Phylogenetic utility of the nuclear genes AGAMOUS 1 and PHYTOCHROME B in palms (Arecaceae): an example within Bactridinae

Abstract

Background and Aims

Molecular phylogenetic studies of palms (Arecaceae) have not yet provided a fully resolved phylogeny of the family. There is a need to increase the current set of markers to resolve difficult groups such as the Neotropical subtribe Bactridinae (Arecoideae: Cocoseae). We propose the use of two single-copy nuclear genes as valuable tools for palm phylogenetics.

Methods

New primers were developed for the amplification of the AGAMOUS 1 (AG1) and PHYTOCHROME B (PHYB) genes. For the AGAMOUS gene, the paralogue 1 of Elaeis guineensis (EgAG1) was targeted. The region amplified contained coding sequences between the MIKC K and C MADS-box domains. For the PHYB gene, exon 1 (partial sequence) was first amplified in palm species using published degenerate primers for Poaceae, and then specific palm primers were designed. The two gene portions were sequenced in 22 species of palms representing all genera of Bactridinae, with emphasis on Astrocaryum and Hexopetion, the status of the latter genus still being debated.

Key Results

The new primers designed allow consistent amplification and high-quality sequencing within the palm family. The two loci studied produced more variability than chloroplast loci and equally or less variability than PRK, RPBII and ITS nuclear markers. The phylogenetic structure obtained with AG1 and PHYB genes provides new insights into intergeneric relationships within the Bactridinae and the intrageneric structure of Astrocaryum. The Hexopetion clade was recovered as monophyletic with both markers and was weakly supported as sister to Astrocaryum sensu stricto in the combined analysis. The rare Astrocaryum minus formed a species complex with Astrocaryum gynacanthum. Moreover, both AG1 and PHYB contain a microsatellite that could have further uses in species delimitation and population genetics.

Conclusions

AG1 and PHYB provide additional phylogenetic information within the palm family, and should prove useful in combination with other genes to improve the resolution of palm phylogenies.

INTRODUCTION

The family Arecaceae (Palmae) comprises approx. 183 genera with a predominantly tropical distribution (Dransfield et al., 2008). The monophyly of the Arecaceae is well established and phylogenetic studies published over the last 20 years have provided good evidence about relationships within the familly (Asmussen and Chase, 2001; Wikström et al., 2001; Davis et al., 2004; Asmussen et al., 2006; Chase et al., 2006; Dransfield et al., 2008). Nevertheless, resolution and support at different taxonomic levels still need to be improved. The genus-level phylogenetic synthesis of the palm family based on supermatrix and supertree methods proposed by Baker et al. (2009) provided a valuable framework supporting a comprehensive classification of palms. Nevertheless, the fact that their study was biased toward plastid genome information suggested that a wider exploration of all plant genomes, in particular the nuclear, could improve resolution and support of the Arecaceae phylogeny.

Nuclear genes are more difficult to optimize as phylogenetic markers than chloroplast genes, due to PCR constraints, heterozygosity and paralogy (Baker et al., 2000a, b; Lewis and Doyle, 2002). However, very early in the history of palm phylogeny studies, low copy nuclear genes became regarded as the largest potential source of phylogenetic markers (Lewis and Doyle, 2001). The first nuclear markers specifically developed for palms were the ADHA and ADHB paralogues of the alcohol dehydrogenase gene family (Morton et al., 1996), but they were not subsequently used. Lewis and Doyle (2001, 2002) then characterized a portion of the malate synthase (MS) gene comprising intron 2 and part of flanking exons 2–3, and a portion of the phosphoribulokinase gene (PRK), paralogue 2, located between exons 3 and 5. Subsequently, the MS marker was little used until the development of new primers to study the phylogeny of Livistona (Crisp et al., 2010). The PRK marker (around intron 4) has been extensively used, either alone or combined with the sequence of intron 23 of RPB2, the gene encoding the second largest subunit of RNA POLYMERASE II. To date, PRK and RPB2 are the only nuclear markers widely used in Arecaceae phylogeny studies (Gunn, 2004; Roncal et al., 2005; Loo et al., 2006; Norup et al., 2006; Savolainen et al., 2006; Thomas et al., 2006; Trénel et al., 2007; Cuenca et al., 2008), although these genes remain less documented in Coryphoideae (Roncal et al., 2008) and little characterized for Calamoideae (Savolainen et al., 2006). A few other genes were recently used in the phylogeny of particular groups, namely the WRKY transcription factor gene loci in the Cocoseae (Meerow et al., 2009), and nuclear intron-spanning loci in Chamaedorea (Bacon et al., 2008). Repetitive nuclear ribosomal DNA cistron markers, although widely used in plant phylogeny in general, have proved problematic in palms (Lewis and Doyle, 2001). Internal transcribed spacers (ITS) of the 18S–26S nrDNA, and the non-transcribed spacer of the 5S nrDNA have been used mostly in the study of one subfamily, Calamoideae, and have posed difficulties in interpreting highly repeated paralogous sequences (Baker et al., 2000a, b). The ITS region has recently, however, attracted new interest (Trénel, 2007; Eiserhardt et al., 2011), despite persistent problems of sequence heterogeneity and some incongruences with other datasets. In one individual study, 18S nrDNA sequences were used to investigate intergeneric and higher-level relationships (Hahn, 2002a).

