1. Introduction

Plastid is a semi-autonomous organelle characteristic of plants that perform photosynthesis and other biosynthesis functions that influence growth and development. Plastid genome (plastome) typically maps as a circular molecule tightly packed with genes encoding core proteins of photosynthetic complexes, proteins necessary for transcription and translation processes in the plastid, rRNAs and tRNAs (Green, 2011; Hu et al., 2015). Plastomes within a plant individual with uniparental inheritance are always thought homogeneous due to vegetative segregation (Birky, 1983; Bendich, 1987; Golczyk et al., 2014). However, plastome heteroplasmy in the same individual has already been found in many species (Moon et al., 1987; Johnson and Palmer, 1989; Chat et al., 2002; Lambertini, 2016; Sun et al., 2019; Lee et al., 2020, 2021; Zhou et al., 2024). The generation of plastome heteroplasmy may involves simple mutations (Greiner, 2012; Scarcelli et al., 2016; Khachaturyan et al., 2024), biparental inheritance (Johnson and Palmer, 1989; Smith, 1989; Reboud and Zeyl, 1994; Lambertini, 2016) and repeat-mediated recombination (Day and Ellis, 1984; Huang et al., 2001; Hsu et al., 2016; Lee et al., 2020, 2021; Zhou et al., 2024). In plastids, recombination events produce structural variations, including deletion, duplication and inversion, which are usually mediated by specific repeated sequences. Among the repeats, canonical large inverted repeats (IRs), which are usually 22–27 kb in angiosperms, can mediate the well-known intramolecular flip-flop recombination, which can create two structural configurations differing in the orientation of single copy regions (Kolodner and Tewari, 1979; Palmer, 1983; Stein et al., 1986; Bausher et al., 2006; Wang and Lanfear, 2019). Intramolecular recombination mediated by small repeats can also produce structural variations, such as inversions when repeats are inverted (e.g. Palmer et al., 1987; Doyle et al., 1996; Kim et al., 2005; Knox, 2014; Martin et al., 2014; de Santana Lopes et al., 2019; Charboneau et al., 2021) and deletions when repeats are direct (e.g. Day and Ellis, 1984; Ogihara et al., 1988; Harada et al., 1991; Kanno et al., 1993; Nimzyk et al., 1993; Shahid Masood et al., 2004; Sinn et al., 2018). Intermolecular recombination mediated by small inverted repeats can cause both duplications and deletions at the same time, which has been found in rice plastome (Hiratsuka et al., 1989), in long-term cultured cell lines and albino plants regenerated by anther culture in rice (Kawata et al., 1997), and in a variegated plant Dianella tasmanica (Zhou et al., 2024).

Variegated plants are excellent materials for studying plastome inheritance and evolution (Baur, 1908; Tilney-Bassett, 1975; Ruhlman and Jansen, 2014; Shrestha et al., 2021; Sakamoto and Takami, 2023), and genetic basis of leaf variegation (Yu et al., 2017; Azarin et al., 2020; Park et al., 2022; Wu et al., 2023; Zhou et al., 2024). Variegation can result from mutations in nuclear and mitochondrial genomes. For example, in the var2 mutant of Arabidopsis thaliana, leaf albinism is caused by defection of a nuclear-encoded, plastid-targeted protein AtFtsH, and in the NCS mutant of maize, the deletion of mitochondrial cox2 gene results in the mitochondrion defect which secondarily affects chloroplast development and causes leaf variegation (Gu et al., 1993; Raghavendra et al., 1994; Rodermel, 2001; Yu et al., 2007). However, mutations in plastomes appear to be the major cause for plant variegation (Greiner, 2012; Massouh et al., 2016; Park et al., 2022). Previous studies show that spontaneous single nucleotide mutations (usually produce premature stop codons) or single short insertions/deletions (indels) of 1–50 bp (usually cause frameshift and premature stop codons) in the coding regions of photosynthesis-related protein genes are predominant modes of plastome variation that lead to variegation (Schaffner et al., 1995; Hirao et al., 2009; Park et al., 2022; Wu et al., 2023), while only one case involves a simple plastome structural variation (Zhou et al., 2024). This structural variation, caused by intermolecular recombination mediated by a pair of 11-bp inverted repeats, generates a substantial expansion of IR to LSC, and a ~7 kb deletion on the boundary of LSC, which eliminates three protein coding genes and one tRNA gene. So relative to single nucleotide mutations or small indels, recombination can cause structural variation with much larger changes.

