Evolution of Linoleic Acid Biosynthesis Paved the Way for Ecological Success of Termites

Abstract Termites are dominant animals of tropical terrestrial ecosystems. Their success is due to their eusocial organization as well as their ability to digest dead plant tissues. While being extremely abundant, the termite diet is poor in crucial nutrients, such as fatty acids. Linoleic acid (LA) is a precursor for many vital biomolecules, and most animals depend on its dietary supply. Termites count among the exceptions known to produce LA de novo, presumably via the action of an unknown Δ12 fatty acyl desaturase (FAD) introducing the second double bond into monounsaturated oleic acid. Here, we search for the evolutionary origin of LA biosynthesis in termites. To this end, we compile the repertoire of FAD homologs from 57 species of termites and their closest relatives, the cockroaches, analyze FAD phylogeny, and identify a potential Δ12 FAD branch, which arose through duplication of a likely Δ9 FAD. We functionally characterize both paralogs and identify the Δ9 activity in the ancestral FAD-A1a and the Δ12 activity responsible for LA biosynthesis in FAD-A1b. Through the combination of homology modeling and site-directed mutagenesis, we pinpoint structural features possibly contributing to the distinct functions, regiospecificities, and substrate preferences of the two enzymes. We confirm the presence of both paralogs in all 36 studied species of the Blattoidea lineage (Blattidae, Lamproblattidae, Cryptocercidae, and termites) and conclude that we identified an evolutionary event important for the ecological success of termites, which took place in their cockroach ancestors roughly 160 My and remained conserved throughout termite diversification into 3,000 extant species.


Introduction
Termites (Isoptera) are the oldest social insects and dominant animals in tropical and subtropical ecosystems. With quantitative abundance estimated to reach nearly 50 Mt of carbon, they represent roughly one-quarter of the total biomass of all terrestrial arthropods (Tuma et al. 2020). The spectacular ecological success of termites is undoubtedly due not only to their eusocial organization (Wilson 1992) but also to their dietary specialization on lignocellulose material originating from dead plant tissues. The ability to digest the recalcitrant but ubiquitous lignocellulose substrates enabled termites to become major players in carbon recycling in warmer regions of the world with important impact on the structure of landscapes and soils; termites are thus sometimes referred to as ecosystem engineers (Jouquet et al. 2011).
While being rich in energy stored in lignocellulose macromolecules, the termite diet is poor in many essential nutrients. Therefore, prior to becoming the main decomposers of cellulose materials, termites needed to evolve adaptations to cope with potential dietary constraints. The termite diet is particularly deficient in fatty acids (FAs), including polyunsaturated FAs (PUFAs), which belong, in the vast majority of animals, among essential nutrients obtained exclusively via dietary supply. PUFAs are key precursors for multiple critical biomolecules within the primary metabolism (triacylglycerols, eicosanoids, phospholipids, and others) that are used for energy storage, as components of cellular and subcellular membranes, and as signaling molecules within intracellular pathways (Pelley 2012). In insects, PUFAs have yet another important function as a part of secondary metabolism, because they serve as precursors for cuticular hydrocarbons, which protect insect bodies against desiccation and also give rise to many insect pheromones, including the MBE greatly diversified repertoire of sex pheromones in moths (Tillman et al. 1999;Symonds and Elgar 2008).
A key step in the biosynthesis of PUFAs is the formation of linoleic acid (Z9,Z12-18:COOH, hereafter also LA) via the introduction of a second double bond in the ubiquitous monounsaturated oleic acid (Z9-18:COOH, OA). Animals were long thought to be incapable of LA biosynthesis and to rely purely on its dietary supply. However, the past few decades firmly established that some animal lineages can synthesize LA de novo using their own enzymatic machinery. LA production has been experimentally demonstrated in nonarthropod taxa such as nematodes or snails and slugs (Gastropoda), as well as in noninsect arthropods such as mites (Acari) (Rothstein and Götz 1968;Weinert et al. 1993;Aboshi et al. 2013;Shimizu et al. 2014). In insects, LA biosynthesis is missing in basal orders (e.g., Zygentoma, Ephemeroptera, and Odonata) and basal Polyneoptera (Dermaptera and Plecoptera). In more derived orders, de novo LA production has been recorded in almost 20 species of Orthoptera, Blattodea, Hemiptera, Hymenoptera, Neuroptera, and Coleoptera (reviewed in Malcicka et al. 2018). However, these lineages also contain species in which the LA biosynthesis has been disproved, suggesting a complex scenario of its evolution, accompanied by multiple origins and/or losses. Interestingly, LA synthesis has not been demonstrated in Lepidoptera and Diptera, two modern and very diversified insect orders representing over one-quarter of insect species richness (Grimaldi and Engel 2005), even though both orders, and especially moths , extensively use PUFA-derived pheromones and are able to desaturate fatty acyl substrates at various positions.
The patchy distribution of LA biosynthesis also applies for cockroaches and termites, together making up the order Blattodea. The American cockroach Periplaneta americana was among the first insects with confirmed LA biosynthesis (Louloudes et al. 1961), which has since been also observed in a few other cockroaches from the Blattoidea lineage, including all three studied species of termites, while being seemingly absent in the Blaberoidea clade (Bade 1964;Mauldin et al. 1972;Blomquist et al. 1982;Cripps et al. 1986;de Renobales et al. 1987;Malcicka et al. 2018). Therefore, the most parsimonious hypothesis based on the current knowledge and modern molecular phylogeny of Blattodea (Evangelista et al. 2019) would be that LA biosynthesis evolved once in basal Blattoidea, a lineage encompassing Blattidae, Lamproblattidae, and Cryptocercidae as a sister group to termites, and termites themselves.
Unsaturated FAs arise via desaturation steps catalyzed by two types of iron-containing fatty acyl desaturases, that is, soluble acyl-acyl carrier protein desaturases restricted to higher plants and structurally unrelated membrane fatty acyl-coenzyme A (CoA) desaturases (hereafter referred to as FADs) ubiquitous in endoplasmic reticulum membranes of Eukaryota (Shanklin and Cahoon 1998;López Alonso et al. 2003). FADs are oxygen-and NADH-dependent enzymes and belong to the superfamily of membrane diiron-containing enzymes, sharing a conserved tripartite histidine-rich motif, coordinating two iron atoms and four transmembrane domains as common features (Sperling et al. 2003). The current knowledge suggests that eukaryotic FADs evolved through an ancestral gene diversification of a canonical Δ9 FAD (López Alonso et al. 2003), which introduces a Z9 double bond to saturated stearoyl-CoA to give rise to oleoyl-CoA. OA is a universally present FA and a dominant component of the FA pool also in insects including termites (Carter et al. 1972;Stanley-Samuelson et al. 1988;Šobotník et al. 2006). Because of their crucial role in generating FA-derived pheromone diversity, insect FADs received much attention, especially in moths, beetles, and hymenopterans, with several dozen being functionally characterized and analyzed with respect to underlying gene evolution. The repertoire of insect FADs consists of multiple gene clusters (six subfamilies according to Helmkampf et al. 2015), subjected to lineage-specific expansions via duplications and neofunctionalizations on the one hand and losses on the other hand (Knipple et al. 2002;Roelofs and Rooney 2003;Helmkampf et al. 2015;Groot et al. 2016;Tupec et al. 2017). The diversity of insect FAD products results not only from different substrate specificities and stereo-and regiospecificities of individual FADs, but also from the versatility of FAD functions: some of them exhibit substrate-dependent specificities, conjugase activities, and other substrate-driven features that can be involved, for example, in the sequential synthesis of multiple double bonds in PUFAs (Moto et al. 2004;Matoušková et al. 2007;Serra et al. 2007;Wang et al. 2010;Ding et al. 2011;Buček et al. 2015).
The first insights into FAD functional determinants have been obtained using inspections of aligned sequences followed by domain swapping and/or site-directed mutagenesis experiments (Vanhercke et al. 2011;Meesapyodsuk and Qiu 2014;Buček et al. 2015;Ding et al. 2016), and recently, they have also been inferred from crystal structures of two mammalian FADs (Bai et al. 2015;Wang et al. 2015;Shen et al. 2020). In C18-specific Δ9 FADs, which are presumably ancestral to other FADs, the enzyme specificity is thought to be maintained by two major structural features: acyl register and tunnel capping (Meesapyodsuk and Qiu 2014;Bai et al. 2015;Wang et al. 2015;Wilding et al. 2017;Cai et al. 2020). In closer detail, the C1 acyl moiety of the substrate forms a hydrogen bond with the acyl register residue (e.g., tryptophan in Δ9 FADs) located near the opening of the substrate tunnel (Bai et al. 2015;Wang et al. 2015;Wilding et al. 2017;Cai et al. 2020). This binding ensures that the desaturation occurs regiospecifically at a certain distance (ΔX) from the acyl moiety. Additionally, the desaturase controls the overall length of the substrate entering the substrate tunnel (i.e., its chain length specificity) by a bulky capping residue located in the distal section of the substrate tunnel (Bai et al. 2015;Cai et al. 2020). When the capping residue appears closer to the active site, it effectively decreases the tunnel length, and as a result, the desaturase prefers shorter substrates, for example, C14 and C16 (Bai et al. 2015;Nagao et al. 2019). Alternatively, MBE the absence of any capping residue can allow substrates as long as C26 to be desaturated (Meesapyodsuk and Qiu 2014;Cai et al. 2020). Multiple studies have tried to pinpoint the exact mutations in FADs leading to a shift in substrate specificity or regiospecificity (Vanhercke et al. 2011;Bai et al. 2015;Shi et al. 2015;Cai et al. 2020). Uncovering the precise structural feature which determines the course of substrate desaturation still remains challenging as it usually depends on multiple residues. Moreover, there are no published structures of desaturases with specificities other than Δ9 (such as Δ11, Δ12, Δ14, ω3, or ω6), which limits the search only to homologous models. However, a few examples show a profound shift in the FAD specificity due to a single residue change in the substrate tunnel (Meesapyodsuk and Qiu 2014;Buček et al. 2015Buček et al. , 2020Ding et al. 2016;Cai et al. 2020).
The previous search for insect Δ12 FADs responsible for de novo LA formation has been motivated by both the crucial significance of LA in general metabolism and its precursor role in the production of pheromones and defensive chemicals. LA-producing Δ12 FADs have so far been identified in four species. They were first recorded in 2008 in the house cricket Acheta domesticus and the red flour beetle Tribolium castaneum (Zhou et al. 2008) and later in the soldier beetle Chauliognathus lugubris (Haritos et al. 2012) and the parasitic wasp Nasonia vitripennis (Semmelmann et al. 2019). While the former two species showed a single Δ12 FAD each, the latter two express two LA-producing paralogs within species-specific FAD expansions. Besides the potential role in primary FA metabolism, the products of some of these FAD paralogs serve as precursors for the biosynthesis of specialized secondary metabolites used for defense (C. lugubris) and communication (N. vitripennis). Most of the listed Δ12 FADs show amino acid sequence similarities with conserved Δ9 FADs and at times even retained residual Δ9 activity (Zhou et al. 2008), suggesting their origin via duplications and neofunctionalizations of the ancestral Δ9 FADs producing OA and/or palmitoleic acid. At the same time, phylogenetic analysis of insect FADs clearly indicates independent origins of Δ12 FADs in individual taxa, as Δ12 FADs may even have evolved from Δ9 FADs situated in different insect FAD subfamilies (see fig. 1).
In this study, we search for the evolutionary origin of de novo LA biosynthesis in termites. We test the hypothesis that LA biosynthesis evolved within the Blattoidea clade

