Characterization of the Complete Zwittermicin A Biosynthesis Gene Cluster from Bacillus cereus

Abstract

Bacillus cereus UW85 produces the linear aminopolyol antibiotic zwittermicin A (ZmA). This antibiotic has diverse biological activities, such as suppression of disease in plants caused by protists, inhibition of fungal and bacterial growth, and amplification of the insecticidal activity of the toxin protein from Bacillus thuringiensis. ZmA has an unusual chemical structure that includes a d amino acid and ethanolamine and glycolyl moieties, as well as having an unusual terminal amide that is generated from the modification of the nonproteinogenic amino acid β-ureidoalanine. The diverse biological activities and unusual structure of ZmA have stimulated our efforts to understand how this antibiotic is biosynthesized. Here, we present the identification of the complete ZmA biosynthesis gene cluster from B. cereus UW85. A nearly identical gene cluster is identified on a plasmid from B. cereus AH1134, and we show that this strain is also capable of producing ZmA. Bioinformatics and biochemical analyses of the ZmA biosynthesis enzymes strongly suggest that ZmA is initially biosynthesized as part of a larger metabolite that is processed twice, resulting in the formation of ZmA and two additional metabolites. Additionally, we propose that the biosynthesis gene cluster for the production of the amino sugar kanosamine is contained within the ZmA biosynthesis gene cluster in B. cereus UW85.

Bacillus cereus strain UW85 was isolated for its ability to suppress disease in alfalfa caused by the plant pathogen Phytophthora medicaginis (17). This antiprotist activity was subsequently found to be associated with the filtrate of fully sporulated B. cereus UW85 (37). Analysis of this filtrate identified two antiprotist antibiotics, zwittermicin A (ZmA) and kanosamine (28, 37). Of the two antibiotics, ZmA has shown the more interesting biological activities, having not only antiprotist activity, but also antibiotic activity against gram-positive and gram-negative bacteria, as well as fungi (32, 38). ZmA was also found to potentiate the activity of the toxin protein of Bacillus thuringiensis against insects (3).

A preliminary chemical structure of ZmA was determined by the Handelsman and Clardy groups (18). More recently, Rogers and colleagues performed a series of elegant structural studies that compared ZmA produced from B. cereus with synthetic ZmA derivatives that had varied stereocenters (32, 33). From this work, the chemical structure of ZmA with the appropriate stereocenters has been determined (Fig. 1). The antibiotic has a number of unusual structural components. First, ZmA is one of only a few linear aminopolyol natural products to be identified. Second, the core of ZmA is formed from ethanolamine and glycolyl moieties that are rarely seen in natural products. Third, the N terminus of ZmA is formed from d-serine (d-Ser), not l-Ser, as initially expected. This suggests that the amino acid either is incorporated as the d isomer or is incorporated as the l isomer and is then isomerized at some point during its biosynthesis. Finally, ZmA is the only natural product that we are aware of that contains an unusual 2-aminosuccinamide moiety. This moiety is likely to come from the amino acid β-ureidoalanine (β-Uda) that has had its carboxylic acid replaced by a terminal amide.

FIG. 1.

Chemical structure of ZmA. Numbers have been added to identify the sites of hydroxyl groups as discussed in the text.

We have been investigating how B. cereus UW85 assembles this antibiotic to gain insights into how production of ZmA can be improved and how the unusual structural components of ZmA are formed. We previously proposed that the biosynthesis of ZmA involves the condensation of the amino acids l-Ser and l-2,3-diaminopropionate (l-Dap), along with the carboxylic acid precursors malonyl-coenzyme A (CoA), (2S)-aminomalonyl-acyl carrier protein (ACP), and (2R)-hydroxymalonyl-ACP (4, 12). The proposal for l-Ser incorporation was made prior to the full elucidation of the stereochemistry of ZmA. The condensation of amino acids and carboxylic acids suggests that ZmA is assembled via a nonribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) megasynthase. These megasynthases are modular enzymes with a set of catalytic domains, or modules, for each precursor incorporated into the natural product (reviewed in references 13 and 40). Support for the involvement of this type of enzymology in ZmA biosynthesis comes from a combination of genetic and biochemical studies. Transposon mutagenesis of B. cereus UW85 identified insertions in genes coding for NRPS and PKS enzymology that abolished ZmA production (12). Sequencing of a 19-kb fragment of the ZmA biosynthesis gene cluster identified genes coding for NRPS and PKS enzymology (12). Furthermore, other groups investigating ZmA biosynthesis in B. thuringiensis strains have identified genes coding for NRPS modules that are essential for ZmA biosynthesis in these strains (35, 49, 50). Finally, we used biochemistry and mass spectrometry to establish the existence of two ACP-linked PKS extender units, (2S)-aminomalonyl-ACP and (2R)-hydroxymalonyl-ACP (4). All of these data support the hypothesis that the backbone of ZmA is assembled by an NRPS/PKS megasynthase.

In addition to the mixed amino acid and carboxylic acid backbone, ZmA also contains a terminal amide (Fig. 1). How these amide groups are formed was investigated by Müller and colleagues and Silakowski and colleagues as they deciphered how myxothiazole is biosynthesized (29, 36). Briefly, the NRPS/PKS megasynthase that assembles the backbone of myxothiazole forms a product that is 1 amino acid longer than myxothiazole. This results in a biosynthetic intermediate that contains a glycyl residue at the C terminus of myxothiazole, while the intermediate remains thioesterified to the peptidyl carrier protein (PCP) domain of the terminal NRPS module. The α-carbon of the glycine is hydroxylated by a flavin-dependent monooxygenase, a modification that results in an unstable intermediate that spontaneously releases the myxothiazole backbone, with the nitrogen of the terminal amide coming from the glycine. The terminal PCP domain contains the glyoxyl group left after C-N bond cleavage, and this product is released from the PCP domain by the neighboring thioesterase (Te) domain. Based on this precedent, the terminal amide of ZmA may be produced by a similar mechanism.

Here, we present the identification of the complete ZmA biosynthesis gene cluster from B. cereus UW85. The biosynthesis gene cluster was identified by locating the previously reported biosynthesis genes and by mapping the locations of transposon insertions that abolished the ability of B. cereus UW85 to produce ZmA. As expected, the gene cluster codes for NRPS and PKS enzymology that is likely to be involved in ZmA assembly from its amino acid and carboxylic acid precursors. Surprisingly, we fiound that ZmA not only is likely to be processed at its C terminus to generate the terminal amide by a mechanism similar to that seen in myxothiazole biosynthesis, but it appears to also be processed at its N terminus. These two processing events potentially lead to the biosynthesis of two additional metabolites besides ZmA. Furthermore, the kanosamine biosynthesis gene cluster appears to be fully contained within the ZmA biosynthesis gene cluster. A mechanism for ZmA production is presented, along with proposals for how three additional metabolites are produced by the enzymes encoded by this unusual gene cluster.

