Previous Page Table of Contents Next Page


Parasite variations and polymorphism


Parasite polymorphisms: Modelling variations in African trypanosomes
Polymorphisms in Theileria parva
Application of modelling to trypanosomiasis chemotherapy
Models for investigating genetic exchange in protozoan populations
Modelling anthelmintic resistance
Session discussion


Parasite polymorphisms: Modelling variations in African trypanosomes

P. Majiwa

International Laboratory for Research on Animal Diseases
P.O. Box 30709
Nairobi, Kenya

Parasitic protozoa are organisms with many interesting and variable phenotypes. The majority are transmitted to their vertebrate hosts by vectors in which they undergo complex cyclical development before acquiring the capacity to re-infect the hosts.

Many studies have been conducted on inherent characteristics of these parasites that enable them to display such variation; in this respect, the salivarian trypanosomes have been one of the most extensively studied. Trypanosomes change their antigenic profiles in the process of antigenic variation, which apparently occurs independent of the host but is qualitatively, and possibly quantitatively, modulated by the host immune responses. Depending on the antigenic repertoire of the parasite (as determined by its genotype), a trypanosome in the host blood stream will display a different number of distinct antigenic types in a semi-defined order. Attempts have been made to estimate empirically and by molecular hybridization using cloned VSG genes the number of different variants which can descend from a single trypanosome. Such analyses can be performed on a carefully selected sample of trypanosome populations. It would be informative to understand by simulation model the extent to which different hosts and host factors influence the dynamics of antigenic variation in a trypanosome of a particular genotype.

Trypanosomes display variability in the severities of the disease they cause in their respective vertebrate hosts; this is particularly true of trypanosomes collectively designated Trypanosoma congolense. The reservoir hosts and some domestic livestock appear not to suffer deleterious consequences upon infection by trypanosomes; on the other hand, exotic breeds of domestic livestock suffer more serious illness often resulting in death unless treated upon infection by the parasites. Evidence to date indicates that both the parasite and host genetic background as well as immune status contribute significantly to the clinical course and outcome of the disease. The extent of genetic polymorphism and proximity among trypanosomes can be estimated by mathematical analysis of randomly amplified polymorphic DNA (RAPD) markers and restriction enzyme fragment length polymorphisms (RFLPs). Similar analyses can be performed on the vertebrate hosts as well. With sufficient experimental data on both the parasite and the hosts, it should be possible to model the course of the disease in a given host infected with a single or multiple parasites of particular genotypes.

It has been established that different species of the African trypanosomes can infect a single tsetse vector and thus a single host. In such a situation, it is envisaged that genetic recombination could occur leading to new or novel genotypes which may display entirely different clinical effects in different hosts. Given that different parasite genotypes can be identified accurately using DNA-based markers, it should be possible to have models which can predict recombination frequencies within parasite species, and the consequences of such events on-parasite genetic repertoires (serodemes) and population types. Such a model could also provide information on how different parasite species interact with each other and with the different species of the vector and host in the presence of host immune responses, treatment of the host with different anti-trypanosomal drugs, different vector control strategies and a generally changing environment. It may also be possible to simulate responses of different parasite types to drugs in common use, for treatment or management of the diseases they cause. Indeed, some of the most useful contributions of such models could include predictions of dynamics of parasite phenotypes such as drug resistance, vector transmissibility, virulence to host and other aspects of genotypic variation in parasite populations.

Polymorphisms in Theileria parva

S.P. Morzaria and R. Bishop

International Laboratory for Research on Animal Diseases
P.O. Box 30709
Nairobi, Kenya

Theileria parva, an apicomplexan parasite of cattle and buffalo, causes a severe disease syndrome in cattle (variously known as East Coast fever, January disease and Corridor disease) and is distributed in large parts of eastern, central and southern Africa. ILRAD's major goal is to develop improved methods of controlling theileriosis, employing novel subunit vaccines. Better understanding of the epidemiology and biology of the parasite will contribute to this goal.

Parasite stocks isolated from the field are heterogeneous and exhibit extensive genotypic and phenotypic diversity. This genotypic diversity has been demonstrated by analysis of genomic DNAs from different stocks of parasites, using a major parasite-specific repetitive sequence. Virtually all isolates so far characterized show restriction fragment length polymorphism. There is also size polymorphism of SfiI fragments in different stocks. Further analysis has revealed that this polymorphism is mainly confined to the telomere-bearing SfiI fragments, suggesting that telomeric or subtelomeric regions may be active sites for the generation of sequence polymorphism within the parasite genome.

Phenotypic differences exist between parasites isolated from buffalo and cattle. The buffalo-derived T. parva parasites are usually more virulent in cattle and undergo limited merogony compared to cattle-derived parasites. A stock of T. parva from Zimbabwe has been isolated that causes mild disease in cattle and yet provides a broad protection against different stocks of parasites.

Antigenic diversity has also been demonstrated using a panel of monoclonal antibodies recognizing epitopes on the surface of T. parva schizonts. The gene encoding the polymorphic schizont antigen exhibits a series of complex polymorphic repeats in its central region. Phenotypic manifestation of the antigenic diversity is demonstrated in in vivo cross-immunity studies. Cattle immunized with one strain of parasite may not be protected if challenged with a different immunological strain. However, the antigens involved in inducing strain-specific immunity have not been identified.

In contrast, characterization of the gene encoding a surface sporozoite antigen p67 of T. parva, which induces protective immunity in cattle, has shown that it is highly conserved among all cattle-derived T. parva and is different from the gene in parasites isolated from buffalo. Thus there is a potential of development of a diagnostic marker for the identification of buffalo-derived T. parva.

Molecular characterization of the T. parva genome reveals four chromosomes. Sexual reproduction has been demonstrated to occur in laboratory experiments indicating a potential for generating novel genotypes during meiosis by independent assortment of chromosomes. As in Plasmodium, a potential for mutation during asexual division in the bovine host may also exist although this has not yet been observed. A complete restriction map of the genome has been constructed using a 'linking clone strategy'. This has generated numerous chromosome specific SfiI linking clones as markers. Additionally, several antigen and house-keeping genes and random cDNAs have been localized in the genome. Thus a large body of data exists on the genome of T. parva and numerous DNA and serological reagents are available for typing field populations of the parasite.

Knowledge of the characteristics of parasite populations in the field is a pre-requisite for developing theileriosis control strategies using novel vaccines. However before field studies are conducted it is necessary to consider the parasite traits which need to be identified. These include strain-specific immunity, virulence and infectivity. Markers for these traits are yet to be developed. If appropriate markers become available then, some of the important questions concerning field populations of parasites are:

· what is the prevalence of different parasite genotypes in nature (especially with regard to different immunological strains, virulence and infectivity to the vertebrate and invertebrate hosts)?

· what is the effect of an obligatory sexual cycle in T. parva on the genetic structure of parasite populations in the field?

· what is the effect of introducing new strains on the creation of novel genotypes bearing in mind that the sexual cycle is obligatory in the life-cycle of the parasite?

· what are the genotypic differences between parasites present in carrier animals and those in animals undergoing acute infections?

In the context of this workshop it is important to consider whether polymorphisms in genes of important biological traits, allelic frequencies of these genes, sexual recombination and population dynamics of these genotypes can be modelled. If so, what relevant data are required to construct these models and, finally, can these models be exploited to predict the effects of various control pressures such as vaccination, chemotherapy and tick control on the evolution of parasite populations in the field?

Application of modelling to trypanosomiasis chemotherapy

A.S. Peregrine

International Laboratory for Research on Animal Diseases
P.O. Box 30709
Nairobi, Kenya

Maintenance of domestic livestock in trypanosomiasis endemic areas is currently carried out in three different manners: controlling the vector tsetse fly, rearing trypanotolerant livestock and administering trypanocidal compounds to domestic livestock.

