The biochemical activities and resultant metabolic by-products of the vaginal microbial community during gestation can provide useful insight into the pathophysiology of preterm birth (PTB), as well as help in identifying women at risk. These metabolic changes leave specific signature fingerprints that can be investigated by Magnetic resonance spectroscopy (MRS). Therefore, we hypothesised that women who ultimately deliver prematurely will have significantly different vaginal microbiota metabolite signatures compared to their term counterparts even in the absence of clinical infection.
In order to characterise and validate the cervicovaginal fluid (CVF) metabolite profiles and determine their predictive capacities for PTB, high-vaginal swabs were obtained from asymptomatic and symptomatic pregnant women sub-classified depending on a previous history of PTB and/or short cervix (< 25 mm) into: asymptomatic low risk (ALR) women with no prior PTB nor short cervix, 20-22 gestational weeks (w), n = 183; and asymptomatic high risk (AHR) women with prior history of PTB and/or short cervix, 20-22 w, n = 186. A subset of these women were assessed again at 26-28 w (due to their high-risk status), n = 129. The fourth cohort comprised women presenting with symptoms of threatened preterm labour (PTL) 24-36 w, n = 89 (SYM).
CVF dissolved in phosphate buffered saline was analysed with a 9.4T MR spectrometer. Metabolites were identified, integrated for peak area and normalised to the total spectrum integral (excluding water signal). Acetate concentrations (AceConc) were also determined from a randomly selected subset of SYM women (n = 57), by a spectrophotometric technique. Additionally, clinical parameters such as cervical length (CL), fetal fibronectin (FFN), and vaginal pH were recorded and correlated to the metabolites. Furthermore, the 16S rDNA of vaginal bacterial species were PCR-amplified and the vaginal cytology was also determined by Gram, Hematoxylin and eosin, and Papanicolaou staining methods.
We observed that acetate normalised integral (N.I.) (P = 0.002), and acetate/lactate ratio (P = 0.002) were higher in the SYM women who delivered preterm. These were also predictive of PTB < 37 w (AUROC: acetate N.I. = 0.75; acetate/lactate ratio = 0.76), < 32 w (AUROC: acetate N.I. = 0.73; acetate/lactate ratio = 0.79), and within 2 weeks of the index assessment (AUROC: acetate N.I. = 0.77; acetate/lactate ratio = 0.78), whilst glutamine/glutamate N.I.s was predictive of PTB < 32 w (AUROC = 0.71), and within 2 weeks of the index assessment (AUROC = 0.68) only.
Also, in the AHR20-22w and ALR women, acetate (AUROC = 0.61) and branched chain amino acids N.I.s (AUROC = 0.75) were predictive of PTB < 37 w respectively. Normalised integrals of succinate, formate, lactate, and glucose did not differ in relation to PTB in any of the groups.
Like the acetate N.I.s, AceConc in the SYM women was higher (P = 0.006) in the preterm-delivered women and was predictive of PTB < 37 w (AUROC = 0.74), and within 2 weeks of the index assessment (AUROC = 0.68), with an optimal cut-off of > 0.53 g/l. AceConc also correlated with acetate N.I. (r = 0.69; P < 0.0001).
PCR revealed a higher prevalence of potentially pathogenic anaerobic bacteria species in the preterm-delivered women across the groups except the ALR women.
Apart from correlating with clinical parameters, the prediction of PTB was improved especially in the SYM women when metabolite N.I.s, CL and FFN were combined.
In conclusion, elevated CVF acetate showed clinically useful discriminative propensity for preterm delivery and delivery within 2 weeks of presentation in symptomatic women. A ratio of acetate to lactate showed similar discriminatory capacity in symptomatic women, whilst branched chain amino acids appeared predictive of preterm delivery in asymptomatic women at low risk of PTB. These metabolite differences were supported with the association of higher prevalence of mixed anaerobes in the vaginal melieu and preterm birth.
