Results and discussion

IS6110 is a transposon-like element identified in the genome of M. tuberculosis that occurs at various locations and at variable copy number. This element is probably most widely used as a polymorphic marker for genotyping and while it has its weaknesses, it will be used as an example to illustrate the concepts concerning genome variability and molecular epidemiology in this discussion. An example where IS6110 was used as a marker in restriction length polymorphism analysis is shown in Fig. 1. This figure shows the salient points of genotype comparison, namely the case where some strains match others (100% similarity index e.g. T2 and R6-R9), others where closely related strains can be seen (e.g. compare U5 with R15, and R23 with R24) and others with no match or no similarity index (compare R30 with R31, R38 and T2). This figure represents strains obtained from a rural community (R) in Mpumalanga, from an urban community (U; Warren et al 1996b) in Ravensmead/Uitsig and from a large drainage area around a metropolitan urban community (T) in the Western Cape Province. No contact occurs or has occurred as far as can be ascertained between the rural and urban communities (approximately 2000 km apart and of quite different economic, a nj^i o ^ w m ^ ui r- CO

tj nj ."v f .■*» r*- i*1 j "*> M ¡c m M

aeuQiaiExEatffnciKdih- « w o iMmif in n> i^co a a — m-

- hi — a} ^ m < — — — a", Y- T- r*. ~ cfl — & ^ pj ^ w ^ T-

tj nj ."v f .■*» r*- i*1 j "*> M ¡c m M

aeuQiaiExEatffnciKdih- « w o iMmif in n> i^co a a — m-

- hi — a} ^ m < — — — a", Y- T- r*. ~ cfl — & ^ pj ^ w ^ T-

FIG. 1. IS61'10-generated DNA fingerprint patterns of strains of Mycobacterium tuberculosis from a rural community (R), from an urban community (U) 2000 km distant from R (Warren et al 1996b) and from a large drainage area surrounding the urban community (up to 500 km distant from R; T). Scale shows similarity index. Figure shows identical matches (e.g. R6—R9), similar (clonal variants R23 and R24) and matching strains where no epidemiological relationship is evident (e.g. R'17 compared to U8).

ethnic and language groups) or between the rural and town communities. However, genotypically identical strains are seen within these communities.

The recovery of genotypically identical strains from apparently unassociated hosts is evidence for a clonal population structure (O'Rourke & Spratt 1994). Figure 1 also illustrates the occurrence of clonal variants (e.g. R17—R21), where there is in all likelihood a common progenitor. Note that strain families (similarity index > 80%) are represented by strains from different communities, thus suggesting the spread of a common ancestor or variants springing from this ancestor. This could perhaps have occurred during the pioneer expansion in the early colonial history of the country. In this context, it is interesting to note that strains which dominate in Asia (van Soolingen et al 1995) have a particular signature pattern that is detectable in many South African strains. This may reflect early colonial trade with Asia or result from the importation of Asian slaves and workers in colonial times. Yet, the majority of strains from the rural community in South Africa do not match any strains from the urban community (the latter is a large database). This is suggestive of at least two routes of infection into the rural community or of sufficient genomic change that a common ancestor can no longer be easily recognized.

It is not possible to study accurately the dynamics of disease in a given area by limited sample selection, and therefore we believe that it is better to focus on a defined group in a given area and attempt full coverage, because sampling rate will influence cluster analysis and the error effects will increase as the sample size or rate decreases. Comprehensive analysis is currently being attempted in a high incidence urban community (Ravensmead/Uitsig; Warren et al 1996b) and results from the strains collected over approximately 3.5 years will be discussed. Identical strains have been grouped into clusters and the results analysed using the formula:

No. of strains in clusters — No. of clusters.

Total number of strains

The results of this analysis are shown in Table 1. This so-called («—1) formula (Small et al 1994) is used to estimate recent transmission. Given that the commonly accepted clinical definition of recent transmission in tuberculosis is transmission occurring over 24 months, it can be seen that an estimate of recent transmission in this community is somewhat less than 50% of the total tuberculosis cases. While it can be seen that the extent of clustering rises between six and 42 months, it can also be ascertained that the rise after 12 months is slow. Yet, it can be seen that the majority of strains in this community (70%) have at least one match within 42 months. The majority of the clusters of strains seen in this high incidence community consists of clusters of two or three individuals (81% of the total). However, it is known that IS6110 alone cannot accurately cluster

