Similar to Illumina sequencing, SMRT is based on the principle of sequencing-by-synthesis, acquiring sequence information during the amplification process of nucleic acid molecules. Before sequencing, both ends of the targeted double-stranded DNA (dsDNA) molecule are ligated with hairpin adapters to form dumbbell-shaped templates (i.e., SMRTbells). These adapters allow DNA polymerases and primers to bind with the SMRTbells ( Figure 1A ). After binding, the SMRTbells are sequenced on a SMRT cell chip. There are thousands of Zero-model waveguides (ZMWs) lined up on each SMRT cell. The ZMWs limit the observation volume to avoid the influence of the fluorescence of uncombined deoxyribonucleoside triphosphates (dNTPs), which allows the detection of a single dNTP. Once the SMRTbells are loaded onto the SMRT cell, a proportion of them fall into the ZMWs. Then, the SMRTbells are fixed on the bottom of ZMWs through the interaction between the biotin on the polymerase and the streptavidin on the glass plate of the ZMWs (Eid et al., 2009). The DNA polymerases catalyze the continuous incorporation of dNTPs labeled by different fluorophores into complementary strands (Flusberg et al., 2010). When the polymerases capture the labeled dNTPs, they emit distinct fluorescence pulses under excitation light ( Figure 1B ). Four classic dNTP types can be recognized from their featured fluorescence signature (Eid et al., 2009).

The methylated bases in the DNA template change the incorporation kinetics of polymerases, which enables SMRT to directly detect methylated bases without chemical modifications (Flusberg et al., 2010). Moreover, because of the circular structure of the SMRTbell and the replacement sequencing ability of the DNA polymerase, inserted DNA templates can be repeatedly sequenced, yielding many copies of both template and complementary strands. Aligning these copies greatly improves the sequencing accuracy. This sequencing strategy is also known as circular consensus sequencing (CCS) (Travers et al., 2010).

Sample and library preparation: In the genome sequencing of medicinal plants, fresh leaves are often used as samples for DNA extraction. High molecular weight genomic DNA (HMW gDNA) should be extracted from the samples because the amount and length distribution of extracted gDNA are important for subsequent library construction. Usually, high fidelity (HiFi) library construction for whole genome sequencing of plants requires at least 1 μg of DNA input per 1 Gb of genome length. DNA molecules ≥10 kb should account for more than 90%, and molecules ≥30 kb should account for more than 50%. Additionally, when this amount of DNA cannot be extracted from the samples, alternative workflows are available (low DNA input workflow and ultra-low DNA input workflow). After quality control of HMW gDNA molecules, they need to be sheared to suitable sizes by the Megaruptor system, followed by cleanup with SMRTbell cleanup beads. Then, the sheared gDNA undergoes end-repair, A tailing, adapters ligation, and nuclease treatment in a thermocycler, followed by size-selection using AMPure PB Beads or cleanup with SMRTbell cleanup beads to form the SMRTbell library (usually 15–18 kb). Finally, DNA polymerases and primers are annealed to the SMRTbell library using Binding kits (e.g., Sequel II binding kits 3.2), and then the final sequencing complexes are constructed ¹ .

Sequencing: After dilution, the SMRTbell library is loaded onto PacBio sequencers with one or more SMRT cells, each of which can yield HiFi reads up to 4 Gb in one run. The runtime is flexible depending on the amount of data needed by the assembly ² .

Primary data analysis: Base-calling and primary filtering analysis are performed on the sequencer. SMRT Link, PacBio DevNet, and other software tools are available to process the raw SMRT data ³ . The complete pipeline of SMRT is shown in Figure 1C .

There are six long-read sequencing instruments based on SMRT sequencing technology: PacBio RS, RSII, sequel, sequel II, sequel IIe, and the newly released Revio. Among them, PacBio RS is the first sequencer commercialized by PacBio. As an early-released instrument, PacBio RS had a relatively low throughput, short average read lengths (~1.5 kb) and a high error rate (~13%) (Quail et al., 2012). With technological advances, the throughput of PacBio sequencers has increased by several hundred folds while the average read lengths underwent a 10-fold increase. The accuracy has also been improved to more than 99% due to the extensive use of the CCS strategy ( Table 1 ). These instruments have been used in many genomic and transcriptomic studies of medicinal plants, among which PacBio sequel is used frequently, probably because of its high throughput and relatively low costs.

