globalchange  > 气候变化事实与影响
DOI: doi:10.1038/nclimate2682
论文题名:
Growth responses of a green alga to multiple environmental drivers
作者: Georgina Brennan
刊名: Nature Climate Change
ISSN: 1758-869X
EISSN: 1758-6989
出版年: 2015-06-15
卷: Volume:5, 页码:Pages:892;897 (2015)
语种: 英语
英文关键词: Climate-change ecology
英文摘要:

One feature of global change is that biota must respond not to single, but to multiple environmental drivers. By growing a model photosynthetic microbe in environments containing between one and eight different drivers, including changes in CO2, temperature, and pH, in different combinations, we show that the number as well as the identities of drivers explain shifts in population growth rates. This is because the biotic response to multiple environmental drivers depends on the response to the single dominant driver, and the chance of a driver of large effect being present increases with the number of drivers. Interactions between drivers slightly counteract the expected drop in growth. Our results demonstrate that population growth declines in a predictable way with the number of environmental drivers, and provide an empirically supported model for scaling up from studies on organismal responses to single drivers to predict responses to large numbers of environmental drivers.

A major challenge facing freshwater and marine biologists is to quantify how aquatic biota will respond to our changing climate. One of the hallmarks of global change is that it is complex; changes in temperature, pH, light levels, carbon dioxide and oxygen concentrations, nutrient availability, salinity and other environmental variables can occur together1, 2. Predicting the action of multiple environmental drivers (MEDs) on population growth is required for understanding how aquatic biota, at all levels from individual genotypes to communities, respond to global change3. Studies in freshwater3, 4 and marine systems5, 6 have historically focused on understanding organismal responses to key environmental drivers alone, such as changing temperature, CO2 levels, or light levels, or investigated MEDs by using pairs or trios of drivers1. This has shown that interactions between the effects of environmental drivers vary with the drivers and the organisms being tested6, but use a small number of environmental drivers relative to the number of drivers in most natural environments2. This leaves open the possibility that when the number of environmental drivers is larger, the effects of interactions between individual drivers may become less important in determining overall organismal responses. The goal of our study is to determine if knowing the interactions between specific environmental drivers at the organismal level is necessary when the number of environmental drivers is large, or whether patterns emerge that allow us to predict organismal responses without knowing particular driver interactions.

Studies on MEDs until now are mainly concerned with understanding interactions between the effects of individual drivers (see ref. 3 for definitions). Driver effects can either be additive, where the response to MEDs is equal to the sum of their individual effects, or multiplicative, where the response exceeds the sum of their individual effects. Interactions that are additive or multiplicative can be further synergistic (having a positive feedback) or antagonistic (having a negative feedback). Antagonistic interactions can thus lead to outcomes where responses to MEDs are less than the sum or product of their individual effects. These definitions must be contextualized in terms of the level of organization they affect, such as cellular processes or community composition. Driver interactions can be studied mechanistically, where the interactions are between drivers themselves (for example, the chemistry that links pH and CO2 levels), or be outcome-based and describe effects on organisms. Here, we use an outcome-based definition of drivers and driver interactions. We focus on the effects of drivers and interactions as population-level organismal responses.

Building an outcome-based prediction of biotic responses to MEDs by understanding specific interactions between key drivers requires that key drivers be identified and the interactions between them be measured. This approach is fruitful when the number of drivers is small. For example, high CO2 and low pH enhance the detrimental effects of ultraviolet irradiation on a key pelagic calcifier, Emiliania huxleyi7, and although many diatom assemblages do not respond to CO2 enrichment alone, CO2 and high light levels interact synergistically to reduce their growth rates8. These experiments can investigate the interactions between drivers, but are difficult to scale up, because measuring interactions between all drivers becomes impossible as the number of drivers increases. This is problematic, because these and similar studies on natural phytoplankton assemblages4, E. huxleyi7, 9, 10, 11, Phaeodactylum tricornutum12, and the freshwater alga Chlamydomonas reinhardtii13, suggest that interactions among drivers are not easily predicted even if they can be explained once observed. If this is the case, then one cannot use studies of pairs or trios of drivers to predict responses to those same pairs when many other drivers are also present (for example, if pH, CO2 and ultraviolet levels change alongside temperature, oxygen levels, and micronutrient levels). One way to reduce the size of experiments is to measure responses to groups of MEDs using combinations of drivers that are likely to change in concert2. This requires knowing how drivers group, and how these groups change on relevant geographic and temporal scales.

