Leaf litter identity rather than diversity shapes microbial functions and microarthropod abundance in tropical montane rainforests

Abstract In tropical forest ecosystems leaf litter from a large variety of species enters the decomposer system, however, the impact of leaf litter diversity on the abundance and activity of soil organisms during decomposition is little known. We investigated the effect of leaf litter diversity and identity on microbial functions and the abundance of microarthropods in Ecuadorian tropical montane rainforests. We used litterbags filled with leaves of six native tree species (Cecropia andina, Dictyocaryum lamarckianum, Myrcia pubescens, Cavendishia zamorensis, Graffenrieda emarginata, and Clusia spp.) and incubated monocultures and all possible two‐ and four‐species combinations in the field for 6 and 12 months. Mass loss, microbial biomass, basal respiration, metabolic quotient, and the slope of microbial growth after glucose addition, as well as the abundance of microarthropods (Acari and Collembola), were measured at both sampling dates. Leaf litter diversity significantly increased mass loss after 6 months of exposure, but reduced microbial biomass after 12 months of exposure. Leaf litter species identity significantly changed both microbial activity and microarthropod abundance with species of high quality (low C‐to‐N ratio), such as C. andina, improving resource quality as indicated by lower metabolic quotient and higher abundance of microarthropods. Nonetheless, species of low quality, such as Clusia spp., also increased the abundance of Oribatida suggesting that leaf litter chemical composition alone is insufficient to explain variation in the abundances of soil microarthropods. Overall, the results provide evidence that decomposition and microbial biomass in litter respond to leaf litter diversity as well as litter identity (chemical and physical characteristics), while microarthropods respond only to litter identity but not litter diversity.


| INTRODUC TI ON
The great majority of plant material enters the soil as litter, in the form of leaves, stems, and roots. Decomposition of these materials is an essential process for nutrient cycling and provides the basal resources of the soil food web (Berg et al., 1993;Berg & McClaugherty, 2008). In addition to providing food resources, leaf litter accumulating on the soil surface forms a variety of microhabitats for soil organisms, with more diverse litter materials increasing habitat variability, but also providing the opportunity for enhanced nutrient acquisition (Bardgett, 2005;Gessner et al., 2010). Therefore, high diversity of leaf litter in mixtures is expected to be an important determinant of the diversity and structure of decomposer communities and, consequently, litter decomposition (Gessner et al., 2010;Trogisch et al., 2016).
Tropical montane rainforest ecosystems harbor an exceptional diversity of plant species (Beck & Ritcher, 2008;Homeier et al., 2008;Myers et al., 2000) and are associated with high numbers of animal species above-and belowground (Brehm et al., 2008;Maraun et al., 2008;Paulsch & Müller-Hohenstein, 2008). However, the effect of plant litter diversity on decomposer communities and decomposition of litter in this ecosystems is little studied Krashevska et al., 2017). Controlled experiments are needed to assess the effect of diversity and composition of litter species in mixtures on litter decomposition and microarthropod abundance.
Differences in leaf litter chemical composition are recognized as the main drivers of decomposition rates at the ecosystem level (Coûteaux et al., 1995;. Studies have reported positive, negative, but also no effects of litter mixtures on decomposition, with mixture effects typically related to variations in litter nutrient concentrations (Gartner & Cardon, 2004;Handa et al., 2014;Makkonen et al., 2012). However, differences in litter chemistry are not the only factors contributing to variations in litter decomposition in mixtures (Hättenschwiler, 2005;Hoorens et al., 2003). Physical leaf litter traits, such as toughness, surface structure, and shape, also contribute to microhabitat diversity and modify microenvironmental conditions of decomposer organisms, resulting in either accelerated or decelerated litter decomposition (Hansen & Coleman, 1998;Kaneko & Salamanca, 1999). Therefore, species identity, which encompasses chemical and physical characteristics, may well explain diversity effects on decomposition.
Commonly, studies investigating effects of litter diversity on litter decomposition focused on microorganisms and detritivore invertebrates (Gessner et al., 2010). Microorganisms are assumed to respond more sensitively to litter diversity than invertebrates as they directly depend on the variety of litter chemical compounds needed for metabolism and growth (Bardgett & Shine, 1999;Chapman et al., 2013). By contrast, the response of invertebrate detritivores, particularly the key decomposer groups Acari and Collembola, more strongly depends on the identity rather than diversity of leaf litter species and varies with the stage of litter decomposition (González & Seastedt, 2001;Illig et al., 2008;Kaneko & Salamanca, 1999;Korboulewsky et al., 2016;Wardle et al., 2006). Indeed, many decomposer microarthropods have the ability to select among co-occurring leaf litter species according to litter palatability and/or the microorganisms colonizing the litter (Klironomos et al., 1992;Korboulewsky et al., 2016;Schneider & Maraun, 2005).
Studies linking microbial-dominated litter decomposition processes and colonization of litter by detritivore invertebrates are needed to uncover the mechanisms responsible for litter diversity effects on the structure and functioning of the decomposer system, particularly in tropical ecosystems characterized by high diversity of plant (tree) species.
In the present study, we investigated the effect of leaf litter diversity and identity on the colonization of litter by microorganisms and microarthropods including Acari and Collembola after 6 and 12 months of incubation in Ecuadorian montane rainforests. We hypothesized that (1) microbial growth and activity increase with litter diversity, but that the abundance of both Acari and Collembola relies more on litter identity. Additionally, assuming that microorganisms are limited by multiple nutrients (Demoling et al., 2007;Krashevska et al., 2010), we hypothesized that (2) nutrient availability increases and microbial stress conditions decrease with time and that (3) the presence of high-quality litter benefits microorganisms. Further, assuming that Acari and Collembola prefer similar food resources and consume both leaf litter tissue and microorganisms (Dhooria, 2016;Ruess & Lussenhop, 2005;Seastedt, 1984), we hypothesized that (4) the abundance of Acari and Collembola increases as decomposition proceeds, particularly in presence of high-quality litter.

