Warming and leaf litter functional diversity, not litter quality, drive decomposition in a freshwater ecosystem

Environment, litter composition and decomposer community are known to be the main drivers of litter decomposition in aquatic ecosystems. However, it remains unclear whether litter quality or functional diversity prevails under warming conditions. Using tank bromeliad ecosystems, we evaluated the combined effects of warming, litter quality and litter functional diversity on the decomposition process. We also assessed the contribution of macroinvertebrates and microorganisms in explaining litter decomposition patterns using litter bags made with different mesh sizes. Our results showed that litter decomposition was driven by litter functional diversity and was increasingly higher under warming, in both mesh sizes. Decomposition was explained by increasing litter dissimilarities in C and N. Our results highlight the importance of considering different aspects of litter characteristics (e.g., quality and functional diversity) in order to predict the decomposition process in freshwater ecosystems. Considering the joint effect of warming and litter traits aspects allow a more refined understanding of the underlying mechanisms of climate change and biodiversity shifts effects on ecosystem functioning.

Warming and leaf litter functional diversity, not litter quality, drive decomposition in a freshwater ecosystem Gustavo H. Migliorini 1* & Gustavo Q. Romero 2 Environment, litter composition and decomposer community are known to be the main drivers of litter decomposition in aquatic ecosystems. However, it remains unclear whether litter quality or functional diversity prevails under warming conditions. Using tank bromeliad ecosystems, we evaluated the combined effects of warming, litter quality and litter functional diversity on the decomposition process. We also assessed the contribution of macroinvertebrates and microorganisms in explaining litter decomposition patterns using litter bags made with different mesh sizes. Our results showed that litter decomposition was driven by litter functional diversity and was increasingly higher under warming, in both mesh sizes. Decomposition was explained by increasing litter dissimilarities in C and N. Our results highlight the importance of considering different aspects of litter characteristics (e.g., quality and functional diversity) in order to predict the decomposition process in freshwater ecosystems. Considering the joint effect of warming and litter traits aspects allow a more refined understanding of the underlying mechanisms of climate change and biodiversity shifts effects on ecosystem functioning.
Warming is predicted to profoundly impact biodiversity and ecosystems in all biomes on Earth 1 . Global warming is expected to alter the individual's metabolic rates 2 , distribution of species and their phenology 3 , the structure of communities 4 , species interactions 5 , and ecosystem functioning [6][7][8][9] . Concomitantly, changes in biodiversity are occurring at unprecedented rates 10 . Thus, understanding the effects of warming and biodiversity shifts should be a priority issue since these two threats are predicted to impact communities and ecosystems, simultaneously 11 .
Although plant species richness and identity directly influence ecosystem processes and food web structure 12 , it has been argued that the species functional traits are stronger predictors of ecosystem functioning 13 . Recent studies suggest that the distribution of species functional traits within communities, i.e., the trait composition, plays a key role in driving ecosystem processes 14,15 . Functional differences may lead to a variety of interactions among species, and because of the complexity of these interactions it may be difficult to predict the effects of losses or introductions of species in the communities 16 .
