A codon-optimized DNA fragment encoding the Imp of ‘Ca. P. aurantifolia’ NCHU2014 without the transmembrane domain was subcloned into the SUMO-pET vector and introduced into Escherichia coli BL21 (DE3). The N-terminal His-SUMO tagged ImpΔN protein was produced at 24°C by isopropyl β-D-1-thiogalactopyranoside induction and purified by Ni²⁺-NTA resin (Qiagen) according to the manufacturer’s instructions. The purified protein was cleaved with Ubiquitin-like-specific protease 1 and then reapplied to Ni²⁺-NTA resin for removing the cleaved His-SUMO tag and uncleaved His-SUMO-ImpΔN. The recombinant ImpΔN was obtained in the flowthrough and prepared for polyclonal antibody production in rabbits.

The antibody-based purification was developed to enrich ‘Ca. P. aurantifolia’ NCHU2014, which was maintained in periwinkle (Catharanthus roseus) by grafting. The stems and leaf veins of symptomatic periwinkle were sampled and gently homogenized with liquid nitrogen using mortar and pestle. Subsequently, 3 g of grinded tissues were suspended with 6 ml PBS buffer (0.137 M NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 6.8) and filtrated with a cell strainer (100 μm) to remove unfragmented tissues. Meanwhile, the anti-Imp antibody-coated beads were prepared by incubating antiserum (containing polyclonal anti-Imp antibodies) and Novex^® Dynabeads Protein A (superparamagnetic beads with recombinant Protein A covalently coupled to the surface). The flowthrough containing homogenized cells was incubated with antibody-coated beads (3 mg) for 30 min at 4°C, and then the unwanted substances were separated from the beads by magnet. The remaining materials containing phytoplasma cells were washed extensively with PBS buffer and collected for genomic DNA extraction.

After the affinity-purified cells enriched from 3 g of grinded tissues were lysed, about 6 μg genomic DNA was obtained from extraction by the Plant Genomic DNA Purification Kit (Gene Mark, Taiwan) according to the manufacturer’s instructions. After further cleaning by the Quick-DNA Plant/Seed Miniprep Kit (ZYMO RESEARCH), approximately 4 μg of DNA was obtained from one sample with 3 g of starting plant materials. A total of 2 μg genomic DNA was used for each library construction by the KAPA Hyper Prep Kit optimization for Nanopore 1D (Oxford Nanopore Technologies, Oxford, United Kingdom) according to the manufacturer’s instructions. To perform long-read sequencing, DNA libraries were loaded on MinION flow cells (R9.4), and the Nanopore reads were base-called from FAST5 files using the ONT Albacore Software (version 2.0.1).

The procedures for genome assembly and annotation were based on those described in our previous studies (Chung et al., 2013; Cho et al., 2020). All bioinformatics tools were used with the default settings unless stated otherwise. Briefly, the MinION reads were mapped to a previously published draft genome (Chang et al., 2015) using Minimap2 v2.15 (Li, 2018). Visual inspection of the mapping result using IGV v2.3.57 (Robinson et al., 2011) produced one circular scaffold representing the chromosome and one circular contig representing the plasmid. Next, an iterative process was used to complete and validate the assembly. In each iteration, the MinION reads were mapped as described, and the Illumina reads from our previous study (Chang et al., 2015) were mapped to the assembly using BWA v0.7.12 (Li and Durbin, 2009). The raw read mapping results were programmatically checked using SAMtools v1.2 (Li et al., 2009) and manually inspected using IGV v2.3.57 (Robinson et al., 2011). The iterative process was repeated until all polymorphic sites were resolved and a complete assembly was obtained. For the assembly process, only the MinION reads from runs A01 to A10 (Table 1) were used. The MinION reads from runs B01 to B03 were only used for checking coverage after the complete assembly was validated.

For annotation, RNAmmer v1.2 (Lagesen et al., 2007), tRNAscan-SE v1.3.1 (Lowe and Eddy, 1997), and Prodigal v2.6.3 (Hyatt et al., 2010) were used for gene prediction. The homologous genes in other representative phytoplasma genomes (Table 2) were identified using OrthoMCL (Li et al., 2003) to provide gene name and product description. Gene encoding putative secreted proteins were identified based on two established procedures, including one that uses SignalP v4.1 (Petersen et al., 2011) as described in Garcion et al. (2021) and another that uses SignalP v5.0 (Armenteros et al., 2019) as described in Cho et al. (2019). Additionally, BlastKOALA (Kanehisa et al., 2016) and GenBank (Benson et al., 2018) sequence similarity searches were used for manual curation of the annotation results.

