Plastome evolution in Santalales involves relaxed selection prior to loss of ndh genes and major boundary shifts of the inverted repeat

Abstract

Background and Aims

Biological aspects of haustorial parasitism have significant effects on the configuration of the plastid genome. Approximately half the diversity of haustorial parasites belongs to the order Santalales, where a clearer picture of plastome evolution in relation to parasitism is starting to emerge. However, in previous studies of plastome evolution there is still a notable under-representation of members from non-parasitic and deep-branching hemiparasitic lineages, limiting evolutionary inference around the time of transition to a parasitic lifestyle. To expand taxon sampling relevant to this transition we therefore targeted three families of non-parasites (Erythropalaceae, Strombosiaceae and Coulaceae), two families of root-feeding hemiparasites (Ximeniaceae and Olacaceae) and two families of uncertain parasitic status (Aptandraceae and Octoknemaceae). With data from these lineages we aimed to explore plastome evolution in relation to the evolution of parasitism.

Methods

From 29 new samples we sequenced and annotated plastomes and the nuclear ribosomal cistron. We examined phylogenetic patterns, plastome evolution, and patterns of relaxed or intensified selection in plastid genes. Available transcriptome data were analysed to investigate potential transfer of infA to the nuclear genome.

Results

Phylogenetic relationships indicate a single functional loss of all plastid ndh genes (ndhA–K) in a clade formed by confirmed parasites and Aptandraceae, and the loss coincides with major size and boundary shifts of the inverted repeat (IR) region. Depending on an autotrophic or heterotrophic lifestyle in Aptandraceae, plastome changes are either correlated with or pre-date the evolution of parasitism. Phylogenetic patterns also indicate repeated loss of infA from the plastome, and based on the presence of transcribed sequences with presequences corresponding to thylakoid luminal transit peptides, we infer that the genes were transferred to the nuclear genome.

Conclusions

Except for the loss of the ndh complex, relatively few genes have been lost from the plastome in deep-branching root parasites in Santalales. Prior to loss of the ndh genes, they show signs of relaxed selection indicative of their dispensability. To firmly establish a potential correlation between ndh gene loss, plastome instability and evolution of parasitism, it is pertinent to refute or confirm a parasitic lifestyle in all Santalales clades.

INTRODUCTION

Parasitism in land plants can be divided into two main categories: mycoheterotrophy, where the plant relies on a fungal connection to its host to mediate uptake of nutrients, and haustorial parasitism, where nutrient uptake is mediated directly through a specialized haustorium (a parasitic root) formed by the plant itself (Leake, 1994; Nickrent, 2020). With increasing levels of heterotrophism and increased carbon uptake from the host, reliance on independent carbon acquisition is reduced. In holoparasites, photosynthesis has been lost completely, whereas hemiparasites have retained a capacity to perform photosynthesis (Westwood et al., 2010).

Haustorial parasitism has evolved independently at least 12 times within the flowering plants, but of these only two lineages, i.e. the family Orobanchaceae and the order Santalales, display the full span of life forms from non-parasites through hemiparasites to holoparasites (Nickrent, 2020). The sandalwood order (Santalales) is the largest group, including more than half of all described haustorial parasitic species, and it is the only group that also includes both root- and stem-feeding parasites (Nickrent, 2020). Thus, Santalales offers an excellent framework for studies of molecular evolution both during the transition from a non-parasitic to a parasitic lifestyle and during the transition between types of parasitism.

The relaxed dependence on photosynthesis is correlated with a degradation of the plastome, including structural changes, elevated substitution rates, pseudogenization, gene loss, and general size reductions of both coding and non-coding regions (Wicke et al., 2016; Graham et al., 2017; Wicke and Naumann, 2018). Degradation of the plastome seems to typically follow a convergent trajectory in heterotrophic or partially heterotrophic lineages, correlated with the degree of host dependence, where hemiparasites and partial mycoheterotrophs tend to have less-degraded plastomes compared to holoparasites and full mycoheterotrophs (Wicke et al., 2016; Graham et al., 2017; Wicke and Naumann, 2018).

In the typically less-degraded plastomes of hemiparasites and partial mycoheterotrophs, functional loss of the plastid-encoded ndh genes seems to be one of the most apparent and unifying traits (Wicke et al., 2011; Graham et al., 2017; Wicke and Naumann, 2018; Chen et al., 2020). However, an exception is the hemiparasitic Krameriaceae with intact ndh genes (Banerjee et al., 2022). The plastid ndh genes code for some of the subunits of the NADH dehydrogenase-like (NDH) complex, whereas additional subunits are nuclear encoded (Shikanai, 2020; Ma et al., 2021). It was proposed that ndh genes lost from the plastome could be transferred to the nuclear (or mitochondrial) genome (Lin et al., 2015), but instead a correlated loss of plastid and nuclear genes suggesting complete loss of the NDH complex has been described (Lin et al., 2017). Curiously, recurrent loss of plastid ndh genes is observed not just in parasitic and other heterotrophic plants, but even in a number of autotrophs (see Graham et al., 2017; Wicke and Naumann, 2018; Strand et al., 2019). While the loss of ndh genes in holoparasites and full mycoheterotrophs is expected due to complete loss of photosynthesis, a similar loss in hemiparasites, partial mycoheterotrophs and carnivorous plants has been associated with the source or assimilation of nutrients (e.g. Wicke et al., 2011; Nevill et al., 2019). For autotrophic plants, a number of hypotheses have been proposed to explain ndh gene loss, but the underlying factors remain obscure (Strand et al., 2019). Thus, it is relevant to ask how, if at all, the loss is related to the evolution of a heterotrophic lifestyle.

Within Santalales, a continuously more comprehensive picture of plastome configuration and evolution is starting to form, although many published papers include data from just a single or a few plastomes (Petersen et al., 2015; Zhang et al., 2015; Su and Hu, 2016; G.-S. Yang et al., 2017; Li et al., 2017; Shin and Lee, 2018; Zhu et al., 2018; Guo and Ruan, 2019a, b; Guo et al., 2019, 2020, 2021a, b; Jiang et al., 2019; S.-S. Liu et al., 2019; Schelkunov et al., 2019; Su et al., 2019, 2021; Y. Yang and He, 2019; Yu et al., 2019; Chen et al., 2020; D. Yang et al., 2020; Ceriotti et al., 2021; Nickrent et al., 2021; X. Liu et al., 2021; Al-Juhani et al., 2022; Darshetkar et al., 2023; Tang et al., 2024). In the most comprehensive study to date (Chen et al., 2020), most genomic changes were found to be associated with the evolution of hemi- and holoparasitism, and only minor differences were found between root- and stem-feeding hemiparasites. Loss of ndh genes was found in all parasites, but not in non-parasites. However, in their study taxon sampling of non-parasitic and root-feeding hemiparasitic lineages was limited and may have resulted in a simplified picture of plastome evolution. Thus, deeper sampling from those clades is strongly called for.

To trace plastome evolution in Santalales, a solid phylogenetic framework is needed. Despite ambiguity in the sister group relationship of the order (APG IV, 2016; Leebens-Mack et al., 2019; Baker et al., 2022, Zuntini et al., 2024), phylogenetic relationships within Santalales are becoming increasingly clear, and the monophyly of clades at the level of family (sensu  Nickrent et al., 2010) is mostly well supported (Su et al., 2015; Nickrent et al., 2019). However, some ambiguity remains about the deepest splits involving non-parasitic and root-feeding, hemiparasitic families (Fig. 1).

Fig. 1.

Four alternative topologies from previous and present phylogenetic analyses of Santalales: (A) Su et al. (2015), 3 plastid + 1 mitochondrial + 3 nuclear genes; (B) Nickrent et al. (2019), 3 plastid + 2 nuclear genes; (C) Nickrent (2020), 3 plastid + 1 mitochondrial + 3 nuclear genes; (D) this study, 76 plastid + 3 nuclear genes, summary tree of maximum likelihood and Bayesian inference analyses. Green terminals are non-parasitic; brown terminals are parasitic; blue terminals have uncertain parasitic status. Other Santalales comprise 13 families including root and stem hemiparasites as well as holoparasites.

Tracing the early evolutionary steps towards parasitism is further complicated by missing data of the potential parasitic status of several taxa (Kuijt and Hansen, 2015). Whereas the parasitic status of holoparasites and stem-feeding parasites is clear, a potential root-feeding parasitic habit remains to be firmly established for some taxa. As delineated here, Strombosiaceae, Erythropalaceae and Coulaceae are generally considered non-parasitic, while the status of Aptandraceae and Octoknemaceae is insufficiently known (Kuijt and Hansen, 2015; Nickrent, 2020) (Fig. 1). Though not all genera within Ximeniaceae and Olacaceae have been investigated, both families are generally considered parasitic based on the presence of secondary haustoria in the roots of Ximenia (Ximeniaceae; DeFilipps, 1969), Olax (Olacaceae; Barber, 1907) and Ptychopetalum (Olacaceae; Anselmino, 1932).

