Synthesis and Activity‐Based Protein Profiling Identifies Aldo‐Keto Reductase 1C3 as Target Protein of Myxoglucamides

ABSTRACT

Myxoglucamides, natural products recently isolated from Cystobacterineae sp., are featured by an unprecedented vinyl‐substituted α‐keto‐γ‐amino acid that is linked to a glycosylated 14‐methyl‐pentadecanoic acid. Thus, they unite elements from three biomolecular classes in a compact glycolipopeptide. To elucidate the biological relevance of this arrangement, we searched for molecular targets by activity‐based protein profiling (ABPP)—although a phenotypic bioactivity has not been reported. An access to the compound class was established through the first total synthesis of myxoglucamide A in 11 steps. A proteome‐wide ABPP study led to the identification of aldo‐keto reductase 1C3 (AKR1C3) as the primary target of myxoglucamides in human cells. AKR1C3 is an oncogenic factor involved in prostaglandin and steroid synthesis, promoting the growth, proliferation, and metastasis of carcinoma cells. The functional inhibition of AKR1C3 by a competitive mechanism (IC₅₀ = 1.61 µm) was validated in vitro, and 20 analogs provided structure‐activity relationships and more potent analogs (IC₅₀ = 181 nm). Biophysical interactions were quantified by thermal shift assays, and essential molecular protein‐ligand interactions were characterized by X‐ray crystallography at 2.0 Å resolution. The study implies that the search for targets of natural products is rewarding even in the absence of an initial phenotypic activity.

Myxoglucamide A—a novel glycolipopeptide isolated from myxobacteria—was synthesized, and activity‐based protein profiling identified aldo‐keto reductase 1C3 (AKR1C3) as its main target in human cells. Structure‐activity relationship studies, together with biophysical characterization and X‐ray crystallography, elucidated the structural basis of AKR1C3 inhibition.

1. Introduction

The study of natural products has been highly beneficial for the progression of biological research [1, 2, 3]. They have not only been a highly prolific source of new drugs until today [4], but also disclosed original cellular targets and mechanisms that were subsequently exploited therapeutically by the natural product itself, or by other molecular formats [5]. The starting point of cellular mode‐of‐action studies is usually a strong phenotypic activity, such as the inhibition of growth of a cancer cell or of pathogenic microbes. However, the absence of a strong phenotype induced by a natural product and the limited availability of phenotypic assays leaves the biological function of the compound undisclosed in the majority of cases—and this uncharted biological space is regarded as one of the most eminent problems in today's natural product research [6].

We followed the hypothesis that an evolutionary optimization of complex natural products, whose production is metabolically costly, must serve a function. Thus, we set out to elucidate the bioactivity of natural products even in the absence of a cell death phenotype. Instead, original chemical structures of metabolites and an underexplored habitat were main selection criteria for a mechanistic study, as recently exemplified by our study of the volatile marine salinilactones [7]. The current study exploits the bioactivity of secondary metabolites produced by myxobacteria, Gram‐negative species predominantly found in soil that are characterized by their complex life cycles, social behaviors such as cooperative hunting, fruiting body formation, and cellular differentiation [8]. These microorganisms express a diverse array of natural products with significant pharmacological potential in multiple indications [9]. To further uncover the enormous genomic potential of myxobacteria [10], Wang et al. followed a genome mining strategy [11] that led to the discovery of a novel family of natural products isolated from Cystobacterineae sp. MCy9003 and named myxoglucamides (Figure 1) [12]. As shown in this issue of Angew Chem, the structure of myxoglucamide A (1) features an unprecedented γ‐amino acid‐containing α‐ketoamide with a vinylic side chain, linked to 13β‐d‐glucopyranosyl‐14‐methyl‐pentadecanoic acid. Because the myxoglucamides combine elements of three major biomolecular classes (lipids, sugars, peptides) in a compact molecule, we aimed to find out the bioactivity exerted by this arrangement by applying an unbiased activity‐based protein profiling (ABPP) approach. We speculated that the myxoglucamides bind covalently to their target(s) through the α‐ketoamide, given that this moiety is a widespread electrophilic warhead in natural and synthetic peptide‐based drugs. Herein, we developed a total synthesis of myxoglucamides that enabled the identification and validation of their mammalian target.

FIGURE 1.

Structure of myxoglucamides.

