Biosynthesis of 2′-Chloropentostatin and 2′-Amino-2′-Deoxyadenosine Highlights a Single Gene Cluster Responsible for Two Independent Pathways in Actinomadura sp. Strain ATCC 39365

ABSTRACT

2′-Chloropentostatin (2′-Cl PTN, 2′-chloro-2′-deoxycoformycin) and 2′-amino-2′-deoxyadenosine (2′-amino dA) are two adenosine-derived nucleoside antibiotics coproduced by Actinomadura sp. strain ATCC 39365. 2′-Cl PTN is a potent adenosine deaminase (ADA) inhibitor featuring an intriguing 1,3-diazepine ring, as well as a chlorination at C-2′ of ribose, and 2′-amino dA is an adenosine analog showing bioactivity against RNA-type virus infection. However, the biosynthetic logic of them has remained poorly understood. Here, we report the identification of a single gene cluster (ada) essential for the biosynthesis of 2′-Cl PTN and 2′-amino dA. Further systematic genetic investigations suggest that 2′-Cl PTN and 2′-amino dA are biosynthesized by independent pathways. Moreover, we provide evidence that a predicted cation/H⁺ antiporter, AdaE, is involved in the chlorination step during 2′-Cl PTN biosynthesis. Notably, we demonstrate that 2′-amino dA biosynthesis is initiated by a Nudix hydrolase, AdaJ, catalyzing the hydrolysis of ATP. Finally, we reveal that the host ADA (designated ADA1), capable of converting adenosine/2′-amino dA to inosine/2′-amino dI, is not very sensitive to the powerful ADA inhibitor pentostatin. These findings provide a basis for the further rational pathway engineering of 2′-Cl PTN and 2′-amino dA production.

IMPORTANCE 2′-Cl PTN/PTN and 2′-amino dA have captivated the great interests of scientists, owing to their unusual chemical structures and remarkable bioactivities. However, the precise logic for their biosynthesis has been elusive for decades. Actually, the identification and elucidation of their biosynthetic pathways not only enrich the biochemical repertoire of novel enzymatic reactions but may also lay solid foundations for the pathway engineering and combinatorial biosynthesis of this family of purine nucleoside antibiotics to generate novel hybrid analogs with improved features.

INTRODUCTION

Nucleoside antibiotics are a large family of important microbial natural products harboring wide-range biological properties and distinctive structural features (1–3). Their biosynthesis generally follows a succinct logic by sequential enzymatic modification of nucleosides or nucleotides originating from primary metabolisms (1). Usually, nucleosides and nucleotides play pleiotropic roles in most fundamental cellular metabolisms; therefore, nucleoside antibiotics are able to target the biosynthesis of diverse biomacromolecules, including nucleic acids, proteins, and glycans (1).

2′-Chloropentostatin (2′-Cl PTN, 2′-chloro-2′-deoxycoformycin) and 2′-amino-2′-deoxyadenosine (2′-amino dA) (Fig. 1A) are purine-derived nucleoside antibiotics concomitantly produced by Actinomadura sp. strain ATCC 39365 (4), and yet they were reported to be individually produced as well by other actinomycetes strains (5, 6). Remarkably, the coproduction phenomenon of the antibiotics 2′-Cl PTN and 2′-amino dA has also been reported in past decades for other antibiotic pairs, including pentostatin (PTN)-arabinofuranosyladenine (Ara-A) (7) and coformycin-formycin (Fig. 1A) (8). Mechanistically, 2′-Cl PTN mimics the transition-state intermediate of the adenosine deaminase (ADA)-catalyzed reaction, and thus, it is a powerful irreversible ADA inhibitor, with a K_i of 1.1 × 10⁻¹¹ M (9). 2′-Cl PTN represents a new group of nucleoside antibiotics with utility in the treatment of hematological cancers (4); however, it is not active against the indicator fungi and bacteria in the presence of the test concentration of 1 mg/ml (6). Moreover, 2′-Cl PTN is able to greatly potentiate the antiviral efficacy of Ara-A, and the acute toxicity of 2′-Cl PTN in mice was less than that of coformycin and PTN (6). Regarding 2′-amino dA, it has been shown to be a selectively effective inhibitor against the replication of some riboviruses, including the measles virus (10, 11). Very interestingly, two other related adenosine analogs designated 2′-amino-2′-deoxyguanosine (2′-amino dG) (12) and 3′-amino-3′-deoxyadenosine (3′-amino dA) (13) have been discovered prior to 2′-amino dA (Fig. 1B). 2′-Amino dG demonstrates antibacterial activity against Escherichia coli and antitumor activity against sarcoma cells (12), and 3′-amino dA shows significant antitumor activity against ascitic tumors of mice (13).

FIG 1.

Chemical structures of relevant antibiotics. (A) Chemical structures of related antibiotic pairs. The upper and lower structures constitute antibiotic pairs which are coproduced by a particular strain. The 2′-Cl pentostatin and 2′-amino-2′-deoxyadenosine (2′-amino dA) pair is concomitantly produced by Actinomadura sp. ATCC 39365, the PTN and Ara-A pair is produced by S. antibioticus NRRL 3238, and the coformycin and formycin pair is produced by Streptomyces kaniharaensis ATCC 21070 or by Nocardia interforma ATCC 21072. (B) Chemical structures of 2′-amino dI, 3′-amino dA, and 2′-amino dG. 2′-Amino dI, 2′-amino-2′-deoxyinosine; 2′-amino dG, 2′-amino-2′-deoxyguanosine; 3′-amino dA, 3′-amino-3′-deoxyinosine.

