Behavioral Ecology and Sociobiology

, Volume 63, Issue 7, pp 1045–1056

Network structure and parasite transmission in a group living lizard, the gidgee skink, Egernia stokesii

Authors

    • School of Biological SciencesFlinders University
  • C. Michael Bull
    • School of Biological SciencesFlinders University
  • Richard James
    • Department of PhysicsUniversity of Bath
  • Kris Murray
    • School of Biological SciencesFlinders University
    • The Ecology Centre, School of Integrative BiologyUniversity of Queensland
Original Paper

DOI: 10.1007/s00265-009-0730-9

Cite this article as:
Godfrey, S.S., Bull, C.M., James, R. et al. Behav Ecol Sociobiol (2009) 63: 1045. doi:10.1007/s00265-009-0730-9

Abstract

Gidgee skinks (Egernia stokesii) form large social aggregations in rocky outcrops across the Flinders Ranges in South Australia. Group members share refuges (rock crevices), which may promote parasite transmission. We measured connectivity of individuals in networks constructed from patterns of common crevice use and observed patterns of parasitism by three blood parasites (Hemolivia, Schellackia and Plasmodium) and an ectoparasitic tick (Amblyomma vikirri). Data came from a 1-year mark-recapture study of four populations. Transmission networks were constructed to represent possible transmission pathways among lizards. Two lizards that used the same refuge within an estimated transmission period were considered connected in the transmission network. An edge was placed between them, directed towards the individual that occupied the crevice last. Social networks, a sub-set of same-day only associations, were small and highly fragmented compared with transmission networks, suggesting that non-synchronous crevice use leads to more transmission opportunities than direct social association. In transmission networks, lizards infested by ticks were connected to more other tick-infested lizards than uninfected lizards. Lizards infected by ticks and carrying multiple blood parasite infections were in more connected positions in the network than lizards without ticks or with one or no blood parasites. Our findings suggest higher levels of network connectivity may increase the risk of becoming infected or that parasites influence lizard behaviour and consequently their position in the network.

Keywords

Social networksLizardsParasite transmissionGroup-living

Introduction

As social contact allows parasite transmission, parasites are considered a constraint on the evolution of social organisation, especially for group living species where close and frequent contact among group members may enhance transmission within groups (Alexander 1974; Moller et al. 1993; Altizer et al. 2003).

Parasites with direct lifecycles that require contact or close association among hosts for transmission probably pose the greatest cost to group living (Cote and Poulin 1995). These parasites can show aggregation within host social units (Boulinier et al. 1996) or positive relationships between host group size and parasite abundance (Brown and Brown 1986). Parasites with indirect lifecycles might also have a stronger effect on group living hosts if the vector or intermediate host has limited mobility (Poulin 1999; Godfrey et al. 2006).

Empirical field studies usually measure the costs of parasites to social organisation at the group level (Brown and Brown 1986; Moore et al. 1988; Arnold and Lichtenstein 1993). But social interactions or associations among group members are rarely homogenous. For example, within a social group, sexual partners or family members may interact more frequently than other group members (Emlen 1997), leading to different opportunities for parasite transmission (Altizer et al. 2003). Similarly, in many animal societies, individuals mix in fission-fusion systems where groups frequently segregate into smaller sub-groups varying in composition (Freeland 1979). ‘Floater’ individuals may move among groups, or sex-biased dispersal may permit transmission of parasites beyond the social group (Altizer et al. 2003; Brown and Brown 2004). These heterogeneities in social contact cannot easily be captured in group-level analyses yet may influence the transmission of parasites within socially structured populations.

For example, “super-spreaders” are individuals (Li et al. 2004), or groups of individuals (Woolhouse et al. 1997; Perkins et al. 2003; Skorping and Jensen 2004) that play a disproportionate role in the transmission of contagious diseases. These can result from a higher level of infectiousness or more frequent contact with other individuals (Galvani and May 2005; Lloyd-Smith et al. 2005). Frequent social contact may increase individual fitness through more mating opportunities whilst also increasing the risk of becoming infected. In the evolution of social organisation, these fitness tradeoffs may limit the extent or frequency of social contact.

Parasites may also influence social organisation by changing host behaviour, either by direct manipulation to enhance transmission opportunities (Moore 2001; Dunn 2005) or by reducing host vigour (Main and Bull 2000). Parasites can reduce the competitive ability of hosts in social conflicts (Schall and Houle 1992; Mougeot et al. 2005), altering association patterns and consequently network structure. These, in turn, may change pathways for parasite transmission. By viewing a population as a social network, where pair-wise behavioural associations form a matrix of social “connections” within a population, we can capture heterogeneities in social contact and trace potential pathways for parasite transmission.

