Conversion of rainforest to oil palm and rubber plantations alters energy channels in soil food webs

Abstract In the last decades, lowland tropical rainforest has been converted in large into plantation systems. Despite the evident changes above ground, the effect of rainforest conversion on the channeling of energy in soil food webs was not studied. Here, we investigated community‐level neutral lipid fatty acid profiles in dominant soil fauna to track energy channels in rainforest, rubber, and oil palm plantations in Sumatra, Indonesia. Abundant macrofauna including Araneae, Chilopoda, and Diplopoda contained high amounts of plant and fungal biomarker fatty acids (FAs). Lumbricina had the lowest amount of plant, but the highest amount of animal‐synthesized C20 polyunsaturated FAs as compared to other soil taxa. Mesofauna detritivores (Collembola and Oribatida) contained high amounts of algal biomarker FAs. The differences in FA profiles between taxa were evident if data were analyzed across land‐use systems, suggesting that soil fauna of different size (macro‐ and mesofauna) are associated with different energy channels. Despite that, rainforest conversion changed the biomarker FA composition of soil fauna at the community level. Conversion of rainforest into oil palm plantations enhanced the plant energy channel in soil food webs and reduced the bacterial energy channel; conversion into rubber plantations reduced the AMF‐based energy channel. The changes in energy distribution within soil food webs may have significant implications for the functioning of tropical ecosystems and their response to environmental changes. At present, these responses are hard to predict considering the poor knowledge on structure and functioning of tropical soil food webs.