Alternatively, it may be possible to make reasonable outcome-based predictions of responses to MEDs based on the number of environmental drivers. Our general reasoning can be explained using environmental tolerance curves (Fig. 1), which usually show the relationship between some aspect of organismal function (for example, growth) and an environmental value (for example, temperature). Here, we consider a tolerance curve showing the relationship between organismal function and the total environment experienced by the organism in a multidriver environment. We assume that organismal function is initially somewhere on the plateau. Changes to one or more randomly chosen drivers will affect organismal function in some unknown way, either as a direct result of one driver, or as a result of interactions among drivers. If a subset of drivers or their interactions have large enough effects to push organismal function off the plateau, but the effects of most drivers are unlikely to be severe (as organisms are generally tolerant of some environmental variability), the chances of at least one of these large-effect drivers occurring grows as the number of drivers and interactions increases. This is analogous to altering organismal function through genetic mutations, where most mutations have small effects on organismal function, but mutations of large effect will eventually occur if enough mutations are sampled14. Here, we instead approach the idea that organismal phenotype is a result of interactions between genotype and environment using environmental ‘mutations rather than genetic ones.

Figure 1: Cartoon of the effects of multiple drivers on organismal function using an environmental tolerance curve.
Cartoon of the effects of multiple drivers on organismal function using an environmental tolerance curve.

Some aspect of organismal function, such as growth, is plotted as a function of the environment experienced by the organism, with the value of ‘Environment being determined by multiple environmental drivers. Initially, organismal function is high (solid black circle). When single drivers change, organismal function changes (patterned filled circles). Although the effect of each driver may be unknown, as an increasing number of drivers occur, the likelihood of at least one driver or driver interaction having a large detrimental effect on organismal function increases. This thought experiment does not require that the population be in its optimal environment, just that, among the environments sampled, the control environment be one where organismal function is high. Figure 2a shows that this is the case here, as the control environment is among the ‘best environments available in this experiment. Note that this cartoon is meant to illustrate our thought process, and not to indicate the quantitative effects of the specific environments used in this experiment. Please refer to Fig. 2 for quantitative data.

Population growth rate declines as the number of drivers in test environments increases (Fig. 2). We see that the number of drivers is the strongest predictor of population growth, explaining approximately 37% of the decrease in growth rate independently of the particular combination of environmental drivers involved, which is in line with our hypothesis that knowing the number of environmental drivers alone is informative (F1,93 = 11.1766, P = 0.001, Fig. 2a; see Supplementary Methods). Regime (the particular drivers in any unique test environment) explains some (32%) of the decrease in population growth rate in test environments, and the overlap in the environmental drivers between regimes also explains some (about 10%) of the variation in growth (F1,93 = 3.877, P = 0.052, Fig. 2a). As expected, extinction is more likely in test environments with a greater number of drivers (F1,93 = 3.310, P = 0.072, Supplementary Fig. 1).

Figure 2: Population growth rate of C. reinhardtii under zero to eight environmental drivers.
Population growth rate of C. reinhardtii under zero to eight environmental drivers.

a, Black data points and bars represent means and standard deviation between regimes for each NED. See Supplementary Table 3 for regimes. Different shapes within each NED indicate individual regimes. Dashed line in a indicates growth in the control environment. bd, Population growth rates (mean and standard deviation) predicted by a model (white triangles) alongside measured values (black circles), followed by goodness-of-fit, for three models: comparative model (r2 = 0.43, P < 0.0001) (b), multiplicative model (r2 = 0.33, P < 0.0001) (c) and additive model (r2 = 0.25, P < 0.0001) (d); extinction (indicated by dashed line in panels bd) is predicted in environments with >5 changes.

In our experiment, test environments become more similar as the number of drivers increases, although this similarity explains less than 11% of the variation in growth. If increases in environmental similarity were driving our results, we would expect that variation among regimes drop as the number of drivers within regimes increases, but this is not the case (correlation between the number of drivers and variance among regimes with the same number of drivers; post hoc fit r2 = 0.06, P = 0.53). Increasing environmental similarity with an increasing number of drivers per test environment is a limit of performing an experiment with a finite number of drivers. To understand how increasing environmental similarity affects our data, we simulated the same experiment using infinite environments with the same distribution of effects on growth for single environmental changes as in our experiment. We found that using a finite number of possible environmental changes in our experiment slightly underestimates growth rates in regimes with many drivers, but the effect is small (Supplementary Fig. 3), confirming that the increase in similarity between regimes with an increasing number of drivers does not explain the overall pattern of our data.

