Whole genome sequencing is more available and less expensive than ever before, yet most scientists continue to rely on targeted, comparative sequencing for phylogenetics and phylogeography. Identifying the most appropriate markers to employ has been challenging. Information for model organisms abounds (e.g., grasses; Saski et al., 2007; Bortiri et al., 2008; Leseberg and Duvall, 2009), and a few studies have specifically screened the same set of markers across a diversity of plant groups, ranking the utility of these markers either explicitly or implicitly (Shaw et al., 2005, 2007, 2014). These studies are exceedingly valuable, demonstrating there is no one-size-fits-all answer to the question “which markers?”. The second critical question to “which markers” is “which primers?”. Hundreds of primer sequences have been published, many designed for specific taxonomic groups. The work presented here was inspired by “The Tortoise and the Hare II” (Shaw et al., 2005), which was the first study to pull together information on a large number of regions commonly in use (at that time) for plant phylogenetics. Our laboratory was also compiling such information, as were many others.
The Tortoise and the Hare II paper was revolutionary in assessing sequence variability for all regions studied across a broad diversity of flowering plants, and providing a ranking of that variability. In the mid-2000s, a small number of complete chloroplast genome sequences were available for land plants and some of those were not annotated (e.g., Medicago truncatula Gaertn. [GenBank NC_003119]; Saski et al., 2005). Grivet et al. (2001) were visionary when they moved beyond analyzing regions commonly being used to design primers for lesser-known and potentially faster-evolving regions of the chloroplast genome. They were the first to take advantage of the new genomic data boom, providing a set of 20 universal chloroplast primers designed around the complete chloroplast data from seven flowering plant species. Around the same time, I developed nondegenerate primers for 36 noncoding regions in the large and small single-copy regions of the chloroplast genome (published here). These near-universal primers were designed based on the complete chloroplast genome sequences of 16 flowering plant species (see Appendix 1).
Grivet et al. (2001) and I designed primers, but Shaw et al. (2007) took an even more applied approach when they examined sequences for three different taxon pairs (Atropa/Nicotiana, Lotus/Medicago, and Saccharum/Oryza), specifically searching for faster-evolving regions. Shaw et al. (2014) go one step further, comparing complete chloroplast genome sequences for 25 (primarily congeneric) sister species pairs. They examined sequence diversity for 107 single-copy noncoding regions, providing the most comprehensive analysis to date.
There are now at least 150 primer pairs available to amplify almost every intergenic, intron, and exon region of the chloroplast genome, including portions of the inverted repeats, thanks to the efforts of Shaw et al. (2005, 2007, 2014) and others (Ebert and Peakall, 2009; Scarcelli et al., 2011; Dong et al., 2012, 2013). Not surprisingly, although all worked independently, many of the same regions were explored (Appendix 2) and, in some cases, identical or nearly identical primers were designed. The push to identify faster-evolving regions was, in part, spurred by groups of organisms with exceptionally slowly evolving chloroplast genomes such as Bromeliaceae (Gaut et al., 1992) and Arecaceae (Asmussen and Chase, 2001). Heinze provided access to a comprehensive database of chloroplast primers in 2007 (Heinze, 2007). The database is periodically updated (last update 18 March 2014) and is available at http://bfw.ac.at/200/2043.html.
In the absence of taxon-specific complete chloroplast genome data, it is possible to mine the wealth of genomic data available in international databases such as GenBank (National Center for Biotechnology Information), EMBL-Bank (European Molecular Biology Laboratory), and DDBJ (DNA Data Bank of Japan). Primer pairs for 130 regions of the chloroplast genome were evaluated relative to representatives of 12 genera, spanning the diversity of flowering plants. Exon regions were avoided because they generally evolve more slowly than intron and intergenic spacer regions. The primers of Shaw et al. (2005, 2007), Scarcelli et al. (2011), and Dong et al. (2012), as well as the primers provided here, were evaluated. Many of the Shaw et al. (2005, 2007) and Scarcelli et al. (2011) primers are degenerate, improving the breadth of taxa they can be used on, but reducing their efficiency during the amplification process. The Dong et al. (2012) primers are primarily used for barcoding, thus amplify a diversity of taxa, but may not target the most quickly evolving regions of the genome. The likelihood of amplification success was estimated based upon the number and position of mismatches between the primer and the target sequence. These data were then evaluated in the context of Shaw et al. (2014) to provide generalizations, by taxonomic group, for primer utility in conjunction with sequence variability.
