Fatty acid amide hydrolase (FAAH) is a conserved amidase that hydrolyzes lipophilic N-acylethanolamines (NAEs) into ethanolamine and free fatty acid^1,2,3,4. For example; the rat, mouse and human FAAHs hydrolyze endogenous anandamide (NAE20:4) into arachidonic acid and ethanolamine^5,6. In mammals, this process is linked to a variety of neurological and physiological processes ranging from pain perception to diet regulation^7,8. In another example, the recombinant FAAH derived from the plant moss, Physcomitrella patens, was also able to hydrolyze NAE20:4 into corresponding products, which appeared to be associated with growth and development of the moss species⁹. Elsewhere, C. elegans was shown to possess a FAAH enzyme which metabolized NAE20:4 in support of regulation of longevity¹⁰. These, and numerous other reports have demonstrated a broadly-conserved FAAH machinery across multiple species for NAE regulation in various biological processes.

In the model plant, Arabidopsis thaliana, in vitro and in vivo studies showed that A. thaliana FAAH (AtFAAH) can cleave the amide bond of endogenous NAEs, including saturated (e.g. NAE12:0), monosaturated (e.g. NAE18:1), and polyunsaturated (e.g. NAE18:3) species, as well as 9-LOX derived NAE oxylipins (9-hydroxy linoleoylethanolamide; NAE-9-HOD)^1,11,12,13. Regulation of NAE contents by FAAH appears to be associated with a variety of biological processes¹⁴. For example, in Arabidopsis, AtFAAH overexpressing lines had lower amounts of endogenous NAEs, and exhibited enhanced differences in seedling growth¹⁵, flowering time¹⁶ and innate immunity¹⁷, compared to wild-type plants. By contrast, Arabidopsis faah-knockouts had elevated endogenous NAE contents, and were hypersensitive to growth inhibition induced by external applications with NAEs (e.g. NAE12:0 or NAE18:2)^15,18,19, alkamides^20,21, N-acyl homoserine lactones (AHLs)²², or the phytohormone ABA¹³. In another report, upland cotton (Gossypium hirsutum L.) seedlings with ectopic overexpression of AtFAAH displayed insensitivity to NAE12:0, or the NAE18:2-derived hydroxide (NAE-9-HOD), thus resembling the outcomes found in Arabidopsis²³. These reports suggest that regulation of N-acylethanolamide content by FAAH may influence a number of physiological processes in plants.

N-Acyl-ʟ-homoserine lactones (AHLs) or N-aryl-ʟ-homoserine lactones (aryl-HLs) are bacteria-derived lipids that are utilized in cell-to-cell communication (quorum sensing; QS) processes (e.g. biofilm production) or interactions with a host (e.g. recognition of symbiotic or pathogenic bacteria)^24,25,26. Due to the structural similarities between NAEs and these QS molecules (polar head group with an amide-linked acyl tail), it was hypothesized that plant FAAHs could hydrolyze AHLs and aryl-HLs²². Further experiments revealed that recombinant AtFAAH hydrolyzes long acyl chain AHLs such as OdDHL (N-(3-oxododecanoyl)-ʟ-homoserine lactone) or OtDHL (N-(3-oxotetradecanoyl)-ʟ-homoserine lactone) better than AHLs with shorter chains such as OHHL (N-(3-Oxohexanoyl)-ʟ-homoserine lactone)²², thus indicating that AtFAAH activity can be influenced by the length/nature of the acyl chain in the substrate. In the same experiment, the aryl-HL, p-coumaryl-HL was a poor substrate for AtFAAH²². Also, AHLs were shown to trigger a "biphasic" growth response in A. thaliana seedlings. In their experiments, wild-type Arabidopsis seedlings fed with AHLs displayed enhanced or reduced growth phenotypes at low (e.g. 0.1 µM) or elevated AHL (e.g. 100 µM) concentrations, respectively²². Then, side-by-side comparisons with Arabidopsis faah-knockouts, revealed that such sensitivity can be modulated in a FAAH-dependent manner, that is, hydrolysis of AHLs by FAAH was required for modulation of the growth signal²².

