Environmental and physiological determinants of carbon isotope discrimination in terrestrial plants




Stable carbon isotope ratios (δ13C) of terrestrial plants are employed across a diverse range of applications in environmental and plant sciences; however, the kind of information that is desired from the δ13C signal often differs. At the extremes, it ranges between purely environmental and purely biological. Here, we review environmental drivers of variation in carbon isotope discrimination (Δ) in terrestrial plants, and the biological processes that can either damp or amplify the response. For C3 plants, where Δ is primarily controlled by the ratio of intercellular to ambient CO2 concentrations (ci/ca), coordination between stomatal conductance and photosynthesis and leaf area adjustment tends to constrain the potential environmentally driven range of Δ. For C4 plants, variation in bundle-sheath leakiness to CO2 can either damp or amplify the effects of ci/ca on Δ. For plants with crassulacean acid metabolism (CAM), Δ varies over a relatively large range as a function of the proportion of daytime to night-time CO2 fixation. This range can be substantially broadened by environmental effects on Δ when carbon uptake takes place primarily during the day. The effective use of Δ across its full range of applications will require a holistic view of the interplay between environmental control and physiological modulation of the environmental signal.

I. Introduction

It has been known for many decades that the stable carbon isotope ratio (δ13C) of terrestrial plant organic material varies among different kinds of plants and among plants growing in different environments (Nier & Gulbransen, 1939; Wickman, 1952). Pioneering work demonstrated that plant δ13C relates to different photosynthetic pathways (Bender, 1968) and to leaf gas exchange characteristics (O'Leary, 1981; Farquhar et al., 1982b). These discoveries have revealed unique insights into the physiology of terrestrial plants.

Plant δ13C analyses are currently used for a diverse range of applications, which can require rather different types of information from the δ13C signal. For example, in order to reconstruct paleoclimate using fossil plant or herbivore remains, it would be ideal if the δ13C signal simply recorded the influence of the environment on plants (Arens et al., 2000; Kohn, 2010). On the other hand, in order to select genotypes that have particular gas exchange characteristics for crop breeding, it would be ideal if the δ13C signal reflected primarily the intrinsic physiology of the plant, without being overly influenced by vagaries of the growth environment. At times, there is a tension between the desire to assign most of the variation in plant δ13C to environmental control and the desire to assign it to species- or genotype-specific physiological set points.

Here, we review environmental and physiological sources of variation in carbon isotope discrimination (Δ) in terrestrial plants, and provide an update on recent developments in the field. For C3 plants, ternary effects have recently been incorporated into a model for Δ, with implications for determining mesophyll conductance from coupled measurements of Δ and leaf gas exchange; for C4 plants, recent work has led to a more refined understanding of the role of bundle-sheath leakiness in modulating Δ; and for plants with crassulacean acid metabolism (CAM), a number of large species surveys have recently been published, revealing new insights into the distribution and evolutionary history of CAM. Our aim is to provide some guidance for the wide variety of applications that make use of the Δ signal, and to stimulate future research that can clarify some of the less well understood patterns in Δ. Because the C3, C4 and CAM photosynthetic pathways show distinct behaviours with respect to Δ, we address each of them separately.

II. Carbon isotope discrimination

Carbon isotope discrimination (Δ) differs from δ13C in that it describes only that change in isotopic composition induced by the plant, eliminating variation as a result of the starting value of the atmospheric CO2 used for photosynthesis. Farquhar & Richards (1984) defined Δ as

display math(Eqn 1)

where Ra is the 13C/12C ratio of CO2 in air, and Rp is that of plant carbon. In the second form of Eqn 1, δa is δ13C of CO2 in air and δp is that of plant carbon. The δ13C is defined with respect to a standard:

display math(Eqn 2)

where δ13Csample is that of the sample of interest, Rsample is its 13C/12C ratio, and Rstd is the 13C/12C ratio of a standard. The internationally accepted standard for expressing stable carbon isotope ratios is PeeDee belemnite (PDB), with a 13C/12C of 0.0112372 (Craig, 1957). In order to avoid working with very small numbers, Δ and δ13Csample are typically multiplied by 1000, and denoted as parts per thousand (‰). When Eqn 1 is multiplied by 1000, this does not affect terms in the denominator. Therefore, if δp were 28‰ in the numerator, 1 + δp in the denominator would still be 1.028.

Plant biomass provides an estimate of Δ integrated over the period of tissue synthesis. On the other hand, Δ can also be measured instantaneously by combining measurements of leaf gas exchange with online analyses of the carbon isotope ratio of CO2 entering and leaving the gas exchange cuvette (Evans et al., 1986):

display math(Eqn 3)

where δe and δo are the δ13C of CO2 entering and leaving the cuvette, respectively. The ξ is defined as

display math(Eqn 4)

where ce and co are the CO2 mole fractions, expressed here for dry air, entering and leaving the cuvette, respectively.

Traditionally, stable carbon isotope ratios have been determined by isotope ratio mass spectrometry (IRMS), which can achieve precisions of better than 0.1‰ for δ13C (Trolier et al., 1996; Vaughn et al., 2004). Recently, absorption spectroscopy techniques, such as tunable diode laser (TDL) spectroscopy and cavity ring-down spectroscopy have become more common. The advantages are high sample throughput, lower cost and suitability for field deployment. The drawback is that analytical precision is often not as good. For TDL, reported precisions for δ13C for near-ambient CO2 concentration samples ranged from 0.03 to 4‰ (Bowling et al., 2003; Griffis et al., 2004; Barbour et al., 2007; Ubierna et al., 2013).

When measurements of δ13C and CO2 mole fraction are combined in Eqn 3, the measurement error in Δ also depends on ξ. For example, a standard error of 0.2‰ in δ13C would scale to 3.2‰ in Δ for an ξ of 15, whereas for an ξ of 3, it would scale to 0.6‰. The value of ξ can be reduced by using larger leaf areas and lower flow rates, thereby increasing the drawdown in CO2 mole fraction in air leaving the leaf cuvette.

III. The C3 photosynthetic pathway

The most widely applied model of Δ in C3 plants (Δ3) is that of Farquhar et al. (1982b). The model relates Δ3 to variation in the ratio of intercellular to ambient CO2 mole fractions (ci/ca), in addition to a number of other parameters. The model has recently been updated to include a ternary correction (Farquhar & Cernusak, 2012). This correction accounts for the influence of transpiration on the diffusion of CO2 between the atmosphere and the intercellular air spaces. ‘Ternary’ refers to three interacting gases, in this case CO2, water vapour and air. The CO2 molecules diffusing into the leaf collide both with air and water vapour molecules. When the leaf is transpiring, this causes the intercellular CO2 mole fraction, ci, to be lower than it would be in the absence of transpiration (Jarman, 1974; Cowan, 1978; von Caemmerer & Farquhar, 1981). This has recently been confirmed experimentally (Boyer & Kawamitsu, 2011). The ternary correction has been applied in calculations of ci in commercial gas exchange systems for many years, but was not previously included in the Δ3 model. Its impact on the latter is on the order of 1–2‰, but varies with leaf-to-air vapour pressure difference (v).

