Background
The role of human genital Ureaplasmas (
Ureaplasma parvum and
Ureaplasma urealyticum) and Mycoplasmas (
Mycoplasma hominis and
Mycoplasma genitalium) in preterm birth (PTB) has been a contentious topic that has resulted in considerable research over the past two decades. A recent review article by Capoccia et al. [
1], summarizing the bulk of this research effort (with the exception of
M. genitalium), showed that the majority of studies with defined case/control subjects found an association with negative pregnancy outcomes, especially for
Ureaplasma spp. For
M. hominis, this association appeared to be increased when detected alongside
Ureaplasma spp. [
1].
M. genitalium, however, is more commonly associated with pelvic inflammatory disease [
2] and has also been linked to bacterial vaginosis (BV) [
3]. Vaginal colonisation by
Candida spp., a common yeast synonymous with vaginal thrush, has also been associated with PTB risk [
4‐
7].
However, it is still unclear why some women who are asymptomatically colonised with these organisms have pregnancies that deliver preterm while others deliver at term; possible explanations include interactions with other microorganisms and inflammatory modulators, the timing and duration of colonisation, differences in virulence between species/strains, and eradication or exacerbation via maternal immune responses. With respect to
Ureaplasma spp., very few studies have ever attempted to refine the data beyond the genus level, despite evidence that numerous other Ureaplasma-associated disease phenotypes are differentially associated with
U. urealyticum and not
U. parvum [
8‐
12]. Further to this, to the best of our knowledge, no studies have ever looked at the relationship that may exist between PTB and the presence of specific
Ureaplasma spp. genotypes (serovars). In addition, few studies have looked at the effect of combined colonisation with these organisms, with most studies instead focusing on either
Ureaplasma spp. and
M. hominis (and not
M. genitalium) or
Candida spp.
PTB is now recognised as the leading cause of death in children under the age of 5 years in the developed world [
13], with intrauterine infection being causally associated with ~40 % of all PTBs [
14]. In particular, the births of the earliest gestational ages, which are associated with the greatest mortality and morbidity, are much more likely to be the consequence of intrauterine infection and/or inflammation (up to 80 %). There is, therefore, a need to be able to predict which women are at a high risk of PTB, based on factors other than previous clinical history alone, so that appropriate treatment may be provided to those who would benefit. This is especially relevant for women in their first pregnancy. Despite the body of research that has been conducted into the presence and significance of genital
Ureaplasma,
Mycoplasma and
Candida spp. colonisation during pregnancy, there is very little information available on the relationship between specific lifestyle practices, such as smoking and sexual activity, and vaginal colonisation. However, it has been well established that tobacco use increases susceptibility to bacterial infection [
15] and a strong association has been documented between smoking and BV [
16‐
19]. Furthermore, some genital
Ureaplasma and
Mycoplasma spp. are deemed to be sexually transmitted infections [
20].
In addition, the vast majority of previous studies that have looked at genital
Ureaplasma,
Mycoplasma and
Candida spp., have focused on collection of clinical samples at a single time point. This approach provides no data on the colonisation dynamics of a particular organism and this is likely to be of significance in terms of implementing diagnostic assays for the detection of women at risk of PTB. It is highly likely that all of the above information combined, will not only increase our understanding of these organisms during pregnancy, but importantly, will allow us to establish useful microbial biomarkers. These, combined with patient-specific qualitative data, will improve our ability to predict women at an increased risk of PTB who would benefit from anti-microbial treatment. Due to previous descriptions of variation in the vaginal microbiota as a result of ethnic, cultural and social factors [
21], it is likely that microbial colonisation status and PTB risk are strongly population-dependent.
The aims of this study, therefore, were to document the presence of U. parvum (including genotype analysis), U. urealyticum, C. albicans, C. glabrata, other non-albicans Candida spp., M. hominis and M. genitalium in the vagina of a cohort of asymptomatic pregnant Australian women at three time points during pregnancy, and to examine any association between their microbiological characteristics, a range of lifestyle factors, and risk of PTB.