In this context, we investigated the phylogenetic usefulness of two nuclear genes not previously used for this purpose in palms, namely AG1, orthologues of the oil palm AGAMOUS homologue Eg-AG1, and a PHYTOCHROME B (PHYB) gene. Both loci belong to low-copy gene families, and unlike the earlier mentioned genes that all have basic and conserved functions in primary metabolism, AGAMOUS and PHYTOCHROME genes have more specific functions related to morphogenesis and response to environmental stimuli. As such, they are prone to selection in relation to particular adaptative responses (García-Gil et al., 2003; Kramer et al., 2004; Ingvarsson et al., 2006; Ikeda and Setoguchi, 2010), and their diversity may in some cases be reactive to environmental changes (Saïdou et al., 2009). These two genes should therefore complement the patterns recovered with previously used nuclear genes in palm phylogeny studies.

AGAMOUS genes belong to the MADS-box family, a group of genes that code for transcription factors in eukaryotes (Messenguy and Dubois, 2003). In plants, MADS-box genes play a key role in floral morphogenesis, regulating both the initiation of flowering and the identity of the inflorescence meristem and floral organs. They are also involved in a number of other processes, including root development (Riechmann and Meyerowitz, 1997; Theissen et al., 2000). In model plants such as Arabidopsis and Antirrhinum, AGAMOUS-type genes (in association with other MADS-box genes) are required for the specification of the identity of stamens, carpels and ovules (C/D functions in the widened ABC model for floral organ identity determination: Coen and Meyerowitz, 1991; Angenent and Colombo, 1996; Pelaz et al., 2000; Honma and Goto, 2001).

Recently, MADS-box genes were characterized in oil palm (Elaeis guineensis) and two AGAMOUS gene paralogues, EgAG1 and EgAG2, were identified in this species (Adam et al., 2006, 2007).

Phytochromes are photoreceptive signalling proteins involved in light-sensitive processes in flowering plants (Balasubramanian et al., 2006; Levskaya et al., 2009). Components A and B of the phytochrome photoreceptor system are present in all seed plants and are involved in the perception and response to red/far-red light (Mathews, 2010). In Arabidopsis thaliana, PHYB is involved in the control of the shade-avoidance response (Nagatani et al., 1991; Whitelam and Smith, 1991). Its natural variation in accessions of this species produces a diversity of responses to light. The PHYB gene is important from a physiological point of view because of its pivotal role linked to the functions of other genes of the flowering receptor network. Together with PHYTOCHROME A (PHYA), it is involved in the regulation of various light processes. PHYA promotes germination in far-red light, whereas PHYB controls germination in the dark. PHYA contributes to the induction of flowering by long days in A. thaliana, whereas PHYB inhibits flowering without obviously affecting the sensing of day length (Reed et al., 1994). In rice, phyB mutants have an earlier flowering behaviour (similar to phyC mutants under long day conditions) but only PHYB hastens the flowering time during short days.

We used the partial sequence of AG1 and PHYB genes to investigate intergeneric and interspecific relationships within a set of species belonging to the Neotropical palm tribe Bactridinae. This group is composed of five or six genera, the status of one of them, Hexopetion, being disputed. Hexopetion has been considered either a valid genus of Bactridinae, composed of two Central American species (Pintaud et al., 2008; Millán and Kahn, 2010; Eiserhardt et al., 2011), or a synonym of Astrocaryum (Dransfield et al., 2008). Three genera of the Bactridinae show important radiations. The genus Bactris (approx. 80 species) is particularly speciose in Amazonia, and to a lesser extent in Central America and the Brazilian Atlantic forest (Henderson, 2000). Diversification in Bactris seems to be driven by niche specialization, in particular adaptation to the deep shade of rain forests, through miniaturization and clonality (Kahn and de Granville, 1992). Aiphanes (approx. 30 species) diversified along and around the Andes over an extended altitudinal gradient (0–3000 m), forming many locally endemic species (Borchsenius and Bernal, 1996). Astrocaryum (40 species) probably diversified allopatrically in the western Amazon region and in inter-Andean valleys in relation to physical barriers produced as a consequence of the Andean orogenesis (Pintaud and Kahn, 2002), as well as in the Cerrado formation of Brazil where it adapted to dry conditions (Kahn, 2008). Astrocaryum still poses several taxonomic problems. One of them is the identity of Astrocaryum minus, a species described from a single collection in western Brazil in the 19th Century, and recently found as scattered individuals in French Guiana (Kahn and de Granville, 1998).

