This is the accepted version of the article: Peguero, Guille; Sardans i Galobart, Jordi; Asensio Abella, Ma. Dolores; [et al.]. «Nutrient scarcity strengthens soil fauna control over leaf litter decomposition in tropical rainforests». Proceedings of the royal society B, Vol. 286, issue 1910 (Sep. 2019), art. 2191300. DOI 10.1098/rspb.2019.1300 This version is avaible at https://ddd.uab.cat/record/218199 under the terms of the license 1 Nutrient scarcity strengthens soil fauna control over leaf litter 2 decomposition in tropical rainforests 3 4 Running head: Fauna control over litter decomposition 5 6 Guille Peguero1,2,3*, Jordi Sardans2,3, Dolores Asensio2,3, Marcos Fernández-Martínez1,2,3, 7 Albert Gargallo-Garriga2,3, Oriol Grau2,3, Joan Llusià2,3, Olga Margalef2,3, Laura Márquez2,3, 8 Romà Ogaya2,3, Ifigenia Urbina2,3, Elodie A. Courtois1,4, Clément Stahl5, Leandro Van 9 Langenhove1, Lore T. Verryckt1, Andreas Richter6, Ivan A. Janssens1, Josep Peñuelas2,3 10 1 11 Centre of Excellence PLECO (Plants and Ecosystems), Department of Biology, University of 12 Antwerp, 2610 Wilrijk, Belgium. 2 13 CSIC, Global Ecology Unit CREAF-CSIC-UAB, 08913 Bellaterra, Spain. 3 14 CREAF, 08913 Cerdanyola del Vallès, Spain. 4 15 Laboratoire Ecologie, Evolution, Interactions des Systèmes Amazoniens (LEEISA), 16 Université de Guyane, CNRS, IFREMER, 97300 Cayenne, French Guiana. 5 17 INRA, UMR EcoFoG, CNRS, Cirad, AgroParisTech, Université des Antilles, Université de 18 Guyane, 97310 Kourou, France 6 19 Department of Microbiology and Ecosystem Science, University of Vienna, 1090 Vienna, 20 Austria. 21 22 * Corresponding author: guille.peguero@gmail.com 23 Abstract 24 Soil fauna is a key control of the decomposition rate of leaf litter, yet its interactions with litter 25 quality and the soil environment remain elusive. We conducted a litter decomposition 26 experiment across different topographic levels within the landscape replicated in two rainforest 27 sites providing natural gradients in soil fertility to test the hypothesis that low nutrient 28 availability in litter and soil increases the strength of fauna control over litter decomposition. 29 We crossed these data with a large dataset of 44 variables characterizing the biotic and abiotic 30 microenvironment of each sampling point and found that microbe-driven Carbon (C) and 31 Nitrogen (N) loss from leaf litter were 10.1 and 17.9 % lower in the nutrient-poorest site but 32 this among-site difference was equalized when meso- and macrofauna had access to the 33 litterbags. Further, on average soil fauna enhanced by 22.6 % the rate of litter decomposition, 34 and this contribution consistently increased as nutrient availability in the microenvironment 35 declined. Our results indicate that nutrient scarcity increases the importance of soil fauna on C 36 and N cycling in tropical rainforests and that is able to equalize differences in microbial 37 decomposition potential thus buffering to a significant extent nutrient shortages at an 38 ecosystem level. 39 40 Keywords 41 Biogeochemistry, extracellular enzyme activity, litter decomposition, nutrients, soil fauna. 42 Introduction 43 More than 90% of the net primary production of global terrestrial ecosystems is channeled into 44 the detrital food web [1], and soils store the majority of the Earth’s organic carbon (C) 45 (Crowther et al., 2016). Identifying the drivers of organic-matter decomposition is therefore 46 crucial to understanding and predicting global ecosystem functioning. Abiotic factors like 47 climate and litter quality, have traditionally been recognized as the dominant controls on 48 decomposition at large spatial scales, while decomposer organisms would operate as 49 additional, but secondary, local agents [3,4]. Recent evidence, however, indicates that the effect 50 size of microbial biomass on decomposition rates can be equivalent to that of soil temperature 51 and litter moisture, suggesting that biotic factors may explain as much or even more variation 52 than climate in multi-site comparisons, thus questioning such a hierarchical model of litter 53 decomposition [5–7]. In addition, soil fauna has recently been reported to consistently increase 54 the rates of litter decomposition across biomes by 37% [8] and losses in their functional 55 diversity are expected to slow global cycling of C and nutrients [9]. Consequently, the role of 56 biota (i.e. microorganisms and soil fauna) should attain a more central position in the emerging 57 biogeochemical models, to emphasize their ability to modulate the effects of the environment 58 and a changing climate on organic-matter decomposition [10–14]. 59 Leaf litter fall is a dominant pathway for returning nutrients to the soil [15], and soil 60 fauna plays a fundamental and often undervalued role in the litter decomposition process 61 [9,16]. Assemblages of soil animals stimulate litter breakdown by a variety of interconnected 62 mechanisms that alter the composition and performance of the microbial community, which 63 ultimately transform complex plant-derived compounds into CO2, mineral and organic 64 nutrients and humus [12,13,16]. Despite their identification as key agents of organic-matter 65 decomposition, the interaction between soil fauna with litter traits, and particularly, with the 66 soil microenvironment have remained elusive so far. A descriptive example is the hypothetical 67 link between litter quality and the contribution of soil fauna to decomposition. Through 68 selective feeding soil invertebrates could preferentially increase the decomposition of litter 69 with a low C to N or C to P ratio (C:N and C:P, respectively), i.e. litter with a high nutritional 70 value [17,18]. Other studies, however, have suggested that the primary effect of soil fauna is 71 precisely to promote the decomposition of low-quality litter [19–23]. Likewise, a landmark 72 study documented that increasing diversity of leaf-litter within a litterbag substantially 73 enhanced the rate of disappearance of the more recalcitrant litter types, but only in presence of 74 soil fauna, suggesting that animals could bolster the effects of litter diversity through a top- 75 down mechanism [24]. Notwithstanding, evidence supporting this hypothesis is still sparse and 76 comes