# Source paper archive

- Source title: Discovery of the β-Carboline Biosynthetic Pathway Enables In Planta Reconstruction of the Amazonian Brew Ayahuasca
- Author(s): Hao, L. K., Cohen, S., Hofer, J., Tuffa, B. A., Khersonsky, O., Kumari, A., Almekias-Siegl, E., Faigenboim, A., Dong, Y., Heinig, U., Meir, S., Rogachev, I., Maoz, I., Fleishman, S.-J., Aharoni, A., Berman, P.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766139v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766139v1.full.pdf
- PDF SHA-256: 4438c9859466f9d66c18a087bcade568dfa62e6cd20227e97154eefb89b7dc1b
- Converter: pymupdf4llm 0.0.27
- Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version.

---

&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

# 1 Discovery of the β-Carboline Biosynthetic Pathway Enables In Planta 2 Reconstruction of the Amazonian Brew Ayahuasca


3


4 Let Kho Hao &#91;1,2&#93;, Shahar Cohen &#91;1&#93;, Janka Höfer &#91;2&#93;, Bekele Abebie Tuffa &#91;2&#93;, Olga Khersonsky &#91;3&#93;, Annu


5 Kumari &#91;4&#93;, Efrat Almekias-Siegl &#91;2&#93;, Adi Faigenboim &#91;5&#93;, Younghui Dong &#91;6&#93; -, Uwe Heinig &#91;6&#93;, Sagit


6 Meir &#91;2&#93;, Ilana Rogachev &#91;2&#93;, Itay Maoz &#91;4&#93;, Sarel-Jacob Fleishman &#91;3&#93;, Asaph Aharoni &#91;2&#93; *, and Paula


7 Berman &#91;1&#93; 

8


9 1 Department of Vegetable Field Crops, Institute of Plant Sciences, Agricultural Research


10 Organization (ARO), Volcani Institute, Rishon LeZion 7505101, Israel.


11 2 Department of Plant and Environmental Sciences, Weizmann Institute of Science, Rehovot


12 7610001, Israel.


13 3 Department of Biomolecular Sciences, Weizmann Institute of Science, Rehovot 7610001,


14 Israel.


15 4 Department of Postharvest Science, Institute of Postharvest and Food Sciences, Agricultural


16 Research Organization (ARO), Volcani Institute, Rishon LeZion 7505101, Israel.


17 5 Bioinformatics Unit, Institute of Plant Sciences, Agricultural Research Organization (ARO),


18 Volcani Institute, Rishon LeZion 7505101, Israel.


19 6 Life Sciences Core Facilities, Weizmann Institute of Science, Rehovot 7610001, Israel.


20 - Current lab address: Haixia Institute of Science and Technology, Fujian Agriculture and


21 Forestry University, Fuzhou 350002, China.


22 &#91;*Corresponding authors. Email: paulab@volcani.agri.gov.il;&#93;(mailto:paulab@volcani.agri.gov.il) &#91;asaph.aharoni@weizmann.ac.il.&#93;(mailto:asaph.aharoni@weizmann.ac.il)


23


24 **Abstract**


25 β-Carbolines are bioactive indole alkaloids with remarkable pharmacological properties and a


26 defining role in ayahuasca, a traditional Amazonian shamanic brew, where their inhibition of


27 monoamine oxidases enables the oral activity of the psychoactive _N,N_ -dimethyltryptamine


28 (DMT). Yet, despite their longstanding medicinal and cultural significance, the biosynthetic


29 origin of β-carbolines in plants remains largely unresolved. Here, we uncover key steps


1


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


30 underlying the biosynthesis of harmine, the major bioactive β-carboline alkaloid in _Peganum_


31 _harmala_, by integrating spatial metabolite and transcriptome profiling with enzyme


32 characterization and pathway reconstruction. We discovered four enzymes that enable _de novo_


33 harmine biosynthesis in _Nicotiana benthamiana_ and revealed how two closely related


34 cytochrome P450s exhibit divergent regioselectivity, directing tryptamine towards either β

35 carboline or serotonin biosynthesis. Structural modeling and targeted mutagenesis pinpoint


36 active-site residues that switch this regioselectivity, revealing a molecular basis for


37 diversification of tryptamine metabolism. By combining distinct tryptamine pathways, we


38 achieved _de novo_ co-production of β-carbolines and DMT in _N. benthamiana_ and _P. harmala_


39 hairy roots without supplementing exogenous precursors. Together, these findings unlock β

40 carboline biosynthesis for future sustainable bioproduction with broad opportunities spanning


41 medicine, crop protection, and biodiversity conservation.


42 **Keywords** : Harmine; _Peganum harmala_ ; biosynthetic pathway; ayahuasca; hairy root


43 transformation; molecular modeling.


44

# 45 Introduction


46 β-Carboline alkaloids are a diverse class of tryptophan-derived specialized metabolites with a


47 broad range of pharmacological activities, including antidepressant, anti-inflammatory,


48 neuroprotective, antidiabetic and antitumor effects (de Osório et al., 2015; Zhang et al., 2020;


49 Aaghaz et al., 2021). In plants, β-carbolines have been implicated in ecological interactions


50 across trophic levels, including anti-herbivore and antimicrobial activities and phytotoxic


51 effects on other plants, such as inhibition of root growth and disruption of cell division (Cavin


52 and Rodriguez, 1988; Bouayad et al., 2012; Shao et al., 2013; Cui et al., 2020; Diehl et al.,


53 2026). They are found across multiple plant lineages and occur in distantly related plant


54 families, raising the possibility that distinct biosynthetic pathways may have evolved to


55 produce structurally related alkaloids (Patel et al., 2012; Dai et al., 2018). Among β-carboline

56 producing plants, _Peganum harmala_ ( _P. harmala_, Nitrariaceae) is native to arid and semi-arid


57 regions of the Mediterranean, Middle East, and Central Asia (Sharifi-Rad et al., 2021). It has a


58 long history of medicinal and ritual use, which has attracted considerable interest for its diverse


59 pharmacological and psychoactive properties (Moloudizargari et al., 2013).


60 Harmine and harmaline were among the earliest characterized β-carboline (harmala) alkaloids,


61 having been isolated from _P. harmala_ seeds in the nineteenth century (Fritzsche, 1848).


2


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


62 Subsequent phytochemical studies identified additional β-carbolines in _P. harmala_, with


63 harmalol, tetrahydroharmol (THH-OH), and tetrahydroharmine (THH) detected predominantly


64 in seeds, and harmol particularly abundant in roots (Herraiz et al., 2010; Fang et al., 2019).


65 These alkaloids are also central to ayahuasca, a traditional Amazonian psychoactive brew, in


66 which β-carboline-rich _Banisteriopsis caapi_ ( _B. caapi_, Malpighiaceae) is combined with _N_, _N_ 

67 dimethyltryptamine (DMT)-producing _Psychotria viridis_ (Santos et al., 2020). In this


68 preparation, β-carboline-mediated inhibition of monoamine oxidase A (MAO-A) enables the


69 oral bioavailability of DMT. _Peganum harmala_ has likewise been used as a β-carboline source


70 in ayahuasca analogue preparations (Kaasik et al., 2021). Growing interest in the therapeutic


71 potential of psychedelics has stimulated clinical investigations of ayahuasca, particularly for


72 treatment-resistant depression and other psychiatric disorders (Palhano-Fontes et al., 2019; Dos


73 Santos and Hallak, 2024). Despite this growing interest, the biosynthetic pathways of β

74 carbolines in plants remain poorly characterized, even after decades of biochemical


75 investigations.


76 Early isotope-labeling studies investigating harmine **1** biosynthesis in _P. harmala_ identified L

77 tryptophan (Trp **2** ) and tryptamine (TAM **3** ) as precursors and demonstrated the incorporation


78 of pyruvate **4** into its β-carboline skeleton (Stolle and Gröger, 1968; **Supplemental Figure**


79 **S1A** ). Subsequent feeding experiments in _P. harmala_ callus cultures demonstrated the


80 oxidation of harmaline **5** to harmine **1**, but failed to detect incorporation of 6

81 hydroxytryptamine (6-HT **6** ), a proposed precursor of harmalol **7**, into β-carboline alkaloids


82 (Nettleship and Slaytor, 1974). Nevertheless, the authors proposed a biosynthetic route


83 involving 6-HT **6**, harmalol **7**, and harmaline **5** as successive intermediates leading to harmine


84 **1**, with a possible alternative route through harmol **8** and its methylation to harmine **1** . They


85 also proposed the biosynthesis of serotonin (5-hydroxytryptamine, 5-HT **9** ) from Trp **2** _via_


86 TAM **3** to explain its accumulation in _P. harmala_ (Nettleship and Slaytor, 1974; **Supplemental**


87 **Figure S1B** ). More recently, Jamieson et al. (2021) proposed that the β-carboline scaffold


88 results from Pictet-Spengler condensation of TAM **3** with pyruvate **4**, yielding 1,2,3,4

89 tetrahydro-β-carboline-1-carboxylic acid (THβC-1-COOH **10** ), followed by oxidative


90 decarboxylation to 1-methyl-β-carboline **11 (Supplemental Figure S1C** ). Similar Pictet

91 Spengler reactions could also proceed with 6-HT **6** instead of TAM **3**, or acetaldehyde **12** could


92 replace pyruvate **4** as the carbonyl donor, bypassing the carboxylated tetrahydro-β-carboline


93 intermediate (Callaway et al., 1994; **Supplemental Figure S1D** ).


3


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


94 In this study, we elucidated key steps in the biosynthesis of harmine **1** and related β-carbolines


95 from Trp **2** in _P. harmala_, addressing unresolved steps from the earlier precursor-feeding


96 studies. We identified two closely related cytochrome P450s (CYPs) that direct TAM **3**


97 hydroxylation toward 5-HT **9** or 6-HT **6**, establishing 6-HT **6** as the entry point to β-carboline


98 biosynthesis. Structural modeling and targeted mutagenesis further identified two amino acid


99 residues whose reciprocal substitution switched the regioselectivities of these CYPs. We next


100 demonstrated that acetaldehyde **12** serves as a carbonyl donor in a nonenzymatic Pictet

101 Spengler-type reaction with 6-HT **6** and identified an _O_ -methyltransferase (OMT) active on


102 multiple β-carboline intermediates. Consequently, we reconstructed β-carboline biosynthesis


103 in _Nicotiana benthamiana_ ( _N. benthamiana_ ), with some steps occurring spontaneously or


104 potentially mediated by endogenous host enzymes, and achieved simultaneous production of


105 β-carbolines and DMT **13** through transient expression. Finally, we established a _P. harmala_


106 hairy-root system engineered to express DMT **13** biosynthetic genes, enabling its accumulation


107 alongside endogenous β-carbolines and recapitulating the principal alkaloid combination found


108 in ayahuasca. This system offers future opportunities to decipher the β-carboline pathway


109 compartmentalization and metabolite transport, while providing a platform for engineering the


110 production of diverse TAM-derived metabolites.


111

# 112 Results


113 **Harmine accumulates predominantly in the epidermis of young** _**P. harmala**_ **roots**


114 We first quantified harmine **1** across tissues of _P. harmala_ and _B. caapi_ by high-resolution


115 liquid chromatography-tandem mass spectrometry (LC-MS/MS). Harmine **1** was detected in


116 all analyzed tissues of both species, with the highest levels found in young roots of _P. harmala_,


117 reaching 2.06 ± 0.25% of fresh weight ( **Figure 1A** and **1B,** and **Supplemental Table S1** and


118 **S2** ). Using matrix-assisted laser desorption/ionization-mass spectrometry imaging (MALDI

119 MSI), we further found that harmine **1** was enriched in the pith and epidermis of _P. harmala_


120 roots, whereas Trp **2** was observed predominantly in the cortex and phloem, suggesting that the


121 conversion of Trp **2** to β-carbolines within roots may be spatially restricted by the differential


122 distribution of biosynthetic enzymes ( **Figure 1C** ). The yellow coloration observed in the


123 optical images coincided with the regions of harmine **1** accumulation and may reflect the


124 optical properties of its conjugated π-electron system. In young and mature stems of _B. caapi_,


4


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


125 harmine **1** was enriched toward the outer bark, with higher accumulation observed in mature


126 stems ( **Supplemental Figure S2** ).