2. Results and Discussion

For an unbiased target identification of myxoglucamides using activity‐based protein profiling (ABPP) with chemical probes [13, 14, 15, 16], a synthetic access to the natural products needed to be established first. We chose myxoglucamide A (1) as the target molecule for a total synthesis. Our retrosynthetic considerations resulted in a convergent strategy in which a glycosylated fatty acid (II) and a peptidyl building block (III) would be coupled in the final steps (Scheme 1A). The peptide fragment was accessible using commercially available, enantiomerically pure starting materials derived from amino acids (IV and V). For the fatty acid building block, we planned to install the chiral hydroxy group via enantioselective synthesis.

SCHEME 1.

(A) Retrosynthetic analysis of myxoglucamide A (1), (B) Synthesis of 1. PG = protecting group, Boc = tert‐butyloxycarbonyl, Ipc = isopinocampheyl, TBDPS = tert‐butyldiphenylsilyl, THF = tetrahydrofuran, DMP = Dess‐Martin periodinane, HMDS = bis(trimethylsilyl)amine, TBAF = tetrabutylammonium fluoride, TMSOTf = trimethylsilyl trifluoromethanesulfonate, HATU = O‐(7‐azabenzotriazol‐1‐yl)‐N,N,N′,N′‐tetramethyluronium hexafluorophosphate, DIPEA = N,N‐diisopropylethylamine.

A Brown allylation [17] of isobutyraldehyde (5) furnished the (S)‐allylic alcohol 6 with an enantiomeric ratio of 82:18, as determined by Mosher ester analysis [18] (Scheme 1B and Supporting Information). After silyl protection of the alcohol, the terminal olefin was subjected to hydroboration‐oxidation to yield the aldehyde 7. A subsequent Wittig reaction with the phosphonium benzyl‐nonanoate 8 extended the carbon chain to the desired length. Following silyl deprotection, glycosylation of the resulting alcohol (9) was achieved using Schmidt's method with trichloroacetimidate 10 to selectively install the desired β‐d‐glycosyl moiety [19]. The synthesis of the fatty acid building block 11 was completed by simultaneous hydrogenation of the alkene and cleavage of the benzyl ester.

The synthesis of the peptide fragment started from commercially available Boc‐protected homoserine lactone (12). Reduction with diisobutylaluminium hydride (DIBAL‐H), followed by Wittig methylenation and Dess‐Martin oxidation afforded aldehyde 13 [20, 21]. Subsequently, the α‐ketoamide 15 was directly accessible via an oxidative coupling with isocyanide 14. We applied a method developed by Grassot et al. [22]. that involved the condensation of N‐methyl‐hydroxylamine with aldehyde 13, followed by an Ugi‐type reaction with isocyanide 14 and acetic acid, that led to an isolable α‐acetoxyamino amide intermediate. Mediated by 4 Å molecular sieves (4 Å MS), β‐elimination and hydrolysis resulted in the formation of α‐ketoamide 15. Finally, amide coupling with the fatty acid fragment 11, followed by global deprotection, afforded myxoglucamide A (1). The synthesis of 1 was achieved in 11 steps in the longest linear sequence with an overall yield of 2.4% and 16 steps in total.

To enable ‘target fishing’ by ABPP, myxoglucamide was linked to biotin. The design of the activity‐based probe 16 (Figure 2A) [15, 16] was built on the hypothesis that the isopropyl group of 1 was least involved in target binding and could be replaced with an alkyne as a handle for click chemistry. The synthesis of the alkynylated fatty acid started with the epoxidation of ethyl 13‐tetradecenoate (17), followed by Jacobsen hydrolytic kinetic resolution, to obtain the enantiopure epoxide 18 [23]. Epoxide opening with lithiated (triisopropylsilyl)acetylene provided the desired (S)‐configured alcohol with >99% ee as confirmed by Mosher ester analysis. After hydrolysis and silyl deprotection, carboxylic acid 19 was directly coupled with amine 15. At this stage, the Schmidt glycosylation proved to be difficult, resulting in very low yield due to incomplete reaction and partial transacetylation. Nevertheless, the synthesis of the activity‐based probe 16 was completed via a copper‐catalyzed azide–alkyne cycloaddition with biotin–PEG_3–azide [24].

FIGURE 2.