Phasm Phasm
FIG. 1. Diversity of FAD genes in Blattodea and other insects. Phylogeny of FADs inferred from 102 protein sequences of 23 insect species, including 9 Blattodea (6 termites and 3 cockroach species). Nonblattodean species were selected to include all previously identified insect Δ12 desaturases producing LA (marked with arrows). Labeling of FAD subfamilies according to Helmkampf et al. (2015). Two subclasses identified for Blattodea within the A1 subfamily are marked as A1a and A1b. The topology and branching supports were inferred using the IQ-TREE maximum likelihood algorithm with the LG+R7 model. The bootstrap values were calculated using ultrafast bootstrap approximation (UFBoot) from 1,000 replicates. Accession numbers of the analyzed FAD sequences are provided in supplementary table S1, Supplementary Material online.

MBE
and that termites thus inherited this capacity from their cockroach ancestors. For this goal, we compile a repertoire of FAD homologs from 57 species of cockroaches and termites, analyze their phylogenetic patterns, identify a candidate FAD branch with potential Δ12 activity which arose through duplication of a likely Δ9 FAD within Blattoidea, and confirm the presence of the duplicated paralog in all subclades of Blattoidea (Blattidae, Lamproblattidae, Cryptocercidae, and termites). We functionally characterize both paralogous FADs using yeast expression system and identify the Δ9 activity in the ancestral FAD and Δ12 activity in its paralogous copy, giving rise to LA from OA. In parallel, using AlphaFold2-derived 3D models of the substrate tunnel and site-directed mutagenesis, we identify a putative acyl register residue needed for Δ9 activity and tunnel capping residues responsible for differential substrate selectivity of the two enzymes. We conclude that we identified an evolutionary event potentially crucial for the dietary specialization of termites allowing them to survive on a poor lignocellulose diet, which took place roughly 160 My and remained conserved in its main functional features throughout termite evolution.

