INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bovine tuberculosis (TB) was endemic in France until the 1960s, with herd prevalence rates of 25% in 1955 (9). From this time onwards, a national program for TB control based on tuberculin skin testing with control of animal movements and total slaughter of infected herds was implemented. This control strategy resulted in a dramatic decrease in bovine tuberculosis leading to a herd prevalence rate of 0.09% in 1998 (2), suggesting that cattle are the most important reservoir, or even the sole reservoir, for Mycobacterium bovis in France. Due to the success of this control strategy, France was declared "officially free of bovine TB" by the European Commission (3).

The very low level of TB in cattle has resulted in the introduction of new control strategies. Consequently, there has been a progressive reduction in the use of skin testing, with an increasing emphasis on systematic sampling of suspect lesions identified at slaughterhouses for M. bovis isolate identification and molecular typing. New laboratory tools were therefore required in order to improve the traceability of the infections and identification of the origin of the outbreak (i.e., persistence in a herd, introduction of new animals from infected herds, or contamination from a neighboring infected herd).

Many new molecular techniques have been developed over recent years to aid the differentiation of isolates belonging to the Mycobacterium tuberculosis complex, which includes M. bovis. Among these techniques, the most widely used are spoligotyping (26), restriction fragment length polymorphism (RFLP) with different probes (IS6110, direct repeat [DR], and polymorphic guanine-cytosine-rich sequence [PGRS]), and variable number of tandem repeat (VNTR) typing (18, 19), which is presently considered a promising technique (28). RFLP with the IS6110 probe has proved to be very useful for the discrimination of M. tuberculosis isolates, which generally harbor a high copy number of this insertion sequence (IS) (21, 25, 39). However, for M. tuberculosis isolates with a low copy number of IS6110 (7, 22) and for the majority of M. bovis isolates from cattle which present only one copy of this IS (5, 6, 13, 26, 28, 37), spoligotyping has been shown to be more discriminating than IS61110 RFLP. DR-RFLP, which detects polymorphism within the same region of the chromosome as spoligotyping, is equally discriminating. PGRS-RFLP is considered to be the most discriminatory of these RFLP techniques for M. bovis (5, 12, 44, 45). But for all the RFLP techniques, DNA extraction is required. In addition, the technique itself is time-consuming and technically demanding, especially for PGRS-RFLP, and problems of reproducibility have been reported (15).

In contrast, spoligotyping is a rapid, simple, and reproducible technique which can be performed on cell lysates and even on clinical specimens (21, 34, 36). It is based on the polymorphism of a region called DR (23). The DR region is unique to bacteria belonging to the M. tuberculosis complex (4, 20, 23, 26, 42) and is constantly present in this group of mycobacteria (26, 30).

In the current protocols, spoligotyping involves the simultaneous detection of the presence or absence of 43 unique short DNA sequences (35 to 41 bp) called spacers. Spoligotyping is considered a very useful technique, at least at the level of a first screening (26, 35), and the results are produced in an intrinsically qualitative form, as the response for each spacer is either present or absent (15).

For these reasons, the first molecular typing technique applied at Agence Française de Sécurité Sanitaire des Aliments (Afssa) (formerly Centre National d'Eudes Vétérinaires et Alimentaires [CNEVA]), for M. bovis typing was spoligotyping. Spoligotyping of the majority of M. bovis isolates in the Afssa collection was performed in order to evaluate the biodiversity and distribution, in space and in time, of M. bovis populations isolated or identified in France. This paper presents our results for 1,349 isolates obtained from 1979 to August 2000.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mycobacterial isolates and strains. (i) M. bovis isolates. Of the 1,349 M. bovis isolates obtained in this study from 1979 to August 2000, 1,266 were from France and 83 were from foreign countries or the French West Indies. Table 1 shows the number of isolates from France for which a spoligotype was obtained compared with the total number of isolates from France that were identified as M. bovis at Afssa by bacteriological methods. No isolates were collected during the years 1980 to 1982. In addition, for some years, like 1984 or 1986, the proportion of isolates available for spoligotyping was low.

View larger version (57K): [in this window] [in a new window]   FIG. 1.   Geographical distribution of the French isolates and spoligotypes. (Left) Number of isolates and number of spoligotypes per department. For each department, the shade of grey indicates the number of isolates for which a spoligotype was identified; the number on the map indicates the number of spoligotypes found in that region. (Right) Dominant spoligotype per department. "None" corresponds to departments where two or more spoligotypes were codominant (i.e., were represented by the same number of isolates).

