Establishment of a Cuscuta campestris‐mediated enrichment system for genomic and transcriptomic analyses of ‘Candidatus Liberibacter asiaticus’

Summary ‘Candidatus Liberibacter asiaticus’ (CLas) is a phloem‐limited non‐culturable α‐proteobacterium associated with citrus Huanglongbing, a highly destructive disease threatening global citrus industry. Research on CLas is challenging due to the current inability to culture CLas in vitro and the low CLas titre in citrus plant. Here, we develop a CLas enrichment system using the holoparasitic dodder plant (Cuscuta campestris) as an amenable host to acquire and enrich CLas from CLas‐infected citrus shoots maintained hydroponically. Forty‐eight out of fifty‐five (87%) dodder plants successfully parasitized CLas‐infected citrus shoots with detectable CLas by PCR. Among 48 dodders cultures, 30 showed two‐ to 419‐fold CLas titre increase as compared to the corresponding citrus hosts. The CLas population rapidly increased and reached the highest level in dodder tendrils at 15 days after parasitizing citrus shoot. Genome sequencing and assembly derived from CLas‐enriched dodder DNA samples generated a higher resolution than those obtained for CLas from citrus hosts. No genomic variation was detected in CLas after transmission from citrus to dodder during short‐term parasitism. Dual RNA‐Seq experiments showed similar CLas gene expression profiles in dodder and citrus samples, yet dodder samples generated a higher resolution of CLas transcriptome data. The ability of dodder to support CLas multiplication to high levels, as well as its advantage in CLas genomic and transcriptomic analyses, make it an optimal model for further studies on CLas–host interaction.


Introduction
'Candidatus Liberibacter asiaticus' (CLas) is a fastidious a-proteobacterium associated with citrus Huanglongbing (HLB, also known as yellow shoot disease), a destructive disease threatening citrus production worldwide (Jagoueix et al., 1994;Bov e, 2006). Due to the current inability to culture in vitro, research on CLas has been mainly relied on analyses of CLas-infected host materials. For instance, CLas-infected citrus leaves midribs and their total DNA extracts were used for genome sequencing analysis of CLas (Garnier and Bov e, 1983;Duan et al., 2009;Hartung et al., 2010;Zhang et al., 2011). Another challenge of CLas research is its uneven distribution and low titre in diseased citrus plants (Tatineni et al., 2008;Li et al., 2009). For these reasons, efforts have been made to obtain host sample with high CLas concentration by the assistance of others amenable plant hosts, such as periwinkle (Catharanthus roseus) (Zheng et al., 2014a).
Dodder (Cuscuta spp.) is a parasitic plant that can form the haustoria to uptake nutrients from the host plant through its haustoria (Kim and Westwood, 2015). The dodder plant was able to transmit the phloem-limited fastidious microbes, e.g. virus, phytoplasma and bacteria, between source and receptor plants (Bennett, 1940;Kunkel, 1952;Ke et al., 1986;Tang and Fan, 1987;Zhou et al., 2007). Early studies had demonstrated that CLas can be transmitted by dodder from citrus to others citrus plants or new hosts, like periwinkle (Catharanthus roseus) and Murraya paniculata (Garnier and Bov e, 1983;Ke et al., 1986;Tang and Fan, 1987;Zhou et al., 2007). In addition to be a microbial transmission tool, the dodder was found to support the multiplication of CLas to high levels (Hartung et al., 2010).
To obtain high quality CLas genome sequence, samples with high CLas titre are required. The insect vector (Asian citrus psyllid, Diaphorina citri) samples with high CLas titre were initially selected for CLas genome sequencing (Duan et al., 2009). However, it is not common to obtain the psyllid with high titre of CLas in the field (Ukuda-Hosokawa et al., 2015), even under artificial feeding conditions in laboratory . Furthermore, high genome sequence variation was observed in CLas strains during psyllid transmission (Katoh et al., 2015). For plant hosts source, the low titre of CLas in samples affected the CLas genome sequencing. Therefore, a process of bacterial DNA enrichment was applied in CLas-infected periwinkle and citrus plants samples to increase the ratio of CLas DNA to total DNA and further improve the CLas genome quality (Zheng et al., 2014a(Zheng et al., , 2014b(Zheng et al., , 2015(Zheng et al., , 2018. In addition, CLas could grow to high titre (Hartung et al., 2010) in dodder, but the use of dodder enrichment for CLas genomic research has not yet been explored.
