Complete plastome sequences from Bertholletia excelsa and 23 related species yield informative markers for Lecythidaceae

Abstract

Premise of the Study

The tropical tree family Lecythidaceae has enormous ecological and economic importance in the Amazon basin. Lecythidaceae species can be difficult to identify without molecular data, however, and phylogenetic relationships within and among the most diverse genera are poorly resolved.

Methods

To develop informative genetic markers for Lecythidaceae, we used genome skimming to de novo assemble the full plastome of the Brazil nut tree (Bertholletia excelsa) and 23 other Lecythidaceae species. Indices of nucleotide diversity and phylogenetic signal were used to identify regions suitable for genetic marker development.

Results

The B. excelsa plastome contained 160,472 bp and was arranged in a quadripartite structure. Using the 24 plastome alignments, we developed primers for 10 coding and non‐coding DNA regions containing exceptional nucleotide diversity and phylogenetic signal. We also developed 19 chloroplast simple sequence repeats for population‐level studies.

Discussion

The coding region ycf1 and the spacer rpl16‐rps3 outperformed plastid DNA markers previously used for barcoding and phylogenetics. Used in a phylogenetic analysis, the matrix of 24 plastomes showed with 100% bootstrap support that Lecythis and Eschweilera are polyphyletic. The plastomes and primers presented in this study will facilitate a broad array of ecological and evolutionary studies in Lecythidaceae.

Lecythidaceae (sensu lato) is a pantropical family of trees with three subfamilies: Foetidioideae, which is restricted to Madagascar; Barringtonioideae, found in the tropical forests of Asia and Africa; and the Neotropical clade Lecythidoideae, which contains approximately 234 of the approximately 278 known species in the broader family (Mori et al., 2007, 2017; Huang et al., 2015; Mori, 2017). Neotropical Lecythidaceae are understory, canopy, or emergent trees with distinctive floral morphology and woody fruit capsules. Among Lecythidaceae species are the iconic Brazil nut tree, Bertholletia excelsa Bonpl.; the oldest documented angiosperm tree, Cariniana micrantha Ducke (dated at >1400 years old in Manaus, Brazil; Chambers et al., 1998); the cauliflorous cannonball tree commonly grown in botanical gardens, Couroupita guianensis Aubl.; and important timber species (e.g., Cariniana legalis (Mart.) Kuntze). Lecythidaceae is the third most abundant family of trees in the Amazon forest, following Fabaceae and Sapotaceae (ter Steege et al., 2013). The most species‐rich genus, Eschweilera Mart. ex DC., with approximately 99 species (Mori, 2017), is also the most abundant tree genus in the Amazon basin (ter Steege et al., 2013), and E. coriacea (DC.) S. A. Mori is the most common tree species in much of Amazonia (ter Steege et al., 2013). Lecythidaceae provide important ecological services such as carbon sequestration and are food resources for pollinators (bats and large bees) and seed dispersers (monkeys and agouties) (Prance and Mori, 1979; Mori and Prance, 1990).

Tools for species‐level identification and phylogenetic analyses of Lecythidaceae could significantly advance research on Amazon tree diversity. However, despite their ease of identification at the family level, species‐level identification of many Lecythidaceae (especially Eschweilera) is notoriously difficult when based on sterile (i.e., without fruit or floral material) herbarium specimens, and flowering of individual trees often occurs only at multi‐year intervals (Mori and Prance, 1987). As a complement to other approaches, DNA barcoding (Dick and Kress, 2009; Dexter et al., 2010) may help to identify species or clades of Lecythidaceae.

A combination of two protein‐coding plastid regions (matK and rbcL) has been proposed as a core plant DNA barcode (Hollingsworth et al., 2009), although other coding and non‐coding plastome regions (psbA‐trnH, rpoB, rpoC1, trnL, and ycf5) and the ITS of nuclear ribosomal genes have been recommended as supplemental barcodes for vascular plants (Kress et al., 2005; Lahaye et al., 2008; Li et al., 2011). However, an evaluation of a subset of these markers (ITS, psbA‐trnH, matK, rbcL, rpoB, rpoC1, and trnL) on Lecythidaceae in French Guiana (Gonzalez et al., 2009) showed poor performance for species identification. Furthermore, the use of traditional markers (plastid ndhF, trnL‐F, and trnH‐psbA, and nuclear ITS) for phylogenetic analysis has produced weakly supported trees (Mori et al., 2007; Huang et al., 2015), indicating a need to develop more informative markers and/or increase molecular sampling.

The main objectives of this study were to (1) assemble, annotate, and characterize the first complete plastome sequence of Lecythidaceae from the iconic Brazil nut tree, B. excelsa; (2) obtain a robust backbone phylogeny for the Neotropical clade using newly assembled draft plastome sequences for an additional 23 species; and (3) develop a novel set of informative molecular markers for DNA barcoding and broader evolutionary studies.

