Parthenogenetic vs. sexual reproduction in oribatid mite communities

Abstract The dominance of sex in Metazoa is enigmatic. Sexual species allocate resources to the production of males, while potentially facing negative effects such as the loss of well‐adapted genotypes due to recombination, and exposure to diseases and predators during mating. Two major hypotheses have been put forward to explain the advantages of parthenogenetic versus sexual reproduction in animals, that is, the Red Queen hypothesis and the Tangled Bank/Structured Resource Theory of Sex. The Red Queen hypothesis assumes that antagonistic predator—prey/ parasite–host interactions favor sex. The Structured Resource Theory of Sex predicts sexual reproduction to be favored if resources are in short supply and aggregated in space. In soil, a remarkable number of invertebrates reproduce by parthenogenesis, and this pattern is most pronounced in oribatid mites (Oribatida, Acari). Oribatid mites are abundant in virtually any soil across very different habitats, and include many sexual and parthenogenetic (thelytokous) species. Thereby, they represent an ideal model group to investigate the role of sexual versus parthenogenetic reproduction across different ecosystems and habitats. Here, we compiled data on oribatid mite communities from different ecosystems and habitats across biomes, including tropical rainforests, temperate forests, grasslands, arable fields, salt marshes, bogs, caves, and deadwood. Based on the compiled dataset, we analyzed if the percentage of parthenogenetic species and the percentage of individuals of parthenogenetic species are related to total oribatid mite density, species number, and other potential driving factors of the reproductive mode including altitude and latitude. We then interpret the results in support of either the Red Queen hypothesis or the Structured Resource Theory of Sex. Overall, the data showed that low density of oribatid mites due to harsh environmental conditions is associated with high frequency of parthenogenesis supporting predictions of the Structured Resource Theory of Sex rather than the Red Queen hypothesis.

