SUPPLEMENTARY INFORMATION
Knowles et al. Stability of within-host parasite communities in a wild mammal system
1. ANTIPARASITE EFFECTS OF IVERMECTIN
Ivermectin is a macrocyclic lactone that kills a broad range of nematodes (1). At the dose used here, Ivermectin effectively kills larval as well as adult infections with Heligmosomoides polygyrus bakeri (2, 3), a laboratory model closely related to H. polygyrus polygyrus, the dominant nematode in these wood mouse populations. It has also been shown to effectively treat H. polygyrus in free-living yellow-necked mice, Apodemus flavicollis (4). Ivermectin is also known to kill some arthropods (1), but efficacy against fleas and ticks (the ectoparasites found on wood mice in these populations) is low (1, 5), and indeed we find no evidence that Ivermectin affected ectoparasites in this study (Fig. S8).
2. PARASITE DETECTION METHODS
2.1 Gastrointestinal parasites
Gastrointestinal (GI) parasites were detected using the salt flotation technique (6). Saturated salt solution was added to formalin-preserved faecal samples, such that eggs and oocysts in each sample could be concentrated on a coverslip, and scanned for parasite detection at 10x magnification. 40x magnification was used for parasite identification and making parasite species-specific egg/oocyst counts. Eimeria species were identified by unsporulated oocyst morphology (7). For each parasite species, the number of eggs (for helminth species) or oocysts (for Eimeria spp.) per gram of faeces in each sample was then calculated. Where multiple samples were present for an individual within a 3-day trapping period, the arithmetic mean egg/oocyst was taken across these days.
2.2 Blood parasites
DNA was extracted from blood samples using a DNAzol BD (Invitrogen) based method designed for small blood volumes (8). DNA concentration was measured using a Qubit dsDNA HS Assay Kit (Invitrogen) for use as a measure of extraction quality. Trypanosomes were detected using a nested PCR targeting a 530bp section of the 18S rRNA gene (9). Primers TRY927F 5’-GAAACAAGAAACACGGGAG and TRY927R 5’-CTACTGGGCAGCTTGGA were used in the first round, and SSU561F 5’-TGGGATAACAAAGGAGCA and SSU561R 5’-CTGAGACTGTAACCTCAAAGC in the second round. Each reaction contained 2ml genomic DNA, 0.1mM each dNTP, 0.2mM each primer, 0.8mM MgCl2, 0.5U Platinum Taq DNA Polymerase (Invitrogen) with the accompanying buffer at 1x concentration. A touchdown PCR profile was used, with initial denaturation at 96°C for 3min, followed by 20 cycles of 96°C for 30s, 60°C for 1min (decreasing by 0.5°C each cycle until 55°C), and 72°C for 90s, followed by a final extension at 72°C for 10min. Bartonella spp. were detected using a semi-nested PCR targeting part of the 16S-23S intergenic spacer region (10). In the first round, primers Big-F (5-TTGATAAGCGTGAGGTC) and Bog-R (5’-TGCAAAGCAGGTGCTCTCCCA) were used, before a second round using Big-F and Big-R (5’-TCCCAGCTGAGCTACG). Each reaction contained 2ml genomic DNA, 0.1mM each dNTP, 0.1mM each primer, 3mM MgCl2, 1U GoTaq DNA Polymerase (Promega) with the accompanying buffer at 1x concentration. Again a touchdown PCR profile was used, with initial denaturation at 96°C for 3min, followed by 20 cycles of 96°C for 10s, 61°C for 10s (decreasing by 0.5°C each cycle until 55°C), and 72°C for 50s, followed by a final extension at 72°C for 10min. In PCRs for both Bartonella and trypanosomes, conditions in the second round were the same as those in the first, except for the addition of 15 cycles with annealing at 55°C, and the use of 2ml product from the first round as template instead of genomic DNA. All PCR plates contained at least five evenly spaced negative controls to detect contamination and one positive control. In the rare event that any negative control was positive, the entire PCR plate was repeated. 5ml of all PCR products were run on 2% agarose gels stained with ethidium bromide and visualized under UV light. For the trypanosome PCR, samples showing an approx. 500bp band were scored as positive. 30 of these positives were sequenced and across the 502bp sequence obtained (excluding primers) all were 100% identical to Trypanosoma grosi (Genbank accession: AB175624). In the Bartonella PCR, product size can indicate species (10). Thus, positive samples were re-run on a higher resolution 4% gel for 2 hours for accurate band sizing. We tested the accuracy of this species identification method by sequencing 93 PCR products from wood mouse samples collected in the same study populations in 2009 that contained a single band. The five most common bands corresponded to sequence types previously described in Manor Wood as B. taylorii, B. birtlesii, B. grahamii, B. doshiae-like and BGA (10). The accuracy of species assignment by this method was 89% (Table S3).
