key: cord-0429006-qf7d827j
authors: Del Rio Flores, Antonio; Twigg, Frederick F.; Du, Yongle; Cai, Wenlong; Aguirre, Daniel Q.; Sato, Michio; Dror, Moriel J.; Narayanamoorthy, Maanasa; Geng, Jiaxin; Zill, Nicholas A.; Zhang, Wenjun
title: Total Biosynthesis of Triacsin Featuring an N-hydroxytriazene Pharmacophore
date: 2021-05-13
journal: bioRxiv
DOI: 10.1101/2021.05.12.443849
sha: 745fb0ecdd1ad239090a549279927be011fd8781
doc_id: 429006
cord_uid: qf7d827j

Triacsins are an intriguing class of specialized metabolites possessing a conserved N-hydroxytriazene moiety not found in any other known natural products. Triacsins are notable as potent acyl-CoA synthetase inhibitors in lipid metabolism, yet their biosynthesis has remained elusive. Through extensive mutagenesis and biochemical studies, we here report all enzymes required to construct and install the N-hydroxytriazene pharmacophore of triacsins. Two distinct ATP-dependent enzymes were revealed to catalyze the two consecutive N-N bond formation reactions, including a glycine-utilizing hydrazine-forming enzyme, Tri28, and a nitrous acid-utilizing N-nitrosating enzyme, Tri17. This study paves the way for future mechanistic interrogation and biocatalytic application of enzymes for N-N bond formation.

The triacsins (1-4) are a unique class of natural products containing an 11-carbon 26 alkyl chain and a terminal N-hydroxytriazene moiety (Fig. 1) . Originally discovered from 27 Streptomyces aureofaciens as vasodilators through an antibiotic screening program 1 , the 28 triacsins are most notable for their use as acyl-CoA synthetase inhibitors by mimicking 29 fatty acids to study lipid metabolism 2-11 . Recently, triacsin C has also shown promise to 30 inhibit SARS-CoV-2-replication, where morphological effects were observed at viral 31 replication centers 12 . Despite potent activities, the triacsin biosynthetic pathway, 32 particularly the enzymatic machinery to install the N-hydroxytriazene pharmacophore, 33 remain obscure.

The N-hydroxytriazene moiety, containing three consecutive nitrogen atoms, is 35 rare in nature and has only been identified in the triacsin family of natural products 13 . In 36 contrast to the prevalence of N-N linkages in synthetic drug libraries, the enzymes for N-37 N bond formation have only started to be revealed recently 14, 15 . Examples include a heme 38 2 enzyme (KtzT) in kutzneride biosynthesis 16 , a fusion protein (Spb40) consisting of a cupin 39 and a methionyl-tRNA synthetase (metRS)-like domain in s56-p1 biosynthesis 17 , an N- 40 nitrosating metalloenzyme (SznF) in streptozotocin biosynthesis 18 , and a transmembrane 41 protein (AzpL) in 6-diazo-5-oxo-L-norleucine formation 19 (Supplementary Fig. 1) . 42 Investigation of the biosynthesis of triacsins that contain two consecutive N-N bonds may 43 determine if known mechanisms are used or if yet new enzymes for N-N bond formation 44 exist. 45 To shed light on triacsin biosynthesis, we identified the triacsin biosynthetic gene 46 cluster (tri1-32) in S. aureofaciens through genome mining and mutagenesis studies 20 . 47 Based on bioinformatic analyses and labeled precursor feeding experiments, we further 48 proposed a chemical logic for N-hydroxytriazene biosynthesis via a key intermediate, 49 hydrazinoacetic acid (HAA) as well as nitrous acid dependent N-N bond formation 20 . In 50 this present study, we scrutinized the activity of 15 enzymes for triacsin biosynthesis 51 through biochemical analyses. We revealed all enzymes required to construct and install 52 the unusual N-hydroxytriazene pharmacophore, highlighting a new ATP-dependent, 53 nitrous acid-utilizing N-nitrosating enzyme.

