Unique Features of a Japanese ‘Candidatus Liberibacter asiaticus’ Strain Revealed by Whole Genome Sequencing

Citrus greening (huanglongbing) is the most destructive disease of citrus worldwide. It is spread by citrus psyllids and is associated with phloem-limited bacteria of three species of α-Proteobacteria, namely, ‘Candidatus Liberibacter asiaticus’, ‘Ca. L. americanus’, and ‘Ca. L. africanus’. Recent findings suggested that some Japanese strains lack the bacteriophage-type DNA polymerase region (DNA pol), in contrast to the Floridian psy62 strain. The whole genome sequence of the pol-negative ‘Ca. L. asiaticus’ Japanese isolate Ishi-1 was determined by metagenomic analysis of DNA extracted from ‘Ca. L. asiaticus’-infected psyllids and leaf midribs. The 1.19-Mb genome has an average 36.32% GC content. Annotation revealed 13 operons encoding rRNA and 44 tRNA genes, but no typical bacterial pathogenesis-related genes were located within the genome, similar to the Floridian psy62 and Chinese gxpsy. In contrast to other ‘Ca. L. asiaticus’ strains, the genome of the Japanese Ishi-1 strain lacks a prophage-related region.


Introduction
Citrus greening (huanglongbing) is a devastating citrus disease that affects crops around the world. The disease was first noted in China in the early 20 th century [1]. Three species of phloemlimited, gram-negative bacteria in the genus 'Candidatus Liberibacter' are associated with greening. 'Ca. L. africanus' is mainly present in Africa [2]; 'Ca. L. americanus' is found in Brazil [3]. A third species, 'Ca. L. asiaticus' is particularly widespread in Asian countries as well as in Sao Paulo, Brazil and Florida, USA. 'Ca. L. asiaticus' is transmitted by phloem-feeding insect vectors, the Asian citrus psyllid Diaphorina citri [4] and the African citrus psyllid Trioza erytreae [5]. A new Liberibacter species, 'Ca. L. solanacearum', was recently associated with the emerging 'zebra chip' disease of potatoes in the U.S. and tomatoes in New Zealand [6].
Zhang et al. [27] reported two highly related, circular bacteriophage-type genes associated with 'Ca. L. asiaticus', named SC1 and SC2. Both were found integrated into the 'Ca. L. asiaticus' Floridian UF506 strain genome as prophages [27]. SC1 was apparently a fully functional, temperate phage with a lytic cycle that was seemingly activated when its host bacterium was present in plants but not when in psyllids [27]. SC2 replicates as an excision plasmid when its 'Ca. L. asiaticus' host is present in either plants or psyllids [27]. These findings suggest the bacteriophagetype genes are important for infection and virulence expression. However, most of the Japanese 'Ca. L. asiaticus' strains lack the bacteriophage-type DNA polymerase gene [18,19]. In Floridian UF506, the bacteriophage-type DNA polymerase gene is flanked by SC1 and SC2. Thus, absence of the bacteriophage-type DNA polymerase gene in Japanese strains suggests they also lack SC1 and SC2. Thus, the Japanese strains have unique genomic features.
In contrast to 'Ca. L. asiaticus' Floridian strains psy62 and UF506, the whole genome sequence of a Japanese 'Ca. L. asiaticus' strain lacking the bacteriophage-type DNA polymerase gene has not been reported. Recently, the complete genome sequence of the Chinese 'Ca. L. asiaticus' strains gxpsy [28] and A4 [29] were reported, although the latter remains in the draft form. Both Chinese 'Ca. L. asiaticus' strains also contained the bacteriophage-type DNA polymerase gene. The results encouraged us to perform whole-genome sequencing of a Japanese strain lacking this gene. Duan et al. [21] obtained a complete circular 'Ca. L. asiaticus' Floridian psy62 strain genome by metagenomic analysis of DNA extracted from a single 'Ca. L. asiaticus'-infected psyllid. We used a similar method to obtain the complete genome of the uncultured 'Ca. L. asiaticus' Japanese strain Ishi-1.

Bacterial strains
Japanese 'Ca. L. asiaticus' strain Ishi-1 was used throughout the study. The strain was originally found in local citrus of unidentified cultivars on Ishigaki Island, Okinawa prefecture, Japan. The infected scion was sent to the NARO Institute of Fruit Tree Science (NIFTS) with permission from the plant quarantine office of Japan, and kept in the isolated greenhouse after grafting on rough lemon (Citrus jambhiri Lush) rootstocks. The strain Ishi-1 induced severe symptoms on rough lemon, yuzu (Citrus junos Tanaka, Figure 1) and other citrus cultivars.