MATERIALS AND METHODS

Sequencing of the B. cereus UW85 genome.

Library construction, random shotgun sequencing, and genome assembly were performed as described previously (31). One large (10- to 12-kb)-insert and one small (4- to 5-kb)-insert random plasmid-sequencing libraries were sequenced for each of these genome projects, with sequencing success rates greater than 85% and average high-quality read lengths greater than 800 nucleotides. The genome-sequencing project of B. cereus UW85 currently consists of approximately 51,000 reads representing approximately 6.5-fold Sanger sequencing coverage. The data have been assembled using the Celera Assembler into 269 putative contigs, and the ZmA cluster was identified based on previous sequence knowledge and analysis of transposon insertion sites that abolished ZmA production. Of the assembled contigs, we identified 214 as chromosomal and 4 as plasmid based on similarity to other B. cereus group plasmids, while 51 contigs of >2 kb could not be placed. We are continuing to work on the B. cereus AH1134 genome; however, a preliminary assembly has been released to the public (GenBank accession number ABDA02000000). Unfortunately, there is no one Sanger sequencing clone that contains the entire ZmA gene cluster in either the B. cereus UW85 or AH1134 sequencing projects due to the size of the cluster (∼65 kb), and as such, we could not examine if the gene cluster could confer the ZmA production phenotype independently of any other genomic factors.

Generation of A domain overproduction clones.

PCR-based cloning was used to introduce the DNA coding for each adenylation (A) domain into an Escherichia coli expression vector. The primers used for PCR amplification are shown in Table 1. Each set of the PCR primers was designed with an NheI recognition site in the primer at the 5′ end of the gene and a SalI recognition site in the primer at the 3′ end of the gene. The amplicons were cloned into the corresponding restriction sites of the pET28b vector (Novagen, Madison, WI). The correct sequence for each cloned gene was verified by sequencing at the University of Wisconsin Biotechnology Sequencing Center (Madison, WI). Each overexpression vector produced a protein with an N-terminal hexahistidine tag that was used for affinity purification.

TABLE 1.

Primers used for PCR amplification of NRPS genes

A domain construct	Orientationa	Primer sequence (5′ → 3′)	Corresponding amino acids
ZmaB-A1	F	GCAGCTAGCATGGTTGAACGAATAGCAAATAAAG	1-802
	R	GCAGTCGACTCACACTACTTTCCATATCTCAGAGAG
ZmaB-A2	F	GCAGCTAGCAGAGAATTTTATCAAGCTTCTTC	876-1906
	R	GCAGTCGACCTAATTATTTTTCTTTTCTGC
ZmaK	F	GCAGCTAGCGATGAGGCAGAAGAAAAGTTTTTGC	421-963
	R	GCAGCTAGCTCATTTTTCTACCCCCAAAACTTCCTTCC
ZmaO	F	GCAGCTAGCAAAACTATTCAAGAATTATTTGAAG	456-974
	R	GCAGCTAGCTCATTCCTTGAGAACTTCACTAACCATTCC
ZmaQ-A1	F	GCAGCTAGCACAGAAAAGAAACAAATATTATATG	251-797
	R	GCAGCTAGCTCAATCTTTCCAAATCTCAACTAATTTTC
ZmaQ-A2	F	GCAGCTAGCATGTTATCTATAGAAGAAGAACATTG	1272-1818
	R	GCAGCTAGCTCAACTACTCCAAATATCAATTAAACTAC

^aF, forward; R, reverse.

Recombinant-protein overproduction and affinity purification.

Each overexpression vector was introduced into E. coli BL21(DE3) for protein overproduction. Cells were grown at 28°C in LB medium (two 1-liter batches) containing 50 μg/ml kanamycin until the optical density at 600 nm reached 0.4 to 0.6. The cells were subsequently shifted to 15°C for 1 h prior to induction with isopropyl-d-thiogalactopyranoside (60 μM final concentration) and grown for 16 h at 15°C. Cells were harvested by centrifugation (10 min at 7,000 rpm; Beckman J2-21 centrifuge; Kompspin KA-9 rotor) and subsequently resuspended in protein purification buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 10% [vol/vol] glycerol). Each overproduction strain resulted in ∼10 g of cells (wet weight), which were resuspended in 20 ml of protein purification buffer. The cell suspension was sonicated (Fisher 550 Sonic Dismembrator; power = 5; 15 min of sonication with 1 s on, 1 s off), and the cell debris was removed by centrifugation (30 min at 15,000 rpm; Beckman J2-21 centrifuge; Beckman JA-25.50 rotor) at 4°C. Imidazole (5 mM final concentration) was added to the cleared lysate, and the lysate was then incubated with 1 ml of Ni-nitrilotriacetic acid resin (Qiagen) for 1 h with gentle rocking. The resin was applied to a column, and protein was eluted using a step gradient with 5 ml of protein purification buffer containing increasing concentrations of imidazole (20, 40, 60, 100, and 250 mM). The elutions were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Coomassie blue staining, and elutions containing pure recombinant protein were combined and dialyzed for 16 h at 4°C in dialysis buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 10% [vol/vol] glycerol). After dialysis, the protein concentration was determined spectrophotometrically at A₂₈₀ by use of the calculated molar extinction coefficients: ZmaB-A1, 72,825 M⁻¹ cm⁻¹; ZmaB-A2, 97,320 M⁻¹ cm⁻¹; ZmaK-A, 57,885 M⁻¹ cm⁻¹; ZmaO-A, 40,865 M⁻¹ cm⁻¹; ZmaQ-A1, 56,075 M⁻¹ cm⁻¹; and ZmaQ-A2, 50,240 M⁻¹ cm⁻¹. The proteins were flash-frozen in liquid nitrogen and stored at −80°C until they were used.

dATP-³²PP_i exchange assays for aminoacyl-AMP formation.