Attempting to trace the early steps of plastome evolution in Santalales, we here target the families of assumed non-parasitic status (Strombosiaceae, Erythropalaceae and Coulaceae) and of unclear parasitic status (Aptandraceae and Octoknemaceae). Further, we add data to two families of root-feeding hemiparasites, Ximeniaceae and Olacaceae, the latter not previously being included in studies of complete plastome evolution. Based on 29 newly sequenced plastomes and the nuclear ribosomal DNA cistron region from Santalales, we complement previous studies of phylogenetic patterns and plastome evolution with an emphasis on the transition to a parasitic lifestyle. In general, we explore whether plastome configuration and gene content is affected by the transition to parasitism, and in particular whether the loss of ndh genes can be correlated with this shift or possibly occurred prior to it. We also investigate whether changes in selection of genes or gene complexes can be related to shifts in lifestyle.

MATERIALS AND METHODS

Taxon sampling

Taxon sampling was guided by previous studies of the order (Malécot and Nickrent, 2008; Su et al., 2015; Nickrent et al., 2019), with a focus on non-parasitic and root hemiparasitic lineages. Following family delineation in Nickrent et al. (2010, 2019), we sampled three species from Erythropalaceae, six from Strombosiaceae, six from Aptandraceae, one from Octoknemaceae, two from Ximeniaceae, three from Olacaceae and seven from the remaining families of the order, thereby allowing a comparative framework that includes all major photosynthetic lineages in the order (see Supplementary Data Table S1). Three previously published Santalalean plastomes were included in the study: Erythropalum scandens (NC036759, Erythropalaceae), Malania oleifera (MK764537, Ximeniaceae) and Loranthus europaeus (MT987630, Loranthaceae). Four plastomes from taxa outside Santalales were downloaded from GenBank to be used as outgroups: Crataegus marshallii (MK920293, Rosales: Rosaceae), Helianthus annuus (NC007977, Asterales, Asteraceae), Myoporum bontioides (NC050956, Lamiales, Scrophulariaceae) and Silene kiusiana (NC048886, Caryophyllales: Caryophyllaceae).

For the nuclear ribosomal cistron (hereafter nrDNA), we downloaded outgroup sequences from GenBank for Crataegus marshallii (MN215977), Helianthus annuus (KF767534), Myoporum tetrandrum (MN411551) and Silene nutans (MT735331). Outgroups were chosen to broadly cover a large section of angiosperm lineages, since the sister group relationship of Santalales is not clear (APG IV, 2016; Leebens-Mack et al., 2019; Baker et al., 2022; Zuntini et al., 2024), and the selection was further guided by data availability of both plastome and nrDNA sequences in GenBank. For ingroup taxa, nrDNA sequences were downloaded for Diogoa zenkeri (MH390535, Strombosiaceae) and Strombosia pustulata (MH390536, Strombosiaceae) to extend incomplete sequences obtained in the present study. The two GenBank sequences were derived from the same specimens sequenced in this study.

DNA extraction and sequencing

Genomic DNA was extracted from leaf material from either herbarium specimens or silica-dried fresh material, using a modified CTAB protocol (Doyle and Doyle, 1987). Library preparation was carried out with TruSeq DNA PCR-free libraries and sequenced as 150- or 250-bp paired-end reads on an Illumina NovaSeq 6000 at SciLifeLab Stockholm Campus Solna or NGS Core at National Taiwan University.

Assembly of plastomes and the nuclear ribosomal cistron (nrDNA)

Assemblies and annotations were performed at three different laboratories following slightly different strategies. For the majority of samples, raw paired-end reads were trimmed for quality (minimum Phred score of 20, minimum length of 50 bp) and adapter contamination with BBDuk, duplicates were removed with Clumpify, and overlapping reads were merged with Tadpole (after extension and recalculation of Phred scores) from BBMap v.38.81 (Bushnell, 2014). This resulted in two sets of reads (merged and unmerged) for downstream processing.

To avoid contamination from other genomic compartments and to get a manageable dataset, trimmed reads were mapped against an appropriate reference (plastome of the putative closest available relative) from GenBank (Supplementary Data Table S1). Reads matching the corresponding k-mer depth were then extracted using KAT v.2.4.2 (Mapleson et al., 2017). This subset of reads was used for de novo assembly in Unicycler v.0.4.9b (Wick et al., 2017). Due to low read numbers in some cases, kmer-filtering was not performed on all read-sets (see Table S1). Bandage v.0.8.1 (Wick et al., 2015) was used to view and evaluate assemblies, and to assess contig connections. When the expected structure of a small and a large contig connected at both ends to a higher-depth linear contig [representing the small and large single-copy regions, connected to the inverted repeat (IR) region, respectively] was found, the plastome was circularized. If not, further extension was carried out with Tadpole (with the settings k = 62, extend left and right = 10 000) on the subset of contigs matching plastome blast hits from a local database (as implemented in Bandage). The extended contigs were used as a reference for mapping reads for another iteration of Unicycler de novo assembly. This procedure was repeated until either the plastome was circularized or no further extension could be achieved. The resulting plastome assemblies were quality checked by mapping the KAT-filtered plastome reads (or the full set of quality trimmed reads for the smaller read-sets) in Geneious Prime v.2019.0.4 (Biomatters, Auckland, New Zealand) with default settings for manual inspection, and a minimum read-depth of 7 (without conflict) in any position was considered reliable.

Most plastome annotations were carried out in GeSeq (Tillich et al., 2017) through the online application (https://chlorobox.mpimp-golm.mpg.de/geseq.html). Settings were left unchanged except for ‘protein search identity’ which was changed to 75, and the third-party programs tRNAscan-SE v.2.0.7. (Chan and Lowe, 2019) and Chloë v.0.1.0. (https://chloe.plantenergy.edu.au) were also used. Annotation duplications were removed, generally favouring Chloë annotations, as they more often resulted in annotations of the expected length and position of the genes (in comparison to previously published plastomes).

For Misodendrum brachystachyum (Misodendraceae), raw reads were trimmed as above and the plastome was de novo assembled and circularized using NOVOPlasty v.2.7.2 (Dierckxsens et al., 2017) using the rbcL sequence from Malania oleifera as seed and a k-mer of 33. Annotations were performed using the Transfer annotations tool in Geneious Prime, where genes were predicted based on amino acid sequence similarity to Malania oleifera (MK764537). Assembly and annotation of Olax imbricata and Scorodocarpus borneensis (Strombosiaceae) plastomes were performed using strategies previously described (Su et al., 2021).

To identify potential structural differences, whole-plastome alignments were performed using Mauve v.1.1.3 (http://darlinglab.org/mauve/mauve.html) as implemented in Geneious Prime 2024.0.3 (https://www.geneious.com).

The nrDNA region was assembled following a procedure similar to the main one described above, with the exception that kmer depth was evaluated based on 18S, 26S, and (when available) internal transcribed spacer (ITS) and/or 5.8S contigs downloaded from GenBank (see Supplementary Data Table S1).

Alignments and phylogenetic analysis

One copy of each (translated) coding sequence (CDS) region from the plastome was extracted and aligned separately using MAFFT v.7.310 (Katoh and Standley, 2013) with the L-INS-I setting applied. After alignment, CDS regions were converted back to their corresponding nucleotide sequences with pal2nal v.14 (Suyama et al., 2006), producing aligned DNA sequences. Sites with more than 50 % gaps were removed from all alignments. After visual inspection in alan alignment viewer v.2.1.0 (https://www.github.com/mpdunne/Alan), two gene alignments (rpl22, rps16) were excluded due to poor alignment before concatenation and partitioning (with separate partitions kept for regions and codon positions). For the nrDNA, 18S, 5.8S and 26S nucleotide sequences were aligned with MAFFT, trimmed as above and partitioned by region, only. The ITS regions were not used because of considerable alignment ambiguity. Scripts for trimming, concatenation and partitioning are available on a Github repository (https://github.com/bmichanderson/scripts).

A maximum likelihood analysis with automatic model selection for all partitions (Kalyaanamoorthy et al., 2017) and partition merging (Lanfear et al., 2012) was implemented in IQ-TREE v.2.1.3. (Minh et al., 2020), executing ten runs (‘--runs 10’) with 10 000 ultrafast bootstrap replicates (Hoang et al., 2018), which re-sampled partitions and sites within them [with the flag ‘--sampling GENESITE’ (Gadagkar et al., 2005; Seo et al., 2005)]. Analyses were carried out separately for the plastome and nrDNA. As no strongly supported conflict (ultrafast bootstrap, UB > 95) was observed between the resulting trees (Supplementary Data Figs S1 and S2), a combined analysis of both plastome and nrDNA data sets was carried out with the same settings as above to improve phylogenetic signal.