Finally, a small number of plant species have sequences available for multiple accessions or different subspecific taxa including Fragaria vesca L. (Rosaceae, N = 2), Gossypium herbaceum L. (Malvaceae, N = 2), Olea europaea L. (Oleaceae, N = 4), and Oryza sativa L. (Poaceae, N = 3). Shaw et al. (2014) specifically excluded species pairs with very low and very high levels of sequence divergence. Very high levels of divergence made alignment difficult, and very low levels provide too few characters for reasonable comparison across all flowering plants. Here I compare the variation at the subspecific level to that of higher-level relationships to determine if the same regions are useful at multiple taxonomic levels.
METHODS
Primers designed here—Sixteen chloroplast genomes, representing a diversity of flowering plants, were downloaded from GenBank (see Appendix 1). Homologous gene sequences were aligned in Se-Al version 2.0a11 (Rambaut, 1996). Primers were designed based on simultaneous viewing of the Se-Al file and an Oligo 4.02 (Rychlik, 2002) file, using a single sequence from the pool. Primers were anchored in coding regions and were designed to have a minimum number of hair-pins and primer-primer interactions, annealing temperatures between 50°C and 64°C, and a 3′ GC clamp if possible, targeting regions 400–1800 bp in length. Primer details are provided in Table 1, and are provided in the order of appearance in the tobacco genome (Nicotiana tabacum L. [GenBank Z00044.1]). The tobacco genome was the genome of choice for describing the location of primers prior to the recent accumulation of genomic data. A total of three different trnS primers were designed, corresponding to the three trnS genes encoded by the chloroplast genome (trnS-GCU, trnS-UGA, and trnS-GGA). Gene order is highly conserved on the chloroplast genome of flowering plants, but does vary and can be highly informative, for example, as in the 22-kb inversion in almost all Asteraceae (Jansen and Palmer, 1987a, 1987b) and the 78-kb inversion in Fabaceae subtribe Phaseolinae (Bruneau et al., 1990). Some primer combinations are not useful in particular groups of plants due to structural rearrangements. In some cases, the downloaded genomes differ in the identification of specific genes.
Primer utility—The chloroplast genomes for species of eight genera (Acorus L., Amborella Baill., Canna L., Ceratophyllum L., Cymbidium Sw., Helianthus L., Magnolia L., and Nelumbo Adans.) and for subspecies of F. vesca, G. herbaceum, O. europaea, and O. sativa were compared to 130 primer pairs published by Shaw et al. (2005, 2007), Scarcelli et al. (2011), Dong et al. (2012), and those designed here. Complete chloroplast genome sequences were downloaded from GenBank (accession numbers, taxonomic identity, and original publication information provided in Appendix 3) and aligned manually in Sequencher (Gene Codes Corporation, Ann Arbor, Michigan, USA). A separate file containing the primer sequences was imported and automatically assembled using the settings “dirty data” and 100% sequence similarity with a minimum overlap of 16 bp. Additional rounds of alignment were conducted with successively lower levels of sequence similarity. Primers that failed to align automatically, or that aligned incorrectly, were realigned manually whenever possible (guided by the GenBank annotations). Alignment of the two Gossypium sequences required inversion of a large region of one taxon (arbitrarily selected as G. herbaceum subsp. africanum (G. Watt) Vollesen) approximately corresponding to bases 115,132–135,355 in the final alignment. The Oryza alignment includes O. nivara Sharma & Shastry because it is a potential progenitor of O. sativa (Li et al., 2006; but see Huang et al., 2012 for an alternative view point).
As mentioned above, degenerate primers provide broader utility, but reduced amplification efficiency. If a mismatch was detected in the last five bases at the 3′ end of the primer, the mismatch was inferred to be fatal (IDT, 2009). If more than three mismatches were detected within any given primer, amplification was inferred to be unsuccessful. These criteria are arbitrary but have worked for me personally and are probably more strict than necessary.
