Small molecule perimeter defense in entomopathogenic bacteria

Abstract

Two Gram-negative insect pathogens, Xenorhabdus nematophila and Photorhabdus luminescens, produce rhabduscin, an amidoglycosyl- and vinyl-isonitrile-functionalized tyrosine derivative. Heterologous expression of the rhabduscin pathway in Escherichia coli, precursor-directed biosynthesis of rhabduscin analogs, biochemical assays, and visualization using both stimulated Raman scattering and confocal fluorescence microscopy established rhabduscin’s role as a potent nanomolar-level inhibitor of phenoloxidase, a key component of the insect’s innate immune system, as well as rhabduscin’s localization at the bacterial cell surface. Stimulated Raman scattering microscopy visualized rhabduscin at the periphery of wild-type X. nematophila cells and E. coli cells heterologously expressing the rhabduscin pathway. Precursor-directed biosynthesis created rhabduscin mimics in X. nematophila pathway mutants that could be accessed at the bacterial cell surface by an extracellular bioorthogonal probe, as judged by confocal fluorescence microscopy. Biochemical assays using both wild-type and mutant X. nematophila cells showed that rhabduscin was necessary and sufficient for potent inhibition (low nM) of phenoloxidases, the enzymes responsible for producing melanin (the hard black polymer insects generate to seal off microbial pathogens). These observations suggest a model in which rhabduscin’s physical association at the bacterial cell surface provides a highly effective inhibitor concentration directly at the site of phenoloxidase contact. This class of molecules is not limited to insect pathogens, as the human pathogen Vibrio cholerae also encodes rhabduscin’s aglycone, and bacterial cell-coated immunosuppressants could be a general strategy to combat host defenses.

Xenorhabdus and Photorhabdus species are enteric mutualistic gammaproteobacteria of their respective Steinernema and Heterorhabditis nematode hosts (1–3). The free-living infective juvenile-stage nematode hunts insect larvae in the soil, invades the insect’s circulatory system, and regurgitates its mutualistic bacteria to initiate insect pathogenesis (entomopathogenesis). The bacteria produce an assortment of small molecules and proteins to regulate the multipartite symbiosis by overcoming the insect’s innate immune system, killing the insect host and microbial competitors, and supporting proper nematode development. A bacteria–nematode pair can efficiently parasitize a range of insect larvae, making the duo an effective biocontrol agent in agricultural applications. The bacteria’s coordination of two vastly different objectives in two different animal hosts provides an excellent model system for investigating mutualistic and pathogenic bacteria–animal interactions (1–3).

We previously found that the bacterial uptake of l-proline, an abundant amino acid in insect hemolymph, enhanced the bacterial proton motive force and antibiotic production in the primary-phase phenotypic variants of X. nematophila ATCC19061 and P. luminescens TT01 (4). Rhabduscin (structure 1), a tyrosine-derived amidoglycosyl- and vinyl-isonitrile product (Fig. 1), was up-regulated in X. nematophila by about an order of magnitude in l-proline–supplemented cultures. A candidate rhabduscin gene cluster was identified in the X. nematophila ATCC19061 genome (5) by homology to the known isonitrile-forming biosynthetic genes isnA and isnB (6, 7) along with a glycosyltransferase, and this analysis was validated by heterologous production of rhabduscin in genetically transformed Escherichia coli (4). The P. luminescens TT01 genome (8) also contained the rhabduscin biosynthetic genes, and rhabduscin production was detected in this genus as well. A structurally related molecule, byelyankacin (structure 2), which was first identified in an Enterobacter species (9) and then later in metagenomic libraries (10), exhibits potent human melanogenesis–inhibitory activity by inhibiting tyrosinase—a phenoloxidase-type enzyme. Taken together, these observations led to the proposal that rhabduscin’s natural function is inhibiting the melanin-producing insect phenoloxidase (PO), a critical enzymatic innate immune response to invading pathogens (4).

Fig. 1.

Rhabduscin (1) and byelyankacin (2) biosynthesis.

