The Subtelomeric khipu Satellite Repeat from Phaseolus vulgaris: Lessons Learned from the Genome Analysis of the Andean Genotype G19833

Subtelomeric regions in eukaryotic organisms are known for harboring species-specific tandemly repeated satellite sequences. However, studies on the molecular organization and evolution of subtelomeric repeats are scarce, especially in plants. Khipu is a satellite DNA of 528-bp repeat unit, specific of the Phaseolus genus, with a subtelomeric distribution in common bean, P. vulgaris. To investigate the genomic organization and the evolution of khipu, we performed genome-wide analysis on the complete genome sequence of the common bean genotype G19833. We identified 2,460 khipu units located at most distal ends of the sequenced regions. Khipu units are arranged in discrete blocks of 2–55 copies and are heterogeneously distributed among the different chromosome ends of G19833 (from 0 to 555 khipus units per chromosome arm). Phylogenetically related khipu units are spread between numerous chromosome ends, suggesting frequent exchanges between non-homologous subtelomeres. However, most subclades contain numerous khipu units from only one or few chromosome ends indicating that local duplication is also driving khipu expansion. Unexpectedly, we also identified 81 khipu units located at centromeres. All the centromeric khipu units belong to a single divergent clade also comprised of a few units from several subtelomeres, suggesting that a few sequence exchanges between centromeres and subtelomeres took place in the common bean genome. The divergence and low copy number of these centromeric units from the subtelomeric units could explain why they were not detected by FISH (Fluorescence in situ Hybridization) although it can not be excluded that these centromeric units may have resulted from errors in the pseudomolecule assembly. Altogether our data highlight extensive sequence exchanges in subtelomeres between non-homologous chromosomes in common bean and confirm that subtelomeres represent one of the most dynamic and rapidly evolving regions in eukaryotic genomes.

Subtelomeric regions in eukaryotic organisms are known for harboring species-specific tandemly repeated satellite sequences. However, studies on the molecular organization and evolution of subtelomeric repeats are scarce, especially in plants. Khipu is a satellite DNA of 528-bp repeat unit, specific of the Phaseolus genus, with a subtelomeric distribution in common bean, P. vulgaris. To investigate the genomic organization and the evolution of khipu, we performed genome-wide analysis on the complete genome sequence of the common bean genotype G19833. We identified 2,460 khipu units located at most distal ends of the sequenced regions. Khipu units are arranged in discrete blocks of 2-55 copies and are heterogeneously distributed among the different chromosome ends of G19833 (from 0 to 555 khipus units per chromosome arm). Phylogenetically related khipu units are spread between numerous chromosome ends, suggesting frequent exchanges between non-homologous subtelomeres. However, most subclades contain numerous khipu units from only one or few chromosome ends indicating that local duplication is also driving khipu expansion. Unexpectedly, we also identified 81 khipu units located at centromeres. All the centromeric khipu units belong to a single divergent clade also comprised of a few units from several subtelomeres, suggesting that a few sequence exchanges between centromeres and subtelomeres took place in the common bean genome. The divergence and low copy number of these centromeric units from the subtelomeric units could explain why they were not detected by FISH (Fluorescence in situ Hybridization) although it can not be excluded that these centromeric units may have resulted from errors in the pseudomolecule assembly. Altogether our data highlight extensive sequence exchanges in subtelomeres between non-homologous chromosomes in common bean and confirm that subtelomeres represent one of the most dynamic and rapidly evolving regions in eukaryotic genomes.