Most of the knowledge on biodiversity-ecosystem function (B-EF) relationships come from experiments focusing on plant productivity, while other fundamental processes such as decomposition are relatively less studied 17 but see 18,19 , especially in aquatic ecosystems. The decomposition of plant organic matter is one of the most important ecosystem processes in freshwater ecosystems since it regulates the cycling of carbon (C) and nutrients, and the efflux of greenhouse gasses such as carbon dioxide (CO 2 ) and methane (CH 4 ), which can have positive feedbacks to climate change 20,21 . Dead leaves that fall from trees provide energy and substrate to a wide variety of organisms in freshwater ecosystems 19 . Chemical and physical traits in these dead leaves determine the nutritional value, as well as the lability and toxicity of the litter for invertebrate detritivores and microbial decomposers 19 . The suite of chemical traits defines the litter quality and has been shown to be the major determinant of litter decomposition at different environments and latitudes 22 . While some litter species are nutrientrich and composed of labile carbon (i.e., high quality litter), which benefit decomposers, others may have high concentrations of some secondary compounds and recalcitrant carbon such as lignin (low quality litter), which may inhibit or hinder the activity of microbial decomposers and invertebrate detritivores 19 . On the other hand, the variation in litter trait values at the community level, i.e., the functional diversity, has been shown to control . We cut branches of each species in October 2014 and air-dried them at room temperature until the leaves fall and we used them for the chemical analyses and the experiment. Carbon (C) and nitrogen (N) concentrations were quantified by dry combustion in an elemental analyzer (Perkin Elmer 2400 II CHN) and phosphorus (P) concentration was determined through colorimetry by the vanadate-molybdate method 49 . We used these data to calculate the carbon to nitrogen ratio (C:N) and nitrogen to phosphorus ratio (N:P) for each litter species (Supplementary Table S1). Also, we quantified lignin concentration with the acid-detergent method 50 , and litter tannins and total phenolics concentrations using the Folin-Ciocalteu method 51 . C and N analyses were performed at the Analytical Center of the Institute of Chemistry-University of Campinas. Lignin, P, tannins and total phenolics were analyzed at the Center of Nuclear Energy in Agriculture (CENA)-University of São Paulo.

Figure 1.
Schematic representation of the experimental design, temperature control system and main predictions. Based on quality hypothesis (a), decomposition should increase with litter quality; however, functional diversity hypothesis (b) assumes that decomposition increases with higher litter trait dissimilarities. Moreover, warming should exacerbate such relationships by accelerating microbial activity and macroinvertebrate consumption rates. (c) Alternatively, warming could affect decomposition rates by interacting with the different levels of quality or functional diversity. (d) Decomposition rates are expected to be higher in the coarse-mesh bags and warming may exacerbate the effect; alternatively, warming could decrease litter decomposition in the coarse-mesh bags by negatively affecting macroinvertebrate composition. (e) Experimental design: ten tank bromeliads per block (five blocks). Warming effect was achieved by increasing ambient temperature in 4 °C, following the projection for the year 2100. (f) Litter treatments (LT) consisted of five mixtures of four out of twelve litter species selected based on their chemical traits, ranging from low to high quality, using coarse and fine-mesh bags. In addition, these litter treatments represented a gradient of functional diversity (see Fig. 2 on how we assessed quality and functional diversity).  Table S1), which allowed us to visually determine five litter treatments (hereafter LT; each one containing four litter species) based on the position of each species (i.e., proximity with traits) in the litter trait space. The LTs represent a gradient of quality, ranging from low to high, which determines the expected rate at which they should decompose (Fig. 2a, (Fig. 2b). www.nature.com/scientificreports/ To assess the role of litter quality, we calculated the community-weighted mean values (CWM) of each trait in each LT assuming that greater trait values would drive the effects of our litter clusters on decomposition 52 . CWM is often used to describe the functional composition of communities which varies according to the abundances of species 53 . Since the abundance of each species was evenly distributed in our LTs, differences in the CWM values were directed to their trait values. CWM values were calculated using the functcomp function implemented in the FD R-package. After obtaining the mean values of each trait in each treatment, we standardized the values using z-scores. Then we conducted a PCA using our mean trait matrix (Fig. 2c). Litter quality was then assessed by retaining the first PCA axis, which explained most of the variance of the mean trait values. The first axis of quality PCA (quality1) was positively loaded by higher concentrations of N and P and negatively loaded by polyphenolics, tannins, and C:N ratio (Fig. 2d).
To determine the importance of litter functional diversity on decomposition, we calculated the Rao's quadratic entropy (Rao's Q) for each trait in each LT, using the dbFD function implemented in FD package (v. 1.0-12). Then, we conducted a PCA using the Rao's Q values previously standardized with z-scores and retained the first axis (Fig. 2c). Although Rao's Q is generally used in multi-trait approaches to access trait variability, single-trait Rao's Q have been used to identify groups of traits driving the variance in the trait matrix 54 . This measure can be used to assess the importance of niche complementarity and assumes that trait differences between litter species drives decomposition through complementary consumption of the different litter species 15 . FD1 axis of litter functional diversity PCA separated our litter treatments mainly by high dissimilarities in C and N concentrations, and in C:N ratios (Fig. 2d). A correlation plot for CWM and Rao values is provided in Supplementary Fig. S1.