For quality check, the raw reads that were not mapped to the phytoplasma genome were collected and mapped to the host (Catharanthus roseus) genome in two steps using Minimap2 v2.15 (Li, 2018). The first step mapped to the chloroplast genome (accession KC561139) (Ku et al., 2013b), and the second step mapped to the host nuclear genome (accession JQHZ01000000) (Kellner et al., 2015).

A codon-optimized DNA fragment encoding SAP11 of ‘Ca. P. aurantifolia’ NCHU2014 without the signal peptide was subcloned into a binary vector pBA002 for expression under the control of the CaMV 35S promoter. Plasmids for expression of the N-terminal FLAG-tagged Arabidopsis TCP transcription factors (SFP-AtT) were obtained as previously described (Chang et al., 2018). Nicotiana benthamiana grown at 26°C was used for transient co-expression assays. A mixture of A. tumefaciens strain ABI carrying the desired constructs of 35S:SAP11 and 35S:SFP-AtTCPs was introduced into N. benthamiana leaves by agroinfiltration. After 2 days, samples prepared from two infiltrated leaves (the third and fourth leaves counting from the top of 4- to 5-week-old plants) were collected for Western blotting analysis.

Collected samples were ground into powder using liquid nitrogen. About 0.1 g sample powder was added to 0.2 ml 2.5X SDS sample buffer (5 mM EDTA, 5% SDS, 0.3 M Tris-HCl, pH 6.8, 20% glycerol, 1% β-mercaptoethanol, and bromophenyl blue) and heated in a boiling water bath for 5 min. After centrifugation, supernatants were obtained as total cell extracts, and proteins were separated by SDS-PAGE. Antibodies against Imp (polyclonal), SAP11 (polyclonal), and FLAG^TM tag (monoclonal) were used to monitor protein amounts. Chemiluminescence signals generated by Amersham ECL reagents were captured using the ImageQuant LAS 4000 mini (GE Healthcare).

The ‘Ca. P. aurantifolia’ NCHU2014 was originally collected from purple coneflower (Echinacea purpurea) in Taiwan and transferred to periwinkle by dodder (Chang et al., 2015). The diseased plants infected with ‘Ca. P. aurantifolia’ NCHU2014 exhibited phyllody, virescence, and witches’ broom phenotypes in which the development of leaf-like structures with the loss of flower pigment and the proliferation of shoots in place of carpels were observed (Figures 1A–C).

As an obligate intracellular parasite, it is challenging to obtain a complete genome sequence of phytoplasma due to the host DNA contamination. To enrich ‘Ca. P. aurantifolia’ NCHU2014 for whole genome sequencing, immunoprecipitation-based purification of phytoplasma cells was conducted using the polyclonal antibody raised against the Imp of ‘Ca. P. aurantifolia’ NCHU2014. The antibody was generated in rabbit using the recombinant protein of Imp without N-terminal transmembrane domain (Figure 1D), which specifically recognized the 16SrII group ‘Ca. P. aurantifolia’ NCHU2014 in symptomatic periwinkle but not the 16SrVIII group ‘Ca. P. luffae’ NCHU2019 associated with loofah witches’ broom disease (Figure 1E and Supplementary Figure 1). For affinity purification of ‘Ca. P. aurantifolia’ NCHU2014, the stems and leaf veins from the symptomatic periwinkle displaying chlorosis, witches’ broom, phyllody, and virescence were used (Figure 1F). Genomic DNA extracted from the immunoprecipitated fraction containing ‘Ca. P. aurantifolia’ NCHU2014 was obtained for PCR examination. The PCR products specific for EpWB_v2c2400, EpWB_v2c2070, EpWB_v2c5130, and EpWB_v2c1780 genes of ‘Ca. P. aurantifolia’ NCHU2014 encoding SAP54/phyllogen, putative secreted protein, Mn/Zn-binding protein, and 50S ribosomal protein L17, respectively, were amplified only with the DNA sample extracted from symptomatic plants but not healthy plants (Figure 1G). The extracted DNA was then further processed for high-throughput DNA sequencing.

To improve the draft genome assembly of ‘Ca. P. aurantifolia’ NCHU2014 (Chang et al., 2015), we used ONT MinION to generate long sequencing reads. A total of 1,385,872 reads containing 6,540,398,289 bp were obtained from the libraries constructed using four DNA samples from our enrichment procedure (Table 1). These reads were obtained from 10 sequencing runs using two MinION flow cells. The first flow cell was used for two libraries (runs A01–A02 and A03–A05, respectively), producing 334,376 reads totaling 1,301,940,688 bp. The second flow cell was used for two additional libraries (runs A06–A07 and A08–A10, respectively), producing 1,051,496 reads totaling 5,238,457,601 bp. The average read lengths ranged from 3.7 to 5.2 kb with a maximum read length of 55.6 kb. Using these newly generated long reads, combined with our previous Illumina data set (Chang et al., 2015), we obtained a complete genome assembly for ‘Ca. P. aurantifolia’ NCHU2014. This assembly contains one circular chromosome with 635,584 bp and one circular plasmid with 4,224 bp (Figure 2). To the best of our knowledge, this is the first complete genome available for 16SrII phytoplasma.

