Spatial distribution of isoprenoid enzymes and MpABCG1 transporter influences sesquiterpene accumulation in Marchantia polymorpha oil bodies

Abstract

Marchantia polymorpha oil bodies (OBs) are specialized cell structures housing a diverse array of C15-terpenes, called sesquiterpenes. These compounds act as herbivore repellents, yet the enzymes responsible for the biosynthesis of their precursors remain poorly characterized. We investigated the localization of isoprenoid biosynthetic enzymes using translational and transcriptional reporters, coupled with confocal microscopy. Most enzymes localized as predicted (e.g., cytosol, plastid and the endoplasmic reticulum), and OB cells were identified as the primary sites of sesquiterpene biosynthesis. To explore OBs as potential storage sites for terpenes, we attempted to produce exogenous but easily identifiable compounds in Marchantia, such as the diterpene taxadiene and the triterpene β-amyrin. Targeting to OB cells resulted in measurable amounts of these compounds, but their yields remained unaffected by the overexpression of key precursor genes, underscoring challenges in redirecting metabolic flux. To further investigate terpene accumulation in OBs, we focused on MpABCG1, an ABC transporter previously reported to localize at the OB membrane. While MpABCG1 overexpression mildly increased endogenous sesquiterpene levels, its disruption via CRISPR dramatically reduced sesquiterpene accumulation. These findings establish that MpABCG1 is necessary for sesquiterpene accumulation in OBs and add to current knowledge of terpene synthesis compartmentalisation in Marchantia polymorpha.

A study maps terpene biosynthesis in Marchantia polymorpha oil bodies and identifies the ABC transporter MpABCG1 as a key factor for endogenous sesquiterpene accumulation in oil bodies, informing future metabolic engineering strategies.

Introduction

In recent years, Marchantia polymorpha has emerged as a useful model plant system¹ for studying terpene biosynthesis and accumulation within specialized storage structures known as oil bodies (OBs)². The OBs in Marchantia are membrane-bound compartments that primarily house terpenes³ and bisbibenzyls⁴, a class of phenylpropanoid derivatives; the mixture confers anti-feedant properties to protect against herbivores⁵. Terpenes are a diverse class of natural products derived from the condensation of isoprene (C₅H₈) units⁶, found across all life, with the greatest structural and functional diversity observed in the plant kingdom. Beyond deterring herbivores as observed in Marchantia, they serve other vital roles in plants, such as preventing fungal infections or attracting pollinators. Additionally, a broad range of bioactive properties makes them valuable in medicine, agriculture, food and other industrial applications.

The sequestration of bisbibenzyls and terpenes into OBs is critical for protecting Marchantia from the potential toxicity of these metabolites when released into other tissues. The unique compartmentalization offered by OBs presents a promising avenue for metabolic engineering, as it provides a natural reservoir for accumulating metabolites that might otherwise be harmful to the plant.

Recent developmental studies have provided significant insights into the mechanisms governing OB cell fate determination and formation, highlighting key transcription factors (TFs) involved in these processes. Specific TFs, such as MpERF13⁷, MpC1HDZ⁵, MpTGA⁸ and MpMYB2^9,10 have been identified as crucial regulators of OB cell differentiation and maturation¹¹. These discoveries have expanded our understanding of how OB cells develop and contribute to the plant's metabolic capacity.

Manipulating these TFs has further revealed the delicate balance between OB cell proliferation and plant fitness. For instance, CRISPR-mediated mutation of MpTGA⁸ or gain-of-function mutants of MpERF13 led to a dramatic increase in OB cell numbers per plant⁷. However, these alterations were accompanied by reduced growth rates or morphological defects, suggesting that OB cell number is tightly regulated in Marchantia to avoid compromising overall plant health and fitness, or that these TFs have pleiotropic effects that influence additional developmental processes.

Early studies on Marchantia’s terpene biosynthetic pathway, which were based on immunolocalization of isoprenoid enzymes, hypothesized that key enzymatic steps might occur at the oil body (OB) membrane³. Today, with the availability of whole-genome sequencing and detailed gene annotations informed by sequence homology with other species, our understanding of the early steps of isoprenoid biosynthesis has significantly improved^12–14 (Table 1 and S1). The mevalonate (MVA) and methylerythritol phosphate (MEP)^15–17 pathways are the two primary metabolic routes responsible for producing the precursors—isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate—for various classes of terpenes (Table 1). The MVA pathway, comprising seven enzymatic steps¹⁸ and typically operating in the cytosol¹⁹, endoplasmic reticulum²⁰ (ER) and peroxisome²¹, generally supplies precursors to sesquiterpenes and triterpenes synthesis^22,23, while the MEP pathway, with seven/eight enzymatic steps and primarily localized in plastids²⁴, commonly provides precursors for monoterpenes and diterpenes^22,23 (Table 1 and S1). In Marchantia, sesquiterpenes are the dominant terpene class found in OBs², produced by several microbial- and fungal-type terpene synthases²⁵. Additionally, Marchantia produces the monoterpene limonene, likely synthesized through the cis-prenyl precursor neryl pyrophosphate rather than the transform geranyl pyrophosphate (GPP), as demonstrated by Kumar et al.²⁶. While cell-type-specific metabolic analyses revealed that only sesquiterpenes are stored inside the OB compartment², plants defective in OB cell differentiation show reduced levels of both sesquiterpenes and monoterpenes⁵, making it unclear whether limonene is specifically stored within OBs.

Table 1.

Overview of MVA, MEP, and terpene scaffold biosynthetic genes in Marchantia polymorpha

	Name	Abbreviation	Putative localization (DeepLoc 2.1)
MEP	1-Deoxy-D-xylulose-5-phosphate synthase 1	MpDXS1	Plastids/soluble
	1-Deoxy-D-xylulose-5-phosphate synthase 2	MpDXS2
	1-Deoxy-D-xylulose 5-phosphate reductoisomerase 1	MpDXR1
	1-Deoxy-D-xylulose 5-phosphate reductoisomerase 2	MpDXR2
	4-Diphosphocytidyl-2-C-methyl-D-erythritol synthase	MpMCT
	4-Diphosphocytidyl-2-C-methyl-D-erythritol 2-phosphate kinase	MpCMK
	2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase	MpMDS
	1-Hydroxy-2-methyl-2-(E)-butenyl 4-diphosphate synthase	MpHDS
	1-Hydroxy-2-methyl-2-(E)-butenyl 4-diphosphate reductase	MpHDR
	Isopentenyl diphosphate isomerase 1	MpIDI1
	Isoprenyl (geranylgeranyl) diphosphate synthase 2	MpIDS2/MpGGPPS	Plastids/soluble
	Isoprenyl diphosphate synthase 3	MpIDS3	Plastids/soluble
	Isoprenyl (solanesyl) diphosphate synthase 4	MpIDS4	Plastids/soluble
	Isoprenyl diphosphate synthase 5	MpIDS5	Mitochondria/soluble
MVA	Acetoacetyl-coA thiolase 2	MpACT2	Peroxisome/soluble
	3-Hydroxy-3-methylglutaryl-coA synthase	MpHMGS	Cytoplasm
	3-Hydroxy-3-methylglutaryl-coA reductase	MpHMG/MpHMGR	ER/transmembrane
	Mevalonate-5-kinase	MpMK	Cytoplasm
	Phosphomevanolate kinase	MpPMK	Cytoplasm
	Mevanolate diphosphate decarboxylase.	MpMVD	Cytoplasm
	Isopentenyl diphosphate isomerase 2	MpIDI2	Cytoplasm
	Farnesyl pyrophosphate synthase	MpIDS1/MpFPS	Mitochondria/soluble
	Squalene synthase 3	MpSQS3	ER/transmembrane
	Squalene/phytoene synthase 1	MpSQS1	Plastids/peripheral
	Squalene/phytoene synthase 2	MpSQS2	Mitochondria/peripheral

This table lists key genes studied or referenced in this work, including those involved in the methylerythritol phosphate (MEP) and mevalonate (MVA) pathways, as well as enzymes responsible for synthesizing terpene scaffolds for different subclasses. Abbreviations correspond to gene names, and their putative protein localizations were predicted using DeepLoc 2.1. Detailed annotations and functional descriptions for each gene are provided in Table S1.

Examination of genome data^12–14,27 along with computational predictive tools for enzyme localization, such as DeepLoc 2.1²⁸, enabled us to identify putative precursor genes involved in prenyl phosphate synthesis for terpene production (Table 1). However, discrepancies remain between earlier theories suggesting that terpene biosynthetic enzymes are localized to OB membranes³ and more recent predictions that place MVA enzymes in the cytosol, ER and peroxisome, and MEP enzymes in plastids (Table 1). These discrepancies prompted us to re-examine the localization of key enzymes using modern techniques such as confocal microscopy combined with translational reporters⁷. This approach allowed us to identify the specific cells expressing these genes and determine the subcellular localization of their corresponding proteins.

