Oikos 000: 001–005, 2014 1 doi: 10.1111/oik.01396 2 © 2014 Th e Authors. Oikos © 2014 Nordic Society Oikos 3 Subject Editor: Dries Bonte. Accepted 4 February 2014 4

5

This is the accepted version of the authors’ manuscript. The final, editor version of this article is 6 available to suscribers at: www.onlinelibrary.wiley.com 7

8

Parasite abundance contributes to condition-dependent dispersal in a wild 9

population of large herbivore 10

11

Lucie Debeffe 1,2*, Nicolas Morellet 1, Hélène Verheyden-Tixier1 , Hervé Hoste 3,4, Jean-12

Michel Gaillard 2, Bruno Cargnelutti 1, Denis Picot 1, Julie Sevila 1,3,4,5 and A.J.Mark 13

Hewison1 14

15

1 INRA, UR35 CEFS, B.P-52627, 31326 Castanet-Tolosan, France 16

2 Université Claude Bernard Lyon1, LBBE UMR5558, 69622 Villeurbanne, France 17

3 INRA, UMR1225 IHAP, F-31076 Toulouse, France 18

4 Université de Toulouse, ENVT UMR1225, F-31076 Toulouse, France 19

5 ANSES, Nancy laboratory for rabies and wildlife, 54220 Malzéville, France 20

21

*corresponding author: lucie.debeffe@gmail.com 22

Phone number: +335 61 28 51 25, Fax number: +335 61 28 55 00 23

Abstract 24

25

Parasite abundance has been shown to have major consequences for host fitness components 26

such as survival and reproduction. However, although natal dispersal is a key life history trait, 27

whether an individual’s decision to disperse or not is influenced by the abundance of parasites 28

it carries remains mostly unknown. Current and opposing hypotheses suggest that infected 29

individuals should either be philopatric to avoid the energetic costs of dispersal (condition 30

dependence) or disperse to escape from heavily parasitised habitats. From intensive 31

monitoring of a roe deer population inhabiting a multi-use and spatially heterogeneous 32

agricultural landscape, we evaluated the link between an individual’s parasite abundance and 33

its propensity to disperse, while accounting for confounding effects of body mass. Dispersal 34

propensity generally decreased with both increasing nematode abundance and with decreasing 35

body mass. Within the dispersing segment of the population, individuals with high nematode 36

abundance left their natal home range later in the season than less parasitised deer. These 37

results clearly show that parasite abundance is an important component of condition-38

dependent dispersal in large herbivores. However, unexpectedly, three individuals that were 39

both heavily parasitised and of low body mass dispersed. We suggest that this “leave it” 40

response to high parasite levels in the natal habitat could represent a last ditch attempt to 41

improve reproductive prospects, constituting a form of emergency life history strategy. 42

43

Key-words: Natal dispersal, Nematode, Roe deer, Ungulate 44

Subject Areas: Animal behaviour 45

Short title: Dispersal and endoparasites in roe deer 46

47

Parasites impact their host in a variety of ways, often inducing immune responses and/or 48

behavioural modifications. For instance, locomotion and activity of infested individuals are 49

reduced in juvenile toads, Bufo bufo (Cameron et al. 1993). In addition, parasites also extract 50

resources from their host and, consequently, may negatively impact their fitness (Poulin 51

2007). High parasite abundances lead to reduced fecundity in mountain hares Lepus timidus 52

(Newey and Thirgood 2004) and reindeer Rangifer tarandus (Stien et al. 2002), lowered 53

survival in cliff swallows Hirundo pyrrhonota (Brown and Brown 1986) and lowered growth 54

rate and reproductive success (ie. presence of young after emergence) in yellow bellied 55

marmots Marmota flaviventris (Van Vuren 1996). However, while natal dispersal is a key life 56

history trait for individual fitness (Clobert et al. 2001) which could potentially be affected by 57

parasitism (Heeb et al. 1999, Boulinier et al. 2001), how parasite abundance influences the 58

propensity of individuals to disperse remains virtually unknown. The few published studies 59

have focused on ectoparasites (Brown and Brown 1992, Sorci et al. 1994, Van Vuren 1996, 60

Heeb et al. 1999), while the only study to date on endoparasites failed to find a link between 61

parasite abundance and natal dispersal in Drosophila melanogaster (Jaenike et al. 1995). 62