However, with only one reported case, it remains unclear whether plastome structural variation can result in variegation in other plants and whether there is any more complex structural variation caused by multiple small repeats-mediated recombination. In this study, we identified a complex plastome structural variation in albino sectors of a widely cultivated shrub Heptapleurum ellipticum (Araliaceae). This structural variation is caused by intermolecular and intramolecular recombination mediated by three pairs of small inverted repeats (7–8 bp each) and the canonical large IRs, leading to a large, rearranged and incomplete (loss of 11 genes) plastome, with dramatic changes in expression of many plastid genes in albino sectors. Moreover, we examined the microscopic characters and chloroplast ultrastructure of albino and green sectors of H. ellipticum to provide anatomical explanation of leaf variegation and further evidence for plastome heteroplasmy.

2. Material and methods 2.1. Sample collection, DNA and RNA extraction, and sequencing

Fresh leaves of two variegated individuals of H. ellipticum were collected in the campus of Sun Yat-sen University, Guangzhou, China. For each individual, albino sectors and green sectors were cut out separately. DNA extraction from these leaf tissues was carried out with a HiPure Plant DNA Mini Kit (Magen, Guangzhou, China). Four shotgun DNA libraries with an insert size of 350 bp were constructed and then sequenced on an Illumina Novaseq 6000 platform, which generated 4.3–5.9 G bases of paired-end reads of 150 bp (Table S2). In addition, one DNA library with an insert size of 100 kb for albino sector of one individual was constructed with the HLS HMW Library System (Sage Science, Beverly, MA, USA) and the Ligation Sequencing 1D Kit (Oxford Nanopore Technologies, Oxford, UK), and then sequenced using the PromethION instrument (Oxford Nanopore Technologies), which generated 21 G bases with read N50 length of 50, 828 bp.

RNA extraction from fresh leaf tissues of albino and green sectors was carried out with Trizol (Invitrogen, Carlsbad, CA, USA), and rRNAs were removed using the Ribo-Zero^TM Magnetic Kit (Epicentre, Madison, WI, USA). Two biological replicates each were done for albino and green sectors. Four transcriptome libraries with an insert size of 350 bp were constructed with the VAHTS® Universal V8 RNA-seq Library Prep Kit (Vazyme, Nanjing, China) according to manufacturer's instructions and then sequenced on an Illumina Novaseq 6000 platform, which generated 6.5–16.2 G bases of paired-end reads of 150 bp (Table S2).

2.2. Chlorophyll content measurement

Fresh leaves of three individuals were collected to measure chlorophyll contents. For each individual, green, transitional and albino sectors were separated and the chlorophyll of each sample was extracted following the manual of Chlorophyll Content Assay Kit (Biosharp, Anhui, China). The absorbance for chlorophyll a and chlorophyll b was measured using Varioskan^TM LUX (Thermo Fisher Scientific, Waltham, Massachusetts, USA) at the wavelengths of 665 nm and 649 nm, with 95% ethanol serving as the blank. The chlorophyll a and chlorophyll b contents of each sample were then calculated according to the manual of Chlorophyll Content Assay Kit. Pairwise t-test was carried out in R to test statistical significance between green, transitional and albino sectors.

2.3. Plastome assembly and annotation

The raw Illumina reads were filtered with Trimmomatic-0.39 (Bolger et al., 2014) with default parameters. For each sample, we used clean Illumina reads for plastome assembly with GetOrganelle v.1.7.5.1 (Jin et al., 2020) with the parameters: –R 30 -w 0.6 and -k 127. The raw Nanopore long reads from albino sector were filtered by NanoFilt-2.2.0 (De Coster et al., 2018) with parameters: -q 10 -l 10000 –headcrop 10 –tailcrop 10 –maxGC 0.49. Plastome assembly was also conducted for albino sector using ptGAUL-1.0.5 (Zhou et al., 2023) with these Nanopore long reads.