Alkamides are fatty acid amides of fungal- or plant-origin^21,27,28 and constitute another group of NAE-like molecules that have been associated with FAAH-mediated hydrolysis²⁰. In wild-type Arabidopsis, exogenous applications of the alkamide, affinin (N-isobutyl head group and acyl chain) resulted in a "biphasic" effect that resembled that of AHLs (see above). Indeed, Arabidopsis wild-type seedlings fed with affinin at concentrations below 28 µM had primary roots that were longer than controls²¹. Conversely, affinin at concentrations above 28 µM led to markedly impaired root development²¹. In another study, Arabidopsis FAAH overexpressors or faah-knockouts demonstrated that alkamide susceptibility was modulated via FAAH-hydrolysis²⁰. Interestingly, the in vitro AtFAAH activity towards affinin or other 12 alkamides of various acyl chain lengths and alkyl head groups was far more inferior compared to that of AtFAAH hydrolysis of NAEs²⁰. Nonetheless, these reports support a connection between FAAH-hydrolysis of alkamides and plant growth effects.

FAAH is characterized by the presence of a conserved Ser-Ser-Lys catalytic triad^1,29,30. The N-acyl amides that reach the active site at the acyl binding channel (ABC) form a covalent intermediate between the substrate and the enzyme. This is achieved by activation of the catalytic serine (nucleophile) which attacks the carbon of the carbonyl group in the substrate³¹. Mammalian and plant FAAHs appeared to have distinct mechanisms for substrate binding. For example, AtFAAH is proposed to undergo conformational changes that lead to the movement of a helix region (residues 531–537) and the closure of its membrane access channel (MAC), by a process termed "squeeze and lock"¹. By contrast, in the rat FAAH, the ligand-enzyme complex is secured by a "dynamic paddle" which closes its narrower MAC and "traps" the substrate in the active site. In both cases, following FAAH catalysis, ethanolamine is released into the cytosol via the cytosolic access channel (CAC) whereas the fatty acid product is released back into the hydrophobic environment of the membrane bilayer^1,29. The resolved crystal structures of both plant and mammalian FAAHs have provided insights into key structural and functional components that could be used as reference for FAAHs from other organisms.

Recently, the comparison of 88 FAAH amino acid sequences from diverse angiosperm species revealed two major phylogenetic groups, FAAH1 and FAAH2³². Plant species in the Brassica family such as A. thaliana or Camelina sativa exhibited only one FAAH isoform (FAAH1) whereas all dicot and monocot species, including Amborella trichopoda (the "angiosperm ancestral species") had both FAAH groups, and many having more than one member³². For example, tomato (Solanum tuberosum) has three FAAH1 isoforms and one FAAH2, whereas rice (Oryza sativa) has one FAAH1 and two FAAH2 isoforms³². Analysis of multiple sequence alignments exhibited residue changes in regions corresponding to the substrate binding pocket of FAAH1 and FAAH2 isoforms. Homology modeling of soybean (Glycine max) FAAHs revealed additional structural and chemical differences between the soybean FAAH1 and FAAH2³². These differences were hypothesized to influence the substrate selectivity of FAAH1 and FAAH2 enzymes, and further suggest that FAAHs may have evolved distinct FAAH machineries to modulate an expanded repertoire of lipophilic signaling lipids^32,33. Although conceptually this notion appears plausible, thus far, experimental evidence to support this hypothesis is lacking, largely because there is not biochemical information related to the FAAH2 group of enzymes. Here, we address this gap in knowledge by combining computational and biochemical approaches to study and characterize two FAAHs from the legume Medicago truncatula, MtFAAH1 and MtFAAH2a, and provide evidence that these two isoforms have reciprocal substrate preferences.

