Land Use Change Increases Wildlife Parasite Diversity in Anamalai Hills, Western Ghats, India

Anthropogenic landscape changes such as land use change and habitat fragmentation are known to alter wildlife diversity. Since host and parasite diversities are strongly connected, landscape changes are also likely to change wildlife parasite diversity with implication for wildlife health. However, research linking anthropogenic landscape change and wildlife parasite diversity is limited, especially comparing effects of land use change and habitat fragmentation, which often cooccur but may affect parasite diversity substantially differently. Here, we assessed how anthropogenic land use change (presence of plantation, livestock foraging and human settlement) and habitat fragmentation may change the gastrointestinal parasite diversity of wild mammalian host species (n = 23) in Anamalai hills, India. We found that presence of plantations, and potentially livestock, significantly increased parasite diversity due possibly to spillover of parasites from livestock to wildlife. However, effect of habitat fragmentation on parasite diversity was not significant. Together, our results showed how human activities may increase wildlife parasite diversity within human-dominated landscape and highlighted the complex pattern of parasite diversity distribution as a result of cooccurrence of multiple anthropogenic landscape changes.

Land use change and habitat fragmentation are two major landscape-level outcomes of human activities that significantly impact biodiversity [1][2][3] . Hence, considerable research on biodiversity change in human-dominated landscape have been conducted, which has resulted in improved understanding of how these two human impacts on landscape can impact biodiversity 1,4,5 . These anthropogenic factors can also modify host-parasite interactions, which, in turn, can lead to either increase or decrease in parasite diversity [6][7][8] . Understanding how these factors may influence parasite diversity is ecologically important for multiple reasons. For instance, parasites regulate host population dynamics 9 , alter species communities 10 and constitute a significant proportion of total biomass of any ecosystem 11 , which is not surprising considering parasites comprise at least 40% of all animal species on earth 12 . Despite their ecological importance, our knowledge on parasite diversity is limited 13,14 , particularly in the context of increasing human impact on environment, underlining a significant research gap 15,16 . The gap is specifically wide for wildlife hosts and urgent research is required in the face of recent increased emergence of novel pathogens of wildlife origin 7,17,18 . It is, thus, crucial to answer how anthropogenic land use change and habitat fragmentation may impact parasite diversity in the wild.
Land use change can affect parasites both directly and indirectly. By altering environment (for example, through pollution), land use change may render transmission of environmentally-transmitted parasites difficult. This is particularly true for parasites that has life stages outside host body. However, land use change can indirectly impact parasite diversity by altering host diversity as it is one of the strongest predictors of parasite diversity [19][20][21][22] . By decreasing host diversity and abundance, land use change can deplete richness of parasites particularly those that require multiple obligatory hosts 23 . This is evident when many host species that are threatened in their natural habitat appear to harbour fewer parasites 24 . On the other hand, land use change can also increase parasite diversity in multiple ways. Land use change can increase parasite diversity by increasing host diversity. For instance, land use change such as agricultural field or land-fill can act as resource traps and amplify host diversity artificially 25 . Land use change can also increase parasite diversity by introducing non-native parasites such as parasites of domestic and feral animals and even from humans 26  It is also important to distinguish between different types of land use change and their effects on parasites 27 . One type of land use change that has not been studied well is the effect of plantation on wildlife parasites 27 . Plantations are usually monocultures of exotic or native plant species grown as timber or fuel wood or as cash crops and have a large and increasing footprint in wildlife habitats worldwide 28 . They can sometime act as refuge to wildlife but usually with a biotic homogenising effect 29,30 . Consequently, plantation may also increase but homogenise parasite community. Plantations are often accompanied by settlement of labourers and livestock foraging 31,32 . These changes within a wildlife habitat can both increase or decrease parasite diversity. Parasite diversity may decrease if wildlife hosts avoid human areas to lessen confrontation with humans and resource competition with livestock. On the other hand, generalist species may actually thrive in human settlements by utilizing novel resources 33,34 . Herbivore species may also prefer to stay closer to human settlements and livestock ("spatial refugia") that may displace predators [35][36][37][38] . Moreover, many wildlife, over time, may actually get habituated to humans and livestock and aggregate near human-dominated landscape 39,40 . These aggregations may eventually increase parasite diversity by increasing contact between native host species. Such situations may also increasingly expose wildlife to humans and human-associated animals, such as livestock and commensals, increasing chance of spillover of non-native parasites to wildlife.