Another interesting topic is the interaction between CLas and its host. This can be primarily achieved by profiling in planta genome-wide transcriptome of CLas. However, few studies have been reported CLas transcriptome profiling by using RNA-Seq, mainly due to low abundance of CLas RNA compared with host RNA. A recent study analysed in citrus and psyllids CLas expression profiles using RNA-Seq, but the analysis lacked enough depth because CLas reads were minoritary (Zuñiga et al., 2020). Therefore, to gain insight into CLas-host interaction during infection, higher resolution of in planta CLas transcripts is required.
Here, we developed a rapid CLas enrichment system using the dodder (Cuscuta campestris) to acquire and enrich CLas from infected citrus shoots. The efficiency of CLas enrichment by dodder was calculated based on CLas quantification. Genome sequencing of CLas strains from citrus shoots and the corresponding parasitized dodder samples were performed to evaluate whether CLas-enriched dodder samples can be used as amenable host sources for CLas genomic study. We compared the genome quality and analysed the possible genomic variation of CLas strains in citrus shoots and the parasitized dodder. To investigate whether there were others microbe can also be transmitted simultaneously with CLas by dodder, the composition of microbial community in CLas-infected citrus shoots and dodder was analysed by metagenomic analysis. In addition, the dual RNA-Seq was generated to compare transcriptome profiling of CLas in citrus and the parasitized dodder. Our data showed that the CLas-enriched dodder can be used as an optimal host for CLas genomic and transcriptomic analyses as compared to the citrus host.

Appearance of dodder grown on hydroponic CLasinfected citrus shoots
The CLas-free dodder was initially maintained on healthy periwinkle and then used to infest hydroponic CLas-infected citrus shoots (Fig. 1). Four representative types of dodder's growth were observed (Fig. 2). In Type I, dodders tendrils, coils and haustoria rapidly declined in three days, these samples were discarded and CLas quantitation analysis was not performed ( Fig. 2A). In Type II (Fig. 2B), dodder rapidly declined at early infesting stage, but the coils and haustoria remained alive for over seven days. Only the viable coils and haustoria of Type II dodder were collected for DNA extraction. In Type III (Fig. 2C), dodder continued to grow but the tendrils stopped elongating after cutting off the connection with periwinkle. The tendrils were maintained without decline between 8 and 24 days. In Type IV (Fig. 2D), dodder continued growing up to 24 days.

PCR quantitation of CLas in citrus shoots and dodders
In the hydroponic assays, we observed that all 48 successful parasitized dodder plants were positive for CLas, with concentration ranging from 42 to 834 072 cells per ng of total DNA.  Fig. 3). Compared to citrus shoots, the CLas concentration in the 30 dodders was enriched between two-and 419-fold (Fig. 3).
For the potted plant experiment, samples were monitored with time and it was observed that CLas population rapidly increased to a high level in the dodder after transmission from citrus plants (Fig. 4). The CLas concentration in dodder tendrils was low or even undetectable at day 1 (Fig. 4), then, increased rapidly and peaked at day 15 with an average of 9.6 9 10 5 CLas cells per ng of total DNA (Fig. 4). In contrast, increase of CLas in citrus plants was significantly slower (Fig. 4). The CLas concentration in dodder tendrils was significantly higher than those in citrus shoots after 10 days (P < 0.05) (Fig. 4). showed that the ratio of CLas reads/total reads in the dodder HiSeq data increased by 33.2-fold, 5.5-fold and  6.1-fold with respect to those of citrus HiSeq data in YNRL13, YNJS7 and GDHZ11 set respectively (Table 2). Mapping to prophage sequences (SC1, SC2 and P-JXGC-3) showed consistent results between citrus and dodder samples ( Table 3). The de novo assembly of dodder HiSeq data generated a higher quality CLas genome with fewer number of CLas contigs and longer N50 as compared to assembly result from citrus HiSeq data (Table 2). In addition, no genomic variation was found in CLas strains from citrus shoots and the parasitized dodder (Fig. S1).

