Reduction of transpiration and altered nutrient allocation contribute to nutrient decline of crops grown in elevated CO2 concentrations

Authors


J. McGrath. E-mail: jmcgrath@illinois.edu

ABSTRACT

Plants grown in elevated [CO2] have lower protein and mineral concentrations compared with plants grown in ambient [CO2]. Dilution by enhanced production of carbohydrates is a likely cause, but it cannot explain all of the reductions. Two proposed, but untested, hypotheses are that (1) reduced canopy transpiration reduces mass flow of nutrients to the roots thus reducing nutrient uptake and (2) changes in metabolite or enzyme concentrations caused by physiological changes alter requirements for minerals as protein cofactors or in other organic complexes, shifting allocation between tissues and possibly altering uptake. Here, we use the meta-analysis of previous studies in crops to test these hypotheses. Nutrients acquired mostly by mass flow were decreased significantly more by elevated [CO2] than nutrients acquired by diffusion to the roots through the soil, supporting the first hypothesis. Similarly, Mg showed large concentration declines in leaves and wheat stems, but smaller decreases in other tissues. Because chlorophyll requires a large fraction of total plant Mg, and chlorophyll concentration is reduced by growth in elevated [CO2], this supports the second hypothesis. Understanding these mechanisms may guide efforts to improve nutrient content, and allow modeling of nutrient changes and health impacts under future climate change scenarios.

INTRODUCTION

A significant fraction of the human population is currently deficient in one or more essential dietary nutrients, even more than the roughly one billion who are defined as food insecure based on calorie consumption. Current estimates indicate that 25% of the global population is iron deficient (http://www.who.int/vmnis), and 31% is zinc deficient (Ezzati et al. 2004). Added to this current challenge is the fact that atmospheric concentrations of [CO2] continue to increase, and experimental data show that even though yield is often improved by growth in elevated [CO2], concentrations of most nutrients are decreased (Loladze 2002; Taub, Miller & Allen 2008). Given the prevalence of micronutrient deficiency and the near certainty of significant further increases in [CO2], concerns have been raised about potential implications for human health (Myers & Patz 2009).

Several studies find that growth in elevated [CO2] decreases nitrogen (N) and mineral concentrations, while increasing carbohydrate concentrations. Nitrogen concentrations are decreased in the grain of wheat, barley and rice (Manderscheid et al. 1995; Kimball 2006; Kobayashi et al. 2006). The response appears to vary depending on the ability of the crop to fix nitrogen, with N concentrations in non-legumes decreasing between 10 and 14% at elevated (540–958 ppm) CO2, and N concentrations in soybean (a legume), decreasing by only 1.5% (Taub et al. 2008). Changes in mineral nutrients are less well studied, and the response varies depending on the mineral and plant species, but several studies find 2–20% decreases in concentration for nutrients such as Mg, Zn and Fe in species including wheat, rice and barley (Manderscheid et al. 1995; Fangmeier et al. 1999). However, at least one study has found increased Ca concentration in rice (Seneweera & Conroy 1997).

While a growing body of literature has documented the experimental effect of elevated [CO2] on crop quality, there is still limited understanding of the mechanisms involved, and there are many biological and physical processes that are potentially affected by growth in elevated [CO2], including photosynthesis and solute translocation through the soil. The biological processes interact with the physical ones because plants have the ability to selectively acquire nutrients. In order to acquire a nutrient from the soil, it must be available in the rhizosphere. Because uptake by the root depletes nutrient concentration in the rhizosphere, nutrients must be translocated from elsewhere, or the rhizosphere must expand to intercept a greater volume of soil. Furthermore, plants do not appear to have perfect control of uptake, because even in cases where translocation exceeds physiological requirements, uptake is still correlated with translocation (Gilliham et al. 2011). Therefore, altering translocation of nutrient to the roots would likely affect nutrient uptake.