Desmoncus is unusual in being the only Neotropical palm genus specialized in the lianescent habit, having developed morphological adaptations such as the cirrus, which parallels those of the unrelated Palaeotropical Calamoid rattans (Dransfield et al., 2008). Acrocomia is a widely distributed Neotropical genus growing in dry, open habitats.

Molecular studies covering some or all of the Bactridinae genera (Hahn, 2002b; Gunn, 2004; Roncal et al., 2005; Asmussen et al., 2006; Couvreur et al., 2007; Baker et al., 2009; Eiserhardt et al., 2011) have produced conflicting topologies. Here we evaluate the usefulness of the AG1 and PHYB genes as phylogenetic markers with a sampling of 22 species of Bactridinae. This study provided some new indications concerning intergeneric relationships and phylogenetic structure within Astrocaryum.

MATERIALS AND METHODS

Biological material

A total of 32 samples representing 22 species and seven genera of the subtribe Bactridinae were used in this study (Table 1). Young leaf tissue was collected and silica gel-dried or lyophilized, then reduced to a fine powder using either a IKA A10 analytical mill (IKA, Staufen, Germany) or a Qiagen TissueLyser II bead mill (Qiagen, Courtaboeuf, France). Total genomic DNA was extracted from 25–40 mg of leaf powder using a DNeasy Plant Mini Kit and a Qiacube robot system (Qiagen). DNA concentrations obtained were measured with a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).

Table 1.

Material studied, including sample identity in the DNA bank of IRD, herbarium voucher number and herbarium where the sample was deposited, GenBank accession numbers of sequences generated for AG1 and PHYB and samples used in the combined analysis of the two genes

Genus and species	IRD DNA bank no.	Herbarium voucher	AG1	PHYB	AG1 and PHYB combined
Acrocomia aculeata	JCP288	Baker 1000 (FTG)	JF414917	–	–
Acrocomia crispa	JCP269	Roncal 79 (FTG)	JF414915	JF422042	+
Aiphanes minima	JCP286	Zona 873 (FTG)	JF414916	JF422043	+
Astrocaryum chambira	B16	Balslev 4812 (K)	JF414909	JF422038	–
Astrocaryum chambira	JCP366	Kahn 2447 (USM)	JF414932	JF812601	+
Astrocaryum gynacanthum	JCP897	Pérez 788 (CAY)	JF414924	JF422053	+
Astrocaryum gynacanthum	JCP1522	Pérez 934 (CAY)	JF414925	JF422057	–
Astrocaryum huaimi	A11/JCP1502	FTBG 347 (FTG)	JF414905	JF422056	+
Astrocaryum jauari	JCP367	Mejía 66 (USM)	JF414922	JF422052	+
Astrocaryum malybo	NDJ109	Jiménez 109 (COL)	JF414929	JF422059	+
Astrocaryum minus	JCP903	de Granville 12921 (CAY)	JF414930	JF422055	+
Astrocaryum minus	JCP1521	de Granville 17659 (CAY)	JF414904	–	–
Astrocaryum minus	JCP1525	de Granville 17660 (CAY)	JF414927	–	–
Astrocaryum minus	JCP1528	de Granville 17661 (CAY)	JF414928	–	–
Astrocaryum murumuru	JCP310	Pérez 718 (CAY)	JF414918	JF422046	+
Astrocaryum paramaca	JCP339	de Granville 5407 (CAY)	JF414919	JF422048	+
Astrocaryum paramaca	JCP900/1523	Pérez 787 (CAY)	JF414926	JF422054	–
Astrocaryum rodriguesii	A65	de Granville 9110 (CAY)	JF414906	–	–
Astrocaryum rodriguesii	JCP341	de Granville 11199 (CAY)	JF414920	JF422049	+
Astrocaryum sciophilum	B40	de Granville 11074 (CAY)	JF414912	–	–
Astrocaryum sciophilum	JCP312	Pérez 786 (CAY)	JF812600	JF422047	+
Astrocaryum standleyanum	JCP1548	Pintaud 605 (G)	–	JF422058	–
Astrocaryum urostachys	B33/JCP356	Balslev 4344 (AAU)	JF414911	JF422050	–
Astrocaryum urostachys	JCP363	Millán 934 (USM)	JF414921	JF422051	+
Astrocaryum vulgare	B43/JCP372	de Granville 10004 (CAY)	JF414913	JF812603	+
Bactris gasipaes	JCP305	de Granville 17288 (CAY)	–	JF422045	+
Bactris major	B1	Balslev 6737 (AAU)	JF414907	–	+
Desmoncus orthacanthos	JCP289	Balslev 6576 (AAU)	JF414931	JF422044	+
Elaeis oleifera	JCP169/179	Balslev 6555 (AAU)	JF414914	JF422041	+
Hexopetion alatum	B15/JCP371	Zona 921 (FTG)	JF414908	JF422040	+
Hexopetion mexicanum	B19	1984-5298 (K)	JF414910	JF422039	–
Hexopetion mexicanum	JCP370	Pintaud 637 (USM)	JF414923	JF812602	+

