Microbiota at multiple body sites during pregnancy in a rural tanzanian population and effects of Moringa-supplemented probiotic yogurt

The nutritional status of pregnant women is vital for healthy outcomes and is a concern for a large proportion of the world’s population. The role of the microbiota in pregnancy and nutrition is a promising new area of study with potential health ramifications. In many African countries, maternal and infant death and morbidity are associated with malnutrition. Here, we assess the influence of probiotic yogurt containing Lactobacillus rhamnosus GR-1, supplemented with Moringa plant as a source of micronutrients, on the health and oral, gut, vaginal, and milk microbiotas of 56 pregnant women in Tanzania. In an open-label study design, 26 subjects received yogurt daily, and 30 were untreated during the last two trimesters and for 1 month after birth. Samples were analyzed using 16S rRNA gene sequencing, and dietary recalls were recorded. Women initially categorized as nourished or undernourished consumed similar calories and macronutrients, which may explain why there was no difference in the microbiota at any body site. Consumption of yogurt increased the relative abundance of Bifidobacterium and decreased Enterobacteriaceae in the newborn feces but had no effect on the mother’s microbiota at any body site. The microbiota of the oral cavity and GI tract remained stable over pregnancy, but the vaginal microbiota showed a significant increase in diversity leading up to and after birth. In summary, daily micronutrient-supplemented probiotic yogurt provides a safe, affordable food for pregnant women in rural Tanzania, and the resultant improvement in the gut microbial profile of infants is worthy of further study. The study protocol was approved and received ethical clearance from both the Medical Research Coordinating Committee of the National Institute for Medical Research (Mwanza, Tanzania), as well as from the Health Sciences Research Ethics Board at Western University (London, Canada). The study was registered with clinicaltrials.gov ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT02021799″,”term_id”:”NCT02021799″}}NCT02021799). Participants provided written consent, and in the case where they could not write, a thumbprint was obtained before sample collection and subsequent analysis. Women who were attending the antenatal clinic at Nyerere Dispensary in Buswelu, district of Ilemela, Mwanza Region, Tanzania, were recruited into the study if they were between the ages of 18 and 40, as well as between the gestational ages of 12 and 24 weeks. Gestational age was determined by last menstrual period as identified in the participant’s clinic records. To confirm, or if last menstrual period was unavailable, approximation of gestational age was based upon measurement of fundal height. Originally, participants were further grouped into states of nutritional status (nourished or undernourished) based upon having a mid-upper-arm circumference (MUAC) of <235 mm if undernourished, and deviations from expected weight for gestational age (25). Since this was a pilot study, sample size was based on participant availability. Subjects were randomly assigned (using a random number generator) to the intervention group (produced fresh daily yogurt containing ∼1010 CFU Lactobacillus rhamnosus GR-1 per 250 g unit confirmed by regular quality control tests, and 4.3 g of dried ground Moringa oleifera leaves), or control group (no intervention). Administration of the yogurt amounted to 13% required daily intake of protein (9.95 g), 39% calcium (385 mg), 35% vitamin A (269 retinol activity equivalents), 81% vitamin B2 (1.1 mg), and 4% iron (1.2 mg). The probiotic group received 250 g of yogurt 6 days a week from the time of recruitment until exiting the study, which occurred 1 week to 1 month postpartum. Due to a lack of cold storage available to most participants, yogurt was delivered daily by a study employee who observed and recorded compliance. Participants in the probiotic group consumed yogurt for an average of 88 days ± 31 the standard deviation (SD), while the control group had no form of intervention. At the initial and monthly follow-up visits from recruitment until birth, which occurred between August 2012 and April 2013, weight was measured to the nearest 0.1 kg using an analogue scale, and the MUAC was measured to 1 mm using a tape measure. A physical examination was also performed at these visits, which included measurement of blood pressure, heart rate, and general medical history. Two vaginal swabs were collected at the midpoint of the vagina using CultureSwab polyester-tipped swabs (BD Biosciences, Mississauga, Ontario, Canada); one was used for evaluation of bacterial vaginosis (BV) by Nugent Scoring (26), while the second was used for microbial DNA extraction. Fecal samples and saliva samples were also collected in sterile collection containers and frozen at −80C until analysis. A midstream urine sample was analyzed for urinary tract infection using a multitest dipstick system (Rapid Response Urine Dipstick; BTNX Urinalysis, Laja Pharmacy, Mwanza, Tanzania). Hemoglobin levels were determined using a HemoCue Hb 201+ analyser (HemoCue, Ängelholm, Sweden). Blood samples were collected from all participants at the first, middle, and final visits in Vacutainer tubes containing potassium EDTA (BD Biosciences, Mississauga, Ontario, Canada). Plasma was separated from the samples before being frozen at −80°C until analysis. Vitamin A levels were determined by high-pressure liquid chromatography, while the total plasma protein levels were determined by a colorimetric biuret assay by the London Health Sciences Laboratory Services Group (London, Ontario, Canada). The first postpartum visits occurred at 3 days after birth, and the final visit occurred 1 week to 1 month after the first postpartum visit. The same information and samples were collected as during the follow-up visits; however, a human milk sample was collected in sterile collection containers from the mother and stored at −80°C until analysis. Anthropometric data were collected from the infant, including weight, as measured on an analogue baby scale to the nearest 0.1 kg, whereas head circumference, length, and MUAC were measured using a tape measure to the nearest 0.1 cm. In