MATEs have been identified in a variety of plant species, including 56 MATEs in Arabidopsis, 55 in rice (Wang et al., 2016). 67 in tomato (Santos et al., 2017), 64 in potato (Huang et al., 2021), 70 in Medicago truncatula (Wang et al., 2017), 72 in cotton (Xu et al., 2019) and 117 in soybean (Liu et al., 2016). In the present study, we identified 42 CsMATEs and investigated their physical-chemical properties, gene distribution, evolutionary relationships, conserved motifs, gene structures and gene expressions. CsMATEs contains only 42 members ( Table 1 ) and is the fewest when compared with other plant species, which suggests CsMATEs may have undergone contraction during evolution. Those 42 CsMATEs were divided into four subfamilies and distributed on 10 chromosomes, with ~50% CsMATEs adjacent to at least one another CsMATEs ( Figure 1 ), thus the expansion of CsMATEs might be largely due to tandem duplication (Wang et al., 2019).

Although cannabinoids synthetic pathways have already been illustrated, the translocation of the cannabinoids and biosynthetic intermediates are not known (Güllck and Moller, 2020). This inevitably hinders cannabinoids heterologous biosynthesis (Güllck et al., 2020). Cannabinoids are synthesized in glandular trichomes and flowers of female C. sative and their biosynthetic processes undergo cellular compartmentalization (Güllck and Moller, 2020). For instance, plastid localized CsaPT4 prenylates cytosolic synthesized OA in chloroplast to yield CBGA (Taura et al., 2009; Gagne et al., 2012; Luo et al., 2019; Güllck et al., 2020). Metabolite transporter proteins that could shuttle from cytosol to chloroplast should exist in C sativa cells. However, subcellular localization prediction indicates that vast majority of CsMATEs locate in vacuole or plasma membrane ( Table 1 ). Although CsMATE39 contains putative chloroplast localization signal ( Table 1 ), it expresses mainly in leaf ( Figure 4A ), the tissue that is not the most abundant cannabinoids accumulate. We assume that transporters from other families might be involved in translocating phenolic OA such as ATP binding cassette transporters which could translocate plant phenolic metabolites (Hwang et al., 2016). THCAS is secreted to apoplastic space of the glandular trichome to carry out oxidative cyclization of CBGA (Sirikantaramas et al., 2005). The plastid produced CBGA thus should be translocated across the plasm membrane. Thirty-five CsMATEs are predicted to localized on plasm membrane ( Table 1 ). Of them, CsMATE03, CsMATE05, CsMATE23, CsMATE28, CsMATE31, CsMATE34 and CsMATE40 had a highest expression in flower but also expressed moderately in stem and leaf (the tissues that have glandular trichomes) according to transcriptome data ( Figure 4A ) and qRT-PCR ( Figure 4B ). Although we don’t have the trichome-specific transcriptional data of those seven genes in variety DK, they uniformly expressed in the trichomes of the other nine C. sativa varieties (Zager et al., 2019, Supplementary Figure S2 ). Moreover, the seven genes exhibited similar expression pattern in different tissues as that of cannabinoid synthetic genes ( Figure 4A ), which further indicates the involvement CsMATEs in biosynthesis of cannabinoids.

Within the seven genes, except for CsMATE23, CsMATE28, CsMATE34 and CsMATE40 all have an unexpressed alternative spliceform in tissues we studied ( Figure 5A ). Light responsive cis-acting elements were abundantly detected in promoters of the representative CsMATEs (CsMATE23, CsMATE28 and CsMATE34), reflects those genes could be regulated by lights. UV-B is known to affect accumulation of cannabinoids (Lydon et al., 1987). In accordance with this, the contents of cannabinoids and corresponding intermediates were changed under UV-B treatment ( Figure 6A ). As expected, the variation trend of CBGA was analogous to its direct downstream THCA, CBDA and CBG ( Figure 6A ), reflects the precursor nature of the CBGA (Güllck et al., 2020). Notably, although overall variation of OA is similar to CBGA, accumulation of OA is opposed to CBGA at two hours treatment where OA content was decreased but CBGA contents was increased compared to zero hours treatment ( Figure 6A ). Meanwhile, GPP did not show any change at two hours treatment ( Figure 6A ). Therefore, increase of CBGA after two hours UV-B treatment could be largely caused by over-usage of OA for synthesizing CBGA either due to enhanced CBGAS activity or more efficient translocation of OA form cytosol to plastid. Unfortunately, none of the CsMATEs we tested was up-regulated after two hours treatment ( Figure 6B ). We reasoned that plastid located ATP binding cassette transporters that translocate phenolic metabolites might responsible for OA transportation (Hwang et al., 2016). Moreover, Although the transcriptional variation of CsMATE23 is analogous to the content of OA ( Figure 6B ), OA is synthesized in cytosol and CsMATE23 is localized in plasma membrane ( Table 1 ). Meanwhile, CsMATE23 is barely transcribed in C sativa root, however its Arabidopsis ortholog AtRHS2 (AtDTX31) located in plasma membrane of root is required for hairy root elongation (Won et al., 2009). Hence, we may also exclude participation of CsMATE23 in transporting OA precursors.

The most interesting finding we observed is the correlation between transcriptional alterations of CsMATEs with the accumulation of CBGA, THCA and CBDA ( Figure 6 ). As the precursor of THCA and CBDA, CBGA content did not show much alteration at 12 hours treatment compared to the control zero hour ( Figure 6A ). However, the contents of THCA, THC, CBDA and CBD were all increased at least more than four times at 12 hours compared to zero hours and transcriptions of CsMATE28 and CsMATE34 were also significantly enhanced at 12 hours ( Figure 6B ). This could be explained by that increased CsMATE28 or CsMATE34 more inefficiently transport CBGA to the apoplastic space to enable its subsequent conversions (Sirikantaramas et al., 2005).