| INTRODUC TI ON
Indonesia is a hotspot of biodiversity on Earth and at the same time a top world producer of agricultural products, such as oil palm (Fitzherbert et al., 2008;Koh & Ghazoul, 2010) and rubber (Marimin, Darmawan, Machfud, Putra, & Wiguna, 2014). Extension of agriculture in Indonesia was associated with deforestation which increased strongly in the last 20 years and this is predicted to continue (Gatto, Wollni, & Qaim, 2015;Koh & Ghazoul, 2010). Since the 1980s, after the transmigration program, large parts of rainforest in Jambi province, Sumatra, have been converted to oil palm (16% of total area) and rubber plantations (12%; Gatto et al., 2015). Thus, Sumatra represents an ideal model region to investigate the effect of rainforest conversion on biodiversity and ecosystem functioning at local and regional scale in Southeast Asia Drescher et al., 2016).
Conversion of tropical rainforests into plantation systems is associated with changes in ecological niches of species and ultimately with the loss of species, and thereby with changes in ecosystem functioning (Barnes et al., 2014;Clough et al., 2016;Fitzherbert et al., 2008;Gilbert, 2012). Rainforest conversion strongly affects environmental processes, including soil organic carbon pools and soil erosion (Guillaume, Damris, & Kuzyakov, 2015), primary production of trees (Kotowska, Leuschner, Triadiati, Meriem, & Hertel, 2015), carbon dioxide and methane fluxes (Hassler et al., 2015), as well as nitrogen cycling and soil fertility (Allen, Corre, Tjoa, & Veldkamp, 2015). However, the effect of conversion of rainforest into plantations on belowground organisms is less well studied. It has been shown that rainforest conversion strongly affects biomass, vitality, and mycorrhizal colonization of roots (Sahner et al., 2015), biodiversity and abundance of microorganisms (Krashevska, Klarner, Widyastuti, Maraun, & Scheu, 2015;Schneider et al., 2015) and trophic-guild composition of litter-dwelling fauna (Barnes et al., 2014;Klarner et al., 2017). However, the effect of rainforest conversion on food resources of soil fauna and the relative importance of energy channels of soil food webs (Moore et al., 2004) is little understood. This is unfortunate, as the way energy is channeled through soil food webs is a major determinant of their stability (Ruiter, Neutel, & Moore, 1995;Moore & De Ruiter, 2012).
Rainforest conversion is likely to alter the flux of energy through soil food webs. For instance, nutrient inputs, changes in soil pH, as well as physical disturbances caused by management practices may shift fungal-and bacterial-based energy channels in soils (Rousk et al., 2010). In temperate regions, more fertile and productive ecosystems foster the bacterial-based energy channel, while less fertile ecosystems foster the fungal-based energy channel (Wardle, 2004). Previous studies showed that conversion of rainforest into oil palm plantations results in a decreased amount of specific bacterial biomarker PLFAs in the litter  and causes a diet shift of generalist predators toward more herbivore prey (Klarner et al., 2017). However, the resulting changes in energy channels of the entire soil food web have not been explored. Soil organisms interact in complex ways in soil with their activity regulating ecosystem processes and delivering ecosystem services (Lavelle et al., 2006). Food web models allow calculating energy and nutrient fluxes based on consumer-resource interactions and the way how energy and nutrients are channeled through soil microbial and animal communities (De Ruiter, Van Veen, Moore, Brussaard, & Hunt, 1993). This channeling has been shown to be altered significantly with land-use intensification (Wagg, Bender, Widmer, & Heijden, 2014). Building on these knowledge we explored changes in the channeling of energy through soil food webs after conversion of rainforest into plantation systems of rubber and oil palm.
Lipids play a vital role in animals both as storage compounds, for providing energy (neutral lipid fatty acids, NLFAs) and as structural component of cell membranes (phospholipid fatty acids, PLFAs; Ruess, Häggblom, Garcıá Zapata, & Dighton, 2002). Biomarker FAs are synthesized only by microorganisms and plants and transferred through food chains without modification, allowing to infer links between basal resources and higher order consumers (Chamberlain, Bull, Black, Ineson, & Evershed, 2005;Ruess & Chamberlain, 2010).
For soil food web, it has been shown that fatty acids (FAs) are transferred from microorganisms to microbivorous invertebrates to higher order consumers (Chamberlain et al., 2005;Ruess et al., 2002;Ruess, Schütz, et al., 2005;Ruess, Tiunov, et al., 2005) as well as from microorganisms and plants to detritivores and their predators (Pollierer, Scheu, & Haubert, 2010). Thus, lipid profiles of consumers may serve as tool for soil food web diagnostics (Kühn et al., 2018). To-date, however, virtually no studies investigated lipid profiles of soil fauna including both meso-and macrofauna in the same community.
Here, we used lipid profiles of soil fauna to track how energy channels in soil food webs are changing due to rainforest conversion into two major agricultural systems, that is, rubber and oil palm plantations, in Sumatra, Indonesia. We hypothesized that (a) different groups of meso-and macrofauna are linked to different basal resources with mesofauna detritivores, that is, Collembola and Oribatida, being closely associated with the fungal energy channel, and macrofauna detritivores such as Lumbricina and Diplopoda being more closely associated with the bacterial and plant energy channel. Further, based on results of previous studies (Klarner et al., 2017;Krashevska et al., 2015), which indicated reduced feeding on bacteria and increased feeding on plants with conversion of rainforest into oil palm plantations, we hypothesized (b) that rainforest conversion strengthens the plant-based energy channel and reduces the bacterial energy channel in soil food webs of plantation systems.

| Study sites
Soil and litter samples were taken in lowland rainforest, rubber (Hevea brasiliansis) plantations and oil palm (Elaeis guineensis) plantations, located in Jambi province, southwest Sumatra, Indonesia. Jambi province stretches from the Barisan mountain range in the west across extensive lowlands toward the southern Malacca Strait in the east. The climate is tropical and humid with rainy seasons from March to December, and a dryer period during July and August (Drescher et al., 2016). Study sites were located at similar altitude varying between 50 and 100 m a.s.l. in Harapan landscape; each system was replicated four times (see Drescher et al., 2016 for more details). Rainforest sites used as reference comprised secondary rainforest close to natural condition that underwent selective logging some 20-30 years ago. Rubber and oil palm plantations were intensively managed monocultures of an average age of 6-16 and 8-15 years, respectively (Drescher et al., 2016). Soils at the study sites were loam acrisols of low fertility .
Management practices in these smallholder monoculture plantations are described in detail by Allen et al. (2015). In the loam acrisol soil, oil palm plantations were established after clearing and burning the previous jungle rubber whereas the rubber plantations were established from previously logged forest. Oil palms are fertilized once in the rainy season and once in the dry season. The most commonly used fertilizers are NPK complete fertilizer (i.e., Phonska and Mahkota), potassium chloride (KCl) and urea (CO(NH 2 ) 2 ). Both manual and chemical weeding took place throughout the year at the rubber and oil palm plantations. The most commonly used herbicides were Gramoxone and Roundup; these were applied at an average rate of 2-5 L herbicide ha -1 year -1 Clough et al., 2016;Kotowska et al., 2015).