Experimental design.
To investigate the mechanisms controlling litter decomposition and the effect of warming on this ecosystem function, we used 50 tank-bromeliads Neoregelia johannis acquired from a greenhouse to ensure all plants had similar sizes and water storage capacities and were virtually free of colonizing organisms. Before starting the experiment, we washed the bromeliads with spring water, a 500 mL solution of antibiotic (Ciprofloxacin; 5 µg/mL), and 5% sodium hypochlorite solution to remove invertebrates and unnatural bacteria [39][40][41] . We planted the bromeliads in the ground of the restinga following a randomized block design (5 blocks) and installed plastic roofs at 1 m over each plant to prevent any input of rain and organic matter like branches and fallen leaves from trees but allowing the entrance of colonizing organisms such as microorganisms and insects. Colonization of bromeliads occurred naturally during the experiment as we wanted this to be shaped by our litter and temperature treatments. To keep the water level constant, we watered the bromeliads with stored rainwater every other day.
The warming effect was simulated with an electronic heating system (Fig. 1e) composed by pre-programmed controllers connected to a Delta human-machine interface (HMI). The heating equipment was composed of five boxes (containing the electronic components) connected. Each component box (block) was responsible for accessing and controlling the temperature of 10 bromeliads (five with ambient temperature and five warmed; Fig. 1e). We used submersible water heaters (1 W; n = 3) in half of the bromeliads of each block to maintain a continuous 4 °C increase above the ambient temperature of unheated bromeliads ( Supplementary Fig. S2). This increase in temperature was achieved by submersible sensors in the unheated, reference bromeliads, which switched on and off the heaters from treatment bromeliads through the controllers. To account for the physical effect of the heaters we used heaters turned off in the unheated bromeliads. The temperature difference used in this experiment followed the projections of temperature increase in the southeastern region of Brazil over the next century 38,55 . We conducted our experiment from April to July 2015.
For each LT, we placed 0.4 ± 0.05 g of air-dried leaves (0.1 ± 0.01 g of each species; mean ± SD) in coarse (3 mm plus six additional 10 mm openings) and fine-mesh litterbags (0.05 mm), (Fig. 1f). Litter in fine-mesh bags was accessible only by microorganisms, whereas coarse-mesh bags also allowed access of macroinvertebrates, like detritivorous insects. We prepared a total of 250 coarse-mesh and 250 fine-mesh litterbags. In each block, we placed one LT per bromeliad, repeating the treatment in warmed and ambient bromeliads. Each bromeliad received five coarse-mesh and five fine-mesh litter bags in five different wells, being one litter bag of each mesh size per well. One pair of litter bags was retrieved from each bromeliad (50 coarse and 50 fine) after 16, 32, 48, 64 and 80 days. In summary, our experiment had 5 blocks × 2 temperatures × 1 bromeliad × 2 mesh sizes × 5 litterbags, totalizing 100 replicates per each of 5 LT. Litter from each bag was taken out and oven-dried at 60 °C and weighed. Before oven-drying, the coarse-mesh litter was carefully inspected and cleaned under water to remove possible insects which were collected and stored in 80% ethanol solution, quantified, and identified using taxonomic keys 56 and photos of previous collected specimens (Supplementary Table S2).

Statistical analyses.