Metagenomic analysis of CLas-infected citrus shoots and the parasitized dodders
In addition to CLas, the citrus chlorotic dwarf associated virus (CCDaV) and 'Candidatus Portiera aleyrodidarum' (CPa) were identified in both citrus shoot and dodder   (Table 4). For the others two sets (GDHZ11 and YNJS7), CLas was the most abundant (> 0.1% of total classified reads) microorganism found in citrus shoot and dodders. Analysis of taxa-specific qPCR in others 27 citrus/dodder sets identified CPa in only six citrus shoots and CCDaV in ten citrus shoots and their parasitized dodders (Table 4, Table S1). However, among ten CCDaV-positive citrus-dodder sample sets, no enrichment of CCDaV was observed in the parasitized dodder (Table S1). For instance, CCDaV was identified in a high concentration in six citrus shoots (with C t values ranged from 12 to 15) but in a lower concentration in their parasitized dodder tendrils (with C t values ranged from 22 to 29) (Table S1).

Efficiency evaluation for CLas gene expression profiling
Illumina HiSeq generated a total of 1.58 9 10 8 and 1.25 9 10 8 reads for citrus and dodder RNA samples respectively. Mapping to CLas A4 genome (CP010804.2) and prophage sequences (SC1, SC2 and P-JXGC-3) identified a total of 55 688 CLas reads in citrus HiSeq data by generating a consensus of 1 021 604 bp with~6.59 coverage, while a total of 137 795 CLas reads were identified in dodder HiSeq by generating a consensus of 1 208 490 bp with 16.2 X coverage. A total of 992 CLas genes (88.2% of total genes) were expressed in both citrus shoot and dodder. Sixty-five CLas genes (5.8% of total genes) were only expressed in dodder tendrils but at a low expression level (Fig. 5A, Table S3). It should be noted that a relative high correlation (r = 0.94) in expression levels of 992 CLas genes was observed between citrus and dodder RNA-Seq data (evaluated by TPM values) (Fig. 5B). Additionally, genome coverage based on reads mapping also revealed similar genome-wide gene expression pattern of CLas in citrus and dodder samples (Fig. 5C). a. N-C: N was the name of sample. N-C represented the CLas strain collected from citrus leave sample and N-D represented the CLas strain collected from dodder sample. b. The chromosomal region of CLas strain A4 (CP010804.2, nucleotide position from 1 to 1 191 963) was used as reference. Type 1 prophage = SC1 (NC_019549.1), Type 2 = SC2 (NC_019550.1) and Type 3 = P-JXGC-3 (KY661963.1). The average coverage of prophage region was counted based on the prophage type-specific region. The prophage type-specific region was identified according to Zheng et al., 2018. NF represented the corresponding type of prophage had not been found, i.e. no unique read mapped to the specific region of the corresponding prophage.  (Table 5). Genes involved in lipid transport and metabolism were most abundant (four gene transcripts) among up-regulated CLas genes in citrus compared to dodder (Table 5). Conversely, the 16S rRNA gene was the only identified DEG that was up-regulated in dodder compared to citrus (Table 5).
In addition, analysis of the top 20 most highly expressed genes in citrus shoot and dodders identified 17 CLas genes that encoded functions related to bacterial cell surface biogenesis and protein folding and assembly (Fig. 6). It was found that ribonuclease P (CD16_05550) and a collagen-like protein (CD16_RS05425) were highly expressed CLas gene in citrus and dodder (Fig. 6). It was also found that a cluster of five genes (from CD16_RS04875 to CD16_RS04895) were highly expressed in both citrus and dodder, though the function of the corresponding products is at present unknown. Interestingly, two Flp family Type IVb pilin genes (CD16_RS02370 and CD16_RS02375), predicted to be involved in intracellular trafficking, secretion and vesicular transport, were also highly expressed in citrus shoot and dodder (Fig. 6).