Innate immunity is a largely conserved feature in animals, and the insect’s immune system parallels vertebrate innate immunity in many regards (11). The insect has pattern-recognition receptors (PRRs) that, upon microbial invasion, recognize pathogen-associated molecular patterns (PAMPs) and mount both humoral and cellular immune responses with high spatial and temporal control (12). In the pro-PO system, PAMP recognition stimulates a serine protease cascade leading to pro-PO cleavage and catalytically competent PO (13, 14). PO is the terminal component in invertebrate innate immunity that accepts phenolic substrates initially to generate short-lived cytotoxic quinone products, which are in turn converted to the long-lived and rigid polymer called melanin. The tightly controlled activation of the melanization complex, which binds to bacterial cells and polysaccharides, leads to localization of cytotoxic products around and melanin deposition directly on the intruding microorganism, encapsulating the pathogen. PO also enhances pathogen phagocytosis by hemocytes and contributes in wound healing by sclerotization, providing a formidable defense against microbial invaders (13, 14).

While P. luminescens was previously shown to produce multipotent bacterial stilbenes that exhibit micromolar PO inhibition against the model Manduca sexta insect larvae and contribute to virulence (15–17), rhabduscin’s natural role has not been experimentally assessed. To assign a specific function to rhabduscin in immunosuppression, we began a detailed investigation of rhabduscin’s cellular localization and biological activity. Here, we demonstrate that rhabduscin harbors a potent nanomolar-level immunosuppressive activity against PO, which is dramatically more potent than the previously described stilbenes. Strikingly, rhabduscin binds at the periphery of bacterial cells, as visualized by stimulated Raman scattering and confocal fluorescence microscopic techniques, where its location and potent PO-inhibitory activity contribute substantially to virulence.

Results and Discussion

Rhabduscin Biosynthesis and its Cellular Localization.

The organization of the biosynthetic genes for rhabduscin (1) in P. luminescens differs dramatically from that found in X. nematophila, and heterologous expression in E. coli was used to clarify the differences. In X. nematophila, we recently demonstrated that a small three-gene cluster (XNC1_1221-XNC1_1223), which includes isnA, isnB, and a glycosyltransferase gene, conferred rhabduscin production in the heterologous host E. coli (4). Genome synteny analysis, using the MicroScope bioinformatics platform (18), showed a similar set of genes in P. luminescens, but a large gap, more than 1,000 genes, separated isnA and isnB (plu2816 and plu2817) from two glycosyltransferase homologs (plu1760 and plu1762) (Fig. 2A). It seemed likely that both glycosyltransferase genes contributed to rhabduscin biosynthesis in the P. luminescens biosynthetic pathway either by transferring different activated sugars onto the vinyl-isocyanide aglycone 3 or by expressing redundant enzymatic functions. We constructed a series of heterologous expression vectors, which included isnA, isnB, GT1760 or GT1762, isnA–isnB, and isnA–isnB-GT1760 or -GT1762, to monitor metabolite production in E. coli. As expected, both isnA and isnB were required for aglycone 3 production (Fig. 2B). E. coli cells producing isnA, isnB, and either GT1760 or GT1762 redundantly produced rhabduscin and a second minor glycoside. High-resolution mass spectrometry (HR–MS) indicated that the minor compound has a molecular formula of C₁₅H₁₇NO₅Na [calculated, 314.1004 (M + Na)⁺; observed, 314.1012]. One-dimensional (¹H) NMR chemical shift and coupling constant analyses in addition to two-dimensional (gCOSY, gHSQC, gHMBC, and NOESY) correlations indicated that the minor glycoside was byelyankacin (9) (2) (SI Appendix, Table S1). The isonitrile functional group is sensitive to hydrolysis, especially in acidic conditions, and the resulting hydrolyzed formamide derivatives (1_H2O and 3_H2O) could also be detected.

Fig. 2.

(A) Rhabduscin biosynthetic gene cluster in Photorhabdus luminescens and heterologous expression of the different combinations of P. luminescens genes. (B) LC/MS ultraviolet traces and products from the heterologous expression of P. luminescens genes in E. coli. (C) Rhabduscin biosynthetic gene cluster in Xenorhabdus nematophila and genetic mutants. (D) LC/MS ultraviolet traces and products from X. nematophila wild-type, ΔGT, and ΔisnAB.