To complement the maximum likelihood analyses, the same alignments were analysed using a Bayesian approach in MrBayes v.3.2.7 (Ronquist et al., 2012). To determine the optimum partitioning and models for the plastid, nuclear and combined alignments, these were run again with IQ-TREE using the options ‘-m TEST -mset mrbayes’ to restrict the search to models that could be implemented in MrBayes. The resulting optimal models were then implemented with priors fixed to the values determined in the IQ-TREE model search. The Bayesian analyses were run until the average standard deviation of split frequencies between the two independent runs per analysis was <0.005 or the run reached 500 000 generations (sampling every 1000 generations), whichever took longer. Estimated sample size (ESS) values for tree length were >300 for all analyses, and the potential scale reduction factors (PSRFs) were all close to 1, indicating a good sampling of the posterior and convergence of the independent runs. A relative burn-in of 25 % was used for diagnostics and sampling parameters and trees. The majority-rule consensus tree for each analysis was displayed using FigTree v.1.4.4 (Rambaut, 2018; https://github.com/rambaut/figtree). To visualize differences between trees in the credible sets of trees, consensus networks (25 % cutoff for displaying splits) were constructed using SplitsTree4 v.4.17.1 (Huson and Bryant, 2006).

Functional gene losses were mapped on the phylogenetic tree derived from the combined maximum likelihood analysis. Mapping was based on the principle of parsimony and the assumption that genes may be lost but not gained, and that pseudogenes do not revert to functional copies.

Selection strength changes

Signs of relaxed or intensified selection pressures in protein coding genes were investigated utilizing RELAX (Wertheim et al., 2015), as implemented in HyPhy v.2.5.48 (Pond et al., 2005). RELAX investigates if the ratio of non-synonymous to synonymous substitutions (dN/dS or ω, where ω > 1 indicates positive selection and ω < 1 indicates purifying selection) in a specified set of test branch/es of the tree differs from a set of reference branch/es. A selection intensity parameter, k, is calculated, where k > 1 indicates relatively intensified selection and k < 1 indicates relatively relaxed selection (Wertheim et al., 2015). This was performed on functional gene groups, as well as for separate ndh genes and other individual genes, which cannot be assigned to a functional group. Terminal taxa were treated as individual test branches, against the four non-Santalalean outgroup taxa assigned as the reference group (remaining taxa left undefined). Alignments from the phylogenetic inference were used, ensuring correct codon positioning and interpretation, with alignments concatenated by each functional gene group (e.g. atpA–I concatenated for the atp group). Trees were inferred for each group or gene separately using the same settings as for the main phylogenetic analyses, except only 1000 ultrafast bootstrap replicates were used and sampled by site only. Branches with <50 % ultrafast bootstrap support were collapsed to polytomies with Newick Utilities (Junier and Zbodnov, 2010). RELAX was run separately for each test terminal (up to 32 per alignment/tree) for 30 gene group alignments/trees (>700 runs) in parallel as a batch job on the Pawsey Setonix supercomputer (Perth, Western Australia). The entire set of runs was repeated five times after detecting issues with convergence, i.e. in a few cases significantly different values of k were obtained. To improve convergence, RELAX was run with the options ‘--starting points 10 --models Minimal’ following advice from the developer (Sergei Pond, pers. comm.). Increasing the number of starting points from one (the default) makes RELAX perform more initial optimizations to estimate a good starting point for complete model optimization. The ‘--models Minimal’ flag makes RELAX skip calculating the parameter-rich General Descriptive model, which in some cases may improve convergence (S. Pond, pers. comm.). For each test, the k value with the highest likelihood across the five runs was reported along with the significance compared to the alternative model.

Transfer of infA to the nuclear genome

To detect a possible transfer of infA from the plastome to the nuclear genome within Santalales, the infA gene from Nuytsia floribunda (MT987640, Loranthaceae) was blasted against a local database of ~210 000 assembled transcripts from Viscum album (Viscaceae; Petersen et al., 2022) using the BLASTN algorithm as implemented in Geneious Prime v.2020.2.2 (Biomatters Ltd). The same infA sequence from Nuytsia was also blasted against the whole-genome shotgun contig library of Santalum album (Bioproject PRJNA422746, Santalaceae) and the transcriptome shotgun assembly library of Taxillus chinensis (Bioproject PRJNA548287, Loranthaceae) available in GenBank. These species were selected because all lack the infA gene in their plastomes (Petersen et al., 2015; Guo et al., 2020; Su et al., 2021). Since the identified nuclear infA sequences (from start to stop codon) were more than twice as long as infA plastome genes from Santalales, we tested if the longer N-terminal region could be a presequence coding for a transit peptide using TargetP-2.0 (Almagro Armenteros et al., 2019). The (translated) nuclear genes (infA-N) and plastome genes (infA-P) of Nuytsia floribunda (accession MT987640) and the newly generated Schoepfia arenaria (Schoepfiaceae) were aligned with MAFFT v.7.310 (with the L-INS-i setting applied) for comparison.

RESULTS

Plastome characteristics

New plastome sequence data were generated for 29 species (Supplementary Data Table S2). The number of trimmed reads varied from 3 697 838 to 146 256 820, resulting in an average read-depth of plastomes of 32–6909.2 and an average read-depth of nrDNA regions of 430.9–14 264.2. For 21 samples, complete, circular plastomes were assembled. The remaining eight samples, where read-depth was typically lower, were assembled into 3–14 contigs (Table S2). The total length of all completely assembled Santalales plastomes varied from 122 120 bp (Schoepfia schreberi) to 161 120 bp (Minquartia guianensis, Coulaceae). They all had the quadripartite structure typical of angiosperm plastomes, with a GC content of ~35–38 % (Table 1). The length of the IR varied from 12 426 bp (Schoepfia schreberi) to 35 256 bp (Malania oleifera) (Table 1). The IR size differences were typically connected to movement of the borders between the IRs and the single-copy regions, causing inclusion or exclusion of genes from the IR (Fig. 2).

Table 1.

Characteristics of plastomes of Santalales. Since complete, circular assemblies were not obtained for all samples, the number of assembled contigs is given. A complete circular assembly has one contig.

Family	Species	Plastome size (bp)	GC (%)	IR size (bp)	Genes			Contigs
					Protein	rRNA	tRNA
Erythropalaceae	Brachynema ramiflorum	>129 255	~36.7	26 160	78	4	28	14
Erythropalaceae	Erythropalum scandens	156 154	38	26 394	79	4	30	1
Erythropalaceae	Heisteria densifrons	>159 896	~37.4	26 046	79	4	30	7
Erythropalaceae	Maburea trinervis	>128 892	~36.6	26 069	791	4	30	5
Strombosiaceae	Diogoa zenkeri	159 584	37.7	26 250	79	4	30	1
Strombosiaceae	Engomegoma gordonii	159 618	37.7	26 245	79	4	30	1
Strombosiaceae	Scorodocarpus bornensis	159 272	37.8	26 136	79	4	30	1
Strombosiaceae	Strombosia pustulata	159 388	37.8	26 215	79	4	30	1
Strombosiaceae	Strombosiopsis tetrandra	160 070	37.6	26 119	79	4	30	1
Strombosiaceae	Tetrastylidium peruvianum	>158 467	~37.9	26 191	79	4	30	5
Coulaceae	Coula edulis	>121 474	~36.8	26 382	79	4	29	3
Coulaceae	Minquartia guianensis	161 120	37.7	26 340	79	4	30	1
Octoknemaceae	Octoknema affinis	160 640	37.5	26 325	79	4	30	1
Aptandraceae	Anacolosa papuana	149 924	36.9	24 090	68	4	30	1
Aptandraceae	Anacolosa pervilleana	150 027	37.1	24 167	68	4	30	1
Aptandraceae	Aptandra tubicina	143 301	36.9	23 592	68	4	28	1
Aptandraceae	Cathedra acuminata	156 432	36.8	28 758	68	4	30	1
Aptandraceae	Harmandia mekongensis	141 146	37.4	24 055	68	4	29	1
Aptandraceae	Phanerodiscus capuronii	>127 374	~35.5	24 234	67	4	30	12
Ximeniaceae	Curupira tefeensis	>122 618	~35.8	31 713	68	4	30	5
Ximeniaceae	Malania oleifera	158 163	36.7	35 256	70	4	30	1
Ximeniaceae	Ximenia americana	156 507	36.8	32 746	69	4	30	1
Olacaceae	Dulacia candida	156 333	35.6	33 606	67	4	30	1
Olacaceae	Olax imbricata	155 798	35.8	33 982	67	4	30	1
Olacaceae	Olax scandens	>156 106	~35.8	33 933	67	4	29	3
Schoepfiaceae	Schoepfia arenaria	122 210	37.9	12 438	69	4	30	1
Schoepfiaceae	Schoepfia schreberi	122 120	37.9	12 426	71	4	29	1
Misodendraceae	Misodendrum brachystachyum	135 241	37.1	24 485	70	4	28	1
Loranthaceae	Loranthus europaeus	123 004	37.1	23 087	66	4	28	1
Thesiaceae	Thesium decaryanum	141 066	36.5	25 031	67	4	30	1
Nanodeaceae	Nanodea muscosa	140 372	37.5	26 317	68	4	30	1
Viscaceae	Viscum trachycarpum	128 593	36.8	21 907	67	4	29	1

¹ ndhF partial.