ONT originated from a brand-new idea of threading a single-stranded nucleic acid molecule through a nanopore protein (Clarke et al., 2009) ( Figure 2A ). The sequencing process of ONT takes place in a container filled with an electrolyte solution. A lipid double-layer membrane embedded with a nanopore protein is placed in the container. Under an applied voltage, a stable current is formed in the nanopore due to the flow of ions. Therefore, when a nucleic acid molecule passes through, the nanopore is partially blocked, and the stable current is interfered with. Because the structures of nucleotides are different, they cause distinct interferences with the current ( Figure 2B ). For this reason, nucleotides of sequenced nucleic acid molecules can be distinguished from their distinctive current variations; in this way, sequence information can be decoded (Clarke et al., 2009; Jain et al., 2016).

Sample and library preparation: Library preparation kits for whole genome sequencing, targeted DNA sequencing, and RNA sequencing are all provided by ONT ⁴ . Here we take genome sequencing on MinION using Ligation Sequencing Kit V14 as an example. The process of library preparation using this sequencing kit takes about 60 min. Fresh leaves are frequently used for DNA extraction in whole genome sequencing by ONT. First, HMW gDNA can be extracted from plant tissues using the NEB Monarch HMW DNA Extraction Kit or other compatible extraction kits. Then, researchers can choose whether to conduct fragmentation or size selection in the pipeline. If not, the yielded read length will equal the input fragment length. Second, the extracted HMW gDNA undergoes DNA repair and end-preparation (end-prep) using NEBNext FFPE DNA Repair Mix and NEBNext Ultra II End Repair/dA-tailing Module reagents. Third, sequencing adapters are ligated to the repaired ends of DNA molecules using a ligation sequencing kit and some other reagents, followed by cleanup, after which the sequencing library is prepared. Finally, about 1 μl of the DNA library is quantified using a Qubit fluorometer, and then the DNA library is produced as 12 µl at 10–20 fmol ⁵ . In addition, an automated device, VolTRAX, released by ONT enables hands-free and standard sequencing library construction ⁶ .

Sequencing: Different library preparation kits are compatible with different versions of flow cells, which should be confirmed before sequencing. After priming the flow cell, 10–20 fmol of the final DNA library is advised to be loaded onto the flow cell. The time for one run is up to 72 h.

Primary data analysis: Data acquisition is usually performed by MinKNOW, and base-calling can be conducted by MinKNOW, GUPPY, and many other algorithms available on GitHub ⁷ . The complete pipeline of ONT is shown in Figure 2C .

Flongle, MinION, GridION, and PromethION are the main sequencing instruments of ONT. Among them, GridION and PromethION are bench-top instruments with high throughput and, therefore, usually used for large-scale sequencing projects ⁸ (Jain et al., 2016) such as whole genome sequencing of humans, mammal animals, and plants. MinION is a portable instrument weighing only 90 g that can be used for small sequencing projects, such as microorganism genomes and rapid sequencing outside the laboratory (and even in space) (Jain et al., 2016; McIntyre et al., 2016). Flongle is a single-use product generating 1–2 Gb of data, which is suitable for even smaller projects, such as plasmid and viral sequencings ⁹ .

Decoding nuclear genomes of medicinal plants with high heterozygosity and a high proportion of repeat sequences using NGS techniques often leads to fragmented genome assemblies. With much longer read lengths, TGS enables researchers to uncover the genomic regions missed by NGS techniques. To date, more than 100 nuclear genomes of medicinal plants have been sequenced using TGS, most of which are assembled to a chromosome level combined with high-throughput chromosome conformation capture (Hi-C) mapping technology (Cheng et al., 2021c) ( Supplementary Table 1 ). Among them, a recent study of Gardenia jasminoides ( Figure 3A ) demonstrated the classic processes of applying TGS to nuclear genomes research of medicinal plants, from library construction to data analysis. Therefore, it is taken as an example here (Xu et al., 2020b).