Two means by which nutrients can be translocated to the rhizosphere are mass flow and diffusion. Mass flow refers to the movement of water and solutes through the soil because of differences in water potential. In plants, mass flow is driven by transpiration, which draws water through the roots and releases it into the atmosphere; thus, the mass flow of water is equal to the amount of water transpired. Mass flow of a nutrient also depends on its solubility in the soil water solution and can be calculated as the nutrient concentration multiplied by the volume of water transpired. The solubility of a substance depends on its interaction with other substances in solution and with soil particles, and solubility of nutrients can vary greatly. Some nutrients, such as Fe, are generally found in insoluble complexes in soil water and are translocated almost negligibly by mass flow because they do not move with water. In contrast, some nutrients, such as Ca are highly soluble in soil water and the amount translocated to the root by mass flow is often in excess of physiological demand. In contrast to mass flow, diffusion is the movement of a molecule by random motion from an area of high concentration to an area of low concentration. The rate of diffusion is determined by the difference in concentrations between the two areas and the diffusivity of the molecule, an experimentally determined constant related to the mass of the molecule and the structure of the medium through which the molecule is diffusing. Because solubility and diffusivity vary depending on the molecule, the relative contributions of mass flow and diffusivity to translocation to the root also vary depending on the molecule (Van Vuuren et al. 1997; Loladze 2002). Growth in elevated [CO2] decreases both stomatal conductance and transpiration at the leaf and canopy scales (Long et al. 2004; Ainsworth & Rogers 2007; Bernacchi et al. 2007; Leakey et al. 2009). Therefore elevated [CO2] could significantly affect mass flow while having small or negligible effects on diffusion.

All of the following have been proposed to explain the changes in nutrient concentrations: (1) dilution of non-carbon compounds by increased carbohydrate concentrations from enhanced photosynthesis (Poorter et al. 1997); (2) reduced mass flow of nutrients through the soil, ultimately resulting from decreased stomatal conductance and transpiration; (3) altered root architecture and physiology; (4) down-regulation of photosynthesis (points 2–4 are reviewed in Taub & Wang 2008); (5) inhibition of nitrate assimilation by decreased photorespiration (Bloom et al. 2002; Searles & Bloom 2003); and (6) altered requirements for minerals as protein cofactors or constituents in other organic complexes. Understanding the relative role of these mechanisms would improve our now limited ability to predict responses for different crops and regions other than those that have been experimentally tested. Furthermore, identification of important mechanisms could help guide breeding and agronomic strategies to adapt crops to higher [CO2].

In this analysis, we use the rich database of primary literature and meta-analysis to test the following two hypotheses explaining mechanisms of reduced nutrient concentrations in elevated [CO2]:

  • 1Reduced transpiration in elevated [CO2] reduces mass flow of nutrients to roots, reducing nutrient uptake.
  • 2Changes in physiological processes differentially affect proteins and their mineral constituents, thus altering allocation and possibly total demand of nutrients.

The rational for testing the first hypothesis is that growth in elevated [CO2] decreases both stomatal conductance and transpiration at the leaf scale (Long et al. 2004; Ainsworth & Rogers 2007; Bernacchi et al. 2007), and generally decreases transpiration at the canopy scale (see Leakey et al. 2009). Because soluble nutrients are translocated with the flow of water, it is therefore possible that observed reductions in nutrient concentrations in plants grown at elevated [CO2] are in part due to decreased transpiration.

The rational for testing the second hypothesis is that nutrient allocation within the plant may be altered by physiological changes that alter production of various compounds and therefore alter requirement for the nutrients used in those compounds. The observed correlation between Zn and protein concentrations (Cakmak, Pfeiffer & McClafferty 2010) suggests that this could play a role, because protein concentration decreases in elevated [CO2].

These hypotheses were examined by testing specific predictions. For the first hypothesis, we test the prediction that there is a greater proportional reduction by growth in elevated [CO2] of nutrients translocated mostly by mass flow compared with nutrients translocated mostly by diffusion. For the second hypothesis, we start by identifying a large pool of a single nutrient in only one or a few compounds that respond in the same direction to elevated [CO2]. Mg was identified a priori as a potential nutrient. Both ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco) and chlorophyll require Mg to function and between 6 and 25% of leaf Mg is in chlorophyll (Marschner 1995). In elevated [CO2], Rubisco decreases on a leaf area basis, and chlorophyll decreases on a leaf mass basis (Ainsworth & Long 2005). Because reductions in chlorophyll and Rubisco are leaf specific, we test the prediction that Mg concentrations should show larger reductions in leaves than other tissues. Furthermore, we test the prediction that within leaves, reduction of Mg concentration should be larger than reductions of other nutrients, but for other tissues, reductions of Mg and other nutrients should be similar.