Group I and II FAAHs of the legume M. truncatula

To examine the FAAH diversity within legumes, we compared FAAH amino acid sequences of ten selected species using AtFAAH (a FAAH1 enzyme) as reference (Fig. 1a; Supplementary Table S1). Data showed that these legume FAAHs clustered in two major groups, designated as FAAH1 and FAAH2 (Fig. 1a). Only one FAAH1 isoform was identified in each legume species. Notably, FAAH2 formed two subgroups in most species, FAAH2a and FAAH2b. With the exception of G. max and Vigna radiata, the rest of legumes (including M. truncatula) had two FAAH2 isoforms. Compared to the founding FAAH1 member, AtFAAH, the percentages of amino acid sequence similarity for MtFAAH1, MtFAAH2a, or MtFAAH2b were approximately 66%, 44%, and 44%, respectively (Supplementary Fig. S1; Supplementary Table S2).

Inspection of the transcript expression profiles of MtFAAH1, MtFAAH2a and MtFAAH2b in the Medicago truncatula Gene Expression Atlas "MtExpress"³⁴ showed that MtFAAH1 or MtFAAH2a were highly expressed in different tissues, temperature conditions, time points, abiotic stresses (e.g. N- starvation) and biotic stresses (e.g. exposure to the arbuscular mycorrhizal fungus Rhizophagus irregularis) whereas in the same conditions MtFAAH2b was expressed at a much lower level (Supplementary Fig. S2). Based on these data, we selected MtFAAH1 and MtFAAH2a for further analysis with both computational and biochemical approaches.

Homology models of MtFAAH1 and MtFAAH2a were predicted based on the crystal structure of AtFAAH (PBD: 6DHV) for comparisons of their structural features (Fig. 1b). The quality of the MtFAAH1 and MtFAAH2a models was supported with GMQE (Global Model Quality Estimate) scores (0.90 for MtFAAH1 and 0.80 for MtFAAH2a), QMEANDisCo (Qmean consensus-based distance constraint) scores (0.88 for MtFAAH1 and 0.79 for MtFAAH2a), and Ramachandran scores (Ramachandran favored; 95.59% for MtFAAH1 and 94.31% for MtFAAH2a) (Supplementary Figs. S3, S4; Supplementary Table S3, S4). Both models are shown as homodimers (Fig. 1b) with similar overall predicted content and organization of alpha helices, beta-sheets, turns, and unstructured coils (Supplementary Fig. S5). Both MtFAAH1 and MtFAAH2a have conserved membrane binding caps (MBCs) and membrane access channels (MACs) at their N-termini (Fig. 1b, Supplementary Fig. S6), which are predicted to anchor FAAH to the membrane¹. Like in AtFAAH crystal structure, the MBCs of both MtFAAHs are characterized by a prevalence of hydrophobic residues (Supplementary Fig. S6). As expected, the amidase signature (AS) domain of AtFAAH is characterized by the presence of a Ser-Ser-Lys catalytic triad as in other FAAH family members, and these residues are conserved here in MtFAAH1 and MtFAAH2a (Fig. 1c,d). Altogether, these data highlight conservation of predicted membrane association and active site residues for MtFAAH1 and MtFAAH2a.

Residue differences between the substrate binding pockets of MtFAAH1 and MtFAAH2a