Building on our findings regarding terpene precursor gene expression and localization, we then explored the potential for metabolic engineering in Marchantia polymorpha’s OBs. Specifically, we aimed to produce two valuable terpenes within OBs: the diterpene taxadiene²⁹, a precursor to the valuable anti-cancer compound TaxolⓇ^30,31, and the triterpene β-amyrin³², a precursor to compounds such as glycyrrhizin³³, a sweet-tasting constituent of liquorice. These compounds were selected due to their distinct chemical signatures, which facilitate easy detection in chromatographic analyses, given the minimal background of these terpene subclasses in Marchantia. To further evaluate the metabolic capacity of Marchantia and investigate the transport of sesquiterpenes, we also examined the production of the sesquiterpene amorpha-4,11-diene³⁴ alongside MpABCG1, an ATP-binding cassette (ABC) transporter highly and exclusively expressed in OB cells³⁵. Prior studies have demonstrated its clear localization to the OB membrane using a translational reporter⁷, suggesting a potential role in metabolite transport. This study provides new insights into the involvement of MpABCG1 in terpene accumulation within OBs and contributes to a broader understanding of the partitioning of terpene metabolism in Marchantia polymorpha.

Results

Translational and transcriptional reporters of terpenoid precursor pathway genes

To determine the expression pattern and subcellular localization of key enzymes involved in Marchantia's terpene biosynthesis, we generated constructs of translational reporters, in which each gene’s native promoter drove its respective coding sequence fused to the fluorescent tag mVenus³⁶. In cases where translational reporters were unavailable due to technical challenges, transcriptional reporters were used instead to determine cell-type expression only. These consisted of Marchantia’s native promoter driving mVenus with a nuclear localization signal³⁷ (mVenus-N7), enabling us to investigate the spatial and temporal expression of key genes. Promoter lengths were selected based on Assay for Transposase-Accessible Chromatin with Sequencing (ATAC-seq)³⁸ data from version 6.1 of the Tak accession in the Marchantia.info database^13,25 (Table S2). ATAC-seq identifies regions of open chromatin, which are more accessible to TFs and associated with regulatory domains that influence gene expression. Since relaxed chromatin is frequently found near the start of a gene, it serves as a useful indicator for selecting promoter regions likely to drive gene expression in their native cellular contexts. To delineate cell boundaries, each construct also included a fluorescent membrane marker, mScarlet-LTI6b fusion^39–41, driven by the strong, constitutive ubiquitin-conjugating enzyme E2 gene promoter and 5’UTR from Marchantia polymorpha (_ProUBE2)⁴¹.

For the identification of different subcellular compartments, such as chloroplasts/plastids, ER, cytosol, Golgi, and mitochondria, we interpreted our findings by comparing the localization patterns observed in our images to those of previously characterized standard reporters in Marchantia^41,42, though distinguishing between the ER and cytosol remained challenging under the resolution constraints of our experimental setup.

For this study, we selected key genes from the MVA and MEP pathways on the basis of their predicted roles as rate-limiting enzymes, their redundancy, or their involvement in the final steps of the pathway. Additionally, we included genes responsible for the formation of linear scaffold precursors for sesquiterpenes, diterpenes, and triterpenes to capture a more comprehensive view of terpene biosynthesis. Among these, five genes were annotated as isoprenyl diphosphate synthases (IDS)²⁷ (Table 1), which integrates functional annotation with phylogenetic relationships via its orthophyloviewer tool²⁷. We used these data, alongside sequence comparisons with Arabidopsis homologs, to refine tentative annotations for IDS genes likely involved in terpene precursor formation for this study. The remaining genes analyzed here were already clearly annotated in Marchantia.info thanks to the extensive curation by Chen et al.²⁷, and did not require further reannotation.

We first analyzed the two annotated 1-deoxy-D-xylulose-5-phosphate synthases (MpDXS1 and MpDXS2) (Table 1), which encode the first committed and rate-limiting enzymes of the MEP pathway⁴³. For MpDXS1, tagged with mVenus and driven by its native promoter and 5’UTR (_ProMpDXS1:DXS1-mVenus), mVenus fluorescence was observed in all cell types throughout the plant, with a stronger signal in OB cells (Fig. 1A). The protein localized to plastids, displaying a punctate pattern in non-OB cells rather than a homogeneous distribution (Fig. 1A). By contrast, _ProMpDXS2:DXS2-mVenus showed a restricted localization pattern, with fluorescence observed only in oil body plastids from Day 0 gemmae to 14-day-old plants (Figs. 1B and S1C).

Fig. 1. Confocal imaging of translational reporters for key Marchantia polymorpha isoprenoid biosynthetic genes.

A _ProMpDXS1:DXS1-mVenus, B _ProMpDXS2:DXS2-mVenus, C _ProBidirectional MpHDR/HDS:HDS-eGFP, D _ProBidirectional MpHDS/HDR:HDR-mVenus, E _ProMpGGPPS:GGPPS-mVenus, F _ProMpFPS:FPS-mVenus, and G _ProMpHMGR:HMGR-mVenus. Each construct is represented by two images: the full gemmae at day 0 (left panels, scale bar: 50 μm) and a higher magnification view showing subcellular localization (right panels, scale bar: 5 μm). H Simplified biosynthetic pathway scheme highlighting enzymatic steps analyzed in this study. The mVenus signal (yellow) or eGFP (green) for MpHDS (C) indicates the subcellular localization of the corresponding enzymes, while mScarlet (purple) delineates the cellular boundaries. For the magnified image in (A), chloroplast/plastid autofluorescence (gray) is included to demonstrate sublocalization of MpDXS1 in the plastids. G includes one whole-gemma image and two high magnification images to illustrate the localization of MpHMGR-mVenus to the ER and Golgi apparatus. White arrow indicates the localizations to the Golgi apparatus.

We examined the final steps of the MEP pathway by generating a translational reporter for (E)-4-hydroxy-3-methylbut-2-en-1-yl diphosphate reductase (MpHDR)^44,45 (Table 1). We also generated a translational reporter for (E)-4-hydroxy-3-methylbut-2-enyl diphosphate synthase (MpHDS)^44,46 (Table 1). MpHDS is adjacent to MpHDR and oriented in the opposite direction on the same chromosome, suggesting they share a bidirectional promoter (Fig. S2A). Previous studies have highlighted the potentially toxic nature of the product of HDS^47,48, which is detoxified in the subsequent enzymatic step by HDR^47,48, making it particularly interesting to study both enzymes together. To our knowledge, this is the first report suggesting that MpHDS and MpHDR genes may be regulated by a shared promoter region. Initial experiments revealed that the 2 kb region upstream of MpHDR was insufficient to drive expression, whereas the 1.9 kb region upstream of MpHDS was adequate (Fig. S2B). Expression of a construct driving MpHDS-eGFP and MpHDR-mVenus bidirectionally (Fig. S2C) demonstrated that, like other MEP enzymes, both proteins localized to plastids (Fig. 1C, D). Their expression was stronger, if not exclusive, to OB cells in Day 0 gemmae (Fig. 1C, D), before becoming more widespread across tissues at later developmental stages (Fig. S1A, B).

Finally, we analyzed the subcellular localization and cell-type expression of the enzyme forming the linear diterpene precursor geranylgeranyl pyrophosphate (GGPP)^49,50. Among the IDS candidates²⁷, MpIDS2 and MpIDS3 share 68% and 62% protein homology, respectively, with GERANYLGERANYL PYROPHOSPHATE SYNTHASE 11 (GGPPS11) from Arabidopsis thaliana, the most highly expressed and functionally dominant GGPPS in Arabidopsis⁵¹, which operates as a homodimer⁵². However, transcriptomic data revealed that MpIDS3 exhibited consistently low expression across all tissues and developmental stages^12–14. The other IDS candidates, MpIDS4 and MpIDS5, share 71% and 57% protein homology with SOLANESYL PYROPHOSPHATE SYNTHASE and GERANYL PYROPHOSPHATE SYNTHASE, respectively, suggesting potential roles in ubiquinone biosynthesis⁵³ (MpIDS4) and trans-monoterpene precursor formation (MpIDS5). MpIDS2 emerged as the strongest candidate for the primary GGPPS in Marchantia, given its higher sequence homology and stronger expression. We therefore re-annotated it as MpGGPPS (Table 1 and S1). The translational fusion to MpGGPPS showed localization to plastids in all cell types, with particularly strong expression in OB cells (Fig. 1E).

