Mycorrhizal association as a primary control of the CO2 fertilization effect

See allHide authors and affiliations

Science  01 Jul 2016:
Vol. 353, Issue 6294, pp. 72-74
DOI: 10.1126/science.aaf4610

Fungi relieve nitrogen limitation

Rising concentrations of atmospheric CO2 stimulate plant growth; an effect that could reduce the pace of anthropogenic climate change. But plants also need nitrogen for growth. So far, experimental nitrogen addition has had equivocal effects on the magnitude of CO2 fertilization. Terrer et al. explain that the impact of nitrogen on plant growth depends on the relationship between nitrogen availability and symbioses with mycorrhizal soil fungi. Only plants with ectomycorrhizal fungi associated with their roots can overcome nitrogen limitation.

Science, this issue p. 72


Plants buffer increasing atmospheric carbon dioxide (CO2) concentrations through enhanced growth, but the question whether nitrogen availability constrains the magnitude of this ecosystem service remains unresolved. Synthesizing experiments from around the world, we show that CO2 fertilization is best explained by a simple interaction between nitrogen availability and mycorrhizal association. Plant species that associate with ectomycorrhizal fungi show a strong biomass increase (30 ± 3%, P < 0.001) in response to elevated CO2 regardless of nitrogen availability, whereas low nitrogen availability limits CO2 fertilization (0 ± 5%, P = 0.946) in plants that associate with arbuscular mycorrhizal fungi. The incorporation of mycorrhizae in global carbon cycle models is feasible, and crucial if we are to accurately project ecosystem responses and feedbacks to climate change.

Terrestrial ecosystems sequester annually about a quarter of anthropogenic carbon dioxide (CO2) emissions (1), slowing climate change. Will this effect persist? Two contradictory hypotheses have been offered: The first is that CO2 will continue to enhance plant growth, partially mitigating anthropogenic CO2 emissions (1, 2), whereas the second is that nitrogen (N) availability will limit the CO2 fertilization effect (3, 4), reducing future CO2 uptake by the terrestrial biosphere (57). Plants experimentally exposed to elevated levels of CO2 (eCO2) show a range of responses in biomass—from large and persistent (8, 9), to transient (6), to nonexistent (10)—leaving the question of CO2 fertilization open. Differences might be driven by different levels of plant N availability across experiments (11), but N availability alone cannot explain contrasting results based on available evidence (7, 12). For instance, among two of the most studied free-air CO2 enrichment (FACE) experiments with trees, eCO2 enhanced biomass production only during the first few years at Oak Ridge National Laboratory (ORNL)–FACE (6), whereas trees in the Duke University FACE experiment showed a sustained enhancement during the course of the experiment (8), despite N limitation. In addition to N limitation, other factors have been suggested as potential drivers of the response of plant biomass to eCO2: age of the vegetation (13), water limitation (14), temperature (15), type of vegetation (12), or even the eCO2 fumigation technology used (11). Although these factors may explain some observations, none has been found to be general, explaining the range of observations globally.

About 94% of plant species form associations with mycorrhizal fungi, an ancient mutualism thought to have facilitated the colonization of land by early plants (16). In this mutualism, the fungus transfers nutrients and water to the plant in exchange for carbohydrates, which are necessary for fungal growth. Mycorrhizal fungi are critical for terrestrial C cycling (17); are known to influence plant growth (18), nutrient cycling (19, 20), and soil carbon storage (21); and respond strongly to elevated CO2 (22, 23). Yet, their impact on the N-dependence of the CO2 fertilization effect has not been tested, despite the increasing evidence that N limitation constrains the CO2 fertilization effect (5). Arbuscular mycorrhizae (AM) and ectomycorrhizae (ECM) are by far the most widespread types of mycorrhizae (24): AM-plants predominate in deserts, grasslands, shrublands, and tropical forest ecosystems, whereas ECM-fungi predominate in boreal and many temperate forests (for example, those dominated by Pinus). ECM can transfer N to the host plant under eCO2 to sustain CO2 fertilization (25), whereas the symbiotic effects of AM fungi in N-limited systems can range from beneficial to parasitic (19). Hence, the association of Liquidambar styraciflua with AM-fungi at ORNL-FACE, and Pinus taeda with ECM-fungi at Duke-FACE, might explain why only trees in the latter could increase N-uptake and take advantage of eCO2 to grow faster for a sustained period (20, 25). We tested the hypothesis that the differences in the nutrient economies of ECM and AM fungi influence global patterns of the magnitude of plant biomass responses to elevated CO2.

We synthesized data (overview is provided in table S1) on total plant biomass (grams per square meter) from 83 eCO2 experiments (fig. S1), separating responses into aboveground biomass (n = 83) (fig. S2) and belowground biomass (n = 82) (fig. S3) in a mixed-effects meta-analysis. As potential drivers of the plant biomass response, we considered the increase in atmospheric CO2 concentration (∆ΔCO2), mean annual precipitation (MAP), mean annual temperature (MAT), age of the vegetation at the start of the experiment, vegetation type (such as grassland or forest), CO2 fumigation technology (such as FACE or growth chamber), length of the study (years), dominant mycorrhizal type (AM or ECM), and N-status [high or low N availability, considering soil characteristics and occasional fertilizer treatments, following the approach by Vicca et al. (17) and assigning all experiments with indications for some degree of N limitation to the “low N” class and experiments that were unlikely N limited to the “high N” class] (supplementary materials, materials and methods, and table S2).