TABLE 1 Cluster analysis in a high incidence community

Time (months)2- Per cent clustering (n-1)h Per cent clustering (n)c

Time (months)2- Per cent clustering (n-1)h Per cent clustering (n)c

TABLE 1 Cluster analysis in a high incidence community

6

12-30

30

12

41

56

24

48

62

42

53

70

aThe estimates for six months depend on the window and are variable. The 12-month estimate is the average of three separate years (1993-1995).

bClustering (n — 1) was calculated according to the formula:

aThe estimates for six months depend on the window and are variable. The 12-month estimate is the average of three separate years (1993-1995).

bClustering (n — 1) was calculated according to the formula:

No. of strains in clusters — No. of clusters Total No. of strains cClustering (n) was simply the proportion of strains in clusters.

strains with five or fewer elements of IS6110 and these should therefore be excluded from the analysis. When this is done the percentage of strains clustered within 42 months drops to 46%. It has also been shown in two studies (Warren et al 1996a, Burman et al 1997) that subcluster analysis will almost certainly decrease the estimate of clustering by about 8-10% and therefore a more accurate reflection of recent transmission is probably 38-40%. This low value was not expected in this high incidence community, nor that it would coincide with estimates of clustering from large metropolitan areas in developed countries (Alland et al 1994, Small et al 1994), low incidence developed and developing countries (Hermans et al 1995), and island communities (i.e. French Polynesia; Torrea et al 1996). This phenomenon remains to be satisfactorily explained; however, it must be acknowledged that subgroups of persons in various communities may show vastly different clustering and therefore recent transmission estimates. Examples of this would be prison communities or HIV communities, where clustering is high. It should therefore be realized that estimates quoted are an average for any given community. Cluster estimates are based on 100% identity in pattern matching, which is a further source of error, because if similar strains are epidemiologically linked, these will be excluded (Mazurek et al 1991). More research is required to clarify this risk, because even strains with 100% matching may not be epidemiologically linked (Braden et al 1997, and as indicated in this chapter).

Pattern matching is problematical in fingerprint analysis, and it has been shown that no single probe is sufficiently accurate for epidemiological matching. There is therefore a need to develop a working system that can be universally applied. A future scenario is one where multiplex PCR may be done on any given strain, so that a number of well-known and highly polymorphic sites can be simultaneously amplified to give a consistent, easily assignable pattern. These patterns which would ideally have exactly the same number of fragments in each case, could be easily loaded and compared by spreadsheet analysis. Some potential sites have been identified (e.g. Goyal et al 1994, Vera-Cabrera et al 1997), but this idea has not yet been tested.

Some of the commonly held dogmas which may be challenged by molecular epidemiology are that high incidence implies high transmission rate (previously shown to be incorrect) and that multiple cases in a household are the result of household transmission. Our results would suggest that this occurs in no more than 50% of the households studied, but with a mother—child relationship carrying a far higher transmission risk. Furthermore, the analysis of drug-resistant cases in our study community suggests that the majority of drug-resistant cases are transmitted and not acquired. These results hold considerable significance for therapy and policy planning.

Clearly, although the estimate of the extent of recent transmission may be similar in different communities, there are quite different risk factors driving the transmission. In a diverse area (e.g. a large city) an average estimate is therefore of less value than that obtained in a limited and well-defined study area or community. In planning the control of tuberculosis, an accurate estimate of clustering is likely to be of considerable importance. Should the cluster estimates obtained thus far prove to be correct, i.e. implying that the majority of cases are due to reactivation, then even if we achieve a 100% cure rate of cases, a substantial reservoir of disease is likely to remain present in communities for decades. This is particularly true if no active case finding or contact tracing and prophylaxis of contacts is done. Furthermore, modelling and prediction of the cause of the disease could be extremely valuable for policy makers and planners (Garcia et al 1997), but accuracy is clearly required. Essentially, one needs to calculate the effective reproductive rate (R) of tuberculosis to predict the course of an epidemic, where if R >1 an epidemic will prevail, but if R < 1 an epidemic will wane. The calculation of R is theoretically relatively simple, since R is equal to the infection rate multiplied by the duration of infection multiplied by the probability that a contact will become infectious (Blower et al 1996). It is clear that an authentic determination of R depends on the accuracy of the various functions used in this calculation which is in turn dependent on the accuracy and reliability ofmolecular epidemiology. The calculations that may be made from data acquired in molecular epidemiology studies will enable an impartial assessment of the efficacy of past, current and future control programmes.

