|Home | About | Journals | Submit | Contact Us | Français|
Phenolic glycolipids (PGLs) are polyketide-derived virulence factors produced by Mycobacterium tuberculosis, Mycobacterium leprae, and other mycobacterial pathogens. We have combined bioinformatic, genetic, biochemical, and chemical biology approaches to illuminate the mechanism of chain initiation required for assembly of the p-hydroxyphenyl-polyketide moiety of PGLs. Our studies have led to the identification of a stand-alone, didomain initiation module, FadD22, comprised of a p-hydroxybenzoic acid adenylation domain and an aroyl carrier protein domain. FadD22 forms the first acyl-S-enzyme covalent intermediate in the p-hydroxyphenyl-polyketide chain assembly line. We also used this information to develop the first small-molecule inhibitor of PGL biosynthesis. Overall, these studies provide new insights into the biosynthesis of an important group of small-molecule mycobacterial virulence factors and support the feasibility of targeting PGL biosynthesis to develop new drugs to treat mycobacterial infections.
Mycobacterium tuberculosis and Mycobacterium leprae, the etiologic agents of tuberculosis and leprosy, respectively, are pathogens with serious impacts on global public health [1, 2]. Tuberculosis is one of the top-ten leading causes of death in the world and is responsible for nearly two million deaths per year. Moreover, the growing incidence of multidrug-resistant (MDR) tuberculosis and the emergence of extensively/extremely drug-resistant (XDR) strains pose a new threat [3, 4]. Although leprosy has been controlled effectively using multidrug therapy, it remains one of the major causes of non-traumatic neuropathy and there are over three million people with leprosy-derived disabilities worldwide [5, 6]. Further, the emergence of MDR strains threatens to compromise leprosy control [7, 8]. Thus, there is a significant need for new antimycobacterial drugs with novel mechanisms of action.
M. tuberculosis and M. leprae produce diesters of long-chain methyl-branched fatty acids (e.g., mycocerosic acids) and long-chain, glycol-containing aliphatic polyketides (e.g., phenolphthiocerols and phthiocerols) that are important small-molecule effectors of virulence (see ref.  for a review). Members of this family of lipid diesters were first established as a bona fide virulence factors by Cox and coworkers  and Camacho and coworkers . These compounds are non-covalently linked to the outer cell wall layer and usually referred to as dimycocerosate esters (DIMs) (Figure 1). DIM variants are produced also by M. bovis, M. kansasii, M. marinum, M. ulcerans, M. microti, M. africanum, M. haemophilum, and M. gastri , all of which are pathogenic to humans [12, 13]. Production of phenolphthiocerol-based DIM variants (PGLs) (Figure 1) has been linked to a hyperlethality phenotype in a mouse model of tuberculosis  and associated with more severe clinical manifestations in a rabbit model of M. tuberculosis meningitis . Conversely, a PGL-deficient M. bovis mutant is attenuated in a guinea pig model of infection . PGL overproduction by M. tuberculosis or addition of purified M. tuberculosis PGLs to macrophages reduces pro-inflammatory cytokine release, whereas loss of PGLs correlates with increased cytokine release by M. tuberculosis-infected macrophages . PGLs of M. bovis also reduce pro-inflammatory cytokine release . M. leprae PGLs induce nerve demyelination, a primary contributing factor to the nerve function impairment characteristic of leprosy , and are involved in bacterial attachment to and invasion of Schwann cells . M. leprae PGLs also interfere with antigen-presenting cell function and suppress the proliferative response of T cells to mitogens . PGLs may also contribute to virulence in other mycobacteria, as evidenced by the recent observation that a DIM-deficient mutant of M. marinum is attenuated in a goldfish model of infection .
Biosynthesis of the phenolphthiocerol moiety of the PGLs involves the Pks15/1-PpsABCDE type I polyketide synthase system [20-25], a trans-acting enoyl reductase , and two tailoring enzymes that convert phenolphthiodiolones to phenolphthiocerols [27-29]. Feeding experiments with radiolabeled p-hydroxybenzoic acid (pHBA) [14, 30] suggest that the p-hydroxyphenyl moiety of the PGLs is derived from pHBA and the p-hydroxybenzoyl-Coenzyme A (pHB-CoA) thioester has been suggested as the donor of pHBA . Despite this progress, the current model for PGL assembly does not contemplate a mechanistic hypothesis for phenolphthiocerol chain initiation. We report herein our investigations of this key step in PGL assembly. Our studies have revealed that a stand-alone, didomain initiation module activates and loads pHBA onto the phenolphthiocerol biosynthesis machinery in a CoA-independent manner. We also report the synthesis and characterization of a small-molecule inhibitor of this module that blocks PGL production.
