Phylogenomics and plastomics offer new evolutionary perspectives on Kalanchoideae (Crassulaceae)

Abstract

Background and Aims

Kalanchoideae is one of three subfamilies within Crassulaceae and contains four genera. Despite previous efforts, the phylogeny of Kalanchoideae remains inadequately resolved with persistent issues including low support, unstructured topologies and polytomies. This study aimed to address two central objectives: (1) resolving the pending phylogenetic questions within Kalanchoideae by using organelle-scale ‘barcodes’ (plastomes) and nuclear data; and (2) investigating interspecific diversity patterns among Kalanchoideae plastomes.

Methods

To explore the plastome evolution in Kalanchoideae, we newly sequenced 38 plastomes representing all four constituent genera (Adromischus, Cotyledon, Kalanchoe and Tylecodon). We performed comparative analyses of plastomic features, including GC and gene contents, gene distributions at the IR (inverted repeat) boundaries, nucleotide divergence, plastomic tRNA (pttRNA) structures and codon aversions. Additionally, phylogenetic inferences were inferred using both the plastomic dataset (79 genes) and nuclear dataset (1054 genes).

Key Results

Significant heterogeneities were observed in plastome lengths among Kalanchoideae, strongly correlated with LSC (large single copy) lengths. Informative diversities existed in the gene content at SSC/IRa (small single copy/inverted repeat a), with unique patterns individually identified in Adromischus leucophyllus and one major Kalanchoe clade. The ycf1 gene was assessed as a shared hypervariable region among all four genera, containing nine lineage-specific indels. Three pttRNAs exhibited unique structures specific to Kalanchoideae and the genera Adromischus and Kalanchoe. Moreover, 24 coding sequences revealed a total of 41 lineage-specific unused codons across all four constituent genera. The phyloplastomic inferences clearly depicted internal branching patterns in Kalanchoideae. Most notably, by both plastid- and nuclear-based phylogenies, our research offers the first evidence that Kalanchoe section Eukalanchoe is not monophyletic.

Conclusions

This study conducted comprehensive analyses on 38 newly reported Kalanchoideae plastomes. Importantly, our results not only reconstructed well-resolved phylogenies within Kalanchoideae, but also identified highly informative unique markers at the subfamily, genus and species levels. These findings significantly enhance our understanding of the evolutionary history of Kalanchoideae.

INTRODUCTION

The stonecrop family (Crassulaceae) is the largest member of Saxifragales, with inclusion of ~1410 taxa from 34 genera (APG IV, 2016; Chang et al., 2021). Renowned for its leaf succulents, this family has a nearly cosmopolitan distribution (Van Ham and Hart, 1998; Mort et al., 2001). Additionally, the family Crassulaceae is widely recognized as a monophyletic group (Chase et al., 1993; Soltis et al., 1997). However, despite its easily recognizable nature, the presence of extreme diversity and frequent interspecific intergradation in morphology has posed challenges in elucidating intrafamilial relationships (Mort et al., 2001). Table 1 provides a summary of the subfamily divisions within Crassulaceae throughout its history. Historically, the stonecrop family has been divided into six (Berger, 1930), three (Thorne, 1983, 1992) and two subfamilies (Hart, 1995; Van Ham and Hart, 1998; Mort et al., 2001). Notably, Thiede and Eggli (2007) proposed the most widely accepted taxonomic system to date by incorporating combined data. Their classification ultimately grouped the seven major clades into three monophyletic subfamilies: Crassuloideae, Kalanchoideae and Sempervivoideae. Among these, the former two subfamilies were distinguished by two separate clades, the Crassula clade and the Kalanchoe clade, respectively, while the remaining clades constituted the last subfamily.

Table 1.

Subfamily division history of Crassulaceae in four selected classification systems.

Berger (1930)	Thorne (1983, 1992)	’t Hart (1995), Van Ham and ’t Hart (1998), Mort et al. (2001)	Thiede and Eggli (2007)
Two lineages	Three subfamilies	Two subfamilies	Three subfamilies
Six subfamilies		Seven clades	Five tribes
			Eight clades
Crassula-lineage:	Crassuloideae	Crassuloideae:	Crassuloideae:
Crassuloideae	Cotyledonoideae	Crassula clade	Crassula clade
Kalanchoideae	Incl. subfamily of Berger:	Sedoideae:	Kalanchoideae:
Cotyledonoideae	Kalanchoideae	Acre clade	Kalanchoe clade
Sedum-lineage:	Sedoideae	Aeonium clade	Sempervivoideae:
Echeverioideae	Incl. subfamilies of Berger:	Kalanchoe clade	Telephieae:
Sempervivoideae	Echeverioideae	Leucosedum clade	Hylotelephium clade
Sedoideae	Sempervivoideae	Sempervivum clade	Umbiliceae:
		Telephium clade	Rhodiola clade
			Semperviveae:
			Sempervivum clade
			Aeonieae:
			Aeonium clade
			Sedeae:
			Leucosedum clade
			Acre clade
Positions of four genera:	Positions of four genera:	Positions of four genera:	Positions of four genera:
Adromischus: in Cotyledonoideae	Adromischus: in Cotyledonoideae	Adromischus: in Kalanchoe clade	Adromischus: in Kalanchoideae
Cotyledon: in Cotyledonoideae	Cotyledon: in Cotyledonoideae	Cotyledon: in Kalanchoe clade	Cotyledon: in Kalanchoideae
Kalanchoe: in Kalanchoideae	Kalanchoe: in Cotyledonoideae	Kalanchoe: in Kalanchoe clade	Kalanchoe: in Kalanchoideae
Tylecodon: not erected yet	Tylecodon: not erected yet	Tylecodon: in Kalanchoe clade	Tylecodon: in Kalanchoideae

Great strides have noticeably been made in illustrating the taxonomic relationships within the subfamilies of Crassulaceae. A recent study focused on Crassuloideae, the only monogeneric subfamily, and identified three major clades supported by molecular data from 103 taxa across all sections of the genus Crassula L. (Bruyns et al., 2019). Sempervivoideae, the most species-rich and complex subfamily, has received extensive phylogenetic attention, resulting in the acceptance of a five-tribe classification based on multiple researches (Mort et al., 2002; Gontcharova et al., 2006; Yost et al., 2013; Zhang et al., 2014; Klein and Kadereit, 2015; Nikulin et al., 2016; de la Cruz-López et al., 2019; Liu et al., 2023). In particular, Messerschmid et al. (2020) conducted a 298-taxon phylogenetic analysis that successfully recovered all the six major clades in Sempervivoideae.

Far less studied, however, is the intergeneric relationships within Kalanchoideae, which currently comprises four genera, Adromischus Lem., Cotyledon L., Kalanchoe Adans. and Tylecodon Toelken (Thiede and Eggli, 2007). While some studies have examined the taxonomic affinities of individual genera, such as Kalanchoe (Gehrig et al., 2001) and Cotyledon (Mort et al., 2005), there is a lack of equivalent phylogenetic inferences focusing on Adromischus and Tylecodon. Nowell (2008) conducted the only phylogenetic work that included multiple Kalanchoideae genera, but the taxonomic relationships within many clades remain unresolved due to the limited resolution of datasets, despite sampling 125 taxa. Consequently, more effort is needed to gain insights into the evolutionary relationships in Kalanchoideae.