Parthenogenetic reproduction has advantages that should, in theory, result in rapid replacement of sexual taxa. The advantages include doubled reproductive potential, no exposure to hazards during mating, easier colonization of new habitats, and maintenance of favorable gene combinations (Maynard Smith, 1978;Mirzaghaderi & Hörandl, 2016). A number of theories have been put forward to explain the "paradox of sex," but no consensus has been reached and some researchers doubt the existence of a unifying theory (Neiman et al., 2017;West, Lively, & Read, 1999). Among theories, those proposing short-term generational advantages of sexual reproduction have received most attention (Lehtonen et al., 2012), while theories based on long-term multigenerational advantages and group selection arguments have lost appeal (Becks & Agrawal, 2011;Bell, 1982;West et al., 1999).
Two major theories based on short-term advantages of sexual reproduction include (a) the Red Queen Theory (RQT), and (b) the Tangled Bank Theory (TBT) and Structured Resource Theory of Sex (SRTS), the latter two being based on similar mechanisms and therefore treated together here (Neimann et al., 2017;Song, Drossel, & Scheu, 2011). The RQT suggests that the genetic diversity generated by sexual reproduction reduces the ability of antagonists (parasites, predators) of species to adapt to host genotypes. It therefore predicts sexual taxa to dominate in habitats where parasite-host and/ or predator-prey interactions control community dynamics (Hamilton, 1980;Jaenike, 1978), which is often the case when predators/ parasites reach high densities (Ladle, 1992). The RQT does not make explicit predictions on the effects of resource availability, but it has been suggested that small populations are less infested by parasites (Arneberg, Skorping, Grenfell, & Read, 1998) thereby weakening Red Queen processes. A number of field and laboratory studies supported the RQT (Gibson & Fuentes, 2015;Haafke, Chakra, & Becks, 2016;Kotusz et al., 2014), but many are based on a single parasite-host system, the snail Potamogyrgus antipodarum and its trematode parasite (Jokela, Dybdahl, & Lively, 2009;King, Delph, Jokela, & Lively, 2009;Lively, 2010).
The TBT (Bell, 1982;Ghiselin, 1974) predicts sexual reproduction to dominate in spatially heterogeneous habitats where diverse niches are more likely to be successfully colonized by genetically diverse offspring produced via sexual reproduction. The TBT does not make explicit predictions on the role of biotic interactions or resource levels of the habitat of sexual versus parthenogenetic species. Building on the TBT, the SRTS explicitly includes resource dynamics and density-dependent population regulation, but its predictions are essentially similar to those of the TBT (Scheu & Drossel, 2007). The SRTS predicts that sexual reproduction should become more frequent in spatially structured habitats where resources are structured and renewed at low rates. The implication of this theory is that density-dependent population regulation is controlling the frequency of sexual and parthenogenetic reproduction. Both SRTS and RQT predict sexual reproduction to dominate when biotic interactions are strong, but the SRTS is based on bottom-up control of population regulation due to resource limitation, whereas the RQT is based on top-down control of prey species by predators and parasites. The SRTS explicitly predicts parthenogenesis to dominate in populations that are structured by harsh abiotic conditions since the advantage of sexual reproduction vanishes in the absence of resource limitation. The RQT makes no explicit predictions about the effects of environmental harshness, but harsh conditions typically reduce predator-prey/ parasite-host interactions thereby weakening the advantages of sexual reproduction. Importantly, the RQT and the SRTS are not mutually exclusive. In fact, their predictions partly overlap and some field observations might be explained by either of them (Neimann et al., 2017). Further, from the perspective of resources, parasites may be viewed as agents reducing the resource supply of the host; as a consequence, Red Queen processes may be viewed as forming part of the SRTS (Scheu & Drossel, 2007;Song et al., 2011).
The predictions of the two theories can be examined by studying the relative frequency of sexual or parthenogenetic reproduction in animal taxa across biomes and environmental gradients. These two theories can be tested in a number of soil animals including oribatid mites, collembolans, nematodes, earthworms, and isopods (Bell, 1982;Song et al., 2011). Arguably, the most powerful model taxon in this respect is oribatid mites, an abundant and diverse group of predominantly detritivorous and fungivorous soil arachnids. They occur in virtually any ecosystem including forests, grassland, tundra, taiga, agricultural sites, fresh water, bogs and marshes, and marine shorelines, and in a very wide range of microhabitats including different soil layers, decaying wood, fungal sporophores, epilithic, and epiphytic plants, bark of trees and suspended soils (Norton & Behan-Pelletier, 2009).
Oribatid mites either reproduce sexually or via thelytoky; other reproductive modes such as arrhenotoky, cyclical parthenogenesis as well as geographic parthenogenesis are unknown (Norton, Kethley, Johnston, & O'Connor, 1993). About 10% of the known species of oribatid mites are assumed to be obligate parthenogenetic, with more than 90% of the individuals of oribatid mite communities in some ecosystems comprising parthenogenetic species (Maraun, Norton and Palmer (1991) highlighted that parthenogenetic species dominate many early-derivative taxa and, rather than each being recent derivatives of an ancestral sexual lineage, most extant parthenogenetic species form part of clusters of parthenogenetic species. These taxa represent phylogenetic groups of different size, from small to speciose, indicating past parthenogenetic radiations (Heethoff, Norton, Scheu, & Maraun, 2009;Pachl et al., 2017;Schaefer, Norton, Scheu, & Maraun, 2010). Furthermore, parthenogenetic species occur in habitats "where they should not be" according to most theories, for example, in climax habitats such as old-growth forests (Huhta et al., 1986) and at low altitudes (Maraun et al., 2013). Finally, Cianciolo and Norton (2006) found a positive correlation between biotic interactions (competition and predation) and asexuality (when only phylogenetically clustered asexuals were considered), but did not propose any mechanistic explanation for these findings.
Considering the many species and supra-specific taxa in soil that reproduce via parthenogenesis, we assume that identifying factors favoring parthenogenesis in these species may help in understanding why sexual reproduction predominates in the animal kingdom.

