Structural Plastome Evolution in Holoparasitic Hydnoraceae with Special Focus on Inverted and Direct Repeats.

Plastome condensation during adaptation to a heterotrophic lifestyle is generally well understood and lineage-independent models have been derived. However, understanding the evolutionary trajectories of comparatively old heterotrophic lineages that are on the cusp of a minimal plastome, is essential to complement and expand current knowledge. We study Hydnoraceae, one of the oldest and least investigated parasitic angiosperm lineages. Plastome comparative genomics, using seven out of eight known species of the genus Hydnora and three species of Prosopanche, reveal a high degree of structural similarity and shared gene content; contrasted by striking dissimilarities with respect to repeat content [inverted and direct repeats (DRs)]. We identified varying inverted repeat contents and positions, likely resulting from multiple, independent evolutionary events, and a DR gain in Prosopanche. Considering different evolutionary trajectories and based on a fully resolved and supported species-level phylogenetic hypothesis, we describe three possible, distinct models to explain the Hydnoraceae plastome states. For comparative purposes, we also report the first plastid genomes for the closely related autotrophic genera Lactoris (Lactoridaceae) and Thottea (Aristolochiaceae).

Plastome condensation during adaptation to a heterotrophic lifestyle is generally well documented and lineage-independent evolutionary models have been derived. Controversially discussed are the putative evolutionary trajectories resulting in structures of plastomes close to a proposed “minimal plastome.” Three evolutionary models are discussed to explain the large repeat content of newly assembled, highly condensed plastomes of holoparasitic Hydnoraceae. One of the models assumes the regain of the quadripartite plastome structure after ancestral loss, which has only been reported for two other seed plant lineages to date. An alternative model discusses lineage-specific inverted repeat condensation and translocation instead. Furthermore, we report the loss of all but one intron in the Hydnoraceae plastomes.

Introduction

The plastid chromosome of land plants is, on average ∼150 kb in length (Daniell et al. 2016; Weng et al. 2017) with extreme size differences between the two ends of the size spectrum. The largest plastid genome to date has been reported for Pelargonium (∼242 kb, Weng et al. 2017) and the smallest for the heterotrophic genus Pilostyles (∼11 kb, Bellot and Renner 2016). Typical plastid chromosomes show a quadripartite structure. Two single-copy regions, namely the large (LSC) and small (SSC) single-copy region are separated by two copies of an inverted repeat (IR) (Palmer 1985; Mower and Vickrey 2018). Around 100–120 unique genes are encoded on plastomes of autotrophic plants, which are mostly related to photosynthesis, carbon fixation, or involved in their transcription and translation (Wicke et al. 2011). Plastomes are considered highly conserved with respect to gene order, content and structure (Palmer 1985), with many extremes and deviations from the norm reported for mycoheterotrophic, and parasitic plants including: (1) record-setting smallest known plastomes sizes exclusively found in parasitic and mycoheterotrophic plants (Bellot and Renner 2016; Petersen et al. 2018; Su et al. 2019), (2) the most extreme plastome A + T nucleotide bias reported in parasitic Balanophora (Balanophoraceae) with only ∼12% G + C nucleotides (Su et al. 2019) compared with an average of 34–39% (Cai et al. 2006), or (3) the plastome of Pilostyles aethiopica (Apodanthaceae) with a total of only five encoded genes (Bellot and Renner 2016). Hypothetically culminating in a complete loss of the plastid chromosome in endoparasitic Rafflesiaceae (Molina et al. 2014; Cai et al. 2021).

Apart from the abovementioned differences in genome size, nucleotide composition, and gene content, structural diversity such as changes of gene order and IR loss can be observed. The loss of one IR copy is not uncommon in highly condensed plastomes (Wicke et al. 2013; Bellot and Renner 2016; Arias-Agudelo et al. 2019; Su et al. 2019; Jost et al. 2020), but not exclusive to the heterotrophic lifestyle (Palmer et al. 1987; Lavin et al. 1990; Wicke et al. 2011; Zhu et al. 2016; Jin et al. 2020). Comparative research on IR-containing and IR-lacking plastomes suggests a positive influence of the IR presence on structural plastome stability (Birky and Walsh 1992; Maréchal and Brisson 2010). IR losses or extreme reduction, on the other hand, seem to correlate with an increased number of rearrangements in some lineages (Palmer and Thompson 1982; Palmer et al. 1987; Wu et al., 2011; Wu and Chaw, 2016). However, not all IR-lacking plastomes show a higher degree of structural changes (Jansen et al. 2008; Sanderson et al. 2015; Blazier et al. 2016), casting doubts on IRs being a major factor preventing plastomes from structural changes. Current studies suggest a putative involvement of short repeat sequence accumulation as an additional factor driving homologous recombination (Wu et al. 2011; Mower and Vickrey 2018). As much as the adaptation to a heterotrophic lifestyle seems to follow lineage-independent patterns (Barrett and Davis 2012; Wicke et al. 2013; Naumann et al. 2016; Wicke et al. 2016; Graham et al. 2017), differences among the spectrum of highly condensed plastomes can be observed.