The generic treatment in Kalanchoideae was uncertain until Hart (1995) proposed the four-genus classification in 1995 (Table 1). Despite this instability, the close relationships among the four genera have been supported by cytology (Baldwin, 1938; Uhl, 1948) and floral vascular patterns (Quimby, 1971) throughout the taxonomic history of Crassulaceae. Cotyledon, currently the smallest genus in Kalanchoideae, has only ten species (Hart and Eggli, 1995; Eggli, 2012). Previously, it was considered a ‘hold-all’ of sympetalous Crassulaceae, but many groups have now been separated from it (Thiede and Hart, 1999; Eggli, 2012). Adromischus and Tylecodon have also experienced this segregation. Adromischus was erected in 1852 with nine transferred taxa from Cotyledon and two newly described taxa based on common morphological characteristics (Smith, 1939). Similarly, the initial description of Tylecodon included 22 taxa from Cotyledon, one from Adromischus and four new species (Tölken, 1978). Presently, Adromischus and Tylecodon comprise 32 and 50 accepted species, respectively (Borsch et al., 2020). Kalanchoe, the largest member of Kalanchoideae, has 170 currently accepted taxa (Borsch et al., 2020). The infrageneric treatment of Kalanchoe has been a subject of debate. One perspective integrated the genera Bryophyllum and Kitchingia into Kalanchoe, leading to a three-subgenus classification (Hamet, 1907; Jacobsen, 1954; Jacobsen, 1974; Boiteau and Allorge-Boiteau, 1995). Another viewpoint treated the two subgenera, particularly Bryophyllum, as a distinct subgenus from subg. Kalanchoe (Berger, 1930; Resende, 1956; Lauzac-Marchal, 1974; Byalt, 2008). Recent studies have supported the former opinion, dividing Kalanchoe into three subgenera (Gehrig et al., 2001; Chernetskyy, 2011; Smith and Figueiredo, 2018; Tian et al., 2021).

Notably, several pending questions have arisen from the existing phylogenetic hypotheses in Kalanchoideae. (1) Regarding Adromischus, Nowell (2008) found a moderate correspondence between unresolved molecular polytomies and Tölken’s five-section classification (Tölken, 1978) based on morphology. What are the internal branching patterns of Adromischus and how does its molecular phylogeny align with the earlier morphological classification? (2) The phylogenetic inference for Tylecodon, provided by Nowell (2008), exhibits overwhelmingly unstructured polytomies with weak support. Is this unresolved phylogeny considered as a soft polytomy and can it potentially be resolved with more sufficient data in our study? (3) Gehrig et al. (2001) conducted the only wide-sampling phylogenetic analysis of Kalanchoe. Despite some reconstructed structured clades, the relationships among and within these clades, especially between the subgenera Bryophyllum and Kalanchoe, remain largely unknown due to some unresolved polytomies. Can our study propose a phylogenetic hypothesis for the genus Kalanchoe with well-supported branching patterns? Additionally, is there any difference between our topology and the three-section division proposed by Boiteau and Allorge-Boiteau (1995)?

To address these issues, it is crucial to incorporate additional datasets, particularly plastome data, into molecular phylogenies. Plastid phylogenomics, also known as organelle-scale ‘barcodes’, has been shown to enhance phylogenetic resolution in various studies (Parks et al., 2009; Schulte et al., 2009; Yang et al., 2013; Han et al., 2022a). However, challenges remain, including the potential for erroneous affinities due to introgression events, such as chloroplast capture (Rieseberg and Soltis, 1991). To mitigate these misinterpretations, recommended strategies include comparing plastid and nuclear gene-based phylogenies and analysing combined datasets (Soltis et al., 2011; Wambugu et al., 2015; Doyle, 2022). Unfortunately, a well-resolved phylogeny with dense sampling and multiple loci is currently lacking for Kalanchoideae, highlighting the need for further efforts.

Plastomes not only demonstrate excellent phylogenetic performance but also hold great potential for plant lineage identification (Yang et al., 2013). Indeed, plastomes are characterized by conserved gene content and genomic organization (Palmer, 1985a; Ravi et al., 2008; Wicke et al., 2011; Wu et al., 2021). However, recent studies have focused on the diversifications among plant plastomes (Plunkett and Downie, 2000; Jansen et al., 2007; Park et al., 2018). Surprisingly, comparative plastomics research has revealed informative plastid markers at various taxonomic levels, such as plastid codon aversion motifs (CAMs), unique distributional patterns of rps19, and family-specific structures of plastomic tRNAs (pttRNAs) (Ding et al., 2022; Han et al., 2022ab, , c). Despite these findings, detailed and comparative exploration of Kalanchoideae plastomes is lacking, with only eight plastomes currently available. Therefore, further investigations are necessary to expand our understanding of taxonomic affinities within Kalanchoideae.

In this study, we present 38 newly reported plastomes of Kalanchoideae, covering all four constituent genera. Through comprehensive analyses, our work aims to achieve two main objectives: (1) resolving the pending phylogenetic questions by using both plastid and nuclear data, and (2) exploring the unique plastomic diversities among Kalanchoideae taxa, with a focus on genomic organizations, pttRNA structures and plastid CAMs. Ultimately, our research hopes to enhance our understanding of the phylogenetic affinities and evolutionary patterns within Kalanchoideae.

MATERIALS AND METHODS

Taxon sampling, DNA extraction and sequencing

A total of 38 taxa of Kalanchoideae were sampled, covering all four genera, with 11 from Adromischus, six from Cotyledon, 15 from Kalanchoe and six from Tylecodon. More specifically, for Adromischus, our sampling strategy encompassed representatives from all five sections classified by Tölken (1978). All subgenera of Kalanchoe were included except for Kalanchoe subg. Kitchingia, which comprises only two species (Smith et al., 2021). All 38 species were grown in the glasshouses of Anhui Normal University, with voucher information given in Supplementary Data Table S1. The specimens were deposited in the herbarium of Anhui Normal University. Total DNA isolation and library construction were carried out respectively using the Plant Genomic DNA kit (Tiangen, Beijing, China) and TruSeq DNA PCR-Free Library Prep Kit (Illumina, San Diego, CA, USA). Based on the generated libraries, raw reads were then obtained by the Illumina HiSeq X Ten platform (Illumina) with a paired-end strategy. All sequencing data have been submitted to the NCBI Sequence Read Archive (SRA) database under accession numbers of SRR27702550–SRR27702587.

Plastome assembly and annotation

The excision of adaptors and low-quality sequences was conducted by Trimmomatic v.0.39 (Bolger et al., 2014), following the parameter settings of Fonseca and Lohmann (2017). Quality checks were performed on all obtained clean reads using FastQC v.0.11.9 (Andrews, 2020). Subsequently, GetOrganelle v.1.7.5 (Jin et al., 2020) and NOVOPlasty v.2.7.1 (Dierckxsens et al., 2017) were employed for plastome assemblies. The reference choices for assembly were Adromischus maculatus (NC_068664) for Adromischus taxa, Kalanchoe fedtschenkoi (NC_053951) for Kalanchoe taxa, and Cotyledon tomentosa (NC_053948) for Cotyledon and Tylecodon taxa. The assembled results were scrutinized and initially annotated by GeSeq (Tillich et al., 2017) and CPGAVAS2 (Shi et al., 2019). After careful comparison, the annotations were then manually corrected through BLAST searches (Johnson et al., 2008) to improve their accuracy. In addition, a circular representation of final plastomes was provided by Chloroplot (Zheng et al., 2020).