127 We further performed targeted profiling of β-carbolines and related metabolites in both species,


128 focusing on previously identified metabolites and proposed intermediates of the β-carboline


129 biosynthetic pathway (Aslam et al., 2014; Santos et al., 2020; Fahmy et al., 2021) ( **Figure 1A**,


130 **Supplemental Figure S3**, and **Supplemental Table S1** ). Both species accumulated the β

131 carbolines harmalol **7**, harmol **8**, harmaline **5**, and harmine **1** . In _P. harmala_, these metabolites


132 accumulated predominantly in roots, whereas in _B. caapi_, most showed a relatively uniform


133 distribution across tissues, with higher levels of harmaline **5** and THH **14** detected in roots and


134 leaves ( **Figure 1A** ). Interestingly, THH **14**, THH-OH **15**, and 6-methoxy-tryptamine (6-MeO

135 TAM **16** ) were detected solely in _B. caapi_ tissues ( **Supplemental Figure S4** and **Supplemental**


136 **Table S1** ). 6-HT **6** was detected in all analyzed tissues of _B. caapi_ except the stem pith, with


137 highest accumulation in the outer stem area (bark), whereas in _P. harmala_, it was detected only


138 in roots and at substantially lower levels ( **Figure 1A** ). In contrast, _P. harmala_ exhibited high


139 levels of 5-HT **9** in the roots, consistent with previous reports (McKenzie et al., 1975; Sasse et


140 al., 1982; Berlin et al., 1993) ( **Figure 1A** ). Together, these results revealed both shared and


141 species-specific features of β-carboline metabolism, potentially reflecting evolutionary


142 divergence in the underlying biosynthetic pathways and their tissue-specific regulation. In this


143 context, β-carboline-producing plants, including _P. harmala_ and _B. caapi,_ are placed in


144 evolutionarily distant lineages, highlighting the broad phylogenetic distribution of these


145 alkaloids and raising the possibility of convergent evolution of the β-carboline pathway


146 ( **Supplemental** **Figure S5** and **Supplemental Table S3** ). Given the substantially higher


147 harmine **1** accumulation and pronounced tissue-specific enrichment in _P. harmala_, we focused


148 subsequent pathway-discovery efforts on this species. Based on our chemical profiling and


149 previously proposed biosynthetic routes, we predicted a biosynthetic pathway for harmine **1** in


150 _P. harmala_, in which Trp **2** is decarboxylated to TAM **3**, followed by hydroxylation, β-carboline


151 scaffold formation, _O_ -methylation, and oxidation steps ( **Figure 1D** and **Supplemental Figure**


152 **S1D** ).


153 **Transcriptome-guided identification of key β-carboline biosynthetic enzymes**


154 To identify β-carboline biosynthetic enzymes, we performed RNA sequencing of _P. harmala_


155 leaves, stems, mature and young roots, matching the tissues used for metabolite profiling


156 ( **Figure 1A** ). We generated _de novo_ assemblies of transcriptomes using complementary TruSeq


157 and Iso-Seq sequencing approaches to obtain comprehensive transcriptomic data in the absence


5


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


158 of an annotated reference genome ( **Supplemental Table S4** ). To identify transcripts with


159 tissue-specific expression patterns corresponding to the metabolite profiles, we performed


160 hierarchical clustering of transcript expression profiles across the analyzed tissues. The


161 resulting clusters revealed distinct tissue-specific expression patterns, including groups


162 enriched in root tissues that corresponded to the preferential accumulation of β-carbolines


163 ( **Figure 2A** ). Based on the predicted pathway ( **Figure 1D** and **Supplemental** **Figure S1** ), we


164 focused on six candidate enzyme families: tryptophan decarboxylases (TDCs), CYP


165 monooxygenases, potential Pictet-Spenglerases (PSs), pyruvate decarboxylases (PDCs), _O_ 

166 methyltransferases (OMTs), and flavin-dependent oxidoreductases. Candidate enzymes were


167 prioritized according to their relative expression patterns across _P. harmala_ tissues, together


168 with predicted protein size, the presence of complete open reading frames (ORFs) and


169 conserved Pfam domains, and sequence identity to previously characterized plant enzymes


170 with similar activities.


171 **Regioselective tryptamine hydroxylation directs β-carboline biosynthesis**


172 We identified a pyridoxal 5'-phosphate (PLP)-dependent decarboxylase candidate ( _TDC1_ ) that


173 was highly expressed in young roots and lowly expressed in aerial tissues ( **Supplemental**


174 **Table S5** ). _In vitro_ assays using _Escherichia coli_ ( _E. coli_ ) lysates expressing recombinant


175 TDC1, supplemented with PLP and Trp **2**, demonstrated that TDC1 (PhaTDC1) catalyzed the


176 decarboxylation of Trp **2** to TAM **3** ( **Figure 2B** ). We further confirmed PhaTDC1 _in vitro_


177 activity using recombinant protein purified by nickel-affinity chromatography **(Supplemental**


178 **Figure S6A-S6C** ). Transient expression of _PhaTDC1_ in _N. benthamiana_ by agro-infiltration


179 resulted in substantial accumulation of TAM **3** from endogenous Trp **2**, without exogenous


180 substrate supplementation ( **Figure 2C** ). Using _PhaTDC1_ as a reference gene, we further


181 performed co-expression analysis to identify candidates for the subsequent biosynthetic steps.


182 Applying a Pearson correlation coefficient ( _r_ ) &gt; 0.8 and false discovery rate (FDR)-adjusted _p_


183 &lt; 0.05 as thresholds, we identified 1,013 Iso-Seq and 176 TruSeq transcripts co-expressed with


184 _PhaTDC1_, including 15 CYPs, 3 PDCs, 3 PSs, 15 OMTs, and 10 oxidases ( **Supplemental**


185 **Figure S6D** and **Supplemental Table S5** ).


186 Several plant CYPs have been shown to catalyze the hydroxylation of TAM **3** to 5-HT **9**


187 (Fujiwara et al., 2010; Negri et al., 2021; Dangol et al., 2022). Among the 15 CYPs identified


188 by co-expression analysis, we selected two root-enriched CYP candidates, _CYP1_ and _CYP2_,


189 annotated as putative tryptamine 5-hydroxylases (T5Hs), for functional characterization as


190 potential TAM **3** hydroxylases ( **Supplemental Table S5** ). The two candidates share 79.3%


6


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


191 amino acid sequence identity and exhibit 51.48% ( _CYP1_ ) and 53.86% ( _CYP2_ ) identity to the


192 characterized rice T5H (OsCYP71P1, LOC_Os12g16720). Phylogenetic analysis placed both


193 candidates within a clade containing functionally characterized plant T5Hs, further supporting


194 their predicted roles as TAM **3** hydroxylases ( **Supplemental Figure S7** and **Supplemental**


195 **Table S6** ). We transiently co-expressed _PhaTDC1_ with either _CYP1_ or _CYP2_ in _N._


196 _benthamiana_, and detected the production of 6-HT **6** and 5-HT **9**, respectively ( **Figure 2D** ).


197 These results establish the distinct regioselectivities of the two CYPs, with CYP1


198 (CYP71BG47, PhaT6H) and CYP2 (CYP71BG48, PhaT5H) catalyzing TAM **3** hydroxylation


199 at the C6 or C5 positions, respectively ( **Figure 2D** ). These findings support the observed


200 accumulation of 5-HT **9** in _P. harmala_ ( **Figure 1A** ) and involvement of 6-HT **6** in the early


201 steps of β-carboline biosynthesis. Unexpectedly, co-expression of _PhaTDC1_ and _PhaT6H_ also


202 resulted in the accumulation of harmalol **7** ( **Figure 2E** ), suggesting that endogenous host


203 enzymes and/or nonenzymatic reactions contribute to β-carboline scaffold formation from 6

204 HT **6** in _N. benthamiana_ ( **Figure 2F** ).


205 **Pyruvate decarboxylation precedes β-carboline scaffold formation**


206 Given the potential role of acetaldehyde **12** in Pictet-Spengler condensation with 6-HT **6**


207 ( **Supplemental Figure S1D** ), we next investigated whether pyruvate **4** decarboxylation


208 contributes to the accumulation of harmalol **7** and related intermediates. We transiently co

209 expressed each of the three identified PDC candidates ( _PDC1_ - _PDC3,_ **Supplemental Table S5** )


210 with _PhaTDC1_ and _PhaT6H_ in _N. benthamiana_ . Although none significantly altered harmalol


211 **7** accumulation ( **Supplemental Figure S8A** ), co-expression of _PDC1_ ( _PhaPDC1_ ) resulted in


212 the production of THH-OH **15** ( **Supplemental Figure S8B** ), a tetrahydro-β-carboline


213 previously detected in _B. caapi_ tissues ( **Figure 1A** and **Supplemental Figure S4A** and **S8B** ),


214 consistent with its generation through a Pictet-Spengler-type reaction between 6-HT **6** and


215 acetaldehyde **12** ( **Supplemental Figure S1D** ). Neither of the other two PDC candidates


216 supported THH-OH **15** accumulation ( **Supplemental Figure S8B** ). To directly evaluate the


217 catalytic activity of PhaPDC1, we next transiently expressed _PhaPDC1_ or _CmPDC1_, a


218 functionally characterized PDC from melon ( _Cucumis melo_ L.; MELO3C009145P1,


219 Melonomics) (Wang et al., 2019), in _N. benthamiana_ leaves, supplementing the infiltrated


220 tissues with pyruvate **4** as the substrate. Using gas chromatography-mass spectrometry (GC

221 MS), we detected a significant increase in acetaldehyde **12** in leaves expressing either PDC


222 compared with leaves infiltrated with an empty vector ( **Figure 3A** ). Transient co-expression of


223 _CmPDC1_ with _PhaTDC1_ and _PhaT6H_ also resulted in the accumulation of THH-OH **15**


7


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


224 ( **Figure 3B** ). Notably, the levels of acetaldehyde **12** were reduced in _N. benthamiana_ leaves


225 co-expressing _PhaPDC1_, _PhaTDC1_ and _PhaT6H_ compared with leaves expressing _PhaPDC1_


226 alone ( **Figure 3A** ). Together, the ability of PDCs from two different plant species to support


227 THH-OH **15** accumulation, coupled with the depletion of acetaldehyde **12** upon co-expression


228 of _PhaPDC1_ with the identified enzymes, supports a role for PDC-derived acetaldehyde **12** as


229 the carbonyl donor in β-carboline scaffold formation ( **Figure 3A** and **3B**, and **Supplemental**


230 **Figure S8B** ).


231 We further co-expressed _PhaTDC1_ and _PhaT6H_, with or without _PhaPDC1,_ in _Saccharomyces_


232 _cerevisiae_ ( _S. cerevisiae_ ), together with the _Arabidopsis thaliana_ NADPH-CYP reductase _AtR1_


233 to support plant CYP activity in yeast (Urban et al., 1997). Expression of _PhaTDC1_ and


234 _PhaT6H_ resulted in accumulation of 6-HT **6**, confirming TAM 6-hydroxylation activity by


235 _PhaT6H_ detected in _N. benthamiana_ leaves ( **Figure 3C** and **Supplemental Figure S8C** ).


236 Notably, THH-OH **15** and harmalol **7** accumulated in yeast only when _PhaPDC1_ was co

237 expressed, consistent with the results obtained in _N. benthamiana_ ( **Figure 3C** and


238 **Supplemental Figure S8C** ). These findings support a role for PDC-derived acetaldehyde **12**


239 in β-carboline scaffold formation.


240 We next tested several proteins implicated in Pictet-Spengler-type or related condensation


241 reactions (Lichman, 2021) for their potential to catalyze β-carboline scaffold formation. Three


242 candidates were selected based on sequence similarity to members of the Bet v 1-like ( _PS1_ ),


243 strictosidine synthase-like ( _PS2_ ), and major latex protein-like (MLP-like, _PS3_ ) families


244 (Maresh et al., 2008; Roddan et al., 2020; Singh et al., 2018) ( **Supplemental Table S5** ).


245 However, transient co-expression of these candidates in _N. benthamiana_ leaves together with


246 _PhaTDC1,_ _PhaT6H_, and _PhaPDC1_ did not result in increased accumulation of 6-HT **6**,


247 harmalol **7**, or THH-OH **15** ( **Supplemental** **Figure S8D** ).


248 To investigate whether the β-carboline backbone of THH-OH **15** can be generated through


249 nonenzymatic Pictet-Spengler-type reactions, we performed cell-free assays in which pyruvate


250 **4**, acetaldehyde **12**, and 6-HT **6** were incubated individually or in combination at pH 7.4 and


251 37 °C for 1 h, following the conditions described in (Callaway et al., 1994). We detected THH

252 OH **15** peaks in reactions containing both acetaldehyde **12** and 6-HT **6** ( **Supplemental** **Figure**


253 **S9A** ), while reactions of pyruvate **4** and 6-HT **6** resulted in harmalol **7** ( **Supplemental** **Figure**


254 **S9B** ). Interestingly, we detected a previously unreported peak with an accurate mass consistent


255 with that of 6-hydroxy-1,2,3,4-tetrahydro-β-carboline-1-carboxylic acid (6-H-THβC-1-COOH


256 **17** ) in reactions containing pyruvate **4** and 6-HT **6 (Supplemental Figure S9C** and **S9D** ). The


8


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


257 detection of this putative intermediate in the cell-free reactions is consistent with a reaction


258 proceeding _via_ a carboxylated tetrahydro-β-carboline intermediate followed by its spontaneous


259 oxidative decarboxylation to harmalol **7** ( **Supplemental Figure S9E** ). Together, our findings


260 provide biochemical evidence that 6-HT **6** can undergo nonenzymatic scaffold-forming


261 reactions with either pyruvate **4** or acetaldehyde **12**, supporting the two proposed routes for β

262 carboline scaffold formation ( **Supplemental Figure S1D** ).