Fig. 2.

Inverted repeat (IR) region in representatives of Santalales. Red bars showing relative length of the IR. Coloured squares indicate a core IR region shared by outgroup representatives and ancestral Santalales. Functional genes in the core IR region are in green, pseudogenes in yellow. Genes of species with expanded IRs are listed upstream or downstream of the core region. Pa = genes partially included in the IR; P = pseudogenes in the non-core IR region.

Gene content of all complete plastomes included in the study varied from 98 (Loranthus europaeus, Loranthaceae) to 113 (most non-parasitic species), with 66–79 protein coding genes, four rRNA genes and 28–30 tRNA genes (Table 1). The gene order was conserved within the order, and no genome rearrangements were observed among the complete plastomes. For the eight species where complete plastomes could not be assembled, rearrangements may exist and a few genes may not have been detected (Supplementary Data Table S3).

The difference in gene numbers observed among the sampled Santalales is mostly due to absence or presence of functional copies of the ndh genes. In Erythropalaceae, Strombosiaceae, Coulaceae and Octoknemaceae, all 11 genes (ndhA–K) are present and presumably functional, but in the remaining taxa only pseudogenes or short gene fragments could be identified (Table 2).

Table 2.

Presence of ndh pseudogenes or gene fragments (•) in species of Santalales.

Family	Species	ndhA	ndhB	ndhC	ndhD	ndhE	ndhF	ndhG	ndhH	ndhI	ndhJ	ndhK
Aptandraceae	Anacolosa papuana	•	•	•	•				•	•	•	•
	Anacolosa pervilleana	•	•	•	•				•	•	•	•
	Aptandra tubicina	•	•		•				•	•	•
	Cathedra acuminata	•	•	•	•			•	•	•	•	•
	Harmandia mekongensis		•						•	•	•
	Phanerodiscus capuronii	•	•		•			•	•	•
Ximeniaceae	Curupira tefeensis		•		•				•
	Malania oleifera	•	•				•		•	•
	Ximenia americana	•	•		•		•		•	•
Olacaceae	Dulacia candida	•	•						•
	Olax imbricata	•	•		•				•
	Olax scandens	•	•		•		•		•
Schoepfiaceae	Schoepfia arenaria	•		•	•			•	•	•	•	•
	Schoepfia schreberi	•		•	•			•	•	•	•	•
Misodendraceae	Misodendrum brachystachyum		•		•						•
Loranthaceae	Loranthus europaeus		•
Thesiaceae	Thesium decaryanum	•	•
Nanodeaceae	Nanodea muscosa	•	•								•
Viscaceae	Viscum trachycarpum		•

Phylogenetic results

The concatenated plastome plus nrDNA matrix contained 72 907 sites of which 48 453 were variable. The maximum likelihood analysis resulted in a phylogenetic tree with strong support for monophyly of all families represented by more than one species (UB values of 100) (Fig. 3). Erythropalaceae is strongly supported as sister to Strombosiaceae (UB 100), and this clade is sister to the rest of Santalales. Coulaceae and Octoknemaceae form a weakly supported clade (UB 72), in a sister group position to the confirmed parasites and Aptandraceae (UB 76), but with weakly supported internal relationships among Aptandraceae, Ximeniaceae, Olacaceae and the remaining core parasites. Comparing the topologies of trees derived from separate analyses of plastid and nuclear data revealed differences, but no strongly supported conflict (Supplementary Data Figs S1 and S2).

Fig. 3.

Phylogeny of Santalales based on maximum likelihood analysis of combined plastome gene and nrDNA sequences. Node values are ultrafast bootstrap support. Green branches correspond to non-parasitic lineages, brown branches to parasitic lineages. Blue branches are lineages with uncertain parasitic status. Asterisks (*) after species names indicate stem parasites. Functional gene losses are marked in red. The scale bar indicates inferred substitutions per site.

Bayesian inference of the combined matrix used 17 partitions with either the GTR + GAMMA or GTR + I + GAMMA models. Average ESS for tree length was 373 and the PSRF was 1.003. The credible set of trees contained two trees. A separate analysis of the plastid data used 16 partitions with either the GTR + GAMMA or GTR + I + GAMMA models. Average ESS for tree length was 376 and the PSRF was 0.999. The credible set of trees contained 11 trees. The nuclear analysis used two partitions with the SYM + I + GAMMA or GTR + I + GAMMA models. Average ESS for tree length was 582 and the PSRF was 1.000. The credible set of trees contained 504 trees. The majority rule consensus trees from combined and separate analyses are shown in Supplementary Data Fig. S3.

Comparing the plastid and nuclear trees from Bayesian inference showed some support (low support in the maximum likelihood trees) for conflicting placement of the clade containing Coula edulis, Minquartia guianensis (both Coulaceae) and Octoknema affinis (Octoknemaceae). The plastid analysis resolved those taxa as a clade (0.98 posterior probability, pp) in a larger clade (1.00 pp) containing the confirmed parasites and Aptandraceae. The nuclear analysis showed little support (0.64 pp) for those taxa as a clade, but resolved them in a larger clade (0.99 pp) containing the autotrophic Erythropalaceae and Strombosiaceae. The combined analysis resolved them as a clade (1.00 pp) sister to the confirmed parasites and Aptandraceae (1.00 pp) as in the plastid analysis. Another grouping that differed between the plastid and nuclear trees is a sister relationship between Ximeniaceae and Olacaceae, which was supported in the plastid analysis (1.00 pp), but not in the nuclear analysis, where Ximeniaceae was resolved sister (0.99 pp) to the core parasites and Olacaceae was resolved sister (0.99 pp) to Aptandraceae.

The topology of the trees from the combined analyses using maximum likelihood compared to a Bayesian approach differed only in the relative branching order of the confirmed parasites and Aptandraceae, with the maximum likelihood providing low support (UB 65) for Ximeniaceae + Olacaceae (UB 74) sister to the core parasites, but the Bayesian analysis with low support (0.93 pp) for Aptandraceae sister instead. The credible set of trees in the Bayesian analysis of the combined alignment contained only two trees, one with Aptandraceae sister and the other with Ximeniaceae + Olacaceae sister to the core parasites (Supplementary Data Fig. S4).

Gene evolution

Phylogenetic relationships suggest a single functional loss of all ndh genes in the common ancestor to a clade including confirmed parasites plus Aptandraceae (Fig. 3), although only a partial sequence of ndhF could be obtained from the incomplete plastome assembly of Maburea trinervis (Erythropalaceae). However, read depth was low near the IR–Small Single Copy (SSC) borders where the gene is located, and a complete copy was probably missed. Additional functional gene loss includes the protein coding genes infA and rpl33 and six tRNA genes (trnA-UGC, trnG-UCC, trnH-GUG, trnI-GAU, trnK-UUU, trnV-UAC). All losses are confined to species within the parasitic clade and Aptandraceae, with repeated losses of infA and trnV-UAC (Fig. 3).

Results from testing for changes in selection pressure are presented in Table 3 as the k values with the highest likelihood obtained across five runs of RELAX. Data from each individual run and a summary of the variation between runs are provided as Supplementary Data Table S4. Compared to running RELAX analyses with default parameters, convergence among runs was increased. Thus, results from individual runs were consistent for the majority of tests, but we observed significant differences in four cases: cemA – Brachynema ramiflorum (Erythropalaceae) and the rps gene group – Coula edulis, Maburea trinervis and Viscum trachycarpum. In the 19 genes or gene groups tested for 32 species, 20 % of the results showed significant levels of relaxed selection and 3 % significant levels of intensified selection (Table 3). In general, the results give a mixed picture with only few groups of taxa showing a consistent pattern of relaxation or intensification of individual genes or functional groups of genes. The ycf1 gene stands out as being under relaxed selection in most species. Significant relaxed selection pressures were also evident in the atp, ndh and rpl complexes of all Erythropalaceae, and Erythropalum scandens showed relaxed selection in almost all photosynthesis-associated gene complexes. A similar trend of relaxed selection was apparent in the photosynthesis-associated genes in stem-feeding parasitic genera (Misodendrum, Loranthus, Viscum), where all complexes show signs of relaxed selection in all species, except rbcL in Viscum trachycarpum. It is remarkable that Loranthus europaeus showed significant relaxation in all the genes tested apart from those coding for protease (clpP) and the conserved open reading frames ycf1, ycf2, ycf3 and ycf4. Though missing in at least three lineages within the order, infA only showed significant signs of relaxed selection pressure in Anacolosa papuana, Harmandia mekongensis (both Aptandraceae) and Loranthus europaeus.