addition, the sex of the infant, the place of birth, and whether the baby was born vaginally or by caesarean section were recorded. Fecal samples were collected from the infant's diaper, and infant saliva samples were collected using a BBL CultureSwab (BD Diagnostics, Mississauga, Ontario, Canada) and stored at −80°C until analysis. Forty-eight-hour dietary recall interviews were conducted by a trained study translator at each visit. Information from these interviews was analyzed for nutritional content and caloric intake using the ESHA Food Processor SQL (v9.8), which provides access to an extensive database of comprehensive nutrition information to allow evaluation of a subject's diet and fitness. Although this program utilizes North American food composition tables, nutritional data for local Tanzanian foods and recipes were inputted into the software from Tanzanian food composition tables prepared by Muhimbli University of Health and Allied Sciences, the Tanzania Food and Nutrition Center, and the Harvard School of Public Health (27). In addition, when serving sizes per se, or sizes of specific food items were not clearly stated in the interview, the average portion quantity, as outlined by the Tanzanian Food Composition Tables, was used. Nutritional data for Moringa were obtained (M. Broin, unpublished data) and put into ESHA. The average intake of calories, protein, fat, carbohydrates, saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, calcium, vitamin A, folate, and zinc were calculated over the 48-h period to determine daily consumption. The same was done for the 48-h dietary recalls collected at the first postpartum and final visits to obtain a value representing the daily average intake of the above-listed nutrients postpartum. DNA was extracted from all samples using the PowerSoil-htp 96 Well Soil DNA isolation kit from MoBio (Carlsbad, CA) according to the manufacturer's protocol, with modifications as outlined by the Earth Microbiome Project (version 4_13). Approximately 250 mg of fecal sample, 500 μl of saliva and 500 μl of breast milk were used for the extractions. Samples were sequenced by amplifying the V4 hypervariable region of the 16S rRNA gene using the bacterial/archaeal primers 515F and 806R according to previously described methods (28) and modified for the Illumina MiSeq platform. Obtained reads were quality filtered and demultiplexed using the open-source software QIIME (split_libraries_fast.py with default parameters for Illumina sequencing) (29). A total of 974 samples were sequenced, yielding 15,596,127 sequences total with an average of 16,012 sequences ± 11,840 SD per sample. Samples that generated fewer than 1,000 sequences after quality filtering were discarded. The remaining sequences were then binned into operational taxonomic units (OTUs) using closed-reference OTU picking based on 97% identity using the May 2013 build from the Greengenes reference database (30). Demultiplexed reads and associated metadata were deposited in the Qiita and EBI databases available through study ID 2024 (http://qiita.ucsd.edu). All statistical analyses were carried out in R using unpaired Welch's t tests with Benjamini Hochberg's false discovery rate (FDR) method with a q < 0.05 cutoff unless otherwise indicated. Analysis of alpha (Shannon diversity index) and beta diversity (weighted UniFrac and Bray Curtis) were performed in QIIME. All plotting was carried out in R with OTUs summarized to genus level and filtered to only include organisms which composed >1% of the total data set (vaginal heat map [see Fig. 6C]) or >0.01% (entire data set [see Fig. 3]). For association of individual genera with yogurt/control treatment in infant fecal samples, the data set was filtered to include only infant fecal samples, summarized to the genus level and genera were only kept if they represented a minimum 0.1% of the total reads. Significantly differentially abundant genera were determined by comparing groups using ALDEx2 (v0.99.2) (31, 32) by Wilcoxon rank-sum test with FDR correction of centered log-ratio transformed data using 1,000 Monte-Carlo instances drawn from the Dirichlet distribution with a minimally informative uniform prior of 0.5. Taxa that had zero counts in all samples were removed. This approach accurately estimates the range of technical variation inherent in microbiome data sets for all taxa, including those where some samples contained a zero, and some where they contained a positive count (31, 32). Heat map of all samples (n = 974) at the genus level. Samples were UPGMA (unweighted pair-group method with arithmetic averages) clustered be weighted UniFrac distances showing strong clustering of samples by body site (as indicated in the lower heat map). (A) Time series of vaginal samples shows increasing diversity and leading up to and after birth. Nugent score based on microscopy (wherein 0 to 3 is considered normal, 4 to 6 intermediate, and 7 to 10 is a state of microbial dysbiosis referred to as BV) is also displayed. (B) Comparison of Shannon’s diversity shows an increase within 10 days after birth. (C) Heat map of genera composing >1% of vaginal organisms shows a loss of lactobacilli and an increase in BV-associated organisms such as Prevotella. The breast milk samples collected at the final visit for all women in the study were analyzed for fat, protein, and lactose content. From the calculated lactose amount, the total carbohydrates were estimated. A total of 12 milk samples were excluded from this analysis due to the provider being HIV positive. Data were obtained for nine participants from the control group and six participants from the probiotic yogurt group. Samples were homogenized for 10 s prior to the removal of a 500-μl aliquot for DNA extraction using a sonicator (VCX 130; Chemical Instruments AB, Sollentuna, Sweden). The remaining sample (approximately 500 μl to 1 ml) was used for nutrient analysis. The lactose content was measured as described by Fusch et al. (33). Fat was extracted from the milk using Mojonnier ether extraction and then gravimetrically analyzed to get the total fat content (34). The protein content of the breast milk was determined as per the methods of Choi et al. (34). The true protein content was determined by subtracting the nonprotein nitrogen from the total nitrogen and multiplying this by 6.25 (35).