from single-site or laboratory-based microcosmic experiments [25,26], which may 77 underestimate the large small-scale variability of decomposition rates in natural conditions 78 [5,27]. Moreover, the nutritional status of the soil and the litter microenvironment may affect 79 microbial communities and interact with soil fauna influencing its contribution to 80 decomposition [28]. For instance, the decomposition of low-quality litter may be bottom-up 81 controlled especially in nutrient-poor environments, thus being more dependent on the 82 fragmentation and the microbial stimulation driven by soil fauna [18,28–31]. Still, multi-site 83 litterbag decomposition studies often fail to incorporate high enough within-site replicates 84 along with data of environmental features like nutrient availability measured at the same spatial 85 and temporal grain, therefore masking underlying local variability and hampering our ability 86 to identify alternate regulatory factors [7]. 87 We hypothesize that low nutrient concentrations in the litter substrate and in the 88 surrounding litter and soil microenvironment should increase the importance of soil fauna 89 promoting decomposition. To test this avoiding the limitations of lab microcosms or single- 90 site studies, we conducted a litterbag experiment at two rainforest sites in the Guiana shield 91 (Table 1), and additionally including a high within-site replication to take into account the 92 natural biogeochemical variability typically associated with the topography in these nutrient- 93 poor ecosystems [32,33]. To determine the contribution to the loss of litter mass by mesofauna 94 alone and by meso- plus macrofauna (i.e., invertebrates with body widths smaller and larger 95 than 2 mm, respectively [3]), we used litterbags with three mesh sizes (hereafter referred to as 96 microbes (< 70 µm), mesofauna (< 2 mm) and macrofauna (< 7 mm) for clarity) and filled 97 them with leaf litter substrates from two native tree species with contrasting C:P ratios and 98 their combination: 1561 ± 337 for Goupia glabra Auble. and 2773 ± 307 for Platonia insignis 99 Mart. [34] (hereafter referred as Goupia and Platonia). We focused on P because recent 100 findings have indicated that this element is the predominant limiting nutrient for microbial 101 decomposers in tropical forests [35,36]. Additionally, we also assessed the dependency of the 102 effect of soil fauna on decomposition on a wide range of biotic and abiotic environmental 103 factors, by compiling a unique data set of 44 variables characterizing the below- and 104 aboveground compartments (see Table S1). These variables included soil and litter elemental 105 compositions, activities of extracellular enzymes associated with CNP stoichiometry as 106 indirect measures of the nutritional status of microbial communities [36,37], community-level 107 metrics of functional foliar traits in tree canopies, and abundance and richness of the main 108 orders of litter-dwelling arthropods. Importantly, all these potential regulatory factors were 109 quantified -where appropriate- at the same spatial scale as our individual experimental 110 observation unit (i.e. each block of litterbags). Therefore, by explicitly including this high 111 heterogeneity at a landscape and at a within-plot scale as a set of continuous covariates, we 112 were able to test our hypothesis across the natural environmental gradient included in our study 113 sites, which ranged from low availability to extreme nutrient scarcity. 114 Materials and Methods 115 Study sites and sampling design 116 This study was conducted in two primary tropical forests in French Guiana near the research 117 stations of Nouragues (04°04′53″N, 52°41′13″W) and Paracou (05°16′38″N, 52°55′38″W). 118 Both sites have a mean annual temperature of 25.2 and 25.8 ºC and a tropical climate, with a 119 wet season typically from December to June and a dry season from August to November. 120 Rainfall at the annual scale is similar (2849 vs 3280 mm y-1) although Paracou has a more 121 pronounced dry season due to a higher evapotranspirational demand (mean precipitation and 122 temperature during the driest quarter are 22.3 mm mo-1 and 26.3 ºC at Paracou vs 29.9 mm mo- 1 123 and 25.7 ºC at Nouragues, respectively; Fig. SM1 in Supporting Information). The bedrock 124 at Paracou and Nouragues is Precambrian schist and Caribbean granite, respectively. Soil 125 texture and biogeochemistry in tropical forests can fluctuate with topography due to variations 126 in drainage capacity and erosion, which are usually associated with topographic position. Soils 127 between hills are nutrient-poor sandy Podzols, with clay minerals (kaolinite) and oxides 128 contents increasing toward the tops where Acrisols dominate (Margalef et al. unpublished 129 results). We established 12 plots of 0.25 ha at each site stratified by three topographic positions 130 to account for this heterogeneity: at the top, at the middle and at the bottom between slopes 131 (henceforth referred to as top, slope and bottom plots). We delimited a central 20-m quadrat in 132 each plot where we marked five evenly spaced sampling points around which we focused all 133 our measurements (Fig. SM2). This design thus contained a total of 120 sampling points (2 134 sites × 3 topographic positions × 4 replicate plots per topography × 5 sampling points in each 135 plot). 136 Litterbag experiment 137 We assessed the contribution of invertebrate meso- and macrofauna (body widths smaller and 138 larger than 2 mm, respectively) to the rates of litter-mass loss using 10-cm square polyamide 139 litterbags differing in mesh size: 70 µm (PA-21-71 SEFAR NYTAL, Heiden, Switzerland) 140 excluding both faunal groups but allowing microbes (i.e. fungi and prokaryotes) to decompose 141 the litter substrates, and 2 mm (06-2000/53 SEFAR NYTEX, Heiden, Switzerland) and 7 mm 142 (PE-01903-013 FIBERCORD, Alicante, Spain) allowing the entry of mesofauna and meso- 143 plus macrofauna, respectively. The bottom layers of the litterbags with the largest opening size 144 was made of 0.5-mm mesh [26] (06-500/38 SEFAR NYTEX, Heiden, Switzerland) to prevent 145 the loss of litter fragments. Each litterbag was filled with 2 g of dried leaf litter in three 146 combinations: 1) only Goupia, 2) only Platonia and 3) equal proportions by weight of both 147 species. These native tree species were chosen due to their contrasting C:P and N:P ratios (1561 148 ± 337 and 36.9 ± 3.1 for Goupia vs 2773 ± 307 and 80.7 ± 1.3 for Platonia; mean ± standard 149 error, data from [34]). 