Each purified enzyme containing an A domain was analyzed for amino acid-dependent dATP-³²PP_i exchange using standard protocols (30, 43). Briefly, exchange assays were conducted for each amino acid individually in a reaction mixture (100 μl) containing 500 μM of the amino acid to be tested, 75 mM Tris-HCl (pH 7.5), 10 mM MgCl₂, 5 mM dithiothreitol, 3.5 mM dATP, 1 mM ³²PPi (2.15 Ci/mol; Perkin-Elmer), and 1 μM enzyme. The A domains of NRPSs are not selective against dATP, in contrast to other adenylyltransferases (30); therefore, we substituted dATP for ATP in these reactions to ensure we were analyzing A domain-catalyzed exchange and not exchange catalyzed by other adenylyltransferase enzymes, such as tRNA synthetases. The reaction mixtures were incubated at 23°C for 30 min. The reactions were stopped by the addition of a quench solution containing 3.5% (vol/vol) perchloric acid, 100 mM Na-PP_i, and 1.6% (wt/vol) activated charcoal. The charcoal pellets were washed twice with quenching buffer lacking charcoal before being counted in a scintillation counter. Assays with each enzyme were performed two or more times for each amino acid. The results are reported as percentages of maximum substrate activation.

ZmA purification from B. cereus AH1134.

One-liter cultures of 0.5× tryptic soy broth were inoculated with 1 ml of a saturated overnight B. cereus AH1134 culture and grown at 30°C for 3 days. The cells were removed by centrifugation (10 min at 7,000 rpm; Beckman J2-21 centrifuge; Kompspin KA-9 rotor), and the culture supernatant was retained. The pH of the supernatant was adjusted to pH 7.0 with HCl and batch bound to 15 ml of Amberlite IRC-50 ion-exchange resin (Acros Organics) preequilibrated with 5 mM NH₄PO₄, for 1 h at 23°C. The resin was applied to a chromatography column and washed with 10 bed volumes of 5 mM NH₄PO₄. ZmA was eluted from the column with 5 ml of increasing concentrations of NH₄OH (250 mM, 500 mM, and 1 M). The elutions were flash frozen in liquid nitrogen and lyophilized for 2 days. The lyophilized material was resuspended in H₂O and neutralized to pH 7.0 with HCl.

ZmA analysis by HPLC and MALDI-TOF MS.

Authentic ZmA and samples purified from B. cereus AH1134 were analyzed by high-performance liquid chromatography (HPLC). The samples were separated on a Vydac SP C₁₈ silica column on a Beckman-Coulter Gold system with a 1-ml/min flow rate. Buffer A was H₂O-0.1% trifluoroacetic acid (TFA), and buffer B was acetonitrile-0.1% TFA. The separation profile was 5 min of isocratic development at 100% A-0% B and a linear gradient over 15 min to 20% acetonitrile plus 0.1% TFA. Metabolite elution was monitored at 210 nm. The metabolite corresponding to ZmA was collected and analyzed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) on a Voyager-DE Pro Workstation (PerSeptive Biosystems). The MALDI-TOF MS data were calibrated with peptides of known mass during each analysis.

Nucleotide sequence accession number.

The gene cluster identified in this study has been submitted to GenBank under accession number FJ430564.

RESULTS AND DISCUSSION

Identification of the ZmA biosynthesis gene cluster in B. cereus UW85.

In collaboration with J. Handelsman at the University of Wisconsin—Madison and the Institute of Genomic Studies at the University of Maryland School of Medicine, we are currently sequencing the genome of B. cereus UW85. As this project moves toward completion, one of our initial interests is to identify the biosynthesis gene cluster involved in the production of ZmA. We and others previously reported the identification of a 19-kb fragment of the B. cereus UW85 genome and an additional gene, zmaN, that are involved in ZmA biosynthesis (4, 12, 42). Based on the common finding that all the genes involved in the production of an antibiotic are clustered in one region of the chromosome (5), we hypothesized that the bioinformatics analysis of the B. cereus UW85 genome surrounding the previously reported genes would identify all the genes involved in ZmA biosynthesis. To this end, we identified a gene cluster consisting of 27 open reading frames (ORFs) covering 62.5 kb of the B. cereus UW85 genome that we propose codes for all the proteins required for the production of ZmA (Fig. 2). We propose that the genes in this cluster be named zmaA through zmaV to reflect their role in ZmA production, with five additional genes being named kabR and kabA through kabD. The details of the distinct gene nomenclature for the kab genes are discussed below.

FIG. 2.

ZmA biosynthesis gene clusters from B. cereus UW85 (top) and B. cereus AH1134 (bottom). For the B. cereus UW85 gene cluster, the genes labeled zma are proposed to be involved in ZmA biosynthesis and those labeled kab are proposed to be involved in kanosamine biosynthesis. Due to space limitations, some of the genes have been abbreviated (e.g., C is equivalent to zmaC). The triangles identify Tn5401 insertion sites that result in the loss of ZmA production, with the exception of UW85Δorf2, which identifies a deletion mutant that was previously constructed. The labeling of the insertion sites uses the mutant strain numbering previously described (12). The H28 arrow denotes the location of the promoter that is induced in the presence of P. aureofaciens (11). For the B. cereus AH1134 gene cluster, the number above each ORF identifies the locus tag. The locus tags have been abbreviated due to space limitations (e.g., C0218 is equivalent to BCAH1134_C0218, while 20 is equivalent to BCAH1134_C0220). The genes from B. cereus AH1134 are shown immediately below their homologs from B. cereus UW85. The genes are color coded based on their established or proposed functions: red, NRPS- or PKS-associated genes; blue, ACP-linked PKS extender unit biosynthesis; orange, ZmA resistance; yellow, ZmA processing; gray, kanosamine biosynthesis; and green, β-Uda biosynthesis.

Our conclusion that we had identified the ZmA biosynthesis gene cluster was based on a comparison of the identified gene cluster with the data from a series of genetic and biochemical studies performed by our group and the Handelsman group on ZmA production by B. cereus UW85 (4, 12, 42). As stated above, we previously reported the genetic analysis of a 19-kb fragment of the ZmA biosynthesis gene cluster covering zmaD through a portion of zmaJ (12). As part of that analysis, a number of Tn5401 insertions in the B. cereus UW85 genome that abolished ZmA production were identified. One Tn5401 transposon was inserted into zmaJ and another into zmaL (Fig. 2). At the time of that study, the genomic locations of the other Tn5401 insertion sites were unknown; however, it was noted that some were located in genes coding for NRPS- and PKS-like proteins. We reanalyzed these insertion sites after obtaining the genomic data and determined that two of these transposons were located in zmaA, two were in zmaB, and one was in zmaO (Fig. 2). Thus, Tn5401 insertions have been identified in five different genes in this biosynthesis gene cluster, and all of them abolish ZmA production by B. cereus UW85. When these data are combined with prior genetic studies that showed that a strain with zmaF deleted no longer produced ZmA (42) and our biochemical work on ZmaD, -E, -G, -H, -I, -J, and -K that showed that these enzymes produce precursors for ZmA production (4), they provide strong support for our conclusion that we have identified the ZmA biosynthesis gene cluster in B. cereus UW85.