The model for Δ3, including the ternary correction, is as follows (Farquhar & Cernusak, 2012):

display math(Eqn 5)

where t is the ternary correction factor defined by Eqn 6, ab is the 13C/12C fractionation for CO2 diffusion across the boundary layer (2.8‰), as is that for diffusion through the stomata (4.4‰), and am is that for dissolution and diffusion from the intercellular air spaces to the sites of carboxylation in the chloroplasts (1.8‰). The term b is the fractionation associated with carboxylation, mainly by Rubisco in C3 plants (c. 29‰), e is the fractionation during day respiration, and f is the fractionation during photorespiration. The magnitudes of e and f are discussed later. The Rd is the rate of day respiration, Vc is the rate of Rubisco carboxylation, and Γ* is the CO2 compensation point in the absence of day respiration. ca, cs, ci, and cc are the CO2 mole fractions in the ambient air, at the leaf surface, in the intercellular air spaces and at the sites of carboxylation, respectively. The terms αb, αe, and αf are defined as 1 + b, 1 + e and 1 + f, respectively. When Eqn 5 is expressed as parts per thousand (‰), only one parameter in each term is multiplied by 1000. Therefore, αb remains as 1.029, and similarly for αe and αf. The ternary correction factor, t, is defined as

display math(Eqn 6)

where E is transpiration rate, gac is the combined boundary layer and stomatal conductance to CO2, and αac is defined as 1 + ā, where ā is the weighted fractionation for diffusion across the boundary layer and stomata in series:

display math(Eqn 7)

The magnitudes of e and f are currently under investigation, and a range of values has been reported. Recent estimates for e range from c. 0 to 5‰ (Tcherkez et al., 2004, 2010, 2011b). Recent estimates for f range from c. 8 to 16‰ (Gillon & Griffiths, 1997; Lanigan et al., 2008; Evans & von Caemmerer, 2013), with a value of 11‰ suggested on theoretical grounds (Tcherkez, 2006). For instantaneous measurements of Δ3, the inferred value of f depends on the assumed value for b. Evans & von Caemmerer (2013) recently found that a value for f of 16‰ fit their data for tobacco best when b was assumed to be 29‰.

An apparent respiratory fractionation can occur when Δ3 is measured instantaneously under an atmosphere with δ13C differing from that which the plant was exposed to in the hours to days leading up to the measurement. In this situation, the respired CO2 can have a different isotopic composition than it would have in the steady state, owing to the change in δ13C of the source CO2 for photosynthesis. Under these circumstances, the value for e can be replaced by eRd + e* to account for the difference between δ13C of assimilate produced during photosynthesis and that of the likely substrate for respiration (Wingate et al., 2007). Here eRd would be the fractionation during day respiration and e* would be δa − Δobs − δsubstrate, where δa is the δ13C of CO2 in the cuvette during online measurements, Δobs is the observed discrimination, and δsubstrate is the δ13C of likely respiratory substrates.

Online determinations of Δ3 have proven useful in recent years for estimating mesophyll conductance, gm, when combined with measurements of leaf gas exchange (Evans et al., 1986; Pons et al., 2009). Observed values of Δ3 are less than predicted for the case where cc = ci; this discrepancy forms the basis for estimating gm. Low values of gm can significantly constrain photosynthesis (Flexas et al., 2012). Manipulating gm may therefore provide a means of increasing photosynthesis and water-use efficiency in crop plants (Barbour et al., 2010).

The gm can be estimated by first defining the predicted Δ3 when cc = ci, termed Δi, which would apply if gm were infinite (Evans et al., 1986; Farquhar & Cernusak, 2012):

display math(Eqn 8)

Eqn 8 differs from Eqn 5 in that cc has been replaced by ci, and cc/Vc has been replaced by (ci − Γ*)/(A + Rd). The latter substitution is made because A is the term that is actually measured by instantaneous gas exchange, rather than Vc. The difference between Δi and the observed Δ3obs) can then be used to estimate gm:

display math(Eqn 9)

where P is atmospheric pressure. The P enters the equation here because it is preferable to derive gm from a draw-down in partial pressure rather than mole fraction, so that it is independent of the temperature sensitivity of the solubility of CO2. The gm calculated in this way has dimensions of mol m−2 s−1 bar−1. The ternary correction can have a significant effect on the calculated gm, particularly at high v (Farquhar & Cernusak, 2012; Evans & von Caemmerer, 2013). The effect of the ternary correction on gm determinations in tobacco is shown for a range of v in Fig. 1.

Figure 1.

Mesophyll conductance (gm) in tobacco calculated without (circles) and with (triangles) the ternary correction plotted as a function of leaf to air vapour pressure difference (v). The source of variation in gm was leaf temperature, but gm is plotted here as a function of v to show the sensitivity of the ternary correction to v. (Redrawn from Evans & von Caemmerer (2013), with permission.)

The derivation of Eqn 5 assumes that CO2 is released from mitochondria in the same compartment within the cell as that in which Rubisco is located. The assumption has been justified based on the observation that chloroplasts generally line mesophyll cell walls adjacent to intercellular air spaces, with mitochondria located deeper inside the mesophyll cells. Thus, CO2 produced by mitochondria would have to diffuse through chloroplasts to reach the intercellular air spaces. For modelling purposes, this is equivalent to the mitochondrial CO2 being evolved within the chloroplasts. Recently, it has been suggested that this assumption may not always hold (Tholen & Zhu, 2011; Tholen et al., 2012; Busch et al., 2013). Evans & von Caemmerer (2013) presented an alternative formulation of Eqn 5, which allows the diffusion resistance between mitochondria and the intercellular air space (rw) to be less than that between the sites of Rubisco and the intercellular air space (rm). Assuming rw = 0.5rm in their alternative formulation of Eqn 5 led to an increase in estimated gm of c. 10% at [O2] of 21% (Evans & von Caemmerer, 2013). The difference between the two estimates of gm varies as a function of [O2] because the rate of mitochondrial CO2 release associated with photorespiration varies with [O2]. To put the issue into context, assuming rw = 0.5rm in the alternative formulation of Eqn 5 had a similar effect on estimated gm as assuming f = 16‰ vs f = 11‰ (Evans & von Caemmerer, 2013).

