HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via Google Scholar Google Scholar Articles by Dale, J. W Articles by Pitman, R. Search for Related Content PubMed PubMed Citation Articles by Dale, J. W Articles by Pitman, R. Agricola Articles by Dale, J. W Articles by Pitman, R.

Origins and properties of Mycobacterium tuberculosis isolates in London

¹School of Biomedical and Molecular Sciences, University of Surrey, Guildford GU2 7XH, UK ²NE London TB Network, Department of Respiratory Medicine, Homerton University Hospital, Homerton Row, London E9 6SR, UK ³HPA Mycobacterial Reference Unit, Guy's, Kings and St Thomas’ School of Medicine, London SE22 8QF, UK ⁴Centre for Medical Microbiology, Royal Free and University College Medical School, London NW3 2PF, UK ⁵Mathematical Modelling and Economics Unit, HPA Centre for Infections, London NW9 5EQ, UK

Correspondence Jeremy W. Dale j.dale{at}surrey.ac.uk

Received November 12, 2004
Accepted February 21, 2005

Using similarities of IS6110 banding patterns, isolates of Mycobacterium tuberculosis from a population-based study in London were assigned to 12 large groups termed ‘superfamilies’ (sfams). Analysis of patient data showed a marked geographical association in the distribution of these sfams. In particular, isolates from patients born in Europe were from different sfams than those born elsewhere, indicating that there had been relatively little transmission of tuberculosis in London from immigrant communities into the endogenous population. Multivariate analysis showed that certain sfams were significantly associated with pulmonary rather than extrapulmonary disease, or with sputum smear negativity, independently of country of birth or ethnicity, suggesting that the properties of the infecting organism play a role in the nature of the disease process.

	INTRODUCTION

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
The use of IS6110 for differentiating isolates of Mycobacterium tuberculosis has proved invaluable in analysing the epidemiology of tuberculosis (TB), especially in outbreak investigations where identical IS6110 patterns can be used to infer a potential epidemiological connection. In population-based studies, the main focus has been on the use of ‘molecular clustering’ (identical IS6110 patterns) to provide an estimate of the level of active transmission, as opposed to reactivation of long-standing infection, on the assumption that cases arising from reactivation are likely to be more diverse. Less attention has been paid to the relationships between isolates showing IS6110 patterns that are similar but not identical. However, some groups (variously referred to as families, lineages or clades) of strains with related patterns have been identified (Kremer et al., 1999), most notably the Beijing family (Van Soolingen et al., 1995) which is especially prevalent in East Asia, and has been claimed to be spreading worldwide, as well as being associated with the development of multi-drug resistance (Bifani et al., 1996; Rad et al., 2003; Borgdorff et al., 2003) and a differential interaction with the host immune system (Lopez et al., 2003; Kremer et al., 2004; Manca et al., 2004). The tendency for such families to show a degree of concordance in independent typing methods (McHugh et al., 2005) not only supports their evolutionary relatedness but also suggests that there may be additional biological and clinical associations that merit further investigation.

The ethnic diversity of the population of London provides a unique opportunity for the study of worldwide TB, and especially for the geographical origins of various strains of M. tuberculosis. From a multi-centre collaborative study of TB in London from 1995 to 1997 (Maguire et al., 2002), IS6110 RFLP patterns, and associated patient data, were available for 2490 isolates. Retrospective analysis of these data enabled the assignment of nearly 90 % of the isolates to one of 12 large groups, with related RFLP patterns, which we term ‘superfamilies’ (sfams), to emphasize the broader nature of these groupings. This paper reports the characteristics of these sfams and analyses their origins and biological properties.

	METHODS

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
Source of isolates.

The M. tuberculosis isolates for this study, and associated patient data, were collected for a population-based study of the epidemiology of TB in London between 1 July 1995 and 31 December 1997 (Maguire et al., 2002).

Molecular typing methods.

IS6110 typing was performed, as described by Maguire et al. (2002), by the international standard protocol (van Embden et al., 1993), in which M. tuberculosis DNA is cut with PvuII and Southern blots are hybridized with a probe from the right-hand portion of IS6110. Blots were normalized with the standard M. tuberculosis reference strain 14323. Typing results were analysed and compared using GelCompar and Bionumerics software (Applied Maths), using the Dice coefficient and UPGMA as described previously (Maguire et al., 2002), with a band tolerance level of 1.2 % and optimization of 1 %. The results of other typing methods for the low copy isolates (spoligotyping and polymorphic GC-rich sequence typing) were taken from Dale et al. (2003).

