All infants aged 17–34 weeks recruited to the RECAMP-OPV trial (Varma et al., 2019) (NCT03460002 on www.clinicaltrials.gov), and living in village clusters distant 2 h-drive from the capital Bissau were eligible to participate in the Micro-OPV cohort study (see the experimental design of Micro-OPV study in Figure 1A and the Guinea–Bissau map with the covering area of the Micro-OPV study in Figure 1B). This age range was planned because infants at those ages have already been eligible for a set of vaccines as part of the routine immunization plan, except MV (World Health Organisation [WHO], 2021a; Figure 1A). Area delimitation was due to logistic reasons concerning the transportation and preservation of biological samples, which are further detailed below. Biosampling took place from August 2018 to December 2018.

In RECAMP-OPV clinical trial, clusters were randomly classified 1:1 as Intervention clusters (with infants revaccinated with an extra dose of OPV in addition to those administrated according to routine immunization plan) or Control clusters (whose infants received no additional OPV) based on externally generated random numbers. All clusters received the same number of visits, at baseline and 2 months post-baseline. Due to the randomization design, recruited infants shared similar major microbiome determinants, such as, at term gestational age, vaginal delivery, and feeding progression. Given the exploratory nature of the Micro-OPV study, we conveniently planned to recruit 60 infants from each cluster to participate. At baseline and 2 months after baseline, in addition to the main trial procedures required by the RECAMP-OPV, including the administration of an extra dose of OPV to infants from the Intervention clusters at baseline, fecal and nasopharyngeal biological content was collected, and clinical data registered. A total of 47 participants out of 60 revaccinated with OPV and 47 participants out of 58 not revaccinated completed the 2 months follow-up in the Micro-OPV study.

Study villages are agricultural-based, served by well water and pit-latrines, and just 30–50% of them have access to electricity supply. Households live very close to each other, often sharing the same kitchen space and kitchenware, and people of the same household share the same food container. Also, people live very close to farm animals (chicken, pig, goat, and cow) and pets, such as dogs and cats.

Procedures in the field under the Micro-OPV study were carried out in the rainy season (August to December) in 2018. After assurance of all ethical aspects described below, parents or legal guardians of selected infants attended a questionnaire-based interview at both time-points. All interviews were conducted in Créole, the lingua-franca, by the same BHP field worker under the researchers’ supervision. Information on major microbiome determinants and possible cofounders of the effect of OPV revaccination, such as housing conditions, mother’s health during pregnancy and delivery, and infant’s health since delivery to present, including delivery conditions, feeding history, previous infections, hospitalizations, and use of medicines were recorded at enrollment. Feeding progression, health problems, hospitalizations, and use of medicines since the enrollment were registered at follow-up.

After the interviews, infants were submitted to a clinical evaluation and measurement of axillary temperature. Anthropometric data [weight, height, and mid-upper-arm-circumference (MUAC)] were also noted. No infant met exclusion criteria at enrollment, namely, fever defined as an axillary temperature ≥38°C, MUAC < 110 mm in an infant older than 6 months or previous history of allergic reaction due to vaccination.

Nasopharyngeal (NP) samples and fecal (F) samples were collected from all participants at enrollment and 2 months later at follow-up using ultra minitip flexible and regular flocked swabs (FLOQSWABS^®, Copan, Brescia, Italy), respectively. Immediately after the collection procedure, swabs were immersed in 1 ml of DNA/RNA Shield™ (Zymo Research, Irvine, CA, USA) in 15 ml Falcon tubes and kept cold until laboratory processing. At enrollment, after sampling procedures, one OPV-dose (two drops) was administered to infants belonging to the Intervention clusters, by the experienced BHP nurse. Infants belonging to the Control clusters received no intervention. No additional OPV dose was administrated between the two time-points to any of the participants.

The raw material was transferred from Falcon tubes to 2 ml Cryotubes in the laboratory facility of the Guinea-Bissau National Institute of Health in sterile conditions, kept at −20°C for 1 month in the BHP premises, transported to the Institute of Hygiene and Tropical Medicine, Universidade NOVA de Lisboa, Portugal (IHMT-NOVA) in dry ice and kept at −80°C until DNA extraction.