METHODS

Data were compiled from a literature search of the ISI Web of Science citation database. Data were taken from tables or digitized using software (Demleitner et al. 2001), and for some studies, additional data were provided through contact with the authors, including the European Stress Physiology and Climate Experiment (ESPACE) data set from A. Fangmeier. Response variables recorded were the concentration of any mineral nutrient (e.g. Fe and Ca) and concentrations of protein or nitrogen. Variables that might affect the response to elevated [CO2] were also recorded, including species and tissue (e.g. leaves). For meta-analysis, an effect size that quantifies the relative response of the two treatments must be calculated. The effect size used here was the natural log of the response ratio (where the response ratio is the response in elevated [CO2]/response in ambient [CO2]; Hedges, Gurevitch & Curtis 1999) and data are reported as percentage change in elevated [CO2]. For studies that had multiple CO2 treatments, the elevated treatment with the concentration closest to 550 ppm was used. Observations from different treatments within the same experiments, for example different fertilizer applications, were assumed to be independent and included as separate observations in the analysis. For studies with time series, only the last observation was used.

The data set includes any species with available data, but here, we analyse only crop species because they are much more widely studied, and because we are primarily interested in how changes in nutrient quality may affect human nutrition. The data set also includes studies that report nutrient concentrations after removing the mass of non-structural carbohydrates. By removing non-structural carbohydrates, much of the effect that elevated [CO2] has on carbon uptake is removed. Thus, this value largely removes the effect of carbohydrate dilution. This can be a useful measure, but because it is a different measurement from the actual concentration, aggregating it with the other values would produce incorrect results. Hence, those values were excluded from the analyses. Although examining nutrient concentrations in low [CO2] studies would be equally useful, we did not find enough studies to perform a meta-analysis.

The relationship between transpiration and nutrient concentration in elevated [CO2] would most obviously be tested by correlating the nutrient concentration with mass flow at multiple CO2 concentrations, but these data are not available. However, another approach can be used by taking into account two other observations. First, mass flow is different for each nutrient because of differences in solubility. Because of this, one can compare the change in concentration of a nutrient with the change in mass flow using only two conditions. That is, instead of asking whether the concentration of Fe in the plant is correlated with transpiration at several CO2 concentrations, one can ask whether the change in Fe concentration in elevated compared ambient [CO2] is smaller than the change in Ca concentration. Second, atmospheric [CO2] should not change the nutrient concentration in the soil, so because mass flow is determined as the volume of water transpired multiplied by the concentration of the nutrient in the soil, the difference in mass flow between ambient and elevated [CO2] will be related to the absolute value of mass flow in ambient [CO2]. Therefore, change in mass flow can be estimated using the absolute value mass flow in ambient [CO2]. Together, these allow one to correlate change in nutrient concentrations when plants are grown in elevated [CO2] to mass flow of the nutrient in ambient [CO2], which is the approach used here.

In order to normalize values across studies, the amount of uptake provided by mass flow was calculated as the amount of nutrient in the plant divided by the mass flow of the nutrient (Supporting Information Table S2). Linear regression was used to correlate the response ratio to percentage of uptake acquired through mass flow and sums of squares were estimated using the following model:

image

where ln(RR) is the natural log of the response ratio; M is the percentage of uptake acquired through mass flow; Ci is level i of category C (either species or tissue); and M × Ci is the interaction between the two.

