Plastid Genome Assembly Using Long-read data

Abstract

Although plastid genome (plastome) structure is highly conserved across most seed plants, investigations during the past two decades have revealed several disparately related lineages that experienced substantial rearrangements. Most plastomes contain a large inverted repeat and two single-copy regions, and a few dispersed repeats; however, the plastomes of some taxa harbour long repeat sequences (>300 bp). These long repeats make it challenging to assemble complete plastomes using short-read data, leading to misassemblies and consensus sequences with spurious rearrangements. Single-molecule, long-read sequencing has the potential to overcome these challenges, yet there is no consensus on the most effective method for accurately assembling plastomes using long-read data. We generated a pipeline, plastid Genome Assembly Using Long-read data (ptGAUL), to address the problem of plastome assembly using long-read data from Oxford Nanopore Technologies (ONT) or Pacific Biosciences platforms. We demonstrated the efficacy of the ptGAUL pipeline using 16 published long-read data sets. We showed that ptGAUL quickly produces accurate and unbiased assemblies using only ~50× coverage of plastome data. Additionally, we deployed ptGAUL to assemble four new Juncus (Juncaceae) plastomes using ONT long reads. Our results revealed many long repeats and rearrangements in Juncus plastomes compared with basal lineages of Poales. The ptGAUL pipeline is available on GitHub: https://github.com/Bean061/ptgaul.

1 |. INTRODUCTION

Plastid genomes (plastomes) are highly conserved, comprising linear, branched or occasionally circular molecules that usually contain a large, inverted repeat (IR) and large and small single-copy regions (LSC and SSC). Due to their conserved structure and low rate of nucleotide substitution, plastome data have made substantial contributions to phylogenetic studies for many plant groups (Jansen & Ruhlman, 2012; Jiang et al., 2022; Liu et al., 2022; Xia et al., 2022; Xu et al., 2022; Yu et al., 2022). Despite the high level of plastome structural conservation in seed plants, rearrangements—including inversions, and expansion and contraction of the IR—and IR loss have occurred in unrelated lineages of gymnosperms and angiosperms (Ruhlman & Jansen, 2021). Many of these same lineages experienced substantial gene loss, with most of these genes functionally transferred to the nuclear genome or substituted by an alternative, nuclear-encoded gene (Ruhlman & Jansen, 2021). Documented transferred/substituted genes include accD, infA, rpl22, rpl20, rpl32, rpl23, rps7, rps16, ycf1 and ycf2.

Genome assembly methods have improved substantially over the past decade (Freudenthal et al., 2020; Twyford & Ness, 2017). novoplasty (Dierckxsens et al., 2017) and getorganelle (Jin et al., 2020) are the two most frequently used pipelines for plastome assembly based on Illumina short reads. However, these assemblers, which rely on either the seed-extend method (Dierckxsens et al., 2017) or the De Bruijn graph approach (Compeau et al., 2011), do not always yield accurate assembly results when confronted with long repeat regions in plastomes, particularly when those repeats are larger than kmer sizes. Sometimes these tools generate outputs with multiple contigs/scaffolds or hundreds of possible assembly results. The high number of uncertain paths can sometimes be corrected using bandage (Wick et al., 2015), a software tool that visualizes the depth of read coverage for each contig/scaffold and orders contigs. However, the final arrangement of the contigs is often not well resolved because the Illumina short reads are insufficient to bridge the repeated sequences and their flanking regions. Short reads with a typical insert size (300–400 bp) are inadequate to obtain a complete plastome assembly for plant species that have large repeats and that may be highly rearranged. So far, few plant systematists have recognized this as an issue, probably because most plants possess relatively conservative plastome structures with limited repeated sequences and because their primary interest is the extraction of coding sequences for phylogenetic analyses. This may shift as the cost of sequencing continues to decline and the use of entire plastomes for analysis becomes more common. Long reads generated by third-generation sequencing methods such as Oxford Nanopore Technologies (ONT) or Pacific Biosciences (Pacbio) platforms may help resolve assembly issues as the longer reads are more likely to span long repeats (Liao et al., 2021).

To date, there are several tools available to assemble organelle genomes using long-read data and hybrid data (both short- and long-read data), including organelle_pba (Soorni et al., 2017), canu (Koren et al., 2017), unicycler (Wick et al., 2017) and flye (Kolmogorov et al., 2019; Syme et al., 2021). However, these pipelines have some drawbacks. organelle_pba was designed exclusively for PacBio data; the Sprai (Miyamoto et al., 2014) and Celera (Miller et al., 2008) assemblers in organelle_pba are no longer maintained, limiting its extension to assembly with hybrid data sets. The approach of Syme et al. (2021) requires an extra step to manually filter a subset of raw reads matching the plastome (~250× coverage) and sometimes generates multiple contigs in the assembly result. canu can yield different results depending on different read coverages (Wang, Schalamun, et al., 2018). unicycler was designed for hybrid data; however, it takes an extremely long time to finish as input data are increased. All the above pipelines can probably assemble the conventional plastome but cannot assemble atypical plastome structures accurately. Alternatively, a “fishing approach” using either shasta (for Nanopore data; Shafin et al., 2020) or hifiasm (for PacBio data; Cheng et al., 2021; Feng et al., 2022) to assemble the raw reads first, followed by fishing out the plastid contigs using the reference genome can be used. Nevertheless, this assembly process is time- and resource-consuming when the input data are large, according to the guidelines of each software. Additionally, the accuracy of the assembly results may be affected by a considerable number of redundant sequences (Zhang et al., 2022).

The angiosperm family Juncaceae contains ~500 species within the seven genera Juncus L., Luzula DC., Distichia Nees and Meyen, Oxychloe Philippi, Patosia Buchenau, Marsippospermum Desv. and Rostkovia Desv. (Drábková, 2010). Juncus is the largest genus and includes ~300 species (Balslev, 2018) and two major subgenera, Agathryon and Juncus (Drábková, 2010; Drábková et al., 2006). Although many species of Juncaceae have been included in phylogenetic studies using plastid gene sequences and the internal transcribed spacer (ITS) region of the nuclear ribosomal repeat (Table S1; Brožová et al., 2022; Drábková, 2010; Drábková et al., 2006), species relationships within Juncus remain unresolved. Brožová et al. (2022) recently incorporated rbcL, trnL, trnL-trnF and ITS1–5.8–ITS2 regions to reorganize Juncus into seven distinct genera: Juncus, Verojuncus, Juncinella, Alpinojuncus, Australojuncus, Boreojuncus and Agathryon.

