Cohort characteristics
Overall, there were 278 mother-infant pairs included in this analysis; 82 were from South Africa, and 196 were Nigerian. Several demographic and socioeconomic characteristics differed by study site (Table 1). At enrolment, Nigerian mothers were older (mean age 31 (standard deviation (SD) ± 5.31) versus 28 (SD ± 5.38) years; P = 0.001) with higher gravidity (median 2 [interquartile range (IQR) 1–4] versus 1 [IQR 1–2]; P < 0.001) and lower body weight (mean 62.87 (SD ± 11.51) versus 72.69 (SD ± 13.86) kg; P < 0.001) than South African mothers. While electricity was equally available for participants from both countries, significantly more mothers in South Africa had a refrigerator and running water at home, and significantly more Nigeria mothers lived in formal housing (all P < 0.001). The weight-for-length z score (wflz) of Nigerian infants was significantly lower than that of South African infants at 15 weeks of age (0.54 versus 0.86, P = 0.023). All Nigerian infants were exclusively breastfed (EBF) until 15 weeks of life, whereas only 58.5% of South African mothers reported still EBF at 15 weeks postpartum (P < 0.001). There were no significant differences in cohort characteristics by HIV exposure, except mothers of iHUU had higher formal education than mothers of iHEU (P = 0.002; Supplementary Table S1).
Table 1
Cohort characteristics by Site
| | South Africa | Nigeria | P |
| | (N = 82) | (N = 196) | |
Maternal characteristics | | | | |
Mother's age at delivery (years; mean (SD)) | | 28 (5.38) | 31 (5.31) | 0.001 |
Education (n; %) | None | 0 (0.0) | 2 (1.0) | < 0.001 |
| Elementary | 5 (6.1) | 65 (33.2) | |
| Secondary | 72 (87.8) | 73 (37.2) | |
| Higher | 5 (6.1) | 56 (28.6) | |
Unemployed (n; %) | | 59 (72.0) | 6 (3.1) | < 0.001 |
Formal housing (n; %) | | 33 (40.2) | 185 (94.4) | < 0.001 |
Electricity (n; %) | | 78 (95.1) | 178 (90.8) | 0.332 |
Refrigerator (n; %) | | 70 (85.4) | 96 (49.0) | < 0.001 |
Running water (n; %) | | 38 (46.3) | 48 (24.5) | 0.001 |
Marital status (n; %) | Married/ living together | 25 (30.5) | 186 (94.9) | < 0.001 |
| Single | 57 (69.5) | 10 (5.1) | |
Gravidity (n; median [IQR]) | | 1 [1, 2] | 2 [1, 4] | < 0.001 |
Mother’s weight at enrollment (kg; mean (SD))a | | 72.69 (13.86) | 62.87 (11.51) | < 0.001 |
Infant characteristics | | | | |
iHEU (n; %) | | 61 (74.4) | 141 (71.9) | 0.787 |
Male (n; %) | | 41 (50.0) | 94 (48.0) | 0.858 |
Gestational age at delivery (weeks; median [IQR]) | | 39.30 [38.02, 40.38] | 39.95 [38.98, 40.62] | 0.011 |
Vaginal delivery (n; %) | | 82 (100.0) | 167 (85.2) | 0.001 |
Wflz at W15 (median [IQR])b | | 0.86 [0.32, 1.90] | 0.54 [-0.64, 1.42] | 0.023 |
Mode of feeding at W15 (n; %) | Exclusive breastfeeding | 48 (58.5) | 196 (100.0) | < 0.001 |
| Mixed feeding | 34 (41.5) | 0 (0.0) | |
IQR, Interquartile range; SD, Standard deviation; iHEU, infants who are HIV-exposed uninfected; W15, 15 weeks of age; Wflz, Weight-for-length z score. |
aMissing data from 5 Nigerian participants; bMissing data from 41 participants (South Africa, n = 16; Nigeria, n = 25). |
Gut microbiota differs substantially between South African and Nigerian infants in the first week of life
Of the 524 samples sequenced, 442 samples passed the quality filtering. Of these, 164 (47 South African and 117 Nigerian) out of the 278 infants had gut microbiota data available at both time points. Gut microbiota composition differed significantly by study site during the first week of life (Fig. 1A). Within-sample microbial diversity (Shannon index) was higher among South African than Nigerian infants (P < 0.0001; Fig. 1B). In addition, microbial community composition was significantly different by geographical location, although site only explained 6% of the community composition (Fig. 1C; PERMANOVA P < 0.001). Geographical location remained significantly associated with α- and β- diversity when adjusted for sequencing batch or demographic factors that significantly differed between countries, namely maternal marital status, weight, age, gravity, education level, occupation, type of house, access to a refrigerator or running water, mode of delivery and infant gestational age (P < 0.001 for both α- and β-diversity). In addition, α- and β- diversity remained significantly different by the geographic location when the comparison was made strictly among samples collected in the first day of life (Supplementary Figure S1).
