Background
Each year, 9 million new cases of tuberculosis (TB) are recorded, of which 2 million result in fatality. Diagnostics, chemotherapy and vaccination are available, however, the disease is far from being eradicated
MTC organisms were also shown to evolve clonally
In this new study, we data-mined an updated SpolDB4 version, which contains 1939 STs representing a total of 39,295 clinical isolates originating from 122 countries. Considering the known diversity of the origin of patients, which was documented in some cases, SpolDB4 is representative of a total of 141 countries. The SpolDB4 update adds 1126 newly defined STs and provides a higher resolution picture of the worldwide MTC genome diversity assessed by spoligotyping. However, the new challenge is to link the genetic variability of MTC with the clinical variability of TB, whatever the setting, whether in high burden countries or in densely populated areas such as India or China.
Results
Classification of spoligotypes
Classification of SpolDB4 spoligotype patterns into meaningful lineages
The listing of ST alleles with their distribution by country of location and a presumed sub lineage/lineage label is provided in
Supplemental Table: SpolDB4 listing of all STs, binary description, octal description, distribution per country of isolation and/or of origin when available, clade/subclade label. Country names were chosen according to the ISO3166-three-letter format. "U" = unknown. Clade/subclade label using spoligotyping only should be taken as presumptive or indicative of a likely clade/subclade belonging but may in some case be misleading and requires in most cases further investigations to confirm the identity of a given isolate. In some instances, mixed patterns (unrecognized) did not unambiguously allow spoligotyping classification, hence an ambiguous final label in this table.
Click here for file
Bioinformatical (62 lineages/sub lineages prototype patterns) and statistical (50 most frequent) classification analysis of SpolDB4
Bioinformatical (62 lineages/sub lineages prototype patterns) and statistical (50 most frequent) classification analysis of SpolDB4. First column ST n°: Shared-type (ST) number of prototype pattern for the linage/sub lineage. Second column: lineage/sub lineage name. Third column: Binary spoligo display with black-white squares for respectively hybridizing-non-hybridizing spacers. Fourth column: Octal code (in red: defining octal rule). Fifth column: total absolute number of isolates of the subclass when variant ST spoligos are included (using SpolNet). Sixth column: same but expressed as percentage of total clustered isolates. * Total number and Frequency for these types are already included in their mother clade if known. Undesignated types are counted within the T1-ill-defined lineage. ** in red: octal rule defining the genotype.
Results obtained by the statistical approach
The 20 most frequent STs totalled 17,701 isolates (49.3% of the clustered isolates). The 50 most frequent STs increased clustering to 61.8% (n = 22,219). These 50 most frequent types are shown in figure
Results obtained by the mixed expert-based/bioinformatical approach
Figure
New genetic lineages within
SpolDB4 defines 62 genetic lineages/sub lineages (figure
New genetic lineages within
Among
Within the Haarlem (H) lineage, a 4th sub lineage (H4) is tentatively added. It is characterized by the absence of spacers 29–31 and 33–36 (prototypes ST127 and/or ST777). More than 60% of ST127 isolates are localized in Armenia, Austria, Finland, Georgia, Iran, and Russia. A likely related pattern (ST777) is found in Saudi Arabia. An hypothesis is that these strains could represent an intermediate genetic link between the previously defined Haarlem-1 (H1, ST 47) and Haarlem-3 (H3, ST50) genotypes. Type ST777, which shows a single spacer difference from ST 127, is also found in Kazakhstan, Russia, and Georgia (n = 26). These isolates are likely to be identical to the recently described "Ural" family of strains
Within the Latin-American-Mediterranean (LAM) lineage, we rebaptised the LAM7 sub lineage as LAM7-Turkey since recent results suggested that ST41 is predominant in Asia Minor
The X genotype family (X1-X3 sub lineages) was initially described thanks to data-mining
Comparison between the supervised and an unsupervised Naïve Bayes statistical approach of classification of spoligotypes
Data-mining and clustering techniques are the focus of intense research in Information and Bio Sciences
Unclassified spoligotypes
454 STs (23.4%) did not correspond to any pattern recognition rule when data were mined automatically using SpolNet. Nonetheless, we attributed a family label to these spoligotypes too by extending the rules defined initially, and subsequently by a visual assessment on each unclassified ST. Thus, 314 more STs were tentatively classified with a family label, and only 131 STs (7% of the total) remained unclassified. These STs harboured patterns with either important blocks of deletion, disseminated deletions, mixed signatures, or patterns, which at this time did not correspond to any known spoligotype-signatures.