All data analyzes were carried out using R version 3.2.2 57 . We measured decomposition using the decay coefficient (k) assuming the exponential decay model m f /m i = e −kt where m f and m i are the final and initial litter dry mass, respectively, t is time in days and k is the decomposition coefficient. The decomposition coefficient (k) was log-transformed after verifying the parametric assumptions of linear models using plots of residuals for normality and homoscedasticity. In order to verify the differences in the decomposition rates among our LTs, we used a linear mixed model (lme function, nlme package v. 3.1-144) with LTs, temperature, mesh size and their interactions. LTs and mesh type were included as fixed predictors and temperature as a continuous variable. Block was included as a random factor. We conducted a pairwise comparison test with Tukey-adjusted p-value using the lsmeans function of the lsmeans package (v. 2.30) to check the differences between the LTs. We conducted additional models to test the effects of warming, quality, and functional diversity of litter on decomposition using the quality1 and FD1 axes as continuous predictor variables, temperature as a continuous variable, and the interactions between quality1 and temperature, and FD1 and temperature. We conducted separate models for fine-mesh decomposition (microorganisms) and coarse-mesh decomposition  Fig. 3a) and were positively correlated with temperature (r = 0.35, p < 0.0001; Table 1; Fig. 3b). However, decomposition rates did not differ between mesh sizes (Table 1; Supplementary Fig. S3), meaning that microorganisms, and not macroinvertebrate detritivores, controlled the detrital processing. No significant interaction effect was detected for LTs, temperature and mesh size. The pairwise test revealed that the decomposition rates differed between LT1 and LT2 (t = − 4.568, p = 0.0001), LT1 and LT3 (t = − 5.665, p < 0.0001), LT2 and LT5 (t = 5.857, p < 0.0001), LT3 and LT4 (t = 3.397, p = 0.0066), LT3 and LT5 (t = 6.957, p < 0.0001), LT4 and LT5 (t = 3.564, p = 0.0037) (Fig. 3a).
In both fine and coarse-mesh bags, litter functional diversity, not litter quality, accelerated the decomposition rates ( Table 2; Fig. 4). Moreover, decomposition increased with increasing temperature (fine-mesh: r = 0.40, p < 0.0001, coarse-mesh: r = 0.30, p < 0.0001; Table 2). The positive effect of litter functional diversity indicated higher decomposition with increasing dissimilarities in C (fine-mesh: F 1,241 = 25.18, p < 0.0001; coarse-mesh: F 1,241 = 29.72, p < 0.0001) and N (fine-mesh: F 1,241 = 18.83, p < 0.0001; coarse-mesh: F 1,241 = 3.89, p = 0.0497) concentrations ( Supplementary Fig. S4a-d). No interaction effect was detected in the models for both mesh sizes. Additionally, the decomposition rates in the coarse-mesh bags were not affected by richness and abundance of detritivorous insects collected inside the litter bags. A comparison between models with FD1 and models with Table 1. Results of linear mixed effect models examining how decomposition rates changed in response to litter treatments (LT), temperature, mesh size (fine versus coarse) and interactions. Marginal R 2 = 0.24, conditional R 2 = 0.25. Significance at p < 0.05. numDF = numerator degrees of freedom, denDF = denominator degrees of freedom.  www.nature.com/scientificreports/  www.nature.com/scientificreports/ dissimilarities of C and N as predictor variables revealed that for microbial decomposition (fine-mesh), the model with litter dissimilarities best explained decomposition (AIC = 273.5 and 268.8, respectively). For coarsemesh, the model including trait dissimilarities showed higher AIC (332.8) compared to the model with FD1 as a predictor variable (AIC = 329.7), which may suggest trait dissimilarities besides C and N to be important for decomposition mediated by microorganisms and invertebrates.