Discussion
Dodder has been used as a tool to transmit CLas from citrus to others citrus varieties or phylogenetically distant plants, such as periwinkle, or vice versa (Garnier and Bov e, 1983;Ke et al., 1986;Hartung et al., 2010;Zhang et al., 2011;Zheng et al., 2014a). Hartung et al., (2010)  carried out a detail study using dodder to study CLas, and they showed that CLas accumulated a high density in dodder tendrils. This was confirmed in our study as CLas density increased up to 419-fold in dodder tendrils compared to the host citrus (Fig. 3). We have further taken advantage of the dodder enrichment capacity to obtain high density of CLas sample that facilitate genomic and transcriptomic analysis. The use of hydroponic CLas-infected citrus shoot as source for dodder parasitizing was simple and efficient ( Fig. 1), although there were still some dodders that failed to enrich CLas from citrus shoot (Table 1). This was likely due to the failure of dodder to establish parasitism relationship with citrus shoots. CLas-infection caused a severe phloem collapse and callose plugging, which block the phloem system and in turn inhibited the transport of photoassimilates (Kim et al., 2009;Etxeberria et al., 2009Etxeberria et al., , 2012Folimonova and Achor, 2010;Achor et al., 2010;Koh et al., 2012;Brodersen et al., 2014). Thus, for some citrus shoots whose phloem system was already under the destructive effect by CLas, the nutrition supplies did not satisfy the dodder nutrient requirements. Preliminary observations suggested that the use of young/viable citrus shoots could increase dodder survival rate.
Early studies used dodder as host source for CLas genome sequencing for reason that it contained less contaminating host chloroplast and mitochondrial DNA (Zhang et al., 2011). Our data, in addition, support another advantage of dodder in CLas genomic studies that is the improvement of CLas genomic data quality ( Table 2, Table 3 and Fig. S1). However, we noted that there were still a very high ratio of host reads in dodder HiSeq data (Table 1), indicating high ratio of host DNA remained in the dodder total DNA. In a previous study (Zheng et al., 2014a), bacterial DNA in periwinkle host was enriched by removing the methylated host DNA using a microbial DNA enrichment kit. It will be of high interest to apply the same technique for further reduction of dodder DNA, i.e. to further enrich CLas in system reported here.
No genomic variation was found in CLas strains after transmission from citrus to dodder with short-term parasiting. In previous studies, it was proposed that the variation of intracellular bacteria was mostly induced by hostdriven selective forces during the host-bacteria interaction (Toft and Andersson, 2010). A previous study also reported the changes in the variable number of tandem repeats in CLas genome via psyllid transmission (Katoh et al., 2015). In this study, CLas was transmitted from citrus to dodder and only maintained for a short period (≤ 24 days), which seems to be not long enough to induce variations. Nutrients for growth of dodder, as well as the growth of CLas, were all from the citrus host. Then, it is likely that the similar nutrient composition in dodder tendrils and the phloem of citrus shoot could also reduce the host-driven selective forces for CLas and make the genetic variation much slower. In addition, the similar CLas gene expression patterns in citrus shoot and its parasitized dodder also suggested that similar CLas-host interaction in citrus and dodder took place (Fig. 6).
Few studies have reported the in planta genome-wide transcriptome profiling of CLas, until a recent study that employed RNA sequencing for analyses of CLas expression profile by using phloem-enriched samples from CLas-infected citrus trees (Zuñiga et al., 2020). However, less than 0.53% of total reads were identified as CLas reads in RNA-Seq data derived from CLas-infected citrus phloem samples (Zuñiga et al., 2020). The use of CLas-enriched dodder RNA sample in this study provided a higher resolution of CLas transcripts, and we observed similar gene expression patterns of CLas both in citrus host and dodders. Dodder was known to facilitate bidirectional movement of mRNAs between host and parasite (Kim et al., 2014). The CLas mRNA in dodder plant could be transferred from citrus host via the haustoria's connection or transcribed by CLas cells in dodder tendrils after transmission from citrus shoot. A previous study had demonstrated that the dodder micro-RNA (miRNAs) can target the host mRNA and regulate the host gene expression during parasitism (Shahid et al., 2018). However, the similar CLas transcriptome profiling in citrus and dodder indicated that the CLas gene expression may have not been affected by the dodder plants. This is consonance with dodder obtaining not only nutrients but also metabolites and proteins from host plants (Kim and Westwood, 2015). Therefore, it is not surprising the CLas transcriptome profiles in dodder and citrus are similar because they share a similar micro-environment that contained similar nutrients or even antimicrobial substance. It should be noted that two Flp family Type IVb pilin genes (CD16_RS02370 and CD16_RS02375) were highly expressed in both citrus and dodder tendrils (Fig. 6). The Type IV pilis are dynamic adhesive filaments in the surface of bacteria involved in host cells adherence, DNA uptake, twitching motility and bacterial interactions (Maier and Wong, 2015;Craig et al., 2019). An early study demonstrated that a CLas Type IV pilin gene flp3 was highly expressed in psyllid compared to in citrus plant and could be involved in attachment to the psyllid midgut (Andrade and Wang, 2019). In CLas-infected citrus, a filamentous-like material was observed in the surface of CLas cells and connected to the plasma membrane at sieve plate pore (Achor et al., 2020). Although it is still unclear whether these filaments are involved in CLas adherences to the plasma membrane and if they are Type IV pili.
