We obtained a protein-family hidden Markov model (HMM) profile for the FMO-like family (PF00743 v22) from the Protein Family (Pfam) database (https://pfam.xfam.org/), and conducted a local HMM search (HMMER v3.3.2; http://hmmer.org/) with significance parameters of E-value < 1e^-2 against the International Wheat Genome Sequencing Consortium (IWGSC) fully-sequenced reference genomes for version RefSeqv2.1, to identify all putative wheat flavin-containing monooxygenase genes (TaFMOs) for the cultivar ‘Chinese Spring’. Results from HMMER were further consolidated with the use of query-based searches in both the online database Wheat Proteome (Duncan et al., 2017) with the search term ‘flavin-containing monooxygenase’ and the search term ‘flavin monooxygenase’ in the UniProt database (https://www.uniprot.org/). Similarly, all FMO candidate genes from other plant species spanning a vast range of green plants and algae, were found through a HMM-based search. Table 1 lists the total number of FMOs per plant species. We used the summary of relationships of major angiosperm lineages available at NCBI Common Tree builder (https://www.ncbi.nlm.nih.gov/Taxonomy/CommonTree/wwwcmt.cgi) to display the distribution of FMOs in vascular plants ( Figure 1 ; for sources of sequence and protein data and gene search details, see Supplementary Tables S1 , S2 ; Appendix S1 , Supplementary Figure S1 ).

The full wheat cDNA, CDS, and peptide sequences of candidate TaFMO genes identified were retrieved from IWGSC (The International Wheat Genome Sequencing Consortium (IWGSC) et al., 2018). Gene Ontology (GO) terms were retrieved from the Uniprot database (https://www.uniprot.org/; accessed January 2023). The structure of exons and introns were determined using the CDS and gene sequences of TaFMOs analyzed with the Gene Structure Display Server (GSDS) version 2.0 (gsds.cbi.pku.edu.cn; Hu et al., 2015). A small subset of putative TaFMOs were represented by splice variants and included for downstream analyses, on the basis that each variant could encode for fully functioning TaFMO proteins acting in different tissues, conditions, and developmental stages.

To assess whether the enzymatic activity of each candidate TaFMO could be carried out, the presence of three previously well-described FMO motifs (Eswaramoorthy et al., 2006; Mishina and Zeier, 2006; Schlaich, 2007; Thodberg and Jakobsen Neilson, 2020) known to be crucial for cofactor binding and required for proper FMO function—the FAD- and NAD(P)H-binding motifs crucial for oxygenation activity, the FMO-identifying motif facilitating in NAD(P)H-binding at the core pocket of the enzyme, and the ATG-containing motif thought to govern hydroxylation activity—were assessed with multiple sequence alignments (MSA) using MAFFT with the E-INS-I setting; see Appendix S2 (Katoh et al., 2019). Candidates that did not possess one or all motifs were retained in downstream phylogenetic analyses but flagged as ‘non-canonical’ (nc) FMOs with the assumption that oxygenation activity may be compromised, as we used conservation of all motifs as a proxy for predicting FMO function. We predicted the putative protein structure of all TaFMOs using the Protein Homology/Analogy Recognition Engine V 2.0 (Phyre²; Kelley et al., 2015) and then visualized them with PyMOL (Version 2.0 Schrödinger and DeLano, 2020). We used the Simple Modular Architecture Research Tool (SMART; https://smart.embl.de/; Letunic et al., 2021) program to gather protein domains for display ( Figure 2 ). Possible transmembrane domains and signal peptides for all TaFMO were predicted using the programs DeepTMHMM and SignalP - 6.0 (Hallgren et al., 2022; Teufel et al., 2022).

We inferred a maximum likelihood (ML) phylogeny to define relationships between total validated TaFMOs and validated FMOs from the model plant A. thaliana, the cereal crop Oryza sativa subsp. japonica, and sister group of all other angiosperms, Amborella trichopoda. We selected a red algae having a single FMO as an outgroup. We first aligned full-length FMO peptide sequences using MAFFT with the E-INS-i setting (Katoh et al., 2019). We validated the full-length alignment of the total FMOs across plant species using the software HOMO v2.0 (Jermiin, 2017) by using it to determine whether all FMO sequences met the phylogenetic assumption of evolution under reversible, stationary, and globally homogenous conditions (Jermiin, 2017; Jermiin et al., 2020). Following this, poorly aligned regions of the MSA were masked using the AliStat program (Wong et al., 2014) to ensure downstream analysis focused on regions where amino acid residues were readily alignable across sequences (see Appendix S2 ) (Wong et al., 2014). We then manually inspected alignments using Mesquite v 3.7 (Maddison and Maddison, 2023) and adjusted accordingly. Lastly, we used IQ-TREE v 2.2.0 to infer the ML tree based on amino acid alignments; we inferred the optimal substitution model using ModelFinder (JTT+I+G4), considering the Bayesian Information Criterion, BIC (Kalyaanamoorthy et al., 2017), and calculated branch support from 1000 standard bootstrapping replicates (Felsenstein, 1985). The tree figures were made with iTOL (Letunic and Bork, 2021).