Table 3.

Patterns of relaxed and intensified selection in 11 individual genes and eight functional gene groups. Reference group: Crataegus marshallii, Helianthus annuus, Myoporum bontioides and Silene kiusiana. Blue boxes indicate significant relaxed selection, red boxes indicate significant intensified selection. Values are the selection intensity parameter k. Significance levels for k: *P < 0.05, **P < 0.01, ***P < 0.001.

Test species	accD	atp	ccsA	cemA	clpP	infA	matK	ndh	pet	psa	psb	rbcL	rpl	rpo	rps	ycf1	ycf2	ycf3	ycf4
Brachynema ramiflorum	0.6	0.6***	0.6	0.3*	0.0***	13.1	0.2*	0.3***	0.9	0.8	0.8	0.0	0.2***	0.9	0.6**	–	0.3	1.2	0.4
Erythropalum scandens	0.7	0.2***	0.5	0.5	0.0***	0.6	0.8	0.5***	0.0*	0.7**	0.8*	0.8	0.7*	0.8	0.9	0.2***	1.2	0.8	0.2*
Heisteria densifrons	1.0	0.7*	1.7	0.0**	5.5*	22.0	0.3*	0.1***	0.8	24.2	0.9	0.3	0.3**	0.3**	0.7	1.2	0.9	0.4	1.3
Maburea trinervis	1.9	0.6*	0.5	0.1	0.0	13.1	0.4*	0,5***	11.5*	1.0	0.6**	1.2	0.3*	0.4**	0.2***	0.4***	0.4	1.1	0.0**
Diogoa zenkeri	0.7	0.9	0.6	1.2	1.0	1.2	0.0	0.2*	0.6	0.6	0.5	3.3	1.5	0.5	0.5	0.0***	0.5	0.0	1.1
Engomegona gordonii	0.7	0.2	0.0	1.2	1.2	1.2	5.4	0.2*	34.5	0.4*	0.8	0.2	0.8	0.4	1.4	0.3***	0.5	1.4	0.0
Scorodocarpus borneensis	1.4	0.5	1.1	0.6	0.0	0.7	0.7	0.5	0.0	1.0	0.8	0.7	0.8	0.7	0.9	0.0***	0.0*	0.5	0.6
Strombosia pustulata	1.1	0.8	0.7	0.0	0.5	1.0	0.9	0.1**	0.5	1.1	0.2**	1.3	0.5	0.7	1.2	0.0***	0.0	1.2	1.1
Strombosiopsis tetrandra	0.7	0.6	0.8	24.6	1.2	1.2	0.7	0.5	0.2	0.9	0.1	0.7	0.8	0.7	0.2	0.1***	0.5	0.0	17.6
Tetrastylidium peruvianum	0.6	1.0	0.8	0.0	1.2	1.2	4.6	0.5	0.4	0.5	35.0	3.6*	0.3	0.7	0.7	0.0***	0.5	11.0	0.0
Coula edulis	0.7	22.6	0.8	0.0	1.2	12.5	1.1	21.1	10.9	1.4	3.4	3.3	0.5	0.6	0.0**	0.8	0.5	1.2	1.2
Minquartia guianensis	0.7	0.1	0.1	50.0	1.2	1.2	0.4	1.1	33.2	0.6	0.5	2.4	0.1	0.4	3.3**	0.0***	1.4	1.2	0.0
Octoknema affinis	0.7	9.4*	0.6	2.3	0.6	14.2	1.1	0.1**	1.1	0.7	0.7	1.1	0.6	0.9	1.0	0.0***	1.0	0.0*	0.5
Anacolosa papuana	0.0	0.5	0.2	0.9	0.0	0.0*	0.3	–	0.3	2.9	50.0	0.6	1.4	0.4	0.8	0.0***	1.1	1.2	0.2
Anacolosa pervilleana	0.7	0.9	1.3	0.0	5.5	1.2	0.6	–	0.3	0.4***	0.5	1.1	0.4	0.4	0.9	0.0***	0.0*	1.2	1.0
Aptandra tubicina	0.0	0.8	0.0*	1.2	0.4	0.8	0.0	–	0.4	0.9	0.9	0.0	0.9	0.6*	0.6	0.0***	0.0**	0.5	0.0*
Cathedra acuminata	2.0	0.6*	1.2	0.0	1.2	1.2	0.1	–	0.5	0.6*	0.9	1.2	0.6	0.8	0.3***	0.0***	3.2**	13.1	0.7
Harmandia mekongensis	0.5	0.6***	1.1	1.0	0.0	0.0*	0.3	–	0.6	0.3***	0.8**	1.7	0.3	0.7	20.4**	0.3***	0.4	0.6	0.7
Phanerodiscus capuronii	0.7	0.9	0.6	0.0	0.4	21.0	1.3	–	1.2	0.6**	0.5***	0.4	0.5	0.7	0.4**	0.3***	0.5	1.0	0.6
Curupira tefeensis	1.0	0.3*	0.7	0.0	1.2	11.0	0.7	–	0.8	1.2	0.4***	0.4	1.6	0.9	0.5	0.1***	1.8*	1.0	0.8
Malania oleifera	0.8	0.5**	0.6	0.2	0.9	0.5	0.9	–	0.0*	0.6***	3.2	0.0	0.8	0.6**	2.1	0.5***	0.5	0.4	0.5
Ximenia americana	0.7	0.5	1.0	0.0	1.0	1.2	0.3	–	0.1	0.8	0.7	1.9	0.6	1.3	1.0	0.1**	3.0***	1.2	31.5
Dulacia candida	1.6	0.6	0.5	1.9	0.0	–	0.0	–	1.1	0.8	0.7	1.9*	1.5	1.0	0.8	0.7	0.5	0.0	1.2
Olax imbricata	0.7	1.0	0.1	0.7	1.0	–	4.4	–	0.6	1.1	0.6	1.6	1.7	25.1	0.3	0.0**	3.2	14.5	0.0
Olax scandens	0.7	1.7	0.7	0.0	11.0	–	0.0	–	0.2	3.8	0.9	1.2	2.7*	0.6	0.7	0.3	2.0*	1.2	0.0
Schoephia arenaria	1.2	0.9	1.2	1.4	6.0	1.2	1.2	–	0.2	0.3	0.0	1.6	1.3	0.5	0.0	0.0***	0.0	1.0	1.2
Schoephia schreberi	4.2	0.3	0.8	1.2	0.0	1.2	0.3	–	0.2	0.4	0.0**	2.2*	1.1	0.3	0.4	0.3***	0.4	11.4	1.2
Misodendrum brachystachyum	1.1	0.3**	0.7	35.1	0.1	1.0	1.4	–	0.6**	0.7***	0.0**	0.0**	0.6	0.9	0.0	31.7	0.6	0.8	1.1
Loranthus europaeus	0.3*	0.4***	0.5*	0.1***	1.6	0.0*	0.2*	–	0.6***	0.3***	0.6***	0.0**	0.2***	0.7***	0.0***	33.5	0.5	0.8	0.7
Viscum trachycarpum	3.2***	0.0***	0.0**	0.7	12.6	–	9.4*	–	0.6***	0.7***	0.3***	33.3	0.7*	0.2***	21.3***	–	2.3***	0.5*	0.9
Nanodea muscosa	0.5	0.5**	0.6	0.3*	50.0*	–	0.4*	–	0.6	0.8	0.9	0.0**	0.7	0.2**	0.7*	0.4	1.2	1.1	0.6
Thesium decaryanum	0.7*	0.8*	0.7	1.1	4.5*	–	1.1	–	1.6	0.6	0.8*	0.0*	0.9	1.0	1.0	0.0***	0.0**	1.0	1.1

In addition to Erythropalaceae, Octoknemaceae and some Aptandraceae showed significantly relaxed selection of the combined group of ndh genes (Table 3). However, when investigating the 11 ndh genes separately for the 13 species that possess these genes, significantly relaxed selection was observed in only 12 % and this was confined to members of Erythropalaceae and Aptandraceae (Supplementary Data Tables S4 and S5). Erythropalum scandens had the highest number of genes (six) with significantly relaxed selection, and ndhF was the gene most often under significantly relaxed selection (in six species).