3. STATISTICAL ANALYSES
3.1 Negative binomial and zero-inflated negative binomial models
Both nematode and Eimeria spp. egg/oocyst count data were highly aggregated, with most individuals shedding zero or very few eggs and a minority showing high egg counts (Fig. S1). Depending on the level of overdispersion, data can be modelled with either a negative binomial model (in this case NB2) or, if there are excessive zeros, by a mixture model (zero-inflated negative binomial model, ZINB), which contains two parts: the first models the count data (eggs/occyst per gram faeces of all individuals, including zero counts), and the second models excess zeros over and above those of the count process (11). We used both model types to inform our modelling strategy for these aggregated parasite data, using the R functions ‘glm.nb’ and ‘zeroinfl’ in packages MASS and pescl respectively. For both parasite groups, we examined the short-term effect of treatment (1-3 weeks post treatment) on both these variables, using both NB2 and ZINB models.
For both types of GI parasite (nematodes and Eimeria spp.), evidence for more zeros than expected under a negative binomial distribution was found, both from plots of expected vs. observed frequency distributions (11), and Vuong tests indicating superior fit of ZINB over NB2 models (nematodes: p=0.008 and Eimeria spp: p=0.003 respectively). Results of the short-term treatment models indicated that for nematodes, treatment predominantly affected the probability of having a zero count (Egg count (treated vs. untreated mice): incidence rate ratio (IRR) = 0.379, normal 95% CI 0.121, 1.183; bias-adjusted bootstrap CI: 0.061, 0.915. Probability of having zero eggs (treated vs. untreated mice): odds ratio (OR) = 10.03, normal 95% CI: 2.30, 43.77, bias-adjusted bootstrap CI: 1.782, 52.199). However, the opposite was true for Eimeria spp., where treatment affected oocyst count but not the probability of having a zero count (oocyst count (treated vs. untreated mice): IRR=16.45, normal 95% CI: 5.02, 53.84; bias-adjusted bootstrap CI: 4.633, 88.681. Probability of having zero oocysts (treated vs. untreated mice): OR= 2.12, normal 95% CI: 0.61, 7.43, bias-adjusted bootstrap CI: 0.600, 8.634). These results, showing differential effects of treatment on counts vs. the probability of having a zero count for both nematodes and Eimeria support our decision to subsequently model infection probability and infection intensity (among infected individuals only) separately for these two parasite groups.
3.2. Comparison of model selection methods
For each analysis, we compared results obtained using backwards stepwise elimination of terms from a full starting model, with an AIC-based model selection approach, in which all possible subsets of a full model including treatment effects and covariates were ranked according to AICc (Akaike’s Information Criterion corrected for small sample size) using the R package MuMIn (12). Model selection method did not affect conclusions; the minimum adequate model arrived at by backwards model simplification frequently matched the most parsimonious model from the AIC-based approach (the simplest model within 2 AICc units of the model with the lowest AICc), with the only differences concerning borderline significant covariate effects in backwards stepwise simplification that were not always upheld by the AIC-based approach.
4. FULL FIELDWORK AND MICROSCROPY ACKNOWLEDGEMENTS
Trevor Jones, Rebecca Barber, Lizzie Huxley-Jones, Alisha Imam, Kimberley Pope, Jan Popiela, Jen Shaw, Elizabeth Hughes, Sophie Martin, Laura Riley, Nikki Childs, Fiona McDougall, Anish Pandey, Lewis Botcherby, Andy Davies, Sian Davis, Rachel Piggott, Adam Phillips, Emma Ashcroft, Kim Graham, Kelly Garner, Martin Grunnill, Louise Pickett, Felix Horns, Victoria Spencer, Mike Kelly, Andy Turner, Sara Blanchard, Peter Sirl, Rebecca Jones, Julia Bell, Sarah Gore, Rachael Smith, Alice Muntzer, Camille Tsang, Lydia Moulden, Lachlan Wilmott, Joy Leng, Stephen Price, Leslie Norris, Christina Norris, Jen Coombs, Jenny Rannard
5. REFERENCES
1. Campbell W, Fisher M, Stapley E, Albersschonberg G, Jacob T (1983) Ivermectin - a Potent New Anti-Parasitic Agent. Science 221:823–828.
2. Wahid F, Behnke JM, Conway D (1989) Factors affecting the efficacy of Ivermectin against Heligmosomoides polygyrus (Nematospiroides dubius) in mice. Vet Parasitol 32:325–340.
3. Njoroge JM, Scott ME, Jalili F (1997) The efficacy of ivermectin against laboratory strains of Heligmosomoides polygyrus (Nematoda). Int J Parasitol 27:439–442.
4. Ferrari N, Cattadori I, Nespereira J, Rizzoli A, Hudson P (2004) The role of host sex in parasite dynamics: field experiments on the yellow-necked mouse Apodemus flavicollis. Ecol Lett 7:88–94.
5. Santora KA, Zakson-Aiken M, Rasa C, Shoop W (2002) Development of a mouse model to determine the systemic activity of potential flea-control compounds. Vet Parasitol 104:257–264.
