, which can include life-history traits, growth strategies, and other non-photosynthetic characters, lead to a certain niche conservatism. Moreover, related taxa tend to occur within the same regions as a function of their biogeography, which increases the likelihood of being found in similar environments. Both precipitation variables are similarly correlated between C₃–C₄ and C₄ relatives, likely explaining previously reported differences among C₄ lineages in aridity preferences (Teeri and Stowe, 1976; Stowe and Teeri, 1978; Taub, 2000; Christin and Osborne, 2014). The influence of evolutionary history on the realized C₄ niche could go beyond precipitation preference, as our data suggest that temperature and soil fertility between C₃–C₄ and closely related C₄ groups are also associated, although this was not significant with our small species sampling.

C₃–C₄ species shift closer to the C₄ niche

In some cases, C₃–C₄ lineages emerged from groups that already inhabited warm climates, as reported in C₄ grasses (Edwards and Smith, 2010), while in others cases, C₃ relatives exist in cold areas (Fig. 3A, C). Independent of C₃ ecology, the C₃–C₄ lineages occupy warm habitats, which might reflect the increased temperature tolerance conferred by the C₃–C₄ physiology (Schuster and Monson, 1990). Despite some C₃–C₄ taxa persisting in cold regions, the convergence of physiological intermediates in warmer areas, whether that be in wet forests or dry deserts, may have increased the likelihood of further transitions to a C₄ state that occupies a similar temperature niche. Therefore, in terms of temperature, the C₃–C₄ state brings lineages into warmer habitats that should promote photorespiration and, thus, may encourage selection for C₄ physiology, thereby representing a true bridge between the ancestral C₃ state and C₄ origins. As more detailed phylogenies and updated lists of C₃–C₄ species become available, further comparative work might be able to distinguish whether this happens via an increase in C₃–C₄ migrations toward warmer habitats or a decrease in their migrations outside of such habitats, since both scenarios would result in a concentration of C₃–C₄ lineages in warmer habitats than their C₃ relatives.

While precipitation preferences vary tremendously across C₃–C₄ lineages as a function of evolutionary history, these intermediate lineages shifted toward habitats with more rainfall seasonality than their close C₃ relatives, yet no consistent shift was observed between C₃–C₄ plants and their C₄ relatives (Table 4). Phylogenetic models in grasses have previously reported that C₄ origins were accompanied by consistent shifts into drier habitats (Edwards and Smith, 2010), a trend that we suggest is initiated in C₃–C₄ taxa. Direct measurements and modelling efforts have failed to identify increases in water-use efficiency in intermediates of Flaveria, which suggests that the C₃–C₄ advantage is mainly linked to carbon gain, not water saving (Monson, 1989; Vogan and Sage, 2011; Way et al., 2014). However, the xylem architecture was altered during the transition from C₃ to C₃–C₄ species in Flaveria, providing protection against cavitation and hence increased drought tolerance (Kocacinar et al., 2008). Such alterations of leaf hydraulics, if consistently associated with the C₃–C₄ type, might explain their observed propensity to migrate to habitats with higher rainfall seasonality, habitats that would promote episodes of water limitations, potentially increasing the pressure for further evolutionary transitions to C₄ photosynthesis (Osborne and Sack, 2012), especially in warm habitats where C₃–C₄ plants tend to occur.