Discussion
Numerous previous studies have described an association between detection of
Ureaplasma spp. in the vagina during pregnancy and subsequent PTB [
1]. Consistent with data from the majority of these studies, our findings confirm a significant association between vaginal colonisation by
Ureaplasma spp. and PTB risk. However, unlike the vast majority of previous research, we have further discriminated between human Ureaplasmas at the species level. Our finding that vaginal colonisation by
U. parvum, but not
U. urealyticum, was associated with an increased risk of PTB, compared to colonisation by
Ureaplasma spp. in general, highlights a major limitation in these previous studies. This is despite the fact that in 1990
U. urealyticum was classified as two distinct biovars, parvo (
U. parvum) and T960 (
U. urealyticum) [
27], and then formally proposed as two separate species in 2002 [
28]. In addition, species (biovar)-level PCR detection assays have been available since 1993 [
29]; hence, the ability to study this organism at the species level has existed for many years. It is important that future studies ensure that species-level analyses are carried out, especially those examining pregnancy outcome, as our data showed that
U. parvum confers the greatest risk of PTB. This is in stark contrast to other disease phenotypes, such as pelvic inflammatory disease [
8], endometritis [
8], non-gonococcal urethritis [
8‐
10,
12] and post-gonococcal urethritis [
11], where
U. urealyticum has been reported to be significantly associated with disease, but not
U. parvum. In previous studies [
30‐
45] that examined the obstetric consequences of vaginal colonisation by
Ureaplasma spp., where there were defined case and control groups, only three discriminated between
U. parvum and
U. urealyticum. Similar to our study, the two most recent of these, Mitsunari et al. [
32] and Kataoka et al. [
42] reported that vaginal colonisation by
U. parvum, but not
U. urealyticum, was significantly associated with PTB. However, in contrast, a study from the late 1990s by Abele-Horn et al. [
34] came to the opposite conclusion, with
U. urealyticum apparently having more adverse effects on birth weight, gestational age and preterm delivery compared to
U. parvum. These are the only authors, to our knowledge, that have reported an association between
U. urealyticum and PTB; however, they employed a unique set of inclusion criteria whereby women were recruited who were solely colonised with
Ureaplasma spp. and no other ‘abnormal’ microorganisms, including
C. albicans,
Gardnerella vaginalis and many others. This may have introduced confounders.
We have previously discussed the history and limitations associated with serotyping/genotyping of
U. parvum and
U. urealyticum [
23]. As a result of this, we developed a HRM PCR assay capable of detecting the four genotypes of
U. parvum directly from clinical samples and suggested that this assay may provide valuable information relating to
U. parvum genotype status and specific clinical conditions [
23]. In the current study, using this assay, we were able to identify single
U. parvum genotypes from 91 % of colonised samples, with a further 3 % resolved to the level of ‘mixed’ genotypes. From 74
U. parvum positive samples, we identified genotypes SV6 (42 %) and SV3 (36 %) as the most common, followed by SV1 (15 %). This genotype distribution is slightly different to that described by Xiao et al. [
8], who reported that in 169 vaginal samples from healthy pregnant women, SV3 (63 %) was the most common, followed by SV6 (30 %), SV1 (24 %) and SV14 (4 %). This may indicate that the geographical location of the cohort influences serovar prevalence. Unfortunately, these authors did not provide any data on potential associations between genotype and adverse pregnancy outcome. However, in another sample set from the same study, Xiao et al. [
8] reported a significant association between detection of
U. parvum genotype SV6 in placental tissue and histologic chorioamnionitis, a condition that is indicative of infection-associated PTB. This corroborates our finding of a significant association between detection of
U. parvum genotype SV6 in the vagina and PTB <37 weeks GA, in addition to our detection of this genotype in 4/6 cases of PTB <34 weeks GA.
To our knowledge, the only other study that has examined
U. parvum genotypes present in vaginal samples is that of De Francesco et al. [
46], who collected cervical, urethral, and vaginal swabs from 806 women. In contrast to our results, these authors reported that genotypes SV3 and SV14 (not separated beyond this level) and SV1 were the most common detected, representing 39 and 37 % of the 158 women who were positive for
Ureaplasma spp. Genotype SV6 was detected in 24 % of cases and was apparently associated with a vaginal microbiota deemed ‘normal’, compared to that associated with SV3/SV14, which was linked to an absence of lactobacilli. However, it is difficult to compare these results to those from our study, as De Francesco et al. [
46] did not differentiate the patient type associated with positive samples, instead compiling women from outpatients for gynaecological health care control, routine screening for pregnancy, infertility problems and those with symptoms of genital infections into one group. As such, the multiple phenotypes present make it impossible to extrapolate phenotype-specific data from this study.
Another study that examined
U. parvum genotype distribution was by Sung et al. [
47] and involved preterm neonates with bronchopulmonary dysplasia (BPD). It reported that SV3 and SV6, either alone or together, accounted for 96 % of
U. parvum isolates detected in endotracheal and/or nasopharyngeal aspirates. However, these authors failed to find an association between genotype and the development of moderate to severe BPD.
