As a first step to develop ABPs for DXPS, we designed a probe based on the first-generation alkylAP scaffold (Figure 2). The probe design incorporates the acetylphosphonate reactive group that mimics the donor substrate pyruvate, and reacts with the ThDP cofactor on active DXPS to form the covalent PLThDP adduct. As PLThDP formation is reversible (Sanders et al., 2017), a crosslinking group is required to ensure that active DXPS can be labeled irreversibly. Sterically demanding substituents can be incorporated into the phosphonyl ester group without significant loss in inhibitor potency (Morris et al., 2013; Sanders et al., 2017; Bartee and Meyers, 2018a; Coco et al., 2024), due to the large active site volume of DXPS, whereas modifications to the reactive acetyl group are not tolerated (Smith, Vierling, and Meyers, 2012). Based on this, we designed alkylAP-based probe 1 bearing the commonly-used 2-(3-(but-3-yn-1-yl)-3H-diazirin-3-yl)ethyl moiety (Li et al., 2013), capable of crosslinking to the DXPS active site and presenting a biorthogonal handle for introduction of a fluorophore or biotin.

ABP 1 was synthesized from the commercially available 1-hydroxy-6-heptyn-3-one 2 (Figure 3). Ketone 2 was converted to diaziridine 3 by sequential treatment with anhydrous ammonia and hydroxylamine-O-sulfonic acid (HOSA). Oxidation of crude 3 using iodine afforded diazirine 4 in 60% yield over two steps. Phosphorylation of 4 via phosphoramidite coupling with dimethyl-N,N-diisopropylphosphoramidite in the presence of tetrazole gave phosphite 5 in 94% yield. Phosphite 5 was converted to acetylphosphonate diester 6 by reaction with acetyl chloride, and 6 was subsequently dealkylated with lithium bromide to give ABP 1 in 53% yield over two steps. As expected, ABP 1 exhibited an absorbance profile consistent with a diazirine ring (λ_max 340 nm) (Supplementary Figure S1).

The phosphonolactylThDP (PLThDP) adduct formed via reaction of an alkylAP with ThDP can be detected by circular dichroism (CD) on pyruvate decarboxylase enzymes, including DXPS (Jordan et al., 2003; Nemeria et al., 2009, 2010; Heflin, 2015; Zhou et al., 2017; Coco et al., 2024). ThDP bound to DXPS exists in the 4′-aminopyrimidine (AP) form (Figure 4A) with a characteristic negative CD signal at 320 nm (Figure 4B, blue line) (Patel et al., 2012). Formation of a stable PLThDP adduct is characterized by disappearance of the negative CD signal and formation of a broad positive CD signal corresponding to the 1′,4′-iminopyrimidine (IP) form of the new PLThDP adduct (Figure 4A) (Zhou et al., 2017). As expected, formation of a broad positive CD signal was observed upon addition of 1 (50 μM) to E. coli DXPS (EcDXPS, 30 μM) in the presence of ThDP (200 μM), supporting active site engagement of 1 with DXPS and formation of the PLThDP adduct (Figure 4B, red line).

Using the DXPS-IspC coupled assay (Coco et al., 2024) to measure initial velocity of DXP formation (Supplementary Figure S2), we assessed the ability of ABP 1 to inhibit DXPS-catalyzed DXP formation. Consistent with the observed formation of PLThDP, ABP 1 inhibits EcDXPS with a K _i of 1.60 ± 0.22 μM (Figure 4C), comparable to the observed low micromolar potencies of other first-generation alkylAPs (Smith et al., 2014; Sanders et al., 2017). To rule out inhibition of the coupling system, ABP 1 was assessed as an inhibitor of IspC and found to be inactive up to 100 μM (Supplementary Figure S2B). To confirm that the potency of 1 is not due to UV-induced crosslinking during the DXPS-IspC coupled assay, the K _i of 1 was determined by measuring initial velocities of DXP formation after exposing mixtures of 1 and DXPS to 340 nm light for 5 min (Supplementary Figure S2C). A comparable K _i of 2.60 ± 0.47 μM was determined, indicating 1 is stable under conditions of the coupled assay.

To evaluate the ability of 1 to label active DXPS, we carried out the workflow shown in Figure 2 in which EcDXPS and 1 were incubated on ice for 10 min, then irradiated at 365 nm (180 W) for 3 min at 4°C. Crosslinked DXPS was then denatured, treated with tetramethylrhodamine (TAMRA)-azide under Cu (I)-catalyzed azide-alkyne cycloaddition (CuAAC) conditions to install the fluorophore, and evaluated by SDS-PAGE. Indeed, labeling of DXPS (3 μM) was observed in the presence of 200 μM 1 and depended upon photochemical activation of the diazirine as well as the CuAAC reaction to incorporate TAMRA, as evidenced by a lack of labeling in the absence of irradiation or Cu (I) catalyst (Figure 5A). Labeling of DXPS with 1 shows dose-dependence (Figures 5B, C), with saturation under these conditions evident at 31.3 μM 1.