N-hydroxytriazene assembly starts before PKS extension. 56 Although we inferred from bioinformatic analysis that the acyl portion of the triacsin 57 scaffold is derived from discrete polyketide synthases (PKSs), it was unclear the reaction hydroxytriazene moiety that is further elongated by PKSs. Alternatively, the N-63 hydroxytriazene moiety may serve as a PKS chain terminating agent rather than a "starter 64 unit". The overall odd number of carbon chain length was also puzzling, which was likely N-hydroxytriazene moiety and the terminal carboxylic acid in 5 argues against the 80 possibility of the N-hydroxytriazene moiety as a PKS chain terminating agent. We thus 81 propose that the N-hydroxytriazene assembly starts before PKS extension (Fig. 1) . 82 To confirm that 5 is a direct substrate for Tri10, a UbiD family of non-oxidative 83 decarboxylase, Tri10 was overexpressed in E. coli and purified for biochemical analysis 84 (Supplementary Fig. 3) . Notably, Tri9 is homologous to UbiX, a flavin prenyltransferase 85 that modifies FMN using an isoprenyl unit to generate prFMN, a required co-factor for the 86 activity of UbiD [21] [22] [23] [24] . As expected, the purified Tri10 alone showed no activity toward 5 87 presumably due to the absence of the required co-factor (Fig. 2b) . In contrast, the purified 88 Tri10 that was co-expressed with Tri9 in E. coli successfully decarboxylated 5 based on 89 HRMS analysis with the concomitant production of CO2 (Fig. 2b, Extended Data Fig.   90 1b). While the full structural characterization of the decarboxylated product of 5 was not 91 feasible due to instability of both substrate and product, the UV profile and HRMS data of 92 the reaction product are consistent with a decarboxylated product 6 ( Supplementary Fig.   93 4). In addition, 6 was also identified as a minor metabolite from the culture broth of the 94 wild-type S. aureofaciens (Fig. 2a, Supplementary Fig. 4) . Considering that only 5 with 95 a fully unsaturated alkyl chain was accumulated in Δtri9-10, we propose that 6 is likely a 96 precursor for all observed triacsin congeners varying in the unsaturation pattern of their 97 alkyl chains (Fig. 1) .

Succinylation is required for HAA activation and loading onto ACP. 99 We previously proposed that hydrazinoacetic acid (HAA), presumably generated 100 through the activities of Tri26-28, is a key intermediate for triacsin biosynthesis 20 .

Specifically, N 6 -hydroxy-lysine (7) that is generated by Tri26 (an N-hydroxylase homolog), 102 reacts with glycine, promoted by Tri28 (a cupin and metRS didomain protein), to form N6-103 (carboxymethylamino)lysine (8), which then undergoes an oxidative cleavage to yield 104 HAA and 2-aminoadipate 6-semialdehyde catalyzed by Tri27, an FAD-dependent D-105 amino acid oxidase homolog. The Tri26-28 coupled biochemical assay successfully 106 produced HAA by comparing to an authentic standard through 2,4-dinitrofluorobenzene 107 (DNFB) derivatization 17,25 , confirming the predicted activity of Tri26-28 (Fig. 3a , 108 Supplementary Fig. 5 ). In addition, Tri26 was found to only hydroxylate lysine but not 109 ornithine, and the coupled assay with Tri26/28 generated 8 as expected, further 110 confirming the activity of each enzyme (Extended Data Fig. 2, Supplementary Fig. 5) . 111 We next dissected Tri28 to obtain insight into the roles of the cupin and metRS 112 domains in forming the N-N bond between 7 and glycine. Each domain was individually 113 purified from E. coli and re-combining each domain in the in vitro assay showed 114 successful, albeit less efficient, 8 formation than the intact didomain enzyme, 115 demonstrating the feasibility of domain dissection (Fig. 3b) . Surprisingly, the metRS 116 domain alone was sufficient in generating 8 without the cupin domain, although addition 117 of the cupin domain increased the titer of the product (Fig. 3b) .