Structurally, 2′-Cl PTN has a 1,3-diazepine ring identical to that of other related antibiotics, including PTN and coformycin, but contains an unusual chlorination at the C-2′ position (4), and 2′-amino dA is an adenosine analog featuring a C-2′ amino group substituted for the corresponding hydroxyl group of adenosine (4). Previous metabolic labeling experiments indicated that adenosine acts as the direct precursor for the biosynthesis of both antibiotics (4). In addition, the C-7 origin of the “fat” 1,3-diazepine ring comes from C-1 of d-ribose by inserting the N-1 and C-6 of the intact purine ring (7). In our recent report, we revealed that a single gene cluster is involved in two independent pathways during PTN and Ara-A biosynthesis, and we further demonstrated that their biosynthesis employs a protector-protégé strategy, with PTN capable of protecting Ara-A from deamination by the host adenosine deaminase (14). Notably, 2′-Cl PTN has been reported as the first naturally occurring nucleoside natural product bearing the chloro group (15). As for the biosynthetic origin of the chlorination, metabolic labeling experiments indicated that the inorganic chloride in the medium could be utilized by Actinomadura sp. ATCC 39365 serving as the chloro group in 2′-Cl PTN (4); however, the precise mechanism on how the chlorination occurs during the 2′-Cl PTN biosynthesis has remained elusive for decades.

In the present report, we reveal that 2′-Cl PTN and 2′-amino dA biosynthesis exploits a “a single gene cluster encoding two independent pathways” strategy, and we further demonstrate that the host adenosine deaminase ADA1 contributes to the deamination of 2′-amino dA to 2′-amino deoxyinosine (2′-amino dI) (Fig. 1B). Moreover, we illustrate that a Nudix hydrolase (AdaJ) governs the initial step of 2′-amino dA biosynthesis. Deciphering of the biosynthetic puzzles for 2′-Cl PTN and 2′-amino dA lays a solid foundation for the rational generation of designer 2′-Cl PTN and 2′-amino dA analogs via synthetic biology strategies, and it undoubtedly expands the chemical diversities concerning the biosynthesis of nucleoside natural products.

RESULTS

Reinvestigation of the target nucleoside metabolites produced by Actinomadura sp. ATCC 39365.

Actinomadura sp. ATCC 39365 has been previously characterized as a 2′-Cl PTN and 2′-amino dA producer (4), but this strain has never been reported to produce other related nucleoside antibiotics, including PTN and 2′-amino dI, which are most likely to be biosynthesized by this strain due to the fact that 2′-Cl PTN potentially originates from PTN by direct chlorination, and 2′-amino dA is prone to deamination by host ADA to 2′-amino dI. To address this question, we reinvestigated the metabolite profiles of the sample of Actinomadura sp. ATCC 39365 by liquid chromatography-mass spectrometry (LC-MS) analysis, and the results indicated that it could generate the obvious characteristic [M+H]⁺ ion at m/z 268.1021 and fragment ions at m/z 136.9485 and 251.0468, in full agreement with those of the 2′-amino dI authentic standard (see Fig. S1A to C in the supplemental material); also, the sample could produce a distinctive [M+H]⁺ ion at m/z 269.1224 as well as fragment series m/z 152.9804, 134.8981, and 251.0840, which correspond well with those of the PTN authentic standard (Fig. 2B and S2A and B). Moreover, as anticipated, LC-MS results showed that the sample of Actinomadura sp. ATCC 39365 could give apparent distinctive [M+H]⁺ ions of 2′-amino dA and 2′-Cl PTN (Fig. 2B and C and S2C to E), confirming the identity of the strain as a 2′-amino dA and 2′-Cl PTN producer, as previously documented. These data unquestionably demonstrated that Actinomadura sp. ATCC 39365 is also a 2′-amino dI and PTN producer.

FIG 2.

Genetic organization and validation of the 2′-Cl PTN and 2′-amino dA biosynthetic gene cluster (ada). (A) Genetic organization of the ada gene cluster. The insertion region of the positive cosmid 3G12 is indicated on the top of the ada gene cluster, and the conserved genes in the shaded region are used for the identification of the ada gene cluster. (B) LC-MS analysis of the target metabolites produced by Actinomadura sp. LG1 mutant. ST, authentic PTN and 2′-amino dA standards; WT, wild type of ATCC 39365. (C) MS analysis of the 2′-Cl PTN ion generated by the sample of Actinomadura sp. ATCC 39365.

Identification of 2′-Cl PTN and 2′-amino dA biosynthetic gene cluster.

For identifying the target gene cluster responsible for 2′-Cl PTN and 2′-amino dA biosynthesis, the genome of Actinomadura sp. ATCC 39365 was sequenced by the Illumina method, which rendered ca. 11.2 Mb of nonredundant bases after assembly of clean reads. The genome data of Actinomadura sp. ATCC 39365 were subsequently annotated by the Glimmer 3.0 software, yielding 10,533 valid open reading frames (ORFs). As 2′-Cl PTN and PTN share the 1,3-diazepine scaffold, implicating that they harbor the same biosynthetic logic, we thus use the three key enzymes, PenA (ATP phosphoribosyltransferase), PenB (short-chain dehydrogenase), and PenC (5-amino-4-imidazole N-succinocarboxamide ribonucleotide-5-phosphate [SAICAR] synthetase) (14), in the PTN biosynthetic pathway as probes to conduct individual BLASTP analysis against the genome of Actinomadura sp. ATCC 39365, which leads to the location of the only one candidate gene cluster encoding enzymes involving ORF02747 (designated AdaC, 58% identity to PenA), ORF02756 (AdaL, 53% identity to PenA), ORF02748 (AdaB, 70% identity to PenB), and ORF02749 (AdaA, 72% identity to PenC) (Table 1). This bioinformatics analysis result strongly suggests that the target gene cluster is most likely involved in 2′-Cl PTN and PTN biosynthesis in Actinomadura sp. ATCC 39365 (Fig. 2A).

TABLE 1.