Social networks provide a quantitative framework to analyse social structure (Krause et al. 2009). Individuals are considered “nodes”, and associations among individuals form “edges” that link nodes together into a social network. Networks can be symmetrical, where associations are considered mutual, or they can be directed, where associations are directed towards one or both of the individuals in a pair. In the study of disease transmission, directed networks can indicate the direction of possible transmission pathways through a population and identify individuals at greater risk of infection (Bell et al. 1999; Christley et al. 2005). Social networks have been used for understanding disease transmission in humans (particularly sexually transmitted diseases; Anderson et al. 1991; Jolly et al. 2001; Eames and Keeling 2002; Potterat et al. 2002; Liljeros et al. 2003; Masuda and Konno 2004) and livestock (Keeling et al. 2001; Webb 2005; Kao et al. 2006; Ortiz-Pelaz et al. 2006; Shirley and Rushton 2006) but less commonly in wildlife (Corner et al. 2003; Cross et al. 2004; Otterstatter and Thomson 2007). In natural host–parasite systems, social networks may advance our understanding of how parasites influence social organisation and how social organisation influences parasite transmission dynamics.

We investigated whether observed networks could predict infection patterns in a long-lived viviparous lizard, the gidgee skink (Egernia stokesii), from arid regions of Australia. In the Flinders Ranges of South Australia, gidgee skinks inhabit rocky outcrops and form stable social aggregations of up to 17 individuals that occupy up to five rock crevice refuges within group territories (Duffield and Bull 2002; Gardner et al. 2007). Group members are highly related and include multiple cohorts of the progeny of breeding adults (Gardner et al. 2001, 2002, 2007). “Core crevices” within the group territory are used exclusively and consistently by group members, especially older permanent members (Duffield and Bull 2002). “Marginal crevices” on the perimeter of the group territory may also be used by adjacent groups and non-group member “floaters” (Duffield and Bull 2002). Group membership remains stable for over 5 years (Duffield and Bull 2002) despite low levels of male-biased dispersal (Gardner et al. 2001).

Gidgee skinks host ectoparasitic ticks (Amblyomma vikkiri; Keirans et al. 1996) and three genera of protozoan blood parasites (Plasmodium, Schellackia and Hemolivia; Stein 1999). Plasmodium has dipteran vectors, probably phlebotamid sand flies (Stein and Dyce 2002), which can fly across lizard population sites. Hemolivia is transmitted by ticks (Telford 1984; Smallridge and Paperna 1997; Smallridge and Bull 1999; Stein 1999) that wait for hosts in rock crevices, with little independent movement between crevices (Duffield and Bull 1996). Schellackia may be transmitted by ticks or directly from mother to offspring (Stein 1999) when juveniles ingest the placenta from infected mothers soon after parturition (Lanham and Bull 2000). We expected ticks and the parasites they vector (Hemolivia and Schellackia) to be more strongly associated with network structure than Plasmodium which has a more mobile vector.

In a previous study, we found no relationships between host social group size and the prevalence of blood parasite infection within groups (Godfrey et al. 2006). However, Hemolivia and Schellackia had aggregated distributions, where more host groups had either high or low infection prevalence of these parasites than Plasmodium (Godfrey et al. 2006). We suggested that this pattern resulted from contact among social group members enhancing transmission within groups. Ticks can survive for extended dormant periods in rock crevices waiting for a host (Duffield and Bull 1996), so non-synchronous crevice use among lizards may also permit parasite transmission. In this way, male-biased dispersal and movement of lizards among marginal crevices may create opportunities for transmission beyond direct social association.

Our aim was to investigate whether networks could predict patterns of parasite infection, assuming higher levels of social connectivity would increase the likelihood of becoming infected. We constructed two different types of networks from a mark-recapture study of four populations of gidgee skinks. To demonstrate how direct social associations might allow parasite transmission, we constructed “social networks” from observations of lizards co-occupying the same crevice refuge on the same day. We expected social networks would be more fragmented, and provide fewer transmission pathways than “transmission networks”, which were an extension of social networks to represent possible transmission opportunities from non-synchronous crevice use. Transmission networks linked lizards that used the same crevice within a time period that may allow parasite transmission (41 days). Edges in this network were directed towards the individual that occupied the crevice last to indicate the possible direction of parasite transfer. We predicted that lizards infected by parasites would be connected to more other individuals in the transmission network than uninfected lizards. We also predicted that individuals adjacent to more infected individuals in the transmission network would be more likely to be infected themselves. We tested these predictions against each of the parasite species, and the number of parasite species a lizard harboured.