After confirming the production of HAA, we next probed the fate of HAA in N-119 hydroxytriazene assembly. Tri26-28 is a part of six-gene operon of tri26-31 that are 120 homologous to spb38-43 in s56-p1 biosynthesis 17 , whereas the functions of tri29-31, 121 putatively encoding an AMP-dependent synthetase, an ACP and an N-acetyltransferase, 122 respectively, remained unknown for both triacsin and s56-p1 biosynthesis. We 123 hypothesized that Tri29 activates HAA through adenylation and then loads it onto Tri30. Supplementary Fig. 6) , suggesting that Tri31 acetylates HAA prior to the catalyzed activation and loading onto ACP. A subsequent assay of Tri31 with HAA and 129 acetyl-CoA, however, did not yield acetyl-HAA (Fig. 3c) . Instead, a new product with 130 mass consistent with succinyl-HAA (9), was detected in a trace amount (Supplementary 131 Fig. 7) . We hypothesized that 9 was generated from succinyl-CoA, which tightly bound to 132 the Tri31 protein purified from E. coli. The LC-HRMS analysis of the purified Tri31 protein 133 solution revealed a trace amount of succinyl-CoA as compared to the standard, and the 134 addition of extra succinyl-CoA to the Tri31 assay increased the yield of 9, consistent with 135 our hypothesis (Fig. 3c, Supplementary Fig. 7) . 136 The preference of Tri31 toward succinyl-CoA over acetyl-CoA was unexpected 137 based on the prediction of Tri31 as an N-acetyltransferase. To exclude the possibility of 138 an artifact due to co-purification of succinyl-CoA with Tri31, we next investigated the acyl-

CoA substrate specificity of Tri31 using various CoA substrates. Tri31 recognized all 140 tested acyl-CoAs except acetyl-CoA, and moderately preferred succinyl-CoA over others 141 based on kinetic analysis (Fig. 3c, Supplementary Fig. 8) . The promiscuity of Tri31 also 142 made possible to confirm that Tri31 acylates the distal nitrogen of HAA, as products such 143 as hexanoyl-HAA were more feasible to be purified from in vitro assays than 9 due to its 144 increased hydrophobicity. The NMR analysis of hexanoyl-HAA showed a 3 J 1 H-15 N HMBC 145 cross-peak between the nitrogen bearing methylene (H 3.69) and the amide nitrogen (N 146 144.7), as well as a 2 J 1 H-15 N HMBC cross-peak between the amide -NH-(H 9.68) and 

The activation and loading of all acyl-HAAs generated by Tri31 were then tested 150 using a coupled assay containing Tri29-31. The succinyl-HAA loaded Tri30 (10) was 151 successfully generated based on both the intact protein HRMS and the 152 phosphopantetheine (Ppant) ejection assay (Fig. 3d) . Interestingly, none of the other 153 acyl-HAAs were loaded onto Tri30 (Supplementary Fig. 9) , indicating that Tri29 acts as 154 a gatekeeper enzyme in the biosynthetic pathway and succinylation is required for HAA 155 activation and loading onto ACP (Fig. 1) .

Hydrazone "starter unit" for PKS extension.