Additional support for this conclusion comes from work that has analyzed ZmA production in B. thuringiensis strains (35, 49, 50). Shao and colleagues recently constructed a deletion of a homolog of zmaC in the ZmA-producing bacterium B. thuringiensis GO3 (35). The resulting strain was no longer able to produce ZmA, confirming that this ZmaC homolog plays a role in antibiotic biosynthesis. Additionally, Zhao and colleagues constructed strains of B. thuringiensis subsp. kurstaki strain YBT-1520 with homologs of zmaU, zmaV, and zmaQ deleted, and these strains also lost the ability to produce ZmA (49, 50). Based on the similarity of the genes from these B. thuringiensis strains and those we identified in B. cereus UW85 (97 to 100% identity) and the close evolutionary relationship between the two species (46), it was reasonable to expect that similar results would be observed in B. cereus UW85.

The identification of the ZmA biosynthesis gene cluster also provided new insights into environmental cues that induce ZmA production. Previous work investigated the genome of B. cereus UW85 for promoters that are induced when the bacterium is exposed to the plant-associated bacterium Pseudomonas aureofaciens (11). The influence of this Pseudomonas species on B. cereus UW85 gene expression was investigated due to both bacteria being isolated from the same plant root segments (11). One of the B. cereus UW85 promoters that is induced in the presence of P. aureofaciens is carried on the clone H28. The promoter from this clone is immediately upstream of a gene that at the time had no known function. We reanalyzed this promoter in comparison to the genes of the ZmA biosynthesis gene cluster and determined the promoter was immediately upstream of zmaA. Thus, one of the natural inducers of the ZmA biosynthesis gene cluster in B. cereus UW85 is a common inhabitant of the same plant root environment.

Analysis of B. cereus AH1134 for ZmA production.

During our analysis of the genes and associated proteins in the ZmA biosynthesis gene cluster, it was noticed that a nearly identical gene cluster (genes showing 97 to 100% identity) is present in the genome of B. cereus AH1134 (Fig. 2). This finding suggested that B. cereus AH1134 was capable of producing ZmA. To test this possibility, we analyzed the supernatant of a sporulated culture of B. cereus AH1134 for the presence of ZmA. A metabolite with the same elution time from HPLC as an authentic sample of ZmA was identified (Fig. 3), and it also had the same UV-visible spectrum (data not shown). This metabolite was collected from the HPLC and analyzed by MALDI-TOF MS, and the metabolite had a mass consistent with that of ZmA (calculated m/z, [M + H]⁺ = 397.2, [M + Na]⁺ = 419.2, [M + K]⁺ = 435.2; observed m/z, [M + H]⁺ = 397.0, [M + Na]⁺ = 419.0, [M + K]⁺ = 435.1). These data support the hypothesis that B. cereus AH1134 produces ZmA.

FIG. 3.

Representative HPLC traces of authentic ZmA (trace A) and metabolites purified from B. cereus AH1134 (trace B). The absorption peak associated with ZmA from the authentic sample is identified. The metabolites eluting with corresponding absorption peaks in trace B were collected and analyzed by MALDI-TOF MS. The y axis represents milliabsorbance units at 210 nm.

While the homologous genes in the two B. cereus strains show a high level of sequence identity, there is one significant difference between the two gene clusters. The B. cereus AH1134 strain lacks the five genes found between zmaS and zmaT in B. cereus UW85 (Fig. 2). In their place is a 789-bp fragment of DNA. We were unable to find homologs for these five genes in the nearly completed genome sequence for B. cereus AH1134 (GenBank accession number ABDA00000000). We were also unable to detect any of these genes by PCR amplification using primers based on the B. cereus UW85 sequence (data not shown). This strongly suggests that these five genes in B. cereus UW85 are not involved in ZmA production.

In further support of this conclusion, these genes are similar to five genes from B. subtilis that have been shown to be involved in the production of the amino sugar antibiotic 3,3′-neotrehalosadiamine (20). This antibiotic has not been detected in cultures of B. cereus UW85, but it is known that the bacterium produces the amino sugar kanosamine (Fig. 4). The relevance of this is that 3,3′-neotrehalosadiamine is a dimer of kanosamine monomers. Based on this, we propose that the five genes between zmaS and zmaT are involved in the regulation, biosynthesis, and export of kanosamine from B. cereus UW85; thus, we have used kab to denote the genes we propose to be involved in kanosamine biosynthesis (Fig. 2 and Table 2). A simple three-step pathway from UDP-glucose to kanosamine can be developed based on the sequence similarities of KabA, -B, and -C to enzymes with known functions (Fig. 4). Further analysis will be required to determine whether these genes are involved in 3,3′-neotrehalosadiamine or kanosamine biosynthesis. In any case, the lack of these genes in B. cereus AH1134 did not disrupt its ability to produce ZmA, allowing us to conclude that the kabRABCD genes are not directly involved in ZmA production in B. cereus UW85.

FIG. 4.

Proposed pathway for the biosynthesis of kanosamine.

TABLE 2.

Predicted proteins involved in kanosamine production

Protein (gene)	Basesa	Size (bp)	Mass (Da)	Putative function	Closest protein homologb
KabR (kabR)	56690-55638	1,053	39,785	Transcriptional regulator	LacI family; B. thuringiensis Al Hakam; YP_895701 (90/96)
KabA (kabA)	56847-58050	1,206	46,058	Aminotransferase	DegT/DnrJ/EryC1/StrS family; B. thuringiensis Al Hakam; YP_895700 (89/96)
KabB (kabB)	58037-58897	861	32,660	UDP-kanosamine hydrolase	HAD superfamily hydrolase; B. thuringiensis Al Hakam; YP_895699 (88/96)
KabC (kabC)	58914-59963	1,050	39,785	UDP-glucose C3 dehydrogenase	GFO_IDH_MocA oxidoreductase family; B. thuringiensis Al Hakam; YP_895699 (85/92)
KabD (kabD)	60100-61236	1,137	42,395	Kanosamine transporter	Fucose permease family; B. anthracis Ames; YP_895699 (81/90)

^aNumbering with respect to GenBank accession no. FJ430564.