The assembly results were visualized in Bandage v.0.8.1 (Wick et al., 2015) to extract plastomes according to the sequencing depths (plastid DNA usually has much higher depth than nuclear and mitochondrial DNAs) and blast results of the assembled contigs. To verify the assembled plastomes, we checked the depth of coverage of the assembled plastomes by mapping Illumina reads to the assemblies (see below), and also did a dotplot analysis using some extracted plastid Nanopore long reads by aligning all the Nanopore long reads to the assembled plastomes using Minimap2 (Li, 2018). The dotplot was drawn with Gepard v.1.30 (Krumsiek et al., 2007), setting word length to 15. To show structural variations between the plastomes assembled from albino and green sectors, we also drew a dotplot with the same procedure.

Protein-coding genes were annotated with GeSeq (Tillich et al., 2017), and rRNA and tRNA genes were annotated with RNammer (Lagesen et al., 2007) and tRNAscan-SE (implemented in GeSeq) with the organelle option, respectively. Gene map of the plastome was plotted in OGDRAW (Lohse et al., 2013). The SeqMan NGen® v.17.6 (Lasergene, Madison, WI, USA) was used to identify the exact position of small inverted repeats flanking the duplication and deletion regions.

2.4. Assessment of plastome heteroplasmy

Illumina reads from albino and green sectors were mapped to the A-type and G-type plastomes, respectively, using BWA-mem (Li and Durbin, 2010) with default parameters. To confirm if there is heteroplasmy and calculate the proportion of the two types of plastome in each sector, we used Samtools v.1.13 (Danecek et al., 2021) and ggplot2 in R v.4.2.2 packages (Wickham, 2016; R Core Team, 2022) to extract and visualize sequencing depth at each position, respectively.

When reads from albino sectors were mapped to the G-type plastome, sequencing depth in the two deletion regions represent the depth of the G-type plastome in the albino sector, while sequencing depth in single copy region for both plastome types, such as nucleotide positions from 54, 000 to 60, 000 we used here, represents the sum of depths of the two plastome types. Therefore, the depth of the A-type plastome in albino sectors can be calculated using the average depth of the single copy region minus the average depth of the deletion regions.

To assess the proportion of the A-type plastome in green sectors, we mapped the reads from green sectors to the A-type plastome. The number of reads that span the junction at nucleotide positions from 61965 to 61966 was treated as the depth of the A-type plastome because reads from the G-type plastome cannot span this junction. Again, the depth of nucleotide positions from 54000 to 60000 represents the sum of depths of the two plastome types. Therefore, the depth of the G-type plastome in green sectors can be calculated using the average depth of the single copy region minus the depth of the A-type plastome.

2.5. Plastid gene expression analysis

Clean RNA-seq reads of each sample were mapped to the G-type plastome with Hisat2 (Kim et al., 2019), and the output files were converted and sorted by Samtools v.1.13. Then StringTie (Pertea et al., 2015) was used to assemble the transcriptome and calculate the transcript count matrix for plastid genes.

To normalize the expression levels of plastid genes among different samples, we de novo assembled the transcriptome by pooling RNA-seq reads of all four samples together using Trinity v.2.1.1 (Grabherr et al., 2011) with default parameter and then removed redundant transcript sequences using CD-hit 4.8.1 (Li and Godzik, 2006). Then the clean RNA-seq reads of each sample were mapped to this transcriptome with the Trinity script align_and_estimate_abundance.pl. We then merged the expression abundances to matrixes by the Trinity script abundance_estimates_to_matrix.pl. Finally, we combined two transcript count matrixes above for normalization and then identified differential expressed plastid genes using DESeq2 (Love et al., 2014). The volcano plot showing differentially expressed genes was drawn with ggplot2 in R.

2.6. Microscopic and transmission electron microscopy observations

To investigate anatomical differences between green and albino sectors in the leaves of H. ellipticum, we prepared free-hand transverse sections of fresh leaves and examined their microscopic characters. Fresh mature leaves were removed from the same individuals used for DNA/RNA isolation. The leaves were cleaned, and then albino, transitional and green sectors were sliced using a double-edged blade. The free-hand sections were then mounted on slides with distilled water, observed and photographed under an optical microscope.