Results
Population genetics
The worldwide distribution of data points was assessed within eight regions. Isolates from Europe and North America represented 65.5% of all entries. Africa, Far-East Asia, Middle East and Central Asia, and South America were equally represented (6.5% to 8.3% of entries) whereas Central America and Oceania were underrepresented (3.6% and 1.1% of entries). The overrepresentation of
Figure
Percentage of main spoligotyping-defined MTC genotype families within SpolDB4 (Beijing, Beijing-like, CAS, EAI, Haarlem, LAM, Manu, X, T), by studied continents and worldwide
Percentage of main spoligotyping-defined MTC genotype families within SpolDB4 (Beijing, Beijing-like, CAS, EAI, Haarlem, LAM, Manu, X, T), by studied continents and worldwide. Abbreviations : AFR = Africa, CAM = Central America, EUR = Europe, FEA = Far-East Asia, MECA = Middle-East and Central Asia, NAM = North America, OCE = Oceania, SAM = South America.
Synthesizing World Maps showing absolute (diameter) and percentage (colour) numbers of 3 genotype families within each country: Beijing; EAI (East-African Indian) CAS (Central Asia)
Synthesizing World Maps showing absolute (diameter) and percentage (colour) numbers of 3 genotype families within each country: Beijing; EAI (East-African Indian) CAS (Central Asia). These maps were built on an updated SpolDB4 on 2005 September 14th, on clusters of the 50 most frequent shared types as shown in Table 1, for a total of n = 17212 isolates (Beijing n = 4042, EAI n = 1684, CAS n = 1022).
The Beijing family of strains is prevalent in Far-East-Asia, but also in Middle-East-Central Asia and Oceania (45.9%, 16.5% and 17.2% respectively) (Figure
The CAS1-Delhi family is essentially localized in the Middle-East and Central Asia, more specifically in South-Asia, (21.2%), and preferentially in India (75%;
Lastly, the X family is highly prevalent in North America (21.5%) and Central American (11.9%) regions. It could be linked to an Anglo-Saxon ancestry, as it has been encountered in English-colonized areas such as in the United Kingdom, United States, Australia, South Africa and in the Caribbean
Analysis of the spreading and epidemiological status of MTC clones
We further analyzed the epidemic status of each spoligotyping-defined clone as shown by the index C1 and C2
A. Definition of the variables (MC, Ar, SI) used in SpolDB4 to define : (1) the geographic index C1 (Endemic, Localized, Ubiquitous) (2) the quantitative index C2 (Epidemic, Common, Recurrent, Rare). B. Distribution of the 1939 Shared-types in 12 classes.
Name
Abbreviation
Type of data, Def.