Repertoire of Blattodean FAD Genes
In the first step, we studied the overall diversity of FAD-coding genes in a subset of three cockroach and six termite species in the context of FADs identified in 14 selected nonblattodean insects, including all those known to possess a functional LA-producing Δ12 FAD. A series of BLAST searches in public databases and our in-house sequencing data generated a set of 198 FAD sequences from the 23 species (supplementary table S1, Supplementary Material online). In the phylogenetic reconstruction ( fig. 1), these sequences were distributed among six FAD subfamilies (A1, A2, and B-E) defined by Helmkampf et al. (2015). LA-producing Δ12 FADs previously identified in A. domesticus, N. vitripennis, C. lugubris, and T. castaneum were situated in three different subfamilies (A1, A2, and B), in most cases within species-specific FAD expansions, suggesting their independent evolution at low taxonomic levels. In most blattodean species, we retrieved eight to ten reliable FAD sequences per species. Only seven were identified in the termite Embiratermes neotenicus and as many as 12 in the cockroach Blattella germanica. Blattodean FADs were represented in all subfamilies as one or two monophyletic Blattodea-specific branches, each mainly containing phylogenetically organized orthologs of all or most studied species, indicating the ancient origin of individual FADs.

Identification of a Candidate Δ12 Desaturase FAD-A1b
A closer look at the phylogeny revealed an interesting branching pattern within the FAD-A1 subfamily ( fig. 1). In this subfamily, the blattodean monophylum contained two FAD sequences for all species except B. germanica, which was represented by the basally situated single sequence. The most parsimonious scenario of the origin of this pattern would be an ancient duplication of the ancestral FAD-A1 and subsequent conservation of its paralogous copy during diversification of blattoid cockroaches and termites. Indeed, A1 topology mostly supported this view since, for most species, each of the two putative copies was situated on one of two divergent subbranches.
The inspection of blattodean FAD-A1 amino acid sequences (figs. 2A and S1, Supplementary Material online) and AlphaFold2-derived 3D model of the substrate tunnel ( fig. 2B) confirmed the hypothesized evolutionary history and split the identified sequences into two groups. All nine species, including the basal B. germanica, possessed one FAD copy sharing the tunnel architecture reminiscent of the known C18-specific Δ9 FADs (e.g., Cai et al. 2020), particularly the tunnel length defined by bulky Tyr82 residue (all amino acid numberings according to the sequence of Prorhinotermes simplex; figs. 2 and S1, Supplementary Material online) and putative acyl register mediated by Trp236. This suggests that this gene, hereafter referred to as FAD-A1a, is a Δ9 FAD and represents the ancestral A1 paralog, because the same sequence and structure features are also shared by the basal B. germanica FAD-A1 as well as the A1 sequence in the sister group, the cricket A. domesticus (figs. 1 and 2A).
In contrast, the paralogous A1 sequences, hereafter referred to as FAD-A1b, have several substitutions in the substrate tunnel that are shared by all nine species. Most importantly, the model indicates tunnel shortening by substituting FAD-A1a Gly86 with bulkier isoleucine residue, and the hypothesized acyl register at Trp236 is lost due to substitution with leucine. These combined substitutions could contribute to a possible shift of desaturation regiospecificity from ω9 to ω6/7, as needed for OA conversion into LA ( fig. 2A and B). Additional frequent substitutions in the internal region of the substrate tunnel were Ala266Thr, Asn122Thr, and Asp130Asn. In the external region, we detected a few more frequent (Arg47Pro, Leu51Met, and Phe52His) or systematic (Leu159Met) substitutions ( fig. 2B).

FAD-A1a Is a Δ9 Desaturase with a Broad Range of Substrates
Having identified FAD candidates for the ancestral Δ9 and derived Δ12 desaturase function, we decided to functionally characterize FAD-A1a and FAD-A1b from the cockroach P. americana and from three species of termites using heterologous expression in desaturase-deficient yeast, supplemented with monounsaturated FAs to ensure their survival. The selection of termite species represented three families and covered the phylogenetic diversity of Isoptera: Zootermopsis nevadensis (Archotermopsidae), P. simplex FAD-A1a showed broad substrate preferences, because its desaturation activity was also observed for palmitic acid, giving rise to Z9-16:COOH, and myristic acid, leading to the formation of Z9-C14:COOH. Results of assays #1-3 suggested the substrate preferences to be C18 > C16 > C14, with conversion rates of 95%, 29%, and 5%, respectively (figs. 4B, 4C, S2, S3, and S5, Supplementary Material online). Likewise, FAD-A1a showed catalytic activity also with saturated FAs longer than C18 since we observed exclusive peaks of monounsaturated fatty acyl methyl esters (FAMEs) having 20, 22, and 24 carbon atoms (supplementary fig. S2, Supplementary Material online). The former was identified as Z9-20:COOMe, whereas in the latter two, the double bond position could not be identified. The substrate preferences were ranked as C18 > C20 > C22 > C24.

FAD-A1a Rescues the Desaturase-Deficient Yeast Cultures
To address the effect of Δ9 desaturation activity of FAD-A1a on the survival of the desaturase-deficient elo1Δole1Δ yeast strain, we performed a rescue experiment with P. simplex FAD-A1a and FAD-A1b, the negative control, and the positive control (Δ9 FAD from the bumble bee Bombus and FAD-A1b desaturases from five blattodean species compared with previously functionally characterized Δ9 and Δ12 FADs from other insects and with FAD-A1 from the cricket A. domesticus (numbered according to the P. simplex FAD-A1a sequence). Arrowheads highlight three positions in the internal region of the substrate tunnel that consistently distinguish the blattodean FAD-A1a and FAD-A1b proteins and that may have an impact on substrate specificity and desaturation regiospecificity (especially positions 86 and 236) based on previous data on FADs. Fully conserved residues are marked by dark shading and residues with similar physicochemical properties by light-gray shading. Sequences were aligned using the MUSCLE v5 algorithm. Complete protein sequences are aligned in supplementary figure S1, Supplementary Material online. (B) 3D model-based diagrams of the substrate tunnels of FAD-A1a and FAD-A1b in the termite P. simplex, fitted with their putative substrates, that is, stearoyl-CoA for the presumed Δ9 activity of FAD-A1a and oleoyl-CoA for the hypothesized Δ12 activity of FAD-A1b. Internal region of the substrate tunnel (right part of the model) and external region (left part). Gray font represents residues fully conserved in the eight studied Blattodea having FAD-A1b in figure 1, and black font denotes residues with some substitutions. Background shading marks the residues consistently substituted and highlighted in the alignment in (A). Hydrogen bonding to the acyl register residue Trp236 is indicated by an arrow, whereas the absence of such bonding with Leu233 is indicated by a cross. AS, active site.