At baseline, most South African infants had gut microbiota consisting of (1) Actinobacteriota, including several Bifidobacterium species (such as B. longum subspecies longum, B. catenulatum, and B. breve) and Collinsella aerofaciens, (2) Firmicutes, including Streptococcus species (such as S. salivarius, S. caprae, and S. lutetiensis) and Veillonella dispar and (3) Proteobacteria which mainly consist of E. coli, which was named “cluster 1” identified by PAM clustering (Fig. 1A). On the other hand, the majority of Nigerian infants’ gut microbiota was classified as community cluster 3, dominated by (1) Actinobacteriota, mainly B. longum subspecies infantis and (2) Firmicutes, including Staphylococcus species (such as S. haemolyticus and S. saprophyticus) and Enterococcus species (such as E. faecalis and E. faecium).
Age is a major driver of microbiota development, but microbial succession differs between sites
We next assessed the gut microbiota longitudinally. The α-diversity in South African infants increased significantly from week 1 to week 15 (P = 0.036), while α-diversity in Nigerian infants significantly decreased (P < 0.0001) (Fig. 2A). There was a clear separation of microbial community composition among Nigerian samples by age, which was less evident for South African infants (Fig. 2B). In agreement, the dominant bacterial changed only marginally from week 1 to week 15 among South African infants, while Nigerian infants experienced a shift from a Firmicutes-dominated microbiota (cluster 3) to one dominated by Bifidobacterium infantis and Streptococcus salivarius (cluster 2) at 15 weeks of age (Fig. 2C; Fig. 3). The significant differences observed between countries in α- and β-diversity remained the same over the 15 weeks when the comparison was strictly among EBF infants or vaginally delivered infants (Supplementary Figure S2-3).
HIV exposure has a subtle effect on the gut microbiota regardless of the geographical location
There were no significant differences in α-diversity (Supplementary Figure S4A), β-diversity (Supplementary Figure S4B), or PAM cluster transition (Supplementary Figure S4C-D) by HIV exposure status in either country. Differential abundance testing using ANCOM-BC analysis of differential abundance in microbiome data was performed adjusting for feeding mode at the week 15 time point [34]. Several bacterial taxa significantly associated with HIV exposure status in South Africa (Table 2). Several Enterococcus species (E. faecalis, E. faecium, E. gilvus, and E. raffinosus) were significantly more abundant in iHEU than iHUU at week 15 (Loge fold change (LFC): 0.61, 0.57, 1.02, and 0.76, respectively). Moreover, Collinsella aerofaciens (LFC: 0.72 at week 1 and 1.18 at week 15) and Klebsiella quasipneumoniae (LFC: 0.84 at both week 1 and week 15), which are known to be pathobionts [37, 38], were consistently more abundant in iHEU during the first 15 weeks of life. In contrast, no bacterial taxa were differentially abundant by HIV exposure in the Nigerian cohort.
Table 2
ANCOM-BC analysis of gut microbiota in South African HIV-exposed versus un-exposed infants.
Taxonomy (Genus, Species) | Taxon ID | LFCa |
At 1 week of age | | |
Klebsiella variicola | ASV46 | 1.22 |
Sutterella (unclassified) | ASV150 | 1.02 |
Holdemanella (unclassified) | ASV53 | 1.00 |
Parabacteroides merdae | ASV101 | 0.98 |
Catenibacterium (unclassified) | ASV218 | 0.96 |
Blautia obeum | ASV59 | 0.93 |
Senegalimassilia (unclassified) | ASV145 | 0.87 |
Bifidobacterium breve | ASV10 | 0.84 |
Klebsiella quasipneumoniae | ASV36 | 0.84 |
Libanicoccus (unclassified) | ASV153 | 0.81 |
Blautia (unclassified) | ASV225 | 0.80 |
Ruminococcus torques group (unclassified) | ASV75 | 0.72 |
Collinsella aerofaciens | ASV25 | 0.72 |
Subdoligranulum (unclassified) | ASV251 | 0.70 |
Bacteroides vulgatus | ASV83 | 0.70 |
Sutterella (unclassified) | ASV496 | 0.67 |
Klebsiella pneumoniae | ASV39 | 0.66 |
Megamonas (unclassified) | ASV169 | 0.65 |
Romboutsia ilealis | ASV93 | 0.64 |
Senegalimassilia (unclassified) | ASV171 | 0.64 |
Faecalibacterium (unclassified) | ASV505 | 0.58 |
Fusobacterium mortiferum | ASV278 | 0.57 |
Enterococcus faecium | ASV7 | 0.57 |
Parabacteroides distasonis | ASV138 | 0.54 |
Actinomyces (unclassified) | ASV668 | -0.54 |
Parabacteroides distasonis | ASV203 | -0.57 |
At 15 weeks of age | | |
Streptococcus gallolyticus | ASV44 | 1.33 |
Collinsella aerofaciens | ASV25 | 1.18 |
Clostridium innocuum group (unclassified) | ASV336 | 1.13 |
Enterococcus gilvus | ASV157 | 1.02 |
Klebsiella quasipneumoniae | ASV42 | 0.84 |
Veillonella atypica | ASV163 | 0.83 |
Enterococcus raffinosus | ASV338 | 0.76 |
Enterococcus (unclassified) | ASV40 | 0.71 |
Bifidobacterium adolescentis | ASV296 | 0.71 |
Enterococcus raffinosus | ASV51 | 0.71 |
Lactococcus lactis | ASV206 | 0.66 |