Parasites have been hypothesised to influence natal dispersal in opposing ways. First, 63

highly parasitised individuals might be more likely to be philopatric or to disperse over 64

shorter distances than healthy individuals (Dufty and Belthoff 2001) so as to minimise 65

dispersal costs (Bonte et al. 2012) (i.e. condition-dependent dispersal). That is, if highly 66

parasitised individuals are unable to cover the energetic costs associated with natal dispersal 67

(Dufty and Belthoff 2001), they should remain philopatric. Furthemore, it may be less costly 68

from a life history perspective for heavily parasitised individuals to remain philopatric if there 69

is local adaptation to a given parasite community (Kaltz and Shykoff 1998, Heeb et al. 1999, 70

Moller and Szep 2011) and/or if the mosaic of parasite strains potentially encountered during 71

dispersal imposes a supplementary cost (Möller and Szep 2011). These kinds of effects of 72

parasitism on natal dispersal behaviour have been reported in a great tit Parus major 73

population, where recruits from infested broods had a lower probability of dispersing, and 74

when they did disperse, they did so over shorter distances (Heeb et al. 1999). Alternatively, if 75

natal dispersal is a way for individuals to escape from highly parasitised environments 76

(Boulinier et al. 2001), dispersal propensity and distance should increase with parasite 77

abundance (Brown and Brown 1992, Van Vuren 1996, Moller et al. 2004). For instance, 78

dispersing yellow bellied marmots had higher ectoparasite loads than their philopatric 79

counterparts (Van Vuren 1996), while cliff swallows with low parasite levels had a higher 80

probability of returning to their natal site and being recruited locally (Brown and Brown 81

1992). 82

We aimed to assess the link between endoparasite abundance, namely gastrointestinal 83

nematodes, and natal dispersal behaviour in a roe deer, Capreolus capreolus, population 84

inhabiting a multi-use, spatially heterogeneous, landscape. Nematode abundance is known to 85

impact body condition in large herbivores (Stien et al. 2002, Body et al. 2011). Hence, 86

because we have previously shown that dispersal behaviour in roe deer is condition-dependent 87

such that both dispersal propensity and distance increase with increasing mass (Debeffe et al. 88

2012), we predicted that both dispersal propensity and distance should decrease with 89

increasing nematode abundance. Indeed, condition dependence is a multi-factorial construct 90

(Ims and Hjermann 2001), potentially involving several components including both body 91

mass and parasitism. Because body mass and parasite abundance are negatively related (Body 92

et al. 2011), in order to disentangle the effects of these two factors on dispersal behaviour, we 93

first controlled for the body mass-parasite abundance relationship. We were thus subsequently 94

able to distinguish between the effects of the body mass component and the parasitism 95

component of condition-dependent dispersal. In addition, because parasitism directly affects 96

individual body condition by increasing energetic expenditure, parasite abundance may 97

influence the timing of dispersal (Dufty and Belthoff 2001). Hence, as individuals require a 98

certain amount of energy in order to disperse, we expected highly parasitised individuals, 99

presumably in lower condition, to delay dispersal during a given season until they gained 100

sufficient condition. 101

102

Methods 103

104

Study site and sampling 105

The data were collected in a heterogeneous agricultural landscape (N 43°27', E 0°85') 106

covering around 10 000 ha in South-West France. We identified three sectors of contrasting 107

landscape structure based on the proportion of woodland: a closed sector of two forest blocks, 108

an open landscape with a few fragmented woodlots, and an intermediate sector with inter-109

connected woodland fragments (Morellet et al. 2011). From 2007 to 2013, we captured 67 110

juvenile roe deer (less than 1 year old) during winter, recorded their body mass and sampled 111

faeces directly from the rectum, before fitting them with a GPS collar and releasing them on 112

site. 113

114

Assessing natal dispersal 115

Natal dispersal corresponds to permanent emigration from the natal range to a distinct 116

adult range such that pre-dispersal locations do not overlap post-dispersal locations (Kenward 117

et al. 2002). As described in Debeffe et al. (2012), to discriminate dispersers from philopatric 118

animals, we examined the degree of range overlap and used the range stability index of 119