Comparative plastomic analyses

To elucidate the heterogeneities among Kalanchoideae plastids, comprehensive comparative analyses were conducted on both the plastomic nucleotide and structural variations. First, EasyFig v.2.2.5 (Sullivan et al., 2011) was utilized to visualize the linear plastome comparison, based on the annotation files and BLASTn results. Then, the gene distributions at the IR boundaries were analysed in IRScope (Amiryousefi et al., 2018), and plotted manually after a careful check. Additionally, DnaSP v.6.12 (Rozas et al., 2017) was used to explore the sequence variations among Kalanchoideae plastomes. Following the parameters from Han et al. (2022a), the sliding window analyses calculated the nucleotide divergence (π) values. The hypervariable regions (HVRs) were measured based on the criteria described by Ding et al. (2022). The structural variations were also investigated for pttRNAs among our samples, using tRNAscan-SE v.2.0.3 (Lowe and Eddy, 1997) under default settings. Furthermore, as highly phylogenetically conserved features, the codon aversion motifs were estimated for each plastid gene by CAM v.1.02 (Miller et al., 2019).

Phylogenetic reconstruction

To gain further insights into the taxonomic affinities among Kalanchoideae taxa, phylogenetic estimation was performed with multiple datasets and methods. The outgroup selection, two taxa from Sempervivoideae, was determined according to the closest relationship between the two subfamilies (Thiede and Eggli, 2007). Two data sampling strategies were applied for phylogenetic inferences: (1) plastid coding sequences (CDSs) and (2) nuclear coding regions.

Retrieval of loci was separately conducted for plastid and nuclear genomes. All plastid CDSs were directly obtained from annotation files using DAMBE v.7.2.25 (Xia and Xie, 2001). In addition, nuclear sequence assemblies were performed using the intronless genes of Kalanchoe fedtschenkoi (GCA_002312845.1) as reference. The mapped reads gathered by HybPiper v.2.1.4 (Johnson et al., 2016) were used for sequence assemblies, which were performed using SPAdes v.3.15.1 (Anton et al., 2012) with a kmer of 21. For each nuclear gene, only the longest assembled contig was retained for each taxon. Each reference gene and the corresponding filtered contigs from all taxa were merged to form a distinct matrix. Then, the matrices with identity >4 % were used for further analyses. Sequence alignment by MAFFT v.7.505 (Katoh and Standley, 2013) and trimming by trimAl v.1.2 (Capella-Gutiérrez et al., 2009) allowed us to obtain the final nuclear sequence matrices.

Afterwards, the two phylogenetic datasets were individually aligned using MAFFT v.7.505 (Katoh and Standley, 2013), and concatenated by SequenceMatrix (Vaidya et al., 2011). Subsequently, phylogenetic reconstructions were separately performed through two algorithms, maximum-likelihood (ML) and Bayesian inference (BI). All ML analyses were conducted by RAxML v.8.2.12 (Stamatakis, 2014), with 50 runs, 1000 bootstrap replicates and a convergence-check command. For BI inference, the optimal model for each partition was evaluated using the Bayesian information criterion (BIC) by ModelTest-NG v.0.1.7 (Darriba et al., 2020). BI trees were then built using two runs of MrBayes v.3.2.7a (Ronquist et al., 2012), with four independent Markov chain Monte Carlo (MCMC) chains for each. After 10 million generations, convergence was assessed in Tracer v.1.7.1 (Rambaut et al., 2018).

RESULTS

Comparisons of the plastomic sizes and organization among Kalanchoideae

Our comparative analyses of the newly reported 38 plastomes of Kalanchoideae revealed both conserved and variable features. These plastomes generally shared high similarities in their quadripartite architecture, and GC and gene contents (Fig. 1A and Table 2). The overall GC content exhibited a narrow range of 37.6–38.3 %, with the standard deviation (s.d.) values for each genus not exceeding 0.05. Moreover, all the Kalanchoideae taxa investigated had the same number of plastid genes, comprising 85 protein-coding genes (PCGs), 44 RNA-coding genes (36 tRNA and eight rRNA genes), and four pseudogenes (Supplementary Data Table S2).

Fig. 1.

Comparisons of plastomic organizations and lengths among four constituent genera in Kalanchoideae. (A) Visualization of plastome comparison in linear format, with green boxes denoting pairs of IRs, arrows indicating genes and colours demonstrating various gene groups. (B) Boxplot of sizes (bp) in various plastome regions underscored considerable heterogeneities among four genera. TL, LSCL, SSCL and IRL denote total, LSC, SSC, and IR lengths, respectively.

Table 2.

Basic features of plastomes in 38 Kalanchoideae species, including accession number, genome size, gene number and GC content.