Understanding heterotrophic lineages and exploring evolutionary events at the cusp of a putative minimal plastome is essential to complement and expand current knowledge. One of these families is Hydnoraceae, with only two plastomes published to date (Naumann et al. 2016; Jost et al. 2020). The plastomes of Hydnora visseri and Prosopanche americana are among the most reduced genomes with only 24 and 26 unique genes, respectively (Naumann et al. 2016; Jost et al. 2020). Their gene order is mostly conserved, along with their genome size and G + C content (Jost et al. 2020). However, P. americana lacks a typical IR but instead contains a direct repeat (DR) with a gene set similar to the IR in H. visseri, indicating putative homology (Jost et al. 2020). In both cases, the repeat gene contents represent a highly condensed version of what is commonly found in IRs of autotrophic flowering plants (Zhu et al. 2016).

Hydnoraceae is a family of holoparasitic root parasites consisting of the two genera Hydnora Thunb. (Thunberg 1775) and Prosopanche de Bary (de Bary 1868) (fig. 1). Hydnora and its eight known species (Bolin et al. 2018) are found exclusively in the Old World, in Central and South Africa and Madagascar (Musselman and Visser 1989). The seven species of genus Prosopanche are found in the New World, in Central and South America (Gómez-Pignataro and Gómez-Laurito 1981; Musselman and Visser 1989; De Carvalho et al. 2021). Hydnora and Prosopanche have an estimated crown group age of ∼55 Ma (Naumann et al. 2013), but could potentially be as old as the separation of the African and South American continent. To date, little is known about the infrageneric relationships of Hydnora and Prosopanche. Tree reconstruction on a set of eight Hydnoraceae species using two phylogenetic markers (ITS, rpoB) recovered distinct clades (Bolin 2009), yet one of the two markers (rpoB) was later shown to be missing from the complete plastomes (Naumann et al. 2016; Jost et al. 2020).

Fig. 1.

The only aboveground plant structures of Hydnorraceae are the flowers. Flowers of (A) Prosopanche americana, (B) Prosopanche bonacinae, (C) Hydnora visseri, and (D) Hydnora triceps. The latter species is one of the only two flowering plants from which it is known that they exclusively flower below ground.

Here, we compare complete plastomes of seven species of Hydnora, six newly sequenced, and three species of Prosopanche, two newly sequenced. For comparative reasons we additionally assembled plastomes of close photosynthetic relatives (Thottea sumatrana, Aristolochiaceae and Lactoris fernandeziana, Lactoridaceae). Using comparative approaches, we studied the extent of plastome condensation and structural variation. In combination with a resolved species-level phylogenetic hypothesis of Hydnoraceae, we explore putative plastome evolutionary trajectories, in particular the inverted and DR evolution.

Results

Size and Gene Content of Hydnoraceae Plastomes and Their Relatives

The plastomes of all ten studied species of Hydnoraceae show a high degree of similarity with respect to gene content as well as genome size (table 1, supplementary fig. 1, Supplementary Material online). The plastomes of the Prosopanche species are on average larger than the ones of the Hydnora species. Within Prosopanche, P. panguanensis has the largest (28,262 bp) and P. americana the smallest (28,191 bp) plastome. Plastomes of Hydnora species vary to a higher degree with respect to genome size, with H. esculenta having the smallest (24,479 bp) and H. triceps having the largest (28,658 bp) (table 1).