263 **A single OMT catalyzes multiple methylation steps in β-carboline biosynthesis**


264 To identify the active OMTs, we selected four genes from the 15 OMT transcripts identified in


265 the _PhaTDC1_ -centered co-expression network based on their expression patterns and sequence


266 similarity to characterized plant OMTs. The four candidates ( _OMT1-4_, **Supplemental Table**


267 **S5** ) were recombinantly expressed in _E. coli_, and their activities were evaluated in _in vitro_


268 assays using the corresponding cell lysates, S-Adenosyl-L-methionine (SAM) as the methyl


269 donor, and either harmalol **7** or harmol **8** as substrates. Only OMT2 (PhaOMT2) catalyzed the


270 conversion of harmalol **7** to harmaline **5** and harmol **8** to harmine **1**, whereas the other


271 candidates showed no detectable activity ( **Figure 4A** ). We next purified the recombinant


272 protein and confirmed its dual-substrate activity ( **Supplemental Figure S10A** and **S10B** ).


273 Transient expression of _PhaOMT2_ in _N. benthamiana_ leaves by agro-infiltration, followed two


274 days later by separate supplementation with harmalol **7** or harmol **8**, resulted in accumulation


275 of harmaline **5** and harmine **1**, respectively ( **Figure 4B** ). Neither product was detected in the


276 corresponding controls, suggesting that _N. benthamiana_ lacks endogenous OMT activity


277 capable of catalyzing these reactions. To test whether PhaOMT2 could also _O_ -methylate 6-HT


278 **6**, we co-expressed _PhaTDC1_, _PhaT6H_, and _PhaPDC1_ with or without _PhaOMT2_ . However,


279 6-MeO-TAM **16** was not detected in any of these combinations ( **Supplemental Figure S10C** ),


280 consistent with its absence from the metabolite profiles of _P. harmala_ ( **Figure 1A** ). On the


281 other hand, we observed PhaOMT2-mediated _O_ -methylation of THH-OH **15** to THH **14**,


282 providing additional support for the tentative structural assignment of THH-OH **15**


283 ( **Supplemental** **Figure S10D** ). Interestingly, replacing _PhaPDC1_ with _CmPDC1_ did not result


284 in detectable THH **14**, although both enzymes supported THH-OH **15** production


285 ( **Supplemental** **Figure S10D** ). These results support a role for PhaOMT2 in β-carboline


286 biosynthesis, demonstrating its selectivity for substrates containing the β-carboline scaffold


287 over non-cyclized tryptamines.


9


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


288 **Reconstruction of β-carboline biosynthesis in** _**N. benthamiana**_


289 Sequential introduction of _PhaTDC1_, _PhaT6H_, _PhaPDC1,_ and _PhaOMT2_ together with a


290 feedback-insensitive mutant of bacterial deoxy-D-arabino-heptulosonate 7-phosphate


291 synthase, to enhance Trp **2** levels (Tzin et al., 2013; Berman et al., 2026) resulted in


292 accumulation of TAM **3**, 6-HT **6**, harmalol **7**, THH-OH **15**, THH **14**, and unexpectedly, also


293 harmaline **5**, and harmine **1** ( **Figure 4C** and **Supplemental** **Figure S11** ). Notably, THH **14** was


294 only observed upon co-expression of _PhaPDC1_ and _PhaOMT2_, which also increased the


295 accumulation of harmaline **5** and harmine **1** ( **Figure 4C** and **Supplemental** **Figure S11** ). These


296 findings demonstrate that β-carboline biosynthesis can be reconstructed in _N. benthamiana_


297 without introducing a dedicated oxidase, suggesting that spontaneous oxidation or endogenous


298 host enzymes can support the remaining oxidation steps.


299 To investigate whether _P. harmala_ encodes oxidases catalyzing the remaining steps of β

300 carboline biosynthesis, we focused on FAD-dependent oxidases of the berberine bridge


301 enzyme-like (BBE-like) family, which mediate diverse oxidative reactions in plant specialized


302 metabolism, including the oxidative dehydrogenation of reduced alkaloids in _Papaver_


303 _somniferum_ (Hagel et al., 2012; Bakam et al., 2026). Six BBE-like candidates were selected


304 based on their high expression in roots and co-expression with _PhaTDC1_ ( **Supplemental Table**


305 **S5** ). In addition, one YUCCA (YUC)-like flavin monooxygenase was also selected for


306 functional testing based on its potential role in oxidative reactions (Zhao et al., 2001). The


307 BBE-like candidates were tested with and without their predicted N-terminal signal peptides


308 to assess enzymatic activity independently of subcellular targeting. Oxidase candidates were


309 transiently co-expressed with _AroG, PhaTDC1_, _PhaT6H_, _PhaPDC1_, and _PhaOMT2_ in _N._


310 _benthamiana_ leaves. However, no increased accumulation of harmalol **7**, THH-OH **15**, THH


311 **14**, harmaline **5,** or harmine **1** was observed under the conditions examined ( **Supplemental**


312 **Figure S12A** and **S12B** ).


313 **Two active-site residues determine CYP regioselectivity for 5-HT and 6-HT production**


314 To understand the molecular basis of the distinct TAM **3** hydroxylation regioselectivities


315 exhibited by the closely related CYPs PhaT6H and PhaT5H, we used AlphaFold3 (Abramson


316 et al., 2024) to model each CYP protein in complex with its respective product, 6-HT **6** or 5

317 HT **9**, and superimposed the predicted structures ( **Figure 5A** ). The predicted active sites of the


318 two CYPs were nearly identical, differing at only two positions: Leu373 and Thr490 in


319 PhaT6H, corresponding to Val371 and Ala488 in PhaT5H ( **Figure 5A** and **Supplemental**


10


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


320 **Figure S13A** ). We hypothesized that the bulkier Leu373 residue in PhaT6H creates steric


321 constraints that disfavor hydroxylation at the C5 position of TAM **3**, thereby favoring C6


322 hydroxylation. To determine whether these residues govern hydroxylation regioselectivity, we


323 generated single and double mutants of each enzyme ( **Figure 5B** ). The single mutants


324 PhaT6H(L373V) and PhaT5H(V371L) produced both 6-HT **6** and 5-HT **9**, whereas


325 PhaT6H(T490A) and PhaT5H(A488T) largely retained the products of their respective wild

326 type enzymes ( **Figure 5B** ). Remarkably, the double mutants PhaT6H(L373V/T490A) and


327 PhaT5H(V371L/A488T) completely reversed their hydroxylation regioselectivity ( **Figure 5B** ).


328 AlphaFold3 modeling further revealed that the active site of the rice T5H (OsCYP71P1) closely


329 resembles that of PhaT5H ( **Supplemental Figure S13B** ). Together, these findings identify two


330 key residues governing the regioselectivity of TAM **3** hydroxylation and suggest that changes


331 in substrate binding orientation within the CYP active site may have facilitated the emergence


332 of 6-HT **6** biosynthesis in _P. harmala_ and its recruitment into the β-carboline pathway.


333 **Engineering ayahuasca-associated alkaloid biosynthesis in a single plant system**


334 We next explored whether the biosynthetic pathways of DMT **13** and harmine **1**, which share


335 TAM **3** as a common precursor but occur naturally in different plant species **Supplemental**


336 **Figure S14A** ), could be functionally integrated within a single plant system. We employed two


337 complementary approaches: transient co-expression of both pathways in _N. benthamiana,_ and


338 engineering of the DMT **13** pathway into β-carboline-producing _P. harmala_ hairy roots _._ In the


339 first approach, we transiently co-expressed _PhaTDC1_, _PhaT6H_, _PhaPDC1_, and _PhaOMT2_ for


340 β-carboline biosynthesis, together with the previously characterized DMT **13** biosynthetic


341 genes from _P. viridis_, _PvTDC2_ and _PvNMT1_ (Berman et al., 2026), using _AroG_ to enhance


342 precursor availability, in _N. benthamiana_ leaves. This resulted in simultaneous accumulation


343 of harmine **1**, DMT **13**, and their biosynthetic intermediates in the same tissue ( **Figure 6A** and


344 **Supplemental Figure S14B** ).


345 Next, we generated transgenic _P. harmala_ hairy roots through _Agrobacterium rhizogenes_ 

346 mediated transformation of cotyledons with constructs expressing _PvTDC2_ and _PvNMT1_,


347 followed by selection of transformed tissues based on RFP fluorescence. RFP-positive calli


348 developed and subsequently formed transgenic hairy roots within approximately two months


349 ( **Figure 6B, 6C**, and **Supplemental Figure S15A** ). Similar to the reconstruction in _N._


350 _benthamiana_, transformed _P. harmala_ tissues simultaneously accumulated harmine **1** and DMT


351 **13**, together with their respective biosynthetic intermediates, whereas control tissues


352 expressing RFP alone exhibited comparable β-carboline profiles but no detectable NMT **18** or


11


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


353 DMT **13** ( **Figure 6D** and **Supplemental** **Figure S15B** and **S15C** ). Notably, THH-OH **15** and


354 THH **14**, which were not detected in native _P. harmala_ roots ( **Figure 1A** ), accumulated in


355 transformed calli and hairy roots ( **Figure 6D** and **Supplemental** **Figure S15B** ), potentially


356 reflecting altered metabolic flux or reduced turnover of these intermediates in transformed


357 tissues. Their accumulation further supports their proposed participation in the biosynthetic


358 pathway, consistent with our functional assays. Together, the simultaneous production of β

359 carbolines and DMT **13** reconstitutes the characteristic combination of β-carbolines and


360 psychedelic tryptamines associated with ayahuasca brew within a single plant system and


361 demonstrates the feasibility of integrating distinct alkaloid biosynthetic pathways through


362 metabolic engineering.


363


364 **Discussion**


365 The reported antimicrobial, antifungal, and anti-herbivore activities of β-carboline alkaloids


366 suggest that these metabolites may serve diverse ecological functions. The spatial distribution


367 of these metabolites may provide further insights into these functions, as their accumulation in


368 tissues exposed to the surrounding environment could facilitate interactions with soil- and


369 surface-associated organisms. In this context, the localization of harmine **1** in the epidermal


370 regions of _P. harmala_ roots and outer stem regions of _B. caapi_ may support a role in chemical


371 defense. The preferential accumulation of β-carbolines in the roots of _P. harmala_, compared


372 with their broader distribution across tissues in _B. caapi_, may reflect differences in ecological


373 adaptation between these species, as _P. harmala_ inhabits arid and semi-arid environments,


374 whereas _B. caapi_ is a tropical Amazonian liana.


375 Based on our findings, we propose a biosynthetic pathway for β-carbolines in _P. harmala_,


376 beginning with PhaTDC1-mediated decarboxylation of Trp **2** to TAM **3**, followed by PhaT6H

377 catalyzed hydroxylation to 6-HT **6** ( **Figure 7** ). The identification of PhaT6H establishes a


378 previously unresolved enzymatic link in β-carboline biosynthesis. The distinct


379 regioselectivities of PhaT6H and its closely related homolog PhaT5H illustrate how limited


380 active-site divergence may enable the emergence of alternative metabolic pathways by


381 directing shared precursors toward chemically distinct products. The competing 5

382 hydroxylation branch may also influence precursor availability for β-carboline biosynthesis,


383 consistent with previous feeding experiments in _P. harmala_ hairy roots, in which exogenous


384 TAM **3** preferentially increased 5-HT **9** rather than β-carboline production (Berlin et al., 1993).


12


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


385 Accordingly, suppressing PhaT5H activity in this system may provide a strategy to redirect


386 metabolic flux toward increased β-carboline production. More broadly, our findings highlight


387 the potential of AlphaFold 3-guided structural modeling and protein design to elucidate how


388 specific active-site residues influence substrate binding and positioning, thereby determining


389 enzyme specificity and regioselectivity, and guide protein engineering to tailor enzyme activity.


