P. knowlesi Genome Sequencing, Analysis, and de Novo Assembly.

After passage through a Plasmodipur filter (EuroProxima) to remove white blood cells, DNA was purified from schizont-stage parasites using the phenol/chloroform extraction method. DNA quality was highest for the A1-H.1 and A1-C.2 lines, so these were selected for the generation of large insert libraries. HiSeq 2000 (Illumina) compatible genomic shotgun libraries (500-bp fragment size) were prepared according to the manufacturer’s protocols (Illumina). Additional PCR-free libraries were prepared and sequenced on MiSeq (Illumina). Poor-quality bases from the raw reads were trimmed using Trimmomatic (www.usadellab.org/cms/?page=trimmomatic). For genome assemblies, 20-kbp insert template libraries were prepared for single-molecule real-time (SMRT) sequencing chemistry at Macrogen, Inc. Samples were sequenced on a PacBio RS-II instrument (Pacific Biosciences). Genome assembly was performed using PacBios’ FALCON tools (https://github.com/PacificBiosciences/FALCON), and consensus scaffolding was achieved through the Genome Consensus suite (https://github.com/PacificBiosciences/GenomicConsensus). PacBio sequencing generated 361,197 and 355,195 reads from three SMRT cells each for PKNA1-C.2 and PKNA1-H.1 genomic DNA, respectively. Reads with a mean length just above 8 kb were generated to obtain a theoretical coverage of 122.6-fold (for PKNA1-C.2) and 124.1-fold (for PKNA1-H.1) of the genomes. Genome assembly resulted in 45 and 37 contigs for PKNA1-C.2 and PKNA1-H.1, respectively, and after PCR-free Illumina read correction the total assembly sizes were 24,393,488 bp and 23,988,125 bp, respectively. Consensus scaffolds were error-corrected using the PCR-free Illumina sequence, and annotation was transferred from the reference genome of the P. knowlesi H strain (24) using the Rapid Annotation Transfer Tool, RATT (41). Gene models were checked in Artemis BAM view where flawed models (containing in-frame stop codons or lacking start or end codons) were corrected using strand-specific RNA-seq datasets from schizont stages of the human RBC-adapted parasite PKA1-H.1. RNA-seq analysis of schizont stages of A1-H.1 and ΔNBPXa parasites was performed to identify differential gene expression (three biological replicates for each, P value of 0.05 corrected for false discovery rate). Significance was determined using cuffdiff and CummeRbund (bioconductor.org/packages/release/bioc/html/cummeRbund.html).

The PacBio sequencing, assembly, and annotation information for both genomes is summarized in Table S1. Genomic deletions were determined by cross-mapping the reads against the reciprocal assembly and against the reference H strain (24) using the Burrows–Wheeler Alignment tool. GATK pipeline (Broad Institute) was used to perform variant calling with the corrected assemblies as references. GATK-based filtering was applied with the following parameters: quality-to-depth ratio of 5 and Fisher strand value of 60 for SNPs and 200 for indels. Variants having an allele frequency of 80% were retained for further analysis. Additional variant filtering and comparisons were performed using VCFtools (https://github.com/vcftools/vcftools). Natural variation (SNP frequency) in the five invasion-related genes NBPXa (PKNA1_H1_1472300), NBPXb (PKNA1_H1_0700200), DBLα (PKNA1_H1_0623500), DBLβ (PKNA1_H1_1400800), and DBLγ (PKNA1_H1_1356900) was determined using the Illumina raw sequence datasets from 48 P. knowlesi infections (25) compared with the reference PKA1-H.1 sequences using the GATK pipeline, and the SNPs were visualized using ggplot2 (ggplot2.org). A collection of three tools was used for computational analysis of copy number variation in addition to manual examination of the Illumina reads mapped onto the genome assembly of P. knowlesi H strain: Control-FREEC (42) was used with a permissive breakpoint threshold of 0.35 and a coefficient of variation of 0.07; CNVnator (43) was used with a window size of 150; and GROM-RD (44) was run with a P value of 0.001, a duplication coverage threshold of 3, and a sampling rate of 15.