The phylogenetic diagram of the studied plants species together with the bar chart of ABCE gene number was created based on NCBI taxonomy with phyloT and visualized with iTOL (Letunic and Bork, 2007; 2019; phylot.biobyte.de).

ABCE sequence data for 55 species was downloaded from the online resource Phytozome portal https://phytozome.jgi.doe.gov/ (Goodstein et al., 2012). ABCE sequence data for additional 18 species was downloaded from Ensembl Plants (Howe et al., 2020). The genome data for C. hirsuta was accessed at http://bioinfo.mpipz.mpg.de/blast/ (Gan et al., 2016). Genome data for N. tabacum and N. benthamiana was downloaded from Sol Genomics Network http://solgenomics.net (Edwards et al., 2017; Kourelis et al., 2019). We downloaded genomic, CDS and translated amino acid sequences for each plant ABCE gene used in the study.

The length of amino acid sequences was calculated with SeqinR package (version 3.6.1) in R 4.0.2 (Charif and Lobry, 2007). In order to calculate their similarities to AtABCE2, all 152 sequences were aligned using the online interface of MUSCLE with default Pearson/FASTA parameters provided by the European Bioinformatics Institute (EBI) (Madeira et al., 2019). Thereafter the percent identity scores were calculated with MUSCLE algorithm for aligned sequences in R 4.0.2 package Bio3D version 2.4-1 (Edgar, 2004; Grant et al., 2021).

Next, sequences aligned to AtABCE2 were inspected for the general protein structure, that is the presence and correct order of the domains, including iron-sulphur (FeS) cluster domain, NBD1, NBD2 and bipartite Hinge domain (Supplementary Figure S1). Sequences lacking critical motifs within these domains (Karcher et al., 2005; Barthelme et al., 2007; Nürenberg and Tampé, 2013; Nürenberg-Goloub et al., 2020) were filtered out from the analysis.

Amino acid sequences were aligned with default parameters of the MUSCLE algorithm, as implemented in the MEGA software package (version 11.0.13) (Tamura et al., 2021). CDS sequences were aligned with the default parameters of MAFFT (version 7.4.9.0) (Katoh and Standley, 2013). As the alignments contained gaps, sites with no data for more than 10% of the sequences were removed with trimAl (version 1.4. rev22) (Capella-Gutiérrez et al., 2009), resulting in an amino acid alignment of 603 positions and a CDS alignment of 1,810 positions.

Phylogenetic trees of ABCE full length amino acid sequences and the corresponding CDS sequences were constructed by IQ-TREE (version 2.0.7) (Minh et al., 2020) with the Maximum Likelihood method and 10,000 rapid bootstrap replicates. Initial tests suggested that the best models to use would be JTTDCMut for amino acids and TIM2e for CDS sequences, both with five degrees of FreeRate heterogeneity (Soubrier et al., 2012). Tree calculation was performed 20 times independently for both CDS and amino acid input, with random seed values ranging from 1 to 20. TreeGraph (version 2.15.0–887) (Stöver and Müller, 2010) was used to root all trees on Chlorophyta and collapse any branches supported by bootstrap values of 50% or less. The trees were then all added to a single file in the order of their log likelihoods (highest to lowest) and annotated in FigTree (version 1.4.3; http://tree.bio.ed.ac.uk/software/figtree/). Final adjustments (fine-tuning the color scheme) were done in Adobe Illustrator.

Unrooted tree supporting suitability of Chlorophyta as an outgroup as well as separate trees based on the amino acid sequences of ABCE domains were constructed by MEGA software. Phylogenetic relationships were inferred using the Maximum Likelihood method and JTT matrix-based model (Jones et al., 1992), selected for each data set based on the lowest BIC scores (Bayesian Information Criterion). A discrete gamma distribution was used to model evolutionary rate differences among sites. Bootstrap analysis was performed with 500 replicates. The trees with the highest log likelihood were published for each analysis.