| MATERIAL AND ME THODS
We collected data from 25 studies that had investigated oribatid mite communities in different ecosystems and habitats (Table S1). We included only studies where all or most oribatid mites were identified to species level, where total density of oribatid mites was given, and from which we could retrieve data on altitude and latitude. In most cases, we contacted the first author of the study and asked for the original data. In few cases, we estimated density from other studies investigating similar habitats. We further included data from unpublished studies (Ricarda Lehmitz, unpubl. data; Mark Maraun, unpubl. data). The selected studies represent the range of environments in which oribatid mites occur and cover various biomes including temperate and tropical forests, meadows, fields, freshwater, salt marshes, and peat bogs, and microhabitats such as deadwood, mesovoid shallow substratum (a system of empty spaces within the stony debris covered by soil), and littoral algal mats.
To visualize patterns of community dissimilarity and identify factors correlating with species composition across different biomes and habitats, we used detrended correspondence analysis (DCA) as implemented in CANOCO 5 (Microcomputer Power, Ithaca, USA, 2012). Altitude, latitude, the percentage of parthenogenetic oribatid mite individuals, the percentage of parthenogenetic oribatid mite species, oribatid mite species number, and oribatid mite densities were included as passive/supplementary variables.
Further, we used linear regression to relate altitude, latitude, species richness, and total density of oribatid mites to the percentage of parthenogenetic individuals and parthenogenetic species.
In these regression analyses, each community represented a single data point. To improve normality and homogeneity of variance, data were log-transformed. Residuals were tested for heteroscedasticity and normality. As the relationships analyzed, at least in part, might be due to phylogenetic relatedness of species, we controlled for phylogenetic relatedness in the analysis of the data. Unfortunately, however, a solid phylogeny of the very wide range of taxa included in this study does not exist precluding the use of standard phylogenetic comparative methods (Felsenstein, 1985;Martins & Hansen, 1997;Swenson, 2014). Confronted with this limitation, we used the taxonomic system of oribatid mites as surrogate of the phylogeny. We assume this to be justified as molecular phylogenies of oribatid mites in general match the established taxonomic system (Pachl et al., 2017;Schaefer et al., 2010). We assigned each species to genus, family, superfamily, and suborder, and assumed that species within the same genus where phylogenetically closer to each other than species of different genera, and assumed that this is also true for higher taxonomic ranks. We then assigned an arbitrary unitary distance to species within the same genus and accordingly created an ultrametric tree based on taxonomic distance. We converted the tree into a cophenetic distance matrix and calculated mean taxonomic distance between each pair of communities (equivalent to phylogenetic beta diversity). We then subjected the resulting community distance matrix to Principal Coordinate Analysis and extracted as many axes as needed to account for at least 2/3 of the variance in the distance matrix. The axes were used as proxies for phylogenetic correlation between the analyzed communities and were thus used to approximate a phylogenetic correction of the variables analyzed with regression.
The analyses were implemented using the R packages ape, phylobase, picante, phytools, and Geiger (R Development Core Team 2018).

| RE SULTS
Generally, the percentage of both parthenogenetic individuals and parthenogenetic species correlated significantly with the total density of oribatid mites (F 1,25 = 4.97, p < 0.045 and F 1,25 = 4.47, p < 0.044, respectively). On a logarithmic scale, a similar proportion of the variance in the number of individuals of parthenogenetic species (r 2 = 0.22) and the number of parthenogenetic species (r 2 = 0.23) was explained by density ( Figure 1). Density was highest in peat bogs, acidic forests and in freshwater systems, and these were the ecosystems in which parthenogenetic species dominate (Table S1). The correlation between total density and frequency of parthenogenesis also held when corrected for phylogenetic relatedness as approximated by the first six axes of PCoA of the taxonomic distance matrix ( Figure S1 and Table   S2). Of these six axes, only axis 2 significantly correlated with density and percentage of parthenogenetic species and individuals (Table S2).
We thus regressed these three variables against PCoA axis 2 and used the residuals as phylogenetically corrected variables ( Figure S1). As expected, the r 2 -values slightly decreased, but the slope of the regression lines remained statistically significant (p < 0.01).   (Table S1) indicating a hump-shaped relationship.
However, this pattern was not consistent because parthenogenetic species dominate in temperate and boreal lowland regions, whereas at low altitudes in the tropics sexual species predominate.
In the following, ecosystems and habitats are discussed and grouped according to the general pattern of reproductive mode of the oribatid mite communities (Figure 2). Ecosystems and habitats dominated by parthenogenetic species (50%-100% of all species) are discussed first, then the ones with intermediate numbers (25%-50%) and with few parthenogenetic species (<25%).

| Ecosystems and habitats dominated by parthenogenetic species
Both the number of parthenogenetic oribatid mite species as well as the number of individuals of parthenogenetic species dominated in freshwater systems, peat bogs, acidic forests and in glacier foreland ( Total oribatid mite density in all these ecosystems was high, typically > 50,000 ind./m 2 .

| Ecosystems and habitats with intermediate numbers of parthenogenetic species
Systems with similar numbers of parthenogenetic and sexual taxa (and individuals) included moss-grass tundra, temperate baserich forests, meadows, fields, tropical mountain rain forests, salt marshes, and algal mats (Table S1). In moss-grass tundra (

| Ecosystems and habitats with low numbers of parthenogenetic species
Both the number of sexual oribatid mite species as well as the num-  (Nae & Bancila, 2017). Oribatid mite densities in all of the above-mentioned habitats are low.