Not many plastome structures have been reported for Juncus (s.l.). To avoid confusion regarding Juncus species names, we did not adopt the generic classification of Brožová et al. (2022) in the present investigation because a more comprehensive study with more markers is necessary to justify this reranking. Plastomes of just eight Juncus species are publicly available in GenBank (Lu et al., 2021; Wu et al., 2021), but there has been little investigation of how the plastome itself has changed structurally in Juncus (s.l.). Wu et al. (2021) focused on the phylogenetic relationships in the Poales using shared plastid protein-coding genes, but did not provide any information on plastome structure. Lu et al. (2021) assembled the plastome of Juncus effusus using velvet (Zerbino, 2010) and novoplasty (Dierckxsens et al., 2017) with gapfiller (Nadalin et al., 2012) without any confirmation by either long-range PCR or long-read data, leaving the final structure uncertain. Recently, two Juncus (J. effusus and J. inflexus) nuclear genomes were assembled by Planta et al. (2022) but no plastomes were reported. Adding more complete plastomes of Juncaceae would allow a deep understanding of plastome evolution inside the family and provide more phylogenetic insights within Juncaceae and Poales.

To assist in assembling potentially complex plastomes and to explore structural variation in Juncus, we created a pipeline, plastid Genome Assembly Using Long-read data (ptGAUL), which assembles plastomes using raw ONT long-read sequencing data. After instantiating this tool, the aims of our study were to: (i) test the reliability of the ptGAUL pipeline using 16 published plastomes; (ii) employ the pipeline to assemble plastomes of two Juncus species (J. validus and J. roemerianus) sequenced in our study and assemble two other species (J. effusus and J. inflexus) from the reads of Planta et al. (2022); and (iii) compare plastome evolution in Juncus to selected members of the Poales.

2 |. MATERIALS AND METHODS

2.1 |. Juncus sample collection and DNA extraction

Young leaves of Juncus roemerianus (voucher number; NCU00441655) and Juncus validus (voucher number: NCU00434802) were collected from North Carolina, USA, and stored in silica gel. Vouchers were deposited in the herbarium of the University of North Carolina at Chapel Hill (NCU). Total genomic DNA extraction of dried leaves was performed using a modified cetyltrimethylammonium bromide (CTAB) protocol described by Cullings (1992) and Xiang et al. (1998). DNA quantity was analysed with Qubit 2.0 (Life Technologies) and quality was measured using a NanoDrop spectrophotometer 2000 (ThermoFisher Scientific) and 1% (w/v) agarose gels. Sequencing was performed at the High-Throughput Sequencing Facility (HTSF) at UNC Chapel Hill. For Illumina sequence libraries (Illumina), ~250 ng of total DNA was utilized. An agilent 2100 Bioanalyzer (Agilent Technologies) was used to select ~450-bp fragments for Novaseq 6000, 250-bp paired-end (PE) sequencing. For the Oxford Nanopore sequencing, ~2000 ng of high-molecular-weight DNA was prepared using the ligation sequencing kit (SQK-LSK109) and sequenced on two R9.4.1 flowcells (Oxford Nanopore Technologies).

2.2 |. ptGAUL pipeline and validation.

We generated a pipeline to facilitate plastome assembly using long-read data, which can be applied to both PacBio and ONT raw reads. The ptGAUL pipeline (Figure 1) includes three major parts: filtering long reads, setting the depth of coverage and assembling the filtered plastid data. Step 1: use minimap2 (Li et al., 2018) to find all reads that map partially or entirely to the closely related reference plastome, followed by filtering all reads using a customized bash script. Then, use seqkit (Shen et al., 2016) to keep long reads greater than a specified length (default is 3000 bp, “–f” in ptGAUL). Step 2: calculate the coverage by assembly-stats (available at https://github.com/sanger-pathogens/assembly-stats). If the coverage is over 50×, apply seqtk (available in https://github.com/lh3/seqtk) to randomly select a subset of data, including about 50× coverage of the plastome (excessive coverage will slow down the assembly and may result in a failure to assemble). Step 3: use flye (Kolmogorov et al., 2019) to assemble the plastome. When only three contigs are detected in the graphical fragment assembly (.gfa) file, we use “combined_gfa.py,” a customized python script, to assemble the plastome into two different paths. Otherwise, the assembly result can be checked manually using bandage. All pipeline code was deposited on GitHub (https://github.com/Bean061/ptgaul). In Steps 2 and 3, we implement seqtk to randomly choose a subset of long reads to minimize the bias of read selection, which can speed up the assembly process. To make it more user-friendly, we added three parameters that included: the expected plastome size (-g) with default value of 160,000 bp, the expected coverage of plastome data for assembly (-c) with a default value of 50× coverage, and the output directory (-o).

FIGURE 1.

ptGAUL workflow. The program starts with an initial filtering step to filter the long reads of the target species using at least one closely related reference plastome (1). Subsequently, the coverage for those filtered long reads is calculated and filtered to make sure it is about 50× (2). Finally, two paths of plastomes were obtained through flye and a customized Python script, combine_gfa.py (3).

After assembly, if short-read sequencing data are available, the FM-index Long-Read Corrector (fmlrc) software (Wang, Holt, et al., 2018) is recommended to polish and improve the accuracy of the final assembled sequences because it can generate a more accurate assembly result (Mak et al., 2023).

Long-read data from 16 published plastomes in NCBI were used to validate the efficacy of ptGAUL for assembly (Table 1). All analyses were run on the longleaf high-performance compute cluster at UNC Chapel Hill. Comparative analyses were performed that included the number of assembled contigs, total genome size (bp) and nucleotide sequence identity between the published results and those obtained with ptGAUL through a pairwise identity alignment using geneious version 2022.2 (Kearse et al., 2012). We also compared the performance (memory usage and running time) of ptGAUL to a “fishing approach” by contrasting it against shasta version 0.10.0 for ONT data (Shafin et al., 2020) and hifiasm version 0.16 for PacBio data (Cheng et al., 2021; Feng et al., 2022). Then, we used minimap2 version 2.24 (Li et al., 2018) to map the assembled contigs and a reference sequence (Table S6) to locate the plastome.

