Sequencing, assembly, and annotation of the C. hualienensis genome

In total, 100.42 Gb (~700 × coverage) of clean data from Illumina and 20.04 Gb (~140 × coverage) of clean data from PacBio Sequel were obtained (S1 and S2 Tables). A schematic diagram of the hybrid assembly is shown in S1 Fig. The assembled C. hualienensis genome was 143,521,810 bp, with a scaffold N50 size of 12,989,427 bp (Fig 2A and Table 1), representing 99.3% of the estimated genome size (144.5 Mb) (S2 Fig). The integrity of the assembled genome was assessed. First, the short reads obtained from the Illumina sequencing data were compared with the assembled genome. The percentage of reads mapped to the draft genome was greater than 96%. Next, BUSCO [32] analysis revealed that 95.6% of the single-copy arthropod orthologues were complete. These results indicated that the C. hualienensis genome assembly was of high quality and coverage. The genome was the smallest among the genome-sequenced Myriapods, millipedes and centipedes, ranging from 181 to 2530 Mbp [33,34] (S3 Fig). As the size of repetitive sequences, genomic sequence comprised of transposable elements (TE) and simple repeats, are known to positively correlate with genome size [35], we analysed the TE in C. hualienensis genome. In the C. hualienensis genome assembly, 25.4 Mb of repetitive sequences, which cover 17.7% of the genome, were identified (S3 Table). The unclassified elements were the most abundant type of repetitive sequence in C. hualienensis, spanning 15.8 Mb (11.0%) of the genome. Class I long terminal repeat retrotransposons, class II DNA transposons, and non-long terminal repeat elements account for 1.55%, 3.26%, and 0.25%, respectively, of the genome. A strong correlation between transposable element content and genome size across C. hualienensis and other millipedes and centipedes genomes was observed (R² = 0.98, P = 7.85e-08; regression line y = 128 + 1.63 x) (S3 Fig), in agreement with the earlier report [33].

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Assembly and annotation statistics of Chamberlinius hualienensis genome.

https://doi.org/10.1371/journal.pgen.1011955.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. C. hualienensis genome annotation.

(A) Characterisation of different elements on the millipede scaffolds. From outer to inner ring, the > 10 kb scaffolds of the assembled C. hualienensis genome, gene density, transposable element density, and GC-skew. (B) Upset plot comparing shared orthogroup in millipedes. (C) Phylogenetic tree. The tree is rooted with Hypsibius dujardini as the outgroup. Polydesmida species are labeled with red circles.

https://doi.org/10.1371/journal.pgen.1011955.g002

Ab initio gene prediction programs and evidence from RNA sequencing (S4 Table) were integrated to annotate the C. hualienensis genome. A total of 14,132 gene models were predicted for the genome (Table 1). BUSCO assessment indicated that the predicted genes covered 93.8% of arthropod-conserved orthologues. Among the predicted proteins, 12,443 (80.3%) were annotated using GenBank, 12,541 (81.2%) using interproscan, 11,702 (75.7%) using eggnog mapper, and 11,038 (71.4%) using pfam.

Orthogroup and species tree

To infer the evolutionary history of C. hualienensis, orthorogroup analysis was performed to identify single-copy orthologous proteins. Across myriapod and other arthropod species: millipedes (Helicorthomorpha holstii, Nipponia nodulosa, Anaulaciulus tonginus, Trigoniulus corallinus, and Glomeris maerens), centipedes (Rhysida immarginata, Lithobius niger, Thereuonema tuberculata, and Strigamia martima), other arthropods (Ixodes scapularis, Daphnia pulex, Tribolium castaneum, and Drosophila melanogaster), and the tardigrade Hypsibius dujardini, 41,810 orthogroups were identified. In total, 3,104 orthogroups contained all species, 18,160 contained a subset of species, and 21,323 were species-specific orthogroups. Six millipede species shared 446 orthogroups, three Polydesmida millipede species shared 107 orthogroups, two cyanogenic millipedes shared 144 orthogroups, and C. hualienensis shared 290 species-specific orthogroups (Fig 2B). Phylogenetic reconstruction was conducted based on 102 single-copy orthologous proteins shared by the abovementioned organisms (Fig 2C). Three Polydesmida species, C. hualienensis, H. holstii, and N. nodulosa, formed a monophyletic clade (Fig 2C).