Model selection analysis, based on corrected Akaike Information Criterion (AICc), showed that the most parsimonious model within two AICc units included N-status, mycorrhizal type, and ∆ΔCO2 (P < 0.001). The relative importance of the predictors (Fig. 1) supported the removal of climate variables, length of the experiment, age of the vegetation, fumigation technology, and system type. Some predictors reduced the CO2 effect on biomass (such as age of the vegetation), whereas others were associated with an increased CO2 effect (such as ECM, ∆ΔCO2, and high N availability) (fig. S4).

Fig. 1 Model-averaged importance of the predictors of the CO2 fertilization effect on total biomass.

The importance is based on the sum of Akaike weights derived from model selection using AICc (Akaike’s Information Criteria corrected for small samples). Cutoff is set at 0.8 (dashed line) in order to differentiate among the most important predictors.

The response of total biomass to an increase of CO2 from 400 to 650 μmol mol−1 was larger (P < 0.001) in ECM (30 ± 3%, P < 0.001) than in AM-dominated (7 ± 4%, P = 0.089) ecosystems (mean ± SE, mixed effects metaregression). The overall response of total biomass was 20 ± 3% (P < 0.001), which is similar to previous meta-analyses (15), with a larger effect under high (27 ± 4%, P < 0.001) than low N availability (15 ± 4%, P < 0.001), as expected (5, 7, 11). Furthermore, we found a strong interaction between mycorrhizal type and N-status (P < 0.001); under low N availability, eCO2 had no effect on total biomass of AM-dominated species (0 ± 5%, P = 0.946) but increased biomass by 28 ± 5% in ECM-dominated species (P < 0.001) (Fig. 2A). Under high N availability, the CO2 effect on total biomass in both AM- and ECM-dominated species was significant: 20 ± 6% (P = 0.002) for AM and 33 ± 4% (P < 0.001) for ECM (Fig. 2A), with no significant differences between the two groups (P = 0.139). Hence, high N availability significantly increased the CO2 effect in AM [post-hoc, Tukey’s honestly significant difference (HSD): adj-P = 0.038] but not in ECM-associated species (adj-P = 0.999).

Fig. 2 Overall effects of CO2 on plant biomass.

(A to C) Effects on (A) total, (B) aboveground, and (C) belowground biomass for two types of mycorrhizal plants species (AM and ECM) in N-limited experiments (low N) or experiments that are unlikely N-limited (high N). Overall means and 95% confidence intervals are given; we interpret CO2 effects when the zero line is not crossed.

The patterns observed for total biomass were reflected in both aboveground and belowground biomass. Under low N availability, eCO2 stimulated aboveground biomass significantly in ECM plants (P < 0.001), with no effect in AM plants (P = 0.584) (Fig. 2B). Similarly, eCO2 enhanced belowground biomass in ECM plants at low N (P = 0.003) but not in AM plants (P = 0.907) (Fig. 2C).

We conducted a sensitivity analysis to ensure that the findings were robust. First, we added an intermediate level of N availability (table S2) by assigning some ecosystems that were initially classified as “low” to a “medium” class (for example, Duke, Aspen, and ORNL) (fig. S5). This enabled testing whether the large CO2 stimulation in ECM plants was driven by experiments with intermediate N availability. Second, we weighted individual experiments by the inverse of the mixed-model variance (fig. S6) so as to ensure that the weights of the meta-analysis did not affect the outcome. Third, we ran a separate meta-analysis with the subset of experiments with trees only (fig. S7). Previous meta-analyses have reported that trees are more responsive to eCO2 than are grasslands (12); as such, our findings could reflect differences of plant growth form rather than mycorrhizal association per se. Because trees are the only type of vegetation that can associate with ECM and AM (or both), an analysis of tree responses to eCO2 can thus be used to isolate the influence of mycorrhizal type from that of vegetation growth form. These three sensitivity analyses confirmed that the CO2 stimulation of total and aboveground plant biomass was significant and large in ECM plants regardless of N availability, whereas the effect was not significant in AM plants under low N availability. The trend was consistent for belowground biomass in ECM plants, although with high variance and low sample size, the effect was not significant (P = 0.244) under low N when the “medium” class was included.

Plant N uptake can be enhanced through mycorrhizal associations or through associations with N-fixing microbes. Some of the CO2 experiments in our study contained N-fixing species, which might have increased N availability (table S3). eCO2 stimulated aboveground biomass in AM species under low N by 8 ± 3% (P = 0.019) in this subgroup of experiments that included N-fixing species, whereas the remaining AM experiments under low N availability showed no biomass response to eCO2 (1 ± 10%, P = 0.893). But even with the additional N input from N2 fixation, the 8% biomass increase in AM plants under low N was considerably smaller than the 28 ± 5% increase found for ECM plants.