The meta-analysis was performed as described by Curtis & Wang (1998) using R statistical software (R Development Core Team 2009). Standard errors were not available from most studies. Therefore, an unweighted analysis was performed using resampling techniques. To test the transpiration hypothesis, regression analysis was used. To test for differences among levels of categories, observations were randomly reassigned to categories without replacement. The actual sums of squares estimate was compared with the distribution of resampled estimates to determine significance (Gurevitch & Hedges 1999). Both species and tissue showed significant variation among levels, so further analysis was performed; for each species and type of tissue (e.g. grain, root), a regression was performed between change in nutrient concentration and percentage acquired through mass flow, using the following model:

image

where β0 and β1 are the intercept and the slope, respectively (see Fig. 1 as an example). Confidence limits were estimated using bootstrapping by randomly reassigning observations to independent variables for 4999 iterations (Efron & Gong 1983). Bias-corrected 2.5 and 97.5 percentiles of the resampled estimates were used as 95% confidence limits (Adams, Gurevitch & Rosenberg 1997), but confidence limits for groups with fewer than five observations are not given because they were not enough unique combinations of observations to generate a reliable bootstrap distribution. Each level of a factor needed to have a range of mass flow estimates larger than 90 and have more than 10 observations to be included in the analysis.

Figure 1.

Example analysis for wheat grain using mass flow estimates from Gregory et al. (1979). The trend line is the predicted linear relationship using individual observations, but for simplicity, data are shown as means with standard errors. Values in parentheses are the number of observations for each nutrient.

Mass flow varies by soil type, soil water content, species and soil nutrient concentrations, but we found no studies that report both nutrient concentrations and mass flow. Thus, in order to perform the analysis, data from studies that measured the response of nutrient concentration to elevated [CO2] were paired with estimates of mass flow from a separate study. For nutrients supplied in excess of demand, values were set to 100%. Pairing mass flow estimates from a single study with all of the elevated [CO2] studies results in non-independence. To examine the effects of this, three independent estimates of mass flow were used (Gregory, Crawford & McGowan 1979; Clarkson 1981; Oliveira et al. 2010). By comparing results when using the three studies, it is possible to get some measure of how important estimates of mass flow are to the final result.

As a further means of accounting for non-independence, a more conservative test was used in which means and standard deviations for the change in nutrient concentration were calculated for each nutrient. Using those means and standard deviations, a weighted regression analysis was performed. Observations within each nutrient were resampled so that each nutrient had the same number of observations in each bootstrap run. This analysis is similar to a hierarchical nested design in which observations are nested within the type of nutrient.

For the allocation hypothesis, categorical analysis was used. Means for each category were determined, and confidence limits for parameters were calculated by randomly choosing studies within a category with replacement. Means were considered different from one another if their 95% confidence limits did not overlap one another, and were considered different from zero if the confidence limits did not overlap zero.

RESULTS

Of the 21 nutrients analysed, 17 showed a decrease in concentration when grown in elevated [CO2] when averaged over all observations. Of these, 12 were significantly lower (P ≤ 0.05, Fig. 2). Only four nutrients showed a mean increase in concentration, and none of those were significant. Of the 21 nutrients in the data set, between 7 and 12, depending on the source, had estimates for mass flow (Supporting Information Table S2), which were then used to examine the effect of transpiration on concentration changes.

Figure 2.

Response of nutrient concentration to growth in elevated [CO2]. Data are means with 95% confidence limits except for nutrients with fewer than five observations, for which bootstrapped confidence intervals are unreliable. The number of observations for each nutrient is given in parentheses.

Slopes and intercepts of the relationship between change in nutrient concentration and percentage of acquisition through mass flow were determined for the data set overall and for species and tissues individually (Fig. 3). In this analysis, the intercept is the change in nutrient concentration in elevated compared with ambient [CO2] when mass flow has no effect on nutrient translocation. The slope indicates the percentage reduction in nutrient concentration for each percent of uptake by which a nutrient is acquired through mass flow. A negative slope suggests that elevated [CO2] reduces mass flow and nutrient acquisition.

Figure 3.

Intercepts and slopes for the regression of change in nutrient concentration versus percentage acquired through mass flow. Regressions were performed on the whole data set or by breaking the data into levels of tissues or species. Estimates of mass flow from three studies were used for comparison (Gregory et al. 1979, square; Clarkson 1981, circle; Oliveira et al. 2010, triangle). Points are the parameter estimate using the original data. Error bars are bias-corrected 95% confidence limits from bootstrapping. Values in parentheses are the number of observations used for the three mass flow studies. These varied because the studies measured different nutrients.

