Author Correction: Genome sequencing of Prototheca zopfii genotypes 1 and 2 provides evidence of a severe reduction in organellar genomes

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

The sequences of Prototheca species currently available in public databases are those of the 18S rDNA (small subunit of rDNA, SSU) and 28S rDNA (large subunit of rDNA, LSU) 2,19 , and those of the Internal Transcribed Spacer regions (ITS), as well as some mitochondrial and plastid genomes. Notably, this information is not available for all species and, full-length sequences are very often missing 12,20,21 : complete sequences of the organellar DNA are currently only available for P. wickerhamii (both mitochondrion and plastid) 22,23 , P. cutis and P. stagnora (plastid only) 24 .
In this paper, we present the complete and manually annotated genomes of mitochondria and plastids of P. zopfii genotypes 1 and 2, and the first draft assembly of the whole nuclear genomes of both P. zopfii subspecies. Our work gives, for the first time, a representative overview of the extreme reduction which occurs within the mitochondrial and plastid genomes of these algae and provides basic information for further investigations.

Materials and Methods
Strains and culture conditions. P. zopfii genotype 1 (SAG 2063) and P. zopfii genotype 2 (SAG 2021) were obtained from the Culture Collection of Algae at Göttingen University ("Sammlung von Algenkulturen der Universität Göttingen", SAG, Göttingen, Germany). The strains were aerobically sub-cultured on Sabouraud Dextrose Agar plates for 48-72 h at 30 °C until DNA isolation was carried out.
Isolation of DNA and RNA. Genomic DNA and RNA were extracted starting from approximately one g of P. zopfii genotype 1 and genotype 2 resuspended in 20 ml wash solution (0.6 M Sucrose, 20 mM Tris, 20 mM MgCl 2 and 1 mM DTT, pH 7.5) and centrifuged at 2200 g for 2 min at 4 °C.
For DNA, washed pellets were resuspended in 500 μl of TRIS EDTA lysis buffer (10 mM Tris-HCl, 10 mM EDTA, 250 mM NaCl, pH 8), supplemented with 25 µl of proteinase K (20 mg/ml) (Sigma-Aldrich, St. Louis, MO, USA), 25 µl of SDS 10% and incubated at 56 °C for 2 h. Next, 25 µl of RNaseA (20 mg/ml) (Sigma-Aldrich) were added and the suspensions were incubated at 56 °C for 30 min. DNA was extracted using an equal volume of 1:1 (v/v) phenol:chloroform and precipitated with one volume of cold isopropanol. DNA was rinsed with 70% (v/v) cold ethanol, air dried, resuspended in 30 µl of ultrapure water and stored at −20 °C until use. DNA concentration and quality were estimated by PicoGreen (Thermo Fisher, Waltham, MA, USA) and by agarose gel electrophoresis.
Total RNA was extracted from washed pellets with TRIzol (Invitrogen, Carlsbad, CA, USA) and purified by NucleoSpin miRNA kit (Macherey-Nagel, Duren, Germany), following the manufacturer's protocol, in combination with TRIzol lysis. RNA concentration (ng/µl) and quality RNA Integrity Number (RIN) were determined by Agilent 2100 Bioanalyzer (Santa Clara, CA, USA). RNA extracts were stored at −80 °C until use.
DNA and RNA library preparation and sequencing. P. zopfii DNA libraries for Illumina sequencing (paired-end and mate-pair sequencing) (Illumina, San Diego, CA, USA) were prepared and sequenced on an Illumina MiSeq instrument, following the manufacturer's instructions as detailed in Supplementary Table 1. DNA library preparation for GS-FLX sequencing was performed using the GS FLX Titanium Rapid Library Preparation Kit (Roche, Basel, Switzerland) as follows: 1 μg DNA for each P. zopfii strain was used in the preparation of shotgun libraries by genomic DNA fragmentation by nebulization and ligation to specific adapters. According to the manufacturer's instructions, libraries were subjected to clonal amplification by emulsion PCR reaction, recovered by isopropanol breaking and enriched for positive reaction beads. Each library was separately loaded onto one region of the GS-FLX PicoTiter Plate and sequenced according to the 454 GS-FLX Titanium XL protocol.
