The addition of Alconox (final concentration 0.75% w/v) to the gradient purification steps resulted in considerably less contamination (Fig. 1). Lower concentrations of Alconox yielded preparations with more contamination (data not shown). The addition of Alconox benefited the purification of oocysts from stool preserved in potassium dichromate, fixed in a zinc–polyvinyl alcohol (Zn–PVA) based fixative, or collected in a transport medium (Cary-Blair) in our study.Fig. 1

Microscopy images of oocysts after gradient purification without (a and b) and with (c and d) Alconox. a and c Light microscopy; b and d UV-fluorescence microscopy

Gradient purified oocysts were separated from remaining contaminants through flow cytometry sorting. Bi-parameter scatter and fluorescence dot plots of representative oocyst preparations are shown in Fig. 2. Scatter plots, oocyst size and internal complexity alone failed to separate oocysts from all debris (see gate region P2). Therefore, fluorescence dot plots were used to improve the separation (see gate region P1). Flow cytometry sorting was only practical for oocyst preparations that had reached a sufficient degree of purity in the gradient steps; preparations that contained too much stool residue would either clog the instrument or take an exorbitant long time to sort in the flow cytometer. For this reason, samples purified without Alconox were generally less successful in the flow cytometry process: in the example illustrated in Fig. 2, the proportion of oocyst events among all events was only 2.1% for oocysts purified without Alconox (panels a and b), but 40.3% for oocysts purified with Alconox (panels c and d).Fig. 2

Representative images of flow cytometry sorting of oocysts purified without (a and b) and with (c and d) Alconox in the gradient purification process. a and c Separation based on scatter; b and d separation based on fluorescence (FITC-A, filter 530/30 nm and PerCP-A, Cy 5.5 filter 695/40 nm). Oocysts (red events in gates P1 and P2) were sorted on logical “AND” function. Both sorted oocysts and sorted oocysts with conflicts were collected in single-cell mode

Genomic DNA was extracted from unsorted oocysts as well as sorted oocysts. DNA was extracted using mechanical disruption using freeze-and-thaw cycles. A majority of oocysts needed up to 25 cycles of freeze-and-thaw to disrupt the tough walls of oocysts and sporocysts (where present). Less than 10% of the oocysts showed disruption after 5 cycles of freeze-and-thaw. Purified oocysts before and after 15 cycles of freeze-and-thaw are shown in Fig. 3.Fig. 3

Composite microscopy image of purified oocysts before (a) and after (b) 15 cycles of freeze and thaw. Black arrows = intact oocysts; black arrowheads = oocysts, partially or completely empty; white arrows = sporocysts; white arrowhead = empty sporocyst. Scale bar = 10 μm

Genomic DNAs from unsorted and sorted oocysts were used to generate Illumina sequencing reads. Trimmed reads were then mapped to the CDC:HCNY16:01 draft genome assembly (Table 1) and analyzed for contaminants by metagenomic profiling (Fig. 5).Table 1

Proportion of trimmed reads mapped to the CDC:HCNY16:01 draft genome assembly using oocysts from three representative samples

Flow cytometry sorting had a major impact on the quality of the resulting Illumina sequencing data. Metagenomic profiling analysis revealed that the majority of reads from unsorted oocysts matched prokaryotic sequences, whereas relatively few prokaryotic sequences were identified in the flow-sorted oocysts (Fig. 5a and b, respectively). Furthermore, less than 25% of the reads from un-sorted oocysts could be mapped to the CDC:HCNY16:01 draft genome assembly, whereas more than 80% of the reads from sorted oocysts could be mapped to the same genome draft (Table 1).

DNA extracted from flow-sorted oocysts were used to obtain draft genome assemblies. Reads from Illumina libraries with different insert sizes were pooled together and assembled de novo. A comparison of these new genome assemblies is summarized in Table 2. The new assemblies had contig numbers between 669 and 2827. The total length of these contigs and the GC content were consistent with published draft genome assemblies of C. cayetanensis strains CDC:HCNY16:01 and CHN_HEN01 (GenBank accession numbers ASM130573v1 and ASM76915v2).Table 2

Summary statistics of de novo assemblies (including scaffolded regions) obtained from C. cayetanensis samples, including the two previously published assemblies (two first rows)

