Massively parallel sequencing of Cannabis sativa chloroplast hotspots for forensic typing

Background Marijuana (Cannabis sativa) is the most commonly used illicit drug in the USA, and the use of DNA barcodes could assist drug trafficking investigations by indicating the biogeographical origin and crop type of a sample and providing a means for linking cases. Additionally, the legality of marijuana in the USA remains complicated with some states fully legalizing marijuana for recreational use while federally marijuana remains completely illegal. Massively parallel sequencing (MPS) offers distinct advantages over capillary electrophoresis (CE), including more comprehensive coverage of target loci, analysis of hundreds of markers simultaneously, and high throughput capabilities. Methods This study reports on the development of a MiSeq FGx® assay targeting seven “hotspot” regions in the Cannabis sativa chloroplast genome that are highly polymorphic and informative in attempts to determine biogeographical origin and distinguishing between marijuana and hemp. Sequencing results were compared to previous studies that used CE-based genotyping methods. Results A total of 49 polymorphisms were observed, 16 of which have not been previously reported. Additionally, sequence data revealed isoalleles at one locus, which were able to differentiate two samples that had the same haplotype using CE-based methods. This study reports preliminary results from sequencing 14 hemp and marijuana samples from different countries using the developed MPS assay. Conclusion Future studies should genotype a more comprehensive sample set from around the world to build a haplotype database, which could be used to provide investigative leads for law enforcement agencies investigating marijuana trafficking. Supplementary Information The online version contains supplementary material available at 10.1186/s42238-022-00123-2.


Background
Massively parallel sequencing (MPS), also called next generation sequencing (NGS), is a high throughput technique capable of collecting DNA sequence data from multiple targets and multiple samples in parallel. It offers several distinct advantages over traditional DNA typing using capillary electrophoresis (CE), including providing more comprehensive coverage of target markers (sequence data in addition to length) and the ability to analyze hundreds or thousands of targets at a time (Bruijns et al. 2018;Moorthie et al. 2011), compared to only about 25 loci by CE for a five-dye short tandem repeat (STR) kit (Lazaruk et al. 1998). Sequence data may elucidate isoalleles, alleles which have the same length and appear identical on CE but actually have different sequences, leading to more discriminatory results. Isoalleles may differ in their repeat structure or contain variants in the flanking regions, and the International Society of Forensic Genetics (ISFG) has reported guidelines to standardize the nomenclature for these sequence variable alleles (Parson et al. 2016). Costs and run times associated with MPS have dropped substantially, making targeted MPS assays a cost-effective approach for Page 2 of 10 Roman et al. Journal of Cannabis Research (2022) 4:13 characterizing samples for genetic individualization or identification (Moorthie et al. 2011). In forensics, MPS assays have been used for human identification purposes, including sequencing autosomal STRs and single nucleotide polymorphisms (SNPs) (Eduardoff et al. 2015;Guo et al. 2017;Kim et al. 2016;Seo et al. 2013;Wang et al. 2017), mitochondrial DNA analysis (Davis et al. 2015), phenotype prediction (Mehta et al. 2016), and other purposes (Bruijns et al. 2018;Budowle et al. 2017). While much of the forensic research on MPS has focused on human DNA, its use for forensic plant science has recently been investigated (Houston et al., 2018a, b). Houston et al. reported an MPS panel for the Ion S5 ™ consisting of twelve autosomal STRs in Cannabis sativa (marijuana). Results showed concordance with CE methods, and isoalleles were found at eight of the loci, providing a higher discriminatory power compared to CE (Houston et al., 2018a, b). MPS has been also been used for DNA barcoding studies in animals and has shown better recovery, reduced costs, and faster processing times compared to traditional Sanger sequencing (Shokralla et al. 2015). Studies involving the use of MPS for DNA barcoding in plants have been limited but do show the advantage of simultaneous analysis of multiple barcodes for enhanced phylogenetic resolution (Parks et al. 2009;Sucher et al. 2012). Recent studies reported chloroplast DNA barcoding markers in C. sativa that were informative for biogeographical origin and crop type prediction (Roman et al. 2019;Roman and Houston 2020). Since these regions represent the most highly polymorphic regions of the C. sativa chloroplast genome, they are referred to as "hotspots. " The polymorphisms were genotyped using CE-based methods, and Sanger sequencing revealed isoalleles at several loci with different repeat sequences or variations in the flanking regions that were not detected by CE (Roman et al. 2019;Roman and Houston 2020). This study seeks to expand upon previous studies by incorporating the "hotspot" barcoding regions into an MPS assay to provide more discriminatory results and a high throughput method for building a database of C. sativa chloroplast haplotypes.
Full chloroplast genome sequences have been reported for several marijuana and hemp cultivars (Matielo et al. 2020;Oh et al. 2016;Vergara et al. 2016). The full genome is 153,871 bp (Carmagnola and Dagestani cultivars) and contains 83 genes (Vergara et al. 2016). In comparison, the human mitochondrial genome is 16,569 bp (Anderson et al. 1981) (about a tenth of the size), and typically mitochondrial DNA analyses only involve sequencing a portion of the genome, usually the hypervariable regions (HV1 and HV2) (Ingman and Gyllensten 2006;Miller and Budowle 2001). Due to the large size of the C. sativa chloroplast genome, sequencing targeted regions (barcoding markers) gives better coverage and increased throughput capabilities compared to whole genome sequencing. The chloroplast genome is AT-rich (63%) and contains numerous homopolymeric stretches (Vergara et al. 2016). The MiSeq FGx ® platform was chosen for sequencing in this study because has been shown to have a higher fidelity when sequencing homopolymeric stretches of DNA (Duke et al. 2015;Loman et al. 2012;McElhoe et al. 2014), and many of the polymorphisms identified in the previous studies (Roman et al. 2019;Roman and Houston 2020;Cheng and Houston, 2021) were homopolymeric STRs (hSTRs).
This study seeks to design an MPS panel for the MiSeq FGx ® consisting of seven highly polymorphic "hotspot" regions in the C. sativa chloroplast (trnK-matK-trnK, rps16, trnS-trnG, ycf3, accD-psaI, clpP, and rpl32-trnL) to provide additional sequence data, discover isoalleles, and provide a high throughput method for creating a haplotype DNA database for hemp and marijuana samples. This assay could provide important investigative leads for law enforcement agencies investigating marijuana trafficking into and within the USA.