Rules for Definition of Qualifiers (C1 and C2)
A Definitions
Matching Code
MC
1–8 digits, built by linking region codes
If 1 digit, then C1 = Endemic (genotype found in one macroregion only)
If 2 digits, then C1 = Localized (genotype found in two macroregions)
If ≥ 3 digits, go to Area section below for further interpretation
Area
Ar
numeric, n° of countries in which a given SIT is found
If MC ≥ 3 digits and Areas ≤ 5; C1 = Localized
If MC ≥ 3 digits and Areas ≥ 6; C1 = Ubiquitous (genotype found in more than three macroregions)
Spreading Index
SI
numeric, mean indicator of spreading independent of geography SI = n/Ar where n is the absolute value for a given shared-type
If SI ≥ 25; C2 = Epidemic
If 10 < SI < 24; C2 = Common
If 3 < SI < 9; C2 = Recurrent
If SI ≤ 2; C2 = Rare
B Distribution of the 1939 genotypes in 12 class
Endemic Epidemic
6
Endemic Recurrent
286
Localized Epidemic
2
Localized Recurrent
207
Ubiquitous Epidemic
6
Ubiquitous Recurrent
176
Endemic Common
31
Endemic Rare
501
Localized Common
14
Localized Rare
587
Ubiquitous Common
20
Ubiquitous Rare
82
NA*
21
Total
1939
*not applicable
Analysis of the geographical Inter and Intra-continental matches between the shared-types found within 1 or 2 settings (endemic types n = 824) and localized types (n = 564), within and between the 8 macro-regions. The geographical analysis of the matches between localized types found in 3 macroregions and 5 settings or more (n = 246) and between the ubiquitous types (found in more than 3 regions (n = 284) was not done. Number of intra and intercontinental matches between STs detected between the 8 macro regions previously defined.
Macroregion (code)
Africa
Americas
Europe
Asia
Oceania
North America
Central America/Caribbean
South America
Middle-East and Central-Asia
Far-East Asia
Africa (1)
73
29
2
7
79
7
3
0
North America (2)
138
8
35
131
34
45
4
Central America (3)
7
0
9
1
2
0
South America (4)
94
69
ND
0
3
Europe (5)
351
56
25
3
Middle-East Asia (6)
99
8
0
Far-East Asia (7)
61
4
Oceania (8)
1
ND = not done
Toward an interactive system of geographic atlas of genotype frequency data of
Figure
Synthesizing World Maps showing absolute (diameter) and percentage (colour) numbers of 3 genotype families within each country:
Synthesizing World Maps showing absolute (diameter) and percentage (colour) numbers of 3 genotype families within each country:
Discussion
In this study, we data-mined an updated international spoligotype database of the
The scaling-up that represents SpolDB4 relatively to SpolDB3 (4×), allowed new sub lineages to be discovered. However, it also showed the limit of the approach of using spoligotyping only to define the precise identity of a given MTC clone, since over fitting was observed. Combined DR, MLVA, SNPs, Region of Differences polymorphic datasets are now required to improve our knowledge of MTC genomes diversity.
Today's observed pattern of phylogeographical diversity of MTC is undoubtedly the result of both a deep ecological differentiation and of a more recent demographic and epidemic history. It is tempting to speculate, especially since the publication of the recent studies done on
Such an endeavour as the building of SpolDB4, and more generally the building of a representative genetic diversity database, should however assume limitations concerning both the quality and representativity of data. We partially eliminated the first problems by carefully double checking many datasets visually, reinterpreting other datasets by asking for the autoradiography results sent as electronic files, or by simply excluding datasets harbouring systematic genotyping errors. The procedure of examining only STs improves data quality by minimizing artefacts analysis. Some internationally agreed upon recommendations to improve the quality of spoligotyping are also in progress and will be published elsewhere. The second problem (representativity) is more chronic and difficult to solve. More financial means should be devoted to improving both mycobacteriology diagnostic and genotyping facilities in countries where the outbreaks prevail. Another limitation using spoligotyping is the study of mixed populations of bacilli
One of the major lessons from this collaborative effort is that new markers such as MLVA, SNPs or others are needed to improve our knowledge on the population structure of MTC. MLVA provides an improved knowledge of MTC clonal complexes, a strategy for the MTC which could even be more effective than the Multi-Locus-Sequence-Typing (MLST) approach
Our efforts to characterize the classification of the tuberculosis strains present within populations have also led to an increased understanding of their global distribution. Our results do not fully agree with recently published results
Linking these results for a clinical benefit for individual patients remains another challenge. In particular, we need to better understand why in certain areas a small number of strains are causing a disproportionate number of cases of the disease and we need to better understand the effects of host's genetic and environmental variability on the presentation of the disease
Conclusion
The SpolDB4 database is by far one of the largest publicly available database on
Methods
Database
A total of 39,295 entries were collected in an Access® database. The Previous STs numbered 1 to 817 were previously described in SpolDB3. ST633, 714, 729, 770, 797, which were missing in the former version were replaced by new genotypes. SpolDB4 is available [see
The distribution by continent and country of isolation is shown below. Macro region name, defining acronym, identification number and total number of patterns (into brackets) are respectively: Africa AFR -1 (n = 3121), Central-America CAM -2 (n = 1353), Europe EUR -3 (n = 13624), Far-East Asia FEA -4 (n = 2624), Middle East and Central Asia MECA -5 (n = 2639), North-America NAM -6 (n = 9153), Oceania OCE -7 (n = 235), South-America SAM -8 (n = 3176).
