Widespread occurrence of the amphibian chytrid fungus in Kenya

Authors


Correspondence
J. Kielgast, Department of Biology, Section for Evolution and Microbiology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark.
Email: jkielgast@snm.ku.dk

Abstract

Amphibians at the global scale are dramatically declining and the pathogenic fungus Batrachochytrium dendrobatidis (Bd) has been suggested to be an important driver in this biodiversity crisis. Increasing evidence points towards the global emergence of Bd being a panzootic caused by pathogen pollution. Africa has been suggested to be the origin of the pathogen but remains one of the least-studied areas. We have conducted the most comprehensive survey on the continent to date focusing on Kenya for investigating taxonomic and environmental components in the distribution of Bd in tropical Africa. Eleven sites along a 770 km transect from the coast up to the border of Uganda were surveyed. Using quantitative PCR, we screened 861 samples from 23 different species in nine genera. The pathogen was confirmed at all studied sites, with an overall prevalence of 31.5%. No dead or symptomatic specimens were found and no declines have been reported in the region so far. Both prevalence and parasite load ranged from the detection limit to some of the highest ever reported. The parasite load showed a significant taxonomic bias and a strong inverse correlation with temperature. Our findings suggest that Bd may be enzootic in the region. We recommend that further research should focus on comparative experimental studies of susceptibility to Bd in African species. Moreover, we stress the need for improved knowledge on the conservation status of the tropical African amphibian fauna to confirm the enzootic nature of widespread Bd infections.

Introduction

Amphibians are the most drastically declining vertebrates on our planet, with nearly one-third of the c. 6400 known species being threatened with extinction under the IUCN Red List (Stuart et al., 2008). This biodiversity crisis is developing globally at a rate only paralleled by mass extinction events in geological time (Mendelson III et al., 2006; Wake & Vredenburg, 2008). Epizootics of the disease chytridiomycosis, caused by the parasitic fungus Batrachochytrium dendrobatidis (Bd), have been proposed to play an important role as the proximate cause of the rapid decline of more than 200 amphibian species (Skerratt et al., 2007). This is supported by well-documented aetiology in focal studies of population decline in North America (e.g. Rachowicz et al., 2006), Central America (e.g. Lips et al., 2006), Europe (e.g. Bosch, Martínez-Solano & García-París, 2001) and Australia (e.g. Schloegel et al., 2006). Bd shows a remarkably low host specificity and has now been detected in more than 400 different anuran and salamander species on all continents on which amphibians occur (Fisher, 2008). However, molecular studies strongly indicate that Bd's current geographic distribution is the result of a recent panzootic spread (James et al., 2009). This is supported by empirical evidence of a wave-like emergence of disease in the Neotropics (Lips et al., 2008).

The large-scale dissemination of Bd has been suggested to be human mediated and data on the international trade of amphibian species underline the potential for pathogen pollution (e.g. Daszak et al., 2006; Fisher & Garner, 2007; Walker et al., 2008). The point of origin is therefore a central question towards an understanding of the epidemiology of this disease. It has been advocated, based on evidence from clawed frogs (genus Xenopus), that an ‘out of Africa’ scenario is plausible (Weldon et al., 2004). This was founded on the oldest known record of Bd, a temporally constant Bd prevalence in museum collections and disease resistance. Moreover, a dissemination pathway exists via worldwide trade in these frogs beginning as early as the 1930s – first used for a pregnancy assay and subsequently as a model animal in biological and medical teaching and research (Weldon et al., 2004; Weldon, De Villiers & Du Preez, 2007).

The ‘out of Africa’ hypothesis has been frequently referred to in the literature. However, research on Bd has so far suffered from a geographic bias, whereby Africa has largely been left unstudied. Owing to the lack of apparent barriers to Bd within the continent, we hypothesize that an ‘out of Africa’ scenario would render widespread occurrence of Bd in suitable habitats. Furthermore, if the host–pathogen system has co-evolved in the African herpetofauna, some intrinsic factors characterizing its interactions are likely to differ from the epizootic range, that is parameters such as virulence and host specificity. Here we examine the host–pathogen system in tropical Africa using landscape-diverse Kenya as a representative. For the first time, we investigate multiple localities and host taxa with quantitative methods and present the most comprehensive survey of Bd conducted in Africa so far.