150 Freshly fallen leaf litter was collected with litter traps placed under trees in monocultured 151 plantations established by the Center for the International Cooperation in Agronomic Research 152 for the Development (CIRAD) in 1983-84 near the Paracou research station. The traps were 153 harvested monthly, and the plant material was dried at 40 °C in a heater to a constant weight. 154 The leaf litter was placed inside the litterbags and visually inspected. Any material in an 155 advanced stage of degradation was discarded. All individually tagged litterbags were closed 156 and fixed to the soil surface with stainless-steel staples and wire. Each block of nine litterbags 157 (3 mesh sizes × 3 litter combinations) was tied with polyamide thread at each sampling point 158 in November 2015 (end of the dry season) and retrieved in June 2016 (end of the wet season) 159 in the same order as they were initially placed. All harvested litterbags were dried at 40 °C in 160 a heater to constant weight, root and soil residues were gently removed, litter fragments were 161 identified to species for the Goupia-Platonia mixture and were then weighed. A subsample 162 representative of all site, topographic, mesh-size and litter-composition combinations, along 163 with five random samples of each litter type, were milled and analyzed to obtain initial and 164 final C and N contents. Losses of these two elements from the litter were calculated as 100 × 165 [(Mi × CNi)-(Mf × CNf)]/(Mi × CNi), where Mi and Mf are the initial and final litter dry masses, 166 respectively, and CNi and CNf are the initial and final C or N concentrations (% of litter dry 167 mass), respectively [9]. Using C loss (%) in addition to total litter-mass loss allowed us to 168 assess the potential effects of any possible inorganic contamination of the litter retrieved from 169 the field [9]. 170 Environmental biotic and abiotic data 171 We compiled data for 44 variables describing the below- and aboveground biophysical and 172 biological components surrounding each sampling point (i.e. block of litterbags) to identify the 173 potential microenvironmental and biotic drivers behind the effect of fauna on decomposition. 174 Briefly, we determined the concentrations of nutrients (C, N, P, K, Ca, Mg and Na) in the litter 175 (organic horizon) and soil (0-15 cm depth) pools at each sampling point by means of coupled 176 plasma/optical emission spectrometry. Additionally, the concentration of available P in the soil 177 was determined by both the Olsen and Bray methods. We also determined the activities of the 178 extracellular enzymes β-glucosidase, leucine and glycine aminopeptidases and acid and 179 alkaline phosphatases (henceforth referred to as βgluc, leu, gly, acidP and alkP, respectively) 180 in the litter and soil at each sampling point by means of colorimetric assays. We sampled the 181 communities of arthropods in the litter surrounding each sampling point by means of 182 Winkler/Moczarsky traps and then classifying each collected specimen into 33 Order or sub- 183 Order taxonomic categories covering all major lineages within Arthropoda. And finally, all 184 trees (diameter at breast height ≥10 cm) within the 0.25-ha plots were mapped, tagged and 185 identified to species or genus with herbarium vouchers for determining the tree species 186 richness, phylogenetic diversity and three complementary indexes of functional trait diversity 187 for each plot (please see Table S1 and supplementary methods for detailed procedural 188 descriptions). 189 Data analyses 190 All statistical analyses were carried out with R v3.4.3 [38]. The variation of litter mass lost 191 from the litterbags after the incubation was assessed using a linear mixed model as 192 implemented in the lme4 package [39], including site, topography, mesh size, litter composition 193 and the interaction between site and mesh size as fixed-effects terms. Sampling point was 194 added as a random intercept term nested within plot, topography and site, thus representing the 195 spatial structure of our experimental design. Higher-order interactions were sequentially 196 removed when not significant (P > 0.05), additionally assessing the Akaike Information 197 Criterion (AIC) and retrieving the coefficients of determination (r2). Parameter-specific P- 198 values for the mixed models were calculated by normal, Satterthwaite and Kenward-Rogers 199 approximations to the number of degrees of freedom, and all approaches yielded qualitatively 200 identical results. The same models were used for C and N losses, although the lower number 201 of samples precluded the inclusion of a random-effects structure. 202 We determined the distribution of all environmental biotic and abiotic variables using 203 a Principal Components Analysis (PCA). We confirmed the apparent differences between sites 204 and across topographic levels for the first and second PCA axes using a linear mixed model 205 with the PC1 and PC2 scores as response variables. Then, we analyzed the variation of the 206 most relevant environmental variables, i.e. those with larger loadings on these first two axes of 207 the PCA. The effects of soil fauna on leaf-litter decomposition were measured as the difference 208 in mass loss between the litterbags with and without fauna access [34]. To visualize these fauna 209 effects within the multivariate environmental space we repeated this PCA including the six 210 corresponding fauna-effect variables (two mesh sizes crossed with three litter combinations). 