Two different approaches were used to test the impact of assembly size, ploidy level, and the number of known ancient whole genome duplications on ABCE copy numbers. A simple linear regression was performed with the lm function available in base R (version 4.3.2) (R Core Team, 2023). In an alternative approach, the R package taxize (version 0.9.100) (Pinheiro et al., 2023) was first used to extract phylogenetic relationships of all species involved from the NCBI databases. Next, the function gls (generalized least squares) from R package nlme (version 3.1–164) (Pinheiro and Bates, 2000; Pinheiro et al., 2023) was used to create additional regression models with phylogenetic signal included as a random effect. Ggpredict from the R package ggeffects (version 1.3.4) (Lüdecke, 2018) was used to obtain prediction intervals for the models. The results were plotted with the packages ggplot2 (version 3.4.2) (Wickham, 2016) and patchwork (version 1.2.0) (Pedersen, 2024), then adjusted in Adobe Illustrator.

Data (SNPs and indels) available for 1135 A. thaliana strains was downloaded from the 1,001 Genomes Project depository (Weigel and Mott, 2009).

Seeds of the selected 21 Arabidopsis ecotypes were acquired from the Nottingham Arabidopsis Stock Centre (NASC). Both AtABCE1 and AtABCE2 full coding sequences were PCR-amplified and sequenced by Sanger sequencing for ecotypes Can-0, Ei-2, IP-Car-1, Kia1 and Pra-6. For the other 15 ecotypes only AtABCE1 full coding sequences was sequenced. Primer pairs used for the PCRs are shown in the Supplementary Table S1. For all the PCR reactions touchdown PCR method with the following conditions was used: 95°C for 15 min; 13 cycles of at 95°C for 15 s, at the gradually decreasing temperature from 60°C to 54°C (the temperature drops by 0.5°C per cycle) for 30 s, at 72°C for 70 s; 15 cycles of 95°C 15 s, 54°C 30 s, 72°C 70 s and the final extension at 72°C for 10 min. Amplified DNA fragments were purified from the agarose gel using GeneJET Gel Extraction Kit (Thermo Scientific) according to the manufacturer’s instructions. Thereafter the purified DNA fragments were Sanger sequenced and aligned respectively to AtABCE1 or AtABCE2. The final results were based on the sequencing of at least two plants for each ecotype. Columbia (Col-0) ecotype was used as a reference.

For the haplotype analysis, first an alignment file with CDS sequences was created in Nexus format. Thereafter the multiple sequence alignment was analyzed with PopART version 1.7 (Population Analysis with Reticulate Trees) (Leigh and Bryant, 2015). The network was constructed with Median Joining Network algorithm (epsilon = 0).

To gain insight into the diversity of ABCE genes in plants, we compiled a selection of ABCE genes from 76 different plant species available in public databases. The selection criteria for including in further analysis was the presence of all known essential structural elements of ABCE proteins (Karcher et al., 2005; Barthelme et al., 2007; Nürenberg and Tampé, 2013; Nürenberg-Goloub et al., 2020). Truncated or aberrant sequences were discarded from further analysis. Altogether 152 plant ABCE genes were included in the study. We found that the amino acid sequence identity among the 152 proteins was 78% or higher when compared to A. thaliana AtABCE2. The selected species represented a wide range of plant groups, including unicellular algae such as Chlamydomonas reinhardtii and Micromonas sp. RCC299, monocots such as Zea mays and Triticum aestivum, Solanum species such as Solanum tuberosum and N. benthamiana, and Brassicaceae such as Brassica napus and A. thaliana (Figure 1; Supplementary Table S2). Our analysis revealed that plant species from the phylum Chlorophyta (green algae) usually possess only a single ABCE gene, except for Ostreococcus lucimarinus, which has two genes. In contrast, most of the analyzed species in the Poaceae family have at least two ABCE genes, and some have as many as eight genes in their genome, as is the case for T. aestivum. Another group of plant species with an above-average number of ABCE genes is the Brassicaceae family. For example, B. napus has eight genes, B. rapa five genes, and Capsella rubella four genes. On the other hand, among Brassicaceae, C. hirsuta and Boechera stricta have only a single ABCE gene. Despite clustering of multi-gene-species in the Poaceae and Brassicaceae families, there was no visual segmentation between the number of ABCE genes and phylogenetic origin in other plant families (Figure 1). Interestingly, only 30 species out of 76 (39.5%) had a single ABCE gene. Similarly, there were 34 species (44.7%) possessing two ABCE genes (Figure 2A). This shows that despite some species having only single functional ABCE gene (containing full set of critical structural elements) ABCE genes in plants can be classified as a low-copy gene family instead of single-copy gene family.

