Comparative analyses of Ion Torrent V4 and Illumina V3-V4 16S rRNA gene metabarcoding methods for characterization of cervical microbiota: taxonomic and functional profiling

Abstract

Next generation sequencing technologies have enabled characterization of microbiota at an unprecedented resolution. Microbiota research methodologies are constantly improving and differing in the primer choice, sequencing technology, and bioinformatics pipeline used. These variations could impact comparability of microbiota results across studies. Here, we used 19 cervical samples to compare the performance of two metabarcoding methods that differ in the primer choice and sequencing technology used. We sequenced the V4 16S rRNA libraries with Ion Torrent PGM (IT-V4 method) and V3-V4 16S rRNA libraries with Illumina MiSeq (IM-V3/V4 method). Microbial communities were analyzed using QIIME and UPARSE. Linear correlation of the abundances of the shared genera between the IT-V4 and IM-V3/V4 datasets were computed using Pearson's correlation (r). Procrustes analysis was used to examine whether beta diversity estimations from the two methods were consistent (closeness of fit (M2) from 1000 Monte Carlo permutations < 0.3 and p < 0.05). Functional modules of the microbiota in the two datasets were predicted and analyzed using PICRUSt and LEfSe, respectively. IM-V3/V4 method yielded longer and 2.4-fold more post-processed reads than IT-V4 method. More IM-V3/V4 reads were assigned taxonomy than IT-V4 reads (family: 99.8% versus 96.1% and genus: 95.9% versus 92.2%). Overall, the IM-V3/V4 and IT-V4 methods showed a high degree of correlation in the average relative abundances of the shared genera (n = 36, r = 0.89, p < 0.0001) and functional modules (n = 265, r = 1.00, p < 0.0001). However, correlation was low for Gardnerella (r = 0.35) and Clostridium (r = 0.15), with their relative abundances being significantly higher with IT-V4 and IM-V3/V4 datasets, respectively. Furthermore, seven metabolic pathways were consistently differential between the two datasets (p < 0.05, LDA score > 2.0). Procrustes analyses showed a statistically consistent taxonomic and functional clustering between the two methods (Count_Better = 0, M2 = 0.2–0.3, p < 0.0001). Our study highlights that the cervical microbiota profiles from the IT-V4 and IM-V3/V4 16S rRNA gene metabarcoding methods are generally comparable.