bioRxivGenomicsPreprint
Sequence and annotation of 42 cannabis genomes reveals extensive copy number variation in cannabinoid synthesis and pathogen resistance genes
Kevin Judd McKernaniD, Yvonne Helbert, Liam T. KaneiD, Heather M. Ebling, Lei ZhangiD, Biao LiuiD, Zachary Eaton, Stephen McLaughlin, Sarah B. KinganiD, Primo A. Baybayan, Gregory T. ConcepcioniD, Mark Jordan, Alberto RivaiD, W. Brad BarbazukiD, Timothy T. HarkinsiD
Abstract
Cannabis is a diverse and polymorphic species. To better understand cannabinoid synthesis inheritance and its impact on pathogen resistance, we shotgun sequenced and assembled a Cannabis trio (sibling pair and their offspring) utilizing long read single molecule sequencing. This resulted in the most contiguous Cannabis sativa assemblies to date. These reference assemblies were further annotated with full-length male and female mRNA sequencing (Iso-Seq) to help inform isoform complexity, gene model predictions and identification of the Y chromosome. To further annotate the genetic diversity in the species, 40 male, female, and monoecious cannabis and hemp varietals were evaluated for copy number variation (CNV) and RNA expression. This identified multiple CNVs governing cannabinoid expression and 82 genes associated with resistance to Golovinomyces chicoracearum, the causal agent of powdery mildew in cannabis. Results indicated that breeding for plants with low tetrahydrocannabinolic acid (THCA) concentrations may result in deletion of pathogen resistance genes. Low THCA cultivars also have a polymorphism every 51 bases while dispensary grade high THCA cannabis exhibited a variant every 73 bases. A refined genetic map of the variation in cannabis can guide more stable and directed breeding efforts for desired chemotypes and pathogen-resistant cultivars. O_FIG O_LINKSMALLFIG WIDTH=160 HEIGHT=200 SRC="FIGDIR/small/894428v1_ufig1.gif" ALT="Figure 1"> View larger version (126K): org.highwire.dtl.DTLVardef@7a1ac3org.highwire.dtl.DTLVardef@731a4dorg.highwire.dtl.DTLVardef@c3e5acorg.highwire.dtl.DTLVardef@11babb4_HPS_FORMAT_FIGEXP M_FIG Sequence and annotation of 42 cannabis genomes reveals extensive copy number variation in cannabinoid synthesis and pathogen resistance genes C_FIG
- organism: plant (other)
- organism: C. elegans
- assay: long-read sequencing
- assay: whole-genome sequencing
- assay: DNA methylation
- research focus: genome assembly
- research focus: variant calling
- research focus: database / resource
- approach: experimental assay
In-depth summary
Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (5 Jan 2020) · CC BY 4.0
Long-read assemblies of an inbred cannabis trio plus whole-genome sequencing of 40 cultivars link cannabinoid synthase gene deletions to chemotype and associate CsTLP1 loss with powdery mildew susceptibility.
Copy number variation in cannabinoid synthase genes had been studied with fragmented references or gene-targeted assays, and the genetics of powdery mildew resistance were undescribed. The paper reports what it calls the first assembly of a Type II plant with functional tetrahydrocannabinolic and cannabidiolic acid synthase genes (THCAS, CBDAS), and combines it with 52x whole-genome data from 40 cultivars to place chemotype (Bt:Bd) alleles and resistance-gene copy number on the genome.
Key results
ShownThree PacBio long-read assemblies of an inbred Jamaican Lion trio (mother, father, F1) reach contig N50s of 1.6 to 3.8 Mb with over 96% BUSCO (benchmarking universal single-copy orthologs) completeness, and are annotated with Iso-Seq full-length transcripts from five tissues.
FALCON-Unzip on 84.8 to 125 Gb of long reads per genome; BUSCO 96.1%, 97.0% and 97.3% complete (father, mother, F1). Purging the maternal assembly gave an 876 Mb haploid size and 3.2 Mb N50, but 12.6% BUSCO duplicates remain. 83,464 Iso-Seq isoforms; 27,664 female and 31,108 male genes in S6 vs 27,644 and 32,108 in S20. · Sequencing, assembly and annotation; Structural variation and gene models; DNA Assembly; Gene annotation
ShownNormalized gene coverage over THCAS, CBDAS and cannabichromenic acid synthase (CBCAS) sorts cultivars into Type I, II and III chemotypes; the most common Bt:Bd alleles are whole-gene deletions of THCAS or CBDAS.