211 The relationship between the contribution of soil fauna to decomposition with the 212 microenvironment was assessed using a linear mixed model with fauna effect as a response 213 variable and replacing site and topographic categorical factors by the scores of each sampling 214 point over the PC1 and PC2 (obtained from the PCA without fauna-effect variables included), 215 as surrogates of variations in nutrient availability associated to the environment. This analytical 216 approach allowed us to synthesize a complex multidimensional scenario of regional and 217 topographically associated variation in the environment into a more tractable and interpretable 218 output [18,40]. Furthermore, by including this environmental heterogeneity as continuous 219 covariates, we were able to assess the effect of soil fauna on decomposition across the natural 220 gradient of nutrient availability encompassed in our study sites. Finally, we additionally 221 explored the potential contribution of the first six PCA axes (which together explained a 222 cumulative proportion of variance of 58%) over the effects of the fauna on decomposition using 223 automated model selection with the dredge function from the MuMIn package [41]. However, 224 the subset of models with the lowest AIC only included PC1, therefore discarding all other 225 axes. 226 Results 227 Loss of litter mass and nutrients. After seven months of incubation, between 68 and 70% of 228 the initial leaf-litter mass was lost when meso- and macrofauna had access to the litterbags. 229 However, in litterbags with the smallest mesh size (microbial decomposition only), litter mass 230 loss dropped to 48% on average in Nouragues, and to only 40% in the relatively nutrient-poorer 231 site at Paracou (Fig. 1 and Table 2, site × size interaction). Models assessing C and N losses 232 yielded qualitatively similar results, although this between-site difference in microbial 233 decomposition potential was even larger for N, being 18% lower at Paracou than at Nouragues 234 (Table 2, site × size interaction). The soil fauna in Paracou was nevertheless able to compensate 235 this lower baseline of microbial decomposition, so that the loss rates of litter mass and nutrients 236 were equalized between sites when both meso- and macrofauna had access to the litterbags 237 (Fig. 1 and Table 2). 238 Additionally, the decomposition rates of the comparatively P-richer litter of Goupia and 239 the P-poorer Platonia were unexpectedly similar, although the mass losses for the combination 240 of the two species was larger (+3.4%), indicating that when mixed both species decomposed 241 faster (Table 2, species). 242 Environmental variation between and within study sites. A principal component analysis 243 (PCA) combining 44 potential regulatory controls with the effect of soil fauna on litter 244 decomposition, measured as the difference in the loss of litter mass between the litterbags with 245 and without faunal access [34], showed that the first two axes comprised 29.7% of the total 246 variation between and within sites, underlining the high environmental heterogeneity at large 247 and small spatial scales (Fig. 2, see Table S1 for descriptions of the variables). Despite this 248 variability, the clear separation of the sampling points at both sites indicated that PC1 captured 249 regional-scale disparities mostly associated to nutrient-related variables in the litter layer. 250 Conversely, PC2 mainly identified within-site soil-related variation linked with topographic 251 position of sampling plots (Fig. S1). All fauna effects appeared to consistently correlate with 252 lower scores on the PC1 (Fig. 2, red vectors). Repeating this PCA excluding the fauna effect 253 variables resulted in very subtle changes but a slight increase in the amount of total variance 254 explained by PC1 and PC2 (32.6%, Fig. S2). Total N concentration in all compartments, foliar 255 C:nutrient ratios in the canopy and litter and phosphatase and aminopeptidase activities in the 256 litter were the most important variables in PC1 (Fig. S3). Overall, the Nouragues site was richer 257 in N in all compartments, from the canopy to the soil (Table 1 and Fig. S4), whereas the higher 258 litter C:nutrient ratios at Paracou suggested that the activity of microbial decomposers could 259 be constrained to some degree. 260 Indeed, we also found that the activities of the extracellular aminopeptidases and 261 phosphatases in the litter were lower at Paracou, indicating either a lower microbial biomass, 262 restricted microbial performance [37], or lower substrate availability [42]. The stoichiometry 263 of extracellular enzymes is a good indicator of the relative nutrient demands of microbial 264 communities [36,37]. The relative allocation between N- and P-acquiring enzymes was similar 265 at both sites, despite the lower activity of all extracellular enzymes at Paracou, suggesting that 266 the microbial communities there were generally nutrient-limited instead of stoichiometrically 267 unbalanced (Fig. S5). In contrast to the organic horizon, enzymatic activity in the topsoil 268 mostly varied across topographic levels, generally increasing toward the top as total nutrient 269 concentrations did in that compartment (Fig. S1). 270 Environmental dependency of the effect of fauna on decomposition. We assessed the 271 relationship between the contribution of soil fauna to decomposition and the microenvironment 272 using a linear mixed model with fauna effect as a response variable and replacing site and 273 topographic categorical factors by the scores of each sampling point on PC1 and PC2 (obtained 274 from the PCA without fauna effect variables included), as surrogates of regional (between- 275 sites) and locally (across topographies) associated variations in the microenvironment. This 276 analytical approach synthesized complex multivariate environmental scenarios into more 277 tractable and interpretable outputs [18,40], but most importantly, it allowed to assess the effect 278 of soil fauna on decomposition across the natural gradient of nutrient availability encompassed 279 within our study sites that ranged from low availability to extreme nutrient scarcity. The effect 280 of soil fauna on decomposition was strongly and negatively correlated with the PC1, but not 281 with the PC2 scores, indicating that the main drivers of the variation in the fauna effect on 282 decomposition were the microenvironmental variables associated with differences in nutrient 283 availability in the litter layer such as total N concentration, C:nutrient ratios and enzymatic 284 activities (Fig. 3 and Table 2). 