Methods

Mark-recapture study

We surveyed populations of E. stokesii from four isolated rocky outcrops in the southern Flinders Ranges near the township of Hawker, South Australia. The study populations were Camel Hill (CAM), Castle Rock (CAS), Castle Rock Ridge (CRR) and East Ridge (ER; Fig. 1; Godfrey et al. 2006). Crevice refuges vary in size in these sites, but crevice openings are on average 5–7 cm high, 35–50 cm wide and can be up to 81 cm deep (Arida 2005). In crevices, lizards are at least within visual and olfactory range to allow social recognition (Bull et al. 2000) and are often in physical contact.
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-009-0730-9/MediaObjects/265_2009_730_Fig1_HTML.gif
Fig. 1

Map of the extent of habitable rock crevices (shaded grey) for each of the four study populations of gidgee skinks and the spatial distribution of rock crevices (triangle) in Camel Hill (CAM)

Each population was surveyed 15 times, first in April 2003, and then at least once a month over the subsequent spring, summer and autumn (Aug 2003–March 2004). Surveys were usually 5 days in duration, with an average of 15.5 (2.46 SE) days between surveys (range 5–40 days). Usually, two populations were each searched for 1 to 3 h each day. Each site was sampled for 27 (ER) to 43 (CAM) days over the whole study (Table 1). We observed no deaths and one dispersal event (from CAS to CRR).
Table 1

Recapture statistics for each population of lizards

Population

N

%recap

Total observ.

Census days (N)

Lizards per census

Crevices (N)

mean

SE

CAM

49

75.5

225

43

5.07

0.53

42

CAS

55

78.2

223

40

5.55

0.65

39

CRR

63

66.7

248

34

6.70

0.88

49

ER

65

76.9

230

27

8.52

0.92

41

N total number of lizards captured, %recap percentage of lizards captured on more than one occasion, Total observ. total number of observations, Census days (N) the total number of census days, Lizards per census (mean) mean number of lizards per census; Crevices (N) the number of active (used by lizards) crevices in each population

Lizards were either manually extracted from crevice refuges or trapped in unbaited Elliot traps left near crevices for three sequential days. Crevices are a confined space and rarely offered an escape route, so lizards were highly detectable and reliably captured. Rock crevices where lizards were caught were individually labeled. Lizards were individually marked with permanent integrated transponder (PIT) tags inserted sub-dermally. Marked lizards could be subsequently identified by inserting a modified microchip reader into occupied crevices. On first capture, we measured snout-vent length (SVL), weight and determined sex of adults by everting the hemipenes of males. The number of ticks attached was recorded for each lizard.

Infection status for Plasmodium, Schellackia and Hemolivia was determined for 171 lizards (73.7% of all lizards) across the four populations (Table 2) from the first available blood sample from each lizard, which were collected throughout the study period. Although blood parasite infections in E. stokesii can vary from 2 to more than 18 months duration (Stein 1999), such that lizards could have changed infection status over the study, we considered our samples to be representative of infection status. We made a thin smear of blood collected from the caudal vein onto a microscope slide. The slides were air dried, stained with a Modified Wright–Giemsa Stain in a Hematek slide-stainer, and examined under 1,000× oil immersion at a cell density of about 100 cells per microscope field. If no parasites were detected after 100 fields, the lizard was considered to be uninfected for that species. Parasite prevalence varied among populations (Table 2). All blood parasite species were, however, present in each of the four populations and lizards could therefore have multiple blood parasite infections.
Table 2

Parasite prevalence (percentage of lizards infected by the parasite) among populations of gidgee skinks

Population

Ticks

Hemolivia

Schellackia

Plasmodium

N

%inf

(95% CI)

N

%inf

(95% CI)

%inf

(95% CI)

%inf

(95% CI)

CAM

49

0

(0–7.3)

34

2.9

(0.5–14.9)

64.7

(47.9–78.5)

20.5

(10.3–36.8)

CAS

54

27.7

(17.6–40.9)

40

72.5

(57.2–83.9)

57.5

(42.2–71.5)

15

(7.1–29.1)

