Genetic variability in microbial eukaryotes reshapes marine biodiversity assessment in the age of amplicon sequencing

PLOS One | https://doi.org/10.1371/journal.pone.0326053 June 20, 2025 22 / 29 The retained skeletons post-DNA-extraction were gently rinsed in deionized water before being placed onto individual glass coverslips. Once moisture had fully evaporated, coverslips were mounted onto glass microscope slides with Canada Balsam in xylene (Refractive Index 1.52). Specimens were observed under transmitted light (Olympus BX51 compound microscope at 100x, 200x, and 400x) so that skeletal remains could be classified to the most precise taxonomic level possible. All taxonomic classifications were performed by ST based on formal descriptions and illustrations in several published literature sources, including: [11,24,38,49,55,57,60,62,65,67]. The curated online taxonomic resource Mikrotax (www.mikrotax.org [69]) was also consulted to confirm species identifications. If specimens lacked a skeleton (i.e., naked Collodaria or juveniles) or if a skeleton could not be identified due to breakage, taxonomic classification was made to the lowest level possible based on initial shipboard photos. The Collodaria specimens in this study were morphologically iden- tified as Acrosphaera murrayanna (6 specimens), Collosphaera huxleyi (51), Collozoum sp. (6), Disolenia zanguebarica (2), Siphonosphera socialis (6), Procyttarium primordialis? (3), Rhaphidozoum sp. (3), Sphaerozoum sp. (8), or Thalassi- colla nucleata (10). Nassellaria specimens were identified as acanthodesmid sp. (1 specimen), Acanthodesmia vinculata

(2), Eucyrtidium hexagonatum (2), Lophophaena hispida (1), Peromelissa thoracites (1), Pterocanium praetextum (3), Pterocorys campanula (1), Pterocorys sp. (4), Pterocorys zancleus (19), or Tetracorethra tetracorethra (1), and one puta- tive juvenile specimen. Spumellaria specimens were comprised of Dictyocoryne profunda (6 specimens), Dictyocoryne sp. (4), Dictyocoryne truncatum (3), Didymocyrtis sp. (6), Didymocyrtis tetrathalamus coronatus (3), Didymocyrtis tetrathala- mus tetrathalamus (7), Euchitonia elegans-furcata (5), Pyloniidae sp. (4), or lithellid sp. 3 (sensu Trubovitz et al. [55]) (1), and three putative juvenile specimens. Data analyses In total, 5,247,250 barcoded circular consensus reads were received from the University of California, Irvine Genomics Research and Technology Hub, with a mean of 22,617 reads per specimen. These reads were put through the DADA2 bioinformatic pipeline [70] to remove primers, filter by length and quality, denoise errors, call ASVs, and remove chime- ras. Sequences were initially checked for primer matches and oriented in the forward direction using cutadapt, with the “linked adapter” setting (-g) and the parameters --revcomp --discard-untrimmed -e 0.1, to account for reverse- complemented reads, remove reads lacking both forward and reverse primer sequences, and limit the allowable error rate to 10%. PacBio Circular Consensus Sequence data must be corrected for orientation because reads can be output in either the forward or the reverse direction. DADA2 removePrimers function was then applied to double-check read

orientation using the orient = TRUE parameter. The removePrimers function was also used to confirm that all reads contained primer matches (using the default maximum mismatch threshold of 2), remove any reads that did not contain primer matches, and then trim primer sequences from all passing reads. Next, the DADA2 filterAndTrim function was used to filter reads for quality and length. We used the parameters maxLen = 1600 and minLen = 1300, based on the expected length range of our target amplicons (1400–1500 bp). Following the filtering recommendations for long-read PacBio amplicon data [71], we required that reads contain zero Ns, have a minimum quality score of 3, and a maximum expected error rate of 2 (maxN = 0, minQ = 3, maxEE = 2). Within the DADA2 learnErrors function, we used the param- eters “errorEstimationFunction=PacBioErrfun” and “BAND_SIZE=32,” as advised for PacBio long-read data [71]. To best detect rare sequence variants that may be shared by multiple specimens, the parameter pool = TRUE was used in executing the dada function. Finally, the removeBimeraDenovo function in DADA2 was used to identify and remove chimeric reads. These steps were designed to separate spurious reads due to technical errors from authentic biolog- ical sequence variants [70]. Although we implemented the recommended parameters for quality filtering and calling ASVs with PacBio long-read circular consensus sequence data [71], there is no set of parameters that is guaranteed

to remove all PCR and sequencing errors while retaining all authentic biological sequence variants. Thus, it is possible that our stringent filtering methods removed some authentic sequence variants or failed to remove some erroneous reads.