39 cannabinoid synthase genes profiled for copy number variation (CNV) across 40 genomes at 52x average coverage, using unique molecular identifier (UMI) libraries (PCR vs PCR-free control coverage r = 0.999). Label HPLC chemotypes for fourteen Massachusetts samples all matched the deletion-based calls; deletion status cannot separate 0.3% from 0.5% THCA Type III plants. · Copy number variation across the genomes; Cannabinoid synthase genes; Chemotyping; Figure legends
SuggestedA roughly 2 Mb cassette of 8 co-oriented CBCAS genes on contig 756 is absent from many Type I plants and also carries pathogen-response genes. The authors propose that breeding hemp below 0.3% THCA could select for the deletion and so remove those genes.
The 8 CBCAS copies are over 99.4% identical; the deleted span holds NPF3, RMT1, PIP1 and NIP1. Most Type III plants carry the contig (one exception), Type II plants are usually heterozygous, and some Type I plants (Black 84, Arcata Trainwreck) carry it. THCA synthesis by CBCAS in planta is not tested here; cited support comes from yeast work. · Cannabinoid synthase genes; Conclusions; Figure legends
SuggestedDeletion of a thaumatin-like protein (TLP) gene, CsTLP1, is common in cultivars reported as powdery mildew susceptible. CsTLP1 expressed in E. coli shows β-1,3-glucanase activity and inhibits Fusarium oxysporum growth on plates.
Copy number of 23 TLPs, 35 chitinases and 24 mildew resistance locus O (MLO) genes compared with cultivator-reported resistance across 40 genomes; endochitinase CH25 and MLO absence also correlated with resistance. CsTLP1 alone did not inhibit Aspergillus flavus or Penicillium chrysogenum; with Trichoderma viride chitinase it inhibited P. chrysogenum and F. oxysporum. The mildew fungus itself was not tested. · Pathogen response genes; Anti-fungal activity; Figure legends
ShownContigs covered only in male genomes, intersected with male-specific Iso-Seq transcripts, place 574 genes in the non-recombining region of the Y chromosome. Whole-genome coverage also indicates an XX structure in monoecious hemp.
40 genomes (9 male, 2 monoecious, 29 female) aligned to the paternal assembly: Y contigs have zero coverage in females and X contigs have double coverage; 118 Mb of putative Y contigs. Monoecious C3 and Fedora show an XX pattern. Y genes include FT, FY and PIN2 homologs; their role in sex determination is proposed, not tested. · Introduction; Identification of the Y chromosome; Conclusions; Figure legends
Methods, data and reproducibility
An inbred Jamaican Lion trio was assembled from PacBio long reads (FALCON-Unzip, Arrow) and annotated with Iso-Seq and MAKER2. Forty cultivars were Illumina-sequenced with UMI adaptors and genotyped with DRAGEN; gene copy number came from normalized BEDtools coverage. CsTLP1 was expressed in E. coli for glucanase and antifungal assays, and EM-seq profiled methylation in four tissues.