390 Several routes have been proposed for the biosynthetic assembly of the β-carboline scaffold,


391 but their biochemical basis has remained unresolved ( **Supplemental Figure S1** ). Our findings


392 support a pathway in which 6-HT **6**, rather than TAM **3**, undergoes Pictet-Spengler-type


393 cyclization through two alternative routes: a putative nonenzymatic Pictet-Spengler reaction


394 between 6-HT **6** and pyruvate **4**, followed by oxidative decarboxylation to harmalol **7** (Module


395 I in **Figure 7** ), or a nonenzymatic Pictet-Spengler condensation of 6-HT **6** with acetaldehyde


396 **12**, generated through PhaPDC1-mediated decarboxylation of pyruvate **4**, to yield THH-OH


397 **15** (Module II in **Figure 7** ). Subsequent oxidation steps may occur spontaneously or involve


398 endogenous enzymes that remain to be characterized. The spontaneous formation of β

399 carboline intermediates demonstrates the chemical feasibility of scaffold assembly without a


400 dedicated PS enzyme. However, spontaneous and enzyme-catalyzed reactions may occur in


401 parallel, and the contribution of endogenous enzymes, including potentially promiscuous PS


402 enzymes, to these reactions cannot be excluded. This raises the possibility that the emergence


403 of β-carboline biosynthetic pathways may depend on the evolution of enzymes that generate


404 suitable reactive intermediates, without requiring dedicated catalysts for every subsequent


405 transformation. The activity of PhaOMT2 toward multiple β-carboline intermediates, including


406 harmalol **7**, harmol **8**, and THH-OH **15**, provides another example of this biosynthetic


407 flexibility ( **Figure 7** ).


408 The involvement of PhaPDC1 in this pathway reveals an unexpected role for an enzyme


409 primarily associated with central carbon metabolism in specialized alkaloid biosynthesis. In


410 addition to the established functions of PDCs in fermentation, the pyruvate dehydrogenase


411 bypass, and volatile aroma biosynthesis (Wang et al., 2019), our findings demonstrate that


412 PDC-derived acetaldehyde **12** can serve as a reactive precursor for spontaneous PS-type


413 condensation with 6-HT **6** . To our knowledge, this is the first example of PDC-derived


414 acetaldehyde 12 serving as a reactive precursor for alkaloid biosynthesis in plants. Notably,


415 PhaPDC1 was predicted to localize to plastids based on the presence of a chloroplast transit


416 peptide (localization probability of 0.97; **Supplemental Table S5** ), whereas CmPDC1 was


417 previously reported to localize to the cytosol (Wang et al., 2019). Despite their predicted


13


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


418 differences in subcellular localization, both enzymes supported THH-OH **15** accumulation in


419 _N. benthamiana_, whereas only PhaPDC1 showed detectable THH **14** accumulation. These


420 differences suggest that precursor availability, subcellular compartmentalization, and


421 downstream metabolic activities may influence the relative contributions of alternative


422 biosynthetic routes and the resulting metabolic profiles.


423 The complementary engineering of _N. benthamiana_ and _P. harmala_ hairy roots demonstrates


424 two strategies for the production of high-value TAM **3** -derived alkaloids. Transient expression


425 in _N. benthamiana_ provides a flexible platform for reconstructing and combining heterologous


426 pathways, paving the way for their stable introduction into engineered plants, either


427 individually or in combination. In contrast, _P. harmala_ hairy roots enable the introduction of


428 new biosynthetic capabilities into a native β-carboline-producing system, where targeted


429 silencing of competing metabolic branches may enhance product accumulation and facilitate


430 the discovery of additional biosynthetic enzymes, transporters, and other pathway-associated


431 factors. This platform could be further expanded by incorporating additional TAM **3** -derived


432 pathways, building on our previous reconstruction of psychedelic tryptamines and their


433 halogenated derivatives in plants (Berman et al., 2026). Such modular engineering strategies


434 offer opportunities to produce diverse natural and new-to-nature alkaloids while investigating


435 precursor competition, pathway compatibility, and the biochemical factors governing their


436 production.


437 The growing pharmacological and therapeutic interest in β-carbolines and DMT **13**, together


438 with their reliance on plant-derived sources, highlights the need for sustainable production


439 strategies. Given the importance of β-carbolines and DMT **13** in ayahuasca preparations, stable


440 engineered plants could provide renewable sources of these compounds and their derivatives,


441 supporting further pharmaceutical and biotechnological research. Such approaches could


442 reduce reliance on wild harvesting, particularly of slowly regenerating plants, alleviate pressure


443 on natural populations, and contribute to the conservation of plant genetic diversity and


444 associated ecosystems.


445


446


447


14


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

# 448 Methods


449 **Chemicals and plant material**


450 All analytical standards used in the study were of ≥ 95% purity. Trp **2** and TAM **3** were obtained


451 from Sigma-Aldrich (Rehovot, Israel). 5-MeO-TAM, THβC-1-COOH **10**, 5-HT **9**, 6-HT **6**,


452 THH **14**, harmalol **7**, harmaline **5**, harmol **8**, and harmine **1** were purchased from Cayman


453 Chemical (Ann Arbor, MI, USA). _Peganum harmala_ seeds were obtained from the Israel Gene


454 Bank, and _B. caapi_ young plants were sourced from a local nursery. Plants were grown in a


455 greenhouse under long-day conditions, with day/night temperatures maintained at 24/18 ºC,


456 respectively. Tissues from _P. harmala_ and _B. caapi_ plants were collected at approximately one


457 year of age for chemical profiling and RNA extraction.


458 **High resolution liquid chromatography-mass spectrometry (LC-MS) chemical analysis**


459 Unless otherwise stated, 100 mg of frozen powdered plant tissue was extracted with 300 µL


460 80% methanol containing 0.1% formic acid. Samples were sonicated for 30 min, agitated for


461 an additional 30 min, and centrifuged at 14,000 g for 10 min. The resulting supernatants were


462 filtered through 0.22 µm syringe filters prior to LC-MS/MS analysis. Metabolite profiling was


463 performed using an ultra-performance liquid chromatography system (UPLC; Waters Acquity)


464 equipped with a diode array detector coupled to either an Orbitrap Exploris 240 MS (Thermo


465 Scientific, Bremen, Germany) or an Orbitrap IQ-X Tribrid mass spectrometer.


466 Chromatographic separation was achieved using a Waters Acquity UPLC HSS-T3 C18 column


467 (100 mm × 2.1 mm i.d., 1.7 µm particle size). The mobile phases consisted of (A) 0.1% formic


468 acid in acetonitrile: water (1:99, v/v) and (B) 0.1% formic acid in acetonitrile. The flow rate


469 was maintained at 0.3 mL min &#91;−1&#93;, and the column temperature was set to 35 °C. The gradient


470 program was as follows: 0% B for 2 min, increased to 75% B over 14 min, raised to 100% B


471 within 0.2 min, maintained at 100% B until 19 min, returned to 0% B within 0.2 min, and


472 equilibrated the system at 0% B until 22 min. Mass spectrometry was performed using


473 electrospray ionization (ESI) in positive ionization mode over an _m/z_ range of 70-1,000. Source


474 parameters were set as follows: sheath gas, auxiliary gas, and sweep gas flow rates of 45, 10,


475 and 1 arbitrary units, respectively; vaporizer and ion transfer tube temperatures of 300 °C; and


476 spray voltage of 3.5 kV. Data were acquired in full MS &#91;1&#93; mode at 60,000 resolution, followed


477 by data-dependent MS/MS (dd-MS &#91;2&#93; ) acquisition at 15,000 resolution. Both acquisition modes


478 used standard normalized automatic gain control (AGC) targets and automatic maximum


479 injection times. The dd-MS &#91;2 &#93; settings included a 1.5 m/z isolation window and stepped


15


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


480 normalized collision energies of 30, 45, and 60. Compounds with analytical standards were


481 identified by comparison of their retention times and MS/MS spectra with those of the


482 corresponding standards, whereas compounds without available analytical standards were


483 tentatively annotated based on accurate mass, relative retention time, and MS/MS


484 fragmentation patterns ( **Supplemental Figure S2** and **S10** ). Since minor differences in


485 fragmentation spectra, retention times, and ion abundance were observed between instruments,


486 metabolites were confirmed by comparison with analytical standards and/or extracts analyzed


487 under matched experimental conditions.


488 **Absolute quantification of β-carbolines**


489 Fresh bark, pith, sap, stem, leaf, and root tissues were collected from _B. caapi_, and young and


490 mature roots, stems, and leaves were collected from _P. harmala_ . For metabolite extraction, 100


491 mg of frozen, powdered tissue was weighed in triplicate and extracted with 1 ml of 80%


492 methanol containing 0.1% formic acid, following the procedure described above. Absolute


493 metabolite quantification was performed using external calibration curves. To ensure that


494 analyte concentrations fell within the linear range of the calibration curves, samples were


495 analyzed at multiple dilutions. LC-MS analyses were conducted on an IQ-X Tribrid mass


496 spectrometer operating in full MS &#91;1&#93; acquisition mode. In addition to absolute quantification,


497 samples were subjected to comparative metabolite profiling, with peak areas used for relative


498 abundance analyses.


499 **MALDI-MSI of** _**B. caapi**_ **stems and** _**P. harmala**_ **roots**


500 For metabolite localization in _B. caapi_ and _P. harmala_ tissues, samples were cryo-sectioned


501 ( _n_ = _2_ biological replicates) and coated with matrix using a TM Sprayer (HTX Technologies,


502 USA). The matrix solution consisted of 2,5-dihydroxybenzoic acid (DHB; 40 mg mL &#91;-1&#93; in 70%


503 containing 0.2% trifluoroacetic acid). Matrix application was performed with the nozzle


504 maintained at 70 °C, using 16 spray passes at a linear velocity of 120 cm min &#91;-1 &#93; and a flow rate


505 of 50 μL min &#91;-1&#93; . Tissue sections were imaged using a Nikon DS-Ri2 microscope. MALDI-MSI


506 was conducted on a 7 T Solarix FT-ICR (Fourier Transform Ion Cyclotron Resonance) mass


507 spectrometer (Bruker Daltonics). Data acquisition and processing were performed using


508 flexImaging v4.1 (Bruker). Spectra were acquired in positive ion mode with lock-mass


509 calibration using the DHB matrix ion (&#91;3DHB+H-3H 2 O&#93; &#91;+&#93;, _m/z_ 409.055408). Data were


510 collected at a laser frequency of 1 kHz and 40% laser power, with 200 laser shots per pixel and


511 a spatial resolution of 50 µm. Mass spectra were acquired over an m/z range of 120-2000 in


16


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


512 broadband mode with an acquisition time domain of 1M, yielding an estimated resolving power


513 of approximately 115,000 at _m/z_ 400. MS/MS spectra were normalized to root-mean-square


514 intensity, and ion images were generated using a mass tolerance of ± 0.005% of the theoretical


515 _m/z_ value with pixel interpolation enabled.


516 **RNA sequencing and annotation of plant transcriptomes**


517 Total RNA was extracted from the same tissue types of _P. harmala_ and _B. caapi_ used for


518 metabolite profiling _,_ using either the Spectrum Plant total RNA kit (Sigma-Aldrich) or the


519 RNeasy plant mini kit (Qiagen). RNA integrity was assessed using a TapeStation instrument.


520 Extracted RNA from pooled tissues was used for both short-read and long-read transcriptome


521 sequencing. Iso-Seq sequencing libraries for _B. caapi_ and _P. harmala_ were generated from


522 pooled RNA samples derived from six and four tissues, respectively. Paired-end Illumina


523 libraries were prepared using the TruSeq RNA Library Preparation Kit (Illumina) following the


524 manufacturer’s instructions and sequenced on an Illumina NovaSeq X Plus platform (PE150),


525 generating approximately 40 million reads per sample. Full-length transcript sequencing was


526 performed using the PacBio Iso-Seq workflow. Libraries were prepared using the SMRTbell


527 Prep Kit 3.0 (Pacific Biosciences) from pooled tissue RNA samples of _B. caapi_ and _P. harmala_


528 and sequenced on the PacBio Sequel II platform using SMRT Cell 8M and the Sequel II


529 Sequencing Kit, according to the manufacturer’s instructions. Iso-Seq data were processed


530 using the SMRT Link command-line tools. Primer sequences were removed and reads were


531 demultiplexed with lima v2.9.0, using the isoseq and peek-guess switches, and clipped


532 sequences were saved with the dump-clips switch. The isoseq refine v4.0.0 was then used with


533 the require-polya switch to obtain full-length, nonconcatemer (FLNC) reads. FLNC reads were


534 converted to FASTQ with BamTools v2.5.2 and clustered using isONclust v0.0.6.1 (Sahlin and


535 Medvedev, 2020) with the isoseq switch and default parameters (k = 15, w = 50, minimum


536 shared minimizers = 5, minimum average read quality = 7). The representative read of each


537 cluster was extracted into a FASTA file per species, which was used for downstream annotation.


538 Raw Illumina reads were quality-filtered and trimmed using Trimmomatic (Bolger et al., 2014),


539 as implemented in Trinity (Haas et al., 2013) with default parameters. The processed reads


540 were subsequently used for _de novo_ transcriptome assembly using Trinity (Haas et al., 2013).