Taxon	Accession number	Size (bp)				Gene number					GC content (%)
		Total	LSC	SSC	IR	Total	PCGs	tRNA	rRNA	Pseudo	Total	LSC	SSC	IR	Coding region
Adromischus
A. caryophyllaceus	OR397134	150 914	83 086	16 936	25 446	133	85	36	8	4	37.7	35.7	31.5	43.0	37.7
A. cooperi	OR397135	150 367	82 776	16 801	25 395	133	85	36	8	4	37.7	35.8	31.5	43.0	37.7
A. cristatus	OR397136	150 438	82 751	16 819	25 434	133	85	36	8	4	37.7	35.8	31.5	43.0	37.7
A. filicaulis	OR397137	150 858	83 040	16 910	25 454	133	85	36	8	4	37.7	35.7	31.5	43.0	37.7
A. hemisphaericus	OR397138	150 446	82 565	16 953	25 464	133	85	36	8	4	37.7	35.8	31.5	43.0	37.7
A. leucophyllus	OR397139	150 802	83 011	16 915	25 438	133	85	36	8	4	37.7	35.7	31.5	43.0	37.7
A. maculatus	OR397140	150 539	82 657	16 954	25 464	133	85	36	8	4	37.7	35.7	31.5	43.0	37.7
A. marianae	OR397141	150 679	82 941	16 924	25 407	133	85	36	8	4	37.7	35.8	31.5	43.0	37.7
A. schuldtianus	OR397142	150 459	82 624	16 939	25 448	133	85	36	8	4	37.7	35.7	31.6	43.0	37.7
A. triflorus	OR397143	150 174	82 312	16 954	25 454	133	85	36	8	4	37.8	35.8	31.5	43.0	37.7
A. trigynus	OR397144	150 633	82 881	16 888	25 432	133	85	36	8	4	37.7	35.8	31.5	43.0	37.7
Cotyledon
C. cuneata	OR397145	149 683	81 934	16 893	25 428	133	85	36	8	4	38.3	36.4	32.4	43.3	38.0
C. eliseae	OR397146	150 100	82 333	16 881	25 443	133	85	36	8	4	38.2	36.3	32.4	43.3	38.0
C. orbiculata	OR397147	149 468	81 807	17 001	25 330	133	85	36	8	4	38.2	36.4	32.4	43.3	38.0
C. papillaris	OR397148	149 474	81 689	16 899	25 443	133	85	36	8	4	38.3	36.4	32.4	43.3	38.0
C. pendens	OR397149	149 758	81 911	17 007	25 420	133	85	36	8	4	38.2	36.3	32.4	43.3	38.1
C. tomentosa	OR397150	149 625	81 826	16 995	25 402	133	85	36	8	4	38.2	36.3	32.4	43.3	38.0
Kalanchoe
K. beharensis	OR397151	150 299	82 667	16 980	25 326	133	85	36	8	4	37.6	35.6	31.4	42.9	37.3
K. blossfeldiana	OR397152	150 511	82 539	17 016	25 478	133	85	36	8	4	37.6	35.6	31.6	42.9	37.7
K. delagoensis	OR397153	150 060	82 135	16 987	25 469	133	85	36	8	4	37.7	35.7	31.5	42.9	37.6
K. eriophylla	OR397154	151 159	83 183	17 004	25 486	133	85	36	8	4	37.6	35.6	31.5	43.0	37.7
K. fedtschenkoi	OR397155	149 941	82 015	17 012	25 457	133	85	36	8	4	37.7	35.7	31.4	42.9	37.6
K. gastonis-bonnieri	OR397156	149 817	81 927	17 026	25 432	133	85	36	8	4	37.7	35.7	31.5	42.9	37.7
K. millotii	OR397157	150 489	82 591	17 040	25 429	133	85	36	8	4	37.7	35.7	31.4	43.0	37.7
K. nyikae	OR397158	149 753	81 895	16 956	25 451	133	85	36	8	4	37.7	35.7	31.5	43.0	37.6
K. orgyalis	OR397159	150 805	82 959	16 998	25 424	133	85	36	8	4	37.7	35.7	31.4	42.9	37.6
K. prolifera	OR397160	149 585	81 694	16 977	25 457	133	85	36	8	4	37.7	35.7	31.6	42.9	37.7
K. rhombopilosa	OR397161	150 716	82 849	17 033	25 417	133	85	36	8	4	37.6	35.6	31.4	43.0	37.6
K. rotundifolia	OR397162	149 687	81 837	16 968	25 441	133	85	36	8	4	37.7	35.7	31.6	43.0	37.7
K. synsepala	OR397163	150 841	82 856	17 061	25 462	133	85	36	8	4	37.6	35.6	31.4	43.0	37.7
K. thyrsiflora	OR397164	149 793	81 926	16 985	25 441	133	85	36	8	4	37.7	35.7	31.5	42.9	37.6
K. tomentosa	OR397165	150 820	82 910	17 050	25 430	133	85	36	8	4	37.6	35.6	31.3	43.0	37.7
Tylecodon
T. buchholzianus	OR397166	150 121	82 266	17 005	25 425	133	85	36	8	4	38.2	36.3	32.2	43.2	38.0
T. paniculatus	OR397167	150 221	82 494	16 885	25 421	133	85	36	8	4	38.1	36.2	32.3	43.2	38.0
T. reticulatus	OR397168	150 273	82 475	16 950	25 424	133	85	36	8	4	38.1	36.2	32.2	43.2	38.0
T. schaeferianus	OR397169	150 220	82 429	16 953	25 419	133	85	36	8	4	38.1	36.2	32.2	43.2	37.9
T. striatus	OR397170	150 311	82 513	16 950	25 424	133	85	36	8	4	38.1	36.2	32.2	43.2	38.0
T. wallichii	OR397171	150 155	82 397	16 932	25 413	133	85	36	8	4	38.2	36.3	32.3	43.2	38.0

Despite this high level of conservation, we observed heterogeneities in plastome lengths. The overall sizes ranged from 149 468 (Cotyledon orbiculata) to 151 159 bp (Kalanchoe eriophylla), with a mean ± s.d. of 150 263 ± 452 bp. Each plastome consisted of one LSC region (81 689–83 183 bp), one SSC region (16 801–17 061 bp) plus two IR regions (25 326–25 486 bp) (Table 2).

Of note, we conducted more in-depth comparisons of plastome size on two levels (Fig. 1B and Supplementary Data Table S3). First, at the inter-region level, we assessed the size diversity of each region across the subfamily. By comparing the coefficients of variation (CV = s.d./mean), the LSC in Kalanchoideae plastomes was the most variable region (CV = 0.0054), followed by SSC (0.0035) and IR (0.0013). Second, at the inter-generic level, Adromischus plastomes had the largest average size of 150 574 bp, followed by Kalanchoe (150 285 bp), Tylecodon (150 217 bp), and Cotyledon (149 685 bp). In terms of total plastome size diversity, Kalanchoe and Tylecodon were the most (CV = 0.0034) and least (CV = 0.0005) diverse groups, respectively. The most variable sizes were found in the LSC regions in Kalanchoe (CV = 0.0060), while the IR regions in Tylecodon (CV = 0.0002) exhibited the smallest range of sizes.

To explore the key factors influencing plastome size, we performed correlation analyses between the total plastome length (TL) and several elements, including the lengths of LSC (LSCL), SSC (SSCL), IR (IRL) and intergenic sequences (IGSL). Across all the sampled Kalanchoideae taxa, plastome sizes were significantly correlated with LSCL (r = 0.98, P < 0.01), moderately correlated with IGSL (r = 0.49, P < 0.01), and not correlated with SSCL (r = 0.02, P > 0.05) or IRL (r = 0.26 P > 0.05) (Supplementary Data Table S4). We also conducted correlation analyses for each of the four constituent genera. The TLs of Cotyledon and Tylecodon, the two closest-related genera, shared the same pattern of significant correlation with only LSCL. By contrast, differences were detected in Adromischus and Kalanchoe, with the former strongly correlated with LSCL and IGSL, and the latter significantly correlated with LSCL, SSCL and IGSL.

Variations of gene distributions at IR boundaries among Kalanchoideae

The gene content at the quadripartite junctions of Kalanchoideae plastomes can be assigned to seven types (Fig. 2). Therein, the gene distributions at LSC/IRb and IRa/LSC of all involved taxa exhibited strong consistency, with the rps19 gene occupying 110 bp of IRb (type A) and the trnH-rps19 cluster overlapping by 3 bp in IRa (type G), respectively. The second conserved junction, IRb/SSC, featured a 44-bp expansion of IRb to ndhF in all plastomes (type B), except for Cotyledon pendens, which showed a 333-bp contraction (type C). After BLAST searches and alignment with other taxa, the ndhF gene in C. pendens was found to possess an ~380-bp premature termination, caused by a 5-bp insertion of ‘TATAG’.

Fig. 2.

Different patterns of gene distributions at all boundaries of Kalanchoideae plastomes. For the four boundaries, a total of seven types were summarized: type A at LSC/IRb; types B and C at IRb/SSC; types D, E and F at SSC/IRa; and type G at IRa/LSC. All these types were marked on the corresponding nodes in the given phylogenetic tree, with different colour boxes.

Notably, informative diversities were revealed in the gene content at SSC/IRa. All investigated Adromischus, Cotyledon and Tylecodon plastomes shared a 1089-bp occupation of ycf1 in IRa (type D), with the only exception being A. leucophyllus, which had 1083 bp (type E). In the case of Kalanchoe plastomes, two groups could be recognized by two different types. Group A featured in type D, while ycf1 genes of group B uniquely extended 1080 bp into IRa (type F).

Assessment of plastomic hypervariable regions

The objective of this study was to examine the nucleotide variations in the plastomes of the subfamily Kalanchoideae, using π values to measure sequence polymorphism (Fig. 3A and Supplementary Data Table S5). As indicated in Table 3, the genus Adromischus harboured the most HVRs from 11 loci, with π falling in the range 0.01300–0.02630, followed by Cotyledon (ten HVRs, 0.01478–0.03031), Kalanchoe (eight HVRs, 0.02603–0.05279) and Tylecodon (six HVRs, 0.01622–0.04461). These 35 HVRs were distributed into 27 types of annotations, predominantly (96.3 %) found in the SC regions, with a minor percentage (3.7 %) in the IR regions (Fig. 3B). Additionally, several loci were shared among multiple genera, including one HVR observed in three genera, and three HVRs each detected in two genera. Most notably, the ycf1 gene was identified as a hypervariable region in all four genera.