Roshier and Reid (2003). Dispersal behaviour was monitored for a total of 67 individuals, of 120

which, 31 dispersed while 36 remained philopatric. For the 31 individuals that dispersed, the 121

dispersal date corresponded to the last date an individual was located inside its natal range, 122

while dispersal distance was calculated as the Euclidean distance between the pre-dispersal 123

range and the most distant location of the movement trajectory. 124

125

Estimating endoparasite abundance 126

All faeces were collected during winter, between early January and mid February, so that 127

confounding seasonal variations in parasite abundance were minimised. As a result, sampling 128

for parasites occurred two to three months before the dispersal period, so that parasite 129

exposure was a priori equivalent for individuals that subsequently were either philopatric or 130

dispersed. From one gram of fresh faeces, we identified the eggs of gastrointestinal parasites: 131

Strongylida, Trichocephalida, Ascaridida, three different orders of worms, were pooled 132

together as “nematodes”. We estimated parasite abundance, defined as the number of 133

individuals of a particular parasite in a single host regardless of whether or not the host is 134

infected (Bush et al. 1997), as the abundance of these nematodes. Nematode abundance was 135

measured using faecal egg counts (Rózsa et al. 2000) which were log-transformed as ln 136

(x+25) to achieve normality. When parasite abundance was estimated to be less than 100 eggs 137

per gram of faeces (corresponding to less that 2 eggs on the slide), that individual was 138

considered to be non-infested. Faecal egg counts provide a reliable estimate of gastrointestinal 139

parasite abundances (Raynaud 1970), and this method has been validated on another roe deer 140

population for Trichuris using 244 roe deer shot during the hunting season (linear regression 141

between faecal egg counts and number of adult worms found at necroscopy: n = 244, r² = 142

0.811, p < 0.001) (see Appendix in Body et al. 2011). Since parasite abundance is negatively 143

correlated with body mass in roe deer (Body et al. 2011), for subsequent analyses, we 144

generated a standardised measure of nematode abundance as the residuals from the linear 145

regression between nematode abundance and body mass (McCullagh and Nelder 1989), 146

allowing us to discriminate between the effects of body mass per se and the effects of parasite 147

abundance on the thee dispersal metrics (propensity, date and distance). 148

149

Data analysis 150

We assessed the influence of parasite abundance on the three dispersal metrics using 151

generalized linear mixed models (GLMM) in R software (version 3.0.0) (R Development 152

Core Team. 2010). We included body mass as a fixed effect in our models, while landscape 153

sector was included as a random factor to account for observed variation in dispersal among 154

sectors (Debeffe et al. 2012). Both body mass and the standardised abundance of nematodes 155

were included in the models to evaluate the contribution of these two components of 156

condition-dependent dispersal. We also included the body mass x nematode abundance 157

interaction in the full model to test for consistency in the effects of parasite abundance on 158

dispersal behaviour across individuals of differing body mass. To simplify interpretation, 159

individuals were assigned to one of three body mass categories: heavy individuals, from 17.9 160

to 20 kg (20% of the data), average individuals, from 15 to 17.9 kg (60% of the data) and light 161

individuals, from 11.5 to 15 kg (20% of the data). Body mass of male and female juveniles 162

did not differ (mean ±SD: 16.74 ± 1.85 kg and 16.12 ± 1.68 kg for males and females 163

respectively, n = 67, r² = 0.03, F1,65 = 2.04, p = 0.16), hence we did not correct individual 164

body mass for any sex-specific difference. We fitted five candidate models for each metric of 165

dispersal behaviour: dispersal propensity, distance and departure date. We selected the models 166

with the lowest AICc value (Akaike Information Criterion corrected for small sample size, 167

Burnham and Anderson 1998)). Analyses using body mass as a continuous variable did not 168

change the results or our interpretations and are provided in the electronic supplementary 169

materials. 170

171

Results 172

173

Dispersal propensity 174

Dispersal propensity was best described by the two-way interaction between nematode 175

abundance and individual body mass (Table 1 and Table 2; Supplementary material Appendix 176

1, Table A1 and Table A2). Dispersal propensity decreased as standardised nematode 177

abundance increased for the majority of individuals, including both heavy deer and deer of 178

average mass, while dispersal propensity increased with body mass over most of the range of 179

observed parasite abundance (ANOVA between best model and constant model: Chisq 5 = 180