Also of interest, and particularly relevant to the clinical use of our HRM assay, we noted that
U. parvum genotype colonisation in our study was typically singular and in nearly all cases was stable throughout pregnancy. Although we are unable to compare this to the work of any previous studies due to omission of details relating to single/multiple genotype detection [
46] or presenting multiple genotype data including
U. urealyticum genotypes [
8,
47], it is interesting to note that studies that have looked at
U. urealyticum genotype distribution generally report that when detected, multiple genotypes are typically present [
8,
47,
48]. However, this may also be a result of problems with cross-reactivity associated with PCR assays for these genotypes [
8].
We did not detect a significant association between the presence of vaginal
Candida spp. alone and PTB, a result that contradicts four previous studies [
4‐
7] and a recent systematic review [
49], although we did observe a trend towards significance. This warrants further investigation in a larger cohort of women. However, when vaginal colonisation by
U. parvum was accompanied by
C. albicans (the most commonly detected
Candida spp. in our study), the odds ratio (OR) for risk of PTB increased marginally from 3.32 (
U. parvum alone) to 3.77 (both organisms). Stratifying
U. parvum by genotype, additional increases in risk were seen when
C. albicans was combined with
U. parvum genotype SV6, with the OR rising from 4.17 to 5.16. Both organisms have previously been associated with increased PTB risk [
1,
5,
50], while colonisation of amniotic fluid by
C. albicans has been linked to severe fetal injury [
51‐
53]; however, no previous studies have looked at the effects of combined asymptomatic vaginal colonisation and pregnancy outcome, particularly regarding
U. parvum genotype. Further research is warranted, firstly to confirm our novel findings in a larger cohort, and second to attempt to uncover what factors may be responsible for this association. If our findings are confirmed, then detection of these two organisms during the early second trimester of pregnancy could prove to be a useful indicator of women at a high risk of infection-associated PTB and therefore targets for antimicrobial therapy.
Although we have discussed the results of our study in terms of presence/absence of specific bacteria at ~21 weeks GA (primarily as this is of major relevance to allow implementation of suitable treatment regimens for prevention of PTB), our study also examined vaginal samples at two additional time points (~28 weeks and ~36 weeks GA). To the best of our knowledge, this is the first study to examine the dynamics of Ureaplasma spp., Mycoplasma spp. and Candida spp. throughout the second and third trimesters of pregnancy. We demonstrated that detection of all organisms (with the exception of C. glabrata) was very stable throughout the three sampling points, particularly so for M. genitalium and U. parvum. In addition, for U. parvum we also demonstrated that in nearly all cases, the genotype detected at recruitment was maintained throughout the second and third time points. These are very important findings as they demonstrate that detection of these organisms early in the second trimester of pregnancy is highly predictive of their presence both early and late in the third trimester, and provides valuable information for future diagnostic assays aiming to predict women at risk of PTB.
Of significant relevance to use of
U. parvum detection assays in future clinical diagnostic applications, we showed near 100 % concordance between detection of
Ureaplasma spp. in clinical samples using either culture-based or PCR methods, in contrast to some previous authors who reported disparities between the two techniques [
54‐
56]. It has been well established that
Ureaplasma sp. cells are fragile and in order to optimise recovery from clinical samples, special attention needs to be paid to sample collection, storage and transportation, in addition to the media used for culture [
57]. Our results are most likely a reflection of our highly stringent sample collection and culture conditions, whereby swabs were immediately placed into UTM media, stored at 4 °C and cultured in 10B broth within 24 h of collection under microaerophilic conditions. This is important, as despite the increased use of molecular methods for detection of
Ureaplasma spp., culture is likely to play a key role alongside these in future studies attempting to document virulence traits of strains associated with specific disease phenotypes.
In addition to culturing
Ureaplasma spp. from our clinical samples, we also documented the titre of organisms present in each. We identified a small, but significant association between the titre of
U. parvum and PTB, with an average
U. parvum titre of 10
6 CCU in cases of PTB vs. 10
5 CCU for term cases. Although this may appear significant statistically, it is unlikely to be clinically relevant. We used eight 10-fold serial dilutions of 10B broth culture for quantitation and expressed our titres in colour changing units (CCU), a technique that has been widely used in
Ureaplasma sp. culture [
58‐
61]. A limitation of the CCU method is that although it is far more suited to quantitation of large sample numbers, it really is semi-quantitative in that the titre values obtained actually represent a range rather than hard values. For instance, 10
6 CCU may actually be 1,000,000–9,000,000 cells. As a result, we believe clinical significance would have only been justified in the case of a difference of two dilution series, as opposed to a single series, such as we observed. Standard solid agar plate quantitation methods are not well suited to
Ureaplasma spp. due to the tiny colony size and associated quantitation difficulties (a microscope with ~ 10–40× magnification is required). Despite this, Abele-Horn et al. [
35] used this method for quantitation and following multivariate analysis of 172
Ureaplasma-colonised and 123 non-colonised pregnant women, they reported that
Ureaplasma sp. vaginal colonisation was an independent risk factor for PTB at 10
5 CFU/mL and for chorioamnionitis at both 10
4–10
5 and 10
5 CFU/mL. Lower colonisation levels had no adverse effects on pregnancy outcome [
35]. Regardless of the method used though, it could be argued that accurate quantitation of any bacteria from a swab-collected sample is confounded by the manner of sample collection, which is likely to have a large effect on the results.