Tri22, an acyl-CoA dehydrogenase homolog, was proposed to oxidize 10 as the 158 next step in N-hydroxytriazene formation. Tri22 was recombinantly expressed in E. coli 159 5 as a yellow protein containing FAD (Supplementary Fig. 10 ). An in vitro assay of Tri22 160 with 10 generated in situ by Tri29-31 resulted in conversion of 10 to a product with mass 161 consistent of 11 (Fig. 4a) . To confirm the proposed structure of 11, the reaction product 162 was subjected to base hydrolysis to release the free acid that was further derivatized by 163 ortho-phthalaldehyde (OPTA) for LC-UV-HRMS analysis 28, 29 . The expected acid product, 164 2-hydrazineylideneacetic acid (2-HYAA), was also chemical synthesized and derivatized 165 by OPTA. By comparison, the structure of 11 was confirmed to be 2-HYAA attached to 166 Tri30 (Extended Data Fig. 3, Supplementary Note 3, Supplementary Fig. 11 ). It is 167 notable that Tri22 showed no activity toward the free acid 9 (Supplementary Fig. 11) , 168 demonstrating the necessity of a carrier protein for Tri22 recognition. that Tri17 may activate nitrite using ATP (Supplementary Fig. 13 ). 183 Although enzyme(s) other than Tri17 were likely responsible for the third nitrogen 184 addition, we reasoned that 11 might not be a correct substrate and the N-hydroxytriazene 185 formation might take place at a later stage, after PKS extension and release of the 186 polyketide intermediate. While the discrete nature and lack of some essential enzymes 187 (likely shared with fatty acid biosynthesis in the native producer) made the in vitro 188 reconstitution of the entire PKS machinery a formidable task, the biochemical assay 189 containing Tri13, a ketosynthase homolog, and Tri20, a free-standing ACP, demonstrated 190 that Tri13 promoted translocation of the 2-HYAA moiety from Tri30 to Tri20 to yield 12 191 (Fig. 4b) . This observation suggests that the PKS extension likely takes place on Tri20 192 with a hydrazone "starter unit". After five rounds of extension to yield 13, the 12-carbon 

We next probed whether nitrous acid, generated enzymatically by Tri16/21, could 222 be utilized by Tri17. A biochemical assay containing Tri16, Tri21 and Tri17 with aspartic 223 acid, NADPH, ATP and 15 resulted in 1 production, and the utilization of 15 N-aspartic acid 224 led to the expected mass spectral shift of 1 (Fig. 5b) , confirming the utilization of nitrous 225 acid generated by Tri16/21 in N-hydroxytriazene biosynthesis. Notably, no spontaneous 226 formation of 1 was observed using either nitrite or enzymatically generated nitrous acid 227 in the absence of Tri17 ( Fig. 5a and 5b) . We thus deduce that in triacsin biosynthesis, 228 Tri17 takes nitrous acid and 14 as substrates, and catalyzes the production of 5 in an 229 ATP-dependent manner (Fig. 1) . Tri17 appeared to be promiscuous toward acyl chain Our extensive bioinformatics, mutagenesis, and biochemical analysis allow a first 238 proposal of the complex biosynthetic pathway for triacsins, particularly the N-239 hydroxytriazene moiety unique to this class of natural products (Fig. 1) . The biosynthesis 240 starts with the production of HAA from lysine and glycine catalyzed by Tri26-28, which is Fig. 17) . Surprisingly, the N-terminal cupin domain is also not required for 8 formation, 276 which is inconsistent with prior mutagenesis results where E. coli producing the Spb40 277 mutant with the substitution of alanine for His45, one of the key residues for metal-binding 278 of the cupin domain that is also conserved in Tri28, abolished the ability to produce 8 17 .

Considering that the cupin domain increased the catalytic efficiency in vitro (Fig. 3b) , we Competing financial interests 341 The authors declare no competing financial interests.

Correspondence and requests for materials should be addressed to W.Z. conditions. Linear gradients of 5-50% acetonitrile (vol/vol) at 2.5 mL/min were used for 502 initial purification. Final purification of 5 was performed using an isocratic gradient at 14% 503 acetonitrile using the same mobile phase, column, and flow rate. All fractions containing 504 the product were kept covered to avoid light exposure and stored under N2 gas whenever 505 practically achievable to avoid chemical degradation. Purified products were dried and 506 analyzed by LC-MS and NMR. All NMR spectra were recorded on a Bruker AVANCE at 507 900 MHz (1H NMR) and 226 MHz (13C NMR). The production and purification procedure 508 of 1 was conducted as previously reported 20 .