Primer design

For the AG1 gene, primers AG1-F1 (CAGGAATTTGATGGGAGAGTC) and AG1-R1 (GCTGATTGCTTTGCATGAG) were designed in exons of the Eg-AG1 gene (GenBank accession no. AY739698), downstream from the 3-kb intron 2 (Fig. 1A): the F1 primer was anchored in exon 4, within the K domain region, and the R1 primer was anchored in exon 7, within the region of the C-terminal domain (Kramer et al., 1998). Primer locations corresponded to regions in Eg-AG1-type genes that were both highly conserved within this paralogue and different from Eg-AG2, to ensure consistent and specific amplification (N. Chabrillange et al., unpubl. res.). Primer design, calculation of melting temperature and PCR conditions were determined using Primer3 plus web interface 21 (Untergasser et al., 2007). The amplicon of approx. 630 bp straddles four exons and three introns (Fig. 1B).

Fig. 1.

Structure of the AGAMOUS gene. (A) Outline of the complete sequence in Arabidopsis (from TAIR database locus AT4G18960). (B) Detail of the portion of AG1 sequenced in palms. Arrows indicate the location of primers.

PHYB sequences were initially amplified using degenerate primers designed within exon 1 for Poaceae (Mathews et al., 2000), giving an amplicon of 1180 bp (Fig. 2A). Once some palm sequences were obtained, specific palm primers were designed using Primer3 (Rozen and Skaletsky, 2000) implemented in the Geneious Pro 5·1 software package (Biomatters Ltd, Auckland, New Zealand), on the basis of a consensus Bactridinae sequence, giving an amplicon of 1010 bp: PhyBact-F, CGCTCCCTGGCGGTGACATC; and PhyBact-R, CCTCCGGGTGATGCTTCGCA.

Fig. 2.

Structure of the PHYB gene. (A) Outline of the complete sequence in rice (Oryza sativa GenBank accession no. AB183525), showing the location of the fragment studied in exon 1. (B) Amino-acid sequence variation in the fragment sequenced. Black curve: mean differentiation (from 0 to 1) between Arecaceae and Poaceae, calculated for successive 20-site segments. Horizontal dotted line: mean differentiation for the whole locus. Dashed curve: mean gap percentage per 20-site segments (locating the locus encoded by the microsatellite). Red curve: mean number of amino-acidic changes per site within Arecaceae for successive 20-site segments. Dotted rectangles: highly conserved (left) and highly variable (right) regions.

PCR reactions and cloning

PCR reactions were performed in a final volume of 25 µL with 40 ng of DNA, 0·4 µL of 25 µm primers, Failsafe Premix E 1× and 1 U of Failsafe polymerase mix (Epicentre Biotechnologies, Madison, WI, USA). The amplification programme for the AG1 gene was as follows: initial denaturation for 3 min followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at 58 °C for 60 s and extension at 72 °C for 45 s. A final extension step at 72 °C for 3 min was added to the programme. The amplification programme for PHYB consisted of an initial denaturation for 3 min followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at 64 °C for 90 s, extension at 72 °C for 60 s and a final extension at 72 °C for 10 min.

PCR products were either sequenced directly or cloned prior to sequencing. Cloning was required for some PHYB products obtained with the degenerate primers, and also for several AG1 products, this gene being often heterozygous with allele length variation due to the presence of a dinucleotidic microsatellite in intron 7. Cloning was performed using 7 µL of PCR products and the pGEMT-easy kit (Promega, Charbonnières-les-bains, France) and JM109 High Efficiency Competent Cells (Promega). Five to ten colonies were screened by PCR to identify successful transformants. Colonies of successfully transformed cells were used as template for a new PCR with M13 primers. PCR products were used for sequencing reactions after purification.

Sequencing

PCR products were purified using Ampure kits (Agencourt Bioscience, Beckman Coulter Genomics, Grenoble, France). Their concentration was measured with a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific Inc.). Sequencing reactions were performed using the BigDye v3·1 Terminator kit (Applied Biosystems, Foster City, CA, USA). Sequence reactions were purified with a CleanSeq kit (Agencourt Bioscience, Beckman Coulter Genomics) and read on an ABI 3130 XL automated sequencer (Applied Biosystems). Forward and reverse sequences for each sample were obtained.

Phylogenetic analyses and statistics on sequences

Sequence contigs and editing, sequence alignment, coding sequence translation and Megablast searches were performed with the Geneious Pro 5·1 software package (Biomatters Ltd). All sequences described here were deposited in GenBank (Table 1).