| Study site
The study area is located in southern Ecuador on the eastern slopes of the Andean Cordillera. The site forms part of the Reserva Biológia San Francisco located on the northern borders of the Podocarpus National Park at 2,000 m a.s.l. (3°58′S, 79°04′W). The region is characterized by a semihumid climate with annual precipitation of about 2,200 mm and average annual temperature of 15.2°C (Bendix et al., 2006;Wullaert et al., 2009). The soil is Gley Cambisol with a soil pH of ~3.5 and a thick organic layer up to 35 cm comprised of mainly fermentation/humus material overlaid by litter material (Moser et al., 2007). The tropical rainforest is mostly undisturbed and holds an exceptionally high diversity of fauna and flora with Rubiaceae, Melastomataceae, and Piperaceae as dominant plant families (Beck & Ritcher, 2008;Brehm & Fiedler, 2005;Homeier et al., 2010;Maraun et al., 2008). increasing C-to-N ratio, see Appendix 1] were collected, dried (60°C for 72 hr), and used to fill 20 × 20 cm and 4 mm nylon mesh litterbags. Initial chemical composition of the litter species is given in Appendix 1. The leaves used had no signs of herbivory, fungal infection or atypical texture or color. Large leaves exceeding the size of the litter bags were cut into ~5 × 5 cm pieces. Single-species litterbags (12 g each) and mixtures with all possible two-(6 g per species) and four-species combinations (3 g per species) were prepared, resulting in a total of 36 litterbag types with three levels of species diversity (1, 2, and 4 leaf litter species). Litterbags were randomly placed in the field on top of the undisturbed litter layer and fixed with nails in four blocks. Minimum distance between the blocks was 20 m. One replicate of each treatment was harvested after 6 and 12 months.