Target identification by ABPP. (A) Synthesis of activity‐based probe (ABP) 16. (B) ABPP workflow. Graphic created with BioRender. (C) Volcano plot of proteins identified by tandem mass spectrometry after enrichment by ABPP in A549 lung epithelial cell lysate (positive control: n = 5, negative control: n = 5). The thresholds for significantly enriched proteins were set to p = 0.01 and log₂ (fold change) = 2 and visualized with dashed grey lines. (D) Competitive ABPP. A549 cell lysate was incubated with 1 (n = 3) or DMSO (n = 3) for 30 min prior to the addition of ABP 16. The resulting log2‐transformed peak areas of the previously identified proteins are listed in (D). Proteins that were significantly enriched and showed binding to 1 in the competitive ABPP are labeled as red dots in (C). Proteins that exhibited a higher enrichment after preincubation with 1 are marked in gray. mCPBA = meta‐chloroperoxybenzoic acid, TIPS = triisopropylsilyl, DMF = N,N‐dimethylformamide, PEG = polyethylene glycol, THPTA = tris(3‐hydroxypropyltriazolylmethyl) amine, DMSO = dimethylsulfoxide.

The anomeric center of 16 was β‐configured, as shown by a doublet at 4.27 ppm with a characteristic coupling constant of J  =  7.9 Hz in the ¹H NMR spectrum.

To search for human protein targets of myxoglucamides, we selected the well‐characterized A549 lung epithelial cell line as a model system for ABPP, also because it is cultivated in a reproducible manner in our hands. Probe 16 was incubated with an A549 cell lysate for 3 h (Figure 2B). In parallel, lysates treated with DMSO served as negative controls to account for nonspecific binding artifacts. A pulldown of probe‐bound targets was performed using streptavidin‐coated Sepharose beads, leveraging the high affinity between biotin and streptavidin. After washing and centrifugation steps to remove unbound material, probe‐bound proteins were eluted under denaturing conditions. For their identification by nano‐liquid chromatography coupled tandem mass spectrometry (nano‐LC‐MS/MS), the samples were prepared using an adapted SP3 (single‐pot, solid‐phase‐enhanced sample preparation) protocol [7, 25]. The resulting MS/MS data were analyzed against a human protein database and visualized in a volcano plot (Figure 2C). Several proteins exhibited significant enrichment compared to the negative control, with the highest observed for aldehyde dehydrogenase 1A1 (AL1A1). The aldo‐keto reductase 1C3 (AKR1C3) demonstrated the strongest statistical significance, reflected by the highest –log₁₀(p) value. Notably, AKR1B10 and AKR1C2, two other members of the aldo‐keto reductase family, were significantly enriched. In addition to these oxidoreductases, several tubulin α‐chains (1A–C, 3C–E, 4A–B, and 8) and keratins (type II Hb1/3/5/6) were identified.

In order to confirm the specific binding of 16 to these proteins, we performed a competitive ABPP. For this purpose, A549 cell lysates were pre‐incubated with an excess of 1 (or DMSO as a negative control) for 30 min. Subsequently, ABP 16 was added to all samples, and the previously described ABPP workflow was carried out. Proteins specifically bound by myxoglucamides were then indicated by a decrease in their peak abundances following pretreatment with 1. The proteins AKR1C3, AKR1B10 and keratins II Hb1/3/6 (KRT81/83/86) were no longer enriched after pre‐incubation, indicating complete saturation of the binding site by 1 (Figure 2D). AL1A1 was detectable in both conditions, but slightly less abundant under pre‐treatment conditions. A specific binding to tubulin‐α chains, except for TBA4A/B, could not be confirmed in the competition experiment.