TABLE 1.

ptGAUL performance on 16 published sequence data sets, including the information of assembled plastome from published papers and the information on assembled plastomes from ptGAUL.

Species	Library preparation and sequencing methods	Raw read no./N50 (bp)	Reference	Plastid size from ptGAUL (bp) (% nucleotide sequence identity to references)	No. of assembled plastid contigs from ptGAUL	Plastome reference used for ptGAUL (reference length from original studies)a
Arctostaphylos glauca	WGS/PacBio	1,814,591/15,245	Huang et al. (2022)	150,241 (NA)	3 (S)	NC_035584.1/NC_042713.1/NC_047438.1/JAHSPW020000272.1 (118,663 bp)
Lepidium sativum	WGS/PacBio	400,322/7277	Zhu et al. (2019)	153,666 (99.9%)	3 (S)	NC_047178.1 (154,997 bp)
Chaetoceros muellerii	WGS/PacBio	87,313/12,921	Li and Deng (2021)	117,304 (99.8%)	1 (S)	MW004650.1 (116,284bp)
Potentina micrantha	WGS/PacBio	28,638/2464	Ferrarini et al. (2013)	159,850 (99.8%)	3 (S)	NC_015206.1 (155,691 bp)
Durio zibethinus	WGS/PacBio	853,182/9670	Shearman et al. (2020)	142,806 (99.95%)	1 (S)	MT321069 (163,974 bp)
Beta vulgaris	WGS/PacBio	96,874/3980	Stadermann et al. (2015)	155,383 (99.9%)	3 (S)	KR230391.1 (149,722 bp)
Eleocharis dulcis	WGS/PacBio	68,167/16,288	Lee etal. (2020)	199,919 (99.5%)	3 (S)	NC_047447.1 (199,561 bp)
Eucalyptus pauciflora	WGS/ONT	705,554/24,988	Wang, Schalamun, et al. (2018)	158,561 (99.0%)	1 (S)	MZ670598.1/HM347959.1/NC_014570.1/AY780259.1/NC_039597.1 (159,942bp)
Leucanthemum vulgare	Long-range PCR/ONT	18,031/7900	Scheunert et al. (2020)	119,593 (NA)	5(F)	NC_047460.1 (150,191 bp)
Oryza glaberrima	Plastid capture/ONT	81,363/4319	Bethune et al. (2019)	124,133 (NA)	4(F)	NC_024175.1 (132,629 bp)
Cenchrus americanus	Plastid capture /ONT	105,760/5580	Bethune et al. (2019)	143,162 (96.6%)	3 (S)	NC_024171.1 (140,718 bp)
Digitana exilis	Plastid capture /ONT	141,250/4028	Bethune et al. (2019)	136,650(96.0%)	3 (S)	NC_024176.1 (140,908 bp)
Podococcus acaulis	Plastid capture /ONT	249,417/2621	Bethune et al. (2019)	81,976 (NA)	2(F)	NC_027276.1 (157,688 bp)
Raphia textilis	Plastid capture /ONT	83,833/2495	Bethune et al. (2019)	60,089 (NA)	2(F)	NC_020365.1 (157,270 bp)
Phytelephas aequatorialis	Plastid capture /ONT	202,925/2551	Bethune et al. (2019)	NA	(F)	NC_029957.1 (159,075 bp)
Picea glauca	WGS/PacBio	563,675/4671	Soorni et al. (2017)	123,476 (98.9%)	1 (S)	NC_021456.1 (124,084 bp)

Abbreviations: F, the samples failed using ptGAUL; NA, low nucleotide sequence identity between assembled plastome between published data and our data; S, the samples are well assembled by ptGAUL.

^aThis column includes the references we used for genome assembly in ptGAUL and the references in bold type were considered as references for comparisons with ptGAUL results.

2.3 |. Assembly and comparison of four Juncus species

The Illumina Novaseq 6000 platform (Illumina) was used to generate 250-bpPE reads for J. roemerianus and J. validus. Reads were de novo assembled using getorganelle version 1.7.5 (Jin et al., 2020) with default settings. Long-read data were also generated using ONT for J. roemerianus and J. validus. Long-read data were assembled using ptGAUL with default parameters (“–f”=3000 bp; “–g”=160,000 bp; “–c” = 50×) with all eight Juncus plastomes from GenBank (Table 2) as references for the filtering step. We verified the assembly graph results (.gfa file) obtained with flye version 2.7 in bandage version 0.8.1 (Wick et al., 2015). Then, we used fmlrc version 1.0.0 to polish the final plastomes (an optional step in the ptGAUL pipeline).

TABLE 2.

Summary of features of the plastid genomes of four Juncus species, including length, GC content and gene numbers.

Genome features	J. effusus	J. effusus	J. inflexus	J. roemerianus	J. validus
Accession no.	NC_059754.1	Present study	Present study	OP235509	OP235510
No. of Illumina read clusters	12,443,053	96,653,565	83,412,073	158,922,322	156,712,430
No. of ONT reads and N50	0	2,960,380/21,529	2,735,792/24,397	427,549/15,998	243,884/14,365
Plastid genome size (bp)	170,612	178,158	181,566	196,852	147,183
LSC length (bp)	81,818	86,497	86,649	82,944	87,215
SSC length (bp)	7522	7539	7509	7902	2046
IR length (bp)	40,636	42,061	43,704	53,003	28,961
Overall GC content (%)	36.0	35.9	35.6	32.2	34.7
GC content In LSC (%)	33.2	33.2	33.3	33.1	31.6
GC content in SSC (%)	26.3	26	26.2	26.5	23
GC content in IR (%)	39.7	39.5	38.7	37.5	39.8
Total no. of genes	129	133	134	136	114
No. of unique genes	105	106	106	106	93
No. of unique PCGs	72	72	72	72	60
No. of unique tRNA genes	29	30	30	30	29
No. of unique rRNA genes	4	4	4	4	4

Note: The plastome data of Juncus effusus were from two different two different sources, this paper and Lu et al. (2021). PCG, protein-coding genes.