In the reference study, ONT, Illumina, and Hi-C are combined to gain a chromosome-level assembly of the G. jasminoides genome, while RNA sequencing (RNA-seq) is used to evaluate assembly quality, predict protein-coding genes, and calculate the expression level of genes.

Genome size estimation: Before ONT sequencing, flow cytometry (Pfosser et al., 1995) and k-mer distribution analysis (Manekar and Sathe, 2018) were used to estimate genome size and heterozygosity. Based on the two methods, the genome of Gar. jasminoides was predicted to have a total size of 550.6 ± 9 Mb and a high heterozygosity of 2.2%, implicating that it is challenging to assemble this genome.

Library preparation: Libraries for ONT, Illumina, RNA-seq, and Hi-C were constructed. In the reference study, the fresh leaves of Gar. jasminoides were pooled for DNA extraction of ONT and Illumina sequencing. Seven organs, including fruits at different maturity stages, of Gar. jasminoides were collected for RNA-seq and the measurement of crocin content. The Hi-C library was constructed with fresh tissue from Gar. jasminoides. For ONT library construction, HMW gDNA was extracted from the pooled leaves and then fragmented, size selected, and purified to get large fragments, after which the large fragments underwent end-prep, adapter ligation, tether attachment, and then an ONT library was constructed.

Sequencing and assembly: The complete ONT library of Gar. jasminoides was sequenced on GridION X5, and the raw data was base-called using Guppy (v1.8.5), generating 2.67 Gb reads with an N50 of 21.6 kb. For assembly, the authors developed a satisfactory package (Canu-SMARTdenovo-3×Pilon) by testing various de novo assembly pipelines. Specifically, using this package, base-called ONT reads were corrected and trimmed by Canu and then assembled with SMARTdenovo, followed by Illumina short reads to polish the Canu-SMARTdenovo contigs with Pilon three times. The final scaffolds were assembled with the polished contigs, and ONT reads were corrected using Canu; heterozygous sequences were eliminated by Purge Haplotigs. Finally, the researchers acquired a 534.1 Mb assembly with a contig N50 of 1.0 Mb ( Figure 3B ). The quality of the assembly was evaluated using Benchmarking Universal Single-Copy Orthologs (BUSCO) and mapped with Illumina short reads from the DNA and RNA libraries of Gar. jasminoides, respectively, which found 95.0% complete BUSCOs. Hi-C technology was used to further improve the quality of the assembly. As a result, 99.5% of sequences from the assembly were scaffolded into 11 pseudo-chromosomes using the Lachesis package. At this stage, the final chromosome-level genome of Gar. jasminoides was successfully constructed. It was 535 Mb in size, with a scaffold N50 of 44 Mb ( Figure 3C ).

Further data analysis: The first step of data analysis for medicinal plant genomes is usually genome annotation, followed by comparative genomic analysis and expression analysis. In the reference study, at first, the chromosome-level genome of Gar. jasminoides was used for genome annotation, including the annotation of repeat elements, the prediction and functional annotation of protein-coding genes, and the annotation of non-coding RNA. This genome was then used for comparative genomic analyses, including synteny analysis between Gar. jasminoides and Coffea canephora, phylogenetic analysis between Gar. jasminoides and ten additional angiosperms, followed by mapping transcriptome reads obtained from RNA-seq of seven organs of Gar. jasminoides to the annotated genes to calculate the relative expression level of genes (fragments per kilobase of exon per million reads mapped, FPKM). Furthermore, genome-wide analysis was conducted after mapping, in which genes from three families related to crocin biosynthesis were identified, including 14 carotenoid cleavage dioxygenases (CCDs) genes, 18 aldehyde dehydrogenases (ALDHs)-like genes, and 237 UDP-glucosyltransferases (UGTs) genes. The Gar. jasminoides crocin biosynthetic was elucidated after expressing 14 candidate crocin biosynthetic genes in Escherichia coli to test their enzymatic activity. In addition, the above-mentioned comparative analysis revealed the evolution of crocin and caffeine biosynthesis genes in Rubiaceae ( Figure 4 ).

