Every base matters: assessing small subunit <scp>rRNA</scp> primers for marine microbiomes with mock communities, time series and global field samples

<jats:title>Summary</jats:title> <jats:p> Microbial community analysis via high‐throughput sequencing of amplified <jats:styled-content style="fixed-case">16S rRNA</jats:styled-content> genes is an essential microbiology tool. We found the popular primer pair <jats:styled-content style="fixed-case">515F</jats:styled-content> ( <jats:styled-content style="fixed-case">515F‐C</jats:styled-content> ) and <jats:styled-content style="fixed-case">806R</jats:styled-content> greatly underestimated (e.g. <jats:styled-content style="fixed-case">SAR</jats:styled-content> 11) or overestimated (e.g. Gammaproteobacteria) common marine taxa. We evaluated marine samples and mock communities (containing 11 or 27 marine 16S clones), showing alternative primers 515F‐ <jats:styled-content style="fixed-case">Y</jats:styled-content> (5′‐ <jats:styled-content style="fixed-case"> GTG <jats:bold>Y</jats:bold> CAGCMGCCGCGGTAA </jats:styled-content> ) and <jats:styled-content style="fixed-case">926R</jats:styled-content> (5′‐ <jats:styled-content style="fixed-case">CCGYCAATTYMTTTRAGTTT</jats:styled-content> ) yield more accurate estimates of mock community abundances, produce longer amplicons that can differentiate taxa unresolvable with <jats:styled-content style="fixed-case">515F‐C</jats:styled-content> / <jats:styled-content style="fixed-case">806R</jats:styled-content> , and amplify eukaryotic <jats:styled-content style="fixed-case">18S rRNA</jats:styled-content> . Mock communities amplified with <jats:styled-content style="fixed-case">515F‐Y</jats:styled-content> / <jats:styled-content style="fixed-case">926R</jats:styled-content> yielded closer observed community composition versus expected (r <jats:sup>2</jats:sup>  = 0.95) compared with <jats:styled-content style="fixed-case">515F‐Y</jats:styled-content> / <jats:styled-content style="fixed-case">806R</jats:styled-content> (r <jats:sup>2</jats:sup>  ∼ 0.5). Unexpectedly, biases with <jats:styled-content style="fixed-case">515F‐Y</jats:styled-content> / <jats:styled-content style="fixed-case">806R</jats:styled-content> against <jats:styled-content style="fixed-case">SAR11</jats:styled-content> in field samples (∼4–10‐fold) were stronger than in mock communities (∼2‐fold). Correcting a mismatch to <jats:styled-content style="fixed-case">T</jats:styled-content> haumarchaea in the <jats:styled-content style="fixed-case">515F‐C</jats:styled-content> increased their apparent abundance in field samples, but not as much as using <jats:styled-content style="fixed-case">926R</jats:styled-content> rather than <jats:styled-content style="fixed-case">806R</jats:styled-content> . With plankton samples rich in eukaryotic <jats:styled-content style="fixed-case">DNA</jats:styled-content> (> 1 μm size fraction), 18S sequences averaged ∼17% of all sequences. A single mismatch can strongly bias amplification, but even perfectly matched primers can exhibit preferential amplification. We show that beyond <jats:italic>in silico</jats:italic> predictions, testing with mock communities and field samples is important in primer selection. </jats:p>