Presence of infA in the nuclear genome

A transcript containing an infA-like sequence was recovered from the Viscum album transcriptome assembly, as well as from Santalum album and Taxillus chinensis. The transcript has a 498-bp-long open reading frame in Viscum, 474 bp in Santalum and 858 bp in Taxillus, i.e. more than twice the length of the gene when present in the plastome of species of Santalales and outgroup taxa (234 bp). The extended sequences in Viscum, Santalum and Taxillus were all found to correspond to N-terminal presequences predicted by TargetP-2.0 to be thylakoid luminal transit peptides with a predicted cleavage site between amino acid residues 63 and 64 (Viscum), 70 and 71 (Santalum), and 74 and 75 (Taxillus) (Fig. 3).

DISCUSSION

Phylogenetic relationships and evolution of parasitism

The clades recovered here at the rank of family (Fig. 3) correspond to the family delineation of Nickrent et al. (2010, 2019) and Su et al. (2015), and the topologies within these clades are also consistent. The current study differs from this previous work by utilizing significantly more sequence data that provided better support for several deep splits in the phylogeny. However, some relationships remain weakly supported.

The combined phylogenetic analyses conducted here strongly support a sister group relationship between Strombosiaceae and Erythropalaceae, and this clade is strongly supported as sister to the rest of the Santalales (Fig. 3; Supplementary Data Fig. S3). In contrast, most previous phylogenetic analyses have placed either Erythropalaceae or Strombosiaceae as sister to the remaining Santalales (e.g. Su et al., 2015; Nickrent et al., 2019; Nickrent, 2020) (Fig. 1). However, Nickrent et al. (2019) used Erythropalum to root the phylogeny, which affected the topology. In Su et al. (2015), combined analysis of seven genes from all genomic compartments very weakly supported the consecutive branching order Strombosiaceae, Erythropalaceae and the remaining order, but their analysis excluding the plastid genes weakly supported a monophyletic Strombosiaceae + Erythropalaceae sister to the rest of the order. A conflicting phylogenetic signal from nuclear versus plastid data was also detected in the present study. Analyses of the plastid data give results congruent with combined analyses, whereas analyses of nuclear data recover a variably supported clade with Erythropalaceae, Strombosiaceae, Coulaceae and Octoknemaceae (Figs S1–S3). In previous studies the phylogenetic position of Octoknemaceae, including only the genus Octoknema, has been ambiguous (see Fig. 1). In the seven-gene analysis of Su et al. (2015), Octoknema was found in an unresolved position in relation to two clades including all families other than Erythropalaceae and Strombosiaceae. In Nickrent et al. (2019), Octoknema was placed as a weakly supported sister group to a clade of confirmed parasitic families, but after adding additional sequence data, a sister group relationship to Strombosiaceae was suggested instead (Nickrent, 2020). In the present study, Octoknema is consistently placed as the sister group to Coulaceae (Figs 3 and S1–S3), but support is mostly weak. At the same time, the conflicting position of the Coulaceae + Octoknema clade recovered by nuclear versus plastid data indicates that further data will be needed to shed light on the relationships. Irrespective of the placement of Octoknema and Coulaceae within a clade including Erythropalaceae and Strombosiaceae or as sister to the confirmed parasites and Aptandraceae, the position is consistent with the contention that Octoknema is non-parasitic.

Despite relatively weak support for a clade including Aptandraceae, Ximeniaceae, Olacaceae and the remaining parasites in the combined maximum likelihood analysis (UB 76, but 1 pp from Bayesian inference), all analyses recover these taxa in a monophyletic group (Fig. 3; Supplementary Data Figs S1–S3). However, internal topological differences can be observed both from analyses of different data partitions and to some extent from maximum likelihood versus Bayesian inference. This is consistent with previous observations demonstrating incongruence between data signal from different genomic compartments, but also by the choice of phylogenetic method, primarily parsimony versus maximum likelihood or Bayesian inference (Su et al., 2015). The topological differences observed here have only weak support from the maximum likelihood analyses with no relationship between any of the four groups having a higher support than UB 76 %. However, Bayesian inference tends to assign numerically higher posterior probabilities to the resolved clades (Cummings et al., 2003), thus inflating conflict between the nuclear and plastid data partitions (Fig. S3). Since the amount of nuclear data used in the current study is limited, additional nuclear data are needed for further exploring the potential intergenomic data conflict and ultimately for resolving relationships within the clade. Additional nuclear data may also help shed light on the biological reason(s) for observed data partition incongruence. Independent lineage sorting, chloroplast capture and introgression are well-known causes of incongruence, but with the current data for Santalales, only speculations are possible.

Topological ambiguity of the Santalales phylogeny obviously affects interpretation of the evolution of parasitism. Tentatively assuming that the phylogenetic hypothesis derived from maximum likelihood analysis of the combined data set (Fig. 3) is correct, parasitism probably evolved once, either in the ancestor to the clade including all confirmed parasitic families, or – if Aptandraceae is parasitic – in the common ancestor to the larger clade including both groups. Although haustoria have never been reported from any species of Aptandraceae, this cannot be used to infer that they are not parasitic (Nickrent, 2020), and parasitism has been proposed based on the position of the family in earlier phylogenetic analyses (Malécot et al., 2004). However, the phylogenetic position with Aptandraceae sister to the confirmed parasites is equally consistent with a non-parasitic status of Aptandraceae as suggested by Kuijt and Hansen (2015). Any alternative position of Aptandraceae within the clade of confirmed parasites will complicate the scenario for evolution of parasitism, but only if the family is non-parasitic. Either parasitism would have evolved twice or Aptandraceae would have reverted to an autotrophic habit. The latter scenario has never been demonstrated in any other clade of parasitic plants (Nickrent, 2020).

The non-parasitic status of Erythropalaceae, Strombosiaceae and Coulaceae has been inferred by the absence of haustoria in those few species investigated (Kuijt and Hansen, 2015; Nickrent, 2020). As above, this does not necessarily imply that all members of the families are non-parasitic, but currently no evidence exists for parasitism in the groups. The same applies to Octoknema, where no haustoria have been reported (Gosline and Malécot, 2011). Potential parasitism has been suggested for this family, but entirely based on a previously inferred phylogenetic position among parasitic taxa (Nickrent et al., 2010). However, both the current phylogenetic hypotheses and the relationships suggested in Nickrent (2020) are consistent with a non-parasitic status.

Further ambiguity about the status of parasitism even applies to Olacaceae and Ximeniaceae. Although we have scored both families as parasitic here, as discussed in Nickrent (2020), haustoria have been documented only in Olax, Ptychopetalum and Ximenia. Thus, tracing the origin(s) of parasitism in Santalales precisely still needs both a stronger supported phylogenetic framework and greater knowledge about parasitism or non-parasitism in members of the families now assumed to be non-parasitic or root parasitic.

Gene losses

In concordance with previous studies of Santalales plastomes (Petersen et al., 2015; Zhang et al., 2015; Su and Hu, 2016; G.-S. Yang et al., 2017; Li et al., 2017; Shin and Lee, 2018; Zhu et al., 2018; Guo and Ruan, 2019a, b; Guo et al., 2019, 2020, 2021a, b; Jiang et al., 2019; S.-S. Liu et al., 2019; Schelkunov et al., 2019; Su et al., 2019, 2021; Y. Yang and He, 2019; Yu et al., 2019; Chen et al., 2020; D. Yang et al., 2020; Ceriotti et al., 2021; Nickrent et al., 2021; X. Liu et al., 2021; Al-Juhani et al., 2022; Darshetkar et al., 2023; Tang et al., 2024) as well as with investigations of other lineages of hemiparasites and photosynthetic mycoheterotrophs in general (Wicke et al., 2016; Frailey et al., 2018; Wicke and Naumann, 2018; Kim et al., 2020), hemiparasitic root parasites in the Santalales have undergone relatively few gene losses (Fig. 3). A notable exception is the extensive gene loss reported in Ximeniaceae by Guo et al. (2020), where not only the ndh genes were reported as lost, but also matK, psaC, psbA,I,K, rpl32, rps15, rps16, ycf1 as well as eight tRNA genes. This conclusion was based on the 125 050-bp-long accession MG799332 of Malania oleifera, originally published in Liu et al. (2019), and the unverified accession MN414175 of Ximenia americana, originally published in Chen et al. (2020). Though gene loss was reported in Guo et al. (2020) for Ximenia, the original publication by Chen et al. (2020) reported the presence of all these genes, but they are not annotated in the available GenBank accession. However, local blastN searches (as implemented in Geneious Prime v.2019.0.4) using our new assembly of Ximenia americana confirm that all genes are present in accession MN414175, as originally reported in Chen et al. (2020). Therefore, the loss reported by Guo et al. (2020) for accession MN414175 of Ximenia americana is probably the result of annotation problems. Although we were unable to locate genes reported as missing in Guo et al. (2020) in accession MG799332 of Malania oleifera, they are present in the much longer accession MK764537 (158 163 bp long) published in Yang and He (2019), which we included in the current study because it shows a higher degree of similarity with that of Ximenia americana (both in our new assembly and in accession MN414175).