It is curious that although most clusters consist of two to four individuals and there is a rapid decline in the number of clusters from two to nine individuals, there are four large clusters identified in our high incidence urban community (Warren et

FIG. 2. Similarity matrix of a collection of strains from a well-defined high incidence tuberculosis community (Warren et al 1996b) collected over 3.5 years. Diagonal represents strains of 100% similarity coefficient with corresponding strains. Grey tones represent less than 100% similarity. Note clustering into some dominant families and clear division into two major groups (clear space).

FIG. 2. Similarity matrix of a collection of strains from a well-defined high incidence tuberculosis community (Warren et al 1996b) collected over 3.5 years. Diagonal represents strains of 100% similarity coefficient with corresponding strains. Grey tones represent less than 100% similarity. Note clustering into some dominant families and clear division into two major groups (clear space).

al 1996b, R. Warren, M. Richardson & S. Sampson, unpublished results 1996). These clusters include 11, 12, 15 and 35 individuals, respectively, to date, and although these four clusters represent only about 4% of the number of clusters identified, they include 20% of the strains found in clusters or 15% of the total number of strains. It is possible that these strains represent those where a higher degree of bacterial fitness may be represented (e.g. transmissibility, virulence, pathogenicity). This may be related to strain uptake (Cywes et al 1997), host resistance (Rhoades & Orme 1997) or many as yet unknown factors. Not only are these strains disproportionately represented, but their clonal variants are also present in large numbers. This is represented in Fig. 2, where every strain in the large database is compared to every other strain and a similarity index plotted. The darker shaded triangles represent strains with an increasing degree of similarity, and they suggest that a few families dominate in the community. This diagram may therefore represent clonal expansion of highly successful strains in this relatively stable community. Neo-Darwinism explains that genotypes that have a higher level of fitness will contribute disproportionately to successive generations (Brookfield 1996, Shapiro 1995). Some mutations arise more frequently because of selection and are therefore useful (e.g. drug resistance; Shapiro 1995), but there may also be gradual changes in phenotype, where mutations are cumulative and each may have only a minor but dominant effect. Figure 2 shows another significant feature, namely a clear segregation between two groups of strains. At this stage the significance of these two independent groups remains obscure.

It is interesting that the largest cluster identified contains 21 IS6110 inserts and consists of 35 individuals. The results of copy number analysis correlated with clustering are also suggestive that fitness may relate to copy number, since at higher copy number the likelihood of changes is enhanced. For example, the results presented in Table 2 show that those strains with a higher copy number of IS6110 (17—23 copies) occur significantly more frequently (p<0.009) in clusters than as unique strains.

Our lack of understanding of much of the biology of M. tuberculosis is partly a consequence of a dearth of information concerning the ecology of the organism. A population of bacteria may be either panmictic (with random association between loci), clonal or some function between these two extremes. The bacteria may be sexual, but because of epidemic status may appear to be superficially clonal. At present, there is no evidence for panmixia in tuberculosis, but rather clonality, with the possibility that all tuberculosis bacteria may be traced to one progenitor (Sreevatsan et al 1997). In the clones (clonal variants) and families of strains seen in M. tuberculosis, there is significant evidence for association between loci (i.e. temporary disequilibrium because of explosive expansion). This still needs to be tested by linkage disequilibrium analysis (non-random association of loci), which should theoretically prove clonality. However, the proof will be complex because

TABLE 2 IS6110 copy number in clustered or unique strains

Number of strains

IS6110 copy number

Clustered (C)

Unique (U)

Ratio (C:U)

6-16

211

136

1.55 : 1

17-23

64

20

3.2 : 1

this equilibrium can also arise by genetic drift. Recent evidence suggests that transposition is biased towards certain sites (S. Sampson, R. Warren, M. Richardson & P. van Helden, unpublished results 1997, M. D. Cave, personal communication 1997), which is another factor that will influence the interpretation of results.