The isonitrile functional group, which occurs rarely in naturally produced molecules, has a characteristic Raman resonance (observed 2,121 cm^-1 for rhabduscin) that should be distinct for rhabduscin in E. coli. Robust production of rhabduscin in the heterologous host combined with its unique Raman shift enabled observation of its cellular localization by stimulated Raman scattering (SRS) microscopy (19, 20). Within the limits of spatial resolution for SRS, it was observed that rhabduscin’s 2,121 cm^-1 Raman signal was localized to the periphery of the overproducing E. coli cells, while the background signal in control cells lacking the gene cluster was low (Fig. 3 A and B). E. coli cells overproducing rhabduscin’s aglycone qualitatively showed less congruency in signal localization, suggesting that the sugar substituent contributes to rhabduscin placement (SI Appendix, Fig. S1).

Fig. 3.

SRS microscopy analyzing the spatial distribution of rhabduscin based on the vibrational resonance of the isonitrile functional group at 2,121 cm^-1. (Scale bar, 10 μm.) In this technique, two laser beams (pump and Stokes) with the frequency difference matching the vibrational frequency of the isonitrile group are utilized to drive the transition coherently. The spectrum is identical to spontaneous Raman spectrum but with much higher acquisition speed, allowing fast imaging (19). Because most naturally produced molecules do not have Raman signatures in this spectral region, the signal is distinctive for rhabduscin. The wild-type X. nematophila signal level of rhabduscin is lower than the E. coli overproducer, indicating lower expression level, but the difference in signal between wild-type and control cells is big enough to measure different cell phenotypes.

To extend these results into a wild-type host, we first constructed scarless-deletion mutants of the vinyl-isocyanide biosynthetic genes, isnA–isnB, and the glycosyltransferase gene in X. nematophila (Fig. 2C). X. nematophila was selected because we had previously found that l-proline up-regulates production of rhabduscin in this host by about an order of magnitude in rich laboratory medium (4), and also because we wanted to investigate rhabduscin’s contribution to the cellular inhibition of phenoloxidase. P. luminescens TT01 was less suitable because it robustly produces additional phenoloxidase inhibitors belonging to the bacterial stilbene class (15), and l-proline supplementation does not stimulate rhabduscin production (4). Deletion of the X. nematophila glycosyltransferase alone led to accumulation of the aglycone intermediate 3, and deletion of isnA–isnB completely destroyed isonitrile biosynthesis (Fig. 2D). Similar to the E. coli results described above, rhabduscin’s distinct Raman signal localized to the periphery of wild-type X. nematophila cells, whereas the control ΔisnAB strain lacking rhabduscin only exhibited weak background signal (Fig. 3 C and D).

Synthesis of Substrate Mimics and Precursor-Directed Biosynthesis.

With the heterologous glycosyltransferase expression constructs in hand, we began to investigate glycosyltransferase substrate promiscuity for the amino acid-derived part of the molecule. We challenged both P. luminescens glycosyltransferases individually expressed in E. coli with a panel of synthetic and commercial aglycone substrate mimics. We first synthesized the sterically conservative nitrile mimic 6, which simply inverts the N─C triple bond (Fig. 4A). We then synthesized the azide derivative 12 (Fig. 4B), which could serve as a glycosyltransferase substrate mimic and allow orthogonal probes to be introduced for further cellular localization studies. The less conservative commercial mimics included para-coumaric acid, 4-hydroxy-benzylideneacetone, (4-hydroxybenzylidene)malononitrile, trans-4-hydroxy-β-nitrostyrene, and umbelliferone. Using a precursor-directed biosynthetic strategy, the synthetic nitrile 6 and the deprotected azide 12 mimics were fed to BL21(DE3) Star pLysS E. coli cells overexpressing GT1760 or GT1762. The products were extracted with ethyl acetate and analyzed by C₁₈ reversed-phase LC/MS. The nitrile mimic was nicely converted to its corresponding rhabduscin (7, 46–71% conversion) or byelyankacin (8, 13–42%) glycoside derivatives (Fig. 4 and SI Appendix, Table S2), and the azide mimic was converted to its rhabduscin glycoside derivative (13, 61–93%). Of the commercial substrates, only benzylideneacetone and umbelliferone incorporated at minor levels (≤ 3%), indicating limited substrate tolerance in glycoside formation.

Fig. 4.