The phylogenetic relationships obtained here support a single functional loss of the 11 ndh genes from the plastome in the common ancestor to the group formed by confirmed parasites and the family Aptandraceae (Fig. 3; Supplementary Data Fig. S3). If Aptandraceae is parasitic (and Octoknemaceae non-parasitic), the loss of the plastid ndh genes would coincide with the transition to a parasitic lifestyle, further supporting previous claims of correlation between some degree of heterotrophism (haustorial parasitism and mycoheterotrophy) and loss of plastid ndh genes (Wicke et al., 2011; Graham et al., 2017; Wicke and Naumann, 2018; Chen et al., 2020). However, if Aptandraceae is not parasitic, loss of ndh genes pre-dates evolution. Although loss of the plastid ndh genes is common to almost all parasitic and mycoheterotrophic lineages, it has been reported in an increasing number of non-parasitic lineages as well (e.g. Martín and Sabater, 2010; Blazier et al., 2011; Sun et al., 2018; Nevill et al., 2019; Morais da Silva et al., 2021). The plastid ndh genes code for some of the subunits of the NDH complex, involved in the minor pathway of photosystem I cyclic electron transport; however, an alternative major pathway is mediated by the entirely nuclear encoded PGR5/PGRL1 protein complex (Shikanai, 2020; Ma et al., 2021). The partial redundancy of both pathways may explain the recurrent loss of plastid ndh genes even in a number of autotrophs (see e.g. Strand et al., 2019). The importance of the NDH-dependent pathway has been associated with numerous factors of environmental stress, such as low, high or fluctuating light conditions, low, high or fluctuating temperatures, drought, and phosphorus deficiency. Although oxidative stress appears to be a key factor, the complete physiological and possibly dual functions of the NDH complex remain unclear (for recent reviews see Yamori and Shikanai, 2016; Strand et al., 2019; Shikanai, 2020; Ma et al., 2021). Repeated loss of ndh genes and probably the entire NDH complex in several autotrophs may suggest that a similar loss in parasites and other heterotrophs is not related to their lifestyle per se. If loss of ndh genes in Santalales pre-dates evolution of parasitism, it could rather be hypothesized that the loss was a driving force towards a heterotrophic lifestyle. However, this is neither the case in Orobanchaceae, where loss of ndh genes appears to be correlated with the transition to hemiparasitism (Zhang et al., 2020), nor in Krameriaceae, where hemiparasitism has evolved without functional loss of ndh genes (Banerjee et al., 2022).

Among angiosperms, the infA gene, coding for translation initiation factor A, has been lost repeatedly from the plastome (Millen et al., 2001; Daniell et al., 2016). Within Santalales, Chen et al. (2020) reported the loss of infA in all investigated species (both root- and stem-feeding parasites) except Erythropalum scandens and ‘early diverging root-(hemi)parasites’ (Malania oleifera, Ximenia americana, Schoepfia fragrans and S. jasminiodora). Although our findings concur with Chen et al. (2020) regarding the presence of infA in Ximeniaceae and Schoepfia spp., we found it in the stem parasitic Misodendrum brachystachyum as well but not in Olacaceae. In Loranthaceae, the functional loss of infA in Loranthus was reported previously (Nickrent et al., 2021). Similar functional losses apply to most investigated genera of Loranthaceae, but intact infA genes were found in Nuytsia and Moquiniella (Nickrent et al., 2021; Tang et al., 2024). Phylogenetic evidence suggests that infA was lost more than once from the plastome in Loranthaceae (Tang et al., 2024). On the general assumption that acquisition after loss is unlikely, loss of infA has occurred repeatedly within Santalales, and no clear parallel can be drawn between loss and a parasitic lifestyle.

In contrast to the completely lost ndh genes, the presence of homologous genes in the nuclear genome has previously been demonstrated for infA in Arabidopsis thaliana (Brassicaceae), Glycine max (Fabaceae), Solanum lycopersicum (Solanaceae) and Mesembryanthemum crystallinum (Aizoaceae) (Millen et al., 2001). They all have nuclear-encoded infA genes with transit peptide presequences comparable to those of Viscum album, Santalum album and Taxillus chinensis, albeit all are highly dissimilar and probably reflect independent events of transfer from the plastome to the nuclear genome. The pattern of infA presence/absence in Santalales (Fig. 2; Nickrent et al., 2021; Su et al., 2021) suggests that the gene was ancestrally present, but subsequently functionally or completely lost from the plastome at least three times: in Olacaceae, in Loranthaceae (where it might have been lost more than once) and in a clade corresponding to Santalaceae s.l. In the current study the last clade is represented by Nanodeaceae, Thesiaceae and Viscaceae, but also Santalaceae s.s., Amphorogynaceae, Cervantesiaceae and Comandraceae (Nickrent, 2020). These losses are consistent with the data from previously published complete plastomes of the Santalales, including sequences from families being underrepresented or not represented at all in the current analysis (Petersen et al., 2015; Zhang et al., 2015; Su and Hu, 2016; G.-S. Yang et al., 2017; Li et al., 2017; Shin and Lee, 2018; Zhu et al., 2018; Guo and Ruan, 2019a, b; Guo et al., 2019, 2020, 2021a, b; Jiang et al., 2019; S.-S. Liu et al., 2019; Schelkunov et al., 2019; Su et al., 2019, 2021; Y. Yang and He, 2019; Yu et al., 2019; Chen et al., 2020; D. Yang et al., 2020; Ceriotti et al., 2021; Nickrent et al., 2021; X. Liu et al., 2021; Al-Juhani et al., 2022; Darshetkar et al., 2023; Tang et al., 2024). The presence of potentially functional nuclear infA genes in Viscum album, Santalum album and Taxillus chinensis suggests at least one event of intracellular gene transfer, but more independent events cannot be ruled out. The higher similarity of the transit peptide presequences of Viscum and Santalum (see Fig. 4), both members of the Santalaceae s.l. clade, makes a common origin for these sequences likely, but nuclear data from more representatives from the clade would be desirable. After the establishment of a functional nuclear gene, the plastid gene copy is expected to become superfluous and experience relaxed selection, pseudogenization and loss. Indeed, pseudogenes and complete losses are observed in three clades. If intracellular transfer and establishment of a functional nuclear gene occurred only once in a common ancestor to the three clades, it could be expected that Schoepfiaceae and Misodendraceae, which are derived from the same ancestor and still possess plastome infA genes, would show signs of relaxed selection. However, this is not the case and may indicate a scenario of several independent transfer events in clades lacking functional plastome infA genes.

Fig. 4.

Comparison between plastid (P) and putative nuclear (N) infA amino acid sequences. Predicted cleavage sites are in bold/red.

Patterns of relaxed selection

The tests for relaxed and intensified selection provide a mixed picture with only a few clades showing a consistent pattern of relaxation of selection for individual genes or functional groups of genes. At the same time, comparison with data from similar studies of the Santalales or other groups of heterotrophic plants are not straightforward because of differences in taxon sampling, differences in how genes are treated (as individual units or functional groups) and differences in how the tests are performed. Here we surprisingly detected issues with convergence of results among individual runs using RELAX. As a random-effects model, RELAX may be sensitive to small sets of test and reference branches (S. Pond, pers. comm.). Adjusting the settings as described above improved convergence, but a few repeated runs still gave significantly different results. To our knowledge this has not been reported in other studies, which usually report a single run.

Overall there is no clear difference in patterns of relaxation between non-parasitic and parasitic taxa. Given that all parasitic species included here are hemiparasites that might not be expected, although Petersen et al. (2015) did find relaxed selection of photosynthesis genes based on a Santalales sampling limited to Viscum and Osyris (Santalaceae). In Orobanchaceae relaxed selection has also been reported for some hemiparasitic taxa (Wicke et al., 2016; Li et al., 2021). Among the hemiparasitic taxa in the current study, the three stem parasites, Misodendrum, Loranthus and Viscum, show a tendency towards a higher degree of relaxed selection, in particular in photosynthesis genes compared to the root parasitic taxa. The same observation was made by Chen et al. (2020) and may be explained by increased acquisition of host carbon and a reduced need for photosynthesis in stem parasites compared to root parasites (Těšitel, 2016).