Eqn 5 predicts Δ3 during photosynthesis. If the δ13C of the CO2 in air is known, then Eqn 5 can be used to estimate the δ13C of carbon taken up by a C3 leaf in the light. Some of this carbon will be incorporated into structural plant material, some stored in nonstructural organic compounds, and some respired back to the atmosphere. The metabolic processes that take place after the photosynthetic carbon reduction cycle provide opportunities for fractionation and for an unequal distribution of the 13C/12C ratio across different plant organs and metabolic compartments. Considerable research effort is currently directed towards teasing apart the intricate details of these processes, including the roles of dark respiration, phloem loading, anaplerotic CO2 fixation, and secondary metabolism (Hobbie & Werner, 2004; Tcherkez et al., 2004, 2011a; Badeck et al., 2005; Bowling et al., 2008; Cernusak et al., 2009a; Werner & Gessler, 2011). Post-photosynthetic fractionations can cause Δ as determined from plant biomass to differ from Δ measured on line during photosynthetic carbon uptake.

Eqn 5 can be simplified to a form that omits effects other than those of diffusion through stomata and carboxylation (Farquhar et al., 1982b). While the simplified form is clearly an approximation, it has been sufficient for many applications:

display math(Eqn 10)

Here a is the diffusional fractionation, taken as 4.4‰, and b is carboxylation fractionation. In this form of the model, b is usually taken as 27‰, which allows for a reduction in fractionation caused by the terms in Eqn 5 that are omitted from Eqn 10, mainly the draw-down in CO2 mole fraction from ci to cc.

Eqn 10 provides a means of estimating ci/ca from measurements of δp, assuming δa is known. For the most part, the CO2 in the atmosphere is well mixed, and δa can be assumed mostly constant, although it may vary in closed canopy forests, glasshouses, and other situations where the turbulent exchange of air between the plant canopy and the free troposphere is impeded. There is a seasonal cycle in tropospheric δa with an amplitude approaching 1‰ at high northern latitudes. The amplitude decreases toward the equator and very little seasonal cycle is observed in the southern hemisphere (Trolier et al., 1996). In addition, the δa of the troposphere has decreased during the industrial period as a result of the combustion of fossil fuels and the consequent release of 13C-depleted CO2 into the atmosphere. For analyses of plant organic material formed over a range of time periods, such as in tree rings, this chronological decrease in δa should be accounted for. This can be accomplished with data from a combination of direct measurements of the δ13C of atmospheric CO2 and measurements of CO2 in air trapped in ice bubbles (Feng, 1998; McCarroll & Loader, 2004).

Eqn 10 predicts that δp should vary as a function of ci/ca in C3 plants if δa is relatively constant. This prediction generally holds. Fig. 2 shows the δ13C of leaf dry matter plotted against instantaneous measurements of ci/ca for seedlings of 44 tree species. The dataset includes both conifer and angiosperm species (Orchard et al., 2010). A geometric mean regression through the dataset yields the following equation: δp(‰) = −12.1–22.0 ci/ca. This is close to the predicted relationship assuming δa = −8‰, a = 4.4‰ and b = 27‰, which would be δp(‰) = −12.4–22.6 ci/ca. The seedlings were grown either in the open air or in well-ventilated shade houses under well-watered conditions. Comparisons between plant dry matter δ13C and instantaneous ci/ca can be problematic if ontogeny or a change in environmental conditions causes ci/ca at the time of instantaneous measurements to differ from the integral over the course of tissue synthesis. This has been avoided in Fig. 2 by selecting species that have a prolonged vegetative stage (trees) and for which no dramatic change in environmental conditions was imposed. The average standard deviation for individuals within a species across the dataset shown in Fig. 2 was 0.6‰.

Figure 2.

Leaf carbon isotope ratio (δ13C) plotted against the instantaneous ratio of intercellular to ambient CO2 mole fractions (ci/ca) during photosynthesis. Data are for well-watered tree seedlings (angiosperms, circles; conifers, triangles), grown either outdoors or in well-ventilated shade houses (Brodribb & Hill, 1998; Cernusak et al., 2007, 2008; Orchard et al., 2010). Each data point represents the average of several individuals of one species. Species' identities are detailed in Orchard et al. (2010). The solid line is a geometric mean regression: leaf δ13C(‰) = −12.1–22.0 ci/ca.

Although some authors have recently speculated that cross-species comparisons of ci/ca based on δp could be problematic (Warren & Adams, 2006; Cernusak et al., 2009b; Salmon et al., 2011), Fig. 2 shows that this is really a matter of how broad a view one takes. There are variations around the regression line relating δp to ci/ca of c. 1–3‰ in δp for a given ci/ca. These could relate to small variations in δa, to temporal or spatial variations in ci/ca, to the terms in Eqn 5 that are omitted from Eqn 10, or to postphotosynthetic processes not included in Eqn 5. However, it is also clear that δp is strongly related to ci/ca, with c. 80% of among-species variation in δp in Fig. 2 explained by ci/ca. Whether one should focus on the regression line in Fig. 2 or on the residuals around it depends on the research question being asked.

Measurement of δp in C3 plants is especially powerful for being able to provide a time-integrated estimate of ci/ca (Farquhar et al., 1982a). The ci/ca is a function of the supply of CO2 to the leaf intercellular air spaces by stomata and the demand for CO2 by photosynthetic capacity within the mesophyll. For a given ca and v, ci/ca relates to photosynthetic water-use efficiency (Farquhar & Richards, 1984), defined as the ratio of photosynthesis (A) to transpiration (E):

display math(Eqn 11)

As a result of the relationship between ci/ca and A/E, δp can be used, for example, to identify C3 crop genotypes that potentially have a high water-use efficiency (Rebetzke et al., 2002, 2008; Richards et al., 2002), to assess the water-use efficiency response of trees to rising ca through measurements of δp in tree rings (Francey & Farquhar, 1982; Marshall & Monserud, 1996; Feng, 1999; Loader et al., 2011), and to test models of optimal stomatal behaviour in response to environmental gradients (Farquhar et al., 2002; Wright et al., 2003; Medlyn et al., 2011).

The supply of CO2 to the leaf intercellular air spaces is mainly controlled by the concentration of CO2 outside the leaf, ca, and by stomatal conductance, gs (Farquhar & Sharkey, 1982). Stomata generally function to allow CO2 to diffuse into leaves when conditions are favourable for photosynthesis, while at the same time preventing water loss from leaves at rates that would lead to excessive dehydration and impaired photosynthetic capacity. But stomata are not simply open or closed; their degree of opening is a continuous variable that more strongly limits the diffusion of CO2 into leaves when the supply of soil water is relatively low and when the evaporative demand of the atmosphere is relatively high (Marshall & Waring, 1984). As a result, variation in Δ3 can be expected in response to the availability of soil moisture and the atmospheric vapour pressure deficit. This can be manifested along precipitation gradients, or, more precisely, along moisture availability gradients that take into account both the local availability of soil water and potential evapotranspiration.