ChuaHNL and its paralogous gene cluster on the genome

HNL is a key enzyme of cyanogenesis in millipedes. This enzyme catalyzes the release of hydrogen cyanide from (R)-MAN [36], which is accumulated in C. hualienensis approximately 570 µg and 596 µg per male and female millipede, respectively (S4 Fig). However, evolutionary origin of the enzyme remains unclear. We investigated whether C. hualienensis harbours ChuaHNL homologous genes (paralogues) by orthogroup analysis. ChuaHNL (CHUA_005138) and its two paralogues (CHUA_005137 and CHUA_005136) were grouped into the orthogroup OG0016952. These protein-coding genes were clustered in the genome (Fig 3A). CHUA_005137 and CHUA_005136 shared 70% and 54% amino acid sequence identities with ChuaHNL, respectively (S5 Fig), and shared all and four of the five catalytic residues [18,20,21], respectively (Figs 3B and S6). Further, seventeen ChuaHNL-like genes were observed within 75 kb of the ChuaHNL gene cluster (Fig 3A). The proteins encoded by these genes were assigned to orthogroups OG0001941, OG0009681, and OG0016953, unlike ChuaHNL, which belonged to OG0016952 (Fig 3A and 3B). However, the amino acid sequences of these proteins shared 11–21% identity at the amino acid level with ChuaHNL (S5 Fig) and eight structurally important Cys residues (S6 Fig), which formed inter- and intra-disulfide bonds in ChuaHNL [21], except CHUA_005132. This gene was partially truncated and is considered malfunctional (S6 Fig). These results suggested that ChuaHNL and the paralogous proteins, except for CHUA_005132, share lipocalin folds, although these paralogous proteins are grouped into four orthogroups. Orthologues of ChuaHNL and its paralogous genes were found in the H. holstii genome. They clustered in the C. hualienensis genome, and their syntheny was conserved (Fig 3A). The N. nodulosa genome only encoded an orthologous protein belonging to OG0016953, not the gene cluster in C. hualienensis and H. holstii (Fig 3A). Lipocalins are a group of widely distributed proteins that can be found in animals, plants, and bacteria [37]. Millipede HNLs and related proteins are phylogenetically distant from lipocalins of other arthropods, plants, and bacteria (S7 Fig), and their fold is distinct from plant and microbial HNLs (S8 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Gene cluster and characterisation of ChuaHNL and its paralogues.

(A) ChuaHNL and its paralogue cluster on the C. hualienensis genome and synteny among Polydesmida millipedes. (B) Phylogenetic tree and catalytic residues, HNL activity, and gene expression (TMM-normalised reads counts) of ChuaHNL and its paralogues.

https://doi.org/10.1371/journal.pgen.1011955.g003

Biochemical characterisation of ChuaHNL and its paralogous proteins

Whether ChuaHNL paralogous proteins exhibit HNL activity, we heterologously produced ChuaHNL and six paralogous proteins (CHUA_005137 and CHUA_005136 from OG0016952; CHUA_005139 from OG0014028; CHUA_005135, CHUA_005127, and CHUA_005124 from OG000814) as N-terminal His-tagged proteins using the yeast Pichia pastoris expression system [38]. These proteins were isolated from 120 mL of the culture supernatant using a nickel affinity column. SDS-PAGE showed that all proteins were secreted into the medium (S9 Fig). Among them, ChuaHNL, CHUA_005137, and CHUA_005136 catalysed (R,S)-MAN degradation, indicating that CHUA_005137 and CHUA_005136 are functional HNLs. However, the others did not exhibit HNL activity (Fig 3B), which agrees with the fact that these paralogous proteins did not have the catalytic residues of millipede HNLs (Figs 3B and S6). We attempted to create variants with full active site residues but were unable to do so in P. pastoris. To compare ChuaHNL and the paralogous proteins having HNL activity, recombinant proteins were purified from 1.5 L of culture supernatant (S5–S7 Tables). ChuaHNL and two paralogous proteins catalysed the degradation of racemic MAN into benzaldehyde and HCN with the following kinetic parameters (S10A Fig): ChuaHNL, V_max = 2950 ± 227 µmol/min/mg, K_m = 13.7 ± 2.35 mM; CHUA_005137, V_max = 916 ± 67.4 µmol/min/mg, K_m = 12.9 ± 2.16 mM; CHUA_005136 V_max = 50.0 ± 0.681 µmol/min/mg, K_m = 1.10 ± 0.0618 mM. ChuaHNL, CHUA_005137, and CHUA_005136 also catalysed its reverse reaction, namely the (R)-MAN synthetic reaction from benzaldehyde and potassium cyanide, with the following kinetic parameters (S10B Fig): ChuaHNL, V_max = 1310 ± 63.6 µmol/min/mg, K_m = 4.31 ± 0.771 mM; CHUA_005137 V_max = 1280 ± 80.1 µmol/min/mg, K_m = 4.50 ± 0.994 mM; CHUA_005136 V_max = 677 ± 93.6 µmol/min/mg, K_m = 2.36 ± 0.696 mM, K_i = 21.1 ± 5.74 mM. When the same activity of the enzyme was used for the reaction (that is, 8 U/mL), the enantiomeric excess of (R)-MAN produced was 92.1%, 91.5%, and 83.8%, respectively (S11 Fig). These results indicated that CHUA_005137 and CHUA_005136 have excellent R-selectivity in cyanohydrin synthesis, similar to ChuaHNL [17].