Fig. 3.

Assessment of hypervariable regions among Kalanchoideae plastomes. (A) Overall sequence diversities of plastomes for four genera, with Cotyledon and Tylecodon shown above the x-axis, Adromischus and Kalanchoe shown below the x-axis. Each genus is denoted by a different coloured line. (B) Comparisons of assessed hypervariable loci among four genera. Above: stacked histogram indicating various loci from various genera, with the height of boxes denoting π values. Below: Venn diagram depicting both genus-unique and shared loci for the four genera. Note that the red line highlights a shared hypervariable region by all four genera, ycf1.

Table 3.

Identified hypervariable regions (HVRs) of plastomes among the four genera of Kalanchoideae.

HVR	Annotations	π	Mutation sites	Region length (bp)
Adromischus
1	trnH-GUG-psbA	0.016125	64	832
2	rps16-trnQ-UUG-psbK-psbI	0.026301	559	2733
3	trnS-GCU-trnR-UCU	0.013000	27	751
4	rpoB-trnC-GCA	0.015523	91	1055
5	petN-psbM	0.015373	76	1023
6	trnE-UUC-trnT-GGU-psbD	0.013580	20	1178
7	psbZ-trnG-GCC-trnfM-CAU-rps14	0.016313	125	1169
8	psbJ-psbL-psbF-psbE-petL	0.023777	101	1063
9	ccsA-ndhD	0.019180	77	802
10	ndhD-psaC	0.013613	70	1046
11	ycf1	0.013210	30	600
Cotyledon
1	rps16-trnQ-UUG	0.015280	43	1024
2	trnS-GCU-trnR-UCU-atpA	0.026250	125	1644
3	rpoB-trnC-GCA	0.014780	39	825
4	trnF-GAA-ndhJ-ndhK	0.015280	42	1576
5	rbcL	0.016275	44	818
6	psbL-psbF-psbE-petL	0.030312	181	1436
7	rps18-rpl20-rps12	0.016500	38	833
8	psbB	0.015110	18	632
9	ccsA-ndhD	0.017055	46	811
10	ycf1	0.017483	71	1000
Kalanchoe
1	trnK-UUU-rps16	0.027945	136	993
2	rps16-trnQ-UUG-psbK-psbI-trnS-GCU-trnR-UCU	0.052793	613	2368
3	trnE-UUC-trnT-GGU-psbD	0.028075	131	908
4	rps14-psaB	0.026030	54	748
5	petL-petG-trnW-CCA-trnP-UGG-psaJ	0.033610	137	1166
6	ndhI-ndhA	0.026710	55	696
7	ycf1	0.026293	194	1036
8	ycf1-trnN-GUU-trnR-ACG-rrn5-rrn4.5-rrn23	0.030111	489	2687
Tylecodon
1	rps16-trnQ-UUG-psbK	0.044612	330	1867
2	rpoB-trnC-GCA	0.016220	19	700
3	rbcL-accD	0.021627	109	1004
4	ycf4-cemA-petA	0.019037	73	1055
5	psbJ-psbL-psbF-psbE-petL	0.026695	133	1233
6	ycf1	0.016560	20	609

Among the ycf1 genes, further analysis revealed highly informative and lineage-specific indels. As depicted in Fig. 4, four indels could serve as unique markers for the involved Adromischus taxa, including indel 2 (6 bp), indel 3 (9 bp), indel 7 (9 bp) and indel 8 (9 bp). Moreover, the 6-bp indel 4 could distinctly differentiate the Cotyledon and Tylecodon species from the other two genera. Also, indels 1 (9 bp) and 5 (30 bp) could collectively represent one clade in the dichotomy of Kalanchoe.

Fig. 4.

Illustration of observed lineage-specific indels in alignments of the ycf1 gene among 38 Kalanchoideae taxa. Adromischus uniquely possessed four indels: indel 2 (a 6-bp deletion), indel 3 (a 9-bp insertion), indel 7 (a 9-bp insertion) and indel 8 (a 9-bp deletion). Indel 4, a 6-bp insertion, was shared by all Cotyledon and Tylecodon species. In addition, one major clade of Kalanchoe featured in indel 1 (a 9-bp deletion) and indel 5 (a relatively large insertion). Indels 6 and 9 were identified at two deeper nodes, respectively.

Diversities of putative secondary structures of pttRNAs

Structural predictions of 36 pttRNAs in each Kalanchoideae taxon displayed ten non-typical structures (Fig. 5). These pttRNAs could be assigned to four types according to the specific locations of structural variations: (1) expanded ANC-loop (trnV-UAC and trnL-UAA), (2) bulky V-arm (trnS-UGA, trnS-GCU, trnS-GGA, trnY-GUA and trnL-CAA), (3) an additional loop in the Ψ-arm (trnC-GCA) and (4) an excess loop in the AC-arm (trnA-ACG, and trnT-UGU).

Fig. 5.

Non-typical secondary structures detected from Kalanchoideae pttRNAs, with colour nucleotides marking key areas. (A) trnV-UAC. (B) trnL-UAA. (C) trnS-UGA. (D) trnS-GCU. (E) trnS-GGA. (F) trnY-GUA. (G) trnL-CAA. (H) trnC-GCA. (I) trnA-ACG. (J) trnT-UGU.

Furthermore, comparative analyses of the ten pttRNAs within the subfamily Kalanchoideae revealed distinct interspecific heterogeneities in three pttRNAs: trnS-GCU, trnS-UGA and trnT-UGU (Fig. 6). Remarkably, genus-unique structures were identified for two genera, Adromischus and Kalanchoe. Adromischus taxa were characterized by a 5ʹ-GG-3ʹ at the AC-loop of trnT-UGU, which differed from the other three genera. Additionally, all investigated Kalanchoe taxa possessed a V-loop of 5ʹ-CUUGUUCA-3ʹ in trnS-UGA, while the other genera featured a 5ʹ-UUUGUUCA-3ʹ V-loop.

Fig. 6.

Significant structural diversities of pttRNAs among Kalanchoideae taxa. Right: structural variations (SVs) of three informative pttRNAs. The complete stem and loop were plotted in each type A, while other subsequent types were simplified with only variable loops. Left: illustration of lineage-specific SVs with each type denoted after the corresponding taxon.

Inter-subfamilial differences in pttRNA structures were further investigated by comparing the structures of all 147 currently available plastomes of Crassulaceae. Importantly, the two genus-unique features mentioned above were confirmed in this larger-scale analysis (Supplementary Data Table S6). For example, all trnT-UGU sequences from Crassulaceae taxa, except for Adromischus, had AC-loops of 5ʹ-CC-3ʹ or 5ʹ-CU-3ʹ. Most notably, the striking unique structure for the subfamily Kalanchoideae was detected in trnY-GUA. Specifically, the V-loops of this particular pttRNA were exclusively 5ʹ-AAAAU-3ʹ in Kalanchoideae, while they were 5ʹ-AUA-3ʹ in Crassuloideae and Sempervivoideae.