In conclusion, we developed a rapid enrichment procedure by using dodder as an amenable host to capture and enrich CLas from the hydroponic excised CLas-infected citrus. The use of CLas-enriched dodder DNA samples generated higher quality CLas genome sequence data than those from citrus host. No genomic variation was observed in CLas strains after transmission from citrus shoots to dodder during short-term parasitizing. Dual RNA-Seq analyses of CLas-enriched dodder samples generated a higher resolution CLas transcriptome data than those obtained in citrus host, and a similar CLas gene expression pattern was found in dodder and citrus. The dodder-mediated CLas enrichment system can benefit both CLas genomic and transcriptomic research.

Experimental procedures
Plant material and water-culture maintaining of citrus shoot HLB-affected citrus shoots that used for dodder's parasitizing were collected from two citrus growing provinces in China, i.e. Guangdong and Yunnan provinces, between March 2018 and December 2018. A total of four citrus cultivars were collected, including Shatangju (Citrus reticulata Blanco cv. Shatangju), Lemon (Citrus limon), Gongkan (Citrus reticulata Blanco cv. Gongkan) and Nianju (Citrus reticulata Blanco cv. Nianju) ( Table 1,  Table S1). The symptomatic citrus shoots were kept moisturized before bring back to the laboratory from citrus orchards. The bottom of citrus shoot, about 1 cm in length, was further cut out. The remaining stem of young shoot was maintained with the distilled water in a 15 ml tube immediately and fixed with cotton to keep the leaves in the air (Fig. 1A). Three citrus leaves from each citrus shoot were sampled for DNA extraction and used for PCR detection of CLas. Citrus shoots, which were detected as CLas-positive and can be maintained in distilled water over three days with most leaves still attached, were further used for the CLas enrichment experiment.

Cultivation of dodder and CLas enrichment procedure
Dodder (Cuscuta campestris) was germinated from the seed and placed on CLas-free periwinkle plant (Catharanthus roseus) (Fig. 1B). Tendrils from vigorously growing dodder were used to infest another CLas-free periwinkle plant to get more dodder tendrils. Approximately a week after initial attachment, new tendrils emerged and elongated until to have 3-5 tendrils for each dodder. Thereafter, they can be used to parasitize the CLas-infected citrus shoots. CLas enrichment was performed on both hydroponic and potted plant experiments.
For the hydroponic assay, each CLas-infected citrus shoot was maintained hydroponically and parasitized by the tendrils from the individual dodder plant (Fig. 1C). After dodder successfully formed the haustoria and parasitized in citrus shoot, the connection between dodder tendrils and periwinkle was immediately cut off (Fig. 1D). The dodder tendrils and three citrus leaves (closed to the parasitizing site) were sampled simultaneously when the dodder tendrils/coils started to decline (e.g. the tendril started to wilt) (Fig. 1F, H). The survival period of dodder on citrus shoots was counted as the time starting at day that the connection between dodder and periwinkle was cut and ending at the day that dodder tendrils/coils began to decline.
