Chloroplast Phylogenomics: Decoding the Plastome for Evolutionary and Synthetic Biology

Meta Intent: A high-resolution technical roadmap for plastid genome assembly, boundary interpretation, and functional analysis in plant research workflows. This article is written for research contexts.

Chloroplast genomes still look deceptively simple in many online resources. The standard summary is familiar: a compact genome, a quadripartite structure, and a conserved gene set linked to photosynthesis and housekeeping. That summary is not wrong. It is just too shallow for modern plastome work. In real projects, the hard questions are different. Where exactly does the IR end? Is the assembly complete, or only apparently circular? Are low-frequency plastid variants real, technical noise, or nuclear carryover? And if a locus is chosen for plastid engineering, will the local architecture remain interpretable during selection and sorting? Recent comparative and structural work makes one point clear: plastomes are conserved enough to compare, but dynamic enough to require an architectural view rather than a fact-sheet view.

That shift matters for both phylogenomics and engineering. For phylogenomics, IR boundary shifts, lineage-specific repeats, and rearrangements can change how genome size, gene dosage, and junction markers are interpreted across taxa. For engineering, the same structural logic shapes target-locus choice, homologous recombination design, and the path from mixed plastid populations toward genetically consistent research lines. The most useful plastome resource in 2026 is not a glossary. It is a decision guide that connects structure, assembly, variation, and downstream experimental design.

CD Genomics supports chloroplast DNA sequencing workflows that fit naturally into projects where plastome recovery is the starting point for comparative analysis, junction validation, or downstream engineering decisions. When plastid signals also need to be interpreted against a broader genomic background, that work can be paired with whole-genome de novo sequencing support or a more focused targeted region sequencing strategy, depending on whether the main challenge is architecture discovery or locus-specific confirmation.

Structural Evolution of the Plastome

The canonical plastome is still the right place to start. In most land plants, the chloroplast genome is organized into a large single-copy region (LSC), a small single-copy region (SSC), and two inverted repeats (IRa and IRb). Across many angiosperms, total size often falls in the rough range of 120 to 160 kb, with a gene complement conserved enough to support comparative genomics, but not so fixed that boundary movement and gene loss can be ignored. That is why a chloroplast genome can look stable at first glance while still carrying lineage-specific structural signals that matter for assembly and interpretation.

The deeper reason this architecture persists is not that plastomes are frozen. The plastome is under layered constraints. Core photosynthetic and gene-expression functions remain strongly conserved. The inverted repeat is not just a separator. It influences local sequence behavior, dosage context, and genome maintenance. Recent experimental work in tobacco showed that removing one copy of the large IR affected gene dosage and increased plastid genome copy number, reinforcing the idea that the repeat is functional architecture rather than decorative symmetry.

That duality explains why old fact-sheet summaries age badly. They treat the plastome as a static circular object. Modern phylogenomics treats it as a structure with preferred organization, repeat-governed behavior, and lineage-specific routes to change. In some clades, repeats accumulate and correlate with inversions or rearrangement hotspots. In others, IR boundary shifts are more informative than total genome size. In still others, gene loss or transfer to the nucleus can occur alongside large boundary movements, making a simple size comparison misleading.

Figure 1. Whole-plastome architecture as a 3D cutaway map. Show the quadripartite plastome inside a semi-transparent chloroplast and make the figure answer one question: where do LSC, SSC, and the two IRs sit as spatial architecture rather than as a flat textbook circle?

A practical reading rule follows from this. When a plastome looks “conserved,” ask conserved at what scale. At the whole-genome scale, quadripartite structure may be intact. At the border scale, one junction may have shifted enough to alter duplicated gene content. At the repeat scale, small lineage-specific repeats may already be setting up future inversions or alternative isoforms. And at the population scale, a clean consensus sequence may be masking low-frequency structural diversity. Each scale supports a different biological conclusion. Each scale also demands different sequencing and validation choices.

The Mechanics of IR Expansion and Contraction

IR expansion and contraction are often described as bookkeeping events. That framing is too weak. In plastome research, IR border movement is one of the most important mechanisms for explaining why closely related taxa can differ in genome size, duplicated gene content, and junction architecture while still retaining the same broad quadripartite plan. The border is not just a boundary. It is a moving interface where recombination, repair, and junction-specific history leave visible genomic signatures.