METHODS

Plant material and DNA library preparation

We performed genomic skimming on 24 Lecythidaceae species, including 23 Lecythidoideae and one outgroup species (Barringtonia edulis Seem.) from the Barringtonioideae. The sampling included all 10 Lecythidoideae genera (Appendix 1). Silica‐dried leaf tissue from herbarium‐vouchered collections was collected by Scott Mori and colleagues and loaned by the New York Botanical Garden. Total genomic DNA was extracted from 20 mg of dried leaf tissue using the NucleoSpin Plant II extraction kit (Machery‐Nagel, Bethlehem, Pennsylvania, USA) with SDS lysis buffer. Prior to DNA library preparation, 5 μg of total DNA was fragmented using a Covaris S‐series sonicator (Covaris Inc., Woburn, Massachusetts, USA) following the manufacturer's protocol to obtain approximately 300‐bp insert sizes. We prepared the sequencing library using the NEBNext DNA library Prep Master Mix and Multiplex Oligos for Illumina Sets (New England BioLabs Inc., Ipswich, Massachusetts, USA) according to the manufacturer's protocol. Size selection was carried out prior to PCR using Pippin Prep (Sage Science, Beverly, Massachusetts, USA). Molecular mass of the finished paired‐end library was quantified using an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, California, USA) and by quantitative PCR using an ABI PRISM 7900HT (Thermo Fisher Scientific, Waltham, Massachusetts, USA) at the University of Michigan DNA Sequencing Core (Ann Arbor, Michigan, USA). We sequenced the libraries on one lane of the Illumina HiSeq 2000 (Illumina Inc., San Diego, California, USA) with a paired‐read length of 100 bp.

Plastome assembly

Illumina adapters and barcodes were excised from raw reads using Cutadapt version 1.4.2 (Martin, 2011). Reads were then quality filtered using Prinseq version 0.20.4 (Schmieder and Edwards, 2011), which trimmed 5ʹ and 3ʹ sequence ends with a Phred quality score <20 and removed all trimmed sequences <50 bp in length, with >5% ambiguous bases, or with a mean Phred quality score <20. A combination of de novo and reference‐guided approaches was used to assemble the plastomes. First, chloroplast reads were separated from the raw read pool by BLAST‐searching all raw reads against a database consisting of all complete angiosperm plastome sequences available on GenBank (accessed in 2014). Any aligned reads with an E‐value <1^–5 were retained for subsequent analysis. The filtered chloroplast reads were de novo assembled using Velvet version 7.0.4 (Zerbino and Birney, 2008) with k‐mer values of 71, 81, and 91 using a low‐coverage cutoff of 5 and a minimum contig length of 300. The assembled contigs were then mapped to a reference genome (see below) using Geneious version R8 (Kearse et al., 2012) to determine their order and direction using the reference‐guided assembly tool with medium sensitivity and iterative fine‐tuning options. Finally, raw reads were iteratively mapped onto the draft genome assembly to extend contigs and fill gaps using the low‐sensitivity, reference‐guided assembly in Geneious. We first assembled the draft genome of B. excelsa; the plastomes of the remaining 23 species were assembled subsequently using the plastome of B. excelsa as a reference. The B. excelsa plastome was annotated using DOGMA (Wyman et al., 2004) with the default settings for chloroplast genomes. Codon start and stop positions were determined using the open reading frame finder in Geneious and by comparison with the plastome sequence of Camellia sinensis (L.) Kuntze var. pubilimba Hung T. Chang (GenBank ID: KJ806280). A circular representation of the B. excelsa plastome was made using OGDraw V1.2 (Lohse et al., 2007). The complete annotated plastome of B. excelsa and the draft plastomes of the remaining 23 Lecythidaceae species sampled were deposited into GenBank (Appendix 1).

Identification of molecular markers

Chloroplast simple sequence repeats (cpSSRs) in B. excelsa were identified using the Phobos Tandem Repeat Finder version 3.3.12 (Mayer, 2010) by searching for uninterrupted repeats of nucleotide units of 1 to 6 bp in length, with thresholds of ≥12 mononucleotide, ≥6 dinucleotide, and ≥4 trinucleotide repeats, and ≥3 tetra‐, penta‐, and hexanucleotide repeats (Sablok et al., 2015). We developed primers to amplify the cpSSRs using Primer3 version 2.3.4 (Untergasser et al., 2012) with the default options and setting the PCR product size range between 100 and 300 bp.

The 24 plastomes were aligned with MAFFT version 7.017 (Katoh et al., 2002) and then scanned for regions of high nucleotide diversity (π; Nei, 1987) using a sliding window analysis implemented in DNAsp version 5.10.1 (Librado and Rozas, 2009) with a window and a step size of 600 bp. Levels of nucleotide diversity were plotted using the native R function “plot” (R Core Team, 2017), and windows with values over the 95th percentile were considered of high π.