One μg of RNA was used for libraries construction using TruSeq ® RNA Sample Preparation v2 Kit (Illumina), according to the manufacturer's instructions using poly(A) enrichment and sequenced on a 2 × 101-cycles Hiseq 2000 run (Illumina).
De novo genome assembly. The whole genomic sequence collections -composed of Illumina paired-end and mate-pair sequences, plus GS-FLX reads -from P. zopfii genotype 1 and 2 were assembled with SPAdes 25 in read error correction and assembling mode. Six K values, ranging from K21 to K127, were automatically selected by the algorithm based on the read length and dataset type. CAP3 26 (-p 96 -o 500) was run on the resulting scaffolds to further assemble contiguous regions. Illumina reads were mapped back to scaffolds by means of bwa mem (v. 0.7.10 27 ), keeping only those having ≤2 mismatches with the reference and Pilon (v 1.22) 28 in assembly improvement mode was employed in order to correct substitutions and short indels.
Organelles assembly. Among the assembled scaffolds, sequences of mitochondrion and plastid were detected by homology with a set of 322 mitochondrial and 605 plastid/chloroplast sequences from related species (i.e.: green algae) and circularized, giving rise to full circular sequences for both organelles in both P. zopfii genotypes. Assembly of the mitochondrial and plastid genomes of P. zopfii was independently confirmed by a custom assembly strategy that, starting from "seeds", implements an iterative procedure aimed at finding reads partially overlapping with the seed and assembling them with the original contig. Then, overlaps among the new contigs were used to generate a "supercontig". Further alignment-assembly-extension steps were performed on each side of the supercontig until an overlap between the 5′ and 3′ ends of the sequence was found, meaning that the whole circular genome had been covered. A full description of this custom assembly procedure is available in Supplementary Methods. Nuclear genome annotation. Gene prediction in nuclear genomes was performed with Augustus 29 using Chlorella variabilis as reference species. A double annotation procedure was carried out on the predicted protein models, employing both BLAST 30 (-evalue 1e −10 ) versus the UniProtKB database (http://www.uniprot. org/uniprot/), and InterProScan 5 31 to provide functional analysis. BLAST comparison (-evalue 1e −10 ) was also SCIENTIfIC RepoRts | (2018) 8:14637 | DOI:10.1038/s41598-018-32992-0 performed between the annotated protein datasets of the two P. zopfii genotypes. Annotation of proteins identified as putative DNA-directed RNA polymerases (NEPs) was double-checked by BLASTp against the NCBI's own non-redundant ("nr") database. Presence of a chloroplast transit peptide (cTP) in genes annotated as NEPs was assessed by using several prediction tools available on the web: PProwler 32 , PredAlgo 33 , TargetP (v1.1) 34 , and predSL 35 . Organelles annotation. Organelles were annotated using the DOGMA webserver 36 . Annotation was manually refined, relying on similarities obtained by BLAST versus the organellar genes of the closest known relatives of P. zopfii (i.e.: C. variabilis, Auxenochlorella protothecoides, Helicosporidium sp. and P. wickerhamii). Transfer RNA (tRNA) and ribosomal RNA (rRNA) sequences were determined by BLASTn alignment, whereas coding DNA sequences (CDS) positions were refined on the basis of protein-protein (BLASTp) matches. Circular representations of the mitochondria and plastids were drawn using Gview 37 . Comparison of CDS among species was performed by BLASTp.