A border shift can pull a gene into the repeat or move it back into a single-copy region. That changes more than annotation aesthetics. A gene duplicated within the IR may experience altered dosage logic, different local mutation dynamics, and different assembly behavior from a homolog sitting near, but outside, the repeat. This is one reason IR-associated genes are so useful in comparative plastome studies. They help distinguish true evolutionary movement from superficial length variation in intergenic sequence. The 2024 IR-deletion study in tobacco pushed this logic further by connecting repeat structure directly to dosage behavior and genome copy-number effects.

For phylogenomics, this matters in a concrete way. Suppose two related taxa differ in total plastome size by only a few kilobases. A shallow reading might call them near-identical. A junction-aware reading asks a better question: did the size shift arise from repeat expansion, from localized indels in noncoding regions, or from sequence capture at the IR/SC border? Those scenarios do not carry the same evolutionary meaning.

Figure 2. IR expansion and contraction at plastome borders. Show canonical, expanded, and contracted border states in one precision layout, and make the figure answer this question: how do IR-LSC or IR-SSC boundary shifts change duplicated content and comparative interpretation?

The same border logic also explains why plastome maps built from short reads alone can look more certain than they really are. A repeat boundary is easy to draw in a polished circular map. It is much harder to resolve in an assembly graph if reads do not span the repeated segment cleanly. This is the first place where a plant biologist needs to think like an assembler. A chloroplast genome is not “finished” because software produced a circle. It is finished when the repeated structure is consistent with the graph, the read support, and the junction validation strategy. GetOrganelle became influential for precisely this reason: it recruits organelle-associated reads, disentangles assembly graphs, and can report alternative circular configurations rather than hiding ambiguity behind one attractive output.

The question becomes even more interesting when non-canonical plastome forms are considered. For years, many resources presented chloroplast DNA as a single stable circular molecule. More recent structural analyses have pushed back against that simplification. Alternative isoforms, substoichiometric states, and branched recombination intermediates complicate the old picture. This does not mean every project requires a full structural-isoform atlas. It does mean the “one perfect circle” model is often a reporting convenience rather than a full biological description.

That is why long-read data has moved from helpful to strategically important. Short-read pipelines can still recover many plastomes well, especially from clean skimming data and standard architectures. But IR boundaries and rearranged repeat landscapes are exactly where longer reads deliver disproportionate value. HiFi-scale reads can bridge junctions, clarify graph paths, and reduce the number of equally plausible repeat solutions. Hybrid strategies also help when a project needs both base-level accuracy and confident repeat traversal. Benchmarking work has shown that both software choice and sequencing coverage can materially change plastome outcomes.

This becomes most important when the biological question depends on exact border placement, repeat orientation, or rearrangement detection. In that setting, a short-read-only design can become limiting. Long-read workflows are especially useful when the target plastome is expected to carry expanded repeats, lineage-specific rearrangements, or ambiguous junctions. In practice, researchers often use Nanopore ultra-long sequencing when the goal is to span difficult repeat structures directly, then combine that structural view with plastome-focused polishing and validation.

Figure 3. Circular, linear, and branched plastome isoforms. Show three suspended DNA conformations without a chloroplast body and make the figure answer this question: why is the plastome often better understood as a population of structural forms rather than one perfect circle?

Figure 4. Long-read versus short-read resolution at IR boundaries. Show graph ambiguity, spanning-read support, and boundary-confidence outcomes under short-read, long-read, and hybrid designs, and make the figure answer this question: why do repeat boundaries remain the first place where weak assemblies fail?

A useful rule for manuscript interpretation is this: treat every IR boundary as both a biological feature and an assembly test. Biologically, it may mark dosage shifts, repeat-mediated constraint, or lineage divergence. Computationally, it is where ambiguous mapping, collapsed repeats, and overconfident circularization often appear first. Good plastome papers now make both layers visible. Great plastome papers make them inseparable.

Plastid Heteroplasmy and Intra-individual Variation

Heteroplasmy is one of the most underdeveloped topics in many chloroplast overviews. The common simplification is that a plant carries a chloroplast genome, and that genome can be represented by one consensus sequence. The more accurate view is that plastid variation exists across levels of organization. A variant can be fixed within a single organelle, mixed across organelles in one cell, uneven across cells in one tissue, and still visible at the scale of the whole plant. That layered view is far more useful for interpreting plastome data and for designing follow-up validation.