Our next goal was to explore how genomic parameters contribute to ABCE gene copy number in plants. We tested the effect of ploidy level, genome size and WGD events during evolution on the ABCE gene family size of a species. WGD data was based on the data of 53 plant species published by the One Thousand Plant Transcriptomes Initiative (Leebens-Mack et al., 2019). The observation of WGT data in the common ancestor of Brassica species was based on the study of Wang and coworkers (Wang et al., 2011). Regression analysis showed positive effect of ploidy level and genome size on ABCE gene copy number. Among selected parameters a degree of ploidy is likely the most suitable as a prediction factor, although there were examples of tetraploid species with a single ABCE gene (e.g., Nicotiana tabacum) and diploids with five ABCE genes (e.g., Brachypodium distachyon, B. rapa) (Figures 2B,C,E; Supplementary Table S2). In contrast, there was no clear correlation between ancient WGD events experienced by a species and ABCE gene family size (Figures 2B,D; Supplementary Table S2). Interestingly, there are examples of species, which have encountered at least five WGDs in their evolutionary history, but still possess only a single ABCE gene, for example, Actinidia chinensis and Glycine max (Figure 1; Supplementary Table S3).

ABCE proteins are composed of four domains: NBD1 and NBD2 forming the ATPase core, bipartite hinge domain that is tightly engaged in twin-NBD cassette arrangement and a unique N-terminal FeS cluster domain (Figure 3A). Additionally, ABCEs embody a helix–loop–helix (HLH) motif in NBD1 that distinguishes it from otherwise superimposable NBD2 (Karcher et al., 2005).

As many as 86.8% out of the 152 analyzed gene sequences encode ABCE of canonical protein length (600—609 amino acids) (Figure 3B). Along with exceptional conservation within functionally critical motifs, all analyzed ABCE sequences are highly similar to AtABCE2 sharing at least 78% of amino acid sequence identity (Figure 3C; Supplementary Table S4).

We also plotted amino acid sequence length to amino acid sequence identity for the studied 152 ABCE proteins. There was a clear clustering of proteins with the length of 605 amino acids (Figure 3D). Proteins with lower sequence identity did not cluster by protein length (Figure 3D). Interestingly, proteins with more than 90% identity to AtABCE2 could be as short as 591 amino acids and as long as 625 amino acids long (Figure 3D). Hence, despite some variance in amino acid sequence length and sequence identity to AtABCE2, the selection of amino acid sequences analyzed here is uniform and represents well the plant ABCE genes.

To understand how ABCE proteins have evolved in the green plant lineage, we constructed 20 Maximum Likelihood (ML) trees of 152 full-length ABCE protein sequences and the corresponding coding DNA (CDS) sequences from 76 species (Figure 4; Supplementary Figures S2–S4). Chlorophyta (green algae), the earliest lineage to have split off from the rest of the green plants, was used as an outgroup for rooting the trees. In an unrooted tree the representatives of Chlorophyta formed a separate cluster with high bootstrap support (Supplementary Figure S5).

The CDS tree (Figure 4; Supplementary Figure S4) proved much more informative than the amino acid tree (Supplementary Figures S2, S3). Most internal branches in the amino acid tree are poorly supported by bootstrap values, likely resulting from the lack of phylogenetic signal in the highly conserved sequences. The CDS tree is more congruent with the known species tree (Figure 1). However, both show that the sequences cluster according to major taxonomic groupings, usually with high bootstrap support. For instance, all ABCEs from Fabaceae (legumes) form a single cluster, as is the case for Solanaceae (nightshades), Poaceae (grasses), Brassicaceae (mustard and cabbage family, including the thale cress A. thaliana), and others. These groupings appear older than any duplication events present in the ABCE family. For example, the AtABCE1 and AtABCE2 genes of A. thaliana are the result of a duplication that happened in Brassicaceae, whereas the multiple copies seen in wheat and other members of Poaceae result from different duplication events. Further examination of the data revealed that all cases with more than one ABCE sequence in a given species can be broadly divided in two. Often the closest neighbor for one of the sequences was a different one from the same species, which is most likely reflective of recent duplication events. However, in other cases much older duplications were found, with two or more gene copies evolving independently across different species of the same plant family or order. This is well known for ABCE genes in Brassicaceae (Navarro-Quiles et al., 2022), and our analysis confirms that many of their family members including A. thaliana encode distinct ABCE1 and ABCE2 (Figure 4). In support of the notion that AtABCE2 preserves the ancestral function (Navarro-Quiles et al., 2022), we show that ABCE2 sequences have fewer mutations and shorter branch lengths compared to ABCE1s in the phylogenetic tree (Supplementary Figure S4). Notably, all Brassicaceae species have at least one ABCE2 protein while ABCE1 can be missing.