ABP 1 is designed to bind within the DXPS active site and undergo reaction with ThDP to form the PLThDP adduct; thus, efficient labeling of DXPS by 1 should be dependent upon the ability of DXPS to activate ThDP to the reactive ylide. To demonstrate this, we assessed the ability of 1 to label denatured wild-type DXPS as well as a catalytically impaired DXPS variant (EcE370A DXPS) (Brammer, 2013; Querol-Audí et al., 2014). While secondary structure and stability of EcE370A DXPS are similar to wild type (Supplementary Figures S6, S7), this variant lacks the conserved glutamate within hydrogen bonding distance of the cofactor N1’, required for activation of ThDP to the reactive ylide during catalysis (Muller et al., 1993; Wikner et al., 1994; Schellenberger, 1998; Schneider and Lindqvist, 1998; Berthold et al., 2005; Jordan and Nemeria, 2005; Xiang et al., 2007; Querol-Audí et al., 2014; White et al., 2016). As expected, labeling of EcE370A DXPS was significantly diminished at concentrations of 1 that fully label active wild-type DXPS (>15.6 μM, Figure 6). Likewise, diminished labeling of denatured EcDXPS was observed relative to active wild-type DXPS. Weak labeling could be detected at [1] > 7.81 μM, indicating low-level non-specific interactions between the probe and inactive DXPS. At [1] > 31.3 μM, more pronounced non-specific labeling is observed (Supplementary Figure S8). Together, these results indicate that reversible binding to the DXPS active site alone is insufficient for productive labeling by 1, and conversion to the PLThDP adduct via reaction of 1 with ThDP is required.

To gain additional evidence that ABP 1 engages the DXPS active site, we conducted competition experiments using previously studied pyruvate-competitive alkylAP-based inhibitors known to act via PLThDP formation on DXPS. Three inhibitors with varying potencies were selected (Figure 7A), including butylacetylphosphonate (BAP, 7), methylacetylphosphonate (MAP, 8), and dibenzylglycine triazole acetylphosphonate (DBGlyTrAP, 9) (Smith, Vierling, and Meyers, 2012; Sanders et al., 2017; Coco et al., 2024). BAP (7) and MAP (8) are first-generation alkylAPs that display low micromolar and submicromolar potencies, respectively, against DXPS enzymes. DBGlyTrAP (9) is a recently-discovered time-dependent bisubstrate analog inhibitor displaying low nanomolar potency against EcDXPS (Coco et al., 2024). EcDXPS (3 μM) was incubated with each inhibitor for 10 min prior to the addition of 1 (Figure 7B). Following a 10 min incubation with 1, mixtures were irradiated for 3 min at 4°C, subjected to CuAAC reaction conditions to install the TAMRA fluorophore, and analyzed by SDS-PAGE, as described above. As expected, the concentration of inhibitor required to block DXPS labeling by 1 decreases with increasing inhibitor potency (Figures 7C, D); BAP (7) is unable to compete effectively with 1 up to 30 μM, whereas MAP (8) and DBGlyTrAP (9) block labeling by 1 in a dose-dependent manner consistent with their relative potencies. These results offer further strong evidence that ABP 1 is acting at the DXPS active site, and demonstrate the utility of 1 for identifying and characterizing inhibitor potency in DXPS inhibitor development.

As noted, alkylAPs bearing sterically demanding phosphonate ester substituents have the potential to selectively target the large active site of DXPS (Smith, Vierling, and Meyers, 2012; Morris et al., 2013; Sanders et al., 2017). To gain preliminary insights into the selectivity of 1 for DXPS, labeling experiments were performed on porcine pyruvate dehydrogenase (PDH) and Saccharomyces cerevisiae pyruvate decarboxylase (PDC), ThDP-dependent enzymes that also catalyze pyruvate decarboxylation. Weak labeling of PDH by 1 relative to DXPS was observed (Figures 8A, B). In agreement with this, 1 displayed weak inhibitory activity against PDH (Figure 8C; Supplementary Table S1). Interestingly, PDC appeared to have intrinsic fluorescence in the absence of 1 (Figures 8A, B). In contrast to PDH, increasing the concentration of 1 in labeling experiments did not lead to an increase in fluorescent labeling of PDC, nor was there evidence of PDC inhibition by 1 up to 1 mM (Figure 8C; Supplementary Table S1). Taken together, these results indicate 1 displays selectivity for DXPS over other ThDP-dependent pyruvate decarboxylase enzymes.