	RESULTS

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
Using IS6110 RFLP analysis, fingerprints were obtained from single isolates from 2490 patients in London during the study period (1 July 1995 to 31 December 1997). Of these, 448 (18 %) were low copy number strains, i.e. they had fewer than five IS6110 bands (Maguire et al., 2002). The multi-copy isolates were initially assigned, empirically, to families with 50 % or more similarity in the dendrogram reported by Bionumerics, ignoring any groups with less than 10 members. The outcome was verified by bootstrapping, using random samples (10 or 20 %) of the database, rejecting groups that failed to cluster and reassigning some isolates manually on the basis of visual inspection. Repeated rounds of verification and reassignment resulted in assignment of nine large groups, termed ‘sfams'. The robustness of these assignments was tested by repeated clustering of random samples of the database (10 %, 20 %, 40 %), assigning groups (blind to the original sfam designation) at 50 % similarity. With the 40 % samples, all the groups consisted of more than 80 % of a single sfam, and the majority were over 90 %. With the smaller samples, some of the groups were too small to be reliable, but in all cases they were still predominantly (>70 %) of a single sfam type. The overall assignments of sfams were consistent in 83–92 % of cases.

This procedure was not appropriate for the low copy isolates. These were assigned previously (Dale et al., 2003) to three groups, I–III, on the basis of the IS6110 insertion site, spoligotype and polymorphic GC-rich sequence patterns. Subsequent comparison of these groups with the high copy sfams led to the reassignment of some 4-banded isolates (especially type IN5030; Dale et al., 2003) to the multi-copy sfam9, resulting in three low copy sfams designated L1–L3. In the final assignment, 1803 (88 %) of the multi-copy isolates and 394 (88 %) of the low copy isolates were assigned to sfams. The remainder of the isolates fell into groups that were too small for reliable analysis by the methods described in this paper.

The number of isolates in each sfam and the range of band numbers are shown in Table 1, and a sample of each is displayed in Fig. 1. For comparison with previously characterized families (Kremer et al., 1999), the Haarlem family maps to sfam1, and the Africa and Beijing families to sfam5. The relationships between the low copy groups and other family designations have been described previously (Dale et al., 2003).

View larger version (22K): [in this window] [in a new window]   Fig. 1. Examples of IS6110 RFLP patterns for each sfam. The definition of the low copy families L1, L2 and L3 includes the identity of the insertion sites as well as the RFLP pattern. Note that the dendrogram shown is merely an illustration of the similarity between the banding patterns and does not represent a true phylogenetic tree.

Patient characteristics

Country of birth information was available for 1330 patients (53 %). Analysis of the data (Table 2) showed marked differences in distribution of sfams amongst patients originating from different parts of the world. In particular, patients born in Europe (predominantly the UK and Ireland) had a significantly (P < 0.01) greater than expected level of the related sfams sfam9 and L3, and of sfam7. sfam5, which includes both Beijing and African families, was amongst the predominant types in patients from East Asia and from East Africa. The strains predominant in patients from Bangladesh were different from those from India or Pakistan. There was little overlap between the different regions of Africa in the predominant strain types; there were too few isolates from patients born in southern Africa to include in this analysis.

Analysis by ethnic group (data not shown) produced similar results, reflecting the fact that most non-white patients were not European born. The major exception to this was the patients described as Black Caribbean; where data were available, 46 % of Black Caribbean patients were born in Europe. Amongst these patients, the distribution of the sfams was similar to that of the white patients, especially in the preponderance of sfam7, which was present in 30 % of the Black Caribbean patients.

Patients infected with sfam9 or L3 strains were significantly older, with median ages of 45 and 46, respectively, than other patients (overall median age = 35, P < 0.01; Table 3), and were especially prevalent in the over-60 age group (32 % and 29 % of each type, respectively, P < 0.01; data not shown). This is mainly due to the age distribution of the European-born patients (in whom sfam9/L3 were significantly more common). However, in a multivariate analysis with age and European birth, there was still an independent association with age (P < 0.05), suggesting that these might be relatively stable strains that have been in circulation longer and/or that they are more likely to cause TB through reactivation.

One other strain type showed a substantial difference in its age distribution: sfam2, with a median age of 29, with a statistically significant (P < 0.01) enhancement of prevalence in the 10–19 age group. This strain type was one of the prevalent types in patients from NE Africa; these patients had a median age of 28 years, suggesting that the age distribution of this sfam is a reflection of the relatively young age of the patients from those countries.