Two meta-analyses have recently summarized Δ3 responses to mean annual precipitation (Diefendorf et al., 2010; Kohn, 2010). Both found decreasing Δ3 with decreasing mean annual precipitation, which explained about half the variation in Δ3 (Fig. 3). Based on these results, it was argued that paleoprecipitation could be reconstructed from ancient plant carbon, preserved, for example, in fossil tooth enamel (Kohn, 2010, 2011). This was disputed on the grounds that such paleoprecipitation estimates would not be meaningful as a result of the high variability in Δ3 among different species at a given mean annual precipitation (Freeman et al., 2011). Such differences among tree species, when grown under well-watered conditions, are demonstrated in Fig. 2. Ultimately, the utility of δ13C analyses of ancient plant carbon for reconstructing paleoprecipitation will depend on the degree of uncertainty that one is willing to tolerate (Kohn, 2011).

Figure 3.

Carbon isotope discrimination (Δ) of C3 plant biomass plotted against mean annual precipitation. Data are from two meta-analyses (Diefendorf et al., 2010; Kohn, 2010). For the Diefendorf et al. data set (triangles), each data point is a species by site combination. For the Kohn data set (crosses), each data point is a site average.

Nutrient availability can influence ci/ca and Δ3 through effects on photosynthetic capacity. Photosynthetic capacity partly determines the demand for CO2 in the leaf mesophyll. The nutrient that has been best studied in this regard is nitrogen, which has a well-known relationship with photosynthesis (Field & Mooney, 1986). Fig. 4 shows an example of variation in instantaneous ci/ca and Δ3 of plant biomass as a function of leaf nitrogen concentration for seedlings of a tropical pioneer tree, Ficus insipida, grown under varying soil fertilities (Cernusak et al., 2007). Leaf δp was also observed to vary as a function of leaf nitrogen concentration within the crowns of mature trees of a range of conifer species (Duursma & Marshall, 2006). Other nutrients, such as phosphorus, would be expected to correlate with variation in Δ3 if they cause variation in photosynthetic capacity (Domingues et al., 2010).

Figure 4.

The ratio of intercellular to ambient CO2 mole fractions (ci/ca) measured instantaneously plotted against leaf nitrogen concentration (a) and carbon isotope discrimination (Δ) of plant biomass plotted against leaf nitrogen concentration (b) for seedlings of Ficus insipida grown under varying soil fertility. Solid lines are least-squares linear regressions. (Redrawn from Cernusak et al. (2007), with permission.)

Irradiance can also affect Δ3. Plants exposed to low photon flux densities show higher ci/ca and higher Δ3 associated with low photosynthesis rates (Ehleringer et al., 1986). At photon flux densities above c. 250 μmol m−2 s−1, ci/ca tends to be independent of irradiance (Wong et al., 1978; Farquhar & Wong, 1984), but Δ3 may still decrease because of an increasing draw-down from ci to cc associated with increasing photosynthesis (Evans et al., 1986; Farquhar et al., 1989).

In addition, ci/ca and Δ3 generally decline with tree height independent of irradiance (Marshall & Monserud, 1996; Ryan & Yoder, 1997; Koch et al., 2004; McDowell et al., 2011), although not always (Barnard & Ryan, 2003). This decline has been attributed to two hydraulic issues: the maintenance of lower water potentials with height above the soil water supply (Koch et al., 2004), and the frictional resistance to water flux, which increases with path length as stems grow taller (Ryan & Yoder, 1997; Becker et al., 2000). Either of these effects would tend to reduce leaf water potential and stomatal conductance, unless there were compensating shifts in leaf area to sapwood area ratio (Barnard & Ryan, 2003). Such potential height effects should be considered when comparing tree ring δp among locations and over time (Marshall & Monserud, 1996; McDowell et al., 2011).

The δp of C3 plant biomass generally increases with increasing elevation above sea level at a rate of c. 1– 2‰ km−1 of elevation gain (Körner et al., 1988, 1991; Marshall & Zhang, 1994; Hultine & Marshall, 2000). This pattern is general, occurring in both herbs (Körner et al., 1988, 1991) and trees (Vitousek et al., 1990; Marshall & Zhang, 1994). The pattern appears to be linked to both decreasing temperature and decreasing atmospheric pressure as elevation increases (Körner et al., 1991). Decreasing oxygen partial pressure reduces ci/ca by increasing the carboxylation efficiency of Rubisco (Farquhar & Wong, 1984). Decreasing temperature causes the viscosity of water to increase. This may slow the transport of water from the soil to the evaporative sites in leaves (Roderick & Berry, 2001), thereby decreasing stomatal conductance and the diffusion of CO2 into leaves, resulting in lower ci/ca. There are also changes in leaf morphology with elevation that could contribute to the trend in Δ3 (Hultine & Marshall, 2000; Zhu et al., 2010).

The preceding discussion demonstrates that Δ3 responds to environmental controls that influence the balance between supply of CO2 to carboxylation sites within the leaf and demand for CO2 by photosynthesis. However, C3 plants have also been observed to maintain a relatively constant ci/ca in the face of a changing growth environment (Wong et al., 1979, 1985; McDowell et al., 2006; Franks et al., 2013). Herein lies the tension between viewing Δ3 as a universal sensor that responds to, for example, water availability (Prentice et al., 2011) or temperature (Körner et al., 1991) and viewing Δ3 as a homeostatic set point that differs among species and genotypes (Comstock & Ehleringer, 1992). The two viewpoints both reflect reality, and at the same time oppose one another. Environmental factors modify Δ3, but internal physiology constrains the response. The balance between environmental forcing on Δ3 and physiological damping of the response appears to vary among species. This is shown in Fig. 5 for three tropical tree species, which were grown under conditions of high and low water supply, and with and without added fertilizer (Cernusak et al., 2009b). Swietenia macrophylla showed a much more pronounced response of Δ3 to the experimental treatments than did Tectona grandis or Platymiscium pinnatum.

Figure 5.

Carbon isotope discrimination (Δ) of plant biomass for seedlings of three tropical tree species grown under high and low water supply and with (black bars) and without (grey bars) added fertilizer (Cernusak et al., 2009b). Bars are means for five plants and error bars are +1 SE.

The response of Δ3 to environmental gradients can be damped by coordination between stomatal conductance and photosynthetic capacity, which reduces the realized range of variation in ci/ca (Wong et al., 1979, 1985; Cernusak & Marshall, 2001; Hetherington & Woodward, 2003; Cernusak et al., 2011). This was the case for several eucalypt species sampled along a rainfall gradient in northern Australia (Fig. 6). The dashed line in Fig. 6 shows the expected relationship between ci/ca and gs, if photosynthetic capacity had remained constant across the range of gs. Instead, photosynthetic capacity decreased with decreasing gs, such that ci/ca showed only a weak relationship with gs. Although such close coordination between gs and photosynthetic capacity has been known for some time (Wong et al., 1978, 1979, 1985), the mechanisms through which it is achieved are still not well understood (Jarvis et al., 1999; von Caemmerer et al., 2004).

Figure 6.