Discussion
Although the effects of litter diversity and their traits on decomposition are well documented in both terrestrial and aquatic ecosystems 28 , the mechanisms underlying litter effects are still poorly understood (but see García-Palacios et al. 54 ), especially in tropical regions. Likewise, little is known on how decomposition will respond to climate changes along with shifts in plant composition 42,58 . As far as we know, this is the first decomposition study that investigated the combined effects of predicted global warming and litter composition, by partitioning litter effects into quality and functional diversity. Here we show that litter functional diversity may be the main driver of decomposition in tropical freshwater ecosystems, increasing complementarity potentially via complementary resource use and nutrient transfer among litter species 25 . Our results indicate that litter decomposition may be more dependent on how different leaf species are combined in terms of traits rather than on the concentration of nutrients or secondary and structural compounds. Additionally, our findings suggest that increasing the homogenization of trees species by loss or changes in the composition may have substantial impacts on the ecosystem functioning of aquatic environments. Moreover, warming independently of the litter functional diversity accelerated the rates of litter breakdown in our study, which was mainly microbial driven. Litter decomposition in freshwater systems provides a substantial amount of CO 2 to the atmosphere, contributing significantly to the global C cycle 59 . Therefore, it is important to understand how climatic changes will affect the processing of organic matter in these systems. Our study demonstrated that microorganisms play a key role in litter decomposition and increasing temperature by 4 °C as predicted for the end of this century may accelerate microbial decomposition in tropical aquatic systems. This is particularly important because decomposition in tropical freshwater ecosystems is mostly driven by microorganisms (i.e., bacteria and fungi) 33 . Although such findings have emerged from studies in streams, they are not limited to these ecosystems. Previous studies using tank bromeliads reported similar patterns in litter decomposition 41,48,60 . Indeed, in our study we did not detect differences in decomposition between the two mesh sizes after 80 days, which reinforces the role of microorganisms as the main decomposers in tank bromeliads. However, we cannot rule out the possibility that due to our design (i.e., fine-mesh and coarse-mesh were in the same tank compartment), fauna effect on litter in coarse-mesh bags may have boosted microbial communities inside fine-mesh bags via microbial and nutrients exchange. In addition, one could argue that the lack of time for colonization by insects before starting the experiment could have influenced the results. However, colonization was fast as it can be observed in Supplementary Table S2. Besides that, in a recent study the authors manipulated tank bromeliads to have only bacteria or to be colonized by bacteria and macroinvertebrates and found no difference in decomposition rates 41 , supporting our findings here.
Interestingly, litter functional diversity arose as the main driver of litter decomposition, regardless of the presence or absence of invertebrate detritivores. Litter functional diversity hypothesis predicts that higher trait dissimilarity increases decomposition through resource complementarity for microbial decomposers and invertebrate detritivores 15,54,61,62 . In our study, decomposition was enhanced by increasing dissimilarity in carbon and nitrogen, in both microbial (fine-mesh) and total (coarse-mesh) decomposition, which is supported by recent studies 15,54 , but see 44 . Contrary to our findings, Frainer et al. 44 found no evidence for effects of litter functional dissimilarity on decomposition and conclude that litter diversity effects are less pronounced in streams. However, they point out this result as being potentially caused by a lack of statistical power to detect dissimilarity effects in their two-species mixtures. In addition, they point to the possibility of effect suppression in the scale at which their study and analyzes were made, that is, the litter bag scale. Among the mechanisms of litter functional diversity effects on decomposition, nitrogen transfer between litter species has been pointed out as the main mechanism in several studies in terrestrial systems 19,25,54 . Nitrogen transfer among litter species, via passive leaching or active microbial transfer (fungal hyphae), is related to niche complementarity effects, where a nutrient-rich litter enhances the nutritional values of a nutrient-poor litter, increasing the resource availability for microbial decomposers and invertebrate detritivores, and hence increasing the overall decomposition 18,25 . Handa et al. 25 suggest that nitrogen transfer from N-fixing litter boosted the decomposition of nutrient-poor litter in a global scale experiment in terrestrial and aquatic systems. Indeed, in our study the litter treatment that showed the highest dissimilarities in C and N (LT3), and that tended to decompose at higher rates, contained the N-fixing species Abarema brachystachya. We suggest that N transfer from A. brachystachya to nutrient-poor species, by leaching or fungal hyphae, may have increased the overall decomposition in these treatments. It is important to mention that decomposition tended to be higher in the intermediate levels of litter quality (LT2 and LT3), which also represented the highest levels of functional diversity. This finding corroborates the results from a recent laboratory study where litter quality was manipulated at different levels and the faster decomposition occurred at the intermediate quality 58 . However, the authors suggest compensatory feeding of detritivores, i.e., enhanced feeding on poor quality food to compensate for the low nutrient intake, as the possible explanation for this result 63,64 . In our study, decomposition was mainly driven by microorganisms, which indicates that compensatory feeding is not the mechanism behind the effect.
Our study identified that litter functional diversity was the main driver of litter decomposition in tank bromeliads. However, it is important to recognize that our design to setup litter treatments did not fully allow disentangling quality from functional diversity. Quality and functional diversity values were both obtained from the same litter treatment, which did not allow testing it in an orthogonal way. Our highest functional diversity www.nature.com/scientificreports/ treatment was also the intermediate litter quality, and this intermediate quality effect could probably be detected by fitting a non-linear model with polynomial predictor for quality. However, in our predictions we expected the decomposition to increase with our quality gradient in a linear way. Thus, our findings support our predictions, and despite the methodological limitations, can be applied in future studies addressing similar questions and using different study systems.