Labeled Precursor Feeding of 1-13 C-glycine. S. aureofaciens was cultured as 510 described in a previous section. 24 hours after the 30 mL inoculation, the culture was 511 supplied with 10 mM 1-13 C-glycine. Compound extraction and LC-HRMS analysis were 512 performed as described above. Table S2 . Oligonucleotides utilized in this study were purchased from Integrated 529 DNA Technologies. All oligonucleotides used in this study are listed in Table S1 and all 

All proteins reported in this study were purified using the HEPES pH 8 buffer system, 557 except for Tri29 for which Tris pH 7.5 was used to avoid precipitation upon concentration. 

Tri29-31, and Tri22 was performed as described in a previous section 695 except for 50 μM of each enzyme was utilized and the assay was performed at a 100 μL 696 scale. After the 20-minute incubation period, the reaction was quenched with two volumes 697 of chilled methanol

The solution was spin filtered using a 2 kDa Amicon spin filter to 701 remove protein residues and isolate the flowthrough. The flowthrough was subjected to 702 OPTA derivatization 28 . The derivatization agent was prepared by mixing 1 mL of 80 mM 703 OPTA with 98 mM of 2-ME (both dissolved in 200 mM NaOH). 100 μL of the flowthrough 704 was combined with 100 μL of the derivatization agent. The reaction mixture was vortexed 705 and allowed to incubate at room temperature for 2 minutes. The mixture was analyzed 706 using LC-UV-MS

Agilent Eclipse Plus C18 column (4.6 x 100 mm). A water/acetonitrile mobile phase with 708

2-HYAA was chemically synthesized and subjected to the same 710 derivatization procedure with OPTA. The mass spectrum, retention time, and UV at were 711 compared to that of the standard. At least three independent replicates were performed 712 for each assay

Analysis of 2-HYAA Translocation to Tri20 with the Aid of Tri13. Reactions 714 were performed at room temperature for 30 minutes in 50 μL of 50 mM Tris pH 7.5, 1 mM 715 succinyl-CoA, 2 mM HAA, 5 mM ATP, 1 mM FAD, 2 mM MgCl2

After the incubation 717 period, the reaction was diluted with 4 volumes of purified water and chilled on ice to slow 718 further reaction progress. The diluted reaction mixture was immediately filtered by a 0.45 719 μm PVDF filter, and 15 μL was promptly injected and analyzed using the same 720 methodology in a previous section with the exception that a

1 x 250 mm) was used. At least two independent replicates 722 were performed for each assay, and representative results are shown

CoA Loading Assays on Tri30 and Tri20 using Sfp

Reactions were performed at room temperature for 30 minutes in 50 μL of 50 mM HEPES 725 pH 8.0, 2.5 mM CoA substrate, 2 mM MgCl2, 100 μM ACP, and 50 μM Sfp. After the 30-726 minute incubation period, the reaction was diluted with 4 volumes of purified water and 727 chilled on ice to slow further reaction progress. The diluted reaction mixture was 728 immediately filtered by a 0.45 μm PVDF filter

Agilent Technologies 6545 Q-TOF LC-MS equipped with a Phenomenex Aeris 3.6 μm 730

1 x 250 mm) and Phenomenex Aeris 3.6 μm Widepore XB-731 C18 column (2.1 x 100 mm) for lauroyl-CoA and hexanoyl-CoA assays

1% (vol/vol) formic acid, analysis was 733 performed with a linear gradient of 30-98% and 30-50% acetonitrile for lauroyl-CoA and 734 hexanoyl-CoA assays, respectively at a flow rate of 0.25 mL/min. The analysis to 735 deconvolute the protein spikes and obtain the Ppant fragment are analogous to a previous 736 section. Hexanoyl-CoA and lauroyl-CoA were loaded to Tri20

Reactions were 740 performed at room temperature for 1 hour in 100 μL of 50 mM HEPES

After the 1-hour 742 incubation period, the reaction was quenched with two volumes of chilled methanol. The 743 precipitated protein was removed by centrifugation (15,000 x g, 5 min) and the 744 supernatant was used for analysis. LC-HRMS analysis was performed using an Agilent 745