For the MVA pathway and sesquiterpene synthesis, we examined the rate-limiting enzyme 3-HYDROXY-3-METHYLGLUTARYL-COA REDUCTASE (MpHMGR)⁵⁴ (Table 1). Expression of _ProMpHMGR:MpHMGR-mVenus revealed a restricted localization to oil body cells in gemmae (Fig. 1G) and 14-day-old plants (Fig. S1D). The enzyme appeared to display a dual localization pattern: a punctate distribution around the OB, indicative of Golgi, and a web-like pattern surrounding the nucleus, consistent with ER localization (Fig. 1G). As a key enzyme downstream of the MVA pathway, we analyzed FARNESYL PYROPHOSPHATE SYNTHASE (MpFPS) (Table 1), which forms farnesyl pyrophosphate (FPP)⁵⁵, the linear precursor to all sesquiterpenes. Annotated as IDS1²⁷ in the Marchantia.info database, MpFPS shares 66% sequence similarity with both FPS homologs from Arabidopsis thaliana. Expression analysis revealed that MpFPS is specifically expressed in OB cells, persisting even after 14 days (Figs. 1F and S1E). Contrary to the prediction of mitochondrial localization by DeepLoc 2.1²⁸ (Table 1), the enzyme appeared predominantly cytosolic, as indicated by its uniform signal surrounding the OB and delineating the plastids (Fig. 1F).

We further investigated the expression pattern⁵⁶ of the promoters driving the genes encoding 1-deoxy-D-xylulose 5-phosphate reductoisomerases^57,58 (MpDXR1 and MpDXR2) and mevalonate diphosphate decarboxylase (MpMVD)⁵⁹ using transcriptional reporters. The corresponding enzymes, MpDXRs and MpMVD, catalyze the second enzymatic step of the MEP pathway and the penultimate enzymatic step of the MVA pathway, respectively (Table 1). _ProMpDXR1:mVenus-N7 showed ubiquitous expression (Fig. 2A), whereas _ProMpDXR2:mVenus-N7 showed specificity to OB cells in Day 0 gemma (Fig. 2B). Similarly, _ProMpMVD:mVenus-N7 exhibited specificity to OB cells, even in 14-day-old thalli (Fig. 2C).

Fig. 2. Confocal imaging of transcriptional reporters for additional isoprenoid biosynthetic genes in Marchantia polymorpha.

A _ProMpDXR1:mVenus-N7, B _ProMpDXR2:mVenus-N7, C _ProMpMVD:mVenus-N7, and D _ProMpSQS3:mVenus-N7. The mVenus signal (yellow), localized to the nucleus, indicates the activation of these promoters in specific cell types, while mScarlet fluorescence (purple) delineates cellular boundaries. Each construct is represented by two panels: day 0 gemmae (left, scale bar: 50 μm) and the meristem area of 14-day-old plants (right, scale bar: 50 μm). White arrows indicate mVenus signals detected in oil body cells.

Finally, given that the expression of some MVA biosynthetic genes and MpFPS appears specific to OB cells, we investigated whether this pattern extends to phytosterol biosynthesis. Phytosterols are essential structural components of plant membranes⁶⁰ and are initially formed through the fusion of two FPP molecules to produce squalene, a reaction catalyzed by SQUALENE SYNTHASE⁶¹ (SQS). We therefore examined the transcriptional reporter of the MpSQS promoter. Among the three squalene synthase-like genes annotated in the Marchantia genome (MpSQS1-3) (Table 1 and S1), we selected MpSQS3 based on its higher protein sequence homology (60%) with Arabidopsis thaliana SQUALENE SYNTHASE. In contrast, MpSQS1 showed greater similarity (73%) to Arabidopsis PHYTOENE SYNTHASE, suggesting a role in carotenoid biosynthesis, while MpSQS2 shared 52% homology with an uncharacterized terpenoid synthase. Plants expressing the _ProMpSQS3:mVenus-N7 constructs show a mVenus signal in meristematic regions in Day 0 gemmae, followed by ubiquitous expression in later developmental stages (Fig. 2D). These findings indicate that sterol biosynthesis is likely to occur broadly throughout the plant, based on the widespread activity of the MpSQS3 promoter.

Although localization data reflect relative protein accumulation rather than absolute presence or absence, our findings provide a comprehensive map of the subcellular localization and expression patterns of key terpene biosynthetic enzymes, offering insights into the spatial organization of biosynthetic intermediates and terpene scaffolds in Marchantia polymorpha.

Exogenous production of taxadiene and β-amyrin in Marchantia polymorpha whole plants vs. oil body cells

Given the observed expression of terpene precursor enzymes predominantly in OB cells, we aimed to harness the metabolic potential of these specialized structures. As proof of concept, we attempted to produce the diterpene taxadiene²⁹, and the triterpene β-amyrin³².

We generated constructs to express Taxus baccata taxadiene synthase (TXS)⁶² or Talinum paniculatum β-amyrin synthase⁶³ (β-AS), driven by either the constitutive _ProUBE2 or the OB cell-specific MpR2R3-MYB2 promoter (_ProMYB2)³⁷, enabling expression throughout the entire plant or specifically within OB cells, respectively. TXS catalyzes the cyclization of the linear diterpene precursor GGPP into taxadiene, while β-AS converts the linear triterpene precursor 2,3-oxidosqualene into β-amyrin.

In plants expressing _ProUBE2:TXS, we detected a prominent new peak in hexane extracts (Fig. 3A), corresponding to taxadiene based on its mass spectrum (MS), as described in prior publications^64,65. While no commercial taxadiene standards were available, we identified the compound by its characteristic ion at m/z 122 and the molecular ion [M + ] at m/z 272 (Fig. S3A). Quantification of taxadiene in four independent transformants expressing either _ProUBE2:TXS or _ProMYB2:TXS revealed significantly higher levels of taxadiene in whole plants compared to those with expression restricted to OB cells (Fig. 3D). Specifically, _ProUBE2:TXS lines produced ~21 µg of taxadiene per g of fresh weight (FW), whereas _ProMYB2:TXS lines produced only 0.7 µg/g FW (Fig. 3D, E), dodecane equivalent. The correct targeting of TXS protein in OB cells was confirmed using a _ProMYB2:TXS-mVenus fusion construct, which showed plastid localization (Fig. 3C).

Fig. 3. Production of taxadiene and β-amyrin in Marchantia polymorpha and subcellular localization of their synthases.

A Total ion chromatograms (TICs) of terpene extracts from wild-type (WT, upper panel) and transgenic plants expressing _ProUBE2:TXS (taxadiene synthase, lower panel), highlighting the taxadiene peak (T). B TICs of trimethylsilyl-derivatized terpene extracts, showing WT (upper), a transgenic line expressing _ProUBE2:β-AS (β-amyrin synthase, middle), and an authentic β-amyrin standard (lower). Peaks: IS (internal standard: coprostan-3-ol), s1 (campesterol), s2 (stigmasterol), and β (β-amyrin). C Subcellular localization of TXS and β-AS fused to mVenus under the _ProMYB2 promoter, showing plastid localization for TXS (upper panel) and ER or cytosol localization for β-AS (lower panel). mScarlet fluorescence (purple) delineates cellular boundaries. Scale bar: 5 μm. D Quantification of taxadiene (upper graph) and β-amyrin (lower graph) in plants expressing _ProUBE2:TXS and _ProUBE2:β-AS, respectively (whole-plant expression) or _ProMYB2:TXS and _ProMYB2:β-AS respectively (oil body cell-specific expression). Data represent four independent lines per condition. E Quantification of taxadiene (left) and β-amyrin (right) in transgenic lines overexpressing precursor genes. For taxadiene, combinations included MpDXS1, MpDXS2, PpDXS1, PpDXS2, MpHDR, and MpGGPPS. For β-amyrin, MpHMGR, Mp-tHMGR, PpHMGR, and Pp-tHMGR were tested. Box plots represent data points from independent transformants. One-way ANOVA with Dunnett's multiple comparison tests (p < 0.05) showed no significant differences (ns) across gene combinations.

Similarly, plants expressing _ProUBE2:β-AS produced a new compound, identified as β-amyrin through its MS profile and comparison with an authentic standard (Figs. 3B and S3B). The _ProUBE2:β-AS lines produced an average of 58 µg of β-amyrin per g of fresh weight (coprostanol equivalent), whereas lines expressing the enzyme specifically in OB cells yielded an average of 1.5 µg/g FW (Fig. 3D, E). In MYB2-driven β-AS-mVenus lines, the protein localized specifically in the cytosol of the OB cells (Fig. 3C).

These data indicate that precursors are still present in non-OB cells and contribute to higher terpene yields in plants with ubiquitous expression. Interestingly, Suire et al.³ reported an approximate ratio of one OB cell to 150 non-OB cells in gemmae. Whether this ratio is maintained in older thalli remains unknown, as this would require measuring cell density in depth. Despite this uncertainty, our quantitative data show that OB-specific expression results in yields only ~30-fold lower for taxadiene and ~40-fold lower for β-amyrin, rather than a 150-fold decrease, which could hypothetically suggest that OB cells may have a higher precursor availability or storage capacity compared to the non-OB cells.

However, possible higher precursor availability in OB cells does not necessarily preclude flux limitations during engineered terpene biosynthesis. To investigate whether precursor supply was constraining the production of taxadiene and β-amyrin in these cells, we co-expressed TXS and β-AS with key rate-limiting enzymes of the MEP and MVA pathways (i.e., MpDXS and MpHMGR, respectively), as our work demonstrated that many of these enzymes appeared to be highly expressed in OB cells.