CRR

62

22.5

(14.0–34.4)

43

6.9

(2.4–18.6)

46.5

(32.5–61.1)

58.1

(43.3–71.6)

ER

65

3.0

(0.9–10.5)

54

20.3

(11.7–32.9)

31.4

(20.7–44.7)

59.2

(45.9–71.3)

N number of lizards sampled for parasites, %inf percentage of sampled individuals that were infected for that species, with 95% confidence intervals

Network construction

We used the mark-recapture records to construct social networks and transmission networks. Crevice use is a good indicator of both social association (individuals in the same social group often share crevices), and parasite transmission (ticks wait in crevices to attach to new hosts). Thus we constructed our networks around common crevice use. Social networks were constructed by assigning an undirected edge between two individuals that used the same crevice on the same day from observations aggregated over the duration of the study.

Our transmission networks reflect the fact that an infected animal A can leave a crevice yet still, within a time period T, potentially infect animal B using the crevice some time later. This is represented by a directed edge from A to B. We take T = 41 days, an estimate of the mean off-host survival period for the tick. This includes a pre-moult period (time from detachment from the reptile host to ecdysis), and the post-moult survival time of ticks. The pre-moult period for A. vikirri nymphs is assumed to be similar to a related tick species with similar environmental preferences, Bothriocroton hydrosauri, which takes 13.6 days (Chilton et al. 2000). The mean post-moult survival period for unfed nymphs of A. vikirri under conditions similar to those experienced in the field is 27.4 days (Duffield and Bull 1996). Hence, we assumed ticks could survive an average of 41 days in the crevice refuge. A two-way edge was placed between lizards if they used the same crevice on the same day or if lizards alternated in crevice occupation within the 41-day period, such that both lizards had opportunities for transmission in both directions. All our networks have binary edges (they are either there or not and are not weighted by association strength or frequency) because association frequency would be influenced by the irregular intervals between our censuses.

Network construction by this method is known as ‘gambit of the group’ (James et al. 2009), where close association is inferred from group membership. Interpretation of networks constructed by this method can be problematic when group sizes are large and it is unlikely that all individuals within the group equally associate (James et al. 2009). However, in our study, the number of lizards using a crevice was small, and we assumed that lizards had equal opportunity of becoming infected by any parasites present in the crevices they used.

Transmission network connectivity and parasite infection

We predicted that higher network connectivity would increase the likelihood of a lizard becoming infected, so lizards infected by parasites would have more network neighbours than uninfected lizards. To measure the connectivity of individuals in the transmission network, we calculated two different network parameters: “in-degree” and “total-degree”. In-degree is the number of nodes connected to an individual node by edges directed towards that node. Total-degree is the number of nodes connected to an individual by edges directed towards and away from that individual and indicates the overall connectivity of an individual.

We also predicted that parasite transmission would be influenced by network structure, so lizards infected by a parasite species would be adjacent to more lizards infected by that parasite species in the network than would uninfected lizards. To test this, we calculated two other degree metrics that measure the connectivity of lizards to infected lizards in the network, “parasite-in-degree” and “parasite-total-degree”. These measures are equivalent to in-degree and total-degree, but consider only the number of infected nodes connected to an individual. Parasite-degree measures were calculated for populations with a prevalence of more than 20% for that parasite (Table 2).

Node-based measures such as degree are derived from relational data, so, within a single population, they violate the assumption of independence in parametric statistical analyses (James et al. 2009). Thus, instead of using individual lizards as our unit of measurement, we treated populations of lizards as our replicate units and used population means in our analyses. We log-transformed node-based measures of lizards to normality and calculated the mean degree measures for each combination of factor levels that we tested within each population.

We calculated mean degree measures for lizards with different infection states (infected or uninfected) among males, females and sub adults within each population. Lizard sex/age classes were included to determine whether they influenced relationships between connectivity and parasitism. We also tested whether the total number of blood parasite species (none, one or multiple blood parasite infections) or tick infestation (infected or uninfected) had an effect on total connectivity (total-degree and in-degree). Because there were multiple means derived from the same sample of lizards for each population, we used a linear mixed model analysis in SPSS 15.0 to account for repeated measures within populations. We included main effects and two-way interactions in the model and checked residuals for normality.