To understand how interactions between focal drivers change when additional drivers are present, we measured the effects on population growth of increased CO2, increased temperature and decreased pH—either alone, in pairs, all together, or all together in the presence of other drivers (Fig. 3). When these focal drivers occur singly, populations grow fastest under CO2 enrichment, slower under low pH, and slowest under high temperature. In pairs, the effect of CO2 enrichment counteracts that of high temperature so that these populations have higher growth rates than those under high temperature alone, whereas the combined effects of CO2 enrichment and low pH reduces growth. Populations grown in low pH and high temperature grow faster than those subjected to either driver alone, and populations subjected to all three drivers together grow faster than any of the paired or single cases. In these test environments, containing between one and three drivers, specific interactions between responses to drivers determine growth effects, and the most informative way to explain changes in growth is by investigating the physiological mechanisms involved47.

In contrast, when elevated CO2, low pH, or high temperature co-occur with other drivers, changes in population growth are predictable from the effects of single drivers. This prediction is more robust when a greater number of drivers are present in the test environments. For example, if CO2, pH and temperature change, decreasing light intensity does not affect growth further, as expected from the small effect of light intensity on growth alone relative to the effect of other drivers already present in the regime. In contrast, the presence of herbicide, which has a drastic effect alone, reduces growth when it is added to a test environment that already contains several other drivers. The addition of nutrient depletion has very little effect on growth and is masked by the dominant effects of herbicide.

These interactions are all expected under the simple comparative model. Interestingly, at high NED, phosphate limitation has an antagonistic interaction when herbicide is present. This is surprising, as both herbicide and phosphate are dominant environmental drivers. The herbicide used here is atrazine, which directly blocks the photosynthetic electron transport chain, reducing photosynthetic efficiency13. Phosphate is a limiting factor in many natural environments, yet it is a necessary macronutrient that photosynthetic organisms such as C. reinhardtii require in large amounts35. Previous work13 suggests protection mechanisms such as nonphotochemical quenching of excess light energy and adjustment of the photosystem stoichiometry to explain the antagonistic interaction observed between atrazine and very high light in C. reinhardtii and arrested growth with no loss in viability in low light conditions. Similar protection mechanisms may be in place here to protect populations from the lethal effects of atrazine under limited resources at high NED. Antagonistic interactions between phosphate depletion and other environmental changes have also been found in a nitrogen-fixing species (Trichodesmium48), where phosphate-limited populations are also CO2-limited, so that high CO2 can increase population growth rate when phosphate is limiting. Our case study broadly supports the observation that elevated CO2 can partly mitigate the drop in growth in test environments with MEDs, including phosphate-limited environments. However, we also find some exceptions where growth is not increased by high CO2, such as when nutrients and phosphate are co-limiting.

Although interactions between drivers increase variation in the reduced data set that excludes high-CO2 test environments (Supplementary Fig. 2) relative to the full data set shown in Fig. 2, the overall relationship between population growth and the number of drivers is the same. When many drivers co-occur, the effects of individual drivers, in particular of the driver with the single largest effect alone, are reasonable predictors of population growth. Our data also show that even if the individual effects of drivers on growth are unknown, the number of drivers offers a good estimate of the expected growth rate when large numbers of drivers co-occur. As with the full data set, this is due to test environments with a greater number of drivers having a higher chance of containing at least one severely detrimental driver so that, generally, growth decreases as the number of drivers increases.