For taxadiene production via the MEP pathway, we co-expressed MpDXS1 or MpDXS2 with TXS. To mitigate potential homology-dependent gene silencing associated with overexpressing native Marchantia genes⁶⁶, we also expressed the homologous PpDXS1 and PpDXS2 genes from Physcomitrium patens, which have recently been re-annotated as PpDXS1A and PpDXS1D⁶⁷. To further enhance precursor flux, we introduced MpHDR and MpGGPPS, two key enzymes previously shown to increase GGPP flux for diterpene production in Nicotiana benthamiana^68,69.

For β-amyrin production through the MVA pathway⁷⁰, we overexpressed the HMGR genes from Marchantia (MpHMGR) and Physcomitrium (PpHMGR), alongside truncated versions designed to remove their negative feedback regulatory domain^71,72, thereby increasing the pool of terpene precursors and allowing unrestrained production of β-amyrin. These truncated variants (Mp-tHMGR and Pp-tHMGR) lack the first 148 and 146 amino acids, respectively, thereby removing both transmembrane domains. This truncation strategy follows the routinely used tHMGR design for triterpene production in N. benthamiana⁷³. All genes were expressed under the MYB2 promoter to specifically target OB cells.

Quantitative analysis of four independent transformants showed no significant increase in taxadiene or β-amyrin levels upon overexpression of these precursor supply genes (Fig. 3E). We confirmed the correct targeting to OB cells and subcellular localization of Marchantia DXS, HDR, GGPPS, HMGR, and tHMGR by fusing these genes with mVenus under the control of the MYB2 promoter (Fig. S4). MpDXSs, MpHDR, and MpGGPPS enzymes specifically localized to plastids of the OB cells (Fig. S4A–D). However, their overexpression to increase precursor availability for TXS did not result in increased taxadiene levels, nor in the level of main endogenous sesquiterpenes (Supplementary Data 1 and 2), which were tentatively identified by MS library matching and by comparison of Kovats retention indices⁷⁴ (Fig. S5 and Table S3). While Mp-tHMGR-mVenus appeared to localize in the ER and Golgi apparatus of OB cells (Fig. S4E), the native version driven by the MYB2 promoter remained undetectable. To address this, we re-cloned MpHMGR under the 2x35S promoter and successfully detected the protein, which accumulated exclusively in the ER of OB cells (Fig. S4F). Interestingly, when the truncated version (Mp-tHMGR) was expressed under the 2x35S promoter, the fusion protein showed expression across all cells (Fig. S4G), suggesting that the negative feedback domain removed in this construct may play a role in restricting localization to OB cells.

Overall, these results demonstrate that overexpression of selected rate-limiting enzymes was insufficient to significantly enhance yields of exogenous terpenes in Marchantia OB cells. This could reflect either that precursor availability was already sufficient to saturate the reactions or that additional regulatory mechanisms may play critical roles in terpene biosynthesis and accumulation in these specialized cells.

Accumulation of terpenes in OBs may require a specific transporter

Although overexpression of presumed rate-limiting enzymes is a standard strategy in terpene engineering, in our hands, it did not enhance exogenous terpene yields when targeted to OB cells. Our initial aim was to produce exogenous terpenes and rely either on their putative passive diffusion into the OB lumen or on the earlier hypothesis that biosynthetic enzymes positioned at the OB membrane, which we assumed could allow newly formed terpenes to enter the lumen. However, after determining the subcellular localization of the enzymes, we realized that these compounds may not naturally accumulate in OBs: taxadiene may primarily localize to plastids, while β-amyrin could accumulate in the cytosol. Furthermore, previous studies demonstrated high and exclusive expression of an MpABCG1 gene in oil body cells³⁵, with its protein localized to the OB membrane when expressed under its native promoter, and to the plasma membrane of other cells when driven by a non-OB-specific promoter⁷. This prompted us to hypothesize that precursor flux in Marchantia predominantly supplies sesquiterpene biosynthesis in the oil body cells and that a specific transporter, potentially MpABCG1, may be required, directly or indirectly, for their accumulation inside OBs.

To investigate whether MpABCG1 could facilitate the accumulation of any sesquiterpenes in Marchantia’s OBs, we expressed Artemisia annua amorphadiene synthase (AMS), a sesquiterpene synthase that converts FPP into amorpha-4,11-diene³⁴. Cyclisation of FPP into this sesquiterpene likely occurs via the bisabolyl cation⁷⁵, similar to the endogenous β-chamigrene⁷⁶, therefore we expected this structural similarity to enable amorpha-4,11-diene transport into OBs. We targeted AMS expression to either the whole plant (_ProUBE2) or specifically in OB cells (_ProMYB2). Additionally, we co-expressed AMS and MpABCG1 in OB cells, alongside MpHMGR, MpDXS2, and MpFPS, with FPS enzyme often referred to as a rate-limiting enzyme in sesquiterpene biosynthesis⁷⁷, or alternatively as a key step in substrate channeling toward FPP. By introducing MpABCG1 and MpFPS, we aimed to overcome potential transport limitations and to test whether this combination could further increase FPP availability for sesquiterpene production. Amorpha-4,11-diene was detected only in transgenic lines expressing the construct _ProUBE2:AMS (Fig. S6), with an average yield of 0.75 µg/g FW in six out of eight independent transformants (Fig. 4A, Supplementary data 3). No amorpha-4,11-diene was detected when expression was targeted to OB cells, either with or without precursor supply genes and MpABCG1 (Fig. S6). The absence of detection is consistent with the low yields observed in whole-plant _ProUBE2:AMS lines, where amorpha-4,11-diene levels were nearly two orders of magnitude lower than taxadiene or β-amyrin in comparable constructs, and may therefore have fallen below the detection threshold in OB-targeted lines. Localization studies of AMS-mVenus and MpFPS-mVenus driven by the MYB2 promoter showed that both proteins localized to the cytosol of OB cells (Fig. 4B).

Fig. 4. Expression of amorphadiene synthase (AMS) in Marchantia polymorpha with or without precursor supply genes and the MpABCG1 transporter, and its effect on endogenous sesquiterpene levels.

A Quantification of amorpha-4,11-diene levels in plants expressing Artemisia annua AMS under the Marchantia UBE2 promoter. The compound was detected in six independent transformants (mean ± SD; n = 6). B Subcellular localization of MpFPS and AMS proteins when expressed under the oil body-specific MYB2 promoter. mVenus fluorescence (yellow) indicates protein localization, with signals observed in oil body cells, while mScarlet fluorescence (purple) marks cell boundaries. Scale bars: 5 μm. C Quantification of endogenous sesquiterpenes, including cis-thujopsene, β-chamigrene, γ-cuprenene, and thujopsan-2α-ol, in lines expressing AaAMS with or without precursor supply genes (MpFPS, MpDXS2, and/or MpHMGR) and with or without MpABCG1, under _ProMYB2. Box plots represent data points from eight independent transformants (n = 8), with significance determined by one-way ANOVA followed by Dunnett's multiple comparison test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns  not significant).

When measuring endogenous sesquiterpene levels in at least eight independent transformants for each gene combination, we observed that, compared to lines expressing _ProMYB2:AMS alone, co-expression of precursor genes significantly increased sesquiterpene levels (Fig. 4C, Supplementary Data 3). Among combinations without MpABCG1, MpFPS + MpHMGR boosted sesquiterpene levels the most, with a 2.8-fold average increase for the major sesquiterpenes compared to a 1.8-fold increase with MpFPS and MpDXS2. The overexpression of MpABCG1 alongside all precursor genes led to the most significant increases in endogenous sesquiterpene levels, with a threefold increase in cis-thujopsene, a 3.2-fold increase in β-chamigrene and γ-cuprene, and a 3.6-fold increase in thujopsan-2α-ol (Fig. 4C, Supplementary Data 3). The observed increase in endogenous sesquiterpene levels when MpHMGR was expressed under the MYB2 promoter also suggests that MpHMGR protein is likely localized to OB cells, despite its undetection in localization studies using the mVenus tag. This discrepancy may reflect interference from the C-terminal mVenus fusion affecting proper folding or stability of MpHMGR, whereas the native untagged version used in metabolic engineering was more efficiently expressed and active.

To further validate the transport role of MpABCG1, we generated CRISPR lines to disrupt its function. We obtained six independent transformants with early stop codons in the MpABCG1 DNA sequence (Fig. 5A). Terpene analysis revealed a dramatic reduction or complete absence of sesquiterpenes in Mpabcg1 mutants (Mpabcg1^ge) compared to Cas9-only controls (Cas9^OE) (Fig. 5B, C). Notably, this reduction was specific to sesquiterpenes, as the levels of phytosterols remained unaffected (Fig. 5B, C). Given the apparent differences in peak abundance observed in the overlaid chromatograms (Fig. 5B), we quantified additional compounds alongside neophytadiene and phytol, such as (Z)-1,3-phytadiene and (E)-1,3-phytadiene, phytol derivatives identified in prior studies^78,79 (Fig. 5B). Quantification revealed no significant changes in these compounds when comparing the five Cas9^OE controls and six Mpabcg1^ge lines (Fig. S7).