Our results provide novel evidence on the effects of climate change and litter diversity on litter decomposition in aquatic systems. Warming affected litter decomposition in both mesh sizes, but the rates did not differ between mesh sizes indicating that microorganisms control litter decomposition. Therefore, we suggest that macroinvertebrates may be less sensitive to the predicted warming in tropical regions, and thus will not increase their consumption sufficiently to overcome the microbial effect. Such findings are in accordance with a previous study in tank bromeliads where the authors found that warming did not affect richness and composition of macroinvertebrates 39 . They suggest that these organisms may be resistant or resilient to changes in temperature as they live in small water bodies and are frequently exposed to temperature variation, being adapted to these environment 4 . However, we cannot rule out the possibility that our coarse-mesh bags may have prevented access to litter for some invertebrate detritivores and thus decreased their effect on overall decomposition, despite previous studies in tank bromeliads also did not find effects of invertebrate detritivores on decomposition 41,48,59 . On the other hand, recent studies have shown that although microorganisms may be the major decomposers of organic matter in bromeliads, macroinvertebrates may have an important role in the interactions with microorganisms 41 . In their study, Bernabé et al. 41 demonstrate that warming may reduce bacterial density by increasing the metabolism of invertebrate detritivores and direct consumption on bacteria biofilm, thus reducing bacterial activity on litter decomposition 65 . It is known that invertebrate detritivores consume the biofilm on litter as it is a nutrientrich resource 66 . Contrary to our results, Ferreira and Canhoto 67 found that the invertebrate-driven decomposition was more responsive to the increasing temperature than the microbial decomposition, which accelerated the total decomposition. However, their study was conducted in a temperate region where, unlike in the tropical areas, microorganisms are less active because of colder temperatures 33,68 . Litter decomposition rates are expected to increase with temperature because elevated temperature accelerates biochemical reactions and metabolism of litter consumers 2,33 . Our results highlight the importance of investigating the impacts of climate changes on species interactions to understand the responses of ecosystems.
Contrary to our expectations, we did not find an interactive effect between warming and our gradients of quality and functional diversity. We expected that warming would exacerbate the effects of quality and functional diversity on decomposition by accelerating the consumption rates of microbial decomposers and invertebrate detritivores, as increasing temperature should accelerate metabolic rates of litter consumers 2 . However, our findings suggest that the effects of warming may not depend on litter type (low-or high-quality, low-or highfunctional diversity). Such results may be supported by a recent meta-analysis where the authors examined the effects of elevated atmospheric CO 2 and temperature on litter decomposition in streams 9 . They concluded that the effects of warming on litter decomposition may not depend on litter chemical differences 9 . On the other hand, a global synthesis study on the temperature sensitivity of leaf litter decomposition in streams and rivers analyzed data from 169 studies and found some evidence that the suite of litter chemical traits determines how decomposition is affected by temperature 37 . However, this response was only detected when a particular species (N-fixing species) was included in their analyses 37 . Other manipulative studies should address the interactive effects of temperature and litter trait diversity considering different components of microbial and detritivore communities, in order to predict the functioning of freshwater ecosystems in future scenarios.
Predictive studies combining the joint evaluation of future temperature changes and litter diversity on decomposition are rare. Our study demonstrates that the degree of variation of litter traits, i.e., the functional diversity, rather than litter quality, may be more important to predict litter decomposition. Moreover, future temperature rise may accelerate decomposition rates in tropical aquatic systems, mainly by stimulating the microbial-driven decomposition. Our work reveals the differential impact that warming may have on decomposer communities and highlight the importance of considering different litter trait aspects in order to predict the effects of biodiversity shifts on the functioning of freshwater ecosystems. Future studies should access the different components of microbial communities (e.g., diversity and abundance) to better understand the effects of global warming on decomposition in tropical freshwater ecosystems.