Sequence variability of the AG1 and PHYB genes was compared with previously published sequences (Eiserhardt et al., 2011) of chloroplast (matK, trnL-F, rps16 intron, trnQ-rps16, trnH-psbA, trnD-trnT) and nuclear (PRK, RPB2 and ITS) regions, for a set of six Bactridinae species: Acrocomia crispa, Aiphanes minima, Desmoncus orthacanthos, Hexopetion mexicanum, Astrocaryum chambira and Astrocaryum gynacanthum. Pair-wise uncorrected p-distance values were computed using PAUP* 10 software (Swofford, 2002) for each locus, differentiating intergeneric and interspecific variability (pairwise comparison between the two Astrocaryum species). Separate and combined phylogenetic analyses were conducted with the sequences obtained for the AG1 and PHYB genes.

The maximum-parsimony (MP) method was used as implemented in the PAUP* program (Swofford, 2002) using an heuristic search with 1000 replicates of random taxon-addition, tree bisection-reconnection (TBR) branch swapping, retaining all equally most-parsimonious trees (MULPARS) and unordered characters. A limit of ten trees was set for each replicate. After completing the replicates, all trees found were then used as starting trees for another round of swapping with a tree limit of 5000. Relative support for each branch was assessed by 1000 bootstrap replications and TBR branch swapping. Strict consensus trees were generated. Maximum-likelihood (ML) analysis was conducted with PhyML (Guindon and Gascuel, 2003), using the GTR model, starting tree build with BIONJ and subsequent NNI branch swapping. One hundred bootstrap replicates were completed to evaluate branch support. A Bayesian phylogenetic analysis was carried out using MR. BAYES v.3·1. (Ronquist and Huelsenbeck, 2003). The program was run under the GTR model with gamma-distributed rate variation for 10 000 000 generations in all cases, sampling a tree every 1000 generations. Burn-in was set at 25 % for AG1 and PHYB analyses, and at 14 % for the combined analysis.

For PHYB analyses, five Poaceae sequences from GenBank were aligned with the obtained palm sequences and served as the outgroup. For the AGAMOUS and combined analyses, Elaeis oleifera was used as the outgroup (Elaeidinae being the sister clade to Bactridinae, Hahn, 2002b). Microsatellite regions, which were present in both genes, were excluded from the phylogenetic analyses. Prior to combining the PHYB and AG1 datasets for matching individuals, congruence among datasets was evaluated by searching for conflicting clades significantly supported in either analysis (Wiens, 1998). Two focused analyses relevant to each dataset were also conducted. With the AG1 matrix, an unrooted distance tree was computed with the program SeaView (Gouy et al., 2010) using the BIONJ method and Jukes–Cantor distance, among all specimens of Astrocaryum subgenus Munbaca, to evaluate the diversity and relationships of Astrocaryum minus. With the PHYB translation matrix, the mean divergence in conceptual amino-acid sequence between Arecaceae and Poaceae was calculated for successive segments of 20 sites to visualize variation in sequence conservation along exon 1.

RESULTS

AG1 gene

Thirty-eight sequences of approx. 630 bp were obtained. After verification by MP analysis that the different clones of a same individual clustered together, a single sequence per specimen was retained for phylogenetic analysis, giving a dataset of 30 samples.

A Megablast search identified Elaeis guineensis AGAMOUS paralogue 1 (Adam et al., 2006) as identical to our Elaeis oleifera AG1 genomic sequence. However, the comparison was limited to the exon segments, as EgAG1 is represented in GenBank by an mRNA sequence only (AY739698). The Megablast search also resulted in a hit on a Euterpe edulis microsatellite clone (EE48, AF328885), with flanking sequences sharing 94·6 % similarity with our E. oleifera sequence.

Within the Bactridinae, the exon and intron components of the fragment sequenced displayed 92 and 86·5 % invariable sites, respectively, 5·8 and 7·4 % parsimony-non-informative variable sites, and 2·2 and 6·1 % parsimony-informative sites (microsatellite excluded).

In the ML analysis [optimal tree: ln(L) = –1286·4), Desmoncus, Acrocomia and Aiphanes formed a clade [bootstrap probability (BP) 92 %], Hexopetion and Astrocaryum (both monophyletic) formed another, weakly supported clade, while the position of Bactris was equivocal. Within Astrocaryum, the subgenus Astrocaryum remained unresolved, the subgenus Munbaca was monophyletic (BP 89 %) but without significant internal structure, and the sections Monogynanthus and Huicungo of the subgenus Monogynanthus were each monophyletic with moderate support, but their relationships with other clades remained unresolved (Fig. 3).

Fig. 3.

Phylogenetic analysis of AG1 dataset. (A) Cladogram representation of the ML tree with ML bootstrap values on branches. (B) Outline of the same tree with branch lengths. Abbreviations: right side, typological classification: A, Astrocaryum subgen. Astrocaryum; Mh, subgen. Monogynanthus sect. Huicungo; Mm, subgen. Monogynanthus sect. Monogynanthus; Mu, subgen. Munbaca. Left side, biogeography–ecology: P, Pacific coast of north-west South America and Central America; WA, western Amazonia; EA, eastern and northern Amazonia; PA, peri-Amazonian–peri-sylvatic species; G, Guyanan region (Guianas and towards central Amazonia).