To assess the quality of the assembly result, we mapped all raw Illumina and ONT reads of each Juncus species to our polished assembly and tested the evenness of the coverage at all sites. If every site shared a similar coverage of raw reads without gaps in coverage, this usually indicated a good de novo assembly result. We used the samtools version 1.9 (Danecek et al., 2021) depth function to record read depth at every site, followed by a dot plot created using python’s matplotlib library (Hunter, 2007). We downloaded the raw whole-genome sequencing data of J. effusus and J. inflexus (both ONT reads: SRR14298760/SRR14298751 and Illumina reads: SRR14298746/SRR14298745 from Planta et al., 2022) to assemble plastomes following the same steps detailed above.

After assembly, we uploaded the plastomes of four Juncus species (J. roemerianus, J. validus, J. effusus and J. inflexus) to geseq online (Tillich et al., 2017) for annotation using chole (Zhong, 2020), hmmer (Finn et al., 2011) and blat (Kent, 2002). We manually checked the start and stop codons of each annotated gene using geneious version 2022.2. The genes not in frame in each Juncus species were either adjusted or removed after a careful comparison with Typha latifolia plastid annotation (NC_013823; Guisinger et al., 2010) by mapping its annotations to our Juncus assemblies. For the uncertain tRNAs, we confirmed the tRNA secondary structures via RNAfold Webserver (Hofacker, 2003). Linear plastome maps were drawn with ogdraw version 1.2 (Lohse et al., 2013). Circular representations were drawn using circoletto (Darzentas, 2010) to visualize the repeats.

2.4 |. Examination of repeats and rearrangement events in Juncus

We removed one copy of the IR region prior to the repeat analyses to avoid counting the repeats from this region. We implemented blast version 2.8.1+ (Altschul et al., 1990) and tandem repeats finder version 4.09.1 (Benson, 1999) to detect the dispersed repeats and tandem repeats, respectively, following the steps from Lee et al. (2020). We manually checked the result, eliminated duplicated blast hits and recorded the total number of distinct dispersed repeats. We also downloaded the complete plastomes of Eriocaulon decemflorum (NCJD44895; Darshetkar et al., 2019) and two early diverging Poales, Typha latifolia (NC_013823; Guisinger et al., 2010) and Ananas comosus (NC_026220; Nashima et al., 2015), for comparison. All the plots were drawn using the matplotlib library (Hunter, 2007) from Python.

We focused on the four confirmed assemblies of Juncus, namely J. roemerianus, J. validus, J. effusus and J. inflexus, for characterizing and comparing the rearrangements in Juncus plastomes. The other eight publicly available (Lu et al., 2021; Wu et al., 2021) Juncus plastomes on GenBank were excluded from the rearrangement analyses (Table 2) because of the unreliable assemblies resulting from short-read data. To eliminate uncertainty in short-read assemblies, we compared the sequence identity between the J. effusus plastome assembled from short-read data (Lu et al., 2021) and the ptGAUL-assembled plastome of J. effusus from long-read data (Planta et al., 2022). To detect rearrangement events within Juncus, whole-genome alignments of J. roemerianus, J. validus, J. effusus and J. inflexus were performed to examine the arrangements of locally colinear blocks (LCBs) using progressive mauve (Darling et al., 2004). One copy of the IR was removed from plastomes prior to mauve alignment to prevent spurious alignments. Typha latifolia was used as a reference, and Ananas comosus and Eriocaulon decemflorum were also included.

3 |. RESULTS

3.1 |. Validation of ptGAUL

Overall, ptGAUL assemblies were successful. Assemblies contained either one or three contigs in 11 of 16 species, with plastome sizes similar to those reported previously (indicated with an “S” in Table 1). The assembly graph results (.gfa files) showing plastome structure were visualized and confirmed in bandage and deposited in GitHub (https://github.com/Bean061/ptgaul). Assembled plastomes had >95% nucleotide sequence identity to the references, but the plastome of Arctostaphylos glauca was 31,578 bp longer (21% total length) than the published data (Table 1). ptGAUL failed to assemble plastomes of five species (indicated with an “F” in Table 1). The ptGAUL pipeline produced consistent and reliable results with a data set of long reads (>5000 bp N50) with ~50× coverage of the plastome.

Our results indicated that different library preparation methods affected plastome assembly, regardless of the long-read sequencing platform (PacBio or ONT) employed (Table 1). Plastomes derived from a whole genomic sequencing approach assembled correctly (either one or three contigs), with a reasonable plastome length and structure (verified in bandage), while the plastomes using plastid capture approaches (i.e., long-range PCR and long-fragment target capture) were more fragmented and had a smaller genome size. For example, Leucanthemum vulgare had a similar N50 value to Lepidium sativum (7900 and 7277 bp, respectively), but the Leucanthemum vulgare library prepared using long-range PCR failed in plastome assembly. All five failed data sets involved the plastid capture approach and most of the raw sequence reads had relatively short length with small N50 values (<5000 bp) (Table 1).

3.2 |. Plastome features of four Juncus species

We generated 158,922,322 and 156,712,430 short reads for J. roemerianus and J. validus, respectively, along with 427,549 ONT reads from J. roemerianus (N50 value: 15,998 bp) and 243,884 ONT reads from J. validus (N50 value: 14,365) (Table 2). The data are accessible at NCBI under the BioProject accession no.: PRJNA865266 (SRR21976089; SRR21976090; SRR21976091; SRR21976092). We also downloaded sequence data (PRJNA723756) of J. effusus and J. inflexus from Planta et al. (2022) (Table 2). The ptGAUL pipeline produced three contigs each for J. validus and J. roemerianus (Figure S1a,b) sequenced in this study, and one contig each for J. effusus and J. inflexus sequenced by Planta et al. (2022) (Figure S1c,d). The final assembled plastomes of J. validus, J. roemerianus, J. effuses and J. inflexus ranged from 147,183 to 196,852 bp, had similar sized LSCs, different sizes of the SSC and large differences in IR size (Table 2).