6. Pritchard M, Kruse G (1982) The collection and preservation of animal parasites (University of Nebraska Press, Lincoln, Nebraska).
7. Nowell F, Higgs S (1989) Eimeria species infecting wood mice (genus Apodemus) and the transfer of two species to Mus musculus. Parasitology 98:329–336.
8. Mackey K, Steinkamp A, Chomczynski P (1998) DNA extraction from small blood volumes and the processing of cellulose blood cards for use in polymerase chain reaction. Mol Biotechnol 9:1–5.
9. Noyes H, Stevens J, Teixeira M, Phelan J, Holz P (1999) A nested PCR for the ssrRNA gene detects Trypanosoma binneyi in the platypus and Trypanosoma sp. in wombats and kangaroos in Australia. Int J Parasitol 29:331–339.
10. Telfer S et al. (2005) Disruption of a host-parasite system following the introduction of an exotic host species. Parasitology 130:661–668.
11. Hilbe, JM (2011). Negative Binomial Regression, 2nd edition (Cambridge University Press, New York).
12. Bartoń, K. MuMIn: Multi-model inference (R package version 1.0.0. http://CRAN.R-project.org/package=MuMIn, 2011).
SUPPLEMENTARY FIGURES
Fig. S1: An outline of the trapping schedule used in the perturbation experiment. Primary trapping (coloured arrows) involved trapping the grid(s) shown for 3 consecutive nights (Mon-Weds) within a week. Treatments were administered during primary, but not secondary trapping sessions. Sample sizes shown are numbers of primary and secondary captures in each month. Grids HW1, HW2 and HW3 are in Haddon Wood, while MW1 and MW2 are in Manor Wood.
Fig. S2: The highly aggregated distributions of (a) nematode eggs per gram faeces (EPG) and (b) Eimeria oocysts per gram faeces, prior to treatment.
Fig. S3: The effect of Ivermectin on the most prevalent nematode in the study populations, Heligmosomoides polygyrus, which closely resembles the effects of treatment on all nematodes (Fig. 1). (A) The effect of Ivermectin treatment within the previous three weeks on infection probability (Drug: c21=15.52, p<0.001, n=42) and (B) changes in H. polygyrus Infection probability over time since first capture with respect to treatment group. Arrows indicate the time-points of Ivermectin treatment for mice in the repeated treatment group (solid arrows), and the single treatment group (dashed arrows). The single treatment is grouped with the repeated group until 4 weeks, at which point this treatment diverges, as single treated mice do not receive a further dose of Ivermectin. In (B), at 5-7 weeks, all 5 captured mice from the single treatment group were H. polygyrus infected, hence there are no error bars on this data point. Means and S.E.M from raw data are plotted.
Fig. S4: Dynamics of nematode infection intensity over time since first capture with respect to treatment. Arrows indicate the time-points of Ivermectin treatment for mice in the repeated treatment group (solid arrows), and the single treatment group (dashed arrows). The single treatment is grouped with the repeated group until 4 weeks, at which point this treatment diverges (as single treated mice do not receive a further dose of Ivermectin). Nematode infection intensity is expressed as log-transformed eggs per gram of faeces, among infected mice only. Means and S.E.M from raw data are plotted.
Fig. S5: Dynamics of Eimeria spp. infection probability over time since first capture with respect to treatment. Arrows indicate the time-points of Ivermectin treatment for mice in the repeated treatment group (solid arrows), and the single treatment group (dashed arrows). The single treatment is grouped with repeated until 4 weeks, at which point this treatment diverges (as single treated mice do not receive a further dose of Ivermectin). Means and S.E.M from raw data are plotted.
Fig. S6: Temporal dynamics of two non-target blood-borne microparasites in response to parasite community perturbation using the anthelminthic drug Ivermectin. Neither the infection probability for (A) Bartonella spp. bacteria nor (B) Trypanosoma grosi significantly responded to Ivermectin treatment over the timescale assessed. Means and S.E.M from raw data are plotted.
Fig. S7: Temporal dynamics of species richness for (A) non-target gut parasites and (B) all parasites examined (gut and blood-borne) in response to perturbation using the anthelminthic drug Ivermectin. The single treatment is grouped with the repeated group until 4 weeks, at which point this treatment diverges (as single treatment mice do not receive a further dose of Ivermectin after this time). Means and S.E.M from raw data are plotted, and in (B) include only monthly captures as blood parasites were not monitored in secondary trapping sessions. All drug and treatment terms p>0.25 at 1-3 weeks after Ivermectin treatment (for gastrointestinal richness), and 1 and 2 months after first capture (for gastrointestinal and total parasite richness).
Fig. S8: Temporal dynamics of (A) tick and (B) flea prevalence following perturbation using the anthelminthic drug Ivermectin, showing no significant effect of treatment. Means and S.E.M from raw data are plotted. The single treatment is grouped with the repeated group until 4 weeks, at which point this treatment diverges (as single treatment mice do not receive a further dose of Ivermectin after this time).