Deduced functions of the open reading frames in the ada gene cluster

Protein	No. of amino acids	Proposed function	Homolog, origin	Identity, similarity (%)	Accession no.
Orf-1	1,061	Hypothetical protein	BN2537_3105, Streptomyces venezuelae ATCC 15439	34, 46	CUM37070
AdaE	479	Cation/H+ antiporter	ADL15_15700, Actinoplanes wajinensis NRRL B-16712	54, 67	KUL34520
AdaD	402	MFS transportera	SAMN05421874_102571, Nonomuraea maritima	72, 82	SDJ65310
AdaC	295	ATP phosphoribosyl-transferase	PenA, Streptomyces antibioticus NRRL 3238	58, 71	AKA87340
AdaB	234	Short-chain dehydrogenase	PenB, S. antibioticus NRRL 3238	70, 82	AKA87339
AdaA	239	SAICAR synthetase	PenC, S. antibioticus NRRL 3238	72, 81	AKA87338
AdaF	425	Aminotransferaase	COCOR_00673, Corallococcus coralloides DSM 2259	53, 70	AFE09038
AdaG	351	Dehydrogenase	PCL1391_5850, Pseudomonas chlororaphis subsp. piscium	28, 46	KZO46392
AdaH	595	ABC transporter, partial	SAMN05421811_12174, Nonomuraea wenchangensis	46, 63	SEU43234
AdaI	592	ABC transporter	Trad_2349, Truepera radiovictrix DSM 17093	49, 68	ADI15458
AdaJ	161	Nudix hydrolase	Caci_7624, Catenulispora acidiphila DSM 44928	54, 66	ACU76448
AdaK	257	Phosphoribosyl isomerase A	HMPREF1486_01538, Streptomyces sp. HPH0547	75, 85	EPD96008
AdaL	288	ATP phosphoribosyl-transferase	ADK55_17455, Streptomyces sp. WM4235	54, 68	KOU52239
AdaM	264	Hydrolase	ADK55_17445, Streptomyces sp. WM4235	62, 71	KOU52238
Orf1	339	ABC transporter substrate-binding protein	BCD48_37375, Frankia sp. BMG5.36	55, 68	OHV64536

^aMFS, major facilitator superfamily.

Further examination of the surrounding region of adaABC results in the revealing of the genes coding for an aminotransferase (AdaF) and a dehydrogenase (AdaG), which are fully consistent with the predicted enzymes required for 2′-amino dA biosynthesis. According to the in silico analysis, adaFGHIJ constitute a transcriptional unit (AdaJ is an annotated Nudix hydrolase), and downstream are the three genes individually encoding a phosphoribosyl isomerase A (AdaK, HisA homolog), an ATP phosphoribosyltransferase (AdaL, HisG homolog), and a hydrolase (AdaM) (Fig. 2A, Table 1). To see if the genes (adaFGHIJ) are needed for 2′-Cl PTN/2′-amino dA biosynthesis, the target region covering adaJKLM was deleted to give the Actinomadura sp. LG1 mutant strain, as confirmed by PCR (Fig. S2F and G), and the mutant (LG1) was then inoculated for further analysis of the metabolites. As expected, the LC-MS results indicated that the sample of the strain LG1 could not generate the characteristic peaks of 2′-Cl PTN, PTN, and 2′-amino dA, which are present in the sample of wild-type strain (Fig. 2B). All of the data demonstrate that the target gene cluster is simultaneously responsible for the biosynthesis of 2′-Cl PTN, PTN, and 2′-amino dA in Actinomadura sp. ATCC 39365.

Engineered production of 2′-Cl PTN and 2′-amino dA in a heterologous host.

To achieve engineered production of 2′-Cl PTN and 2′-amino dA in a heterologous host, a positive cosmid, 3G12, housing the probable complete ada gene cluster, was screened from the genomic library of Actinomadura sp. ATCC 39365 by a narrow-down PCR strategy (16) and introduced into the heterologous host Streptomyces aureochromogenes CXR14 (17). Subsequently, the recombinant strain (CXR14::3G12), confirmed by PCR, was fermented for further metabolic analysis, and the LC-MS results indicated that the targeted [M+H]⁺ ions of 2′-Cl PTN (m/z 303.0825), PTN (m/z 269.1220), and 2′-amino dA (m/z 267.1175) could be clearly detected from the sample of the CXR14::3G12 recombinant. In addition, tandem MS (MS/MS) analysis of the target peaks indicated that their individual major fragment ions are fully consistent with the fragmentation patterns of those antibiotics produced by Actinomadura sp. ATCC 39365 (Fig. S3A to C). However, we could not detect the [M+H]⁺ ions of 2′-Cl PTN, PTN, and 2′-amino dA from the sample of the strain without the target gene cluster (CXR14::pJTU2463b, negative control) (Fig. 3). All of the data have demonstrated that the CXR14::3G12 recombinant is conferred with the ability to produce the antibiotics 2′-Cl PTN, PTN, and 2′-amino dA, and it was determined that the cosmid 3G12 contains the complete gene cluster essential for the biosynthesis of 2′-Cl PTN and 2′-amino dA.

FIG 3.

Genetic investigation of the 2′-Cl PTN and 2′-amino dA biosynthetic pathways. Extract ion chromatography (EIC) analysis of the metabolites produced by S. aureochromogenes CXR14::3G12 and its variants was performed. ΔadaE refers to the sample from the S. aureochromogenes CXR14 containing 3G12/ΔadaE, in which adaE was in-frame deleted via a PCR-targeting strategy; likewise, other related samples are correspondingly assigned. ST, authentic PTN and 2′-amino dA standards; WT, wild-type strain of ATCC 39365; 3G12, S. aureochromogenes CXR14 containing cosmid 3G12; 2463b, S. aureochromogenes CXR14 containing pJTU2463b as a negative control.

Minimal 13-gene cluster is essential for 2′-Cl PTN and 2′-amino dA biosynthesis.