Overall, the regressions between percent of uptake acquired through mass flow and change in nutrient concentration tended to result in negative slopes and intercepts, but estimates varied by species and tissue (Figs 1 & 3, Table 1). Three analyses were performed using estimates of mass flow from different studies (Gregory et al. 1979; Clarkson 1981; Oliveira et al. 2010) as a robustness check, with all three resulting in a negative slope and intercept for wheat (Triticum aestivum), but with variable results for other species. Among tissues, stems had a consistently negative slope, and the slope for grains was negative in two regressions, but slopes for tubers, stover and roots were statistically insignificant. Estimates of the intercept were negative for all tissues except tubers and roots, regardless of which study was used for mass flow estimates. Using the more conservative analysis, estimates were similar (Supporting Information Fig. S3), but only some of the differences were significant, and the slopes of species did not vary significantly from each other (Supporting Information Table S2).

Table 1. Between-group heterogeneity (Qb) for CO2 effect size versus percentage of uptake acquired through mass flow across species or portion of plant
Study usedkMass flowSpeciesSpecies × mass flow
  1. The analysis was performed three times for each categorical variable (mass flow or portion of plant), using estimates of mass flow from three different studies. * P < 0.10; ** P < 0.05; *** P < 0.001.

Categorical variable: species.
 Gregory11151.03***1.7***0.64**
 Clarkson15920.92***1.53***0.47
 Oliveira15630.76***1.63***0.96**
   Portion of plantPortion of plant ×mass flow
Categorical variable: portion of plant.
 Gregory11031.02***2.57***0.54**
 Clarkson15840.79***2.22***0.46
 Oliveira15540.74***2.05***0.77*

To evaluate the effects of allocational changes, concentrations of Mg were compared in different tissues. In tissues other than stems and leaves, the reduction of Mg concentration was similar to changes of other nutrients, but the Mg reduction in stems and leaves was larger than the reduction of any other nutrient in any tissue: approximately an 18–30% mean reduction compared with a 2–15% mean reduction (Fig. 4).

Figure 4.

Percentage change of Mg (circle) or all nutrients measured excluding Mg (square) for several tissues. Points are the estimate using the original data. Error bars are bias-corrected 95% confidence limits from bootstrapping. Values in parentheses are the number of observations for Mg or nutrients excluding Mg.

To compare our results with previous findings (Loladze 2002), mean changes of nutrients in grain and leaves were calculated (Supporting Information Figs S1 & S2). For this comparison, in order to replicate the previous methodology, for grain, only observations from wheat were included, for leaves, all species were included, and data corrected for non-structural carbohydrates were included. With the exception of a single nutrient, none of the mean changes were significantly different from previous findings.

DISCUSSION

Many nutrients showed a decline in concentration when grown in elevated [CO2] (Fig. 2), similar to findings from previous studies (Loladze 2002; Taub & Wang 2008). When compared directly with a previous study, the mean changes in concentration were very similar (Supporting Information Figs S1 & S2). These results suggest an overall decline in nutrient concentrations when plants are grown in elevated [CO2]. Several mechanisms to explain this decline have been proposed, and the role of dilution by enhanced carbohydrate production has been established (Taub & Wang 2008; Taub et al. 2008). In addition to dilution by increased carbohydrate content, the findings in this study suggest that two other mechanisms play a role in nutrient concentration decreases. Altered physiological requirements change partitioning of nutrients among tissues and possibly alter uptake of nutrient by the roots. In addition, decreased transpiration reduces flow of nutrients to the roots, decreasing uptake.

Here, we show a statistically significant correlation between percentage of uptake acquired through mass flow and reduction in nutrient concentration, supporting the hypothesis that reduced transpiration in elevated [CO2] contributes to the decline of nutrients. Although it has already been suggested that reduced transpiration is a likely cause for decreased nutrient concentrations in elevated [CO2] (Taub & Wang 2008), this is the first direct analysis of that hypothesis.