| Sampling
Samples were taken in September 2017 from three 5 × 5 m subplots within 50 × 50 m plots with a minimum distance of 200 m between plots established at each study site (Drescher et al., 2016). Each sample measured 16 × 16 cm and included the litter layer and underlying top soil to a depth of 5 cm. Since only few soil fauna were collected in oil palm plantation we performed an additional nonquantitative sampling of litter and soil in this system. Large soil fauna (>2 mm in body length; macrofauna) were collected by hand after sieving the litter through 2 cm mesh; small soil fauna (<2 mm; mesofauna) were extracted by heat (Kempson, Lloyd, & Gheraldi, 1963) Table S1 for the full list of taxa analyzed). To get enough material for lipid analysis, small fauna from different subplots and layers were bulked (in few cases, animals from two plots within the same system were combined); one sample consisted of 1-54 specimens of 0.5-2 mg fresh weight. After identification, animals were immediately transferred into pure methanol and stored at −20°C before fatty acid extraction (Zieger & Scheu, 2018

| Analysis of fatty acids and biomarker assignment
NLFAs were extracted as described in Ruess, Schütz, et al. (2005)), Ruess, Schütz, et al. (2005)). In brief, animals were placed in 5 ml single phase extraction solvent (chloroform, methanol, 0.05 M phosphate buffer at a ratio of 1:2:0.8; pH 7.4) and NLFAs were extracted overnight on a shaker. The solvent was then transferred to new tubes, and the extraction was repeated by shaking for 1-2 hr with an additional 2.5 ml of extraction solvent. Extraction solvents of both steps were combined, 0.8 ml of distilled water and 0.8 ml of CHCl 3 were added, and samples were centrifuged at 7°C in a multi centrifuge (402,48 g) for 5 min. Then, samples were allowed to stand to separate phases. The top two phases were removed and the chloroform fraction was transferred to a silica gel (SiOH) column (0.5 g, mesh size 100-200 µm). Lipids were eluted with 5 ml of chloroform, and the solvent was reduced by evaporation in a vacuum centrifuge.
NLFAs were saponified and methylated following the procedure given for the Sherlock Microbial Identification System (MIDI Inc., Newark, Del). For saponification a solution of sodium hydroxide/ methanol was used and the samples incubated at 100°C for 30 min followed by acid methanolysis in HCl-methanol at 80°C for 10 min.
The resulting fatty acid methyl esters (FAMEs) were stored at −20° until analysis.
FAMEs were transferred into vials, capped and analyzed by gas chromatography (Clarus 500, Perkin Elmer, Waltham, USA). Helium was used as carrier gas and NLFAs were identified by a flame ionization detector (capillary column: 30 m × 0.32 mm i.d., 0.25 mm film thickness; PE-5, Perkin Elmer). The temperature program started with a temperature of 60°C (held for 1 min) and increased by 30°C per min to 160°C followed by 3°C per min to 260°C. The injection temperature was 250°C and helium was used as carrier gas. On the basis of retention time and comparison with standard mixtures, 37 different FAMEs ranging from C11 to C24 and 26 different BAMEs ranging from C11 to C20 were identified. Hexadecadienoic acid methyl ester and methyl-hexadecatrienoate were used as algal biomarkers (SigmaeAldrich, St. Louis, USA).

| Statistical analysis
Statistical computations were done in R version 3.4.0 (R Core Team, 2017) with R Studio interface (R studio Inc.). To normalize the data across different taxa and sites, we analyzed proportions of total (%) of individual NLFAs instead of the initial peak areas.
We first calculated mean proportions for each NLFA for each group on each plot. In cases where more than one sample of the same group was analyzed per plot, we calculated mean proportions across these samples. We used the mean values for each taxon on each plot as replicates in statistical analyses (Table S2).