To examine plastid ultrastructural differences between green and albino cells observed in the microscope, we carried out transmission electron microscopy (TEM) observations of ultrathin leaf sections from one individual. Sections were obtained from green, transitional and albino sectors, and each section was cut into 3× 3 mm pieces. All samples were vacuumized in 5% glutaraldehyde and 4% paraformaldehyde solution for 48 h, washed three times in 0.1 mmol/L PBS buffer (pH = 7.2) for 20 min each, and then post-fixed with 2% Osmium tetroxide overnight, followed by three additional washes. The post-fixed samples then underwent a gradual dehydration process in an acetone series (20 min per step), before being infiltrated with Spurr's resin. Ultrathin sections ranging from 70 to 100 nm were prepared using a Leica EM UC7 ultramicrotome (Leica Microsystems, Wetzlar, Germany) and placed on copper grids. Sections were stained with uranyl acetate, washed three times with distilled water, further stained with Reynold's lead citrate, and then subjected to three more washes. Finally, the samples were examined and photographed using a JEOL JEM-1400Flash Electron Microscope.

3. Results 3.1. The assembled plastomes differ between green and albino sectors

In variegated plants of Heptapleurum ellipticum, more or less albino sectors occur in the leaflets of a compound leaf (Fig. 1). The albinism is characterized by a significant decline of chlorophyll contents, including chlorophyll a and chlorophyll b (Fig. S1). Using Illumina reads, we assembled the plastomes of one green sector and one albino sector from each of two individuals of H. ellipticum, respectively. For each sector, no sequence variation was found for the assembled plastomes of the two individuals. The assemblies were supported by relatively even Illumina sequencing depth (Fig. S2a and S2b), and the assembly of albino sectors was also validated by nanopore long reads (Fig. S3). The assembled plastome for green sectors is a circular molecule of 156, 831 bp, including a large single copy region (LSC) of 86, 742 bp, a small single copy region (SSC) of 18, 147 bp, and a pair of IRs of 25, 971 bp each (Table 1 and Fig. S4). The plastome structure and size are typical for most angiosperms. The assembled plastome for albino sectors is also a circular molecule, but with a much larger size of 219, 181 bp, which is caused by its substantially expanded IRs (72, 841 bp each), while its LSC (54, 325 bp) and SSC (15, 954 bp) are in fact smaller than those of the assembled plastome for green sectors (Table 1). We denoted the two assembled plastomes for green and albino sectors as G-type and A-type plastomes, respectively. All but one annotated genes of the G-type plastome are identical to a previous study of this species (Zong et al., 2015). The exception is pbf1 (formerly psbN) (Krech et al., 2013), which has not been annotated in the previous study.

We found four major changes in the A-type plastome compared with the G-type plastome: 1) two areas (area Ⅰ: 7760 bp and area Ⅱ: 17, 803 bp) on the boundaries of the LSC were moved from the LSC to the IR, resulting in expansion of the IR and contraction of the LSC; 2) the IR was further expanded due to duplications of nearly whole region; 3) the LSC was further reduced by a deletion of 6, 975 bp (deletion Ⅰ) adjacent to area Ⅱ; and 4) the SSC was reduced by a deletion of 3, 504 bp (deletion Ⅱ) on the boundary of SSC (Fig. 2). Deletion Ⅰ eliminates 10 genes, i.e. pafII, psaI, petA, petG, petL, cemA, psbE, psbF, psbJ, and psbL, and deletion Ⅱ eliminates most of one gene, ndhF (2140 of 2226 bp), leaving only 86 bp at the 5′-end of this gene (Fig. S4). Thus, 11 genes are lost in the A-type plastome relative to the G-type plastome.