Origin of spoligotypes
The spoligotypes were either obtained at the Pasteur Institute of Guadeloupe using the 43-spacers format and home-made membranes
Description
The description of SpolDB4, per country of isolation of the clinical isolates, or in very rare cases, by country of origin of the patients, is as follows.
AFR, (n = 3121): Angola (n = 4), Burundi (n = 18), Benin (n = 4), Botswana (n = 1), Burkina Faso (n = 1), Centrafrican republic (n = 154), Ivory Coast (n = 84), Cameroon (n = 498), Congo Democratic (former Zaire) (n = 7), Congo (n = 20), Djibouti (n = 5), Algeria (n = 137), Egypt (n = 60), Eritrea (n = 8), Ethiopia (n = 153), Gabon (n = 2), Ghana (n = 1), Guinea (n = 7), Gambia (n = 1), Guinea-Bissau (n = 217), Kenya (n = 71), Libya (n = 54), Morocco (n = 127), Mali (n = 5), Mozambique (n = 28), Mauritania (n = 4), Malawi (n = 122), Namibia (n = 79), Niger (n = 1), Nigeria (n = 6), Rwanda (n = 8), Sudan (n = 41), Senegal (n = 69), Sierra Leone (n = 3), Somalia (n = 226), Tunisia (n = 10), Tanzania (n = 9), Uganda (n = 11), South Africa (n = 564), Zambia (n = 43), Zimbabwe (n = 247), East-Africa (not specified, n = 15).
CAM, (n = 1353): Netherlands Antilles (n = 1), Barbados (n = 6), Costa Rica (n = 1) Cuba (n = 239), Commonwealth of Dominica (n = 1), Guadeloupe (n = 232), Honduras (n = 7), Haiti (n = 375), Mexico (n = 386), Martinique (n = 105).
EUR, (n = 13624): Albania (n = 5), Austria (n = 1456), Belgium (n = 517), Bulgaria (n = 2), Bosnia and Herzegovina (n = 3), Czech Republic (n = 393), Germany (n = 574), Denmark (n = 281), Spain (n = 374), Estonia (n = 115), Finland (n = 347), France (n = 2265), United-Kingdom (n = 1523), Greece (n = 1), Hungary (n = 62), Ireland (n = 1559*, mainly
FEA, (n = 2624): China (n = 145), Indonesia (n = 344), Japan (n = 138), Korea (n = 4), Myanmar (n = 20), Mongolia (n = 19), Malaysia (n = 598), Philippines (n = 237), Singapore (n = 4), Thailand (n = 302), Vietnam (n = 789), Far-east-Asia unspecified (n = 24).
