Bacterial vaginosis (BV) is one of the most common reasons for women to seek medical attention; the underlying cause of BV is controversial. Women with BV are at higher risk for preterm delivery, pelvic inflammatory disease (PID) and acquisition of HIV [1–5]. Several reports have demonstrated that women with BV have elevated levels of sialidase activity [6–9], although the source has not been demonstrated. Studies have demonstrated that treatment of HIV-1 or lymphocytes with bacterial sialidase increases the infectivity of the virus [10, 11]. A number of different bacteria have been associated with BV, including Gardnerella vaginalis [12–14]. G. vaginalis can be isolated from women without any symptoms and recovered from sites which are usually sterile [15, 16]. Studies have also shown G. vaginalis biotype 1 fractions are capable of stimulating HIV-1 production [17]. G. vaginalis is a fastidious organism requiring subculture to fresh media every two days. Isolates identified as G. vaginalis may be further characterized by β-galactosidase and lipase activity and hippurate hydrolysis resulting in 8 biotypes [18]. According to one study biotypes 1–4 produce lipase and in a longitudinal study were significantly associated with BV. After successful treatment, the predominant G. vaginalis biotypes shifted to non-lipase producing types 5–8 [6]. Other studies did not find a relationship between BV and biotype or genotype [15]. Piot et al. [18] defined biotypes using egg yolk agar (EY) to test for lipase while Briseldon and Hillier [6] used 4-methylumbelliferyl-oleate (MUO). Since the use of MUO had not been validated, we compared the results of lipase detection using egg yolk to those obtained with MUO. Because G. vaginalis sialidase could play an important role in both BV and HIV infection we also tested our strains for sialidase activity.

Lipase activity was detected in 21 of 31 strains tested (68%) using egg yolk agar but using the MUO spot test only 12 of 31 strains tested positive (39%) (Table 1). To assess the performance of the MUO based lipase test, the egg yolk reactions were used as the true values in the statistical comparison as described previously by Moncla et al. [20]. The values obtained for the MUO test were: sensitivity (32%), specificity (33%), positive predictive value (54%) and negative predictive value (17%). The alternate method for lipase detection (see above, mixing equal volumes of buffer and liquid substrate) demonstrated a lack of reproducibility and the results from these assays are not presented.

Sialidase was detected in one or more strains of biotypes 1, 2, 4, 5 and 7 but not in strains from biotype 3, (Table 1). Biotypes 6 and 8 were not found among the strains examined. Most of the strains studied were biotype 1 (32.2%), followed by biotypes 2, 7, 4, 5, and 3 (22.5%, 19.5%, 9.6%, 9.6%, 6.5% respectively). Overall, the 39% of the strains tested demonstrated sialidase activity.

GVA provides an inexpensive alternative to the long term cultivation of G. vaginalis, compared to blood agar or HBT on which the organism requires passage every day or two. Some isolates survived for up to three weeks on the GVA. HBT is an important medium commonly used to test the hemolysis characteristics and to maintain the organism; however, it is too expensive for long term passage of the organism.

Sialidase activity is an important feature in bacterial vaginosis. Accordingly, we concurrently tested our strains for the enzyme. Previous reports had noted the enzyme in 10% of the isolates [21]. We observed higher rates among our isolates 39% (table 1). About one third of our isolates were biotype 1, of which 40% of the isolates were positive. The presence of sialidase was detected in strains from all biotypes tested except biotype 3; however, we only had three isolates identified as biotype 3. We did not identify any isolates as biotypes 6 or 8.

The sialidase activity associated with BV is most likely from bacterial sources, although the specific organisms have not been identified. We report a higher incidence of the enzyme than reported by others [21] the significance is difficult to access since we only examined 31 strains. Because they are found in low frequency we did not test biotype 6 or 8. Other bacteria associated with BV have a much greater percentage of isolates that produce sialidase for example; all isolates of Prevotella bivia produced the enzyme [19]. While P. bivia isolates all produce sialidase, only about 3% of the activity is released into the medium; however, we do observe surprisingly large amounts of sialidase in the culture supernatants of sialidase producing strains of G vaginalis (Moncla unpublished observations).

Because of the relationship between G. vaginalis biotype and the stimulation of HIV replication [12, 17]; we attempted to biotype our strains using the MUO method of Briselden and Hillier. G. vaginalis ATCC 14018 is biotype 1; however because of a negative lipase reaction we consistently identified it as biotype 6. As the MUO method of Briselden and Hillier had not been validated we compared the results of lipase detection using egg yolk agar to those obtained with MUO. The choice of lipase substrate had a dramatic effect on the biotype. For example, using EY lipase results there are 10 strains listed in Table 1 as biotype 1; however, using MUO as the substrate, only 2 of the 10 isolates demonstrated lipase activity. The 8 strains that were negative using MUO would have been incorrectly identified as biotype 6. Overall 20 of the tested isolates would have yielded a different biotype depending on the method used for testing lipase activity. The MUO method had poor sensitivity and specificity when calculated as described previously [20]. Using the EY data for biotyping as presented in Table 1, we observed a distribution of biotypes that is roughly similar to that reported by Piot et al.

Lipases are specialized esterases; therefore, which synthetic substrates, if any, are appropriate is controversial. Studies suggest synthetic substrates such as MUO detect non-specific esterase activity [22–27]. Our data would support this concept. When other 4-methylumbelliferyl fatty acids were used, we observed all strains give a positive test results with MU-heptonate but none with 4-methylumbelliferyl palmitic acid, indicating the assays are measuring esterase activity [28, 29]. These data would tend to negate the observations of others regarding the correlation of G. vaginalis biotype with BV. Briselden and Hillier's observation of a reduction of lipase producing biotype 1–4 after successful treatment could be reinterpreted as an association of non-specific esterase activity in G. vaginalis with BV [6]. Our results demonstrate the importance of lipase activity in the typing of G. vaginalis and that lipase activity should be tested using EY plates or other lipase assay methods such as titration. Further, our work suggests the reports of biotypes using the MUO or other 4-methyumbelliferone substrates in lipase spot tests are not accurate. Other differences exist in the methodologies reported, Piot et al. and Briselden and Hillier grew cultures anaerobically while the other groups mentioned grew organisms aerobically with enriched CO₂. Lipase reactions on EY often take up to 7 or more days, and all these groups used only 3 days or less for reactions. Our observations suggest all isolates should be cultured anaerobically, and EY plates should be incubated for 7 days before they can be interpreted as lipase negative.

In summary a medium was described that allows survival of G. vaginalis isolates for at least one week and longer in some cases. Sialidase activity was observed in 40% of the strains tested but was not restricted to any particular biotypes. The synthetic lipase substrate 4-methylumbelliferyl-oleate did not reliably detect lipase activity compared to egg yolk plates.

Our data suggests the relationship of BV and G. vaginalis biotype should be reexamined, since our study demonstrates that 4-methylumbelliferyl-oleate and other 4-methylumbelliferyl- derivatives should not be used for the detection of lipase activity as a tool for bacterial identifications. The Gardnerella vaginalis agar allows extended viability of the cultures, therefore the time and costs of frequent subculture is greatly reduced. We cannot rule out an association of G. vaginalis, sialidase, BV and increases in HIV acquisition rates among women with BV.