To elucidate the expression pattern of ChuaHNL and its paralogous genes in C. hualienensis, we performed RNA-seq analysis of four tissues, including the antennae, paranota-containing defensive glands, paranota without defensive glands, and gut (S8 Table). ChuaHNL was specifically expressed in the paranota-containing defensive glands (Fig 3B) in accordance with previous gene expression analyses [17]. CHUA_005137 and CHUA_005136 were expressed in traces, indicating that these proteins do not accumulate in the defensive glands, although they exhibit HNL activity. A few paralogous proteins, such as CHUA_005124 and CHUA_005127, were highly expressed in the antennae (Fig 3B). ChuaHNL belongs to the lipocalin family [21] and the lipocalin proteins often functions as a carrier protein for small hydrophobic molecules [37]. The expression of ChuaHNL paralogues in the antennae is consistent with that of lipocalin genes in the antennae of several insect species [39].

CYPome and identification of CYPs involved in (R)-MAN biosynthesis

CYPs are key biosynthetic enzymes of cyanohydrins in cyanogenic plants and burnet moths (Fig 1B) [15,40]. However, the CYPs in Myriapoda are not well understood. We searched for CYPs in millipede and centipede genomes and found 96 C. hualienensis genes that encode putative functional CYPs (Fig 4A). The Polydesmida millipedes H. holstii and N. nodulosa contain 119 and 80 CYPs, respectively. These numbers were greater than those of other millipedes (32–64) and centipedes (52–63) (S12A Fig). Phylogenetic analysis revealed that C. hualienensis CYPs (ChuaCYPs) clustered into the CYP2, CYP3, CYP4, CYP20, and mitochondrial clans (Fig 4A). CYP clans are deep gene clades observable on phylogenetic trees [41]. CYP2 clan members comprised the largest number of CYPs in C. hualienensis (Fig 4A). Similarly, CYP2 clan members were most abundant in other Myriapoda species (S12 Fig). However, the CYP3 and CYP4 clans are predominant in insects [42]. The difference in clan proportions in arthropods is thought to be caused by the dynamics of gene births and deaths over evolutionary time, affecting the CYPome [42].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Identification of (R)-mandelonitrile biosynthetic enzymes from C. hualienensis.

(A) Phylogenetic tree of cytochrome P450s (CYPs) from C. hualienensis and previously characterised cyanohydrin biosynthetic CYPs from the burnet moth and plants. Bar indicates 40% divergence. (B) Identification of CYPs catalysing the conversion of (E/Z)-PAOx into PAN. Yeast cells carrying CYPs from C. hualienensis expression plasmids or an empty vector were incubated with (E/Z)-PAOx. PAN formation was analysed using ultra-performance liquid chromatography. Reaction product peaks are indicated by red arrows. (C) Phylogenetic tree of FMOs from C. hualienensis, arthropods, plants, and microorganisms. Bar indicates 60% divergence. FMOs from C. hualienensis were labelled with red circles. (D) Identification of the (E/Z)-PAOx-producing FMO. Escherichia coli cells carrying FMOs from C. hualienensis expression plasmids or an empty vector were cultured. The accumulation of (E/Z)-PAOx was analysed using gas chromatography–mass spectrometry. Reaction product peaks are indicated by red arrows. (E) (R)-Mandelonitrile biosynthetic pathway in C. hualienensis and gene expression (TMM-normalised read counts) of biosynthetic enzyme genes.

https://doi.org/10.1371/journal.pgen.1011955.g004

To narrow down the CYP candidates involved in (R)-MAN biosynthesis, we searched for CYPs specifically expressed in paranota with defensive glands, which accumulate (R)-MAN, by analysing expression using RNA-seq. However, we were unable to identify such candidate CYPs. Therefore, we performed functional CYPome analysis to identify (R)-MAN biosynthetic CYPs. Of these 88 microsomal CYPs, 63 were cloned. To evaluate the activity of CYPs in yeast, the redox partner cytochrome P450 reductase (CPR) must be co-expressed. In C. hualienensis, there is one CPR gene (ChuaCPR, CHUA_001042). We generated a yeast strain, S. cerevisiae ChCR11, in which ChuaCPR was genomically introduced. The strain exhibited a higher PAN-metabolising activity than the parent strain INVSc1 when expressed CYP3201B1 (S13 Fig), which catalyses the hydroxylation of PAN to (R)-MAN [23].

We evaluated the activity of 63 ChuaCYPs towards l-Phe and (E/Z)-PAOx using whole-cell biocatalysis, which can detect the catalytic activities of CYPs despite their production levels in yeast being below the detection limit of the CO difference-spectrum assay [43]. Although no (E/Z)-PAOx-producing CYPs were detected, two CYPs, CHUA_008368 and CHUA_0014123, produced PAN from (E/Z)-PAOx (Figs 4B and S14). CHUA_008368 and CHUA_0014123 were designated CYP4GL4 and CYP30008A2, respectively, based on the standard CYP nomenclature. CYPs were assigned to the same family and subfamily when the amino acid level sequence identity was > 40% and >55%, respectively [44].