Habitat fragmentation may lead to higher parasite diversity because heavily fragmented habitats may disrupt wildlife dispersal and increase host diversity in smaller fragments. Such increase in host diversity in a smaller patch may alter host characteristics such as home range, abundance and intra and interspecific contacts thus increasing overlap among host species making host individuals exposed to higher parasite infections 41,42 . These effects are likely to be greatest in the smallest and most isolated of the fragments 3,43 . By disrupting host dispersal, fragmentations can also adversely affect parasite diversity. This could be especially true for parasites who require multiple host species to complete its life cycle, such as those that are transmitted trophically 44 . So far, many studies looked into this effect but the results have been mixed 6,41,[45][46][47] .
The Anamalai (Elephant hills in Tamil) hills of southern India is a highly biodiverse rainforest habitat of Western Ghats, which holds about 30% of India's plant and vertebrate species diversity in less than 6% of the country's area 48 . It is also one of the most altered natural habitats in India and typifies different levels of land use change and habitat fragmentation rampant in Indian wildlife habitats. Large section of the habitat is highly modified due to land use change, bordered by large, relatively undisturbed tropical rainforests. The landscape is a matrix of over 40 rainforest fragments (1-2,500 ha in size),often surrounded by plantations (coffee, tea and cardamom), roads, hydroelectric dams and settlements 49 . Highly-modified fragments contain within them human settlements and have higher livestock pressures than other remote, less disturbed fragments. In spite of such high levels of land use change and habitat fragmentation, the Anamalai hills still harbour a large number of wildlife whose ranges often unavoidably overlap with humans and livestock [50][51][52][53] . In fact, large number of wildlife species are regularly observed within human-dominated habitats and this concurrence with humans often precipitates into wildlife-human conflicts 49,50,[54][55][56] . It is possible that many of the wildlife are important reservoirs of multiple environmentally-transmitted parasites. In fact, recent studies have recorded important parasite groups within certain host species populations that may cause Ascariasis, Trichuriasis and Strongylodiasis in humans 45,46,57,58 .
To assess the effect of land use change (plantation, livestock foraging and human settlements) and habitat fragmentation on parasite diversity, we studied gastrointestinal parasites of wild mammalian hosts across rainforest fragments in Anamalai hills. We employed statistical models to test these effects. We predicted a positive impact of land use change on parasite diversity due to increased host diversity and an increased exposure of wildlife to humans and livestock. For habitat fragmentation too, we predicted an increase in parasite diversity with decrease in habitat size and increase in habitat isolation. Our alternative predictions were that land use change and habitat fragmentation could actually deplete parasite diversity by decreasing host diversity in disturbed fragments. Finally, it is also possible that land use change and habitat fragmentation may not significantly impact parasite diversity either by not impacting host community or by not spillover from non-native hosts such as livestock and humans.  59 . As a result, the Anamalais today consists of both a relatively undisturbed, large (958.59 km 2 ; ~100,000 ha) tropical rainforest within the protected Anamalai Tiger Reserve (ATR; 10°12′-10°35′N and 76°49′-77°24′E) and about 1,000 ha highly degraded Valparai Plateau (Fig. 1). The plateau consists of many tea estates and other plantations, which are surrounded by four protected areas-ATR in Tamil Nadu state and three others in Kerala state. The major vegetation types include scrub forests in the rain-shadow areas in the eastern foothills, dry and moist deciduous forests (<800 m), mid-elevation tropical wet evergreen forest (600-1,500 m) and high-altitude shola-grassland ecosystems (>1,500 m) 60 . Although a large part of the tropical wet evergreen forests occurs within ATR, many of the smaller (<200 ha) fragments are found in private estates in the Valparai plateau. These small fragments are highly degraded and disturbed due to fuel-wood collection and livestock grazing. Valparai town also is a part of the plateau and around 200,000 people live across the town and plantations 60 . Due to the ongoing habitat fragmentation, the whole landscape is a matrix of over 40 rainforest fragments, ranging 1 ha-2,500 ha in size and often surrounded by plantations (coffee, tea and cardamom), roads, hydroelectric dams and settlements 49 . Based on size range (2-2,500 ha), level of perceived human disturbance and access, we selected 19 forest fragments in total for sampling -16 mid-elevation tropical rainforest and three low-elevation dry and moist deciduous forest fragments (Fig. 1).