DNA was extracted from 500 μl of each raw sample through ZymoBIOMICS™ DNA Miniprep Kit (cat. no°.: D4300, Zymo Research, Irvine, CA, USA) following manufacturer’s orientations. Mechanical lysis was standardized through a ZymoBIOMICS™ Microbial Community Standard II (Log Distribution) (cat. n°.: D6310, Zymo Research, Irvine, CA, USA) using a Mini-BeadBeater device (model EUR16, BIOSPEC Products, Inc., Bartlesville, OK, USA). All DNA samples presented A₂₆₀/A₂₈₀ ratio >1.8 measured by a NanoDrop™ 1000 Spectrophotometer device (Thermo Fischer Scientific, Waltham, MA, USA) and 100 ng of DNA per sample, measured in a Qubit Fluorometer (Thermo Fischer Scientific, Waltham, MA, USA), were used to prepare the library.

Paired (F and NP) extracted DNA samples of 94 participants who completed the 2 months of follow-up (47 OPV revaccinated infants and 47 controls) were deeply characterized by 16S rRNA amplicon sequencing to identify different changes in bacterial diversity and composition of gut (G) and upper respiratory (R) bacterial microbiotas over time. For that, paired-ended amplicon deep sequencing of V3 and V4 variable regions of the gene that codifies the 16S rRNA was performed using previously evaluated primers (341F: 5′-CCTACGGGNGGCWGCAG-3′; 805R: 5′-GACTACHVGGGTATCTAATCC-3′), preceded by heterogeneity spacers (Klindworth et al., 2013). Library construction and sequencing were performed on an Illumina MiSeq instrument (Illumina Inc., San Diego, CA, USA), using a 600 cycle V3 kit.

Demultiplexing was done with the bcl2fastq Conversion Software (Illumina Inc., San Diego, CA, USA). Cutadapt (version 2.3) was used to trim off heterogeneity spacers and primers at an error rate of 8% (allowing for one mismatch per primer) (Martin, 2011). The resulting sequences were quality-filtered, truncated, denoised, merged, and chimera-filtered through the DADA2 pipeline (version 1.12.1). High-resolution amplicon sequence variants (ASVs) were inferred from trimmed reads per run with default settings, except for truncation lengths (forward reads: 270 bp, reverse reads: 210 bp) (Callahan et al., 2016). Consensus removal of chimeras was performed on data from all runs. Samples with a read count <10,000 after quality filtering were re-sequenced. Taxonomic assignment of the resulting ASVs was done by means of dada2’s assignTaxonomy() and addSpecies() functions with the Silva reference database and species-level training set (version 132), respectively (Callahan, 2018). Subsequently, the ASV count table and the taxonomic table were integrated with the R package phyloseq (McMurdie and Holmes, 2013) to perform ecological and statistical analyses of microbiota data.

The data set was split into F and NP samples and the R package decontam was used to identify and remove contaminants (Davis et al., 2018). For this, controls used in DNA extraction, the DNA/RNA Shield™ (Zymo Research, Irvine, CA, USA) solution and the controls from PCR were used. The decontam method “either” was used in F and NP samples data set adjusted for each one (frequency threshold 0.075, prevalence threshold 0.5 and frequency threshold 0.05, prevalence threshold 0.5, respectively). After computational contaminant removal, read counts of re-sequenced samples were merged. From the F sample data set, 67 ASVs were identified as contaminants and removed. In addition, 40 ASVs classified as Archaea, Cyanobacteria, Planctomycetes, Chloroflexi, Rhizobiales, Rhodobacterales, or Rhodospirillales were removed manually, as our major interest was to observe the behavior of the bacterial microbiota of gut and upper respiratory tract. Moreover, 61 ASVs that were not classified to at least order-level were excluded. Twenty-seven ASVs from the NP sample data set that were identified as contaminants were excluded. In addition, 99 ASVs that were not classified to at least order-level were excluded, as well as 224 ASVs classified as Eukaryota, Archaea, Cyanobacteria, Planctomycetes, Chloroflexi, Deinococcus-Thermus, Kiritimatiellaeota, Rhizobiales, Rhodobacterales, Rhodospirillales, or Oceanospirillales. A number of 2,373 ASVs were obtained from the 94 F samples, while 770 ASVs were obtained from the 94 NP samples, both distributed by seven taxonomic ranks. The average read count per sample after all pre-processing steps was 37,897 (F samples) and 31,139 (NP samples).