3.2. Intermolecular and intramolecular recombination can explain the formation of the A-type plastome

We noticed that there are several pairs small repeats of 7–8 bp each, either direct or inverted, flanking the areas of duplications and deletions mentioned above in the A-type plastome (Fig. 2). To explain the formation of the A-type plastome, the modes of repeat-mediated recombination in the plastome were divided into four categories. First, intramolecular recombination by one pair of repeats can lead to a deletion when they are direct (Fig. S5a) or an inversion when they are inverted (Fig. S5b). Second, let's consider two pairs of small inverted repeats (SIRs) in the plastome that can mediate recombination and divide the plastome into four segments. Here the only purpose of using "small" is to distinguish them from the canonical large IRs in the plastome. Based on their positions, the two pairs of SIRs can exist in an alternate or non-alternate manner. Intermolecular recombination mediated by one pair of SIRs can generate a large circular molecule consisting of two original plastomes. Note that after intermolecular recombination, the other pair of SIRs becomes small direct repeats (SDRs) in the large recombined molecule. When the two pairs of SIRs exist in an alternate manner, intramolecular recombination mediated by the other pair of SIRs within the large molecule can generate two circular molecules, with one containing two duplicated, alternate segments (red and green on the left, or yellow and blue on the right), and the other containing only two other segments (Fig. S5c). When the two pairs of SIRs exist in a non-alternate manner, intramolecular recombination mediated by the other pair of SIRs within the large molecule can also generate two circular molecules, both possessing one duplicated segment (green on the upper left) but missing another segment (red on the upper left) (Fig. S5d).

As shown in Fig. 3, the formation of the A-type plastome can be completed by three steps. First, recombination mediated by SIR₁ and SIR₂ generates a molecule containing a deletion (deletion Ⅰ) at nucleotide positions 61, 965–18, 939, and a duplication at nucleotide positions 132, 297–156, 831-1-7, 759 (recombination mode shown in Fig. S4d). Second, recombination mediated by SIR₂ and IR generates a molecule containing a deletion at nucleotide positions 130, 859–132, 297 and a duplication at nucleotide positions 68, 939–112, 713 (recombination mode shown in Fig. S4d). Given the nature of inverted repeats, we then move the sequence at the nucleotide positions 111, 404–112, 713 in the IR_B to fill the gap at nucleotide positions 130, 859–132, 297 in the IR_A. Third, recombination mediated by SDR₃ produces a deletion (deletion Ⅱ) at nucleotide positions 109, 374–114, 907 (recombination mode shown in Fig. S4b). Other intermediate products are generated during these steps, but they are not likely able to survive because of large-scale deletions. The process of the formation of the A-type plastome is also shown in Fig. S6, in a circular manner. Note that the above steps could occur in any order and the resulting recombinants are the same. In addition, there are other possible scenarios that can generate the A-type plastome by inter- and intra-molecular recombination mediated by these repeats, and some of them were shown in Fig. S7.

3.3. Plastome heteroplasmy exists in both green and albino sectors

To assess plastome heteroplasmy in the two sectors, we calculated the proportion of the G- and A-type plastomes using the depth information of a part of the LSC (nucleotide positions from 7761 to 61, 964) based on its presence in both plastomes as a single-copy region, and that of deletion Ⅰ (nucleotide positions from 61, 965 to 68, 939) based on its presence and absence in the two plastomes, respectively (See Material and Methods). The average depths of the G- and A-type plastomes in green sectors of one individual are ~2, 063× and ~40×, respectively, and those of the other individual are ~773× and ~25×, respectively. In albino sectors, ~1451× for the A-type plastome and ~193× for the G-type plastome were detected for one individual, and ~836× and ~162× for the other individual. Thus, both green and albino sectors of the two individuals show plastome heteroplasmy, and the A-type plastome exists at a very low frequency (1.9–3.1%) in green sectors, while the G-type plastome takes up a much higher proportion (11.7–16.2%) in albino sectors.

3.4. Dramatic changes in expression of plastid genes in albino sectors

Significantly decreased expression (Fold change > 2, p_adj < 0.05) was found for 37 plastid protein-coding genes in albino sectors compared with green sectors, including all 11 genes lost in the A-type plastome (Fig. 4, Table S1). Six of the 11 genes are the most significant ones. Meanwhile, one gene (rpl32) shows significant increase in expression in albino sectors. All 37 genes showing decreased expression are crucial components of photosynthesis, while the only gene with increased expression is related to translation apparatus (Table S1).