In addition, several other proteins unrelated to ThDP-dependent enzymes were subjected to labeling conditions for preliminary assessments of non-specific labeling by 1. These included reductoisomerase IspC, the coupling enzyme used to measure DXP forming activity of DXPS (Supplementary Figure S2), glyceraldehyde 3-phosphate dehydrogenase (GADPH), alcohol dehydrogenase (ADH) and bovine serum albumin (BSA). In all cases, negligible labeling was observed up to 31.3 μM 1 (Supplementary Figure S11), conditions under which purified EcDXPS is fully labeled (Figure 5). The absence of IspC labeling by 1 is also consistent with the lack of inhibitory activity of 1 against IspC (Supplementary Figure S2). These results suggest minimal non-specific interactions of 1 under these conditions.

As a first step to evaluate 1 as a probe of DXPS activity in a complex proteome, we assessed labeling of DXPS by 1 in lysate from DXPS-overexpressing E. coli. Bacterial lysates were prepared from E. coli BL21 (DE3) cells harboring the dxs-pET37b expression construct for inducible expression of DXPS, the strain used for production and purification of recombinant DXPS (Brammer and Meyers, 2009). In lysate prepared from isopropyl β-D-1-thiogalactopyranoside (IPTG)-induced cultures, DXPS overexpression was observed, and labeling of DXPS by 1 is evident (Figure 9A, lane 3), compared to a lack of overexpression and DXPS labeling in lysate prepared from uninduced cultures (Figure 9A, lane 1). Further, incubation with 9 led to a reduction in labeling by 1 (Figure 9A, lane 4; Figure 9B). Taken together, these results suggest a potential utility of 1 to probe DXPS activity in complex proteomes and as a tool for DXPS inhibitor discovery and development.

Chemicals were purchased from Millipore Sigma (Sigma-Aldrich) and used as received, unless otherwise stated. TAMRA-azide (CCT-AZ109) was purchased from Vector Laboratories (Click Chemistry Tools). E. coli DXPS and IspC were overexpressed, purified using an AKTA-GO fast protein liquid-chromatography (FPLC) system, and characterized as previously reported (Coco et al., 2024). Bovine serum album (BSA), glyceraldehyde 3-phosphate (GAPDH, rabbit muscle), alcohol dehydrogenase (ADH, S. cerevisiae), pyruvate dehydrogenase (PDH, porcine heart) and pyruvate decarboxylase (PDC, S. cerevisiae) were purchased from Millipore Sigma (Sigma-Aldrich). The E. coli BL21 (DE3) cell line harboring the dxs-pET37b plasmid (Brammer and Meyers, 2009), for inducible DXPS overexpression, was used in experiments to investigate labeling of DXPS by 1 in complex lysate. DXPS inhibitors 7–9 were prepared as previously described (Smith, Vierling, and Meyers, 2012; Sanders et al., 2017; Coco et al., 2024). A BioTek Epoch 2 microplate reader was used at 25°C for aerobic spectrophotometric analyses. A Li-COR Odyssey CLx was used for imaging Coomassie-stained SDS-PAGE gels. Fluorescent gels were scanned using an Amersham Typhoon (Cytiva) biomolecular imager (excitation 554 nm; emission 566 nm). Circular dichroism (CD) experiments were performed on an Applied Photophysics Chirascan V100 CD spectrometer (Surrey, United Kingdom). A Spectroline Model FC-100 lamp with longwave ultraviolet light (365 nm, 180 W) was used for crosslinking studies. Protostain Blue (colloidal Coomassie Blue G-250) was used to stain proteins in SDS-PAGE analysis.

DXPS (30 μM) was diluted to a final volume of 1.5 mL in enzyme buffer (1 mM MgCl₂, 100 mM NaCl, 200 μM ThDP, 50 mM HEPES pH 8) on ice in a conical tube. Immediately prior to CD experiments, protein solutions were equilibrated at 25°C for 10 min. Sample was then transferred from the conical tube to a 1 cm quartz cuvette and a CD scan was recorded at 25°C from 280–400 nm with a 2 nm step and 0.5 s averaging time. ABP 1 was added to a final concentration of 50 μM (3 μL of 25 mM stock), the cuvette was inverted gently 3 times to mix, and a second CD scan was recorded on the mixture at 25°C using the same parameters as the initial scan. Experiments were performed in duplicate.