Parasite infection patterns

Demographic factors, such as lizard sex or age class, that might influence the position of a lizard in the network may also influence parasite infection. The number of crevices a lizard used could also influence infection likelihood by exposing the lizard to more different potential sources of infection. We used a generalized linear model (GLM) in R (R Development Core Team 2007) to determine whether the likelihood of parasite infection was influenced by any of these factors. We included population and lizard sex (male, female or sub-adult) as factors, and lizard SVL and the number of different crevices it used as covariates in the model. A separate analysis was conducted for each parasite species (ticks, Hemolivia, Schellackia and Plasmodium), with infection status as the dependent variable (0 = uninfected or 1 = infected) and a binomial error distribution. The total number of blood parasite species infecting a lizard was normally distributed, so we fitted the same model with Gaussian error distribution. We began with a maximal model including all main effects and all two-way interactions among main effects, then reduced the model by removing non-significant terms until only significant terms remained. Changes in deviation between each model reduction were tested with an ANOVA F test for models with a Gaussian error distribution and an ANOVA χ2 test for models with a binomial error distribution. Changes in deviation of deleted terms, terms included in the final model and associated P values, are reported in the “Electronic supplementary material”.

Results

We made 926 observations of 232 lizards across the four populations. Lizards (74.1%) were recaptured on one or more occasions (Table 1). Although the mean number of lizards detected per census day varied among populations (Table 1), there was no significant difference in mean observation frequency of lizards among populations (mixed model analysis, P = 0.108). We made 1–17 observations (mean 4.0, SE 0.23) on each lizard and recorded individuals in 1–7 different crevices (mean 2.1 SE 0.08). There were 1–8 lizards per crevice (mean 2.19, 0.11 SE) within any 41-day period.

Social and transmission networks

Social networks, constructed from observations of lizards using the same crevice on the same day, formed 31 network fragments across the four populations (Fig. 2a, c, e, g; Table 3). The largest fragment in each population varied from 5–12 lizards (Table 3). They were not spatially isolated, with 43–86% of network fragments within populations overlapping in crevice use by an average of 27.3% to 70% of crevices (Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-009-0730-9/MediaObjects/265_2009_730_Fig2_HTML.gif
Fig. 2

Diagrams of social networks (SN) and transmission networks (TN) for each population: a CAM-SN, b CAM-TN, c CAS-SN, d CAS-TN, e CRR-SN, f CRR-TN, g ER-SN, h ER-TN. Letters located on the right of nodes (circles) denote the fragment those individuals belong to in the social network and where this fits into the transmission network. Social networks are symmetrical (no arrows are shown), and transmission networks are directed with arrows showing the direction of the association

Table 3

Summary table of network fragments for social and transmission networks in each population

Population

Social network

Transmission network

Nf

Sizemax

%ove

%crev

Nf

Sizemax

%ove

%crev

CAM

6

6

83

51.9

6

14

67

42.4

CAS

7

12

86

70.0

6

37

50

68.7

CRR

11

5

73

68.1

8

26

63

65.8

ER

7

12

43

27.3

6

15

50

14.5

Nf number of fragments, Sizemax size (number of nodes) of the largest network fragment, %ove percentage of fragments within the population that overlapped in crevice use with another fragment, %crev mean percentage of crevices within a fragment that were used by other fragments within that population

Transmission networks, constructed from observations of non-synchronous occupancy of crevices, formed 26 network fragments across the four populations (Table 3; Fig. 2b, d, f, h). Transmission networks were less fragmented than social networks, with up to five social network fragments linked in the transmission networks. This was expected since derived social networks were a sub-set of transmission networks. It indicates that non-synchronous crevice use may extend transmission opportunities beyond direct crevice sharing. However, transmission network fragments were not spatially distinct, with 50–67% of fragments within populations overlapping in crevice use by an average of 14.5% to 68.7% of crevices (Table 3).

Transmission network connectivity and parasite infection

Tick infestation status was significantly associated with overall connectivity in the transmission network. Lizards infested by ticks had a higher mean number of network neighbours than lizards without ticks (Table 4; Fig. 3a). The number of blood parasite species infecting a lizard was significantly influenced by total-degree and in-degree. Lizards infected by two or more blood parasite species were in more connected positions in the network than lizards infected by one or no blood parasite species (Fig. 3b). There was a significant interaction between tick infestation and the number of blood parasites on total-degree and in-degree (Table 4). Lizards with ticks had higher degree measures than those without ticks, but the difference was greatest in lizards with two or more blood parasites (Fig. 3c). However, there were no significant effects of parasite infection status by any of the blood parasite species separately or of lizard sex on total-degree or in-degree measures (Table 5). Tick infestation status had a significant effect on tick-total-degree and tick-in-degree, with lizards infested by ticks being connected to more other infested lizards than lizards without ticks (Table 6, Fig. 4). There were no significant effects of sex or parasite infection for any of the other parasites on parasite total-degree or in-degree (Table 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs00265-009-0730-9/MediaObjects/265_2009_730_Fig3_HTML.gif
Fig. 3