We hypothesized that PGL assembly requires formation of a p-hydroxybenzoyl-AMP (pHB-AMP) species that would either undergo transesterification to a pHB-CoA thioester intermediate under the canonical polyketide biosynthesis pathway [31, 32] as suggested previously  or, alternatively, acylate an aroyl carrier protein (ArCP) domain thiol directly. The latter case would be in analogy to the biosynthetic pathways of salicylate-derived mycobacterial siderophores  and the 3-amino-5-hydroxybenzoate-derived antibiotic rifamycin B . We performed protein similarity searches (http://www.ncbi.nlm.nih.gov/blast) using HbaA, a pHB-CoA ligase from Rhodopseudomonas palustris , as the query to probe the genomes of M. tuberculosis, M. leprae, and M. bovis for possible pHB-AMP and/or pHB-CoA ligases. FadD22 produced the most significant alignment and is conserved in the three species and annotated as a putative fatty acyl-CoA ligase. Encouragingly, fadD22 is located between pks15/1, proposed to encode a type I polyketide synthase that elongates pHBA to form p-hydroxyphenylalkanoate precursors for PGL biosynthesis , and a chorismate lyase gene, encoding the enzyme that catalyzes the formation of pHBA from chorismate . FadD22 orthologues are also found in M. marinum  and M. ulcerans Agy99 . The FadD22 orthologues have 701–708 aa and high sequence identity (73–100%) (Figure S1).
Each FadD22 orthologue has an adenylation (AMP-binding, Pfam00501) domain and a 100-150 aa C-terminal extension relative to known acyl-CoA ligases that contains a putative 4′-phosphopantetheine attachment site (Pfam00550; Figure S1), a hallmark of acyl, aroyl, and peptidyl carrier protein domains. This suggested that the adenylation domain forms pHB-AMP and that this intermediate is converted directly to a pHB-S-FadD22 thioester by attack of the phosphopantetheine thiol of the ArCP domain (Figure 2A). Thus, we hypothesized that phenolphthiocerol biosynthesis initiation involves FadD22, a stand-alone didomain initiation module that catalyzes formation of a pHB-S-ArCP domain thioester, without the intermediacy of pHB-CoA.
Based on our bioinformatic analysis, we predicted that FadD22 is essential for PGL biosynthesis. Conversely, this protein would not be required for the biosynthesis of non-pHBA-derived DIM variants based on phthiocerol and phthiodiolone (referred to herein as PDIMs and PNDIMS, respectively) (Figure 1). To investigate these hypotheses, we constructed a M. bovis strain with a fadD22 deletion. To examine the effect of the deletion on DIM production, we used [14C]-pHBA feeding to label PGLs or [14C]-propionate feeding to label both PGLs and PDIMs/PNDIMs (Figure 3A). Analysis of labeled PGLs and PDIMs/PNDIMs from the parental strain and ΔfadD22 isolates revealed that deletion of fadD22 abrogates PGL production but not PDIM/PNDIM production (Figure 3B). The ΔfadD22 strain transformed with pJAM2-FadD22tb (a plasmid expressing M. tuberculosis fadD22, identical to M. bovis fadD22) produced PGLs, thus ruling out the possibility that the deletion prevents expression of pks15/1 or other neighboring gene required for PGL production (Figure 3B). Conversely, transformation with the control vector pJAM2 failed to restore PGL biosynthesis (Figure 3B). In concordance with the fact that PGLs are not required for growth in vitro, no growth difference was observed ex vivo between the parental and mutant strains (Figure S2).
To begin assessing the catalytic competence of FadD22, we explored whether non-phosphopantetheinylated (apo) FadD22 (C-terminally hexahistidine tagged M. marinum orthologue; Figure S3) was able to synthesize pHB-AMP. In radiolabeling experiments with both [14C]-pHBA (Figure 4A) and [α-32P]-ATP (not shown), FadD22 catalyzed the ATP-dependent formation of labeled pHB-AMP as detected by TLC (Rf ~ 0.5) and confirmed by LC-MS (Figure 4A and LC-MS analysis not shown; see Experimental Methods). Based on these analyses, we concluded that FadD22 has pHBA adenylation activity. M. marinum FadD22(S576A), a protein variant with the predicted Ser-576 phosphopantetheinylation site substituted by Ala (Figure S3), also catalyzed pHB-AMP formation (Figure 4A-C), indicating that the conserved Ser is not essential for adenylation activity.