Statistical significance levels were set at <0.05 for all tests used in analyses of epidemiological or sequencing data unless otherwise stated.

No major differences in participant profiles were observed between the 47 (60% female) OPV revaccinated infants (participants in the Intervention group) and the 47 (49% female) controls (participants in the Control group) at enrollment (Table 1). Although a significantly higher mean length was observed for OPV-revaccinated infants at enrollment, it may reflect the slightly higher age of the infants in this group and, possibly missed length measures for six participants in the Control group and for two OPV-revaccinated participants. This difference was no longer observed at follow-up when complete data were recorded from all infants. No major differences in clinical conditions anticipated to affect the microbiome at the time of sampling and during the 2 months follow-up were observed between groups (Tables 1, 2).

Participant profiles after excluding infants that used antibiotics during follow-up (seven OPV revaccinated infants and three controls) were similar to that shown for the 47 infants in each group (Supplementary Tables 6, 7).

Infants in both groups were similar in characteristics of major microbiome determinants at enrollment, such as, environmental, birth, diet, and clinical factors that influence the microbiome development (see Supplementary Tables 1–4). Supplementary Table 3 describes, specifically, the feeding history of participants at enrollment and follow-up. No significant differences were observed between groups. Breastfeeding coverage was very high among all participants at both time-points. The same was observed regarding porridge feeding and water administration, including median number of doses administrated per week. Regarding another sources of nutrients, fruits were introduced at follow-up for both groups of infants, however, to a lower extent. Significant differences were observed regarding a lower number of residents per house younger than 15 years old (5 vs. 6, p = 0.02) in the group with OPV revaccinated infants, together with a relatively higher use of tap water (5 vs. 0, p = 0.03) and charcoal to cook (6 vs. 0, p = 0.03) in this same group, although most infants in both groups lived in households which used well water and wood to cook. All deliveries in both groups came from at term pregnancies of newborns born vaginally and almost 50% at home. Few mothers and infants in both groups had been tested for human immunodeficiency virus (HIV), and no positive cases had been detected. Chemoprophylaxis to malaria in pregnancy was taken by almost 80% of mothers and few of them reported malaria in pregnancy. Most infants were exclusively breastfed up to the 5th month of life, and then started to receive porridge and water. Few infants had malaria before enrollment, while a higher proportion had already experienced diarrhea episodes. More than 50% of infants in both groups had used medicines up to 1 month before enrollment, mainly Paracetamol and equally few infants in both groups had used antibiotics previously, mainly Amoxicillin. No significant differences were observed regarding the proportion of infants in both groups with complete doses of recommended vaccines according to routine immunization schedule until 14 weeks of age (World Health Organisation [WHO], 2021a).

The distribution of Shannon (H) and Inverse of Simpson (InvSimpson) indexes in both groups at enrollment and follow-up is shown in Figure 2.

First, we showed that infants in both groups presented similar indexes for both microbiotas at enrollment as observed in Figures 2A,B (for G microbiota) and Figures 2C,D (for R microbiota), evidencing similar exposition to major microbiomes determinants between groups, as previously shown. Then, both indexes significantly increased from enrollment to follow-up in G microbiota of OPV revaccinated infants (H p = 0.001, InvSimpson p < 0.001) and controls (H p = 0.03, InvSimpson p = 0.03), whereas in R microbiota, the H index only significantly increased in OPV revaccinated infants (p = 0.011). Changes in InvSimpson index in R microbiota were not significant for both groups in the same period.

Next, we compared the magnitude of changes over time for both microbiotas in each group to test whether they might still be greater in OPV revaccinated infants due to revaccination. A similar magnitude of changes in alpha-diversity over time [comparison (3) in Figure 2] was found in G (Figures 2A,B) and R microbiotas (Figures 2C,D) in infants from both groups.

Similar results were obtained after excluding infants in both groups who used antibiotics during follow-up period (data not shown).