MBE
lapidarius, Blap_NPVE). The results, shown in figure 4A, clearly prove the ability of FAD-A1a and Blap_NPVE to rescue the cultures deficient in Z9-unsaturated FAs, in contrast to the negative control and FAD-A1b cultures. This indicates that FAD-A1a efficiently supplies OA and other Z9 unsaturated FAs to the cultures, and that FAD-A1b has a different function.
Mutation of Gly86 (FAD-A1a) to Ile86 (FAD-A1b) Shortens the Substrate Tunnel Length and Shifts the Desaturation Closer to ω-End of the Substrate To address the predicted impact of the substrate tunnel capping residues on the differential function of FAD-A1a and FAD-A1b, we performed functional assays #6 and #7 comparing the activity of wild-type FAD-A1a from P. simplex with its Gly86Ile mutant, that is, a substitution systematically occurring in FAD-A1b ( fig. 4B, 4C, and supplementary table S2, Supplementary Material online). The substitution of glycine residue with the bulky isoleucine residue resulted in a dramatic drop in conversion rates of stearoyl and palmitoyl chains into corresponding Z9 fatty acyls (95% to 2% and 29% to 1%, respectively). By contrast, the conversion of myristoyl chain into myristoleoyl chain (Z9-14:) increased from 5% to 16%. This suggests that Ile86 effectively shortens the substrate tunnel, and in combination with the presumed substrate-binding residues (namely, the acyl register Trp236), it limits the binding of longer fatty acyls (C16 and background analytes exclusive analytes

MBE
C18), while being preferentially accessible to shorter substrate chain lengths. Such spatial limitation then leads to a shift in the position of the ω-end of the accepted substrate closer to the active site, that is, from ω-9 optimized for stearoyl substrate in wild-type FAD-A1a to ω-5 (myristoyl) in FAD-A1a-86Ile. This is in line with the prerequisite shift in the desaturation position of C18 substrate from ω-9 in FAD-A1a to ω-6 in FAD-A1b.
Trp236 in FAD-A1a Is Responsible for Its High Δ9 Activity In another mutagenesis experiment, we tried to verify our model predictions attributing to Trp236 the role of acyl register residue, presumably responsible for the proper localization of the C 9 -C 10 bond respective to the active site, giving rise to the observed highly specific Δ9 activity of FAD-A1a. A comparison of wild-type PsimFAD-A1a with its Trp236Leu mutant revealed a major decrease in Δ9 desaturation activity, manifested by the drop in conversion rates of C14, C16, and C18 saturated substrates to Z9 homologs, that is, from 5% to 0, from 29% to 2%, and from 95% to 18%, respectively. This clearly shows that Trp236 is crucial for the high efficiency of wild-type FADA1a, especially with its preferred C18 substrate.
Termite FAD-A1b Is a Δ12 Desaturase Specialized in the Production of LA from OA FAD-A1b showed a different regiospecificity and substrate preferences from FAD-A1a. In cultures supplemented with OA, FAD-A1b of all four studied blattodean species acted as a Δ12 FAD and converted OA into LA, as evidenced by the presence of an abundant peak of Z9,Z12-18:COOMe (figs. 3B, S3, S5, and S6, Supplementary Material online). We did not observe any other C18 unsaturated product. Thus, FAD-A1b was revealed to be the LA-producing enzyme predicted by our 3D model.
When offered Z9-16:COOH, all four studied FAD-A1b enzymes produced a trace peak of a 16:2 FAME, whose

MBE
double bond localizations could not be assigned due to very low amounts of the analyte (figs. 3B and S3, Supplementary Material online). Interestingly, while this trace compound was the only C16 desaturation product in the three termite species, PameFAD-A1a from P. americana produced two additional C16 peaks, found in all out of the five independent cultivations. These were identified as FAMEs of two stereoisomers of a 9,11-diunsaturated We did not detect any desaturation product across the range of the naturally present saturated FAs, suggesting that FAD-A1b requires monounsaturated substrates (figs. 3, S2, S3, S5, and S6, Supplementary Material online), which is in line with the observed low viability of FAD-A1b-containing yeasts in a rescue experiment without the supplementation with unsaturated FAs (fig. 4A).
Taken together, these observations indicate that FAD-A1b evolved a relatively strict substrate specificity and regioselectivity for OA conversion into LA in termites while having a broader range of possible substrates and products in the cockroach P. americana by effectively accepting also C16 monounsaturated substrate (Z9-16: COOH).

Additional Unsaturated Products Due to Yeast Elongase Activities
In addition to the desaturation products of FAD-A1a and FAD-A1b, we observed a range of small peaks of unsaturated FAMEs corresponding to elongation products of yeast FA elongases. They had a double bond at Z11 (sometimes even Z13) and arose from Z9 precursors (figs. 3, S2-S4, Supplementary Material online), as confirmed in assay #5 using deuterium-labeled Z9-18:COOH tracer and yielding labeled Z11-20:COOMe and Z13-22:COOMe (supplementary fig. S7, Supplementary Material online). This indicates the activity of yeast elongase 2 and/or 3, since elongase 1 has been knocked out in the used strain.
Gene, Transcript, and Protein Structure Confirm the Common Evolutionary Origin of FAD-A1a and FAD-A1b We studied the gene, transcript, and protein structure of FAD-A1a and FAD-A1b using the genomic assembly of the termite P. simplex ( fig. 5A-C). We identified two genomic loci corresponding to FAD-A1a and FAD-A1b, respectively. Both genes consist of six exons, five of which are protein coding and share highly similar structures with a high level of homology and fully conserved exon junctions. Likewise, the protein structures of FAD-A1a and FAD-A1b are highly homologous and contain all functionally important FAD domains. These observations independently confirm the common evolutionary origin of FAD-A1a and FAD-A1b and their paralogous relationship.

FAD-A1a and FAD-A1b Show a Ubiquitous Expression in Termite Castes and Tissues Suggesting Their Participation in Primary Metabolism
The comparison of transcript abundances of FAD-A1a and FAD-A1b estimated from RNA-Seq analysis in the termite P. simplex revealed a ubiquitous expression of both genes in all four studied caste phenotypes (workers, soldiers, primary kings, and primary queens) and both studied tissues (head and abdomen with removed digestive tube). As shown in figure 5D and supplementary table S3, Supplementary Material online, FAD-A1a had consistently higher expression than FAD-A1b, roughly by one order of magnitude. In most castes, expression of both genes was slightly higher in abdominal tissues than in heads, and we did not observe any significantly caste-biased upregulation in either of the two genes. Altogether, these results show that FAD-A1a and FAD-A1b expressions are not restricted to any particular caste or organ, suggesting the participation of both genes in primary FA metabolism rather than a derived role in secondary metabolism, such as pheromone or defensive compound biosynthesis.

