Methodological Challenges and Ecological Insights in Characterising the Semen Microbiome: a Low Microbial-Biomass Niche

Semen-associated microbiota have been hypothesised to influence semen quality and male factor infertility, but their characterisation is challenging due to the low microbial-biomass in semen samples. This thesis proposes that the inconsistent links between semen-associated microbiota and male factor infertility reported to date arise from inaccurate community characterisation.

Using boar semen as a model, a sequencing pipeline for the characterisation of semen-associated microbiota was developed and optimised. For DNA extraction, lysis of both Gram-positive and Gram-negative bacteria was evaluated when accounting for a complex sample matrix. qPCR targeting the V4 region of 16S-rRNA was used to quantify bacterial load within boar semen, which were estimated to contain between 1.2 × 10^5 and 2.4 × 10^7 bacterial cells / mL, dependent on sample-type, confirming their definition as low microbial-biomass.

Culture-based methods and qPCR were run in parallel to contextualise sequencing results, whilst Cell- and DNA-based mock communities were constructed to evaluate the sensitivity and accuracy of pipelines. In pilot datasets, mock communities provided evidence suggesting that boar semen was highly susceptible to technical artefacts, including sequencing-level cross-talk arising from the use of combinatorial dual indexes, platforms utilising ExAmp chemistry, and lane-sharing with high microbial-biomass samples.

Comparisons of V4 copy numbers with a boar GAPDH amplicon enabled approximation of a host:bacteria ratio in semen samples as >99%. This informed attempts to deplete semen of host DNA prior to microbial DNA extraction, which ultimately proved unsuccessful. Metagenomic sequencing on a subset of 16S-rRNA samples enabled cross-validation of the 16S-rRNA approach, although low-depth samples and the resulting fragmented assemblies limit the usefulness of this approach. Host depletion remains an important consideration, as off-target amplification of non-specific host DNA led to low-depth sampling in a final 16S-rRNA dataset. Despite this, for host-associated low microbial-biomass samples, 16S-rRNA remains the most feasible strategy.

In the final dataset, adoption of unique dual indexing and sequencing utilising MiSeq chemistry, reduced the frequency of apparent technical artefacts. In addition, inclusion of broadened sampling of adjacent environmental niches allowed for comparison of microbial communities by sample-type. High-abundance taxa detected within semen samples were shared across all sample-types, with beta-diversity metrics showing that the microbial community composition of semen samples was most closely matched to those originating from boar penis skin swabs. This comparative approach also enabled the detection of low-abundance taxa that were unique-to-semen, though these showed high inter-sample variability. The most prevalent unique-to-semen taxa were Fastidiosipila, a strict anaerobe that favours protein substrates with the potential to be resident within anaerobic micro-niches within the male reproductive tract, and Actinobaculum, an organism known to colonise the urogenital tract of boar. Nonetheless, the sparse ecological literature on Fastidiosipila and the ambiguity in identifying the precise provenance of taxa within a multi-organ tract when interrogating semen limits firmer conclusions.

Finally, comparisons of parameters of semen quality showed no statistical association with microbial community composition, with the largest effects resulting from differences in handling practices. Limited statistical power and the dominance of environmental noise remain important caveats to these conclusions, but the possibility remains that any true biological effects of microbiota on semen quality are small or indirect. In the absence of compelling data to the contrary, semen microbiome profiles cannot be recommended as clinical biomarkers of semen quality or fertility.

This thesis sets out a rigorous methodological framework for studying semen-associated microbiota, and by extension, other low microbial-biomass niches. Mock communities, reagent blanks, and environmental sampling controls are essential for identifying inherent biases. Future work should assess overlap with adjacent niches and incorporate environmental sampling as standard. Tissue localisation studies such as fluorescent in-situ hybridisation could strengthen evidence of tract residence. Through careful consideration of study design, it should become possible to move from descriptive sequencing towards genuinely informative biological insight.