Enterococcus faecalis | ASV5 | 0.61 |
Granulicatella (unclassified) | ASV681 | 0.59 |
Dorea formicigenerans | ASV229 | 0.59 |
Faecalibacterium prausnitzii | ASV103 | 0.52 |
Staphylococcus (unclassified) | ASV41 | 0.51 |
Lactobacillus rhamnosus | ASV191 | -0.51 |
Klebsiella michiganensis | ASV266 | -0.51 |
Lactobacillus gasseri | ASV161 | -0.53 |
Prevotella copri | ASV405 | -0.55 |
Megasphaera elsdenii | ASV167 | -0.58 |
Olsenella (unclassified) | ASV97 | -0.58 |
Prevotella (unclassified) | ASV176 | -0.83 |
Bacteroides caccae | ASV552 | -0.93 |
Olsenella (unclassified) | ASV120 | -1.89 |
Ruminococcus torques group (unclassified) | ASV75 | -2.43 |
aAbundance in iHEU in relation to iHUU. ANCOM-BC: Analysis of Compositions of Microbiomes with Bias Correction; LFC, Loge fold change; ASV, amplicon sequence variants; iHEU, infants who are HIV-exposed uninfected; iHUU, infants who are HIV-unexposed uninfected. |
Maternal HIV status and infant gut microbes influence infant TT vaccine response
In Nigeria, it is recommended that pregnant women receive TT booster vaccinations, whereas this is not policy in the Western Cape, South Africa [39]. Therefore, not surprisingly, infant anti-TT IgG titers in the first week of life, representing maternally transferred antibodies, were significantly lower among South African infants than Nigerian infants (median 1.0 versus 1.5 IU/ml, adj P = 0.002; Supplementary Figure S5A). In contrast, titers did not differ between South African and Nigerian infants at 15 weeks of age (median 1.9 versus 1.6 IU/ml, adj P = 0.280). We investigated the correlation of TT vaccine response between mother and infant pairs living in Nigeria. Anti-TT IgG titers were strongly correlated at week 1. However, iHEU mother-infant anti-TT IgG titers showed a lower Pearson’s correlation coefficient compared to iHUU (R: 0.72 versus 0.95) (Fig. 4A). The correlation between maternal and infant anti-TT IgG titers was no longer evident by 15 weeks of age in either iHEU and iHUU (Supplementary Figure S5B). We did not see any difference in anti-TT IgG titers among mothers by their HIV status (Supplementary Figure S5C). However, iHEU had significantly lower TT IgG concentrations than iHUU at 15 weeks of life (P = 0.016), and this remained significant after adjusting for multiple comparisons (adj P = 0.031; Fig. 4B). However, the difference between iHEU and iHUU at week 15 was no longer statistically significant when infants were compared separately by study site (adj P = 0.290 in South Africa and adj P = 0.180 in Nigeria; Supplementary Figure S5D).
Since gut microbiome is thought to modulate the development of the immune system [15], we intended to investigate the relationship between infant gut microbiota and TT vaccine response at week 15. We did not see consistent correlations between 15-week anti-TT IgG titers and Shannon diversity of either time point (Fig. 5A).
To further explore factors associated with infant TT vaccine response at week 15, we conducted Lasso regression analysis. Rank-transformed top 50 ASVs at either week 1 or week 15, HIV exposure status, and anti-TT IgG titers at week 1 were included as explanatory variables to investigate the predictor, TT vaccine response at 15 weeks of age. In South Africa, infant HIV exposure status showed a strong negative association with 15-week TT vaccine response (β-coefficient = -0.44), and the rank-transformed bacterial taxon abundance at week 1, including Streptococcus salivarius (β-coefficient = 0.038), Bacteroides dorei (β-coefficient = 0.016), Collinsella aerofaciens (β-coefficient = 0.015), and Sutterella wadsworthensis (β-coefficient = -0.011) were independently associated with the vaccine response, albeit with weaker β-coefficients than HIV-exposure (Fig. 5B; Supplementary Table S2). In contrast, no variables were selected as predictors of the TT vaccine response in the Nigerian cohort. Previously, it has been shown that passively transferred maternal antibody interferes with infant TT vaccination response [40]. Since Nigerian infants showed significantly higher maternal antibodies than South African infants at week 1 (Supplementary Figure S5A), we speculated that these maternal TT antibodies may have masked any associations underlying the infant TT vaccine response. For this reason, we re-assessed the Lasso regression without including week 1 anti-TT IgG data in the explanatory variables (Supplementary Figure S6; Supplementary Table S3). Although there was no change in the result for the South African infants (Supplementary Figure S6A), HIV exposure and several bacteria present at 15 weeks of age, including S. salivarius, were independently associated with the TT vaccine response in Nigerian infants (Supplementary Figure S6B). However, the β-coefficients for all selected predictors were small, including HIV exposure status.