To further define the minimal gene cluster for the biosynthesis of 2′-Cl PTN and 2′-amino dA, we determined the inserted foreign fragment of 3G12 cosmid via terminal sequencing (Fig. 2A). On the basis of bioinformatic analysis, adaABCDE comprises a transcription unit, and orf−1 in 3G12 is incomplete, suggesting that this gene is the left boundary of the ada gene cluster (Fig. 2A). We then investigated the downstream component of the ada gene cluster, and adaLM compose a transcriptional unit, implicating that orf+1 is likely unrelated to 2′-Cl PTN and 2′-amino dA biosynthesis. To validate the assumption, we then mutated this gene directly on 3G12 cosmid and introduced the resultant 3G12 derivative into S. aureochromogenes CXR14 to see if it is necessary for 2′-Cl PTN and 2′-amino dA biosynthesis. The LC-MS results showed that the sample of the mutant (CXR14::3G12/Δorf1) is capable of generating the apparent 2′-Cl PTN, PTN, and 2′-amino dA [M+H]⁺ ions, confirming the role of this gene being unrelated to the biosynthesis of the target antibiotics (Fig. 3). These combined data unambiguously established that the 13 genes (adaA–M) spanning a ca. 14.4-kb region constitute the minimal ada gene cluster for 2′-Cl PTN and 2′-amino dA biosynthesis (Fig. 2A and 3).

Mutational analysis of the ada gene cluster reveals that 2′-Cl PTN and 2′-amino dA utilize independent biosynthetic pathways.

For systematic insight into the genetic roles of the ada genes during 2′-Cl PTN and 2′-amino dA biosynthesis, we directly conducted in-frame deletions of all the target individual genes on 3G12 via PCR-targeting technology (18), and the resulting 3G12 derivatives, verified by PCR, were independently conjugated into S. aureochromogenes CXR14 (Fig. S4). After fermentation for 6 days, the samples of the related recombinants were subjected to LC-MS analysis. The results indicated that mutation of adaA, adaB, adaC, or adaL completely abolishes 2′-Cl PTN, PTN, and 2′-amino dA production (Fig. 3), demonstrating their essential functional roles for biosynthesis of the three antibiotics; also, mutation of the structural genes adaF, adaG, or adaJ results in the nonproduction of 2′-amino dA, but interestingly, the production of 2′-Cl PTN and PTN remains almost unaffected, suggesting that they are specially required for 2′-amino dA biosynthesis. Moreover, we revealed that individual mutation of the transporter genes, including adaD, adaH, or adaI, prevents 2′-amino dA production (Fig. 2A and 3), implying that the gene products act as specific transporters of 2′-amino dA.

Furthermore, we found the production of 2′-amino dA and 2′-Cl PTN remains almost intact for the ΔadaK or ΔadaM mutant (Fig. 3), suggesting that there might be other genes that encode alternative functional products capable of performing the same roles as adaK or adaM. Curiously, as for the ΔadaE mutant, LC-MS analysis reveals that the production of 2′-Cl PTN and 2′-amino dA is completely abrogated, while that of PTN is barely unaffected, implicating that this AdaE enzyme might play a role in the chlorination of PTN to produce 2′-Cl PTN; however, the precise mechanism on how this reaction occurs has remained enigmatic. Altogether, our genetic data suggest that 2′-amino dA and 2′-Cl PTN arise from independent biosynthetic pathways, and 2′-amino dA biosynthesis is strictly dependent on the production of 2′-Cl PTN (Fig. 3).

In silico analysis of the 2′-Cl PTN and 2′-amino dA biosynthetic gene cluster.

Sequence analysis indicates that the G+C content of the minimal ada gene cluster is relatively high (73.64%) but similar to that of the typical genome of actinomycetes. In the ada gene cluster, adaABCDE, adaFGHIJ, and adaLM constitute individual transcriptional units according to bioinformatic analysis, whereas adaK is a standalone gene which is located in the middle of the ada gene cluster (Fig. 2A).

As suggested by genetic investigation and in silico analysis, adaABCEKL are proposed to be involved in 2′-Cl PTN biosynthesis. Of their products, AdaA, AdaB, and AdaC correspond to the individual enzymes PenC (SAICAR synthetase, 72% identity), PenB (short-chain dehydrogenase, 70% identity), and PenA (ATP phosphoribosyltransferase, 58% identity) in the PTN biosynthetic pathway from Streptomyces antibioticus NRRL 3238 (Fig. 2A and Table 1). Another enzyme, AdaL, also indicates significant homology to PenA with 53% identity, and it is deduced to be needed for 2′-Cl PTN biosynthesis. As for adaK, it encodes a protein showing 75% identity to HMPREF1486_01538 annotated as phosphoribosyl isomerase (HisA) from Streptomyces sp. strain WM4235 (Fig. 2A and Table 1). Very attractively, AdaE exhibits 54% identity to ADL15_15700, an annotated cation/H⁺ antiporter, from Streptomyces venezuelae ATCC 15439 (Fig. 2A and Table 1), but how this protein participates in 2′-Cl PTN biosynthesis is currently unknown.

Four other structural genes, adaFGJM, are likely to be required for 2′-amino dA biosynthesis (Fig. 2A and Table 1). adaF codes for a protein showing 53% identity to an aminotransferase, COCOR_00673 from Corallococcus coralloides DSM 2259, and this enzyme is directly responsible for the transfer of an amino group during 2′-amino dA biosynthesis. AdaG has moderate homology (28% identity) to PCL1391_5850 (a predicted dehydrogenase) of Pseudomonas chlororaphis, and AdaJ shows significant homology (54% identity) to Caci_7624, a putative Nudix hydrolase from Catenulispora acidiphila. AdaJ may trigger the biosynthesis of 2′-amino dA by hydrolyzing ATP to form AMP. In the ada gene cluster, three genes, adaDHI, encode putative transporters, which are particularly responsible for transportation of the product 2′-amino dA.