| Food resources of different groups of soil fauna
Twenty-three biomarker FAs were grouped into seven categories representing different basal resources and two indices were calculated: plant-to-fungi and fungi-to-bacteria (Table 1; Figure 1).
In general, oleic acid (18:1ω9) as relative biomarker FA for plants and linoleic acid (18:2ω6,9) as relative biomarker for fungi were the most common fatty acids in soil fauna across ecosystems and taxa (Table S3)

| Effect of land-use change on food resources of soil fauna
Land use affected the NLFA indices-based profiles, if all groups of soil fauna were analyzed together (MANOVA: F 2,14 = 1.82, p = .0469; Figure 2). The first axis of LDA, explaining 63% of distinction between land-use systems, was related to specific and nonspecific bacterial biomarker FAs with the proportion of the former being higher in animals from rainforest and the latter being higher in animals from plantation systems (Table S4)

| Food resources of different groups of soil fauna
The plant biomarker 18:1ω9 and the fungal biomarker 18:2ω6,9 were the most common biomarker FAs in the majority of soil fauna groups, which is in line with other studies (Pollierer, Dyckmans, Scheu, & Haubert, 2012;Ruess et al., 2002). Soils contain high amounts of plant residues which are colonized by fungi and serve as basal resources/food for many groups of detritivore soil fauna. Via predator-prey interactions, these resources are transferred to higher trophic levels (Pollierer et al., 2010;Ruess & Chamberlain, 2010).
Based on FA analysis, Ferlian, Klarner, Langeneckert, and Scheu (2015) reported plant, fungi, and bacteria as major resources for Collembola in a temperate ecosystem. Plant and fungal biomarker FAs represented 34% and 27% of total NLFAs in Collembola, respectively (Haubert et al., 2004), and up to 80% of total NLFAs in Oribatida (Pollierer et al., 2012). However, Ngosong, Raupp, Scheu, and Ruess (2009)   F I G U R E 4 Channeling of energy from basal resources into consumers in rainforest, rubber and oil palm plantations as indicated by fatty acid analysis. The width of channels to respective land use systems reflects the relative importance of the channels as indicated by the mean proportion of biomarker neutral lipid fatty acids (NLFAs) of total NLFAs in soil fauna in the respective land-use systems (data bulked across animal groups). The width of boxes/pictures of basal resources reflects the relative importance of these resources across land-use systems; note that it ignores potential differences in production and assimilation rate of different NLFAs among producer and consumer groups. Colors represent land-use systems: F-rainforest (gray), R-rubber plantation (brown), O-oil palm plantation (orange). Figure prepared using the R package bipartite feed on algae, with on average algal biomarker FAs comprising 6% of total NLFAs in Collembola and 4% in Oribatida. Photoautotrophic algae and bacteria may represent a considerable portion of the diet of Collembola in various habitats (Potapov, Korotkevich, & Tiunov, 2018;Schmidt, Dyckmans, & Schrader, 2016). A high palatability of algae for Collembola and Oribatida was demonstrated in laboratory experiments (Brückner, Schuster, Smit, & Heethof, 2018;Buse, Ruess, & Filser, 2013Scheu & Folger, 2004) and field stable isotope-based studies from temperate forest ecosystems suggested that algae serve as food resource for certain species of Collembola and Oribatida (Schneider et al., 2004;Maraun et al., 2011;Potapov, Semenina, Korotkevich, Kuznetsova, & Tiunov, 2016). However, field data on algal biomarker FAs from temperate ecosystems are scarce since biomarkers for algae only were suggested recently (Buse, Ruess, & Filser, 2013. Overall, the results indicate that the diet of Collembola and