Materials and
Host sampling. Between Oct 2013 and Oct 2015, faecal samples were collected from populations of mammalian wildlife. We collected fresh faecal samples non-invasively during the day on transects (400 m-3 km in length). For large and medium herbivores and primates, we followed individuals and collected fresh faeces when animals defaecated. For elusive species such as carnivores, we identified home range based on secondary information and faecal samples were identified based on morphology and also using nearby secondary signs such as pug-marks or hoof-prints. To avoid sampling the same individual repeatedly, only one sample of a host species was collected from each spot and the sample source was either marked or removed whenever possible. However, if removal was not possible (for example, due to large quantity), then we marked. To avoid contamination from soil, samples were collected from the inside of the bolus or only top pellet was collected from a heap. We immediately fixed each sample in 10% formaldehyde solution (50 ml), labelled the containers with the information of origin (fragment name, date, time and host species) and stored them at room temperature until parasitological screening. Differences in sampling effort can confound the comparison of diversity among replicates. We accounted for differences in number of host species encountered by calculating richness estimates with the assumption that each faecal sample represents single individual. We used bootstrap, which is a resampling method for estimating the whole sampling distribution of richness by sampling with replacement from the original sample, that can offer greater precision than jackknife estimates, especially when sample sizes are small 61 . parasite sampling. Employing both the flotation and sedimentation techniques (NaNO 3 solution), we screened the faecal samples for the presence of helminth eggs, larvae and protozoan cysts 62 . For each concentration technique, we examined two slides per sample. Slides were examined under a light microscope (400X). Eggs and cysts were first examined at 10× magnification and then their size was measured with a micrometre eyepiece (0.1 μm) at 40× magnification. To facilitate identification of parasite eggs, we often added a drop of Lugol's iodine solution to the slides to highlight detailed structures. In addition, photographs of each parasite species have been archived and are available for examination on request. We identified parasites to the lowest possible taxonomic level using published keys 63,64 . Differences in sampling effort can confound the comparison of diversity among replicates. We accounted for differences in number of parasite taxon encountered by calculating richness estimates with the assumption that each faecal sample represents single host individual. We used bootstrap, which is a resampling method for estimating the whole sampling distribution of richness by sampling with replacement from the original sample. Bootstrap can offer greater precision over jackknife estimator, especially when sample sizes are small 61 . This method is particularly recommended for parasite richness estimation 65 . www.nature.com/scientificreports www.nature.com/scientificreports/ Land use data. In Anamalai hills, land use change manifests in largely three forms-presence of human settlements, plantations and livestock foraging. There are only few large (>1000 ha) fragments that are legally protected and thus undisturbed. Many of the studied fragments share more than one type of land use change. For instance, some fragments with human settlements may also have livestock present. For the current study, we identified 18 fragments with land use change, out of which 15 (83.3%) had plantation, in contrast to three (16.7%). Eleven (61.1%) of the fragments have significant livestock foraging pressure, in contrast to seven (38.9%) fragments without livestock. Finally, ten (55.6%) of the fragments had human settlements within them, in contrast to eight (44.4%) without settlements.
Habitat fragmentation data. To measure effect of habitat fragmentation, we used fragment size and isolation distance between fragments. According to the equilibrium theory of island biogeography, organism dispersal probability declines as distance between islands increases, reducing rates of immigration and, in turn, reducing diversity [66][67][68] . Assuming each forest fragment as an island, their isolation was summarized with an isolation index which was calculated as the sum of the square root of the distances to the nearest equivalent (no smaller than 80% of size) or larger fragment (Dahl, 2004). Data on fragment size, distance between fragments and presence of human settlements, plantations and livestock were collected from earlier studies from Anamalai hills 45,60 . Data analyses. To assess the effects of land use change and habitat fragmentations on bootstrap estimate of parasite taxon richness, we created two different linear mixed effects models 69 . Each model included random effects of host species and fragments to account for multiple observations within each fragment (across host species) and across fragments. In the land use model, the predictor variables (fixed effects) were presence of plantation, human settlement and livestock. The predictor variables for the habitat fragmentation model were fragment size and fragment isolation index. In both the models, we incorporated both bootstrap estimates of host species richness and host body mass as co-predictors as these were known to effect parasite richness. We retrieved host body mass data from online ecological database 70 . Given the potential role of host density as a covariate to parasite richness (discussed in Introduction), we also planned collection of host density data. This exercise, however, became logistically and financially impractical due to the large number of host species (>20) studied, 19 rainforest patches sampled (totaling >12,000 ha in size), often lack of visibility within the rainforest patches and limited time and resources available to us. Consequently, we dropped this covariate from our study and decided instead to investigate the specific hypotheses born out of the current study (discussed in Discussion) in future, based on a logistically manageable subset of study sites and hosts.