In summary, alpha-diversity for both microbiotas were similar between groups at enrollment, i.e., before infants in the Intervention group have received OPV revaccination. Two months later, we observed that alpha-diversity increased similarly in both groups over time, despite infants in the Intervention group have received the OPV boost.

Visualization of differences in bacterial composition between groups and time-points for G and R microbiota samples were performed using PCoA based on Bray–Curtis dissimilarities (Supplementary Figures 1A,B). After ensuring homogeneity in whitin-group variations by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses for G and R microbiotas showed that differences between bacterial composition were based on group and time point for G (p < 0.01) and R (p = 0.001) bacterial microbiotas. When analyses were deepened to identify whether differences were attributable to time point, group or an interaction of both, we identified that differences for G bacterial microbiota were more attributable to time point (p < 0.01) than group (p > 0.05) or an interaction of both (p > 0.05), whilst for R bacterial microbiota differences were attributable to time-point (p < 0.05) and group (p = 0.001) but not due to an interaction of both (p > 0.05).

To assess differences in bacterial composition for both microbiotas over time and to test if they were larger in OPV revaccinated infants than controls, firstly, data were split according to time-point and G and R samples from OPV revaccinated and controls at enrollment were analyzed. After Hellinger transformation of data and ensuring homogeneity in within-group variations by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses showed no differences between groups regarding bacterial communities in G (p > 0.05) and R (p > 0.05) microbiotas at enrollment, i.e., at baseline, before infants in the Intervention group being revaccinated with an extra-dose of OPV. Next, we split up phyloseq object by group and tested if there was a difference in within-group bacterial composition between samples at enrollment and samples collected 2 months later. After Hellinger transformation of data and testing for homogeneity in whitin-group variation by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses showed significant differences over time in bacterial composition in R microbiota samples of OPV revaccinated infants (p < 0.05) and controls (p = 0.01), indicating changes over time in R bacterial microbiota samples in both groups. Using the same methodology, PERMANOVA analyses showed that the bacterial composition of G microbiota samples of controls was different between time points (p < 0.05). Heterogeneity in within-group variation over time in G microbiota samples of OPV revaccinated infants avoided PERMANOVA analyses to verify differences in bacterial composition over time. This fact suggests that changes over time had occurred in G microbiota of OPV revaccinated infants, at least in some samples.

To assess whether changes over time in bacterial composition were greater in OPV revaccinated than controls, we carried out pairwise comparisons of within- and between-group Bray–Curtis dissimilarities on Hellinger transformed data. Data for G and R microbiotas are shown in Figures 3A,B, respectively. First, we observed that bacterial communities in G microbiota in both groups at enrollment (first black box, Figure 3A) were no more different from each other than within the two groups at enrollment [first (p_adjsted = 0.524) and third gray boxes (p_adjusted = 0.540), Figure 3A]. The same was found for bacterial communities in R microbiota [first and third gray boxes (p_adjusted = 0.199), Figure 3B]. Then, the Control group was used as a reference for changes in OPV revaccinated infants over time, showing that bacterial communities in G microbiota changed significantly more over time in OPV revaccinated infants than controls (blue box vs. second black box, respectively, p_adjusted < 0.001, Figure 3A). Results for R microbiota samples were similar, also showing a significantly larger change over time in OPV revaccinated infants than controls (blue box vs. second black box, respectively, p_adjusted = 0.018, Figure 3B).

In summary, results suggest that changes over time in the composition of G and R bacterial microbiotas occurred in both groups and were greater in OPV revaccinated infants. The large within-group variation found in G microbiota samples in the group of OPV revaccinated infants at follow-up may suggest favorable growth of specific taxa in certain samples. Similar results were found after excluding antibiotic users in both groups (data not shown).

Differences in the relative abundance of the 50 most abundant genera and higher taxonomic levels for both microbiotas and groups over time were visualized in differential taxonomic heat trees shown in Supplementary Figures 2A–D, fully described in the same Supplementary material.