541 Transcript abundances were estimated using RSEM (Li and Dewey, 2011) and normalized


542 using DESeq2 (Love et al., 2014). Functional annotation of predicted transcripts from the


543 Trinity assemblies and the Iso-Seq cluster sequences of both species was performed as


544 described previously (Berman et al., 2024), with additional analyses. Subcellular localization


17


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


545 was predicted using TargetP 2.0 (Armenteros et al., 2019) and DeepLoc 2.0 (Thumuluri et al.,


546 2022). Enzymatic functions were predicted with ECPred (Dalkiran et al., 2018) for the


547 Illumina-based transcriptomes and with DeepEC (Ryu et al., 2019) for the Iso-Seq sequences.


548 Completeness of the Iso-Seq transcript sets was assessed with BUSCO v5.7.1 (Manni et al.,


549 2021) in transcriptome mode using the eudicots_odb10 dataset (n = 2,326). Hierarchical


550 clustering was performed using Euclidean distance and complete linkage on trimmed mean of


551 M-values (TMM)-normalized TruSeq transcript expression values converted to counts per


552 million (CPM). Transcripts with CPM &gt; 25 in at least one of 12 sample tissues were retained,


553 resulting 10,905 of 307,954 transcripts (3.54%). Expression values were averaged across three


554 biological replicates for each of four tissues and transformed as log 2 expression relative to the


555 median expression across tissues. For co-expression analysis, Pearson correlation coefficients


556 were calculated between the bait transcript ( _PhaTDC1_ ) and each isoseq transcript across the


557 analyzed samples. Transcripts with _r_ - 0.8 and FDR-adjusted _p_ &lt; 0.05 were considered


558 positively co-expressed with _PhaTDC1_ . The resulting co-expression network was visualized


559 using Cytoscape (Shannon et al., 2003).


560 **Sequence alignments and phylogenetic analysis**


561 Protein sequence comparisons were performed using Clustal Omega, and sequence identity


562 was quantified as the proportion of identical residues within the aligned regions. For the


563 phylogenetic analysis of β-carboline-producing plants, amino sequences of maturase K


564 (MatK), encoded by the chloroplast _matK_ gene, were retrieved from GenBank. The


565 corresponding protein accession numbers are listed in **Supplemental Table S3** . For the CYP


566 phylogenetic analysis, amino acid sequences of putative _P. harmala_ CYPs and functionally


567 characterized plant CYPs were selected from previously published studies of TAM


568 hydroxylases and related enzymes (Fujiwara et al., 2010; Negri et al., 2021; Dangol et al.,


569 2022). Sequences were retrieved from the plant CYP database, and the complete list of protein


570 IDs is provided in **Supplemental Table S5** . A multiple-sequence alignment was generated


571 using MUSCLE, and a maximum-likelihood phylogenetic tree was constructed in MEGA 12


572 with 1,000 bootstrap replicates. The species-level tree was rooted.


573 **Recombinant protein expression in** _**E. coli**_


574 The coding sequence of _PhaTDC1_ was amplified by PCR from cDNA of _P. harmala_, and


575 _PhaOMT2_ was synthesized by Twist Biosciences. Each gene was individually cloned into the


576 EcoRI-digested pET28a vector using Gibson Assembly (Gibson et al., 2009). All constructs


18


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


577 were sequenced-verified and transformed into _E. coli_ BL21(DE3) pLysS competent cells.


578 Starter cultures were grown overnight in LB medium supplemented with kanamycin (50 µg


579 mL &#91;-1&#93; ) at 37 ºC, then diluted 1:100 into fresh LB containing kanamycin and incubated at 37 ºC


580 until reaching an optical density at 600 nm (A600) of 0.6. Protein expression was induced with


581 500 μM isopropyl-1-thio-β-d-galactopyranoside (IPTG), and cultures were further incubated


582 overnight at 16 °C. Cells were harvested by centrifugation at 4,300 rpm for 10 min, and pellets


583 were lysed by sonication in lysis buffer consisting of 50 mM Tris-HCl (pH 8.0), 10% glycerol,


584 0.1 mg mL &#91;-1&#93; lysozyme (Sigma-Aldrich), 0.01% benzonase nuclease (Sigma-Aldrich), protease


585 inhibitor cocktail (Sigma-Aldrich), and 0.5 mM phenylmethylsulfonyl fluoride (PMSF, Sigma

586 Aldrich). The clarified lysate was either used directly for activity assays or subjected to protein


587 purification. Hexahistidine-tagged proteins were purified using a Capturem His-tag


588 purification 24-well plate (Takara Bio, USA) following the manufacturer’s instructions.


589 Elution was performed with 500 mM imidazole in buffer containing 20 mM NaH 2 PO 4 (pH 7.6)


590 and 500 mM NaCl. Desalting was carried out by repeated centrifugation at 4 °C using Amicon


591 Ultra 10 kDa cutoff filters (Merck Millipore) in Tris-HCl (pH 8.0). Protein purity was evaluated


592 by SDS-PAGE on 10% polyacrylamide gels. Protein purification and enzymatic assays were


593 performed sequentially on the same day.


594 **TDC and OMT enzyme assays**


595 TDC activity assays were carried out in a total volume of 50 μL containing 100 mM Tris-HCl


596 buffer (pH 8.0), 1 mM pyridoxal-5′-phosphate (PLP), 2 mM Trp **2,** and either 25 μL of crude


597 lysate or 5 μL of purified enzymes. OMT enzyme assays were similarly performed in 50 μL


598 reactions composed of 100 mM Tris-HCl buffer (pH 8.0), 1 mM S-adenosyl-L-methionine


599 (SAM), and 25 μL of crude enzyme lysate. Reactions were incubated at 30 °C for 2 h and


600 terminated by the addition of 70 μL ice-cold methanol with 0.1% formic acid. Samples were


601 vortexed and centrifuged at 13,000 g for 10 min. The supernatants were filtered and analyzed


602 by LC-MS.


603 **Gas chromatography-mass spectrometry (GC-MS) analysis of acetaldehyde**


604 Leaves of _N. benthamiana_ infiltrated with _PhaPDC1_ or _CmPDC1_, or with combinations of


605 _PhaTDC1_ and _PhaT6H_ with or without _PhaPDC1_, were ground into a fine powder, and 1 g of


606 tissue was transferred into GC-MS glass vials containing 1 g of NaCl and 2 mL of 20% NaCl


607 solution (w v &#91;-1&#93; ), followed by addition of 20 µL of 2-octanol (5 mg L &#91;-1&#93; ) as an internal standard.


608 The vials were sealed with magnetic screw caps filled with silicone/PTFE septa (Thermo Fisher


19


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


609 Scientific, Waltham, MA, USA) and shaken for 30 min at 37 °C and 125 rpm. After shaking,


610 acetaldehyde **12** was extracted from the leaf tissue using solid-phase microextraction (SPME)


611 using a 50/30-µm DVB/CAR/PDMS StableFlex fiber (2 cm; Supelco, Bellefonte, PA, USA).


612 GC-MS analysis was performed on an Agilent 7890A gas chromatograph coupled to an Agilent


613 5977B mass spectrometer using an HP-5MS column (5% phenyl methyl polysiloxane; 30 m x


614 0.25 mm x 0.25 µm film thickness; Agilent Technologies, Santa Clara, CA, USA). The SPME


615 fiber was desorbed at 250 °C for 5 min in splitless mode, with helium used as the carrier gas.


616 The temperature program was as follows: initial temperature at 40 °C for 1 min; ramp to 90 °C


617 at 5 °C/min; and held at 90 °C for 2 min. The mass spectrometer was operated in electron


618 ionization mode at 70 eV, with an acquisition range of _m/z_ 40-280. Acetaldehyde **12** was


619 identified by matching its mass spectral fragmentation pattern to reference spectra in the NIST


620 library, version 14. Four biological replicates from each treatment were analyzed.


621 **Cell free assays**


622 Cell-free reactions were performed in a total volume of 50 μL containing 100 mM Tris-HCl


623 buffer (pH 7.4) and 1 mM each of pyruvate **4**, acetaldehyde **12**, 6-HT **6**, added individually or


624 in the indicated combinations. Reactions were incubated at 37 °C for 1 h and quenched with


625 70 μL ice-cold methanol with 0.1% formic acid. Samples were vortexed and centrifuged at


626 13,000 g for 10 min. The supernatants were filtered and analyzed by LC-MS.


627 _**Agrobacterium**_ **-mediated transient expression in** _**N. benthamiana**_


628 The coding sequences of _PhaTDC1, PhaT6H, PhaT5H, PS1, PS3, PhaPDC1, PDC2-3, BBE1,_


629 _BBE4, BBE5_ were amplified by PCR from cDNA of _P. harmala_, whereas _OMT1, PhaOMT2,_


630 _OMT3, OMT4,_ the _PhaT5H_ and _PhaT6H_ single and double mutants _, PS2, BBE2_T, BBE2_iso,_


631 _BBE3, YUC1,_ and _CmPDC1_ were synthesized by Twist Biosciences. The genes were


632 individually cloned into the pAlpha2-NPT II-Ubq10-CCD-Ter10 vector, which had been


633 digested with ApaI, using Gibson Assembly (Gibson et al., 2009). All constructs were


634 sequenced-verified and subsequently transformed into _Agrobacterium tumefaciens_ ( _A._


635 _tumefaciens_ ) strain GV3101 by electroporation. Recombinant _A. tumefaciens_ carrying


636 overexpression constructs were cultured overnight at 28 °C in LB medium supplemented with


637 kanamycin and gentamicin (50 µg mL &#91;-1&#93; each). For expression of the mutated _AroG_


638 (AroGF209A, carrying the F209A substitution), previously generated GV3101 stocks kindly


639 provided by Prof. Gad Galili, Prof. Vered Tzin, and Dr. Moran Oliva were used (Tzin et al.,


640 2013). Following overnight incubation, bacterial cells were harvested by centrifugation,


20


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


641 washed, and resuspended in infiltration buffer (2 mM Na 3 PO 4, 50 mM MES, 10 mM MgCl 2,


642 0.5 % w v &#91;-1&#93; glucose, and 100 µM acetosyringone) to an OD 600 of 1.0. Strains carrying different


643 constructs were combined in equal or defined volumes to achieve the desired gene


644 combinations. The resulting suspensions were infiltrated into the abaxial side of 4-5-week-old


645 _N. benthamiana_ leaves using a 1-mL needleless syringe. Leaves were harvested 5-7 days post

646 infiltration for metabolite analysis. Samples were flash-frozen and extracted as previously


647 described followed by LC-MS/MS analysis.


648 **Heterologous expression in yeast**


649 The coding sequences of _PhaTDC1_, _PhaT6H_, _PhaPDC1_, and _AtR1_ were individually cloned


650 into pESC vector backbones (Agilent Technologies, Santa Clara, CA, USA) to enable dual gene


651 expression under bidirectional, galactose-inducible _GAL1_ and _GAL10_ promoters. _PhaTDC1_


652 and _PhaT6H_ were cloned into pESC-URA under the control of _GAL1_ and _GAL10_ promoters,


653 respectively, whereas _PhaPDC1_ and _AtR1_ were cloned into pESC-HIS under the corresponding


654 promoters. Vectors were linearized by PCR, and constructs were assembled _via_ In-Fusion


655 cloning. Construct integrity was confirmed by Sanger sequencing before transformation into _S._


656 _cerevisiae_ strain BY4743 using the Frozen-EZ Yeast Transformation II Kit (Zymo Research,


657 Irvine, CA, USA) according to the manufacturer’s instructions. Yeast co-transformed with


658 empty pESC-HIS and pESC-URA vectors served as the control. Transformants were selected


659 on Synthetic Defined (SD) minimal medium lacking uracil and histidine (SD-URA-HIS),


660 prepared using Yeast Nitrogen Base without amino acids (Sigma-Aldrich, Rehovot, Israel) and


661 appropriate Synthetic Complete drop-out supplements (Formedium, Hunstanton, UK)


662 supplemented with 2% (w v &#91;-1&#93; ) D-glucose. Positive colonies were screened and verified by


663 colony PCR. For expression studies, primary cultures were inoculated into 10 mL of SD


664 minimal medium containing 2% (w v &#91;-1&#93; ) D-glucose and grown for 24 h at 30 °C with shaking


665 at 200 rpm. To induce heterologous expression, cells were harvested, washed, transferred to


666 SD minimal medium supplemented with 2% (w v &#91;-1&#93; ) D-galactose, and incubated for an


667 additional 24 h at 30 °C with shaking at 200 rpm. Cultures were then centrifuged at 4,000 × _g_


668 for 5 min, and cell pellets were collected and weighed. For metabolite extraction, two stainless


669 steel beads (4 mm diameter) and 400 µL of cold methanol were added to the pellets. Samples


670 were homogenized by brief vortexing, sonicated for 20 min, and centrifuged at 15,000 × _g_ for


671 10 min. The resulting supernatant was filtered through a 0.22-µm membrane prior to LC

672 MS/MS analysis.


673 **Hairy root transformation of** _**P. harmala**_


21


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


674 The _PvTDC2_ and _PvNMT1_ genes were amplified from constructs containing the respective


675 genes from Psychotria viridis, as described in our previous study (Berman et al., 2026).