Oribatida in tropical ecosystems predominantly consists of bacteria and to some extent on algae, contrasting temperate and boreal ecosystems where these animals predominantly feed on fungi and dead organic matter. This shift in diet may be related to low food quality of litter in tropical ecosystems (Hättenschwiler, Coq, Barantal, & Handa, 2011;Illig, Reinhard, Norton, & Scheu, 2005;Krashevska et al., 2018;Marian, Sandmann, Krashevska, Maraun, & Scheu, 2017. Further, low quality of litter associated with low availability of fungi as food might be responsible for the lack of primary decomposers (Illig et al., 2005) and low abundance of Collembola and Oribatida in tropical ecosystems (Marian, Sandmann, Krashevska, Maraun, & Scheu, 2018).
Among the soil fauna groups studied, Lumbricina contained the highest amount (on average 22%) of C20 carbon polyunsaturated FAs (C20 PUFAs;20:4ω6,9,12,15 and 20:5ω3,6,9,12,15) which are assumed to be of animal origin (Albro, Schroeder, & Corbett, 1992;Sampedro, Jeannotte, & Whalen, 2006). As Lumbricina are detritivores that live in and feed on soil organic matter these FAs presumably are synthesized predominantly by the animals themselves as suggested earlier (Chamberlain et al., 2005;Petersen & Holmstrup, 2000) rather than originate from the diet. However, Sampedro et al. (2006) also found high concentrations of C20 PUFAs in the digestive tract of Lumbricina suggesting that at least in part they may also originate from the diet, possibly by feeding on protozoa and other soil microfauna (Aira, Monroy, & Dominguez, 2003;Bonkowski & Schaefer, 1997;Domínguez, Parmelee, & Edwards, 2003;Monroy, Aira, & Domínguez, 2008;Pokarzhevskii, 1997;Sampedro et al., 2006). Bacterial and fungal biomarker FAs were found to comprise up to 14%, whereas plant biomarker FAs only 6% of total NLFAs in Lumbricina. This also is in line with results of Sampedro et al. (2006) Zhang, Locati, Rouland, & Lavelle, 1997;Lavelle et al., 1987) also revealed that digestion system in association with soil microflora is very efficient and to digest soil organic matter through a mutualist Lumbricina microflora-digestion system and the intestinal mucus was supposed to play a central role in the process of digestion.
The similarity in lipid profiles and gut content of Lumbricina in our study and published data from temperate and tropical ecosystems suggests that Lumbricina occupy a similar trophic niche across biomes and function as trophic level omnivores feeding on detritus, microorganisms, and microfauna.
Araneae and Chilopoda are generalist predators which commonly feed on Collembola (Eitzinger, Rall, Traugott, & Scheu, 2018;Rusek, 1998;Voigtländer, 2011). Unexpectedly and contrasting our first hypothesis, the biomarker FA composition of both Araneae and Chilopoda was different from that of Collembola, and comprised high proportions of plant and fungal biomarker FAs.
Since Araneae are generalist predators, they can consume prey from both the decomposer and herbivore system (Oelbermann, Langel, & Scheu, 2008;Wise, 2004Wise, , 2006. The similarly high concentrations of plant biomarker FAs in Araneae and Chilopoda suggest that, in contrast to boreal and temperate systems, these predators heavily rely on herbivorous prey in tropical systems. Conform to this finding, generalist predators have been shown to be able to regulate populations of pest species in tropical agricultural systems (Panizzi & Corrêa-Ferreira, 1997;Sigsgaard, 2000).
In addition, predators contained higher amounts of C20 PUFAs than decomposer animals, suggesting that in addition to synthesizing these FAs they also incorporate significant amounts from their diet which makes C20 PUFAs a potential trophic level biomarker (but there are some remarkable exceptions, such as Lumbricina).