Parsimony analysis resulted in more than 10 000 trees of length L = 220, consistency index (CI) = 0·868 and retention index (RI) = 0·866. In the strict consensus tree (not shown), Hexopetion was resolved as monophyletic (BP 95 %) and formed a weakly supported clade with Astrocaryum. High sequence variation was observed among Astrocaryum minus samples.

The Bayesian phylogenetic analysis recovered a clade including Desmoncus, Aiphanes and Acrocomia (100 % posterior probability), and a clade grouping the three Hexopetion samples (100 %), but other intergeneric relationships remained unresolved (tree not shown).

Distance analysis within subgenus Munbaca grouped individuals of section Munbaca (A. minus and A. gynacanthum) and individuals of section Munbacusu (A. paramaca and A. rodriguesii). There was significant heterogeneity among samples of A. minus, which are intermixed with those of A. gynacanthum. Two plants from Mont Grand Matoury (JCP 1521 and JCP 1525) grouped together (an adult and a putative juvenile offspring), while an isolated adult individual from Montagne Tortue (JCP 1528) was rather distant (Fig. 4).

Fig. 4.

Unrooted BIONJ phylogram computed from the AG1 dataset using Jukes–Cantor distance, depicting relationships among individuals of Astrocaryum subgenus Munbaca and A. sciophilum (Guianan clade).

The four individuals of A. minus and the two individuals of A. gynacanthum sequenced uniquely shared an amino-acid change: a P (proline) vs. L (leucine) in other species in position 124 with respect to the Eg-AG1 translation (accession AY739698).

A highly variable dinucleotide microsatellite (TC/GA)_n was located in intron 7. This microsatellite is generally perfect in Astrocaryum and Acrocomia. Astrocaryum vulgare and A. malybo samples displayed the longest microsatellites, with 26 repeats of the motif, whereas Acrocomia aculeata had the shortest, with nine repeats. The motif is often followed by a stretch of C/G homopolymer, and was seen to be altered in Elaeis, Desmoncus, Aiphanes and Bactris. Astrocaryum minus showed allelic diversity. The AG1 microsatellite has been serendipitously captured in an enriched (TC/GA)_n genomic library of Euterpe edulis (Gaiotto et al., 2001). The clone in question, EE48, aligned perfectly with our sequences over a 280-bp region at the 3′ end of the segment studied, and had 27 perfect repetitions of the motif. It was the longest motif and the most variable locus (with 19 alleles among 66 individuals) of the 18 microsatellite loci characterized for this species (Gaiotto et al., 2001).

PHYTOCHROME B gene

Twenty-five sequences of approx. 1 kb within the PHYB exon 1 were obtained.

Our Megablast search of GenBank data identified Poaceae PHYB sequences as the most similar to our PHYB Bactridinae sequences (there were no other PHYB monocotyledon sequences found in GenBank). Similarity values between Bactridinae and Poaceae ranged from 79·4 to 75·4 % at the nucleotide level and from 87·3 to 85·4 % in the corresponding amino-acid sequence (Fig. 2B).

Translated protein alignments revealed variations in sequence conservation along exon 1 (Fig. 2B). Some potentially informative amino-acid changes and corresponding non-synonymous substitutions were detected. Desmoncus was particularly divergent, although it still displayed a typical PHYB sequence structure (78 % similarity with the Hakonechloa macra PHYB sequence). Desmoncus showed four autapomorphic amino-acid changes, compared with one for Elaeis, two for Acrocomia, Bactris and Hexopetion, three for Aiphanes, but none for Astrocaryum. Desmoncus also displayed two potential symplesiomorphies shared with Elaeidinae and Poaceae but not with the other Bactridinae.

Within Bactridinae, 89·6 % of the sites were invariable, 7·6 % were parsimony-non-informative (of which Desmoncus orthacanthos alone accounted for 3 %) and 2·8 % were parsimony-informative (microsatellites excluded).

MP analysis of the PHYB nucleotide matrix resulted in 8000 trees of length L = 627, CI = 0·815 and RI = 0·868, and recovered a weakly supported Hexopetion–Bactris–Astrocaryum clade containing a highly supported Hexopetion clade (99 % bootstrap support) and a monophyletic Astrocaryum with low clade support and little internal resolution (strict consensus tree not shown). The Bactridinae were not resolved as monophyletic because of the divergent Desmoncus sequence that appeared in the basal position of the Arecaceae clade.

The optimal ML tree had a log likelihood of –4317·5, and a topology similar to that of the MP trees, but with better support for a monophyletic Astrocaryum (BP 92 %). In relation to the typological classification of Astrocaryum (Kahn, 2008), species of the subgenus Monogynanthus sect. Huicungo grouped together with high support (BP 95 %), the subgenus Astrocaryum was polyphyletic and there was a weakly supported clade grouping all species of the subgenus Munbaca and subgenus Monogynanthus sect. Monogynanthus (Fig. 5).

Fig. 5.