The assemblies for the four Juncus species were verified by mapping both Illumina and ONT reads back to the plastome assembly. All mapping results showed a high and even coverage of four species (Figure 2c–f; Figure S2a,b,d,e). There were no gaps in the assemblies regardless of the sequencing platform. Annotation of the four Juncus plastomes revealed that they contained 114–136 genes, 93–106 of which were unique. There were 60–72 unique protein coding genes (PCGs), 29–30 tRNA genes and four rRNA genes (Table 2). J. roemerianus had the greatest gene number (136), which is similar to J. effusus (133), J. inflexus (134), and basal Poales ancestors such as Typha latifolia (133), Ananas comosus (132) and Eriocaulon decemflorum (135) (Table S2). J. effusus and J. inflexus shared a highly similar gene content while J. validus lacked 11 ndh genes, rps15 and trnT-GGU (Figure S3; Table S2).

FIGURE 2.

Plastome structural maps and read coverage graphs of Juncus roemerianus and J. validus. (a,b) Linear maps of the J. roemerianus and J. validus plastome, respectively, were drawn by ogdraw (Lohse et al., 2013). Genes that belong to different functional groups are colour-coded. Small single copy (SSC), large single copy (LSC) and inverted repeats (IRa, IRb) are indicated for both plastomes. Circular representations of the two Juncus plastomes were used to show locations of repetitive DNA using circoletto (Darzentas, 2010). The blue lines represent dispersed repeats in the plastome, while red regions represent the IR regions, (c–f) Read coverage plots of J. roemerianus and J. validus using Illumina reads and ONT reads, respectively, showing the good quality of the assemblies. The x-axis represents the position in the plastome, while the y-axis represents the coverage.

3.3 |. Verification of published J. effusus plastome

We compared the J. effusus published assembly based on short-read data (Lu et al., 2021, MW366789) with our new ptGAUL assembly employing the long-read data from Planta et al. (2022). The result indicated that the short-read assembly generated by Lu et al. was >7.5 kb shorter than our long-read assembly (170,612 vs. 178,158 bp). The mapping results showed that our assembly was well supported by both long- and short-read data from Planta et al. (2022) (Figure S2a,b). Yet, it was unsupported by the Illumina reads from Lu et al. (2021) with 777 positions with less than 10× coverage, including 295 positions that had no read coverage (Figure S2c). The previous short-read assembly of J. effusus (MW366789) was not supported by the long-read data from Planta et al. (2022). Based on these results, we removed the eight publicly available Juncus plastomes assembled with short-read data prior to the comparative analyses of plastomes as we thought that these would unfairly bias the comparisons in favour of ptGAUL.

3.4 |. Repeats in Juncus plastomes

Repeat analyses identified many dispersed and tandem repeats in the four Juncus plastomes (17.2%–24.3% of the genome without IRa) in comparison with basal Poales and Eriocaulon (1.8%–3.3% of the genome without IRa) (Table 3). The combined length of both dispersed and tandem repeats in Juncus plastomes ranged from 22,577 bp (J. validus) to 34,027 bp (J. roemerianus), which was far greater than for Typha (4436 bp), Ananas (3552 bp) and Eriocaulon (2227 bp) (Table 3). When dispersed repeats were parsed into five different size classes, Juncus plastomes contained more dispersed repeats than basal Poales and Eriocaulon (Figure 3; Table S3). Larger repeats (>201 bp) were found only in Juncus (Figure 3; Table S3). Among four Juncus plastomes, J. effusus and J. validus had more abundant dispersed repeats, yet J. roemerianus was the only one with a repeat larger than 1 kb. Juncus plastomes also experienced substantial accumulation of tandem repeats (Table 3). Tandem repeat accumulation was higher than that of the dispersed repeats in J. inflexus and J. roemerianus. All four Juncus plastomes contained exceptionally expanded tandem repeats, ranging from 4.6 to 6.6 kb, some of which contain clpP (Table S4).

TABLE 3.

Statistics of dispersed and tandem repeats in Typha, Ananas, Eriocaulon and Juncus plastomes.

	Typha latifolia	Ananas comosus	Eriocaulon decemflorum	Juncus effusus	Juncus inflexus	Juncus roemerianus	Juncus validus
Genome size (no IRa)	134,642	132,862	125,164	136,097	137,859	143,849	118,221
GC (%)	35.5	36.3	34.2	34.8	34.6	34.3	33.5
Dispersed repeat (DR)
Length of DRs	1210	1495	1418	15,117	13,229	14,712	14,714
GC (%)	33.7	36.3	33	35	36	35.7	34
GC (% without DR)	35.5	36.3	34.2	34.7	34.6	34.1	33.3
Percentage of DR in genome	0.9	1.1	1.1	11.1	9.6	10.2	12.4
Tandem repeat (TR)
Length of TRs	3270	2057	859	12,248	15,783	22,978	8797
GC (% of TRs)	8.8	18.4	20	34.2	33.4	32.1	32.1
Genome size without TRs	131,372	130,805	124,305	123,849	122,076	120,871	109,424
GC (% without TRs)	36	36.6	34.3	34.8	34.8	34.8	33.6
Percentage of TRs in genome	2.4	1.5	0.7	9.0	11.4	16.0	7.4
Total repeat
Length of total repeats	4436	3552	2227	23,345	26,451	35,027	22,577
GC (% of total repeats)	12.6	21.5	27.5	34.7	34.7	33.6	33.5
GC (% without total repeats)	36	36.6	34.3	34.8	34.8	34.5	33.4
Percentage of total repeats in genome	3.3	2.7	1.8	17.2	19.2	24.3	19.1

FIGURE 3.

Bar plot of dispersed repeats in plastomes from seven Poales species, including four newly assembled Juncus species.

3.5 |. Rearrangement of Juncus plastomes

Whole-genome alignment using progressivemauve (Figure 4) detected 27 LCBs from seven complete plastomes (Typha latifolia, Ananas comosus, Eriocaulon decemflorum, Juncus effusus, J. Inflexus, J. roemerianus and J. validus). The plastomes of the two basal Poales and Eriocaulon were colinear, whereas all Juncus species have many breakpoints (BPs) relative to the reference, T. latifolia (Figure 4; Table 4). When compared with basal Poales plastomes, the BP and reversal distances were 15 and 19 in J. effusus and J. inflexus, respectively. J. roemerianus has the largest BP (17) and reversal distances (20), and J. validus has the smallest BP (14) and reversal distances (17). Among the four Juncus, 27 LCBs were identified (Figure S4). While J. effusus and J. inflexus shared the same gene order, widespread rearrangements were detected in the other two species (J. roemerianus and J. validus).