We constructed separate ML phylogenies between all TaFMO and FMOs from closely-related species in the grass order, Poales—specifically, H. vulgare (barley), T. urartu (red einkorn wheat) and O. sativa ssp. japonica (rice) to improve our understanding of the relatedness of TaFMO in the three major clades A, B, and C ( Supplementary Figures S2 - 5 ). For detailed notes on alignment and modelling, see Appendix S2 .

We retrieved chromosomal locations of all TaFMOs in wheat from the RefSeqv2.1 gff3 file. We inferred homeologs based on well-supported phylogenetic clustering; separate sub-phylogenies were inferred for the three major partitions (clades A, B, and C) and a 70% bootstrap support value was used as a threshold to acknowledge clades with moderate support. Gene duplication and syntenic gene blocks of all TaFMOs (including ‘nc’ TaFMOs) were inferred through the MCSanX algorithm (Wang et al., 2012) via TBtools software v1.116 (Chen et al., 2022) and NOTUNG (Chen et al., 2000) to resolve tandem duplications and relationships between TaFMO homoeologous groups. Genes with no discernible relatives were denoted as ‘orphan’ TaFMO ( Tables 2 , 3 ). A comprehensive synchronized map of all TaFMO homeologs ( Figure 3 ) connected by designated group colors assigned in Supplementary Figures S2 - 4 was generated using the shinyCircos software in RStudio (Yu et al., 2018; R Core Team, 2021). See Appendix S3 for more details. A de novo search for putative conserved protein motifs among all TaFMOs, HvFMOs, OsjFMOs, AtFMOs, AmTriFMOs, and PpFMOs was conducted with MEME Suite (Bailey et al., 2015): see Appendix S4 .

We searched promoter regions 1.5 kb upstream from the start codon of validated TaFMOs for plant cis-acting regulatory elements (PlantCARE: http://bioinformatics.psb.ugent.be/webtools/plantcare/html/; Lescot, 2002) and visualized putative cis-acting regulatory elements (CAREs) governing the different clades of TaFMOs ( Supplementary Tables S3 , 4 , Appendix S5 ). We also compiled a list of putative transposable elements (TEs) detected by the CLARI-TE program (https://github.com/jdaron/CLARI-TE) either spanning introns and exons of TaFMOs, or flanking TaFMOs within a 2 kb region on either side ( Table 4 ), from the TE annotation file provided in IWGSC RegSeqv2.1 (International Wheat Genome Sequencing Consortium et al., 2018; Zhu et al., 2021). The TE positions flanking or inside genes were transposed into TaFMO gene structure analysis. Refer to Appendix S6 for details.

Using public transcriptomic data across a range of conditions, we extracted candidate TaFMO gene expression information, obtaining transcript values in FPKM for five general tissues (root, stem, leaves, spike, and grain) and proteome data (peptide spectral counts) for different tissues from the developmental atlas in the Wheat Proteome database (Duncan et al., 2017). We acquired transcript data for a broad range of different conditions in values of log₂ transcripts per million (tpm) for TaFMO expression under various abiotic stressors (cold, drought, heat) and biotic stressors (i.e., infection by Septoria blotch, Fusarium graminearum, or Puccinia striiformis f.sp. tritici) from the Wheat Expression Database (Ramírez-González et al., 2018) in order to survey the differential expression of TaFMO candidates, and infer their possible biological functions (see Supplementary Tables S5 - 8 for an overview of studies and sources used).