Sequence variability within species—The sequences of F. vesca, G. herbaceum, O. europaea, and O. sativa were examined manually to assess the variation of the 130 regions. Length of the inferred amplicon was noted along with the number of mismatched bases (aka inferred substitutions; excluding primer regions), the number of insertion/deletion (indel) events, and the number of inversions. These data provided an estimate of the utility of the regions for inferring phylogeny among closely related subspecies, and potential for application to phylogeographic studies. Shaw et al. (2014) specifically avoided these types of comparisons due to the very small number of parsimony informative characters. Sequence diversity was estimated using three criteria calculated as: (1) [(number of substitutions*2)+(number of indels)+(number of inversions)]/ amplicon length, (2) number of substitutions+indels+inversions, and (3) sequence diversity (number of substitutions/sequence length). The first criterion (criterion 1) is a weighted rank, and includes information on the number of inferred substitutions (weighted twice as heavily as the other two components), indels, and inversions. Substitutions were weighted more heavily because chloroplast indels may be more homoplasious (Kelchner and Clark, 1997), especially among closely related taxa. Inversions are often low in homoplasy (Graham et al., 2000) and thus could be weighted more heavily, but are relatively rare so weighting was not employed. The 10 most variable regions for each species were identified, as measured under each criterion. Frequency of any specific “top 10” primer pair was summed across the four species.
RESULTS
Primers designed here—The 72 primers targeted noncoding regions of the chloroplast genome with amplicon sizes of 500– 1800 bp. Degenerate primers were avoided because they were assumed to decrease priming efficiency, as were mismatches within the last five bases at the 3′ end of the primer. Only two primers required degenerate bases: one primer with two degenerate bases and another primer with one degenerate base. None of these degeneracies were located within the last five bases. In contrast, 17 of the Scarcelli et al. (2011) primers have at least one degenerate base in the last five bases at the 3′ end of the primer, and so are assumed to fail for at least some taxa.
Table 1.
Region, primer name, primer sequence, amplicon position, and amplicon length for plastid noncoding regions relative to the Nicotiana tabacum L. (GenBank Z00044.1) genome.
Continued.
Primer evaluation—Three of the four sets of primers examined here were equally likely to amplify target chloroplast regions (81–85% should work; see Table 2). The Dong et al. (2012) primers were least likely to work based on the 12 species examined here (65% on average) and were particularly poorly matched to the Oryza genome (29% amplification success predicted), and only moderately suited for Amborella (52%), Cymbidium (52%), and Helianthus (57%). However, the Dong et al. (2012) primer pair trnH-psbA was not expected to work on any of the target species, possibly due, in part, to an extra “A” near the 3′ end of the published sequence for the trnH primer. The primers designed here were poorly matched to three of the four monocots (Cymbidium, Oryza, and Canna; 61%, 64%, and 67%, respectively), despite being a good match for Acorus (81%). Scarcelli et al. (2011) primers were designed with monocots in mind and did an exceptional job matching the monocot genomes examined here, with amplification success ranging from 82–97%. They were almost equally good for the dicots examined here, with amplification success of 72–93%. The Shaw et al. (2005, 2007) primers were useful across the angiosperm phylogeny, with all anticipated amplification success percentages above 78%.
On average, the Shaw et al. (2005, 2007) and Scarcelli et al. (2011) primers are more degenerate, yet they were only slightly more likely to amplify the target sequences than the nondegenerate primers designed here, at least for nonmonocot taxa. With so many different primers available, most regions could be amplified in almost all target taxa provided an appropriate primer pair was selected. Indeed, many primer pairs should work in all 12 species examined here. Details of the inferred priming success are provided in Appendix S1 (apps.1400085_s1.xlsx), and species-specific notes on primer/sequence mismatches are provided in Appendix S2 (apps.1400085_s2.xlsx).
Primer utility × sequence variability—Shaw et al. (2014) conveniently summarized sequence variability across the chloroplast genome including the identification of the 13 fastest-evolving regions for six taxonomic groups (magnoliids, monocots, eurosids I, eurosids II, euasterids I, and euasterids II). Summing across these major groups, 28 different regions were identified as the most variable. Primers to amplify those 28 regions are detailed in Table 3, along with the Shaw et al. (2014) rank for each region (in bold typeface above each primer region), for each taxon examined here. Multiple primer pairs are available for each of the 28 regions except the trnT-trnL (Shaw et al., 2005 only), ycf4-ycf10 (or cemA; current study only), and ndhD-psaC (none of the publications examined). The ndhD-psaC region was ranked 10th fastest for eurosids I, but as there are no primers to be evaluated this region will not be discussed further. Primers are available for each of the remaining 27 regions.