Table 1

Plastome Size and Nucleotide Composition of Hydnoraceae in Comparison to Outgroup Lineages

Accession	Total Length (bp)	LSC (bp)	SSC (bp)	IR (bp)	Total GC (%)	GC LSC (%)	GC SSC (%)	GC IR (%)
Aristolochia fimbriata	160,529	90,080	19,583	25,433	38.5	36.9	33.3	43.4
Thottea sumatrana	158,610	88,665	19,053	25,446	38.3	36.3	33.4	43.5
Lactoris fernandeziana	159,739	81,254	18,529	29,978	38.5	37.1	32.6	42.2
Prosopanche bonacinae	28,246	—	—	—	23.9	—	—	—
Prosopanche americana	28,191	—	—	—	20.4	—	—	—
Prosopanche panguanensis	28,262	—	—	—	20.8	—	—	—
Hydnora esculenta	24,479	—	—	—	24.0	—	—	—
Hydnora arabica	24,790	15,734	6,740	1,158	23.1	24.4	20.1	23.0
Hydnora abyssinica	24,672	15,793	6,947	966	24.1	25.5	21.2	23.0
Hydnora africana	26,957	22,233	1,386	1,669	22.3	23.0	22.1	17.4
Hydnora triceps	28,658	22,126	1,420	2,556	21.8	23.1	21.3	16.2
Hydnora longicollis	26,714	22,329	1,539	1,423	23.4	23.9	23.3	19.7
Hydnora visseri	27,233	22,751	1,550	1,466	23.4	24.0	23.0	19.0

note.—Size and %GC differences are estimated based on complete plastomes as well as on plastome regions (LSC, SSC, and IR).

The two Hydnoraceae genera share a set of genes, consisting of 15 genes coding for ribosomal subunits (rps, rpl), one gene coding for the beta subunit of the acetyl CoA carboxylase (accD), two genes involved in protein translocation through plastid membranes (ycf1, ycf2), four ribosomal RNAs (rrn4.5, rrn5, rrn16, rrn23), as well as three tRNAs (trnE-UUC, trnI-CAU, trnfM-CAU). Prosopanche species contain two additional tRNAs, that is, trnW-CCA and trnY-GUA. The Hydnoraceae plastomes have lost the genes generally containing group IIa introns with the exception of rpl2, rpl16, and rps12. However, the only remaining intron in the Hydnoraceae plastomes is intron 1 of the transspliced rps12.

All Hydnoraceae protein-coding genes show open reading frames (ORFs), except for rps19 in P. americana and rps7 in H. visseri. For the latter, however, a site was identified where potential posttranscriptional editing could create an in-frame start codon. Generally, the Hydnora rps7 5′-region is highly divergent between accessions regarding length and nucleotide composition. In contrast, the 3′ end is highly conserved, including H. visseri. In total, the rps7 ORF in Hydnora varies greatly in length (546–1,065 bp from H. triceps to H. longicollis, respectively) with the majority of length differences stemming from the 5′ region. The longest ORFs are more than twice the size of autotrophic relatives (1,065 bp in H. longicollis compared with 468 bp in Aristolochia fimbriata, L. fernandeziana, and T. sumatrana). The rps19 gene shows a high degree of sequence similarity in Prosopanche although no ORF could be identified in P. americana owing to a missing in-frame start codon. In Prosopanche bonacinae, the ORF is expanded, partially overlapping with the ycf1 gene.

The secondary structure predictions for the Hydnoraceae tRNA genes result in typical cloverleaf folds for all tRNAs (supplementary fig. 3, Supplementary Material online). Additionally, the anticodon as predicted by sequence similarity is seen in all cases, except for trnW-CCA in P. americana which showed ACA instead.

The nucleotide composition of the Hydnoraceae plastomes is highly biased toward A and T nucleotides, compared with outgroup taxa (table 1). The average GC content (%) of the latter is 38.4, compared with an average of 23.1 in Hydnora (21.8 in H. triceps—24.1 in Hydnora abyssinica) and 21.7 in Prosopanche (20.4 in P. americana—23.9 in P. bonacinae).

For comparative reasons, we assembled plastomes of two close photosynthetic relatives of Hydnoraceae. Thottea sumatrana (Aristolochiaceae) and L. fernandeziana (Lactoridaceae) plastomes show the typical quadripartite structure consisting of two single-copy regions (LSC and SSC) separated by two copies of an IR. For details, refer to table 1 and supplementary fig. 2, Supplementary Material online.