Among the non-parasitic clades, Erythropalaceae stands out as having considerably more genes and gene groups showing significantly relaxed selection compared to Strombosiaceae and Coulaceae (and Octoknemaceae, although the parasitic status is not known). In contrast to the three stem parasites, relaxed selection does not involve all photosynthesis-related genes. However, significantly relaxed selection can be observed for the combined group of ndh genes in Erythropalaceae, Octoknemaceae and some Strombosiaceae. Testing for relaxed selection in the individual ndh genes showed only a few significant changes, most involving ndhF. This might suggest that ndhF could be the first gene to be affected by relaxed selection, but with the longest coding region among the 11 ndh genes, it may be that it is easier to detect significant changes for ndhF. The effect of gene length can be seen for Octoknema affinis, where significant relaxed selection was observed for the combined group of ndh genes, but none of the individual gene tests gave significant results.

Although relaxed selection of the ndh gene complement is not observed for all non-parasites, Coulaceae and some Strombosiaceae being exceptions, the results are consistent with a hypothesis that relaxed selection of the ndh genes could lead to their loss in the hemiparasitic clade (including Aptandraceae). In a comparable study of a clade of orchids, Neottieae, including both autotrophs, and partially and fully heterotrophic species, no significantly relaxed selection of ndh genes was detected in the autotrophic species prior to complete or functional loss of the genes in the heterotrophs (Lin et al., 2017; Lallemand et al., 2019). However, relaxed selection may apply to the entire clade of orchids compared to their monocot relatives.

In addition to the loss of ndh genes, three independent losses of infA were observed. As for the ndh loss, none of these losses are pre-dated by relaxed selection of infA. Since infA was probably transferred to the nuclear genome, rather than lost completely, relaxed selection would only be expected to apply to a remaining plastid gene copy following a functionally successful transfer event. Thus, the lack of relaxed selection of infA in all but one of the included representatives of Santalales may support transfer as multiple independent events.

IR size and boundary shifts

Whereas little variation in IR size and boundaries is found among the confirmed non-parasitic species and Octoknema affinis, expansion and contraction of the IR have occurred within the clade including confirmed parasites and Aptandraceae (Fig. 2). Among non-parasites and Octoknema, the size of the IR ranges between 26 046 bp (Heisteria densifrons, Erythropalaceae) and 26 394 bp (Erythropalum scandens), with the IR–LSC (Long Single Copy) junction placed between trnN-GUU and ycf1, and with the IR–SSC junction between rpl2 and rps19 (Fig. 2), which seems to be the ancestral state in angiosperms (Zhu et al., 2016). In the clade including parasites and Aptandraceae, the IR ranges between 12 426 bp (Schoepfia schreberi) and 35 256 bp (Malania oleifera), exhibiting major border shifts and inclusion or exclusion of several genes. The greatest deviations in size, border and gene content of the IR are observed in some root parasitic lineages (Ximeniaceae, Olacaceae, Schoepfiaceae) but not in the stem parasites included here (Viscum, Loranthus, Misodendrum), where dependency on parasitism might possibly be more strongly manifested. Despite our limited sample size of stem parasites, additional support for this observation can be found in Chen et al. (2020) where a much denser sampling of stem parasites revealed only minor differences in IR size and boundaries.

The major expansions of the IR in Ximeniaceae and Olacaceae appear to result from two independent events: a possible shared ancestral shift of the IR–SSC border and a shift of the IR–LSC border in Olacaceae, only. Whereas both these events have resulted in a larger IR gene content, a shift of the IR–LSC boundary in Schoepfiaceae caused a major IR size reduction and several genes are no longer included in the IR region. Within Aptandraceae both contractions and expansions, mostly at the IR–SSC boundary, are also observed but far less pronounced than in the three confirmed stem parasitic families.

Among the representatives of Santalales included here, variation in IR size is largely explained by boundary shifts and only to a minor extent by loss of genes (e.g. ndhB). Similar instability of IR boundaries, mostly leading to IR expansion, has been observed in hemiparasitic Orobanchaceae (Frailey et al., 2018; Li et al., 2021). Orchidaceae also shows IR instability prior to complete loss of one copy in a few fully mycoheterotrophic taxa (Kim et al., 2020). Complete loss of one IR copy was even observed in the hemiparasites Cassytha (Lauraceae) and Cuscuta subgenus Cuscuta (Convolvulaceae) (Wu et al., 2017; Banerjee and Stefanović, 2020; Yu et al., 2023). Owing to a copy-dependent repair mechanism, genes located within the IR regions have lower synonymous substitution rates compared to genes in the single-copy regions (Zhu et al., 2016). Thus, expanding IR boundaries can lead to increased conservation and boundary contraction to decreased conservation of affected genes. In Santalales this probably explains the retention of some ndh pseudogenes in Ximeniaceae and Olacaceae, which are otherwise completely lost. Whether IR instability can be linked to a heterotrophic lifestyle is not clear, and both instability and loss are reported also from autotrophic taxa (see Zhu et al., 2016). Major IR boundary shifts in Santalales are only observed in confirmed parasitic clades, but are not a characteristic of all. Some shifts are even observed in Aptandraceae, but the uncertain parasitic status precludes any firm conclusions. However, IR instability is correlated with the functional loss of ndh genes in the ancestor to Aptandraceae plus confirmed parasites.

CONCLUSION

With the addition of 29 new plastome assemblies, we provide a framework for the study of plastome configuration and evolution in non-parasitic and hemiparasitic Santalales. Phylogenetic patterns indicate a single loss of the plastid ndh genes, and signs of relaxed selection of these genes can be observed prior to the functional loss. Loss of ndh coincides with an apparent instability of the IR borders in the group formed by confirmed parasites and Aptandraceae. However, correlation with the transition to a parasitic lifestyle remains uncertain, as the parasitic status of Aptandraceae and Octoknemaceae is yet to be determined. Although loss of the ndh gene complex is one of the most common and unifying traits of plastomes from parasitic and mycoheterotrophic plants (Wicke et al., 2011; Graham et al., 2017; Wicke and Naumann, 2018; Chen et al., 2020), similar losses are continuously being reported in autotrophic plants (e.g. Martín and Sabater, 2010; Blazier et al., 2011; Sun et al., 2018; Nevill et al., 2019; Morais da Silva et al., 2021). For these reasons, determining the parasitic status of Aptandraceae is therefore crucial for interpreting the potential correlation between parasitism and ndh loss in Santalales. The same applies to potential correlation between parasitism and plastome instability in general.

Additionally, we conclude that repeated loss of infA from the plastome is best explained by a pattern of (repeated) transfer to the nuclear genome rather than repeated functional loss of the gene. Although these transfer events have taken place in parasitic lineages, there is no evidence they are connected to parasitism.

The relatively few gene losses observed in hemiparasites is consistent with most previous findings, indicating that, except for the loss of the plastid encoded subunits of the NDH complex, the function of the plastid remains relatively unaltered in hemiparasites within Santalales, and that efficient photosynthesis is probably maintained and important.

In addition to uncertainty about parasitism in some lineages of Santalales, some of the deeper splits in the phylogeny lack solid support. Additional taxon sampling and molecular data (particularly from the nuclear genome) are desirable to further test phylogenetic hypotheses and potential data partition incongruence.

SUPPLEMENTARY DATA

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

Table S1: Origin of samples and information about reference sequences used for assembly. Table S2: Number of reads, average read depth (ARD) and GenBank accession numbers of new sequences. Table S3: Genes obtained in partially assembled plastomes. Table S4: Detailed results from five runs of RELAX showing relaxed or intensified selection in functional gene groups and individual genes. Table S5: Patterns of relaxed and intensified selection of individual ndh genes. Fig. S1. Phylogeny of Santalales based on ML analysis of plastome gene sequences. Node values are ultrafast bootstrap support. Fig. S2. Phylogeny of Santalales based on ML analysis of nrDNA sequences. Node values are ultrafast bootstrap support. Fig. S3. Bayesian majority rule consensus tree from analysis of the plastid/nuclear/combined alignment. Fig. S4. SplitsTree4 consensus network of the credible set of trees from Bayesian analysis of the plastid/nuclear/combined alignment.

FUNDING

This work was supported by the National Genomics Infrastructure (NGI) in Stockholm and the Swedish National Infrastructure for Computing (SNIC) in Uppsala. SNIC is partially funded by the Swedish Research Council through grant agreement no. 2018-05973. This work was also supported by resources provided by the Pawsey Supercomputing Research Centre with funding from the Australian Government and the Government of Western Australia. The work was further supported by the Department of Ecology, Environment and Plant Sciences, Stockholm University.