Plant secondary metabolites, also known as plant specialized metabolites, are often specific to certain lineages, making their biosynthesis pathways highly evolvable compared to those of primary metabolites. Nevertheless, our understanding of the evolution of these biochemicals is still limited (Weng et al., 2012). One such secondary metabolite is artemisinin, an endoperoxide sesquiterpene lactone that is effective against chloroquine-resistant strains of Plasmodium falciparum. Artemisinin is exclusively produced by Artemisia annua L. (Asteraceae) (Chen and Xu, 2016; Ikram and Simonsen, 2017; Kung et al., 2017; Wang et al., 2019). On the other hand, a group of structurally diverse sesquiterpene lactones, including guaianolides, eudesmanolides, and germacranolides, which feature an α-methylene γ-lactone moiety, are produced by other plants in the Asteraceae family. These sesquiterpene lactones are collectively known as Asteraceae sesquiterpene lactone (ASTL) (de Kraker et al., 2002). It has been suggested that the biosynthesis of artemisinin is derived from the ASTL pathway (Nguyen et al., 2010, 2019), providing an opportunity to study the dynamics of metabolic pathway evolution.

The biosynthetic pathways of artemisinin begin with amorpha-4,11-diene synthase (ADS), which converts farnesyl diphosphate (FPP) into the amorphadiene skeleton (Bouwmeester et al., 1999; Chang et al., 2000; Mercke et al., 2000; Huang and Fang, 2021). Subsequently, a cytochrome P450-dependent monooxygenase known as CYP71AV1 performs a three-step oxidation of amorpha-4,11-diene, resulting in the generation of minor products of artemisinic alcohol and artemisinic aldehyde and a major product of artemisinic acid in yeast (Ro et al., 2006). However, in an in vitro enzymatic assay using microsomes containing CYP71AV1 and the substrate amorpha-4,11-diene, only a major product of artemisinic alcohol was found, as well as a minor product of artemisinic aldehyde sometimes (Teoh et al., 2006). This enzyme is referred to as amorpha-4,11-diene oxidase (AMO) ( Figure 1 ). Notably, there are two additional enzymes also involved in these oxidations in A. annua, namely, an alcohol dehydrogenase (ADH1) that catalyzes the conversion of artemisinic alcohol to artemisinic aldehyde and an aldehyde dehydrogenase (ALDH1) that transforms artemisinic aldehyde to artemisinic acid (Liao et al., 2022). Since the Δ11(13) double bond in the above precursors is replaced by a methyl group in artemisinin, artemisinic aldehyde Δ11(13) reductase (DBR2) catalyzes the reduction of artemisinic aldehyde to dihydroartemisinic aldehyde, which is subsequently oxidized to dihydroartemisinic acid by ALDH1 (Liao et al., 2022). Finally, a light-dependent non-enzymatic spontaneous autoxidation converts dihydroartemisinic acid to artemisinin (Czechowski et al., 2018).

The biosynthesis of ASTL shares a conserved core pathway involving germacrene A synthase (GAS) and germacrene A oxidase (GAO), both of which collectively transform FPP to germacrene A acid, a key intermediate of ASTL biosynthesis. GAO catalyzes a similar three-step oxidation reaction to convert germacrene A to a single product, germacrene A acid, in yeast ( Figure 1 ) (Bennett et al., 2002; Nguyen et al., 2010; Cankar et al., 2011; Ramirez et al., 2013; Eljounaidi et al., 2014; Nguyen et al., 2019). Interestingly, the GAOs from other plants of Asteraceae exhibit high homology with AMO, and their corresponding proteins can utilize amorpha-4,11-diene as a substrate to produce artemisinic acid as a single product in yeast. Conversely, the AMO protein is inactive toward germacrene A (Nguyen et al., 2010; Cankar et al., 2011; Nguyen et al., 2019). Based on the fact that the activity of ADS is exclusively found in A. annua, a prevalent theory postulates that AMO evolved from GAO, gradually acquiring specificity for amorpha-4,11-diene in A. annua (Nguyen et al., 2019). This theory provides an explanation for the promiscuous activity of GAO toward amorpha-4,11-diene and the loss of ancestral activity on germacrene A during the specialization of AMO. Consequently, the absence of a GAO copy in the genome of A. annua resulted in the elimination of germacrene A acid and ASTL production, despite the presence of germacrene A and functional GAS expression in glandular secreting trichomes (GSTs) (Bertea et al., 2006) where artemisinin is synthesized and accumulated.

In the A. annua genome, two copies of CYP71AV1 can be found. One is a long version with a seven-amino acid extension at the N-terminus of the protein sequences, which was initially isolated and highly expressed in a low-artemisinin production (LAP) chemotype. This long version is referred to as AMOLAP (GenBank: PWA40082.1). The other copy is a shorter version that was primarily cloned and dominantly expressed in a high-artemisinin production (HAP) chemotype, which is named AMOHAP (GenBank: PWA47004.1) (Ting et al., 2013; Shen et al., 2018; Liao et al., 2022). When expressed in Nicotiana benthamiana, AMOLAP demonstrated higher enzymatic activity compared to AMOHAP in amorpha-4,11-diene (Ting et al., 2013). This suggests that the differential high expression of these two variants in LAP and HAP strains contributes to the different metabolomic profiles of the two chemotypes (Czechowski et al., 2018). However, in another high-artemisinin content chemotype (HAN1), the transcript level of AMOLAP was higher than that of AMOHAP, which contradicts the previous findings (Liao et al., 2022). It is important to note that, in the above experiment, the AMOLAP form of GAO was selected to examine its cross-activity on germacrene A (Nguyen et al., 2010). However, the activity of AMOHAP on germacrene A was not evaluated. Therefore, the real function of AMOHAP remains to be elucidated.