Importantly, addition of CoA to the reactions had no impact on pHB-AMP formation, nor did it result in the formation of additional radiolabeled products detectable by TLC. In particular, pHB-CoA formation was not detected by TLC (Figure 4B) or LC-MS (not shown). In contrast, pHB-CoA (Rf = 0.37) formation was detected in controls with a Rhodopseudomonas palustris benzoyl-CoA ligase  that accepts pHBA as a substrate (Figure 4B and LC-MS analysis not shown; see Experimental Methods). As expected, benzoyl-CoA ligase-dependent accumulation of the pHB-AMP intermediate was observed in absence of CoA (Figure 4B). Overall, these observations support the view that FadD22 is not a pHB-CoA ligase.
Time-courses of pHB-AMP formation by apo-FadD22 and FadD22(S576A) revealed a maximum pHB-AMP accumulation corresponding to a pHB-AMP/enzyme ratio of ~5, suggesting only 5 catalytic turnovers and that adenylation is inhibited by the acyl-AMP intermediate (Figure 4C). However, the binding of the intermediate to the enzyme does not appear to be tight enough to stop adenylate accumulation at 1:1 stoichiometry. Overall, this is consistent with the fact that mechanistically related acyl adenylate-forming enzymes bind (non-covalently) their cognate acyl-AMP intermediates 2–3 orders of magnitude more tightly than their carboxylic acid and ATP substrates [39-42]. It is likely that pHB-AMP is sequestered in the adenylation domain until transesterification to the ArCP domain liberates the adenylation domain for another catalytic cycle.
To investigate whether apo-FadD22 retains bound pHB-AMP, we examined the enzyme and [14C]-pHB-AMP elution patterns obtained when [14C]-pHB-AMP-formation reactions were subjected to size-exclusion chromatography. pHB-AMP co-eluted with the enzyme, rather than following the small-molecule elution profile delineated by [14C]-pHBA (Figure 4D). Equivalent results were obtained with FadD22(S576A) (not shown). We verified that the acyl adenylate intermediate is stable in solution for at least 2 h at 37 °C (Figure S4), thus excluding the possibility that the lack of a significant amount of free, unbound intermediate is due to rapid hydrolysis of released pHB-AMP. Reasonable stability of pHB-AMP and comparable intermediates in aqueous solutions has been reported . Thus, the elution pattern of pHB-AMP is in agreement with the formation of a stable, non-covalent FadD22·pHB-AMP complex.
The second step in our proposed initiation mechanism is transesterification of the p-hydroxybenzoyl moiety from pHB-AMP to the phosphopantetheinylated (holo) ArCP domain (Figure 2A). We developed a phosphopantetheinylation assay based on FlashPlate technology  (Supplementary Methods) to investigate whether FadD22 and FadD22(S576A) are targets for phosphopantetheinylation. In this assay we used the phosphopantetheinyl transferase Sfp  and probed for incorporation of [3H]-phosphopantetheinyl group into these proteins. These studies demonstrated that apo-FadD22 is a substrate for phosphopantetheinylation while FadD22(S576A) is not (Figure 5A). These results support the presence of a carrier protein domain in FadD22 and the role of Ser-576 as the modification site. We also found that FadD22 co-expressed with Sfp (to obtain holo-FadD22) could not be [3H]-phosphopantetheinylated in vitro (not shown), suggesting that Sfp-dependent phosphopantetheinylation of FadD22 in E. coli proceeds with ~100% efficiency.