Precursor-directed biosynthesis of glycoside derivatives. (A) (E)-3-(4-hydroxyphenyl)acrylonitrile (6) was obtained via a Heck reaction between 4-iodophenol (4) and acrylonitrile (5) (29), which resulted in a 15/85 cis/trans mixture. The P. luminescens glycosyltransferases (GTs) both glycosylated the aglycone mimic to produce glycosides 7 and 8. (B) (E)-4-(2-azidovinyl)phenol (12) was obtained in three steps starting from 4-ethynylphenyl acetate (9); 10 was prepared by hydroboration of 9 (30); 11 was obtained via a copper-catalyzed cross-coupling between sodium azide and boronic acid 10 by adapting a previously described procedure (31); 11 was deprotected with potassium carbonate to afford phenolic compound 12, which was accepted by the GTs to afford glycoside 13. SI Appendix provides further details and spectral data.

Confocal Fluorescence Microscopic Analyses of Metabolite Localization.

Because azide-aglycone mimic 12 was accepted by the glycosyltransferase in vivo to generate the azide-glycoside 13 (Fig. 4B), their azide chemical handles could be used to introduce bioorthogonal probes in vivo via click chemistry (21). To evaluate whether the small molecule inhibitors are accessible at the bacterial cell surface, rather than embedded in the cell membrane, we used exogenous addition of green fluorescent protein (GFP) as an easily detectable surrogate of insect phenoloxidase. In addition, to avoid competition with the natural isonitrile small molecules, we selected the X. nematophila strain disabled in isonitrile biosynthesis (the ΔisnAB strain). The glycosyltransferase was left intact in this mutant so that the synthetic mimics could be used for precursor-directed glycoside biosynthesis, bypassing vinyl-isonitrile biosynthesis. Stationary-phase X. nematophila mutant cultures supplemented with the synthetic azide 12 or the biosynthetic azide-glycoside 13 were first reacted with a biotin–alkyne conjugate, washed, and then followed by the binding of a streptavidin–GFP conjugate. Similar to the SRS results, the fluorescence localized to the periphery of the cells within the limits of resolution only in the presence of the inhibitor mimics (Fig. 5), supporting that the small molecule inhibitors would be accessible to inhibit the extracellular phenoloxidase response of an insect’s innate immune system. To determine if this peripheral localization resulted from covalent linkage to the cell’s peripheral lipopolysaccharide (LPS), the LPS was purified (Gentaur LPS extraction kit) and separated over a polyacrylamide gel. The fluorescent signal partitioned to an initial phenol/chloroform wash fraction, and no fluorescence was detected in the denaturing LPS gel, suggesting that localization is physical, not covalent. Further supporting a physical interaction, the isonitrile-containing molecules are present in both cell preparations and in the cleared spent medium as a diffusion product.

Fig. 5.

Confocal fluorescence microscopy. (A) X. nematophila ΔisnAB fed with the two azide mimics 12 and 13 compared to PBS control. Brightness and contrast were adjusted identically for compared image sets. The azide mimics were first reacted with a biotin-alkyne conjugate followed by the binding of a streptavidin-GFP conjugate. (B) Enlargement of X. nematophila ΔisnAB fed with 13. The enlarged region is indicated in A by the white rectangle.

Metalloenzyme Inhibition: Tyrosinase and Phenoloxidase.

The peripheral localization and hemolymph- or l-proline-induced up-regulation of rhabduscin suggested that it had a functional role in bacterial pathogenesis. In an earlier screen for skin-whitening agents, the α- l-rhamnopyranoside byelyankacin (2) was reported to be a potent tyrosinase inhibitor, the key enzyme in human melanin formation (9). Establishing whether rhabduscin plays a similar role first required establishing the absolute stereochemistry of the attached sugar, because only the relative configuration for rhabduscin was reported previously (4). Mosher ester analysis (22) indicates that rhabduscin (1) contains a β-l-acetamidopyranoside (SI Appendix, Figs. S2 and S3 and Table S3).