Chemotherapy and chemoprophylaxis of trypanosomiasis in cattle, sheep and goats is dependent upon the salts of three chemical compounds; diminazene (an aromatic diamidine), homidium (a phenanthridine) and isometamidium (a phenanthridine aromatic amidine). While isometamidium is primarily used as a prophylactic agent and diminazene is used only as a therapeutic agent, homidium is used both as a therapeutic and a prophylactic agent. All three compounds have been used in the field for over 30 years and the occurrence of resistance toeach of the compounds (especially isometamidium and homidium) has been associated with their usage at sites across Africa. However, the mechanism(s) by which resistance arise(s) in the field are not clearly understood. Laboratory studies indicate that subcurative treatment dosages could be a precursor to the development of resistance to isometamidium and homidium. Furthermore, reciprocal cross-resistance is associated with the two compounds. However, the same does not appear to be true for diminazene since resistance to this compound is extremely difficult to induce when one treats infected animals with subcurative dosages. Furthermore, development of diminazene-resistance is not associated with cross-resistance to either isometamidium or homidium. Thus, diminazene is often used as part of a 'sanative pair', in conjunction with either isometamidium or homidium, to prevent the development of drug resistance.

Since the cost of developing new trypanocides is now prohibitively expensive, it is unlikely that new trypanocides will be forthcoming during the next decade. It is therefore important to maintain the efficacy of the existent trypanocides. To this end, the development of field-usable assays that will rapidly quantify the drug-resistance phenotypes of trypanosome populations in large numbers of cattle is required. Such quantitative information, when used in conjunction with data concerning the pharmacokinetics of the individual trypanocides in domestic livestock, should enable field workers to decide whether recommended dosage regimens for any of the three trypanocides will control the disease, whether de novo therapeutic regimens with any of the three trypanocides would be efficacious, whether alternative measures such as tsetse fly control are indicated, or whether a multifactorial integrated control strategy is required. It seems feasible that models may be valuable in integrating these variables and presenting them in a decision-making format. Ideally, such models should consider the long-term sustainability of the chosen control strategy. However, the decision-making process is partly dependent on assays that will rapidly quantify the drug-resistance phenotype of large numbers of trypanosome field isolates. Since such assays do not currently exist, their development is one of the goals of ILRAD's trypanosomiasis chemotherapy project. Finally, the decision process is also dependent on a comprehensive understanding of the epidemiology of drug-resistant trypanosomes in the field. Because there is very little information on this subject, it is important that experimental protocols are designed to determine the factors responsible for emergence, maintenance and disappearance of drug resistance in the field.

Models for investigating genetic exchange in protozoan populations

C. Dye

Department of Medical Parasitology
London School of Hygiene and Tropical Medicine
Keppel Street, London WC1E 7HT, UK

Abstract

This first part this paper deals with the mode of reproduction of Leishmania, and its epidemiological consequences. The second part is about the evolution and measurement of virulence or pathogenicity.

The recent debate about whether parasitic protozoa are generally clonal is discussed. The arguments and the evidence are summarized as follows.

· A critique of population genetic methods for assessing the mode of reproduction of parasitic protozoa, mainly those proposed by M. Tibayrenc. The key question for protozoa has been framed in terms of the relative frequency of automixis and amphimixis. There are two main problems: (i) arguments have often been based on weak tests, particularly those which amount to the search for linkage disequilibrium, (ii) few data are available with which to carry out any test properly; sample sizes are very limited and samples from different countries and continents have been lumped together when it is inappropriate to do so.

· One practical consequence of clonality (high coefficient of inbreeding) in Plasmodium populations: the evolution of resistance to more than one drug. I use a stochastic population model for Plasmodium falciparum to confirm and extend the earlier findings that less inbreeding tends to slow the rate of evolution of resistance to a drug mixture.

· The available, equivocal evidence for Leishmania, which points towards occasional genetic exchange.

A second reason as to why we need to know about the mode of reproduction of Leishmania is identified and discussed: it is that pathogenicity may be associated with certain genotypes, as has been suggested for both American and African trypanosomes. Among samples of L. infantum, the analysis of isoenzymes and of kDNA endonuclease fragment patterns separates strains responsible for cutaneous and visceral disease. In the southern republics of the former USSR, human cutaneous lesions apparently occur following infection with one member of the L. major group, but not another. In the Peruvian Andes, clones and strains of L. peruviana have shown reproducible difference in virulence in vivo and in vitro. I describe a model of the L. major system which highlights the question of whether we are seeing a balanced polymorphism, or simply different but sympatric transmission systems.

The final problem is concerned with measurement of the preponderance of virulent parasites in a Leishmania population. A compartmental model of cutaneous leishmaniasis demonstrates the difficulty of using a simple index such as the ratio of person scar positive:skin test positive.

Introduction

Plasmodia reproduce sexually, whilst trypanosomatids reproduce asexually. For Plasmodium, this long-established view has not been seriously challenged. For Trypanosoma, however, at least T. brucei, recent laboratory crossing experiments have shown repeatedly that genetic exchange between parasites can occur in tsetse flies, although it is not obligatory. Tait and Turner (1990) summarized the results of six crosses in which recombinants were produced on 14-45% of occasions.

These experiments have not yet been successfully repeated with any Leishmania species, or with South American T. cruzi. The best laboratory evidence that genetic exchange might occur in Leishmania is only partial evidence: G. Lanotte and J.-A. Rioux (personal communication) have observed and recorded, by videomicroscopy, promastigote fusion leading to the production of a synkaryon. It has yet to be demonstrated that fusion leads to the production of viable recombinant progeny.

For Leishmania, there is indirect evidence for hybridization between parasite species in field samples, though putative hybrids could be mutants or common ancestors. Isoenzyme analysis and molecular karyotyping have shown that an apparent hybrid of L. major and L. arabica, isolated in an area where both parasites have the same vector and reservoir host, does indeed have characteristics of both species (Kelly et al., 1991).

All this is qualitative analysis. However convincingly it demonstrates that genetic exchange can occur, we should like to assess its frequency of occurrence in natural populations. The first part of this paper explores some of the quantitative methods which have been used to investigate this question for Trypanosoma and Leishmania. The answer has implications for the rate of evolution of drug resistance, the distribution of virulent or pathogenic genotypes in the population and antigenic variation. The second part of the paper illustrates the first in this list by making use of a population genetic model of the malaria parasite R. falciparum. The choice of an obligately sexual parasite in this context underlines the point that discussions about genetic exchange are as much about inbreeding (or, conversely, outcrossing) as they are about sexuality.

Quantitative analysis of field populations

Single-locus analyses for Trypanosoma and Leishmania

Genetic exchange between parasites will tend to bring alleles within and between loci into association equilibrium. A lack of genetic exchange may be due to asexual reproduction, or, in a sexual population, to inbreeding. For segregation of alleles at any one locus, we look for Hardy-Weinberg equilibrium, reached in a panmictic population after just one generation of random mating. The three possible genotypes at a diallelic locus are expected to be seen in the proportions p2:2pq:q2, where p and q are the respective frequencies of two alleles, say A and a. Departures from Hardy-Weinberg equilibrium due to non-random segregation lead to a deficit of heterozygotes, and this deficit can be used to estimate the coefficient of inbreeding, F = 1 - PAa/2pq, where PAa is the observed frequency of heterozygotes.

Tests for departures from the Hardy-Weinberg ratio have their difficulties. One problem of working with parasites in vectors is small sample size. Infection rates in mosquito-like vectors, including tsetse flies and phlebotomine sandflies, are often very low - under 5% even in an area classed as highly endemic on the basis of infection in the vertebrate host population. Smaller samples are more likely to show genotypes at frequencies which are statistically indistinguishable from the Hardy-Weinberg ratio. In other words, there is a greater probability, b, of making a Type-II error, which accepts the Hardy-Weinberg null hypothesis when it should be rejected. Really, b £ 0.05 is needed before conformation with Hardy-Weinberg becomes convincing.

How many samples are required to avoid making a Type-II error? In the case of T. brucei some answers have been provided in a careful appraisal by Cibulskis (1988). Commenting on Tait's (1980) analysis of genotype frequencies for a diallelic locus and a sample of 17, Cibulskis calculated as shown in Figure 1. With this sample size and three genotypes, 120 genotype frequency distributions are possible. Every distribution is equally likely so is just the proportion of distributions that yield non-significant c 2 values. In this case it is big at 72/120 = 0.6.

A much larger sample of 220 T. brucei stocks was collected by Gibson and Wellde (1985) from the Lambwe Valley in Western Kenya. The isolates from flies were particularly striking with b in the range 0.07-0.205, depending on the precise method of analysis, and this range was lower than found in human isolates. These b 's associated with the tsetse fly isolates point to conformation with Hardy-Weinberg, but are not low enough to draw firm conclusions.