The FMO gene family is highly expanded in the vascular plants, as previously reported (Eswaramoorthy et al., 2006; Mishina and Zeier, 2006; Schlaich, 2007; Thodberg and Jakobsen Neilson, 2020). Figure 1 exhibits the general number of potential FMO protein-encoding genes among select vascular plant reference genomes. The three most commonly present types of domain architectures that represent these plant FMOs are a single FMO-like domain (FMO-like), two consecutive FMO-like domains (FMO-like x 2), and a pyruvate redox 3 (Pyr3) domain. Some FMOs are predicted as fusion proteins, such as found in both wheat (Ta) and barley (Hv) with a tyrosine protein kinase conjugated to an FMO, alluding to a possible function in environmental sensing and downstream signaling coupled with oxidation processes (Tichtinsky et al., 2003; Shumayla and Upadhyay, 2023). The phenomenon of FMO fusion proteins does not present for lycophytes, moss, or algae, where total numbers of FMOs in the genome are substantially reduced relative to flowering vascular plant species. The total number of FMOs in plant genomes also appears to be largely reflective of species ploidy levels for recently diverged taxa, as exemplified in Triticum urartu (2n, 41 putative TuFMOs), T. dicoccoides (4n, 84 putative TdFMOs) and T. aestivum (6n, 170 putative TaFMOs). However, in plants considered diploid, the number of FMOs varies substantially, such as in Arabidopsis (29 AtFMOs), rice (31 OsjFMOs), and barley (59 HvFMOs), up to 64 potential FMOs in sunflower (H. annuus), which has double the amount of most diploid species in the eudicots (lower, green clade in Figure 1 ). We report several additional FMO candidates for rice compared to a recent report of 28 OsjFMOs by (Gaba et al., 2023), and for barley previously reported having 41 HvFMOs (Thodberg and Jakobsen Neilson, 2020), highlighting the complexity of conducting genome-wide searches for highly expanded gene families.

An HMM search conducted against the RefSeqv2.1 wheat reference genome for cultivar ‘Chinese Spring’ yielded a total of 171 high confidence (HC) TaFMO genes, and 82 low confidence (LC) TaFMO genes. We selected 170 HC TaFMO genes including those deemed ‘unmapped’ to specific chromosomes after consolidation with other search methods (see Supplementary Figure S1 , Appendix S1 for additional information). A total of 34 TaFMO candidate genes (40 including transcript splice variants) are denoted as ‘non-canonical’ (nc), based on alterations in—or the absence of—protein motifs previously reported as crucial for proper FMO function, discussed below ( Table 2 ). The ‘nc’ TaFMOs possessing alterations in important protein motifs but were still found in synteny with other TaFMO homoeologous triads, were retained in downstream analyses; further experimental validation is required to prove their functional and biological relevance as an FMO in wheat. Other more highly truncated ‘nc’ TaFMOs unlikely functional in the canonical sense, could still highlight trends in motif diversification/loss for certain subclades when included in a phylogenetic analysis. The same logic was applied to sort for FMOs of Arabidopsis thaliana (At), Oryza sativa ssp. japonica (Osj), Hordeum vulgare (Hv), and Triticum urartu (Tu).

A maximum-likelihood phylogeny was constructed with protein sequences representing the total 170 TaFMO gene candidates (198 independent splice variants); FMOs from Amborella trichopoda (Amtri), rice Oryza sativa sp. Japonica (Osj), and A. thaliana (At) were included to probe for phylogenetic partitioning of TaFMOs among flowering plants. A red microalgae (G. sulphuraria) served as an outgroup, whose genome encodes for only one known FMO ( Figure 1 ). We observed three major phylogenetic groupings, indicated as A, B, and C in Figure 2 .

Domain architecture analysis ( Figure 2 , circle I) shows that Group A FMOs mostly share the same domain architecture type as the FMO from the outgroup red algae (‘FMO-like x 2’), with two TaFMOs as exceptions having the ‘pyruvate redox 2’ architecture (TraesCS3B03G0026700.1, TraesCS3B03G0026800.2). Group B (purple) and Group C (green) FMOs form a clade across vascular plants with Group B mostly composed of an ‘FMO-like’ architecture, whereas group C has a mix between the ‘FMO-like’ and ‘pyruvate redox 3’ architectures. The distribution of total FMOs across an array of vascular plants (or within one species) has been documented by Thodberg and Jakobsen Neilson (2020); Gaba et al. (2023), and Yoshimoto et al. (2015), and the phylogenies presented in these studies also find Group B (AtFMO1-containing clade) and Group C (‘YUCCA’ clade) to be more closely related to each other than to Group A (the AtFMO GSOX-containing clade).

We found that TaFMOs span all seven (and unmapped) chromosomes and sub-genomes (A, B, and D) derived from progenitor grass species Triticum urartu, Aegilops speltoides, and Aegilops tauschii, respectively (Ramírez-González et al., 2018; Zhu et al., 2021). Chromosomes 2, 4 and 7 are enriched with TaFMOs in the telomeric regions ( Figure 3 ). Translocations of TaFMOs are inferred to have occurred most frequently between chromosomes 4A, 7A, and 7D, which are reported to be hot spots for homoeologous gene translocation events for many genes in wheat (Zhou et al., 2020) ( Appendix S3 ). Other prominent translocations of TaFMO occurred between chromosomes 4A, 5A, and 5B in Group C TaFMO, and between chromosomes 5A, 4B, and 4D in Group A TaFMO ( Figure 3 , Table 2 ).