3.5. Albino sectors contain both abnormal and normal plastids

Transverse sections of green, transitional and albino sectors of variegated leaves were examined using optical and transmission electron microscopies. Mesophyll cells in green sectors harbored normal plastids, evident from their green color and well-developed grana lamellae and starch granules (Fig. 5a, c, e). In contrast, albino sectors exhibited a layer-specific difference in plastid morphology: most mesophyll cells contain abnormal plastids characterized by large vacuolated bodies and a lack of grana lamellae, while only the first layer of spongy mesophyll cells beneath the lower epidermis contain normal plastids (Fig. 5b, d, f). In transitional sectors, several layers of albino cells were observed between the multilayered upper epidermis and the first layer of spongy mesophyll cells beneath the lower epidermis, coexisting with green mesophyll cells (Fig. S8a and S8b). Despite these differences, plastids in the guard cells of the lower leaf epidermis have consistent morphology across all sectors, noted by the light green color, loosely arranged thylakoids, and large prominent starch granules (Figs. 5c, 5d, S7c and S7d).

4. Discussion 4.1. Intermolecular and intramolecular recombination leads to unusual large, rearranged and incomplete plastome in Heptapleurum ellipticum

Both spontaneous single nucleotide mutations and single small indels can lead to plant variegation (Schaffner et al., 1995; Hirao et al., 2009; Azarin et al., 2020; Park et al., 2022; Wu et al., 2023). In addition, a simple plastome structural variation, which is caused by intermolecular recombination mediated by one pair of 11-bp inverted repeats, eliminates four genes due to a ~7 kb segment deletion, is responsible for leaf variegation in Dianella tasmanica (Zhou et al., 2024). In this study, a complex plastome structural variation, involving three pairs of small repeats (7-8 bp each) that mediate both intermolecular and intraspecific recombination, is identified in albino sectors of H. ellipticum. This structural variation results in two deletions, two duplications and multiple rearrangements in the mutated (A-type) plastome relative to the wild-type (G-type) plastome. The most striking influence in function of the A-type plastome should be the two large deletions, one at ~7.0 kb and the other at ~3.5 kb in length, which eliminate 11 genes mostly related to photosynthesis and should be responsible for leaf variegation in H. ellipticum. The substantial IR expansion to the LSC may make genes moved from the LSC more stable, because a mutation in one of the IRs will be purged because of gene conversion from the other IR (Palmer 1983).

The specific mechanisms for these small repeats-mediated recombination should be microhomology-mediated end joining (MMEJ) and/or microhomology-mediated break-induced replication (MMBIR) (Kwon et al., 2010; Ottaviani et al., 2014; Sahoo et al., 2014; García-Medel et al., 2019), which is consistent with the modes shown in Fig. S5. The mechanism for the recombination events mediated by large canonical IRs should be homologous recombination (HR) (Aldrich et al., 1985; Li and Heyer, 2008; Guo et al., 2014; Qu et al., 2017; Wang and Lanfear, 2019).

It is well known that SSC isomerization in the plastomes is caused by intramolecular recombination mediated by the large canonical IRs (Palmer, 1983; Stein et al., 1986; Liu et al., 2013; Martin et al., 2013; Zhang et al., 2014; Wang and Lanfear, 2019; Barloy-Hubler et al., 2025). Many other inversions of different length in plastomes are also caused by small IR-mediated intramolecular recombination (Kim and Lee, 2005; Ban and Jansen, 2006; Knox, 2014; Martin et al., 2014; de Santana Lopes et al., 2019; Charboneau et al., 2021). In addition to inverted repeats, direct repeats can also mediate intramolecular recombination and result in deletions (Day and Ellis, 1984; Palmer et al., 1987; Ogihara et al., 1988; Harada et al., 1991; Kanno et al., 1993; Nimzyk et al., 1993; Shahid Masood et al., 2004; Sinn et al., 2018). Such intramolecular recombination generates either inversions or deletions, without any duplications.

Intermolecular recombination via one pair of IRs will produce a large circular dimer and subsequent intramolecular recombination via this pair of IRs within the dimer will produce two circular monomers each with an inversion between the IRs. The outcome of this recombination process is equivalent to intramolecular recombination, such as the example in rice (Hiratsuka et al., 1989), and more generally the flip-flop recombination mediated by the canonical IRs. When two pairs of IRs are involved in intermolecular recombination, the final products will contain duplications and/or deletions, as shown in Fig. S5. Plastome structural variation in Dianella tasmanica is such a case, in which a pair of 11 bp small IRs mediates intermolecular recombination and then the canonical IRs mediates intramolecular recombination (Zhou et al., 2024). Plastome structural variation in this study, involving three pairs of small repeats and causing multiple deletions, duplications and rearrangements, is more complex and has never been reported in plants before. However, as shown in Fig. 3, the complex process can be divided into multiple single recombination events, and the final product is a combination of the consequences of these events. As small direct and inverted repeats are not uncommon in plant plastomes, and rearranged plastomes can form by recombination involving either two or more pairs of small repeats, plastome structural variation of this kind may not be rare in albino sectors of variegated plants. The role of plastome structural variation in plant variegation can be evaluated by characterizing more variegated plants.