For the potted plant experiment, six potted CLas-infected two-year citrus seedlings, which contained at least four symptomatic shoots, were used as receptor plants for dodder's parasitizing. The CLas-free dodder tendrils (grew on CLas-free periwinkle) were used to infest the CLas-infected citrus seedlings. The connection between dodder and periwinkle was cut off after the dodder successfully formatted the stable haustoria in citrus shoots/ branches. As a control, the dodder tendrils and three citrus leaves near the parasitized site were collected at first day when cut off the connection with periwinkle. Therefore, both citrus leaves and dodder tendrils from same branch were sampled every five days until 20 days. Dodder was removed from citrus plants as needed if the dodder grew excessively and began to damage the host. All dodder and citrus leaves samples were immediately put into the liquid nitrogen when sampled and brought back to the laboratory for both DNA and RNA extraction.

DNA extraction and CLas quantitation
For citrus leaves samples, total DNA was extracted from the leave midribs. For dodder samples, only fresh tissue (tendrils, coils or haustoria), which still showed green or light orange, were collected when it started to decline and used for DNA extraction. 100 mg of fresh tissue from citrus midribs or dodder was extracted by E. Z. N. A. HP Plant DNA Kit (OMEGA Bio-Tek Co., Guangdong, China).
Quantitation of CLas was performed by TaqMan â Real-time PCR developed by Bao et al. (2020) with primer-probe set, CLas4G/HLBp/HLBr. The plasmid PCLas4G, which harboured the CLas4G-HLBr specific region, was constructed by cloning a 78-bp fragment that was amplified using CLas4G/HLBr (Table S2) into pEASY-T1 vector (TransGen Biotech Co., Beijing, China). A standard equation, y = À3.201x + 42.7 (R 2 = 0.999), was developed to quantify CLas copy number in each DNA sample. Briefly, the quantification of bacterial populations was considered as cells per nanograms of total DNA. The concentration of plasmid PCLas4G and each DNA sample was determined using Qubit (Thermo Fisher Scientific Inc., Waltham, MA, USA). The copy number of plasmid PCLas4G was calculated using the formula: number of copies = (amount in nanograms 9 Avogadro's number)/(length in base pairs 9 1 9 10 9 9 650). The Avogadro's number is 6.022 9 10 23 , and the average weight of a base pair is assumed to be 650 daltons. The quantitative TaqMan â Real-time PCR was performed in CFX Connect Real-Time System (Bio-Rad, Hercules, CA, USA). The TaqMan â PCR reaction mixture contained 10 ll of Bestar â qPCR Master Mix (DBI â Bioscience, Shanghai, China), 1 ll of DNA template (~25 ng), 0.2 ll of PCR probe (10 lM), 0.4 lL of each forward and reverse primer (10 lM) in a final volume of 20 ll under the following procedure: 95°C for 2 min, followed by 40 cycles at 95°C for 10 s and 60°C for 30 s, with fluorescence signal capture at the end of each 60°C step. A standard equation, y = À3.201x + 42.7 (R 2 = 0.999), was developed based on a previous study by Bao et al. (2020). All data were analysed using Bio-Rad CFX Manager 2.1 software with automated baseline settings and threshold.
The enrichment fold of CLas concentration by dodder was calculated using the formula, enrichment fold = the copy number of CLas (cells/ng of total DNA) in dodder sample/the copy number of CLas (cells/ng of DNA) in the corresponding citrus samples. The copy number of CLas between citrus samples and the parasitized dodder tendrils was analysed by independent-sample t-test under the SPSS Statistic package (v19.0; IBM, Armonk, New York, NY, USA).