The key process here is often called sorting out. A mixed plastid population does not stay mixed forever at every scale. During growth, development, and transmission, plastid genotypes can drift toward fixation or remain mosaic depending on the number of transmitted organelles, the effective bottleneck, the developmental context, and the selective effect of the variant. This is why heteroplasmy should be treated as a dynamic state, not a label. A plastome that looks homogeneous in bulk leaf DNA may still have reached that state through a history of strong sorting, and a plastome that looks mixed may reflect ongoing partitioning rather than sequencing noise.

That distinction matters for both natural variation and plastid engineering. In natural populations, low-frequency plastid variants can inform mutation, repair, transmission, and tissue-level mosaicism. In transplastomic workflows, the same logic becomes operational. Early transformants are often mixed. The central practical question is whether selection and regeneration will push the population toward homoplasmy or leave a persistent mosaic. Recent plastid engineering reviews stress this transition explicitly, because interpretable expression readouts under research conditions depend on more than successful insertion. They depend on population purification within the organelle compartment.

Figure 5. Plastid heteroplasmy mosaic across a leaf tissue section. Show chloroplast-to-chloroplast and cell-to-cell variation inside one tissue slice, and make the figure answer this question: how can one leaf still contain multiple plastome states even when a bulk sample suggests one consensus genome?

The bottleneck concept helps make this intuitive. When only a limited subset of plastids, plastid genomes, or plastid-bearing lineages contributes to the next developmental stage or generation, random sampling can sharply change allele frequencies. That narrowing is then followed by expansion. The result can be rapid fixation, retention of mosaic states, or divergent outcomes across tissues or descendants.

For researchers, the main lesson is methodological. Heteroplasmy is easiest to erase on paper and hardest to erase in biology. A consensus FASTA file suppresses low-frequency complexity by design. Variant calling thresholds suppress it again. Bulk extraction mixes cells. Repeats blur mapping. And if plastid-like reads also exist as nuclear insertions, apparent minor alleles can become even harder to classify. So when a project cares about low-frequency plastid variation, the discussion must move beyond generic “deep sequencing.” It must include read origin, local repetitiveness, duplicate handling, depth distribution, strand support, and orthogonal confirmation strategy.

In practice, the right sequencing design depends on what needs to be resolved. A broad whole genome sequencing dataset may be enough for plastome recovery in a clean sample. But low-frequency heteroplasmy questions often need more local support and cleaner interpretability. In those cases, researchers may strengthen boundary or hotspot confirmation with amplicon sequencing services or narrow the analytical field with targeted region sequencing when a specific plastid locus is already under suspicion.

There is also a biological trap worth stating clearly. A low-frequency plastid variant is not automatically evidence of active plastome diversification. It may reflect true intra-individual mosaicism. It may reflect tissue admixture. It may reflect mapping ambiguity across the IR. Or it may come from plastid-derived fragments now resident in the nucleus. Strong projects do not collapse those possibilities too early. They rank them, test them, and only then interpret the variant biologically.

Figure 6. Bottleneck effect and sorting-out across transmission. Show narrowing, stochastic sampling, re-expansion, and divergent fixation outcomes, and make the figure answer this question: how does a mixed plastid population move toward fixation in one lineage but remain mosaic in another?

Detection Sensitivity: How Low Can Plastid Variant Calling Go?

The next question is not whether heteroplasmy exists. It is how confidently it can be detected when variant allele frequency drops below 0.5%. That range is uncomfortable for a reason. It is low enough that true plastid variants may still matter biologically, but also low enough that mapping ambiguity, duplicate inflation, local sequence context, and read-origin uncertainty can distort the signal. In chloroplast projects, that problem is amplified by two recurring features: repeated regions and plastid-like sequences outside the plastome itself. A minor allele near an IR boundary is never interpreted the same way as a minor allele in a unique, well-covered single-copy locus.

In 2026, the best practice is not to advertise a universal sensitivity threshold. The better practice is to define a validated operating range. A stronger paper makes that range explicit through five filters:

1. Locus uniqueness: Can the site be mapped without obvious IR or Nupts ambiguity?
2. Effective depth after duplicate policy: Is the usable depth still sufficient after duplicate handling and quality trimming?
3. Strand concordance: Do both strands support the same minor signal?
4. Replicate recovery: Does the same site recur across technical or biological replicates?
5. Orthogonal confirmation: Can the signal be confirmed by targeted follow-up rather than bulk mapping alone?