FAD-A1b Arose via Duplication of FAD-A1a and Remained Conserved over 160 My
Following the discovery of the Δ12 FAD responsible for LA biosynthesis in termites and cockroaches, we decided to map in detail the suggested pattern of FAD-A1b evolution via duplication of FAD-A1a. We performed phylogenetic reconstruction including FAD-A1 protein sequences from 57 Blattodea, that is, 28 termite species from 7 families and 29 cockroach species from both Blaberoidea and Solumblattodea clades, with the beetle T. castaneum as an outgroup (supplementary table S4, Supplementary Material online). The resulting phylogenetic tree, depicted in figure 6A, unambiguously confirmed the hypothesized evolutionary scenario. FAD-A1a sequences were retrieved in all 57 included blattodean species, whereas the FAD-A1b sequences were found exclusively in the monophyletic branch of Blattoidea within Solumblattodea, encompassing Blattidae, Lamproblattidae, Cryptocercidae, and termites. It was present in all 36 Blattoidea species, including all 28 termites, while being absent in all three species from the sister family Corydiidae and the 18 studied Blaberoidea. A closer look at the protein alignments revealed that the two residues experimentally confirmed as important for Δ9 activity of FAD-A1a are systematically substituted during FAD-A1b evolution across Blattoidea (97% for Gly86Ile and 100% for Trp236Leu, supplementary table S5, Supplementary Material online).
Tangible functional differences between FAD-A1b of termites and that of the cockroach P. americana prompted the question what the PameFAD-A1b sequence features are that may be responsible for this difference, and whether they represent an isolated peculiarity of P. americana or a common characteristic of Blattidae or Blattidae + Lamproblattidae that would reflect the evolutionary distance between these cockroach families and the monophylum Cryptocercidae + termites. Indeed, the FAD-A1b protein phylogeny revealed sequence differences between termites + Cryptocercus and the remainder of cockroaches, manifested by the long branch separating termites + Cryptocercidae, indicating that termite ancestors have acquired a considerable number of sequence modifications after their separation from Lamproblattidae ( fig. 6A). A detailed analysis of the amino acid sequences of the FAD-A1b substrate tunnel supports this view and splits the studied species into two groups, that is, Blattidae + Lamproblattidae versus all termites + their sister cockroach genus Cryptocercus. All Blattidae + Lamproblattidae possess Val238 residue inside the tunnel cavity, inherited from FAD-A1a, while in termites + Cryptocercus, this amino acid is substituted with threonine (27 out of 29 species). In the external region of the substrate tunnel, all Blattidae + Lamproblattidae have Tyr228 from FAD-A1a substituted

Discussion
In the present study, we attempted to reconstruct the evolutionary history of FAD genes underlying the ability of termites (Isoptera) to convert OA into LA, which is crucial for their survival while being scarce in their specialized lignocellulose diet. In the quest for the responsible Δ12 FAD, we identified a candidate duplication within the FAD-A1 subfamily. The ancestral gene, FAD-A1a, is present in all studied species of Blattodea (29 cockroaches + 28 termites) as well as in other representatives of Polyneoptera and belongs among Δ9 FADs as judged from functionally important residues of the substrate tunnel. In contrast, the newly evolved FAD-A1b is restricted to the monophyletic clade Blattoidea, encompassing Blattidae, Lamproblattidae, Cryptocercidae, and termites, while being absent in all other 21 studied cockroaches from the sister lineages Corydiidae and Blaberoidea. FAD-A1b was distributed in a one-to-one orthologous pattern in all 36 studied representatives of Blattoidea and revealed conserved modifications in the substrate tunnel that we predicted to allow for Δ12 desaturase activities. A common origin of FAD-A1a and FAD-A1b was also apparent from their gene structures mapped on the genome assembly of the termite P. simplex, in which both paralogs shared a high level of homology and full conservation of exon-intron boundaries.
Functional characterization of FAD-A1a and FAD-A1b originating from the cockroach P. americana and three