21.70; p < 0.001) (Fig 1; Supplementary material Appendix 1, Fig A1). Among the lightest 181

individuals, dispersal propensity was generally lowest, but this was not the case for three 182

highly infested and light individuals that all dispersed. 183

184

Dispersal distance 185

Dispersal distance was best described by the effect of body mass such that dispersal distance 186

increased with increasing body mass (Table 1; ANOVA between best model and constant 187

model: Chisq 2 = 8.36; p = 0.015), while nematode abundance did not influence dispersal 188

distance. 189

190

Dispersal date 191

Dispersal date was, by definition, available only for dispersing individuals (n = 31). Dispersal 192

date was best described by the effect of nematode abundance (Table 1; ANOVA between best 193

model and constant model: Chisq 1 = 5.62; p = 0.018). While dispersal timing appeared to be 194

independent of body mass, deer dispersed increasingly later with increasing nematode 195

abundance (n = 31, slope ±SE: 5.52 ±2.20; Fig 2). 196

197

198

Discussion 199

200

This is, to our knowledge, the first report showing that endoparasites can influence the natal 201

dispersal behaviour of their host, with potentially important consequences for the dynamics 202

and functioning of the host population (Möller et al. 2004). Specifically, we found that 203

nematode abundance was linked to the natal dispersal behaviour of roe deer, both in terms of 204

dispersal propensity and departure date. Note that because parasite abundance was estimated 205

several months prior to the dispersal period, parasite exposure at sampling was a priori 206

equivalent for philopatric and dispersing individuals. Overall, our results support the view that 207

parasites impose additional energetic expenditure on their host (Poulin 2007), preventing 208

highly parasitised individuals from dispersing as often or as far as healthy individuals because 209

dispersal is costly (Bonte et al. 2012). However, the propensity of individuals that were in 210

particularly poor condition, with both low body mass and high parasite abundance, to disperse 211

was much higher than expected based on this condition-dependent dispersal framework (Fig 212

1; Supplementary material Appendix 1, Fig A1). Dispersal can thus have different proximate 213

causes (Clobert et al. 2001, Bowler and Benton 2005). In particular, these findings suggest 214

that the triggers determining the dispersal decision for individuals in especially poor condition 215

markedly differ from those of other individuals. 216

We found clear evidence that parasite abundance contributes to condition-dependent 217

natal dispersal in a large herbivore, acting over and above the previously reported effect of 218

body mass in this roe deer population (Debeffe et al. 2012). For all but the lightest 219

individuals, the probability of dispersing was influenced by additive effects of nematode 220

abundance (negative) and body mass (positive), suggesting that energetically stressed 221

individuals that are unable to pay the costs of dispersal (Debeffe et al. 2012) should be 222

philopatric. Moreover, parasitised individuals that dispersed did so later in the season, 223

presumably once they had accumulated sufficient body condition to offset the combined costs 224

of dispersal and anti-parasite defense. Hence, condition-dependent dispersal is a multi-225

factorial process in which parasite abundance may play an important role. 226

In direct contrast to the above general pattern, the high dispersal propensity of three 227

individuals with both high nematode counts and low body mass was, at first sight, 228

contradictory to the condition-dependent dispersal of roe deer. We suggest that because these 229

poor condition individuals have an especially low reproductive value in their natal 230

environment, they shift to an emergency life history tactic (Wingfield 2003). That is, despite 231

the associated costs and the difficulty for these low quality individuals to pay them, they have 232

nothing to lose by dispersing away from poor quality and highly parasitised habitat in a last 233

ditch attempt to improve their reproductive prospects (Ims and Hjermann 2001). This result 234

requires confirmation on a larger sample size, but could represent a “leave it” response (from 235

the ‘take it or leave it’ emergency life history strategy (Wingfield 2003)) to the direct 236

perturbation posed by a high level of parasites in the natal habitat. If so, it should be 237

accompanied by physiological modifications (e.g. increased secretion of glucocorticosteroids 238

(Wingfield 2003)), providing an interesting avenue for future research on this question. 239

Interpreting the relationship between a host and its parasites is, however, complex, as 240

parasites also have the ability to manipulate their host, potentially altering their behaviour. 241

Indeed, parasites may increase their transmission rate by modifying the host’s social or 242

exploratory behaviour (Moore 2002). For instance, rats Rattus norvegicus that were infected 243

by Toxoplasma gondii had a higher propensity to explore novel stimuli in their environment 244