Tables 2 , 3 summarize the synteny status of Groups A, B, and C TaFMO homeologs and homeolog sub-genome ratios (denoted as a TaFMO ratio of 1: 1: 1 from A, B, and D sub-genomes), respectively. Group A TaFMOs show substantially less syntenic triads with only 8.2% of triads in Group A exhibiting an even ratio of 1: 1: 1; even more, around 42% of triads in Group A were missing either an A-, B-, or D-copy, a higher frequency than the 13.2% expected for all wheat proteins in the genome ( Table 3 ). Group C, subclade C1, showed the most conserved ‘clean’ homoeologous triads (56.3%) such that there were rarely any singleton/orphan or ‘nc’ TaFMOs, and each TaFMO triad had corresponding orthologues from other grass species ( Supplementary Figure S4 ); the high evolutionary conservation of these TaFMO triads is perhaps due to functional constraints influencing gene retention. In contrast, TaFMOs in subclade C2 (along with barley FMOs) appear to have large, uneven homeolog expansions via tandem duplications and intra- or interchromosomal dispersion of genes ( Supplementary Figure S4 ).

In Group B, 36.4% of TaFMOs belong to even triads which more closely resemble the anticipated percentage (35.8%) of all predicted wheat proteins belonging to even homoeologous triads ( Table 3 ) (Ramírez-González et al., 2018). Group B appears to have many instances of sub-genome homeolog expansion and uneven tandem duplications ( Table 2 ). Subclade B1 appears to contain three pairs of TaFMO WGD homeologs that underwent multiple duplications, with more HvFMO orthologues present than OsjFMO and TuFMO, ( Supplementary Figure S3 ), while subclade B2 shows FMO expansions for all grass species surveyed, but none for Arabidopsis ( Supplementary Figure S3 ).

Such divergent patterns of gene retention and expansion among homoeologous triads between the three major groups of TaFMOs might implicate differing biological roles (Birchler and Yang, 2022). make a point that genes involved in defense mechanisms are more likely to diverge after duplication events; the retention of the active sites for some of the highly duplicated Group B triads could infer a role in defense-related functions.

We detected fifteen of the most conserved protein motifs in FMOs across Arabidopsis and the grasses rice (Osj), barley (Hv), red einkorn (Tu) and wheat ( Supplementary Figure S5 ). FMOs in plants are reported to contain three crucial protein motifs essential for ‘canonical’ FMO activity; these are the FAD-binding motif (GxGxxG), the NAD(P)H-binding motif (GxGxxG), and the FMO-identifying motif [FxGxxxHxxxY/F; (Eswaramoorthy et al., 2006; Hansen et al., 2007; Huijbers et al., 2014)]. While the FAD- and NAD(P)H-binding sites are active sites surrounding the enzyme pocket where hydroxylation takes place, the FMO-identifying motif is considered a linker region between the two active sites (Malito et al., 2004). The last and less widely talked about motif is the ‘F/LATGY’ motif, more recently referred to in some studies as the ‘ATG-containing’ or ‘TGY’ motif (Yoshimoto and Saito, 2015; Gaba et al., 2023); this motif is thought to be a crucial factor governing N-hydroxylation (Stehr et al., 1998; Fennema et al., 2016).

The FAD-binding motif at the N-terminus is very well conserved with regards to three glycine residues across all groups ( Figures 4A, B ). However, the residues surrounding this motif for Group A and subclade B2 from Group B is recorded to be distinct from that in subclade B1 and Group C ( Figures 4A, B ; Supplementary Figures S2 - 4 ); the FAD-binding site variant 1 (for Group A and B2) and the other variant are highlighted in orange in Figure 4A, B . An extra copy of the FAD-binding variant 1 also exists between the NAD(P)H-binding site and the ATG-motif in most of the examined grass species (Hv, Tu, Ta) in subclade C2, but not for rice. This may indicate a novel function for FMOs specific to Triticeae. The NAD(P)H-binding motif across all FMOs ( Supplementary Figure S5 ) is well-conserved for the first glycine residue, but not for the other two glycine residues (Gxgxxg; Figure 4A ). The FMO-identifying motif for FMOs in Group A has the first residue swapped from the canonical F (Phenylalanine) for W (Tryptophan), which is the case for most of the plant species examined ( Figure 4A ). Additionally, the ATG-containing motif for Group A FMOs across all flowering plants surveyed are highly conserved, represented as HCTGY or YCTGY. The differences of the Group A ATG-motif and FMO-identifying motifs relative to Groups B and C may allude to specificity for S-oxygenation of compounds and specialized metabolites yet uncharacterized in cereals, where Group A FMOs from A. thaliana (FMO-GSOX enzymes) are well-characterized for their S-oxygenation activity during synthesis of specialized metabolites (Hansen et al., 2007). The ATG-containing motif is most varied for Group B FMOs, where in addition to the canonical ‘(F/L)ATGY’ residues reported by (Schlaich, 2007), other variations are present (LATGF, FLATGF, FATGY, LATGY, FGTGF) in the B2 expanded subclade comprising other grass species (Tu, Hv, Osj). By contrast, Group C FMOs mostly present with ‘LATGY’ motifs (and some ‘MATGY’ motifs in barley and wheat) in subclade C1 and ‘FATGY’ motifs in subclade C2 ( Figure 4A ).