Abundant unique markers from codon aversion patterns of Kalanchoideae

Among all the available Kalanchoideae plastomes, the codon aversion patterns were identified and compared across 53 CDSs over 300 bp. In total, the CAM was detected in 48 CDSs, excluding five that were longer than 2000 bp (rpoB, rpoC1, rpoC2, ycf1 and ycf2). The comparison showed a high variability among the Kalanchoideae codon aversions, with only five genes (accD, ndhB, psaA, psbB and rps12) showing strictly conserved CAM patterns (Supplementary Data Table S7). Significantly, 24 CDSs underscored a total of 41 lineage-specific unused codons ( Table S7). Among these, the CAM from each of the three CDSs, atpE, rps2 and ndhC, could clearly differentiate three certain groups ( Table S8). The remaining genes exhibited distinct CAM patterns, with 16 (66.67 %) showing a unique CAM for only one group, and five (20.83 %) shared by two groups.

More importantly, it was striking to observe the substantial presence of lineage-specific CAM patterns that spanned all four constituent genera of Kalanchoideae. As presented in Fig. 7 and Supplementary Data Table S8, Adromischus taxa harboured the most specific patterns of 11 CAMs, followed by (Cotyledon + Tylecodon) with eight unique motifs, Kalanchoe (six unique CAMs), Cotyledon (four unique CAMs), Tylecodon (two unique CAMs) and (Cotyledon + Kalanchoe) with only one specific motif. Interestingly, the CAM patterns within Kalanchoe taxa distinctly depicted two subgroups, which could be recognized by five and four characteristic CAMs, respectively.

Fig. 7.

Venn diagram in network format presented considerable highly informative codon aversion motifs among 24 CDSs in Kalanchoideae plastomes. A total of five lineages, i.e. Adromischus (A), Cotyledon (C), two major clades of Kalanchoe (K1 and K2), and Tylecodon (T), are marked with small circles in the centres of networks. Green and purple large circles denote lineage-specific and shared CAMs, respectively. Within large circles, the number placed before the codon represents the corresponding gene, as defined in the genes list.

Phylogenetic inferences in Kalanchoideae

Our phylogeny strongly supported the monophyly of Kalanchoideae based on the dataset of plastomic CDSs from all available Saxifragales (Supplementary Data Figure S1). We then performed further phyloplastomic reconstructions within Kalanchoideae. Using a 68 039-bp alignment of 79 plastomic CDSs, the ML and BI inferences produced highly consistent and well-resolved cladograms at both inter- and infra-generic levels (Fig. 8A). Importantly, our sampled taxa indicated that all four genera were monophyletic. A distinct dichotomy existed between the Adromischus clade and another clade comprising the remaining three genera. Within the latter clade, Kalanchoe was strongly supported as sister to (Cotyledon + Tylecodon) [bootstrap support (BS) = 100 %, posterior probability (PP) = 1.0]. Moreover, the deep-node phylogenetic relationships within the four genera were individually described as follows.

Fig. 8.

Phylogenetic inferences for Kalanchoideae. (A) Phyloplastomic tree based on a 79-PCG matrix with the ML and BI methods, with dots in different colours denoting BS and PP values. Note that circles adjacent to Adromischus and Kalanchoe taxa highlight the classifications by previous studies of Tölken (1978) and Gehrig et al. (2001), respectively. (B) Phylogram obtained based on 1054 nuclear genes of Kalanchoe taxa. Bootstrap values are given at the nodes (*BS = 100 %). The section Bryophyllum and Eukalanchoe marked on the right side of the diagram correspond to the classification of Gehrig et al. (2001).

• (1) Adromischus. Within this genus, two main clades were clearly resolved. Notably, our molecular phylogenies moderately corresponded to the morphological five-section classification of Tölken (1978). As shown in Fig. 8A, three taxa from section Longipedunculati (5) formed one distinct clade with strong support (BS = 100 %, PP = 1.0). This clade was found to be the sister group to another distinct clade, containing the remaining species from the complicated sections, i.e. sections Adromischus (1), Boreali (2), Brevipedunculati (3), Incisilobati (4) and Longipedunculati (5). Significantly, therein, the non-monophyly was identified for all five sections, except for section 3 with only one taxon sampled. For example, our results placed A. leucophyllus (section 5) in a sister clade to all the other taxa of section 5. Besides, A. hemisphaericus, a section 1 taxon, was clearly nested within a subclade of section 4 while not clustered with A. filicaulis (section 1).

• (2) Cotyledon. Our phyloplastomic reconstruction distinctly depicted two fully supported clades (BS = 100 % and PP = 1.0) in this genus. Within one clade, C. orbiculata clustered with the species pair of C. tomentosa and C. pendens. The second clade comprised three taxa, with C. cuneata sister to (C. papillaris + C. eliseae).

• (3) Tylecodon. Species of Tylecodon formed two well-resolved clades. One clade was composed of T. striatus and T. reticulatus (BS = 100 %, PP = 1.0). Within the other clade, a highly supported pairing between T. buchholzianus and T. schaeferianus was reconstructed (BS = 100 %, PP = 1.0), which was robustly sister to (T. paniculatus + T. wallichii) with full support.

• (4) Kalanchoe. All relationships among the involved Kalanchoe taxa were successfully recovered. The topology obtained clearly underscored two clades. The first clade harboured a species pair, K. synsepala + K. eriophylla, strongly supported and forming a sister relationship with the other five taxa (BS = 100 %, PP = 1.0). Of the five species, three formed a highly supported cluster (BS = 93 %, PP = 1.0), which was clearly sister to (K. orgyalis + K. rhombopilosa) (BS = 100 %, PP = 1.0). Within the second clade, K. blossfeldiana was placed as sister to two structured subclades. One subclade comprised K. nyikae + (K. rotundifolia + K. thyrsiflora), and another consisted of (K. prolifera + K. gastonis-bonnieri) + (K. delagoensis + K. fedtschenkoi). Each relationship received strong support (100 % BS and 1.0 PP). Most interestingly, according to the sections classification of Gehrig et al. (2001), the section Eukalanchoe taxa were found to be non-monophyletic, as four of its members clustered with section Bryophyllum taxa (as marked in Fig. 8A).

Additionally, to increase the reliability of our phylogenetic inferences, we assembled nuclear genes for all sampled Kalanchoideae taxa using K. fedtschenkoi (GCA_002312845.1) as the reference genome. However, only our Kalanchoe data generated sufficient nuclear loci to establish a robust cladogram, which may be due to the substantial genetic differences in nuclear genes among genera within the family Crassulaceae. As a result, we focused the nuclear phylogeny analyses exclusively on the interspecific relationships within Kalanchoe. Through our strict filtering strategy, we obtained a total of 1054 nuclear genes (data available at https://doi.org/10.6084/m9.figshare.24945594.v1). The nuclear phylogram in Fig. 8B showed that our Kalanchoe samples were divided into two clades, each with the same constituent taxa as with the phyloplastomic results. Nonetheless, the branching patterns varied at several deep nodes. For example, differing from the plastomic phylogenies, the four Eukalanchoe taxa that grouped with Bryophyllum formed a distinct branch in the nuclear phylogenies, namely K. nyikae, K. thyrsiflora, K. rotundifolia and K. blossfeldiana. Furthermore, K. thyrsiflora was more closely related to K. nyikae (in the nuclear tree) rather than to K. rotundifolia (in the chloroplast tree). Most importantly, the non-monophyly of section Eukalanchoe was highly supported by both the plastomic and nuclear trees.

