General Data Exploration
A total of 97 samples were analysed for bacterial 16S rRNA and 4006 unique Operational Taxonomic Units (OTUs) were identified according to 97% sequence similarity after the quality control and the removal of identified contaminants. The remaining sequences adequately depicted the total microbial composition per sample with the maximum being reached between 20000 to 30000 sequences. The sample sequencing library was normalized to allow for unbiased multiple comparisons, and this reduced the unique OTUs to 2141 revealing that 47% of the bacteria identified were represented by low number of sequences.
Assessment of Microbiome Pre-Treatment
We used a cross-sectional approach to first compare pre-treatment samples according to their hookworm infection status. Grouping of pre-treatment samples gave rise to 32 hookworm positives and 21 hookworm negative samples. Hookworm positive samples showed a higher number of unique OTUs (759) in comparison to hookworm negative samples (301) but both groups shared a total of 814 core OTUs (Fig. 2).
At the phylum level, the uninfected group showed Bacteroidota to have the greatest abundance, with Firmicutes, Proteobacteria and Actinobacteriota respectively, among taxa with a relative abundance greater than 1%. The hookworm positive sample group showed an increased relative abundance of both Firmicutes and Actinobacteriota, whilst Bacteroidota and Proteobacteria both saw reductions (Fig. 3).
Abundance distribution at the genus level revealed that Prevotella, Succinivibrio and Collinsella, accounted for the greatest proportion of taxa within the Bacteroidota, Proteobacteria and Actinobacteriota phyla, respectively. Firmicutes on the other hand demonstrated a wider distribution, with six genera having relative abundances greater than 1%. Among these genera Blautia, Subdoligranulum, Dorea and Clostridium_sensu_stricto_1 had increased relative abundances in the positive group (Fig. 4).
Positive samples recorded higher median values across all 3 alpha diversity indices (Species richness, Shannon and Simpson) (Fig. 5).
However, a significant difference in alpha diversity between positive and negative samples was also observed when measured using the species richness index (p = 0.0062). Comparing the beta diversity of hookworm positives as against negatives using weighted unifrac distances, revealed a substantial dissimilarity across the two groups (ADONIS: p = 0.0072, R2 = 0.06326). Through principal coordinate analysis using Bray-Curtis dissimilarity, it was observed that the positive group and the negative group formed distinct clusters clearly separated from each other based on compositional dissimilarities implying a difference in microbiome structure across the two groups (Fig. 6).
Employing linear discriminant analysis (LDA), deferential microbial testing was performed to identify potential microbial taxa driving dissimilarities between infection groups. Clostridia was identified to be the main taxa group accounting for dissimilarities within positive samples (LDA score > 4.75). On the other hand, Prevotella was identified as a significant taxon driving dissimilarities within negative samples (LDA score > 4.75) (Fig. 7).
Effect of Albendazole Treatment on Gut Microbiome
The microbiomes of individuals were assessed in stool samples collected before and 10–14 days after treatment to provide insight into the effects of deworming.
Pre- and post-treatment samples for successful therapy (pre-ST vs post-ST) and the failed therapy (pre-FT vs post-FT) groups were compared separately to observe how the microbiome is altered after either successful or sub-optimal treatment. Beta diversity analysis showed a reduction in microbiota composition after drug administration in both successful therapy and failed therapy groups. However, after successful clearance, the microbiota composition differed significantly from the pre-treatment infected state (ANOSIM: R = 0.12; p = 0.02) (Fig. 8) whilst the reduction between pre- and post-treatment samples after failed therapy showed no statistical significance (ANOSIM: R = 0.008; p = 0.35) (Fig. 9).
We proceeded to compare the state of the microbiome based on treatment outcomes. We analyzed the microbiome structure and composition of individuals who cleared infection (post-ST) against individuals who remained infected (post-FT) after treatment using linear discriminant analysis. When compared, the microbiota composition of post-FT samples was represented most significantly by genera within the class Clostridia such as Blautia and Ruminococcus. Post-ST samples, on the other hand were represented by an increased abundance of Prevotellaceae (LDA score > 4) (Fig. 10).
From this result, we noticed a similarity between microbiota representing negative samples before treatment and those representing successful treatment groups post albendazole administration. This led us to hypothesize that clearing the infection (post-ST) restructured the microbiome to a composition similar to the pre-treatment non-infected state (NTC). Conversely, a significant difference in the microbiota composition existed between the two groups, implying the microbiome did not revert to a non-infected state (ANOSIM: R = 0.23; p = 0.01) (Fig. 11).
When compared to non-infected samples (NTC), the T1B samples still demonstrated a microbiome structure similar to the positive pre-treatment state with an increase abundance of Collinsella and genera within the class Clostridia such as Subdoligranulum, Clostridium_sensu_stricto_1 and Blautia (Fig. 12).
Predictive Function of Gut Microbiome in Successful and Failed Treatment Groups
We also investigated microbiota function on the basis of predicted metagenomes using PICRUSt2. We compared the differences between successful and failed albendazole treatment groups (post-ST vs post-FT) in the Kyoto Encyclopaedia of Genes and Genomes (KEGG) orthology (KO). Among 559 affiliated KEGG pathways, 65 were statistically significant in T1B and T0B groups (p < 0.05). Among these, 63 pathways were associated with the successful treatment group whilst 2 pathways were associated with the failed treatment group (p < 0.05) (Fig. 13).
Interestingly, pathways mainly associated with metabolism, biosynthesis of cofactors and membrane transporters were enriched in the successful treatment group whilst the failed treatment group showed enrichment of pyruvate ferredoxin oxidoreductase (PFOR) subunit pathways.
Gut Microbiome, Chronicity and Post-Treatment Reinfection
Individuals who demonstrated a pattern of reinfection (SR) were also analysed for association of microbial biomarkers during each time-point. These individuals included those who presented with hookworm at baseline, cleared the infection following treatment but were infected again at the 9-month follow-up study. Both alpha and beta diversity analysis showed no association in terms of microbial composition and diversity across samples within the successive reinfection group (p > 0.05).
Across the larger cohort, one individual maintained a persistent hookworm infection over all sample collection events (CTF). The microbiome state of this individual was also analysed to assess the effects of progressive treatment failure and chronicity of hookworm infection. When analysed, no association in terms of microbial composition and diversity across samples in the Constant Therapy Failure group was identified.