Biochemical characterization of AdaJ as an ATP Nudix hydrolase.

In silico analysis indicated that AdaJ contains a highly conserved 23-residue Nudix motif (GX₅EX₇REUXEEXGW) (Fig. S5), which functions as a metal binding and catalytic site; therefore, this enzyme presumably functions as a Nudix hydrolase. To see if AdaJ executes such a functional role, it was overexpressed in E. coli with a fusion maltose binding protein (MBP) tag and purified to near-homogeneity (Fig. 4A). 2′-Amino dA is an adenosine analog, implicating that ATP is most likely the substrate of AdaJ; moreover, Nudix hydrolases usually require a divalent cation, such as Mg²⁺ or Mn²⁺, for their activity. We therefore tested its activity in vitro using ATP as the substrate and Mg²⁺ as a divalent cation; as expected, the results showed that AdaJ is capable of converting ATP to form a new peak at a retention time (RT) of 21.0 min, which is consistent with that of AMP authentic standard (RT, 21.3 min) (Fig. S6A). Further LC-MS analysis exhibited that the new peak could produce a characteristic [M+H]⁺ ion at m/z 348.0704, with fragment at m/z 135.9938, which fully agrees with those of the AMP authentic standard (Fig. S6B and C). However, the negative control could not generate the characteristic AMP peak.

FIG 4.

Biochemical characterization of AdaJ as an ATP Nudix hydrolase. (A) SDS-PAGE analysis of the purified AdaJ. (B) Relative activity of AdaJ with ATP as the substrate and different divalent cations as a metallic cofactor. The arrow and number correspondingly indicate AdaJ and its size. (C) HPLC traces of AdaJ-catalyzed reaction with ATP as the substrate and Co²⁺ as a cofactor. (i) AdaJ-catalyzed reaction with ATP as the substrate and Co²⁺ as cofactor; (ii) the negative control without enzyme added; (iii) the authentic standard of AMP; (iv) the authentic standard of ATP. (D) Evaluation of the AdaJ activity against different substrates. Co²⁺ was utilized as the metallic cofactor. The error bars represent the standard deviations from at least three different experiments.

Next, we evaluated the influence of divalent cation on AdaJ activity. Of all divalent cations selected, including Co²⁺, Mg²⁺, Mn²⁺, Fe²⁺, Cu²⁺, and Zn²⁺, surprisingly, we found that Co²⁺ is capable of maintaining the maximal activity for AdaJ, thereof suggesting that AdaJ is a noncanonical Nudix hydrolase using Co²⁺ as the most preferred metallic factor (Fig. 4B and C). We subsequently tested the substrate flexibility of the enzyme, and we found that AdaJ could also consume dATP and GTP as the substrate, but it could not recognize ADP as a substrate (Fig. 4D). Taken together, our biochemical data verified that AdaJ functions as a distinctive Nudix hydrolase that initiates the biosynthesis of 2′-amino dA by hydrolysis of ATP to AMP.

Host adenosine deaminase ADA1 contributes to 2′-amino dA deamination and is relatively insensitive to PTN.

We found that Actinomadura sp. ATCC 39365 is capable of producing 2′-amino dI in a lower yield (only detectable by LC-MS), despite the fact that 2′-Cl PTN and PTN are both powerful adenosine deaminase inhibitors. In addition, why does the adaE mutant abolish the capability to produce 2′-amino dA? To address which adenosine deaminase of the host contributes to the phenotypes, we thus selected a versatile Ara-A/adenosine deaminase from S. antibioticus, SanADA3 (accession no. KT591401) (14), as a probe to conduct BLASTP analysis against the genome of Actinomadura sp. ATCC 39365, which leads to the identification of four homologs. Of them, ADA1 exhibits the highest homology (62% identity) to SanADA3 (Fig. S7A), implicating that this enzyme is most likely to govern the deamination of 2′-amino dA.

To test the assumption, ADA1 was overexpressed in E. coli and purified to near-homogeneity (Fig. 5A), and we then tested its activity in vitro. As anticipated, the results established that ADA1 converts adenosine/2′-amino dA to inosine/2′-amino dI, confirming its functional role as an adenosine/2′-amino dA deaminase (Fig. 5B to D and S7B and C). To further see if PTN is capable of protecting 2′-amino dA from deamination by ADA1, the reactions containing mixtures with PTN (final concentration, 0.01 mM, 0.1 mM, or 1 mM) and the substrates (adenosine or 2′-amino dA) were initiated by adding ADA1. High-performance liquid chromatography (HPLC) results indicated that PTN is not able to protect adenosine from deamination at all, even in the presence of 1 mM PTN (Fig. 5C and S7D), while 2′-amino dA can be partially protected from deamination in the presence of PTN (0.1 mM, or 1 mM) (Fig. 5D and S7D). These data suggest that ADA1 functions as an adenosine/2′-amino dA deaminase and is relatively insensitive to inhibition by PTN.

FIG 5.

In vitro characterization of ADA1 as the adenosine/2′-amino-dA deaminase, which is not very sensitive to PTN. (A) SDS-PAGE analysis of Actinomadura sp. ATCC 39365 adenosine deaminase 1 (abbreviated ADA1). The arrow and number correspondingly indicate ADA1 and its size. (B) Schematic of ADA1-catalyzed reaction. (C) HPLC traces of ADA1-catalyzed reaction with adenosine as the substrate. PTN is not an effective ADA1 inhibitor. (i) The authentic standard of inosine; (ii) ADA1 reaction with 1 mM PTN (final concentration) added; (iii) ADA1 reaction with 0.1 mM PTN (final concentration) added; (iv) ADA1 reaction with 0. 01 mM PTN (final concentration) added; (v) ADA1 reaction without PTN added; (vi) negative control without enzyme and PTN added; (vii) negative control without enzyme but with 1 mM PTN added. (D) HPLC traces of ADA1-catalyzed reaction with 2′-amino dA as the substrate and PTN as an inhibitor. (i) The authentic standard of 2′-amino dI; (ii) ADA1 reaction with 1 mM PTN (final concentration) added; (iii) ADA1 reaction with 0.1 mM PTN (final concentration) added; (iv) ADA1 reaction with 0.01 mM PTN (final concentration) added; (v) ADA1 reaction without PTN added; (vi) negative control without enzyme and PTN added; (vii) negative control without enzyme but with 1 mM PTN added.