Pathogenic bacteria may require some degree of recombination in order to achieve variation (e.g. cell surface epitopes) to protect against (periodic) selection. The reason for the apparently large variation in M. tuberculosis still remains obscure, given that much of the genome is quite possibly invariant and highly conserved (Sreevatsan et al 1997). The strain heterogeneity seen in any given geographic area or population group may then be due to heterogeneity in hosts and could be locally biased due to ethnic differences (the relative proportion of susceptible compared to resistant hosts). The degree of heterogeneity could also reflect the stage of the epidemic, since in a low incidence community there will be few hosts and possibly little opportunity for change, whereas in a high incidence community the opportunity for diversification is probably high. If we do not understand the rate of change (diversification), it is possible that our estimate of clustering and understanding of molecular epidemiology will be faulty. In this context, it could be that unique strains (often detected in older patients and thought to be associated with latency) may be unique owing to hypermutability induced by starvation (Yarmolinsky 1995) such as is experienced in a moribund state (Rosenberg 1994).

Thus, for many possible reasons, we have a number ofvariants that may exist in order to preserve diversity. The concept that there may be phenotypic variations with different degrees of virulence is commonly held (see Fig. 3, model A). However, this model probably reflects an oversimplified situation and the population biology of tuberculosis is possibly more accurately depicted by model B in Fig. 3. A three-dimensional figure would enhance the complexity and almost certainly provide a more accurate depiction ofthe population biology. In this case, the third dimension could represent host susceptibility or resistance, which will also be variable and dependent on both genetics and environment. By availing itself of the opportunity to exist in different forms, the tuberculosis bacterium can best ensure its survival, because clearly in a highly virulent form there may be less chance for long-term survival. Similarly, with no variation and in a limited population of hosts, a rapidly progressing form may disappear (Rhodes & Anderson 1996).

By operating in different modes, the bacillus is able to ensure its survival by (periodic) explosive expansion (Chevrel-Dellagi et al 1993) in local or susceptible populations, as well as ensuring its long-term survival in the event of the failure of the above strategy. Thus, tuberculosis is likely to be endemic in both large and small (isolated) population groups. The 'rapids' allow for rapid clonal expansion,

FIG. 3. Conceptual models for strain phenotype.

but the 'persisters' do it more invidiously by possibly interfering with the host immune function, resulting in dormancy. Thus, it is unlikely that the genome of M. tuberculosis is a simple structure with individual non-redundant genes coding for a single virulence factor, rather virulence (or other qualitative characteristics) will be a function of degree, probably determined by multiple genes and products. Unfortunately, genetics alone cannot predict the fitness of genotype as yet, but the analysis of genetic variation is an exciting and important area for study.

A population of bacteria is a natural experiment, where during natural recombination (knockout) mutants will probably be generated. Some of these will be lethal and not found in nature. Any natural knockout found in a patient will thus invalidate the knockout gene as a target for chemotherapeutics. Conversely, certain knockouts may dominate (e.g. a large cluster), by providing the organism with a natural advantage. Thus, genes found to be involved in pathology, virulence, transmissibility or other functions can be validated as targets prior to drug design and development.

The synthetic generation of recombinant knockouts will complement this natural experiment, but has the inherent problem that in vivo testing will be extremely complex and time consuming. A careful combination of recombinant experimentation together with analysis of strains found in the natural experiment of life may provide a synergistic answer towards discovery of new drugs so urgently needed to combat tuberculosis.

I thank the Medical Research Council of South Africa, the University of Stellenbosch, Tygerberg Hospital, Glaxo Wellcome Action TB Initiative and National Institutes of Health grant RO1 A135265—03 for financial support. The European Union contract Biomed1-BMH1-CT93— 1614 assisted with computer analysis. I would also like to thank my colleagues for many hours of stimulating discussion and collaboration, particularly R. Warren, J. Hauman, I. Wiid, W. Bourn, T. Victor, N. Beyers, A. van Rie, P. Donald, M. Richardson, S. Sampson and C. Classen. Finally, I thank Eileen van Helden for many forms of assistance, not the least of which is her editorial assistance in completing this chapter.