676 Respective binary constructs encoding RFP or RFP::PvNMT1::PvTDC2 were individually


677 assembled into the pAlpha2-NPT II-Ubq10-CCD-Ter10 vector, each under the control of a 35S


678 promoter and NOS terminator, using GoldenBraid cloning (Sarrion-Perdigones et al., 2011).


679 To generate transgenic hairy roots of _P. harmala_, approximately two-week-old cotyledons were


680 infected with _Rhizobium rhizogenes_ ( _R. rhizogenes_ ) strain MSU440 harboring the respective


681 binary constructs. Electrocompetent MSU440 cells were transformed by electroporation and


682 cultured in LB medium supplemented with the appropriate antibiotics. Overnight cultures were


683 transferred to LB containing 100 µM Acetosyringone and incubated at 28 °C overnight with


684 shaking (160 rpm). Cells were then pelleted, resuspended in co-cultivation or transformation


685 (CT) medium containing D-glucose (22 g L &#91;-1&#93; ) and KH 2 PO 4 (0.17 g L &#91;-1&#93;, pH 5.5), supplemented


686 with 50 µM Acetosyringone, and adjusted to an OD 600 of 0.5. Cotyledons were immersed in


687 the _R. rhizogenes_ suspension for 30 min, bottled dry, and placed on CT medium containing 100


688 µM Acetosyringone and 6 g L &#91;-1 &#93; Plant Agar. Explants were co-cultivated with _R. rhizogenes_ for


689 2 days at 23 ± 1°C in the dark. Subsequently, explants were transferred to Murashige and Skoog


690 (MS) medium containing cefotaxime (250 mg L &#91;-1&#93; ) and 8 g L &#91;-1&#93; Plant Agar and incubated at 25


691 ± 1 °C under partial light. Explants were sub-cultured onto fresh MS medium with appropriate


692 antibiotics every two weeks until transgenic hairy roots emerged, which were then transferred


693 to liquid Gamborg’s B5 medium supplemented with 4% sucrose and the appropriate antibiotics.


694 The B5 medium was refreshed every two weeks.


695 **Structural modeling and rationale-guided mutagenesis of PhaCYPs**


696 PhaT6H and PhaT5H proteins were modeled using AlphaFold3 (Abramson et al., 2024) with


697 the substrates 6-HT **6** and 5-HT **9**, respectively. AlphaFold3 was installed locally following the


698 &#91;developers’ instructions (https://github.com/google-deepmind/alphafold3). Predictions were&#93;(https://github.com/google-deepmind/alphafold3)


699 performed using multiple random seeds, and the plDDT scores of the obtained AlphaFold3


700 models were in the range of 0.96 to 0.97. Structural models were visualized in PyMOL, and


701 comparison of the predicted active sites identified two divergent residues: Leu373 in PhaT6H


702 versus Val371 in PhaT5H and Thr490 versus Ala488, respectively. These residues were


703 selected for rationale-guided site-directed mutagenesis to assess their contribution to TAM


704 hydroxylation regioselectivity. Mutant genes were synthesized by Twist Biosciences, and


705 individually cloned into the pAlpha2-NPT II-Ubq10-CCD-Ter10 vector, which had been


22


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


706 digested with ApaI, using Gibson Assembly (Gibson et al., 2009). The resulting mutants were


707 functionally characterized by agroinfiltration into _N. benthamiana_ as described above.


708 **Statistical analysis**


709 All statistical analyses were conducted using GraphPad Prism (version 8.0.1). Data are


710 presented as median ± standard deviation (SD) based on at least three independent biological


711 replicates. For comparisons between two independent groups, statistical significance was


712 assessed using a two-tailed unpaired Student’s t-test assuming unequal variances. For analyses


713 involving more than two groups, a one-way ANOVA followed by Tukey’s post hoc multiple

714 comparison test was applied. When ANOVA could not be performed due to zero variance in


715 one or more groups, significance was instead evaluated using a two-tailed unpaired Student’s


716 t-test with unequal variance. The number of biological replicates, statistical tests applied, and


717 significance thresholds are indicated in the corresponding figure legends.

# 718 Data and code availability


719 All raw next-generation sequencing data will be deposited in the European Nucleotide Archive


720 (ENA), and the sequences of the active genes reported in this manuscript will be deposited at


721 NCBI GenBank. All data and sequences will be made publicly available upon acceptance of


722 the manuscript.

# 723 Supplemental information


724 Supplemental Figures, Tables and References are provided as Supplemental Information files.

# 725 Author contributions


726 L.K.H., A.A., and P.B. conceptualized the study. P.B. and J.H. performed the chemical profiling


727 with the aid of I.R. E.A.-S. performed the Iso-Seq transcriptome assembly and annotation, and


728 A.F. performed the TruSeq transcriptome assembly and annotation. L.K.H., J.H., S.C. and P.B.


729 performed all the molecular biology work, _in vitro_ enzyme assays, protein purification, and


730 transient expression of genes in _N. benthamiana_ . L.K.H. analyzed LC-MS data with the


731 assistance of P.B. L.K.H. and A.K. performed GC-MS analysis of PhaPDC1 activity in _N._


732 _benthamiana_ . S.C. cloned the genes, performed yeast transformation, and conducted the


733 enzymatic activity assays. Y.D. prepared samples and performed the MALDI imaging


734 experiments. O.K. performed the molecular modeling, analyzed the data, and designed the site

735 directed mutants. L.K.H. cloned, expressed, and tested the site-directed mutants with the help


736 of S.C. S.C. cloned the genes for the hairy root transformation, B.A.T. performed


23


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


737 transformations and generated calli, and L.K.H. maintained the cultures and generated hairy


738 roots. L.K.H. wrote the original draft, and L.K.H. and P.B. reviewed and edited the manuscript


739 with input from all coauthors. S.M., U.H., S.J.F., and I.M. contributed materials and guidance


740 to specific parts of the study. P.B. and A.A. supervised the study and acquired funding. All the


741 authors read and approved the final version of the manuscript.

# 742 Acknowledgments


743 L.K.H was partially supported by a Faculty Dean fellowship from Faculty of Biochemistry at


744 the Weizmann Institute of Science. A.A. is the incumbent of the Peter J. Cohn Professorial


745 Chair. This study was supported by the Israel Science Foundation (P.B.); and the Adelis


746 Foundation, the Leona M. and Harry B. Helmsley Charitable Trust, the Jeanne and Joseph


747 Nissim Foundation for Life Sciences, Tom and Sondra Rykoff Family Foundation Research,


748 Ron Sklare and the Raymond Burton Plant Genome Research Fund (A.A.). We thank Prof.


749 David Nelson (University of Tennessee, USA) for the systematic naming of the CYPs


750 characterized in this study.


751

# 752 References


753 **Aaghaz, S., Sharma, K., Jain, R., and Kamal, A.** (2021). β-Carbolines as potential anticancer


754 agents. _Eur. J. Med. Chem._ **216** :113321.


755 **Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O.,**


756 **Willmore, L., Ballard, A. J., Bambrick, J., et al.** (2024). Accurate structure prediction


757 of biomolecular interactions with AlphaFold 3. _Nature_ **630** :493–500.


758 **Armenteros, J. J. A., Salvatore, M., Emanuelsson, O., Winther, O., Von Heijne, G.,**


759 **Elofsson, A., and Nielsen, H.** (2019). Detecting sequence signals in targeting peptides


760 using deep learning. _Life Sci. Alliance_ **2** :e201900429.


761 **Aslam, N., Wani, A. A., Nawchoo, I. A., and Bhat, M. A.** (2014). Distribution and Medicinal


762 importance of _Peganum_ _harmala_ -A review. _Int. J. Adv. Res._ **2** :751–755.


763 **Bakam, V. M., Mérindol, N., Germain, H., and Desgagné-Penix, I.** (2026). Berberine bridge


764 enzyme-like oxidases in plant specialized metabolism: evolution, catalytic diversity and


765 opportunities for metabolic engineering. _Phytochem. Rev._ Advance Access published July


766 11, 2026, doi:10.1007/s11101-026-10286-y.


24


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


767 **Berlin, J., Rügenhagen, C., Greidziak, N., Kuzovkina, I. N., Witte, L., and Wray, V.**


768 (1993). Biosynthesis of serotonin and β-carboline alkaloids in hairy root cultures of


769 Peganum harmala. _Phytochemistry_ **33** :593–597.


770 **Berman, P., de Haro, L. A., Cavaco, A. R., Panda, S., Dong, Y., Kuzmich, N., Lichtenstein,**


771 **G., Peleg, Y., Harat, H., Jozwiak, A., et al.** (2024). The biosynthetic pathway of the


772 hallucinogen mescaline and its heterologous reconstruction. _Mol. Plant_ **17** :1129–1150.


773 **Berman, P., Höfer, J., Mehlman, H., Almekias-Siegl, E., Khersonsky, O., Dong, Y.,**


774 **Heinig, U., Sulimani, L., Hao, L. K., Cohen, S., et al.** (2026). Complete biosynthesis of


775 psychedelic tryptamines from three kingdoms in plants. _Sci. Adv._ **12** :eaeb3034.


776 **Bolger, A. M., Lohse, M., and Usadel, B.** (2014). Trimmomatic: a flexible trimmer for


777 Illumina sequence data. _Bioinformatics_ **30** :2114.


778 **Bouayad, N., Rharrabe, K., Lamhamdi, M., Nourouti, N. G., and Sayah, F.** (2012). Dietary


779 effects of harmine, a β-carboline alkaloid, on development, energy reserves and α-amylase


780 activity of _Plodia_ _interpunctella_ Hübner (Lepidoptera: Pyralidae). _Saudi J. Biol. Sci._


781 **19** :73–80.


782 **Callaway, J. C., Gyntber, J., Poso, A., Airaksinen, M. M., and Vepsäläinen, J.** (1994). The


783 pictet‐spengler reaction and biogenic tryptamines: Formation of tetrahydro‐β‐carbolines


784 at physiological pH. _J. Heterocycl. Chem._ **31** :431–435.


785 **Cavin, J. C., and Rodriguez, E.** (1988). The influence of dietary β-carboline alkaloids on


786 growth rate, food consumption, and food utilization of larvae of _Spodoptera_ _exigua_


787 (Hubner). _J. Chem. Ecol._ **14** :475–484.


788 **Cui, G., Yuan, H., Jiang, Z., Zhang, J., Sun, Z., and Zhong, G.** (2020). Natural harmine


789 negatively regulates the developmental signaling network of _Drosophila_ _melanogaster_


790 (Drosophilidae: Diptera) _in_ _vivo_ . _Ecotoxicol. Environ. Saf._ **190** :110134.


791 **Dai, J., Dan, W., Schneider, U., and Wang, J.** (2018). β-Carboline alkaloid monomers and


792 dimers: Occurrence, structural diversity, and biological activities. _Eur. J. Med. Chem._


793 **157** :622–656.


794 **Dalkiran, A., Rifaioglu, A. S., Martin, M. J., Cetin-Atalay, R., Atalay, V., and Doğan, T.**


795 (2018). ECPred: A tool for the prediction of the enzymatic functions of protein sequences


796 based on the EC nomenclature. _BMC Bioinformatics_ **19** :334.


797 **Dangol, A., Shavit, R., Yaakov, B., Strickler, S. R., Jander, G., and Tzin, V.** (2022).


25


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


798 Characterizing serotonin biosynthesis in _Setaria viridis_ leaves and its effect on aphids.


799 _Plant Mol. Biol._ **109** :533–549.


800 **de Osório, F. L., Sanches, R. F., Macedo, L. R., dos Santos, R. G., Maia-De-Oliveira, J.**


801 **P., Wichert-Ana, L., de Araujo, D. B., Riba, J., Crippa, J. A., and Hallak, J. E.** (2015).


802 Antidepressant effects of a single dose of ayahuasca in patients with recurrent depression:


803 A preliminary report. _Rev. Bras. Psiquiatr._ **37** :13–20.


804 **Diehl, K. A., Ocampo-Dallas, S., Hazelwood, O. S., Demura-Devore, J., Muro-Villanueva,**


805 **F., Nett, R. S., and Ashraf, M. A.** (2026). Harmala alkaloids regulate cell division planes


806 in plants. _bioRxiv_ Advance Access published June 2, 2026,


807 doi:10.64898/2026.05.29.728883.


808 **Dos Santos, R. G., and Hallak, J. E. C.** (2024). Ayahuasca: Pharmacology, safety, and


809 therapeutic effects. _CNS Spectr._ **30** :e2.