Previously, a report highlighted several key residue differences that altered the predicted shape and properties of the substrate binding pockets (SBPs) of soybean FAAH1 and FAAH2³². In order to corroborate that such findings are conserved in a different legume species, we analyzed the SBPs of M. truncatula FAAHs, MtFAAH1 and MtFAAH2a. Data revealed multiple residue differences especially within the cytosolic access (CAC) and acyl binding (ABC) channels (Fig. 2; Supplementary Table S5). The CAC of MtFAAH1 has several polar (e.g. Thr²⁹⁹, Glu³³⁷) whereas these residues in the CAC of MtFAAH2a are nonpolar amino acids (e.g. Val³⁰⁶, Trp³⁴¹) (Fig. 2a–e). Further, MtFAAH1 has positioned the less bulky residue, Gly³³⁴ in its CAC, this appeared to provide a more opened cavity compared to that of its FAAH2 counterpart (Fig. 2a–e). Conversely, MtFAAH2a has the large aromatic residue, Trp³⁴¹ in its CAC, and this results in a more restrictive CAC (Fig. 2a–e). The ABC of MtFAAH1 is characterized by nonpolar/hydrophobic residues such as Leu⁴⁴¹, Phe⁴⁷⁵, Met⁵³¹ whereas in MtFAAH2a those residues have been replaced by three tyrosine residues (Tyr⁴⁴⁴, Tyr⁴⁷⁷, and Tyr⁵³³) which provide more neutral, polar and aromatic environments (Fig. 2a–e). Further, the tyrosine residues in MtFAAH2a resulted in a more closed, shorter ABC (Fig. 2a–e). By contrast, the residues in MtFAAH1 yielded a more opened ABC, as shown in their aromatic and hydrophobic surface profiles (Fig. 2d,e) more similar to the ACB of AtFAAH1¹. Notably, several of the residues found in the SBPs of MtFAAH1 or MtFAAH2 are conserved among the FAAH1 or FAAH2 groups of multiple legume species (Supplementary Fig. S7). Together, these data suggest that MtFAAH1 and MtFAAH2a have distinct structural and physiochemical properties in their SBPs. Therefore, these different SBPs are likely to accommodate different acylamide structures.

Molecular docking of MtFAAH1 and MtFAAH2a

To test how the differences in the SBPs of MtFAAH1 and MtFAAH2a could alter substrate binding and accommodation, we carried out molecular docking experiments (Fig. 3). To geometrically stabilize both MtFAAH1 and MtFAAH2a prior to the docking experiments, we conducted molecular dynamic simulations (MDS) of the apo forms of MtFAAH1 and MtFAAH2a for 100 ns. The resulting MD trajectories were studied via Root-Mean-Square Deviation (RMSD) and Root-Mean-Square-Fluctuation (RMSF) calculations. Both MtFAAHs appeared stable throughout the entire simulation with RMSD values of approximately 1.5–1.7 Å for MtFAAH1 and 1.8–2.2 Å for MtFAAH2a (Supplementary Fig. S8). The RMSF values showed similar fluctuation profiles for both MtFAAHs, except for a sequence region in MtFAAH2a comprised by residues 366–387 where a higher fluctuation was observed when compared to MtFAAH1 (Supplementary Fig. S8).

For docking studies three structurally different acylamides (Fig. 3a) were compared– NAE18:2, NAE12:0, or p-coumaryl-HL. The MD stabilized MtFAAH1 or MtFAAH2a homology models were docked with these three ligands (Fig. 3b–g). Docking revealed the potential for covalent bonds between the side-chain oxygen (Oγ) of Ser³⁰⁴ or Ser³¹¹ and the carbon of the carbonyl group of the substrates with bond distances that ranged from 2.1 to 2.6 Å. The polar ethanolamine head groups of NAEs (NAE18:2 or NAE12:0) were supported by multiple hydrogen-bonds in the SBPs of both MtFAAH1 (Fig. 3b,d) and MtFAAH2a (Fig. 3c,e); this involves residues such as Gly³⁰¹, Gly²⁵⁴, Thr²⁹⁹ and Asp²⁰⁶ in MtFAAH1, and residues such as Gly³⁰⁸ and Gly³⁰⁹ in MtFAAH2a. Inspection of MtFAAHs bound to p-coumaryl-HL revealed some interesting similarities and differences. Both MtFAAH1 and MtFAAH2a required glycine residues (Gly³⁰¹ for MtFAAH1 and Gly³⁰⁸ for MtFAAH2a) for hydrogen bonding with the head group of p-coumaryl-HL (Fig. 3f,g). However, the homoserine lactone (HL) head group of p-coumaryl-HL was far more supported in the SBP of MtFAAH2a with additional hydrogens bonds mediated by Glu²¹³, Met³⁴⁵ and Val³⁰⁶.Further, MtFAAH1 positioned Thr²⁹⁹ for hydrogen bond interactions with p-coumaryl-HL head group whereas MtFAAH2a had the aromatic side-chain of Trp³⁴¹ and the pyrrolidine loop of Pro³³⁸ for hydrophobic interactions with the HL head (Fig. 3f,g). These data indicate similar organization of the catalytic triad residues to achieve catalysis, and suggest similar and distinct interactions to stabilize the different polar "head regions" of the substrates in the active sites of MtFAAHs.