^bThe homologs include the species name, the locus, and, in parentheses, the percent identity/percent similarity.

Bioinformatics and biochemical analyses of ZmA biosynthesis enzymes.

The results outlined above strongly suggested we had identified the complete ZmA biosynthesis gene cluster. Our next goal was to develop a hypothesis for how ZmA was assembled based on the proteins encoded by zmaA through zmaV. As outlined below, the most surprising finding of this analysis was that ZmA formation is likely to arise from the processing of a larger hybrid nonribosomal peptide-polyketide metabolite.

(i) Biosynthesis of ZmA precursors.

We proposed that ZmA biosynthesis would involve the condensation of five precursors (12). Two of these precursors, l-Ser and malonyl-CoA, would be readily available for ZmA biosynthesis due to their being common primary metabolites. The three remaining precursors would be unique to ZmA biosynthesis and would require specific enzymes for their formation. We had previously established that ZmaG, ZmaD, ZmaE, and ZmaN produce the precursor (2R)-hydroxymalonyl-ACP (Fig. 5A and Table 3), while the concerted actions of ZmaG, ZmaH, ZmaI, and ZmaJ form the precursor (2S)-aminomalonyl-ACP (Fig. 5B and Table 3) (4). We also proposed that l-Dap was a potential precursor for ZmA based on the amino acid being incorporated into the antibiotics viomycin and capreomycin (12, 14, 44). Consistent with this hypothesis, homologs of the enzymes proposed to generate l-Dap for viomycin (44) and capreomycin (14) biosynthesis are encoded by the ZmA biosynthesis gene cluster. These enzymes, ZmaU and ZmaV, are proposed to cooperatively catalyze the formation of l-Dap from l-Ser or O-acetyl-l-Ser and an amino group donor, such as l-ornithine (Fig. 5C; Table 3). Zhao and colleagues recently showed that deleting homologs of zmaV and zmaU from the genome of B. thuringiensis subsp. kurstaki strain YBT-1520 resulted in a strain that would not produce ZmA unless l-Dap was added to the culture medium (50). This supports our initial hypothesis that this amino acid is required for ZmA biosynthesis. However, these data did not eliminate the possibility that β-Uda, the amino acid that forms the terminal amide of ZmA, was the true precursor for incorporation into ZmA by the hybrid NRPS/PKS. The nonproteinogenic amino acid β-Uda would arise from the carbamoylation of l-Dap by ZmaT (Fig. 5C and Table 3).

FIG. 5.

Schematic of the ZmA precursor biosynthesis pathways. (A) Formation of (2R)-hydroxymalonyl-ACP. (B) Formation of (2S)-aminomalonyl-ACP. (C) Formation of l-Dap and β-Uda. Carb∼P, carbamoylphosphate; PLP, pyridoxal phosphate.

TABLE 3.

Predicted proteins involved in ZmA production

Protein (gene)	Basesa	Size (bp)	Mass (Da)	Established functionb or protein homolog
ZmaD (zmaD)	21953-22216	264	10,206	ACP involved in HM-ACP formation
ZmaE (zmaE)	22213-23361	1,149	42,061	Glyceraldehyde-S-ZmaD dehydrogenase (HM-ACP formation)
ZmaR (zmaR)	23358-24485	1,128	43,488	ZmA resistance gene (acetyltransferase)
ZmaG (zmaG)	25763-26611	849	31,624	Glycerol-S-ZmaD and Seryl-S-ZmaH dehydrogenase (HM-ACP and AM-ACP formation)
ZmaH (zmaH)	26642-26887	246	9,576	ACP involved in AM-ACP formation
ZmaI (zmaI)	26887-28080	1,194	43,692	2-Aminomalonaldehyde-S-ZmaH dehydrogenase (AM-ACP formation)
ZmaJ (zmaJ)	28077-29654	1,578	58,769	Seryl-AMP synthetase involved in AM-ACP formation
ZmaL (zmaL)	37336-38412	1,077	40,468	Alkanesulfonate monooxygenase; Pseudomonas putida GB-1; YP_001669573 (29/49)
ZmaM (zmaM)	38428-41511	3,084	116,965	Amino acids 1-350: class C β-lactamase and d-Ala/d-Ala peptidase; Bacillus sp. strain 14905; EA_Z83790. Amino acids 350-1028: putative cyclic peptide transporter; Paenibacillus larvae subsp. larvae BRL-230010; ZP_02328117 (46/66)
ZmaN (zmaN)	41508-42587	1,080	41,893	Glyceryl-S-ZmaD synthase involved in HM-ACP formation
ZmaP (zmaP)	47121-47840	720	26,848	Type II thioesterase; Bacillus licheniformis ATCC 14580; YP_077643 (41/64)
ZmaS (zmaS)	54807-55505	699	27,601	Phosphopantetheinyltransferase; B. cereus ATCC 14579; NP_832219 (42/63)
ZmaT (zmaT)	61649-62581	933	34,145	Carbamoyltransferase; Streptomyces avermitilis MA-4680; NP_824818 (59/76)
ZmaU (zmaU)	62702-63679	978	36,415	Dap synthase; P. larvae subsp. larvae BRL-230010; ZP_02328109 (61/77)
ZmaV (zmaV)	63707-64675	969	36,343	Ornithine cyclodeaminase; P. larvae subsp. larvae BRL-230010; ZP_02328108 (62/79)

^aNumbering with respect to GenBank accession no. FJ430564.

^bEstablished functions have been shown for ZmaD, -E, -G, -H,- I, -J, and -N (4). The protein homologs include the species name, the locus, and, in parentheses, the percent identity/percent similarity.

(ii) PKSs involved in ZmA biosynthesis.