- PacBio long-read genomes of the Jamaican Lion trio (mother, father, F1 offspring JL5) · Cannabis sativa · n = 3 genomes (84.8 to 125 Gb each) · NCBI PRJNA575581; SUB6635057 (project-level)
- Illumina NovaSeq whole-genome sequencing of hemp and drug-type cultivars, including the trio parents and 6 F1 offspring as controls · Cannabis sativa · n = 40 genomes (9 male, 2 monoecious, 29 female) · NCBI PRJNA575581 (project-level)
- Iso-Seq full-length mRNA from male flowers plus four female tissues (unseeded and seeded flowers, leaves, roots) · Cannabis sativa · n = 5 tissues; 83,464 isoforms
- EM-seq CpG methylation libraries · Cannabis sativa · n = 4 tissues, in duplicate
- Hi-C reads on Jamaican Lion (scaffolding results not shown) · Cannabis sativa · n = over 600 million reads
- Cannabis samples with label HPLC cannabinoid measurements, used to check genotype-based chemotypes · Cannabis sativa · n = Fourteen
- Fungal cultures for CsTLP1 antifungal plate assays · Fusarium oxysporum, Penicillium chrysogenum, Aspergillus flavus · ATCC #18476 (P. chrysogenum); ATCC #16870 (A. flavus)
- Methods & tools
- FALCON-Unzip v0.3.0, Arrow, Purge Haplotigs, BUSCO 3.0.2, MAKER2, Iso-Seq, DRAGEN, SNPeff, PBSV, BEDtools, EM-Seq (NEB), Megazyme malt β-glucanase assay
- Data & accessions
- https://genomevolution.org/https://kannapedia.net/
- Reagents & environment
- DRAGEN HostSoftware 05.011.281.3.2.5 (commands printed); DRAGEN methylation mapper 3.4.5; FALCON-Unzip v0.3.0 (full configuration printed); Quast 5.0; BUSCO.py 3.0.2, Augustus 3.3.21, Hmmer 3.2.1, Blast 2.7.1, eudicotyledons_odb10; IGV 2.4.14.
- Reproducibility gaps
- No analysis code is released; the only code link is to the third-party PBSV repository. The assembly-specific CoGe link is printed with apparent typos (genomevo1ution.org). Purge Haplotigs trials and Hi-C scaffolding results are 'data not shown'.
Limitations (6)
- The sample is small, and the related Jamaican Lion family may create synthetic associations, so cloning is needed to confirm CsTLP1. Mildew susceptibility is not binary and may appear only under stress, and several modes of resistance are possible. · stated by the authors
- Iso-Seq came from parental tissues without biological replicates and the libraries look unsaturated, so tissue expression comparisons are tentative. · stated by the authors
- Haplotig contamination is hard to tell apart from true copy number variation (12.6% BUSCO duplicates remain). The assembly is left unscaffolded because Hi-C gave few long-range links and disagreed with Oxford Nanopore scaffolding. · stated by the authors
- THCA production by CBCAS or other synthases in planta is unconfirmed. Short-read genotyping had an estimated 4.2% false-positive variant rate. · stated by the authors
- The abstract's figure of a polymorphism every 51 bases in low-THCA cultivars rests on one hemp sample (Merino vs Jamaican Lion). S5 gives 77-68 bases for dispensary cannabis, not the abstract's 73. · noted in this summary
- Mildew resistance labels are cultivator reports, and the CNV associations are shown as heatmaps and PCA with no statistical test. The antifungal plate assays report no replicate counts. · noted in this summary
Read this if you breed or genotype cannabis or hemp and need gene-level copy number maps of cannabinoid synthase, terpene synthase and candidate mildew resistance genes, or a male and female long-read Type II reference.
- Citations
- 71
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Gene copy number is associated with phytochemistry in Cannabis sativa
Daniela Vergara, Ezra L. Huscher, … Nolan C. Kane·AoB PlantsPeer-reviewed·Nov 1, 2019
- Domesticated cannabinoid synthases amid a wild mosaic cannabis pangenome
Ryan C. Lynch, Lillian K. Padgitt‐Cobb, … Todd P. Michael·NaturePeer-reviewed·May 28, 2025
- Single molecule sequencing of THCA synthase reveals copy number variation in modern drug-type Cannabis sativa L.
Kevin J McKernan, Yvonne Helbert, … Douglas Smith·bioRxiv·Oct 8, 2015
- A complete Cannabis chromosome assembly and adaptive admixture for elevated cannabidiol (CBD) content
Christopher J. Grassa, Jonathan P. Wenger, … George D. Weiblen·bioRxiv·Dec 11, 2018
- Genome-wide characterization of the MLO gene family in Cannabis sativa reveals two genes as strong candidates for powdery mildew susceptibility
N. Pepin, F. O. Hebert, D. L. Joly·bioRxiv·Jul 16, 2021
- Accumulation of somatic mutations leads to genetic mosaicism in Cannabis
K. Adamek, A. M. P. Jones, D. Torkamaneh·bioRxiv·Feb 11, 2021