Phylogenetic analysis of the PHYB dataset. (A) Cladogram representation of the ML tree with ML bootstrap values on branches. (B) Outline of the same tree with branch lengths. Abbreviations as in Fig. 3.

Bayesian phylogenetic analysis produced the same topology with a posterior probability value of 64 % for the Hexopetion–Bactris–Astrocaryum clade, 64 % for the Astrocaryum clade and 99 % for the Hexopetion clade (tree not shown).

A composite trinucleotide microsatellite (AGC)_0–3(GGC)_1–2(GAC)_2–4(GGC)_1–10 was found in this exon. Interestingly, this structure in incipient in Poaceae with a synonymous motif (AGT)₂(GGT)₂(GAT)_2–3(GGC)₂. In palms, the last motif shows considerable tandem repeat extension (up to ten units) but no intraspecific variation. However, the microsatellite showed high interspecific variability, allowing the separation of closely related species pairs such as A. minus from A. gynacanthum and A. huaimi from A. vulgare. As all the motifs are trinucleotidic, the variation in number of tandem repeats does not alter the reading frame, and results in a variable length amino acid motif S_1–3G_1–2D_2–4G_1–10 in both Arecaceae and Poaceae, but with few repeats and little variation in the latter.

Combined analysis

No strongly conflicting relationships were detected between the individual AG1 and PHYB analyses (e.g. Desmoncus sister to all other species, 35 % BP in the PHYB ML analysis; Acrocomia–Desmoncus clade 22 % BP in AG1 ML analysis).

Twenty individuals each representing a different species were then used in the combined AG1 and PHYB analysis. MP analysis resulted in three most-parsimonious trees of length L = 191, CI = 0·911, RI = 0·837. Desmoncus formed a basal branch followed by an Aiphanes–Acrocomia clade and then a Hexopetion–Bactris–Astrocaryum clade. Hexopetion was monophyletic (BP 100 %) as well as Astrocaryum (BP 64 %), with A. minus sister to A. gynacanthum (BP 70 %).

ML analysis [ln(L) = –4129·4] gave a similar topology except that there was a Desmoncus–Aiphanes–Acrocomia clade instead of an isolated basal branch for Desmoncus. The combined ML analysis resulted in a higher branch support for Astrocaryum (BP 92 %), and also for a Guyanan clade associating subgenus Munbaca and section Monogynanthus of subgenus Monogynanthus, including A. minus sister to A. gynacanthum with 90 % BP support (Fig. 6).

Fig. 6.

Combined ML analysis of AG1 and PHYB data for one individual per species (see Table 1), *except Bactris represented by B. major for AG1 and B. gasipaes for PHYB. Phylogram representation of the ML tree with branch lengths and ML bootstrap values on branches. Abbreviations as in Fig. 3.

The Bayesian phylogenetic analysis of the combined dataset resulted in the same topology as in the ML analysis, with strong support for the Hexopetion–Bactris–Astrocaryum clade (99 % posterior probability). There was moderate support for a sister relationship between Astrocaryum and Hexopetion (78 %), while each genus was strongly supported individually (99 % for the Astrocaryum clade and 100 % for the Hexopetion clade).

Comparison of gene variability

Uncorrected p-distance values showed that the chloroplast sequences were overall less variable than the nuclear sequences, with PHYB less variable than the other nuclear sequences studied, AG1 showing the same amount of intergeneric differentiation as PRK, and both PHYB and AG1 being slightly more variable than RPB2 at the interspecific level, when comparing Astrocaryum chambira with A. gynacanthum (Fig. 7).

Fig. 7.

Comparison of sequence divergence among genera of Bactridinae (open circles) and among species of Astrocaryum (solid squares), as revealed by different markers used in palm phylogeny including AG1 and PHYB from the present study.

DISCUSSION

AG1 and PHYB as phylogenetic markers

The AG1 and PHYB genes amplified consistently in palms with the primers used, and Megablast checks confirmed expected sequence affinities. The levels and patterns of nucleotide diversity obtained complement those of previously studied genes. The combination of AG1 and PHYB improves resolution and branch support compared with individual gene trees. These results confirm the usefulness of AG1 and PHYB as new nuclear phylogenetic markers for palms. The phylogeny obtained with a set of species of Bactridinae is in agreement with most relationships recovered by published molecular studies based on plastid and nuclear genes (Eiserhardt et al., 2011) but also provides some new insights, such as a possible sister relationship between Astrocaryum subgenus Munbaca and section Monogynanthus of subgenus Monogynanthus.

The major discrepancy observed was the divergent PHYB sequence of Desmoncus. Such isolated inconsistencies have been observed with other nuclear genes. For example, Astrocaryum paramaca failed to cluster with the other species of Astrocaryum subgenus Munbaca in a multigene analysis of Bactridinae due to a highly divergent RPB2 sequence for this species (Eiserhardt et al., 2011), whereas a Munbaca clade including A. paramaca is obtained in the present analysis. These occasional inconsistencies could be buffered by combining several genes.

