Sensitivity and specificity of different culture techniques for the detection of GBS
In our study, only the combination of Lim broth and subculture on Granada agar enabled detection of all carriers, whereas direct inoculation onto Columbia CNA agar achieved a sensitivity of 40% and subculture of the Lim broth onto Columbia CNA agar detected 81% of the carriers.
Gil et al.  showed that in different studies the sensitivity of Granada agar ranged from 88.5 to 91.1%, that of Columbia CNA agar from 83.9 to 94.3% and that of Lim broth (to which 5% horse serum was added) was 63.5% when subcultured on Granada agar and 75% when subcultured on Columbia CNA agar. Bosch-Mestres et al.  showed that the use of direct inoculation onto Granada agar allowed fast detection for about 87% of carriers, where as the combination of Todd-Hewitt broth and subculture on Granada agar or Columbia CNA agar allowed detection of more than 99% of GBS carriers. Elsayed et al.  reported 79% sensitivity for direct inoculation onto Columbia CNA agar and incubation during 48 hours compared to 100% sensitivity for Lim broth with subculture on blood agar. It can be concluded from this study and others that the combination of broth enrichment (Todd-Hewitt or LIM broth) with subculture on solid medium, yields higher sensitivity than direct inoculation onto solid media.
Although no definite conclusions can be drawn with regard to the sensitivity of the different solid media, Granada agar has the advantage that it makes possible to provide overnight results by visual inspection of the presence of red orange colonies, that are produced exclusively by GBS. The red orange colonies are easily observed, even when few colonies are present or when GBS is mixed with other microorganisms, which are mainly other streptococci  . The colonies are so characteristic and unique that identification by antigen detection or the CAMP test is unnecessary. In this study, all red orange colonies were CAMP positive (data not shown). Claeys et al.  missed only two CAMP positive isolates on a total of 310 tested, which were not red orange pigmented on Granada. Blanckaert et al.  used a combination of Granada and Columbia blood agar for GBS screening and demonstrated that 6% of the samples positive for GBS on Columbia blood agar lacked red orange colonies on Granada agar. Notably, Granada agar does not detect non-pigmented isolates, and on blood-agar these non-hemolytic isolates are difficult to detect as well. As a result of that, non hemolytic, nonpigmented strains may have been missed. In conclusion, in our hands, the use of combination of Lim broth and subculture on Granada agar provided high sensitivity and specificity for detecting GBS in vaginal and rectal swabs from pregnant women.
The high prevalence of S. agalactiae colonization, i.e. 24%, as established in this study, is in accordance with results from other European studies that report colonization rates between 10% and 36% [38–43]. Thinkhamrop et al.  reported a prevalence of 7.1% (Myanmar) to 19.1% (Philadelphia), using Lim broth culture. Toresani et al.  found a prevalence of only 3.2% among 531 Argentinan pregnant women. Brimil et al.  showed equal carrier rates of 16% for GBS among pregnant and nonpregnant women and, based on the prevalence of GBS carriage, these authors concluded that strict adherence to the guidelines for prevention of GBS neonatal infection results in peripartal antibiotic prophylaxis in up to 20% of all deliveries.
Our data, are also in correspondence with other results on GBS prevalence in our country. Rectovaginal colonization with group B streptococci in Belgium is 13-25%. These data are based on different studies carried out by the Belgian reference laboratory for GBS in collaboration with the section of epidemiology of the Scientific Institute for Public Health (ISP-WIV, Brussels) . For example, Blanckaert et al.  compared the results of GBS screening on Granada agar with those obtained using standard Columbia blood agar at two participating centers in Belgium. They reported GBS-positive culture results of 10-30% of pregnant women. The Flemish Study Centre for Perinatal Epidemiology evaluated GBS prevalence in Flanders and found an average colonization rate of 16% among Flemish pregnant women .
Manning et al.  found that the prevalence of GBS colonization was equally high among 241 women (34%) and 211 men (20%) living in a college dormitory. Sexually experienced subjects had twice the colonization rates of sexually inexperienced participants. Van der Mee-Marquet et al.  reported that the prevalence of carriage was 27% in women and 32% in men. The major positive body site was the genital tract (23% in women and 21% in men) and skin, throats, and anal margins were also positive in 2%, 4%, and 14%, respectively.
Comparison between vaginal and rectal carriage
Brimil et al.  reported that for a total of 34 GBS positive pregnant women, 32% carried GBS only vaginal, 24% only rectal and 44% both rectal and vaginal, which compares well with our results, i.e. resp. 24%, 22% and 53%.