FIGURE 4.

Whole plastome alignment of seven Poales species, including four newly assembly Juncus and Typha latifolia, Ananas comosus and Eriocaulon decemflorum. The local colinear blocks (LCBs) were identified by progressivemauve with the Typha plastome as the reference. The corresponding LCBs among seven plastomes are shaded and connected with a line of the same colour. LCBs that are flipped indicate inversions. Numbers on the upper x-axis are genome map coordinates in basepairs (bp).

TABLE 4.

Summary of breakpoint and reversal distances for plastomes of Juncus, Eriocaulon and basal Poales.

Species	Typha latifolia	Ananas comosus	Eriocaulon decemflorum	J. effusus	J. inflexus	J. roemerianus	J. validus
Typha latifolia	-
Ananas comosus	0/0	-
Eriocaulon decemflorum	0/0	0/0	-
J. effusus	15/19	15/19	15/19	-
J. inflexus	15/19	15/19	15/19	-	-
J. roemerianus	17/20	17/20	17/20	7/9	7/9	-
J. validus	14/17	14/17	14/17	8/10	8/10	11/13	-

4 I. DISCUSSION

4.1 |. ptGAUL application and suggestions for sequencing approach

The ptGAUL pipeline generated either one or three contig(s) for 11 publicly available data sets using PacBio or ONT data (Table 1). However, it failed to assemble the data from five species generating more than three contigs and predicted a much smaller plastome size, which is less than optimal (Table 1). In successful cases, the assemblies were highly similar to the published short-read assemblies at a basepair level with over 96%–99% nucleotide sequence identity. The lower percentage identity between Cenchrus americanus and Digitaria exilis and their reference assemblies may be due to different sequencing approaches between the Mariac et al. (2014) combined plastid capture method and Illumina sequencing and our long-read approach. For Arctostaphylos glauca, we used the read mapping method to verify that our assembly was more reliable than the result of Huang et al. (2022) as it showed more proportional coverage across the entire plastome (Figure S5). This difference could be caused by the selection of a distantly related reference genome (Camellia taliensis) from another family by Huang et al. (2022).

We found that the five failed samples had some features in common. For example, the sequencing approaches in the failed assemblies were different from the whole-genome sequencing method in those that were successful. In the Leucanthemum vulgare study, long-range PCR was implemented to generate amplicons that were then sequenced to produce a set of long reads that had an N50 value of ~8000 bp (Scheunert et al., 2020). In the remaining failed assemblies, plastid capture was utilized (Bethune et al., 2019). The PCR processes in both studies can greatly increase the bias among different plastome regions; for example, AT- and GC-rich regions do not amplify as efficiently as other regions (Quail et al., 2012). This could lead to underrepresentation/unevenness in read coverage of different regions resulting in many fragmented assemblies/contigs. Furthermore, the probes were designed based on the plastome data from distantly related species (Bethune et al., 2019), which may be unable to capture all plastome fragments for the target nonmodel species due to the divergence between the probe regions and the genome being captured. Additionally, the sequences obtained from PCR methods tend to be much shorter than the reads generated from sequencing total genomic DNA (see N50 values in Table 1). The low N50 values could also result from degraded DNA caused by poor storage, the use of silica-dried or herbarium material and/or DNA extraction method. For example, the Qiagen DNEasy Plant kits can generate high-quality DNA for short-read sequences because the column shreds the DNA to a maximum of ~25-kb fragments (Qiagen, Crawley, UK). CTAB, SDS or other methods that can produce much higher molecular weight (HMW) DNA are preferred for third-generation sequencing (Jung et al., 2019; Mayjonade et al., 2016), emphasizing the importance of sample preparation. Likewise, the assembly approaches, parameter combinations, read coverage, and the presence of nuclear genome and/or mitogenome contaminants could impact the completeness of an assembly (Jung et al., 2019; Scheunert et al., 2020).

Overall, considering the read length and read coverage, ptGAUL performs well for HMW samples using total genomic sequencing, resulting in high N50 values. Therefore, we recommend using HMW DNA extraction methods to isolate highly intact DNA, followed by long-read sequencing and subsequent assembly using ptGAUL.

4.2 |. Long-read data for plastome assembly

We found that short-read data alone may be insufficient to accurately assemble plastomes in species with many long-dispersed repeats. This phenomenon has been seen in several lineages, including Eleocharis (Lee et al., 2020) and Monsonia (Ruhlman et al., 2017). Plastome assembly using getorganelle for 11 Juncus species (12 accessions) failed using Illumina short reads only, including two samples in this study (Figure S6). Overall, the average assembly time for each Juncus is about 1 hr (Table S5). All Juncus plastome assemblies Indicated many fragmented contigs or assembly paths (Figure S6). This is because the numerous long-dispersed repeats present in the Juncus plastomes are longer than the kmer size/length of short reads. Based on our J. effusus plastome comparison, the final assembly length and the total number of genes based on short-read data are smaller than those assembled from long-read data (Table 2; Table S2), which might be caused by the random selection of one of the paths as the final assembly when using short-read data. Other studies demonstrated that a three-step approach can resolve this issue: (i) by comparing different contigs from short-read assemblers (e.g., spades, velvet), (ii) by manually checking through the contigs when contrasted with closely related species, and (iii) using long-range PCR to confirm assemblies (Lee et al., 2020; Ruhlman et al., 2017). This approach requires considerable time and effort.

Our ONT data resolved the plastome structure of four Juncus, confirming previous work (Lee et al., 2020; Ruhlman et al., 2017), showing that long-read data vastly improve the assembly of plastomes with many long repeats. Based on our study and that of Scheunert et al. (2020), ~50× mapping coverage of long-read data can result in an accurate plastome assembly. In our research, long reads of plastid origin represented 5%–6% of reads generated from the total genomic DNA of Juncus. Assuming 5% plastid DNA content from whole-genome HMW extractions, generating ~50× coverage of a 160,000-bp plastome requires only ~160 Mb reads per sample. Currently, one chip of ONT generates ~10 Gb of sequence data, enabling multiplexing up to 64 samples at consumable cost of roughly $1000 USD.