Bar-chart of back-transformed estimated means of a mixed model analysis comparing mean total-degree of populations (n = 4) between a lizards infested by ticks and without ticks, b lizards infected by none, one or multiple (2+) blood parasite species and c an interaction between tick infestation status and the number of blood parasite species infecting the lizards, with error bars showing 95% confidence intervals

https://static-content.springer.com/image/art%3A10.1007%2Fs00265-009-0730-9/MediaObjects/265_2009_730_Fig4_HTML.gif
Fig. 4

Bar-chart of back-transformed estimated means of a mixed model analysis comparing the mean tick-total-degree of populations (n = 4) between lizards infested by ticks and without ticks, with error bars showing 95% confidence intervals

Table 4

Results of mixed model analyses comparing the population means of total-degree and in-degree between lizards infected by ticks or not, the number of blood parasite species infecting a lizard, and an interaction between these effects

 

Total-degree

In-degree

df

F

P value

F

P value

Ticks

1.5

45.543

0.001

30.054

0.003

Parasite sp.

2.5

21.415

0.004

5.675

0.052

Ticks*Parasite sp.

2.5

6.476

0.041

7.510

0.031

Table 5

Results of mixed model analyses comparing the population means of total-degree and in-degree between sexes, lizards infected by each blood parasite and an interaction between these effects

 

Hemolivia

Schellackia

Plasmodium

df

F

P value

df

F

P value

df

F

P value

Total

Sex

2.5

0.628

0.571

2.15

0.734

0.497

2.9.3

0.171

0.846

Parasite

1.5

0.229

0.652

1.15

0.299

0.593

1.9.3

0.311

0.590

Sex*parasite

2.5

0.368

0.710

2.15

1.290

0.304

2.9.3

1.599

0.253

In

Sex

2.5

1.588

0.292

2.15

0.604

0.559

2.9.3

0.319

0.735

Parasite

1.5

0.129

0.734

1.15

0.259

0.618

1.9.3

0.189

0.674

Sex*parasite

2.5

0.045

0.957

2.15

1.026

0.382

2.9.3

1.649

0.244

Table 6

Results of mixed model analyses comparing the population means of parasite-degree measures (the number of infected lizards an individual is directly connected to) among sexes, parasite infection status for that parasite and the interaction between these effects for total-degree measures and in-degree measures for each parasite species

 

Tick-degree

Hemolivia-degree

Schellackia-degree

Plasmodium-degree

df

F

P value

df

F

P value

df

F

P value

df

F

P value

Total-degree

Sex

2.5

1.232

0.367

2.5

1.327

0.345

2.15

0.058

0.944

2.9

0.035

0.966

Parasite

1.5

19.350

0.007

1.5

0.982

0.367

1.15

0.035

0.855

1.9

0.087

0.775

Sex*Parasite

2.5

0.131

0.880

2.5

0.940

0.450

2.15

0.078

0.925

2.9

2.530

0.134

In-degree

Sex

2.5

0.823

0.491

2.5

1.459

0.317

2.15

0.145

0.867

2.9

0.722

0.512

Parasite

1.5

12.127

0.018

1.5

0.006

0.942

1.15

0.015

0.903

1.9

0.026

0.874

Sex*parasite

2.5

0.547

0.610

2.5

0.008

0.992

2.15

0.026

0.975

2.9

2.672

0.123

Parasite infection patterns

Infection likelihood was positively related to lizard size for ticks (χ2 = 7.542, df = 1, P = 0.006, odds ratio 1.02 (1.01–1.04 95% CI)) and Schellackia (χ2 = 4.926, df = 1, P = 0.026, odds ratio 0.99 (0.97–1.01 95% CI). Likelihood of Hemolivia infection also showed a positive relationship with lizard size but only in ER (χ2 = 8.501, df = 1, P = 0.004, odds ratio 1.03 (1.01–1.06 95% CI). There was no effect of lizard sex on the likelihood of infection for any of the parasite species separately, but the total number of blood parasites infecting a lizard was significantly higher in females (mean 1.39, 0.08 SE) than males (mean 1.07, 0.09 SE) or sub-adults (mean 0.92, 0.09 SE; F2,165 = 7.898, P < 0.001). The number of different crevices used by a lizard was positively related to infection likelihood by Schellackia in CRR only (χ2 = 6.396, df = 1, P = 0.008, odds ratio 2.46 (1.24–5.53 95% CI). There was no effect of crevice number on any of the other parasite species or the total number of blood parasite species. Non-significant terms removed from these models are reported in “Electronic Supplementary material”.