285 The effect of the soil fauna was also larger in the mixed litter treatment (+3.8%) and 286 was marginally larger (+2.9%) in the relatively P-poor litter species (Platonia, Fig. 4a and 287 Table 2, species). The relationship between this fauna impact on decomposition and the 288 variation of the microenvironment (PC1 scores), however, had a smoother, less negative slope 289 for the mixed litter treatment, indicating that the combination of different litter substrates may 290 have weakened the context-dependency of fauna effects on decomposition (Table 2, PC1 × 291 species, and Fig. S6). Finally, as anticipated in the analysis of litter mass loss, the net effect on 292 decomposition was larger (+4.1%) for the complete community of soil fauna (i.e. meso- plus 293 macrofauna) than for the mesofaunal component only, irrespective of the microenvironment 294 and in all litter combinations (Fig. 4b and Table 2). 295 Discussion 296 We here demonstrate that the strength of soil fauna control on litter decomposition is linked 297 with its biotic and abiotic environment. The net contribution of soil fauna to litter mass loss 298 increased as the conditions for microbial decomposition were more adverse, specifically when 299 nutrient concentrations, and N in particular, were lower, not only in the litter substrate within 300 each litterbag but also in the surrounding litter pool. This was consistent with the reduction of 301 the activity of N- and P-acquiring extracellular enzymes in the litter layer, which were 302 associated to stronger fauna effects on decomposition, thus providing additional support to the 303 view that when the microbial communities inhabiting the organic horizon are relatively nutrient 304 limited the facilitating role of soil fauna acquires a greater importance. Therefore, we found 305 that soil fauna was able to minimize differences in litter decomposition buffering ecosystem- 306 level nutrient shortages at regional scales. This supports recent findings challenging the long- 307 standing view that biotic controls on decomposition would be subordinate to regional and 308 global-scale features such as climate [6,7], and support propositions of local-scale variables 309 regulating microbial activity as predominant drivers of decomposition [5]. 310 Microbes are the ultimate agents responsible for the transformation of dead organic 311 matter, mineralization to CO2 and inorganic nutrients, and humus formation [12,13,16]. 312 Nutrient availability rather than abundance of detritus per se is a main limitation to microbial 313 growth and so of litter decomposition [35,36]. Microbial communities inhabiting environments 314 differing in nutrient availability may face contrasting stoichiometric imbalances that can 315 restrict their ability to decompose organic matter [35,43]. In low-nutrient environments (e.g. 316 with high C:N ratios) microbes can adjust their metabolism to reduce their C-use efficiency 317 while increasing their nutrient-use efficiency (i.e. the ratios of growth over organic C or 318 nutrient uptake) to cope with the physiological challenges of resource imbalance [44,45]. Many 319 direct and indirect animal-mediated processes may enhance nutrient supply, potentially 320 stimulating microbial activity [12,13]. For example, the fragmentation and comminution of 321 litter increases its surface area to mass ratio, making it more readily attacked by microbes 322 (Chapin, Matson, & Mooney, 2002; Joly, Coq, Coulis, Nahmani, & Hättenschwiler, 2018). The 323 translocation and redistribution of freshly fallen litter across soil surfaces and depths together 324 with modifications of aggregation properties and pore structure may likewise accelerate 325 nutrient release [12,13]. Microbial inoculation and the preconditioning of litter during transit 326 through animal guts may also facilitate decomposition [12,13], and importantly, this effect can 327 be directly associated with initial litter quality (Joly, Coulis, Gérard, Fromin, & Hattenschwiler, 328 2015; Joly et al., 2018). In fact, Joly and collaborators found that the lower the initial litter 329 quality the greater the magnitude of microbial stimulation after invertebrate gut passage (Joly 330 et al., 2015), and that the positive effect of soil fauna was mainly related with greater N release 331 from faeces than from litter where this nutrient is more rapidly immobilized (Joly et al., 2018). 332 Direct grazing by soil fauna on living fungal hyphae, bacterial mat and microbial necromass 333 may also alter density-dependent community functions such as substrate, enzyme and nutrient 334 diffusion and exploitative and interfering competitive interactions affecting species 335 coexistence and thus the composition and performance of microbial communities 336 (Buchkowski, Bradford, Grandy, Schmitz, & Wieder, 2017; Crowther, Boddy, & Jones, 2011). 337 The nutrients acquired by soil animals generally exceed their demands, and the surplus 338 is excreted in easily available forms such as urea, ammonia, phosphate and other derivative 339 forms (Chapin et al., 2002). At a macroscopic scale, it is well-known that, through their dung 340 and flesh, megafauna increases nutrient diffusion across the landscape with strong impacts on 341 ecosystem functioning [51]. Likewise, soil fauna could improve the movement of nutrients 342 across the litter-soil interface. Indeed, nutrient transfer between litter types, from the N pool in 343 the soil or from microbial fixation, has been suggested as a widespread mechanism behind the 344 diversity-function effects on decomposition [9,29,52]. We argue that soil fauna may play a role 345 in these phenomena because they could locally enrich low-quality litter substrates by increasing 346 nutrient diffusivity, thereby relaxing the stoichiometric constraints that may hinder their 347 breakdown. If so, a low nutrient concentration in a particular litter substrate and in the 348 associated microenvironment should increase the importance of the facilitation of nutrient 349 mobility by soil fauna. 350 Previous studies have reported that soil fauna can strengthen the diversity-function 351 effects on litter decomposition, increasing the rates of loss of litter mixtures with higher 352 diversity [24,26]. Our results also indicated that soil fauna had a larger effect in the litter 353 mixture treatment. The stoichiometric heterogeneity of complex litter mixtures could better 354 match the nutritional demands of litter-feeding animals, thereby stimulating its activity [25]. 