Beyond the documented protein motifs discussed above, one novel motif is reported for each Group A and Group C FMOs (referred to here as motif-A1 and motif-C1, respectively; see Supplementary Figure S5 for these and other motifs referred to in this section). Meanwhile, Group B FMOs have four clade-specific motifs (motif-B1-4) ( Figure 4 ). Groups A, B, and C FMOs all share one novel motif each with each other (motif-AB1, motif-AC1, motif-BC1). Additionally, there is a novel motif common to almost all FMOs located at the N-terminus following the FAD-binding site for most FMOs, reported here as motif-ABC1 ( Figure 4A ). A look at predicted protein models highlights that many of these novel motifs may present as-yet-unclassified linker regions contributing to the integrity of the enzyme, or motifs that govern substrate specificity and binding affinity ( Figure 4B ). We hypothesize that differences in tertiary protein structures, including these clade-specific motifs, likely play a crucial role in substrate specificity (Li et al., 2008; Zhu et al., 2019). While no plant FMOs have been crystalized to date, the use of 3D protein homology modelling and ligand-docking simulations—in corroboration with functional analyses—may shed some light on which compounds TaFMO, and other cereal FMOs, may be acting upon.

Group A FMOs contain all the Arabidopsis FMO-GSOX (glucosinolate oxidase) genes. These FMOs have been experimentally validated to participate in glucosinolate (GSL) biosynthesis by catalyzing the S-hydroxylation of methylthioalkyls to methylsulfinylalkyls, where GSLs are a class of specialized sulphur-containing metabolites involved in plant-herbivory defenses (Hansen et al., 2007; Li et al., 2011; Kong et al., 2016; Cang et al., 2018; Schall et al., 2020). GSLs are predominantly found in the mustard seed family (Brassicaceae), although there is evidence that 500 neighboring eudicot species outside this family may also contain one or more of the 120 documented GSLs (Flamini, 2012). The possible metabolites that these Group A FMOs help modify in cereals are not well-understood, as cereals are not reported to produce GSLs (Hansen et al., 2007; Thodberg and Jakobsen Neilson, 2020). More recently, other A. thaliana FMOs have been documented to synthesize trimethylamine N-oxide (TMAO) (Fennema et al., 2016), the levels of which are increased in plants under low temperatures, drought, and high salt stress conditions (Catalá et al., 2021). Group A FMOs segregate into two major subclades, A1 and A2 ( Figures 2 , Supplementary Figure S2 ). While little is known for Group A FMOs in barley and red einkorn, it is reported that five rice FMOs from subclade A1 (Os01t0368000-nc, Os02t0580600-nc, Os07t0111700, Os07t0111900, Os07t0112000) are co-expressed with WRKY13 (a transcription factor involved in pathogen defense) upon attack by rice sheath-infecting fungi (John Lilly and Subramanian, 2019). For TaFMOs in subclade A1, the literature supports involvement in root drought tolerance, tiller growth, and nutrient accumulation in the grain ( Table 2 ; Supplementary Figure S6 ). The rice FMO Os10t0553800 in subclade A2 may play an important role in rice germination and seedling establishment during floods, via epigenetic methylation responses to anaerobic conditions (Castano-Duque et al., 2021). By contrast, several TaFMOs in subclade A2 are implicated in disease resistance against both biotrophic (Navathe et al., 2022) and necrotrophic pathogens (Nussbaumer et al., 2015), in addition to root drought tolerance ( Supplementary Figure S6 ). The amino oxidase (AOX) protein domain which is seen in almost all FMOs in Group A at the N-terminus but rarely for Group B (and which is missing from Group C), is larger in subclade A2 AtFMOs (~150 amino acid residues), but not for wheat (~60 residues) or other plant FMOs ( Figure 2 ). The AOX domain in Group A TaFMOs appears to encompass the motif-A1 mentioned previously unique to Group A and may participate in substrate binding specificity for novel compounds in wheat.