The ratio of intercellular to ambient CO2 mole fractions (ci/ca) plotted against stomatal conductance for several eucalypt species growing along the north Australian tropical transect. The solid line is a least-squares linear regression through the data. The dashed line is the predicted relationship between the two parameters if the maximum Rubisco carboxylation velocity, the electron transport rate and the mesophyll conductance had remained constant over the range of stomatal conductance. (The figure is from Cernusak et al. (2011), with permission.)

A second process that can damp the response of ci/ca and Δ3 to environmental gradients, particularly of water availability, is adjustment of the ratio of leaf area to water-conducting tissue (Farquhar et al., 2002). An example of this type of adjustment occurs along the north Australian tropical transect, a rainfall gradient in northern Australia (Hutley et al., 2011). Little change was observed in the Δ3 of plant biomass in trees along the transect (Schulze et al., 1998; Miller et al., 2001; Cernusak et al., 2011), in contrast to what would be expected based on the global meta-analyses shown in Fig. 3 and observations along other rainfall gradients (Stewart et al., 1995; Midgley et al., 2004; Prentice et al., 2011). Instantaneous gas exchange measurements confirmed that ci/ca changes little along the transect, even towards the end of the dry season when the north–south gradient in water availability is most pronounced (Cernusak et al., 2011). Eucalypts along the transect reduce their leaf area to sapwood area ratios from wet to dry seasons, thereby reducing the total leaf area for a given amount of water transporting tissue, allowing for maintenance of relatively constant ci/ca (Eamus et al., 2000; Cernusak et al., 2011). This leaf area adjustment may be possible owing to the highly predictable nature of the wet season, and this could explain why the northern Australian rainfall gradient is associated with much less variation in Δ3 than other rainfall gradients associated with more stochastic precipitation regimes.

Manipulative experiments provide further evidence for the role of leaf area adjustment in constraining variation in ci/ca and Δ3 (Cernusak & Marshall, 2001; McDowell et al., 2006). In one example, Pinus ponderosa stands were thinned to seven densities and maintained over a 40 yr period. The ratio of leaf area to sapwood area increased as a result of decreased basal area in thinned stands. There was an initial increase in Δ3 after thinning. However, after one decade, Δ3 returned to control values in thinned stands as the increase in leaf area to sapwood area ratio compensated for increased water availability associated with reduced basal area (McDowell et al., 2006).

It is clear that there are elements of both environmental control and genetically determined physiological set points in Δ3 (Ehleringer, 1993a,b; Zhang & Marshall, 1994; Ehleringer & Cerling, 1995). This can present a challenge for interpretation, depending on the application for which Δ3 measurements are employed. Research priorities for gaining further insight into this interplay should be to unravel the mechanisms that lead to coordination between gs and photosynthetic capacity, to determine how the statistics of rainfall regimes influence the response of Δ3 to water availability, including the role of leaf area adjustment, and to determine the specific phenotypic traits associated with genotypic variation in Δ3 (Masle et al., 2005; Liang et al., 2010).

IV. The C4 photosynthetic pathway

The C4 photosynthetic pathway is more recently derived than the ancestral C3 pathway (Edwards et al., 2010; Sage et al., 2012). It occurs mainly in grasses, in some dicotyledonous herbs, shrubs, and in a small number of trees (Pearcy & Troughton, 1975; Sage et al., 2011). The C4 pathway concentrates CO2 around Rubisco, thereby greatly reducing photorespiration relative to C3 plants. C4 plants often have Kranz anatomy (Haberlandt, 1914), meaning that Rubisco is compartmentalized within the bundle-sheath cells, although C4 photosynthesis can also occur within a single cell (Voznesenskaya et al., 2001).

C4 plants have lower Δ (Δ4) than C3 plants. The difference can be used to identify species that employ the C4 photosynthetic pathway (Bender, 1968). The lower Δ4 is a consequence of the processes and enzymes involved in C4 photosynthesis. Initially, bicarbonate is fixed in the mesophyll cells by PEP (phosphoenolpyruvate) carboxylase to form four-carbon acids (Hatch et al., 1967). Bicarbonate is enriched in 13C compared with CO2 and discrimination against 13C by PEP carboxylase is much less than by Rubisco, the primary carboxylating enzyme in C3 plants. The C4 acids fixed by PEP carboxylase then move to the bundle-sheath cells where they are decarboxylated, raising the [CO2] around Rubisco. Some of the carbon that is fixed by PEP carboxylase leaks out of the bundle-sheath cells, with the proportion termed leakiness (ϕ). The extent of ϕ is determined by the bundle-sheath conductance to CO2, which depends on physical properties, such as the presence of a suberized lamella (Hattersley, 1982; Hatch et al., 1995), and on the [CO2] gradient between the bundle-sheath and mesophyll cells, which in turn depends on the balance between PEP carboxylase and Rubisco activities (Peisker & Henderson, 1992). The ϕ plays an important role in determining Δ4 variations in C4 plants, as it controls the extent to which Rubisco fractionation is expressed.

A model to describe Δ4 was first published by Farquhar (1983) and has recently been updated to include ternary effects (Farquhar & Cernusak, 2012):

display math(Eqn 12)

where b4 is the effective fractionation by PEP carboxylase (−5.7‰ at 25°C), b3 is the fractionation by Rubisco (c. 29‰), and s is fractionation during diffusion of CO2 out of the bundle-sheath cells (1.8‰). The cm and cbs are the CO2 mole fractions in the mesophyll cytosol and in the bundle-sheath, respectively. Elaborations to take into account day respiration, photorespiration and incomplete equilibration between CO2 and bicarbonate can also be included (Farquhar, 1983).

As with Eqn (Eqn 5), Eqn (Eqn 12) can be simplified to a more manageable form (Farquhar, 1983):

display math(Eqn 13)

Eqn (Eqn 13) neglects ternary effects, assumes CO2 is in equilibrium with bicarbonate in the mesophyll cytoplasm, assumes negligible fractionation during photorespiration and dark respiration, assumes gm is infinite, and assumes the CO2 partial pressure inside the bundle-sheath cells is much larger than in the mesophyll cells. The impact of these simplifications should be carefully considered when the formulation is used to calculate ϕ (Ubierna et al., 2011).

Variation in Δ4 in response to environmental and genetic drivers is small but significant. Differences among plants are often in the range of 1–3‰, much smaller than for C3 plants (Fig. 2). Interpreting the variation in Δ4 is challenging because it cannot be attributed to a single major factor, as is generally the case with ci/ca in C3 species. In C4 species, even in the simplest case scenario presented in Eqn (Eqn 13), Δ4 depends on both ci/ca and ϕ (Fig. 7a). There is either a positive or negative correlation between Δ4 and ci/ca, depending on whether ϕ is larger or smaller than c. 0.37 [≈ (− b4)/(b3 − s)]. Moreover, the linear relationship between Δ4 and ci/ca described by Eqn (Eqn 13) is a simplification, and the full model relating Δ4 to ci/ca, expressed in Eqn (Eqn 12), has some curvature (Fig. 7b).