Figure 1. The distributions of genotype frequencies for a locus with 2 alleles and a sample of 17 isolates. Solid squares indicate those genotype frequencies which are significantly different from the Hardy-Weinberg ratio by c 2 test (p < 0.05). Open squares indicate genotype frequencies which are not significantly different. From Cibulskis (1988).

However, they are interesting in the context of a second general problem associated with testing for departures from Hardy-Weinberg equilibrium. It is that disequilibrium can arise from processes other than asexual reproduction and non-random mating. In particular, selection may have occurred between the time of 'zygote' formation and the time of sampling. So the ideal analysis will look at the genotypes of the zygotes themselves. For Plasmodium, these are found in Anopheles mosquitoes and it is now possible actually to genotype the zygotes by making use of the polymerase chain reaction, PCR (Ranford-Cartwright et al., 1991). For Trypanosoma too, the fusion products of gametes are found in the vectors, so samples taken from tsetse flies should be less biased by postzygotic selection.

Do samples of Leishmania parasites conform with the Hardy-Weinberg ratio? Leishmania are probably diploid (Bastien et al., 1992) and, although no formal segregation analysis has yet been carried out on a large sample from a single population, Tibayrenc et al. (1990) have persuasively argued that some segregation genotypes which ought to occur have simply never been found in field samples. Among some Old World Leishmania, no heterozygotes have been found at all.

Joint-locus analyses for Trypanosoma and Leishmania

Recombination between loci leads to linkage equilibrium (somewhat misleadingly named since loci may be on different chromosomes in the same genome, i.e. unlinked). Given the pairs of alleles at two loci, A/a and B/b, the overall departure from the expected joint-locus frequencies, linkage disequilibrium, can be calculated from D = PabPAB - PaBPAb, where the P's denote the observed frequencies of the four possible haplotypes. In making calculations (as for Plasmodium below), dependence of D on allele frequency is conveniently removed by expressing it as a fraction of the maximum disequilibrium possible, D'. Furthermore, since D' can be either positive or negative, we use the modulus of its value, D' (in other words, always make the sign positive), in measuring the average departure from equilibrium. Otherwise the average of all the positives and negatives is zero. Extreme linkage disequilibrium is manifested in the widespread occurrence of identical genotypes and, conversely, the absence of certain recombinant genotypes (Tibayrenc et al., 1990).

When there are numerous genotypes in relatively few samples, the expected genotype frequencies are too low to test against a c 2 distribution. We can ask instead how many joint-locus combinations are expected in a sample of given size. In T. brucei collected from tsetse flies, Tait (1980) found 51 out of the 90 genotypes possible with five diallelic loci. The probability of obtaining different numbers of joint-locus combinations under the hypothesis of random assortment of genotypes is shown in Figure 2 (Cibulskis 1988). The probability of obtaining as few as 51 is very small, about 1 in 5000. Essentially the same results were obtained with a second set of data for fly infections (Gibson and Wellde 1985). They were also obtained for human infections in the same area (Tait et al., 1985) and in analyses by Cibulskis (1988). Tibayrenc and colleagues (1986) have also applied the same sort of argument to populations of T. cruzi, in which linkage disequilibrium is extreme.

By contrast, there is little sign of linkage disequilibrium in the population of Leishmania infantum studied by Blaineau et al. (1992). They looked at the distribution of eight, four and three variants of chromosomes I, II and V in 22 L. infantum isolates taken from a small area on the Franco-Spanish border. Again the sample size is small, and refers to isolates rather than clones, but the chromosome size variants are roughly in linkage equilibrium. For example, using the results shown in Figure 3, we can calculate the expected number of triplets which should be found one, two and three times. These expected numbers are 13.51, 2.96 and 0.65, which compare with the observed 16, 3 and 0. But the similarity could easily be due to chance. And there are other explanations: for example, the diversity of genotypes could be explained by a high rate of mutation (Blaineau et al., 1992). As for trypanosomes, we need both laboratory experiments and further careful field studies with large sample sizes to resolve the issue.

Figure 2. The expected number of joint-locus combinations in a sample of T. brucei from tsetse flies, compared with the observed number of combinations (51). The distribution of numbers expected was calculated assuming random reassortment of genotypes. From Cibulskis (1988).

Cladistic Analysis of Genotype Diversity

Cibulskis (1988) also used cladistic analysis to determine whether genetic exchange is likely to be important in T. brucei populations. In a population of asexual organisms, all new genotypes arise by mutation, and it is possible to calculate the minimum number of mutational steps, M, required to generate all observed genotypes in a sample. This shortest mutational pathway is called a Wagner network. For sexual organisms, mutation is only required to generate new alleles. With all the necessary alleles, all genotypes can arise by recombination. The shortest mutational pathway for sexual organisms will have length S (< M)

The difference R = M - S can be summarized as 'recurrent mutations'. Recurrent mutations do not give rise to new alleles, but to new combinations of alleles, and therefore serve the same function as recombination. In fact, since a recurrence of any mutation is thought to be unlikely, most of R could be ascribed to recombination. Although there are caveats (Cibulskis, 1988), large R suggests that some of the diversity of genotypes in a field sample has arisen by genetic exchange.

Figure 3. The distribution of chromosome size variants I, II and V in a sample of 22 strains of L. infantum. Modified from Blaineau et al. (1992).

In one sample of T. brucei from the Lambwe Valley, isoenzyme analysis identified 28 genotypes using five loci, giving R ³ 2 22. Cibulskis (1988) points out that this is almost three times greater than observed for the parthenogenetic weevil Polyhydrosus mollis, for which all variation is attributed to mutation. It is also comparable to R for one parthenogenetic population of the cladoceran Daphnia pulex, where sexual reproduction in recent history is thought to be responsible for some of the genotypic diversity. In sum, observed variation in the Lambwe Valley population of T. brucei cannot easily be explained by mutation alone.

Rate of genetic exchange and the evolution of drug resistance: The example of Plasmodium.

Given two novel antimalarial drugs, are they best used as a mixture or in sequence? The problem was first investigated quantitatively by Curtis and Otoo (1986), and the answer depends on the frequency of genetic exchange in the population in question.

Curtis and Otoo used a simple deterministic model employing the usual beanbag genetics (Haldane, 1964). In every generation of sexual reproduction, gametes were allowed to meet at random. They assumed either that recombination never occurred, or that recombination always occurred. With no recombination, there was no difference between the two strategies. With recombination, mixtures effectively delayed resistance, particularly when genes for resistance to both drugs were initially rare, and when drugs were available to a small fraction of people.

In fact, parasites probably do not live in randomly mating populations, and it is worth asking whether the Curtis and Otoo result holds when they are assumed not to do so. A significant feature of vector-borne parasites is that they live as semi-isolated sub-populations, confined to their vertebrate and invertebrate hosts. How great the isolation is will depend on the frequency with which a mosquito taking a bloodmeal acquires parasites of different genotypes, which depends in turn on the frequency with which vertebrates are superinfected, and the duration of the infection arising from each inoculation. It also depends on how many zygotes (oocysts) a mosquito can support. As isolation becomes greater, linkage disequilibrium is generated by random genetic drift between the subpopulations, and departures from Hardy-Weinberg equilibrium arise because inbreeding leads to a deficiency of heterozygotes.

To represent this patchiness, we need a model which keeps track of the Plasmodium population in each individual human and mosquito host. To capture the effect of genetic drift, the model needs to be stochastic. Figure 4 is a summary flow diagram of such a model; more details are given in Dye (1991).

We first compare the sensitivity of the two principal measures of the frequency of genetic exchange. In Figure 5, from left to right, linkage disequilibrium and the coefficient of inbreeding change as parasite subpopulations become increasingly isolated. In case A, mosquitoes are unrealistically permitted to acquire parasites more or less at random from the entire parasite population. Each infected mosquito is assumed to support just one oocyst, and each of the two necessary gametes (called gametocytes in the human host) is selected first by choosing an infectious person at random, and then by randomly selecting one genotype from that person's subpopulation. In case B, pairs of mating gametes come from the same randomly chosen infectious host, and the frequency distribution of oocysts in mosquitoes is overdispersed with mode 1, precisely as found by Collins et al. (1984) in The Gambia. Case C is a hybrid of A and B: gametes come from the same infectious person, and each mosquito supports just one oocyst.

Figure 4. Flow chart summarizing a simple, stochastic model of P. falciparum.

