At the time of this study, no male C. sativa genome assembly was publicly available. The only available genome from a male individual, ‘JL_father’ (GCA_013030025.1), was assembled at the scaffold level, making it impossible to easily identify Y chromosomal sequences.

To obtain sequences which likely originated from the Y chromosome, we previously described a computational pipeline (Shi et al. 2025). The pipeline uses RNA-seq data and hence does not rely on genome assemblies containing a Y chromosome. Briefly, we used RNA-seq samples from male and female C. sativa plants from the same developmental stage and used the following criteria to identify putative Y chromosomal transcripts.

Transcripts were required to

1. not map to the female CBDRx (GCA_900626175.2) genome.
2. have a sequence signature (k-mer of 16 nucleotides) shared by all male samples but different from all female samples.
3. be male biased in their expression.

The pipeline resulted in the initial identification of 379 transcripts putatively originating from the Y chromosome (Shi et al. 2025).

To build on the Shi et al. 2025 method and further reduce the inclusion of false positives, a strict TPM expression level filter (average TPM > 10 in male samples, and TPM = 0 in all female samples) was applied. This allowed us to identify 25 transcripts that originated from the Y chromosome with relatively high certainty (Supplementary Data S1). From those transcripts, the genomic sequence was retrieved through a BLAST search of the ‘JL_father’ assembly. Even though this genome is not assembled at the chromosome level, blasting the mRNA sequences of the 25 putatively Y chromosomal transcripts allowed us to identify scaffolds that are most likely part of the Y chromosome. The genomic sequences of the 25 genes were then aligned to putatively corresponding sections on the X chromosome identified through a BLAST search of the CBDRx genome assembly (GCA_900626175.2). The genes with the highest sequence divergence between X and Y that were deemed the most suitable for a PCR assay to distinguish XX and XY genotypes were FE.chrY.t5, FE.chrY.t9, FE.chrY.t122, FE.chrY.t25, FE.chrY.t42 and FE.chrY.t47 (Supplementary Data S2, transcript naming after Shi et al. 2025). Of those, we focussed on FE.chrY.t9.

FE.chrY.t9 is homologous to the PRECOCIOUS DISSOCIATION OF SISTERS 5 (PDS5) gene from Arabidopsis thaliana and was therefore renamed CsPDS5. This gene is positioned at 80.6 Mbp in chromosome X in what we believe is an ancient stratum (Toscani et al. 2025). PDS5 proteins play a crucial role in regulating genome architecture by influencing the 3D chromatin organisation; the A. thaliana PDS5 genes encode cohesin cofactors, the depletion of which compromises development, fertility and homologous recombination (HR) during DNA repair mechanisms, and it also causes subtle meiotic alterations (Pradillo et al. 2015).

Although both the X and Y chromosome versions of CsPDS5 are diverged, there is a highly conserved region, which allowed the design of a common pair of primers that amplified a 419 bp region from the X as well as from the Y chromosome. Importantly, the Y version contains an EcoRI restriction enzyme recognition inside the amplicon, which is not present in the X allele. Thus, a CAPS assay in which the PCR products are subjected to digestion with EcoRI results in a 157 bp and a 262 bp band for Y chromosome amplicons, while X chromosome amplicons remain at their original 419 bp length (Fig. 1). Hence, visualisation via an agarose gel results in only one band at 419 bp for female and monoecious (XX) samples, while the male (XY) samples display three bands, one corresponding to the amplicon derived from the X chromosome at 419 bp and two smaller bands at 157 bp and 262 bp, corresponding to the digested amplicon derived from the Y chromosome (Fig. 1). We term this assay CsPDS5-CAPS hereafter.

The assay was initially tested using the dioecious cultivar ‘FINOLA’. Leaf samples collected at the flowering stage, when sexual dimorphism was fully apparent, were used to verify that the sex as determined by CsPDS5-CAPS agreed with the phenotypic sex. Subsequently, ‘FINOLA’ samples at different developmental stages (from germination to flowering) and from different tissues (cotyledon and leaf) were tested, and the sex as determined by CsPDS5-CAPS was compared to the phenotypic sex. In total, 312 ‘FINOLA’ samples were tested, and the CsPDS5-CAPS genotyped sex matched the phenotypic sex in all cases.