Serotypic and genotypic diversity among GBS isolates
We studied the rectal and vaginal colonization with GBS of pregnant women attending the Ghent University hospital (Belgium) and the serotypic and genotypic diversity among the GBS isolates. Strong genotypic and serotypic heterogeneity was observed between women and for individual women, e.g., we found a total of 66 genotypes among 118 isolates (from 32 women), of which 4 women had up to 4 different genotypes in total (rectal + vaginal) and 14 women had up to 2 different genotypes vaginally.
Manning et al.  genotyped GBS isolates from vaginal-rectal swabs of women at two visits and documented a turnover in 8.3% of 126 women colonized both at 35-37 weeks of gestation and 6 weeks after delivery. Unfortunately, only one isolate was genotyped per visit. Taking into consideration the genotypic diversity per subject, as observed in our study, a turnover of 8.3% might be an overestimation. When the same subject, carrying e.g. genotypes a, b and c, is sampled at two different moments, and whereby on each occasion only one colony is picked, the detection of a different genotype may be interpreted as turn over, but it may be that a genotype a strain has picked at the first visit but a genotype b or c strain at the second visit.
Moreover, their conclusion that some clones are more likely to be lost should be interpreted with care, since their apparent disappearance might be explained by the presence of these clones in relatively lower numbers (compared to other clones), which reduces their chance of being picked at two separate occasions. To our knowledge, only one group studied genotypic diversity within individual women, by genotyping 15 randomly picked isolates for each of 30 women, albeit from mixed vaginal-rectal specimens . In opposition to our results, these authors found a high degree of genotypic and serotypic homogeneity, i.e. for 29 of 30 women, all 15 isolates from each woman had the same serotype, and for 27 of 30 women, all 15 isolates had the same chromosomal SmaI-DNA restriction digest fingerprint. For the three women with different PFGE types, one had three different genotypes and two different serotypes. Another PFGE based study reported high stability of the GBS type for each woman, followed up to two years . In accordance with our results, these authors found 30 different GBS isolates among 32 women, with only two women carrying isolates with the same genotype.
The high level of heterogeneity established in our study, compared to other studies, may be due to the different culture media that were used and to the use of separate swabs for sampling vagina and rectum for each individual woman. We found that strains of the same serotype recovered from different women were heterogeneous in DNA profiles. In accordance, SmaI-restriction digestion of chromosomal DNA and PFGE revealed high genotypic heterogeneity among both Zimbabwean serotype III and serotype V isolates , another study also reported different SmaI restriction types within serotypes II and V , and Fasola et al.  and Savoia et al.  showed that several genotypic lineages are present within the different serotypes.
Serotype switching is believed to occur within genotypic lineages [7, 24, 25], presumably by horizontal transfer of genes of the cps locus, i.e. of genes encoding the GBS capsular polysaccharide structure , and may be an explanation for our observation of the presence of two isolates in two different women with the same genotype, but with a different serotype.
Although MLST has become the standard method to study the population structure of GBS [22, 24], more rapid and less expensive and laborious methods remain useful for carrying out single centre studies. We used RAPD-analysis in combination with high resolution capillary electrophoresis, which also makes possible immediate digitization of the fingerprints. Chatellier et al.  found that the simplest typing scheme of S. agalactiae was obtained by the combination of RAPD typing and serotyping (discriminatory index 0.97). Zhang et al.  found congruence between RAPD analysis and serotyping, on a limited number of strains. The findings of Toresani et al. , who found a total 16 RAPD profiles among 21 GBS isolates, from 17 women, and of Chatellier et al. , who identified 71 RAPD types among 54 unrelated S. agalactiae strains isolated from cerebrospinal fluid samples from neonates, point to the same genotypic diversity as observed in our study.
General serotype distribution
The capsule of S. agalactiae has long been recognized as one of the most important virulence factors. Variations of the capsular polysaccharide structure allow the antigenic distinction of 13 different S. agalactiae serotypes, of which 9 are of clinical importance (Ia, Ib, II-VIII). Studies from the US and Europe show that the serotypes Ia, II, III, and V are found in 80-90% of all clinical isolates [4, 48].
Serotype distribution among GBS isolates from pregnant women in our study was compared with that reported by others. Several studies indicate that serotype III is globally the most prevalent serotype, e.g. 29.2% of Israelian isolates , 28% of German isolates , 24.3% of Swedish isolates  and 33.2% of Czech isolates . In accordance, we observed a frequency of 20.8% for serotype III isolates in our study, as the most frequent serotype.
The finding that two out of 8 women carried isolates with different susceptibility to clindamycin indicates that when testing susceptibility for clindamycin several colonies should be tested, since colonies with different susceptibility may be simultaneously present.