Group C contains the “YUCCA” (YUC) genes that are known to participate in auxin (idole-3-acetic acid, IAA) biosynthesis (Kendrew, 2001; Cao et al., 2019), where auxin is involved in many plant development processes (Schlaich, 2007; Chandler, 2015). We identified 74 TaFMOs in Group C, which can be split into two main subclades, C1 and C2. Several TaFMOs in subclade C1 are implicated in the response to drought stress ( Table 2 ; Supplementary Figure S7 ), found to be upregulated during root drought treatment (Grzesiak et al., 2019), or associated with an increased reactive oxygen species (ROS, H₂O₂) burst in response to root drought conditions (Kamruzzaman et al., 2022). In subclade C2, several TaFMOs are implicated in plant development, including in grain (Mangini et al., 2021) and seed development (Li et al., 2014), and tiller number increase (Yu et al., 2021). To date, only three homoeologous genes involved in seed development have been functionally characterized via gene cloning (Li et al., 2014). A catalogue of 63 ‘TaYUC’ gene candidates involved in auxin biosynthesis by Yang et al. (2021) included several TaFMO genes outside Group C (dispersed throughout subclade B2; Supplementary Table S4 ), hinting towards a flexibility in TaFMO involvement in auxin biosynthesis-regulated plant development. In the absence of a full gene family survey and experimental characterization, adopting a more wheat-specific nomenclature for the FMO gene family via the guidelines put forward by the Wheat Initiative community (Boden et al., 2023), may minimize confusion on gene names in wheat to streamline future research efforts.

Group B FMOs can be split into two distinct subclades, B1 and B2 ( Figure 2 ). This group contains the Arabidopsis AtFMO1 and AtFMO2 genes. AtFMO1 hydroxylates the N atom of an L-lysine catabolic product to form N-hydroxylated pipecolic acid (NHP), crucial for establishing systemic acquired resistance (SAR) to combat microbial pathogen invasions (Conrath, 2006; Mishina and Zeier, 2006; Chen et al., 2018; Hartmann et al., 2018), and a lack of NHP (due to defects in its biosynthesis) results in immunocompromised Arabidopsis plants. Various studies have recently contributed to better understanding the mechanism of NHP-mediated SAR regulation in other plants, including cereal crops, by searching for functional orthologs to AtFMO1 and their possible involvement in disease resistance (Holmes et al., 2019; Lenk et al., 2019; Schnake et al., 2020; Vlot et al., 2021; Zhang et al., 2021).

In subclade B1, AtFMO1 and AtFMO2, the only Group B FMOs in Arabidopsis, form a clade with 18 TaFMOs, where only three candidates are ‘nc’ ( Table 2 , Figure 4B ). All FMOs from subclade B1 possess the most minimal number of known protein domains via SMART analysis, whereas multiple protein domains are present for TaFMO from all other clades ( Figure 2 ). Of note, AtFMO1 harbors an additional K-oxygenase domain (IPR025700) in the first half of its protein sequence, which is not seen in any of the B1 TaFMOs or AtFMO2 but is present in most B2 TaFMOs and TaFMOs in Groups A and C. The K-oxygenase domain is involved in siderophore biosynthesis, and in ornithine and lysine hydroxylation (Visca et al., 1994; Krithika et al., 2006); therefore, presence of this K-oxygenase may play a critical role in hydroxylation of L-lysine and/or L-ornithine derived compounds.

While subclade B1 TaFMOs show little to no transcriptional activity in the conditions surveyed, certain subclade B2 TaFMOs exhibit higher transcriptional activity ( Figures 2 , 5 – III/IV). The TaFMO TraesCS5D03G0799800.1 in subclade B2 (B2β-3b) was detected as the most transcriptionally upregulated “FMO1-like” TaFMO in response to treatment of wheat coleoptiles with NHP (Zhang et al., 2021). Two homoeologous triads in B2α-2 ( Table 2 ) show transcriptional upregulation in root, stem, and leaf tissues ( Figure 2 ), and peptide abundance of these triads was also detected in a wide variety of wheat tissues ( Figure 5 – II). These TaFMOs in B2α-2 undergo transcriptional activation in wheat seedlings during infection by the fungal pathogens Zymoseptoria tritici (Zt) (Yang et al., 2013; Rudd et al., 2015) and Puccinia striiformis f.sp. tritici (Pst) (Cantu et al., 2013; Dobon et al., 2016), but have very little to no transcriptional activation during older developmental stages of wheat, such as seen in controls during infections by Fusarium graminearum (Fg) in the wheat grain head (Schweiger et al., 2016) and Magnaporthe oryza (Mo) at the grain-filling stage (Islam et al., 2016) ( Figure 5 - III).