Genome sequencing and assembly
Three sets of total DNA samples extracted from CLas-infected citrus shoots and the parasitized dodders were selected for whole genome sequencing with Illumina HiSeq platform (Illumina, Inc., San Diego, CA, USA). Illumina sequencing was carried out by a commercial source. The citrus HiSeq data were initially mapped to the Citrus sinensis genome (AJPS00000000.1), Citrus clementina genome (AMZM00000000.1), Citrus sinensis mitochondrion genome (NC_037463.1) and Citrus sinensis chloroplast genome (DQ864733.1) by using Bowtie2 software (Langmead and Salzberg, 2012). For dodder HiSeq data, the Cuscuta australis genome (NQVE00000000.1) and Cuscuta campestris (OOIL00000000.1) were used as reference for reads mapping with Bowtie2 software (Langmead and Salzberg, 2012). All mapped reads were removed, and only the unmapped reads were retained for CLas assembly. A previously developed procedure (Zheng et al. 2016) was followed to perform de novo assembly and referenced-based assembly. Briefly, the de novo assembly of CLas genome was conducted with Velvet 1.2.10 (kmer = 75, min_contig_lgth = 1000) (Zerbino and Birney, 2008). Contigs from de novo assembly were blast against with CLas strain A4 genome (CP010804.2) and three prophage sequences (SC1, NC_019549.1; SC2, NC_019550.1 and P-JXGC-3, KY661963.1) by using Standalone BLASTn software (word_size = 28, evalue = 1e-5) (Camacho et al. 2009). The hit contigs from blast result were identified as the candidate CLas contigs and retrieved. All candidate CLas contigs from de novo assembly were further used to blast against NCBI nucleotide database by web BLAST with the default setting. The chimeric contigs were removed based on the web BLAST result, and the rest contigs composed the draft genome sequence of CLas. The draft CLas genome was further used as reference to remapped with HiSeq reads by using Bowtie2 software (Langmead and Salzberg 2012). All mapped reads were collected and designed as CLas reads. For reference-based assembly, the published CLas genomes (strain A4: CP010804.2, strain Psy62: CP001677.5 or strain JXGC: CP019958.1) and three prophage sequences (SC1: NC_019549.1, SC2: NC_019550.1 and P-JXGC-3: KY661963.1) were used as references to guide the assembly by CLC Genomic workbench 9.5 (length fraction = 0.95, similarity fraction = 0.95). The detail mapping report of each sequencing data was created, and the mapping consensus was extracted by CLC Genomic workbench 9.5. The total number of mapped reads of each HiSeq data was directly obtained from the mapping report. The average coverage of prophage region was counted based on mapping result with prophage type-specific region, which was identified according to Zheng et al. (2018).
The initial version of CLas genome was obtained by combining with the assembly contigs from both de novo assembly and referenced-based assembly based on sequence overlap using Standalone BLASTn software (word_size = 28, e-value = 1e-5). The optimized version of CLas genome was optimized by gap-closure PCR and Sanger sequencing. Genome annotation was performed on the RAST server (http://rast.nmpdr.org/) (Aziz et al. 2008).

Genome comparison of CLas from citrus and the parasitized dodder
Compared with the dodder-origin CLas genome (named CLas-D), the average coverage of CLas genome sequenced from citrus (named CLas-C) was lower (< 15x). To evaluate the possible genomic variation of CLas strains after transmission from citrus to dodder, the high quality CLas-D genome was used as reference for mapping with citrus HiSeq data by CLC Genomic Workbench v9.5 (length fraction = 0.95, similarity fraction = 0.95). The mapping result was further used for quality-based variation detection by CLC Genomic workbench v9.5 (the required variant reads count was set as the average coverage of whole genome sequence) to identify the possible genomic variation between CLas-D and CLas-C genome. All possible variations (single nucleotide polymorphisms, SNPs or Insertion/Deletion, In/Del) were further confirmed by using the variant-supported reads to blast against with both CLas-C and CLas-D genome through a web BLASTn-based method. Briefly, when the variation-contained reads can match the multiple copies loci in CLas chromosomal region or different types of prophage sequence with > 95% coverage and > 95% identity, the reads-covered variation was identified as the false-variation due to the high similar region between repeat loci in the CLas genome or among different types of prophages. The web BLASTn was performed with the default setting (expect threshold = 0.05, word size = 28).

Metagenomic analysis
HiSeq data of three sets of CLas-infected citrus shoots and their parasitized dodders were initially used for metagenomic analysis. All unmapped reads from dodder and citrus HiSeq data that retained after filtering with host genome (citrus or dodder) were used for metagenomic analysis with Kaiju software (Menzel et al., 2016). The NCBI non-redundant protein database was used for taxonomic classification by Kaiju software. Only taxa that comprised at least 0.1% of total classified reads were considered for further analyses. Each unique taxa in the taxa list of each sample was filtered by removing the microbe which was known as pathogen of human/animal or the common contaminating microbe in DNA extraction and other laboratory regents (Salter et al., 2014). An in silico evaluation was performed to initially confirm the existence of each microbe in the taxa list. Briefly, the representative genome sequence for each filtered taxa was downloaded from NCBI Nucleotide database and used as reference for reads mapping with filter reads. The covered of each reference genome from mapping result was calculated. The taxa were regarded as the false-taxa and removed from the taxa list if the covered of the corresponding reference genome and the consensus coverage were significantly low (less than 5% covered and 59 depth). In addition to in silico evaluation for each candidate taxa, the taxa-specific Real-time PCR primer set (Table S2) was designed based on consensus sequence (from mapping result) of each microbe by using Primer 3 software (Untergasser et al., 2012) and further used to test in 27 citrus-dodder sample sets (including CLas-enriched and non-enriched sets). The possible transmission of each taxa between citrus and dodder was evaluated based on the PCR result.