Analysis of the interactions between the acyl chains of NAEs or p-coumaryl-HL and the residues of the SBPs of MtFAAHs revealed certain similarities and differences. The acyl chain of NAE18:2 is supported in MtFAAH1 by hydrophobic interactions with Leu⁴⁴¹ and an array of three threonine residues (Thr²⁵⁷, Thr²⁵⁸, Thr⁵²⁹). The geometric positioning of these residues in MtFAAH1 appears to provide a flexible environment for accommodation of long acyl chains (Fig. 3b; Supplementary Fig. S9) similar to that reported for AtFAAH1¹. Binding of NAE18:2 in MtFAAH2a was supported by predicted hydrophobic interactions with Thr²⁶⁴, Ile⁴⁴⁷, Ile⁵⁴⁰, and two tyrosine residues, Tyr⁵³³ and Tyr⁴⁴⁴ (Fig. 3c). Compared to MtFAAH1, the surface of MtFAAH2a binding pocket appeared to be shaped differently to accommodate NAE18:2 (Fig. 3c; Supplementary Fig. S9). It appears that the acyl chain of NAE18:2 may be structurally more compressed in MtFAAH2a, presumably, to adjust to a more restrictive-sized SBP in MtFAAH2a (Fig. 3c; Supplementary Fig. S9). The acyl chain of the shorter NAE (NAE12:0) docked in MtFAAH1 or MtFAAH2a is supported in both cases by van der Waals interactions with the aliphatic amino acid isoleucine (Ile⁴⁴⁴ for MtFAAH1 or Ile⁵⁴⁰ for MtFAAH2a) (Fig. 3d,e). Further, MtFAAH2a has positioned the aromatic residue tyrosine (Tyr⁵³³) for interactions at positions C11 and C12 of NAE12:0 tail whereas MtFAAH1 displays Leu⁴⁴¹ for interactions at C10 of the acyl chain of NAE (Fig. 3d,e). Unlike MtFAAH1, MtFAAH2a appears to possess a more aromatic surface to accommodate NAE12:0 into its SBP (Fig. 3d,e; Supplementary Fig. S9). Finally, MtFAAH1 accommodates p-coumaryl-HL by predicted van der Waals interactions between Thr²⁵⁷ and the aromatic tail of the aryl-HL (Fig. 3f) whereas MtFAAH2a, besides the interactions with similar residues (Ile⁴⁴⁷), also has positioned an aromatic residue, Tyr⁴⁷⁷, for additional hydrogen bonding and π-π interactions with the phenolic acyl moiety of p-coumaryl-HL (Fig. 3g; Supplementary Fig. S9). These data suggest that MtFAAH1 or MtFAAH2a can accommodate NAEs or aryl-HLs via residues with distinct physicochemical properties. Indeed, MtFAAH1 appears to generally utilize hydrophobic residues with aliphatic side chains whereas MtFAAH2a consistently relies on neutral residues with aromatic side chains.

Molecular dynamic simulation experiments of ligand bound MtFAAH1 and MtFAAH2a complexes

To further dissect additional residues and mechanisms that could be potentially involved in the accommodation of substrates in MtFAAHs, we carried out molecular dynamic simulation (MDS) on MtFAAH1 or MtFAAH2a docked with NAE18:2 or p-coumaryl-HL during 100 ns (Figs. 4, 5; Supplementary Figs. S10, S11; Supplementary Table S6; Supplementary Videos 1–12). The MDS trajectories for MtFAAH1 and MtFAAH2a revealed that NAE18:2 or p-coumaryl-HL can be accommodated through distinct predicted interactions.