Fig. 5. Analysis of endogenous terpene profiles in CRISPR-generated Mpabcg1^ge mutants compared to Cas9^OE controls.

A DNA sequence alignment of the Mpabcg1 genomic region in wild-type (WT) and CRISPR-generated Mpabcg1^ge mutant lines. Dashes indicate deletions, red sequences represent insertions, and purple nucleotides correspond to premature stop codons introduced by the deletions and/or insertions. B TICs of terpene extracts from a Cas9^OE control (upper chromatogram) and Mpabcg1^ge mutant line (lower chromatogram). Peaks corresponding to quantified compounds are labeled: sesquiterpenes (1: cis-thujopsene, 2: β-chamigrene, 3: γ-cuprenene, 4a: thujopsan-2α-ol), fatty acids/diterpenoids (5: neophytadiene, 6: phytol, a: (Z)-1,3-phytadiene, b: (E)-1,3-phytadiene), and phytosterols (7: campesterol, 8: stigmasterol). C Quantification of endogenous terpene levels in Mpabcg1^ge (n = 6) compared to Cas9^OE (n = 5). The left graph shows the amounts of the four major sesquiterpenes (1 to 4a) labeled in (B). The right graph displays the levels of phytosterol (7 and 8) labeled in (B). Box plots display individual data points, with bars representing the median and interquartile range. Statistical significances are denoted by asterisks (unpaired t tests or Welch’s tests, depending on variance equality, p < 0.05).

To assess whether the reduced sesquiterpene levels in Mpabcg1^ge could be attributed to changes in OB characteristics, we performed BODIPY staining on three pieces of newly formed meristematic tissues from primary transformants after the second round of selection. Confocal microscopy revealed no detectable differences between Mpabcg1^ge lines and Cas9^OE in OB number (Fig. S8A) or average fluorescence intensity (Fig. S8C, Supplementary data 5). In contrast, OBs in Mpabcg1^ge were significantly smaller in size compared to controls (Fig. S8B, Supplementary Data 5). Since OBs in M. polymorpha also contain abundant non-terpenoid metabolites such as bisbibenzyls, the BODIPY signal may primarily reflect lipidic and aromatic components other than terpenes. The reduction in OB size may therefore reflect impaired terpene accumulation, which restricts OB expansion without affecting overall fluorescence intensity.

These findings suggest that MpABCG1 plays a selective role, directly or indirectly, in promoting the accumulation of endogenous sesquiterpenes, with limited evidence of an effect on other terpene subclasses or on overall OB composition.

Discussion

In this study, we used translational and transcriptional reporters to map both the cell-type specificity and subcellular localization of enzymes involved in the early steps of terpene biosynthesis. These findings corroborate earlier transcriptomic analyses in OB-defective plants^5,7, as well as co-expression network analysis¹⁴ and single-cell RNA sequencing data³⁵, all of which indicate that these enzymes are OB cell-specific.

Enzymatic steps of the MEP pathway and diterpene biosynthesis, including MpDXS, MpHDS, MpHDR, and MpGGPPS, were localized to chloroplasts/plastids—referred to as chloro-amyloplasts in OB cells by Suire et al.³—consistent with their predicted roles in producing isoprenoid precursors for plastid-derived terpenes. In contrast, steps of the MVA pathway and sesquiterpene biosynthesis, represented by MpHMGR and MpFPS, were localized to the ER, Golgi, and cytosol, supporting their involvement in cytosolic terpene precursor production. With the exception of MpFPS, all studied enzymes localized as predicted by DeepLoc 2.1²⁸. However, our experimental conditions allowed us to localize steady-state accumulation of proteins and did not account for any dynamic behavior due to protein trafficking⁸⁰. Our findings corroborate earlier studies³ that reported high or exclusive expression of isoprenoid biosynthetic enzymes in OB cells. Given that OBs are formed through redirection of the secretory pathway⁷, some enzymes like MpHMGR—which localizes to both the ER and Golgi depending on the OB cell or plant being imaged—may transit to other compartments during OB development, supporting the possibility of their transient presence around the OB membrane³. For each translational or expression pattern reporter, we selected promoter lengths based on a single ATAC-seq³⁸ peak, ensuring coverage of the putative core promoter and upstream regulatory regions. In some cases, promoters (including the 5'UTR) were shorter than the typically advised 2.5 kb length^37,41, balancing the need for regulatory coverage with constraints related to sequence synthesis. For instance, a promoter as short as 970 bp like _ProMpGGPPS was sufficient to drive strong and ubiquitous signal in chloro-amyloplasts. However, in other instances, such as _ProMpDXR2, promoter functionality required extending the length from 1.7 kb to 2.7 kb to achieve detectable expression. Altering promoter lengths could include or exclude positive or negative regulatory elements affecting cell-type specificity; therefore, further systematic studies would be needed to draw concrete conclusions about the relationship between promoter structure and regulatory outcomes. Nevertheless, our results confirm the specialized role of OB cells in terpene biosynthesis and highlight the importance of carefully selecting promoters to ensure accurate expression and localization of biosynthetic enzymes.

With respect to precursor supply, our study suggests that the MVA pathway predominantly provides precursors for sesquiterpene synthesis in Marchantia polymorpha, supported by the apparent exclusive expression of MpHMGR and MpMVD in OB cells and the significant increases in endogenous sesquiterpene levels observed when MpHMGR and MpFPS were co-expressed. Notably, MpFPS—also a potential OB-specific enzyme—emerged as a critical limiting step in sesquiterpene biosynthesis, consistent with previous findings in Nicotiana tabacum⁸¹. The MEP pathway’s contribution to sesquiterpene synthesis cannot be ruled out, given the specific expression of MpDXS2 and MpDXR2 in OB cells and the fact that a previous study using GC-MS analysis of physically extracted OB content detected sesquiterpenes² as the only subclass of terpenes. If OBs also store monoterpenes like limonene²⁶ or specific diterpenes, their levels remained undetectable or unchanged despite the targeted expression of precursor supply genes. Furthermore, overexpression of key MEP genes and MpGGPPS had no effect on taxadiene yield, despite co-localization of these enzymes with TXS in plastids. This suggests that a significant portion of the precursors is redirected toward sesquiterpene synthesis instead. Supporting this, co-expression of MpDXS2 with MpFPS, MpHMGR, and MpABCG1 mildly increased the sesquiterpene levels compared to the same combination of genes without MpDXS2, demonstrating a modest effect of MpDXS2 on sesquiterpene yield. Dual contribution from the MVA and MEP pathways in Marchantia polymorpha would parallel findings in Artemisia annua, where inhibitor assays demonstrated a mixed origin of precursors⁸². In contrast, MpDXS1 and MpDXR1 are ubiquitously expressed and likely play broader roles in synthesizing essential metabolites, such as phytyl diphosphate for chlorophyll production. Supporting this, ChIP-seq studies in Marchantia have shown that the chloroplast biogenesis regulator GLK binds to the promoter of MpDXR1⁸³, further highlighting its role in primary metabolism rather than specialized metabolite production. This divergence in expression patterns between MpDXS1 and MpDXS2 is consistent with the functional distinction between Class I and Class II DXS isoforms, typically associated with primary and secondary metabolism, respectively⁸⁴. The presence of such functional specialization in an early-diverging land plant is in line with recent findings in Physcomitrium patens⁶⁷. The specific localization of MpHMGR, MpMVD, and MpFPS protein fusions in OB cells raises intriguing questions about how essential metabolites, such as sterols, are produced in non-OB cells. Despite the ubiquitous expression of squalene synthase MpSQS3, which catalyzes the formation of squalene as a precursor to sterols, the source of precursors for sterol synthesis in non-OB tissues remains unclear. Notably, sterol levels in Mpabcg1^ge were unaffected, suggesting that OBs are not the primary site of sterol synthesis or storage for the rest of the plant. This is further supported by the observation that sterol levels remain unchanged in plants defective in OB cell differentiation⁵. This raises the possibility that other, as-yet-unannotated MVA enzymes may supply precursors for sterol biosynthesis in non-OB cells. Alternatively, it is plausible that the genes studied here are expressed ubiquitously but accumulate at levels too low to be detected in other cell types with our confocal microscopy studies. These findings highlight the complexity of precursor flux and compartmentalization in Marchantia, accentuating the need to further investigate how primary metabolite synthesis is maintained across tissues.