Similarly to the ABCE1 of Brassicaceae, Poaceae (grasses) also include one set of ABCEs that have acquired more mutations than the others. It is present in all analyzed species of rice (Oryza), in foxtails (Setaria), in sorghum and in the common grasses Panicum hallii and Brachypodium distachyon. All of those also have at least one slower evolving copy. Wheat and related species (Triticum) do not have a direct homolog of the fast-evolving ABCE copy, but they have multiple ABCEs regardless. In any case, the fast-evolving ABCE of Poaceae is not the direct homologue of the ABCE1s in Brassicaceae, and it is much closer to other ABCEs in Poaceae instead. Thus, both Poaceae and Brassicaceae have at least two distinct ABCE lineages that appeared in the ancestors of the respective families. In both cases there is evidence of rapid accumulation of mutations in one of the genes (long branches in the phylogeny), likely reflective of neo- or subfunctionalization. In amino acid trees (but not CDS trees) those two unrelated groups stemming from long branches are typically grouped together, which is likely due to long branch attraction. In most such cases, one of the three genes from the plum Prunus persica also tends to group together with those two sets of sequences (as seen from supplementary trees presented in the associated GitHub repository: https://github.com/jsuurvali/abce152). As expected, the plum gene originates from a longer branch than the other two genes in that species.

In contrast to Brassicaceae and Poaceae, based on our dataset no such ancestral subtype separation was found in other plant clades. For example, the model species N. benthamiana has two ABCEs, but some Solanaceae have only one, and all ABCEs in Solanaceae are closer to each other than to any of the ones in Brassicaceae or Poaceae (Figure 4; Supplementary Figure S4). However, in CDS tree the placement of some groups in relation to each other could not be resolved from ABCE sequences alone and was both incongruent with their known positions in the Tree of Life and poorly supported by bootstrap analyses. This is the case for the relationship between Brassicaceae and other representatives of Rosids (including Citrus sp.). The placement of ABCE sequences from Amborella trichopoda, a single extant member of a sister lineage to all other angiosperms, did not match the species tree as well (Figure 4; Supplementary Figure S4). However, the effect of those artifacts was reduced by collapsing poorly supported branches in the tree, and showing then clearly that the exact relationship between Brassicaceae and other eudicots, or A. trichopoda and other angiosperms cannot be fully resolved based on the data. In the current version, those branches and groupings appear as part of polytomies in a multifurcating tree.

In addition, we separately analyzed the three main domains of ABCE proteins, FeS cluster domain, NBD1, and NBD2, by realigning the corresponding amino acid sequences and constructing a ML-tree for each (Supplementary Figures S6–S8). The topology of the resulting trees was different for each domain, but none of those were well supported by bootstrap analysis.