In summary, the chemoproteomics data implied that myxoglucamides may target keratin and in particular oxidoreductases. We decided to focus further investigations on the latter. In order to validate the ABPP results, we probed whether 1 had any functional effect on the identified oxidoreductases. This was tested by six enzymatic in vitro assays that measured the catalytic conversion of model substrates in the presence of 1 by quantifying the cofactor NADPH. Aldehyde dehydrogenase 1A1 was not inhibited by 1 without pre‐incubation or after just 1 h of pre‐treatment. However, an IC₅₀ value of 210 µm was observed after 6 h of pre‐incubation (Table 1). These results were consistent with the competitive ABPP data, where only a moderate reduction in enrichment was seen following the short pre‐incubation period. Since several aldo‐keto reductases have been identified by ABPP, we also investigated the inhibitory activity of 1 against the structurally related family members AKR1B1, AKR1C1, although these were not detected in the ABPP experiment. AKR1B1 and AKR1B10 were only inhibited by very high concentrations of 1, while AKR1C1 and AKR1C2 showed no inhibition even at the highest tested concentration (IC₅₀> 250 µm). In contrast, AKR1C3 was strongly inhibited by 1 with an IC₅₀ value of 1.61 µm. Aldo‐keto reductase family 1 member C3 (AKR1C3; aka 17 β‐HSD5) is an NADP⁺/NADPH‐dependent oxidoreductase that plays a crucial role in steroid metabolism by catalyzing the final reduction step in the biosynthesis of testosterone and 17β‐estradiol (Scheme S1) [26]. Additionally, AKR1C3 functions as a prostaglandin F2α synthase [27]. Overexpression of AKR1C3 has been observed in various types of cancer (e.g., prostate, breast, lung cancer, leukemia) and acts as an oncogenic factor, promoting the growth, proliferation, and metastasis of carcinoma cells [28].

TABLE 1.

Evaluation of inhibitory activity of 1 against identified target proteins.

Protein	IC50 [µm]
AL1A1| Aldehyde Dehydrogenase 1 A1	210 ± 7 a
AKR1B1| Aldo‐Keto Reductase 1 B1	221 ± 36 b
AK1B10| Aldo‐Keto Reductase 1 B10	348 ± 136 b
AKR1C1| Aldo‐Keto Reductase 1 C1	> 250 c
AKR1C2| Aldo‐Keto Reductase 1 C2	> 250 c
AKR1C3| Aldo‐Keto Reductase 1 C3	1.61 ± 0.14 c

^a NAD⁺/MTT‐coupled oxidation of acetaldehyde, 6 h preincubation with 1.

^b NADPH‐coupled reduction of dl‐glyceraldehyde.

^c NADP⁺‐coupled oxidation of (S)‐tetrahydronaphthol. MTT = (3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazoliumbromid).

AKR1C3 has also been investigated as a treatment of endometriosis [27]. Therefore, the discovery and development of selective inhibitors has become an area of intense research that led to the identification of natural products, steroids, nonsteroidal anti‐inflammatory drugs (NSAIDs), fatty acids, organic ruthenium complexes, off‐label drugs, other small‐molecule heterocycles, or proteolysis‐targeting chimeras (PROTACs) [28, 29, 30, 31]. Clinical studies with the steroidal BAY1128688 and the indol ASP9521 were initiated, but discontinued due to off‐target toxicity or lack of efficacy [28].

Based on the selectivity of 1 for AKR1C3 and its relevance as a therapeutic target, we focused our efforts on this protein to establish structure–activity relationships (SAR) of myxoglucamids (Table 2). We found that the glucose moiety was not essential for inhibition, as the corresponding aglycon 20 was more potent. Consequently, the role of the fatty acid substituent was investigated further. Complete removal of the alkyl chain as in 21 or substitution with only short alkyl groups as in 22 significantly reduced activity.

TABLE 2.

Structure‐activity relationship study of aldo‐keto reductase 1 C3 (AKR1C3).

#	R	IC50 [µm] b	CC50 [µm] c	SI d	#	R	IC50 [µm] b	#	R	IC50 [µm] b	CC50 [µm] c	SI d
1	R1	1.61	> 100	> 62	33	R2	22.8	37	R1	0.501	> 100	> 199
20	R2	0.947	> 100	> 106				38	14:0 a	0.314	40.4	129
21	C(O)Ot‐Bu	> 128	n.d.
22	6:0 a	15.6	n.d.
23	10:0 a	0.855	> 100	> 116	2	R1	> 64
24	14:0 a	0.344	55.9	163	34	14:0 a	14.0
25	18:0 a	0.573	25.7	45
26	16:1 cis −9 a	0.181	53.6	296
27	18:1 cis −9 a	0.208	32.5	156
28	22:1 cis −13 a	1.54	21.8	14	35	R1	47.5
29	18:2 cis− 9,12 a	0.242	51.6	213
30	18:3 cis −9,12,15 a	0.196	59.3	303
31	20:4 cis− 5,8,11,14 a	0.432	44.2	102
32	22:6 cis− 4,7,10,13,16,19 a	0.246	42.3	172	36	14:0 a	22.6

^a Fatty acid—number of carbon atoms: double bonds  ^{Configuration ‐Location.}

^b Mean of n  =  3, AKR1C3: 132 nm, (S)‐tetrahydronaphthol: 1 mm, NADP⁺: 0.5 mm.