Psyllid treatment
All experiments using live individuals of D. citri were performed in insect-proof growth chambers at 25uC with a 16L:8D photoperiod at the Kuchinotsu Citrus Research Station, NIFTS (Otsu 954, Kuchinotsu, Minamishimabara, Nagasaki 859-2501, Japan). Healthy fifth instars of psyllids were transferred to an HLB-affected rough lemon tree (Citrus jambhiri, approximately 40 cm in height) with a high titer of 'Ca. L. asiaticus' bacteria. After acquisition feeding for 20 days on the infected plant, nine emerged adults were reared individually for 20 days on healthy Citrus junos seedlings for incubating the HLB bacteria, and they were stored at 250uC.

DNA extraction and quantitative real-time PCR
Total DNA was purified from the entire body of single psyllids using the DNeasy Blood and Tissue Kit (Qiagen, Tokyo, Japan) and a plastic homogenizer pestle (As One, Tokyo, Japan) The Complete Ca. L. asiaticus Japanese Ishi-1 Genome PLOS ONE | www.plosone.org according to the manufacturer's instructions, and eluted in 150 mL.
Individual psyllid DNA samples were analyzed for 'Ca. L. asiaticus' populations by quantitative real-time PCR analysis as described by Inoue et al. [30]. Samples containing copies of 'Ca. L. asiaticus' genomic DNA were selected by real-time PCR (data not shown). Whole-genome amplification was performed with Illustra GenomiPhi V2 (GE Healthcare, Buckinghamshire, England) according to the manufacturer's instructions. DNA concentration was estimated with the Qubit 2.0 instrument and the Qubit dsDNA HS Assay (Life Technologies, Invitrogen, California).

Genome sequencing and mapping
Sequencing was performed at the Bio Dragon Genomics Center (Takara Bio Co. Ltd. Mie, Japan). DNA libraries with 300,350 bp inserts were constructed according to manufacturer's instructions (Illumina GaIIx platform) and 75-bp paired-end reads were generated on an Illumina HiSeqTM 2000 platform. Reads were mapped to the 'Ca. L. asiaticus' Floridian psy62 genome using BWA [31] and Bowtie [32]. Mapping results were visualized with Integrative Genomics Viewer (IGV) version 2.3 [33].

Polymerase chain reaction for whole genome mapping confirmation
After initial genome mapping results were obtained, ambiguous sequences were determined by PCR amplification and conventional sequencing on an ABI 31306l instrument. Total DNA was extracted from the leaf midrib tissue of citrus trees infected with the 'Ca. L. asiaticus' Japanese Ishi-1 strain. Total DNA was extracted with the DNeasy plant minikit (Qiagen, Valencia, CA) according to manufacturer's instructions with minor modifications: approximately 0.2 g of the leaf midrib was placed in 400 mL AP1 buffer in a mortar and ground with a pestle until the leaf midrib became a fine green liquid.
Many In/Dels and SNPs were found by mapping the sequence reads of Ishi-1 to the complete sequence of the pathogenic 'Ca. L. asiaticus' Floridian psy62 (1.23 Mb) strain, and primers were designed from the surrounding sequences (Primer3, http://frodo. wi.mit.edu/primer3/) (Table S1). Other primers were selected from Duan et al. [21]. PCR was performed with the Gene Amp PCR System 9700 (Applied Biosystems, Foster City, CA) in 20-ml reactions containing 1 ml DNA template, 0.1 mM each primer, 200 mM dNTPs, 16 PCR buffer, and 2.5 units of Ex Taq DNA polymerase Hot Start Version (TaKaRa, Shiga, Japan) under the following cycling conditions: initial denaturation at 92uC for  Figure 3 A. Green arrows indicate CDS. B, Deduced amino acid sequences of the hypothetical protein at CLIBASIA_03230 of psy62, WSI_02190 of gxpsy, and CGUJ_03230 of Ishi-1 aligned by CLUSTAL W [48] and identical residues are indicated with asterisks. Databank accession numbers are CP001677 for psy62 [21], AP014595 for Ishi-1, and CP004005 for gxpsy [28]. doi:10.1371/journal.pone.0106109.g002 The Complete Ca. L. asiaticus Japanese Ishi-1 Genome 2 min; 35 cycles of denaturation at 92uC for 30 s, annealing at 54uC for 30 s, and extension for 1 min/kb of the desired product at 72uC. Long-range PCR of products above 3.0 kbp was performed with Tks Gflex DNA polymerase (TaKaRa). Each 50-ml reaction contained 1 ml DNA template, 0.1 mM each primer, 26 Gflex PCR buffer (Mg 2+ , dNTP plus), and 1.25 U Tks Gflex DNA polymerase. Cycling conditions were as follows: 30 cycles of denaturation at 98uC for 10 s, annealing at 52uC for 15 s, and extension for 30 s/kb of the desired product at 68uC. DNA sequences were aligned using GENETYX-windows ver. 11 (Software Development, Tokyo, Japan), and homology analysis was performed as recommended by the DNA Data Bank of Japan (http://www.ddbj.nig.ac.jp/Welcome-j.html).