CYP4GL4 and CYP30008A2 belong to the CYP4 and CYP2 clans, respectively. Microsomes harbouring CYP4GL4 and CYP30008A2 catalysed the formation of PAN from (E/Z)-PAOx (Fig 4B). CYP4GL4 and CYP30008A2 exhibited optimum pH and temperature of 8.0 and 35°C and 7.0 and 40°C, respectively (S15 Fig). The V_max and K_m towards (E/Z)-PAOx were as follows: CYP4GL4, V_max = 615 ± 11.6 pmol min⁻¹ mg⁻¹ (microsome), K_m = 8.13 ± 1.06 µM; CYP30008A2, V_max = 106 ± 4.54 pmol min⁻¹ mg⁻¹ (microsome), K_m = 164 ± 29.5 µM. Both P450s act on 4-hydroxyphenylacetaldoxime (4HPAOx) and indole-3-acetldoxime (IAOx) to produce the corresponding nitriles (S16 Fig), indicating that the two nitrile-synthesising CYPs act on multiple amino acid-derived aldoximes.

Flavin-dependent monooxygenases catalysing the first (R)-MAN biosynthesis step

Next, we identified 10 putative functional flavin-dependent monooxygenases (FMOs) in the C. hualienensis genome (Fig 4D and S9 Table) as the candidates for PAOx-producing enzymes. These FMOs had the following FMO signature sequences: FAD-binding motif (GXGXXG), FMO motif (FXGXXXHXXXYK), and NADPH-binding motif (GXGXXG) (S17A Fig), and they shared 23–63% sequence identity (S17B Fig). These FMOs, except for CHUA_011563, were predicted to have a transmembrane region at their C-terminus (S9 Table) and were localised to the endoplasmic reticulum (S9 Table), similar to mammalian FMOs involved in xenobiotic metabolism [45]. However, the insect FMOs that have been characterised are soluble proteins [46–48]. FMOs are divided into six groups (A–H) based on their structures and properties [49], and group B FMOs are further divided into four clades (I–IV) [50]. Phylogenetic analyses showed that the C. hualienensis FMOs belonged to clade I in group B, except for CHUA_011563, which belonged to clade II in group B (Fig 4D). The other millipede and centipede genomes contained 2–14 genes encoding putative functional FMOs (S18A Fig). These FMOs clustered into clade I or II of group B, as in the case of C. hualienensis FMOs (S18B Fig) and were not phylogenetically related to fern FOS1 (clade IV) or microbial FMOs (Figs 4B and S18).

All C. hualienensis FMOs were expressed, but CHUA_003298 was expressed at trace levels and not in a tissue-specific manner (S10 Table). Therefore, all cDNAs encoding FMOs were heterologously produced in Escherichia coli to determine the (E/Z)-PAOx-producing activity of the FMOs. Among them, E. coli harbouring CHUA_003298 accumulated 128 ± 51 µM (E/Z)-PAOx in the culture (Figs 4D and S19), whereas E. coli harbouring the other FMOs did not (S19 Fig). Given that CHUA_003298 did not show enzymatic activity after the disruption of E. coli cells, biochemical characterisation of the enzyme was not performed. The accumulation of 4HPAOx and IAOx derived from l-tyrosine and l-tryptophan, respectively, was not detected (S20 Fig), indicating that CHUA_003298 has a narrow substrate specificity for producing l-Phe-derived aldoxime. We designated the PAOx-producing enzyme CHUA_003298 as “millipede aldoxime synthase (MOxS)”.

We investigated the expression of biosynthetic enzyme genes using RNA-seq (Fig 4E and S11 Table). ChuaMOxS was expressed at a basal level. In contrast, other enzyme genes (CYP4GL4, CYP30008A2, and CYP3201B1) were clearly expressed, and their expression was not restricted to paranota-containing defensive glands but also observed in other tissues.

Distribution of MAN biosynthetic enzyme genes among Myriapod species

Genes encoding cyanogenesis-related enzymes were identified in two Polydesmida millipedes, H. holstii and N. nodulosa, but not in genome-sequenced non-cyanogenic millipede species from Julida, Spirobolida, and Glomerida nor centipedes (Fig 5). ChuaMOxS (CHUA_003298) shared 82.2% and 70.6% identity towards Hho_019011 from H. holstii and Nno_0338811 from N. nodulosa, respectively (S21 Fig). H. holstii had CYP4GL4 (Hho_020879) and CYP30008A2 (Hho_011653) orthologues, but the CYP30008A2 orthologue lacked the N- and C-terminal region coding sequences (S22 Fig). The coding protein is likely malfunctional or the deletion may be due to gene annotation and/or sequencing errors. CYP4GL4 and Hho_020879 shared 67% identity (S23 Fig). However, N. nodulosa harboured the CYP30008A2 orthologue (Nno_033394) but not the CYP4GL4 orthologue. CYP30008A2 and Nno_033394 shared 72% identity (S22 Fig). Thus, the nitrile synthesis step likely differs between millipede species that utilise either one or both PAN-producing CYPs (Fig 5). H. holstii and N. nodulosa harbour an orthologue of CYP3201B1 that catalyses the conversion of PAN to (R)-MAN: Hho_006485 (86%) and Nno_043294 (62%), respectively. These (R)-MAN biosynthetic enzymes share relatively high amino acid sequence identity (>65%) among C. hualienensis, H. holsti, and N. nodulosa but not cyanogenesis-related enzymes from other cyanogenic plants or burnet moths. This indicates that Polydesmida millipedes independently evolved cyanogenesis-related enzyme genes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Distribution of cyanogenesis-related enzymes in millipedes.