^c Cytotoxicity in A549 cells determined by Alamar Blue assay. Mean of n = 3.

^d Selectivity Index.

In contrast, longer fatty acid chains markedly improved potency, reaching nanomolar IC₅₀ values (23‐25). The strongest inhibition was observed with unsaturated fatty acid substituents (26‐32), consistent with prior reports that AKR1C3 is potently inhibited by unsaturated fatty acids [32]. We next examined the importance of the carboxylic acid functionality. Its conversion into a methyl ester (33) led to a >20‐fold decrease in activity. A similar trend was observed for the ketoamide moiety: reduction (2, 34, 35) or removal (36) of the carbonyl group resulted in a significant loss of potency.

Interestingly, saturation of the vinyl group had no observable effect on inhibitory activity (37, 38).

Next, we evaluated the cytotoxic effect of myxoglucamide A and its analogs on A549 cells. While 1 showed no cytotoxicity up to a concentration of 100 µm, several of its derivatives displayed CC₅₀ values ranging from 20 to 60 µm. However, the ratio between the concentration required for enzymatic inhibition and for cellular cytotoxicity, termed selectivity index (SI) here, was still high. The only analogs with an SI below 100 were 25 and 28, which feature the longest alkyl chains with the lowest degree of unsaturation. In the enzyme inhibition assay, these derivatives achieved only 60–70% inhibition, even at very high concentrations, suggesting potential solubility issues. These solubility limitations are likely also a contributing factor to their increased cytotoxicity. Moreover, the fact that the potency of AKR1C3 inhibition did not correlate with cytotoxicity speaks against an on‐target effect, but for a lipophilicity‐related off‐target phenomenon.

To gain a deeper understanding of the mechanism of AKR1C3 inhibition, we conducted an enzyme kinetic study. Myxoglucamide A exhibited competitive inhibition relative to the substrate (S)‐tetrahydronaphthol, as evident from the Michaelis‐Menten and Lineweaver‐Burk plots (Figure 3A,B). Additionally, we determined the substrate‐independent inhibition constant (K_i) for 1 and the most active analogs 24, 26, and 30. As these compounds operate in the tight‐binding regime, conventional Michaelis–Menten‐based analyses are not applicable. Therefore, K_i values were determined using the Morrison equation, yielding inhibition constants in the low‐nanomolar range, with compound 26 displaying a K_i value of only 20 nm (Figure 3F).

FIGURE 3.

Michaelis‐Menten enzyme kinetic study and thermal shift assays for the binding of 1 to AKR1C3. (A) Michaelis‐Menten enzyme kinetics study for the inhibition of AKR1C3 by 1. Competitive inhibition fit performed using GraphPad Prism 10.3.0. Conditions: AKR1C3: 132 nm, (S)‐tetrahydronaphthol: 0.25 – 2 mm, 1: 0 – 16 µm, NADP⁺: 0.5 mm. (B) Lineweaver‐Burk plot of enzyme kinetics measurement from (A). (C) Thermal shift assay (nanoDSF) for T_M determination of binary and ternary complexes between AKR1C3, 1 and NADP⁺. AKR1C3 (12 µm) was incubated with or without 200 µm of 1 and NADP⁺, each. (D) Thermal shift assay of AKR1C3 (13 µm) with increasing concentrations of 1. (E) Fraction of unfolded protein was calculated from D) using “FoldAffinity” [33, 34] at a selected temperature of 57°C and plotted against the concentration of 1. K_d values were estimated and are listed in F) as the mean of n = 3. (F) Inhibition constants K_i and dissociation constants K_d for selected myxoglucamide analogs. K_i values (mean of n  =  2) were determined from Michaelis‐Menten kinetics and fitted with GraphPad Prism 10.3.0 using the Competitive Inhibition model for 1, and the Morrison K_i model for inhibitors 24, 26 and 30. Conditions: AKR1C3: 132 nm, (S)‐tetrahydronaphthol: 0.25 – 2 mm, NADP⁺: 0.5 mm. P = AKR1C3, I = Inhibitor, N = NADP⁺.