Compared with the previous Illumina-only draft assembly of the same strain (Chang et al., 2015), 23 chromosomal regions ranging from 214 to 32,404 bp in size are newly assembled; the plasmid is unchanged. One 193-bp region (positions 354,755–354,947 of the chromosome) is shown as a gap based on the ONT reads but is supported by the Illumina reads. One possible explanation is that this region was deleted between our sample collections for Illumina (in 2014) and ONT sequencing (in 2019). Additionally, by examining the mapping results of Illumina reads to this complete assembly, we found that 15 chromosomal regions (mostly 36–128 bp in size, totaling 1,822 bp) are imperfect repeats and have no Illumina reads mapped.

As a quality evaluation, we also performed MinION sequencing to generate 708,299 reads (B01–B03; three runs totaling 2,831,035,013 bp using one MinION flowcell) using the libraries constructed with three DNA samples obtained without the immunoprecipitation-based enrichment. Compared with the enrichment protocol from which 22.5 to 36.0% of the raw reads were from phytoplasma, only 8.0 to 23.2% of the raw reads were from phytoplasma without the enrichment (Table 1). Importantly, the reads obtained with the enrichment procedure can cover all base pairs of the assembled phytoplasma genome even when only 14,718 reads were obtained in run A04 (Table 1). In contrast, up to approximately 37 kb of the phytoplasma genome were not covered by reads obtain from unenriched samples (B01). Based on these results, our enrichment procedure provides substantial improvements applicable to phytoplasma genome research.

A phylogenetic tree based on 16S ribosomal RNA (16S rRNA) genes of representative phytoplasmas was inferred to illustrate the evolutionary relationships (Figure 3). As the first complete genome sequence available for 16SrII group phytoplasma, ‘Ca. P. aurantifolia’ NCHU2014 shares 100% sequence identity of the 16S rRNA gene with ‘Ca. P. aurantifolia’ NTU2011. ‘Ca. P. aurantifolia’ NTU2011 was originally collected from infected peanut plants associated with peanut witches’ broom (PnWB) disease in Taiwan. So far, only a draft genome was available for ‘Ca. P. aurantifolia’ NTU2011, which contains 13 chromosomal contigs with a combined size of 562,473 bp and a plasmid with 4,221 bp (Chung et al., 2013). Comparative genomic analysis between two closely related strains revealed that the average nucleotide identity (ANI) is 99.72% across 90.79% chromosomal segments. This result reflects a high genomic similarity between these two strains in Taiwan. However, the draft genome of NTU2011 lacks two large segments corresponding to positions 232,370–246,973 and 313,259–348,010 of the NCHU2014 chromosome (Supplementary Figure 2). Further analysis by the pairwise genome alignment revealed that the chromosomes of these two strains are largely collinear although some rearrangements, including inversions and translocations, were observed (Figure 4A). The rearranged regions were confirmed experimentally by PCR analysis (Figure 4B).

The plasmids found in these two strains are nearly identical with 99.13% sequence identity. Only four protein-coding genes were found, including those corresponding to one replication protein, one DNA primase, one threonine synthase, and one hypothetical protein. Our BLASTN search against the NCBI Nucleotide Collection (nt) database did not find any similar sequence available for comparative analysis.

The ‘Ca. P. aurantifolia’ NCHU2014 genome consists of one circular chromosome with 635,584 bp and one plasmid with 4,224 bp. Based on the annotation, the chromosome encodes six rRNA genes, 24 tRNA genes, 471 protein-coding genes, and 35 pseudogenes (Table 2). Notably, ‘Ca. P. aurantifolia’ NCHU2014 has a low coding density (66.3%) with the lowest numbers of tRNA and protein-coding genes compared with other phytoplasmas with complete genome sequences available (Table 2). Among the protein-coding genes, only 352 (75%) genes were assigned to COG categories with specific functions (Figure 2). There were 119 (25%) genes annotated as hypothetical proteins without COG functional category assignments. Similar to other phytoplasmas with complete genome information, ‘Ca. P. aurantifolia’ NCHU2014 has a small chromosome with low GC content (24.5%) and lacks the genes for many metabolic pathways although it has 33 genes annotated as transporters. The comparative analysis further revealed that ‘Ca. P. aurantifolia’ NCHU2014 share 284–313 homologous gene clusters with other lineages surveyed (Figure 5A). Among them, a conserved core of 204 homologous gene clusters was found in all lineages surveyed (Figure 5B). This is consistent with other obligate parasitic bacteria that pose a high level of genomic plasticity (Andersen et al., 2013).