Among the basal dicot grade (Amborella and Magnolia), successful primers are available for all 27 regions. Primer selection is more challenging for Amborella than for Magnolia. The top ranked region was the rpl32-trnL intergenic spacer (IGS). Shaw et al. (2007) primers will work for both taxa; Dong et al. (2012) primers will not. In contrast, rps16-trnQ, the second highest ranked region, has three sets of primers available (Shaw et al., 2007; Scarcelli et al., 2011; and Dong et al., 2012), all of which should work.
Among the monocots sampled (Acorus, Cymbidium, Oryza, and Canna), Acorus was the least difficult sequence to match and Oryza the most difficult. Structural rearrangements are the primary reason for failure to amplify across all available primers (e.g., rbcL-accD in Oryza and petA-psbJ in Cymbidium). One region cannot be amplified in Acorus—the accD-psaIIGS, despite the availability of four different primer pairs. In all, four regions cannot be amplified in Cymbidium with the primers studied here: petN-psbM, psbM-trnD, atpB-rbcL, and petA-psbJ. The ndhA region can be amplified in only some species of Cymbidium due to fatal substitutions in some species for all three primer pairs evaluated here. In Oryza, the trnS[GCU]-trnG[GCC], trnT-psbD, rbcL-accD, accD-psaI, and rps15-ycf1 cannot be amplified using any primer pair. In Canna, ndhF-rpl32 will not amplify with either of the available primer pairs. Unfortunately, according to Shaw et al. (2014), ndhF-rpl32 is the most variable and psbM-trnD is the third most variable region for monocots.
Basal eudicots were not evaluated by Shaw et al. (2014) in detail, so direct comparisons cannot be made here. Fortunately, at least one primer pair was successful for each of the 27 fastest-evolving regions, with the exception of the ycf4-ycf10 region. The only available primers for this region were designed here, and they will not work for Ceratophyllum. In general, Ceratophyllum was more difficult to match than was Nelumbo.
Shaw et al. (2014) detailed variability of higher eudicots for four major groups: eurosids I, eurosids II, euasterids I, and euasterids II. Only a single species representing each group was included here. Fragaria (eurosids I) could not be amplified for a single region, the ycf4-ycf10 IGS. According to Shaw et al. (2014), the fastest region for this clade was the ndhA intron. Both the Shaw et al. (2007) and Scarcelli et al. (2011) primers should work, but the Dong et al. (2012) primers will not. The second fastest region was the trnS[GCU]-trnG[GCC], which should amplify with any of the primer pairs (Shaw et al., 2005; Scarcelli et al., 2011; or Dong et al., 2012).
The sole representative of eurosids II and euasterids I (Gossypium and Olea, respectively) could successfully be amplified by at least one pair of primers studied here. The fastest region for eurosids II was the ndhF-rpl32 IGS. The Shaw et al. (2007) primer pair should work, but the Scarcelli et al. (2011) primer pair likely will not. The second most variable region was the psbZ-trnG IGS. For this region, both the Scarcelli et al. (2011) and Dong et al. (2012) primers should work, but the Shaw et al. (2005; as trnfM-trnS) primers will not. In euasterids I, the fastest region was the rps16-trnQ IGS. For Olea, the Shaw et al. (2007) and Scarcelli et al. (2011) primers should work, but not so the Dong et al. (2012) primers. The next-fastest region was the rpl32-trnL IGS. Both the Shaw et al. (2007) and Dong et al. (2012) primers should work.
Table 2.
Summary of amplification success probability for 130 pairs of chloroplast primers.
Primer failure in Helianthus (euasterids II) was primarily due to structural rearrangements (e.g., trnS[GCU]-trnG[GCC], rpoB-trnC, trnE-trnT, rbcL-accD). rpl32-trnL IGS was the fastest region according to Shaw et al. (2014), and either the Shaw et al. (2007) or Dong et al. (2012) primers should successfully amplify this region. The adjacent ndhF-rpl32 IGS was the second most variable region. Both the Shaw et al. (2007) or the Scarcelli et al. (2011) primers should work.