| Analytical procedures
After harvest, material in each litterbag was separated into two subsamples of equal weight, disturbing the fauna as little as possible but ensuring that all litter types were present in both halves. One half was used for microarthropod extraction and the other for analysis of microbial parameters. Microarthropods were extracted by heat over one week using a modified high gradient extractor and then stored in 70% ethanol (Kempson et al., 1963;Macfadyen, 1961). Microarthropods were determined to group level [Collembola (Insecta), Oribatida, Mesostigmata, and Prostigmata (Acari)] using Schaefer (2018). The dry litter was sorted to species, weighed and used to measure litter chemical composition.
Microbial basal respiration (BR) and microbial biomass (C mic ) were determined using an automated respirometer system (Scheu, 1992). BR (μl O 2 g −1 dry weight hr −1 ) was measured at 22°C and calculated as mean of O 2 consumption rates 10 to 20 hr after attachment of the samples to the respirometer system. C mic was measured by the substrate-induced respiration method (SIR; Anderson & Domsch, 1978;Beck et al., 1997). The maximum initial respiratory response (MIRR; µl O 2 g −1 dry weight hr −1 ) was measured at 22°C after the addition of glucose to saturate the catabolic activity of microorganisms. MIRR was calculated as the average of the lowest three readings within the first 10 hr, and C mic was calculated as C mic = 38 × MIRR (mg/g dry weight). Respiration rates between the lowest (usually 3-6 hr after glucose addition) and highest reading were taken to calculate the slope of microbial growth (+C Slope ). Data were ln-transformed, and the slope determined by linear regression. The microbial metabolic quotient (qO 2 ; μl O 2 mg −1 C mic hr −1 ) was calculated by dividing BR by C mic .
Leaf litter mass loss (M loss ) was calculated as M loss (%) = (m 0 -m 1 /m 0 ) × 100, where m 0 is the initial dry weight and m 1 the dry weight of leaf litter at harvest. To measure chemical composition, leaves from each of the six species were dried (65°C for 72 hr) and milled to particles <1 mm. Carbon (C) and nitrogen (N) were measured using a CN elemental analyzer (Vario EL III, Elementar).
Total element analysis was measured by an ICP-OES system (ICP-OES, Optima 5300 DV, Perkin Elmer). Lignin and cellulose concentration were measured based on the methanol-chloroform-water (2:2:1) extraction method detailed in Allen et al. (1974). For litter mixtures, the proportion of elements per litterbag was calculated by proportionally summing the amount of the respective elements in the individual litter species. The chemical concentrations of elements, lignin and cellulose, were expressed as milligram per gram litter dry weight (dw).

| Statistical analyses
Analyses were performed using R version 3.6.0 (R Core Team, 2014).
In each model, the fixed factors litter diversity (LD; 1, 2, and 4 litter species), time of exposure (6 and 12 months), and the presence/ absence all leaf litter species (litter identity; 1,0; CA, DL, MP, CZ, GE, and Cs), as well as the interactions (time × LD and time × litter identity), were fitted in a hierarchical design. Block was fitted first as random factor followed by the fixed factors litter diversity, time, interaction between litter diversity and time, and litter identity. To assess the relative importance of the six leaf litter species, analyses were repeated changing the order of fitting individual litter species and their interactions. F-and p-values for individual litter species in the text and tables refer to those when fitted first (Schmid et al., 2002(Schmid et al., , 2017. Differences between means were inspected using Tukey's honestly significant difference test (package "emmeans"). Values presented in text are means ± SD of non−transformed data. Pearson correlation coefficients were calculated to investigate relationships between C-to-N ratio, C mic , qO 2 and M loss , and the abundance of Collembola and Acari (package "stats").

| Initial litter chemistry
Initial N concentrations were highest in C. andina, followed by D.

| Mass loss
Generally, M loss was higher after 12 than after 6 months of incubation with averages of 52.6% ± 7.1% and 41.8% ± 6.9% of initial, respectively (Table 1). M loss varied significantly with species diversity but the effect depended on time (Table 1; Figure 1a); after 6 months M loss was lower in single species (average of 29.6% ± 6.9%) compared to the two and four litter species treatments (43.1% ± 3.8 and 44.9% ± 3.6%, respectively), while after 12 months decomposition was similar in each of the litter diversity treatments. Further, M loss varied significantly with litter species identity; however, this depended on time, with the effect generally being restricted to the first sampling date and to four of the six litter species (Table 1). At the first sampling date, M loss increased in presence of C. andina from 39.7% ± 7.4% to 44.4% ± 5.1%, in presence of C. zamorensis from 40.5% ± 7.9% to 43.2% ± 5.3%, in presence of G. emarginata from 39.4% ± 7.6% to 44.8% ± 4.2%, and in presence of Clusia spp. from 39.6% ± 7.3% to 44.6% ± 5.1%. M loss positively correlated with C mic , BR, qO 2 , +C Slope, and the abundance of Collembola and Oribatida, but negatively with the litter C-to-N ratio (Pearson correlation coefficients; Table 2).