Results
Whole genome re-sequencing Sequencing yielded 2,721,927,150 bp of DNA from 36,292,362 pair-end reads of 75 bp. As a result of mapping using BWA, Bowtie against the psy62 strain reference, reads of 14.6% of the 2.7 Gbp were mapped. Coverage to the reference was 96.9%. The sequence reads of Ishi-1 were not mapped near the nucleotide position of 0 to 7,803 and after nucleotide position 1,195,171 of the linear genomic map of psy62, indicating that Ishi-1 lacks a large genomic fragment. PCR amplification of the Ishi-1 template with the LJ754r and LJ764f primers, which are separated by about 35 kbp on the psy62 sequence, yielded a 2.6-kbp product. Sequence analysis showed that the 2.6 kbp fragment filled the 35-kbp gap in the Ishi-1genome. The results clearly demonstrated that Ishi-1 lacks about 33 kbp, corresponding to both ends of the linear genomic map of psy62 ( Figure 2). Likewise, by mapping the candidate SNPs/InDels and PCR verification, the draft genome of Ishi-1 was obtained.
In order to confirm the whole genome sequence of Ishi-1 strain, six primers published by Duan [21] were selected, and 128 new primers were designed for conventional and long PCR (Table S1). All amplicons generated from these primers were directly sequenced on an ABI 31306l. In total, we re-sequenced over 40,000 bp by Sanger sequencing. These efforts generated a circular chromosome sequence consisting of 1,190,853 bp ( Figure 3).
General features of the 'Ca. L. asiaticus' Japanese Ishi-1 genome The calculated GC content of the Ishi-1 genome is 36.32%, similar to other 'Ca. L. asiaticus' strains (Table 1). After annotation, the newly confirmed CDS regions were compared to those of other 'Ca. L. asiaticus' strains. Then, the tRNAs and rRNA of Japanese Ishi-1 were compared to those of Floridian  The Complete Ca. L. asiaticus Japanese Ishi-1 Genome PLOS ONE | www.plosone.org psy62. These analyses revealed 1,075 coding sequences and 975 ribosome binding sites. We also found 44 tRNA genes that were shared with the Floridian psy62 and Chinese gxpsy strains, as well as 13 rRNA operons and 313 hypothetical proteins (Table 1). Our comparison of 'Ca. L. asiaticus' Japanese Ishi-1 and Floridian psy62 revealed 291 base substitutions (Table 2). We also confirmed 122 in/del loci ( Table 2). The five SSR loci were also polymorphic between the two strains ( Table 2).
Bacteriophage-type polymerase and other genes As described above, the biggest difference in Japanese Ishi-1 is the absence of the 33-kbp fragment (Figure 3 A). In psy62, this fragment encodes 40 CDS, including the bacteriophage-type polymerase gene between the SC1 and SC2 genes in the prophage region ( Table 3). Most of the 40 CDS are shared between psy62, UF506, and Gxpsy ( Table 3). None of these 40 CDS, including the bacteriophage-type polymerase gene, were found elsewhere in the genome of Ishi-1. Thus, Ishi-1 lacks the bacteriophage-type DNA polymerase gene found in Floridian strains psy62 and UF506, and the Chinese gxpsy strain ( [21], [27], [28], shown by a vertical red line in Figure 3A). Another bacteriophage-type DNA polymerase is encoded in the middle of the linear schematic representation of the psy62 genome (shown by a vertical yellow line in Figure 3A). This bacteriophage-type DNA polymerase is also encoded in the corresponding region of the Ishi-1 genome ( Figure 3A). Thus, it became clear that Ishi-1 carries a single bacteriophage-type DNA polymerase gene, whereas psy62 has two. In contrast, Chinese gxpsy and Floridian UF506 carry three bacteriophage-type DNA polymerase genes (WSI_05345, 05570, 05770, UF506_015, SC1_gp210, SC2_gp210).