Relationships Within Hydnoraceae

Constrained and unconstrained phylogenetic tree reconstructions using maximum likelihood recovered identical relationships within holoparasitic Hydnoraceae, including full support (100 bootstrap support, BS) for monophyletic Prosopanche and Hydnora (fig. 2 and supplementary figs. 4 and 5, Supplementary Material online). The constrained reconstruction is based on generic relationships published in Jost et al. (2021). Within Prosopanche, P. bonacinae is recovered as sister to Prosopanche clade I (P. americana, P. panguanensis) with full BS support. The sister relationship within the latter clade scored BS 93–100. Within Hydnora, H. esculenta is recovered as sister to all other species (BS 100). Hydnora clade I (H. africana, H. triceps, H. longicollis, H. visseri) is recovered sister to Hydnora clade II (H. arabica, H. abyssinica). Within Hydnora clade I (BS 99–100), H. longicollis plus H. visseri are sister to H. africana plus H. triceps (BS 100, fig. 2, supplementary figs. 4 and 5, Supplementary Material online). Extremely elevated branch lengths compared with the outgroup sampling are found leading to Hydnoraceae as well as to Hydnora and Prosopanche specifically (supplementary fig. 4, Supplementary Material online). The longest branch lengths within Hydnoraceae are found leading to Hydnora clade II, whereas they are generally much longer between all parasitic members compared with their autotrophic relatives (supplementary figs. 4 and 5, Supplementary Material online).

Fig. 2.

Tree reconstruction recovers distinct clades within Hydnoraceae. Phylogenetic tree reconstruction using RAxML shown as phylogram. Only Hydnoraceae taxa are shown. For full phylogenetic relationships, refer to supplementary figs. 4 and 5, Supplementary material online. The tree reconstruction is based on an 83 gene matrix, applying a gene partition approach and without restriction of relationships. Support values are based on 1000 bootstrap replicates and are shown for nodes with

Structural Hydnoraceae Plastome Comparison

Structurally, all species of Hydnora contain an IR with varying gene content (fig. 3), except for H. esculenta, which has lost its IR (fig. 3, supplementary fig. 1, Supplementary Material online). The three sampled Prosopanche species do not contain an IR, instead two copies of a DR were identified (fig. 3, supplementary fig. 1, Supplementary Material online). The latter repeats contain a complete copy of trnI along with a partial copy of rpl2 and, in P. bonacinae additionally a partial ycf2 copy (figs. 3 and 4). The latter repeat is identical to the IR gene content found in Hydnora clade I (ycf2-trnI-rpl2). The IR of H. africana, unlike the other Hydnora clade I repeats, additionally contains the 3′-portion of trnE (figs. 3 and 4). In general, the Hydnoraceae repeat content is a highly condensed version of what is found in Aristolochiaceae and Lactoridaceae (fig. 4, supplementary fig. 2, Supplementary Material online). PCR and Sanger sequencing have confirmed the IRs and their positioning for H. abyssinica and H. arabica.

Fig. 3.

Linear plastome comparison highlights structural similarities and dissimilarities within Hydnoraceae (Hydnora [H.] and Prosopanche [P.]). Linear comparison of plastome diversity is shown with genes depicted as arrows, with their direction indicating their orientation in relation to one another. The colors correspond to different gene groups: accD (purple), tRNA (blue), rps (light brown), rpl (dark brown), ycf (gray), and rRNA (red). IRs are highlighted by green boxes, DRs by light blue boxes. Synteny between neighboring plastomes shows rearrangements.

Fig. 4.

Hydnoraceae repeats show varying gene content and size. Gene content of the Hydnora IR and Prosopanche DR copies in comparison to the IR of autotrophic Thottea sumatrana as representative for the autotrophic sister lineages. Taxa with identical repeat gene content are grouped. Repeats are not drawn to size; repeat lengths are displayed as number of base pairs.

Multiple events of gene inversion and translocation are observed in Hydnoraceae. The rps7–rps12 region in Prosopanche is inverted compared with the majority of Hydnora species (H. esculenta and Hydnora clade I) as well as the outgroup (figs. 3 and 4). In Hydnora clade II, rps7 is found in opposite orientation compared with the remainder of Hydnora accessions. Lastly, the rps12 exon 2 is inverted in H. abyssinica compared with H. arabica, but identical with the other Hydnoraceae species (fig. 3). For the sister group H. longicollis and H. visseri, trnI is inverted compared with the other sampled members of Hydnoraceae (fig. 3) as well as to outgroup taxa L. fernandeziana and T. sumatrana. Additional structural differences to the remainder of the sampling can be found in Hydnora clade II (fig. 3). Here, the rpl14 gene is translocated and inverted, placing it in the accD and rps3 spacer. Rpl16–rps3 has been subjected to a similar event, whereas also containing the second IR copy in Hydnora clade II. Although neighboring one another, both events are to be considered independent from each other, based on gene order.

Simple and Tandem Repeat Analyses

Mononucleotide repeats were the most common type in all studied taxa, accounting for at least half of the total SSRs (supplementary table 1, Supplementary Material online). In Hydnoraceae, the more complex repeat types account for much less of the total number of SSRs, with penta- and hexanucleotide repeats either missing or having the smallest share. In Hydnora arabica, no tetra-, penta, or hexanucleotide repeats have been found (supplementary table 1, Supplementary Material online).