DISCUSSION

Earlier metabolic feeding experiments have established that 2′-Cl PTN and PTN biosynthesis involves ring expansion with an additional one-carbon-unit (C-7) insertion between C-6 and N-1 of the purine scaffold, and the one-carbon unit has been previously demonstrated to be derived from C-1 of ribose (7, 19). More than that, the 2′-Cl PTN/PTN pathway was deduced to be closely related to that of the primary l-histidine (20). The results of this study are consistent with previous work for PTN pathway in S. antibioticus NRRL 3238 (14). Five enzymes, AdaA (SAICAR synthetase), AdaB (short-chain dehydrogenase) AdaC (ATP phosphoribosyltransferase), AdaK (phosphoribosyl isomerase), and AdaL (ATP phosphoribosyltransferase), are assigned as being involved in PTN pathway (Fig. 6A). The PTN biosynthesis would start with condensation of dATP and PRPP (phosphoribosyl pyrophosphate) to form compound 1 by two enzymes, AdaC and AdaL, and we propose that the two of them would collaborate to fulfill the catalytic function accounting for the fact that mutation of each one, either AdaC or AdaL, leads to the abolishment of PTN production. Subsequently, compound 1 will be converted to compound 2 through sequential reactions by three enzymes, HisI (phosphoribosyl-AMP cyclohydrolase), HisE (phosphoribosyl-ATP pyrophosphatase), and HisA (AdaK homolog, phosphoribosyl isomerase), from the histidine pathway. Interestingly, we find that the 2′-Cl PTN/PTN pathway contains a HisA homolog (AdaK) as well, which is likely to perform an identical function as HisA, as confirmed by our genetic investigation. Compound 2 will be catalyzed to compound 5 via the deduced intermediate compounds 3 and 4 by unique reactions that are proposed to be sequentially catalyzed by AdaA (SAICAR synthetase), which was previously uncharacterized in nucleoside antibiotic biosynthesis (Fig. 6A). After dephosphorylation by AdaM or an unknown enzyme, compound 6 is dehydrogenated by AdaB (PenB homolog) to produce the target product, PTN (Fig. 6A) (14).

FIG 6.

Proposed biosynthetic pathways to 2′-Cl PTN (A) and 2′-amino dA (B). In the proposed 2′-Cl PTN pathway, the initial step is consistent with the previous metabolite labeling experiments, and the following steps are proposed according to in silico analysis.

2′-Cl PTN was documented as one of the few naturally occurring nucleosides that contained a chloro group (21), but the biosynthetic origin and molecular mechanism of the chlorination have remained obscure for decades (21). Previous metabolic labeling studies have established that the inorganic chloride could be directly utilized during 2′-Cl PTN biosynthesis (4). Furthermore, it was proposed that several alternative chlorination mechanisms would be employed for such chlorination, including (i) stereo-selective insertion of a chloro group, involving a chloro-peroxidase-catalyzed reaction, (ii) conversion of 2′-amino dA to 2′-deazaadenosine, whose diazonium group can be displaced by a nucleophilic attack by a chloride ion, and (iii) a mechanism involving a cation radical enzyme-catalyzed reaction (4). However, based on the in silico analysis, we cannot identify an obviously classical halogenase in the 2′-Cl PTN pathway. Incredibly, our genetic studies indicated that adaE (encoding cation/H⁺ antiporter) is likely related to the chlorination, which is further supported by the identification of the candidate homologous enzymes in the pathways of ascamycin (AcmU) and nucleocidin (NucU); thus, we speculate that AdaE should be related to the tailoring chlorination step (Fig. 6A), and there might be a haloperoxidase enzyme in the both original and heterologous expression hosts that assists in the chlorination. Related research is now in intensive progress in our laboratory.

Previous metabolic labeling experiments have demonstrated that adenosine is the direct precursor for the biosynthesis of 2′-amino dA, and adenosine is first dehydrogenated to form a putative 2′-keto adenosine, which is then transaminated to generate the end product 2′-amino dA (4). In the present study, this is essentially correct. Four enzymes, including AdaJ (Nudix hydrolase), AdaM (hydrolase), AdaG (dehydrogenase), and AdaF (aminotransferase), are defined to participate in the biosynthesis of 2′-amino dA. We propose that the biosynthesis of 2′-amino dA is initiated by AdaJ, which catalyzes the hydrolysis of ATP to AMP with release of a pyrophosphate, and the catalytic reaction has been well characterized in our present study. The intermediate AMP is then dephosphorylated to adenosine by AdaM or an alternative enzyme(s), which undergoes the following dehydrogenation to form the 2′-keto adenosine intermediate (even though it could not be detectable due to its poor production or other unknown reasons). Finally, 2′-keto adenosine is catalyzed by a transamination step to accomplish the biosynthesis of the end nucleoside 2′-amino dA (Fig. 6B).