Subspecific sequence variability—Intraspecific sequence variation was evaluated in four species: F. vesca, G. herbaceum, O. europaea, and O. sativa. This represents a tiny fraction of angiosperm diversity, but is the first analysis of subspecific diversity across the entire chloroplast genome for multiple species, in the context of available primer resources. Appendix S3 (apps.1400085_s3.xlsx) identifies the fastest-evolving regions among the four species, under three different criteria. On average, only five inversions per chloroplast genome were detected here and the distribution across species was very different. Gossypium and Oryza each had 10 inversions, Fragaria none, and Olea only one. Details of subspecific comparisons for all regions are provided in Appendix S2 (apps.1400085_s2.xlsx).
No single genic region was identified as the top 10 fastest for all four species. Pooling data across all three criteria, the most frequently identified genic region was the psbZ-trnfM IGS with eight occurrences out of a maximum of 12 possible, followed by the trnS(GCG)-trnG(GCC) IGS, with six occurrences, rps16-trnQ IGS and trnT(GGU)-psbD IGS each with five, and rps12-psbB IGS and rps4-trnT(UGU) IGS each with four occurrences. Data for individual species have limited general application, but are provided below.
Oryza sativa, the only monocot in this comparison, showed highest variation, based on rank, for clpP-psbB (0.0195, 924 bp), atpB-rbcL (0.0168, 1070 bp), and psbM-trnD(GUC) (0.0150, 523 bp). Two of the same regions were identified as fastest under criterion 2, atpB-rbcL (12 characters, 1070 bp) and clpP-psbB (11 characters, 924 bp), plus rbcL-accD (13 characters, 1824 bp). Sequence divergence was highest in and around the clpP region including what would be the clpP intron 2 (1.9455%, 257 bp), clpP intron 1 (1.0050%, 199 bp), and clpP-psbB (0.7576%, 924 bp). In contrast, the three fastest regions per Shaw et al. (2014) for monocots were ndhF-rpl32 (rank 1), ndhC-trnV (rank 2), and psbM-trnD (rank 3).
The highest variation for Fragaria under criterion 1 was for trnW(CCA)-psaJ (0.0101, 789 bp), trnT(GGU)-psbD (0.0098, 1527 bp), and trnP(UGG)-rps18 (0.0090, 1563 bp). Under criterion 2: trnT(GGU)-psbD (eight characters; 1527 bp), trnP(UGG)-rps18 (eight characters, 1563 bp), and petN-trnD (seven characters, 2504 bp). Under criterion 3, the top three regions were trnT(GGU)-psbD (0.4584%, 1527 bp), psbB-psbH (0.4451%, 674 bp), and rps4-trnT(UGU) (0.4435%, 451 bp). Shaw et al. (2014) eurosids I top three regions were ndhA intron (rank 1), trnS(GCU)-trnG(GCC) (rank 2), and rps16 intron (rank 3).
Table 3.
Amplification success prediction for the 28 fastest Shaw et al. (2014) regions.a
Continued.
In Gossypium, the most informative regions under criterion 1 were psbZ-trnfM(CAU) (0.0534, 1179 bp), trnH(GUG)-psbA (0.0444, 496 bp), and rps4-trnT(UGU) (0.0425, 635 bp). Criterion 2 fastest regions were trnS(UGA)-trnG(GCC) with 39 variable characters over 1673 bp, followed by psbZ-trnfM(CAU) with 37 characters for 1179 bp, and trnT(UGU)-trnL(UAA) with 33 characters over 1470 bp. Sequence divergence (criterion 3) was highest for psbZ-trnfM(CAU) (2.2053%, 1179 bp), then trnS(UGA)-trnG(GCC) (1.6736%, 1673 bp), and finally the rps16 intron (1.6181%, 927 bp). Eurosids II top three regions for Shaw et al. (2014) were ndhF-rpl32 (rank 1), psbZ-trnG (rank 2), and trnT-trnL (rank 3).
For Olea, the most informative regions under criterion 1 were psbC-psbZ (0.0411, 1045 bp), trnS(UGA)-trnfM (0.0333, 1203 bp), and clpP intron 2 (0.0313, 702 bp). The highest number of variable characters (criterion 2) were found in rps16-trnQ (29 characters, 2739 bp), psbC-psbZ (22 characters, 1045 bp), and trnS(UGA)-trnfM (21 characters, 1203 bp). Criterion 3 (percent sequence divergence) was highest in the same three regions as under criterion 1: psbC-psbZ (2.0096%, 1045 bp), trnS(UGA)-trnfM (1.5794%, 1203 bp), and clpP intron 2 (1.4245%, 702 bp). Shaw et al. (2014) euasterids I top three included rps16-trnQ (rank 1), rpl32-trnL (rank 2), and ndhC-trnV (rank 3).