In the SBP of MtFAAH1, NAE18:2 appears to require accommodation of its acyl chain via Thr⁵³⁴ and Met⁵³¹ (Fig. 4a; Supplementary Table S6, Supplementary Video S1, S2). These residues are within a helix region (residues 530 to 536) known to be necessary for the "squeeze" and "lock" mechanism, as reported for AtFAAH elsewhere¹. At 10 and 100 ns, van der Waals interactions were observed between Thr⁵³⁴ and several carbons of the acyl chain of NAE18:2, namely, C14 to C16 and C12–C13, respectively. Additionally, the residue Met⁵³¹ which is located at the end of the helix region, appears to be highly flexible, and could potentially form similar interactions with NAE18:2 at the C10 position, as shown at 100 ns (Fig. 4a). Close inspection at the SBP revealed that the conformation of the acyl tail of NAE18:2 changed over time. Indeed, NAE18:2 tail was highly flexible during the simulation especially from C12 to C18, and oriented itself from C1 to C9 in a straight line at 100 ns (Fig. 4a; Supplementary Video S1, S2). Thus, suggesting a highly flexible SBP suitable for accommodation of acyl amides with long acyl chains. Furthermore, analysis of the MAC revealed an inward movement of the α1 helix (residues 51–58). This motion seems to lead to the closing of the MAC following binding to NAE18:2 (Fig. 4c), which may suggest a "lock" mechanism to confine the substrate in the binding pocket for catalysis (Supplementary Video S3). By contrast, analysis of MtFAAH1 bound to p-coumaryl-HL suggested that this aryl-HL is poorly accommodated in the SBP of MtFAAH1 with few potential interactions (Fig. 4b; Supplementary Video S4, S5). Indeed, the helix region (residues 530–536) did not seem to interact at all with the p-coumaryl-HL substrate. Further, residues such as Met⁵³¹ and Thr⁵³⁴ were far away to interact with the aromatic tail of p-coumaryl-HL (Supplementary Video S5), suggesting protein–ligand interaction(s) in this region of the SPB are highly unlikely for p-coumaryl-HL. Lastly, no dramatic changes were observed in either the conformation of the p-coumaryl-HL structure or in the movement of the α1 helix of MtFAAH1 MAC, resulting in an opened or partially opened MAC during the course of the simulation (Fig. 4b,d; Supplementary Video S6).

Unlike the case with MtFAAH1, the helix region (residues 532–538) of MtFAAH2a showed no dramatic movement toward NAE18:2 substrate upon binding (Fig. 5a; Supplementary Table S6; Supplementary Video S7, S8). Nevertheless, van der Waals interactions were found between Tyr⁵³³ and the acyl chain of NAE18:2 at C13 and C16 (10 ns), and at C11 and C12 positions (100 ns) (Fig. 5a). No other residue in this region seemed to be involved or in close contact with the acyl chain of NAE18:2. Further, the conformation of the acyl chain of NAE18:2 from C1 to C9 appeared to be highly compressed and with few motions over the course of the simulation. This ligand exhibited multiple contortions at different carbon positions of its tail at the end of the simulation (Fig. 5a; Supplementary Video S7, S8). Different from the more open SBP of MtFAAH1, NAE18:2 fit within the smaller SBP of MtFAAH2a seemed to be much more limited and restrictive. A very slight outward movement of the α1 helix (residues 52–59) was recorded during the course of the simulation; however, this movement did not coincide with the opening or closing of the MAC of MtFAAH2a MAC (Fig. 5c; Supplementary Video S9). Binding of MtFAAH2a to p-coumaryl-HL was supported by multiple interactions including the residue Tyr⁵³³ (Fig. 5b; Supplementary Video S10, S11). Over the simulation time, there were consistent close interactions between this tyrosine residue and the aromatic tail of p-coumaryl-HL. These likely π- π interactions could be fixating this ligand in place for catalysis (Fig. 5b; Supplementary Video S11). Furthermore, we also noted consistent hydrophobic interactions between the head HL ring and the residue Trp³⁴¹ located at the cytosolic access channel (CAC) of the SBP of MtFAAH2a (Fig. 5b; Supplementary Video S10). Finally, although some slight outward displacements were detected in the α1 helix (residues 52–59) of the MAC of MtFAAH2a (Fig. 5b), no substantial changes were found and the MAC remained in a closed conformation during the 100 ns simulation (Fig. 5d; Supplementary Video S12).