The results in Figure 5 need to be interpreted comparatively. A model of this kind cannot say how much linkage disequilibrium and inbreeding we expect to see in natural populations. A model to do this would have to take account of both the immune response to different genotypes, and of the mutation rate. The present model deals with neither, so conclusions about the relative magnitude of F and D will be much more robust.

Recall that when mating is random, Hardy-Weinberg equilibrium is reached in just one generation. By contrast, linkage disequilibrium is only halved after each generation of random mating (Maynard Smith, 1989). So, under the simplest assumptions, we may expect a joint-locus analysis to be more revealing than a single-locus analysis. Figure 5 actually suggests that, in this structured population, F is a more sensitive measure of non-random mating than D. The medians are higher and variation around them smaller.

Now imagine resistance to two drugs DA and DB to be conferred by two genes a and b, which are initially rare and at unlinked loci. Individuals taking drug DA are susceptible to parasites of genotype a, but parasites of genotype a are killed before they can be transmitted. Likewise, only parasites of genotype b can survive treatment with DB. Those taking the mixture DA + DB are susceptible only to parasites of genotype ate. Alongside F and D in Figure 5, is plotted t50, the median time for genotype a (and hence b, approximately) to rise to a frequency of 0.5 from an initially low value. More outcrossing effectively breaks up linkage disequilibrium and, in consequence, significantly delays the build-up of resistance to a mixture. Qualitatively, this is the same result as obtained by Curtis and Otoo (1986). It is just one example of the importance of knowing how much genetic exchange (inbreeding or outcrossing) effectively occurs in parasite populations.

Conclusions

Three methods have been used to investigate genetic exchange using samples collected from the field: (1) single-locus analysis (looking for Hardy-Weinberg equilibrium); (2) joint-locus analysis (looking for linkage equilibrium); and (3) cladistic analysis of genotype diversity.

For T. brucei, (3) suggests that the diversity of genotypes in field samples cannot easily be explained without postulating that genetic exchange occurs at least some of the time, but (2) indicates that recombination is nowhere near frequent enough to break down all linkage disequilibrium. These results are consistent with those from laboratory experiments, which have demonstrated directly that genetic exchange does occur, but not obligately. The results of (1) are so far inconclusive because of small sample sizes. By contrast, natural populations of T. cruzi show extreme linkage disequilibrium, indicating that genetic exchange occurs very infrequently, if at all.

Method (3) has not been applied to Leishmania. Applying (2) to a single population of L. infantum shows conformity with linkage equilibrium, but again this is suspect with small samples. Careful segregation analyses (1) have not been carried out on large samples, but genotypes (particularly heterozygotes) missing from some samples suggest large departures from Hardy-Weinberg equilibrium. In sum, genetic exchange may well be more frequent among T. brucei than among T. cruzi or Leishmania, but it is too early to say so definitively.

Figure 5. Linkage disequilibrium (D', squares), coefficient of inbreeding (F, diamonds), and the time taken for the frequency of resistance genes to reach 0.5 (t50, circles), under different regimes of mating (see text). All points are medians, with 95% c.i.

The frequency of genetic exchange affects, among other things, the rate of evolution of resistance to drug mixtures. In general, resistance to a mixture is expected to arise more quickly when effective recombination occurs less frequently, that is, when there is more inbreeding or infrequent sexual reproduction. All the quantitative methods of analysis described in this paper suggest that future field work should concentrate on collecting large samples from single populations. Only with more data of this kind will we be able to accurately assess the frequency and consequences of genetic exchange in natural populations of parasitic protozoa.

References

BASTIEN, P., BLAINEAU, C. and PAGES, M. 1992. Leishmania: sex, lies and karyotype. Parasitology Today 5: 174-177.

BLAINEAU, C., BASTIEN, P. and PAGES, M. 1992. Multiple forms of chromosome I, II and V in a restricted population of Leishmania infantum contrasting with monomorphism in individual strains suggest haploidy or automixy. Molecular and Biochemical Parasitology 50: 197-204.

CIBULSKIS, R.E. 1988. Origins and organization of genetic diversity in natural populations of Trypanosoma brucei. Parasitology 96: 303-322.

COLLINS, F.H., ZAVALA, F, GRAVES, P.M., COCHRANE, A.H., GWADZ, R.W., AKOH, J. and NUSSENZWEIG, R.S. 1984. First filed trial of an immunoradiometric assay for the detection of malaria sporozoites in mosquitoes. American Journal of Tropical Medicine and Hygiene 33: 538-543.

CURTIS, C.F. and OTOO, L.F. 1986. A simple model for the build-up of resistance to mixtures of anti-malarial drugs. Transactions of the Royal Society of Tropical Medicine and Hygiene 80: 889-892.

DYE, C. 1991. Population genetics of non-clonal, non-randomly mating malaria parasites. Parasitology Today 7: 236-240.

GIBSON, W.C. and WELLDE, B.T. 1985. Characterization of Trypanozoon stocks from the South Nyanza sleeping sickness focus in Western Kenya. Transactions of the Royal Society of Tropical Medicine and Hygiene 79: 671-676.

HALDANE, J.B.S. 1964. A defense of beanbag genetics. Perspectives in Biology and Medicine 7: 343-359.

KELLY, J.M., LAW, J.M., CHAPMAN, C.J., Van EYS, G.J.J.M. and EVANS, D.A. 1991 Evidence of genetic recombination in Leishmania. Molecular and Biochemical Parasitology 46: 253-264.

MAYNARD SMITH, J. 1989. Evolutionary Genetics. Oxford: Oxford University Press, 325 pp.

RANFORD-CARTWRIGHT, L., BALFE, P., CARTER, R. and WALLIKER, D. 1991. Genetic hybrids of Plasmodium falciparum identified by amplification of genomic DNA from single oocysts. Molecular and Biochemical Parasitology 49: 239-244.

TAIT, A. 1980. Evidence for diploidy and mating in trypanosomes. Nature 287: 536-538.

TAIT, A. and TURNER, C.M.R. 1990. Genetic exchange in Trypanosoma brucei. Parasitology Today 6: 70-75.

TAIT, A., BARRY, J.D., WINK, R., SANDERSON, A. and CROWE, J.S. 1985. Enzyme variation in T. brucei ssp. II. Evidence for T. b. rhodesiense being a set of variants of T. b. brucei. Parasitology 90: 89-100.

TIBAYRENC, M., KJELLBERG, F. and AYALA, F.J. 1990. A clonal theory of parasitic protozoa: the population structures of Entamoeba, Giardia, Naegleria, Plasmodium, Trichomonas and Trypanosoma and their medical and taxonomical consequences. Proceedings of the National Academy of Sciences 87: 2414-2418.

TIBAYRENC, M., WARD, P., MOYA, A. and AYALA, F.J. 1986. Natural populations of Trypanosoma cruzi, the agent of Chagas' disease, have a complex multiclonal structure. Proceedings of the National Academy of Sciences 83: 115-119.

Modelling anthelmintic resistance

I.A. Barger*, E.H. Barnes+ and R. J. Dobson+

* CSIRO Division of Animal Health
Pastoral Research Laboratory
Armidale, New South Wales 2350, Australia

+ CSIRO McMaster Laboratory
Glebe, New South Wales 2037, Australia


Abstract
Introduction
Quantitative analysis of field populations
Rate of genetic exchange and the evolution of drug resistance: The example of Plasmodium.
Conclusions
References


Abstract

Anthelmintic resistance in nematode parasites of sheep and goats is now a firmly established phenomenon, particularly in warm temperate or tropical regions of the world. The evolution of resistance by nematodes to broad-spectrum anthelmintics is of particular concern, as there are currently only three different chemical families of such drugs. Resistance to two of these is already ubiquitous, with several reports of resistance to the third. With no new broad-spectrum anthelmintics on the horizon, it is vital that we learn to manage resistance and conserve susceptibility to these valuable drugs for longer than we have in the past.

A crucial factor in the evolution of anthelmintic resistance in a worm population is the extent to which survivors of drug treatment contribute their genes to future generations of worms. This contribution is influenced by frequency and timing of anthelmintic treatment, drug efficacy, life-expectancy and fecundity of adult worms, and current and future rates of larval intake. Larval intake, in turn, is determined by previous egg deposition, grazing management and weather. The acquired immune response of the host is of central importance through its effects on worm establishment, fecundity and death rate, and provides density-dependent regulation of worm populations.