The fate of C₃–C₄ lineages lacking C₄ relatives

Since all of the taxa included in this study still naturally occur in the wild, they have all persisted in a C₃–C₄ state since their early emergence from C₃ ancestors, which is estimated to be as recent as 2 and as old as 20 Ma, depending on the group (Christin et al., 2011a). However, most of the known C₃–C₄ lineages are related to some C₄ groups, which prove that their ancestors had the ability, at least at some point, to produce C₄ descendants. Clear exceptions include the closely related groups Diplotaxis and Moricandia, which belong to a large family completely lacking C₄ taxa (Brassicaceae). While three other C₃–C₄ groups (Steinchisma, Mollugo verticillata, Parthenium) belong to families with C₄ origins, which are included here for other C₃–C₄ groups (Poaceae, Molluginaceae, Asteraceae), they are sufficiently distant from any C₄ group in their phylogenies that one cannot be sure whether their ancestors were able at any point to produce C₄ descendants (Christin et al., 2011b; Grass Phylogeny Working Group II, 2012). It is therefore reasonable to ask whether some attributes of these groups decreased the likelihood of C₄ evolution. While genomics, anatomy, and physiology might play a role (Christin et al., 2013; Bräutigam and Gowik, 2016), the ecology might also affect these evolutionary trajectories. For instance, C₃–C₄Moricandia occurs mainly in colder climates, which might decrease pressure for C₄ evolution. Three of the other four C₃–C₄ groups lacking close C₄ relatives are among the most widespread geographically (see Supplementary Figs S1 and S2), and these groups tend to occur in habitats with relatively high minimum precipitation and fertile soil. While none of these factors should prevent C₄ evolution in itself, it is possible that the realization of the C₃–C₄ phenotype in these groups was successful enough to limit selective pressures for further transitions in photosynthesis.

Conclusions

In this study, we present the first systematic description of the geographical and ecological distribution of C₃–C₄ intermediates. Our investigations reveal that C₃–C₄ taxa are found in a very large range of conditions and habitats, from dry deserts to tropical rainforests and cold wetlands. This variation is partially explained by evolutionary history, with C₃–C₄ lineages tending to grow in habitats with similar precipitation to their C₃ relatives, a conservatism that is further reported onto C₄ lineages. However, C₃–C₄ taxa inhabit warm climates, independent of the ancestral condition, and shift toward more seasonal rainfall habitats. Our findings indicate that the C₃–C₄ condition moves lineages into environments that promote photorespiration and, as such, may facilitate the evolution of a full C₄ pathway. There is, in our dataset, no clear difference between C₃–C₄ and C₄ in any of the environmental preferences. However, different C₄ groups might shift in various directions or extend their niche in ways that are not universal across flowering plants as, for example, it has been suggested that C₄ evolution was linked to different pressures in grasses and chenopods (Osborne and Freckleton, 2009; Kadereit et al., 2012). While group-specific detailed analyses might reveal peculiarities of each lineage, our angiosperm-wide joint analysis of C₃, C₃–C₄, and C₄ groups helps to disentangle the ecological changes that happened during consecutive phases of C₄ evolution. Indeed, shifts toward drier and warmer habitats occurred in C₃–C₄ lineages, but others, such as geographic expansions, might be specific to the C₄ state. When detailed phenotype information becomes available for a larger number of taxa, similar analyses might identify the changes linked to each individual C₄ component, bringing together anatomy, biochemistry, physiology, and evolutionary ecology.

Supplementary data

Supplementary data are available at JXB online.

Dataset S1. Occurrence and environmental data for C₃–C₄ taxa and their close C₃ and C₄ relatives used in this study.

Fig. S1. Distribution of C₃ sister (blue), C₃–C₄ (green), and C₄ sister (red) taxa in eudicot comparison groups.

Fig. S2. Distribution of C₃ sister (blue), C₃–C₄ (green), and C₄ sister (red) taxa in monocot comparison groups.

Table S1. Details on the environmental data used in this study.

Table S2. Details of C₃–C₄ species used in this study and the C₃ and C₄ sister taxa within each comparison group.

Abbreviations:

Abbreviations:
 
• CEC

  cation exchange capacity

 
• FRI

  fire return interval

 
• MAP

  mean annual precipitation

 
• MAT

  mean annual temperature

 
• OC

  organic carbon

 
• TEB

  total exchangeable bases

 
• PCA

  principal component analysis.

Acknowledgements

This work was supported by a European Research Council grant (ERC-2014-STG-638333) and a Royal Society University Research Fellowship (URF120119).