(A) Cyanogenesis pathway in C. hualienensis. (B) Red box indicates the presence of orthologues of C. hualienensis cyanogenesis-related enzyme genes.

https://doi.org/10.1371/journal.pgen.1011955.g005

Animals

C. hualienensis was collected from Kagoshima Prefecture, Japan, in November 2017 and 2018. The millipedes were reared in litter derived from the Japanese cedar Cryptomeria japonica D. Don at 22°C until use, as described previously [23].

DNA preparation and genome sequencing

Genomic DNA was prepared from male millipedes for library construction. DNA samples were shipped to the Beijing Genomics Institute, Shenzhen, China, and sequenced using Illumina and PacBio Sequel sequencers. For detailed information about hybrid assembly, annotation, and other bioinformatic analysis, see S1 Text.

Phylogenetic analysis

A comparative genome sequence analysis of C. hualienensis was performed using 14 genomes. Orthologous groups were identified using Orthofinder2 (v. 2.5.4) with the ‘-M msa -T raxml-ng’ flag.

Purification of ChuaHNL and its paralogous proteins with HNL activity

P. pastoris transformants were inoculated into 5 mL of YPD (1% yeast extract, 2% peptone, and 2% d-glucose). After 24 h culture at 30°C with shaking at 300 rpm, cells were harvested via centrifugation at 3,000 × g and 4°C for 10 min, resuspended in 500 mL of BMGH in a 2-L baffled flask, and grown at 30°C with shaking at 150 rpm for 14–18 h. Then, cells were harvested via centrifugation at 3,000 × g and 4°C for 10 min and resuspended in 500 mL of buffered minimal methanol medium, and 1% methanol was added as an inducer every 24 h. After 5 days of culture at 28°C with shaking at 150 rpm, the culture was centrifuged at 8,000 × g and 4°C for 15 min. The supernatant was recovered, and the pH was adjusted to 7.5 by adding aqueous 1 M K₂HPO₄. Resultant insoluble materials were removed via centrifugation at 13,000 × g and 4°C for 30 min and directly applied to a Ni Sepharose 6 fast flow column (Cytiva, Little Chalfont, UK), which was equilibrated with 20 mM KPB, pH 7.5, containing 300 mM NaCl and 20 mM imidazole. The column was washed with the same buffer, and the absorbed proteins were eluted using a linear imidazole gradient (20–500 mM) in the same buffer. The active fractions were concentrated and desalted using an Amicon Ultra-15 centrifugal filter device with a 10,000 nominal molecular weight limit cutoff (Merck Millipore, Billerica, MA, USA). It was loaded onto a Mono Q 5/50 column (Cytiva) equilibrated with 20 mM tricine-NaOH (pH 8.5) and eluted using a linear gradient of NaCl (0–500 mM) in the same buffer. The active fractions were pooled and concentrated using a centrifugal filtration device (Amicon Ultra 0.5 Centrifugal Filter Unit with ultracel-10 membrane; Merck Millipore).

HNL assay

Enzyme activities for (R)-MAN synthesis and MAN cleavage were assayed using high-performance liquid chromatography and spectrophotometry, respectively. HNL activity during (R)-MAN synthesis was measured as previously reported [21], with slight modifications. In brief, the reaction (total volume = 0.25 mL) was initiated by adding potassium cyanide (100 mM) in citrate buffer (400 mM, pH 4.0) containing benzaldehyde (50 mM; 12.5 μL of 1 M benzaldehyde dissolved in dimethyl sulfoxide), and enzyme solution (0.1–1.0 U/mL). The final dimethyl sulfoxide concentration in the mixture was 5% (v/v). The mixture was incubated at 25°C for 5 min, and then aliquots (50 μL) of reactant were mixed with a nine-fold volume (450 µL) of a n-hexane:2-propanol (85:15) mixture containing 5 mM p-xylyl cyanide as an internal standard.