Phytoplasmas possess the Sec secretion system for transportation of effectors into the host cell cytoplasm (Sugio et al., 2011b; Oshima et al., 2013). Based on the prediction of N-terminal signal peptide by SignalP-5.0, 28 putative secreted proteins were identified (Supplementary Table 1). Surprisingly, when SignalP-4.1 was used, one of these 28 (EPWB_v2c3230) was excluded, and 70 additional putative secreted proteins were identified. Our manual inspection of these prediction results found that many of the putative secreted proteins identified by SignalP-4.1 are likely to be false positives (e.g., ribosomal proteins). In comparison, only one of the 28 candidates identified by SignalP-5.0 was an obvious false positive (EPWB_v2c2520; ATP-dependent Zn protease). Based on these findings, 27 putative secreted proteins predicted by SignalP-5.0 (excluding EPWB_v2c2520) were used for downstream analysis.

Among these 27 candidates, only one (EpWB_v2c2530) with uncharacterized function was found in PMU (Figure 6), which is different from the previous report that the many of the 56 secreted AYWB protein (SAP) genes in ‘Ca. P. asteris’ were associated with PMUs (Bai et al., 2009). This result may be explained by the fact that only one PMU was found in ‘Ca. P. aurantifolia’ NCHU2014, which contains PMU-associated genes (dnaG, dnaB, tmk, smc, hflB, himA, ssb, and rpoD) and is closely related to those of ‘Ca. P. luffae’ NCHU2019 and ‘Ca. P. ziziphi’ Jwb-nky in gene content and organization (Figure 6). In comparison, the chromosome of ‘Ca. P. asteris’ AY-WB harbors at least five PMU regions (Bai et al., 2006, 2009). Nevertheless, two thirds of putative secreted protein genes found in ‘Ca. P. aurantifolia’ NCHU2014 formed clusters and closely located on both sides of PMU (Figure 2). In addition, 12 out of the 13 SAP homologs in the NCHU2014 genome are located in these clusters. The only exception is the homolog of SAP11 (SAP11_EpWB; EPWB_v2c3970), which is located far from these clusters. Comparison of putative effector gene content revealed that ‘Ca. P. aurantifolia’ NCHU2014 harbors SAP11 (Bai et al., 2009) and SAP54 (MacLean et al., 2011) homologs while lacking homologs o SAP05 (Huang et al., 2021) or TENGU (Hoshi et al., 2009; Figure 7 and Supplementary Table 2).

Previously, it has been demonstrated that SAP11 has the ability to destabilize CYC/TB1 (CYCLOIDEA/TEOSINTE-BRANCHED1)-TCPs (TCP12 and TCP18), leading to the proliferation of axillary meristems (Chang et al., 2018; Wang N. et al., 2018; Pecher et al., 2019). The SAP11_EpWB identified in ‘Ca. P. aurantifolia’ NCHU2014 shares a high degree of amino acid sequence identity (98.8%) with SAP11_PnWB but only has 36.7% sequence identity with SAP11_AYWB (Figure 8A). To understand the ability of SAP11_EpWB in destabilizing TCP transcription factors, co-expression of SAP11_EpWB and FLAG-tagged A. thaliana TCPs were conducted in N. benthamiana using agroinfiltration. Similar to SAP11_PnWB, SAP11_EpWB exhibited a strong ability to destabilize class II CYC/TB1-TCPs (AtTCP12 and AtTCP18) but did not destabilize class II CIN-TCPs (AtTCP2, AtTCP13, and AtTCP24) as well as class I PCF-TCP (AtTCP20) (Figure 8B). As a control, the AtTCP12 and AtTCP18 were not decreased in abundance with the presence of the vector alone. Despite the fact that SAP11_EpWB display one amino acid difference with SAP11_PnWB in the C-terminal end, the potential biochemical activities in destabilizing CYC/TB1-TCP transcription factors were similar to each other. This is consistent with the fact that SAP11 lacking C-terminus (SAP11ΔC) still has the abilities for TCP binding and destabilization (Sugio et al., 2014). Because another phytoplasma effector, TENGU, with the ability to enhance the proliferation of axillary meristems, was not found in ‘Ca. P. aurantifolia’ NCHU2014, SAP11_EpWB would be the core virulence factor to induce witches’ broom symptom in diseased plants (Figures 1A–C).