Taking into account that DNA barcodes can also be used in phylogenetic analyses and because regions with high π do not necessarily have high phylogenetic signal (e.g., unalignable hypervariable regions), we employed a log‐likelihood approach modified from Walker et al. (2017) to identify phylogenetically influential regions. First, we inferred a phylogenetic tree with the plastome alignment (including only one inverted repeat) by performing 100 independent maximum likelihood (ML) searches using a GTRGAMMA model with RAxML version 8.2.9 (Stamatakis, 2014). Those searches resulted in the same topology that was subsequently annotated with the summary from 100 bootstraps using “sumtrees.py” version 4.10 (Sukumaran and Holder, 2010). We then calculated site‐specific log‐likelihoods in the alignment over the plastome phylogeny and calculated their differences site‐wise to the averaged log‐likelihood per site of 1000 randomly permuted trees (tips were randomly shuffled). Log‐likelihood scores were calculated with RAxML using a GTRGAMMA model. The site‐wise log‐likelihood differences (LD) were calculated using 600‐bp non‐overlapping windows with a custom R script (see below). We interpreted greater LD as an indication of greater phylogenetic signal, and windows with an LD above the 95th percentile were considered to have exceptional phylogenetic signal.

Primers flanking the top 10 regions with high π were designed using Primer3 with default program options. We employed a maximum product size of 1300 bp because lower cutoff values (i.e., 600 bp) made the primer design extremely challenging due to the lack of conserved regions. Primers were designed to amplify across all 23 Neotropical species without the use of degenerate bases. However, primers with a small number of degenerate bases were permitted for some regions where primer development otherwise would not have been possible due to high sequence variability in the priming sites. We investigated the potential of our markers to produce robust phylogenies by calculating individual gene trees in RAxML version 8.2.9 in an ML search with 100 rapid bootstraps (option “‐f a”) using the GTRGAMMA model. To evaluate the number of markers needed to obtain a resolved tree with an average of ~90 bootstrap support (BS), we first concatenated the two markers with the highest π and inferred a tree; subsequently, we added the marker with the next highest π score. We iterated this process until we obtained a matrix with each of the 10 markers developed. For every tree obtained, we calculated its average BS and its Robinson–Foulds distance (RF; Robinson and Foulds, 1981) from the plastome phylogeny using a custom R script employing the packages APE (Paradis et al., 2004) and Phangorn (Schliep, 2011). The scripts and alignments used in this study can be found at https://bitbucket.org/oscarvargash/lecythidaceae_plastomes.

RESULTS

Lecythidaceae plastome features

The sequenced plastome of B. excelsa contained 160,472 bp and 115 genes, of which four were rRNAs and 30 were tRNAs (Fig. 1, Table 1). The arrangement of the B. excelsa plastome had a typical angiosperm quadripartite structure with a single‐copy region of 85,830 bp, a small single‐copy region of 16,670 bp, and two inverted repeat regions of 27,481 bp each. Relative to C. sinensis var. pubilimba, we found no gene gain/losses in B. excelsa. The only structural difference we found is that B. excelsa contains the sequential genes trnH‐GUG, rps3, rpl22, and rps19 in the inverted repeat region, whereas C. sinensis var. pubilimba contains these genes in the large single‐copy region. Similarly, no gene gain/losses were found when B. excelsa was compared to other Neotropical Lecythidaceae plastomes assembled herein (Table 2). In addition to B. excelsa, the plastome of Eschweilera alata A. C. Sm. was also completely assembled; the coverage for the remaining plastomes ranged between 85% and 99.60% (Appendix 1).

Figure 1.

Plastome map of the Brazil nut tree, Bertholletia excelsa. Genes outside of the circle are transcribed clockwise; genes inside of the circle are transcribed counterclockwise. Gray bars in the inner ring show the GC content percentage.

Table 1.

Genes contained within the chloroplast genome of Bertholletia excelsa

Function	Gene group	Gene name
Self‐replication	Ribosomal proteins (large subunit)	rpl2, rpl14, rpl16, rpl20, rpl22, rpl23, rpl32, rpl33, rpl36
	Ribosomal proteins (small subunit)	rps2, rps3, rps4, rps7, rps8, rps11, rps12, rps14, rps15, rps16, rps18, rps19
	RNA polymerase subunits	rpoA, rpoB, rpoC1, rpoC2
	Ribosomal RNAs	rrn4.5, rrn5, rrn16, rrn23
	Transfer RNAs	trnA‐UGC, trnC‐GCA, trnD‐GUC, trnE‐UUC, trnF‐GAA, trnG‐GCC, trnG‐UCC, trnH‐GUG, trnI‐CAU, trnI‐GAU, trnK‐UUU, trnL‐CAA, trnL‐UAA, trnL‐UAG, trnfM‐CAU, trnM‐CAU, trnN‐GUU, trnP‐UGG, trnQ‐UUG, trnR‐AGC, trnR‐UCU, trnS‐GCU, trnS‐GGA, trnS‐UGA, trnT‐GGU, trnT‐UGU, trnV‐GAC, trnV‐UAC, trnW‐CCA, trnY‐GUA
Photosynthesis	Photosystem I	psaA, psaB, psaC, psaI, psaJ
	Photosystem II	psbA, psbB, psbC, psbD, psbE, psbF, psbH, psbI, psbJ, psbK, psbL, psbM, psbN, psbT, psbZ
	NADH dehydrogenase	ndhA, ndhB, ndhC, ndhD, ndhE, ndF, ndhG, ndhH, ndhI, ndhJ, ndhK
	Cytochrome b/f complex	petA, petB, petD, petG, petL, petN
	ATP synthase	atpA, atpB, atpE, atpF, atpH, atpI
	RuBisCO large subunit	rbcL
Other genes	Subunit of acetyl‐CoA‐carboxylase	accD
	Envelope membrane protein	cemA
	Protease	clpP
	c‐type cytochrome synthase	ccsA
	Translational initiation factor	infA
	Maturase	matK
Unknown function	Hypothetical chloroplast reading frames	ycf1, ycf2, ycf3, ycf4, yc15

Table 2.