Finally, the organic phase was analysed using a high-performance liquid chromatography instrument (UFLC Prominence Liquid Chromatograph LC-20AD, Shimadzu, Kyoto, Japan) equipped with a chiral column (CHIRALCEL OJ-H, Daicel, Osaka, Japan; 250 mm length × 4.6 mm inner diameter (i.d.), particle size 5 µm). The amount of (R)-MAN was estimated using a standard curve obtained from the peak areas and concentration ratios of the authentic compounds and the internal standard. One unit of synthesis activity was defined as the amount of the enzyme that synthesises 1 μmol of (R)-MAN from benzaldehyde and potassium cyanide per minute.

For the MAN cleavage activity assay, the reaction was initiated by adding the enzyme sample to a reaction mixture (1 mL) consisting of (R,S)-MAN (10 mM) in citrate buffer (100 mM, pH 5.0). The reaction mixture contained 5% (v/v) dimethyl sulfoxide at the final concentration. The reaction velocity of benzaldehyde formation was monitored using an Evolution 201 ultraviolet-visible spectrophotometer (Thermo Fisher Scientific) at 25°C and 280 nm for 1 min (extinction coefficient of benzaldehyde = 1,352 L mol⁻¹ cm⁻¹) [21]. One unit of cleavage activity was defined as the amount of enzyme that produces 1 μmol of benzaldehyde from (R,S)-MAN per minute.

Collection of ChuaCYP sequences from myriapod genomes

CYPs from C. hualienensis and genome-sequenced myriapods were searched via BLASTP using CYP3201B1 (CYP2 clan, BAV93938.1), CYP3193A1 (CYP3 clan, BAV93917.1), CYP4GQ1 (CYP4 clan, BAV93949.1), CYP20A1, (CYP20 clan, BAV93936.1), and CYP302A1 (mitochondrial clan, BAV93914.1) from C. hualienensis. The E-value threshold was 1.0 × 10⁻¹⁰. CYPs with sequences that were too short (<400 aa) or too long (>600 aa) were eliminated from the dataset as putative malfunctioning proteins because animal P450s are 500 amino acids in length [70]. CYPs from S. martima were derived from a previous study [42].

Heterologous production of ChuaCYPs in S. cerevisiae and identification of aldoxime- and nitrile-producing CYPs

Expression vectors were constructed to co-express cDNAs encoding ChuaCPR and ChuaCYPs. cDNAs encoding ChuaCPR (CHUA_001042) and ChuaCYPs were amplified via PCR and cloned into the pYeDP60 vector [71] using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Ipswich, MA, USA). The inserted DNA sequences were confirmed via Sanger sequencing. ChuaCPR with the GAL10-CYC1 promoter and PGK terminator region was amplified via PCR, and the amplicon was cloned into the SphI site of pAUR101 (Takara, Shiga, Japan) to generate pAUR-ChuaCPR. The plasmid was linearised using StuI and integrated into the genome DNA of S. cerevisiae INVScI (Thermo Fisher Scientific) through homologous recombination. The transformants were selected on a YPD-agar plate containing 0.4 µg/mL of aureobasidin A at 30°C for 3 days. Genome integration of ChuaCPR was confirmed through PCR, and the selected strain was named ChCR11 for subsequent experiments.

pYeDP60 vectors carrying cDNAs encoding CYPs were used to transform ChCR11. The transformants were selected on SGI-agar plates (0.67% yeast nitrogen base without amino acids, 0.1% casamino acid, 40 µg/mL l-tryptophan, 2% d-glucose, and 1.5% agar) at 30°C for 2 days. Transformants were inoculated into 0.5 mL of SGI medium (0.67% yeast nitrogen base without amino acids, 0.1% casamino acid, 40 µg/mL l-tryptophan, and 2% d-glucose) and cultured at 30°C for 24 h. Cells were harvested from 0.1 mL culture and resuspended in 0.5 mL of 2 × SLI medium (1.34% (w/v) yeast nitrogen base without amino acids, 0.2% (w/v) casamino acid, 80 µg/mL l-tryptophan, and 4% (w/v) galactose) supplemented with 0.2% (w/v) raffinose. After culturing at 30°C for 24 h, the cells were harvested and resuspended in 0.2 mL of 2 mM l-Phe, 1 mM (E/Z)-PAOx, or 1 mM PAN in 50 mM KPB (pH 7.0) containing 20 mM d-glucose. The reaction was performed at 30°C with shaking at 1,200 rpm for 180 min.