Two regions containing tandem repeats in the Hydnoraceae plastomes are noteworthy. First, the shared tandem repeats in the rps3–rpl16 spacer. Some of which seem to be homologous based on sequence alignments, but often vary in length and copy number, others being unique for specific taxa. With Hydnoraceae having lost the rpl16 intron, the length of the spacer in Hydnora often equals the length between rps3 and rpl16 exon 2 in Aristolochiaceae and Lactoridaceae. In Prosopanche, the spacer is much shorter, yet tandem repeats were also identified here. Second, tandem repeats in the Hydnora rps7 5′-region have been identified. Based on sequence similarity, the repeat sequences seem to be homologous, yet their exact length and copy number varies between accessions, resulting in different rps7 ORF lengths. In Prosopanche, rps7 contains two distinct tandem repeats, both in the center of the gene as well as near the 3′ end. The repeat copy numbers here are more conserved, resulting in less variation in gene length. Generally, the number of genes affected by tandem repeats is slightly higher in Hydnoraceae compared with Aristolochiaceae and Lactoridaceae.

In nearly all accessions, tandem repeats make up a significantly larger portion of the plastomes compared with SSRs (with the exception of L. fernandeziana and H. esculenta, supplementary table 1, Supplementary Material online).

Discussion

In Hydnoraceae, a high degree of plastome similarities within and among the two genera are reported for many categories such as genome size, gene order, as well as gene and GC content, despite the split of Prosopanche and Hydnora ∼55 Ma (crown age 36–74 Ma, Naumann et al. 2013). However, major differences are observed with respect to inverted and DR structure and content among Hydnoraceae species.

Possible Evolutionary Trajectories Resulting in the Current Plastome States

A large fraction of genes retained in the Hydnoraceae plastomes is commonly found in the IRs of autotrophic relatives (supplementary figs. 1 and 2, Supplementary Material online, Naumann et al. 2016; Jost et al. 2020). Although this fraction is generally high among highly condensed plastomes (Bellot and Renner 2016; Lim et al. 2016; Roquet et al. 2016; Arias-Agudelo et al. 2019; Su et al. 2019; Yudina et al. 2021), comparatively those genes appear to occupy the most space in the Hydnoraceae plastomes. In contrast to the studied species of Hydnora (excluding H. esculenta), plastomes of other highly condensed holoparasitic plants generally lack their quadripartite structure, due to IR loss (Bellot and Renner 2016; Roquet et al. 2016; Arias-Agudelo et al. 2019; Su et al. 2019). The plastomes of mycoheterotrophic Thismia on the other hand contain two IR copies, occupying a large portion of the total plastomes sequence (Lim et al. 2016; Yudina et al. 2021). Currently, there are no highly condensed plastomes for which Prosopanche-like DR structures have been reported. Mapping evolutionary events to the Hydnoraceae phylogeny leads to three possible scenarios (models) for plastome evolution (fig. 5), based on the presence or absence of distinct repeats, their gene content, as well as their differing positioning (fig. 6). Comparison of gene order and orientation between Hydnoraceae and the outgroup taxa (Thottea plastome shown as representative) reveals that the same repeat copy (hereafter named IRb/DRb) is the only one affected by either translocation or loss, whereas the second copy (hereafter named IRa/DRa) positioning is homologous to autotrophic relatives (fig. 6). Assuming homology between the plastomes of the outgroup and Hydnoraceae, one would expect IRb/DRb to be located between rps19 and ycf1 (considering gene losses in comparison to the outgroup, fig. 6). However, the repeats are found at distinct positions for each clade, resulting from independent events of translocation (fig. 6).

Fig. 5.

Three models to explain repeat (IR, DR) presence in Hydnoraceae. Phylogenetic tree reconstruction of the Hydnoraceae with mapped evolutionary events, separated by model. IR1 corresponds to the putative ancestral IR of the Hydnoraceae MRCA, IR2 to the IR found in Hydnora clade II, and IR3 to the IR of Hydnora clade I. DR corresponds to the DR found in Prosopanche. Model-dependent events on the branches are highlighted in red.

Fig. 6.