Figure 7.

Predicted carbon isotope discrimination (Δ) in C4 plants as a function of the ratio of intercellular to ambient CO2 mole fractions (ci/ca) during photosynthesis. Different lines represent the relationship for different values of leakiness (ϕ). Relationships in (a) are for the simplified model presented in Eqn (Eqn 13), and relationships in (b) for the nonsimplified model presented in Eqn (Eqn 12). Full details of the calculations are provided in Ubierna et al. (2011). Bundle-sheath conductance was assumed to be 0.01 mol m−2 s−1 bar−1.

In C4 species, δp of leaf dry matter has been shown to range from −9.2‰ to −19.3‰ (Hattersley, 1982), but most values are concentrated in a narrow band between −11‰ and −14‰, with a global average of c. −12.5‰ (Cerling et al., 1997). For comparison, the global average for C3 plants is c. −28‰ (Kohn, 2010). C4 grasses are grouped into three subtypes depending on the major enzyme used to decarboxylate the C4 acids: NAD-ME (NAD-malic enzyme), NADP-ME (NADP-malic enzyme) and PEPCK (phosphoenolpyruvate carboxykinase). When plants were grown under controlled conditions to minimize environmental variation, leaf δp spanned c. 3‰, with small but significant differences across subtypes: NAD-ME, −12.7‰; PEPCK, −12.0‰; NADP-ME, −11.4‰ (Hattersley, 1982).

Currently, there is no satisfactory explanation for differences in Δ4 across subtypes. As described earlier, Δ4 depends on both ϕ and ci/ca. The ϕ has been shown to differ across subtypes, with variation related to morphological features such as the presence of suberized lamellae around the bundle-sheath cells. The ϕ was estimated to be higher in the NAD-ME subtype that lacks suberized lamellae (Hattersley, 1982; Hatch et al., 1995). The lack of suberized lamellae in the NAD-ME subtype might be counterbalanced by other morphological features; for example, this subtype has elongated chloroplasts surrounding the mitochondria (Hattersley & Browning, 1981). Furthermore, the presence of suberized lamellae cannot be the only factor accounting for isotopic differences among subtypes, as both NADP-ME and PEPCK have suberized lamellae and yet differed in δp (Hattersley, 1982). The NADP-ME subtype has reduced grana in the bundle-sheath chloroplasts (Hatch, 1987; Pfündel & Neubohn, 1999), and consequently less photosystem II activity and O2 evolution than the other subtypes (Yoshimura et al., 2004; Gowik & Westhoff, 2011). Photorespiration is not always an unwanted process; under stress conditions it serves as an energy sink to prevent overreduction of the photosynthetic electron transport chain (Osmond & Grace, 1995). Interestingly, the NADP-ME subtype, with reduced O2 evolution from the bundle-sheath cells, occurs in mesic environments, whereas the NAD-ME subtype occurs in drier areas (Hattersley, 1982; Schulze et al., 1996). Cousins et al. (2008) showed that NAD-ME leaves had lower δp than did NADP-ME leaves, but instantaneous Δ4 did not differ between subtypes. This observation suggests that postphotosynthetic fractionations might play a role in determining δp.

The response of Δ4 to drought can be positive, negative, or insignificant, depending on ϕ, which contrasts with the consistent directional response in C3 plants. Under most environmental conditions, ϕ is < 0.37, and Δ4 is expected to increase with decreasing water availability as a result of decreasing ci/ca (Fig. 7). Buchmann et al. (1996) and Ghannoum et al. (2002) reported increases in Δ4 of < 1‰, with increasing drought stress in all biochemical subtypes. Increased Δ4 in leaf dry matter with increasing drought stress has also been found at large geographical scales (Tieszen & Boutton, 1989; Liu et al., 2005; Murphy & Bowman, 2009) and throughout different microhabitats (Wang et al., 2005). There are also contrasting observations; for example, no relationship between δp of leaf dry matter and rainfall was reported across several sites in Africa (Swap et al., 2004).

As in C3 species, C4 plants exhibit a relationship between Δ4 and water-use efficiency, although the trend is generally muted and in the opposite direction to that for C3 plants. Ghannoum et al. (2002) showed that increases in water-use efficiency with drought translated into significantly more depleted δp of leaf dry matter (0.5‰) for 18 grasses from the NAD-ME and NADP-ME subtypes. There was a negative correlation between water-use efficiency and δp of leaf dry matter in 30 lines of Sorghum bicolor (Henderson et al., 1998). Depleted δp values in C4 grasses have also been correlated with increased photosynthesis, growth and yield (Bowman et al., 1989; Hubick et al., 1990; Meinzer et al., 1994; Buchmann et al., 1996).

The largest reported variations in Δ4 are in response to light. The increase in observed Δ4 from high to low irradiance was c. 3‰ in Zea mays (Kromdijk et al., 2010; Ubierna et al., 2013) and Miscanthus giganteus (Kromdijk et al., 2008; Ubierna et al., 2013), and as much as 8‰ in Flaveria bidentis (Pengelly et al., 2010; Ubierna et al., 2013). Buchmann et al. (1996) found that in all C4 subtypes, reduced light during growth resulted in about a 2‰ increase in Δ4. The large Δ4 at low irradiances has been interpreted as high ϕ (0.6–0.9) and inefficient functioning of the C4 photosynthetic pathway (Cousins et al., 2006; Tazoe et al., 2006, 2008; Kromdijk et al., 2008, 2010; Pengelly et al., 2010). Recently, Ubierna et al. (2011, 2013) demonstrated that these reports of large ϕ might have resulted from the simplifications made to the theoretical model of Δ4. If Eqn (Eqn 13) was used to solve for ϕ, it resulted in very large values at low irradiances. However, when no simplifying assumptions were included in the Δ4 calculations (Eqn (Eqn 12)), ϕ only slightly increased from high to low irradiance.

In fact, if ϕ is mostly constant and insensitive to environmental variation, then Δ4 depends largely on ci/ca, which can be related to environmental variables in a predictable way. Unfortunately, ϕ cannot be measured directly, which has resulted in somewhat contradictory reports on its values and sources of variation. Leakiness has been estimated using 14C labelling (Hatch et al., 1995), mathematical modelling based on observations of quantum yield, O2 inhibition of photosynthesis, or the inorganic carbon pool size (Farquhar, 1983; Jenkins et al., 1989; He & Edwards, 1996), and coupled measurements of Δ4 and leaf gas exchange (Evans et al., 1986; von Caemmerer et al., 1997). The latter is the most commonly used method, especially since the development of laser technologies for isotopic measurements. Initial studies used δp of leaf dry matter as a proxy for Δ4 during photosynthesis. This approach resulted in differences in ϕ with environmental parameters such as drought (Saliendra et al., 1996), across subtypes (Buchmann et al., 1996), and overall larger ϕ values (0.3–0.5) than when it was derived from online Δ4 measurements (0.2–0.3, Henderson et al., 1992). Given the generally small δp variations among C4 plants, using δp of leaf dry matter to estimate ϕ could be problematic in light of possible postphotosynthetic fractionations. The problem is further complicated by different integration times between leaf tissues and instantaneous gas fluxes.