Based on the structure of ZmA, we previously hypothesized that three PKS modules are involved in its biosynthesis (12). These modules would incorporate the precursors malonyl-CoA, (2S)-aminomalonyl-ACP, and (2R)-hydroxymalonyl-ACP. Analysis of the proteins encoded by the ZmA gene cluster identified the three expected PKS modules contained on ZmaA, ZmaF, and ZmaK (Table 3). The N terminus of ZmaK contains an NRPS module, while the C terminus contains a complete PKS module. Based on the structure of ZmA, it was reasonable to propose that the PKS module of ZmaK was involved in malonyl-CoA incorporation due to the N-terminal NRPS module. The presence of a ketoreductase domain within this PKS module is consistent with the formation of the C-13 hydroxyl group of ZmA (Fig. 1). ZmaA contains two PKS modules with a C-terminal condensation (C) domain of an NRPS module. Based on the structure of ZmA, ZmaA would function downstream of ZmaK to incorporate (2S)-aminomalonyl-ACP and (2R)-hydroxymalonyl-ACP. The first PKS module is an “acyltransferase (AT)-less” PKS module because it lacks an AT domain to incorporate the associated precursor. When AT-less PKS modules are observed, the AT domain is commonly encoded by another gene within the biosynthesis gene cluster (6, 45). Consistent with this observation, an AT domain homolog is encoded by zmaF, and this enzyme is the “missing” AT domain of the first PKS module of ZmaA. The second PKS module of ZmaA would incorporate the precursor (2R)-hydroxymalonyl-ACP. Both of the PKS modules of ZmaA contain ketoreductase domains to generate the hydroxyl groups on C-11 and C-9 (Fig. 1). Biochemical evidence that supports the proposed roles of ZmaA and ZmaF in ZmA biosynthesis will be presented elsewhere. The combination of ZmaF, ZmaA, and the C terminus of ZmaK account for all the PKS enzymology needed for ZmA biosynthesis.

(iii) NRPSs involved in ZmA biosynthesis.

The structure of ZmA suggests that two NRPS modules are required to incorporate the amino acids on either side of the central polyketide core. Consistent with this hypothesis, the N terminus of ZmaK contains an NRPS module and the C terminus of ZmaA contains a C domain of NRPSs. NRPS modules contain the necessary catalytic domains to recognize and incorporate an amino acid into a natural-product backbone, while a C domain is the portion of an NRPS module that catalyzes amide bond formation between two amino acids tethered to neighboring PCP domains (reviewed in reference 13). These observations suggest that an amino acid will be introduced before and after the PKS components discussed above (Table 4). The unusual aspect of the NRPS module of ZmaK is that it contains a C domain at the N terminus, suggesting that an amino acid is condensed to the first amino acid of ZmA. Additionally, when we analyzed the remaining ORFs in the ZmA gene cluster, we identified not just one NRPS module to incorporate the final amino acid, but four additional NRPS modules contained on ZmaB, ZmaC, ZmaO, and ZmaQ (Table 4). These observations suggested that ZmA biosynthesis is more complicated than is implied by its chemical structure.

TABLE 4.

NRPS and PKS components encoded by the biosynthesis gene cluster

Protein (gene)	Basesa	Size (bp)	Mass (Da)	Proposed domain organization (aa residues)a
ZmaA (zmaA)	3190-13200	10,011	377,827	KS (13-614)-KR (1005-1183)-ACP (1288-1352)-KS (1373-1799)-AT (1897-2226)-KR (2582-2690)-ACP (2811-2875)-C (2894-3179)
ZmaB (zmaB)	13227-18950	5,724	218,676	A (269-709)-PCP (796-858)-C (9875-1165)-A (1351-1752)-PCP (1839-1868)
ZmaC (zmaC)	19269-21920	2,652	102,337	Pr (1-354)-PCP (355-419)-C (435-717)
ZmaF (zmaF)	24505-25722	1,218	45,579	AT
ZmaK (zmaK)	29629-37317	7,689	291,152	C (12-291)-A (469-869)-PCP (949-1012)-KS (1036-1456)-AT (1551-1822)-KR (2170-2371)-ACP (2465-2529)
ZmaO (zmaO)	42569-47065	4,497	172,467	C (12-291)-A (483-880)-PCP (986-1029)-E (1049-1345)
ZmaQ (zmaQ)	48084-54620	6,537	249,755	A (253-783)-PCP (789-844)-C (870-1292)-A (1296-1826)-PCP (1832-1896)- Te (1918-2167)

^aAmino acid (aa) numbering with respect to GenBank accession no. FJ430564. KS, ketosynthase; KR, ketoreductase; Pr, C39A protease.

To begin to decipher how ZmA would be assembled from the enzymes coded by the ZmA biosynthesis gene cluster, we analyzed the substrate specificity of each A domain from the various NRPS components. A domains are viewed as the “gatekeepers” of NRPS modules because they select the amino acid to be incorporated by that specific module, activate the amino acid as an aminoacyl-AMP intermediate, and subsequently tether the amino acid onto the associated PCP domain (8, 39). By determining the amino acids activated by all the A domains of an NRPS, a reasonable proposal can be made for how the associated metabolite is biosynthesized. Each of the A domains from ZmaB, ZmaK, ZmaO, and ZmaQ was heterologously produced in E. coli, purified using nickel-chelate chromatography (Fig. 6A), and analyzed for its amino acid specificity (Fig. 6B to G). All of the A domains proved to be functional based on their abilities to activate a specific amino acid using standard (d)ATP/PP_i exchange assays (30, 43). We tested each A domain with all 20 standard proteinogenic amino acids, along with l-Dap, d-Ser, and β-Uda.

FIG. 6.

(A) Analysis of purified proteins using 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis stained with Coomassie blue. (B to G) Representative amino acid-dependent dATP-³²PP_i exchange reactions using purified proteins and amino acids.

ZmaB has two A domains, requiring that each A domain be characterized separately. We refer to them here as ZmaB-A1 and ZmaB-A2 to reflect the first and second A domains of the protein, respectively. ZmaB-A1 was produced with a 248-amino-acid domain at the N terminus of ZmaB that showed sequence similarity with the C-terminal halves of C domains from other NRPS systems. ZmaB-A2 was overproduced as a three-domain construct, C-A2-PCP, since it was the only construct containing ZmaB-A2 that resulted in soluble protein when overproduced in E. coli. ZmaB-A1 failed to activate any of the 20 proteinogenic amino acids, d-Ser, or l-Dap. Instead, the enzyme was found to activate β-Uda (Fig. 6B). ZmaB-A2 specifically activated l-Ala (Fig. 6C). The remaining A domains were all found to be soluble and active, and each was selective for one of the 20 proteinogenic amino acids based on dATP/PP_i exchange assays. The ZmaK A domain activated l-Ser, while the A domain of ZmaO activated l-Asn (Fig. 6D and E). The finding that the A domain of ZmaK was specific for l-Ser and did not activate d-Ser suggests that l-Ser is incorporated by the NRPS and the conversion of the stereochemistry to the d isomer must occur at a point after amino acid activation. ZmaQ contained two A domains, and each was overproduced and purified independently. ZmaQ-A1 activated l-Leu, while ZmaQ-A2 activated l-Met (Fig. 6F and G).