Materials and methods

From 16 September to 22 October 2006, fieldwork was carried out at 11 localities along a 770 km transect from the Kenyan coast up to the border of Uganda covering semi-humid and humid habitats at an altitudinal range of 159–3100 m a.s.l. (Fig. 1, Appendix S1). The time of sampling targeted the beginning of the ‘short rains’ of the bimodal precipitation pattern. No information on the seasonality of Bd infection dynamics in Africa is available but sampling took place corresponding well with the recommendations given for sampling in Australia, for example, focusing on the cold season (Kriger & Hero, 2006; Skerratt et al., 2008). Approximately 1500 specimens were examined in situ during the survey. A minimum of 30 individual diagnostic samples for Bd per species and locality was targeted to enable 95% probability of detecting a positive assuming an underlying prevalence of 10% (following Cannon & Roe, 1982). Only post-metamorphic and adult anurans were sampled (as caecilians were not found and salamanders are absent). Recommended disinfection and containment procedures were followed to avoid transmission and dissemination of the pathogen (see Speare et al., 2004). Specimens were caught by hand or, in a few cases, by dip-nets and placed individually into plastic bags. Transmission of Bd between individuals was eliminated by collecting the specimens wearing a latex glove or a plastic bag on the hand. Diagnostic sampling was carried out using a diagnostic fine-tip dry swab (Medical Wire & Equipment, MW-100) by comprehensively swabbing each specimen's dorsum, ventrum, both lateral sides, the dorsal and ventral surface of hind limbs, toes and toe webbing. A fresh pair of latex gloves was used for every specimen to avoid contamination of samples. Voucher specimens to validate the taxonomic identification (following Channing & Howell, 2006) of all sampled species (Appendix S2) were deposited at National Museums of Kenya (NMK), Nairobi and Zoologisches Forschungsmuseum Alexander Koenig (ZFMK), Bonn. Swabs were stored as cool as possible in the field and thereafter at −20 °C until processing.

Figure 1.

 Study sites in Kenya sampled for Batrachochytrium dendrobatidis: 1, Saiwa Swamp National Park; 2, Mt. Elgon National Park; 3, Kakamega Forest National Reserve; 4, Thompson Falls; 5, Aberdares National Park (moorlands); 6, Aberdares National Park (Salient); 7, Tigoni Dam; 8, Nairobi (Karens); 9, Taita Hills (Mwundanyi); 10, Taita Hills (Mwatate); 11, Shimba Hills National Parc (see Appendix S1).

For Bd analysis, 861 samples from 23 different species in nine genera (Appendix S2) were analysed by quantitative PCR, following the protocol of Boyle et al. (2004), but running each sample in two replicates. Samples were identified as positive for Bd if a clear log-linear amplification was observed for both replicates and genomic equivalents (GE) quantified according to standards yielded above 0.5. If between-replicate standard deviation on the GE was higher than either of the quantified replicates, or only one replicate amplified above the detection threshold, the sample was re-run. If this also yielded equivocal results, the sample was identified as negative. The mean GE of positive samples was regarded as an index for parasite load.

Fisher's exact test was used for comparing Bd prevalence in the sampled genera. The parasite loads were compared by a rank-based generalized linear model (using SAS 9.1 Statistical package). Detected GE loads within a locality were percentile ranked by partitioning them into 100 groups in which the smallest received a value of 0 and the largest value received a value of 99. Thus, the influence of sample size and locality-specific parasite abundance was removed, enabling comparison of interspecific parasite load across localities in a generalized linear model. The GE loads present in the investigated genera were further compared by the distribution-free mood's median test using Quantitative Parasitology 3.1 (http://www.zoologia.hu/qp/qp.html).

In order to assess the relationships between climate conditions during the surveys and Bd prevalence among sites, we extracted information on current climate (minimum and maximum temperature and precipitation) and topography from the Worldclim database, version 1.4 (http://www.worldclim.org). This is a climate model based on the weather conditions recorded between 1950 and 2000 with a grid cell resolution of 30 arc sec (Hijmans et al., 2005) and was created by interpolation using a thin-plate smoothing spline of observed climate at weather stations with latitude, longitude and elevation as independent variables (Hutchinson, 1995; 2004). The relationships between locality-level prevalence and environmental factors were assessed with simple linear regressions calculated with Xlstat 2009 (http://www.addinsoft.com).