Physical experimentation with such a complex system is difficult, expensive and above all time-consuming. Further, results are usually specific to the site and type of animal management used in the experiment. Because of the large number of climatic, biological and management variables that interact to determine the size and genetic constitution of a worm population, we believe that the only practical way to explore the system is with the aid of a model. Examples of the types of models used to investigate anthelmintic resistance are presented. The CSIRO model UNIVERSE is used to examine the consequences of some common management practices for evolution of anthelmintic resistance in the ruminant parasite Trichostrongylus colubriformis.

The problem

Anthelmintic resistance in nematode parasites of sheep, goats and horses is now a firmly established phenomenon, particularly in warm temperate or tropical regions of the world (Waller and Prichard, 1986). The evolution of resistance by nematodes to broad-spectrum anthelmintics is of particular concern, as there are currently only three different chemical families of such drugs. Resistance to two of these, the benzimidazoles and levamisole/morantel is already common in sheep and goat nematodes, with several reports of resistance to the third, the avermectins (Prichard, 1990). With no new broad-spectrum anthelmintic families readily apparent, and only these three developed in the last 30 years, it is vital that we manage resistance and conserve susceptibility to these drugs for longer than we have in the past.

A critical factor in the evolution of anthelmintic resistance in a worm population is the extent to which survivors of drug treatment contribute their genes to future generations of worms. Given that their initial domination of the post-treatment worm burden is progressively diminished by establishment of less-selected larvae from pasture (Martin, 1990), this contribution is influenced by frequency and timing of anthelmintic treatment, drug efficacy, life-expectancy and fecundity of adult worms, and current and future rates of larval intake. Larval intake, in turn, is determined by previous egg deposition, grazing management and weather. The acquired immune response of the host (Wakelin, 1987) is of central importance through its effects on worm establishment, fecundity and death, and provides density-dependent regulation of worm populations.

Physical experimentation with, or even monitoring of, such a complex system is technically difficult, expensive and, above all, time-consuming. Current technology is unable to detect anthelmintic resistance by any means short of full "treat-and-slaughter" trials until resistant worms comprise 25-50% of the worm population (Martin et al., 1989; E. Lacey, unpublished data). Further, results are usually specific to the particular site and type of animal management observed. This is borne out by the common observation, in Australia at least, that the extent and type of broad-spectrum anthelmintic resistance varies widely among sheep farms within the same climatic region, sometimes reflecting varying patterns of previous anthelmintic use and flock management (Edwards et al., 1986), and sometimes not (Waller et al., 1988). Because of the large number of climatic, biological and management variables that interact to determine the size and genetic constitution of a worm population, we believe that the only practical way to explore the system of grazing hosts, helminth parasites and anthelmintics is with the aid of a mathematical model.

Models of anthelmintic resistance


Abstract
The problem
Models of anthelmintic resistance
Implications of models for resistance management
Will models help solve the problem?
Acknowledgements
References


Although the use of biocides for control of both arthropods and helminths has led to similar problems of biocide resistance, Smith (1990) has suggested that the theoretical and modelling literature on the evolution of pesticide resistance in arthropods may not be directly applicable to superficially similar questions about anthelmintic resistance. There are three differences between arthropod and helminth populations that account for this difficulty of extrapolation. First, helminth generations are overlapping rather than discrete and contemporary generations can vary widely with respect to frequencies of resistance genes. Secondly, treatment of hosts with anthelmintics exposes only the parasitic stages of the helminth population to the selective effects of treatment; the numerous free-living stages on pasture escape selection at that time (Martin, 1985). Finally, nematodes are much more immobile than arthropods and rely overwhelmingly on their hosts for transport. Nevertheless, an examination of resistance management tactics used by entomologists reveals little that is conceptually new to helminthologists; the most effective tactic so far has been the reduction of pesticide use (Roush, 1990).

There have been few attempts to model the evolution of anthelmintic resistance. Models range in complexity from simple spreadsheet models used to examine components of the entire system, such as the combined effects of drug efficacy and the size of the free-living fraction of the worm population examined by Martin (1990), to the complex simulation models described by Gettinby et al. (1989) and Barnes and Dobson (1990). Models of the genetics of anthelmintic resistance have been described by Anderson (1983) and Dobson et al. (1987). These sub-system models are valuable in exploring and understanding their restricted domains, but recommendations to manage resistance need to be tested either within the wider context of more comprehensive whole-system models, or in the field.

To capture the essential features of the interactions between parasitic and free-living populations, and the selective effects of anthelmintic treatment, an anthelmintic resistance model must at least include components that model the dynamics of both populations and the genetics of resistance. Details of model structure and examples of predictions have been published for three models of the evolution of anthelmintic resistance in nematode parasites of grazing animals. Gettinby et al. (1989) and Gettinby (1989; 1990) described a site-specific simulation model for Ostertagia circumcincta in sheep, which allowed examination of the parasite population dynamics and evolution of anthelmintic resistance in ewe-lamb management systems under various combinations of anthelmintic treatment and grazing management. While climatic factors influenced development and survival of free-living stages, and thus permitted an assessment of the effect of climate (Gettinby, 1990), the model in its current form recycles one year's meteorological data from its site annually for the duration of its simulation. The best strategies for minimizing selection of the worm population were associated with reduced frequencies of treatment and moving lambs to clean pastures at weaning (Gettinby et al., 1989; Gettinby, 1990).

The model of Barnes and Dobson (1990) allows any sheep management system involving up to five flocks grazing up to five paddocks to be simulated. Based on extensive data on the population dynamics of the parasitic stages of Trichostrongylus colubriformis, the parasitic stage sub-model (Dobson et al., 1990) is combined with a pasture sub-model which uses daily temperature, rainfall and evaporation to predict the population dynamics of free-living stages (Barnes et al., 1988). Although site-specific, in the sense that it may be used to simulate results for a specific flock or farm, the model can be used to simulate results at any site for which daily climatic data are available. It has been used successfully for this purpose for several sites in Australia and Fiji.

Smith (1990) has presented the most general model of the evolution of anthelmintic resistance. It is neither site nor species-specific, and was used to examine the consequences of several drug management strategies for the evolution of single-gene resistance to one or two drugs. Population dynamics of the free-living and parasitic stages were modelled as a pair of linked differential equations for each worm genotype. These equations were solved numerically so that integration could be stopped when required for the application of anthelmintic treatment. Acquired immunity was not modelled explicitly, but worm populations were regulated by density-dependent mortality of parasitic stages. There was no provision for climate to influence development or survival of the free-living stages, nor for variations in grazing management involving pasture rotations. Effects of stocking rate could be examined indirectly by altering the parameter describing net infection rate.

While these restrictions may be limitations in any attempt to use Smith's model in a site- or species-specific manner (a purpose for which it was not intended), the strength of this approach lies in its generality and mathematical economy. The biologically complex models of Gettinby et al. (1989) and Barnes and Dobson (1990) necessarily contain a profusion of equations and parameters, many of which have been questioned by the authors themselves, let alone by critics. In contrast, Smith's formulation of the population dynamics of the parasitic and free-living stages reduces the system to its bare but probably sufficient essentials; the only disputation that seems possible might be over the values chosen for the seven parameters. The model's major use will be in exploration of the interplay between nematode population dynamics, genetics and anthelmintic management: It will also be valuable as an aid in ranking of anthelmintic management options for conserving drug susceptibility. Comparison of predictions from Smith's model and those of the two simulation models could also be fruitful. Similar outcomes would indicate consequences arising mainly from population dynamics and the genetics of anthelmintic resistance. Different outcomes may point to grazing management or climatic factors being primarily responsible.

Implications of models for resistance management

Management recommendations aimed at slowing the rate of selection of nematode parasites for anthelmintic resistance were urgently required before any of the three models described here were available to assist in their formulation. Of necessity, these recommendations (Prichard et al., 1980; Anon., 1989; Coles and Roush, 1992) were made largely by analogy with measures adopted to manage pesticide resistance in arthropods or on the basis of limited experimental evidence with nematode parasites. Recommendations made in one or more of these papers that are particularly amenable to investigation with models are (i) reduce frequency of treatment, (ii) avoid under-dosing, (iii) slow (annual) rotation of anthelmintic groups and (iv) integration of anthelmintic treatment and grazing management.