The more reliable estimates of ϕ from instantaneous measurements of gas exchange and Δ4 show it to be relatively constant over different CO2 concentrations and temperatures (Henderson et al., 1992), light intensities (Ubierna et al., 2011, 2013), moderate drought stress (Williams et al., 2001), C4 subtypes (Cousins et al., 2008), and genotypes (Henderson et al., 1998). Thus as technology for making online determinations of Δ4 has improved and calculations have been refined, the emerging trend is for a relatively small (< 0.3) and constant ϕ under a wide range of conditions. Under these circumstances, Δ4 is primarily influenced by ci/ca (Fig. 7).

C4 grasses contribute significantly to the production of both food and biofuels, and therefore play an important role in human society. Determination of Δ4 provides a tool to probe C4 photosynthetic performance in relation to environmental and genotypic variation. To make full use of this tool, future research should investigate the mechanisms that cause variation between Δ4 measured online and that measured in leaf dry matter, sources of variation of ϕ, and the extent to which and under what conditions Δ4 of leaf dry matter reflects ci/ca. More studies using combined measurements of instantaneous gas exchange and Δ4, together with fully parameterized models for Δ4, should be useful for describing ϕ under varying conditions.

V. Crassulacean acid metabolism

Crassulacean acid metabolism (CAM) is a water-conserving mode of photosynthesis exhibited by c. 6% of vascular plant species (Smith & Winter, 1996). The CAM cycle involves the uptake of atmospheric CO2 via PEP carboxylase in the dark and the overnight storage of the carbon in malic acid. During the following light period, the malic acid is decarboxylated and the CO2 is refixed by Rubisco (Borland et al., 2011). In addition to the CAM cycle, most plants with CAM can also fix atmospheric CO2 directly via C3 photosynthesis in the light. The relative proportion of CO2 fixation in the dark and the light is species dependent, plant- and tissue-age dependent, and is modulated by the environment.

Carbon isotopic discrimination in CAM plants (ΔCAM) is the result of this combination of C4 and C3 photosynthetic processes. The daytime fixation into carbohydrates of nocturnally produced malic acid would not introduce any further 13C discrimination if all the malic acid were to be consumed and all the liberated CO2 refixed by Rubisco. CAM plants can also take up atmospheric CO2 during the day, especially in the afternoon when the nocturnally fixed malic acid pool has been consumed (Osmond, 1978). This CO2 is usually fixed by Rubisco, although PEP carboxylase can also be involved (Griffiths et al., 1990).

The ΔCAM can be calculated as a photosynthesis-weighted average of nocturnal and daytime CO2 uptake (Farquhar et al., 1989):

display math(Eqn 14)

where A is photosynthesis rate, ∫Ddt and ∫Ldt are the time integrals in the dark and in the light, respectively, and Δ4 and Δ3 are carbon isotope discriminations during the C4 and C3 phases of carbon fixation, respectively. An expression for Δ4 can be derived from Eqn (Eqn 12) when ϕ = 0:

display math(Eqn 15)

An expression for Δ3 can be obtained from Eqn (Eqn 5), with b now representing the flux-weighted average discrimination of ribulose bisphosphate and PEP carboxylations (Farquhar et al., 1989). Several simplifications can be applied to these formulae, namely negligible ternary effects, large boundary layer and mesophyll conductances, and negligible respiratory and photorespiratory fractionations. This results in the following simplified expression (Farquhar et al., 1989):

display math(Eqn 16)

Elaborations can be made to these equations to account for processes such as continuing decarboxylation of malic acid in the afternoon when stomata open, and leakage of CO2 during the middle of the day, when CO2 released from malic acid is refixed (Griffiths et al., 2007). It is not known to what extent these processes may influence ΔCAM in the field.

In contrast to C3 and C4 plants, δp of plants with CAM can vary by > 20‰ (Silvera et al., 2005). The range is broad, owing to the variable contributions of light and dark fixation, in addition to other environmental and internal influences that affect fractionation (Bender et al., 1973; Osmond et al., 1973). Any factor that alters the contributions of light and dark CO2 uptake to carbon gain or alters the relative limitations of carboxylation and diffusion to CO2 uptake, for example stomatal or mesophyll conductance, will influence the isotopic signal (Farquhar et al., 1989).

Whole-tissue δp cannot be used a priori to distinguish plants in which CAM is weakly expressed from C3 plants, or strong CAM plants from C4 plants. Where CAM δp values overlap with values typical for C3 or C4 plants, supplementary characteristics such as nocturnal acidification and leaf gas exchange are required to confirm the operation of the CAM cycle (Fig. 8a).

Figure 8.

(a) Bimodal distribution of δ13C values (blue line) derived from 506 species from nine plant families containing C3 and crassulacean acid metabolism (CAM) species. The blue line represents a histogram with 0.25‰ bin widths that has been smoothed (see Winter & Holtum, 2002, for sources). The red line indicates the relationship between δ13C of plant carbon and the contribution of nocturnal CO2 fixation to the production of this carbon (after Winter & Holtum, 2002, with permission). (b) The δ13C values of 87 orchid species from Panama that exhibited statistically significant nocturnal increases in titratable acidity, based on data from Silvera et al. (2005). (c) The relationship between gravimetrically determined transpiration ratio and δ13C for C3 and CAM plants growing at a tropical outdoor research facility in Panama (from Winter et al., 2005 and Cernusak et al., 2007, with permission).

The extensive range in δp of plants with CAM was illustrated in a study of orchids from Panama (Silvera et al., 2005). The 87 species with increases in nocturnal acidity characteristic of CAM exhibited δp values between −11.8 and −32.3‰, a span of 20.5‰ (Fig. 8b). The least negative isotopic values were observed in plants with high capacities for nocturnal acidification. Forty of the species with low but significant amounts of nocturnal acidification had isotopic values that overlapped with those of 86 C3 orchid species that lacked nocturnal acidification.

In a detailed analysis of the processes that shape ΔCAM in succulent tissues of CAM plants, Griffiths et al. (2007) highlighted the importance of gm. This constraint increases Δ4 during night-time PEP carboxylation and decreases Δ3 during daytime CO2 uptake, consistent with observations that short-term Δ4 in the dark is more negative than predicted by models based on stomatal and carboxylation limitations alone (Holtum et al., 1983; O'Leary et al., 1986; Griffiths et al., 1990; Roberts et al., 1997; Winter & Holtum, 2005). Discrimination within the mesophyll probably contributed 1.4–2.5‰ to the nocturnal online Δ4 signal in three massively succulent columnar cacti, Trichocereus atacamensis, Carnegiea gigantea and Stenocereus thurberi, which exhibited internal conductances among the lowest values reported for vascular plants (Williams et al., 2012).