Although our sample size was too small to accurately document any association between detection of vaginal
M. genitalium and PTB, we observed a trend towards significance, with detection rates of 12 and 2 % at recruitment in women delivering preterm and term, respectively. This result is particularly interesting in that it conflicts with the findings of four previous studies, all of which found no association between vaginal
M. genitalium and PTB; interestingly, they reported detection rates (0.7–8 %) somewhat lower than those reported here [
62‐
65]. One of these was by Labbe et al. [
65] who reported rates of 8 % in preterm deliveries (16/183) and 6 % in women delivering at term (36/564) in a population of women from a developing country, many of whom also screened positive for other known sexually transmitted infections, including human immunodeficiency virus (HIV). In comparison, Hitti et al. [
66], reported detection rates of only 4 % (29/661) and 2 % (12/667) in Peruvian women who delivered preterm and term, respectively. This finding was significant, accompanied by an odds ratio of 2.5 (95 % CI: 1.2–5.0). Of particular interest, both Oakeshott et al. [
63] and Hitti et al. [
66] documented that vaginal
M. genitalium was more commonly observed in women of a younger maternal age (<20 years and mean 21.2 years, respectively). This was also true in our cohort, where the mean age of women in which vaginal
M. genitalium colonisation was detected was 24.3 years (in comparison to the overall cohort mean age of 30), and perhaps suggests that maternal age needs to be taken into account when assessing microbiological-associated PTB risk.
Many previous studies that have examined associations between vaginal bacterial colonisation and PTB have not considered the effect of lifestyle factors on the species present. By administering questionnaires to our participants at each sampling point (asking about such lifestyle factors as smoking before/during pregnancy and the amount of sexual intercourse during pregnancy), we sought to establish if there was a relationship between these factors and organism presence. For both
Ureaplasma spp. and
Mycoplasma spp., we observed significantly higher detection rates in women who reported having sexual intercourse greater than or equal to 3 times per week during the course of their pregnancy. Considering that both of these organisms are known colonisers of both the female and male genital tracts [
67], and are also considered sexually-transmitted infections in certain circumstances [
20], this result is not surprising. However, of particular interest, we also observed associations between women who either smoked prior to or during their pregnancy and increased detection rates of
Ureaplasma, Mycoplasma and/or
Candida spp. Of statistical significance, higher detection rates of
Ureaplasma spp. (37 vs. 17 %) and
Mycoplasma spp. (44 vs. 24 %) were observed in women who smoked prior to their pregnancy and for
Candida spp., colonisation rates were significantly higher in women who continued to smoke throughout their pregnancy (18 vs. 7 %). A similar phenomenon has been reported previously in relation to smoking and vaginal microbiology in the case of human papillomavirus (HPV) infection [
68] and also for bacterial vaginosis [
15]. Although the association between BV and smoking may help to explain the association between
Ureaplasma spp. and
Mycoplasma spp., both of which have been previously reported as BV-associated agents [
3],
Candida spp. are not typically associated with BV. In addition, whereas it has been reported that the anti-estrogenic effect of smoking [
69] may predispose a woman to BV [
16], it has been previously documented that
Candida spp. are more prevalent during pregnancy as a result of the increased amounts of estrogen present [
70]. One possible explanation could be that benzo[a]pyrene, found in trace amounts in the vaginal secretions of smokers, has been shown to significantly induce
Lactobacillus spp. prophages [
71]. This would result in a decrease in numbers of vaginal
Lactobacillus spp. due to cell lysis from lytic phages, potentially providing a more favourable environment for
Candida spp. and especially BV-associated organisms, as
Lactobacillus spp. are critical for maintenance of an acidic vaginal pH [
72]. Detailed metagenomics studies would be required to address this hypothesis.