A raw claim that “deep sequencing detected 0.2% plastid heteroplasmy” means very little unless the paper also explains whether the site sits in a unique region, whether read families were collapsed appropriately, and whether plastid-derived nuclear fragments were excluded from the analysis. Short reads still provide excellent depth and are often the most practical route for screening low-frequency variants. But longer reads add context when a candidate site sits inside structurally confusing territory. Short reads answer “how many supporting molecules do I see?” Longer reads help answer “where do those molecules truly belong?” In difficult cases, those are not competing questions. They are sequential ones.

A useful rule is to treat sub-0.5% plastid calls as hypotheses under validation, not final biological facts. If the site is reproducible across replicates, supported on both strands, stable under stricter mapping, and independent of obvious Nupts interference, confidence increases. If the signal disappears when repetitive loci are masked or when only uniquely anchored reads are retained, the interpretation should narrow accordingly.

Advanced Functionality: The Plastid as a Metabolic Bioreactor

The chloroplast is not only a photosynthetic organelle. It is also a dense biochemical workspace with growing importance in synthetic biology. That is why plastid engineering has regained momentum. The plastome combines compact organization, prokaryote-like gene-expression logic, high genome copy number, and the ability to support coordinated expression of engineered functions. Those features make the chloroplast attractive as a chassis for specialized metabolite production, pathway tuning, and experimental redesign of organelle performance. Recent reviews now frame chloroplast synthetic biology as a platform problem in vector design, regulatory part selection, and metabolic routing rather than as a niche extension of plant transformation.

One reason this platform view matters is that chloroplast function is deeply connected to the nucleus. Plastids constantly report their physiological state through retrograde signaling. When chloroplast development stalls, redox balance shifts, protein homeostasis breaks down, or photooxidative stress accumulates, the nucleus responds by changing transcriptional programs. That distinction between developmental signaling and stress signaling helps explain why the same organelle can behave as both an energy center and a signaling hub.

That signaling layer intersects with chloroplast RNA metabolism more than many overview articles admit. In land plants, C-to-U RNA editing in chloroplast transcripts remains one of the most important post-transcriptional correction systems in the organelle. Editing can restore conserved codons, influence translation competence, and modulate the effective output of chloroplast genes. Recent reviews continue to link chloroplast RNA editing to plant stress response and organelle function. The same logic now extends into broader chloroplast RNA modification biology, where m6A, m5C, and other marks are increasingly discussed as regulators of RNA stability, processing, and translation.

Figure 7. Retrograde signaling and chloroplast RNA editing control map. Show plastid stress cues, RNA editing events, signal flow to the nucleus, and transcriptional response, and make the figure answer this question: how do chloroplast state and chloroplast transcript processing reshape nuclear response and photosynthetic output?

This is why the phrase “plastid as a metabolic bioreactor” should be used carefully. It is true, but only if “bioreactor” includes regulation, not just expression. A chloroplast is attractive because it can accumulate products at high levels and support polycistronic logic more naturally than the nucleus. But it is not a passive container. It is an active compartment embedded in broader cellular control.

Synthetic Biology Logic: Why Homologous Recombination Still Sits at the Center

Despite new editing concepts and expanded engineering toolkits, site-specific integration by homologous recombination remains the backbone of most practical plastid transformation logic. The basic workflow is familiar: build a transformation cassette, flank it with plastome-homologous regions, deliver it into plastids, select for integration, and then drive the population toward homoplasmy. But the details decide whether this works smoothly or becomes a prolonged sorting problem. Recent plastid engineering reviews still emphasize vector design, homologous arms, regulatory elements, and species-specific transformation constraints as decisive variables.

The first design decision is the target locus. An insertion site is never just an address. It carries a local transcriptional environment, neighboring gene context, repeat structure, and selection history. A locus that is easy to target is not always the best locus for interpretable output. A locus that supports strong output in one species may be poorly behaved in another because plastome organization, regulatory context, or regeneration behavior differs. That is why species expansion in plastid transformation remains such an active area. A 2025 study reported efficient plastid transformation protocols across five Nicotiana species, reinforcing that the technology is transferable but not automatically uniform across backgrounds.

The second decision is cassette logic. Homology arms must be long and specific enough to guide correct insertion. Regulatory elements must fit the plastid expression system rather than simply borrowing nuclear assumptions. Multigene designs must also respect plastid transcription and intercistronic processing rules. In other words, plastid engineering is modular, but it is not plug-and-play in the casual sense.