After fitting these models to the data, we also compared and selected the best fit model using lowest AIC value 71 . At the end, diagnostics were run to check distribution of the residuals for each model. This analysis was conducted in the lme4 package 72 . We also assessed the effects of land use change and habitat fragmentation on bootstrap estimates of host species richness using two linear models. In the land use model, the predictor variables were presence of plantation, human settlement and livestock. The predictor variables for the habitat fragmentation model were fragment size and fragment isolation index. We followed the same strategy as described above for model fitting, fitting diagnostics and model selection. Finally, we tested whether land use change homogenized the composition of the parasite community. We used a multivariate nonparametric Analysis of Variance (per-mAnoVa; 1,000 permutations) based on the Jaccard dissimilarity index for a matrix of parasite presence/absence. We calculated the variance of homogeneity of parasite communities within each fragment based on disturbed vs. undisturbed divisions using the betadisper function of the vegan package in R 73 .

Results
Sample diversity. From  In total, seven protozoa (18.42%) and 32 helminth (81.58%) species were recorded, including five trematodes, five cestodes and 20 nematodes. At least seven different parasites, belonging to different parasite groups, were recorded in ≥20 different host species-protozoa Coccidia sp. Host and parasite diversity and disturbance. For parasite diversity analysis, the human disturbance model was the best fit (Table 2). Parasite diversity was significantly driven by presence of plantation (estimate = 4.779, CI Profile = 0.326-9.232, t = 2.103, p < 0.05). Presence of livestock had a substantial but not significant positive effect (estimate = 3.209, CI Profile = −0.052-6.366, t = 1.992, p > 0.05). Effects of settlement, host richness and host body mass on parasite richness were not significant (Fig. 2). For host diversity analysis the human disturbance model was again the best fit ( Table 3). Presence of plantation was the only predictor that had a significant positive effect on host diversity (estimate = 10.798, CI Profile = 2.302-19.294, t = 2.726, p < 0.05 (Fig. 3)-almost half of all host species occur in plantations. Although presence of livestock did not have a significant effect, its wide confidence interval was mostly on the positive side suggesting potential positive impact-limited by sample size-on host richness (estimate = 5.602, CI Profile = −0.639-11.843, t = 1.925, p > 0.05). Similarly, presence of human settlement did not significantly affect host richness, however, the substantial effect was mostly on the negative side, suggesting potential negative effect on host diversity (estimate = −4.112, CI Profile = −10.717-2.492, t = −1.335, p > 0.05).  www.nature.com/scientificreports www.nature.com/scientificreports/ We recorded 12 parasites (ten helminths and two protozoa) that occurred only in plantations. Six of the ten helminths were nematodes (60%), while rest were trematodes (30%) and one cestode (10%). Fragments without plantations did not harbour any parasite taxon exclusively, which means parasites in those undisturbed fragments also occured in plantations. Fragments with livestock harboured three parasite taxa (two nematodes and one cestode) exclusively relative to their undisturbed counterpart. However, only one parasite taxon (Taenia sp.) exclusively occurred in livestock disturbed fragments, while other two nematodes also occurred in the plantations. Its counterpart undisturbed fragments only harboured one taxon exclusively (Paragonimus sp.), which however also occurred in plantations. Finally, settlements harboured three nematode taxa exclusively in comparison to their undisturbed counterpart. Only one of these taxa (Uncinaria sp.) were exclusive to settlements across all fragments. Undisturbed counterpart of settlements harboured only one parasite taxon (Sarcocystis sp.) exclusively. parasite and host homogeneity. Parasite communities within disturbed forest fragments were not significantly more homogeneous than the undisturbed ones due to presence of either plantations (F = 2.58, p > 0.05), livestock (F = 0.04, p > 0.05) or settlements (F = 3.55, p > 0.05). Host communities within plantations (TukeyHSD; p < 0.05; Fig. 4a) and human settlements (TukeyHSD; p < 0.05; Fig. 4b) were, however, significantly more homogeneous than undisturbed fragments. Finally, we did not find any of the disturbance variables to significantly alter the parasite community composition between undisturbed and disturbed fragments.