The ability to obtain holo-FadD22 enabled us to explore whether FadD22 is competent for pHB-S-FadD22 formation (Figure 2A). This autoacylation was assayed by probing for covalent incorporation of a [14C]-pHBA-derived label into purified holo-FadD22. We also examined [14C]-pHBA loading onto apo-FadD22 and FadD22(S576A). Holo-FadD22 was readily autoacylated in an ATP-dependent and CoA-independent manner (Figure 5B). Maximum autoacylation reached 100% (Figure 5D), in agreement with the 100% conversion of apo- to holo-protein inferred from the phosphopantetheinylation analysis. Conversely, neither apo-FadD22 nor FadD22(S576A) were capable of autoacylation (Figure 5B). These observations indicate that FadD22 is not phosphopantetheinylated to a meaningful degree in E. coli in the absence of Sfp. Apo-FadD22 and FadD22(S576A), like holo-FadD22, are able to form pHB-AMP (Figure 5C), and we attribute their lack of starter unit loading to the absence of the phosphopantetheine prosthetic group. Overall, our studies of FadD22 demonstrate its pHBA adenylation activity, its apo- to holo-protein conversion by phosphopantetheinylation, and its self-loading with pHBA.
We and others have recently described a non-hydrolyzable mimic of salicyl-AMP, salicyl-AMS (5′-O-[N-salicylsulfamoyl]-adenosine), as a tight-binding inhibitor of salicylic acid adenylation domains [46-48]. Based on this precedent, we postulated that non-hydrolyzable pHB-AMP analogs would inhibit the adenylation domain of FadD22. To evaluate this idea, we synthesized the pHB-AMP analog 5′-O-[N-(4-hydroxybenzoyl)sulfamoyl]-adenosine (2) (pHB-AMS) (Figure 6A and Figure S5), and tested the activity of this compound in pHB-AMP and pHB-S-FadD22 formation assays and DIM production assays. pHB-AMS inhibited both pHB-AMP formation (IC50 = 6 μM; Figure 6B) and pHB-S-FadD22 formation (IC50 = 11 μM; Figure 6C) with IC50 values that are <10 × [enzyme], a characteristic of tight-binding inhibitors . When tested in PGL and PDIM/PNDIM production assays with M. tuberculosis, M. marinum, M. kansasii, and M. bovis (not shown), pHB-AMS strongly inhibited PGL production in all four species at up to 14- to 48-fold reduction (Figure 7A). Interestingly, this dramatic inhibition of PGL production correlated with a slight (1.3- to 3.4-fold) increase in PDIM/PNDIM production (Figure 7A), probably due to augmented availability of enzymes and building blocks for PDIM/PNDIM assembly due to metabolic flux redirectioning upon PGL biosynthesis shutdown. We also demonstrated that increasing the intracellular concentration of FadD22 or its adenylation domain reduced the effectiveness of pHB-AMS (Figure 7B). This result is in line with an expected inhibitor titration effect and provides additional evidence that the adenylation domain of FadD22 is the cellular target of pHB-AMS in the PGL pathway.
Since PGLs are dispensable for ex vivo growth, pHB-AMS is not expected to have antimicrobial activity in vitro. Thus, to assess the selectivity of this compound, we examined its activity against M. tuberculosis, M. bovis, M. marinum, or M. kansasii. As expected, pHB-AMS (up to 800 μM) had no effect on growth in cultures started with the high inoculum used in the PGL inhibition experiments above (OD580nm = 0.6). Very modest growth inhibition was detected at the maximum compound concentration in M. tuberculosis and M. bovis cultures started with low inoculum (OD580nm = 0.001), which affords a more sensitive setting for antimicrobial activity detection. This suggests that the inhibitor has only a marginal off-target effect, which is detectable at a considerably high compound to cell ratio. These results provide important support for the selectivity of pHB-AMS in targeting mycobacterial PGL biosynthesis.
Several mycobacterial pathogens produce PGLs, a family of surface-exposed bioactive glycolipids that function as small-molecule effectors in the host-pathogen interplay and contribute to virulence. The work presented herein provides the first insights into the molecular mechanism of phenolphthiocerol biosynthesis initiation, a fundamental step in PGL assembly. Our studies led to the identification of FadD22 as a stand-alone, didomain loading module that is integral to the phenolphthiocerol biosynthesis machinery. FadD22 is responsible for CoA-independent activation and charging of a pHBA starter unit to form pHB-S-FadD22, the first committed acyl-S-enzyme covalent intermediate in the phenolphthiocerol chain assembly line.