Intergeneric relationships in the Bactridinae

The two Hexopetion species formed a highly supported monophyletic group with both genes that was separate from all other species of the currently recognized Astrocaryum genus. The latter formed a well-defined monophyletic group that was weakly supported as a sister clade to Hexopetion. These results, together with those obtained by Eiserhardt et al. (2011) and Millán and Kahn (2010), support the separation of Hexopetion as a distinct genus (Pintaud et al., 2008).

Overall inter-generic relationships are still uncertain in the Bactridinae. The three comprehensive studies carried out on this group to date led to different topologies: a supermatrix analysis (Baker et al., 2009) gave (Acrocomia ((Aiphanes Desmoncus) (Bactris, Astrocaryum + Hexopetion))); a combined cpDNA, ITS, PRK and RPB2 analysis (Eiserhardt et al., 2011) gave ((Acrocomia Desmoncus) (Aiphanes(Bactris(Astrocaryum Hexopetion)))); whereas the PHYB + AG1 analysis (present study) gave ((Desmoncus(Acrocomia Aiphanes)) (Bactris(Astrocaryum Hexopetion))). Overall, a tendency to resolve Acrocomia and/or Desmoncus in the basal position can be noted. The conservation of some putative plesiomorphies not found in other Bactridinae in the PHYB amino-acid sequence of Desmoncus argues in favour of this topology. Bactris, Hexopetion and Astrocaryum are grouped in all combined analyses. Finally, Aiphanes has the most unstable position, differing in all three analyses.

The tree topology obtained by Eiserhardt et al. (2011) suggests an early radiation of Bactridinae into four lineages approx. 35 Mya (Astrocaryum–Hexopetion, Bactris, Aiphanes and Acrocomia–Desmoncus). These deep short branches, possibly associated with incomplete lineage sorting, partly explain the poor resolution of intergeneric relationships. The inclusion of a larger number of genes and the use of analytical methods specifically addressing this evolutionary pattern (Maddison and Knowles, 2006; Carstens and Knowles, 2007) will be needed to obtain a better understanding of the group.

Phylogenetic structure within Astrocaryum

The PHYB data identify the Central American/western Andean clade A. malybo–A. standleyanum recovered by Eiserhardt et al. (2011) and also group the two peri-Amazonian species growing in woodlands, A. vulgare and A. huaimi, in agreement with the close affinity of these two species suggested by Henderson et al. (1995). They form a clade with A. jauari, another species of forest margins (river banks). The PHYB analysis revealed another biogeographically consistent clade, grouping A. sciophilum (subgenus Monogynanthus sect. Monogynanthus) with subgenus Munbaca, all these species being centred on the eastern Guianan shield. This grouping also appeared in a similarity analysis of leaf anatomical characters by Millán and Kahn (2010), but these authors interpreted this association of two different infrageneric taxa as the result of a morphological convergence. In the AG1 analysis, the clades recovered conform to the morphological classification (subgenus Munbaca, subgenus Monogynanthus sect. Monogynanthus and sect. Huicungo). Although the signal is not very strong, a tendency can be noted for PHYB phylogenies to be generally more consistent with the biogeography and ecology of the species and AG1 phylogenies with the classical taxonomy. It is interesting to note in this connection, although it is not necessarily significant, that the morphological classification is primarily based on pistillate flower structure, in which the AG1 gene is likely to exert a regulatory function, whereas the PHYB gene is most probably involved in the environmental–ecological responses of the species. Overall, it can be seen that the combined AG1–PHYB analysis consolidates the Guianan and peri-Amazonian clades.

The status of Astrocaryum minus

The single sample of Astrocaryum minus used in the PHYB analysis has autapomorphic non-synonymous substitutions and a unique microsatellite sequence. This sample is resolved as sister to A. gynacanthum in the combined analysis with significant support (BP 90 %), in accordance with the morphological classification (they together form the section Munbaca of subgenus Munbaca). When several individuals of A. minus are incorporated, however, as is the case in the AG1 analysis, they appear to show marked differences and together form a complex with A. gynacanthum. This complex is supported by a synapomorphic non-synonymous substitution. It is still difficult to interpret the nature of the relationships between these two species. They are clearly distinct morphologically (Kahn and de Granville, 1998), but A. gynacanthum is widespread and abundant whereas A. minus is extremely rare and scattered (Kahn, 2008), the latter very rarely setting fruit (J.-J. de Granville, Herbier de Guyane (CAY), French Guiana, pers. comm.,). Population genetic studies will be needed to obtain a better understanding of this complex.

Conclusions

The AG1 and PHYB genes appear to be good candidates to increase phylogenetic resolution at various taxonomic levels in palms, due to their structural diversity (exons; introns; microsatellites), potentially different patterns of molecular evolution (neutral; potentially submitted to selection), and contrasting functions, which might allow them to provide complementary data. They are a useful addition to the panel of phylogenetic markers now available for the palm family.