| Microbial parameters
Parallel to M loss , the microbial parameters C mic , BR, qO 2, and +C Slope significantly increased from 6 to 12 months (Table 1; for means see Appendix 2). Among microbial parameters, only C mic varied with litter diversity. Unlike M loss , the effect of litter diversity was restricted to the second sampling date, decreasing in the order one > two > four litter species (Figure 1b). Further, C mic also varied with litter species identity, but the effect was restricted to treatments with G. emarginata and depended on time. At the second sampling date, C mic decreased from 15.23 ± 11.74 to 11.58 ± 7.37 mg C mic g −1 dw in litterbags without and with G. emarginata, respectively.
The other microbial parameters only were significantly affected by litter species identity, with the effects in part varying with time (  the litter C-to-N ratio. BR positively correlated with M loss , C mic, qO 2 , +C Slope, and the abundance of Oribatida, but negatively with the abundance of Mesostigmata and the litter C-to-N ratio. qO 2 positively correlated with M loss and BR, but negatively with C mic and the abundance of Mesostigmata. +C Slope positively correlated with M loss , C mic , and BR, but negatively with the litter C-to-N ratio (Table 2).

| Microarthropods
The number of Collembola, Oribatida, and Prostigmata significantly increased from 6 to 12 months, but the abundance of Mesostigmata decreased (Table 3; Figure 2; for means, see Appendix 3). None of the soil microarthropod taxa investigated varied with litter diversity, although they did vary significantly with litter species identity (  (Table 3); in the presence of these species, the reduction was most pronounced after 12 months (from 60 ± 42 to 123 ± 132 and from 62 ± 38 to 124 ± 135 ind. 10 g −1 litter dw, respectively).

| Litter diversity
Contrary to our first hypothesis, C mic decreased rather than increased with increasing litter diversity after one year of exposure in the field ( Figure 1). Leaves of tropical forest trees are of low nutritional quality and contain high concentrations of structural compounds and secondary metabolites, typically higher than those in trees of temperate forests (Cárdenas et al., 2015;Coley & Barone, 1996;Hallam & Read, 2006). Secondary metabolites, particularly polyphenols known to suppress microorganisms by inhibiting enzyme activity (Hättenschwiler & Vitousek, 2000;Hoorens et al., 2003), are important drivers of decomposition processes particularly in tropical rainforests (Coq et al., 2010 (Butenschoen et al., 2014;Marian et al., 2017).
Similar to C mic , M loss significantly increased in single litter species treatments after one year of exposure underscoring the correlation between (Table 2). Changes in the chemical composition of litter material throughout the decomposition process alter the structure and functioning of microbial communities and thus affect the rate at which litter material is decomposed (Berg & McClaugherty, 2008). Notably, M loss increased with litter diversity after 6 months of exposure; however, the effect was no longer present after 12 months. Presumably, this reflects reliance of the early microbial community on labile litter compounds, which were more abundant in leaf litter mixtures (Pérez Harguindeguy et al., 2008;Rinkes et al., 2014). However, as decomposition proceeded, the remaining more recalcitrant compounds accumulated and their decomposition was independent of litter diversity.
In contrast with C mic and M loss , the abundance of microarthropods was not affected by litter diversity (Table 3). Some previous studies found mixtures to promote the abundance of microarthropods (Hansen, 2000;Hättenschwiler & Gasser, 2005;Migge et al., 1998;Schädler & Brandl, 2005), while others did not find evidence that litter diversity beneficially affects microarthropods (Bluhm et al., 2019;Ilieva-Makulec et al., 2006;Korboulewsky et al., 2016;Patoine et al., 2020;Scheu et al., 2003). Our results agree with the latter findings and support the results of Marian et al. (2018) suggesting that litter diversity in this tropical rainforest neither improves habitat conditions nor the availability of resources for microarthropods, at least during early stages of decomposition. Indeed, detritivore microarthropods are considered to comprise predominantly generalist feeders colonizing a range of forest types and therefore are rather insensitive to changes caused by litter mixing (Ball et al., 2014;Gergócs & Hufnagel, 2016;Patoine et al., 2020;Wardle et al., 2006). However, even though litter diversity did not affect microarthropod abundance, it may still have fostered the diversity of microarthropods, as has been shown for other soil organisms, such as testate amoebae at our study site .