Comparison for plastome subunits for the samples for which the inverted repeats were completely assembled.a

Species	LSC length (bp)	SSC length (bp)	IR length (bp)	GC content (%)	Protein‐coding genes	rRNAs	tRNAs
Allantoma decandra	85,269	18,738	27,618	36.9	81	4	30
A. lineata	85,119	18,756	27,635	36.9	81	4	30
Bertholletia excelsa	85,840	18,950	27,841	36.4	81	4	30
Corythophora amapaensis	85,861	18,778	27,638	36.7	81	4	30
C. labriculata	85,673	18,759	27,594	36.7	81	4	30
Couratari macrosperma	83,785	18,728	27,614	37.0	81	4	30
C. stellata	85,547	18,491	27,576	36.9	81	4	30
Eschweilera alata	85,056	18,721	27,635	36.6	81	4	30
E. caudiculata	84,713	18,759	27,638	37.0	81	4	30
E. congestiflora	84,815	18,167	27,715	37.1	81	4	30
E. integrifolia	84,688	18,796	27,592	36.9	81	4	30
E. micrantha	85,286	18,719	27,668	36.8	81	4	30
E. wachenheimii	85,378	18,815	27,603	36.8	81	4	30
Lecythis pneumatophora	85,506	18,845	27,622	36.7	81	4	30

IR = inverted repeat; LSC = large single‐copy region; SSC = small single‐copy region.

^aLength and GC content of the large single‐copy and small single‐copy regions in partial plastomes are estimates only.

Identification of molecular markers

Within the plastome of B. excelsa we found 23 cpSSRs, 22 of which were in non‐coding regions and one in the ndhD coding region. We designed 19 primer pairs with an acceptable product length, annealing temperature, and GC content for cpSSRs located in non‐coding regions (Table 3). π exceeded the 95th percentile for nine 600‐bp windows (Fig. 2, Tables 4 and 5). Similarly, 13 windows were over the 95th percentile for LD (Fig. 2, Tables 4 and 5), indicating high phylogenetic signal. Although most of the informative windows were in non‐coding regions, two consecutive regions were positioned in the ycf1 gene. Six windows contained both high π and LD. As expected, high π and greater LD largely agreed. Based on the rank of the windows obtained for π, we developed primers for the following regions (ordered from high to low π): ycf1, rpl16‐rps3, psbM‐trnD, ccsA‐ndhD, trnG‐psaB, petD‐rpoA, psbZ‐trnfM, trnE‐trnT, and trnT‐psbD (Table 6).

Table 3.

Primers for the amplification of simple sequence repeats in the plastome of Bertholletia excelsa. All primer pairs amplify non‐coding sequences with the exception of ndhD.a