A water/acetonitrile mobile phase with 0.1% (vol/vol) formic acid with a 747 linear gradient of 2-98% acetonitrile at a flow rate of 0.5 mL/min was utilized

Assays with hexanoyl-CoA and lauroyl-CoA were utilized with Tri20, while lauroyl-CoA 750 was utilized solely for Tri30. At least three independent replicates were performed for 751 each assay

Griess Test for Nitrous Acid Production by Tri21 and Tri16. Reactions were 753 performed at room temperature in 100 μL of 50 mM Tris (pH 7.5) containing 0.1 mM FAD, 754 1 mM NADPH or NADH, 1 mM aspartate, 20 μM Tri21, and 20 μM Tri16. The reaction 755 was prepared with all reagents except NAD(P)H and aliquoted identically into 12 wells

Reactions were initiated simultaneously by the addition of NAD(P)H with a multichannel 757

Starting with an initial time point, one reaction well was quenched every 4 minutes 758 by the addition of one volume of Griess reagent purchased from Cell Signaling 759 Technology

After the incubation period, the reaction was 765 quenched with two volumes of chilled methanol. The precipitated protein was removed 766 by centrifugation (15,000 x g, 5 min) and the supernatant was used for analysis. LC-767 HRMS analysis was performed using an Agilent Technologies 6545 Q-TOF LC-MS 768 equipped with an Agilent Eclipse Plus C18 column (4.6 x 100 mm). A water/acetonitrile 769 mobile phase with 0.1% (vol/vol) formic acid with a linear gradient of 2-98% acetonitrile 770 at a flow rate of 0.5 mL/min was utilized. 1 isolated from WT S. aureofaciens was utilized 771 as a standard to compare the retention time, mass spectrum, and UV profiles between 772 the biochemical assays. At least three independent replicates were performed for each 773 assay, and representative results are shown

Coupled Tri16, Tri21, and Tri17 Activity Assay. Reactions were performed at room 777 temperature for 1 hour in 100 μL of 50 mM Tris pH 7.5, 0.1 mM FAD, 1 mM NADPH, 1 778 mM aspartate (or 2 mM 15 N-aspartate), 20 μM Tri16, and 20 μM Tri21. After the incubation 779 period, 100 μL of 50 mM Tris pH 7.5, 0.5 mM 15, 5 mM ATP, 2 mM MgCl2, and 100 μM

Tri17 was added to the first reaction. The coupled assay was incubated at room

After the incubation period, the reaction was quenched with two 782 volumes of chilled methanol. The precipitated protein was removed by centrifugation 783 (15,000 x g, 5 min) and the supernatant was used for analysis. LC-HRMS analysis was 784 performed using an Agilent Technologies

acid with a linear gradient of 2-98% acetonitrile at a flow rate of 0.5 mL/min 787 was utilized. 1 isolated from WT S. aureofaciens was utilized as a standard to compare 788 the retention time, mass spectrum, and UV profiles between the biochemical assays. At 789 least three independent replicates were performed for each assay

Tri17 assays utilizing nitrate, 2-HYAA, or 12-792 aminododecanoic acid were analyzed via LC-HRMS using an Agilent Technologies 6545

A 794 water/acetonitrile mobile phase with 0.1%

98% acetonitrile at a flow rate of 0.5 mL/min was utilized. Peak picking and comparative 796 metabolomics were performed using MSDial with peak lists exported to Microsoft Excel

AMP Detection from Tri17 Assay. Reactions were performed at room temperature for 798 30 minutes as described in the Tri17 Activity Assay on 15 section. After the incubation 799 period, the reaction was quenched with two volumes of chilled methanol. A similar assay 800 was performed using 11 (generated enzymatically as described in a previous section) as Detection of CO2 production from the Tri10 9 biochemical assay. GC-MS chromatograms 906 extracted with m/z=44, demonstrating CO2 production from a Tri10 9 assay containing 5 907 as compared with an authentic standard. A 10-ppm mass error tolerance was used for 908 each trace. At least two independent replicates were performed for each assay, and 909 representative results are shown.