355 The variation of their contribution to the decomposition of the richest mixture, however, was 356 less dependent on the microenvironment than for the single-species litterbags. From our point 357 of view, this finding implies that more complementary litter mixtures would be less reliant on 358 a potential animal-mediated mechanism of nutrient transfer, which delivers nutrients from the 359 pool in the microenvironment. Additional support to this hypothesis may come from an 360 experimental fertilization experiment, where synergistic diversity effects on decomposition 361 correlated with the stoichiometric dissimilarity of the litter mixture only in the presence of soil 362 fauna, while this relationship disappeared when the nutrient pool available in the 363 microenvironment was experimentally increased [26]. The same authors concluded that 364 microbial activity was subsidized by nutrient uptake coming from other sources than the litter 365 present in the litterbags. In light of our findings, we also suggest that soil fauna may be a key 366 facilitator of this external flow of resources, which could be increasingly important as nutrient 367 content in the microenvironment decreases or the litter mixtures become poorer or more 368 unbalanced. 369 Acknowledgements 370 This research was supported by the European Research Council Synergy grant ERC-2013-SyG 371 610028-IMBALANCE-P, the Spanish Government grant CGL2016-79835-P and the Catalan 372 Government grant SGR 2017-1005. G.P was supported by a postdoctoral fellowship from the 373 Ramon Areces Foundation. We thank the staff of the Nouragues station, managed by USR 374 mixte LEEISA (CNRS; Cayenne), and the Paracou station, managed by UMR Ecofog (CIRAD, 375 INRA; Kourou). Both research stations received support from “Investissement d’Avenir” 376 grants managed by Agence Nationale de la Recherche (CEBA: ANR-10-LABX- 25-01, 377 ANAEE-France: ANR-11-INBS-0001). We thank Vincent Freycon for the identification and 378 characterization of each topographical position and Pascal Petronelli for the botanical 379 determination. 380 Authors’ contributions 381 G.P, I.A.J and J.P designed the study. G.P, D.A, A.G-G, O.G, J.L, L.M, O.M, R.O, I.U, E.L.A, 382 C.S, L.V, L.T.V performed field and/or lab work. G.P compiled and analyzed the data with 383 advice of J.S, M.F-M and J.P. G.P wrote the manuscript with substantial inputs of J.S, A.R, 384 I.A.J and J.P, and revisions of all co-authors. 385 Data accessibility 386 All data supporting the results presented in this contribution will be archived in an appropriate 387 public repository and the data DOI will be included at the end of the article upon acceptance 388 of the manuscript. 389 References 390 1. Cebrian J. 1999 Patterns in the fate of production in plant communities. Am. Nat. 154, 391 449–468. 392 2. Crowther TW et al. 2016 Quantifying global soil carbon losses in response to 393 warming. Nature 540, 104–108. (doi:10.1038/nature20150) 394 3. Swift MJ, Heal OW, Anderson JM. 1979 Decomposition in terrestrial ecosystems. 395 Berkeley, USA: University of California Press. 396 4. Cornwell WK et al. 2008 Plant species traits are the predominant control on litter 397 decomposition rates within biomes worldwide. Ecol. Lett. 11, 1065–1071. 398 (doi:10.1111/j.1461-0248.2008.01219.x) 399 5. Bradford MA, Warren RJ, Baldrian P, Crowther TW, Maynard DS, Oldfield EE, 400 Wieder WR, Wood SA, King JR. 2014 Climate fails to predict wood decomposition at 401 regional scales. Nat. Clim. Chang. 4, 625–630. (doi:10.1038/nclimate2251) 402 6. Bradford MA et al. 2017 A test of the hierarchical model of litter decomposition. Nat. 403 Ecol. Evol. 1, 1836–1845. (doi:10.1038/s41559-017-0367-4) 404 7. Bradford MA, Berg B, Maynard DS, Wieder WR, Wood SA. 2016 Understanding the 405 dominant controls on litter decomposition. J. Ecol. 104, 229–238. (doi:10.1111/1365- 406 2745.12507) 407 8. García-Palacios P, Maestre FT, Kattge J, Wall DH. 2013 Climate and litter quality 408 differently modulate the effects of soil fauna on litter decomposition across biomes. 409 Ecol. Lett. 16, 1045–1053. (doi:10.1111/ele.12137) 410 9. Handa IT et al. 2014 Consequences of biodiversity loss for litter decomposition across 411 biomes. Nature 509, 218–21. (doi:10.1038/nature13247) 412 10. Schmidt MWI et al. 2011 Persistence of soil organic matter as an ecosystem property. 413 Nature 478, 49–56. (doi:10.1038/nature10386) 414 11. Soong JL, Nielsen UN. 2016 The role of microarthropods in emerging models of soil 415 organic matter. Soil Biol. Biochem. 102, 37–39. (doi:10.1016/j.soilbio.2016.06.020) 416 12. Filser J et al. 2016 Soil fauna: Key to new carbon models. Soil 2, 565–582. 417 (doi:10.5194/soil-2-565-2016) 418 13. Grandy AS, Wieder WR, Wickings K, Kyker-Snowman E. 2016 Beyond microbes: 419 Are fauna the next frontier in soil biogeochemical models? Soil Biol. Biochem. 102, 420 40–44. (doi:10.1016/j.soilbio.2016.08.008) 421 14. Kaiser C, Franklin O, Dieckmann U, Richter A. 2014 Microbial community dynamics 422 alleviate stoichiometric constraints during litter decay. Ecol. Lett. 17, 680–690. 423 (doi:10.1111/ele.12269) 424 15. Schlesinger WH, Bernhardt ES. 2013 Biogeochemistry: an analysis of global change. 425 Waltham, USA: Academic Press. 426 16. Bardgett RD, van der Putten WH. 2014 Belowground biodiversity and ecosystem 427 functioning. Nature 515, 505–511. (doi:10.1038/nature13855) 428 17. Zimmer M, Kautz G, Topp W. 2005 Do woodlice and earthworms interact 429 synergistically in leaf litter decomposition ? Funct. Ecol. 19, 7–16. 430 (doi:10.1111/j.0269-8463.2005.00926.x) 431 18. Fujii S, Cornelissen JHC, Berg MP, Mori AS. 2018 Tree leaf and root traits mediate 432 soil faunal contribution to litter decomposition across an elevational gradient. Funct. 433 Ecol. 32, 840–852. (doi:10.1111/1365-2435.13027) 434 19. Van Wensem J, Verhoef HA, Van Straalen NM. 1993 Litter degradation stage as a 435 prime factor for isopod interaction with mineralization processes. Soil Biol. Biochem. 