4.2. Gene loss due to two deletions in the plastome accounts for albinism in leaves of Heptapleurum ellipticum

In the A-type plastome, two deletions eliminate 11 genes (pafII, psaI, petA, petG, petL, cemA, psbE, psbF, psbJ, psbL and ndhF), and most of which encode proteins playing important roles in photosynthesis. Among these genes, psaI and pafII encode subunit Ⅷ of photosystem Ⅰ (PSI)and PSI assembly protein YCF4, respectively, involving the stability and efficiency of PSI (Scheller et al., 1989; Xu et al., 1995; Boudreau et al., 1997; Krech et al., 2012; Plöchinger et al., 2016); psbE, psbF, psbJ and psbL encode subunits of PSII, which are components of the reaction center of PSII (Wada and Arnon, 1971; Hird et al., 1986; Zheleva et al., 1998; Regel et al., 2001); petA and petG encode subunits of cytochrome b₆-f complex, which mediates electron transfer between PSII and PSI (Berthold et al., 1995; Choquet et al., 1998; Cramer et al., 2021); cemA encodes K⁺/H⁺ antiporter CemA, which controls proton extrusion and homeostasis in chloroplasts in a light-dependent manner to modulate photosynthesis (Harada et al., 2019; UniProt Consortium, 2022); and ndhF encodes subunit 5 of NAD(P)H-quinone oxidoreductase, involving the cyclic photophosphorylation induced by NDH complex (Martín et al., 2009; Su et al., 2022). Except cemA and ndhF, all these genes encode proteins directly associated with photosynthesis.

Many studies suggest that loss of function of a single gene related to photosynthesis in the plastome is responsible for albinism in plants (Schaffner et al., 1995; Kumari et al., 2009; Mozgova et al., 2012; Pfannschmidt et al., 2015; Azarin et al., 2020; Park et al., 2022). Meanwhile, loss of function of a single gene related to transcription (rpoA, rpoB, rpoC1 or rpoC2) can also lead to albinism in variegated plants, because these transcription-related genes can influence the expression of other plastid genes, including genes related to photosynthesis (Chateigner-Boutin et al., 2008; Canonge et al., 2021; Park et al., 2022; Wu et al., 2023). This suggests changes in gene expression can also cause leaf variegation. In albino sectors of H. ellipticum, the expression levels of all 11 genes lost in the A-type plastome are significantly decreased, and therefore, the loss of the 11 genes and accompanied expressional decline of 37 genes can account for the loss of photosynthesis, and thus albinism in the leaves of H. ellipticum.

Interestingly, only one gene, rpl32, is significantly upregulated in albino sectors of H. ellipticum. Deletions, duplications or rearrangements identified in the A-type plastome are beyond the range of the coding region of rpl32. The increased expression of this gene in albino sectors may be due to the use of an alternative promoter for transcription in albino sectors, where the primary promotor doesn't work normally, as characterized in the plastids of a non-photosynthetic line of Nicotiana tabacum, which also shows increased expression of rpl32 (Vera et al., 1992, 1996). Surprisingly, the alternative promoter for the non-photosynthetic tobacco line lies in the coding region of ndhF gene, whereas that for albino sectors of H. ellipticum may be located in the ndhF-rpl32 intergenic spacer region based on the position of transcription start site inferred by mapping transcriptome reads to this region.