Characteristics of 'Ca. L. asiaticus' Japanese Ishi-1 strain marked by large In/Del variations Several large In/Dels are shown in the simplified schematic presentation of the genome (Figure 3 A). The 147-bp deletion at nucleotide positions 507106 through 507252 of the Floridian psy62 strain was detected in the genome of 'Ca. L. asiaticus' Japanese Ishi-1 (Figure 3 B). This deletion reduced the hypothetical protein sequence at CGUJ_03230 by 49 amino acids in comparison to CLIBASIA_03230 of psy62 and WSI_02190 of Chinese gxpsy (Figure 3 B). In contrast, the 2,108 bp insertion between nucleotide positions 983990 and 983991 (Figure 3 C), an untranslated region in psy62, was detected in Ishi-1. This insertion carries a prophage anti-repressor at CGUJ_04441, and two hypothetical proteins at CGUJ_04442 and CGUJ_04443 were newly confirmed. The deduced amino acid sequence of the prophage anti-repressor at CGUJ_04441 was identical to that of Chinese gxpsy (WSI_04270), but different from those of Floridian psy62 and UF506. The hypothetical protein at CGUJ_04442 was also identical to that of Chinese gxpsy (WSI_04275). The hypothetical protein at CGUJ_04443 locus shared 99% amino Data are based on the genome sequence of 'Ca. L. asiaticus' Floridian psy62 strain. The accession number is CP001677 [21]. b Data are based on the genome sequence of 'Ca. L. asiaticus' Floridian UF506 strain. The accession number is HQ377374 [27]. c Data are based on the genome sequence of 'Ca. L. asiaticus' Chinese gxpsy strain. The accession number is CP004005 [28]. acid sequence identity with the putative WSI_04280 in Chinese gxpsy. These two hypothetical proteins shared no identity with any of the hypothetical proteins from Floridian psy62 and UF506. Another insertion around nucleotide position 1081791 (psy62) was detected in Ishi-1 (Figure 3 D), encoding a hypothetical protein at CGUJ_04911 within the 1,660-bp span. The deduced amino acid sequence of the hypothetical protein shares 48% identity with the hypothetical protein CKC_03455 from 'Ca. L. solanacearum' CLso-ZC1, a pathogen of zebra chip. Ishi-1 also carries a 149 bp-long insertion that correspond to the nucleotide positions 1182471 and 1182472 of psy62 ( Figure 3A). No reading frames were found in the insertion.
Because of a single base insertion, Ishi-1 has two copies of the malic enzyme gene at CGUJ_00080 and CGUJ_00081, while 'Ca. L. asiaticus' Floridian psy62 (CLIBASIA_00080) and Chinese gxpsy (WSI_00005) each carry a single copy.
The genome of Ishi-1 also encodes a non-heme ferritin-like protein (CGUJ_03035), just like psy62 [38,39]. This ferritin-like protein is also found in 'Ca. L. solanacearum' [37], but is absent from the genomes of all other Rhizobiaceae. The ferritin superfamily of proteins includes several diverse members that are typically involved in iron storage and detoxification [40,41,42]. Lin et al. hypothesized that this ferritin-like protein may play a critical role in the survival and/or virulence of 'Ca. L. solanacearum' and 'Ca. L. asiaticus' [37]. The genome of 'Ca. L. asiaticus' Chinese gxpsy also encodes a ferritin-like protein (WSI_2370). These sequences were identical but for a single amino acid substitution in Japanese Ishi-1 ( Figure S1). In contrast, Floridian UF506 does not encode ferritin, indicating that this protein is dispensable.
Lin et al. reported that the 'Ca. L. solanacearum' genome encodes three known proteins involved in DNA replication and repair, all of which are absent from 'Ca. L. asiaticus' Floridian  [44,45,46,47]. The argB that was not encoded by 'Ca. L. asiaticus' Floridian psy62 but encoded by Japanese Ishi-1 and Chinese gxpsy is indicated in red letters. doi:10.1371/journal.pone.0106109.g004 psy62: LexA, DnaE, and RadC [37]. The genome Japanese Ishi-1 does not encode LexA, but does encode DnaE at CGUJ_03631 and RadC at CGUJ_03976. Chinese gxpsy also encodes DnaE and RadC, at WSI_03515 and WSI_03815, respectively (Table 4). However, a nucleotide sequence identical to that encoding RadC on the other 'Ca. L. asiaticus' strains is also carried in psy62 (Table 4), while LexA and DnaE are absent. This difference might be an annotation error. The hypothetical protein at CGUJ_03991 was newly confirmed because of a one-base substitution in the untranslated region of the psy62 genome. The hypothetical proteins shared no similarity to other proteins of 'Ca. L. asiaticus'. These proteins are listed in Table S2.