Of these recommendations, only (i) has been universally accepted as being almost self-evident, and has been confirmed by several experimental studies (e.g. Barton, 1983; Martin et al., 1984; Waller et al., 1989) end by models (Gettinby et al., 1989; Smith, 1990). Concern has been expressed about the implications of "dose-and-move" strategies for worm control on the grounds that they may select more powerfully for resistance (Le Jambre, 1978; Martin, 1985; Coles and Roush, 1992). This concern arises from consideration of a conceptual model - resistant worms surviving anthelmintic treatment and contaminating a clean pasture with eggs. When the conceptual model is broadened to allow for the reduction in treatment frequency permitted by such strategies, the concentrations of larvae on clean, or safe, pastures being small but not zero, and an anthelmintic efficacy of less than 100% against susceptible worms, the outcome seems less clear-cut. Moving sheep to lightly contaminated pastures after treatment did not select more rapidly for resistance than the same number of treatments given to set-stocked sheep in an experiment over five years reported by Waller et al. (1989). Simulations with computer models by Gettinby (1990) and Barnes and Dobson (1990) also indicated that dose and move strategies can provide more effective control with a lesser risk of anthelmintic resistance than set-stocking of ewes and lambs. We should point out, however, that results of such simulations are sensitive to assumptions made about the initial contamination of the safe pasture and the efficacy of the anthelmintic against susceptible and heterozygous resistant worm genotypes. The latter point was discussed extensively by Smith (1990) in relation to his model and is considered below in relation to treatment efficacy.

There has been less general agreement among parasitologists about the effectiveness of recommendations concerning the avoidance of under-dosing (administration of less than the manufacturer's recommended dose) and slow (annual) rotation between unrelated anthelmintic groups. Reservations about these recommendations have been largely confined to verbal arguments after conference dinners, probably because of the paucity of evidence on either side. Waller et al. (1989) reported results with Haemonchus contortus, but not T. colubriformis, that slightly favoured slow rotation over rapid (every treatment) alternation between thiabendazole and ivermectin when treatments were given eight times per year. No such difference was evident when sheep were treated three times per year.

Sequential, Rotational and Mixture Strategies

Smith (1990) examined resistance management strategies assuming resistance to any drug was determined by a single major gene with two alleles conferring resistance or susceptibility. He found little difference between sequential and rotational strategies using two drugs in their effects on evolution of resistance to both drugs. Simultaneous administration of both drugs was much more effective in delaying resistance, but only if the frequency of resistance genes to either drug was initially low.

In the following simulations with the Barnes and Dobson (1990) model, we examine the consequences of these strategies for preserving susceptibility to two unrelated anthelmintics. The sheep production system used as the basis of all simulations was one in which 100 12-week-old lambs were purchased each year on 1 January and grazed the same 10 ha paddock for 11 months. They were then notionally sold and replaced with a new flock of lambs one month later. All lambs were given anthelmintic treatment on 1 January, 1 March and 1 May, in accordance with current strategic control recommendations for the Armidale District. Daily climatic data from the CSIRO Pastoral Research Laboratory at Armidale for the years 1959 to 1968 were used in all simulations, which were run for a 10-year period. It was assumed that the sheep entered the simulation on 1 January each year with a mean burden of 5,000 T. colubriformis, and that on their farm of origin no anthelmintic resistance had developed. Resistance to each drug was assumed to be controlled by a single locus on different chromosomes. For both loci there were two alleles, S denoting susceptibility and R denoting resistance. All worm genotypes were regarded as equally fit in the absence of anthelmintic treatment. Initial frequency of R alleles for both drugs was set at 0.001 on the simulated paddock, unless otherwise stated, and on the farm of origin. Efficacies of both drugs against their SS, RS and RR worm genotypes were set at 0.99, 0.5 and 0.1 respectively, i.e. resistance was assumed to be incompletely dominant. Individual lambs were regarded as dead when their adult worm burdens exceeded 50,000.

Figure 1 shows simulated mean adult worm burdens, faecal egg counts, infective larvae on pasture and R allele frequency in infective larvae on pasture over the 10-year period. Two treatment regimes are compared; no treatment, or treatment three times per year with the same drug on all occasions as outlined above. Over the ten years involving 1,000 simulated lambs under the nil treatment regime, the model predicted 208 deaths from trichostrongylosis, compared with one death in the 3-drench program. Under this program, with the same drug used every year, R allele frequency in larvae on pasture approached 90% by the end of year 10 (Figure 1 (d)). The sharp decline and recovery in R allele prevalence near the beginning of each year was caused by the influx of SS genotypes in the 1% of susceptible worms surviving the 1 January treatment.

Figure 1. Results of simulation over ten years of the effect of no treatment, or three treatments per year on (a) adult worm burdens, (b) faecal egg counts, (c) concentration of infective larvae on pasture and (d) prevalence of alleles for resistance in the population of larvae on pasture. Initial frequency of R alleles was 0. 1%, efficacy of the anthelmintic was 0.99, 0.5 and 0.1 against SS, RS and RR worm genotypes respectively. Sheep management is described in the text.

Figure 2 shows the effect on R allele prevalence of three different ways of managing two unrelated drugs in the same 3-drench program shown in Figure 1. In Figure 2 (a), drug 1 was used for the three treatments for the first five years, then drug 2 was used solely for the second five years. Resistance to drug 1 increased exactly as for the drug in Figure 1 (d) for five years, then declined when use of this drug was suspended, under the influence of the annual influx of S alleles carried in by the replacement lambs. In the second five-year period, resistance to drug 2 appeared, and by the end of the ten-year simulation the prevalence of R alleles to both drugs was in the 30% to 60% range.

Figure 2. Simulated effects on prevalence of R alleles in infective larvae on pasture of three strategies for managing two unrelated anthelmintics. (a) Drug 1 used exclusively for the first five years, followed by exclusive use of drug 2. (b) Drugs 1 and 2 used in alternate years. (c) Drug 1 used for the first two treatments in each year and drug 2 used for the third treatment in each year. Initial R allele frequency, drug efficacies and sheep management as for Figure 1.

When the two drugs were rotated annually (i.e. drug 1 used in year 1, drug 2 in year 2 and so on), R allele prevalence to both drugs after ten years was in the range of 40% to 60% as shown in Figure 2 (b). Differences between drugs 1 and 2 in trajectories of R allele prevalence were due to effects of different weather in the years the two drugs were used. The predictions of this model are therefore consistent with the predictions of Smith's (1990) model, in that there is no clear advantage of rotational over sequential use of two drugs. In Figure 2 (c), drug 1 was used for the first two treatments each year, while drug 2 was used only for the third treatment each year. Under this fast rotation regime, resistant survivors of drug 1 treatment on days 1 and 60 each year are largely removed by treatment with drug 2 on day 120. Worms surviving drug 1 therefore have, at most, 120 days each year to contaminate pastures, while worms surviving drug 2 have up to 245 days. These differences in relative opportunities to contaminate pasture are reflected in R allele prevalence to the two drugs over the ten years. There was little resistance to either drug for the first seven years and resistance to drug 1 remained low over the whole period. Resistance to drug 2 increased rapidly in the last three years to reach levels similar to those resulting from sequential or slow rotation strategies.

When the three treatments per year consisted of simultaneous administration of drugs 1 and 2, appearance of resistance to either drug was dramatically delayed. In order to show any significant appearance of resistance on the same set of axes used in Figure 2 we had to re-run the simulation with the initial frequency of R alleles set at 0.05, rather than 0.001 as in Figures 1 and 2. Results of a comparison between annual rotation between drugs 1 and 2, and a mixture of the two drugs are shown in Figure 3. Again, the predictions are consistent with those of Smith (1990).

Effect of Anthelmintic Efficacy

Four simulations were run using the same drug over ten years for the 3-drench program described previously. Initial frequency of R alleles was 0.001. Four dose rates of the drug were simulated, with results shown in Figure 4. Dose A gave the same efficacies against the SS, RS and RR genotypes as used in all previous simulations, namely 0.99, 0.5 and 0.1. Dose B simulated a minor reduction in dose rate, which, given the sigmoidal shape of the usual dose-response curve, was assumed to result in efficacies against the SS, RS and RR genotypes of 0.99, 0.1 and 0.1. Efficacies against SS and RR genotypes were thus unchanged, with the only effect of the reduced dose rate being reduced efficacy against heterozygous resistant worms. Dose C simulated an increased dose rate, with efficacies of 0.99, 0.9 and 0.1 against SS, RS and RR worms. Finally, dose D simulated more extreme under-dosing, with an efficacy set of (0.90, 0.1, 0.1), i.e. reduced efficacy against SS genotypes.