Altogether these data suggest distinct means for fitting of acylethanolamide substrates into the SBPs of MtFAAH1 or MtFAAH2a enzymes. Our MDS experiments highlight potential residues that could be involved in the preferential accommodation of NAEs or aryl HL substrates by these two enzyme isoforms.

Recombinant production of MtFAAH1 and MtFAAH2a

To test the functional differences between MtFAAH1 and MtFAAH2, we cloned MtFAAHs (Supplementary Fig. S12; Supplementary Table S8) into expression vectors as His-tagged fusions for production of recombinant proteins in E. coli. To purify the recombinant proteins, we utilized immobilized metal (nickel) affinity chromatography (IMAC) and size exclusion chromatography (SEC) techniques (Fig. 6). We observed prominent bands at ≈ 70 kDa representing MtFAAH1 and MtFAAH2a from the IMAC-purified and SEC-purified samples (Fig. 6b,d). These bands were equivalent to the calculated molecular weights (MW) of MtFAAHs proteins. To assess whether the oligomerization states of each purified MtFAAH differ from that of AtFAAH¹, the elution volumes from the chromatograms of gel filtration standards and AtFAAH were used as reference for MtFAAH calculations (Fig. 6a; Supplementary Fig. S13). We observed an average elution volume of 10.08 mL (at pH 9.0) and 10.14 mL (at pH 7.5) for MtFAAH1 and MtFAAH2a, respectively (Supplementary Fig. S13). In solution, the MW of MtFAAH1 was estimated as ≈ 346 kDa and MtFAAH2a as ≈ 371 kDa (Fig. 6c,e; Supplementary Fig. S13). Considering the molecular weights of individual subunits of MtFAAH1 and MtFAAH2 as 69.6 or 70.1 kDa, respectively (Fig. 6; Supplementary Fig. S13), and the reported MW of DDM micelles as ≈ 70 kDa³⁵, we estimated that MtFAAH1 and MtFAAH2a were each purified as tetramers (Supplementary Fig. S13). The uncropped/unedited SDS-PAGE gels used for Fig. 6b,d can be found in Supplementary Figs. S19, and S20, respectively whereas the original gels used for Supplementary Fig. S13 can be found in Supplementary Figs. S21–S23.

Additionally, we utilized liquid chromatography with tandem mass spectrometry (LC/MS/MS) to confirm the identity of purified MtFAAH and AtFAAH recombinant proteins (Supplementary Figs. S14, S15, S16). MtFAAH1, MtFAAH2a, and AtFAAH had coverage values of 77% (43 unique peptides), 37% (17 unique peptides), and 100% (243 unique peptides), respectively. Taken together, these results demonstrate the production and purification of MtFAAH1 and MtFAAH2a proteins, and suggest that both purified MtFAAH protein complexes are larger than the AtFAAH dimers¹, likely organized as homotetramers. The uncropped/unedited gels used for Supplementary Figs. S14–S16 can be found in Supplementary Figs. S24–S26.