Our in silico analysis pointed to the orphan protein FadD22, previously annotated as a putative fatty acyl-CoA ligase, as the lead candidate for the initiation module required for phenolphthiocerol biosynthesis. Our genetic analysis has shown that fadD22 is essential for PGL production and our in vitro analysis of FadD22 has validated the functionality of its adenylation and ArCP domains. The catalytic partnership of these two domains leads to formation of pHB-S-FadD22. We propose that this charged module primes phenolphthiocerol biosynthesis by presenting the starter unit for the first extension cycle of the p-hydroxyphenyl-polyketide formation process. Starter unit extension is then carried out by the next enzyme in the phenolphthiocerol assembly line, probably Pks15/1. Our ability to obtain pHB-S-FadD22 will facilitate future enzymological studies to decipher the functional cooperation of this initiation module with Pks15/1 or other potential enzyme partners in the phenolphthiocerol assembly line.
The CoA-independence of the phenolphthiocerol biosynthesis initiation mechanism parallels that observed in the biosynthetic pathway of rifamycin B . The rifamycin B system initiation module consists of an adenylation domain for 3-amino-5-hydroxybenzoate and its partner ArCP domain for formation of a 3-amino-5-hydroxybenzoyl-S-ArCP domain intermediate (Figure 2B). The mechanism of phenolphthiocerol biosynthesis initiation is also reminiscent of that observed in the biosynthesis of the salicyate-capped, hybrid non-ribosomal peptide-polyketide (carboxy)mycobactin siderophores produced by M. tuberculosis and other Mycobacterium spp. . During the biosynthesis of these siderophores, MbtA catalyzes adenylation of salicylic acid and subsequent transesterification onto a holo-ArCP domain located at the N-terminus of the non-ribosomal peptide synthetase MbtB (Figure 2C). The parallel aryl acid-specific adenylation-ArCP didomain composition of the initiation machinery of the phenolphthiocerol and siderophore pathways suggests an intriguing possible evolutionary relationship between the PGL and the siderophore biosynthesis machinery. Interestingly, the adenylation-ArCP didomain architecture differs between these three aryl acid-specific loading systems. While FadD22 is a stand-alone didomain module, the rifamycin B system has a didomain initiation module that is part of a larger protein (RifA) containing three polyketide synthase modules, while the siderophore pathway features an in trans loading strategy with a stand-alone adenylation domain and an ArCP domain fused to an additional module in MbtB (Figure 2).
The mechanistic insights into phenolphthiocerol biosynthesis initiation provided us with a framework to design a small-molecule inhibitor of PGL biosynthesis, pHB-AMS. The inhibitory activity of pHB-AMS in enzymatic and cellular assays affords additional support for our model of phenolphthiocerol biosynthesis initiation. These studies also provide proof-of-principle for the drugability of the PGL pathway and for inhibition of the biosynthesis of a bona fide polyketide virulence factor. It has been suggested that drugs that inhibit the biosynthesis or the function of virulence factors may be combined with classical antimicrobials to afford more efficient treatments against infections [50-53]. It is interesting to speculate that drugs that block PGL biosynthesis may reduce the adaptive fitness of PGL-producing M. tuberculosis strains in the human host by eliminating PGL-dependent immunomodulatory effects. These drugs may also diminish the ability of M. leprae to invade Schwann cells and produce nerve function impairment. Thus, pHB-AMS represents an initial lead compound for exploring the potential therapeutic value of PGL biosynthesis inhibitors in animal infection models in the future.
Mycobacterial PGLs are a family of surface-exposed bioactive glycolipids that function as small-molecule effectors in the host-pathogen interplay and contribute to virulence. Understanding the biosynthesis of these and other effectors of mycobacterial virulence is an important goal as it may lead to new therapeutics. The studies reported herein illuminate the mechanism of chain initiation required for the assembly of the phenolphthiocerol moiety of PGLs. Our results support a model in which a stand-alone, didomain initiation module comprised of a pHBA adenylation domain and an aroyl carrier protein domain forms the first acyl-S-enzyme covalent intermediate in the p-hydroxyphenyl-polyketide chain assembly line. The stand-alone status of this initiation module sets it apart from related aryl acid primer unit didomain loading systems characterized previously. Moreover, to our knowledge, FadD22 is the first characterized bona fide pHBA-specific initiation module implicated in a polyketide biosynthetic pathway. Lastly, the insights gained on the mechanism of phenolphthiocerol biosynthesis initiation allowed us to develop an inhibitor of PGL assembly with potent activity in several mycobacterial pathogens. Overall, these studies advance our understanding of the biosynthesis of an important group of small-molecule effectors of mycobacterial virulence and provide important support for the feasibility of targeting PGL biosynthesis to develop new drugs to treat mycobacterial infections.