DNA extraction and quantification
Total genomic and organelle DNA was extracted using the DNeasy ® Plant Mini kit (QIAGEN, Hilden, Germany) (N=8 samples) or nexttec ™ 1-Step DNA Isolation Kit for Plants (nexttec, Hilgertshausen, Germany) (N=3 samples) according to the manufacturers' protocols. Extraction from USA-Mexico marijuana was performed on-site at CBP, and extraction from hemp was performed at Sam Houston State University. Chilean samples were provided by collaborators in the form of DNA extracts. Chloroplast DNA was quantified using a real-time PCR method reported by Houston et al. (Houston et al., 2018a, b).
Following amplification, samples were quantified using the Qubit ™ dsDNA HS Assay Kit ™ (Invitrogen, Carlsbad, CA) on a Qubit ™ 2.0 fluorometer (Invitrogen). The seven PCR targets were then diluted and pooled to a final concentration of 1 ng/μL. A 25 μL aliquot of the mixed PCR products was moved to a new tube and incubated with 2 μL exonuclease I (10 U/μL, Invitrogen) at 35 °C for 72 min and 75 °C for 15 min to remove excess primers.

Library preparation and sequencing
The Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA) was used for tagmentation and indexing of libraries. A 1 ng input of amplified, cleaned DNA was used (1 μL), and libraries were prepared according to the manufacturer's instructions with the exception that 15 cycles of PCR were used for indexing instead of 12 cycles to ensure adequate library quantity (Illumina 2018). Sample libraries were checked on the 2100 Bioanalyzer (Agilent, Santa Clara, CA) using the High Sensitivity DNA kit (Agilent).
Following manual normalization, all 14 libraries and the NTC were pooled in equal amounts and denatured according to the manufacturer's instructions, resulting in a 15 pM library (Illumina 2019). A denatured PhiX control (20 pM) was spiked in at 5% volume. Sequencing on the MiSeq FGx ® (Verogen, San Diego, CA) was performed using the MiSeq FGx ® Reagent Micro Kit (Verogen).

Data analysis
Data analysis was performed using the Miseq Control Software v1.4.0.0, MiSeq Reporter v2.5.1.3, and Real-Time Analysis v.1.18.54.0 software installed on the instrument. Variants were reported at a minimum coverage of 10 reads, Q30 variant score, and a minimum variant percentage of 20%. Sequences were compared to a C. sativa chloroplast reference genome (Yoruba Nigeria cultivar, GenBank accession NC_027223.1). Variants to the reference genome were reported in a variant call file, and .bam files were viewed in Integrative Genomics Viewer (IGV) 2.8.0 (Robinson et al. 2011).