Linear comparison of Hydnoraceae repeat content and positions compared with autotrophic Thottea. Linear comparison of the distinct repeat sequences of P. bonacinae, H. longicollis (exemplary for Hydnora clade I) and H. arabica (exemplary for Hydnora clade II) compared with Thottea as an example for the autotrophic relatives. Repeat copies are named based on the Thottea IRs and their putative origin, for example, Hydnora IRb is assumed homologous to the Thottea IRb, whereas Hydnora IRa’ is a repeat directly derived from Hydnora IRa. Dotted lines represent Hydnoraceae plastome sequence corresponding to the Thottea IRs, as well as their expected positioning based on gene order. The green area highlights the actually repeated sequence in the Hydnoraceae plastomes and their corresponding region on the Thottea plastomes. Arrowheads indicate repeat copy orientations. Black boxes on the Thottea plastomes indicate genes missing from the Hydnoraceae plastomes.

The contraction model (figs. 5 and 6) assumes the Hydnoraceae repeats (IR or DR) to be condensed copies of the IR present in autotrophic relatives. A total of three independent events of repeat contraction and a single repeat loss have to be assumed in Hydnoraceae (fig. 5), considering the different gene content of the Hydnora clade I and II IRs (figs. 4 and 6). Additionally, this model requires three independent events of repeat copy translocation (IRb/DRb) among Hydnoraceae taxa. The Prosopanche DRs therefore result from the inverse insertion of the translocated IRb copy (fig. 6). Small fluctuations of the inverted repeat borders are not uncommon (Zhu et al. 2016), however, the degree of contraction as assumed under this model has not been reported so far to our knowledge.

The second model, which requires fewer steps compared with the previous model and is, therefore, more parsimonious, is called the ancestral loss model. This model has to assume an IR loss (IRb) before Hydnoraceae diversification (fig. 5), followed by three independent, secondary repeat regain events (DRa’ in Prosopanche and two IRa’ in Hydnora depending on the clade, fig. 6). These events have to be considered independent based on the significantly different plastome structures and repeat gene contents. One regain is postulated in Prosopanche (DR) and one in each Hydnora clade (IR2 and IR3) (fig. 5). In this model, the repeat gene content and the repeat position, for each independent regain event, are simply by chance. In case of duplication events, chances are high that parts of the originally IR-located sequences are duplicated, given that ∼40–55% of the Hydnoraceae plastomes comprise content that is commonly found in IRs of autotrophic relatives (fig. 6). In general, secondary IR gain after an ancestral loss is extremely rare and has only been reported in two lineages so far. Twice in legumes (Choi et al., 2019; Wu et al. 2021) and once in heterotrophic Parasitaxus (Qu et al. 2019). Mechanisms underlying the regain have not been studied extensively, although the involvement of double-strand break repair mechanisms has been discussed. The repair of double-strand breaks via formation and solution of Holliday junctions and synthesis-dependent strand annealing are proposed to trigger potential IR regain (Choi et al. 2019).

The following repeated loss model is as parsimonious as the latter model (fig. 5). In contrast to the aforementioned ancestral loss model, the repeat content in Hydnora clade I (IR) and Prosopanche is assumed homologous as the independent gain of repeat structures with (nearly) identical gene content is unlikely. In both lineages, the repeat consists of trnI, and partial copies of ycf2 and rpl2 (fig. 4). However, Prosopanche clade I lost the partial ycf2 copy and the IR in H. africana contains an additional partial copy of trnE (fig. 4). An additional translocation step is required to explain the inverted second repeat copy and positioning in Prosopanche (fig. 6 and Jost et al. 2020). The model further requires the IR loss on the branch leading to H. esculenta and a loss leading to Hydnora clade II. The latter is followed by a secondary IR gain in Hydnora clade II as the IRs of Hydnora clades I and II are clearly distinct from one another (fig. 5). The IR loss in Hydnora clade II could have been the trigger for the observed structural alternations (rps3–rpl16, rpl14, rps7, fig. 5).

Under parsimony criteria, the ancestral loss and repeated loss model are to be favored as they imply fewer steps than the contraction model. However, the contraction model might be intuitively favored, given that the Hydnoraceae repeat content could represent highly condensed versions of the IRs of autotrophic relatives. Although the second repeat copies are never in the location expected for an IRb contraction (fig. 6).

A model-independent observation is related to the Hydnora clade II IR copy (IRa′/IRb, fig. 6) contained within translocated rps3–rpl16 region. This has possible implications for the IR creation within this clade. If the second repeat copy was integrated into this cluster before its translocation, it was so as a DR. Translocation and subsequent inversion could therefore have created the Hydnora clade II IR based on an ancestral DR. In general, we did not find any other structural rearrangements near the translocated repeat copies (fig. 3).