References

AllandD, Kalkut GE, Moss AR et al 1994 Transmission of tuberculosis in New York City — an analysis by DNA fingerprinting and conventional epidemiological methods. New Engl J Med 330:1710-1716

Blower SM, Small PM, Hopewell PC 1996 Control strategies for tuberculosis epidemics: new models for old problems. Science 273:497-500 Braden CR, Templeton GL, Cave MD et al 1997 Interpretation of restriction fragment length polymorphism analysis of Mycobacterium tuberculosis isolates from a state with a large rural population. J Infect Dis 175:1446-1452 Brookfield J 1996 Population genetics. Curr Biol 6:354-356

Burman WJ, Reves RR, Hawkes AP et al 1997 DNA fingerprinting with two probes decreases clustering of Mycobacterium tuberculosis. Am J Respir Crit Care Med 155:1140-1146 Chevrel-Dellagi D, Abderrahman A, Haltiti R, Koubaji H, Gicquel B, Dellagi K 1993 Large-scale DNA fingerprinting of Mycobacterium tuberculosis strains as a tool for epidemiological studies of tuberculosis. J Clin Microbiol 31:2446-2450 Cywes C, Hoppe HG, Daffe M, Ehlers MRW 1997 Non-opsonic binding of Mycobacterium tuberculosis to complement receptor type 3 is mediated by capsular polysaccharides and is strain dependent. Infect Immun 65:4258-4266 De la Salmoniere Y-OG, Li MH, Torrea G, Bunschoten A, van Embden J, Gicquel B 1997 Evaluation of spoligotyping in a study of the transmission of Mycobacterium tuberculosis. J Clin Microbiol 35:2210-2214 Dubos R 1960 Mirage of health. Allen & Unwin, London

Garcia A, Maccario J, Richardson S 1997 Modelling the annual risk of tuberculosis infection. Int

J Epidemiol 26:190-203 Goyal M, Young D, Zhang Y, Jenkins PA, Shaw RJ 1994 PCR amplification of variable sequences upstream of the katG gene to subdivide strains of Mycobacterium tuberculosis complex. J Clin Microbiol 32:3070-3071 Haas WH, Butler WR, Woodley CL, Crawford JT 1993 Mixed-linker polymerase chain reaction: a new method for rapid fingerprinting of isolates of the Mycobacterium tuberculosis complex. J Clin Microbiol 31:1293-1298 Hermans PWM, Messadi F, Guebrescabher H et al 1995 Analysis of the population structure of Mycobacterium tuberculosis in Ethiopia, Tunisia and the Netherlands: usefulness of DNA typing for global tuberculosis epidemiology. J Infect Dis 171:1504-1513

Mazurek GH, Cave MD, Eisenach KD, Wallace RJ, Bates JH, Crawford JT 1991 Chromosomal DNA fingerprint patterns produced with IS6110 as strain-specific markers for epidemiology study of tuberculosis. J Clin Microbiol 29:2030-2033 O'Rourke M, Spratt BG 1994 Further evidence for the non-clonal population structure of Neisseria gonorrhoeae: extensive genetic diversity within isolates of the same electrophoretic type. Microbiology 140:1285-1290 Rhoades ER, Orme IM 1997 Susceptibility of a panel of virulent strains of Mycobacterium tuberculosis to reactive nitrogen intermediates. Infection Immunity 65:1189-1195 Rhodes CJ, Anderson RM 1996 A scaling analysis of measles epidemics in a small population.

Philos Trans R Soc Lond B Biol Sci 351:1679-1688 Rosenberg SM 1994 In pursuit of a molecular mechanism for adaptive mutation. Genome 37:893-899

Shapiro JA 1995 Adaptive mutation: who's really in the garden? Science 268:373-374 Small PM, Hopewell PC, Singh SP et al 1994 The epidemiology of tuberculosis in San Francisco.

N Engl J Med 330:1703-1708 Sreevatsan S, Pan X, Stockbauer KE et al 1997 Restricted structural gene polymorphism in the M. tuberculosis complex indicates evolutionarily recent global dissemination. Proc Natl Acad Sci USA 94:9869-9874

Torrea G, Levee G, Grimont P, Martin C, Chanteau S, Gicquel B 1995 Chromosomal DNA fingerprinting analysis using the insertion sequence IS6110 and the repetitive element DR as strain-specific markers for epidemiological study of tuberculosis in French Polynesia. J Clin Microbiol 33:1899-1904 Torrea G, Offredo C, Simonet M, Gicquel B, Berche P, Pierre-Audigier C 1996 Evaluation of tuberculosis transmission in a community by 1 year of systematic typing of Mycobacterium tuberculosis clinical isolates. J Clin Microbiol 34:1043-1049 van Embden JDA, Cave MD, Crawford JT et al 1993 Strain identification of Mycobacterium tuberculosis by DNA fingerprinting: recommendations for a standardized methodology. J Clin Microbiol 31:406-409 van Soolingen D, Qian L, de Haas PEW et al 1995 Predominance of a single genotype of