MBE
termite species from different phylogenetic positions using desaturase-deficient yeasts confirmed the in silico predicted functions of both enzymes. FAD-A1a is a Δ9 FAD introducing a Z9 double bond into a broad range of saturated FAs (C14-C24), with preference for stearic acid. In contrast, FAD-A1b acts as a Δ12 FAD converting OA into LA and has a relatively strict substrate specificity for OA in termites, while acting as a potentially multifunctional enzyme in the cockroach P. americana by also producing substantial quantities of conjugated FAs Z9,E11-16:COOH and Z9,Z11-16:COOH from Z9-C16:COOH. Whether these FAs are native to P. americana and what would be their biological significance are unknown.
All these observations bring us to conclude that the ability of extant termites to synthesize LA evolved together with FAD-A1b roughly 160 My in the common ancestor of Blattoidea, was inherited by termites, and remained conserved across 140 My of their evolution and diversification into 3,000 extant termite species (Krishna et al. 2013;Buček et al. 2019;Evangelista et al. 2019). The universal presence of complete FAD-A1b transcripts, high level of conservation of functionally important amino acid residues across the phylogenetic diversity of termites, and ubiquitous expression in different castes and body parts invite us to speculate that FAD-A1b has a significant role in termite primary metabolism. Historical focus on insect FADs involved in the biosynthesis of FA-derived pheromones has led to functional characterizations and evolutionary analysis of multiple tens of FADs with various substrate specificities and stereo-and regiospecificities, especially in numerous lepidopterans and a few hymenopteran (Knipple et al. 2002;Roelofs and Rooney 2003;Helmkampf et al. 2015;Groot et al. 2016) and dipteran species (Dallerac et al. 2000;Legendre et al. 2008;Fang et al. 2009). In addition to gaining important insights into pheromone evolution, this search was motivated by the applied potential in controlling moth pest species and biotechnological applications (Tupec et al. 2017). The amassed body of information contrasts with the scant knowledge on FADs in other insect taxa and on FADs involved in primary metabolism leading to the formation of PUFA as crucial building blocks for a variety of biomolecules. It is also the case of LA-producing Δ12 FADs, which have only been characterized in four species, in spite of the complex pattern of their repeated independent evolution from the ancestral Δ9 FADs, suggested by the few available identifications (Zhou et al. 2008;Haritos et al. 2012;Semmelmann et al. 2019) and knowledge on LA biosynthesis distribution across Insecta (Malcicka et al. 2018). These reports indicate a notable plasticity in LA biosynthesis evolution and de novo Δ12 FAD occurrence at low taxonomic levels of individual species or genera, even though this image may be biased by low sampling in the previous studies. In contrast, using extensive mining of termite and cockroach sequences, we report here the origin of a Δ12 FAD via an ancient Δ9 FAD duplication and neofunctionalization taking place deep in the history of Blattodea and its conservation over a long evolutionary period with a single paralog per species and with conserved sequence patterns. The distribution of FAD-A1b in cockroaches and termites corroborates the previous observations on blattodean LA biosynthetic capacities restricted to Blattoidea species (Louloudes et al. 1961;Bade 1964;Mauldin et al. 1972;Blomquist et al. 1982;Cripps et al. 1986;de Renobales et al. 1987;Malcicka et al. 2018) and thus confirms our initial hypothesis on single LA biosynthesis origin in this clade.
Besides the participation in primary metabolism, the de novo biosynthesis of LA may also be important for secondary metabolism. Termites, as social insects, extensively use pheromone communication to coordinate the social tasks in their colonies. Some of these pheromones are derived from unsaturated FAs, including PUFAs. It is especially the case for trail-following pheromones, used to mark the foraging trails, and sex-pairing pheromones, allowing the future kings to localize the future queen during the dispersal flights. Chemical diversity of both of these pheromone types is dominated by C12 fatty alcohols Z3-dodecenol, Z3, Z6-dodecadienol, and Z3,Z6,E8-dodecatrienol, which are used as pheromones in virtually all species of the derived clade Kalotermitidae + Neoisoptera, encompassing more than 98% of extant termite species (Bordereau and Pasteels 2011;Krishna et al. 2013). Though their biosynthesis is not yet described, they are likely to originate from the ubiquitous FAs OA and LA and proceed through three cycles of β-oxidation followed by the reduction to corresponding alcohols by fatty acyl-CoA reductases, a transformation order analogous to the origin of FA-derived alcohol pheromones in other insects Lienard et al. 2010;Jurenka et al. 2017;Tupec et al. 2019). The resulting pheromones are then expected to be stored in the form of wax esters prior to their release as free alcohols (Tokoro et al. 1990(Tokoro et al. , 1992Jirošová et al. 2016;Kyjaková et al. 2017). The availability of LA independently of dietary sources may facilitate the production of these pheromones. While the expected biosynthesis of the former two alcohols appears as relatively trivial, the formation of the latter would require an additional stereospecific desaturation at ω4 position of LA (or its shortened homolog) to give rise to the unusual 3Z,6Z,8E double bond topology of the final pheromone. The repertoire of FAD sequences assembled in the present study may be an initial step in the search of the FAD responsible for the introduction of this third double giving rise to the unique structure of Z3,Z6,E8-dodecatrienol, the most frequent termite pheromone, known to be naturally produced only by termites and the wood-degrading basidiomycete fungus Gloeophyllum trabeum (Esenther et al. 1961;Matsumura et al. 1968).
Our functional characterizations of cockroach and termite FADs point at the significance of structural aspects of the desaturase substrate tunnel. We used the insights obtained by previous studies on animal FAD structures and sequence features (e.g., Bai et al. 2015;Wang et al. 2015;Wilding et al. 2017;Nagao et al. 2019), summarized in supplementary figure S9, Supplementary Material online, to predict two possible mechanistic relationships for Δ9 and Δ12 functions. First, we verified the hypothesized role of the Trp236 residue in substrate binding specific MBE to Δ9 function. This residue has been recorded at the opening of the substrate tunnel in multiple Δ9 FADs while being absent in some FADs of other specificities (e.g., Wilding et al. 2017). It is presumed to stabilize the thioester moiety of the substrate at a defined distance from the histidine-rich active site via hydrogen bonding. Indeed, we showed here that this residue is of great importance, because its absence dramatically decreases the Δ9 activity, yet it does not render the enzyme inactive. Notably, certain lowering of Δ9 activity was reported in a long chain-specific mutant of mouse SCD1 when its acyl register Trp258 was changed to phenylalanine or tyrosine residue, while the register mutants with other residues, including leucine, were completely inactive (Cai et al. 2020). This suggests that the acyl register is not vital for Δ9 regiospecificity of FADs, but it rather adds to the overall substrate binding mediated by CoA-binding residues in the external region of the substrate tunnel ( fig.  2B) and on the surface of the enzyme. In all of our studied cockroach and termite FAD-A1b Δ12 desaturases, Trp236 register is lost, which may contribute to its differential function compared with FAD-A1a. However, the loss of this Δ9-specific acyl register does not appear to be universal and indicative for all Δ12 enzymes, since in some of them, it remains conserved (see fig. 2A). In fact, the genuine acyl register in Δ12 FADs might be significantly shifted out of the substrate tunnel in order to compensate for the shorter substrate tunnel length in these enzymes compared with the ancestral Δ9 FADs (fig. 7). Such a Δ12-specific acyl register has not yet been identified.
Second, based on previous experiments and modeling studies suggesting substrate preference shifts resulting from substrate tunnel capping (e.g., Meesapyodsuk and Qiu 2014;Cai et al. 2020), we aimed at testing the possible role of the shorter substrate tunnel observed in FAD-A1b compared with FAD-A1a. The capping residues should determine the length of accepted substrates and localization of the double bond introduction relative to the substrate ω-end. Shortening of the FAD-A1a substrate tunnel length using the substitution consistently appearing in FAD-A1b really leads to exclusion of longer substrates (C16 and C18) in favor of shorter fatty acyls (C14), thus resulting in the shift from the dominant ω-9 to ω-5 desaturation; such a shift towards the ω-end is needed for the Δ12 activity of FAD-A1b. We compared this observation with the models of Δ12 FADs previously characterized in insects and found a similar distal tunnel architecture, as evidenced in figure 7. While the tunnel capping mechanism is shared by the independently evolved insect Δ12 FADs, the overall substratebinding pattern of Δ12 desaturases in general remains to be elucidated and will require complex mutagenesis experiments and/or structure determination. Nevertheless, the pattern of substrate tunnel capping may be used for the preliminary prediction of non-Δ9 functions.
In light of the partial functional differences between FAD-A1b in cockroaches on the one hand and termites + their sister genus Cryptocercus on the other hand, we identified a few sequence differences characteristic for the cockroach and termite FAD-A1b variants. What is the mechanistic role of the pinpointed differences remains unknown, as well as whether the FAD-A1b products additional to LA in P. americana (conjugated C16 FAs) are of adaptive significance. Even so, the ability to convert OA into LA is shared by both FAD-A1b variants and in both of them, the capacity to accept saturated substrates has been lost, as shown by both the functional assays and rescue experiments. Loss of the original desaturation function along the evolution from Δ9 to Δ12 FADs is reported in some insect desaturases, whereas in some others, the Δ9 catalytic function on saturated substrates has been in part preserved (Zhou et al. 2008).
Dietary specialization of termites and their sister group Cryptocercidae on lignocellulose substrates has been preceded by the acquisition of multiple metabolic adaptations facilitating the cellulose digestion and compensating for the deficiency in various nutrients. The establishment of obligatory symbiosis with cellulose-degrading intestinal flagellates was among the most important ones (Brune 2014), together with the acquisition of nitrogen-fixing bacteria, allowing for compensation of the low nitrogen content in a cellulose diet by fixation of atmospheric nitrogen (Breznak and Brill 1973). Since the energetic and nutrient flow between the intestinal symbionts and the termite host is primarily mediated by the microbial production of acetate (Odelson and Breznak 1983), the termite lipids are predominantly synthesized de novo from acetate in the fat bodies. Symbiont-independent LA origin in the fat body has been confirmed by previous radiotracer experiments (Mauldin et al. 1978;Blomquist et al. 1982) and is in line with our observations of ubiquitous expression of both studied genes in both the heads (lower expression) and abdomens (higher expression), both of which contain the fat body tissue. In contrast, termite dietary gains of unsaturated FAs including PUFAs are low and may essentially be due to the digestion of the symbionts themselves by the termite host. The evolution of LA biosynthesis via FAD-A1b in omnivorous termite ancestors might have been yet another critical preadaptation for the switch to the specialized lignocellulose diet and ultimately for the undisputed ecological success of termites in warmer terrestrial ecosystems.

Origin of Insects
Beside publicly available sequencing data, we used de novo generated transcriptomes of one lower termite species, P. simplex (Rhinotermitidae), and three species of higher termites (Termitidae), E. neotenicus (Syntermitinae), Spinitermes trispinosus, and Inquilinitermes inquilinus (both Termitinae). Multiple P. simplex colonies collected in Cuba andFlorida (United States) between 1964 and2009 are kept in the laboratory cultures of the authors. The authors collected the remaining three sequenced species from field colonies in the primary forest of French Guiana along the road to Petit Saut (5°04.250′N 52°58.770′W-5°04.650′N 361 53°01.360′W).