Although several assembly tools have been developed, several issues persist. Some pipelines/software are no longer maintained (i.e., sprai, celera assembler, organelle_pba). The assemblers of Syme et al. (2021), canu and hinge (Wang, Schalamun, et al., 2018) cannot generate a consistent plastome assembly result with one contig when using different data coverage, unicycler (Wick et al., 2017) is computationally intensive and does not produce well-resolved assemblies when dealing with complicated plastomes with many long repeats. The “fishing approach” associated with the assembly process used in shasta and hifiasm assemblers requires a considerable utilization of either time (hifiasm) or memory (shasta) (Table S6). These “fishing approaches” can find many relatively short plastid contigs, for example 16 contigs in Eucalyptus pauciflora with a maximum of 26,790 bp matching to the reference and 13 contigs in Arctostaphylos glauca with a maximum of 85,406 bp matching to the reference (Table S6). Compared to currently published pipelines for plastome assembly, ptGAUL can help generate accurate plastome assemblies in less than 10 min with ~16 Gb memory when the raw sequence data are less than 10 Gb (Table S6), making it highly convenient and typically significantly faster than other tools (Tables S5 and S6). Thus, ptGAUL should greatly facilitate plastome assembly of long-read data for phylogenetic and molecular evolutionary studies, especially in plastomes with a significant fraction of long repeat regions. Although ptGAUL can expedite plastome assembly, researchers still need to pay close attention to species with multiple plastome types, such as Eleocharis (Lee et al., 2020) and Monsonia (Ruhlman et al., 2017).

4.3 |. Juncus plastome organization

While many Poales genera contain plastomes with conserved gene order and content (Jones et al., 2007), including Typha (Guisinger et al., 2010), Ananas (Redwan et al., 2015) and Eriocaulon (Darshetkar et al., 2019), the data from the four Juncus examined here suggest that at least some species in this group contain plastome features atypical to most angiosperms. A limited number of complete plastome sequences are available from Juncus or other Juncaceae, but recently assemblies of two Eleocharis plastomes, in the sister family Cyperaceae (Hochbach et al., 2018), revealed accumulated duplications, gene losses, gene order rearrangements and intra-individual structural heteroplasmy (Lee et al., 2020). Similar phenomena contributed to size variation in the four Juncus plastomes, which ranged from 147,183 to 196,852 bp (Table 2). Many long repeats, including an unusually high number of dispersed repeats of 61–200 and 201–1000 bp, were present in the four Juncus with the greatest accumulation in J. effusus. Repeats >1000 bp were detected only in J. roemerianus (Table S3; Figure 3). Accumulation of large repeats may predispose plastome rearrangements in addition to contributing to overall size expansion (Tables 2 and 3, Figure 4), yet at present it is not clear if repeat accumulation predicated rearrangement or vice versa (Lee et al., 2021).

Similar repeat accumulation and plastome rearrangement occur in other taxonomic groups. In Trachelium caeruleum, gene-order changes, along with gene duplication, pseudogenization and loss were identified, as well as an abundance of variously sized repeats (Haberle et al., 2008). A relationship between repeat accumulation and rearrangement was suggested (Kim & Lee, 2005); studies of Pelargonium (Chumley et al., 2006), Jasminum, Mendora (Lee et al., 2007) and Trifolum (Cai et al., 2008) plastomes show early support for the theory. Many of the repeated sequences, when plotted onto the assembled plastid chromosomes, clustered at rearrangement endpoints. The relationship is also supported by findings in bacterial genomes where repeated sequences lead to gene order rearrangements (Rocha, 2003). Reconfiguration of the ancestral angiosperm plastome through repeat-mediated recombination has now been reported in several groups (Choi et al., 2019, 2020; Ruhlman et al., 2017; Schwarz et al., 2015; Sloan et al., 2014; Weng et al., 2014). We speculate that the recombinogenic potential of long repeats identified in the Juncus plastomes contributed to the diversification of gene order.

The observation of slight variations in IR length between Nicotiana species was explored in seminal work that focused on the IR/LSC boundary in closely related groups. This work ultimately inferred recombination-mediated gene conversion between poly-A tracts that gave rise to a > 12-kb expansion at the N. acuminata J_LB (IR_B/LSC boundary), placing the new J_LB near clpP and duplicating the 12-kb sequence now included In the IR (Goulding et al., 1996). Although the details of the mechanism have been clarified and refined over the years, repeat-mediated gene conversion appears to be at the heart of it (Maréchal & Brisson, 2010; Oldenburg & Bendich, 2015; Ruhlman & Jansen, 2021).

Plastomes that contain a large number of long repeats can experience extensive rearrangement of gene order and both loss and gain of plastome sequence, including genes, introns and noncoding sequences alike. Expansion and contraction at both LSC and SSC boundaries contributed to variation in Juncus plastome size. Photosynthetic seed plant plastomes and IRs range from ~120 to 170 kb and 20 to 30 kb, respectively, but most IR-containing angiosperms sequenced to date display highly similar gene arrangement and plastome size (~150 kb; IR, ~25 kb; Ruhlman & Jansen, 2021). Total plastome size in some groups is strongly influenced by IR expansion, yet in other lineages the association is loose at best. For example, a study of five Cyperus plastomes revealed the largest plastomes had the smaller IRs (i.e., C. esculentus; 186,255/37,438 kb), and the smaller plastomes contained the larger IRs (i.e., C. difformis; 167,974/38,427 kb) (Ren et al., 2021).

While total plastome size scaled with IR size (Table 2) and total repeat content (Table 3) in the four Juncus, the myriad events that altered each plastome relative to a shared ancestor with a more conserved structure remain elusive. The smallest of the four plastomes, in J. Validus, would appear to be a typical plastome based on the overall plastome and IR size (~147 and ~29 kb). However, the assembly and annotation show that it is not always size that matters. This plastome has probably experienced/is experiencing an ongoing series of IR boundary migrations resulting in a novel organization relative to the other taxa evaluated here. The near total elimination of the NDH gene suite, predominantly situated in the SSC in typical angiosperm plastomes, was unique to J. validus and suggests that IR boundary migration into the SSC played a role it their eventual loss. Although retained by the three other taxa, NDH sequences appear in alternate loci, and several have been duplicated by IR inclusion (Figure 2; Figure S3), suggesting migration at the SSC boundaries. Indeed, the gene order arrangement proximal to IR/LSC boundaries display little rearrangement across all four Juncus (Figures 2 and 3; Figure S3).