(ii) Reference strains. The reference strains used in this study were M. tuberculosis H37Rv and M. bovis BCG P3 or ATCC 19210.

Spoligotyping. The protocol developed by Kamberbeek et al. (26), as described by Aranaz et al. (6), was used for the spoligotyping of M. bovis isolates. Biodyne C membranes (Pall Gelman, Easthill, N.Y.) were prepared for reverse line blotting by the addition of 43 aminolink oligonucleotides by following the procedures recommended by the manufacturer of the oligonucleotides (Isogen Bioscience, Maarssen, The Netherlands). The isolates were tested as cell lysates. For the reference strains, we usually used purified DNA extracted according the technique described by Wilson et al. (43). In some cases, we used cell lysates obtained by heat treatment. The reference strains were systematically included in each hybridization assay.

Data processing. The data were collected in Microsoft Access. The maps were constructed using the MapInfo geographical information system (Claritas ADDE, Boulogne, France).

Classification methods. As discussed by Brittain et al. (D. Brittain, R. A. Skuce, and S. D. Neill, Abstr. 5th Int. Conf. Mycobacterium bovis, abstr., 2000), the "direction of evolution" of the DR region is very probably unidirectional, i.e., it occurs only by deletion of spacer regions, with the lack of contiguous spacers resulting from a succession of events or from a unique event. The objective of classification methods is to define groups and for molecular epidemiology studies, such groups should be stable over time. This requirement is better satisfied if the definition of a group is grounded on evolutionary arguments (16). Therefore, we have combined a classical cluster analysis method (which aim is only to identify groups of isolates on the basis of the degree of similarity of the DR area and does not allow any phylogenetic interpretation from the data), with a phylogenetic analysis using a parsimony-based method, which takes into account the unidirectional way of evolution of the DR region. The phylogram obtained using these methods can then be used to evaluate the significance (in evolutionary terms) of the groups defined using cluster analysis. Cluster analysis was conducted using the program NEIGHBOR from the PHYLIP package (17) with the option UPGMA, the distance matrix being computed using Dice coefficients. Phylogenetic reconstruction was conducted using the PAUP* software package (D. L. Swofford, Phylogenetic Analysis Using Parsimony [*and other methods], version 4; Sinauer Associates, Sunderland, Mass.), the lack of a spacer being considered a character. All changes were weighted equally and all character changes were considered irreversible (Irrev.Up), with the presence of a spacer being considered plesiomorphic. The heuristic method was selected with random sequences additions (100 replicates) and TBR branch swapping (options: DELTRAN optimization, MULPARS, and COLLAPSE 0-length branches). The trees were rooted with a BCG-like spoligotype (which presents the maximum number of spacers, i.e., 35 spacers) as an outgroup as a consequence of the irreversibility of the transformation of the character's state from presence to absence.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Global results. We adopted the following nomenclature to name the different profiles obtained. In general, the letter F (for France) was used followed by a number, which was different for each different spoligotype obtained. Two exceptions were introduced: first, we used GB (for Great Britain, United Kingdom) plus a number when the profiles had been previously described in Great Britain; second, the term BCG-like was adopted for field isolates with a profile identical to the spoligotype of the vaccine M. bovis BCG strains.

View larger version (59K): [in this window] [in a new window]   FIG. 2.   Spoligotype phenogram obtained from 1,349 M. bovis isolates and corresponding patterns. For each spoligotype, the spacer deletions are indicated. Because all spoligotypes share the BCG-like spoligotype deletions (spacers 3, 9, 16, 39 to 43), these deletions are not included in each spoligotype description. Therefore, for example, the AN5 description (5, 11-12, 22) indicates the deletions of spacers 3, 5, 9, 11 to 12, 16, 22, and 39 to 43. The phenogram was generated using the UPGMA clustering method, with the between-spoligotypes distances being Dice coefficients. The scale indicates dissimilarity proportions. Due to the length of the phylogram, it has been fractionated into two parts, the upper right part representing the continuation of the lower left part.

Spatial distribution. Figure 1, left, shows the number of spoligotypes found in each of the metropolitan French departments (French administrative subdivisions), and Fig. 1, right, shows the most frequent spoligotype per department. The most prevalent spoligotype, BCG-like, was widely distributed in France, with an apparent predilection for the southeastern areas, where it was the predominant spoligotype in most departments (Fig. 1, right). The next most prevalent spoligotype, GB54, was also present in many regions, as was GB35 (the next most frequently observed spoligotype). Initially, GB54 appeared to have been restricted to the north and northwest regions of France, and then it appeared to have extended into other regions of the country over time (data not shown). The majority of the spoligotyping patterns that have been observed in both Great Britain and in France were mainly restricted to northern French departments (Fig. 1, right). Conversely, the fourth most frequent spoligotype, F004, was isolated mostly in southern France. Some isolates were obtained from cattle imported into France or from samples sent from the French West Indies or from foreign countries. As shown in Table 2, the majority of the spoligotypes of these isolates were also observed in metropolitan French isolates (20 out of 28).