Given the reported activity and insect phenoloxidase’s critical role in innate immune defense, we tested the compounds for inhibition against a model tyrosinase from mushroom and the phenoloxidase from waxmoth larvae (Galleria mellonella), a model insect host. The aglycone intermediate 3, byelyankacin, and rhabduscin all showed low nanomolar-level inhibition (IC₅₀) against both mushroom tyrosinase and insect phenoloxidase, a level roughly three orders of magnitude more potent than the commercial whitening agent kojic acid (Table 1 and SI Appendix, Fig. S4). The isonitrile-containing natural products are also roughly three orders of magnitude more potent than the bacterial stilbenes from Photorhabdus species, which have been linked to phenoloxidase inhibition in the model tobacco hornworm larvae, Manduca sexta (15). Strikingly, conservative replacement of the isonitrile with the nitrile in the synthetic aglycone mimic 6 largely destroys tyrosinase inhibition, demonstrating the isonitrile’s crucial role for inhibitory activity. The azide mimic 12 also exhibited weak activity. Phenoloxidases are copper-containing enzymes (23), and it is possible that the isonitrile’s ability to bind copper, and other metal ions, plays a role in enzyme inhibition.

Table 1.

Inhibition (IC₅₀) of tyrosinase from mushroom and hemolymph-activated phenoloxidase from Galleria mellonella (waxmoth larvae)

Compound	Tyrosinase inhibition [nM]	Phenoloxidase inhibition [nM]
Rhabduscin (1)	15.1 ± 1.0	64.1 ± 2.0
Byelyankacin (2)	37.1 ± 2.5	184.9 ± 14.5
Aglycone-NC 3	7.9 ± 0.8	61.7 ± 6.2
Aglycone-CN 6	> 400,000*	ND
Aglycone-N3 12	> 400,000*	ND
Kojic acid	22,173 ± 2,924	ND

^*Low inhibitory activity precluded an accurate IC₅₀ measurement.

When a bacterium invades an insect host, the insect deposits melanin onto the periphery of the invading cell, which exhibits a bactericidal activity and traps the bacteria in nodules (24). To determine whether the cells themselves can inhibit tyrosinase, we subjected washed X. nematophila cells to tyrosinase-inhibitory assays. X. nematophila wild-type (initial OD₆₀₀ = 0.71, 64% inhibition) and the glycosyltransferase deletion mutant (initial OD₆₀₀ = 0.86, 65% inhibition) both inhibited tyrosinase activity at similar levels, whereas the isonitrile isnAB biosynthetic gene deletion (initial OD₆₀₀ = 0.77, 0% inhibition) destroyed the cell’s inhibitory activity. The inhibitory activity remained persistently associated with the cells under physiologically relevant conditions, as iterative phosphate-buffered saline (PBS) washes did not flush the activity away (SI Appendix, Fig. S5). Taken together, these results identify rhabduscin as a potent immunosuppressant that localizes to the periphery of the bacterial cells producing it and enables them to inhibit phenoloxidase, thereby suppressing one of the insect’s primary innate immune defense strategies.

Rhabduscin is a Virulence Factor.

Having obtained evidence for rhabduscin’s nanomolar-level activity against phenoloxidase and its localization at or near the periphery of the bacterial cell, we wanted to determine the compound’s importance in pathogenesis in an animal model. X. nematophila wild-type, ΔGT, ΔisnAB, and a PBS control were individually injected into the hemolymph of ten Galleria mellonella larvae. Wild-type bacteria killed all of the larvae within 50 h, and PBS killed none of the larvae throughout the 96 h experiment (SI Appendix, Fig. S6). The ΔGT mutant was less virulent than wild-type, but this bacterium grew more slowly than wild-type, presumably because of the toxicity of accumulated aglycone. The ΔisnAB bacteria only killed half of the larvae during the experiment, and the dead larvae were generally darker compared to the wild-type–infected dead larvae because of heightened melanization. These semiquantitative results demonstrate that rhabduscin contributes to an effective pathogenesis.

Isonitrile Biosynthesis in Human and Other Pathogens.

The rhabduscin biosynthetic pathway is not limited to the insect pathogens X. nematophila and P. luminescens. For example, sequence analysis indicates that the related insect pathogen Xenorhabdus bovienii contains a homologous gene cluster (XBJ1_1355-XBJ1_1353). Like P. luminescens, two gene sets separated by over 1,000 genes reside in Photorhabdus asymbiotica, an emerging human pathogen also capable of infecting insects. P. asymbiotica appears to contain three adjacent copies of the glycosyltransferase (PAU_02755-PAU_02757) in addition to the isnA and isnB genes (PAU_01721-PAU_01720). The conservation across multiple Xenorhabdus and Photorhabdus species and the redundant glycosyltransferase copies in P. luminescens and P. asymbiotica emphasize the pathway’s important role.