(Berdoy et al. 1995). Nematodes are known to affect both host activity and microhabitat 245

choice (Poulin 1994), hence, the dispersal of the three highly infested deer of low body mass 246

could alternatively be the result of host manipulation by their parasites. Increasing dispersal 247

propensity of their host could be beneficial for the parasites if a host which is in poor 248

condition constitutes a poor quality environment. In this case, selection may favour parasites 249

which are capable of manipulating their host to move to a more beneficial environment with 250

more and/or higher quality resources (hosts). 251

To conclude, our study has demonstrated a strong influence of endoparasite abundance 252

on the natal dispersal behaviour of their hosts. At the individual level, the dispersal outcome is 253

likely determined by the interaction between the condition-dependent tactic of the deer and 254

the interests of the parasite in terms of host environment quality. In any case, by modifying 255

their host’s natal dispersal behaviour, endoparasites likely have an important impact on the 256

dynamics and functioning of their host populations (Möller et al. 2004). 257

258

Acknowledgements 259

The study was supported by an INRA grant. We thank the local hunting associations and all 260

volunteers for their assistance during capture and Eric Pardo for carrying out fecal analyses. 261

We thank Atle Mysterud and 2 anonymous referees for comments on a previous version of 262

this manuscript. 263

264

References 265

266

Berdoy, M. et al. 1995. Parasite-altered behaviour: Is the effect of Toxoplasma-Gondii on 267

rattus-norvegicus specific? - Parasitology 111: 403–409. 268

Body, G. et al. 2011. Population density and phenotypic attributes influence the level of 269

nematode parasitism in roe deer. - Oecologia 167: 635–646. 270

Bonte, D. et al. 2012. Costs of dispersal. - Biol. Rev. 87: 290–312. 271

Boulinier, T. et al. 2001. Dispersal and parasitism. - In: Clobert, J. et al. (eds), Dispersal. 272

Oxford university press, pp. 169–179. 273

Bowler, D. E. and Benton, T. G. 2005. Causes and consequences of animal dispersal 274

strategies: relating individual behaviour to spatial dynamics. - Biol. Rev. 80: 205–225. 275

Brown, C. and Brown, M. 1986. Ectoparasitism as a cost of coloniality in cliff swallows 276

(Hirundo pyrrhonota). - Ecology 67: 1206–1218. 277

Brown, C. R. and Brown, M. B. 1992. Ectoparasitism as a cause of natal dispersal in cliff 278

swallows. - Ecology 73: 1718–1723. 279

Burnham, K. P. and Anderson, D. R. 1998. Model Selection and Inference. - Springer-Verlag. 280

Bush, A. O. et al. 1997. Parasitology Meets Ecology on Its Own Terms: Margolis et al. 281

Revisited. - J. Parasitol. 83: 575. 282

Cameron, P. G. et al. 1993. Effects of body size and parasite infection on the locomotory 283

performance of juvenile toads, Bufo bufo. - Oikos 66: 129–136. 284

Clobert, J. et al. 2001. Dispersal. - Oxford university press. 285

Debeffe, L. et al. 2012. Condition-dependent natal dispersal in a large herbivore: heavier 286

animals show a greater propensity to disperse and travel further. - J. Anim. Ecol. 81: 287

1327–1337. 288

Dufty, A. M. and Belthoff, J. R. 2001. Proximate mechanisms of natal dispersal: the role of 289

body condition and hormones. - In: Clobert, J. et al. (eds), Dispersal. Oxford 290

university press, in press. 291

Heeb, P. et al. 1999. Ectoparasite infestation and sex-biased local recruitment of hosts. - 292

Nature 400: 63–65. 293

Ims, R. A. and Hjermann, D. O. 2001. Condition-dependent dispersal. - In: Clobert, J. et al. 294

(eds), Dispersal. Oxford university press, pp. 203–216. 295

Jaenike, J. et al. 1995. Parasite-induced mortality in mycophagous drosophila. - Ecology 76: 296

383–391. 297

Kaltz, O. and Shykoff, J. A. 1998. Local adaptation in host-parasite systems. - Heredity 81: 298