Finally, our findings point to an important role for the ABC transporter MpABCG1 in determining terpene accumulation within oil body cells. Our study demonstrated that exogenous terpenes, such as amorpha-4,11-diene, taxadiene, and β-amyrin, could be produced throughout the whole plant. However, achieving useful yields within OB cells proved less straightforward, even with the overexpression of key biosynthetic genes. This led us to consider whether transport into OBs might be a limiting factor for the accumulation of exogenous compounds. Supporting this, CRISPR-mediated disruption of MpABCG1 resulted in a sharp reduction in endogenous sesquiterpene levels, demonstrating that MpABCG1 plays a crucial role in terpene accumulation. However, whether MpABCG1 directly transports sesquiterpenes or acts indirectly to facilitate their accumulation in the OB lumen remains unclear, and its exact function therefore remains an open question. Some ABCG transporters have been experimentally characterized through transport assays to identify their specific substrates⁸⁵. Among those sharing higher protein sequence similarity with MpABCG1 (~55%), some exhibit broad substrate specificity, such as Nicotiana tabacum NtPDR1, which can accommodate chemically diverse terpenes like the sesquiterpenes sclareol and capsidiol, and the diterpene cembrene⁸⁶. In contrast, others are more selective, such as Arabidopsis thaliana ABCG29, which transports only p-coumaryl alcohol⁸⁷ or Oryza sativa ABCG36, which exports cadmium and cadmium conjugate⁸⁸. This raises the question of whether MpABCG1 could directly transport the 35+ endogenous sesquiterpenes detected by our GC-MS analysis and previous studies²—a number that likely underestimates the total—as these methods primarily detect compounds with limited hydroxylation. An alternative hypothesis is that MpABCG1 may influence the transport of FPP into the OB lumen rather than that of the sesquiterpene end-products. Supporting this, we observed that a translational reporter of the fungal terpene synthase-like 2 (FTPSL2), an endogenous sesquiterpene synthase characterized by Kumar et al.²⁶, was detected within the OB lumen in two independent lines, but only in one lobe out of four in 14-day-old plants (Fig. S9). Because fluorescent protein fusions may occasionally affect trafficking, targeting or organelle morphology^1,89–91, these technical factors cannot be fully excluded as contributors to the restricted MpFTPSL2–mVenus signal. Nevertheless, the observed localization provides preliminary support for OB lumen functioning as a biosynthetic compartment³ rather than solely storage, while the limited localization pattern detected underscores the need for further confirmation. To date, no ABC transporter has been reported to transport phosphorylated isoprenoid intermediates. Indeed, previous research demonstrated that the transport of shorter-chain prenyl diphosphates, such as IPP, occurs across chloroplast membranes via a proton-symport mechanism rather than an ATP-dependent ABC transporter system⁹². However, the OB membrane cannot be directly compared to the chloroplast double membrane, and FPP, being three times larger than IPP, may have different transport requirements. Interestingly, FPP shares amphiphilic properties and a sesquiterpene backbone with abscisic acid (ABA), a known substrate of several ABCG transporters⁸⁵. For example, Triticum aestivum ABCG36⁹³ and Arabidopsis ABCG40⁹⁴, both sharing ~45% protein sequence similarity with MpABCG1, are specific ABA transporters. Moreover, the ability of Arabidopsis ABCC5 to transport inositol hexakisphosphate⁹⁵ suggests that phosphate-containing metabolites can indeed be substrates for ABC-type transporters. If this provisional hypothesis proves correct, it would significantly reshape strategies for engineering terpene biosynthesis in OBs. In this scenario, prenyl phosphate precursors would need to be transported into the OB lumen, and terpene synthases precisely targeted to the lumen or OB membrane to enable localized production of desired compounds. In line with this, we present a tentative map of terpene biosynthesis in Marchantia OB cells (Fig. 6). This map summarizes our findings on the potential accumulation and subcellular localization of key enzymes involved in terpene precursor synthesis, and integrates speculative elements regarding the remaining MVA and MEP pathway steps within OB compartments, as well as the putative role of MpABCG1 (Fig. 6). Overall, our work provides a comprehensive view of terpene synthesis in Marchantia polymorpha OB cells by corroborating predicted and experimentally observed subcellular localization of key isoprenoid biosynthetic enzymes. While the exact role of MpABCG1 remains to be fully elucidated, our results suggest that this transporter is involved in the accumulation of endogenous sesquiterpenes and highlight its potential relevance in metabolic compartmentalization. Our study also clarifies which engineering strategies may be limited in Marchantia, particularly when targeting the production of exogenous terpenes in OBs. Beyond Marchantia, these insights offer a foundation for developing more targeted approaches to achieve efficient terpene biosynthesis in specialized plant compartments.

Fig. 6. Proposed overview of the isoprenoid precursor pathways in Marchantia polymorpha oil body cells based on previous works and findings from this study.

Biosynthetic scheme illustrating the mevalonate (MVA) and methylerythritol phosphate (MEP) pathways in Marchantia polymorpha. Enzymes shown in bold and underlined were localized using translational reporters, confirming their subcellular localization. Enzymes shown in underlined text only were analyzed with promoter expression constructs, indicating activity in oil body cells. Details of enzymatic steps and annotations can be found in Table 1 and Table S1. Abbreviations: Ac-CoA, acetyl coenzyme A; G3P, glyceraldehyde-3-phosphate; CPT, cis-prenyl transferase; MTPSL2, microbial terpene synthase-like 2, functionally characterized in yeast as a limonene synthase (Kumar et al.).

Methods

Plant material and growth conditions

Marchantia polymorpha subspecies ruderalis accessions Cam-1 (male) and Cam-2 (female) were used in this study⁹⁶. Plants were cultivated on solid 0.5× Gamborg B–5 basal medium (#G398, PhytoTech Labs, Lenexa, Kansas, USA) adjusted to pH 5.7–5.8 and solidified with 1.2% (w/v) agar micropropagation grade (#A296, PhytoTech Labs). They were maintained under continuous light at 22 °C with a light intensity of 150 μmol/m²/s. For general propagation of the lines, plants were grown in 94 × 16 mm Petri dishes (#633181, Greiner Bio-One, Kremsmünster, Austria). For imaging purposes, plants were cultivated on gridded 65 × 14.5 mm Gosselin™ Petri dishes (#BB64-01, Corning, Corning, NY, USA). Several gemmae, originating from four independent transformants, were grown on these plates and imaged at developmental stages corresponding to Days 3, 5, 7, and 14 after germination. For gemma cup production and the collection of material for terpene analysis, the media were supplemented with 0.5% (w/v) sucrose. Plants destined for terpene extraction and quantification were grown in 100 × 25 mm Petri dishes (#D943, Phytotech Labs) containing 50 mL of medium. Spore production was carried out in 1L Microbox micropropagation containers (#O119/140 + OD119/140, Sac O₂, Deinze, Belgium), initially maintained at 21 °C under continuous illumination (150 µmol m⁻² s⁻¹), and subsequently transferred after ~1 month to long-day conditions (16 h light/8 h dark) with the same light intensity supplemented with far-red radiation (peak emission ~730–740 nm) provided by Philips GreenPower far-red LEDs, following a previously established protocol⁴¹.

Protein sequence analysis

Protein sequences of the selected MpIDS and MpSQS candidates from Marchantia polymorpha were analyzed using NCBI BLASTP (Basic Local Alignment Search Tool for Proteins, https://blast.ncbi.nlm.nih.gov/Blast.cgi) against the Arabidopsis thaliana protein database. Sequence similarity was assessed based on percentage identity and coverage to infer functional homology. To assess relatedness among ABCG transporters from different plant species, we retrieved the protein sequences of experimentally characterized ABCG with known substrate specificity from Do et al.⁸⁵, when available. These were aligned with the MpABCG1 sequence using the UniProt⁹⁷ online tool to evaluate percentage identity and phylogenetic grouping.

Plasmid construction for overexpression studies

DNA fragments used to generate mVenus and metabolic constructs were synthesized by Genewiz (Azenta, Burlington, Massachusetts, USA) either as linear fragments or in pUAP1 vector⁹⁸. Linear fragments included overhangs compatible with LguI (#ER1931, ThermoFisher Scientific, Waltham, Massachusetts, USA) cloning into L0 acceptor vectors following the Loop protocol⁴¹. Linear fragments and sequences synthesized in pUAP1 vectors, contained overhangs enabling direct cloning into L1 or pBy01³⁷ vectors using BsaI-HF ^®v2 enzyme (#R3733S, New England Biolabs, Ipswitch, MA, USA). MpFPS and MpHDS sequences were cloned from the Marchantia transcriptome as they did not require the removal of BsaI and LguI internal restriction sites, incompatible with the Loop cloning system. Total RNA was first extracted from Marchantia tissue using the RNeasy Plant Mini Kit (#74904, Qiagen, Hilden, Germany) and reverse-transcribed into cDNA with Superscript IV (#18090010, Invitrogen, Waltham, Massachusetts, USA) using random hexamer oligos (#N8080127, Invitrogen) following the protocol from the manufacturer. Target sequences were subsequently amplified from the cDNA using VeriFi® Polymerase Mix (#PB10.43-01, PCR Biosystem, London, UK). PCR products were purified using the QIAquick PCR Purification Kit (#28104, Qiagen) before being used in the cloning steps. DNA sequences exogenous to Marchantia were domesticated when necessary and codon-optimized using the Genewiz codon-optimization tool prior to ordering. NCBI accession numbers for each exogenous gene are as follows: TXS (AY424738), β-AS (MG492000), PpDXS1 (XM_024524883), PpDXS2 (XM_024533934), PpHMGR (XM_024507706) and AMS (AY006482). Certain sequences were re-amplified to remove stop codons and introduce the appropriate overhangs for fusion with mVenus or eGFP tags. Primers used in this study were ordered from IDT (Integrated DNA Technology, Coralville, Iowa, USA) (Table S4). Vectors, promoters, terminators, tags, and pre-assembled cassettes used to generate the constructs were sourced from the OpenPlant toolkit⁴¹ or adapted from Romani et al.³⁷. Final plasmids (L2, L3, or pBy01) were sequenced at Plasmidsaurus (Eugene, Oregon, USA) to confirm proper assembly and integration of all transcription units. Details of the sequences and vectors syntax are available in Supplementary data 6.