A single isnA–isnB fusion protein (VC1949) was reported in the related gammaproteobacterium Vibrio cholerae (10), the causative agent of cholera, but to date, no small molecule has been reported for this Vibrio pathway. We heterologously expressed a codon-optimized version of this isnA–isnB fusion protein in E. coli, but a biosynthetic small molecule product was not detected. Closer inspection of the genes surrounding VC1949 in the V. cholerae O1 biovar El Tor N16961 genome revealed an additional isnB homolog (VC1944) (SI Appendix, Fig. S7A). Coexpression of VC1949 with VC1944 or VC1944-VC1945 in E. coli led to aglycone 3 production (SI Appendix, Fig. S7B). Both the vinyl-isonitrile 3 and its hydrolysis product 3_H2O were observed. While it is not clear from these experiments whether 3 is further modified, based on the studies above, it is reasonable to speculate that this related pathway in V. cholerae contributes to pathogenesis by inhibiting metalloenzymes in the human innate immune system.

Conclusions

Gram-negative bacterial pathogens often modify their outermost LPS layer to modulate the effects of their target host’s immune system (25, 26). Additionally, the LPS layer’s hydrophobicity minimizes hydrophobic small molecules, including many antibiotics, from reaching the inside of the cell. Rhabduscin, byelyankacin, and their common aglycone represent another protective strategy: one in which they serve as substrate mimic inhibitors for phenoloxidase, a copper-dependent metalloenzyme that plays a central role in invertebrate immunity by depositing melanin onto the cell surface of invading microorganisms. Decorating the bacteria’s outermost layer with these inhibitors provides a spatially appropriate and high local concentration to defend effectively against this terminal phenoloxidase response and thereby support the bacteria’s pathogenic lifestyle.

Materials and Methods

Construction and Metabolite Analysis of Deletion Mutants.

Scarless-deletion mutants of genes isnA–isnB and glycosyltransferase Xn_1223 in X. nematophila, and subsequent product analysis of organic extracts by LC/MS, were prepared and analyzed using similar conditions as previously described (4) (SI Appendix, Table S4). Aliquots of stationary-phase cultures were visualized directly on poly-l-lysine-coated slides by SRS microscopy without manipulation (SI Appendix).

Heterologous Expression of P. luminescens and V. cholerae Genes.

All genes except VC1949 were cloned into pETDuet-1 (Novagen) and expressed in BL21(DE3) Star pLysS E. coli using standard molecular biological procedures (SI Appendix, Table S4). E. coli cells harboring the P. luminescens rhabduscin pathway or harboring the pETDuet-1 control vector were used for SRS microscopic experiments (SI Appendix), and aliquots of stationary-phase cells were visualized directly. A codon-optimized VC1949 in pJExpress401 was obtained from DNA2.0.

Precursor-Directed Biosynthesis of Glycoside Derivatives.

BL21(DE3) Star pLysS E. coli containing the P. luminescens glycosyltransferase genes (either pEGT1760 or pEGT1762) were grown overnight at 37 °C and 250 rpm as LB suspension cultures with 0.1 mg/mL ampicillin. Fifty-μL aliquots of overnight stationary-phase culture were used to inoculate 5-mL portions of M9 minimal medium supplemented with 0.5% (m/vol) casamino acids and 0.1 mg/mL ampicillin. The cultures were grown at 37 °C and 250 rpm to an OD₆₀₀ = 0.5–0.6. Glycosyltransferase gene expression was induced with IPTG (1 mM final) and supplemented with small molecule precursors (100 μM) in DMSO. The induced cultures were incubated at 25 °C and 250 rpm to a final time of 24 h. The whole cultures were rigorously extracted with 5 mL of ethyl acetate, separated by centrifugation; the top 4 mL of ethyl acetate was transferred to a glass vial and dried, and the resulting residue was dissolved in 500 μL methanol for reversed-phase LC/MS analysis.

Chemoenzymatic Synthesis of Rhabduscin-Azide (13).