M. tuberculosis in countries of East Asia. J Clin Microbiol 33:3234-3238 Vera-Cabrera L, Howard ST, Laszlo A, Johnson WM 1997 Analysis of genetic polymorphism in the phospholipase region of Mycobacterium tuberculosis. J Clin Microbiol 35:1190-1195 Warren R, Richardson M, Sampson S et al 1996a Genotyping of M. tuberculosis with additional markers enhances accuracy in epidemiological studies. J Clin Microbiol 34:2219-2224 Warren R, Hauman J, Beyers N et al 1996b Unexpectedly high strain diversity of Mycobacterium tuberculosis in a high-incidence community. S Afr Med J 86:45-49 Yarmolinsky MB 1995 Programmed cell death in bacterial populations. Science 267:836-837

DISCUSSION

Davies: If the adenylate cyclase gene of Streptomyces, which like mycobacteria is an actinomycete, is disrupted the organism develops an unusual growth defect. It grows to stationary phase, but fails to undergo a particular transition in growth phase normally associated with the switching on of certain metabolic pathways. Does a natural adenylate cyclase knockout of Mycobacterium tuberculosis grow normally, and are its properties as a pathogen affected?

vanMelden: I can't answer that specific question because we haven't gone back to that strain and looked at it from any of those points of view. The gene encoding adenylate cyclase is a multicopy gene in M. tuberculosis, so just one gene knockout may not have any effects.

Mi^rahi: There's at least one copy that is eukaryotic like and is not related to the Streptomyces gene.

Anderson: I didn't quite follow the argument about the estimated rate of reactivation versus active transmission. Are you totally excluding the possibility that somebody can be infected twice with the same strain?

van Melden: They can probably be infected twice with the same strain. If you analyse two cultures from a person, you don't know whether the person was not sterilized the first time round and then reactivated with the same strain, or whether they were reinfected by the same strain.

Anderson: So, there is uncertainty about recent transmission versus reactivation? vanMelden: Yes. The n — 1 formula is useful in a low incidence environment but perhaps not in a high incidence environment.

Mopewell: I don't understand the concept of families, as defined by restriction length fragment polymorphism typing, and how they develop, or at least how strains develop identical patterns from non-related sources. All the data suggest divergence in patterns rather than convergence. Unless you have a completely closed community where there are no mutations of the IS6110 patterns, then divergence rather than convergence is more likely, so how do families develop?

van Melden: We don't know enough to be able to say whether convergent or divergent changes occur. We have certain strains that are relatively stable, and we have a few that appear to be unstable, and they could be the ones that drive divergence. I would rather say that the diversity of M. tuberculosis has arisen over the last 6000—10 000 years by clonal expansion. However, this diversity could also be explained by multiple sources of import, i.e. from Asia, Europe and America.

Mopewell: About 25% of people with repeated positive cultures over time (i.e. a positive culture more than 90 days after the initial positive) have at least a one band change, i.e. either an addition, deletion or a shift.

vanMelden: The amount of change we see in our community is much less than you see in San Francisco. In our collection we have about five strains that demonstrate changes following repeat cultures.

Mopewell: Different strains may behave differently in terms of pattern shifts. Anderson: I also didn't totally understand your point about fitness and virulence. Fitness is simply reproductive success of the pathogen, in the strictest Darwinian sense. However, I didn't follow the argument about which out of the virulent, avirulent or moderate strains would be the most fit.

van Melden: But all strains are virulent. It's a matter of degree.

Anderson: Virulence can only be defined, in terms of transmission success, by the likelihood of transmission, which can be measured by the average duration of infection and the probability that it's going to be reactivated later on, etc. It's not clear to me whether evolution would drive the pathogen towards more or less virulence.

vanHelden: Sreevatsan et al (1997) proposed that the furthest evolved strains are the least virulent, but this has not been tested functionally or biologically.

Steyn: As a general rule, the more virulent organisms, and by that I mean the ones that cause disease, also have to transmit rapidly. You left me with the impression that you had a different relationship between virulence and transmission.

vanHelden: I don't want to discuss the definition of virulence. I am using the term loosely. If you could provide me with a better definition I would be grateful.

Steyn: A virulent organism would be one that caused a more severe disease, a more rapid disease progression and a more rapid transmission.