Twenty-one human stool specimens were selected for this study due to the presence of relatively high numbers of oocysts (large volume and/or high parasite load as estimated by UV-fluorescence microscopy). We included twelve specimens collected from seven separate U.S. outbreaks: Washington D.C. in 1996 (HCDC004_96); Rhode Island in 1997 (HCRI001_97); Nebraska in 2016 (HCNE181_16); and Texas in 2013 (HCTX119_13 and HCTX365_13), 2014 (HCTX535_14 and HCTX569_14), 2015 (HCTX542_15 and HCTX547_15) and 2016 (HCTX460_16, HCTX495_16 and HCTX503_16). Also included were 9 specimens collected in four different countries: Guatemala (HCGM002_97, HCGM011_97 and (HCGM012_97), Nepal (HCNP016_97), Indonesia (HCJK001_14, HCJK008_15, HCJK011_15 and HCJK015_15), and Mexico (HCMX010_16). Specimens collected prior to 2013 and specimens collected in countries outside the U.S. were preserved in 2.5% (w/v) aqueous potassium dichromate and stored at 4 °C following collection (n = 11). Specimen HCNE181_16 was collected in Cary-Blair transport medium. The remaining specimens (n = 9) were collected in Zn–PVA.

Gradient-purified C. cayetanensis oocysts were sorted by flow cytometry using a BD FACSAria III (BD Biosciences, San Jose, CA) equipped with blue (488 nm) and red (633 nm) lasers. Oocysts were diluted with sheath fluid appropriately for a sorting efficiency rate ≥ 45% and processed in the single-cell sorting mode using a 70 µm nozzle at 70 psi. Propidium iodide (PI) was added to the oocyst preparation at a final concentration of 1.0 µg/ml to label “dead” cells by binding to their DNA, thus increasing the shift away from oocysts because oocysts do not take up PI due to their intact oocyst wall [21]. Oocysts were identified by their size (approximately 8–10 µm) by forward scatter (FSC), their internal complexity by side scatter (SSC), and their autofluorescence emission properties when excited by 488 nm light using fluorescence filters appropriate for fluorescein isothiocyanate (FITC), tandem fluorochrome peridinin chlorophyll protein, and cyanine 5.5 (PerCP-Cy5.5).

Twenty-five cycles of freeze-and-thaw (freeze for 1 min by immersion in an ethanol/dry ice bath followed by thawing at 95 °C for 1 min in a heat block) was used to mechanically break open the purified oocysts. Genomic DNA was extracted from ruptured oocysts using DNeasy Blood & Tissue Kit (Qiagen, Germantown, MD) following the manufacturer’s instructions. The concentration of extracted genomic DNA was determined using a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific Inc., Cleveland, OH). The integrity of genomic DNA was analyzed by electrophoresis using Genomic DNA Analysis ScreenTape on a 2200 TapeStation (Agilent Technologies, Santa Clara, CA).

Genomic DNA (10 ng) was sheared in an M220 Focused-ultrasonicator™ (Covaris Inc., Woburn, MA) using settings for an average fragment of 300, 500, or 700 base pairs. Genomic libraries were constructed for Illumina sequencing using Ovation^® Ultralow Library Systems V2 (NuGEN Technologies Inc., San Carlos, CA) [22]. Libraries were barcoded to facilitate pooling during subsequent sequencing runs. Size distribution and concentration of genomic DNA libraries were analyzed by electrophoresis using Genomic DNA Analysis ScreenTape and D1000 ScreenTape on a 2200 TapeStation (Agilent). Barcoded genomic DNA libraries were pair-end sequenced using Illumina MiSeq Reagent v3 (600 cycles, 2 × 300 bp) kits (Illumina Inc., San Diego, CA).

Raw sequence data were assessed for quality using FASTQC v0.11.5. AdaptorRemoval v2.2.2 [23] was used to remove adaptor sequences from reads and to merge overlapping paired reads into consensus sequences. Genome assembly was performed using SPAdes v3.12.0 [24]. Contigs derived from prokaryotic human gut microbiota were removed with BBMap v35.82 [25]. CLC Genomics Workbench Map to Reference assembler (QIAGEN) was used to map reads to the draft genome assembly of strain CDC:HCNY16:01 (GenBank Accession No. ASM130573v1).

Metagenomic analysis was performed to estimate the degree of contamination in the sequences obtained at various steps of oocyst purification. Sequencing reads were profiled for taxonomic diversity by assessing the percentage of data arising from certain contaminating organisms using MetaPhlAn [26] and signature database version 2.0. This database does not contain signatures from Cyclospora species.