810 **Fahmy, S. A., Issa, M. Y., Saleh, B. M., Meselhy, M. R., and Azzazy, H. M. E. S.** (2021).


811 _Peganum_ _harmala_ alkaloids self-assembled supramolecular nanocapsules with enhanced


812 antioxidant and cytotoxic activities. _ACS Omega_ **6** :11954–11963.


813 **Fang, X., Yu, H. Y., Han, L. F., and Pang, X.** (2019). N-containing compounds from seeds


814 of _Paganum_ _harmala_ . _Zhongguo Zhongyao Zazhi_ **44** :1601–1606.


815 **Fritzsche; J.** (1848). Bestandtheile der Samen von Peganum harmala. _Justus Liebigs Ann._


816 _Chem._ **64** :360–369.


817 **Fujiwara, T., Maisonneuve, S., Isshiki, M., Mizutani, M., Chen, L., Ling Wong, H.,**


818 **Kawasaki, T., and Shimamoto, K.** (2010). Sekiguchi lesion gene encodes a cytochrome


819 P450 monooxygenase that catalyzes conversion of tryptamine to serotonin in rice. _J. Biol._


820 _Chem._ **285** :11308–11313.


821 **Gibson, D. G., Young, L., Chuang, R. Y., Venter, J. C., Hutchison, C. A., and Smith, H.**


822 **O.** (2009). Enzymatic assembly of DNA molecules up to several hundred kilobases. _Nat._


823 _Methods_ **6** :343–345.


824 **Haas, B. J., Papanicolaou, A., Yassour, M., Grabherr, M., Blood, P. D., Bowden, J.,**


825 **Couger, M. B., Eccles, D., Li, B., Lieber, M., et al.** (2013). De novo transcript sequence


826 reconstruction from RNA-seq using the Trinity platform for reference generation and


827 analysis. _Nat. Protoc._ **8** :1494–1512.


828 **Hagel, J. M., Beaudoin, G. A. W., Fossati, E., Ekins, A., Martin, V. J. J., and Facchini, P.**


26


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


829 **J.** (2012). Characterization of a flavoprotein oxidase from opium poppy catalyzing the


830 final steps in sanguinarine and papaverine biosynthesis. _J. Biol. Chem._ **287** :42972–42983.


831 **Herraiz, T., González, D., Ancín-Azpilicueta, C., Arán, V. J., and Guillén, H.** (2010). β

832 Carboline alkaloids in _Peganum_ _harmala_ and inhibition of human monoamine oxidase


833 (MAO). _Food Chem. Toxicol._ **48** :839–845.


834 **Jamieson, C. S., Misa, J., Tang, Y., and Billingsley, J. M.** (2021). Biosynthesis and synthetic


835 biology of psychoactive natural products. _Chem. Soc. Rev._ **50** :6950–7008.


836 **Kaasik, H., Souza, R. C. Z., Zandonadi, F. S., Tófoli, L. F., and Sussulini, A.** (2021).


837 Chemical Composition of Traditional and Analog Ayahuasca. _J. Psychoactive Drugs_


838 **53** :65–75.


839 **Li, B., and Dewey, C. N.** (2011). RSEM: Accurate transcript quantification from RNA-Seq


840 data with or without a reference genome. _BMC Bioinformatics_ **12** .


841 **Lichman, B. R.** (2021). The scaffold-forming steps of plant alkaloid biosynthesis. _Nat. Prod._


842 _Rep._ **38** :103–129.


843 **Love, M. I., Huber, W., and Anders, S.** (2014). Moderated estimation of fold change and


844 dispersion for RNA-seq data with DESeq2. _Genome Biol._ **15** :1–21.


845 **Manni, M., Berkeley, M. R., Seppey, M., Simão, F. A., and Zdobnov, E. M.** (2021). BUSCO


846 Update: Novel and Streamlined Workflows along with Broader and Deeper Phylogenetic


847 Coverage for Scoring of Eukaryotic, Prokaryotic, and Viral Genomes. _Mol. Biol. Evol._


848 **38** :4647–4654.


849 **McKenzie, E., Nettleship, L., and Slaytor, M.** (1975). New natural products from _Peganum_


850 _harmala_ . _Phytochemistry_ **14** :273–275.


851 **Moloudizargari, M., Mikaili, P., Aghajanshakeri, S., Asghari, M., and Shayegh, J.** (2013).


852 Pharmacological and Therapeutic effects of Peganum harmala and its main alkaloids.


853 _Pharmacogn. Rev._ **7** :199–212.


854 **Negri, S., Commisso, M., Avesani, L., and Guzzo, F.** (2021). The case of tryptamine and


855 serotonin in plants: A mysterious precursor for an illustrious metabolite. _J. Exp. Bot._


856 **72** :5336–5355.


857 **Nettleship, L., and Slaytor, M.** (1974). Limitations of feeding experiments in studying


858 alkaloid biosynthesis in Peganum harmala callus cultures. _Phytochemistry_ **13** :735–742.


27


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


859 **Palhano-Fontes, F., Barreto, D., Onias, H., Andrade, K. C., Novaes, M. M., Pessoa, J. A.,**


860 **Mota-Rolim, S. A., Osório, F. L., Sanches, R., Dos Santos, R. G., et al.** (2019). Rapid


861 antidepressant effects of the psychedelic ayahuasca in treatment-resistant depression: A


862 randomized placebo-controlled trial. _Psychol. Med._ **49** :655–663.


863 **Patel, K., Gadewar, M., Tripathi, R., Prasad, S. K., and Patel, D. K.** (2012). A review on


864 medicinal importance, pharmacological activity and bioanalytical aspects of beta

865 carboline alkaloid “ Harmine.” _Asian Pac. J. Trop. Biomed._ **2** :660–664.


866 **Pinkas, Z., Khersonsky, O., Berman, P., Kuzmich, N., Rogachev, I., Fleishman, S. J., and**


867 **Aharoni, A.** (2025). Glycosylated cannabinoids in _Cannabis_ _sativa_ and enzyme design to


868 modulate their synthesis. _Proc. Natl. Acad. Sci. U. S. A._ **122** :e2515688122.


869 **Ryu, J. Y., Kim, H. U., and Lee, S. Y.** (2019). Deep learning enables high-quality and high

870 throughput prediction of enzyme commission numbers. _Proc. Natl. Acad. Sci. U. S. A._


871 **116** :13996–14001.


872 **Sahlin, K., and Medvedev, P.** (2020). De Novo Clustering of Long-Read Transcriptome Data


873 Using a Greedy, Quality Value-Based Algorithm. In _Journal of Computational Biology_,


874 pp. 472–484. J Comput Biol.


875 **Santos, B. W. L., de Oliveira, R. C., Sonsin-Oliveira, J., Fagg, C. W., Barbosa, J. B. F.,**


876 **and Caldas, E. D.** (2020). Biodiversity of β-carboline profile of _Banisteriopsis_ _caapi_ and


877 ayahuasca, a plant and a brew with neuropharmacological potential. _Plants_ **9** :1–14.


878 **Sarrion-Perdigones, A., Falconi, E. E., Zandalinas, S. I., Juárez, P., Fernández-del-**


879 **Carmen, A., Granell, A., and Orzaez, D.** (2011). GoldenBraid: An iterative cloning


880 system for standardized assembly of reusable genetic modules. _PLoS One_ **6** :e21622.


881 **Sasse, F., Heckenberg, U., and Berlin, J.** (1982). Accumulation of β-Carboline Alkaloids and


882 Serotonin by Cell Cultures of _Peganum_ _harmala_ L: I. CORRELATION BETWEEN


883 PLANTS AND CELL CULTURES AND INFLUENCE OF MEDIUM


884 CONSTITUENTS. _Plant Physiol._ **69** :400–404.


885 **Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D., Amin, N.,**


886 **Schwikowski, B., and Ideker, T.** (2003). Cytoscape: A software Environment for


887 integrated models of biomolecular interaction networks. _Genome Res._ **13** :2498–2504.


888 **Shao, H., Huang, X., Zhang, Y., and Zhang, C.** (2013). Main alkaloids of _Peganum_ _harmala_


889 L. and their different effects on dicot and monocot crops. _Molecules_ **18** :2623–2634.


28


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


890 **Sharifi-Rad, J., Quispe, C., Herrera-Bravo, J., Semwal, P., Painuli, S., Özçelik, B.,**


891 **Hacihasanoǧlu, F. E., Shaheen, S., Sen, S., Acharya, K., et al.** (2021). _Peganum_ spp.:


892 A Comprehensive Review on Bioactivities and Health-Enhancing Effects and Their


893 Potential for the Formulation of Functional Foods and Pharmaceutical Drugs. _Oxid. Med._


894 _Cell. Longev._ **2021** :5900422.


895 **Singh, A., Massicotte, M. A., Garand, A., Tousignant, L., Ouellette, V., Bérubé, G., and**


896 **Desgagné-Penix, I.** (2018). Cloning and characterization of norbelladine synthase


897 catalyzing the first committed reaction in Amaryllidaceae alkaloid biosynthesis. _BMC_


898 _Plant Biol._ **18** :338.


899 **Stolle, K., and Gröger, D. D. D.** (1968). Untersuchungen zur Biosynthese des Harmins. _Arch_


900 _Pharm_ **301** :561.


901 **Thumuluri, V., Almagro Armenteros, J. J., Johansen, A. R., Nielsen, H., and Winther, O.**


902 (2022). DeepLoc 2.0: multi-label subcellular localization prediction using protein


903 language models. _Nucleic Acids Res._ **50** :W228–W234.


904 **Tzin, V., Rogachev, I., Meir, S., Zvi, M. M. Ben, Masci, T., Vainstein, A., Aharoni, A.,**


905 **and Galili, G.** (2013). Tomato fruits expressing a bacterial feedback-insensitive 3-deoxy

906 D-arabino-heptulosonate 7-phosphate synthase of the shikimate pathway possess


907 enhanced levels of multiple specialized metabolites and upgraded aroma. _J. Exp. Bot._


908 **64** :4441–4452.


909 **Urban, P., Mignotte, C., Kazmaier, M., Delorme, F., and Pompon, D.** (1997). Cloning,


910 yeast expression, and characterization of the coupling of two distantly related _Arabidopsis_


911 _thaliana_ NADPH-cytochrome P450 reductases with P450 CYP73A5. _J. Biol. Chem._


912 **272** :19176–19186.


913 **Wang, M., Zhang, L., Boo, K. H., Park, E., Drakakaki, G., and Zakharov, F.** (2019).


914 PDC1, a pyruvate/α-ketoacid decarboxylase, is involved in acetaldehyde, propanal and


915 pentanal biosynthesis in melon ( _Cucumis melo_ L.) fruit. _Plant J._ **98** :112–125.


916 **Zhang, L., Li, D., and Yu, S.** (2020). Pharmacological effects of harmine and its derivatives:


917 a review. _Arch. Pharm. Res._ **43** :1259–1275.


918 **Zhao, Y., Christensen, S. K., Fankhauser, C., Cashman, J. R., Cohen, J. D., Weigel, D.,**


919 **and Chory, J.** (2001). A role for flavin monooxygenase-like enzymes in auxin


920 biosynthesis. _Science (80-. )._ **291** :306–309.


29


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


921


922 **Figure 1. Tissue-specific distribution of β-carbolines. (A)** Liquid chromatography-mass


923 spectrometry (LC-MS) relative abundance of β-carbolines across _Banisteriopsis_ _caapi_ ( _B._


924 _caapi_ ) and _Peganum harmala_ ( _P. harmala_ ) tissues (non-detected metabolites are shown in


925 white; _z_ -score; mean, _n_ = 3 biological replicates). Metabolites were identified by LC-MS using


926 analytical standards based on accurate mass, retention time, and MS/MS fragmentation patterns


927 ( **Supplemental Figure S3** ). Tetrahydroharmol (THH-OH **15** ) and 6-methoxy-tryptamine **(** 6

928 MeO-TAM **16** ) were tentatively identified based on accurate mass and comparison of their


929 retention times and MS/MS fragmentation patterns with those of the structurally related


930 compounds tetrahydroharmine (THH **14** ) and 5-methoxy-tryptamine ( **Supplemental Figure**


931 **S4** ). **(B)** Representative photographs of _B. caapi_ and _P. harmala_ plants and tissues used for


932 metabolite profiling. **(C)** Matrix-assisted laser desorption/ionization-mass spectrometry


933 imaging (MALDI-MSI) of _P. harmala_ root sections ( _n_ = 2 biological replicates) showing the


934 spatial distribution of harmine **1** (&#91;M+H&#93; &#91;+&#93;, m/z 213.102 ± 0.01 Da) and tryptophan (Trp **2**,


935 &#91;M+H&#93; &#91;+&#93;, m/z 205.097 ± 0.01 Da). **(D)** Simplified predicted biosynthetic pathway of β

936 carbolines in _P. harmala._ Several arrows represent multiple reaction steps. The continuation of


937 **Figure 1** appears on the next page.