LEfSe analysis was used to identify taxa groups that significantly changed in both microbiotas and groups over time (LDA score ≥4). Figure 4A illustrates which taxa groups significantly decreased over time in OPV revaccinated infants (yellow bars), i.e., that were higher at enrollment, and which ones significantly increased over time (green bars). B. longum, Escherichia/Shigella, Campylobacter, and Clostridium butyricum are the taxa that most decreased in OPV revaccinated (range of p_adjusted values from 0.0006 to 0.016), whereas Prevotellaceae and Streptococcus were the ones that most increased over time (range of p_adjusted values from 0.001 to 0.03). A small number of taxa (dark red bars in Figure 4B) significantly decreased in G microbiota in the Control group, mainly B. longum (p_adjusted = 0.01), whereas a greater number of taxa significantly increased over time (blue bars in Figure 4B), mainly Prevotella_9, Bifidobacterium kashiwanohense and Campylobacter (range of p_adjusted values from 0.0007 to 0.02).

Corynebacterium_1 was the taxon that most decreased in R microbiota in OPV revaccinated (yellow bars in Figure 4C, range of p_adjusted values from 0.0003 to 0.0006), whereas Haemophilus (p_adjusted = 0.006) and Staphylococcus (p_adjusted = 0.002) were those that most increased over time (green bars in Figure 4C). Corynebacterium_1, Staphylococcus, and Fusobacterium were those that most decreased in the Control group, mainly the former (p_adjusted = 0.0002, dark red bars in Figure 4D), whereas Haemophilus (p_adjusted = 0.03) and Streptococcus p_adjusted = 0.001) were the ones that most increased over time (blue lines in Figure 4D).

When comparing changes over time among taxa with relative abundance greater than 1%, we can observe more taxa groups in OPV revaccinated (Figure 5A) than in controls (Figure 5B) that significantly changed in G microbiota. The opposite occurred in R microbiota, for OPV revaccinated (Figure 5C) and controls (Figure 5D). A significant decrease in the median proportion of B. longum over time was observed both in OPV revaccinated (25–14.3%, p_adjusted = 0.0006) and controls (25.3–11.6%, p_adjusted = 0.01) (Figure 5A). A significant increase in the median proportion of Prevotellaceae (7.2–17.4%, p_adjusted = 0.005) family was observed in OPV revaccinated infants (Figure 5A), whereas a significant increase in the median proportion of Prevotella_9 genus was observed in the Control group (1.4–7.0%, p_adjusted = 0.02) (Figure 5B). Additionally, a significant decrease in Escherichia/Shigella (5.8–3.4%, p_adjusted = 0.01) was observed within OPV revaccinated group over time (Figure 5A), with a significant decrease of an undetermined Escherichia/Shigella species (4.9–3%, p_adjusted = 0.02). On the other hand, the median proportion of Campylobacterales (0.9–2.6%, p_adjusted = 0.03) and Lachnospiraceae (0.8–2.2%, p_adjusted = 0.007) significantly increased over time in G microbiota in the Control group (Figure 5B). The main change in R microbiota of the OPV revaccinated infants over time (Figure 5C) was the significant reduction in the median proportion of a Corynebacterium_1 species (3.4% to 0.8%, p_adjusted = 0.0006). The same was also observed in the Control group (4.8–1%, p_adjusted = 0.005) (Figure 5D), but simultaneously with an increase in the median proportion of other several taxa groups (Figure 5D), such as Firmicutes (from 19.2–22.8%, p_adjusted = 0.04), Streptococcus genus (2.9–11.8%, p_adjusted = 0.001) and an undetermined Streptococcus species (2–5.9%, p_adjusted = 0.01), followed by the increase in Pasteurellaceae family (11.4–20.5%, p_adjusted = 0.03), including organisms from Haemophilus genus (11.3–20.5%, p_adjusted = 0.03).

In summary, taxonomic composition of G and R microbiotas changed differently over time in OPV revaccinated infants vs. controls. A more proportional alternance over time between the most abundant commensals (B. longum and Prevotellaceae) was observed in G microbiota of OPV revaccinated infants. Also, it was observed a more abundant and diversified community of Prevotellaceae in G microbiota of OPV revaccinated infants, followed by a reduction in the median proportion of Escherichia/Shigella and limited increase in the median proportion of Campylobacterales. In R microbiota, it was observed a limited increase in the median proportion of Streptococcus and Haemophilus in OPV revaccinated infants, whereas in controls these taxa were present among those with relative abundance greater than 1% at follow-up. Similar results were found after excluding antibiotic users in both groups (data not shown).