One of the B2α-2 homoeologous triads (TraesCS5A03G0836200.1, TraesCS5B03G0873900.1, TraesCS5D03G0799100.1) shows greater upregulation of transcription after Pst infection at one day post-infection (dpi), which is sustained through 11 dpi; TraesCS5A03G0836200.1 is also reported by (Vikas et al., 2022) in a genome-wide association study (GWAS) to be likely involved in seedling resistance to the biotrophic fungal pathogen Puccinia triticina (Pt), the wheat leaf rust pathogen closely related to Pst. Another TaFMO in sub-clade B2β-4c(2), TraesCS7B03G1279500.1, might be involved in Puccinia graminis f.sp tritici (Pgt, stem rust) resistance (Sahu et al., 2021). In a QTL analysis by (Matros et al., 2017), the two triads in subclade B2α-2 were also detected as significantly enriched in SNPs associated with ornithine metabolism in wheat. Ornithine is an amino acid bearing great resemblance to lysine and has functional significance in plant abiotic stress tolerance (Kalamaki et al., 2009) through its role in osmoregulation during drought and salinity stress (Roosens et al., 1998; Xue et al., 2009). These two triads are among the only Group B TaFMOs to be transcriptionally responsive in the abiotic conditions presented, downregulated during drought stress, heat stress, and combined drought + heat stress at one hour, but bounce back to nearly control levels after six hours of treatment for certain splice variants ( Figure 5 – IV) (Liu et al., 2015). Transcripts of the TaFMO triad in B2α-1 (except for the ‘nc’ 5D homeolog) have been detected in wheat roots ( Figure 2 , Table 2 ) (Grzesiak et al., 2019). reported that genes of this triad were upregulated in response to water deprivation in wheat root tissues, though no upregulation of transcription in leaf tissues was detected in the drought study by Liu et al. (2015) ( Figure 5 – IV), highlighting a need for careful interpretation of tissue-specific expression patterns of TaFMOs. Furthermore, a GWAS analysis indicates that the B-copy gene of this triad, TraesCS5B03G0877300.1, influences root-growth and biomass yield-related traits (Zhao et al., 2023). Trends for other candidates of interest are found in Appendix S7 .

In summary, several lines of evidence hint at roles in responsiveness to various microbial diseases for several TaFMOs in subclade B2. Additionally, certain subclade B2 TaFMOs may operate during plant development (tiller increase) and drought, heat, and salinity tolerance. Thodberg et al. (2020) showed that a novel Group B FMO in fern (FOS1) participates in both the N-hydroxylation of a novel class of metabolites not previously reported in vascular plants and in the biosynthesis of cyanogenic glycoside. Likewise, the exploration and characterization of these B2 TaFMOs may identify novel functions and metabolites towards plant development, biotic, and abiotic stressors.

Untranslated regions in transcripts (UTRs) have been reported in plants as important regulatory regions often harboring important cis-acting regulatory elements (CAREs) that influence gene expression, modulate translational efficiency, and increase the coding capacity of genes that have different splice variations (Mignone et al., 2002; Srivastava et al., 2018). We found a wide variety of CAREs in the promoter regions 1.5 kb upstream of the start codon of TaFMO genes ( Figures 6 ; Supplementary Figures S8 , S9 ). The presence and sizes of 5’- and 3’-UTRs among TaFMOs varied; we observed that for all TaFMOs, the 3’-UTRs were longer than the 5’-UTRs, when UTRs were present ( Figures 7 ; Supplementary Figures S10 , S11 ). Srivastava et al. (2018) reported elongated UTRs (particularly 3’-UTRs) in rice compared to Arabidopsis. In wheat, 3’-UTRs have also been reported to have a critical involvement in mediating drought stress (Ma et al., 2023) and in establishing resistance against stripe rust (Zhang et al., 2019), possibly through regulating mRNA stability. Of note, all the tandemly duplicated wheat TaFMOs in subclade C2 lack any UTRs ( Supplementary Figure S11 ) and show virtually no (or low) transcript responsiveness to the biotic and abiotic conditions explored ( Supplementary Figure S7 – III, IV), with minimal presence of expressed peptides detected in various tissue types.