Mycobacteria were grown in Middlebrook 7H9 (Difco) supplemented with 10% Albumin Dextrose Complex (Difco) and 0.1% Tween-80 or Middlebrook 7H11 plates (Difco) containing 10% Oleic Acid Albumin Dextrose Complex (Difco). M. tuberculosis Canetti, M. bovis BCG, and M. kansasii were cultured at 37 °C. M. marinum was cultured at 28 °C. When required, kanamycin (30 μg/mL), hygromycin (50 μg/mL) and/or sucrose (2%) were added to the media.
Cultures grown for 8 d (M. tuberculosis and M. bovis) or 7 d (other species) were diluted in fresh medium to OD580nm = 0.6 and loaded in 12-well plates (1 mL/well). Wells were treated (in triplicate) with 0.8% DMSO (controls) or pHB-AMS at different concentrations and 0.8% DMSO. [1-C14]-propionate (specific activity [sp act] = 54 mCi/mmol; ARC, Inc.) was added to each well at 0.2 μCi/mL and plates were incubated for 12 h before the OD of the cultures was measured and cells were harvested for apolar lipid extraction. Apolar lipids were extracted with a biphasic mixture of methanolic saline and petroleum ether (PE) as described [27, 54]. PGLs in apolar lipid extracts were analyzed by TLC on aluminum-backed 250 μ silica gel plates (EM Science) with CHCl3/MeOH (95:5 v/v) eluent as reported [20, 28, 55]. PDIMs and PNDIMs were analyzed by TLC with PE/diethyl ether (Et2O) (9:1 v/v) eluent as reported [27, 54]. Plates were exposed to phosphor screens, which were scanned using a Typhoon Trio Imager (Amersham Biosciences). PGL, PDIM, and PNDIM signals were quantified using ImageQuant v1.2 (Molecular Dynamics). Dose-response graphs for these and other experiments herein were generated using KaleidoGraph 4.0. IC50 values were calculated by fitting the data to the sigmoid equation , Eq. (1), where: Si and Sc are the background-corrected signals of the lipids in the inhibitor-treated cultures and DMSO controls respectively; a and b are the top and bottom of the curve respectively; and s is the Hill coefficient.
Cultures grown for 7 d were diluted in Middlebrook 7H9 to OD580nm of 0.6 or 0.001 and loaded in 96-well plates (200 μL/well). Wells were treated (in triplicate) with 0.8 % DMSO (controls) or pHB-AMS at different concentrations and 0.8% DMSO. Plates were incubated for growth for 10 d and OD580nm was monitored daily in a Spectra Max Plus reader (Molecular Dynamics).
The ΔfadD22 mutant was engineered using the p2NIL/pGOAL method as reported  (Supplementary Methods). Briefly, we constructed a ΔfadD22 cassette-delivery suicide vector carrying a deletion cassette that contains a 5′ arm (1.2-kb region upstream of fadD22 and fadD22's start codon) and a 3′ arm (fadD22's last ten codons and the downstream 810-bp segment) (Figure S2). The C-terminal codons were preserved due to sequence overlap with pks15/1. This vector was introduced into M. bovis and mutants with fadD22 replaced by the ΔfadD22 cassette via double-crossover were selected and identified by reported methods [54, 56].
pJAM-FadD22tb, pJAM-FadD22tbAd, and pJAM-FadD22ml were based on the mycobacterial expression vector pJAM2  and express the mycobacterial genes under control of the acetamidase promoter. pCP0-FadD22tb, pCP0-FadD22tbAd, and pCP0-FadD22ml were based on the mycobacterial expression vector pCP0 and express the mycobacterial genes under control of the hsp60 promoter . See Supplementary Methods for plasmid construction.