Phylogenetic analysis. The protein sequences inferred from nine conserved plastid ribosomal genes (RPL2, RPL5, RPL14, RPL16, RPL20, RPS8, RPS11, RPS12, RPS14) were retrieved from P. zopfii genotype 1 and 2, from related species (i.e.: C. variabilis, A. protothecoides, Helicosporidium sp., P. wickerhamii, P. cutis and P. stagnora), and from other Trebouxiophyceae species available at NCBI, for a total of 40 species. Only ribosomal proteins whose sequences were available for each considered species were taken into account. Sequences were concatenated and aligned to produce a super-alignment with Clustal-Omega 38 , which was manually inspected and used to infer a Maximum Likelihood phylogenetic tree with the program PhyML 39 using four substitution rate categories, the cpREV substitution model, estimated gamma shape parameter, 1000 bootstraps, and core Trebouxiophyceae as outgroup.
RNA-Seq data analysis. RNA-Seq reads from P. zopfii 1 and 2 were mapped to the respective draft assemblies with STAR aligner (v 2.5.3a) 40 . Alignments were filtered retaining, for each genotype, only reads aligning for ≥80% of their length and having ≤2 mismatches. The number of reads mapping within the predicted gene models was assessed with BEDtools (v 2.26.0) 41 , requiring a read to map for at least 90% of its length within a gene to be counted. Counts were then converted to reads-per-kilobase-per-million (RPKM) values.

Results
The genomes of P. zopfii genotype 1 and 2 were sequenced using a combination of different approaches (Supplementary Table 1 Mitochondrial structure and annotation. The mitochondrial genomes of both P. zopfii genotype 1 and 2 are sized at about 38-39 Kb (Fig. 1A, Supplementary Fig. 1A). They are extremely compact, with only about 32% of non-coding DNA and are characterized by a substantial loss of any intron-exon structure in their genes. Only P. zopfii genotype 2 shows a single intron (length: 777 bp) in the long ribosomal subunit (lrn) gene, whereas the other species belonging to the Chlorellales (i.e.: C. variabilis, A. protothecoides, Helicosporidium sp., and P. wickerhamii) display a more complex structure, with intron length reaching 4,000-8,000 bp. A putative LAGLIDADG homing endonuclease, a class of restriction enzymes directly involved in the DNA cutting process 42 , is encoded within the intron of lrn gene. The annotation revealed a number of coding genes for both P. zopfii species similar to those observed in other Chlorellales species. All the genes encoding for cytochrome units, NADH dehydrogenase, ATP synthases and ribosomal proteins, as well as all the tRNAs and the ribosomal units, have a conserved structure (Table 1, Fig. 2A). As expected, mitochondrial protein similarity between P. zopfii genotype 1 and 2 is above 90%, while similarity with Helicosporidium sp., C. variabilis, A. protothecoides and P. wickerhamii is around 60%.
Plastid structure and annotation. The structure of the plastid genomes of P. zopfii genotype 1 and 2 ( Fig. 1B, Supplementary Fig. 1B) is very similar to that of other non-photosynthetic algae belonging to the Trebouxiophyceae class, being extremely compact. The genomes are sized only about 28.7 Kb for both genotypes and are the shortest plastid genomes within their class. Both P. zopfii genotypes possess only 19 CDS, and as a result are simpler than those of Helicosporidium sp. (26 CDS), P. wickerhamii, P. cutis (both possessing 40 CDS) and P. stagnora (28 CDS). On the other hand, the tRNAs and rRNAs are conserved among all these species and other photosynthetic algae (i.e.: C. variabilis and A. protothecoides) (Table 1, Fig. 2B). Plastid protein similarity between genotype 1 and 2 is about 85% while similarity with other organisms is around 41-43%, with the best match being with P. stagnora (i.e.: 49.9% and 49.5% similarity to P. zopfii 1 and 2, respectively) (Fig. 3A).
Comparison with other related organisms shows that the plastid genomes of P. zopfii genotype 1 and 2, like those of Helicosporidium sp., P. wickerhamii, P. cutis and P. stagnora, lack the genes associated with photosynthesis (photosystem I and II, chlorophyll biosynthesis and cytochrome components), whereas ATP synthase genes were  Table 2, Supplementary Table 2). Our transcriptome analysis demonstrated that most plastid genes encoding rRNA and proteins were expressed in P. zopfii (Supplementary Fig. 2A), although three of them (i.e.: rps4, rps7, rpl20) had low RPKM values, suggesting that nuclear-encoded counterparts could compensate for the loss of plastid-encoded RNA polymerases.