In vitro testing of guide RNA and generation of CRISPR-based mutants

In vitro testing of candidate gRNAs was performed prior to plant transformation, following a protocol adapted from an open-access method on protocols.io⁹⁹, with modifications described below. Genomic DNA was extracted from Marchantia polymorpha using the InnuPrep Plant DNA Kit (#845-KS-1060050, IST Innuscreen, Berlin, Germany). A genomic region of 860 bp containing the six candidate gRNA target sites was amplified using primers spanning from the 5′UTR to the first coding exon of the ABCG1 sequence. For sequences of the six gRNAs and their target positions, see the table in Fig. S10A. To amplify the 860 bp genomic sequence, the forward primer AAGACTCCGGATCCGAGGG and reverse primer CCTCAACTTTGGGCAGGTCA were used. PCR amplification was performed using VeriFi® Polymerase Mix, and products were purified using the QIAquick PCR Purification Kit. For each of the six candidate gRNAs designed with the CasFinder tool¹⁰⁰, a DNA template for in vitro transcription was generated by PCR. Forward primers consisted of the T7 promoter sequence, followed by the 20-nt gRNA target sequence, and the first portion of the trans-activating CRISPR RNA (tracrRNA), in the following format: GAAATTAATACGACTCACTATAGG<gRNA sequence>GTTTTAGAGCTAGAAATAGC. The reverse primer AAAAGCACCGACTCGGTGCCAC corresponded to the remaining portion of the tracrRNA scaffold. PCR templates were amplified using the single-guide acceptor CRISPR vector from Sauret-Guëto et al.²³ containing the gRNA scaffold. PCR products were purified using the QIAquick PCR Purification Kit and quantified prior to in vitro transcription. PCR templates for each of the six candidate gRNAs were transcribed into RNA using the MEGAshortscript™ T7 Transcription Kit (#AM1354, ThermoFisher Scientific) following the manufacturer’s protocol, including treatment with TURBO DNase. RNA was precipitated by adding 15 µL RNase-free water and 15 µL LiCl Precipitation Solution (#10498254, Fisher Scientific) and processed according to the manufacturer’s instructions. In vitro cleavage reactions were performed using the Cas9 Nuclease (#M0646T, New England Biolabs) following the supplier’s protocol, with 300 nM in vitro–transcribed sgRNA and 200–300 ng of the genomic 860 bp amplicon as substrate. Reactions were incubated at 37 °C for 1 h, then treated with 1 µL Proteinase K (#11588916, Fisher Scientific) at 65 °C for 10 min. Cleavage products were resolved immediately on an agarose gel alongside a 100 bp DNA ladder (#N3231S, New England Biolabs). Among the six tested guides (Fig. S10A), gRNA #1 showed a slightly reduced intensity of the full-length 860 bp band (Fig. S10B), suggesting possible in vitro activity, whereas gRNA #5 produced a clear cleavage pattern with two distinct fragments and a reduced intensity of the 860 bp band. Both gRNAs #1 and #5 were cloned into the single-guide acceptor vector following Sauret-Guëto et al. protocol²³. Numerous mutants were obtained only with gRNA #5, while no mutants were recovered for gRNA #1.

Agrobacterium-mediated transformation of Marchantia spores and selection of independent transformants

Marchantia spores were sterilized and transformed following previously described protocols, with slight modifications. Briefly, a single archegoniophore, dried and stored in silica beads, was mixed in 1.5 mL of a chlorinated solution consisting of one Milton sterilizing tablet (troclosene sodium; Boots, UK) dissolved in 10 mL of sterile water. Spores were released into the solution, filtered through a 40 µm cell strainer (#542040, Greiner Bio-One) and incubated in the sterilizing solution for 30 min. After sterilization, spores were harvested by centrifugation, resuspended in sterile water, and spread on Gamborg medium agar plates (without sucrose). After 5 days of germination, the sporelings were co-cultured with Agrobacterium tumefaciens GV3101 carrying the constructs of interest. Transformation of Agrobacterium with the constructs was performed using the freeze–thaw method¹⁰¹. Sporelings and Agrobacterium were co-cultured for 2 days in a 4 mL solution of 0.5× Gamborg B–5 plus supplements, as previously described⁴¹, and incubated at 22 °C under continuous shaking and light. Following co-culture, sporelings were collected on 70 µm cell strainers (#542070, Greiner Bio-One), rinsed thoroughly with sterile water, and plated on 90 mm Petri dishes containing 0.5× Gamborg B5 medium supplemented with 100 µg/mL cefotaxime (#BIC0111, Apollo Scientific, Bredbury, UK) to eliminate Agrobacterium and 20 µg/mL hygromycin (#10687010, Invitrogen) to select for transformants. After two weeks on the first selective plates, eight transformants were typically transferred to a second plate containing the same selection agents to confirm successful transformation and eliminate residual Agrobacterium. For lines expressing fluorescent proteins (mScarlet, mVenus or eGFP), transformants were pre-selected under a Leica stereo microscope (#MDG41, Leica, Wetzlar, Germany) using suitable filter sets to confirm the presence of fluorescence signals. For lines expressing exogenous terpenes, eight plants were randomly chosen, grown for one month on the second selective plates, and subsequently extracted to confirm the presence of the terpene of interest. The four best lines producing the desired terpene were then transferred to sucrose-supplemented plates to quantify endogenous and exogenous terpene levels. For lines co-expressing AMS and precursor supply genes, a single vector carrying up to six transcription units could not be constructed due to cloning or vector size limitations. Attempts to co-transform using two Agrobacterium strains carrying distinct plasmids—one with hygromycin B as the selection agent and the other with chlorsulfuron—resulted in an insufficient number of transformants. Therefore, transformants were selected based on hygromycin B resistance for AMS (with or without MpABCG1) on one vector, and the presence of an mScarlet-Lti6b fluorescent signal from the second vector carrying MpFPS, MpHMGR, and/or MpDXS2. All eight independent lines were retained to measure amorphadiene and endogenous terpene levels. To genotype CRISPR-generated Mpabcg1^ge mutants and Cas9^OE lines, genomic DNA was extracted using the InnuPrep Plant DNA kit. For Mpabcg1^ge lines, a 500 bp DNA fragment spanning the gRNA target site was amplified using Q5® High-Fidelity DNA Polymerase (#M0491S, New England Biolabs). The resulting PCR products were purified and sequenced by Genewiz to confirm the disruption or not of the MpABCG1 coding sequence. The presence of the Cas9 cassette in control lines was confirmed by amplification with specific primers (listed in Table S4) and subsequent sequencing. A total of 16 lines were genotyped for the Mpabcg1^ge ones, and eight for the Cas9^OE controls.

Laser scanning confocal microscopy

For transgenic lines expressing fluorescent proteins, images were acquired on an upright Leica SP8X confocal microscope equipped with a 460–670 nm supercontinuum white light laser, two continuous wavelength laser lines of 405 nm and 442 nm and a five-channel spectral scanhead (four hybrid detectors and one photomultiplier). Imaging was conducted using either a 25× water immersion objective (Fluotar VISIR 25×/0.95 WATER) for whole gemmae or meristematic area imaging, or a ×40 water immersion objective (HC PL APO CS2 ×40/1.10 WATER) for higher magnification images, with an additional digital zoom applied up to a factor of 5× to enhance visualization of subcellular localizations. Excitation laser wavelength and fluorescence emission bandwidth windows were as follows: 515 nm and 525–550 nm (for mVenus); 570 nm and 591–621 nm (for mScarlet); 442 nm and 645–664 nm (for chlorophyll autofluorescence). Each channel was acquired separately using a hybrid detector, with a sequential scan for each channel performed on the Leica LAS X software. To image the mVenus, eGFP and mScarlet fluorescences on the translational reporter lines carrying the bidirectional construct of MpHDS-eGFP and MpHDR-mVenus, excitation and emission bandwidth windows were adjusted as follows to prevent signal crosstalk: 487 nm and 497–513 nm (for eGFP); 521 nm and 530–550 nm (for mVenus); 570 nm and 588–598 nm (for mScarlet). Day 0 gemmae were imaged by mounting them on a glass slide with perfluorodecalin¹⁰² (#130040250, ThermoFisher) and covering them with a glass coverslip. For plants aged 3–14 days, a 15 × 16 mm gene frame (#AB0577, ThermoFisher) was placed on the glass slide to prevent compression of the plant material under the coverslip. Whole gemmae and overviews of the meristematic region in older plants were imaged using Z-stack scans, which were processed in Fiji¹⁰³ to generate maximum intensity projections of the Z-stacks. High magnification images were acquired as single scans to better display subcellular features.