DISCUSSION

This study reported 38 newly sequenced plastomes of Kalanchoideae. Through comprehensive analyses, we aimed to deepen our understanding of the taxonomic affinities within Kalanchoideae. Plastomic comparisons were further conducted to provide additional molecular evidence for interspecific differences, such as genome organizations, gene distributions at IR junctions, secondary structures of pttRNAs, codon aversion motifs and phylogenetic inferences. Most importantly, our findings significantly enhance current insights into the phylogeny and molecular evolution across Crassulaceae.

Plastomic size diversities among Kalanchoideae

Known as an iconic characteristic of plants, the chloroplast has been widely recognized to retain its own unique DNA. The overwhelming majority of plant plastomes fall within a constrained size range of 120–160 kb (Palmer, 1985b). Importantly, plastomes are less prone to recombination, especially compared to the variable size range of mitochondrial and nuclear genomes (Alexeyev et al., 2004; Greilhuber et al., 2005). These features of plastomes are advantageous for exploring diversities and the evolution of genome size (Zheng et al., 2017).

Currently, plastome length variations have been attributed to the differences in three main aspects: (1) intergenic regions (Shahid Masood et al., 2004; Tang et al., 2004; Wu et al., 2011), (2) IR regions (Glöckner et al., 2000; Lin et al., 2003) and (3) gene content (Wolfe et al., 1992; Wakasugi et al., 1994). A most recent study on Poaceae plastomes (Wang et al., 2023) proposed additional influencial factors, containing repeat sequences and LSCL. In particular, for closely related taxa, IGSL was considered a crucial factor for TL variations (Zheng et al., 2017). Our study indicated that LSCL and IGSL mainly accounted for the total size variations of Kalanchoideae plastomes. Interestingly, IRL and gene number, two important factors, had little or no influence on plastome sizes. This may be due to the close relationships and highly conserved nature of the involved Kalanchoideae taxa. In fact, at different sampling scales, the analyses for the driving factors of plastome size variations could yield conclusions of uncertain generality. Thus, to improve our understanding of the evolution of plastome size in Kalanchoideae, more samples are needed for further explorations.

Implications from the gene distribution patterns at IR junctions

As classical DNA secondary structures in plastome, IRs have been found to contribute significantly to influencing genetic stability (Lobachev et al., 2007) and gene splicing (Smith et al., 2000; Lacomme et al., 2003). Moreover, the variations of IRs could commonly occur by contraction/expansion over evolutionary time, which have been applied as phylogenetic characteristics (Zhu et al., 2016; Weng et al., 2017; Dong et al., 2018; Park et al., 2018; Wei et al., 2021).

For Kalanchoideae plastomes, the seven types of gene content flanking IRs revealed important evolutionary implications. Type A, detected in all involved Kalanchoideae taxa, reaffirmed the family-specific marker of Crassulaceae, a 110-bp expansion of IRb into rps19, as proposed by our previous research (Ding et al., 2022; Han et al., 2022a, b). More importantly, the unique type F cluster is a valid phylogenetic group, and it provides strong support for the relationships between the taxa in group B of Kalanchoe. Type C was found only in Cotyledon pendens, characterized by a 380-bp deletion in the ndhF gene, which has not been previously reported in Crassulaceae. However, it has been observed in other angiosperms, such as Gymnospermium (Berberidaceae, Ranunculales) (Song et al., 2022) and Anathallis (Orchidaceae, Asparagales) (Mauad et al., 2019). The premature termination of ndhF is a well-known plastomic genetic evolutionary event in angiosperms (Fu et al., 2017; Wang et al., 2020; Zhang et al., 2020). The ndhF gene is essential for photosynthesis (Martín and Sabater, 2010), and its deletion can result in the gene becoming non-functional, a process known as pseudogenization (Mauad et al., 2019). Overall, the findings of this study shed light on the importance of IR junction patterns in lineage identification and understanding evolutionary relationships among Crassulaceae.

Evolutionary insights and unique markers from plastomic HVRs of Kalanchoideae

To our knowledge, the uneven accumulation of mutations in the plastome, with some regions experiencing a greater number of mutations than others, sheds light on the heterogeneous nature of plastome evolution. These regions of high mutational activity are known as HVRs (Henriquez et al., 2020; Shahzadi et al., 2020). The π values, which serve as a widely used tool for inferring HVRs, can reflect the evolutionary rates of certain sequences (Cui et al., 2019; Liu et al., 2020). Notably, the findings of the present study support the suggestion that HVRs are more common in SC regions of the plastome than in IR regions. This is clearly evident in the Kalanchoideae plastomes, where the number of identified HVRs in SC regions overwhelmingly outnumbered those in IR regions. This remarkable finding strongly implies faster evolutionary rates in SC regions than in IRs, which aligns with the conclusions of numerous previous studies (Curtis and Clegg, 1984; Perry and Wolfe, 2002; Guo et al., 2017; Liu et al., 2018; Xu et al., 2019). Similarly, the Kalanchoe plastomes generally featured higher π values than the other three genera, suggesting this genus might evolve at a faster rate. The ycf1 gene, known for its phylogenetic reliability and considered a promising plastomic barcoding marker, has been identified as a hotspot in abundant taxa (Dong et al., 2015; Kong et al., 2021; Zhang et al., 2021). In our work, the ycf1 gene was the only shared HVR among the investigated Kalanchoideae taxa. Most interestingly, the highly informative indels of ycf1 genes could serve as specific markers for various lineages, providing new evidence for the capability of ycf1 in distinguishing taxa. In summary, these findings contribute to a deeper understanding of the evolutionary process of Kalanchoideae.

Striking unique markers from secondary structures of certain pttRNAs

tRNAs are crucial in the process of translating mRNA into proteins. Since their first identification in 1956 (Holley et al., 1965), it has been established that their cloverleaf-like secondary structure is largely conserved across tRNAs (Zhao et al., 2021). These stable structures are vital for the interplay between tRNAs and various molecules (Lorenz et al., 2017). However, some flexibility is also required for optimal interactions, primarily provided by the D-loop and V-loop (Goodenbour and Pan, 2006; Zhong et al., 2021). Notably, previous research has suggested two general types of tRNA structures based on the variable-region size (Brennan and Sundaralingam, 1976). The most common, type I, harbours a small V-loop (4–5 nt), while type II, found in leucine, serine or tyrosine (Dock-Bregeon et al., 1989), has a larger variable region with a loop (3–5 nt) and additional stem (3–7 bp).

Our investigations of Kalanchoideae predicted type II structures in five pttRNAs from the three isotypes mentioned above. Notably, the large variable region has been found to have a vitally important role in the catalytic reaction (Redlak et al., 1997). According to a model proposed by Dock-Bregeon et al. (1989), the type II V-arm can contribute to a more planar L-shape, facilitating the binding of tRNA and ribosome.

We also detected non-typical structures in several type I pttRNAs of Kalanchoideae. Among all cellular RNA families, tRNAs undergo the most post-transcriptional modifications, which have critical physiological implications (Zhang et al., 2022). For instance, the expanded 9-nt ANC-loop found in trnV-UAC and trnL-UAA within our samplings is inferred to affect the translocation step size, potentially leading to a 3- or 4-nt reading frame (Curran and Yarus, 1987). Additionally, an extra loop was detected at the AC-arm in trnA-ACG and trnT-UGU. The AC-arm structure, a hallmark of tRNAs, may play a crucial role in maintaining catalytic activity (Redlak et al., 1997). Importantly, sequence variations of the AC-arm are thought to significantly impact catalytic efficiency, mainly by directing the 3ʹ end toward the catalytic core (Meinnel et al., 1993).