Figure 4 shows that in comparison with the default efficacy set (0.99, 0.5, 0.1), mild under-dosing with reduced efficacy against RS worms (dose B) hastened selection for resistance for the first six years of the simulation. The sudden decline in R allele frequencies in year 6 was produced by unusually small concentrations of larvae on pasture at the beginning of that year. Dose C, representing an increased dose with greater efficacy against RS worms, and dose D, representing gross under-dosing with reduced efficacy against SS worms both resulted in weak selection for resistance compared with dose A. These results, which seem surprising at first, can be explained in terms of gene frequencies among survivors of the four dose rates. Doses A, B and C differed only in their efficacies against RS genotypes, and their apparent effectiveness in selecting for resistance increased from C to A to B, which was inversely related to their efficacies against RS worms. Survivors of dose B included a greater proportion of RS worms (0.9) than did survivors of dose A (0.5), which in turn represented a greater proportion of RS worms than did survivors of dose C (0.1). For dose D, which simulated extreme under-dosing with a reduction in efficacy against SS worms from 0.99 to 0.90, there were now ten times as many SS worms among the survivors, but only five times as many RS worms, when compared with the default dose A. The net effect has therefore been a reduced intensity of selection for resistance.

Figure 3. Comparison between effects of (a) annual alternation and (b) simultaneous administration of two unrelated anthelmintics on evolution of resistance. Initial frequency of R alleles 5%, drug efficacies and sheep management as for Figure 1.

This biphasic response, with increasing evolution of resistance associated with increasing anthelmintic efficacy against SS genotypes, followed by decreasing resistance with increasing efficacy against RS genotypes was also noted by Smith (1990), and was attributed to dilution of the very small number of RS survivors by incoming larvae from pasture once efficacy against SS worms reached 100%. While dilution effects are important, we believe that the phenomenon is more directly attributable to the effects of differential anthelmintic efficacy against the worm genotypes.

Figure 4. Effects of four dose rates of anthelmintic on evolution of resistance. Initial R allele frequency and sheep management as for Figure 1. Dose A killed 0.99, 0.5 and 0.1 of SS, RS and RR worm genotypes respectively. Dose B killed 0.99, 0.1 and 0.1, dose C killed 0.99, 0.9 and 0.1 and dose D killed 0.9, 0.1 and 0.1 of SS, RS and RR genotypes.

Considering the simplest case, of resistance attributable to variation at a single locus, where R allele frequency is p and S allele frequency is q, p + q = 1, and the population in Hardy-Weinberg equilibrium, if:

a is proportion of SS genotype surviving treatment
b is proportion of RS genotype surviving treatment
t is proportion of RR genotype surviving treatment

then p after treatment as a multiple of p before treatment can be shown to be:

This ratio is a measure of the selective effect of the treatment and is tabulated in Table 1 for a range of efficacies against SS and RS worm genotypes. As efficacy against SS increases down the columns of Table 1, so R allele frequency in surviving worms after treatment also increases. As efficacy against RS increases across the rows, R allele frequency after treatment declines, although for the low value of p = 0.001 chosen for this example the values in the table are more sensitive to SS efficacy than to RS efficacy.

Spatial Rotations

One of the major differences between arthropods and nematodes that has not been exploited in resistance management strategies is the relative immobility of nematode parasites. Barnes and Dobson (1990) noted the possibility of maintaining susceptibility to different anthelmintics in different paddocks of a farm. A simple example is presented in Figure 5, where a farm, or part of a farm, is divided into three paddocks of similar carrying capacity. Using the same notional production system used in previous simulations, purchased weaners were treated with drug 1 on day 1 as they entered paddock 1, drug 2 on day 120 when they were moved to paddock 2, and drug 3 on day 240 when they were moved to paddock 3. To make the management system more commercially realistic, a flock of mature wethers was grazed behind the weaners, in the paddock they had last vacated, so that at any time two-thirds of the area was being grazed and one-third spelled. These wethers were relatively resistant to nematode infection and received no anthelmintic treatment. Simulations over 20 years with Armidale climatic data, an initial frequency of R alleles to each drug of 0.001 and the default anthelmintic efficacy set (0.99, 0.5, 0.1) showed that mild resistance to drug 1 evolved on paddock 1, to drugs 1 and 2 on paddock 2, and to all three drugs on paddock 3. This did not lead to problems with nematode control, as the specific drug used on lambs before entering a specific paddock retained its initial efficacy for the full 20 years of the simulation against worms picked up from the previous paddock. When the same grazing management system was simulated with the three drugs used in annual, rather than spatial rotation, intense resistance to all drugs evolved on all paddocks, with adverse consequences for worm control.

Table 1. Ratio of R allele frequency after treatment to R allele frequency before treatment, for a range of treatment efficacies against SS and RS worm genotypes. R allele frequency before treatment was 0.001, efficacy against RR worms was 0.1.

Efficacy against SS (1-a)

Efficacy against RS (1- b)


0.000

0.100

0.200

0.500

0.900

0.990

0.900

9.8

8.9

7.9

5.0

1.0


0.950

19.3

17.4

15.5

9.8

2.0


0.980

45.5

41.4

37.1

23.9

5.0


0.985

58.9

53.7

48.3

31.3

6.6


0.990

83.5

76.4

69.1

45.6

9.9

1.1

0.995

143.1

132.6

121.4

83.5

19.4

2.2

0.998

250.3

237.2

222.5

167.0

45.9

5.4

0.999

333.6

321.8

308.1

250.5

84.1

10.7

Will models help solve the problem?

Given the extremely slow observed rates of reversion to anthelmintic susceptibility once use of an anthelmintic is discontinued (Martin et al., 1988), it is difficult to see how modelling, or indeed any other approach, can help to restore susceptibility in parasites that have already evolved substantial levels of resistance. Van Wyk and van Schalkwyk (1990) have experimentally overwhelmed resistant H. contortus in the field with a laboratory-bred susceptible strain, but the logistics of doing this on thousands of farms for even one nematode species are daunting.

Figure 5. Diagrammatic representation of a simple spatial rotation of three anthelmintics. The three paddocks constitute a whole farm, or part of a farm where susceptible animals are grazed. Drug treatment is only given at times of movement of hosts to new paddocks in the direction shown by the arrows. Larvae on paddock 1 remain susceptible to drugs 2 and 3, and larvae on paddock 2 remain susceptible to drug 3, thus a drug with high efficacy was always available, despite the development of resistance.

Modelling of anthelmintic resistance enables evaluation of strategies for conserving susceptibility to new or existing drugs. Currently available models indicate that reductions in treatment frequency, particularly when allied with grazing management, judicious choice of dose rates and combinations of unrelated anthelmintics can materially extend the useful life of effective anthelmintics. Spatial rotation of anthelmintic groups among the various paddocks of a farm also shows promise of delivering sustainable anthelmintic control programs. The biological and management flexibility of the Barnes and Dobson (1990) model has also led to its use in preliminary evaluation of non-chemical control technologies such as genetically resistant sheep (Barger, 1989; Windon, 1990), biological control and vaccination. Although models may not solve our current problems of anthelmintic resistance they will certainly help us to avoid exacerbating them, and to avoid repeating our past mistakes with future anthelmintics.

Acknowledgements

The financial support of the Australian Wool Research and Development Corporation is gratefully acknowledged.

References

ANDERSON, R.M. 1983. Reproductive success and fitness. In: Borgesteede, F.H.M., Henriksen, Sv.Aa. and Over, H.J., eds. Facts and Reflections IV. Resistance of Parasites to Anthelmintics. Lelystad: Central Veterinary Institute, pp. 51-58.

ANONYMOUS. 1989. Anthelmintic Resistance. Report of the Working Party for the Animal Health Committee of the Standing Committee on Agriculture. SCA Technical Report series; No. 28. Melbourne: CSIRO, pp. 16-17.

BARGER, I.A. 1989. Genetic resistance of hosts and its influence on epidemiology. Veterinary Parasitology 32: 21-35.