Natural variation among A. thaliana ecotypes has been well documented by the 1,001 Genomes Project (Weigel and Mott, 2009). We analyzed the ABCE gene sequences of all 1135 A. thaliana ecotypes reported in that project and found 35 and four non-synonymous SNPs in AtABCE1 and AtABCE2, respectively (Supplementary Table S5). Only four reported non-synonymous SNPs in AtABCE2 indicate a low degree of natural variation, which is consistent with its fundamental, conserved role in growth and development. On the other hand, 35 non-synonymous SNPs annotated for AtABCE1 show relatively higher natural variation. This finding is in agreement with the results from the transspecies phylogenetic analysis of Brassicaceae ABCE1 and ABCE2. From the previously reported SNPs in AtABCE1, we selected 18 that cause amino acid substitutions at conserved and important positions or that were present in combination with other SNPs of interest. Therefore, 21 ecotypes were included in the further study and resequencing, together with Col-0 (Table 1; Supplementary Table S5). Two SNPs causing amino acid substitution could not be detected by resequencing (Pro399Thr in Grivo-1 and Gly182Ser in IP-Cot-0). Instead, one SNP previously undocumented in the 1,001 Genomes Project database (Leu253Phe in Grivo-1) was identified. Figure 5A shows the positions of the amino acid substitutions caused by the 17 SNPs sequenced in the AtABCE1 gene. In our resequencing analysis, the most frequent SNPs in AtABCE1 caused the changes His561Leu and Ala441Thr (Table 1; Supplementary Table S5). Noteworthy, histidine at the position 561 seems to be characteristic to Col-0, as all the other analyzed ecotypes had leucine at this position. Next, we performed haplotype analysis with CDS sequences on the PopART platform and found that the most conserved sequence of AtABCE1 is most probably the one identical to Ei-2, Kia1, Pra-6, IP-Car-1 and Can-0. Eight SNPs out of 17 appear as single SNPs in the AtABCE1 of Kly4, Toufl-1, Col-0, IP-Ezc-2, IP-Vis-0, IP-Moz-0, IP-Hoy-0, IP-Loz-0, IP-Cot-0 and Lebja-1. Interestingly, a substitution of Ala441Thr can appear both as the consequence of a single SNP in Leska-1-44 and together with other SNPs such as in Cvi-0 or Qar-8a (Figure 5B). Some amino acid changes, like Pro129Gln, Ala549Gly and His561Leu, are always grouped (Table 1). Pro129Gln appears only together with at least two other SNPs, e.g., in Grivo-1 or Qar-8a (Figure 5B).

From the previously reported four non-synonymous SNPs in AtABCE2, we were able to reconfirm three SNPs causing amino acid changes (Gly44Ser, Met379Thr and Gly405Ala), which were located in FeS cluster domain and NBD2 at non-conserved positions (Figure 5C; Supplementary Table S5). Asp189Glu in AtABCE2 was not possible to reverify due to unavailable seed material. The verified SNPs were present in five different ecotypes (Table 1). The haplotype map of AtABCE2 SNPs shows that Col-0 has the most conserved sequence and the three different SNPs root from it (Figure 5D).

In this study we could not find any correlation between the presence of non-synonymous SNPs in AtABCE genes and the geographical origin of the ecotype (Table 1). Visual rosette phenotype of the studied ecotypes matched with characterization available in the public databases (Supplementary Figure S9).

In the case of 18 out of 21 ecotypes, all SNPs were confirmed as reported earlier. For three ecotypes only part of the SNPs was validated: Leska-1-44 did not exhibit Pro129Gln, IP-Cot-0 did not exhibit Gly182Ser and Grivo-1 did not exhibit Pro399Thr amino acid changes in AtABCE1 (Supplementary Table S5; Table 1). More interestingly, we verified Gly125Glu in IP-Cot-0 and Pro129Gln, Leu253Phe, Ala441Thr and His561Leu in Grivo-1. Leu253Phe had not been annotated in any A. thaliana ecotype in the 1,001 Genomes Project database (Table 1).

Surprisingly, we noticed some SNPs affecting highly conserved amino acid residues in AtABCE1. These include arginine residues from R cluster of Hinge domains (Arg323Cys and Arg572Leu of IP-Moz-0 and Toufl-1, respectively), and Arg86Gln of IP-Ezc-2 ecotype that locates to the Y-loop I (Figure 5A; Table 1). According to the 1,001 Genomes Project database the latter SNP is present in 19 A. thaliana ecotypes (Supplementary Table S4).

Although, the length of ABCE proteins in A. thaliana ecotypes is very conserved, in a single ecotype, namely, Kly-4, we found a SNP in AtABCE1 causing premature stop codon that makes the protein 14 amino acid residues shorter. Despite this deletion, the cluster of arginine residues remains intact in Hinge II subdomain (Figures 5A,B; Supplementary Figure S1).