The production of (E/Z)-PAOx from l-Phe was analysed using liquid chromatography–mass spectrometry (LC–MS) (Nexera HPLC coupled with an LCMS-2020, Shimadzu), equipped with a COSMOSIL 3C18-MS-II packed column (100 mm × 2.0 mm i.d., particle size 3 µm; Nacalai Tesque, Kyoto, Japan). The separation conditions were as follows: column oven temperature, 40°C; mobile phase A, 0.1% formic acid in water; mobile phase B, acetonitrile; 10–60% linear gradient of B for 7.5 min and 98% B for 2.5 min, delivered at 0.4 mL/min. MS was simultaneously performed in the negative-ion mode using an LCMS-2020 apparatus (Shimadzu) via electrospray ionisation. (E/Z)-PAOx ionised in the positive-ion mode was monitored in the extracted ion m/z 136 [M + H]⁺. The production of PAN from (E/Z)-PAOx and MAN from PAN was analysed using an ACQUITY UPLC H-Class system (Waters, Milford, MA, USA) equipped with a COSMOCORE 2.6 C₁₈ column (50 mm × 2.1 mm i.d., particle size 2.6 µm; Nacalai Tesque) under the following conditions: column oven temperature, 40°C; mobile phase A, 0.1% formic acid in water; mobile phase B, acetonitrile; 10–60% linear gradient of B for 4 min and 60% B for 0.5 min, delivered at 0.4 mL/min. Aldoximes and nitriles were quantified at 210 nm.

Preparation of microsomes harbouring CYP4GL4 and CYP30008A2 from yeast

Microsomes harbouring CYP4GL4 (CHUA_008638) and CYP30008A2 (CHUA_014123) were prepared from S. cerevisiae ChCR11, following a previously described method [63]. The yeast cells were cultured in 2 mL SGI medium at 30°C for 24 h. The culture was transferred to 50 mL SGI medium at 30°C for 16 h. Yeast cells were harvested via centrifugation at 5,000 × g and 4°C for 10 min and resuspended in 500 mL of 2 × SLI medium supplemented with 0.2% (w/v) raffinose in a 2-L baffled flask at 30°C for 24 h. Yeast cells were harvested via centrifugation (5,000 × g, 10 min, 4°C), washed with 50 mM HEPES-NaOH (pH 7.6) containing 20% glycerol, and weighed to obtain the wet cell weight. The cells were then resuspended in 1 mL of 50 mM HEPES-NaOH (pH 7.6) containing 20% glycerol, 1 mM DTT, and an appropriate amount of proteinase inhibitor cocktail (complete Mini EDTA-Free; Merck KGaA, Darmstadt, Germany) per gram of wet cells. The cells were disrupted using a Multi-beads Shocker system (Yasui Kikai, Osaka, Japan) and glass beads (diameter 0.5 mm), as described previously [18]. The cell debris and beads were precipitated via centrifugation (10,000 × g, 10 min, 4°C), and the supernatant was ultracentrifuged (150,000 × g, 60 min, 4°C) to precipitate the fraction of intact microsomes. The microsomes were resuspended in 50 mM HEPES-NaOH (pH 7.6), containing 20% glycerol and 1 mM DTT and stored at −80°C until further use. The protein concentration of the microsomes was determined using the TaKaRa Bradford Protein Assay Kit (Takara) with bovine serum albumin as the standard.

Aldoxime dehydratase assay

CYP4GL4 and CYP30008A2 activity was evaluated at optimal pH and temperature. A 100 µL reaction mixture containing the microsomal fraction harbouring CYP4GL4 (5 mg/mL), 50 mM KPB (pH 8.0), 1 mM NADPH, and 1 mM aldoxime was incubated at 35°C for 30 min. After the pre-incubation at 35°C for 5 min, the reaction was started by adding the microsomes. A 100 µL reaction mixture containing the microsomal fraction harbouring CYP30008A2 (5 mg/mL), 50 mM KPB (pH 7.0), 1 mM NADPH, and 1 mM aldoxime was incubated at 40°C for 30 min. After pre-incubation at 40°C for 5 min, the reaction was initiated by adding microsomes. The reaction was terminated by adding 0.1 mL of 0.2% formic acid in 40% acetonitrile. The resulting insoluble materials were precipitated via centrifugation (21,500 × g, 10 min, 4°C). The supernatant was analysed using an ACQUITY UPLC H-Class system (Waters) equipped with a COSMOCORE 2.6 C₁₈ column (50 mm × 2.1 mm i.d., particle size 2.6 µm; Nacalai Tesque) under the following conditions: column oven temperature, 40°C; mobile phase A, 0.1% formic acid in water; mobile phase B, acetonitrile; 10–60% linear gradient of B for 4 min and 60% B for 0.5 min, delivered at 0.4 mL/min. Aldoximes and nitriles were quantified at 210 nm. The nitriles were quantified using standard curves generated from authentic compounds (PAN, Tokyo Chemical Industry, Tokyo, Japan; 4-hydroxyphenylacetonitrile, Tokyo Chemical Industry; indole-3-acetonitrile, Sigma-Aldrich, St Louis, MO, USA). To determine K_m and k_cat, the reactions were conducted under standard assay conditions using 10–2000 µM (E/Z)-PAOx. Aldoximes were chemically synthesised as previously described [63]. K_m and k_cat were determined by curve fitting the data using the drc package (version 3.0-1) [72] in R (4.1.2) [73] and the Michaelis–Menten equation as previously described [74].

