Crescentia cujete L. (Bignoniaceae) is a diploid species (2n = 40) that produces non-edible fruits that have been of great importance to many indigenous and traditional communities of tropical America since pre-Columbian times, especially as drinking cups and storage vessels. Its wild geographic distribution is unknown, but it is found in many areas in the Neotropics in close contact with wild relatives in quite different environments.

There are two hypotheses of its origin of domestication. Gentry (1980) hypothesized an origin in Mesoamerica, where wild populations are found in seasonally flooded savannas. This hypothesis was not confirmed with chloroplast microsatellites in the eastern Yucatán of Mexico (Aguirre-Dugua et al., 2012). Ducke (1946) hypothesized that C. amazonica Ducke (described in 1937) gave rise to the cultivated C. cujete. This Amazonian species is also found in the Orinoco Basin and elsewhere in northern South America (Gentry, 1980; Wittmann et al., 2006; Díaz, 2009), where it is common in floodplain forests. The distributions of the other four accepted species of Crescentia L. are restricted to Central America and the Antilles, leading Gentry (1980) to comment on C. amazonica's distribution outside of the distribution of the other species. Contrary to Ducke (1946), Gentry (1980) suggested that C. amazonica was derived from cultivated C. cujete “when human selection for large fruits is relaxed.” However, using amplified fragment length polymorphism markers and a single accession of C. amazonica from the Orinoco Basin, Arango-Ulloa et al. (2009) found no relationship with C. cujete from Colombia.

Identification of the origin of domestication of treegourd and its routes of dispersal in the Neotropics remains unclear, and requires a molecular genetic analysis of a broader geographic sample. Using C. amazonica and C. cujete collections widely distributed along major rivers of the Brazilian Amazon Basin and the assembly of the chloroplast genome, we aim to identify single-nucleotide polymorphisms (SNPs) to compare chloroplast diversity between C. cujete and C. amazonica in order to evaluate the two hypotheses about the relationships between these species and better understand the domestication history of treegourd.

METHODS AND RESULTS

DNA was extracted from dried leaves of 36 samples of C. cujete and C. amazonica from the Brazilian Amazon Basin (Appendix 1). We used the cetyltrimethylammonium bromide (CTAB) 5% extraction protocol (Doyle and Doyle, 1990) with minor modifications; instead of cold isopropanol, NaCl was added to precipitate pellets. Barcoded libraries were constructed following the protocol of Mariac et al. (2014). Briefly, 0.5–1 µg of DNA were fragmented in a Bioruptor Pico sonicator (Diagenode, Liège, Belgium) using a standard protocol including six cycles and on/off conditions set to 30/90 s to reach a target size distribution of 300 bp. After sizing, end repair, ligation, and Bacillus stearothermophilus (Bst) DNA polymerase treatment, libraries were amplified with the KAPA HiFi HotStart Real-Time PCR Kit (KAPA Biosystems, Wilmington, Massachusetts, USA) with eight cycles to extend Illumina adapters and quantified by using the KAPA SYBR FAST LightCycler 480 qPCR Kit (KAPA Biosystems). Paired-end sequencing (2 × 150) was conducted on an Illumina MiSeq version 3 and HiSeq 2500 (Illumina, San Diego, California, USA) at the CIRAD facilities (Montpellier, France) and at Genotoul (Toulouse, France), respectively. Twelve picomoles of the bulked libraries with 1% PhiX were loaded in the flow cell. Mean passing filter among the different runs was 84.3%, producing 13 million clusters. The percentage of bases having a quality score above Q30 was 93.7%.

Fig. 1.

Circular map of the chloroplast genome of Crescentia cujete from Amazonas, Brazil (5.34°S, 60.44°W), deposited in GenBank (accession no. KT182634). Genes drawn within the circle are transcribed clockwise, while genes drawn outside are transcribed counterclockwise. Genes belonging to different functional groups are color coded. Dark bold lines indicate inverted repeats (IRA and IRB) that separate the genome into large (LSC) and small (SSC) single copy regions. Drawn using OrganellarGenomeDraw (Lohse et al., 2013).