Much attention has been paid to the Beijing family of strains, and in particular the relatively young age of patients infected with these strains in some studies has been used as evidence that this family of strains is spreading worldwide (Anh et al., 2000). Amongst the subjects reported in this study, the median age of patients with sfam5 isolates (which includes the Beijing family) was 34, which is close to the overall median age (35). However, when the Beijing strains (n = 130) were separated from the others, there was a greater difference (median age = 30). This difference was only statistically significant (P < 0.01) for the 20–29 age group, and may simply reflect the relatively high proportion (>40 %) of patients born in East Asia who come into this age group; over 30 % of patients with Beijing isolates were born in East Asia. The proportion of Beijing isolates was actually lower than expected in patients under 20 and those between 30 and 59 years.

Biological differences

The data available in this study enabled an assessment of the possible relationship between strain type and two key disease parameters: the proportion of infections that were pulmonary or extrapulmonary, and the frequency of sputum smear positivity. Compared to the unclassified strains, several sfams (sfam1, sfam6, sfam7, L3 and, marginally, sfam5) were more likely to be associated with pulmonary rather than extrapulmonary disease (Table 4). A multivariate analysis, including age and ethnic group in the model, for 1531 isolates for which these data were available supported the independent association of sfam5, sfam6, sfam7 and L3 with pulmonary disease (Table 5). All ethnic groups other than Bangladeshi and Black Caribbean were more likely than the white patients to have extrapulmonary disease, as were the older patients (>50 years).

A similar multivariate analysis for the 534 cases where sputum smear results were available (Table 6) showed that sfam2 and sfam7 were independently associated with sputum smear negativity, as were the Indian and Chinese patients. There was no association with patient age, which was therefore not included in the model.

Some types of M. tuberculosis, especially the Beijing family, have been associated with enhanced levels of drug resistance. In this study, sfams L2, L3 and 1 were associated with resistance to isoniazid [odds ratio (OR) 2.7, 95 % CI 1.2–5.9], rifampicin (OR 3.6, 95 % CI 1.02–12.3) and streptomycin (OR 3.0, 95 % CI 1.5–6.1), respectively; multivariate analysis showed that these associations were independent of age, ethnicity or country of birth. sfam5 (which includes the Beijing strains) showed no significant association with drug resistance. A separate analysis of specifically Beijing strains showed a marginal association with streptomycin resistance (OR 1.9, 95 % CI 1.04–3.4), but not with other resistances (results not shown)

	DISCUSSION

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
The high discriminatory power of IS6110 typing is presumed to reflect the generation of variants at a rate that is relatively rapid, on an evolutionary time scale, but at the same time slow enough for epidemiologically linked isolates to show identical patterns. However, for certain strains, especially the low copy isolates, IS6110 typing shows much less discrimination, and therefore presumably these strains have a slower rate of variation for this marker. It is likely that there are also differences, although less marked, in the rate of variation amongst multi-copy isolates, and even from band to band in these isolates. These presumed differences in the rate of variation, added to the inevitable imprecision of assessing similarities in band patterns, are a limiting factor in any attempt to use the degree of similarity of IS6110 patterns to assess evolutionary relationships between strains. Nevertheless, the wealth of data that are available through the extensive use of IS6110 typing makes the attempt worthwhile. This is supported by the data reported by McHugh et al. (2005) showing a degree of concordance between relationships inferred from IS6110 similarity and those obtained by other typing methods. In this study, it was possible to assign about 90 % of the isolates to one of 12 large groups (sfams) – 9 containing multi-copy isolates and 3 containing low copy isolates. Superficially, this assignment of sfams, and the relationships amongst the low copy isolates as described previously (Dale et al., 2003), appear to be incompatible with the conclusions of Gutacker et al. (2002), based on single nucleotide polymorphisms (SNPs), that strains with different RFLP patterns may be clonally related and that high copy and low copy strains do not form genetically distinct populations. However, closer examination of the data suggests that the groupings themselves are not necessarily incompatible, provided that one does not attempt to infer true phylogenetic relationships from the IS6110 patterns alone. Further work will be needed to map the sfams onto groupings obtained by methods such as SNPs and deletion mapping.