Collection of FMO sequences from myriapod genomes

FMOs from C. hualienensis and genome-sequenced myriapods were searched using BLASTP using FMO1 from Homo sapiens (group B clade I, UPI0003EAECD6), SNO from Tyria jacobaeae (group B, clade II, Q8MP06), FOS1 from Phlebodium aureum (group B, clade IV, QNT35807), and SCO7468 from S. coelicolor (Q8CJJ9). The E-value threshold was 1.0 × 10⁻¹⁰. The FMOs with short (<350 aa) and long (>550 aa) sequences were eliminated from the dataset because group B FMOs were 420–520 amino acids in length [50]. The transmembrane region and subcellular localisation were predicted using Phobius [75] and DeepLoc 2.1 [76], respectively.

Identification of ChuaFMOs catalysing (E/Z)-PAOx formation in E. coli

Each E. coli BL21(DE3) transformant carrying pGro7 (Takara) and ChuaFMO expression plasmid was inoculated into LB containing 1% (w/v) glucose, kanamycin (50 µg/mL), and chloramphenicol (34 µg/mL) and cultured overnight at 37°C. Ten microlitres of culture was transferred to 2 mL of a TB-based autoinduction medium containing 2 mg/mL l-arabinose and cultured at 37°C for 2 h and further cultured at 16°C for 22 h. The culture was centrifuged at 5,000 × g and 4°C for 15 min, and the supernatant was extracted twice with 1 mL of ethyl acetate. The organic layer was collected and evaporated under a nitrogen atmosphere. The samples were dissolved in 100 µL of methanol, and a portion of the solution (1 µL) was analysed using an Agilent 7890A GC equipped with an HP-5ms column (30 m × 0.25 mm i.d.; 0.25 μm film thickness, Agilent Technologies). The column oven was programmed from 60°C for 2 min, then at 10°C min⁻¹ to 290°C, and maintained for 5 min. Helium was used as the carrier gas at a flow rate of 1.0 ml min⁻¹. MS was performed simultaneously using a GC system coupled with an Agilent 5975C inert XL EI/CI MSD equipped with a triple-axis detector (Agilent Technologies). All mass spectra were acquired in electron impact mode (ionisation energy: 70 eV). (E/Z)-PAOx was identified from the mass spectra, retention times, and comparisons with authentic compounds.

Substrate specificity of ChuaMOxS (CHUA_003298)

E. coli BL21(DE3) carrying pGro7 (Takara) and p28-ChuaMOxS was inoculated into LB containing 1% (w/v) d-glucose, kanamycin (50 µg/mL), and chloramphenicol (34 µg/mL) and was cultured overnight at 37°C. Twenty microlitres of the culture was transferred to 2 mL of a TB-based autoinduction medium containing 2 mg/mL l-arabinose and cultured at 37°C for 2 h and further cultured at 16°C for 22 h. A portion (500 µL) of culture was mixed with an equal volume of 40% acetonitrile containing 0.2% formic acid and centrifuged at 21,400 × g and 4°C for 15 min. The supernatant was collected, and the presence of PAOx, 4-hydroxyPAOx, or IAOx in the medium was detected using an LC–MS system equipped with a COSMOSIL 3C18-EB packed column (100 × 2.0 mm i.d., particle size 3 µm; Nacalai Tesuque). The separation conditions were as follows: column oven temperature, 40°C; mobile phase A, 0.1% formic acid in water; mobile phase B, acetonitrile; 20–98% linear gradient of B for 7.5 min and 98% B for 2.5 min, delivered at 0.4 mL/min. MS was simultaneously performed in positive-ion mode, using an LCMS-2020 apparatus (Shimadzu), via electrospray ionisation (interface voltage, 3 kV; interface temperature, 300°C; DL temperature, 250°C; heat block temperature, 400°C; nebulising gas, 3 L/min; drying gas, 10 L/min; heating gas, 10 L/min). Aldoximes ionised in positive-ion mode were monitored in the extracted ions: PAOx, m/z 136 [M + H]⁺; 4HPAOx, m/z 152 [M + H]⁺; and IAOx, m/z 175 [M + H]⁺. The (E/Z)-PAOx accumulated in the medium was quantified using standard curves generated from authentic compounds.

Phylogenetic analysis

Phylogenetic analyses of HNL, P450s, and FMOs were performed. Multiple sequence alignments were performed using MAFFT software [77]. A phylogenetic tree was constructed using the maximum-likelihood method and RAxML-NG [78] with the best-fit amino acid substitution model (HNL; WAG + I + G4 + F, P450; LG + I + G4 + F, FMO; LG + G4 + F) determined based on ModelTest-NG [79] and the Akaike Information Criterion. The tree was evaluated using bootstrap analysis with 500 (P450) or 1000 (HNL and FMO) replicates.

Statistical analysis

Statistical analysis was performed using the multcomp package [80] in R (version 4.12), and differences were analysed using Welch’s t-test or Tukey’s honest significant difference test. Differences were considered statistically significant at P < 0.05.