In this study, we demonstrate that AMOHAP possesses the ability to catalyze the conversion of germacrene A into germacrene A alcohol, providing evidence that AMOHAP indeed functions as a GAO rather than an AMO. Furthermore, the promoter regions of AMOLAP and AMOHAP exhibit distinct cis-elements, indicating divergent transcriptional regulation between these two genes. Therefore, these findings strongly imply that the inability of A. annua to produce ASTL is primarily attributed to the altered transcriptional regulation of GAO rather than the loss of GAO function.

Costunolide, the most basic sesquiterpene lactone found in Asteraceae, is widely distributed among various species within this family. The biosynthesis of costunolide starts from FPP and is subsequently catalyzed by GAS and GAO (Nguyen et al., 2010). On the other hand, A. annua does not produce costunolide, but contains artemisinin, which is synthesized through the catalytic activities of ADS and AMO ( Figure 1 ). It is worth mentioning that GAO exhibits promiscuous activity as it can utilize amorpha-4,11-diene as a substrate to generate artemisinic alcohol, artemisinic aldehyde, and artemisinic acid in yeast. Conversely, AMO shows no activity toward germacrene A, suggesting its functional specialization and loss of ancestral activity (Nguyen et al., 2010; Huang et al., 2021). Previous studies have proposed the existence of two copies of the CYP71AV1 gene (i.e., AMOHAP and AMOLAP) in the genome of A. annua. Both copies have the ability to produce artemisinic alcohol, artemisinic aldehyde, and artemisinic acid when transiently expressed in N. benthamiana (Ting et al., 2013; Shen et al., 2018; Liao et al., 2022). However, the enzymatic activity of AMOHAP on germacrene A has not been analyzed to date. To elucidate the true function of AMOHAP, we conducted an in vitro enzymatic assay using both AMOHAP and AMOLAP on germacrene A. Our findings revealed that only AMOHAP, but not AMOLAP, was capable of producing germacrene A alcohol. This result strongly suggests that AMOHAP is indeed a GAO rather than an AMO, and its activity toward amorpha-4,11-diene is a promiscuous activity shared by other GAO enzymes. Therefore, we propose renaming AMOHAP as A. annua GAO to accurately reflect its function.

The emergence of a new metabolic pathway can sometimes result in the loss of the ancestral pathway, as is evident in the case of the artemisinin/costunolide pathways in A. annua. Germacrene A and functional GAS were found to be present in A. annua, with the expression of GAS being particularly enriched in the trichome, bud, and young leaf, which is consistent with the expression pattern of the artemisinin biosynthetic genes. This observation further supports the idea that costunolide has been replaced by artemisinin in A. annua. Therefore, the absence of costunolide biosynthesis in A. annua may be attributed to the lack of GAO activity, which is a result of the loss of the GAO copy proposed in a previous study (Nguyen et al., 2019). Our findings provide direct evidence that a change in regulation, specifically the reduction in GAO expression, leads to the abolishment of costunolide biosynthesis. Importantly, a natural loss-of-function event is an essential part of the evolutionary landscape; however, previous investigations have primarily focused on the loss of protein-coding genes (Olson, 1999; Sharma et al., 2018; Hecker et al., 2019; Xu and Guo, 2020). Our study highlights that variations in the non-coding upstream region can also result in a reduction of gene expression and loss of function.

Clearly, the real innovation in the emergence of the artemisinin pathway is the advent of ADS activity, which is exclusively found in A. annua. Interestingly, the production of amorpha-4,11-diene is a relatively simple innovation. Previous studies conducted by ourselves and others have demonstrated that a single residue switch in α-bisabolol synthase (Fang et al., 2017) and (E)-β-farnesene synthase (Salmon et al., 2015) from A. annua can generate amorphadiene analogs, such as amorpha-4,7(11)-diene, which only differs from amorpha-4,11-diene in terms of the double bond position. Therefore, the potential ability of sesquiterpene synthase to produce amorphadiene and the promiscuous activity of GAO toward amorphadiene serve as the evolutionary start points for the advent of the artemisinin metabolic pathway in A. annua.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

XF: Writing – original draft, Writing – review & editing, Conceptualization, Data curation, Funding acquisition, Project administration, Supervision. F-YC: Data curation, Investigation, Writing – original draft, Writing – review & editing. Q-YM: Data curation, Investigation, Writing – review & editing. B-YX: Investigation, Writing – review & editing. Y-CL: Investigation, Writing – review & editing. H-YL: Investigation, Writing – review & editing.