MBE
Cold-anesthetized termites were dissected under stereomicroscope, the dissected tissues transferred into RNase free collection tubes, kept at 4 °C overnight in 1 mL of RNAlater (Thermo Fisher Scientific), and transported to the laboratory in Prague at −10 °C.

Next Generation Sequencing and Bioinformatics
Total RNA was isolated from the abdominal cavity (with removed digestive tube) and heads of 3 primary queens + 2 primary kings + 12 workers + 12 soldiers of P. simplex, 10 female nymphs + 10 neotenic queens of E. neotenicus, PsimFAD-A1a EneoFAD-A1a ZnevFAD-A1b PameFAD-A1b AdomD12Des PsimFAD-A1b Ala108 Leu86 Leu67 Ile80 Asn104 FIG. 7. Substrate tunnels of Δ9 and Δ12 FADs. Cross-section views of the substrate tunnels in blattodean FAD-A1a and FAD-A1b and previously functionally characterized FADs from mouse (Mmus), cricket (Adom), parasitoid wasp (Nvit), and beetles (Clug and Tcas), illustrating the different tunnel capping architectures in Δ9and Δ12-specific enzymes. The model of mouse SCD1 is based on its X-ray structure (Bai et al. 2015), while the rest of the models were predicted by AlphaFold2 (Mirdita et al. 2022). The active site is located at the top of each image (the ball indicating one of the metal ions). Model of stearoyl-CoA (a substrate of canonical Δ9 FADs) is fitted in all tunnels based on the structural alignment with mouse SCD1. The residues homologous to the tunnel capping residues inferred in blattodean FAD-A1b (e.g., Ile83 in PsimFAD-A1b) are represented as sticks. The star indicates a potential steric clash of stearoyl-CoA (if bound in Δ9 FAD-specific manner) with the tunnel capping residue in Δ12 FADs.

MBE
16 soldiers + 16 workers of S. trispinosus, and 16 soldiers + 12 workers of I. inquilinus using the acid guanidinium thiocyanate-phenol-chloroform extraction method. Briefly, deep-frozen samples pretreated with RNAlater were homogenized in TRI reagent solution (Thermo Fisher Scientific) using polypropylene pestles. RNA was collected from aqueous fraction after addition of chloroform and centrifugation according to manufacturer's documentation and precipitated with isopropanol (1:1), and the pellet was washed with 75% ethanol, dried at room temperature, and dissolved in 10 mM Tris-HCl buffer, pH 8.0, with 0.1 mM EDTA. Quality control and quantification of isolated samples were performed on Nanodrop ND-1000 UV/VIS spectrophotometer (Thermo Fisher Scientific) and 1% agarose gel in Tris-EDTA-acetate buffer prestained with ethidium bromide. Preparation of poly(A)-selected strand-specific RNA libraries and Illumina sequencing was conducted by external service providers -Novogene Co. (Beijing, China), providing 414 million of 2 × 150 bp paired-end reads in P. simplex; Eurofins Genomics (Ebersberg, Germany), providing 220 and 178 million of 2 × 150 bp paired-end read in S. trispinosus and I. inquilinus, respectively; and GeneCore facility, EMBL (Heidelberg, Germany), providing in total 103 million of 2 × 100 bp paired-end reads in E. neotenicus.

FAD Sequence Prediction, Gene Structure, and Expression
Candidate FAD sequences were first retrieved from our inhouse transcriptome assemblies of P. simplex, E. neotenicus, S. trispinosus, and I. inquilinus and publicly available assemblies of Z. nevadensis (GCA_000696155.1; ZooNev1.0; Terrapon et al. 2014) and Cryptotermes secundus (GCA_002891405.2; Csec_1.0; Harrison et al. 2018) upon multiple protein BLAST searches using a reference data set combining all sequences used in (Buček et al. 2013(Buček et al. , 2015 and Helmkampf et al. (2015). BLAST hits were then filtered based on e values (≤0.001), alignment length (≥200 bp), and similarity cutoffs (≥40%), and final candidate sequences were retrieved manually in a series of multiple sequence alignments and phylogenetic analyses in SeaView v4 (Gouy et al. 2010). These candidates were later used for FAD discovery in other termites, cockroaches, and comparative insect species in Transcriptome Shotgun Assembly database (TSA, NCBI) based on homology mapping using protein BLAST.
Gene structure of PsimFAD-A1a and PsimFAD-A1b was reconstructed upon BLAST-based local alignment of predicted transcript sequence in our P. simplex in-house genome assembly (Koubová et al. 2021) and manual revision of exon-intron boundaries. The fragments per kilobase million (FPKM) normalized read abundances were estimated in P. simplex queens, kings, soldiers, and workers after bowtie2 mapping of the RNA-Seq reads on transcriptome assembly based on RNA-Seq by Expectation-Maximization (RSEM) method using the align_and_estimate_abundance.pl script from Trinity v2.1.1 package.
The second analysis aimed at a detailed view on the evolutionary pattern of FAD-A1 branch (according to Helmkampf et al. 2015), a subfamily identified by the previous phylogeny and amino sequence analysis to contain putative Δ12 FADs evolved from an ancestral Δ9 FAD. It included 29 species of cockroaches from both Blattoidea and Blaberoidea lineages and 28 species of termites from 7 families. List of species and FAD sequences is provided in supplementary table S4, Supplementary Material online.
Sequence alignments were generated with two to four iterations of the MUSCLE algorithm (Edgar 2004). In case of the first analysis, highly variable N-and C-terminal parts were trimmed and only the relevant conserved region covering all functional domains and corresponding to alignment positions 50-329 in P. simplex FAD-A1b protein sequence was kept for tree building (see supplementary table S1, Supplementary Material online). Maximum likelihood trees were inferred in IQ-TREE (Nguyen et al. 2015) using LG+R7 as the best fitting substitution model, and branch supports were based on 1,000 ultrafast bootstrap replicates (Hoang et al. 2018).