| Exposure time
Generally, M loss increased with time parallel to microbial parameters.
Litter decomposition at our study site can be divided into three phases, with the early phase lasting for about 12 months . This early phase of decomposition is characterized by the loss of labile C compounds via leaching and by the growth of opportunistic microorganisms that form new soluble compounds (Berg & McClaugherty, 2008), and this likely explains the close link between M loss and microbial activity and growth (Table 2). However, contrary to our second hypothesis, the increase in qO 2 values between 6 and 12 months of exposure indicates that microorganisms increasingly suffered from stress conditions later during exposure. Stress conditions result in less efficient use of C compounds and increased investment into maintenance metabolism (Ndaw et al., 2009;Yan et al., 2003). Presumably, toward the end of the early litter decomposition stage microorganisms increasingly competed for resources as easily decomposable leaf litter compounds vanished (Fontaine et al., 2003;Poll et al., 2008;Rinkes et al., 2011). The parallel increase in the +C Slope with time suggests that this was associated with less efficient nutrient capture by microorganisms pointing toward a switch from predominant limitation by nutrients early during exposure to the limitation by easily available carbon resources later (Laganière et al., 2010;Sall et al., 2003). Early stages of litter decay in the studied tropical montane rainforest might be associated with high abundance of mycorrhizal fungi . The C input that mycorrhizal fungi obtain from plants may allow them to efficiently compete with saprotrophic fungi for nutrients, even though their enzymatic capability is typically inferior to that of saprotrophic fungi (Camenzind & Rillig, 2013;Hodge et al., 2001). Indeed, the assumption that mycorrhizal and saprotrophic fungi interact antagonistically early during litter decomposition at our study site is supported by earlier studies (Marian et al., 2019;Sánchez-Galindo et al., 2019).
Parallel to microbial parameters, the abundance of all microarthropod taxa studied increased with time, with the exception of Mesostigmata. Mesostigmata commonly hunt in the litter for other microarthropods, particularly Collembola, Astigmata and weakly sclerotized Oribatida (Koehler, 1997;Schneider & Maraun, 2009).
Although variations in the abundance of Mesostigmata were closely linked to the abundance of Collembola and Oribatida (Table 2), the fact that their abundance decreased with time likely reflects that Mesostigmata in the litterbags were not only feeding on microarthropods, but also on other organisms, presumably Nematoda, insect larvae and eggs. Indeed, some species of Mesostigmata may preferentially colonize certain microhabitats to hunt for prey such as Nematoda (Heidemann et al., 2014;Klarner et al., 2013).
The increase in the abundance of the microarthropod decomposers Collembola and Oribatida with time indicates that changes during the initial stages of decomposition influence both groups in a similar way. Surprisingly, Collembola and Oribatida abundance was not closely associated with microbial biomass (Table 2) even though microorganisms are their major food resource (Dhooria, 2016;Maraun et al., 2003;Scheu et al., 2005). Rather, the stage of litter decomposition within the early decomposition phase (i.e., 6 vs. 12 months) appears to be the more important driver of the abundance of microarthropod decomposers. Indeed, litter material that is highly colonized by microorganisms becomes more palatable for microarthropods (Bardgett, 2005;Das & Joy, 2009), which at least in part is due to the reduction in plant secondary compounds such as phenols (Asplund et al., 2013;Coulis et al., 2009). Overall, our results support earlier findings at this study site in that the role of litter resources for the nutrition of decomposer microarthropods increases with litter decomposition (Marian et al., 2018). Moreover, the parallel increase in the abundance of Prostigmata suggests that the increase in the abundance of decomposer microarthropod prey benefitted higher trophic levels.