Next, we investigated the binding affinity of 1 to AKR1C3. For this purpose, thermal shift assays using nano differential scanning fluorimetry (nanoDSF) were employed that measured the intrinsic fluorescence of the protein at increasing temperatures. Ligand binding typically stabilizes the protein, resulting in an increase in the denaturation temperature (T_M). We determined the T_M shift of the binary complexes 1 – AKR1C3 and NADP⁺—AKR1C3 as well as the ternary complex. The T_M shift of the ternary complex was larger than the sum of the individual T_M shifts, indicating a positive cooperative effect on protein stabilization (Figure 3C and Table S6).

To further study this effect, we estimated the dissociation constant (K_d), by measuring the thermal shift with increasing ligand concentrations (Figure 3D). By fitting the fraction of unfolded protein, obtained by the FoldAffinity method developed by Niebling et al. [33], against the initial ligand concentration, the K_d value can be roughly estimated at temperatures near the T_M. From the thermal shift assay of the binary complex of 1 with AKR1C3, we obtained a K_d value of 40 µm (Figure 3E) [34]. The same methodology was applied for the ternary complex and resulted in an estimated K_d value of 0.33 µm. Thus, a tight binding of 1 is only achieved in the presence of the cofactor. In line with the SAR study, the K_d values of the most active derivatives 24, 26, and 30 were approximately two to four times lower than that of 1 (Figures 3F and S7).

Finally, we were able to obtain an X‐ray cocrystal structure of AKR1C3 in complex with 1 and NADP⁺ at a resolution of 2.0 Å (Figure 4). The protein crystallized as a dimer with myxoglucamide bound in both monomers. While the electron density of the peptide part was well‐defined and consistent in both chains, the glucose moiety was observed only in chain B with reasonable electron density (Figures 4A and S8). This is likely due to the high flexibility of the alkyl linker, and it reflects our observation that a glycosylated chain is not required for activity. AKR1C3 features a large and very hydrophobic active site that has been divided into the following subsections: the oxyanion site formed by the catalytic residues Tyr55, His117 and NADP⁺, a steroid channel (Tyr24, Leu54, Ser129, Trp227) and the three sub‐pockets SP1 (Ser118, Asn167, Phe306, Phe311, Tyr319), SP2 (Trp86, Leu122, Ser129, Phe311) and SP3 (Try24, Glu192, Ser221, Tyr305) [35]. The carboxylic acid of 1 was located at the oxyanion site, forming key hydrogen bonds to the catalytic residues Tyr55 (2.6 Å) and His117 (3.0 Å). This is in line with the observation that the carboxylic acid was essential for AKR1C3 inhibition (Figure 4B). The α‐ketoamide did not participate in any hydrogen bonding or covalent interactions. However, it rigidified the peptidyl moiety by favoring an almost planar conformation due to the antiperiplanar alignment of the adjacent carbonyl groups (Figure 4B). The vinyl group and the adjacent amide extended into the SP1 pocket, with the amide nitrogen atom forming a hydrogen bond to Ser118. Compared to the other AKR1C isoforms, the SP1 site of AKR1C3 is significantly larger, primarily due to a shift in the backbone conformation of residues 305–311 (Figure 4C).

FIGURE 4.

X‐ray crystal structure of AKR1C3 in complex with 1 and NADP⁺ (PDB: 9SWH). (A) 2F_o–F_c electron density map of 1 and NADP⁺ contoured at 1.0 σ, shown as blue mesh within the AKR1C3 surface of chain B; 1 in green and NADP⁺ in yellow sticks. (B) Peptidyl binding site. The residues of the subsections are color‐coded: cyan = oxyanion site, orange = SP1. Hydrogen bonding interactions are shown as dashed lines. (C) Superposition of the SP1 pockets of AKR1C1 (PDB: 3NTY) [36], AKR1C2 (PDB: 4XO7) [37], and AKR1C3. (D) Localization of the glycosylated alkyl moiety. The residues of the subsections are color‐coded: magenta = SP2, purple = steroid channel (SC).