Subsequently, we assessed the performance of CsPDS5-CAPS for 14 different hemp-type cultivars (Figure 2, Table 1), six dioecious (‘FINOLA’, ‘Kompolti’, ‘CSR-1’, ‘CFX-2’, ‘Estica’ and ‘Enectarol’) and eight monoecious (‘Felina 32’, ‘Bialobrzeski’, ‘Santhica 27’, ‘Earlina 8 FC’, ‘Fedora 17’, ‘Ferimon’, ‘Futura 75’ and ‘Henola’). For all dioecious cultivars, our genotyping distinguished male and female individuals (Fig. 2a). Monoecious cultivars are typically reported to carry two X chromosomes (Faux et al. 2014) and consequently appeared genotypically female in our assay (Fig. 2b).

To test the suitability of CsPDS5-CAPS for breeding projects, we further tested the assay on progeny plants resulting from crosses between different hemp-type cultivars. For this purpose, 137 F2 individuals from a previously established ‘FINOLA’ × ‘Felina 32’ cross (Dowling et al. 2024) were tested. As ‘FINOLA’ is dioecious and ‘Felina 32’ is monoecious, the F2 population comprises male, female and monoecious individuals (Dowling et al 2024). In all 137 cases tested, our genotyping results were in agreement with the phenotypic sex (Fig. 2c). We next tested F1 individuals from five additional crosses (‘FINOLA’ × ‘Earlina 8 FC’, ‘FINOLA’ × ‘Ferimon’, ‘FINOLA’ × ‘Santhica 27’, ‘FINOLA’ × ‘Fedora 17’ and ‘Felina 32’ × ‘Estica’). Also in this case, the CsPDS5-CAPS results were in perfect agreement with the phenotypic sex.

In total, over 500 different samples were used to test our genotyping assay (Table 1). In all cases, the genetic and phenotypic sex matched, showing that the CsPDS5-CAPS is a highly reliable assay and produces robust results independently of the cultivar, crossing, developmental stage or tissue.

To improve the CsPDS5-CAPS assay, we adapted it into a high-throughput TaqMan assay. The method is based on a pair of primers that amplify a ≈200 bp fragment of the CsPDS5 gene region, in both the X and Y versions, which contains the EcoRI restriction site. A male-specific probe was designed to bind at the restriction site present in the Y chromosome, while a female-specific probe binds to the corresponding sequence in the X chromosome. Each probe is labelled with a different reporter dye: the Fem probe with VIC and the Male probe with FAM. Hence, in this TaqMan PCR setup, XX samples produce only a VIC signal, while XY samples yield both VIC and FAM signals. We validated this assay with 22 ‘FINOLA’ samples (11 male and 11 female) previously genotyped using CsPDS5-CAPS, and the results were consistent between both approaches, as shown in the allelic discrimination plot (Fig. 2d).

We next analysed whether the CsPDS5 polymorphism utilised in our assay is also conserved in other hemp and marijuana cultivars that were not tested by PCR here but for which the genome sequences are available. We aligned the X and Y chromosomal versions of CsPDS5 regions from different cultivars: the marijuana cultivars ‘Ace High 3–2’ and ‘SourDiesel’, the high CBD hemp cultivars ‘Pink Pepper’ and ‘GoldenRedwood’ and a US hemp landrace variety named ‘BooneCounty’ (Lynch et al. 2025), and compared to the CsPDS5 sequence of ‘Kompolti’, for which we successfully conducted the CsPDS5-CAPS assay (Fig. 2a). The alignment shows that the CsPDS5 primer binding sites are conserved in all sequences, while the EcoRI restriction recognition site in CsPDS5 is present on all analysed Y chromosomal but absent in X chromosomal sequences (Fig. 3). This demonstrates that the selected region is extremely conserved across the different cultivars and that the assay introduced here is potentially useful for a wide variety of C. sativa cultivars and landraces.