Focusing on the Group B TaFMOs, analysis of the CAREs revealed involvement in responsiveness to major phytohormones and conditions ( Figure 6 ; Appendix S5 ). Jasmonic acid (JA)-responsive CAREs were present in all TaFMO promoters at one or multiple sites; JA is a phytohormone known for involvement in many plant processes ranging broadly from photosynthesis to heat stress-responsiveness (Per et al., 2018), but most widely documented for defense against necrotrophic pathogens (Qi et al., 2016) and induced systemic resistance (Heil, 2002). In wheat, JA-signaling has also been implicated in seedling salt-stress tolerance (Qiu et al., 2014). Other notable phytohormone-responsive CAREs in Group B are for salicylic acid (SA), gibberellic acid (GA) and ethylene (ETH) ( Supplementary Tables S3 , 4 , Figure 6 ). SA is a major phytohormone involved in SAR and pathogen defense signaling pathways (Dey et al., 2014; Ding et al., 2018). SA-responsive CAREs were found in higher abundance and closer to the start codons of the triads from B2α-2 ( Figure 6 , Supplementary Table S4 ), whose genes were more transcriptionally active in Zt- and Pst-infected wheat ( Figure 5 – III); this may indicate that number and location of these CAREs influence which TaFMOs are activated under pathogen stress. GA is a phytohormone that was documented in wheat for its involvement in enhancing disease resistance against F. graminearum infection, by modulating both primary and specialized metabolism during early plant defense signaling (Buhrow et al., 2021); GA-responsive CAREs are found predominantly in subclade B2, further supporting that TaFMOs in this subclade are likely involved in microbial disease resistance. Auxin (IAA)-responsive CAREs are found to be abundant among many Group B TaFMOs ( Supplementary Table S4 ), challenging the idea of Group C TaFMOs—the ‘YUCCA’ FMO clade—being largely responsible for auxin biosynthesis and growth regulation in plants for underreported crop species (Kendrew, 2001).

We explored transposable element (TE) distribution in or near the identified TaFMO genes and affecting intron-exon structure to find support for relationships among TaFMO family members ( Figure 7 , Supplementary Figures S10 , 11 ; Appendix S6 ). Focusing again on the Group B TaFMOs, we observed much variation in the number and spatial distribution of introns and exons; high similarity of gene structures was found among certain syntenic triads highlighting distinct TaFMO subfamilies that possibly remain conserved for functional importance. In contrast, extensive intron-exon shuffling following gene duplication events (Kolkman and Stemmer, 2001; França et al., 2012) may indicate lower selective pressure and neofunctionalization, such as in the case for subclades B2β-4c1 and B2β-4c2 ( Figure 7 ).

Nearly all TaFMOs (90%) are flanked by TEs within a 2 kb region upstream and/or downstream of the start or stop codons ( Figure 7 , Table 4 ) in accordance with a report of TEs representing over 80% of the wheat genome (Choulet et al., 2014; Bariah et al., 2020). 44.7% of all TaFMO genes show disruption by TE insertion, mostly occurring inside introns, inside 5’- or 3’-UTRs, or, very rarely, inside exons ( Figure 7 ). TE insertions are present in 56.1% of Group B TaFMO, with the largest TE interruption in TraesCS2B03G0193800 increasing its total gene length to upwards of 29 kb ( Figure 7 ). Most TE disruptions in subclade B2 occur in the B- and D-copy homeologs and appear to be correlated with differential expression patterns among homoeologous triads. Varying locations of TE insertions in the highly tandemly duplicated B-copy TaFMOs in B2β-4c(1) correlate with no transcriptional expression compared to the A- and D-copies, and only an A-copy TaFMO in B2β-4c(2) lacking TE insertions was transcriptionally active ( Figure 5 – III, IV; Figure 7 ). In contrast, one of the triads in B2α-2 (TraesCS5A03G0836200, TraesCS5B03G0873900, TraesCS5D03G0799100) shares the same TE insertion in its three members, which are all more transcriptionally active than gene members in the other B2α-2 triad lacking this TE insertion, particularly upregulated towards Pst infection and cold stress responses ( Figure 5 – III, IV; Figure 7 ).

Yang et al. (2021) reported on the gene structure of a candidate ‘YUCCA’ FMO in wheat, TaYUCCA5-D (TraesCS1D03G0254900.1, Group B), containing a first intron size of approximately 13 kb. We now know this is due to a TE insertion ( Figure 7 ), underscoring a need for expanding on factors driving gene structure diversity in large gene-family analyses. We see that TE insertions in the intronic and 3’-UTR regions may play a role in influencing gene expression of triads on a sub-genome level, possibly leading to eventual neofunctionalization of FMO family members. Several studies indicate that TEs are important drivers in polyploid plant evolution (Vicient and Casacuberta, 2017; Gill et al., 2021), such as in wheat where distinct TE families are correlated with up- or down-regulation of genes in their vicinity (Bariah et al., 2023). Thus, exploring classes of TE elements that lie in or around TaFMOs could aid in predicting gene expression regulation in the different wheat sub-genomes.