The structures of pttRNAs, in addition to their important physiological significance, also harbour highly informative diversities for demonstrating taxonomic affinities. Our recent findings have suggested that several lineage-unique structures could serve as specific markers at both high and low taxonomic levels. For example, two Macaronesian genera (Crassulaceae) (Han et al., 2022a) and several families of Saxifragales (Han et al., 2022b) have been identified. For pttRNAs of Kalanchoideae, our structural comparisons recognized one subfamily- and two genus-specific (Adromischus and Kalanchoe) characteristics. The potential of these structures to be unique markers was confirmed by a denser sampling analysis. However, the general significance of these findings requires further assessment due to limited data. Collectively, our findings underline the striking differences in pttRNA structures among Kalanchoideae and provide novel insights into Kalanchoideae and Crassulaceae phylogenies.

Contributions of codon aversion to understanding Kalanchoideae relationships

Codon usage patterns in certain genes have significant impacts on mRNA folding and translation efficiency, thereby indirectly influencing gene expression (Quax et al., 2015). Additionally, codon aversions, first described by Miller et al. (2017), are unevenly distributed within genes and are informative and conserved across phylogenies (Miller et al., 2017, 2019, 2020). This makes them a valuable tool for gaining insight into taxonomic relationships. Our previous research has demonstrated the important role of codon aversion in understanding these relationships, as exemplified by Han et al. (2022c) on Bletilla (Orchidaceae). Their study revealed that B. foliosa, recently proposed as the new genus Mengzia (Huang et al., 2022), harboured a distinctly different CAM compared to other Bletilla taxa.

In the present study, the identification of 41 specific non-used codons provided unique markers at both inter- and intra-generic levels and clearly illustrated the phylogenetic relationships within Kalanchoideae. For instance, the finding that Adromischus had the most unique CAMs can be linked to its basal sister relationship to other genera. Furthermore, the second highest number of motifs was found in (Cotyledon + Tylecodon), indicating a close relationship between these two genera, which aligns with our phylogenetic inferences. To date, the mechanisms that dominate the phylogenetic conservation of codon aversion remain unclear. However, two possible explanations have been proposed: selection on translation efficiency, and neutral processes (e.g. GC-bias gene conversion) (Miller et al., 2019). In conclusion, our findings provide new molecular evidence for the relationships within Kalanchoideae and strongly support the use of codon aversion in phylogenomic and evolutionary analyses.

Insights into the pending phylogenetic questions of Kalanchoideae

The subfamily Kalanchoideae has received limited phylogenetic scrutiny at the inter-generic level in previous work. Despite efforts to explore the inter-specific relationships among the four constituent genera (Gehrig et al., 2001; Mort et al., 2005; Nowell, 2008), challenges such as poor support, unstructured topologies and polytomies have persisted. Here, our study is committed to addressing these pending phylogenetic questions of Kalanchoideae taxa, as summarized in the Introduction.

For Adromischus, our results successfully clarified the positions of three taxa (A. caryophyllaceus, A. hemisphaericus and A. triflorus), which were previously classified as polytomic branches by Nowell (2008), with relatively high support. Notably, we identified striking non-monophyly in four of the five Adromischus sections classified by Tölken (1978), confirming the findings of Nowell (2008). Despite this progress, more sampling efforts for both molecular data and taxa are still needed to achieve a comprehensive understanding of this genus.

Within the subfamily Kalanchoideae, Cotyledon and Tylecodon have been found to be closely related based on both cytology (Uhl, 1948) and secondary chemistry (Van Wyk and Winter, 1995). As expected, our cladogram clearly shows a sister relationship between the two genera. In Cotyledon, our phyloplastomic reconstruction using 79 PCGs generated a well-established dichotomy, which is highly consistent with earlier inferences from ITS and three plastid loci (Nowell, 2008). This congruence further supports the reliability of the dichotomy within the genus. Moreover, in earlier studies, overwhelmingly unresolved polytomies pervaded Tylecodon phylogenies, leading to poor resolution (Nowell, 2008). However, in our study, two distinct clades were recognized within Tylecodon, and the positions of all six involved taxa were well resolved with robust support.

As mentioned earlier, the classification of Kalanchoe has undergone a rather complicated history of subdivision. Despite the current widespread acceptance of the three-section treatment after decades of debate, a clearly supported cladogram within Kalanchoe remains unavailable. Significantly, this study’s analyses of both plastomic and nuclear data identified two major clades among the included Kalanchoe taxa. Of particular note, two interesting findings emerged when comparing these results to the previous subdivisions.

First, in earlier studies, a controversy converged upon the position of K. thyrsiflora, an African taxon. Gehrig et al. (2001) nested it within the branch of Malagasy section Kitchingia, while Jacobsen (1981) treated it in African section Eukalanchoe. Here, our phylogenetic result provides strong evidence for the latter viewpoint, with a distinct branch of all three African taxa (K. thyrsiflora, K. nyikae and K. rotundifolia) receiving full support.

Second, and most strikingly, our results clearly indicate the non-monophyly of section Eukalanchoe in both the plastomic and nuclear trees – an observation not previously reported. The underlying reasons for such deviations have yet to be ascertained, but the insights of Eggli (2012) may provide some possible explanations. As he proposed, while some members of Eukalanchoe are relatively stable, others are more variable, probably due to their rapid radiation history. Section Bryophyllum is also considered to be a group with rapid evolutionary development. These factors may have made earlier morphological classifications difficult, heterogeneous and artificial (Hamet, 1907; Berger, 1930; Boiteau and Allorge-Boiteau, 1995).

In this study, we have shown that the current section Eukalanchoe is not monophyletic, as evidenced by both phylogenies and the distinct differences between the two clades in IR junction distributions, ycf1 indels, tRNA structures and CAMs. However, considering the lack of morphology and population data, our current results are not yet sufficient to propose a revised subdivision circumscription. To reach a better understanding of Kalanchoe, further investigations with more samples are still essential.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following.

Figure S1: ML tree of Saxifragales inferred from the concatenated 79 plastomic CDSs from 301 samples, with the BS values shown on the branches. Table S1: Voucher information of plant materials in the present study. Table S2: List of genes found in the 38 new Kalanchoideae plastomes in this study. Table S3: In-depth comparison of the diversities of plastome size in different regions among Kalanchoideae taxa. Table S4: Correlation analyses between plastome sizes and lengths of various regions across the Kalanchoideae. Table S5: Evaluations of sequence diversities among the chloroplast genomes of four genera in Kalanchoideae. Table S6: Large-scale comparisons among the 2nd structures of trnS-UGA, trnT-UGU and trnY-GUA across the whole family Crassulaceae. Table S7: Identified informative aversion codons among Kalanchoideae plastid genes. Table S8: Distinguished lineages by CAMs among Kalanchoideae plastid genes.

FUNDING

This work was supported by the Opening Foundation of National Engineering Laboratory of Soil Pollution Control and Remediation Technologies, and Key Laboratory of Heavy Metal Pollution Prevention & Control, Ministry of Agriculture and Rural Affairs (Grant no. NEL&MARA-003); the Basic Research Program of Natural Science Foundation of Jiangsu Province of China (Grant no. BK20211078); and the Doctoral Enhancement Program of Suzhou Polytechnic Institute of Agriculture (Grant no. BS2101).