Features Putatively Affecting Plastome Stability

Hydnora and Prosopanche share an identical set of genes, which is slightly expanded in Prosopanche (additional trnW and trnY copies). Considering the evolutionary time since the switch to full heterotrophy (assumed before lineage speciation), nonessential genes are assumed lost from the plastome. Therefore, the retained set of genes represents the assumed lineage-dependent essential gene set of Hydnoraceae. Lineage-dependent essential gene sets encode for gene products that cannot be otherwise compensated via nuclear import (Schelkunov et al. 2015; Wicke et al. 2016). Retention of tRNAs W and Y in Prosopanche could represent lineage-specific special needs. However, the functionality of trnW in P. americana cannot be assumed, as the secondary structure prediction shows an ACA instead of the expected CCA anticodon (supplementary fig. 3, Supplementary Material online). Whether the putatively nonfunctional status of trnW can be translated to the other Prosopanche species is uncertain. The loss of tRNAs Y and W in Hydnora, as opposed to a gain in Prosopanche, is supported by the general consensus that gene loss during heterotrophic plastome condensation is presumed to be irreversible and therefore one-directional (Barrett and Davis 2012; Graham et al. 2017; Yudina et al. 2021).

Apart from the varying large repeats (IRs, DRs), Hydnoraceae reveal a strikingly low degree of structural rearrangements (fig. 3). We did not find a correlation between IR loss and an increase of the latter (Prosopanche, H. esculenta, fig. 3), as was shown for certain other lineages (Palmer and Thompson 1982; Palmer et al. 1987). Whether the DR in Prosopanche exhibits the same criteria as known for typical IRs is uncertain. In IR-lacking, autotrophic lycophytes, however, IR-like copy-dependent DNA repair is acting upon the retained DR copies, along with lower substitution rates of genes contained within (Mower et al. 2019). IR/DR presence in combination with a lack of large dispersed repeats could help explain the structural conservation as seen in Hydnoraceae. Large dispersed repeats have been shown to coincide with more frequent rearrangements (Mower and Vickrey 2018). Structural plastome stasis, a hypothesized phase of decelerated evolution that followed the phase of accelerated evolution during the major phase of plastome condensation and adaptation to parasitic lifestyle may be considered as additional driver in maintaining the genomic stability (Naumann et al. 2016; Wicke et al. 2016).

Hydnoraceae Plastomes have Lost All But One Intron

Three genes generally known to contain intron sequences are retained on the highly reduced plastomes of Hydnoraceae (rpl2, rpl16, and rps12, Jansen et al. 2007). The loss of the rpl2 intron in a single Hydnora and Prosopanche species was previously reported (Naumann et al. 2016; Jost et al. 2020). Intron losses are not exclusive to plants following a heterotrophic lifestyle, but are generally rather rare (Downie et al. 1991, 1996; McPherson et al. 2004; McNeal et al. 2009; Su et al. 2019). With the additional Hydnoraceae accessions sequenced, we are now confident to report that the remaining two genes have also lost their group IIa introns. As a result, only the transspliced rps12 intron remains. The intron losses did not affect the presence of ORFs for these genes, indicating putative functionality. The loss of all group IIa introns is in line with the loss of maturase K, which is essential for their splicing (Liere and Link 1995; McNeal et al. 2009).

Materials and Methods

Plant Material, DNA Extraction, and Sequencing

Genomic DNA of Hydnoraceae as well as T. sumatrana (Aristolochiaceae) were isolated and sequenced via whole genome shotgun sequencing (supplementary table 2, Supplementary Material online). Hydnoraceae taxa included: H. abyssinica, H. africana, H. arabica, H. esculenta, H. longicollis, H. triceps, P. bonacinae, and P. panguanensis. All DNAs were extracted from silica dried plant material using the protocol of Doyle and Doyle (1987), modified to include RNAse A (Thermo Scientific, Waltham, MA, USA) treatment (10 mg/ml). DNA concentration and quality were measured using a Qubit 3 Fluorometer (Thermofisher Scientific, Waltham, MA, USA) and the Agilent Technologies 12-capillary Fragment Analyzer (Agilent Technologies, Santa Clara, CA, USA) using the Genomic DNA 50 kb kit. Hydnoraceae sequencing was carried out on rapid mode HiSeq 2,500 flow cells (Illumina, San Diego, CA, USA) as 150 bp paired-end reads for 2 × 150 cycles targeting about 50 million reads per sample. Sequencing of Thottea aimed for 5 million 300 bp paired-end reads instead.