Two 4-L Erlenmeyer flasks, each containing 1 L of M9 minimal medium supplemented with casamino acids (0.5%) and ampicillin (0.1 g/L), were inoculated with 5 mL of an overnight BL21(DE3) Star pLysS E. coli culture harboring pEGT1762 (LB, 0.1 mg/mL of ampicillin). The cultures were grown at 37 °C and 250 rpm to an OD₆₀₀ = 0.5–0.6. Expression of the cluster was induced by the addition of IPTG (1 mM final), and, at the same time, synthetic (12) (32 mg, 0.20 mmol) dissolved in DMSO (5 mL) was added to each culture. The cultures were grown at 25 °C and 250 rpm for a total time of 24 h. The combined whole cultures were extracted with ethyl acetate (three times, 2 L total). The combined organic fractions were dried (Na₂SO₄), filtered, and concentrated under reduced pressure to afford an orange solid (220 mg). The residue was purified by preparative HPLC (Luna C₁₈, 5 μm, 250•21.2 mm, Phenomenex) with an acetonitrile∶water gradient at 10 mL/ min: 0–15 min, 30% to 45% acetonitrile; 15–16 min, 45% to 100% acetonitrile; and 16–21 min, 100% acetonitrile. The combined fractions (retention time = 13 min) were concentrated under reduced pressure to afford (13) as a pale yellow compound (9.0 mg, 0.026 mmol, 6.5%). ¹H-NMR: (600 MHz, CD₃OD) δ 7.28 (d, J = 8.8 Hz, 2H), 7.00 (d, J = 8.8 Hz, 2H), 6.78 (d, J = 13.8 Hz, 1H), 6.21 (d, J = 13.8 Hz, 1H), 4.85 (m, 1H), 4.29 (dd, J = 4.7, 1.4 Hz, 1H), 3.92 (qd, J = 1.4, 6.4 Hz, 1H), 3.78 (dd, J = 10.0, 4.7 Hz, 1H), 3.67 (dd, J = 10.0, 7.6 Hz, 1H), 2.05 (s, 3H), 1.17 (d, J = 6.4 Hz, 3H). ¹³C NMR: (151 MHz, CD₃OD) δ 174.8, 158.4, 131.0, 128.0, 126.9, 120.1, 117.9, 102.7, 73.6, 72.3, 71.2, 54.9, 22.5, 17.0. HR-MS: (ESI) calculated for C₁₆H₂₀N₄O₅Na 371.1331 (M + Na)⁺, found 371.1335.

Mushroom Tyrosinase-Inhibition Assay.

Mushroom tyrosinase was purchased from Sigma–Aldrich (T3824). The assay was adapted from the Pomerantz method (27), using 96-well plates (costar 3370) with the absorbance being monitored in a Molecular Devices FlexStation 3. Fifty μL of a 2-mM DOPA solution in 0.1-M phosphate buffer (pH 6.8) was added to 50 μL of the inhibitor (varying concentrations) dissolved in 10% DMSO in water. The resulting solution was incubated for 10 min at room temperature before the addition of 50 μL of a 135-U/mL mushroom tyrosinase solution in phosphate buffer. The 96-well plate was agitated for 3 s before starting to measure absorbance at 475 nm every 20 s. Kojic acid was used as a positive control and a 10% DMSO water solution was used as a blank. The initial rate was obtained by linear regression of the first 3 min of dopachrome formation. The fractional activity was obtained by dividing the inhibited rate by the uninhibited rate obtained from the blank reaction. The inhibition curves and IC₅₀ values were calculated using nonlinear regression sigmoid GraphPad with Prism 5 for Windows.

Mushroom Tyrosinase Inhibition Using X. nematophila Cells.

The cell-inhibition assay was conducted similarly to the small molecule-inhibition assay described above, but a suspension of cells was used as the inhibitor. X. nematophila wild-type, ΔisnAB, and ΔGT mutants were grown overnight in suspension (5 mL of LB) from a single colony at 30 °C and 250 rpm. Five-mL cultures (2-g tryptone, 5-g yeast extract, and 10-g NaCl per L supplemented with 0.1 M final of l-proline) were inoculated with 100 μL of the overnight culture and grown at 30 °C and 250 rpm for 48 h. The cells were centrifuged and washed once with PBS before being diluted to an OD₆₀₀ of approximately 0.75 for use in the assay described above. To assess inhibitor persistence on the cells, a subsequent experiment was conducted with iterative washing steps. Here, X. nematophila was grown and analyzed exactly the same, except cellular inhibition was measured after each successive wash cycle. A 15-mL bacterial suspension was washed 14 times with PBS by centrifugation (800  × g, 5 min) and cell resuspension. At each washing round, an aliquot of 500 μL of cell suspension was removed for analysis, and the volume of PBS added for the next washing round was reduced by 500 μL to minimize cell dilution.

