Comparison of ITS1 and LSU for metabarcoding of Fungi

Metabarcoding is an important tool for understanding fungal communities. The internal transcribed spacer (ITS) rDNA is the accepted fungal barcode but has some known problems. The large subunit rDNA (LSU) has been used in tandem with ITS in order to investigate differences in the fungal community detected by these markers. Many studies have found that ITS and LSU detect similar community structure but differ in taxonomic composition. Numerous LSU primers have been developed for metabarcoding, but these were mostly designed to target Dikarya with little attention paid to early diverging fungi (EDF). We wanted to investigate the extent to which methodological biases impact the recovery of EDF from environmental samples using metabarcoding approaches. We focused on the Zoopagomycota as an example of EDF by creating a mock community comprised of taxa from this phylum. Using specialized culturing techniques, some Zoopagomycota species are abundant in environmental samples but are often in low abundance or absent from metabarcoding studies. We compared three different primer sets (ITS1F/ITS2, LROR/LR3, and a modified version of LR22 paired with LR3) used to amplify the mock community and a set of over 100 environmental samples. We collected soil, mud, water, and microinvertebrate environmental samples from five sites in California and two sites in Florida and sequenced them with the three primer sets using Illumina MiSeq. We performed bioinformatic analyses in AMPtk, including assembly and taxonomic identification of operational taxonomic units (OTUs) using two different methods. Fungal community composition was evaluated with non-metric multi-dimensional scaling plots, betadisper and Adonis analyses, Mantel tests, and sample richness differences between site and sample type. Our results show that all primer sets had challenges with amplifying and identifying EDF sequences, but LSU recovered some EDF taxa that were missing in the ITS1 data. In particular, target fragment length plays an important role both in PCR amplification from mixed samples as well as post-sequencing processing of reads. Target fragment length differences likely exacerbate PCR amplification bias and affect the downstream taxonomic assignment of OTUs. Our results indicate that the shorter LR22F/LR3 target fragment is a better general marker for some EDF than ITS1 and LROR/LR3 due in large part to these length biases.