Discussion
Our results reveal that rainforest fragments with plantations (and potentially with livestock) in Anamalai hills harbour significantly higher parasite diversity than undisturbed fragments. Interestingly, some of the disturbed fragments (at least, fragments with plantations) also has significantly more host diversity than the undisturbed fragments, however host diversity was not found to significantly affect parasite diversity.   www.nature.com/scientificreports www.nature.com/scientificreports/ In Anamalai hills, plantations (coffee, tea and cardamom) had more mammalian wildlife species richness than the undisturbed fragments. This was not particularly a surprising result because studies have reported similar high richness in vertebrate species from plantation within wildlife habitats 74,75 . In fact, earlier studies from Western Ghats also found high vertebrate richness within or around plantations with large variations depending on plantation types, from open tea to more shaded coffee and cardamom plantations 30,76,77 . The reason for such increased host diversity is thought to be an increase in habitat heterogeneity within plantations. Increased habitat heterogeneity is thought to generate greater diversity of niches consequently facilitating cooccurrence of many species 78,79 . However, such increase in species richness is often accompanied by more generalist and wide-ranging species being more abundant within the plantations and a loss of community heterogeneity relative to undisturbed habitats [80][81][82] . We found similar loss of heterogeneity for host species in disturbed habitats with plantations and settlements (Fig. 4).
Effect of livestock presence on host species richness was positive but not statistically significant at α = 0.05. The effect, however, was significant at α = 0.10, which suggested potential, but weak effect that was reflected by the almost equal number of wildlife species recorded from these two groups of fragments (n Livestock = 20 and n Undisturbed = 22). Interaction between livestock and wildlife is complicated. For instance, while a number of studies found evidence of competitive exclusion between livestock and large herbivore 83 , many other recorded resource sharing between these two groups 84,85 . Yet still, many other studies did not find any relationship between the two 86 . The outcome of the interaction may depend on the ecological similarity between the two groups (Niche overlap), availability of natural resources that may vary between habitats (between low to high productivity) and also degree of behavioural habituation by the wildlife. The wildlife community that we studied was an ecologically broad one consisting of wildlife with very different ecology. Therefore, while some of the species-such as spotted deer Axis axis and sambar deer Rusa unicolor, who were found only in the undisturbed fragments-may face resource competition from livestock grazing, others (for example, small carnivores and primates) may not face any competition. In addition, many large herbivores, such as gaur B. gaurus and Indian elephants Elephas maximus indicus, who despite resource competition, may still use the disturbed fragments as corridors contributing to host richness. These processes together may explain almost similar host species richness between fragments with and without livestock grazing.
We did not find any significant effect of human settlement on host diversity but the trend is negative (Fig. 3). While human settlement may attract and facilitate generalist and weedy species with high tolerance for disturbance (for example, rodents, which were not sampled in the present study), many elusive species such as carnivores may be adversely affected and may prefer to avoid fragments with settlements 87 . Still, we recorded overall a large host species richness (host richness Settlement = 19, host richness Undisturbed = 23) from around the settlement in Anamalai hills. This could be explained by the facts that many of these settlements may attract wildlife with unintentionally supplemented resources such as planted fruit trees 60 . Additionally, the high level of fragmentation of the landscape meant large herbivores and carnivores may not have much choice but to disperse through human settlements 54,56 . We did not find evidence of habitat fragmentation (fragment size and isolation) influencing host species richness in Anamalai hills (Fig. 3). This is in line with findings from across studies that reported effects of fragmentation on species communities are often weak 88 . Effects of habitat fragmentation on species diversity is also highly context-specific and varies considerably between animal groups, ecosystems and kinds of human activities prevalent in the landscape [88][89][90][91][92] . In Anamalai hills, habitat fragmentation is widespread, which likely www.nature.com/scientificreports www.nature.com/scientificreports/ disrupts animal movement to some extent but, in the absence of hunting, perhaps not substantially. For instance, studies recorded use of certain plantation as corridors to connect with isolated undisturbed habits 51,52,77 . However, the adverse outcome of these movement through human-dominated habitats is the increase in wildlife-human conflict 54,56 .