A number of authors have speculated that genetic differences between strains influence the nature of the disease (Sreevatsan et al., 1997; Kremer et al., 1999; Kato-Maeda et al., 2001a; Tsolaki et al., 2004; Brosch et al., 2001). In particular, Kato-Maeda et al. (2001b) reported a strain-associated variation in the proportion of patients with pulmonary cavitations, and Fleischmann et al. (2002) showed an association of pulmonary/extrapulmonary disease and sputum smear positivity/negativity with strain characteristics defined by large sequence polymorphisms and SNPs. However, it is important to take account of the possible confounding effect of the association of strain type with ethnicity and/or country of birth of the patient. The multivariate analysis in this study showed that four sfams (sfam5, sfam6, sfam7 and the low copy group L3) were associated with pulmonary rather than extrapulmonary disease, independently of ethnicity or country of birth. This suggests a degree of tropism associated with these strains. Two sfams (sfam2, sfam7) were associated with sputum smear negativity. These results indicate that disease characteristics are in part determined by the properties of the bacteria.

Several studies have shown that strains of the Beijing family, which are included in sfam5, are associated with a higher frequency of drug resistance (Glynn et al., 1999; Rad et al., 2003). Some studies, notably in Vietnam (Anh et al., 2000), have also shown a higher proportion of Beijing isolates in younger patients, this effect being used as evidence that these strains are spreading outwards from their presumed focus in East Asia. We did not detect any clear evidence for either effect in this study. There was a marginal association with streptomycin resistance, but not with resistance to either isoniazid or rifampicin. Although the median age of patients with Beijing isolates was slightly lower than for non-Beijing isolates, the effect was limited to those in the 20–29 age group, which contained a high proportion of the patients who were born in East Asia; the proportion of Beijing isolates was lower than expected in patients under 20 years old. A more marked difference in age distribution was seen with sfam9 and L3 strains, which were especially prevalent in older patients. These strains were also more common amongst patients born in Europe, and the effect may therefore reflect the different age distribution of such patients. However, even amongst patients born in Europe, the median age of patients infected with these strains was higher than average, which suggests that these may be relatively stable strains that have been in circulation longer than others or that they are more likely to cause TB through reactivation.

The most dramatic differences between the sfams was seen in the analyses of the country of birth and ethnic group of the patients. In particular, about half the patients from India and Pakistan had strains belonging to sfam3, while patients from Bangladesh showed a marked difference in the preponderant strains, having few isolates of sfam3, and the majority being sfam4 (rare amongst patients from India or Pakistan) or unclassified. Strains of sfam3 were also amongst the predominant types in patients born in East Africa; over 60 % of these patients had strains belonging to sfams 3, 5 or 6. Amongst ethnically Indian patients born in East Africa, sfam3 was predominant (>40 %), but this sfam was also a major contributor to disease amongst Black African patients from all parts of Africa. European-born patients had a more diverse set of strains, but there was a significant association with sfam7, sfam9 and the low copy type L3 (which is closely related to sfam9). The association of L3/sfam9 with European-born patients has been reported previously (Dale et al., 2003). These types were all much less common in patients born outside Europe, a difference that was reinforced by comparing white patients to other ethnic groups. A characteristic global distribution of TB strains has been observed using spoligotype data (Filliol et al., 2003), and specific families of strains (defined by IS6110 similarities and/or spoligotype), such as the Beijing, Haarlem and Africa families, show considerable differences in prevalence in patients from different parts of the world (Kremer et al., 1999; Brudey et al., 2004). The results reported here demonstrate that IS6110 typing is capable of detecting and differentiating deep-rooted relationships in the overall spectrum of M. tuberculosis strains, extending far beyond the previously characterized families. The geographical origin of these sfams provides evidence of the historic routes of spread of the disease.

The marked difference in the prevalence of these strain types amongst different sectors of the population in London indicates that there had, at the time of the study, been little spread of TB from the immigrant communities to the endogenous population, which is consistent with our analysis of the clustering data for this study (Maguire et al., 2002). Similar observations, of limited transmission from foreign-born patients to USA-born individuals, have been reported from smaller-scale molecular cluster analyses in San Francisco (Chin et al., 1998; Jasmer et al., 1997), and more recently by Hirsh et al. (2004) using deletion analysis. Our results could reflect the likelihood of transmission having occurred in the country of origin prior to entry to the UK, or the closer contact that occurs within particular communities facilitating the spread of specific strains within those communities. However, we cannot exclude the possibility that, as previously suggested (Hirsh et al., 2004), adaptation of strains of M. tuberculosis to different host populations plays a role in restricting the spread of the organism between different groups.

	ACKNOWLEDGEMENTS

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
We are grateful for support from the NHS Executive London Research and Development Programme, and from the European Union under grants BMH4-CT97-91202 and SMT4-CT96-2097 (provision of GelCompar and Bionumerics software).

	REFERENCES

TOP INTRODUCTION METHODS RESULTS DISCUSSION ACKNOWLEDGEMENTS REFERENCES