436 25, 1175–1183. (doi:10.1016/0038-0717(93)90212-T) 437 20. Tian G, Brussaard L, Kang BT. 1995 Breakdown of plant residues with contrasting 438 chemical compositions under humid tropical conditions: Effects of earthworms and 439 millipedes. Soil Biol. Biochem. 27, 277–280. (doi:10.1016/0038-0717(94)00182-Z) 440 21. Schädler M, Brandl R. 2005 Do invertebrate decomposers affect the disappearance rate 441 of litter mixtures? Soil Biol. Biochem. 37, 329–337. 442 (doi:10.1016/j.soilbio.2004.07.042) 443 22. Yang X, Chen J. 2009 Plant litter quality influences the contribution of soil fauna to 444 litter decomposition in humid tropical forests, southwestern China. Soil Biol. Biochem. 445 41, 910–918. (doi:10.1016/j.soilbio.2008.12.028) 446 23. Riutta T, Slade EM, Bebber DP, Taylor ME, Malhi Y, Riordan P, Macdonald DW, 447 Morecroft MD. 2012 Experimental evidence for the interacting effects of forest edge, 448 moisture and soil macrofauna on leaf litter decomposition. Soil Biol. Biochem. 49, 449 124–131. (doi:10.1016/j.soilbio.2012.02.028) 450 24. Hattenschwiler S, Gasser P. 2005 Soil animals alter plant litter diversity effects on 451 decomposition. Proc. Natl. Acad. Sci. 102, 1519–1524. 452 (doi:10.1073/pnas.0404977102) 453 25. Vos VCA, van Ruijven J, Berg MP, Peeters ETHM, Berendse F. 2013 Leaf litter 454 quality drives litter mixing effects through complementary resource use among 455 detritivores. Oecologia 173, 269–280. (doi:10.1007/s00442-012-2588-1) 456 26. Barantal S, Schimann H, Fromin N, Hättenschwiler S. 2014 C , N and P fertilization in 457 an Amazonian rainforest supports stoichiometric dissimilarity as a driver of litter 458 diversity effects on decomposition. Proc. R. Soc. B Biol. Sci. 281, 20141682. 459 27. Waring BG. 2012 A Meta-analysis of Climatic and Chemical Controls on Leaf Litter 460 Decay Rates in Tropical Forests. Ecosystems 15, 999–1009. (doi:10.1007/s10021-012- 461 9561-z) 462 28. Milcu A, Manning P. 2011 All size classes of soil fauna and litter quality control the 463 acceleration of litter decay in its home environment. Oikos 120, 1366–1370. 464 (doi:10.1111/j.1600-0706.2010.19418.x) 465 29. St John MG, Orwin KH, Dickie IA. 2011 No ‘home’ versus ‘away’ effects of 466 decomposition found in a grassland-forest reciprocal litter transplant study. Soil Biol. 467 Biochem. 43, 1482–1489. (doi:10.1016/j.soilbio.2011.03.022) 468 30. Perez G, Aubert M, Decaëns T, Trap J, Chauvat M. 2013 Home-Field Advantage: A 469 matter of interaction between litter biochemistry and decomposer biota. Soil Biol. 470 Biochem. 67, 245–254. (doi:10.1016/j.soilbio.2013.09.004) 471 31. Lin H, He Z, Hao J, Tian K, Jia X, Kong X, Akbar S, Bei Z, Tian X. 2017 Effect of N 472 addition on home-field advantage of litter decomposition in subtropical forests. For. 473 Ecol. Manage. 398, 216–225. (doi:10.1016/j.foreco.2017.05.015) 474 32. Luizao RCC, Luizao FJ, Paiva RQ, Monteiro TF, Sousa LS, Kruijt B. 2004 Variation 475 of carbon and nitrogen cycling processes along a topographic gradient in a central 476 Amazonian forest. Glob. Chang. Biol. 10, 592–600. (doi:10.1111/j.1529- 477 8817.2003.00757.x) 478 33. Werner FA, Homeier J. 2015 Is tropical montane forest heterogeneity promoted by a 479 resource-driven feedback cycle ? Evidence from nutrient relations , herbivory and litter 480 decomposition along a topographical gradient. Funct. Ecol. 29, 430–440. 481 (doi:10.1111/1365-2435.12351) 482 34. Coq S, Souquet J-M, Meudec E, Cheynier V, Hättenschwiler S. 2010 Interspecific 483 variation in leaf litter tannins drives decomposition in a tropical rain forest of French 484 Guiana. Ecology 91, 2080–2091. (doi:10.1890/09-1076.1) 485 35. Fanin N, Fromin N, Buatois B, Hättenschwiler S. 2013 An experimental test of the 486 hypothesis of non-homeostatic consumer stoichiometry in a plant litter-microbe 487 system. Ecol. Lett. 16, 764–772. (doi:10.1111/ele.12108) 488 36. Camenzind T, Hättenschwiler S, Treseder KK, Lehmann A, Rillig MC. 2018 Nutrient 489 limitation of soil microbial processes in tropical forests. Ecol. Monogr. 88, 4–21. 490 (doi:10.1002/ecm.1279) 491 37. Sinsabaugh RL, Follstad Shah JJ. 2012 Ecoenzymatic Stoichiometry and Ecological 492 Theory. Annu. Rev. Ecol. Evol. Syst. 43, 313–343. (doi:10.1146/annurev-ecolsys- 493 071112-124414) 494 38. R Core Team. 2016 R: A Language and Environment for Statistical Computing. 495 39. Bates D, Mächler M, Bolker B, Walker S. 2014 Fitting Linear Mixed-Effects Models 496 using lme4. ArXiv pre-print 497 40. Ochoa-Hueso R. 2016 Nonlinear disruption of ecological interactions in response to 498 nitrogen deposition. Ecology 97, 2802–2814. (doi:10.1002/ecy.1521) 499 41. Barton K. 2018 MuMIn: multi-model inference. R package v.1.40.4. 500 42. Margalef O et al. 2017 Global patterns of phosphatase activity in natural soils. Sci. 501 Rep. 7, 1–13. (doi:10.1038/s41598-017-01418-8) 502 43. Sterner RW, Elser JJ. 2002 Ecological stoichiometry: the biology of elements from 503 molecules to the biosphere. Princeton University Press. 504 44. Mooshammer M et al. 2014 Adjustment of microbial nitrogen use efficiency to 505 carbon:Nitrogen imbalances regulates soil nitrogen cycling. Nat. Commun. 5, 1–7. 506 (doi:10.1038/ncomms4694) 507 45. Manzoni S, Taylor P, Richter A, Porporato A, Ågren GI. 2012 Environmental and 508 stoichiometric controls on microbial carbon-use efficiency in soils. New Phytol. 196, 509 79–91. (doi:10.1111/j.1469-8137.2012.04225.x) 510 46. Chapin III F, Matson P, Mooney H. 2002 Principles of terrestrial ecosystem ecology. 511 New York, USA: Springer-Verlag. 512 47. Joly FX, Coq S, Coulis M, Nahmani J, Hättenschwiler S. 2018 Litter conversion into 513 detritivore faeces reshuffles the quality control over C and N dynamics during 514 decomposition. Funct. Ecol. 32, 2605–2614. (doi:10.1111/1365-2435.13178) 515 48. Joly FX, Coulis M, Gérard A, Fromin N, Hattenschwiler S. 2015 Litter-type specific 516 microbial responses to the transformation of leaf litter into millipede feces. Soil Biol. 517 Biochem. 86, 17–23. 518 49. Crowther TW, Boddy L, Jones TH. 2011 Outcomes of fungal interactions are 519 determined by soil invertebrate grazers. Ecol. Lett. 14, 1134–1142. 