In spite of the complexities of diffusional and enzymatic influences on ΔCAM, it is remarkable that δp values in plants with CAM correlate linearly with the proportion of CO2 fixed during the day and night (O'Leary, 1988; Winter & Holtum, 2002), such that each 10% contribution of dark fixation results in about a 1.8‰ less negative δp signal (Fig. 8a). Measurements of whole-tissue δp for well-watered plants growing inside a gas-exchange chamber indicated a range from −8.7‰ for strong CAM plants that obtained 100% of their carbon from dark fixation, to −26.9‰ for plants that obtained 100% of their carbon during the light (Winter & Holtum, 2002).

This span of 18.2‰ for the full photosynthetic spectrum from 0 to 100% CAM is smaller than the range of δp observed in situ in species-rich taxa that contain both C3 and CAM plants. In 1873 species of C3 and CAM bromeliads, δp varied by 28.2‰, from −8.9 to −37.1‰ (Crayn et al., 2004), and δp of 1002 species of C3 and CAM orchids varied by 25.1‰, from −11.4 to −36.5‰ (Silvera et al., 2010). In these surveys, the most 13C-depleted specimens showed values considerably more negative than demonstrated by Winter & Holtum (2002) for plants that exclusively fix CO2 during the light. Such extreme values are often associated with plants from shaded and humid environments where the diffusional limitation of CO2 uptake is reduced and the source CO2 is 13C-depleted. On the other hand, C3 plant material collected from certain arid environments, where CO2 uptake is strongly diffusion-limited, may have δp values as high as −19.8‰ (Ehleringer et al., 1998). A similarly high value was reported for a high-altitude C3 species of the Rapataceae (Crayn et al., 2001). The large variation in the isotope signal of confirmed C3 plants suggests that the slope of the relationship between δp and the relative contributions of dark and light fixation is sensitive to the environment (Winter & Holtum, 2005). The least negative values observed in bromeliads and orchids coincide with the value of −8.9‰ predicted for 100% CAM in Fig. 8(a), demonstrating that the isotopic signal representing full CAM is less sensitive to the environment than is the C3 isotopic signal, or that the environments inhabited by plants with full CAM share similarities.

An example of the different sensitivities of C3- and CAM-type isotopic signals to the environment can be seen in bromeliads collected from different altitudes. Between sea level and 5000 m, δp of bromeliads with values more negative than −20‰ became less negative by 1.4‰ km−1 (D. Crayn et al., unpublished). By contrast, bromeliads with CAM-type δp (less negative than −20‰) exhibited no significant δp change with increasing altitude.

The distribution of δp values in taxa known to contain C3 and CAM species is bimodal, with a peak containing strong CAM species c. −12 to −16‰, a peak containing C3 and weak-CAM species c. −24 to −32‰, and only few species in between (Kluge et al., 1991; Pierce et al., 2002; Winter & Holtum, 2002; Crayn et al., 2004; Silvera et al., 2005, 2010), as shown in Fig. 8(a). This bimodality implies that both strong CAM and weak CAM are selected for and that intermediate behaviour is not favoured. Bimodality may reflect biochemical optimization and anatomical tradeoffs between the C3 and CAM pathways (Nelson & Sage, 2008; Borland et al., 2011). Since δp is a time-integrated signal, the isotopic value alone does not distinguish between the continuous contribution of a weak CAM signal and the short-term contributions of strong CAM to a mainly C3-type signal. The latter pattern could occur in some tropical species of Clusia with facultative CAM (Holtum et al., 2004; Winter et al., 2008, 2009). A wide spectrum of δp is observed in annuals such as Mesembryanthemum crystallinum (Winter et al., 1978) and Calandrinia polyandra (Winter & Holtum, 2011), which undergo a gradual seasonal shift from C3 to CAM. In M. crystallinum, δp ranges from c. −27 to −14‰ (Winter et al., 1978; Bloom & Troughton, 1979). The least negative value probably underestimates the contribution of dark fixation to carbon gain at the end of the growing season as a result of dilution by previously fixed carbon with more negative δ13C.

Overall, the information embedded in the ΔCAM signal is rich. Isotopic surveys have demonstrated that the CAM pathway is much more widespread among species than was considered previously (Rundel et al., 1979; Winter, 1979; Winter et al., 1983; Earnshaw et al., 1987; Kluge et al., 1991; Smith & Winter, 1996; Zotz & Ziegler, 1997; Silvera et al., 2005, 2010). ΔCAM has provided characters for phylogenetic analyses (Crayn et al., 2004; Silvera et al., 2009), established the contributions of CO2 fixation in the light and dark, and provided information about transpiration ratios (Fig. 8c) and palaeoclimates (English et al., 2007). Measurements of instantaneous and short-term ΔCAM have been fundamental in establishing our physiological and biochemical concepts of CAM (O'Leary & Osmond, 1980; Holtum et al., 1983, 1984; O'Leary et al., 1986; Griffiths et al., 1990, 2007; Griffiths, 1992; Roberts et al., 1997). Future research should focus on further understanding the processes that may influence the relationship between ΔCAM and the relative engagement of the CAM cycle; for example, the extent to which leakage of CO2 from photosynthetic CAM tissues during the daytime occurs and may alter ΔCAM.

VI. Conclusion

Stable carbon isotope ratios of terrestrial plants have the potential to provide unique insights into physiological processes and interactions between plants and the environment. Continued research into environmental and physiological determinants of Δ will further increase this potential. For C3 plants, a priority should be to understand the mechanisms that lead to coordination between photosynthetic capacity and stomatal conductance, thereby muting the response of Δ3 to the environment under some conditions. This is also important for engineering plants that have both high water-use efficiency and high photosynthetic capacity. For C4 plants, future research should focus on understanding the interactions between ϕ and ci/ca, and when and to what extent Δ4 reflects these determinants. For CAM plants, future studies on how growth environment influences the relationship between δp and the relative contribution of dark and light CO2 fixation will improve the interpretation of δp signals in the field. Insights will likely emerge as a more refined understanding of ΔCAM is pursued, for example the role of daytime leakage of nocturnally fixed CO2 in influencing ΔCAM. The enhanced capacity of laser systems for high-throughput analyses of online Δ will continue to advance our understanding of the more subtle determinants of Δ, for example, the distribution of mitochondria in mesophyll cells in relation to chloroplasts in C3 plants (Tholen et al., 2012; Busch et al., 2013). Across the full range of applications that employ Δ measurements, it will be helpful to maintain a holistic view of how both the environment and internal physiology influence the Δ signal.


L.A.C. was supported by a Future Fellowship from the Australian Research Council (FT100100329). N.U. and G.D.F. also acknowledge support from the Australian Research Council (DP1097276).