Evolution with time. The number of spoligotypes apparently increased with time, but this increase appeared to be related to the total number of isolates that were tested. There is a strong correlation between the number of isolates per year and the number of spoligotypes per year (Table 1). This was observed for most of the departments (Fig. 1, left), as departments from which many isolates were obtained were also those where spoligotype diversity was highest. Analysis of the change in frequency of the four most prevalent spoligotypes with time (data not shown) revealed that there was a slight decrease in the frequency of the spoligotype BCG-like and a corresponding increase in the frequency of GB54 over time.

Distribution by animal species. All spoligotypes were isolated from cattle (Table 3). However, striking differences in species distribution were observed for spoligotypes GB54 and F040. Spoligotype GB54 was isolated from a wide range of animal species. Conversely, almost all isolates harboring the spoligotypes F040 were isolated from goats (14 out of 16).

Cluster analysis. The distribution of the spoligotypes into groups by using a clustering method (UPGMA) is presented in Fig. 2. This phenogram shows that, apart from one minor group, the majority of the spoligotypes can be included in one major BCG-like group of similarity, which contains the BCG-like spoligotype. This highlights the high degree of homogeneity of the DR region among the isolates in our study.

Phylogenetic analysis. PAUP* did find four most parsimonious trees 312 steps long (CI = 0.11, RI = 0.70). The strict consensus of these trees (Fig. 3) was not resolved for nearly two-thirds of types, but in the remaining types, some subgroups were noticeable. For example, the F009 subgroup was found by this method (Fig. 4B1). Moreover, a detailed analysis of the four most parsimonious trees (data not shown), in conjunction with epidemiological data (i.e., the geographical distribution of the spoligotypes), allowed identification of a further subgroup, the F015 subgroup, composed of the F015, F100, F145, F061, and F023 spoligotypes. By assuming that the primary evolutionary event was the loss of one spacer (or of a block of contiguous spacers), we could identify a plausible evolutionary path for the F015 subgroups (Fig. 4B2). In this pathway, new strains were created by successive deletions occurring at distinct spacer regions from the parental state (F015), which harbors two deletion sites distinct from those of the BCG-like strain. The geographical distribution of the spoligotypes belonging to the F015 group was restricted to the South of France. A careful examination of the spoligotyping profiles and of the geographical distribution of the corresponding isolates (Fig. 4C2) allowed us to include two other spoligotypes, F007 and F041, in this subgroup. This was done on the basis of phylogenetic evidence (both spoligotypes share two distinct deletions with F015 and harbor an extra deletion at distinct sites) and of epidemiological evidence (both spoligotypes share the same geographical distribution as the one of the phylogram-derived F015 subgroups). Finally, the F015 spoligotype deletions include the loss of spacer 33. F004, which differs from BCG-like by the loss of spacer 33, is the fourth most common spoligotype in France; as for the other members of the F015 group, they were geographically restricted to southern French regions (Fig. 1, right). This suggests that the whole F015 subgroup could have evolved from the F004 spoligotype. Three of the four trees support divergence of the F004 branch just before that of the F015 subgroup.

View larger version (23K): [in this window] [in a new window]   FIG. 3.   Spoligotype phylogram obtained from 1,349 M. bovis isolates. Due to the length of the phylogram, it has been fractionated into two parts, the upper right part representing the continuation of the lower left part. For each spoligotype, the spacers' deletions can be found in the phenogram (Fig. 2). The phylogram was generated using PAUP*.