Most CO2 experiments have been carried out in the Northern Hemisphere (fig. S8), where N, rather than phosphorus (P), is limiting. AM fungi transfer large quantities of P to the plant and hence are more likely mutualistic in P-limited ecosystems (19). Tropical forests are typically associated with P limitations and dominated by AM-fungi and could potentially show enhanced biomass under eCO2. The role of nutrients on the CO2 fertilization effect in these P-limited forests has yet to be explored (26).

Responses of plants to rising CO2 are thus well explained by a simple interaction between N and microbial mutualists: When N availability is limited, only plant species that associate with ECM-fungi show an overall biomass increase due to eCO2. Several mechanisms could explain these responses. First, ECM-associated plants typically allocate more C to support mycorrhizae than do AM plants, particularly under eCO2 (23). Moreover, because ECM fungi, unlike AM fungi, produce extracellular enzymes that degrade organic N compounds (27), increased allocation to ECM fungi under eCO2 may supply host plants with the N needed to sustain their growth response to eCO2. This may explain why eCO2 often stimulates priming effects in ECM-dominated ecosystems (28, 29). Second, differences in litter quality between ECM and AM plants may influence how much N is available to be primed or decomposed. Several studies have reported that AM plants produce litters that decompose faster than those of ECM plants (20, 30). Given emerging evidence that fast decomposing litters promote the formation of stable mineral-associated organic matter (31, 32), much of the organic N in AM-dominated ecosystems may be inaccessible to AM plants or their associated mycorrhizae (20). And whereas slow-degrading ECM litters may reduce N availability in the short term, most of the N exists in particulate forms, which should be accessible to most microbes (including ECM fungi). Therefore, AM fungi are equipped with less specialized enzymes for N acquisition than are ECM and occur in soils in which N is more tightly protected. Both factors would presumably limit the enhancement of AM plant growth in response to eCO2.

Mycorrhizal symbioses are not accounted for in most global vegetation models (24). Thus, the projected CO2 fertilization effect by “carbon-only models” (1) is likely overestimated for AM-dominated ecosystems, which cover ~65% of the global vegetated area (24), albeit only when N-limited. On the other hand, global models that consider N limitation to constrain the CO2 fertilization effect (4) likely underestimate responses of ECM plants to eCO2, an area that encompasses ~35% of the vegetated area of the earth (24), most of which is considered N-limited by these models. Our framework reconciles the apparent discrepancy between widespread N limitation (3), which is assumed to limit C sequestration on land (4), and the observed increase over time of the terrestrial C sink (1, 2), which is thought to be driven primarily by CO2 fertilization (33). These results may also partly explain past findings that forests (commonly ECM) show stronger responses to eCO2 as compared with grasslands (AM) (12). We propose that the CO2 fertilization effect be quantified on the basis of mycorrhizal type and soil nitrogen status, and that large-scale ecosystem models incorporate mycorrhizal types to account for the differences in biomass enhancement by eCO2. Mycorrhizae are ubiquitous and sort predictably with plant functional type (24, 34), making feasible their inclusion in models to capture this microbial influence on global biogeochemistry. Accounting for the influence of mycorrhizae will improve representation of the CO2 fertilization effect in vegetation models, which is critical for projecting ecosystem responses and feedbacks to climate change.


Materials and Methods

Figs. S1 to S8

Tables S1 to S4

References (35139)


  1. Acknowledgments: We thank A. Talhelm, A. Finzi, L. Andresen, I. Kappel, C. Calfapietra, B. Sigurdsson, J. Dukes, P. Newton, D. Blumenthal, B. Kimball, J. Heath, P. Reich, R. Norby, C. Körner, N. Chiariello, C. Field, and M. Schneider, who provided additional data and advice. This research is a contribution to the Imperial College initiative Grand Challenges in Ecosystems and the Environment and the AXA Chair Programme in Biosphere and Climate Impacts. C.T. is supported by an Imperial College Ph.D. studentship within this program. C.T. and S.V. acknowledge support of ClimMani COST Action (ES1308). S.V. is a postdoctoral fellow of the Research Foundation–Flanders (FWO) and acknowledges support from the European Research Council grant ERC-SyG-610028 IMBALANCE-P. B.A.H. was supported by the Biological and Environmental Research program, Office of Science, U.S. Department of Energy (DOE) grant DE SC0008270. R.P. acknowledges support from NSF (Ecosystem Studies Program 1153401) and DOE (Environmental System Science Program). R.P. and C.T. thank the Royal Netherlands Academy of Arts and Sciences, DOE, INTERFACE, and the New Phytologist trust for funding the Workshop “Climate models revisited: the biogeochemical consequences of mycorrhizal dynamics.” The data reported in this paper are available online as supplementary materials. C.T. conceived the initial idea, collected the data, and conducted the data synthesis and meta-analysis. S.V. developed the nitrogen classification. B.A.H. refined model selection, meta-analysis, and framework. All authors contributed to the development of the conceptual framework and to the writing of the manuscript.
View Abstract

Stay Connected to Science

Navigate This Article