Anderson: The literature on this is large. There is a precise quantitative trade-off between transmissibility and virulence, in other words average survival or duration of infectiousness. Evolution for a variety of pathogens has gone in different directions. Sometimes it drives towards higher virulence, and sometimes to lower, but it is possible stably to maintain a variety of strains that differ in virulence. The molecular epidemiology areas of mycobacteria transmission, and genetic diversity within populations, how it's maintained and how it's changing, is about to open, and our notions about this are going to change dramatically in the coming years.

Ehlers: I don't understand how you can reach any conclusions about reactivation versus recent transmission from your data. You have shown clearly that there are four or five clusters. The members of those clusters may not be identical but they are highly related. This suggests to me that a limited number of strains are being exchanged within this community. It is possible that with each transmission from one host to the next there is a brief new acquisition of a polymorphism. You don't know the rate of change within a high incidence, high transmission community, so I don't think you can say anything about transmission versus reactivation.

vanHelden: I agree. We have to study the evolution and obtain a measure of rate of change. If we don't know the rate of change estimates of transmission could be wrong. This may be why we don't understand why in a high incidence community we do not detect a high degree of recent transmission.

Hopewell: This is understandable because most of the evidence suggests that acquisition of a new infection is much more difficult once someone becomes infected. Therefore, in a high incidence community people acquire their infection early on and then develop reactivation disease later.

van Helden: On the other hand, the contact tracing we have done within the clusters shows that we can only determine realistic personal contact for someone who has an identical strain. We have not been able to demonstrate personal contact in people, even within a family, who have minor differences. This is a circular problem, and we can only solve it by doing the molecular biology and sociology together.

'Kaplan: You have a community that's separated from the outside world by crossing the street. If you compared this community with the one four streets away, for example, would you lose reactivation? Your estimate of clusters versus non-clusters is lower than it is in reality, simply because, even though you have a large number of strains, it is a small sample within the entire community.

vanHelden: We picked that study community because it is bounded on three sides by non-suburban areas (industrial areas and a cemetery). Therefore, crossing the street for this suburb is only one of the four sites. Secondly, the people don't move in and out of the community to any large extent.

Kaplan: But they do come into contact with people across the street.

van Helden: Absolutely. And I don't deny that this is a weak point in the study.

Kaplan: Would the estimate have therefore increased if you had studied a larger area?

van Helden: Possibly, but we don't have the resources to do this.

Fourie: I would like to take this one step further. This study area should not be seen as an island, in terms of behavioural contacts, i.e. with whom people work and socialize. The population dynamics in Cape Town, as I understand, have changed dramatically over the past 50 years, and I believe that you will not understand the diversity in this local area unless you understand the strain diversity within the Eastern Cape, where there has been a high rate of migration by people seeking work. We need to set up a national database, as matter of urgency.

van Helden: I agree that we need a national database. With respect to understanding the Eastern Cape, I am not sure that this would have much impact on the study community because the study community didn't exist prior to about 1947. In the early 1960s, about half of the population were forced to settle in Ravensmead by the Group Areas act. Therefore, we have a young study community comprising an ethnic group that probably had little contact, if any, with people from the Eastern Cape.

Fourie: This confirms my own feeling that you're looking at active transmission and not reactivation disease. You're looking at strains being introduced all the time, and you don't know where they're coming from.

Fine: I find your interpretation of the apparently low proportion of reactivation disease in this study community to be intuitively unreasonable. National databases may help, but some basic descriptive epidemiology would be more helpful. I would guess that you would see some interesting trends if you broke down the community simply by age, by residence time and by migration history.

vanHelden: An anthropologist is currently looking into the history of the area, so we are trying to get a handle on this information.

Bateman: I would like to address the question of reinfection in a patient who already has tuberculosis. There are now examples of multidrug resistance arising in patients who are admitted to hospital with sensitive M. tuberculosis, and who pick up other strains in the hospitals. It's not difficult to imagine that the same situation occurs in the community. I suspect that the full picture will not be obtained from setting up a national database. The answers will come from looking at the microenvironment: at person-to-person transmission within limited geographic, domestic or work-place environments.

Hopewell: In the study by Small et al (1993), those patients who went to hospital and acquired a drug-resistant organism, went to hospital because they had endstage AIDS, and so were severely immunocompromised.

Bateman: I know of three other cases, one from the Western Cape and two from another province, where multidrug-resistant strains were acquired in hospital. I don't know if any of these patients had AIDS.

0 0

Post a comment