Additionally, Ser118 is replaced by phenylalanine in the other AKR1C isoforms, eliminating the potential for hydrogen bonding to the amide group. These differences likely contributed to the high selectivity of 1 toward AKR1C3. The alkyl chain was partially solvent‐exposed as it extended through the SP2 pocket and the steroid channel (Figure 4D). The glucose moiety was located outside the binding site and formed a hydrogen bond to the amide carbonyl of Ser129 at the protein surface, which explains its minimal impact on AKR1C3 inhibition.

3. Conclusion

The unique glucolipopeptide structure of myxoglucamides prompted us to shed light on their biological role. To identify its molecular target in A549 human lung epithelial cells, a synthetic access to the compound class was a prerequisite. It was accomplished by the first total synthesis of myxoglucamide A in 11 steps in the longest linear sequence with an overall yield of 2.4%, employing a convergent strategy with a late‐stage coupling between the peptidic and the glycolipidic building blocks. The synthesis enabled activity‐based protein profiling with a myxoglucamide‐biotin conjugate that led to the identification of several oxidoreductases as well as structural proteins as binders. The aldo‐keto reductase 1C3 (AKR1C3) was validated as a target of 1 through functional, biophysical and structural studies.

The synthetic access also allowed studying the structure‐activity relationship of AKR1C3 inhibition with the help of 20 analogs. The glucose moiety was not essential for inhibition. It might therefore rather function as an enhancer of solubility in the lipophilic part of 1 or influence chemical stability, cellular uptake, transport, or interactions with biological targets in the native host or competing microorganisms [38]. The lipid part, whose truncation and unsaturation led to an eightfold enhanced potency, occupied the hydrophobic SP2 pocket and extended through the steroid channel.

Our initial assumption that the α‐ketoamide functions as a covalent warhead was incorrect, as AKR1C3 is lacking a reactive cysteine residue in its active site. However, the SAR study indicates that the ketoamide is essential for biological activity. Given the lack of bonding interactions, it might rather enhance the structural rigidity of the peptidyl fragment in a defined conformation [39]. Also, the carboxylic acid was found to be essential for inhibition by binding to parts of the catalytic tetrad (His117 and Tyr55); similar interactions have been observed for synthetic AKR1C3 binders as well. The peptidic part of 1 thus was essential to occupy the catalytic site of the enzyme, stabilizing the ternary complex formed between AKR1C3, 1, and NADP⁺. These structure‐activity relationships map the roles of the peptide, lipid, and sugar parts of the myxoglucamides with respect to AKR1C3 inhibition. The findings further illustrate that an α‐ketoamide moiety is not necessarily associated with covalent binding, as demonstrated by other natural products such as rapamycin, complestatin, or plitidepsin.

Because AKR1C isoforms show variable and partially undesired bioactivities, a high selectivity of inhibitors for AKR1C3 is regarded as important [29]; this criterion is fulfilled by the myxoglucamides. They have not reached the low nanomolar potency of the best, optimized AKR1C3 inhibitors yet, but provide a new and original starting point against this well‐validated target.

While AKR1C3 has been identified as an important mammalian target of myxoglucamides, the effect(s) of the compound in their natural habitat and associated targets in microbial communities remain to be deciphered in future studies. Given the lack of a strong growth inhibition phenotype, a role in microbial signaling or interspecies interactions is conceivable. On a general notion, the study demonstrates the merit of investigating natural products even in the absence of strong phenotypes such as cell death induction. The use of unbiased chemoproteomic methods was a viable strategy to alleviate the pronounced darkness of biological space targeted by natural products that appear ‘inactive’ at first sight.

Author Contributions

Thomas Siemon: writing – original draft, investigation, formal analysis, methodology, conceptualization. Vivek K. Mishra: investigation, writing – review and editing. Mingming Zhao: investigation, writing – review and editing. Konrad Büssow: investigation, writing – review and editing. Alexander Popoff: writing – review and editing, investigation. Chengzhang Fu: writing – review and editing, investigation. Rolf Müller: writing – review and editing, conceptualization. Lothar Jänsch: supervision, investigation, writing – review and editing, resources, validation. Wulf Blankenfeldt: supervision, formal analysis, resources, validation. Mark Brönstrup: supervision, writing – original draft, writing – review and editing, conceptualization, formal analysis, resources, validation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The authors have cited additional references within the Supporting Information [40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57].Supporting File: anie73999‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.