Assembly was performed using the chloroplast of Tanaecium tetragonolobum (Jacq.) L. G. Lohmann (NC_027955) as a guide sequence for MITObim 1.7 (Hahn et al., 2013). First, MITObim mapped reads to the reference genome using MIRA version 4 (Chevreux et al., 1999) and an initial set of contigs was built ( Appendix S1 (apps.1600048_s1.pdf)). Then, a second mapping was done on these contigs. Contigs were extended if there was at least a 31-bp overlap with a given read. This process was iterated until a complete de novo genome was achieved. For the assembly, we used the two-step strategy pioneered by Li et al. (2013), because the repetitive nature of the inverted repeat (IR) regions (i.e., IRA and IRB) was difficult to assemble (Li et al., 2013). We first performed an assembly using the sequence large single copy (LSC), IRA, and small single copy (SSC), followed by a second independent assembly using the sequence SSC, IRB, and LSC from T. tetragonolobum (NC_027955). From the initial 3,106,862 shotgun reads, 268,499 reads were useful for the de novo chloroplast assembly. The SSC region showed a pairwise identity of 99.6% between the two assemblies, and the LSC region showed 99.7%. The slight differences observed are mainly locally close to repeat regions (mini- and microsatellite), and thus difficult to assemble. The IRs showed a 99.1% pairwise identity. The two fractions were manually aligned using the software Genious Pro 4.8.5 (Drummond et al., 2009), and a consensus C. cujete chloroplast sequence was built. The final assembly has a low number of N positions (46 Ns), and 96.7% of reads were properly paired, meaning that both read R1 and R2 were properly mapped. The mean depth of coverage of the sequence was 165×, meaning that for each position we have an average of 165 aligned reads. The final chloroplast genome size was 154,662 bp (Fig. 1, Appendix 1).

The C. cujete chloroplast genome was aligned with reference annotated genomes using the mauve algorithm implemented in Geneious Pro 4.8.5 (Drummond et al., 2009). For annotation, we used T. tetragonolobum (NC_027955; Bignoniaceae) as reference, and complemented it with Olea europaea L. (NC_013707; Oleaceae) and Capsicum chinense Jacq. (NC_KU041709; Solanaceae) to validate some tRNA orientations and add some introns lacking in the Tanaecium genome (rpl16, rps12). The correspondences of gene positions were identified and annotated manually. The C. cujete chloroplast sequence was deposited in GenBank (accession number KT182634). The assembled chloroplast genome of C. cujete was used to map another 30 C. cujete samples from home gardens and six C. amazonica samples from flooded forests with BWA 0.6.2 (Li and Durbin, 2009). Using SAMtools 0.1.7 (Li et al., 2009) and VarScan 2.3.7 (Koboldt et al., 2012), we generated and filtered the variant call format (VCF) files, following the Scarcelli et al. (2016) pipeline. The average number of chloroplast mapped reads was 54,747, equivalent to a 54× depth of coverage (Appendix 1). The minimum coverage was 15,324 reads, so even for this sample, each nucleotide was sequenced 15 times. We only have 0.7% of missing data in our 36 samples. Diversity analysis of the 30 C. cujete and six C. amazonica samples was done using DnaSP 5.10.1 (Librado and Rozas, 2009).

The size of the reconstructed chloroplast genome of C. cujete is 154,662 bp, structurally divided into four distinct regions: large single copy region (LSC: 84,788 bp), small single copy region (SSC: 18,299 bp), and a pair of inverted repeat regions (IR: 51,575 bp) (Table 1, Fig. 1). We identified 88 coding genes, of which nine were duplicated within IR regions, four rRNAs duplicated in IRA and IRB, 30 tRNAs, of which six were duplicated within IR regions. The C. cujete chloroplast genome size (bp) and GC content are comparable to T. tetragonolobum (Table 1), and within the variation observed in the order Lamiales, where genome lengths vary from 153,493 to 155,889 bp and GC content from 37.6% to 38.3% (Nazareno et al., 2015). The rps19 and rpl2 gene positions duplicated in the boundaries of IR (Fig. 1) agree with expectations from other angiosperms (Wang et al., 2008).

Table 1.

Comparison of chloroplast genomes between two species of Bignoniaceae.

We found 66 SNPs in 30 individuals of C. cujete with 24 haplotypes, and 68 SNPs in six individuals of C. amazonica with six haplotypes. Haplotype diversity (h) was 0.98 and 1.00, nucleotide diversity (π) was 1.1 × 10^-3 and 3.5 × 10^-3, and Watterson's estimator per site ( _W) was 2.3 × 10^-3 and 4.1 × 10^-3 for C. cujete and C. amazonica, respectively. Diversity was about twice as high in C. amazonica compared to C. cujete. If C. amazonica was simply derived from C. cujete, as suggested by Gentry (1980), diversity should be comparable or potentially even slightly lower. Consequently, we rule out the hypothesis that C. amazonica is derived from C. cujete. However, at this point we cannot rule out either that domestication of C. amazonica led to C. cujete or that C. cujete is derived from other wild species from Central America.

CONCLUSIONS

Next-generation sequencing provided data to have a sufficient number of reads to perform a de novo assembly of the C. cujete chloroplast genome, the first assembled chloroplast in the Crescentia genus. The reconstructed C. cujete genome allowed the identification of SNPs in C. amazonica and C. cujete that produced diversity estimates that refuted the hypothesis that C. amazonica is derived from C. cujete, and will be useful in further studies about the origin, diversity, and spread of treegourds in the Neotropics.