30


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


938 TAM **3**, tryptamine; 6-HT **6**, 6-hydroxy-tryptamine; 5-HT **9**, 5-hydroxy-tryptamine; THβC-1

939 COOH **10**, 1,2,3,4-tetrahydro-β-carboline-1-carboxylic acid; THH **14**, tetrahydroharmine;


940 THH-OH **15**, tetrahydroharmol; 6-MeO-TAM **16**, 6-methoxy-tryptamine.


31


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


941


942 The legend of **Figure 2** appears on the next page.


943


32


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


944 **Figure 2. Functional characterization of tryptamine biosynthesis and hydroxylation in**


945 _**Peganum harmala**_ **. (A)** Hierarchical clustering analysis was performed on trimmed mean of


946 M-values (TMM)-normalized TruSeq transcripts using expression values converted to counts


947 per million (CPM), with Euclidean distance and complete linkage. Transcripts with a CPM


948 expression value greater than 25 in at least one of 12 tissue samples were retained (10,905 out


949 of 307,954 total transcripts). Transcripts encoding enzymes characterized in this study are


950 labeled with their names to illustrate their co-expression patterns. **(B)** Extracted ion


951 chromatograms (EICs) of tryptamine (TAM **3** ) following _in vitro_ enzyme assays using


952 _Escherichia coli_ ( _E. coli_ ) lysates expressing _PhaTDC1_ with PLP and tryptophan (Trp **2** ) as


953 substrates ( _n_ = 3). **(C)** Ion abundance peak areas of TAM **3** following transient expression of


954 _PhaTDC1_ in _Nicotiana benthamiana_ ( _N. benthamiana_ ) leaves (median ± SD; _n_ = 4 biological


955 replicates). **(D)** EICs of 5-hydroxy-tryptamine (5-HT **9** ) and 6-hydroxy-tryptamine (6-HT **6** )


956 following transient co-expression of _PhaTDC1_ with _PhaT6H_ or _PhaT5H_ in _N. benthamiana_


957 leaves ( _n_ = 4). Plus signs (+) indicate biosynthetic genes expressed in each experiment, with


958 boxes colored according to the last enzyme added in a set of coexpressed genes. Corresponding


959 chemical structures are included for reference. **(E)** EICs of harmalol **7** following transient co

960 expression of _PhaTDC1_ with _PhaT6H_ ( _n_ = 4 biological replicates). **(F)** Schematic


961 representation of 6-HT **6** biosynthesis and its conversion to harmalol **7** in _N. benthamiana_


962 through a spontaneous or host enzyme-mediated reaction, as indicated by a question mark. All


963 the metabolites were identified by LC-MS/MS using analytical standards based on accurate


964 mass, retention time, and MS/MS fragmentation patterns. Chromatograms were normalized to


965 the highest values. EV, empty vector; Std, standard.


966


33


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


967


968 **Figure 3. Functional characterization of PhaPDC1 reveals an acetaldehyde-dependent**


969 **route to β-carboline scaffold formation. (A)** Ion abundance peak areas (median ± SD; _n_ = 4


970 biological replicates) of acetaldehyde **12** analyzed by gas chromatography-mass spectrometry


971 (GC-MS) following transient co-expression of _PhaTDC1_, _PhaT6H_, and _PhaPDC1_ or


972 _CmPDC1_ (pyruvate decarboxylase from _Cucumis melo_ ) in _Nicotiana benthamiana_ leaves.


973 Pyruvate **4** was supplied two days after agroinfiltration, and samples were harvested 5 days


974 after infiltration. Statistical significance was assessed by one-way ANOVA followed by


975 Tukey’s HSD test; letters denote statistically significant differences ( _p_ &lt; 0.05). Acetaldehyde


976 **12** was identified by matching its mass spectral fragmentation pattern to reference spectra in


977 the NIST library. **(B)** Extracted ion chromatograms (EICs) of tetrahydroharmol (THH-OH **15** )


978 following transient expression of _PhaPDC1_ or _CmPDC1_ together with _PhaTDC1_ and _PhaT6H_ .


979 The continuation of **Figure 3** appears on the next page.


980


34


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


981 **(C)** EICs of 6-hydroxy-tryptamine (6-HT **6** ), THH-OH **15**, and harmalol **7** following


982 heterologous co-expression of _PhaTDC1_ and _PhaT6H_, with or without _PhaPDC1_, together


983 with _AtR1_ in _Saccharomyces cerevisiae_ ( _n_ = 4 biological replicates). Chromatograms were


984 normalized to the highest values. All metabolites were identified using authentic analytical


985 standards, except for THH-OH **15**, which was tentatively identified by comparison of its


986 retention time and MS/MS fragmentation pattern with those of the corresponding metabolite


987 detected in a reference extract of _Banisteriopsis caapi_ ( **Supplemental Figure S8C** ). Plus signs


988 (+) indicate biosynthetic genes expressed in each experiment, with boxes colored according to


989 the last enzyme added in a set of coexpressed genes. Corresponding chemical structures are


990 included for reference. EV, empty vector; Std, standard.


991


35


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


992


993 **Figure 4. Functional characterization of PhaOMT2 enables heterologous reconstruction**


994 **of harmine biosynthesis in** _**Nicotiana benthamiana**_ **.** Extracted ion chromatograms of


995 harmaline **5** and harmine **1** following **(A)** _in vitro_ enzyme assays using _Escherichia coli_ ( _E._


996 _coli_ ) lysates expressing individually OMT1-OMT4 proteins, with S-Adenosyl methionine


997 (SAM) and harmalol **7** or harmol **8** as substrates ( _n_ = 3 biological replicates), or **(B)** transient


998 expression of _PhaOMT2_ in _Nicotiana benthamiana_ ( _N. benthamiana_ ) leaves, ( _n_ = 4 biological


999 replicates). The continuation of **Figure 4** appears on the next page.


36


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1000 Leaves were supplemented with harmalol **7** or harmol **8** as substrates 2 days after


1001 agroinfiltration and harvested 3 days later. Chromatograms were normalized to the highest


1002 values. **(C)** Ion abundance peak areas of β-carbolines and intermediates detected following


1003 transient expression of different combinations of _PhaTDC1_, _PhaT6H_, _PhaPDC1_, and


1004 _PhaOMT2_ in _N. benthamiana_ leaves together with the bacterial _AroG_ mutant which increases


1005 the accumulation of aromatic amino acids in _N. benthamiana_ leaves (Berman et al., 2026).


1006 Pyruvate **4** was supplied 2 days after agroinfiltration, and samples were harvested 3-5 days


1007 later. Data are presented as median ± SD ( _n_ = 4 biological replicates). Plus signs (+) indicate


1008 biosynthetic genes expressed in each experiment, with boxes colored according to the last


1009 enzyme added in a set of coexpressed genes. Corresponding chemical structures are included


1010 for reference. All metabolites were identified using authentic analytical standards, except for


1011 THH-OH **15**, which was tentatively identified by comparison of its retention time and MS/MS


1012 fragmentation pattern with those of the corresponding metabolite detected in a reference extract


1013 of _Banisteriopsis caapi_ ( **Supplemental Figure S11** ). Trp **2**, tryptophan; TAM **3**, tryptamine; 5

1014 HT **9**, 5-hydroxy-tryptamine; 6-HT **6**, 6-hydroxy-tryptamine; THH **14**, tetrahydroharmine; EV,


1015 empty vector; Std, standard.


1016


37


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1017


1018 **Figure 5. Structural and functional characterization of hydroxylation regioselectivity in**


1019 _**Peganum harmala**_ **cytochrome P450 (CYP) enzymes. (A)** Overlay of AlphaFold3-predicted


1020 structures of PhaT6H in complex with 6-hydroxy-tryptamine (6-HT **6**, cyan) and PhaT5H in


1021 complex with 5-hydroxy-tryptamine (5-HT **9**, wheat). The heme groups, bound ligands, and


1022 two differing active-site residues are shown as sticks. Leu373 and Thr490 in PhaT6H,


1023 corresponding to Val371 and Ala488 in PhaT5H, are proposed to influence substrate


1024 positioning and hydroxylation regioselectivity. **(B)** Extracted ion chromatograms of 5-HT **9**


1025 and 6-HT **6** following transient co-expression of _PhaTDC1_ with wild-type _PhaT6H_ or _PhaT5H_


1026 (highlighted in yellow), or their respective mutants in _Nicotiana benthamiana_ leaves ( _n_ = 4


1027 biological replicates). Amino acid substitutions are indicated for each mutant. The continuation


1028 of **Figure 5** appears on the next page.


38


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1029 Samples were collected four days after infiltration. 6-HT **6** and 5-HT **9** were identified using


1030 analytical standards. Chromatograms were normalized to the highest values. EV, empty vector;


1031 Std, standard.


1032


39


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1033


1034 **Figure 6. Combined production of β-carbolines and** _**N**_ **,** _**N**_ **-dimethyltryptamine** ( **DMT) in a**


1035 **single plant. (A)** Representative image of _Nicotiana benthamiana_ ( _N. benthamiana_ ), ion


1036 abundance peak areas of harmine **1**, DMT **13**, and their biosynthetic intermediates following


1037 transient expression of harmine **1** biosynthetic genes ( _PhaTDC1_, _PhaT6H_, _PhaPDC1_, and


1038 _PhaOMT2_ ), DMT **13** biosynthetic genes ( _PvTDC2_ and _PvNMT1_ ), or their combination. The


1039 continuation of **Figure 6** appears on the next page.


40


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1040 The bacterial _AroG_ mutant was co-expressed in all combinations to enhance the accumulation


1041 of aromatic amino acids in _N. benthamiana_ leaves. Data are presented as median ± SD ( _n_ = 4


1042 biological replicates). Samples were harvested seven days after infiltration. Plus signs (+)


1043 indicate biosynthetic genes expressed in each experiment, with blue, purple, and combined


1044 blue-purple boxes representing β-carboline, DMT **13**, and combined biosynthetic genes,


1045 respectively. **(B-C)** Representative bright-field (BF), green fluorescence protein (GFP) or red


1046 fluorescence protein (RFP) reporters, and merged images of transformed _Peganum harmala_ ( _P._


1047 _harmala_ ) hairy roots generated from _P. harmala_ cotyledons following _Rhizobium rhizogenes_


1048 (MSU440)-mediated transformation with a construct expressing _PvTDC2_ and _PvNMT1_ from


1049 _Psychotria viridis_ (Berman et al., 2026), and RFP as a reporter. **(D)** Ion abundance peak areas


1050 (median ± SD; _n_ = 3 biological replicates) of harmine **1**, DMT **13**, and their intermediates


1051 detected in transgenic hairy roots. Hairy roots transformed with an RFP-expressing empty


1052 vector (EV) served as the control. All metabolites were identified using authentic analytical


1053 standards, except for THH-OH **15**, which was tentatively identified by comparison of its


1054 retention time and MS/MS fragmentation pattern with those of the corresponding metabolite


1055 detected in a reference extract of _Banisteriopsis caapi_ . Corresponding chemical structures are


1056 included for reference. 5-HT **9**, 5-hydroxy-tryptamine; 6-HT **6**, 6-hydroxy-tryptamine; THH


1057 **14**, tetrahydroharmine; NMT **18**, _N_ -methyltryptamine.


1058


41


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1059


1060 **Figure 7. Proposed β-carboline biosynthetic pathway in** _**Peganum harmala**_ **.** Modules I and


1061 II illustrate two proposed routes for β-carboline scaffold formation through Pictet-Spengler


1062 (PS) reactions. Yellow arrows indicate the proposed biosynthetic pathway in _Peganum harmala_


1063 ( _P. harmala_ ), and dashed arrows experimentally unconfirmed steps. Enzymes functionally


1064 characterized from _P. harmala_ are highlighted in yellow or red. Enzyme classes marked with


1065 question marks indicate reactions for which specific enzymes remain unidentified. The


1066 hydroxyl groups of 6-hydroxytryptamine (6-HT **6** ) and 5-hydroxytryptamine (5-HT **9** ) are


1067 highlighted in cyan and wheat, respectively, to indicate their distinct hydroxylation positions.


1068 The continuation of **Figure 7** appears on the next page.


42


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766139; this version posted November 19, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.12.766139)

preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.


1069 The _N_, _N_ -dimethyltryptamine (DMT **13** ) biosynthetic pathway discovered previously in


1070 Psychotria viridis (Berman et al., 2026), including its enzymes, metabolites, and reaction


1071 arrows, is shown in purple to illustrate its biosynthetic relationship with harmine **1** . Trp **2**,


1072 tryptophan; TAM **3**, tryptamine; THH-OH **15**, tetrahydroharmol; THH **14**, tetrahydroharmine;


1073 NMT **18**, _N_ -methyltryptamine.


43