Cloning and Expression of Candidate Genes in Yeast
Functional characterization was performed for FAD-A1a and FAD-A1b from the cockroach P. americana (Pame) and three termite species, that is, Z. nevadensis (Znev, Archotermopsidae), P. simplex (Psim, Rhinotermitidae), and E. neotenicus (Eneo, Termitidae). Synthetic FAD sequences codon optimized for yeast (see supplementary table S6, Supplementary Material online) were custom cloned (GenScript) into copper ion-inducible pYEXTHS-BN vector (Holz et al. 2002), equipped with an N-terminal hexahistidine (His 6 ) sequence. After verifying the cloned FAD sequences MBE by Sanger sequencing (Eurofins Genomics) using primers 5′-AATATCATATAGAAGTCATCG and 5′-TTTGCAGC TACCACATTG, the vectors were transformed into desaturase 1-and elongase 1-deficient Saccharomyces cerevisiae strain elo1Δole1Δ (MATa elo1:HIS3 ole1:LEU2 ade2 his3 leu2 ura3) (Schneiter et al. 2000). The yeasts were cultured in a synthetic complete medium lacking uracil (SC−Ura), supplemented with 0.5 mM copper sulfate and 0.1% (w/v) tergitol (type NP-40, Sigma). To allow the growth of the mutant host strain lacking Δ9 desaturase, the medium was supplemented with 0.1 mM Z9-monounsaturated fatty acid(s) (UFA, see below). Heterologous protein expression was monitored by western blot analysis of the whole-cell extracts. First, the harvested cells were incubated for 5 min in 5% (v) β-mercaptoethanol and subsequently for 5 min in 0.4 M NaOH to partially disintegrate the cell walls and then treated with regular sample buffer for SDS-PAGE. The His 6 -tagged proteins were detected using mouse antipolyhistidine-HRP conjugate antibody (Sigma-Aldrich, A7058, 1:2,000 dilution) and Ultra Science Femto Western Substrate kit (Bio-Helix). As a negative control, we used a strain bearing an empty pYEXTHS-BN vector, and as positive controls, we used Blap_NPVE, a Δ9 FAD from the bumble bee B. lapidarius (Buček et al. 2013).
The complete protein-coding sequences (i.e., including the His 6 tag) of newly characterized blattodean FADs were deposited to GenBank under accession numbers OP575963-OP575970.

Site-Directed Mutagenesis
To confirm the functional significance of the amino acid residues expected to codetermine the Δ9 activity of FAD-A1a, we performed functional assay #6 comparing the activity of wild-type PsimFAD-A1a from P. simplex with that of its mutants PsimFAD-A1a-86Ile and PsimFAD-A1a-236Leu. Both mutants were generated by whole plasmid amplification (from 25 ng template pYEXTHS-BN_PsimFAD-A1a) using Q5 high-fidelity polymerase (New England Biolabs) and mutagenic primers listed in supplementary table S7, Supplementary Material online. The amplification was carried out for 30 s at 98 °C, followed by 3 cycles of 20 s at 98 °C, 20 s at 51 °C, and 4.25 min at 72 °C, followed by 25 cycles of 10 s at 98 °C, 20 s at 60 °C, and 4.25 min at 72 °C, and terminatedby 10 min at 72 °C. After amplification, the template was digested with DpnI nuclease (New England Biolabs) at 37 °C for 3, and 10 μL of the digested mixture was transformed into Escherichia coli strain DH5α (Invitrogen). The mutant constructs were then amplified in bacteria, isolated using Zyppy Plasmid Miniprep kit (Zymo Research) and Sanger sequenced (Eurofins Genomics) prior to the transformation into yeast.

Desaturase Functional Assays
Yeasts were cultured for 4 days at 30 °C in 20 mL of medium with variable UFA supplementations to cover the range of possible substrates while still allowing for yeast growth.

Lipid Extraction and Transesterification
The lipids from the yeast cultures were extracted and transesterified as previously described (Matoušková et al. 2007;Tupec et al. 2019) with slight modifications. Briefly, the lyophilized sample was shaken vigorously with 1.2 mL of dichloromethane (DCM)/methanol 2:1 (v/v) and glass beads (0.5 mm) for 1 h. After short centrifugation to sediment the cellular debris, 1 mL of supernatant was evaporated under nitrogen, and the residue was shaken with 0.2 mL of 0.5 M sodium methoxide in methanol for 0.5 h. The mixture was neutralized by adding 0.2 mL of solution of disodium phosphate and monopotassium phosphate (0.25 M each) and 35 μL of 4 M HCl. The obtained FAMEs were extracted with 600 μL of hexane, and a drop of 0.6 M trimethylsilyldiazomethane in hexane was added to the extract, which was then analyzed by gas chromatography (GC).

MBE
0.1 mm, and 0.1 μm film thickness) columns. The SSL injector was heated to 250 °C, the oven programmed to 50 °C (1 min)-8 °C/min-320 °C (5 min) with the modulator set 15 °C higher and the secondary oven 10 °C higher than the primary column. The instrument was operated either in 1D or 2D setup depending on the required peak resolution. We used electron ionization (70 eV), time-of-flight mass analyzer, at acquisition rate 15 spectra/s (1D) or 200 spectra/s (2D).

FAME Derivatization
To identify the double bond positions in the analyzed FAMEs, we performed two types of microderivatizations, that is, conversion to dimethyloxazoline derivatives (DMOX) (Fay and Richli 1991) and methylthiolation using dimethyl disulfide (DMDS) (Dunkelblum et al. 1985). The DMOX derivatization was carried out by adding 500 µL of 2-amino-2-methylpropanol to the FAME samples, which have been rid of the original solvent (hexane) under a stream of nitrogen at room temperature. The samples were then heated with the reagent overnight at 180 °C. After cooling, the reaction mixture was dissolved in 5 mL of DCM and washed twice with 2 mL of water. The DCM solution was subsequently dried with disodium sulfate and evaporated under a stream of nitrogen at room temperature. The residue was dissolved in 200 µL of hexane and injected into the GC system.
For DMDS derivatization, 100 µL of DMDS and 5 µL of 5% (w/v) iodine in diethyl ether were added to 100 µL of FAME samples in hexane. The reaction mixture was kept at 40 °C overnight. The solution was decolorized with two drops of 5% (w/v) sodium thiosulfate solution and the products were extracted into hexane (twice 300 µL). The organic layer was redissolved in hexane and injected into the GC system.

Rescue Experiment
We tested the potential rescuing ability of FAD-A1a and FAD-A1b in the elo1Δole1Δ yeast strain. To do so, a cell suspension (OD 600 0.1) was stepwise diluted to OD 600 10 −5 and the suspensions were applied as drops to SC−Ura plates with 0.5 mM copper sulfate using a replica plater (Merck, 96-well plate 8 × 12 array); growth control plates with 0.1 mM palmitoleic acid and OA (each) and 0.1% (w/v) tergitol were also prepared. The plates were incubated for 4 days at 30 °C and subsequently photographed. The experiment was performed with FAD-A1a and FAD-A1b from P. simplex (PsimFAD-A1a and PsimFAD-A1b), a positive control (Blap_NPVE), and a negative control (yeast transformed with an empty plasmid).

Structure Prediction
The structural models of FAD-A1a and FAD-A1b from P. americana, Z. nevadensis, P. simplex, and E. neotenicus and of Δ12 FADs from A. domesticus (UniProt B7SB91), T. castaneum (B7SB92), C. lugubris (K7PD28), and N. vitripennis (A0A7M7GAP9) were predicted using ColabFold (Jumper et al. 2021;Mirdita et al. 2022) with default parameters (pdb70 template mode). The PDB files with the models used in this study are available in supplementary archive 1, Supplementary Material online. The predicted FAD models were manually trimmed of low-pLDDT terminal regions and aligned to a complex of mouse SCD1 with stearoyl-CoA (PDB code 4YMK) in PyMOL Viewer (Schrodinger). The individual substrate tunnels were subsequently inspected using UCSF ChimeraX 1.4 (Pettersen et al. 2021).

Supplementary Material
Supplementary data are available at Molecular Biology and Evolution online.