No published studies that simultaneously examined mass flow and nutrient concentration for plants grown in elevated [CO2] were available. Therefore, testing the transpiration hypothesis required pairing mass flow data from studies unrelated to elevated [CO2], which results in non-independence among studies. Furthermore, properties of plants and soils can influence mobility. To examine the effects of non-independence and imprecise estimates, we used nutrient uptake estimates from three independent studies as a robustness check. Although some mass flow estimates were nearly identical across the three studies, such as for P which was almost entirely provided by diffusion, estimates for other nutrients varied more across studies. Thus, seeing a similar correlation between mass flow and nutrient response to elevated [CO2] regardless of the mass flow data used indicates a real response that is large enough to be seen even if precise data are not available. Similarly, comparing the results from the three studies can be used to determine the effects of non-independence from pairing mass flow estimates from a single study with all of the elevated [CO2] studies. The three studies often produce similar estimates when used in the regression. This result indicates the problem of non-independence may not be too important because the relative contribution of mass flow to nutrient uptake is likely similar across studies (e.g. Ca is usually transported by mass flow more than Fe), even though the absolute values may be different. The more conservative analysis produces very similar trends, but with most estimates closer to zero and with greater variance. The similar results and conservative nature of this analysis suggest that reduced mass flow does contribute to reduced nutrient concentrations, but the effect is difficult to see without measuring mass flow directly. The only way to properly deal with this issue is to simultaneously measure mass flow and nutrient concentration in elevated [CO2]. However, no studies have yet done that, and this method provides an alternative means of examining an important and unanswered question using existing data.

The estimates for wheat were similar regardless of the estimates used for mass flow. For other species, there was no clear correlation, partly because of only a few studies or many studies of only a limited number of nutrients. The high variability for most species makes it hard to determine whether some species respond differently to elevated [CO2]-induced changes in mass flow, but the response of wheat supports the hypothesis that decreases in transpiration have a role in decreasing nutrient content.

Dilution by enhanced production of carbohydrates is already known to decrease nutrient concentration in elevated [CO2] independent of effects of mass flow (Poorter et al. 1997). In this analysis, the intercept can be used to examine the effects of elevated [CO2] on nutrient concentration that are independent of effects of mass flow, because the intercept is the effect size for nutrients that are not transported by mass flow. With the exception of two estimates, all of the intercepts were negative, often significantly. This indicates that concentrations of nutrients that are not transported by mass flow are reduced in elevated [CO2]. However, the intercept is also an indication of the smallest reduction in nutrients one could expect if it were possible to maintain the same rates of mass flow in elevated compared with ambient [CO2], in this case an approximately 9% reduction overall.

Although previous studies found correlations between water use and ash content (a measure of nutrient content) when differences between water use were due to genetic variation (Masle, Farquhar & Wong 1992), other studies fail to support the transpiration hypothesis, showing large nutrient reductions with minimal transpiration changes, or large transpiration changes with minimal nutrient reductions (Muenscher 1922; Wong 1979; Tanner & Beevers 1990; Conroy 1992; Rudmann, Milham & Conroy 2001). Therefore, it had been concluded that altering mass flow by altering environmental conditions has little impact on plant nutrient acquisition. However, these studies did not account for the fact that different nutrients will be affected differently by changes in mass flow. Thus, they may not have the same power as the analysis used here. Furthermore, many previous studies use hydroponics or potted plants in greenhouses in which water dynamics may be altered. The results from this study suggest that there are cases in which there is a relation between transpiration and nutrient acquisition, and point to a need to reevaluate this topic using more realistic environmental conditions and different analyses.

The larger decrease of Mg in stems and leaves compared with other tissues, and the similar responses of other nutrients across all tissues suggest that differences in physiology are altering the demand for Mg. To our knowledge, this finding has not been demonstrated before, but could have important dietary consequences if the edible portion of the plant is affected. Whether other minerals are altered by such a mechanism is an important question, but only Mg in leaves was examined here. The allocational response of a given mineral is dependent on where and how it is used in the plant and whether those characteristics change in elevated [CO2], so it seems unlikely that every mineral would show a response, and determining the contribution of this mechanism requires examining each mineral individually. Furthermore, from a physiological standpoint, it should be noted that these results do not address whether plants alter nutrient uptake through the roots in response to physiological changes. They address only whether differences in physiology alter allocation among tissues.