Data Mining from Public Repositories

To expand the sampling within Piperales, as well as to sample outgroup lineages from magnoliids and ANA grade taxa for phylogenetic tree reconstruction, the NCBI nucleotide data base was mined for complete plastomes. In addition, sequence raw data for L. fernandeziana (Lactoridaceae) from the Kew Tree of Life data release 0.1 were downloaded from the Sequence Read Archive (ERX4143470). The complete sampling list for this study can be found in supplementary table 2, Supplementary Material online.

Raw Data Assembly and Plastome Reconstruction

Raw read data were assembled using the de novo assembly function in CLC Genomics Workbench (v. 11.0, Qiagen, MD, USA), allowing for automatic calculation of optimal word and bubble sizes for each sample. The assembly was blasted (BLASTn, default settings with an E-value cut-off at 1e−10) for contigs of potential plastid origin, using a database containing >40 complete plastomes representing major angiosperm lineages. The latter contigs were imported into Geneious (v. 11.1.5, Biomatters, Ltd., New Zealand) and the correctness of the assembly verified by analyzing the corresponding read-mapping using Tablet (v. 1.21, Milne et al. 2013).

To verify specific plastome regions of interest, and bridge assembly gaps, primers were derived from the flanking regions (supplementary table 3, Supplementary Material online). Both strands of the resulting PCR products were then Sanger sequenced. Manual scaffolding to create complete, circular plastomes was done in Geneious (v. 11.1.5, Biomatters, Ltd., New Zealand). Gene annotation was done using Geneious (v. 11.1.5, Biomatters, Ltd., New Zealand) based on published, closely related plastomes of H. visseri (NC_029358, Naumann et al. 2016), P. americana (MT075717, Jost et al. 2020), and Piper nigrum (NC_034692, Zhang et al. unpublished). Annotations were manually inspected and adjusted where necessary based on ORFs. Circular and linear plastomes were visualized using OGDraw (Greiner et al., 2019), the latter additionally using EasyFig (v. 2.2.5, Sullivan et al. 2011). Secondary structure of tRNAs was predicted using tRNAscan-SE (v. 2.0, Chan et al. 2021) (supplementary fig. 3, Supplementary Material online). For the plastome visualizations, the genome start for all Hydnoraceae was set to the rps4-accD spacer.

Phylogenetic Analyses

Single gene alignments for 83 plastid loci (79 protein-coding genes, 4 ribosomal RNAs) were created using the MAFFT algorithm (v. 7.450, Katoh et al. 2002; Katoh and Standley 2013), implemented in Geneious (v. 11.1.5, Biomatters, Ltd., New Zealand) and manually adjusted in AliView (v. 1.20, Larsson 2014). All genes were concatenated using SequenceMatrix (v. 1.8, Vaidya et al. 2011) and inferences were calculated using RAxML (v. 8.2.12, Stamatakis 2014) implemented on Cipres Science Gate (Miller et al. 2010) applying a gene partitioning approach, after estimating the best-fitting nucleotide substitution model using PartitionFinder2 (Lanfear et al. 2014, 2017). In addition to the 83 plastid gene matrix, inferences were estimated based on a plastid matrix reduced to the genes present in the Hydnoraceae plastomes (21 gene matrix). Node support was calculated by 1,000 BS replicates. Inferences for the concatenated 83 loci data set were estimated twice, once without predefined tree topology and once by restricting the relationships of Piperales genera to the ones estimated in Jost et al. (2021). Tree files were then visualized using TreeGraph 2 (Stöver and Müller 2010).

Repeat Analysis

Two types of repeat sequences were studied in the Hydnoraceae plastomes as well as in autotrophic relatives (A. fimbriata, L. fernandeziana, T. sumatrana) (supplementary table 1, Supplementary Material online). Simple sequence repeats were analyzed using the MIcroSAtellite identification tool (MISA, Beier et al. 2017) setting the minimum number of repetitions to ten for mononucleotide repeats, to five for dinucleotide repeats, to four for trinucleotide repeats, and to three for tetra-, penta-, and hexanucleotides. Tandem repeats were identified using Tandem Repeats Finder (Benson 1999) with the recommended settings. Repeat analysis was done on complete plastomes, excluding the second IR copy.

Additionally, we studied dispersed repeats using Geneious (v. 11.1.5, Biomatters, Ltd., New Zealand), setting the minimum repeat length to 40 bp and allowing for up to 10% of mismatches between the repeat copies.

Supplementary Material

Supplementary data are available at Genome Biology and Evolution online.

Click here for additional data file.