BODIPY staining and oil body analysis

Meristematic tissue pieces were collected from 2–3-week-old plants and stained with BODIPY™ 493/503 (#11540326, Fisher Scientific). A 5 mM stock solution was prepared in DMSO and diluted to a 2 µM working solution in water. Tissues were incubated in this solution for 30 min in the dark, rinsed twice with water, and gently blotted dry on absorbent paper before mounting in perfluorodecalin, as described above for confocal imaging of plants aged 3–14 days. BODIPY fluorescence was detected using 488 nm excitation and a narrow emission window of 500–505 nm, alongside chlorophyll autofluorescence recorded under the settings described above.

Z-stacks were processed in Fiji. For OB size measurements, the BODIPY fluorescence channel was converted to grayscale, thresholded (values 0–20), and binarized with watershed separation. OBs were then quantified using the “Analyze Particles” function with a minimum size cutoff of 3 µm². For fluorescence intensity measurements, regions of interest (ROIs) corresponding to OBs were generated from the binary image and transferred to the original grayscale image. The mean gray intensity and area of each ROI were then extracted using the “Set Measurements” and “Measure” functions. Data were compiled and plotted in PRISM software version 10 (GraphPad, La Jolla, California, USA).

Extraction of terpenes and analysis by gas chromatography-mass spectrometry (GC-MS)

Approximately 200 mg of frozen tissue from 1-month-old Marchantia polymorpha plants was extracted with 1 mL of cold methanol containing 5 mM NaCl to quench enzymatic activity, following the method of Kumar et al.²⁶. To enable quantification, 5 µg of dodecane (#297879, Sigma-Aldrich) was included as an internal standard. The tissue was disrupted in the solvent using a 3-mm stainless steel ball (#2205, Durston, High Wycombe, UK) and homogenized with a TissueLyser II (Qiagen). Samples were subsequently agitated on a Vibrax® shaker (#0002819002, IKA, Staufen, Germany) at 2000 rpm for 2 hours, and the resulting extracts were centrifuged to remove plant debris. The methanolic phase was then extracted once with hexane to isolate non-polar and medium-polar terpenes from the upper organic layer. Hexane extracts (200 µL) were analyzed on a Trace 1300 gas chromatograph (ThermoFisher) coupled to an ISQ 700 mass spectrometer (ThermoFisher) with a Zebron CD-5MS column (30 m × 0.25 mm × 0.25 μm; ThermoFisher). A splitless injection (1 µL) was conducted at 230 °C, and the GC-MS oven parameters were adapted from Kumar’s work²⁶ as follows: the oven temperature was initially set to 70 °C and held for 3 min, followed by a ramp of 20 °C/min to 90 °C, a second ramp of 3 °C/min to 180 °C, a third ramp of 5 °C/min to 240 °C, and a final ramp of 20 °C/min to 300 °C, with a 6-min hold at this temperature. The MS began data acquisition after a 5.5-minute solvent delay, with the transfer line and ion source temperatures set at 250 °C and 270 °C, respectively. Scanning was conducted in full-scan mode (scan time: 0.17 s) over a mass range of 40–600 atomic mass units. Helium was used as the carrier gas, with a flow rate of 1.2 mL/min for sesquiterpenes in the taxadiene and β-amyrin datasets. For amorphadiene detection, the carrier flow rate was reduced to 0.9 mL/min. Chromatograms were processed using Chromeleon™ software (ThermoFisher), and terpene quantification was performed relative to the internal standard dodecane. Tentative identification of the major sesquiterpenes was achieved by comparing mass spectra to published data and by calculating retention indices using a single sample run with a C8–C40 alkane standard mixture (#40147-U, Superlco, Bellefonte, Pennsylvania, USA) as described by Adams⁷⁴. Note that the endogenous terpene composition in CAM accessions may slightly differ from that reported for other accessions, such as Tak, in previous studies. The authentic standard of amorpha-4,11-diene was provided by Tomasz Czechowski and Ian Graham (University of York). To achieve more accurate quantification of β-amyrin and sterols, the extraction protocol and GC-MS method were modified. The same amount of frozen material was extracted with cold methanol supplemented with 10 µg of coprostan-3-ol (#C7578, Sigma-Aldrich) as an internal standard. The samples were extracted twice with hexane to obtain a total hexane volume of 1.5 mL, which was then dried under nitrogen flow in a Genevac™ concentrator EZ-2 (#EZ3P-23050-NN0, Genevac Ltd, Ipswich, UK). Extracts were derivatized with 50 µL 1-trimethylsilyl-imidazole (TMS; #92718, Superlco) and resuspended in 300 µL hexane. For GC-MS analysis, 200 µL of the derivatized samples was transferred into vials for direct injection. The method used for sterol/triterpene analysis was as follows: 130 °C hold for 2 min; ramp of 30 °C/min until 220 °C, then 2 °C/min until 300 °C, hold for 10 min. Solvent delay of 1 min before acquiring the MS data. The authentic standard of β-amyrin was provided by James Reed and Anne Osbourne (John Innes Center, Norwich, UK). For terpene quantification in Mpabcg1^ge and Cas9^OE lines, extractions were performed on genotyped plants from the second selection stage, as loss of the CRISPR genotype was observed during propagation from gemmae. Given this observation, we considered the possibility of chimerism^104,105 and extracted as much biomass as was reasonably available from individual plants. This approach aimed to minimize variability, homogenize the material, and ensure accurate results. Box plots showing individual data points for each quantified terpene were created using PRISM software version 10 (GraphPad, La Jolla, California, USA), and the results of statistical comparisons were integrated into the visualizations. For lines expressing exogenous terpene synthases, statistical analyses were performed using one-way analysis of variance (ANOVA) to assess differences across gene combinations. Homogeneity of variances was evaluated using the Brown-Forsythe and Bartlett's tests, with Dunnett's multiple comparison test applied post-hoc to compare each gene combination to the respective control (_ProMYB2:β-AS, TXS, or AMS). The significance threshold (alpha) was set at 0.05. Four biological replicates expressing the constructs were analyzed to quantify both exogenous (β-amyrin and taxadiene) and endogenous compounds (n = 4), while eight biological replicates expressing the constructs were analyzed for amorphadiene lines, focusing on the quantification of endogenous terpenes since amorphadiene was not detected (n = 8). For comparisons of terpene levels between abcg1 mutants (n = 6) and Cas9 controls (n = 5), Welch’s t test was used when variances were unequal, and unpaired t test was used when variances were equal, as determined by PRISM. Both tests accounted for differences in sample size between the two groups. Asterisks were added to the box plots to indicate statistically significant differences (P < 0.05) between Mpabcg1^ge mutants and Cas9^OE lines, as determined by Welch’s or unpaired t tests.

Statistics and reproducibility

Statistical analyses were performed on data obtained from independent biological replicates, defined as independently generated transgenic lines or independently genotyped mutant plants. For experiments involving metabolite quantification, four to eight independent primary transformants were analyzed per construct, as indicated in the corresponding figure legends. For CRISPR-generated mutants, all available genotyped transformants were included. Statistical analyses were performed using GraphPad Prism. Where applicable, statistical significance was assessed using one-way ANOVA followed by Dunnett’s multiple comparisons test, unpaired two-tailed Student’s t tests, or Welch’s t tests, as specified in the figure legends. Quantitative data are presented as box plots showing individual data points.

For imaging experiments lacking formal statistical analysis, transformations were performed in independent rounds and multiple plants were analyzed per construct to confirm reproducibility of localization patterns.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.C.F.F. designed the work. F.R. provided DNA parts and early insights. E.C.F.F., P.A., I.B., and E.F. carried out the work. J.H. acquired funding and provided overall project supervision. E.C.F.F. wrote the manuscript with input from all authors. All authors approved the final version of the manuscript.

Peer review

Peer review information

Communications Biology thanks Hiroyoshi Kubo, Matthew E. Bergman, Xiaodong Wang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: George Inglis and David Favero.

Data availability

Source data for terpene quantification and oil body measurements using BODIPY staining are provided as Supplementary Data 1–5. Sequences and vectors used to generate the constructs for plant transformation are provided in Supplementary Data 6. Original Leica.lif files containing confocal microscopy images are available from the corresponding authors upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-09508-4.