In addition to the software used in the reference study, many new tools suitable for TGS techniques have recently been developed. Ratatosk (Holley et al., 2021) was developed for hybrid error correction while CONSENT (Morisse et al., 2021) was designed for self-correction of TGS. For genome assembly, new assemblers are available, such as hifiasm (Cheng et al., 2021a) and Nextdenovo (https://github.com/Nextomics/NextDenovo). And for scaffolding, RegScaf (Li and Li, 2022) was developed to resolve large genomes and repeat regions, while YaHS (Zhou et al., 2023) is suitable for chromosome-scale scaffold construction using Hi-C data. Regarding annotation, PhyloCSF++ (Pockrandt et al., 2022) is a newly updated tool for differentiating protein-coding and non-coding regions, and TransposonUltimate (Riehl et al., 2022) is a newly developed tool for transposon classification, annotation, and detection. Advanced bioinformatics software have also greatly facilitated the genome research of medicinal plants.

High-quality reference genome assemblies have provided valuable genetic information for investigating the biosynthesis of secondary metabolites, such as triptolide, morphinan, and icaritin. By integrating the genome, transcriptome, and metabolome of Tripterygium wilfordii, a cytochrome P450 (CYP728B70) in T. wilfordii was identified to catalyze the oxidation of a methyl to the acid moiety of dehydroabietic acid in triptolide biosynthesis, providing clues for elucidating the biosynthetic pathway of triptolide (Tu et al., 2020). (S)- to (R)-reticuline (STORR) gene fusion is key for morphinan biosynthesis in Papaver somniferum. CYP450 and oxidoreductase genes that combined to form the gene fusion were also identified by paralog analysis using a chromosome-level Papaver somniferum genome assembly (Guo et al., 2018). In addition, an important flavonoid prenyltransferase (EpPT8) in Epimedium pubescens was proven to be involved in the biosynthesis of icaritin and its derivatives by whole genome search using a chromosome-level genome assembly of E. pubescens (Shen et al., 2022a). Notably, the biosynthetic pathways of iridoids in Rehmannia glutinosa (Ma et al., 2021) and crocin in Gardenia jasminoides (Xu et al., 2020b) were successfully elucidated.

The availability of high-quality reference genomes also facilitates the research of the evolutionary history of important gene clusters, biosynthetic pathways, and species in different families. For example, the convergent evolution of CYP82D and CYP706X members in Lamiaceae and Asteraceae (Gao et al., 2022) and the divergent evolution of caffeine and crocin biosynthetic pathways were revealed based on TGS genome assemblies. Information on the evolution of species in Euphorbiaceae (Wang et al., 2021b), Asteraceae (Shen et al., 2018), and Magnoliid (Shang et al., 2020) has been provided as well in recent studies. These are valuable resources for subsequent functional genomics, molecule-assisted breeding, and synthetic biology research (Xin et al., 2019).

Although many chromosome-level assemblies of medicinal plant genomes have been completed, gaps and highly repetitive regions remain to be resolved, such as the centromere and telomere regions. Recently, by combining ONT ultra-long reads, PacBio HiFi reads, and Hi-C technology, researchers successfully obtained truly gapless telomere to telomere (T2T) reference genome assemblies of Hordeum vulgare (Navrátilová et al., 2022), Arabidopsis thaliana (Wang et al., 2022a), Citrullus lanatus (Deng et al., 2022), and several other plants. Compared to chromosome-level assemblies, T2T assemblies are more complete, can be used to discover almost all genomic variations, and enable research into centromere and telomere regions. However, T2T assembly of medicinal plants is still rarely reported.