MECA, (n = 2639): Afghanistan (n = 3), Armenia (n = 119) Azerbaijan (n = 71), Bangladesh (n = 676), Comoro Islands (n = 14), Georgia (n = 272), India (n = 483), Iran (n = 110), Iraq (n = 5), Israel (n = 15), Kazakhstan (n = 55), Lebanon (n = 2), Sri Lanka (n = 16), Madagascar (n = 395), Mauritius (n = 21), Nepal (n = 6), Pakistan (n = 90), Reunion Island (n = 16), Saudi Arabia (n = 99), Turkey (n = 170), Yemen (n = 1).
NAM, (n = 9153): Canada (n = 266), Greenland (n = 4), USA unspecified (n = 1690), USA (Alabama, n = 3), USA (New-York, n = 5948), USA (Texas, n = 1242).
OCE, (n = 235): Australia (n = 36), New Zealand (n = 151), French Polynesia (n = 2), USA (Hawaii, n = 46).
SAM, (n = 3176): Argentina (n = 1150*, mainly
Data format
All spoligotypes were converted into the octal format within Excel spreadsheets
Combined automatized-expert based classification of spoligotypes
To be classified by SpolNet, spoligotypes must be expressed under the form of binary vectors of 43 bits. Using clade and Principal Genetic Groups (PGG) clustering
Secondly, an algorithm establishes the hierarchic links from the entry data file. This algorithm is based on a comparison of vector. The model (assumption) relies on an evolution that proceeds by deletion of a unique block of one to
Definition of indices
The global distribution of spoligotyping patterns was assessed within and between the eight studied continental regions: Africa, Central America, Europe, Far East Asia, Middle East and Central Asia, North America, Oceania, and South America. A slight modification in definition of C1 and C2 was introduced, i.e. a spreading Index SI ≥ 25 instead of SI ≥ 30 is now required to define a clone as "epidemic". Briefly, the Spreading Index (SI) represents a mean number of occurences of a clone, independent of the setting (total n° of occurrences for this ST divided by the number of geographical areas where it is found). The reader should refer to Figure
Availability
SpolDB4 listing [see
Quality control (QC)
During the SpolDB4 project, we faced increased quality control problems due to an increased recruitment rate and had to exclude some datasets. A common problem was the systematic absence of one spacer on some membranes. These data sets were systematically excluded. Most of these problems were linked to manufacturing defaults in the commercialized membranes. All data sets were checked individually and sometimes audited on source results (autoradiography). Most investigators, to check their procedures, completed a QC form. International Guidelines to increase spoligotyping quality are in progress and will be reported elsewhere. As in all databases, we assume a reasonable (2 to 5% maximum) error rate in data points. However, the problem of QC in high throughput genotyping technique and database science technology is an emerging issue
Authors' contributions
All authors contributed to the spoligotyping data contained in the database at various levels, by locally isolating, identifying clinical isolates, preparing DNA and genotyping clinically isolates of MTC and sending their results to the Pasteur Institute in Guadeloupe. Their relative contribution would be to tedious to mention here, however, they are all part of the paper because such population-based studies could not be done without adequate mycobacteriological diagnostics, i.e., isolation of clinical isolates, identification, drug-susceptibility testing, DNA extraction, genotyping and ultimately by sharing of data, hence without adequate reward of the numerous persons and labs involved in such analysis.
KB managed the SpolDB4 Information System, controlled data quality, did part of the supervised analysis, used SpolNet for Bioinformatical classification, compiled data synthesis. CS initiated the SpolDB project, established many of the contacts, recruited many of the investigators with NR, writing e-mails, checked the Information System developments helped by numerous Computer scientists (Philippe Leremon, Christel Delfino, Philippe Abdoul, Georges Valétudie, are warmly acknowledged). CS did part of the mining with KB and wrote the paper. JRD managed the PHRI New-York database and sent his data regularly to CS and KB. LR and WP were the two most "big account" data providers, together with AG. BL produced the maps (P. Waniez, Philcarto®, IRD, France) and contributed to the statistical analysis.