Global change involves many environmental drivers, but biotic responses are often studied using few environmental drivers, so it is vital that we explore if and how studies using few environmental changes inform predictions of biotic responses to higher numbers of drivers. Mechanistically understanding all interactions between the relevant drivers in aquatic systems47 cannot be tackled experimentally—with current methods, full factorial experiments are simply too large to carry out. That being said, we can make a tradeoff between a mechanistic understanding of interactions between specific drivers and predicting overall biotic reactions to MEDs. One well-established way to do this is by using scenarios2 where suites of environmental variables are changed in concert and organismal responses measured. Here, we propose a complementary method suitable for situations where a larger number of drivers is considered, based on data showing that average changes in population growth in a model microalga are largely predictable from either the number of environmental drivers, or the effect of the single most detrimental driver, in cases where a large number of environmental drivers occur together. As with scenarios, our approach trades mechanistic understanding for predictive power. Although the ideal solution to understanding organismal responses to MEDs may be to replace ‘black box approaches such as ours with a mechanism-based understanding that allows prediction, this may not be realistic given current knowledge. Our approach is appropriate when constructing scenarios of environmental change carries significant uncertainty, because of uncertainty in predicting the intensities of individual drivers, of correlations between changes in drivers, or even in the identity of the particular drivers involved at the relevant geographical and temporal scales for focal organisms. It is also useful when data on responses to drivers or scenarios cannot be gathered for all organisms where it is needed. Another use of our method is in making between-species or between-genotype comparisons by uncovering differences in sensitivities to particular drivers. If the effect of many individual drivers is measured on different species or genotypes, then studies can be used to both understand differences in responses between species or genotypes, and to predict the likely range of responses to MEDs within communities containing many species or genotypes.

We show that specific interactions between drivers determine growth responses when only a few drivers change, but these interactions do not need to be taken into account to predict average growth responses when many drivers change. This provides hopeful evidence that continuing to build our understanding of how single drivers affect population growth is indeed informative for understanding population-level responses to MEDs.

Experimental design.

All populations were founded from a single cell of C. reinhardtii (CC-2931, mt-; Chlamydomonas Resource Center, University of Minnesota), grown in sterile Sueokas high salt medium, buffered with Tris-HCl (HSMT; ref. 39), under continuous rotation (50 r.p.m.) at 25 °C and constant light at 32 μmol m−2 s−1 photon flux density (Fisher Scientific Traceable Dual-range Light Meter), at 420 ppm CO2 (Supplementary Tables 1 and 2). These variables were controlled using incubators (Infors AG CH-4103). This strain of C. reinhardtii is from a culture collection, and has been grown in our lab for over seven years—this medium, temperature and light levels represent the usual benign growth conditions for this strain.

Experimental environments.

Experimental populations were grown for approximately three generations in replicate test environments that differed from the benign control environment (430 ppm CO2, pH 7.2, temperature 25 °C, full light and nutrients, no herbicide and no ultraviolet), by between one to eight of the following parameters: increased CO2 to 2,000 ppm, temperature to 26 °C, decreased pH to 6.5, light levels to 18 μmol m−2 s−1, reduced phosphate to 1.69 mM, general nutrient depletion by 75%, and added 0.5 μM of the herbicide atrazine. In addition, test environments with ultraviolet were exposed to a dose 8.1 kJ m−2 ultraviolet radiation once a week as part of the batch culture protocol (Supplementary Tables 1 and 2). There are 96 test environments in total in this study and 288 populations (3 independent replicate populations per test environment ×96 test environments, Supplementary Table 3). The large size of this experiment motivates using C. reinhardtii as the model alga, as it grows easily in small volumes in media that it is already adapted to that have sufficient buffering capacity to control pH when CO2 is varied, has a wealth of information available on responses to the individual drivers used in our study, and is a common model system in algal physiology and evolution. Cultures were grown in 48-well plates containing 1.6 ml of culture media. Each population was acclimated to its test environment for seven days (three generations), and then transferred to fresh test environment medium for each regime.

Details of how individual drivers were manipulated and our reasoning behind specific manipulations are below. In general, driver intensities were kept in line with future climate change scenarios where possible17, 49, 50, but modified to accommodate logistics, the starting point of the benign lab environment, the need that each driver affect grow

URL: http://www.nature.com/nclimate/journal/v5/n9/full/nclimate2682.html
Citation statistics:
资源类型: 期刊论文
标识符: http://119.78.100.158/handle/2HF3EXSE/4697
Appears in Collections:气候变化事实与影响
科学计划与规划
气候变化与战略

Files in This Item:
File Name/ File Size Content Type Version Access License
nclimate2682.pdf(559KB)期刊论文作者接受稿开放获取View Download

Recommended Citation:
Georgina Brennan. Growth responses of a green alga to multiple environmental drivers[J]. Nature Climate Change,2015-06-15,Volume:5:Pages:892;897 (2015).
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[Georgina Brennan]'s Articles
百度学术
Similar articles in Baidu Scholar
[Georgina Brennan]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[Georgina Brennan]‘s Articles
Related Copyright Policies
Null
收藏/分享
文件名: nclimate2682.pdf
格式: Adobe PDF
此文件暂不支持浏览
所有评论 (0)
暂无评论
 

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.