Forward primer sequence (5′–3′)	Reverse primer sequence (5′–3′)	Repeat unit	Location	Region	No. of repeats	Product size (bp)
CCAAAATCATGAACTAACCCCCA	ACCAAGAGGGCGTTATTGCT	A	396–409	trnH‐psbA	14	226
TGAAGTCGTGTTGCTGAGATCT	CTGTTGATAAGTTTGCCGAGGT	C	3686–3702	trnK intron	17	197
GAGGTTTTCTCCTCGGACGG	ACCACTCATTAAACGAAATGCCT	A	5680–5691	rps16 intron	12	244
GTCCACTCAGCCATCTCTCC	AGCCCGGCCATAGGAATAAA	AAAG	9396–9407	trnS‐trnG	3	297
TTTATTCCTATGGCCGGGCT	TGCATTGTTTAAGAATCCATAGTTTCA	A	9769–9780	trnS‐trnG	12	246
TTTTCCCCACACTTCCCCTC	TGTCCGGTCATTTGATTTGGT	A	17,925–17,938	rps2‐rpoC2	14	192
AAGAGAGGAGAAGTTTTTAGGCA	CCTTACCACTCGGCCATGTC	A	29,392–29,403	rpoB‐trnC	12	232
GGGATGCGAGAAAGAGACTT	CAAAAGTATATCTTTCTACGGGTCG	AAAG	34,775–34,786	trnT‐psbD	3	250
TACCGGTTTTCAAGACCGGG	TCACAAATGGGCATGCTGGA	AAAAT	38,160–38,174	trnS‐psbZ	3	201
ACCCATCAATCATTCGATTCGT	GAAAGATCTTTCCTTGGGGGA	AAAG	47,627–47,638	ycf3‐trnS	3	168
No suitable primers found	No suitable primers found	AAAT	49,610–49,625	trnT‐trnL	4	NA
No suitable primers found	No suitable primers found	AATT	50,016–50,027	trnT‐trnL	3	NA
CCACTGAACAAGGGAGAGCC	ACCAAGGCAAACCCATGGAA	AAAATT	75,475–75,492	clpP‐psbB	3	128
TGAATCACTGCTTTTCTTTGACTCT	AGGCGGTTCTCGAAAGAAGA	AAAAT	77,155–77,169	psbB‐psbT	3	183
TTCAATCTCGGGATTCTTTGAGA	TCGCCTGCGAAAACTTAACT	A	85,073–85,085	rpl16‐trnH	13	246
TCGATCAATCCCTTTGCCCT	CGTACTCCTCGCTCAATGAGA	AAAT	102,172–102,183	rps12‐trnV	3	248
TGGAGCACCTAACAACGCAT	AGACCTCCGGGAAAAGCATG	A	106,208–106,219	trnL intron	12	119
AGAGTAAACACAAGATACAAGGGT	GTGGGTTAGGTCAATCGGGA	AACTT	117,345–117,359	rpl32‐trnL	3	194
AGTCAACGTCAAAATTAATGAATGGT	AGGTTGAACGCGAGCGATAT	AT	117,609–117,622	rpl32‐trnL	7	177
AAATAACTCCCGCGGTCCAG	GCTTCTCTTGCATTACCGGG	AAAT	119,729–119,740	ndhD	3	240
No suitable primers found	No suitable primers found	AAT	122,820–122,831	ndhG‐ndhI	4	NA
No suitable primers found	No suitable primers found	AAT	122,843–122,854	ndhG‐ndhI	4	NA
AACCCGCTTCAAGCCATGAT	AAACGGCTTATAAATTCGCAGT	AATC	125,271–125,282	ndhA intron	3	130

NA = not applicable.

^aSequences have been deposited to GenBank (BioProject SUB2740669).

Figure 2.

Sliding 600‐site window analyses on the Lecythidaceae plastome alignment of 24 species showing nucleotide diversity (π) (top) and alignment site‐wise differences in log‐likelihood (LD) calculated from the chloroplast topology versus the average scores of 1000 random trees (bottom). Regions with π and LD above the 95th percentiles are indicated with dashed lines. Continuous vertical lines indicate the boundaries, from left to right, among the large single copy, the inverted repeat, and the small single copy.

Table 4.

Regions of the plastome alignment (windows of 600 sites) with significantly high (above the 95th percentile) nucleotide diversity and/or site‐wise log‐likelihood score differences.b

Location in the alignment	Bertholletia plastome location	Closest flanking expressed region		Region	π	LD
		5′	3′
1–600	1–490	trnH	psbA	LSC		a
5401–6000	4885–5373	trnK‐UUU	rps16	LSC		a
34,801–35,400	30,925–31,450	petN	trnD‐GUC	LSC		a
35,401–36,000	31,451–31,967	psbM	trnD‐GUC	LSC	a	a
37,201–37,800	33,027–33,573	trnE‐UUC	trnT‐GGU	LSC	a	a
39,601–40,200	34,893–35,433	trnT‐GGU	psbD	LSC		a
43,801–44,400	38,798–39,254	psbZ	trnfM‐CAU	LSC	a	a
44,401–45,000	39,255–39,744	trnfM‐CAU	psaB	LSC	a	a
61,201–61,800	54,771–55,275	trnV‐UAC	atpE	LSC		a
78,601–79,200	70,230–70,771	psaJ	rps18	LSC		a
89,401–90,000	80,536–81,103	petD	rpoA	LSC	a
95,401–96,000	85,455–85,906	rpl16	rps3	LSC	a
131,401–132,000	119,237–119,759	ccsA	ndhD	SSC	a
140,401–141,000	127,827–128,402	rps15	ycf1	SSC		a
144,001–144,600	131,283–131,868	ycf1	ycf1	SSC	a	a
144,601–145,200	131,869–132,446	ycf1	ycf1	SSC	a	a

π = nucleotide diversity (see main text); LD = log‐likelihood score differences; LSC = large single copy; SSC = small single copy.

^aSignifies regions with high (above the 95th percentile) nucleotide diversity or site‐wise log‐likelihood score differences.

^bCoding regions are indicated in windows that have the same 5ʹ‐ and 3ʹ‐expressed flanking regions in column 3. Notice that no regions are reported for the inverted repeat (IR). Coordinates are given on the alignment and the Bertholletia excelsa plastome that are assembled with the standard LSC‐SSC‐IR structure.

Table 5.