Discussion

Gidgee skinks live in large social aggregations that should enhance opportunities for parasite transmission among group members. Previously, we found aggregated infection by Hemolivia and Schellackia at the group level, suggesting high rates of transmission of tick-borne parasites within groups (Godfrey et al. 2006). In the current study we used network techniques to describe heterogeneity in the transmission opportunities resulting from non-synchronous use of refuges. Transmission networks, constructed from common use of crevices within an estimated transmission period, provided more extensive opportunities for parasite transmission than social networks derived from direct crevice sharing. We found empirical evidence that the connectivity of individuals in the transmission networks predicted parasite infection patterns.

Our first main finding was that lizards infested by ticks were in more connected positions in the network than lizards without ticks, and were connected to significantly more tick infested lizards. These results supported our hypothesis that the transmission of ticks is influenced by network structure, with the likelihood of a lizard becoming infected by ticks increasing with the number of lizards it shared crevices with, and with the number of infected lizards it shared crevices with.

Our second main finding was that lizards harbouring multiple blood parasite infections were more highly connected in the transmission network than lizards infected by one or no blood parasite species. This relationship was amplified by tick infestation. Lizards infected by multiple blood parasite species and ticks were twice as connected as lizards infected by fewer blood parasite species and no ticks. There was no relationship between the number of different crevices used by a lizard and either infection by ticks or the number of blood parasite species infecting a lizard. So, the likelihood of a lizard becoming infected by parasites did not simply increase with the number of different crevice refuges it used, so much as the extent to which crevices used by a lizard were also used by other lizards. Both tick infestation status and the number of blood parasite species infecting a lizard were also influenced by the sex or age of lizards, but those factors were not significantly associated with degree measures on the network. So we considered that demographic variation in parasitism played a lesser role in transmission dynamics than network connectivity. These findings were consistent with our hypothesis that network connectivity predicted parasite infection, with lizards in well connected network positions being at a greater risk of attaining parasites.

An alternative explanation of our empirical observations is that network structure resulted from infections, rather than that infection patterns resulted from the network structure. There is substantial experimental evidence that parasites can reduce competitive ability in territorial disputes (Schall and Houle 1992; Hoodless et al. 2002) and manipulate host behaviour, altering social organisation. For example, experimental reduction of nematode infection in red grouse (Lagopus lagopus scotius) significantly increased the likelihood of winning territorial contests (Mougeot et al. 2005). Consequently, parasites could influence the position of individuals in a social network. Although demographic and familial structuring probably dominates the position of gidgee skinks in social groups (Duffield and Bull 2002), parasites could reduce host fitness and cause lizards to be displaced from their social position. Fenner and Bull (2008) found the time gidgee skinks spent moving and basking increased when nematode loads were experimentally reduced. Main and Bull (2000) found ticks reduced the vigour of Tiliqua rugosa, a closely related skink species. However, we do not know how blood parasite infections impact the fitness and social structure of gidgee skinks.

If infected hosts are displaced from a social group, the impact of parasites on network structure may depend on where lizards move to. If displaced lizards move to isolated crevices that are rarely used by other lizards, network connectivity of infected lizards would be lowered. Alternatively, if displaced lizards move to marginal crevices which may be used, intermittently, by more lizards, then connectivity of infected individuals would be higher, as reported in this study. Experimental manipulation of parasite infection of individuals in network positions with specified levels of connectivity would provide insights into whether parasites influence network position and in what way. Tracking infection dynamics from experimentally infected individuals would also allow us to determine whether infection dynamics depend on network connectivity.

Future models would also benefit from the use of weighted networks where edges are weighted according to the strength or frequency of associations, particularly where reduced infectivity of a parasite may dilute relationships between network structure and infection patterns. For example, Kao et al. (2007) modeled the transmission of two diseases that had different levels of infectiousness on networks of livestock movement in the UK. They found that scrapies had a weaker signal on contact networks than the more infectious foot-and-mouth disease (Kao et al. 2007). Similarly, in Hemolivia, the tick vector does not always become infected from feeding on an infected lizard, and a lizard that ingests an infected tick does not always become infected (Smallridge and Bull 1999). This could explain why we observed stronger relationships for ticks than for individual parasites that require a vector or intermediate host for transmission.