As was previously mentioned, in most animal and in yeast species the ABCE gene family is represented by a single gene that encodes the vital ABCE1 protein. In plant kingdom, ABCE gene family size across different lineages is more variable. Based on the data from the public databases and our analysis we were able to reconfirm the same number of ABCE genes for a selection of plant species. For example, there is a single gene in C. hirsuta (Kougioumoutzi et al., 2013), in Chlamydomonas reinhardtii (Li et al., 2022), in Citrus sinensis and in Theobroma cacao (Navarro-Quiles et al., 2022). Similarly to previous studies, we reverified two genes in Z. mays (Pang et al., 2013), in Solanum lycopersicum (Ofori et al., 2018), in Oryza sativa, and in Populus trichocarpa (Navarro-Quiles et al., 2022). The same was true for five ABCE genes from Brassica rapa (Navarro-Quiles et al., 2022). Intriguingly, Zhang and others found three ABCE genes in Hordeum vulgare, whereas our study identified only two fully intact ABCE sequences with all canonical subunits (604 and 611 amino acids long) (Zhang et al., 2020). Moreover, for C. rubella we identified four ABCE genes as opposed to two sequences analyzed earlier (Navarro-Quiles et al., 2022). These discrepancies might be due to the strict filtering of non-functional protein sequences performed in our study. In addition, fast-developing sequencing technologies and implementation of novel annotation tools contribute to higher accuracy of the newer genome versions. Taken together, plant species present in our dataset possess one to eight ABCE genes encoding complete proteins, suggesting that plant ABCE genes do not comprise a single-copy gene family, but rather should be classified as a low-copy gene family.

Most plant species have experienced at least two ancient WGD events, followed by additional rounds of genome doubling in many lineages (Qiao et al., 2022). Thus, we were tempted to speculate that WGD could impact ABCE family size variation. However, in this study we did not notice significant correlation between the number of ancestral WGD events obtained from the literature and ABCE gene copy number in a species. This might be due to the rapid genome downsizing following WGD event that is leading to chromosomal rearrangements and extensive loss of duplicated copies (Wang et al., 2021). In addition, the WGD-derived duplicates are often subject to relaxed selection, meaning rapid mutation resulting in defunctionalisation (Qiao et al., 2019). For example, Glycine max and A. chinensis retained a single functional ABCE gene after five documented WGD events. Interestingly, the overexpression of ABCE1 in yeast causes growth inhibition (Dong et al., 2004), meaning that the amount of ABCE present—and therefore probably also the number of hypothetical redundant genes—is critical for the well-functioning of translation, a crucial process.

However, when higher expression of a particular gene is beneficial, its duplicate might be retained in the genome. This could be the case for the two ABCE paralogues in Z. mays that are located close to each other in our phylogenetic analysis (Figure 4) and share the same expression pattern profiles (Pang et al., 2013). Alternatively, as a result of faster evolution, gene duplicates may obtain novel functions or specialized expression patterns (Prince and Pickett, 2002). In Arabidopsis, AtABCE1 and AtABCE2 exhibit partial functional redundancy. In contrast to AtABCE2, which is ubiquitously expressed, AtABCE1 is mostly present in generative organs and at relatively low levels (Klepikova et al., 2016; Navarro-Quiles et al., 2022; Yu et al., 2023). This could mean an ongoing process of pseudogenization or subfunctionalization, where the paralogues acquire specific roles. There is growing evidence regarding ribosomal heterogeneity and the existence of specialized cell-type-specific ribosomes (Xue and Barna, 2012; Barna et al., 2022), suggesting that AtABCE1 is involved in the regulation of translation in generative tissues. Paralogous ABCE genes in plants may serve to provide specificity in fine-tuning translation and controlling cellular translatome (Gerst, 2018).

We also noticed slight positive effect of ploidy level on ABCE gene copy number in a species. In the future, the determinants of ABCE copy number can be further elucidated by including more species from diverse lineages and using statistical modelling that takes phylogenetic structuring of the data also into account. These models could also potentially incorporate other information from the species that was not used for the present study, such as whether the species is annual or perennial, their preferred mode of reproduction, or what kind of environments do they grow in.