Organelle genomes are also important genetic resources for medicinal plant utilization. To date, the chloroplast genomes (cp-genomes) of more than 20 important medicinal plants ( Table 2 ) have been sequenced using TGS. Further, the complete mitochondrial genomes of several important medicinal plants have been sequenced by TGS ( Table 3 ). Most of the obtained cp-genomes were determined as circular molecules with quadripartite structures consisting of a pair of inverted repeat regions (IRs), a large single-copy region (LSC), and a small single-copy region (SSC), while a few of which were also identified as tripartite or bipartite structures. Here we take the research of Salvia miltiorrhiza (Chen et al., 2014) as an example to introduce the achievements of this research and the classic workflow when applying TGS to the sequencing of medicinal plant cp-genomes.

Sample and library preparation: The sample and library preparation of cp-genomes is similar to that of nuclear genomes. In the selected study, fresh leaves were prepared from S. miltiorrhiza for gDNA isolation. The gDNA was extracted using a plant genomic DNA kit (Tiangen, China). Libraries consisting of inserted fragments 1 kb and 10 kb in size were prepared and used for subsequent SMRT sequencing.

Sequencing and assembly: The SMRT sequencing of gDNA was conducted under the guidance of the manufacturer provided by PacBio and the raw sequences were preprocessed using the SMRT Analysis workflow. Regarding assembly, first, more than 200 cp-genomes were downloaded and blasted against the cp-genome of S. miltiorrhiza. Similar sequences in the cp-genome of S. miltiorrhiza were isolated and used as the basis of genome assembly. Second, the cp-genome of Sesamum indicum was selected for guiding the order of contigs because of its highest similarity with the S. miltiorrhiza cp-genome. Third, to fill the gaps in the assembly, isolated sequences and contigs were used to repeatedly search against SMRT reads of S. miltiorrhiza gDNA. Then, an initial assembly was obtained by extending the contigs, adding new reads, and conducting reassembly. Finally, the regions of junction between IRs and LSC (or SSC) were amplified and sequenced by Sanger sequencing and the final assembly was obtained by integrating the Sanger sequences into the initial assembly using Seqman (DNASTAR, WI). Strand-specific RNA sequencing was also conducted to determine the expression level of genes in the cp-genome of S. miltiorrhiza.

Further data analysis: The RNA-seq reads were mapped to the final assembly of S. miltiorrhiza cp-genome using Tophat to identify polycistrons and non-coding RNA (ncRNA) and to determine the content of protein-coding transcripts (cRNA) and ncRNA. Strand-specific real-time quantitative PCR (ss-qPCR) was also conducted to validate the results of RNA-seq. DNA modifications were predicted using the SMRT Portal software (v1.3.2). As a result, the authors identified 19 polycistronic transcripts containing 71 genes, which consisted of 58 protein-coding genes, four rRNA, and nine tRNA. Furthermore, 136 ncRNA transcripts were identified and classified into two categories, intergenic ncRNA and antisense ncRNA (asRNA). Using SMRT Portal 1.3.2, two DNA modification motifs and 2687 DNA modification sites were predicted. Interactions between asRNA and cRNA, DNA modification and gene expression were also analyzed. The results showed that the expression level of protein-coding genes was positively associated with that of asRNA, and the DNA modification was correlated with higher expression of ncRNA.

In addition to the software mentioned above, many kinds of newly developed software, such as GetOrganelle (Jin et al., 2020), Fast-Plast (https://github.com/mrmckain/Fast-Plast), CPGAVAS2 (Shi et al., 2019), and CPGview (Liu et al., 2023), are also available for de novo assembly, annotation, analysis and visualization of chloroplast genomes.

The mitochondrial genomes of medicinal plants are more complex than cp-genomes, which usually contain multiple conformations (isoforms) instead of circular molecules (Kozik et al., 2019; Wang et al., 2021c). Previously, many mitochondrial genome assemblies failed to obtain all the isoforms of the mitochondrial genome of medicinal plants because of the limitations of the methods (Kozik et al., 2019). With the help of TGS techniques, researchers successfully captured various conformations of the mitochondrial genomes of Coriandrum sativum (Wang et al., 2021c), Scutellaria tsinyunensis (Li et al., 2021b), and several other valuable medicinal plants, providing more complete and precise references for the further utilization of mitochondrial genome sequences.