We note that our data on ZmaQ are not consistent with data recently published by Zhao and colleagues (50). In their study, Zhao and colleagues analyzed the second A domain of a homolog of ZmaQ and presented data showing that it activated l-Dap, but at only slightly higher levels than l-Ser, l-Cys, and l-Ala. They did not test any other amino acids. Based on their results, they proposed that this A domain is the one that incorporates l-Dap into the ZmA structure. However, it is not clear what role the first module of ZmaQ would then play. We took the approach that there were clearly too many A domains for the incorporation of only l-Ser and l-Dap into ZmA; thus, we needed to test activation with more amino acids than just those seen in the ZmA structure. The second A domain of ZmaQ clearly preferred l-Met over l-Dap (Fig. 6G). In fact, we failed to detect any A domain that preferred l-Dap over other amino acids. However, we did find that ZmaB-A1 is specific for β-Uda, which is carbamoylated l-Dap (Fig. 6B). Therefore, from our analysis of the NRPS components of the ZmA biosynthesis pathway, l-Dap is not a precursor used by the NRPS components for ZmA biosynthesis. Instead, l-Dap is carbamoylated first by ZmaT to generate β-Uda (Fig. 2), and this is the amino acid incorporated by the NRPS.

Finally, ZmaC is an NRPS component with unusual domain architecture. The protein has three domains, with the second and third domains showing sequence similarity to PCP and C domains, respectively. The N-terminal domain of ZmaC is homologous to the C39A subfamily of C39 cysteine peptidases. This subfamily of peptidases is commonly involved in the processing of bacteriocin antibiotics by cleaving the leader peptide from the antibiotics as they are exported out of the cell (2, 10). While the N-terminal domain of ZmaC is homologous to these types of peptidases, the domain does not contain all of the essential catalytic residues. ZmaC contains the conserved His residue that is part of the catalytic diad and the Gln residue that contributes to the oxyanion hole that stabilizes the enzyme intermediate. However, ZmaC lacks the catalytic Cys residue that functions as the attacking nucleophile and site of the covalently linked acyl intermediate. This suggests that the domain is not functioning as a protease but rather plays some other role in ZmA biosynthesis. A potential role for the domain is discussed in more detail below.

Proposal for the biosynthesis of ZmA and two additional metabolites.

The bioinformatics and biochemistry discussed above provided a summary of all the likely catalytic steps that are involved in ZmA assembly. The next issue was to generate a scheme for how all of these enzymes work coordinately to generate the antibiotic.

(i) Functional order of the PKS and NRPS components.

As discussed above, the order in which the PKS components function during ZmA biosynthesis is straightforward, with the C terminus of ZmaK incorporating malonyl-CoA and the concerted actions of ZmaF and ZmaA incorporating (2S)-aminomalonyl-ACP and (2R)-hydroxymalonyl-ACP (Fig. 7A). The next issue was to determine how the NRPS components are coordinated around the PKS components. With the identity of the amino acids activated by each of the A domains defined by biochemical analysis, it is possible to place ZmaB immediately downstream of ZmaA. This is based on the finding that the first A domain of ZmaB activates β-Uda, the amino acid tethered to the glyoxyl moiety incorporated by the C-terminal PKS module of ZmaA (Fig. 7A). ZmaQ can be placed as the final component of the megasynthase because it contains a C-terminal Te domain. These domains are well characterized and catalyze release of the nonribosomal peptide or polyketide from the megasynthase as the final step in synthesis (24).

FIG. 7.

Proposed biosynthesis scheme for the production of ZmA and two additional metabolites (metabolites A and B). The solid bars at the top of panel A identify the 10 different NRPS or PKS modules (M1 to M10). Modules M1 to M7 are involved in the incorporation of l-Asn, l-Ser, malonate, (2S)-aminomalonate, (2R)-hydroxymalonate, β-Uda, and l-Ala, respectively, while modules M9 and M10 incorporate l-Ile and l-Met, respectively. Each circle represents a catalytic domain of the NRPS or PKS component. C, condensation; A, adenylation; E, epimerization; KS, ketosynthase; KR, ketoreductase; Pr, C39A protease. The NRPS and PKS modules and the precursors they incorporate have been color coded to reflect the metabolite(s) they are involved in producing (green, metabolite A; red, ZmA; blue, metabolite B). The enzymes ZmaL and ZmaM are represented by orange and purple ovals, respectively, and are located at the sites of their proposed functions.

With the functional locations of ZmaB and ZmaQ set, the next issue was to determine where ZmaO and ZmaC function. There were a few reasons to place ZmaO upstream of ZmaK and ZmaC between ZmaB and ZmaQ (Fig. 7A). First, placement of ZmaO between ZmaB and ZmaQ was unlikely because there would be no domain to catalyze the condensation of the amino acids tethered to the PCP domains of ZmaO and ZmaQ; thus, placement of ZmaO prior to ZmaK was more likely. Second, the C domain of ZmaO was most similar to type III C domains that are found at the N terminus of NRPSs. Miao and colleagues have divided C domains from NRPSs into three types (27). The type I domains condense l amino acids, the type II C domains condense d amino acids, and the type III domains are at the N terminus of NRPSs that generate lipopeptides. We aligned all of the C domains from the ZmA system with those from bacitracin (23), surfactin (9), gramicidin (25), fengycin (26), lichenysin A (48), and lichenysin D (22). From this analysis, the C domain of ZmaO was most similar to the type III C domains from the NRPSs that make the lipopeptides surfactin, fengycin, lichenysin A, and lichenysin D (see Fig. S1 in the supplemental material). Importantly, all type III C domains identified to date initiate nonribosomal-peptide synthesis by tethering a fatty acid to the α-amino group of the first amino acid of a lipopeptide. This strongly suggests the N terminus of ZmA initially contained a fatty acid, along with d-Asn. Finally, Chiocchini and colleagues and Hahn and Stachelhaus have shown that E domains and C domains that interact with each other contain communication domains that function as sites of specific protein-protein interactions between E and C domains on different NRPS polypeptides (7, 15, 16). Analysis of the C terminus of the E domain of ZmaO and the N terminus of ZmaK identified sequences consistent with communication domains (data not shown). This supports our contention that ZmaO functions immediately prior to ZmaK. Based on these analyses, ZmaC is proposed to function between ZmaB and ZmaQ to complete the NRPS/PKS megasynthase organization (Fig. 7A).