The attractive phenomenon of the concomitant production of the purine nucleoside pairs, as exemplified by PTN–Ara-A, coformycin-formycin, and 2′-Cl PTN-2′-amino dA, is more widely distributed than we imagined (14). Utilization of the three conserved enzymes (AdaA, AdaB, and AdaC) as valuable probes could lead to the discovery of additional pathways of potential PTN-related antibiotic pairs from the reservoir of sequenced microbial genomes. Notably, the advent of rapid and affordable DNA sequencing will certainly accelerate the traditional process for the discovery of new drugs related to PTN-related antibiotic pairs. Moreover, the enzymatic reactions of the halogenation reaction in the biosynthesis of 2′-Cl PTN, ascamycin, and nucleocidin have remained obscure for a half century (22, 23), and our identifying a probable cation/H⁺ antiporter, AdaE, that plays a potential role in the chlorination in 2′-Cl PTN will not only open a door for further illumination of the entirely unknown mechanism for halogenation chemistry, but it may also be used as an enzyme probe for the discovery of more halogenated natural products.

It is interesting how the adaM mutant also retains the ability to produce 2′-amino dA, and we conclude that the functional role of AdaM (AMP hydrolase) might be replaced by other alternative homologs. We also report a seemingly paradoxical finding, namely, that the adaE mutant lacks the ability to synthesize 2′-amino dA, as this gene was merely proposed to be essential for 2′-amino dA. We tentatively propose that once the tailoring chlorination step terminates during 2′-Cl PTN biosynthesis, the metabolic flux of the cell factory can only turn to PTN biosynthesis; however, the host adenosine deaminase is not sensitive to this nucleoside analog. As a result, adenosine, once synthesized, will be immediately deaminated to inosine for purine recycling.

In summary, we report the discovery and functional analysis of a 13-gene cluster essential for 2′-Cl PTN and 2′-amino dA biosynthesis. We further determine that these two nucleosides arise from independent biosynthetic pathways and provide biochemical proof that the adenosine deaminase ADA1 is capable of catalyzing the deamination of 2′-amino dA, but this enzyme is not highly sensitive to the inhibition of PTN. We have also illustrated that AdaJ (Nudix hydrolase) governs the initial step in 2′-amino dA biosynthesis. We anticipate that uncovering the precise logic underlying the biosynthesis of 2′-Cl PTN and 2′-amino dA will be of great potential for the combinatorial biosynthesis of this group of nucleoside antibiotic pairs with modified activity and selectivity.

MATERIALS AND METHODS

Strains, plasmids, primers, enzymes, chemicals, and general methods.

The strains and plasmids used in this study are described in Table 2, and the relevant PCR primers are listed in Table S1. All of the restriction enzymes and other enzymes used in this study were purchased from New England BioLabs. The standards, including 2′-amino dA, 2′-amino dI, and PTN, were purchased from Aladdin Biotech, Hongene Biotechnology, and Shanghai Qifa Biotechnology, respectively. All of other chemicals were purchased from Sigma-Aldrich, Thermo Scientific, or J&K Scientific. Standard protocols used to manipulate E. coli or Streptomyces were based on those of Green and Sambrook (24) or Kieser et al. (25).

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid	Relevant characteristicsa	Reference or source
Strains
Actinomadura sp. ATCC 39365	Wild-type strain	4
LG1	adaJ, adaK, adaL, adaM genes deleted in ATCC 39365	This study
S. aureochromogenes
CXR14	Industrial polyoxin producer with the entire polyoxin gene cluster deleted	17
CXR14::3G12	CXR14 strain containing 3G12	This study
CXR14::3G12lΔadaE	CXR14 strain containing 3G12lΔadaE	This study
CXR14::3G12lΔadaD	CXR14 strain containing 3G12lΔadaD	This study
CXR14::3G12lΔadaC	CXR14 strain containing 3G12lΔadaC	This study
CXR14::3G12lΔadaB	CXR14 strain containing 3G12lΔadaB	This study
CXR14::3G12lΔadaA	CXR14 strain containing 3G12lΔadaA	This study
CXR14::3G12lΔadaF	CXR14 strain containing 3G12lΔadaF	This study
CXR14::3G12lΔadaG	CXR14 strain containing 3G12lΔadaG	This study
CXR14::3G12lΔadaH	CXR14 strain containing 3G12lΔadaH	This study
CXR14::3G12lΔadaI	CXR14 strain containing 3G12lΔadaI	This study
CXR14::3G12lΔadaJ	CXR14 strain containing 3G12lΔadaJ	This study
CXR14::3G12lΔadaK	CXR14 strain containing 3G12lΔadaK	This study
CXR14::3G12lΔadaL	CXR14 strain containing 3G12lΔadaL	This study
CXR14::3G12lΔadaM	CXR14 strain containing 3G12lΔadaM	This study
CXR14::3G12lΔorf1	CXR14 strain containing 3G12lΔorf1	This study
E. coli
DH10B	Cloning host	Gibco-BRL
BW25113/pIJ790	λ Red(gam beta exo) cat araC rep101	18
BL21(DE3)/pLysE	F− ompT hsdSB(rB− mB−) gal dcm(DE3), pLysE (Cmlr)	Stratagene
Rosetta(DE3)/pLysS	F− ompT hsdSB(rB− mB−) gal dcm λ(DE3 [lacI lacUV5-T7 gene1 ind1 sam7 nin5]), pLysS	Novagen
ET12567 (pUZ8002)	dam dcm hsdM hsdS hsdR cat tet neo; helper strain for intergeneric conjugation	25
EPI300-T1R	Cosmid library host cell	Epicentre
Plasmids
pEASY-Blunt	pUCori lacZ f1 ori neo bla	TransGen Biotech
pJTU2463b	int aac(3)IV oriT RK2 phiC31 attP	16
pOJ446	aac(3)IV, SCP2, reppMB1*, attФC31, oriT	26
pET28a	neo reppMB1, T7 promoter	Novagen
pSJ8	lac MBP f1 ori bla
3G12	Cosmid containing entire ada gene cluster	This study
pLG001	pOJ446 derivative with insertion of 1.9-kb XbaI-BglII-engineered PCR fragment of left arm	This study
pLG002	pLG001 derivative carrying a BglII-engineered PCR fragment containing 2.5 kb of right arm	This study
pLG003	pLG002 derivative with insertion of a BglII fragment containing neo	This study
pET28a/ADA1	pET28a derivative carrying a NdeI-EcoRI fragment containing ada1 encoding 357 aa	This study
pSJ8/adaJ	pSJ8 derivative carrying a EcoRI-HindIII fragment containing adaJ encoding 161 aa	This study

^aCml^r, chloramphenicol resistance; aa, amino acids.