INTRODUCTION
Common bean (Phaseolus vulgaris) is a major source of protein for human consumption in many parts of the world (FAO 1980), especially in developing countries such as tropical areas of Latin America and Eastern Africa where common bean is one of the major staple crops (Pastor-Corrales and Tu, 1989;Broughton et al., 2003). Together with sorghum, millet, groundnut, cowpea, chickpea, pigeonpea, cassava, yam, and sweet potato, common bean is often referred to as an "orphan crop." Indeed, even if common bean is an important crop in developing countries, it is not extensively traded and receives less attention from researchers compared to crops such as maize, rice, and wheat (Varshney et al., 2012). Common bean has a small diploid genome (2n = 22) of 588 Mb (Bennett and Leitch, 1995) including a large amount of repeated sequences (Schlueter et al., 2008;Pedrosa-Harand et al., 2009) compared with other legume species with larger genome sizes, such as Trifolium repens (956 Mb; Bennett and Smith, 1991) and soybean (1,103 Mb; Bennett and Leitch, 1997). Recently, the revolution in sequencing technologies has allowed the establishment of full genome sequencing programs for orphan crops like common bean and the full genome is now available (since July 2012 1 ; Jackson et al., in preparation). The selected common bean genotype is "G19833," an Andean landrace for which a BAC library was used to construct a draft physical map (Schlueter et al., 2008).
Satellite DNA can be defined as highly reiterated non-coding DNA sequences, organized as long arrays of head-to-tail linked repeats located in the constitutive heterochromatin (Plohl et al., 2008). Despite their ubiquity in eukaryotic genomes, the function of such repeats is poorly understood. Early hypotheses considered them to be non-functional "selfish" DNA that proliferate for their own sake or as useless genomic elements accumulated as "junk" with no selective advantage to the organism (Ohno, 1972;Orgel and Crick, 1980). More recently, identification of satellite DNA at structurally important parts of chromosomes, such as centromeres, has suggested functional roles of satellite DNA (Ma and Jackson, 2006).
Satellite DNA is an important component of the knobs, which are cytologically visible regions of highly condensed chromatin (heterochromatin) that are distinct from pericentromeric regions in pachytene chromosomes (Fransz et al., 2000). In common bean, a 528-bp subtelomeric satellite repeat, referred to as khipu, has been identified . Khipu is present on most chromosomal terminal knobs and is specific to the Phaseolus genus Geffroy et al., 2009). Subtelomeric satellite repeats have been reported in different plant and animal species. Indeed, cytologically confirmed subtelomeric satellite repeats have been identified in various plant species, including potato (Torres et al., 2011), rice (Cheng et al., 2001), tomato (Lapitan et al., 1989), maize (Li et al., 2009), barley (Brandes et al., 1995), tobacco (Kenton et al., 1993;Chen et al., 1997), rye (Vershinin et al., 1995), Silene latifolia (Buzek et al., 1997), and Beta species (Dechyeva and Schmidt, 2006). The subtelomeric locations of these repeats were confirmed mostly by FISH (fluorescence in situ hybridization) experiments and were not based on sequence analysis. Except in rice, where sequencing and characterization of the structure of the subtelomeric TrsA sequences were conducted (Ohmido and Fukui, 1997;Mizuno et al., 2006Mizuno et al., , 2008, sequence-based analysis of the molecular organization and evolution of subtelomeric repeats is rare.
In the present paper, we conducted genome-wide analysis to investigate the physical organization and the evolution of khipu sequence based on the complete genome sequence of common bean genotype G19833.

DATA SOURCES
We used the "Phaseolus vulgaris v1.0" genome sequence of the Andean common bean genotype G19833 2 and sequenced BAC clones from G19833 reported in Innes et al. (2008) that correspond to ∼1 Mb of the Co-2 cluster, located at the end of the long arm of chromosome 11, and from Chen et al. (2010), corresponding to 239 kb located at one end of chromosome 5, referred to as PvA05A.