Hemolymph Phenoloxidase–Inhibition Assay.

The assay was a modification of the procedure described by Eleftherianos et al. (28). Waxmoth larvae were chilled on ice for 15 min before sterilizing their surface with 70% ethanol. The larvae were then decapitated and bled into a Falcon tube on ice. The collected hemolymph was diluted in a 3∶1 (vol/vol) ratio with PBS (50 mM, pH = 6.5) and centrifuged at 14,700  × g and 4 °C for 10 min to obtain the plasma. Activation of the phenoloxidase was performed directly in the 96-well assay plate. Ten μL of diluted hemolymph plasma were added to 2 μL of E. coli LPS (5 mg/mL; Sigma-Aldrich) for activation and 115 μL of PBS (50 mM, pH = 6.5), and the solution was kept at room temperature for 1 h. Then, 50 μL of the inhibitor dissolved in 10% DMSO in PBS followed by 25 μL of 20-mM 4-methyl catechol were added. The 96-well plate was agitated for 3 s before starting to measure absorbance at 490 nm every min for 1 h. IC₅₀ values were obtained as described for the tyrosinase-inhibitory assays.

Preparation of Cells for Confocal Fluorescence Microscopy.

X. nematophila ΔisnAB was grown overnight in suspension (5 mL of LB) from a single colony at 30 °C and 250 rpm. Five-mL cultures (2-g tryptone, 5-g yeast extract, and 10-g NaCl per L, and 0.1 M of l-proline) were inoculated with 150 μL of the overnight culture and were grown at 30 °C and 250 rpm for 6 h before adding 25 μL of DMSO stock solution of 12, 13, or blank (100 μM final). The cultures were grown for 72 h at 30 °C and 250 rpm, where 25 μL of DMSO or azide 12 or 13 stock solutions were added after 29 and 56 h. Aliquots of 500 μL were centrifuged at 4,500  × g for 5 min and the cells were then washed with 500 μL PBS and centrifuged. The cells were then reacted with PEG4 carboxamide-propargyl biotin (3 μM final, Invitrogen B10185) using Click-iT Cell Buffer Kit (Invitrogen C10269) following the supplier’s procedure. The only modification to the supplier’s procedure was that the cells were resuspended in half the volume of buffer before adding the second half of the buffer mixed with the active compounds. After 30 min of reaction time, the cells were centrifuged at 1,500  × g for 4 min and washed three times with 500 μL of PBS before being resuspended in 500 μL PBS. A solution of 100 μL PBS and 10 μL streptavidin-Renilla reniformis GFP (Avidity, 320 mg/mL) was added to a 100-μL aliquot, and the mixture was reacted for 15 min at room temperature. The cells were then centrifuged at 1,500  × g for 4 min and washed two times with PBS before being resuspended in 100 μL PBS for microscopy.

Waxmoth Larvae Survival After X. nematophila Infection.

Waxmoth larvae (Galleria mellonella) were obtained from Berkshire Biological. Virulence assays were adapted from Eleftherianos et al. (28). X. nematophila wild-type, ΔisnAB, and ΔGT were grown overnight in liquid culture (5 mL, LB) at 30 °C and 250 rpm. One hundred μL of the overnight cultures were centrifuged (6 min at 3,300  × g), the spent medium was discarded, and the cells were resuspended in 1 mL of PBS. The quantity of cells was estimated with OD₆₀₀ (estimation: 10⁹ cells/mL for OD₆₀₀ = 1). The suspension was diluted to obtain an approximate concentration of 100, 1,000, and 10,000 cells per 10 μL. Prior to injection, the rear ends of the larvae were washed with 70% ethanol. The larvae were then injected into the hemocoel with 10 μL of cell suspension using 31-gauge disposable insulin syringes (BD). PBS-injected insects served as a control set. Larvae were kept at room temperature with food, and mortality (no reaction when poked) was monitored for 100 h. Ten larvae were used for each condition.