Nucleotide diversity and differences in log‐likelihood scores of the informative windows identified in this study and of previously proposed barcode markers

Regiona	πb	LDb
ccsA‐ndhD	0.0258	247.12
matK	0.0153	136.92
petD‐rpoA	0.0246	260.79
petN‐trnD	0.0228	361.07
psaJ‐rps18	0.0176	309.10
psbM‐trnD	0.0292	330.41
psbZ‐trnfM	0.0246	373.97
rbcL	0.0105	95.03
rpl16‐rps3	0.0345	275.89
rpoB	0.0097	120.53
rpoC1	0.0103	178.60
rps15‐ycf1	0.0212	284.57
trnE‐trnT	0.0241	522.51
trnfM‐psaB	0.0254	375.76
trnH‐psbA	0.0126	310.47
trnK‐rps16	0.0164	350.44
trnL	0.0106	192.27
trnT‐psbD	0.0239	291.15
trnV–atpE	0.0128	379.77
ycf1 (1)	0.0273	462.53
ycf1 (2)	0.0469	313.12

π = nucleotide diversity; LD = differences in log‐likelihood scores.

^aInformative windows identified in this study are indicated in bold.

^bHigh values (above the 95th percentile) for π and LD are indicated in bold.

Table 6.

Primer sequences designed to amplify the 10 most polymorphic Lecythidaceae plastome regions, as sorted by decreasing nucleotide diversity

Window in the alignment	π	Region	Forward primer sequence (5′–3′)	Reverse primer sequence (5′–3′)	Length (bp)a
144,103–145,487	0.04691	ycf1(1)	AGAACCTTTGATTATGTCTCGACG	AGAGACATGCTATAAAAATAGCCCA	1186
95,034–95,741	0.03446	rpl16‐rps3	AGAGTTTCTTCTCATCCAGCTCC	GCTTAGTGTGTGACTCGTTGG	1014
35,585–36,413	0.02920	psbM‐trnD	CCGTTCTTTCTTTTCTATAACCTACCC	ACGCTGGTTCAAATCCAGCT	1093
143,235–144,102	0.02733	ycf1(2)	TGATTCGAATCTTTTAGCATTAKAACT	KCGTCGAGACATAATCAAAGGT	1189
131,180–132,054	0.02576	ccsA‐ndhD	CCGAGTGGTTAATAATGCACGT	GCTTCTCTTGCATTACCGGG	1180
44,398–45,132	0.02537	trnG‐psaB	TCGATYCCCGCTATCCGCC	GCCAATTTGATTCGATGGAGAGA	883
89,032–89,688	0.02464	petD‐rpoA	TGGGAGTGTGTGACTTGAACT	TGACCCATCCCTTTAGCCAA	824
43,412–44,397	0.02456	psbZ‐trnfM	TCCAATTGRCTGTTTTTGCATTAATTG	CCTTGAGGTCACGGGTTCAA	706
37,444–38,345	0.02409	trnE‐trnT	AGACGATGGGGGCATACTTG	CCACTTACTTTTTCTTTTGTTTGTTGA	1324
38,346–40,085	0.02391	trnT‐psbD	GGCGTAAGTCATCGGTTCAA	CCCAAAGCGAAATAGGCACA	1717

π = nucleotide diversity.

^aThe product size (length) references the Bertholletia excelsa plastome.

Phylogenetics of the plastomes and the developed markers

The ML analysis of the plastome alignment for Lecythidaceae (145,487 sites) yielded a fully resolved phylogeny with high BS for all clades (Fig. 3). Of the genera in which the sampling included multiple species, Eschweilera and Lecythis Loefl. were polyphyletic, whereas Allantoma Miers, Corythophora R. Knuth, Couratari Aubl., and Gustavia L. were monophyletic (Bertholletia is monospecific, and only one species each of Couroupita Aubl., Cariniana Casar., and Grias L. were included in the analysis). The trees obtained from individual markers with high π had an average BS of 73 throughout their nodes, whereas the trees obtained from two or more concatenated regions had an average BS of 89 (Fig. 4, Appendix S1). None of the gene trees, single or combined (Appendix S1), recovered the topology obtained using the complete plastome matrix (none of the gene trees obtained an RF = 0; Fig. 5). In general, matrices with concatenated markers (mean RF = 6) outperformed single markers (mean RF = 13.8; Fig. 5).

Figure 3.

Maximum likelihood phylogeny inferred from plastomes of 23 Neotropical Lecythidaceae. Numbers at nodes indicate bootstrap support.

Figure 4.

Average bootstrap support for trees inferred from either independent or concatenated regions with high nucleotide diversity, sorted in ascending order.

Figure 5.

Robinson–Foulds distances (RF) for trees inferred from either independent or concatenated regions with high nucleotide diversity, sorted in descending order. Lower RF distances, which measure the number of different taxa bipartitions from the complete plastome topology, indicate better accuracy.

DISCUSSION

Genetic markers from the Lecythidaceae plastome

We are publishing the first full plastome for Lecythidaceae, including high‐depth coverage of the Brazil nut tree (B. excelsa) and 23 draft genomes representing all Lecythidoideae genera and a Paleotropical outgroup taxon. We found no significant gene losses or major rearrangements when the plastome of B. excelsa was compared with that of C. sinensis var. pubilimba, a closely related plastome (Theaceae).