Reaction contained 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 50 μM [14C]-pHBA and, depending on the specific reaction indicated, combinations of the following components: 1 mM ATP, 100 μM CoA, 0.5 μM Sfp, 5 μM FadD22, 5 μM FadD22(S576A), or 5 μM BZLRp. Reactions were incubated (2 h, 30 °C) and product formation was analyzed by radiometric TLC using Al Sil G/UV TLC plates (Whatman) and ethyl acetate (EtOAc)/isopropyl alcohol/acetic acid/water (70:20:25:40 v/v) eluent. Plates were exposed to phosphor screens. The screens were scanned using a Typhoon Trio Imager and signal intensity was quantified using ImageQuant v1.2 software. For time-course experiments, pHB-AMP-formation reactions (100 μL) containing 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 1 mM ATP, 50 μM [14C]-pHBA, and 2 μM of either FadD22 or FadD22(S576A) were incubated at 30 °C. Reaction samples were taken at different times for TLC analysis and pHB-AMP quantification. Concentrations of [14C]-pHB-AMP (or [14C]-pHB-CoA) were calculated using calibration curves obtained by linear regression fitting to calibration data generated by spotting known amounts of [14C]-pHBA on the plates after chromatography. The identity of pHB-AMP and pHB-CoA products was confirmed by LC-MS analysis of the (unlabeled) compounds isolated by preparative TLC. pHB-CoA: [M+H]+m/z observed 888.1432, calculated 888.1436; [M−H]-m/z observed 886.1288, calculated 886.1280; (not shown). pHB-AMP: [M+H]+m/z observed 468.0916 (Figure 4A), calculated 468.0915; [M−H]-m/z observed 466.0769 (not shown), calculated 466.0769. The proteins used in these and other experiments herein were expressed in E. coli and purified as described in Supplementary Methods.
pHB-AMP and pHB-CoA were purified from pHB-AMP and pHB-CoA formation reactions, respectively, by preparative TLC. The product was eluted with methanol and the extract was subjected to LC-MS analysis. Other samples for LC-MS analysis were prepared as follows. Reactions (100 μL) containing 5 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP and, depending on the reaction, 1 mM ATP, 1 mM pHBA, 1 mM CoA, and 10 μM of either FadD22 or BZLRp, were incubated 2 h at 30 °C in Mini Dialysis Units (Slide-A-Lyzer, cutoff: 3.5 kDa, Pierce). Each unit was kept in contact with 100 μL of a reaction mixture lacking enzyme so that released products accumulated in the enzyme-free mixture, which was then subjected to LC-MS analysis. Samples were analyzed on an Agilent Technologies 6210 high resolution time-of-flight mass spectrometer connected to an Agilent Technologies 1200 capillary HPLC system. See Supplementary Methods for details.
pHB-AMP-formation reactions containing 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 1 mM ATP, [14C]-pHBA, and 15 μM FadD22 were incubated 10 min at 30 °C. After incubation, reaction aliquots (50 μL) were applied to Gel Filtration G-50 Macrospin Columns (The Nest Group, Inc.) equilibrated with 75 mM MES pH 6.5. Columns were centrifuged (3 min, 110 g) and flowthrough fractions were collected. The columns were then successively washed with aliquots of 75 mM MES pH 6.5 (50 μL) and the flowthrough fractions were collected. For each flowthrough fraction, a sample (5 μL) was subjected to TLC analysis and pHB-AMP quantification and a sample (10 μL) was analyzed by SDS-PAGE (12.5%). Gels were stained using GelCode Blue Stain (Pierce) and photographed. Densitometric analysis to quantify protein bands was performed using Quantity One 4.5.2 software (Bio-Rad Laboratories).
Reactions (50 μL) containing 75 mM Tris·HCl pH 7.5, 10 mM MgCl2, 1 mM TCEP, 1 μM Coenzyme A (4′-phosphopanetheine-3H; 11 Ci/mmol; ARC, Inc.), 0.25 μM Sfp (absent in negative controls), and 0.75 μM of either FadD22, FadD22(S576A), FadD22 co-expressed with Sfp, or FadD22(S576A) co-expressed with Sfp were incubated for 1 h at 37 °C. After incubation, reactions were diluted with 150 μl of PBS containing 20% MeOH, 1 mM CoA and 0.1% BSA, and the mixtures were transferred to 96-well Ni2+-chelate coated FlashPlate Plus plates (Perkin Elmer). After tagged protein binding (overnight, 4 °C), wells were washed with PBS (3 × 300 μL) and well-bound counts were quantified in a Wallac Microbeta counter. Counts were converted to amounts of phosphopantetheinylated protein using the sp act of [3H]-CoA. For time-dependent phosphopantetheinylation experiments, reactions (50 μL) containing 75 mM Tris·HCl pH 7.5, 10 mM MgCl2, 1 mM TCEP, 0.3 μM [3H]-CoA, 0.25 μM Sfp, and 1 μM of either FadD22 or FadD22(S576A) were diluted as above at different times and transferred to the plates for protein binding and phosphopantetheinylation quantification.