Results

All 11 sampled localities were Bd positive, with a prevalence from 4 to 71% and an overall mean of 31.5% (Table 1, Appendix S2). Parasite load ranged from the detection threshold up to more than one million GE. The highest parasite load coincided with the highest genus-specific prevalence at a stream habitat in the moorlands of Aberdares National Park (Fig. 1), where only a single species, Amietia wittei (Ranidae), was sampled. The distribution of Bd parasite load was heavily aggregated in a small subset of samples with extremely high GE loads as illustrated by the variance/mean ratio (Table 1). However, a high parasite load was generally unusual throughout the survey, as indicated by the median mirroring central tendency for GE load in the infected fraction of samples (Table 1). Again, the sampled A. wittei from Aberdares National Park displayed unusually high values with a median in the parasitized fraction of 3966 GE. Evaluating potential taxonomic bias in the distribution of Bd, we found a significant difference in the proportion of infected individuals across all sampled genera (exact two-sided, P<0.001). The parasite load similarly showed a significant effect of genus (n=861, d.f.=8, χ2=67.57, P<0.001). Hence, we can conclude that our results strongly indicate a taxonomical component to the abundance and intensity of infection at the time of sampling. There was no significant difference in the median parasite load across all genera containing more than a single infected individual (exact two-sided, P=0.077). Regression analysis on climate variables indicates a negative correlation between temperature and prevalence during the sampling period. This pattern strongly co-varies with altitudinal distribution and shows an increase in the frequency of Bd infection with altitude. At the same time, the general precipitation pattern was not found to correlate with prevalence (Fig. 2).

Table 1.   Summary statistics for the distribution of Batrachochytrium dendrobatidis (Bd) across localities (see Fig. 1, Appendix S1) and genera surveyed
 Specimens
sampled
Bd positivePrevalence (CI)aMax GEMean GEMedian GEVariance/mean
ratio
  • a

    95% confidence intervals on Bd prevalence in parentheses were constructed using Sterne's exact method (Reiczigel, 2003).

  • CI, confidence intervals; GE, genomic equivalents.

Locality
 Aberdares National Park (moorlands)31.0022.000.71 (0.53–0.84)1 003 737.0095 146.003966.00739 406.00
 Aberdares National Park (Salient)111.0043.000.39 (0.30–0.48)120 350.007144.0048.0073 771.00
 Kakamega Forest National Park72.0012.000.17 (0.10–0.27)52 232.006059.0096.0040 414.00
 Mt. Elgon National Park11.005.000.46 (0.20–0.74)14 610.002938.006.0014 512.00
 Nairobi (Karens)109.0027.000.25 (0.17–0.34)48 945.002543.0012.0039 526.00
 Saiwa Swamp National Park126.0062.000.49 (0.40–0.58)227 779.005145.0024.00176 607.00
 Shimba Hills National Park150.006.000.04 (0.02–0.09)9.006.006.007.00
 Taita Hills (Mwatate)21.004.000.19 (0.07–0.40)980.00296.00101.00828.00
 Taita Hills (Mwundanyi)116.0034.000.29 (0.21–0.38)10 395.00587.007.006185.00
 Thompson Falls71.0047.000.66 (0.54–0.76)20 336.001124.0065.009780.00
 Tigoni Dam50.009.000.18 (0.09–0.31)565.0083.0012.00443.00
Genus
 Afrixalus (Hyperoliidae)18.002.000.11 (0.02–0.33)4.004.004.003.00
 Amietia (Ranidae)81.0049.000.61 (0.49–0.71)1 003 737.0049 327.00133.00680 415.00
 Amietophrynus (Bufonidae)13.003.000.23 (0.07–0.52)5.005.005.004.00
 Hyperolius (Hyperoliidae)418.00180.000.43 (0.38–0.48)120 350.002579.0030.0056 624.00
 Kassina (Hyperoliidae)30.001.000.03 (0.00–0.18)2.002.002.002.00
 Leptopelis (Arthroleptidae)20.000.000.00 (0.00–0.17)0.000.000.000.00
 Phrynobatrachus (Phrynobatrachidae)27.001.000.04 (0.00–0.18)200.00200.00200.00200.00
 Ptychadena (Ptychadenidae)126.0031.000.25 (0.18–0.33)48 945.002216.0015.0037 812.00
 Xenopus (Pipidae)128.004.000.03 (0.01–0.08)18.007.004.0014.00
Figure 2.

 Simple linear regressions illustrating the relationships between the mean Bd prevalence at the studied sites and (a) minimum temperature (r2=0.825, P<0.001), (b) maximum temperature (r2=0.666, P=0.002), (c) precipitation (r2=0.002, P=0.904) and (d) altitude (r2=0.709, P<0.001); 95% confidence intervals are indicated as vertical bars. The grey lines represent the 95% confidence limits for the mean of the prediction of a given value (dotted line) and the 95% confidence limits on a single prediction for a given value (solid line).