View larger version (23K): [in this window] [in a new window]   FIG. 4.   F009 and F015 groups. Trees for the F009 group (A1) and the F015 group (A2) were obtained using cluster analysis as done for Fig. 2. (B1 and B2) Trees were obtained using PAUP*. (B1) The tree for the F009 group was done as the one for Fig. 3. (B2) F015 group. The proposed evolutionary tree was derived from the phylogram (Fig. 3) and from geographical distribution. The parental spoligotype is indicated in plain letters (excluding typical M. bovis deletions); successive deletions are indicated in italics. (C1 and C2) Geographical distributions of the isolates of the F009 group (C1) and the F015 group (C2) are shown. Each dot corresponds to one isolate randomly positioned inside the department from which the isolate originated.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. It is difficult to know if our sample, for which spoligotyping has been possible, taken globally, is representative of the population of M. bovis in France. However, apart from some years (like 1984 and 1986), it is representative of the M. bovis population isolated at Afssa. From 1988, the proportion of M. bovis isolates which could be typed increased, and the population used for this study is representative of the population of M. bovis strains isolated at Afssa.

Spoligotyping technique. Our results confirm that spoligotyping is a very easy and rapid technique and does not require DNA extraction. For all our isolates, we used cell lysates. Our work has also allowed us to confirm the repeatability of the test, as we tested 183 isolates (13.5%) at least two times, with the same spoligotype obtained in all cases.

Classification methods. Parsimony-based methods assume that each character (the presence or absence of each spacer) evolves independently (the elementary evolutionary event here being the loss of one spacer). This is of course not true if we consider that spacers may be lost directly by blocks. Therefore, before drawing conclusions, we evaluated the robustness of both phenogram- and phylogram-derived groups by using the geographical location of the spoligotypes (a given group being more or less restricted to a given area). In the case of the F015 subgroup, such an analysis led us to exclude one spoligotype of the phenogram-derived group and to add two extra spoligotypes to the phylogram-derived F015 subgroup (Fig. 4B2 and C2).

Spatial distribution. For the most prevalent spoligotypes, like BCG-like, GB54, and even GB35, which have a rather wide distribution, there is obviously a need for additional techniques based on independent markers in order to allow epidemiological studies. In the case of frequent spoligotypes that have a limited geographical extension, like F004, the appearance of isolates with this spoligotype in areas from where they were absent can to a certain extent help to trace the origin of the infection. But the possibility of a switch from BCG-like, still to be demonstrated, could limit the value of this apparent "geographical advantage." The precise identification of a close association between an area and a spoligotype could be particularly helpful for epidemiological purposes. Such a geographical clustering has already been described in England for some spoligotypes (16). Our data show for example a very close association between F041 and Department 47 (Lot-et-Garonne) and between F064 and Department 16 (Charente).

Comparison with spoligotypes obtained in other countries. The absence of a bank of spoligotype data makes it difficult to compare our profiles with those obtained in other countries. However, the spoligotypes in France are rather different from those obtained in Great Britain, as only 10 spoligotypes out of 161 were also observed in Great Britain. This represents only 6.2% of the spoligotypes observed in France and approximately 16% of the spoligotypes known in Great Britain. The predominant spoligotype in Great Britain, corresponding to our GB09 spoligotype, is present in France at a very low level (0.7% of the isolates, with only 11 cases). Conversely, it is dominant in countries that have had traditional links with Great Britain, such as Australia, Canada, Iran (12), the Republic of Ireland (11), and countries from South America, especially Argentina (45).

Animal species. The case of GB54 suggests that isolates with this spoligotype may have a wider host range than isolates with different spoligotypes. This is suggested by the fact that even if the prevalence of BCG-like isolates is higher in France, this spoligotype is less frequent in the cattle-plus-goats group than in groups of the other animal species, compared to the prevalence of GB54, with a very significant difference (² = 54, P < 0.001). This difference was still highly significant (² = 30.5, P < 0.001) when we removed the isolates belonging to known outbreaks in order to suppress the possible bias linked to an overestimation of one category of species.

Clustering and phylogenetic results. The most striking feature of the spoligotype profiles of M. bovis isolates in France is the high level of diversity of the spoligotypes, which is associated with a high degree of homogeneity of the main group designed by clustering. This is illustrated by two elements. First, the total number of spoligotypes is very high, relative to the total number of typed isolates, when compared to many other locations, especially islands like Great Britain (10) or Australia (12) and developing countries like Cameroon (31) or Tanzania (27). Second, the frequency of the two main spoligotypes (BCG-like and GB54) is low compared to studies in other countries. For example, the two main spoligotypes in Great Britain account for 70% of the isolates (10; R. G. Hewinson, unpublished data). The most frequent spoligotype was found in 52% of the isolates in the Republic of Ireland (11), in 39.3% (Ferretti, personal communication) to 66.6% (37) of the isolates in Italy, and in 46% of cattle isolates in Spain (6). In a study conducted with 273 isolates from Australia, Canada, and Iran, 88% of the isolates belonged to the same spoligotype (12).