Sequencing metrics
All samples (n=14) were successfully sequenced. Sequencing quality was high, with 81.2% of base calls having a quality score of 30 or higher, which indicates at least 99.9% accuracy at each base. The error rate in the PhiX control was below 3%. The yield was 1.48 Mb, cluster density was 1045 ± 26 K/mm 2 , and 99.33% ± 0.26% of clusters passed the filter. Phasing/prephasing rate was 0.139/0.036. Coverage varied within each of the amplicons. The trnK-matK-trnK and rps16 amplicons consistently had the lowest coverage, and ycf3 consistently had the highest. The clusters passing filter (PF) and clusters aligned to the reference genome for each sample are shown in Supplemental Table 1.
Due to the homopolymeric nature of many of the repeat units (hSTRs), forward and reverse stutter was observed. However, based on the sequence coverage, confident allele calls were made. Some reads aligned to portions of the genome outside of the hotspot regions and were not interpreted, and additionally, several reads misaligned within hotspot regions. These misalignments had low read depth and did not affect the interpretation of sequencing results.

Sequence data
There were 33 known polymorphisms within the "hotspot" amplicons (Roman et al. 2019 andCheng andHouston, 2021), and 16 more were discovered, bringing the total to 49 polymorphisms. Newly discovered Table 2 Genotypes for polymorphisms in the trnK-matK-trnK region (2233-4337 bp)     Tables 2,  3, 4, 5, 6, 7, and 8. Novel polymorphisms and new variants were reported to GenBank (Accession numbers: MW010378-MW010419). The trnS-trnG and rpl32-trnL hotspots had the most new polymorphisms with 12 and 11, respectively, and ycf3 had the least with only three. Analysis of a higher number of polymorphic loci is expected to show increased differences between samples from different populations. However, it is important to note that the genetic boundaries between hemp and drug-type marijuana have been extensively blurred due to the increased demand for high cannabidiol (CBD) strains, resulting in marijuana and hemp cross-breading. In addition, the relationship between cannabinoid synthase gene diversity and cannabinoid content is complex and not fully understood (Grassa et al. 2021). The genotypes at each locus for 14 samples are displayed in Tables 2, 3, 4, 5, 6, 7, and 8. Samples H8-1 (USA hemp) and 12-A7 (USA-Mexico marijuana) were the only two that produced the same haplotype. These two samples were analyzed previously with CE-based methods and also shown to have the same haplotype (Roman et al. 2019;Roman and Houston 2020). Samples 10-A1 (USA-Mexico marijuana) and MedMJ 10 (Chile medical marijuana) were also shown to have the same haplotype in previous studies. However, using MPS, they were distinguished by their sequences at rpl32-trnL hSTR3; 10-A1 has a 6-bp allele with the sequence TAA AAA , and MedMJ10 has a 6 bp allele with the sequence AAA AAA . Since they are the same size, these two isoalleles could not be distinguished in previous CE-based studies (Roman et al. 2019).

Concordance
Previously, 30 polymorphisms within the seven hotspot regions were analyzed in our laboratory using CE-based methods (Roman et al. 2019;Roman and Houston 2020;Cheng and Houston, 2021). The genotypes obtained by MPS are fully concordant with the CE-based genotypes with exceptions at the trnS-trnG hSTR2 and rpl32-trnL hSTR2 loci. In four samples (NT H5-4, 16-B1, 35, and 41), the sequence genotypes at both loci appeared to be off by 1 bp from the CE-based genotypes (indicated in Tables 4 and 8). The sequence data showed that the CE fragment assays for both of these loci amplified regions containing the locus of interest as well as an additional hSTR locus that was unknown at the time. Variation at this new hSTR locus explains the discrepancy between the CE and sequencing genotypes for all four samples at both loci. Sequence data for the two alleles observed at trnS-trnG hSTR2 were submitted to GenBank (accession numbers: MW010378-79), and these sequences include the newly observed hSTR.
Previously, rps16 hSTR and clpP hSTR3 alleles were reported by their bp size due to unclear results using Sanger sequencing (Roman and Houston 2020). However, as expected, the MPS method was able to elucidate the sequences of all alleles (GenBank accessions: rps16 hSTR (MW010380-83) and clpP hSTR3 (MW010383-86), and the fragment size and sequence data were determined to be concordant. Additionally, the 11 and 15 alleles at the rpl32-trnL hSTR3 locus were unable to be confirmed by Sanger sequencing in the previous study (Roman et al. 2019), but MPS was able to provide sequence data for these alleles (GenBank accessions: MW010387-88).

Conclusions
The MPS assay developed in this study provided an effective method for genotyping seven chloroplast regions previously shown to be informative for intra-species variation of C. sativa. It provided multiple benefits over previous CE-based assays, including simultaneous analysis of all seven regions in multiple samples, higher confidence for haplotype calls, and better discrimination through sequencing more polymorphisms and identifying isoalleles. A preliminary set of 14 samples was sequenced, and a total of 49 polymorphisms were observed, 16 of which have not been previously published. The sequence data