We inferred a robust backbone phylogeny for Lecythidoideae using the 24 aligned plastomes. All nodes in our topology had 100% BS except for a node that connects three closely related species of Eschweilera (Fig. 3). The topology agreed with previous but weakly supported (<50% BS) Lecythidaceae phylogenies based on chloroplast and nuclear ITS sequences (Mori et al., 2007; Huang et al., 2015), indicating that Eschweilera and Lecythis are polyphyletic. Although the polyphyly of these two genera is well supported with all available data, some inferred species‐level relationships may change with increased taxonomic sampling and the inclusion of nuclear genomic data.

We measured π and a proxy for phylogenetic signal using an LD modified from Walker et al. (2017). These calculations helped us to evaluate the performance of specific chloroplast regions as potential phylogenetic markers. The core plant DNA barcodes matK and rbcL did not exhibit high π or LD in our analysis (Table 5). Of the secondary plastome barcodes mentioned in the literature (rpoC1, rpoB, trnL, and psbA‐trnH; Kress et al., 2005; Lahaye et al., 2008; Hollingsworth et al., 2009; Li et al., 2011), only psbA‐trnH showed high LD (Table 5), although it did not exhibit exceptionally high values of π. In contrast, the regions ycf1, rpl16‐rps3, psbM‐trnD, ccsA‐ndhD, trnG‐psaB, petD‐rpoA, psbZ‐trnfM, trnE‐trnT, and trnT‐psbD displayed the highest values of π and LD and therefore outperformed all of the previously proposed plant DNA barcodes.

Phylogenetic trees calculated from concatenated marker sets (based on the π rank) outperformed single regions in terms of support (BS) and accuracy (RF; Figs. 4, 5). In fact, tree topologies using single markers deviated from the complete plastome tree (mean RF = 13.8). The most well‐performing concatenated matrix contained all 10 regions for which we developed primers. However, the combination of ycf1 and rpl16‐rps3 produced an average BS of ~90 (Fig. 4) with reasonable accuracy (RF = 4, Fig. 5); we conclude that these two regions, amplified in three PCRs (Table 6), are promising markers for DNA barcoding, phylogeny, and phylogeography in Lecythidaceae. Although barcoding efficiency in species‐rich clades (i.e., Eschweilera/Lecythis) might decline with the addition of more samples, ycf1 and rpl16‐rps3 effectively distinguished between three closely related species within the Eschweilera parvifolia Mart. ex DC. clade (see branch lengths in Appendix S1), suggesting that these markers might effectively distinguish between many other closely related species. Our results and conclusions agree with those of Dong et al. (2015), who proposed ycf1 as a universal barcode for land plants.

The 19 cpSSR markers developed for noncoding portions of the B. excelsa plastome provide a useful resource for population genetic studies. Because of their fast stepwise mutation rate relative to single‐nucleotide polymorphisms, cpSSRs can also be used for finer‐grain phylogeographic analyses (e.g., Lemes et al., 2010; Twyford et al., 2013). This may be especially useful for species that exhibit little geographic structuring across parts of their ranges. Because they are maternally transmitted and can be variable within populations, the cpSSRs may also be used to track the dispersal of seeds and seedlings relative to the maternal source trees.

Because of their high level of polymorphism and phylogenetic signal content, we anticipate using the cpDNA markers presented here to study the phylogeography of widespread Lecythidaceae species such as Couratari guianensis Aubl. and Eschweilera coriacea, which range from the Amazon basin into Central America.

Barcoding of tropical trees

The DNA barcoding of tropical trees has been useful for several applications (Dick and Kress, 2009), including community phylogenetic analyses (Kress et al., 2009), inferring the species identity of the gut content (diet) of herbivores (García‐Robledo et al., 2013), and for species identification of seedlings (Gonzalez et al., 2009). The power of DNA barcodes to discriminate among species should be high if the studied species are distantly related; for example, Kress et al. (2009) were able to discriminate 281 of 296 tree and shrub species from Barro Colorado Island using standard DNA barcodes, but they were not able to discriminate among some congeneric species in the species‐rich genera Inga Mill. (Fabaceae), Ficus L. (Moraceae), and Piper L. (Piperaceae). Gonzalez et al. (2009) encountered similar challenges with Eschweilera species in their study of trees and seedlings in Paracou, French Guiana. The latter study tested a wide range of putative DNA barcode regions (rbcLa, rpoC1, rpoB, matK, trnL, psbA‐trnH, and ITS) but did not include the markers presented in this article.

Limitations of plastome markers for phylogeny and species identification

These newly identified plastome markers are not free of limitations. First, plastome‐based phylogenies should be interpreted with caution, as they can disagree with nuclear markers and species trees as a result of introgression and/or lineage‐sorting issues (Rieseberg and Soltis, 1991; Sun et al., 2015; Vargas et al., 2017). These same processes limit the cpDNA for species identification. For example, cpDNA haplotypes of Nothofagus Blume, Eucalyptus L'Hér., Quercus L., Betula L., and Acer L. were more strongly determined by geographic location than by species identity because of the occurrence of localized introgression within these groups (Petit et al., 1993; Palme et al., 2004; Saeki et al., 2011; Premoli et al., 2012; Nevill et al., 2014; Thomson et al., 2015). To date, the occurrence of haplotype sharing in closely related Lecythidaceae species has not been examined at a large scale and it is therefore not possible to conclude to what extent introgression or incomplete lineage sorting might affect this group. We suggest that future studies utilizing cpDNA barcodes for Neotropical Lecythidaceae test species from several shared geographic localities to examine to what extent haplotypes tend to be shared among species at the same localities. Nuclear DNA markers may also be used to examine phylogenetic incongruence and to identify cases where introgression might have occurred.