Without biochemical analysis, the physiological changes resulting in decreased Mg concentration are unknown, but Rubisco and chlorophyll are likely causes for the decrease, because they both require Mg, and concentrations of both are reduced by growth in elevated [CO2] (Ainsworth et al. 2002; Ainsworth & Long 2005). Estimating the decrease in Mg using observations of the average decrease in chlorophyll concentration can be useful for determining whether changes in chlorophyll concentration could account for the observed decrease of Mg. In this study, the average decrease in Mg concentration in leaves was about 20% and for other nutrients combined was 10%. Using the simplifying assumption that the 10%-decrease represents dilution (which is not completely true, so this estimate is only approximate), and the 20% reduction of Mg is from both reduction of content and dilution, then there was an 11% reduction of Mg content in leaves (one cannot simply take the difference because concentration involves division). Based on the mass flow experiment in this study, between 3 to 10% of Mg reduction is caused by reduced transpiration, which leaves 1 to 8% due to other factors. Based on the 17% average decrease of chlorophyll concentration (on a mass per mass basis) in elevated [CO2] (Ainsworth & Long 2005) and 6 to 25% of leaf Mg bound to chlorophyll (Marschner 1995), there is a 1 to 5% decline of leaf Mg concentration in elevated [CO2]. Thus, a 1 to 4.25% reduction of Mg caused by reduction of chlorophyll content could reasonably account for most of the 1 to 8% reduction of leaf Mg concentration seen here.

We had predicted that Mg would show a larger decrease in leaves than other tissues, but the Mg decrease in stems was also large. The similar response of stems and leaves may seem unusual, and seemingly contrary to our prediction. However, almost all of the observations of stems in this data set are from wheat due to the inclusion of the ESPACE data set, which has many observations from different sites and years, and nearly half of the observations in that data set are from stems. In wheat, stems are photosynthetically active, with photosynthetic rates of some cultivars approaching one-fourth of rates in the flag leaf (Rawson & Evans 1971). Thus, wheat stems likely also have a large pool of Mg that would be expected to decrease in elevated [CO2].

Determining the relative contributions of these mechanisms to reduction of nutrients would help prioritize approaches to maintain or improve nutrient concentrations in elevated [CO2]. For nutrients that move almost entirely by diffusion, reductions in mass flow would not contribute to reduced concentrations. Using the estimate here of 0.1% reduction for each 1% of uptake acquired through mass flow (the slope of the overall response in Fig. 3), nutrients entirely provided by mass flow have an additional 10% reduction on top of the approximately 9% reduction (based on the intercept) from other factors. Thus, mass flow contributes to about half of the reduction for those nutrients. Using the same assumptions about Mg concentrations mentioned earlier, it is estimated that 50% of the reduction in leaves is due to changes in allocation. Without knowing the total uptake and allocation among tissues, this value is only a rough estimate, but it indicates that in some cases, changes in allocation can explain a large portion of the change in nutrient concentrations. This cannot be generalized to all situations, however, because the relative contributions of these factors will vary with growing conditions, since for instance, as Mg availability decreases, the proportion of Mg in chlorophyll increases (Marschner 1995), which could increase the relative importance of nutrient allocation.

Most of these data are from open-top chamber, growth cabinet and glasshouse experiments. These environments likely have artificial conditions that alter responses and can lead to incorrect conclusions (Arp 1991; McConnaughay, Berntson & Bazzaz 1993; McLeod & Long 1999; Long et al. 2004, 2006). An issue concerning the transpiration hypothesis is altered water dynamics in pots. Because of the small volume, all nutrients supplied to the plant are easily available to the roots, regardless of whether transpiration is reduced. However, this would be expected to decrease any correlation between mass flow and nutrient concentration, making it harder to see a relationship if there was one, but not affecting conclusions if there were no relationship. Free-air gas concentration enrichment (FACE) studies with plants rooted in the ground would likely give better estimates because roots would have unrestricted volumes, and water dynamics would be nearly identical to typical field conditions. There are, however, few FACE studies with measurements of plant nutrient concentrations or mass flow.

ACKNOWLEDGMENTS

We thank Elizabeth Ainsworth for helpful comments while writing the paper. Funding for this work was provided by the Rockefeller Foundation. The authors have no conflicts of interest to declare.

Ancillary