| Leaf litter identity
The presence of specific plant leaf litter species in mixtures might increase or decrease the rate at which the litter decomposes (Hector et al., 2000;Hoorens et al., 2003Hoorens et al., , 2010. Variation can be attributed predominantly to differences in litter quality among the component species in mixtures (Gartner & Cardon, 2004;. Indeed, litter decomposition and colonization of the litter by microarthropods in our study were related to the initial chemical composition of the litter species. Our third hypothesis was supported by the beneficial effects of high-quality C. andina litter. Presence of this litter species significantly decreased qO 2 values and increased the abundance of Collembola and Prostigmata. C. andina had high initial N and P concentrations, and low lignin content (see Appendix 1), providing readily available nutrients, reducing nutrient stress for microorganisms, and thereby contributing to an increase in C mic . Increased microbial C use efficiency may also have resulted from a shift in microbial community composition toward highenergy-efficient species (Dilly & Munch, 1996), for example, from opportunistic bacteria to fungi able to break down complex litter compounds (Chapman et al., 2013). Changes in microbial community composition probably were driven by increasing concentrations of recalcitrant litter compounds favoring saprotrophic fungi able to degrade these compounds, which in turn beneficially affected decomposers, such as Collembola and Oribatida, feeding on these fungi and the litter materials degraded by them.
The high qO 2 and the +C Slope values after 12 months of exposure reflected the low quality of D. lamarckianum, C. zamorensis, and G. emarginata litter, and presumable scarcity of easily accessible C resources to microorganisms. All these litter species were characterized by low initial N and P concentrations, and high concentrations of lignin and cellulose (Appendix 1). The concentrations of lignin and cellulose serve as indicator of litter quality and as predictor of litter decomposition (Berg, 2014;Fioretto et al., 2005). Cellulose not entrapped in lignin degrades rapidly during early stages of decomposition, and this contributes to the release of N and P, typical elements limiting microbial growth (Berg, 2014;Berg & McClaugherty, 2008;Hobbie et al., 2012). However, during this stage, labile compounds are commonly used by opportunistic microorganisms (Cornelissen et al., 1999;Fioretto et al., 2005), impeding the growth of microorganism able to degrade recalcitrant litter compounds (Ilieva-Makulec et al., 2006). Therefore, by the end of the early stage of litter decomposition, structural compounds become relatively more abundant and reduce resource quality, which differentially affects microorganisms and microarthropods, as indicated by the lower abundance of Collembola in litter of C. zamorensis and D. lamarckianum. Interestingly, the decrease in C mic after 12 months in litterbags containing G. emarginata was associated with high abundance of decomposer microarthropods, suggesting that there is no close relationship between decomposer microarthropods and bulk microbial biomass in litter. This conclusion is also supported by the lack of significant correlations between C mic and decomposer microarthropod abundances ( Table 2).
The correlation between the abundance of Collembola and Oribatida and litter M loss presumably reflects that these microarthropods benefited from both higher quality litter and by microorganisms colonizing the litter at later stages of decay. The significant negative correlation between Collembola abundance and litter C-to-N ratio (Table 2) indicates that Collembola heavily rely on litter quality. However, contrary to our fourth hypothesis, the differential responses of microarthropods to litter species suggest that leaf litter chemical composition alone is insufficient to explain variations in the abundance of soil microarthropods, as has been suggested in earlier studies (González & Seastedt, 2001;Hoorens et al., 2010;Kaneko & Salamanca, 1999). This is most strongly supported by the greater abundance of Oribatida in litterbags containing Clusia spp. litter, which was of particular low quality. This indicates that physical litter characteristics such as toughness and structure might play a more important role in driving soil microarthropod abundance than litter chemistry and the degree of microbial colonization.

| CON CLUS IONS
The results of our study showed that higher levels of litter diversity may negatively affect soil microbial biomass and mass loss in the studied tropical montane rainforest, presumably due to the accumulation of recalcitrant compounds and the generally low quality of the leaf litter material. Notably, the response of microbial parameters and microarthropod abundance to litter identity was more pronounced than to litter diversity, with the differential responses of soil biota to litter identity in part being due to differences in the initial chemical composition of litter species. Generally, the results indicate that both microarthropods and microorganisms benefit from larger amounts of easily available litter resources during early stages of decomposition, highlighting the importance of litter quality as driver of the abundance and activity of decomposer organisms. However, the results also indicate that litter traits, related to the physical structure of litter, may be more important to decomposer invertebrates than litter chemistry and gross microbial characteristics of litter such as microbial biomass. Overall, our findings indicate that litter species identity functions as major driver of the abundance and activity of soil organisms, and thereby exert distinct effects on ecosystem processes such as decomposition and nutrient mobilization.

ACK N OWLED G M ENTS
We thank the Deutsche Forschungsgemeinschaft (DFG; FOR816) for financial support. Further, we thank the Ministerio de Ambiente del Ecuador and the Universidad Técnico Particular de Loja (UTPL) for the research permits and the center Naturaleza y Cultura Internacional (NCI) for access to the San Francisco reserve. We thank student helpers for the establishment of the experiment in the field and Christina Lucas for her help in the laboratory.

CO N FLI C T O F I NTE R E S T
None declared. Writing-original draft (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are available as electronic supplementary material. A P P E N D I X 2 (Continued)