KHIPU ANNOTATION
The khipu satellite DNA was recovered using hmmsearch 3 (Eddy, 1998) with a khipu profile previously defined on 92 khipu . In order to work with a "clean" set of khipu, we excluded the first and last khipu element from blocks of tandemly organized khipu. In addition, we excluded khipu < 500 pb and having "n"s in their sequence. The resulting data was imported into the annotation platform Artemis for manual analysis (Rutherford et al., 2000).
Centromeric positions in individual pseudomolecules were identified by BLASTN using centromere satellite repeats; CentPv1 for chromosomes 01, 02, 03, 04, 07, 08, 09, and 10 and CentPv2 for chromosomes 05, 06, and 11 (Jackson et al., in preparation). Each khipu sequence extracted from the genome of G19833 was named Pvxxyk#####, with Pv referring to Phaseolus vulgaris, "xx" referring to pseudomolecule (01-11), "y" corresponding to the location on the chromosome (S for short arm, C for centromere, and L for long arm), "k" referring to "khipu," and a 5-digit number referring to the khipu order on the pseudomolecules from the start (5 ) to the end (3 ). khipu elements on each pseudomolecule were sequentially numbered in increments of 10. Each khipu extracted from BAC sequences was named PvAxxzk##### (Axxz is the name of the contig), with Pv referring to Phaseolus vulgaris, A referring to Andean (these BAC come from the Andean genotype G19833), xx referring to chromosome (05 or 11), z is a letter given to the different contigs from the same chromosome, k referring to khipu, and 5-digit number referring to the khipu order on the contig in increments of 10. For example, khipu named Pv01Sk00010 is the first khipu unit on pseudomolecule 01, located on short arm and PvA05Ak00010 is the first khipu unit of the BAC contig A from chromosome 05.
To determine the coordinates of BACs on the pseudomolecules, we performed a BLASTN of the entire BAC sequence against the genome sequence. BLAST results were inspected manually to set the start and end positions of the BAC on the pseudomolecules. BAC sequences and pseudomolecules were aligned and visualized using Mauve, a genome alignment tool, using the minimal match seed weight value 4 (Darling et al., 2004).
A Maximum-Likelihood tree was made with FastTree 2.1.3 program (Price et al., 2010) with the Jukes-Cantor model of nucleotide evolution. Bootstrap values were computed with the consensus of 100 random trees using the Phylip's consense program (Felsenstein, 1989) and the random trees were calculated with the "−n" option of the FastTree program over a list of bootstrapped sequences generated from the original sequence alignment using Seqboot in the PHYLIP package. The resulting phylogenetic tree ( Figure S3 in Supplementary Material) was displayed using MEGA version 5 (Tamura et al., 2011). One representative khipu sequence from each major clade of the phylogenetic tree, is provided in Figure S2 in Supplementary Material.

CYTOGENETIC ANALYSIS
The khipu probe was generated using a pool of five subclones from BAC clones ( Table A1 in Appendix). Three subclones come from different khipu blocks spread over sequenced BAC clones from the long arm of chromosome 11 (Innes et al., 2008), one additional subclone come from a subtelomeric BAC from the short arm of chromosome 5 (Chen et al., 2010), and the 1H04 subclone come from the B4 locus (short arm of chromosome 4 from the Mesoamerican BAT93 genotype) described in David et al. (2009). Pachytene chromosomes were prepared from young flower buds of G19833 and JaloEEP558 fixed in ethanol:acetic acid (3:1, v/v). Buds were macerated in 2% cellulase/2% pectolyase/2% cytohelicase in 0.01 M citric acid-sodium citrate buffer, pH 4.8, for 3 h at 37˚C, incubated in 60% acetic acid up to 2 h, and squashed after removal of petals and sepals and flaming. Slide selection and pretreatment, chromosome and probe denaturation and hybridization, posthybridization washes, detection, and image analyses were performed according to Fonseca et al. (2010).

GENOME-WIDE IDENTIFICATION OF THE KHIPU SATELLITE REPEAT IN THE PSEUDOMOLECULES OF COMMON BEAN G19833
To identify khipu sequences in the common bean genome, we used the P. vulgaris genome v1.0 from Phytozome 6 . A total of 2766 khipu units were identified. After size selection of khipu units > 500 bp and discarding khipu units with "n"s, we had 2460 khipu units for further analysis. The distribution of these 2460 khipu units in the common bean genome is presented in Table 1. The short arm of chromosome 01 (chr01S), Chr04S, Chr04L (L = long arm), Chr05S, Chr10L, and Chr11L, had the largest number of khipu units, with 193, 349, 261, 169, 435, and 555, respectively. Chr08S and Chr09L, however, had fewer than 10 khipu units and Chr06S and Chr09S were devoid of khipu. These data are in general agreement with the cytogenetic distribution of BAC 63H6, which contains khipu (K.G.B. dos Santos, personal communication) in G19833 (Altrock et al., 2011), except that Chr10S and Chr11S seem to have large amounts of khipu (as estimated based on the intensity of FISH signals) and no khipu signal could be detected in Chr09L. In agreement with previous FISH analysis on BAT93 showing the subtelomeric distribution of khipu , khipu units were mainly located in the first or last five megabase-pairs of the pseudomolecules. These khipu were organized in tandem arrays with varying numbers units, referred to as khipu blocks. For example, chr04S had 28 khipu blocks containing fewer than 13 units and nine blocks containing more than 13 units, within the first 4.7 Mb of the pseudomolecule. In rice, similar organization in discrete clusters of 3-103 copies in a chromosome specific manner was also observed for the TrsA subtelomeric repeats (Mizuno et al., 2008). In common bean, the largest khipu blocks were found on Chr04S and Chr11L, where khipu blocks had 45 and 55 khipu units, respectively. Notably, Chr04S and Chr11L contain the B4 ) and Co-2 disease resistance gene clusters (Innes et al., 2008;David et al., 2009;Chen et al., 2010) suggesting a possible link between the evolution of resistance clusters and the khipu sequences. As expected for a subtelomeric repeat, few or no khipu (<13) were identified in centromeric regions, with the exception of chromosome 8 centromere which had 55 khipu sequences.