In planta genome-wide gene expression of CLas in citrus and the parasitized dodder To ensure the sufficient sequencing data for gene expression analysis of pathogen, the dual RNA-Seq typically required a relatively large amount of host cells be infected by pathogen. Therefore, citrus leaves samples with a high CLas titre were selected for RNA extraction. The dodders tendrils with highest CLas concentration collected after 15 days' parasitizing ( Fig. 4), as well as three leaves from the host citrus shoot, were used for RNA extraction and dual RNA-Seq analyses. Three biological replicates of citrus leaves and dodder tendrils samples were collected from different citrus/dodder experiment sets. The three replicates were further mixed as one sample for RNA extraction. RNA was extracted by using Plant RNA Kit (OMEGA Bio-Tek Co.). The quality of total RNA sample was evaluated by Qubit 2.0 (Thermo Fisher Scientific Inc.) and Agilent 2100 (Agilent Technologies Inc., Santa Clara, CA, USA). The qualified RNA samples were further used for library preparation with a TruSeq RNA library Prep Kit (Illumina, San Diego, CA, USA) by removing rRNA from total RNA. Sequencing was carried out on an Illumina HiSeq 3000 system with 150-bp paired-end reads by a commercial source. All HiSeq reads from citrus leaves and dodder tendrils samples were mapped to CLas strain A4 genome (CP010804.2) by CLC Genomic workbench v9.5 (QIAGEN Bioinformatics, Aarhus, Denmark) (length faction = 0.95; similarity fraction = 0.95). Reads mapped to each CLas gene were then summarized into count tables of 'Total Gene Reads'. The Transcripts Per Kilobase Million (TPM) method was used for normalization of RNA-seq, i.e. TPM = A910 6 9 1/∑(A), where A = total reads mapped to gene 9 10 3 /gene length in bp. Differentially expressed genes (DEGs) between citrus leaves HiSeq data and dodder HiSeq data were identified by GFOLD V1.1.4 with cut-off values setting as Log 2 Fold change ≥ │1│ (Feng et al., 2012). Functional annotation and orthologs assignment of identified DEGs were performed by eggNOG-mapper (Huerta-Cepas et al., 2017). The top 20 most highly expressed CLas genes that evaluated by TPM value from citrus leaves data and dodder data were manually retrieved and compared. Heat map for comparison of CLas gene expression (by TPM value) between citrus leaves and dodder was generated in TBtools software . The gene-specific Real-time PCR for 20 selected DEGs was used to verify the gene expression result based on RNA-Seq. All primers were listed in Table S2. RNA-Seq data used in this study had been deposited in the NCBI Sequence Read Archive (SRA) database (BioProject: PRJNA675299 and Submission ID: SUB8484188).

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Evaluation of the genomic variation of 'Candidatus Liberibacter asiaticus' by the quality-based variation detection based on reads mapping from citrus samples and its parasitized dodder samples. Reads mapping was generated by using dodder-origin CLas genome (including the chromosomal region, strain-C and prophage region, Pstrain-1, 2, or 3) as reference for mapping with citrus HiSeq data. The quality-based variation detection of each read mapping was performed with CLC Genomic workbench v9.5. Table S1. Quantification of 'Candidatus Liberibacter asiaticus' in citrus and the parasitized dodder and taxa-specific PCR result. Table S2. General information of PCR primers used in this study. Table S3. Genes expression profiling of 'Candidatus Liberibacter asiaticus' in citrus and its parasitized dodder.