In this study we analyzed 152 ABCE sequences from 76 plant species. This included the most well studied plant ABCE gene—AtABCE2, which is thought to preserve the ancestral functions of ABCE proteins (Navarro-Quiles et al., 2022). Phylogenetic trees of full-length ABCE sequences confirmed previously reported clustering into ABCE2 and ABCE1 groups for Brassicaceae (Navarro-Quiles et al., 2022) but also demonstrated that the multiple copies observed in several other plant lineages originate from separate duplication events. The results suggest that Brassicaceae and Poaceae families have undergone independent lineage-specific splits of the ancestral ABCE gene. Pooideae, the largest Poaceae subfamily that includes barley and wheat, appears to have had further duplication events and its members have additional ABCE genes. In addition to Brassicaceae and Poaceae, many other plant taxa have also gained ABCE gene copies, most likely because of more recent duplications. Interestingly, it was recently shown that one of the four ABCE gene copies from Prunus dulcis, a close relative of P. persica, originated from tandem duplication. ABCE genes share strong collinearity between these species, suggesting that this duplication preceded speciation event (Zhang et al., 2024). We can therefore postulate that in contrast to species which possess a single ABCE gene and are sensitive to copy number changes (Dong et al., 2004), many plants have gained additional ABCE copies through WGD or local duplication and likely have evolved to benefit from higher numbers of this essential translational factor. Gene copies may arise from different events including WGD, tandem- and transposon-related duplications, but the precise source of ABCE subfamily expansion in plants remains to be investigated.

Usually, essential genes are subject to strong evolutionary pressure and thus, non-synonymous SNPs in conserved regions of gene sequences are rare (Castle, 2011; Pang et al., 2016). In AtABCE1 we found three SNPs that could potentially impact the protein’s function (Table1; Figure 5A). SNPs causing the substitutions Arg323Cys (in IP-Moz-0) and Arg572Leu (in Toufl-1) located at the Hinge domain I and II, respectively, could be of importance, since these domains are essential for NBD-twin cassette assembly in the case of ABCE1 in other organisms (Karcher et al., 2005). In addition, AtABCE1 of IP-Ezc-2 ecotype contains an amino acid substitution at position Arg86Gln, which is an exceptionally conserved site across archaea as well as eukaryotes and locates to Y-loop I. In the context of Y-loop with consensus sequence H R YGVNAF, the arginine residue has been shown to mediate interaction between FeS cluster domain and NBD1 in the sole ABCE1 gene of Pyrococcus abyssi (Karcher et al., 2008).

The His561Leu amino acid change was reported to be present in 997 out of 1135 A. thaliana ecotypes (Supplementary Table S5), which suggest that histidine at this position of AtABCE1 might be characteristic only to a small subset of ecotypes including Col-0. Thus, it seems that leucine is the most conserved residue at position 561 in AtABCE1 among Arabidopsis ecotypes.

As expected, in AtABCE2 gene, known to be essential for the viability of an organism (Navarro-Quiles et al., 2022; Yu et al., 2023) only four non-synonymous SNPs residing in non-conserved regions were reported among 1135 A. thaliana ecotypes (Supplementary Table S5). Importantly, the AtABCE2 gene seems to be hard to mutate, since up to now there is no T-DNA homozygous line available and only one viable, hypomorphic allele has been recently isolated after ethyl methanesulfonate mutagenesis (Navarro-Quiles et al., 2022). Interestingly, the only non-synonymous SNP present in more than one ecotype in the case of AtABCE2 is leading to Gly44Ser substitution in the FeS domain. According to the 1,001 Genomes Project database this mutation is present in 54 ecotypes, three of them were confirmed in the current study (Table 1; Figure 5; Supplementary Table S5). The same position is able to incorporate 12 different amino acid residues among other plant ABCEs studied herein, alanine, glycine and serine being the most common ones. Moreover, the SNP M379T verified only in IP-Car-1 poses a promising material for further mutational analysis as this site is highly conserved among plant ABCEs with only leucine as a rare alternative (Figure 5D; 152_pepseq_MUSCLE.fa from the associated GitHub repository).

Taken together, this study has shown the surprisingly high number of ABCE genes among the plant kingdom. We hypothesize that plants have developed a number of specialized ABCEs with more specific functions compared to species carrying a single copy of ABCE gene such as humans, fruit fly or yeast.