Discussion
As described previously, 'Ca. L. asiaticus' Japanese Ishi-1 lacks a bacteriophage-type DNA polymerase gene [19]. Our study showed that Ishi-1 lacks a large fragment of about 33 kbp that contains the bacteriophage-type DNA polymerase gene. It is noteworthy that this strain is found only in Japan; despite having the smallest genome of all 'Ca. L. asiaticus' strains, Floridian UF506 carries the large 33-kbp fragment [27]. We suggest the large 33-kbp fragment is associated with neither pathogenicity nor transmissibility, because Ishi-1 induced severe symptoms on citrus and propagated to a high titer in the vector insect. This is in sharp contrast to the discussions of Zhang et al. [27] regarding UF506, where the SC1 and SC2 genes flanking the bacteriophage-type DNA polymerase gene are suspected to be important for infection and virulence expression. It is likely that Ishi-1 carries different virulence factors. Most Japanese strains also lack the bacteriophage-type DNA polymerase gene [19]. Thus, the large 33-kbp fragment encoding the bacteriophage-type DNA polymerase gene may be absent from other Japanese strains, although confirmation by sequencing is needed. Another bacteriophage-type DNA polymerase is encoded in the middle of the Floridian psy62 and Japanese Ishi-1 genomes ( Figure 3A), while Chinese gxpsy and Floridian UF506 carry two additional polymerases. These differences also suggest Ishi-1 (and perhaps other Japanese strains) are distinct from other strains from the US and China.
Zhou et al. identified two related and hypervariable genes (hyv I and hyv II ) in the large 33-kbp fragment of the psy62 genome [25]. Although all DNA samples were obtained from symptomatic tissue and tested positive by 16S rRNA gene-based real-time PCR, neither the hyv I nor the hyv II gene was amplified from eight Indian citrus DNA samples and six Philippine psyllid DNA samples using the same primer sets [25]. These 14 strains likely lack the large 33kbp fragment as Japanese Ishi-1 does. Thus, 'Ca. L. asiaticus' lacking the large fragment are not limited to Japan but are widespread in South Asia, pending confirmation by genome sequencing. We conclude it is best not to use primer sets specific to the large 33 kbp fragment for PCR detection of 'Ca. L. asiaticus', because some strains might escape detection.
Two malic enzyme genes were identified in the genome sequence of Ishi-1 in contrast to 'Ca. L. asiaticus' Floridian psy62 and Chinese gxpsy. The malic enzyme of the microaerophilic protist Entamoeba histolytica decarboxylates malate to pyruvate [43]. In 'Ca. L. asiaticus', a phloem-limited bacterium [2], the enzyme might play a similar role to that of E. histolytica. However, both malic enzyme genes of Ishi-1 are shorter than those of other 'Ca. L. asiaticus' strains. The Ishi-1 enzymes might not maintain their original function, suggesting a possible limitation of the Ishi-1 fermentation pathway.
In conclusion, whole genome sequencing of Japanese 'Ca. L. asiaticus' strain Ishi-1 revealed unique genomic features and suggested novel expression of virulence and establishment in host plant as well as distinct molecular evolution. We hope this study will advance our understanding of 'Ca. L. asiaticus' and facilitate efforts to control this devastating disease in the citrus industry. Figure S1 Comparison of deduced amino acid sequences of ferroxidase from 'Ca. L. asiaticus'. Ferroxidase sequences were aligned by CLUSTAL W [48], and identical residues are indicated with asterisks. Databank accession numbers are CP001677 for psy62 [21], AP014595 for Ishi-1, and CP004005 for gxpsy [28], respectively. (TIF)

Supporting Information
Table S1 Conventional and long-distance PCR primers used for 'Ca. L. asiaticus' Japanese Ishi-1 strain sequence confirmation and gap closeure in this study. (XLSX)