Among the different types of land use change in Anamalai hills, plantations had the strongest positive effect on parasite diversity (Fig. 2). Increase in number of parasite taxa in modified fragments ranged between one to ten, with eight parasite taxa that were recorded exclusively in these fragments (Table 4). However, this increased richness in disturbed fragments were likely not driven by host richness as host richness had a small and statistically not significant effect on parasite richness in the disturbance model (Fig. 2). This is in contrast to the predominant patterns across most studies on parasite diversity that found host richness to be the strongest predictor of parasite richness [19][20][21][22] . However, there could be potential deviations from this rule, particularly due to human impacts 21,93,94 . For instance, many human parasites may spillover to wildlife (anthropozoonoses) as humans regularly come in contact with wildlife [95][96][97][98][99][100] . In addition to livestock, humans may also introduce many non-wildlife species such as feral dogs and cats into wildlife habitats and these species may share parasites with wildlife 26 . In such cases, parasite richness in wildlife would be more than in the undisturbed fragments. Indeed, all but one (Schistosoma sp.) of the parasites that we found exclusively in plantations also occurred in cattle (Table 4). Surprisingly, wildlife parasite taxa that were present in the livestock foraging fragments did not occur in cattle samples from the same fragments. This was also the case for the wildlife parasites that only occurred in settlement but not in undisturbed fragments. We did not find any significant effect of host body mass on parasite diversity (Fig. 2). This is in contrast to many studies that found a significant relationship between these two variables 101,102 . On the other hand, many other empirical studies did not find any relationship between body mass and parasite richness when accounting for host phylogenetic relationships 103,104 . Such contradictory results may suggest that relationship between host body mass and parasite diversity is a factor of body mass and life history traits, which vary between ecologically different groups of hosts 22 . Based on these observations, the broad ecological diversity among host species in the present study might have confounded this relationship.
Our results did not find any significant effect of habitat fragmentation on parasite richness (Fig. 2). This was expected as we did not find any effect of fragmentation on host richness either. This lack of relationship between fragmentation and parasite diversity could also be an outcome of large home ranges and low habitat specialisation of most of the host species in our study. Many of the species that we sampled were large herbivores or carnivores (e.g., Elephas maximus, Bos gaurus, Panthera tigris, Panthera pardus) with larg home range and they disperse across fragments. The level of fragment isolation (Median distance = 30.2 km) may not be a deterrent to  www.nature.com/scientificreports www.nature.com/scientificreports/ their dispersal. Similarly, many host species in the study community, such as bonnet macaque Macaca radiata, wild pig Sus scrofa, small Indian civet Viverricula indica, are habitat generalists. These species, according to the distribution-abundance relationship hypothesis 6 , may be adapted to smaller, fragmented habitats. These hosts may then spread parasites across habitats, independent of the level of habitat fragmentation.

conclusion
In the present study, we demonstrated that human-driven land use changes increased parasite diversity in a rainforest habitat and presence of potential spillover of parasites from livestock to wildlife. We also showed that the observed pattern of parasite diversity was not driven by habitat fragmentation.
One of the limitations of this study was that it could not test the effect of land use change and habitat fragmentation on the relationship between host density and parasite diversity. Host density is an important predictor of parasite diversity and in nature, host density is linked to host ecology (e.g., home range). However, land use change can unpredictably change host density, which may have a complex outcome for parasite diversity. It will thus be worthwhile in future to explore this question in the present system. Additionally, with the present evidence of potential anthropozonosis, it will be important in future to compare parasite from the present study to samples from humans, livestock and commensal animals in the fragments. Finally, as far as land use change and habitat fragmentation of wildlife habitats in India are concerned, the present study represents a case study with particular relevance for tropical rainforest habitats. However, there exists a large diversity in habitats and levels of disturbance in India. Given the increased threat to wildlife health from anthropogenic environmental change, it will thus be crucial for wildlife conservation biologists to study the patterns of parasite diversity in other types of habitats, especially those with already threatened wildlife.

Data Availability
The datasets generated and analysed during the current study are available from the corresponding author on request.