The results from each blood parasite species separately were less closely related to network connectivity. Plasmodium has a mobile dipteran vector, and we did not expect strong associations with network structure. For Schellackia, transmission from mother to offspring (Stein 1999) may obscure relationships between network structure and infection patterns. Transmission of Hemolivia by ingestion of infected ticks (Smallridge and Bull 1999) might be more frequent when ticks are in the engorged stage, soon after detachment. Thus, our transmission period may be over-estimated, and transmission networks may have been too dense to detect patterns of Hemolivia transmission. Our ability to detect associations between blood parasites and network position may have also been limited by missing blood parasite samples for some individuals. Ticks were more easily sampled, so we had a more complete picture of their distribution on the network. Despite this, the strong relationships we found between network connectivity and the diversity of blood parasite infection suggests that there is a link between network structure and blood parasite transmission.

When interpreting relationships between network structure and parasite infection, another factor to consider is the potential bias in network construction. Derived network structure and associated node-based measures are sensitive to sampling frequency, with more observations likely to yield more connections (James et al. 2009). In our study, more frequent recaptures of individuals is likely to inflate degree measures, although this may be buffered to an extent by the high levels of crevice fidelity of gidgee skinks (Duffield and Bull 2002). However, unless the likelihood of recapture is biased by parasite infection, we do not expect this to impact our overall findings. Our sampling periods were irregularly spaced, and there was variation in observation frequency of lizards within each population; however, we had a similar sampling regime among populations that were similar in size. Thus, we believe it was appropriate to compare different classes of lizards within populations and use the four populations of lizards as replicate units because each population was sampled with a similar bias.

Although several studies have suggested contact network structure plays an important role in the transmission of disease within host populations, few have compared network results with empirical data on parasite prevalence. Corner et al. (2003) found that highly connected possums (Trichosurus vulpecula) were more likely to become infected by bovine tuberculosis when the disease was experimentally introduced to other highly connected individuals. Otterstatter and Thomson (2007) also found contact rates with experimentally infected bumblebees (Bombus impatiens) strongly predicted infection probability of individual bees within hives. Our study is one of the first to use network techniques to examine natural patterns of parasitism within wild animal populations and is consistent with these experimental studies, showing that more connected individuals are more likely to be infected.

Although we cannot confirm causality, our findings are consistent with the general hypothesis that group living may increase rates of parasitism (Moller et al. 1993; Cote and Poulin 1995). However, instead of investigating relationships between group size and infection prevalence, we have shown that individual levels of connectivity are associated with an increased likelihood of parasite infection. If parasites are costly to host fitness, then our findings suggest a cost for lizards to associate with a larger number of other lizards because of increased risk of infection. Alternatively, if parasites reduce host fitness and consequently alter social structure, then our findings could suggest that parasite infection results in increased connectivity of individuals. This then raises the prospect that parasites could regulate social structure through interactions between host fitness, social position and changes in transmission pathways. Network analysis provides a promising new method to explore how structure of host populations can impact the transmission dynamics of parasites and how parasites may influence the evolution of social organisation.

Acknowledgements

We thank two anonymous reviewers for constructive comments on an earlier version of this manuscript. This research was funded by grants from the Australian Research Council. The ARC/NHMRC Research Network for Parasitology Travel Award supported a research exchange to the University of Leeds and the University of Bath. We thank Professor Jens Krause and members of his research group for valuable discussions about this research. The study was conducted according to the guidelines of the Flinders University Animal Welfare Committee in compliance with the Australian Code of Practice for the use of animals for scientific purposes. All procedures carried out in this study conformed to the current laws of Australia.

Supplementary material

265_2009_730_MOESM1_ESM.doc (63 kb)
ESMSummary table of GLM analysis of factors influencing parasite infection status of ticks, Hemolivia, Schellackia, Plasmodium and the total number of blood parasite species (no. parasite species). Wald χ2/F values, df and P values in bold were remaining in the final minimal model after backward stepwise deletion of non-significant terms. All other non-significant values given were at the time of their deletion from the model (DOC 61.0 KB).

Copyright information

© Springer-Verlag 2009