The DXPS-IspC coupled assay (Coco et al., 2024) was used to measure the rate of DXPS-catalyzed DXP formation in the presence or absence of 1. DXPS (100 nM) and IspC (2 μM) were pre-incubated with 1 (0–100 μM) in buffer (2 mM MgCl₂, 5 mM NaCl, 1 mM ThDP, 100 mM HEPES pH 8, 200 μM NADPH) at 25°C for 10 min. Enzyme reactions were initiated by the addition of substrates (50 μM pyruvate and 500 μM D-GAP). The change in absorbance of NADPH at 340 nm was monitored at 25°C and used to calculate the initial velocity of DXP formation. Initial velocities were plotted as a function of [1]. Data were fit to the Morrison equation (Morrison, 1969) (Eq. 1) to calculate K _i. Non-linear regression analysis was performed using GraphPad Prism version 10. Error bars represent standard deviation. Standard deviation was calculated from three replicate K _i determinations (K _i ± SD). v i v 0 = 1 − E T + I T + K i − E T + I T + K i 2 − 4 E T I T 2 E T (1)

ABP 1 was incubated with purified protein (3 μM) in buffer (2 mM MgCl₂, 5 mM NaCl, 1 mM ThDP, 100 mM HEPES pH 8) for 10 min on ice. The solutions were then irradiated (365 nm, 180 W) for 3 min in a cold room (4°C). The mixture (10 μL) was subjected to denaturing conditions by addition to 10% sodium dodecyl-sulfate (SDS, 10 μL) following by vortexing and heating (5 min at 95°C). Solutions were cooled to ambient temperature, and a 7.5× CuAAC reaction stock (4 μL, prepared immediately prior to addition) was added to achieve final concentrations of 0.1 mM tris ((1-benzyl-4-triazolyl)methyl)amine (TBTA), 1 mM CuSO₄, and 1 mM tris(2-carboxyethyl)phosphine (TCEP). TAMRA-azide (6 μL of 5 mM stock in DMSO) was then added to initiate the reaction. The reaction mixture was incubated for 1 h at ambient temperature, covered by foil. SDS-PAGE loading dye (10 μL of 4× solution containing 200 μM Tris-HCl, 400 μM dithiothreitol, 277 mM SDS, 6 mM bromophenol blue, and 4.3 M glycerol) was added, the mixture was vortexed, heated (5 min at 95°C), and a 15 μL aliquot was analyzed by SDS-PAGE (10% acrylamide). Gels were first scanned to detect in-gel fluorescence and then stained with Protostain blue (colloidal Coomassie Blue G-250) to visualize total protein. Gel images were generated and quantified using ImageJ. For experiments in which labeled enzyme was quantified, pixel densities of fluorescently labeled protein were determined using ImageJ. Briefly, the HiLo threshold command in ImageJ was employed, and contrast was adjusted such that no part of the image exceeded the maximum threshold. Vertical rectangular boxes were drawn to encompass each protein band, and pixel density across the band within the box was plotted. Pixel density was determined by integrating the plotted signal. Fluorescently labeled protein was then normalized to Coomassie-stained protein (Eq. 2). N o r m a l i z e d   f l u o r e s c e n c e = p i x e l   d e n s i t y T A M R A   g e l   b a n d   p i x e l   d e n s i t y C o o m a s s i e   g e l   b a n d (2)

PDH (0.01 U/mL) was preincubated with 1 (0–1,000 μM) for 10 min at 25°C in buffer containing 100 mM HEPES pH 8, 2 mM MgCl₂, 5 mM L-cysteine, 1 mM ThDP, 0.3 mM TCEP, 2.5 mM NAD⁺, and 100 μM Coenzyme A. The enzyme reaction was initiated by addition of pyruvate (60 μM) at 25°C. Initial velocity was determined by measuring NADH formation at 340 nm, and normalized to initial velocity in the absence of 1 to calculate % activity which was plotted as a function of [1] using GraphPad Prism. Error bars represent standard deviation of the fit from three replicates.

Inhibitory activity of 1 against PDC was determined as described above with the following modifications. PDC (0.05 U/mL) activity was assayed in buffer containing 100 mM 2-(N-morpholino)ethanesulfonic acid (MES) pH 6, 5 mM MgCl₂, 5 mM ThDP, 0.17 mM NADH, and 16 U/mL alcohol dehydrogenase as the coupling enzyme to detect formation of acetaldehyde product. The enzyme reaction was initiated by the addition of pyruvate (1 mM). Initial velocity was determined by measuring NADH depletion at 340 nm.