DATA ACCESSIBILITY

DNA sequences have been deposited to GenBank (accession no. MF359935–MF359958 and BioProject SUB2740669). Plastome alignment, gene alignments, trees, and R code are available at https://bitbucket.org/oscarvargash/lecythidaceae_plastomes .

Supporting information

 

APPENDIX 1.

Lecythidaceae species sequenced with their voucher, assembly information, and GenBank accession number. All voucher specimens are deposited at the New York Botanical Garden Herbarium (Bronx, New York, USA).

Species	Voucher	No. of reads	% of ref seqa	Mean coverage	No. of contigs	Average length of assembled contigs (bp)	Minimum contig length (bp)	Maximum contig length (bp)	N50	GenBank accession no.
Allantoma decandra (Ducke) S. A. Mori, Ya Y. Huang & Prance	Mori 25640	527,449	99.20	213	9	157,957	1052	42,447	22,733	MF359949
A. lineata (Mart. & O. Berg) Miers	Chevalier 10101	697,746	99.60	295	8	158,449	400	34,633	32,463	MF359941
Barringtonia edulis Seem.	Tsou 1552	519,377	96.10	230	10	152,805	2636	46,991	32,608	MF359956
Bertholletia excelsa Bonpl.	Mori 25637	1,036,874	100	646	14	160,472	461	160,472	160,472	MF359948
Cariniana estrellensis (Raddi) Kuntze	Nee 52828	759,042	85	292	70	130,037	278	22,237	3803	MF359938
Corythophora amapaensis Pires ex S. A. Mori & Prance	Mori 24148	690,545	99.60	302	4	159,222	6643	75,002	42,750	MF359955
C. labriculata (Eyma) S. A. Mori & Prance	Mori 25518	606,728	99.60	260	5	158,819	6596	74,896	42,691	MF359946
Couratari macrosperma A. C. Sm.	Janovec 2506	340,696	99	144	9	156,981	1107	45,975	42,740	MF359944
C. stellata A. C. Sm.	Mori 24093	493,777	99.40	211	6	158,312	1374	109,344	109,344	MF359936
Couroupita guianensis Aubl.	Mori 25516	503,417	96.50	314	12	154,792	1071	47,693	18,977	MF359935
Eschweilera alata A. C. Sm.	Prévost 4607	851,683	100	358	2	158,981	61,051	97,930	97,929	MF359940
E. caudiculata R. Knuth	Cornejo 8185	273,053	98.30	116	11	156,630	1117	37,154	21,598	MF359957
E. integrifolia (Ruiz & Pav. ex Miers) R. Knuth	Cornejo 8211	440,144	99	187	9	157,206	1105	42,497	21,591	MF359942
E. micrantha (O. Berg) Miers	Mori 25410	289,775	98.90	120	11	157,890	1143	42,551	18,021	MF359958
E. pittieri R. Knuth	Cornejo 8208	160,625	97	166	8	154,547	551	74,876	24,636	MF359954
E. wachenheimii (Benoist) Sandwith	Prévost 4252	367,631	98.90	151	11	157,757	1179	37,141	21,748	MF359939
Grias cauliflora L.	Aguilar 7961	520,480	94.90	326	41	150,768	314	28,189	8102	MF359952
Gustavia augusta L.	Mori 24255	761,640	95.50	476	33	152,601	358	28,353	11,657	MF359943
G. serrata S. A. Mori	Cornejo 8184	534,143	98.90	334	10	157,746	1035	35,586	22,762	MF359947
Lecythis ampla Miers	Cornejo 8229	606,518	94.10	241	39	149,464	348	28,759	6527	MF359951
L. congestiflora Benoist	Molino 2019	1,073,567	96.90	405	27	154,882	302	21,264	11,580	MF359937
L. corrugata Poit.	Mori 24265	544,831	90.70	243	50	143,859	317	28,741	4496	MF359950
L. minor Jacq.	Tsou 1542	666,355	94.80	416	26	151,568	354	31,188	14,768	MF359945
L. pneumatophora S. A. Mori	Mori 25728	690,202	99.50	301	4	158,832	11,782	75,019	49,184	MF359953

^aPercentage of the sequence recovered in relation to Bertholletia excelsa.

Thomson, A. M. , Vargas O. M., and Dick C. W.. 2018. Complete plastome sequences from Bertholletia excelsa and 23 related species yield informative markers for Lecythidaceae. Applications in Plant Sciences 6(5): e1151.