Complete ablation of the plastid-encoded NDH (NADH dehydrogenase-like) gene suite was reported for several unrelated seed plant lineages (Ruhlman et al., 2015). The NDH complex of plant and algal plastids participates in cyclical electron flow (CEF) (Shikanai et al., 1998) and comprises a multisubunit, plastid-localized complex that incorporates imported nuclear-encoded factors. The plastid genes encoding the NDH complex are highly conserved across Streptophyta (Hori et al., 2014), suggesting an essential function in photosynthesis (Ifuku et al., 2011). Using plastome sequencing and nuclear transcriptomics revealed that taxa lacking the plastid genes encoding constituents of NDH concomitantly lacked the relevant nuclear-encoded factors. Probing nuclear transcriptomes revealed that regardless of the state of the plastid NDH gene suite, genes encoding the alternate PGR5-dependent CEF pathway (Shikanai, 2014) were present in the nucleus of all examined taxa (Ruhlman et al., 2015). Loss of the NDH suite from the J. validus plastome is unique among examined Poales plastomes and suggest that an active PGR5-dependent pathway accounts for CEF in this species.

Apart from the loss of NDH genes, gene losses were shared by all four Juncus examined and included other genes that were lost from plastomes of diverse lineages (Ruhlman & Jansen, 2018). The plastid-localized acetyl-coenzyme A carboxylase (ACCase; prokaryotic) is another multisubunit protein complex that incorporates nuclear-encoded polypeptides and participates in fatty acid metabolism (Ohlrogge & Browse, 1995). The plastid accD encodes one subunit of the four-unit complex and was lost in numerous taxa, often those that experienced other gene loss and pseudogenization events (Ruhlman & Jansen, 2018). Because plastid ACCase activity was thought to have an essential function (Kode et al., 2005), accD loss in several groups suggested that it may be expressed from a functional transfer to the nucleus or substituted by a redundant, nuclear-encoded enzyme (Konishi et al., 1996). In Trifolium, which lacks plastid accD, a functional transfer to the nucleus was uncovered (Magee et al., 2010). Further investigation failed to detect any remnant of the accD sequence in the plastomes of T. repens or T. pratense, while mutated copies were identified in T. aureum and T. grandiflorum (Sabir et al., 2014). The 15-amino-acid (aa) C-terminal catalytic domain of the ACCD protein, which is minimally required for prokaryotic ACCase function (Lee et al., 2004), was identified in the mutated copies and may indicate functionality. Probing nuclear transcriptomes from T. repens and T. pratense revealed that, as in T. subterraneum (Magee et al., 2010), a putatively functional ACCD protein was being expressed from a fusion sequence that included the ACCD catalytic domain (~270 aa) fused to the plastid target peptide from nuclear-encoded, plastid-targeted LPD1 (493 aa). Probing transcriptomes of related legumes that contained intact plastid accD was able to detect high-identity copies of the ACCD core sequence, suggesting that incorporation at nuclear loci pre-dated the degradation of plastid accD (Sabir et al., 2014). Functional redundancy was demonstrated for prokaryotic ACCase (Babiychuk et al., 2011; Rousseau-Gueutin et al., 2013) and other gene products through transfer or substitution in different lineages (Ueda et al., 2007, 2008).

The fate of accD sequences and both the prokaryotic and the single-polypeptide eukaryotic ACCase in Poales has been a matter of investigation for some time. Morton and Clegg (1993) identified a recombination hotspot in seven Poaceae plastomes in the region between rbcL and psal (i.e., the locus containing accD sequences in non-Poaceae plastomes; Harris et al., 2013). Exploiting the fact that both the eukaryotic and prokaryotic ACCases contain biotinylated polypeptides, Konishi et al. (1996) were able to identify which form of the enzyme was active in plastids from across the diversity of the green plant lineage, including two nonphotosynthetic representatives. Differentiating the two enzymes by molecular weight revealed that only one group examined did not contain the 35-kDa peptide that represented the prokaryotic holoenzyme: Poaceae. Closer examination of Poales using PCR product sequencing combined with Southern blots probed with plastid accD from Commelinaceae taxa demonstrated pseudogenization or deletion in representatives of three families, Restionaceae, Joinvilleaceae and Poaceae (Harris et al., 2013). Extending the loss of accD to include the Cyperaceae (Cyperus; Ren et al., 2021; Elocharis, Lee et al., 2020) and now Juncaceae suggests either extreme lability of the coding sequence in Poales or that this gene was transferred or substituted by a nuclear-encoded activity in a common ancestor. Differential nuclear retention, expression and transport of the gene product back to plastids among the various lineages could result in a relaxed selection of the plastid gene (Park et al., 2017; Ueda et al., 2007).

The opportunity to sample deeply across and within lineages reveals that the unusual variation identified by early Southern blots and more recent plastome sequencing suggests that these “unusual” structural changes are not unique. The suite of plastid genes susceptible to pseudogenization or loss appears consistent across photosynthetic seed plants. Understanding phylogeny, inherent to evolutionary studies, requires deep sampling, high-quality sequencing, assembly and alignment to infer relationships. As next-generation sequencing and single-molecule long-read sequencing platforms expand and become more accessible, reads will be generated for many diverse taxa. Where long sequence repeats exceed insert sizes in next-generation systems, long reads will be able to “bridge the gap.” The ability to translate raw sequence reads into usable data for evolutionary and functional inquiries depends on advanced computational tools that provide fast, flexible platforms without vast computational demand. Facilitating this effort, the ptGAUL pipeline provides a fast and easy tool for assembling plastomes from long-read data, which will enable the characterization of repeat-rich, highly rearranged plastomes.

DATA AVAILABILITY STATEMENT

Demultiplexed sequence data of short- and long-read data are available for download from the NCBI Sequence Read Archive (SRA) (BioProject PRJNA865266; SRR21976089, SRR21976090, SRR21976091 and SRR21976092). The accession numbers of J. roemerianus and J. validus are OP235509 and OP235510, respectively. Information related to ptGAUL can be retrieved from GitHub (https://github.com/Bean061/ptgaul).