Both P. zopfii genotypes plastids contain the genes of most of the ribosomal proteins, with the exception of RPL12, RPL19, RPL23, RPL32, RPS2, RPS18 and RPS9. The same proteins, with the exception of RPL19 and RPL32, are also absent in P. stagnora, which, apparently, is the most similar organism. The phylogenetic tree inferred from the super-alignment of 9 shared ribosomal proteins confirms that Helicosporidium sp., P. stagnora and P. zopfii are closely related genera, whereas P. wickerhamii is more distant, closer to A. protothecoides than to other Prototheca species (Fig. 3B).
Nuclear genome annotation. From the genome assembly procedure, 6,956 and 4,555 scaffolds with lengths >1 Kb representing the nuclear genome were obtained for P. zopfii genotype 1 and 2, respectively, indicating a genome size of about 26.5 Mbp and 24.7 Mbp for genotypes 1 and 2, respectively, a size between those of Helicosporidium sp. (12.4 Mb) and C. variabilis (46.2 Mb) and comparable to the genome of A. protothecoides (~23 Mb) and P. wickerhamii (~29 Mb) ( Table 1).
The maximum scaffold length was 97,625 and 57,068 bp, with 1,289 and 1,708 contigs exceeding 5 Kb in length and a N50 value of 6,686 and 7,940 for P. zopfii genotypes 1 and 2, respectively. Augustus gene prediction led to the individuation of 6,884 and 6,381 gene models for the two genotypes. 56.5% and 62.0% of P. zopfii genotype 1 gene models were annotated versus UniProtKB and InterPro databases, respectively, while the corresponding percentages for genotype 2 were 59.4% and 67.2%. The main features of the nuclear genome, such as gene number and coding density, were consistent for both genotypes, and similar to their counterparts in A. protothecoides genome (i.e.: gene density: 0.26 genes/Kb and 0.32 genes/Kb, for both P. zopfii genotypes and A. protothecoides, respectively). On the contrary, the average exon and intron sizes were higher compared to those previously observed in related species (Supplementary Table 3). BLAST comparison of the predicted proteins of the two P. zopfii genotypes resulted in a set of 6,134 common entities, representing a core set of homologous proteins conserved in the two genotypes. Among the predicted genes and transcripts, we found evidence of the presence and expression of nuclear-encoded polymerases (NEPs) ( Supplementary Data 1 and 2). P. zopfii genotype 1 and 2 possess 21 and 19 genes annotated as NEPs, respectively, and both showed a RNA-Seq signal indicating active transcription of many of them ( Supplementary Fig. 2B,C). Prediction of target peptides highlighted at least one NEP per genotype as a high-confidence candidate for containing a chloroplast transit peptide (cTP) (genes g5108 and g2780 for genotypes 1 and 2, respectively, for which all the four prediction software employed were concordant); moreover, PredAlgo also suggested two more genes (g3914 and g4216 for genotypes 1 and 2, respectively) to be plastid-directed NEPs. In addition to that, we found evidence of some mitochondrial targeting peptides (mTPs) in more gene models (Supplementary Table 4).

Discussion
In this paper, we describe the complete, manually annotated, circular sequence of both mitochondrial and plastid organellar DNA of P. zopfii genotype 1 and genotype 2, as well as a first draft of the complete genome, by whole genome shotgun sequencing. Structure of the mitochondrial genomes of both P. zopfii genotypes was revealed to be smaller in size and extremely condensed when compared to that of some related organisms, i.e.: P. wickerhamii 22 and Helicosporidium sp. 43 , but similarly functional, with the size reduction mostly due to the lack of intron-exon structures.