The third decision is how the line will reach homoplasmy. This is the most underestimated phase in many summaries. Successful insertion is not the end point. It is the beginning of a population problem. Early transformants usually contain a mixture of transformed and untransformed plastid genomes. Selection, regeneration, tissue context, and developmental bottlenecks then determine whether the transformed genome fixes or remains mosaic. This is exactly why heteroplasmy and engineering should not be written as separate topics. In real plastid work, they are two views of the same compartmental genetics problem.

Researchers working through difficult integration or purity questions often need more than one data layer. For example, chloroplast DNA sequencing can clarify locus structure and transformed plastome composition, while amplicon sequencing services or Sanger sequencing can help confirm specific junctions during line screening. When expression-level consequences also matter, pairing the genome view with RNA-Seq or full-length transcripts sequencing (Iso-Seq) becomes a natural extension rather than a separate project.

Figure 8. Plastid transformation by homologous recombination as an engineering pipeline. Show cassette design, homology-arm targeting, plastid delivery, site-specific integration, and enrichment toward homoplasmy, and make the figure answer this question: where does plastid transformation fail if insertion succeeds but compartmental sorting does not?

The most useful mental model is simple: plastid engineering succeeds when the insert fits the local plastome, the expression logic fits the organelle, and the population dynamics fit the selection scheme. If one of those three layers is mismatched, the line may still be produced, but it is less likely to remain well-characterized, reproducible, and interpretable under research conditions.

Assembly Challenges: Distinguishing PtDNA from Nupts

Among all plastome pitfalls, Nupts remain one of the easiest to mention and one of the hardest to handle well. Nuclear plastid DNA sequences are fragments of plastid origin now embedded in the nuclear genome. They are common enough to confuse read recruitment, variant calling, and even consensus plastome reconstruction if the pipeline assumes every plastid-like read belongs to the chloroplast genome. Once a study moves from simple skimming toward fine-scale structure or low-frequency variation, Nupts become a central interpretive risk.

The trap usually begins early. Reads are recruited with plastome seeds or references. Plastid-like reads accumulate, assembly proceeds, and a circular output appears. If the dataset is clean and the species architecture is standard, this may work well. But if nuclear inserts share sufficient similarity with the plastome, especially in conserved loci, false support can be carried forward into mapping, polishing, and minor-variant interpretation. The result is not always a catastrophic assembly failure. More often, it is a subtly overconfident one.

The right response is not to reject plastid-like reads aggressively. That can remove real signal. The better response is to rank evidence. True PtDNA support tends to show coherent depth, consistent graph placement, and structural agreement across the plastome. Nupts often reveal themselves through inconsistent coverage relative to the main plastome, conflicting placement when nuclear context is considered, or support patterns concentrated in conserved fragments rather than full plastome continuity. In difficult cases, longer reads are especially helpful because they can anchor suspect plastid-like sequence to broader nuclear neighborhoods or, alternatively, support uninterrupted plastid paths. This is one reason Nupts handling and repeat resolution belong in the same assembly discussion.

That logic also shapes software choice. In comparative studies, GetOrganelle has often performed more robustly than many alternatives because it emphasizes graph-aware assembly and disentangling of organelle paths from whole-genome data. NOVOPlasty remains useful in many contexts, especially when users need a lightweight seed-and-extend style organelle assembler, but in repeat-rich or ambiguity-prone plastomes, graph clarity and post-assembly scrutiny matter more than convenience alone. Benchmarking work has shown that software choice and sequencing coverage can materially change plastome outcomes, which is exactly why assembly papers should report pipeline logic, not only final circular sequences.

Not every plastid-like signal requires the same response.

Researchers who need to separate plastid signal from a broader genomic background often benefit from combining plastome-centered work with whole genome sequencing or variant calling pipelines that explicitly support cross-compartment interpretation. When the challenge is structural continuity rather than only depth, long-read-heavy designs can make the nuclear-plastid boundary much easier to resolve.

Figure 9. Nupts filtering and assembler comparison dashboard. Show true PtDNA paths, Nupts-like confounders, graph behavior, and confidence differences between assembler outputs, and make the figure answer this question: how do you tell real plastome continuity from plastid-like nuclear carryover when both can produce plausible-looking assemblies?

The final lesson is broader than chloroplast assembly alone. A plastome is easiest to misread when researchers force it into one of two oversimplified models: either a perfectly stable circular genome or a purely technical assembly object. It is neither. It is a biological structure with evolutionary memory, repeat-governed behavior, and engineering consequences. Good plastome work keeps all three layers visible at once.