SPREAD OF KHIPU TO NON-HOMOLOGOUS CHROMOSOME ENDS
To study khipu satellite evolution in the common bean genome, a phylogenetic tree was constructed (Figure 1). A multiple alignment including the 2460 khipu units from the G19833 genome plus 201 units from G19833 BACs (Innes et al., 2008;Chen et al., 2010) were first screened for recombination events using RDP4 and recombinant sequences removed (76 and 4 khipu units from G19833 genome and BACs, respectively). Khipu units fell into eight major clades (A-H) with Bootstrap support >75% (Figure 1). Except for the small clade F that contains only khipu units from Chr11L, Chr04S and Chr10L, each major clade contains khipu units from most chromosome ends. This indicates that phylogenetically related khipu units were spread among the chromosome ends, suggesting frequent exchanges between nonhomologous subtelomeres. Within a chromosomal cluster, khipu units were found across the tree, indicating that phylogenetically distant khipu units are physically close to each other. This is particularly striking for Chr10L (blue) where khipu elements are spread across the eight clades. The distribution of khipu units coming from BAC clones reinforces these results. For example, khipu units coming from the BAC from Chr11L (∼1 Mbp; Innes et al., 2008) are spread across 24 subclades belonging to five of the eight major clades (black asterisks in Figure 1). Thus, almost www.frontiersin.org Note that centromeric khipu repeats (black) are found only in two small subclades highlighted by brackets. Clades comprising khipu repeats from previously sequenced BAC contigs from chromosome 11 long arm or chromosome 5 short arm are highlighted by black or red asterisks, respectively. Clades comprising khipu repeats used for FISH experiments are highlighted by FISH written in red.
the entire khipu diversity is represented in a single genomic region as small as ∼1 Mbp. Additionally, BAC PvA05A (239kbp; Chen et al., 2010) from Chr05S contains a khipu block bearing phylogenetically distant khipu units (Figure 1; Figure A1 in Appendix). Indeed, this khipu block is composed of 11 complete units from major clade E (light blue, PvA05Ak00020-PvA05Ak00120) followed by an array of 27 khipu units from major clade D (green, PvA05Ak00160-PvA05Ak00430).
Together these results indicate that each subtelomere contains a patchwork of phylogenetically distant khipu units that is likely the result of shuffling between non-homologous chromosome ends. However, most subclades contain numerous khipu units from only one or few chromosome ends (Figure 1). Thus, in addition to khipu spreading between non-homologous loci, local duplication is driving khipu expansion. In the human genome, extensive cytogenetic and sequence analyses revealed that subtelomeres are hot Frontiers in Plant Science | Plant Genetics and Genomics spots of interchromosomal recombination and segmental duplications (Linardopoulou et al., 2005). This exceptional dynamic activity of subtelomeres has been reported in such diverse organisms as yeast and the malaria parasite Plasmodium (Louis, 1995;Freitas-Junior et al., 2000. As expected for a plastic region of the genome subject to reshuffling through recombination events, subtelomeres exhibit unusually high levels of within-species structural and nucleotide polymorphism (Mefford and Trask, 2002). In plants, this plasticity of subtelomeres was not found in Arabidopsis thaliana (Heacock et al., 2004;Kuo et al., 2006) and, to our knowledge, has not been reported for any other sequenced plant species.