Assessment of MtFAAH1 and MtFAAH2 enzyme activities

As an initial characterization step, the enzymatic activities of MtFAAH1 or MtFAAH2a towards NAE12:0 were measured at different pH or temperature conditions (Supplementary Fig. S17). SEC-purified proteins were incubated with 100 µM NAE12:0 in reaction buffers with different pH (6.0, 7.0, 8.0, and 9.0) or temperature (20, 30, 40, and 50 °C) conditions. A flexible, fluorescamine-based assay was utilized for detection of reaction products, as reported elsewhere^22,36. Enzyme activity was calculated as μmol of ethanolamine produced per minute per milligram of protein based on a standard curve for ethanolamine. MtFAAH1 activity was the highest at pH 8.0 or 9.0 (Supplementary Fig. S17), whereas MtFAAH2a operated maximally at pH 7.0 or 8.0 (Supplementary Fig. S17). Notably, the activity of MtFAAH2a dramatically decreased at pH 9.0. Temperate dependencies were similar for both MtFAAH1 and MtFAAH2a, with optimal enzyme activities at temperatures between 30 and 40 °C (Supplementary Fig. S17). GC–MS further confirmed conversion of NAE12:0 into its corresponding free fatty acid (FFA) (Supplementary Fig. S18) product for both MtFAAH enzymes and the positive control, AtFAAH; boiled ("denatured") enzymes were used as negative controls and showed no conversion of substrate to product. The identity of NAE12:0 peak was confirmed by retention time and diagnostic ions (Supplementary Fig. S18) as reported elsewhere²³. The identity of FFA 12:0 (dodecanoic acid) was confirmed by diagnostic ions (Supplementary Fig. S18) and via NIST library comparison (≈ 98% match). In the reactions with boiled MtFAAH1, MtFAAH2a, or AtFAAH, only the peaks corresponding to NAE12:0 substrate appeared in the chromatograms (Supplementary Fig. S18).

To evaluate the progress of MtFAAH1 or MtFAAH2a purification, enzyme activities were measured in cell lysates, IMAC-, and SEC-purified fractions towards NAE12:0 (100 µM) (Supplementary Table S7). Purification levels and yields were calculated relative to the crude lysates. MtFAAH1 and MtFAAH2a were estimated to be enriched over the cell extracts by 1162-fold and 3263-fold, respectively. SDS-gel electrophoresis confirmed the purity of each protein in the SEC-fractions (Fig. 6; Supplementary Fig. S13).

Given the key residue differences between the SBPs of MtFAAH1 and MtFAAH2a, and the fact that docking experiments suggest differential accommodation of N-acyl substrates, we hypothesized that such differences could result in distinct substrate preferences for MtFAAH1 and MtFAAH2a. Increasing concentrations of NAE12:0, NAE18:2, NAE-9-HOD, OdDHL, OtDHL, p-coumaryl-HL, p-coumaryltyramine, and affinin were incubated with purified MtFAAH1 or MtFAAH2a at 30 °C for 30 min (Fig. 7). Fluorescamine-based assays were used for detection of reaction products. Initial velocity (V) of the reactions are presented as μmol of amine produced per minute per milligram of protein. The Kcat/Km ratios derived from Michaelis and Menten plots were used to compare the catalytic efficiencies of the enzymes for their substrates (Fig. 7). MtFAAH1 had the highest Kcat/Km ratios for long chain NAEs (NAE-9-HOD or NAE18:2), or AHLs (OdDHL or OtDHL). Then, compared to OtDHL, Kcat/Km ratios decreased by ≈ 2- to threefold for NAE12:0 or affinin. The Kcat/Km values for p-coumaryl-HL or p-coumaryltyramine were the lowest among all substrates tested for MtFAAH1 (≈ 7- to 12-fold lower compared to NAE-9-HOD) (Fig. 7a,c). By contrast, MtFAAH2a had the highest Kcat/Km values for affinin, p-coumaryltyramine, or p-coumaryl-HL. Compared to p-coumaryl-HL, the Kcat/Km of MtFAAH2a decreased by ≈ twofold for NAE12:0, OdDHL or OtDHL. Notably, MtFAAH2a had the lowest Kcat/Km ratios for NAE18:2 or NAE-9-HOD (≈ 4- to ninefold lower than that of affinin) (Fig. 7b,d). These data indicate that under these in vitro conditions, MtFAAH1 prefers lipophilic substrates with longer acyl moieties whereas MtFAAH2a performs best with substrates with short or aromatic acyl moieties (summarized in Fig. 8).