Discussion

Using landscape-diverse Kenya as a representative, we have examined the Bd host–pathogen system in tropical Africa. Bd was detected at all investigated sites, indicating that it is ubiquitous in the sampled habitat types and may be widespread in tropical Africa. We furthermore found that the prevalence and intensity of Bd infection correlated with the taxonomic and climatic patterns in the sampled region.

Our results suggest that there is a taxonomic component to Bd susceptibility in the anuran communities studied. Susceptibility appeared to be highest in the genus Amietia (Ranidae), which consistently showed a high prevalence and parasite load. The genera Ptychadena (Ptychadenidae) and Hyperolius (Hyperoliidae) exhibited intermediate infection levels and members of the genus Xenopus (Pipidae) were seldom infected. Interestingly, the median parasite load of infected individuals was not taxonomically biased, indicating that the extreme ends of the intensity distribution constitute the difference.

During our survey, despite extensive collection efforts, no dead or moribund anurans were encountered, nor were any clinical symptoms of chytridiomycosis observed. Even the most heavily infected individuals carrying a parasite load of over one million GE appeared to be in a good body condition. The high prevalence of Bd and the large number of sub-clinically infected individuals may suggest that the pathogen is enzootic in and possibly native to the studied region. However, considering the cryptic disease progression characteristic for chytridiomycosis, it is not possible to exclude that it may be causing mortality in the studied populations undetected. Sub-clinical infections can build up to a threshold, with symptoms only occurring in the terminal phase of disease (see e.g. Carey et al., 2006; Voyles et al., 2007), and field observations equivalent to ours have even been made just before mass die-offs (Woodhams et al., 2007). Furthermore, an enzootic state of Bd may be a consequence of populations being in a post-decline phase and that declines were merely not observed as they happened. There are now numerous examples of enzootic Bd infection in amphibian populations globally (Retallick, McCallum & Speare, 2004; McDonald et al., 2005; Longcore et al., 2007; Brem & Lips, 2008; Woodhams et al., 2008; Padgett-Flohr & Hopkins, 2009), and recent compelling evidence suggests that Bd can have a marked impact even decades after becoming enzootic (Murray et al., 2009). Hence, the apparently enzootic state of Bd found in the present study should be regarded as a conservation concern and requires further investigation.

Existing accounts of Bd in African amphibians are equivocal regarding pathogenicity and so far little data are available. In tropical Africa, Bd has been reported to be present in Nigeria (Imasuen et al., 2009), The Democratic Republic of Congo (Greenbaum et al., 2008), Uganda (Goldberg, Readel & Lee, 2007) and Tanzania (Weldon et al., 2004), while a single study reports not detecting Bd in Cameroon (Doherty-Bone et al., 2008). In Tanzania, Bd was suggested to be linked to the decline and disappearance of the Kihansi spray toad Nectophrynoides asperginis, supported by finding two out of four dead individuals collected during the decline to be Bd positive (Lee et al., 2006). The species was categorized under ‘rapid enigmatic declines’ in the IUCN Global Amphibian Assessment 2004 (Stuart et al., 2004), which have been suggested to cover over Bd-driven extinction processes (Skerratt et al., 2007). However, the Kihansi spray toad was endemic to a single waterfall system (c. 2 ha.) that was radically changed by the construction of a power plant before declines (Krajick, 2006). The aetiology of this decline is therefore highly confounded by habitat change and the role of Bd is unclear. All other studies from tropical Africa have not been coupled with observations of mortality or decline. Goldberg et al. (2007) conducted a single site survey with methodology similar to ours and report congruent findings with 22% prevalence in 109 specimens and no mortality events noted in the region. The remaining reports are accounts of small-scale sampling and single Bd-positive individuals.

In the South African region, Bd has been found to be widespread but no associated population declines have been observed (Hopkins & Channing, 2003; Weldon, 2002; Weldon et al., 2004; Smith et al., 2007). However, Bd-associated mortality in frogs has been detected twice without presuming an aetiological link. The findings of the diagnostic characteristics of chytridiomycosis (Lane, Weldon & Bingham, 2003), and patterns of mortality and infection resembling a die-off (Hopkins & Channing, 2003), call for further investigations, but do not contradict an endemic presence of Bd in the region.