DISCUSSION
A large number of “universal” primers have been published for amplification of various chloroplast regions. Some are more degenerate than others, presumably to be more widely applicable. Degeneracy is not required, however, and may not lead to greater success in the laboratory. On the other hand, nondegenerate primers with poor fit are likely to fail, and some primers published as “universal” are not necessarily so. The universal barcoding primers of Dong et al. (2012) were the least likely to be useful across the 12 taxa examined here, with an average success rate of 65%, and a very poor 29% success rate in Oryza. In contrast, the primers designed by Scarcelli et al. (2011) specifically for monocots were exceedingly well-matched to the monocots sampled (97% in Acorus, 93% in Cymbidium, 92% in Oryza, and 88% in Canna), and a good match across all angiosperms.
Unlike previous analyses, this study used published genomes and primer sequences to infer the likelihood of amplification success. Only a small number of published primers were evaluated, and additional primers will be added to future analyses. Indeed, as mentioned in the introduction, Ebert and Peakall (2009) and Dong et al. (2013) have primers that could be evaluated as well as those of Doorduin et al. (2011) designed for species of Asteraceae. The evaluation conducted here shows parallels to prior studies in that general conclusions or recommendations are difficult to distill. For each region, there may be a number of primer pair options. Which primer pair is best is highly variable and depends upon the taxon being investigated. Scarcelli et al. (2011) primers are the best option for monocots in general, but will fail in specific combinations (e.g., trnH-psbA for Canna, atpF intron/exon for Cymbidium, and trnD-trnT for Oryza). Dong et al. (2012) primers are generally less successful, but they are the only primers that will work for psbM-trnD in Amborella and Magnolia. In several instances, a primer will work for some, but not all species in a genus, like the Scarcelli et al. (2011) matK primers in Cymbidium or the trnK-rps16 primers in Helianthus. Table 3 provides a quick summary of primer match for the top regions according to Shaw et al. (2014).
Prior studies have done an excellent job assessing variability of various noncoding regions across a diversity of angiosperms, particularly the recent work of Shaw et al. (2014). Those studies focused on infrageneric or even intergeneric comparisons. Here I compare sequence variability within species to see if the same markers are identified as the most variable, under slightly different criteria. This comparison was specifically avoided by Shaw et al. (2014) due to the small number of variable characters. The fastest regions identified here for Oryza were (depending upon criterion) clpP-psbB, atpB-rbcL, psbM-trnD, and rbcL-accD. In contrast, Shaw identified ndhF-rpl32, ndhC-trnV, and psbM-trnD as the fastest regions for monocots, with only one region of overlap between the two. For Fragaria (eurosids I), the list has no overlap at all. Olea (eurosids II) and Gossypium (euasterids I) each only overlap for a single region between the two studies. The lack of consensus over which region is the most variable at lower taxonomic levels has been pointed out by a number of papers including Särkinen and George (2013) for Solanum, and for 19 species pairs as demonstrated by Shaw et al. (2014). The comparison made here only adds to the argument that there is an acute need for additional comparative information.
Shaw et al. (2014) provided a solid foundation for which markers evolve the most quickly in major angiosperm clades, yet the fastest regions identified here for subspecies comparisons share little overlap with Shaw's regions. This finding suggests the need for a thorough exploration of markers prior to undertaking a large comparative sequencing project. The methods employed here to examine expected primer utility can easily be applied to any taxon, provided complete chloroplast genomic data are available. When complete genome data are lacking, the results presented here can provide a rough estimate of the “best primers,” but this remains a work in progress.
LITERATURE CITED
Appendices
Notes
[1] This work was inspired by phylogenetic combs obtained for Guzmania (Bromeliaceae) and Monardella (Lamiaceae), and by the work of Joey Shaw. Some plant material or DNA samples were provided by The Desert Botanic Garden, S. Eliason, E. A. Friar, Y.-L. Qiu, S. Vanderplank, and G. Wallace. Cris Martinez and Bill Waggoner assisted with primer screening. Harith Alappat assisted with whole chloroplast genome and primer alignments. Rancho Santa Ana Botanic Garden provided financial support.