Sequencing and annotation of the genome of Actinomadura sp. ATCC 39365.

Genomic DNA of Actinomadura sp. ATCC 39365 was prepared according to the standard protocol (25), genome sequencing was performed using the Illumina HiSeq 2500 sequencing system, and the sequence data were then assembled using Velvet software and annotated using the Glimmer 3.0 software. The online programs FramePlot 4.0beta (http://nocardia.nih.go.jp/fp4/) and 2ndFind (http://biosyn.nih.go.jp/2ndfind/) were exploited for the accurate analysis of the 2′-Cl PTN and 2′-amino dA gene cluster.

Construction of the Actinomadura sp. LG1 mutant strain.

For the construction of LG1 mutant, the left arm (1.9 kb) and right arm (2.5 kb) were amplified with the primer pairs JKLM LarmF/LarmR and JKLM RarmF/RarmR. Afterwards, the left arm was digested with XbaI and BglII and cloned into the corresponding sites of pOJ446 to produce pLG001. Later, the right arm, cleaved by BglII, was cloned into the BglII-HpaI site of pLG001 to generate pLG002, and then the kanamycin resistance gene (neo) was cloned into the BglII site of pLG002 to form pLG003, which was subsequently introduced by conjugation into Actinomadura sp. ATCC 39365. After that, the standard methods were conducted in the screening of the LG1 mutant (2).

Production, purification, and LC-MS analysis of related antibiotics.

Production of related antibiotics by Actinomadura sp. ATCC 39365 was performed according to the methods of Tunac and Underhill (9). After fermentation, the broth was centrifuged at 8,000 rpm for 10 min, and the supernatant was extracted with an equal volume of n-butyl alcohol three times. After concentration, the residue was redissolved in water for further analysis. The LC-MS analysis of 2′-Cl PTN and 2′-amino dA was performed on a Thermo LTQ-Orbitrap electrospray ionization (ESI) high-resolution MS (HRMS) machine equipped with a C₁₈ reversed-phase column (5 μm, 4.6 × 250 mm; GL Sciences) in an elution gradient of 5% to 30% methanol/0.15% trifluoroacetic acid (TFA) over 30 min at 0.5 ml/min, and the elution was monitored at 254 nm with a diode-array detector (DAD).

Genomic library construction and screening for Actinomadura sp. ATCC 39365.

For the construction of pJTU2463b-derived genomic library for Actinomadura sp. ATCC 39365, the standard method was performed, using the EPI300-T1^R as suitable host cells, and the narrow-down PCR screening strategy (39365cluster-idF/R, 2463b-1F/1R, and 2F/2R) was employed (3) to screen the positive cosmid 3G12 from the genomic library.

In-frame deletion of the target ada genes by PCR-targeting strategy.

For targeted inactivation of the genes in the ada gene cluster, a kanamycin resistance cassette (neo) was amplified using corresponding primers 39365pcrtgtF/R (Table S1) and then recombined into the target gene in 3G12 by PCR-targeting strategy to give 3G12::neo (18). The neo cassette was then deleted to produce a series of 3G12/Δada derivatives (Fig. S4) (14). The unmarked deletions were subsequently confirmed by PCR using a related pair of primers (Table S1).

Expression and purification of AdaJ and ADA1 in E. coli Rosetta(DE3)/pLysS.

Using the primers listed in Table S1, pSJ8/adaJ and pET28a/ada1 were constructed and subsequently transformed into E. coli Rosetta(DE3)/pLysS cells, according to the standard protocols (24). For adaJ specifically, its codon was optimized in advance based on E. coli codon usage (see supplemental material). Expression and purification for both His₆-tagged proteins were employed according to the method of Wu et al. (14). For AdaJ, it was expressed with a fusion MBP tag, after digestion by tobacco etch virus (TEV), the target purified protein proteins were then concentrated and stored with protein stock buffer (25 mM Tris [pH 8.0], 150 mM NaCl, and 10% glycerol) using Amicon Ultra filters.

Biochemical and relative activity assays of AdaJ.

For the AdaJ activity assay, the reaction mixture consisting of 50 mM Tris-HCl buffer (pH 7.5), 1 mM ATP, 100 mM KCl, 50 mM divalent ion (Mg²⁺, Co²⁺, Mn²⁺, Zn²⁺, Fe²⁺, and Cu²⁺), and 20 μg AdaJ was performed at 30°C for 4 h and then terminated immediately by the addition of an equivalent volume of methanol. Following centrifugation to remove protein, the products were subsequently analyzed by HPLC (Shimadzu LC-20A) and LC-HRMS (Thermo LTQ-Orbitrap). Equipped with a reversed-phase C₁₈ column (Inertsil ODS-3, 4.6 by 250 mm, 5 μm) and monitored by a DAD at 260 nm, HPLC was performed at a flow rate of 0.5 ml/min, with the elution gradient of 5% to 20% methanol/10 mM trimethylammonium acetate (TEAA; pH 7.0) over 20 min, 5% methanol/10 mM TEAA for 22 min, and continued for 30 min. LC-HRMS/MS was operated in an ESI-ion trap mass spectrometer under the positive-ion mode, with drying gas at 275°C, 10 liters/ml, and nebulizer pressure of 30 lb/in.². The activity assays were performed with the general protocol of the EnzChek pyrophosphate assay kit (Thermo).

Accession number(s).

The DNA sequence of the ada gene cluster is available in the GenBank database under accession numbers KP025768 and KY373246.