FAQ

1. Why are chloroplast genomes still so useful for phylogenomics if they are structurally dynamic?

Because most plastomes remain conserved enough to support alignment and comparison across related taxa, while still carrying informative variation at IR borders, repeat landscapes, and rearrangement hotspots. That combination gives researchers both stability and signal.

2. What is the biggest mistake in plastome assembly?

The most common mistake is accepting a circular output as proof of correctness. A plastome assembly should be trusted only after repeat structure, IR boundaries, and suspicious junctions are checked against read support and graph behavior.

3. Why is the inverted repeat so important?

The IR affects plastome stability, local sequence behavior, gene duplication state, and assembly difficulty. In many studies, the most important plastome signal sits at or near an IR boundary rather than in total genome size.

4. Is chloroplast heteroplasmy rare?

Not necessarily. It is often underdetected rather than absent. Bulk sequencing, consensus assembly, and routine filtering can mask organelle-level or tissue-level mosaicism.

5. Can low-frequency plastid variants below 0.5% be trusted?

They can be informative, but only with strong validation. Confidence depends on locus uniqueness, effective depth, strand balance, replicate recovery, and orthogonal confirmation.

6. Why is plastid transformation still difficult if homologous recombination is precise?

Because precision at the insertion step does not guarantee success at the population level. After integration, transformed and untransformed plastid genomes often coexist, and the line must still pass through selection, regeneration, and sorting toward homoplasmy.

7. What role does RNA editing play in chloroplast function?

C-to-U RNA editing can restore conserved coding information and influence chloroplast gene output. More broadly, chloroplast RNA processing affects translation, stress response, and photosynthetic performance.

8. When should researchers worry about Nupts?

Whenever the project depends on fine-scale plastome interpretation rather than only rough organelle recovery. Nupts are especially important in low-frequency variant analysis, conserved-locus mapping, and repeat-heavy assemblies.

9. Which assembler is better, GetOrganelle or NOVOPlasty?

There is no universal winner for every dataset. GetOrganelle is often favored when graph-aware resolution and repeat interpretation are important. NOVOPlasty can still work well for simpler plastome recovery.

10. Why does this topic matter for plant synthetic biology?

Because plastid engineering depends on structural understanding. Target-locus choice, cassette design, expression logic, and homoplasmy all rely on correct interpretation of plastome architecture.

References

1. Dobrogojski J, Adamiec M, Luciński R. The chloroplast genome: a review. Acta Physiol Plant. 2020;42:98. DOI: 10.1007/s11738-020-03089-x
2. Ahmad N, Michoux F, McKenzie MJ, Nixon PJ. Plastid Transformation: How Does it Work? Can it Be Applied to Crops? What Can It Offer? Int J Mol Sci. 2020;21(14):4854. DOI: 10.3390/ijms21144854
3. Jin JJ, Yu WB, Yang JB, et al. GetOrganelle: a fast and versatile toolkit for accurate de novo assembly of organelle genomes. Genome Biol. 2020;21:241. DOI: 10.1186/s13059-020-02154-5
4. Giorgashvili E, Reichel K, Caswara C, Kerimov V, Borsch T, Gruenstaeudl M. Software Choice and Sequencing Coverage Can Impact Plastid Genome Assembly—A Case Study in the Narrow Endemic Calligonum bakuense. Front Plant Sci. 2022;13:779830. DOI: 10.3389/fpls.2022.779830
5. Krämer C, Boehm CR, Liu J, et al. Removal of the large inverted repeat from the plastid genome reveals gene dosage effects and leads to increased genome copy number. Nat Plants. 2024. DOI: 10.1038/s41477-024-01709-9
6. Raison d’être of the plastid repeat. Nat Plants. 2024. DOI: 10.1038/s41477-024-01710-2
7. Emerging roles of the C-to-U RNA editing in plant stress responses. Plant Sci. 2024;349:112263. DOI: 10.1016/j.plantsci.2024.112263
8. The Roles of RNA Modifications in Regulating Chloroplast Performance and Stress Responses. Int J Mol Sci. 2024;25(22):11912. DOI: 10.3390/ijms252211912
9. Recent trends and advances in chloroplast engineering and transformation methods. Front Plant Sci. 2025. DOI: 10.3389/fpls.2025.1526578
10. Luo Q, Obst S, Chang S, Ruf S, Bock R. Development of chloroplast transformation for five species in the genus Nicotiana. Plant J. 2025;e70542. DOI: 10.1111/tpj.70542