520 (doi:10.1111/j.1461-0248.2011.01682.x) 521 50. Buchkowski RW, Bradford MA, Grandy AS, Schmitz OJ, Wieder WR. 2017 Applying 522 population and community ecology theory to advance understanding of belowground 523 biogeochemistry. Ecol. Lett. 20, 231–245. (doi:10.1111/ele.12712) 524 51. Doughty CE, Wolf A, Malhi Y. 2013 The legacy of the Pleistocene megafauna 525 extinctions on nutrient availability in Amazonia. Nat. Geosci. 6, 761–764. 526 (doi:10.1038/ngeo1895) 527 52. Vitousek PM, Hobbie S. 2000 Heterothophic Nitrogen fixation in decomposing litter: 528 Patters and Regulation. Ecology 81, 2366–2376. (doi:10.1890/0012- 529 9658(2000)081{$[$}2366:HNFIDL{$]$}2.0.CO;2) 530 531 532 TABLE 1 Characteristics of the study sites Nouragues Paracou 04°04′53″N 05°16′38″N Coordinates 52°41′13″W 52°55′38″W Soil type Sandy Podzols Sandy Podzols (FAO) and Acrisols and Acrisols MAT (°C) 25.2 25.8 MAP (mm) 3280 2849 Aboveground 423 ± 44 371 ± 20 biomass (t/ha) Litter 1259 ± 40 1265 ± 54 pool (g m-2) Foliar N (%) 2.05 ± 0.01 1.93 ± 0.01*** Litter N (%) 1.49 ± 0.03 1.32 ± 0.18** Foliar C:N 25.21 ± 0.09 26.28 ± 0.12*** Litter C:N 33.53 ± 0.80 37.14 ± 0.70** Litter AlkP 73.73 ± 4.75 33.58 ± 2.29*** activity Arthropod 477 ± 28 536 ± 32 density (id m-2) Tree species 38 ± 2 32 ± 1** richness Tree functional -0.09 ± 0.12 -0.11 ± 0.08 richness Values are means ± standard errors (n=120, except n=24 for tree-community data). Elemental ratios are mass-based. AlkP refers to maximum potential activity of alkaline phosphatase in litter (µmol pNP g-1 h-1). Tree species richness refers to mean number of species per plot while functional richness is a unitless standardized effect size of the convex hull volume defined by six foliar traits. Between-site differences are based on linear mixed-effect models, with site and topography as fixed factors and sampling point within each plot as a random effect. *, ** and *** denote P<0.05, 0.01 and 0.001, respectively. See text and supplementary materials for further details. 533 534 TABLE 2 Coefficients, significance and r2 for the linear mixed models used to evaluate the controls on losses of litter mass and nutrients and faunal effects on decomposition Model Variables Litter-mass C loss N loss Fauna effect loss Intercept 65.0 ± 2.3 71.8 ± 3.3 57.8 ± 3.9 22.6 ± 1.7 Site Paracou 2.9 ± 2.4 -1.9 ± 3.5 3.4 ± 4.2 Topography Slope 0.6 ± 2.4 -0.5 ± 2.5 3.1 ± 2.9 Top 4.1 ± 2.4 4.3 ± 2.4 5.9 ± 2.9 Size Mesofauna -5.3 ± 1.6 -6.1 ± 3.5 -4.4 ± 4.1 -4.1 ± 1.2 Microbes -19.8 ± 1.6 -19.9 ± 3.5 -20.6 ± 4.2 Species Platonia 0.8 ± 1.1 -2.9 ± 2.4 0.9 ± 3.3 2.9 ± 1.5 Platonia + Goupia 3.4 ± 1.1 5.2 ± 2.4 9.1 ± 2.9 3.8 ± 1.5 PC1 (nutrient availability) -3.1 ± 0.5 PC2 (topography) -0.7 ± 0.5 Site × Size Paracou-Mesofauna 2.4 ± 2.2 6.1 ± 4.9 3.7 ± 5.8 Paracou-Microbes -10.3 ± 2.2 -10.1 ± 4.9 -17.9 ± 5.9 PC1 × Species Platonia 0. 8 ± 0.5 Platonia + Goupia 1.3 ± 0.5 Model r2m/ r2c 29.5/50.2 44.1 45.0 12.6/43.3 Losses of litter mass, carbon (C) and nitrogen (N) are percentages from initial dry mass and C and N contents, respectively. The fauna effect on decomposition is the difference between the loss of litter mass from the litterbags with meso- and macrofauna relative to the losses from the corresponding microbial-only litterbag (mesh sizes of 2 and 7 mm vs 70 µm, respectively; see Methods). Intercept group- level is Nouragues-Bottom-Macrofauna-Goupia for the models of litter-mass loss (n=1080) and C and N losses (n=206), and the intercept for the fauna effect model (n=720) is Macrofauna-Goupia. The factor species denotes three litter combinations based on two species with contrasting C to phosphorus ratios. PC1 and PC2 are the scores of each sampling point for the first and second PCA axes, which encompass gradients of nutrient availability and topographic microenvironmental variation (see Figs. 3, S1 and S2). Models are linear mixed models, with sampling point as a random intercept nested within plot, topography and site, except for models of C and N loss, for which the lower number of samples precluded the inclusion of a random term. When applicable, marginal r2 (r2m) values are associated to fixed factors while the conditional r2 (r2c) additionally retain the random effects structure. Significant (P<0.05) and marginally significant (P<0.1) parameter coefficients are highlighted in bold and italics, respectively. 535 536 537 FIGURE 1. Variation in the loss of litter (as a percentage of initial dry mass) by site and 538 litterbag mesh size. Different uppercase letters denote significant differences between sites for 539 the same mesh size, and lowercase letters denote significant differences among mesh sizes 540 within the same site and points indicate outliers. Among-group comparisons are Tukey post- 541 hoc tests based on marginal means estimated from a linear mixed model. See Table 2 for model 542 output. 543 544 545 FIGURE 2. Principal component analysis (PCA) showing the distribution of all sampling 546 points at Nouragues (blue) and Paracou (green) and the loadings of the 44 biotic and abiotic 547 environmental variables (gray vectors). The contribution of soil fauna (mesofauna and meso- 548 plus macrofauna) on the decomposition of three litter combinations are included in this analysis 549 and highlighted in red for visualization. PC1 axis was mainly defined by nutrient-related 550 variables in the litter layer. Labels for the environmental vectors with the lowest loadings have 551 been removed for clarity. See Methods and Table S1 for variable descriptions and 552 abbreviations. 553 554 555 FIGURE 3. Relationship between the effect of soil fauna on decomposition (as the difference 556 between the litter mass loss in the litterbags with meso- and macrofauna relative to the 557 corresponding loss in the litterbags with only microbial access) with the PC1 scores of each 558 sampling point as a proxy of the relative nutrient availability in the litter microenvironment. 559 See Table 2 for model outputs. 560 561 562 FIGURE 4 (a) Differential effect of soil fauna on the decomposition of three litter combinations 563 differing in their C:P ratio. (b) Differential effect of soil mesofauna alone (<2 mm body width) 564 versus the combined effect of the meso- plus macrofauna. In both panels, the distribution of 565 fauna effects is modeled as a density function with highest or widest points having greater 566 probabilities within each categorical group. See Table 2 for model outputs. 567