| D ISCUSS I ON
The proportion of parthenogenetic species and individuals correlated positively with oribatid mite density being high in peat bogs at high and low altitude, in freshwater systems and in acidic forests (Huhta et al., 1986;Maraun & Scheu, 2000 Altitude did not correlate with reproductive mode in oribatid mites. This was due to the fact that low-altitude forests in the tropics harbor few parthenogenetic species whereas acidic temperate forests harbor many (Table S1). Furthermore, the percentage of sexual species (as well as individuals) decreases with increasing altitude up to about 3,000 m in tropical as well as temperate regions (Fischer, Meyer, & Maraun, 2014;Maraun et al., 2013), but in the Andes at 4,000-5000 m it is lower than at 3,000 m (Covarrubias, 2009;Hense, 2016;Maraun et al., 2013). The dominance of parthenogenetic species at very high altitudes in the Andean mountains probably is due to the fact that peatland vegetation dominates at these sites which are colonized by oribatid mite communities similar to those of peat bogs from lowland sites at high latitudes (Lehmitz & Maraun, 2016;Mumladze, Murvanidze, & Behan-Pelletier, 2013;Seniczak, Seniczak, Maraun, Graczyk, & Mistrzak, 2016). The reduced frequency of sexual species at very high altitude conforms to predictions of both the RQT as well as the SRTS since they suggest that harsh environmental conditions foster parthenogenesis by reducing parasite-host interactions and by reducing density-dependent population regulation, respectively. From the perspective of the RQT, the increase in the dominance of sexual species up to 3,000 m argues for increased parasite-host/ predator-prey interactions up to this altitude which is unlikely, in particular considering that above 3,000 m the pattern is reversed. From the perspective of the SRTS, it argues for increased resource limitation and/or more simplified habitat structure at higher altitude which appears to be more plausible. In fact, the density of oribatid mites declines in both temperate and tropical regions up to about 3,000 m indicating increased resource shortage (Fischer et al., 2014;Maraun et al., 2013;Marian, Sandmann, Krashevska, Maraun, & Scheu, 2018).
Latitude also did not correlate with the proportion of parthenogenetic individuals or species. This was due to the fact that there are sites at low latitude with few (high Andes) and many sexual species (lowland tropical sites), and also sites at high latitude with few (peatland) or many (dead wood, forest canopy) sexual species.
Presumably, this is due to the fact that abiotic conditions at low latitudes (e.g., peat bogs in Germany and Poland) and high altitudes (e.g., high Andes) resemble each other, arguing that habitat-specific factors, such as those, for example, in peatlands, rather than latitude drives the reproductive mode of oribatid mites. From the perspective of the RQT and SRTS, the lack of correlation between latitude and reproductive mode in oribatid mites suggests that neither para-  (Caruso, Noto La Diega, & Bernini, 2005), and this presumably also is true for those living on deadwood or in the canopy of trees (Wehner, Heethoff, & Brückner, 2018 Maraun, Augustin, Müller, Bässler, & Scheu, 2014), and these species typically reproduce sexually. As feeding on living food resources may be associated with increased bottom-up population regulation, for example, due to defense mechanisms of the consumed species, the predominance of sexual reproduction is consistent with assumptions of the SRTS, whereas feeding on living resources is unlikely to be related to stronger top-down processes and therefore unlikely to support the RQT.
Most species and individuals of oribatid mites in caves are sexual (Ducarme, Wauthy, André, & Lebrun, 2004;Ivan & Vasiliu, 2010); however, close to the cave entrance or in flooded caves there are more parthenogenetic species (Maslak & Barczyk, 2011 Densities of oribatid mites in caves and in mesovoid shallow substratum are generally low (Ducarme et al., 2004;Ioana Nae, unpubl. data). The same applies to deep soil where densities are low and sexuals dominate (Ducarme et al., 2004). Again, these patterns argue for resource-based processes rather than parasite-host/ predator-prey processes favoring sexual reproduction as low density suggests low resource supply, and the limited access of prey species in caves and deep in soil suggests top-down control to be of limited importance.

| CON CLUS IONS
In sum, (a) parthenogenetic oribatid mites dominate in habitats where densities are high, that is, where resources are plentiful and/ or are easy to access, such as peat bogs and acidic forests. deep soil). All these cases are consistent with the assumption of the SRTS that sexual reproduction is favored if food resources are in limited supply. Notably, the predominance of sexual reproduction in habitats where oribatid mite densities are low is surprising from the perspective of mating/sperm transfer as this is more difficult at low density and therefore should favor parthenogenetic reproduction, but argues that even when rare sexual species are not sperm limited.

ACK N OWLED G M ENTS
We thank Roy A. Norton for comments on an earlier version of the paper.

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
All authors contributed to the conception of the study, the acquisition of data, and the data analysis; they all drafted and revised the article critically and approved the final version to be published.