4.3. Developmental explanation for the formation and maintenance of leaf variegation in Heptapleurum ellipticum

Albinism always involves abnormal development of plastids (Vaughn et al., 1980; Caredda et al., 2004; Lin et al., 2008; Kumari et al., 2009; Yang et al., 2015; Silva et al., 2020; Gajecka et al., 2021; Yan et al., 2022; Wang et al., 2023; Zhou et al., 2024). The A-type plastome, predominantly (83.8–88.3%) existing in albino sectors with plastids lacking grana lamellae while approaching zero (1.9–3.1%) in green sectors, roughly matches the presence of albino tissues/cells in the two sectors. This pattern suggests that the A-type plastome is confined to albino tissues/cells in the leaves of H. ellipticum. As shown in Fig. 5, the presence of the G-type plastome at a low proportion in albino sectors should be attributed to the existence of normally developed chloroplasts in the first layer of spongy mesophyll cells beneath the lower epidermis and guard cells from the lower epidermis.

The absence of some intermediate recombination products (marked "lost" in Fig. S5) may be explained by a sorting out process, which can be driven by both natural selection and random genetic drift (Vanwinkle-Swift, 1980; Hirai et al., 2007; Broz et al., 2024). Like what we observed in this study, when a plastome structural variation generated by intermolecular recombination initially arises, there exists different types of plastomes, including the G-type, A-type, and intermediate products, as shown in Fig. S6. During subsequent cell divisions, intermediate products can be sorted out extremely rapidly (Broz et al., 2022), and at last only the A-type plastome may persist over time in albino cells. This process likely occurred before the plant was recognized as a valuable variety and propagated through cuttings. This vegetative propagation mode allows the plant to escape strong selection against cells containing deficient chloroplasts. Consequently, the A-type plastome could become fixed in a cell, expand by cell division, and transmit across generations by vegetative propagation.

Variegation patterns in plants are determined by chimeras in shoot apical meristem (SAM) (Tian and Marcotrigiano, 1993; Marcotrigiano and Bernatzky, 2003; Frank and Chitwood, 2016). The SAMs of most seed plants are organized into clonally distinct cell layers, viz. the outer "tunica" and the inner "corpus" layers (Rösler, 1928; Satina et al., 1940). In most angiosperms, SAM comprises two tunica layers (L1 and L2) and one corpus layer (L3) (Popham, 1951), which contribute to different tissues in the lateral organs developed along the flanks of the SAM (Frank and Chitwood, 2016). During leaf development, L1 produces the colorless epidermal layer, L2 gives rise to the sub-epidermal palisade mesophyll and abaxial spongy mesophyll tissue, and L3 forms deep mesophyll and vascular tissues (Tilney-Bassett, 1986).

In Heptapleurum ellipticum, albino cells are observed in both palisade and spongy mesophyll tissues, indicating that the mutant cells with the A-type plastome occur in both L2 and L3 of the SAM. As shown in Fig. 1, albino sectors on the leaflets can occur on any part of the leaflet or even the entire leaflet. According to the classification of Frank and Chitwood (2016), the variegation pattern of H. ellipticum likely originates from sectorial chimeras. In the SAM, mutant cells form random patches in both L2 and L3, which eventually develop into albino cells in both palisade and spongy mesophyll tissues. L1 has no mutant cells, and thus forms guard cells with normal chloroplasts on the lower epidermis. Interestingly, the first layer of spongy mesophyll cells beneath the lower epidermis invariably harbors normal plastids in all sectors. This observation challenges the sectorial chimera hypothesis involving L2, since the first mesophyll layer beneath the lower epidermis is commonly believed to originate from L2. A plausible explanation is that the first mesophyll layer originates from L1 rather than L2, via periclinal division of the lower epidermis. In dicots, periclinal divisions in L1 are uncommon, with this layer typically remaining distinct from other layers and forming only the epidermis. However, several studies have demonstrated that this layer is capable of periclinal division, giving rise to internal mesophyll tissues in some plants (Stewart and Burk, 1970; Klekowski et al., 1996; Marcotrigiano, 2001; Beardsell and Norden, 2004). Further studies are needed to test this hypothesis in H. ellipticum.

Acknowledgments

We thank Shiquan Liang and Qi Zeng for help with plant sampling. This study was supported financially by the National Natural Science Foundation of China (31811530297 and 32170217).

CRediT authorship contribution statement

Kainan Ma: Investigation, Methodology, Visualization, Writing- Original draft preparation. Shuaixi Zhou: Investigation, Data curation. Ying Liu: Conceptualization, Methodology, Writing - Review & Editing. Renchao Zhou: Conceptualization, Methodology, Resources, Writing - Review & Editing, Supervision, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.09.003.