Information on the susceptibility of African amphibians to Bd under controlled conditions is limited to pipids and indicates susceptibility in the western African species Silurana tropicalis but resistance in Xenopus laevis (Reed et al., 2000; Parker et al., 2002; Fisher & Garner, 2007). This has not yet been verified by experimental infection challenges but is solely based on observations of mortality and aclinical infection in captive populations.

It has repeatedly been shown that the pathogenicity of chytridiomycosis is context dependent and particularly affected by environmental parameters, for example temperature and precipitation (Bosch et al., 2001; Daszak, Cunningham & Hyatt, 2003; Berger et al., 2004; Kriger & Hero, 2006; Kriger, Peregolou & Hero, 2007; Kriger, 2009; Longo, Burrowes & Joglar, in press). Furthermore, the environmental suitability for a species at a given site may affect the capacity of the host's immune system affecting the prevalence and intensity of Bd infections (Raffel et al., 2006; Fisher, 2007). Our results indicate that prevalence decreases with higher monthly minimum and maximum temperatures (Fig. 2), thus supporting previous findings (i.e. Berger et al., 2004; Piotrowski, Annis & Longcore, 2004; Kriger & Hero, 2006; Rödder, Veith & Lötters, 2008). In our study, precipitation patterns did not correlate with Bd prevalence (Fig. 2). The species sampled were a priori the most abundant; hence, it is understandable that species compositions varied among sampling sites. The correlation detected between temperature and prevalence may therefore either reflect varying degrees of climatic suitability of the study sites for Bd (e.g. coincident with the results shown by Piotrowski et al., 2004; Kriger et al., 2007; Rödder et al., 2008) or differences in the susceptibility of the sampled amphibian communities or both.

The question remains open as to whether the Bd panzootic has its origin in Africa. If this is the case, the observed patterns of high prevalence and apparently low virulence found in the present study can be explained in a co-evolutionary context, that is evolution of host resistance, avirulence or attenuation of the local strain of Bd (McCallum, 2005). There is accumulating evidence that Bd exists in various strains exhibiting markedly different levels of virulence (Berger et al., 2005; Retallick & Miera, 2007; Fisher et al., 2009) although recent evidence points towards no correspondence between the genetic lineage of strain and virulence (James et al., 2009). Comparative clinical studies investigating host–pathogen dynamics by experimental exposure of African and non-African amphibian hosts and Bd isolates may be a way forward in testing the ‘out-of-Africa’ hypothesis. Here, the conditional nature of pathogenicity should be evaluated, by including a range of climatic conditions, to determine the relative importance of the intrinsic taxonomic components of susceptibility and climatic factors. Moreover, the ‘out-of-Africa’ hypothesis needs to be addressed by analysing the phylogenetic patterns and global genetic diversity in Bd including a broader coverage of Bd isolates from Africa than are currently available (James et al., 2009).

It should be underlined that disease surveys as such have a limited capability of detecting declines and proving pathological causality. This requires clinical experimental studies of host–pathogen dynamics (including fulfilment of Koch's postulates) and long-term disease surveillance coupled with studies of host population dynamics. There is an urgent need to improve our knowledge of the conservation status of tropical Africa's amphibian fauna. A crucial step towards this is a qualified assessment of the risk imposed by widespread Bd infections found in the present study. As a priority, we urge the initiation of population monitoring focusing on susceptible amphibian communities in the highlands of tropical Africa.

Acknowledgements

We are grateful to Trent Garner and Andrew Cunningham for making available the facilities and expertise of the Institute of Zoology, London. This extraordinary helpfulness was completely essential in realizing the present study. We thank The National Museums of Kenya and staff for fruitful cooperation. Susanne Schick, Felix Müller, George Kennedy and Beryl Bwong were of invaluable assistance during fieldwork. The first author would like to express deepest gratitude to Mads Frost Bertelsen and Copenhagen Zoo for decisively contributing to the study and Mogens Andersen and Peter Gravlund for practical support. Finally, we would like to extend many thanks to two anonymous reviewers whose helpful comments greatly improved the paper. The project was financed through the Copenhagen Zoo/Center for Zoo and Wild Animal Health, The Danish WWF and Aase og Ejnar Danielsens Fond, BIOLOG-BIOTA from the Federal Ministry of Education and Research (BMB+F, Germany), H.R. Frederiksen og Grete Siim Frederiksen Fond, Fonden Kjebi. D.R. is grateful to the ‘Graduiertenförderung des Landes Nordrhein-Westfalen’ for financial support. The Kenya Wildlife Service, KWS, kindly granted permission for this work (No. KWS/RP/5001).

Conflicts of interest: None.

Ancillary