| Effect of land-use change on food resources of soil fauna
Despite the majority of NLFAs in soil fauna in all studied land-use systems originated from plants and fungi, the relative proportions of several NLFA biomarkers varied systematically among the systems. Due to the large spatial scale of the study and the limited number of replicates analyzed variability in NLFA, data were high. However, the fact that differences between land-use systems were significant suggests that changes would even be more pronounced if more data would have been available. The high abundance of plant biomarker FAs and high plant-to-fungi NLFA ratio suggests that the soil fauna is closely linked to plant basal resources in oil palm plantations. Supporting this conclusion, based on changes in stable isotope ratios, Klarner et al. (2017) found generalist litter-dwelling predators (Chilopoda) in oil palm plantations to more heavily rely on plant-based food chains than in rainforest. Our results suggest that this not only applies to generalist predators but also to the entire soil animal community of oil palm plantations. Presumably, this is due to the well-developed herb layer in oil palm plantations and occasional weeding increasing the input of high-quality plant material to the belowground system. Overall, the results indicate that conversion of rainforest to oil palm plantations increases the plant-based energy channel in soil food webs.
The AMF biomarker NLFA showed different trends in plant litter  and in soil fauna. Investigating the same study sites, Krashevska et al. (2015) found the AMF biomarker NLFA in litter to be lowest in rainforest litter and not to differ significantly between land-use systems in soil, whereas we found the proportion of AMF biomarker NLFAs in soil fauna to be highest in rainforest.
The Acrisol soils at our study sites are poor in nutrients (Allen, Corre, Kurniawan, Utami, & Veldkamp, 2016), potentially forcing AMF to exploit nutrients in litter. The AMF biomarker NLFA was low in fauna from rubber plantations indicating that soil fauna little feed on AMF in this land-use system. Low AMF abundance in rubber plantations likely is related to high root phosphorus concentration (Bolan, 1991;Li, Smith, Holloway, Zhu, & Smith, 2006) and reduced energy flow from leaves to roots due to cutting the phloem for collecting rubber.
Our results suggest that the root-based energy channel in soil food webs of rubber plantations is reduced (Figure 4).
In line with the study of Krashevska et al. (2015), we found the abundance of nonspecific bacterial biomarker FAs to be highest in oil palm plantations while the abundance of specific bacterial biomarker FAs decreased. Based on metagenomics, Schneider et al. (2015) also found certain bacteria taxa to be more abundant in oil palm plantations as compared to rainforest. The increase likely is due to more beneficial environmental factors, in particular increased soil pH (Nacke et al., 2011;Rousk et al., 2010), base saturation (Schneider et al., 2015) and/or fertilization (Shen, Zhang, Di, & He, 2012). The increase in soil pH in plantation systems as compared to rainforest presumably is due to ashes from plant biomass burning (van Straaten et al., 2015) and liming Krashevska et al., 2015).
The more open canopy and improved light conditions in oil palm plantations as compared to rainforest may also favor photosynthetically active bacteria (Schneider et al., 2015).  (Pollierer et al., 2010). Olsson (1999) also reported that AMF can synthesize vaccenic type FA 16:1ω7. Although soil pH and base saturation is higher in rubber and oil palm plantations, rainforest has higher soil water content, soil C and N concentrations and amount of litter van Straaten et al., 2015), resulting in more favorable conditions for bacteria and thereby strengthening the bacterial energy channel (Wardle et al., 2004; Figure 4). Conform to these findings, the absolute biomarker FAs for Gram+ (a15:0, i16:0, i17:0) and Gram-bacteria (cy17:0) in soil fauna significantly declined from rainforest to rubber to oil palm plantations (Table S5). This decline mirrors the decrease in biomarker PLFAs cy17:0 and i17:0 in litter of rubber and oil palm plantations as compared to rainforest , suggesting that soil meso-and macrofauna nonselectively feed on bacteria.

| CON CLUS IONS
The results suggest that different high-rank taxonomic groups are linked to different resources. Fungi appear to contribute only little to the nutrition of detritivore mesofauna taxa, in particular Collembola, contrasting temperate and boreal ecosystems. fueling soil food webs appears to be particularly low in rubber plantations, presumably due to alleviated limitation by phosphorus and by cutting the phloem for harvesting rubber. These alterations suggest that land-use change alters food-web structure and may disrupt interaction networks resulting in changes in ecosystem functioning and food-web resilience that are hard to predict with the current limited knowledge on tropical belowground systems.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
AP designed the study and performed field sampling. WS analyzed fatty acid composition of fauna and wrote the manuscript. All authors contributed critically to drafts and gave their final approvement for publication.