QUALITY OF KHIPU SEQUENCES WITHIN THE G19833 GENOME
Genomic regions containing highly repetitive sequences, especially tandem arrays of satellite repeats, constitute a challenge for short-read, whole-genome shotgun sequencing and are thus considered to be error-prone regions of whole-genome sequencing projects (Jackson et al., 2011). In order to check the quality and the consistency of the G19833 pseudomolecules for khipu, we compared pseudomolecules sequence data with six G19833 BACs (sequenced by the classical Sanger method) containing khipu sequences and coming from two distinct genomic regions (Innes et al., 2008;Chen et al., 2010, unpublished data). Here we present in details the results from two BAC contigs: PvA11D is a 200 kbp clone corresponding to the coordinates 47.3-47.5 Mb of the long arm of pseudomolecule 11 (Chr11L) (Figure 2) while PvA05A is a 239 kbp clone corresponding to the coordinates 1.0-1.2 Mb of the pseudomolecule 5 (Chr05S) (Figure A1A in Appendix). In these regions, apart from khipu blocks, sequences share over 98% nucleic identity between BAC contigs and pseudomolecule sequences (data not shown). A combination of phylogenetic and genomic analyses shows that most khipu units are identical in position and sequence in both the pseudomolecules and BACs (blue areas in Figure 2; Figure A1A in Appendix). For example, khipus from PvA11D and its pseudomolecule counterpart are nearly identical (Figure 2). Similar results were obtained with four other BAC contigs containing khipu units (data not shown), confirming that genome data is of high quality even for khipu containing regions. However, PvA11D contains khipu blocks of only 14 khipu units (∼7400 bp) and a different situation is observed for PvA05A which bears a larger khipu block of more than 35 khipu units (Figure A1 in Appendix). Of these, six (PvA05Ak00020-PvA05Ak00070) are completely identical with corresponding region of Chr05 pseudomolecule, but 21 (PvA05Ak00080, PvA05Ak00120, PvA05Ak00300, PvA05Ak00310, PvA05Ak00350, PvA05Ak00400, PvA05Ak00410, PvA05Ak00430, and PvA05Ak00160-PvA05Ak00280) are completely absent in the whole-genome sequence, resulting in a ∼10 kbp gap (Figure A1A in Appendix). Moreover, three khipu units (PvA05Ak00090, PvA05Ak00100, and PvA05Ak00110) from this Chr05 BAC contig share 100% nucleotide identity with three khipu units from pseudomolecule 8 centromeric region (Pv08Ck00100, Pv08Ck00090, and Pv08Ck00080) (Figures A1A,B in Appendix). Because we were unable to find corresponding khipu units from Chr05S, it is possible that these three Chr08 centromeric khipu are the result of errors in the assembly. Even Khipu with name written in small gray correspond to khipu shorter than 500 bp discarded from the analyses. The blue areas represent correspondences between khipu from BAC sequence and khipu from genome sequence according to the phylogenetic tree.
though BACs are considered to be stable cloning vectors, it has been reported that tandem repeats can be unstable in BAC clones (Song et al., 2001). In conclusion, comparisons between BACs and the genome sequence data suggest that small arrays of satellite units can be well-resolved, while larger satellite blocks may www.frontiersin.org be slightly more error prone. However, based on six independent BACs/genome comparisons, we conclude that genome data is of high quality to conduct genome-wide analysis o khipu sequences.