Reactions (60 μL) containing 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 50 μM [14C]-pHBA and, depending on the reaction indicated, 1 mM ATP, 100 μM CoA and 10 μM of either FadD22, FadD22(S576A), FadD22 co-expressed with Sfp, or FadD22(S576A) co-expressed with Sfp were incubated for 1 h at 30 °C. After incubation, reaction aliquots (10 μL) were diluted with PBS (15 μL) containing 2.5 mM pHBA and 20% MeOH. The mixtures were transferred to nitrocellulose membranes (pore size 0.2 μm; Protran, Whatman) for protein binding using Minifold I Slot Blot System (Whatman). Membranes were washed sequentially with PBS 20% MeOH (3 × 25 mL) and 10 % TCA (25 mL), and then exposed to phosphor screens. Phosphor screens were scanned and signal intensity was quantified as above. Amounts of [14C]-pHB-protein thioester in the samples were calculated using calibration curves obtained by linear regression fitting to calibration data generated from known amounts of [14C]-pHBA spotted on the membranes. An additional aliquot (4 μL) was taken from each reaction and subjected to TLC and pHB-AMP quantification as above. For time-course experiments, aliquots from pHB-protein thioester formation reactions (75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 1 mM ATP, 50 μM [14C]-pHBA, and 5 μM FadD22 co-expressed with Sfp) taken at different times were diluted and analyzed for [14C]-pHB-protein thioester formation as above.
Reactions (20 μL) containing 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 1 mM ATP, 50 μM [14C]-pHBA, 2 μM FadD22 and 0.15% DMSO (controls) or 0.15% DMSO and pHB-AMS at different concentrations were incubated for 1 h at 30 °C. After incubation, reaction aliquots were subjected to TLC analysis and pHB-AMP quantification as noted above. Sets of dose-response data were fitted to the Morrison equation for tight-binding inhibitors , Eq. (3), where E = enzyme; I = inhibitor; Vi and Vc = velocities in the presence and absence of inhibitor, respectively . IC50 values were calculated with the equation for tight-binding inhibitors , Eq. (2), using values derived from the curve fit . The inhibitor used in these and other studies herein was synthesized by 5′-O-sulfamoylation of a protected adenosine derivative, followed by N-acylation of the sulfamate with p-acetoxybenzoic acid and removal of protecting groups (see Supplementary Methods; Figure S5).
pHB-protein thioester formation reactions (20 μL) containing 75 mM MES pH 6.5, 0.5 mM MgCl2, 1 mM TCEP, 1 mM ATP, 50 μM [14C]-pHBA, 5 μM FadD22 co-expressed with Sfp, and 0.15% DMSO (controls) or 0.15% DMSO and pHB-AMS at different concentrations were routinely incubated for 40 min at 30 °C. After incubation, 40 μL of PBS containing 20 % MeOH and 1 mM pHBA were added to each reaction and 30 μL of each mixture was transferred to nitrocellulose membranes for quantification of [14C]-pHB-protein thioester in the samples as noted above. Sets of dose-response data were fitted to Eq. (3) and IC50 values were calculated Eq. (2) as above.
We thank Albert Morrishow (WMC) and Dr. George Sukenick, Hui Fang, Hiu Liu, and Sylvi Rusli (MSKCC) for mass spectral analyses. Generous financial support has been provided by the National Institutes of Health (grant AI069209 to L.E.N.Q.), Stavros S. Niarchos Foundation (L.E.N.Q.), NYSTAR Watson Investigator Program (D.S.T.), William Randolph Hearst Foundation (L.E.N.Q. and D.S.T.), William H. Goodwin and Alice Goodwin and the Commonwealth Foundation for Cancer Research (D.S.T.), and MSKCC Experimental Therapeutics Center (D.S.T.).
Competing Interests Statement: The authors declare that they have no competing financial interests.
Author Contributions: L.E.N.Q. was responsible for designing the study, directing the project, and writing the manuscript. D.S.T. oversaw the synthetic aspects of the project. D.S.T., X.L., J.-S.R., J.A.F., K.L.S., and C.E.S. edited the manuscript. J.A.F. contributed to the experimental design and conducted the enzyme characterization and inhibition experiments. K.L.S. contributed to the experimental design and conducted the genetic analysis and PGL inhibition experiments. C.E.S. performed the MS analysis. X.L. and J.-S.R. developed and executed the synthesis of pHB-AMS.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.