BARNES, E.H. and DOBSON, R.J. 1990. Population dynamics of Trichostrongylus colubriformis in sheep: computer model to simulate grazing systems and the evolution of anthelmintic resistance. International Journal for Parasitology 20: 823-831.

BARNES, E.H., DOBSON, R.J., DONALD, A.D. and WALLER, P.J. 1988. Predicting populations of Trichostrongylus colubriformis infective larvae on pasture from meteorological data. International Journal for Parasitology 18: 767-774.

BARTON, N.J. 1983. Development of anthelmintic resistance in nematodes from sheep in Australia subjected to different treatment frequencies. International Journal for Parasitology 13: 125-131.

COLES, G.C. and ROUSH, R.T. 1992. Slowing the spread of anthelmintic resistant nematodes of sheep and goats in the United Kingdom. Veterinary Record 130: 505-510.

DOBSON, R.J. GRIFFITHS, D.A., DONALD, A.D. and WALLER, P.J. 1987. A genetic model describing the evolution of levamisole resistance in Trichostrongylus colubriformis, a nematode parasite of sheep. IMA Journal of Mathematics Applied in Medicine and Biology 4: 279-293.

DOBSON, R.J., DONALD, A.D., BARNES, E.H. and WALLER, P.J. 1990. Population dynamics of Trichostrongylus colubriformis in sheep: model to predict the worm population over time as a function of infection rate and host age. International Journal for Parasitology 20: 365-373.

EDWARDS, J.R., WROTH, R., de CHANEET, G.C., BESIER, R.B., KARLSSON, J., MORCOMBE, P.W., DALTON-MORGAN, G. and ROBERTS, D. 1986. Survey of anthelmintic resistance in western Australian sheep flocks. 2. Relationship with sheep management and parasite control practices. Australian Veterinary Journal 63: 139-144.

GETTINBY, G. 1989. Computational veterinary parasitology with an application to chemical resistance. Veterinary Parasitology 32: 57-72.

GETTINBY, G. 1990. Computer models applied to drug resistance in parasites. In: Boray, J.C., Martin, P.J. and Roush, R.T., eds. Resistance of Parasites to Antiparasitic Drugs. Round Table Conference ICOPA VII Paris 1990. Rahway: MSD Agvet, pp. 213-219.

GETTINBY, G., SOUTAR, A., ARMOUR, J. and EVANS, P. 1989. Anthelmintic resistance and the control of ovine ostertagiasis: a drug action model for genetic selection. International Journal for Parasitology 19: 369-376.

Le JAMBRE, L.F. 1978. Anthelmintic resistance in gastrointestinal nematodes of sheep. In: Donald, A.D., Southcott, W.H. and Dineen, J.K., eds. The Epidemiology and Control of Gastrointestinal Parasites of Sheep in Australia. Melbourne: CSIRO, pp. 109-120.

MARTIN, P.J. 1985. Nematode control schemes and anthelmintic resistance. In: Anderson, N. and Waller, P.J., eds. Resistance in Nematodes to Anthelmintic Drugs. Melbourne: CSIRO/Australian Wool Corporation, pp. 29-40.

MARTIN, P.J. 1990. Ecological genetics of anthelmintic resistance. In: Boray, J.C., Martin, P.J. and Roush, R.T., eds. Resistance of Parasites to Antiparasitic Drugs. Round Table Conference ICOPA VII Paris 1990. Rahway: MSD Agvet, pp. 129-139.

MARTIN, P.J., ANDERSON, N. and JARRETT, R.G. 1989. Detecting benzimidazole resistance with faecal egg count reduction tests and in vitro assays. Australian Veterinary Journal 66: 236-240.

MARTIN, P.J., ANDERSON, N., BROWN, T.H. and MILLER, D.W. 1988. Changes in resistance of Ostertagia spp. to thiabendazole following natural selection or treatment with levamisole. International Journal for Parasitology 18: 333-340.

MARTIN, P.J., ANDERSON, N., LWIN, T., NELSON, G. and MORGAN, T.E. 1984. The association between frequency of thiabendazole treatment and the development of resistance in field isolates of Ostertagia spp. of sheep. International Journal for Parasitology 14: 177-181.

PRICHARD, R.K. 1990. Anthelmintic resistance in nematodes: extent, recent understanding and future directions for control and research. International Journal for Parasitology 20: 515-523.

PRICHARD, R.K., HALL, C.A., KELLY, J.D., MAETIN, I.C.A. and DONALD, A.D. 1980. The problem of anthelmintic resistance in nematodes. Australian Veterinary Journal 56: 239-251.

ROUSH, R.T. 1990. Genetics and management of insecticide resistance: lessons for resistance in internal parasites? In: Boray, J.C., Martin, P.J. and Roush, R.T., eds. Resistance of Parasites to Antiparasitic Drugs. Round Table Conference ICOPA VII, Paris, 1990. Rahway: MSD Agvet, pp. 197-211.

SMITH, G. 1990. A mathematical model for the evolution of anthelmintic resistance in a direct life cycle nematode parasite. International Journal for Parasitology 20: 913-921.

Van WYK, J.A. and Van SCHALKWYK, P.C. 1990. A novel approach to the control of anthelmintic-resistant Haemonchus contortus in sheep. Veterinary Parasitology 35: 61-69.

WAKELIN, D. 1987. The role of the immune response in helminth population regulation. International Journal for Parasitology 17: 549-557.

WALLER, P.J. and PRICHARD, R.K. 1986. Drug resistance in nematodes. In: Campbell, W.C. and Rew, R.S., eds. Chemotherapy of Parasitic Diseases, New York: Plenum Publishing Corporation, pp. 339-362.

WALLER, P.J., DOBSON, R.J. and AXELSEN, A. 1988. Anthelmintic resistance in the field: changes in resistance status of parasitic populations in response to anthelmintic treatment. Australian Veterinary Journal 65: 376-379.

WALLER, P.J., DONALD, A.D., DOBSON, R.J., LACEY, E., HENNESSY, D.R., ALLERTON, G.R. and PRICHARD, R.K. 1989. Changes in anthelmintic resistance status of Haemonchus contortus and Trichostrongylus colubriformis exposed to different anthelmintic selection pressures in grazing sheep. International Journal for Parasitology 19: 99-110.

WINDON, R.G. 1990. Selective breeding for the control of nematodiasis in sheep. Revue Scientifique et Téchnique Office International des Epizooties 9: 555-576.

Session discussion

The issue of drug resistance in several different parasites was discussed and the view was expressed that before modelling the phenotype of drug resistance in trypanosomes, data on the biological basis of the resistance should be obtained. It was suggested that once the mechanism of drug resistance was identified, models could be developed considering whether the resistance was due to point mutation, gene amplification or deletion. If resistance was due to epigenetic events, the approach to modelling would be different.

It was pointed out, however, that the model presented on anthelmintic resistance was based on the assumption that there is a genetic basis for resistance, since there are no biological data available on the mechanisms of resistance to anthelmintics.

The question was raised as to whether anthelmintic resistance models could be applied to study acaricide resistance in ticks. It was the opinion of several of the modellers that they may not be directly applicable as the current anthelmintic models are based on the assumption that there is a single locus for anthelmintic resistance in nematodes and that this may not be true for acaricide resistance in ticks. It was pointed out that although there had not been many studies of the genetic mechanisms of anthelmintic resistance, it had been found that there are generally one or a small number of genes involved, hence the provision for up to three genes controlling resistance in the model presented.

The issue of whether the model presented allowed for reversion of anthelmintic resistance was brought up. In replying, the modeller stressed that there was little information on this subject and the opinion was expressed that reversal of resistance appears to be very slow.

The anthelmintic resistance model will be mainly used by field veterinarians and extension workers, not by farmers, because a knowledge of parasitology is required to operate it.

Subsequent to a discussion on the role of models in analysing drug kinetics in trypanosomes, it was pointed out that a modelling approach can be useful, even if the model is wrong, in order to dissect and understand the processes involved.

As a result of a question regarding necessary vaccine coverage levels to achieve population immunity, a general discussion developed about the level of immunological protection or chemotherapeutic cover required for different diseases and epidemiological situations. It was emphasized that one must consider the levels of benefits considered to be essential or acceptable and pointed out that control strategies that did not provide 100% protection were by no means useless.


Previous Page Top of Page Next Page