361–370. 299

Kenward, R. E. et al. 2002. From marking to modelling: dispersal study techniques for land 300

vertebrates. - In: Bullock, J. M. et al. (eds), Dispersal ecology. Blackwell Publishing, 301

pp. 50–71. 302

McCullagh, P. and Nelder, J. A. 1989. Generalized linear models. - Chapman and Hall. 303

Möller, A. P. and Szep, T. 2011. The role of parasites in ecology and evolution of migration 304

and migratory connectivity. - J. Ornithol. 152: 141–150. 305

Möller, A. P. et al. 2004. Parasitism, host immune defence and dispersal. - J. Evol. Biol. 17: 306

603–612. 307

Moore, J. 2002. Parasites and the behaviour of animals. - Oxford university press. 308

Morellet, N. et al. 2011. Landscape composition influences roe deer habitat selection at both 309

home range and landscape scales. - Landsc. Ecol. 26: 999–1010. 310

Newey, S. and Thirgood, S. 2004. Parasite-mediated reduction in fecundity of mountain hares. 311

- Proc. R. Soc. B-Biol. Sci. 271: S413–S415. 312

Poulin, R. 1994. Meta-analysis of parasite-induced behavioural changes. - Anim. Behav. 48: 313

137–146. 314

Poulin, R. 2007. Evolutionary Ecology of Parasites. - Princeton University Press. 315

R Development Core Team. 2010. R: a language and environment for statistical computing. - 316

R Foundation for Statistical Computing,Vienna, Austria. URL http://www.R-317

project.org/. 318

Raynaud, J. 1970. Etude de l’efficacité d’une technique de coproscopie quantitative pour le 319

diagnostic de routine et le contrôle des infestations parasitaires des bovins, ovins, 320

équins et porcins. - Ann. Parasitol. Hum. Comparée 45: 321–342. 321

Roshier, D. A. and Reid, J. R. W. 2003. On animal distributions in dynamic landscapes. - 322

Ecography 26: 539–544. 323

Rózsa, L. et al. 2000. Quantifying parasites in samples of hosts. - J. Parasitol. 86: 228–232. 324

Sorci, G. et al. 1994. Maternal parasite load increases sprint speed and philopatry in female 325

offspring of the common lizard. - Am. Nat. 144: 153–164. 326

Stien, A. et al. 2002. The impact of gastrointestinal nematodes on wild reindeer: experimental 327

and cross-sectional studies. - J. Anim. Ecol. 71: 937–945. 328

Van Vuren, D. 1996. Ectoparasites, fitness, and social behaviour of yellow-bellied marmots. - 329

Ethology 102: 686–694. 330

Wingfield, J. C. 2003. Control of behavioural strategies for capricious environments. - Anim. 331

Behav. 66: 807–815. 332

Supplementary material (available online as Appendix at <www.oikosoffice.lu.se/appendix>). 333

Appendix 1 (Table A1: Performance of the 5 candidate generalized mixed linear models of 334

dispersal metrics, with body mass included as a continuous variable; Table A2: Parameter 335

estimates from the best generalized linear mixed model describing variation in dispersal 336

propensity as a function of body mass and nematode abundance, with body mass included 337

as a continuous variable; Figure A1: Influence of the two-way interaction between 338

nematode abundance and individual body mass on dispersal propensity with body mass 339

included as a continuous variable). 340

Table 1. Performance of the 5 candidate generalized mixed linear models of dispersal 341

propensity, distance and departure date, including landscape sector as a random factor. The 342

selected models occur in bold. k refers to the number of model parameters. “NematodesStd” 343

refers to the standardised nematode abundance. Body mass (“BM”) was included as a 344

categorical variable with three classes (heavy, average and light individuals, see text for 345

further details). 346

347

Dispersal behaviour Propensity (n = 67) Distance (n = 31) Departure date (n = 31) model k AICc ∆AICc AICcWt AICc ∆AICc AICcWt AICc ∆AICc AICcWt

NematodesStd X BM 7 86.2 0.00 0.96 102.07 6.89 0.02 256.35 5.15 0.05 NematodesStd + BM 5 94.72 8.52 0.01 98.10 2.92 0.14 254.64 3.45 0.12 NematodesStd 3 97.85 11.65 0.00 99.17 3.99 0.08 251.19 0.00 0.66 BM 4 93.78 7.58 0.02 95.19 0.00 0.60 257.36 6.17 0.03 Constant 2 96.19 9.99 0.01 97.91 2.72 0.15 254.16 2.97 0.15