An extremely compact and simplified structure was also observed in P. zopfii plastid genomes, which showed a substantially reduced size (about 28.7 Kb), smaller than those of all other algae belonging to the class of Trebouxiophyceae. As previously observed in the genomes of non-photosynthetic algae belonging to this class, P. zopfii plastids lack all the genes for the synthesis of the proteins involved in the photosynthesis process 23,44 , and for the RuBisCO large subunit. Fundamental plastid-related functions, however, seem to have been preserved, as indicated by the presence of a RNA-Seq signal on the genes of plastid-encoded ribosomal proteins. The low expression of some of them, however, cannot preclude their presence in the genome as pseudogenes in P. zopfii. However, further experiments should be carried on in order to confirm this observation. More interestingly, the entire set of rpo genes (i.e. rpoA, rpoB, rpoC1 and rpoC2), which codify for the plastid-encoded RNA polymerases (PEPs), was lost in P. zopfii, an unprecedented observation within this class of algae. Loss of PEPs was previously reported for other non-photosynthetic parasitic plants, such as Cuscuta obtusiflora 45 and Rhizanthella gardneri 46 , but not in apicomplexan and algae: plastid genomes of Plasmodium falciparum 47 , A. protothecoides, P. wickerhamii 23 , P. cutis, P. stagnora 24 and Helicosporidium sp. 44 have all retained the complete set of rpo genes. We found no evidence of PEP sequences either in plastid assemblies or in the nuclear draft genomes, whereas nuclear genome contigs contained evidence of 21 and 19 DNA-driven, nuclear-encoded polymerases (NEPs) for P. zopfii genotype 1 and 2, respectively, and at least one of them, per genotype, was predicted to contain plastid-targeting signal peptides. It is therefore possible that P. zopfii codes for other NEPs able to target the plastid, making it possible to transcribe its genetic information.
As previously suggested 23,24,48,49 , considering the degree of similarity between the few structural genes preserved in its plastid genome and the evidence from the phylogenetic analysis, P. zopfii seems to be more closely related to P. stagnora and to Helicosporidium sp., rather than to P. wickerhamii or P. cutis. Moreover, it is noteworthy that P. wickerhamii appears not to be closely related to P. zopfii, but instead to A. protothecoides, strengthening the evidence that P. wickerhamii is only loosely related to other Prototheca spp, as previously revealed by plastid genome comparison 23 and supporting the proposal of either moving P. wickerhamii into Auxenochlorella genus or creating a new genus 48 .
Nuclear genome assemblies of P. zopfii genotype 1 and 2 had a size estimated at about 25-26 Mb for both, consistent with that reported in a previous work 50 . Although further studies are certainly needed to elucidate the structure of nuclear genomes of P. zopfii, this work adds information to the growing body of genome resources for the plant kingdom, being, although a preliminary draft, the first report of the assembly of nuclear DNA of P. zopfii.
In conclusion, we believe that the information reported herein will be important for the understanding of the evolution and genomic organization of Prototheca spp., with a particular focus on the progressive loss of functions of plastids in the shift from autotrophic, photosynthetic, to obligate, heterotrophic, parasitic algae.

Data Availability
The complete genome sequencing project has been registered in the NCBI BioProject portal (https://www.ncbi. nlm.nih.gov/bioproject/) under the accession number PRJNA388740. Raw DNA sequence reads for P. zopfii genotypes 1 and 2 have been deposited into the NCBI Short-Read Archive (SRA, https://www.ncbi.nlm.nih.gov/sra/) under the accession numbers SRR6319956-SRR6319964. RNA-seq reads are saved under the accession numbers SRR7091517-SRR7091518. Full sequences of mitochondria and plastids are available in GenBank, under accessions MF197533, MF197534, MF197535, and MF197536. The Whole Genome Shotgun project has been deposited (as non-annotated contigs) at DDBJ/ENA/GenBank under the accessions PEIA00000000 and PGFX00000000. The versions described in this paper are PEIA01000000 and PGFX01000000. Sequences annotated as nuclear-encoded polymerases (NEPs) are available as amino acid FASTA files in Supplementary Data 1 and 2.