CENTROMERIC KHIPU UNITS: DIVERGENT KHIPU SEQUENCE AND/OR ERRONEOUS ASSEMBLY?
Except for the three potentially false centromeric khipu units (identical to subtelomeric khipu units from Chr05S, Figures A1A,B in Appendix), all 78 centromeric khipu units belong to a single well-defined subclade that also includes subtelomeric units from Chr06L, Chr03L, and Chr02L (Figure 1; Figure A2 in Appendix) which indicates that they are less diverse than the subtelomeric units. It is surprising that no centromeric signals were found in previous FISH experiments on mitotic chromosomes, using either a khipu-specific probe in the BAT93 common bean genome Geffroy et al., 2009) or a khipu-bearing subtelomeric BAC clone in G19833 (Altrock et al., 2011). These conflicting results raise the question of whether the centromeric khipu units are real centromeric sequences or misassembled sequences.
To try to solve this puzzle, we performed FISH on pachytene chromosomes from G19833, using a khipu FISH probe containing a wide diversity of khipu units (highlighted by "FISH" written in red in Figure 1; Table A1 in Appendix). We detected khipu signals on 17 chromosome ends, but found no evidence of centromeric signals in G19833 chromosomes (Figure A3 in Appendix). This result is similar to khipu distribution at 17 chromosome ends in the BAT93 genotype  suggesting that khipu distribution has been stable since the split between Andean and Mesoamerican gene pools. Interestingly, during these FISH experiments heterochromatin connections between non-homologous chromosomes were observed ( Figure A4 in Appendix), providing indirect evidence that subtelomeres are hot spot of interchromosomal recombination. Similar attachments were observed in rye (Gonzalez-Garcia et al., 2006). In common bean, according to whole-genome data Chr08C comprises 55 khipu (Table 1). This begs the question: why was not it possible to detect these 55 centromeric khipu units by FISH?
There are four potential explanations. The first hypothesis is that khipu units were grouped at Chr08C due to biases during assembly of the pseudomolecule. A comparison to BAC contig PvA05A allowed us to find three khipu units from Chr05S subtelomeric region that were assembled at Chr08C by mistake during pseudomolecules assembling (Figures A1A,B in Appendix). What about the other centromeric khipu? As previously discussed, the genome assembly is of high quality and the major differences to BAC sequences are numbers of khipu units within a khipu block rather than a wrong location of khipu blocks (Figure 2; Figure A1 in Appendix). In addition to Chr08C, we also found khipu units in centromeres of Chr03, Chr04, Chr07C, Chr09, and Chr10 (Figure A2 in Appendix). It is unlikely that these khipu units, clustered in a single subclade, would be misassembled at the centromere of various pseudomolecules. Moreover, it is not unusual to find non-centromeric defined repeats in centromeric regions in eukaryotes. Notably, detection of DNA sequences that are partially homologous to telomeric repeats in centromeres has been reported in many species including Drosophila melanogaster (Abad et al., 2004;Mendez-Lago et al., 2009), maize (Alfenito and Birchler, 1993;Jin et al., 2005), and in potato (Tek and Jiang, 2004;He et al., 2013). Moreover, in rice, a satellite DNA present both in subtelomeric and centromeric regions has also been identified (Lee et al., 2005;Bao et al., 2006). A second hypothesis to explain the absence of khipu centromeric FISH signal is that khipu units from Chr08C are spread across a wide region of the centromere (13.8 Mb), thus diluting the signal in FISH. However, according to the pseudomolecule, the khipu units are grouped at the very center of Chr08C, with one block of 27 khipu units which should have been detected by FISH (Figure A5 in Appendix). The third hypothesis is that centromeric khipu are too divergent from the khipu units used for FISH experiments (Figure A6 in Appendix). Even if our FISH probe presents a wide diversity of khipu units (as shown in Figure 1 with "FISH" written in red), the closest khipu unit used in FISH (red arrow in Figure A6 in Appendix) shares only 72% nucleotide identity with the khipu units from the subclade comprising centromeric repeats, while it shares more than 85% identity with the khipu units from other major clades (data not shown). Consequently, it is possible that our probe missed the centromeric khipu units considering the stringency used. The fourth hypothesis is that the chromatin architecture in the centromeric region may make these sequences inaccessible to FISH.

CONCLUSION
Satellite DNA is an enigmatic part of eukaryotic genomes. Subtelomeric satellite repeats have been reported in many eukaryotic chromosomes but their function remains largely unknown. A genome-wide analysis based on the complete genome sequence of the common bean genotype G19833 of the subtelomeric khipu satellite, revealed extensive sequence exchanges between non-homologous chromosomes in subtelomeric regions and also suggests sequence exchange between subtelomere and centromere.

ACKNOWLEDGMENTS
We thank Mireille Sévignac for stimulating discussions. The research was supported by INRA, CNRS, IFR87, and IDEEV, France. Andrea Pedrosa-Harand was supported by a grant from CNPq, Brazil.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found   in red for khipu from PvA05A sequence, bold red and underlined for khipu from the pseudomolecule 5 and bold black for centromeric khipu of pseudomolecule 8. Khipu in small gray type correspond to khipu shorter than 500 bp that were discarded from analysis. The blue area represents the correspondence between khipu from BAC sequence and khipu from genome sequence according to the phylogenetic tree. (B) and (C) Phylogenetic tree clades of khipu present on (A). (B) Part of the clade called "E" (cf. Figure 1) (C) Part of the clade called "D" (cf. Figure 1) www.frontiersin.org www.frontiersin.org