(ii) ZmA biosynthesis and processing.

The analysis of the NRPS/PKS megasynthase strongly suggested that ZmA was biosynthesized as part of a larger molecule that was processed during its biosynthesis. Based on our analysis of the NRPS/PKS components and the other enzymes encoded by the biosynthesis gene cluster, the following biosynthesis scheme is proposed. ZmaO activates l-Asn, condenses it with a fatty acid, and epimerizes the l-Asn to d-Asn to form the fatty acyl-d-Asn-S-ZmaO intermediate (Fig. 7B). The next five NRPS/PKS modules form the backbone of ZmA by condensing the fatty acyl-d-Asn intermediate with l-Ser, malonylate, (2S)-aminomalonate, (2R)-hydroxymalonylate, and β-Uda (Fig. 7B). It is expected that at some point during the condensation of these precursors, the stereochemistry of the l-Ser residue is changed to d-Ser to account for the stereochemistry in the final product. It is not yet clear how this occurs, because an epimerase (E) domain is not found in the NRPS module that incorporates l-Ser. However, the module immediately upstream of the l-Ser-associated module contains an E domain. One possibility is that this domain catalyzes the epimerization of l-Ser, as well as l-Asn. Another possibility is that the C domain associated with l-Ser is involved in this conversion. This C domain does not cluster with type II C domains as would be expected of a C domain associated with an upstream E domain (see Fig. S1 in the supplemental material). Potentially this C domain has an additional epimerase function analogous to that observed with bifunctional C/E domains (1). Further analysis of ZmA biosynthesis will be required to differentiate between these hypotheses.

The finding of an additional NRPS module on ZmaB suggests the growing fatty acyl-peptide/polyketide is extended by an additional amino acid, l-Ala, to form a covalently linked intermediate on the second PCP domain of ZmaB (Fig. 7C). At this stage, the backbone of ZmA has a fatty acyl-d-Asn at its N terminus and l-Ala at its C terminus, both of which must be removed to make the antibiotic. Insights into the processing at the C terminus come from the work of Müller and colleagues on the formation of the terminal amide of the natural product myxothiazole, as discussed above. Using the same logic, the l-Ala introduced onto the C terminus of the ZmA backbone would be hydroxylated to release a ZmA derivative containing a terminal amide, leaving a pyruvyl-S-PCP intermediate tethered to the second PCP domain of ZmaB (Fig. 7D). The enzyme that would catalyze this hydroxylation is ZmaL, a protein that shows sequence similarity to flavin-dependent monooxygenases (Table 3).

There are two issues for ZmA that still need to be addressed. First, the pyruvyl moiety must be removed from the NRPS to allow additional turnovers of the NRPS/PKS megasynthase. Second, the N-terminal fatty acyl-d-Asn moiety must be removed from the N terminus of ZmA. The removal of the pyruvyl moiety is likely to be accomplished by the continued transfer of the PCP-tethered intermediate down the NRPS/PKS megasynthase. The first step in this process is the transfer of the pyruvyl moiety from ZmaB to the PCP domain of ZmaC by the N-terminal C39A peptidase domain of ZmaC. As stated previously, this domain lacks the active-site cysteine but retains the histidine residue for deprotonation of a thiol and the glutamine residue to stabilize the oxyanion intermediate for this class of cysteine proteases. We propose that the thiol from the PCP domain of ZmaC replaces the active-site cysteine thiol, resulting in the domain catalyzing a thiol exchange of the pyruvyl moiety from the PCP domain of ZmaB to the PCP of ZmaC (Fig. 7D). This is then condensed with the l-Leu and l-Met residues tethered to ZmaQ (Fig. 7E), with the Te domain of ZmaQ releasing “metabolite B” from the megasynthase (Fig. 7F). It is not clear at this time what role this metabolite might play in B. cereus physiology.

The fatty acyl-d-Asn-ZmA intermediate released from the NRPS/PKS megasynthase by the action of the ZmaL-catalyzed hydroxylation must also be processed to generate ZmA. The only candidate enzyme encoded by the biosynthesis gene cluster to catalyze such a peptidase activity is ZmaM. The C terminus of ZmaM is homologous to ABC-type transporters involved in metabolite efflux, while the N terminus is homologous to β-lactamases and d-alanyl-d-alanine carboxypeptidases (Table 3). Thus, we propose that as the acyl-d-Asn-ZmA is exported out of the cell through ZmaM, the N terminus of this transporter cleaves the metabolite between the d-Asn and d-Ser, releasing ZmA and the fatty acyl-d-Asn (Fig. 7F, metabolite A). We cannot eliminate the possibility that some other endogenous peptidase catalyzes this cleavage. We are currently analyzing the culture supernatants of B. cereus UW85 for the presence of the predicted metabolites A and B.

The three remaining enzymes encoded by the ZmA biosynthesis gene cluster have functions in resistance or are likely to be involved in forming functional NRPS/PKS components and keeping these components functional. ZmaR has been shown to be the ZmA resistance enzyme that catalyzes the N-acetylation of the α-amino group of the d-Ser moiety of ZmA (41). ZmaS is a homolog of phosphopantethenyltransferases that catalyze the conversion of carrier proteins (ACPs and PCPs) from their apoforms to holoforms by transferring the 4′-phosphopantetheinyl moiety of CoA to a conserved serine of the carrier proteins (47). ZmaP is a homolog of type II Tes that are involved in proofreading the NRPS/PKS megasynthases to ensure stalled intermediates are removed (19, 21, 34). It is reasonable to propose that ZmaP performs similar proofreading functions during ZmA biosynthesis.

Conclusions.

We have presented genetic, bioinformatics, and biochemical evidence in support of our hypothesis that we have identified the ZmA biosynthesis gene cluster in B. cereus UW85. We have also provided evidence that B. cereus AH1134 produces ZmA and have identified the putative biosynthesis gene cluster in the bacterium. We have provided molecular evidence that ZmA is biosynthesized in an unusual manner that involves processing of both its N and C termini, potentially resulting in the production of two additional metabolites besides ZmA (Fig. 7). Finally, we propose that the gene cluster identified in B. cereus UW85 also codes for enzymes involved in the biosynthesis of the antibiotic kanosamine.