348

349

350

Table 2. Parameter estimates from the best generalized linear mixed model describing 351

variation in dispersal propensity as a function of body mass and nematode abundance (n = 352

67). “NematodesStd” refers to the standardised nematode abundance. Body mass (“BM”) was 353

included as a categorical variable. 354

355

Coefficient Se Z value Pr (>|z|)

Intercept -1.4298 0.7893 -1.811 0.0701 NematodesStd 1.1553 0.7704 1.5 0.1337 Heavy BM category 6.5356 3.6507 1.79 0.0734 Average BM category 1.2851 0.8537 1.505 0.1323 NematodesStd : Heavy BM category -6.6904 3.7909 -1.765 0.0776 NematodesStd : Average BM category -1.4633 0.832 -1.759 0.0786 356

357

358

Figure 1. Influence of standardised nematode abundance and body mass on dispersal 359

propensity of roe deer (n = 67). Individuals were assigned to three categories in relation to 360

their body mass: heavy individuals, from 17.9 to 20 kg (black circles), average individuals, 361

from 15 to 17.9 kg (light grey circles) and light individuals, from 11.5 to 15 kg (grey 362

triangles). Lines correspond to the relationships described by the best model (with associated 363

SE), including landscape sector as a random factor. 364

365

-2 -1 0 1 2

0.0

0.2

0.4

0.6

0.8

1.0

Standardised nematode abundance

Dis

pers

al p

rope

nsity

light

average

heavy

366

367

Figure 2. Dispersal date of roe deer in relation to standardised nematode abundance (n = 31). 368

The line corresponds to the best model (with associated SE), including landscape sector as a 369

random factor. 370

-2 -1 0 1 2

80

90

100

110

120

130

140

150

Standardised nematode abundance

Dis

pers

al d

epar

ture

dat

e

371

372

Electronic Supplementary Material (SM) 373

374

Supplementary material Appendix 1, Table A1. Performance of the 5 candidate generalized 375

mixed linear models of dispersal propensity, distance and departure date, including landscape 376

sector as a random factor. “NematodesStd” refers to the standardised nematode abundance. 377

Body mass (“BM”) was included in the models as a continuous variable. The best models 378

occur in bold. k refers to the number of model parameters. 379

380

Dispersal behaviour Propensity (n = 67) Distance (n = 31) Departure date (n = 31) model k AICc ∆AICc AICcWt AICc ∆AICc AICcWt AICc ∆AICc AICcWt NematodesStd X BM 5 87.8 0.00 0.66 95.77 3.52 0.11 256.11 4.92 0.05 NematodesStd + BM 4 92.84 5.09 0.05 95.06 2.82 0.16 253.69 2.50 0.16 NematodesStd 3 97.85 10.1 0.00 99.17 6.93 0.02 251.19 0.00 0.57 BM 3 91.4 3.65 0.11 92.25 0.00 0.66 254.85 3.65 0.09 Constant 2 96.19 8.44 0.01 97.91 5.66 0.04 254.16 2.97 0.13

381

382

383

Supplementary material Appendix 1, Table A2. Parameter estimates from the best 384

generalized linear mixed model describing variation in dispersal propensity as a function of 385

body mass and nematode abundance (n = 67). “NematodesStd” refers to the standardised 386

nematode abundance. Body mass (“BM”) was included as a continuous variable. 387

388

Coefficient Se Z value Pr (>|z|)

Intercept -6.8877 2.769 -2.487 0.0129 NematodesStd 5.2475 2.4523 2.14 0.0324 BM 0.4182 0.1703 2.455 0.0141 NematodesStd : BM -0.3376 0.1513 -2.232 0.0256 389

390

Supplementary material Appendix 1, Figure A1. Influence of the two-way interaction 391

between nematode abundance and individual body mass on dispersal propensity (n = 67) with 392

body mass included as a continuous variable. Isoclines and grey shading on the 3D plot 393

represent the associated dispersal probability. Light grey indicates low dispersal probability 394

while dark grey indicates high dispersal probability (ANOVA between best model and 395

constant model: Chisq 3 = 15.23; p = 0.0016) 396

-2 -1 0 1 2

12

14

16

18

20

Standardised nematode abundance

Bo

dy m

ass

(kg)

397

398 399 400