In vitro Characterization of LmbK and LmbO: Identification of GDP-D-erythro-α-D-gluco-Octose as a Key Intermediate in Lincomycin A Biosynthesis

Abstract

Lincomycin A is a clinically useful antibiotic isolated from Streptomyces lincolnensis. It contains an unusual methylmercapto-substituted octose, methylthiolincosamide (MTL). While it has been demonstrated that the C₈ backbone of MTL moiety is derived from D-fructose 6-phosphate and D-ribose 5-phosphate via a transaldol reaction catalyzed by LmbR, the subsequent enzymatic transformations leading to the MTL moiety remain elusive. Here, we report the identification of GDP-D-erythro-α-D-gluco-octose (GDP-D-α-D-octose) as a key intermediate in the MTL biosynthetic pathway. Our data show that the octose 1,8-bisphosphate intermediate is first converted to octose 1-phosphate by a phosphatase, LmbK. The subsequent conversion of the octose 1-phosphate to GDP-D-α-D-octose is catalyzed by the octose 1-phosphate guanylyltransferase, LmbO. These results provide significant insight into the lincomycin biosynthetic pathway, because the activated octose likely serves as the acceptor for the installation of the C1 sulfur appendage of MTL.

Carbohydrates are indispensable for all living organisms and play important roles in determining the biological activities of diverse glyco-conjugates including glycosylated secondary metabolites.¹ Thus, alteration of the glycosylation pattern of targeted natural products by manipulating the corresponding sugar biosynthetic machinery holds promise to enhance or vary the biological properties of the parent molecules. To achieve this goal, it is essential to know how unusual sugars are biosynthesized. Recent research efforts in this area have advanced our understanding of deoxyhexoses formation in nature.² This knowledge has made it possible for glycodiversification of natural products, taking advantage of substrate promiscuity found for many glycosyltransferases.³ However, information about the construction, activation, and modification of high-carbon sugars (> 7 Cs), is scarce.⁴ For example, the biosynthesis of naturally occurring octoses remains unexplored except for KDO8P (3-deoxy-D-manno-2-octulosonate-8-phosphate), which is derived from arabinose 5-phophate and phosphoenolpyruvate (PEP).⁵

Lincomycin A (1), isolated from Streptomyces lincolnensis,⁶ is a clinically useful antibiotic against Gram-positive bacteria.⁷ It interacts with the peptidyltransferase domain of the 50S ribosomal subunit due to its structural resemblance to the 3′ end of L-Pro-Met-tRNA and deacetylated-tRNA. This, in turn, inhibits the bacterial protein synthesis.⁸ The structure of lincomycin A consists of an amino acid, N-methyl-4-propyl-L-proline (2), and an unusual methylthiolincosamide (MTL, 3) moiety.⁹ Similar thiooctoses are also found in antimicrobial agents, Bu-2545 (4) and celesticetin (5), produced by different Streptomyces strains (Figure 1).¹⁰

Figure 1.

Examples of natural products containing thiooctose moiety.

Recently, we reported that the C₈ backbone of MTL¹¹ is generated via a transaldol reaction catalyzed by LmbR using D-fructose 6-phosphate or D-sedoheptulose 7-phosphate as the C₃ donor and D-ribose 5-phosphate as the C₅ acceptor (Scheme 1).¹² Subsequent isomerization catalyzed by LmbN converts the resulting octulose 8-phosphate (7) to octose 8-phosphate (8).¹² The fact that several genes homologous to those found in various NDP-deoxyhexose pathways¹³ exist in the lincomycin gene cluster (designated as lmb) prompted us to hypothesize that an NDP-activated octopyranose is a possible intermediate in the MTL (3) biosynthetic pathway.¹² However, this hypothesis lacks literature precedence because NDP-octose has never been reported to be an biosynthetic intermediate. Thus, investigations of the existence of this putative NDP-octose intermediate and its transformation from octose 8-phosphate (8) have been the focus of our research effort. In this report, we describe the identification of GDP-D-erythro-α-D-gluco-octose (GDP-D-α-D-octose, 11) as the key intermediate in the MTL biosynthetic pathway.

Scheme 1.

Proposed biosynthetic pathway for MTL (3)

In a typical NDP-sugar biosynthetic pathway, the sugar precursor is first converted to the corresponding sugar 1-phosphate prior to nucleotidylyltransfer to yield the NDP-sugar product (e.g. 14 → 15 to make NDP-α-D-glucose, Scheme 2). For the biosynthesis of NDP-deoxyhexoses, generation of the hexose-1-phosphate precursor is achieved either by direct C-1 phosphorylation or via C-6 phosphorylation followed by a mutase-catalyzed 6→1 phosphoryl migration (13 → 14, Scheme 2).² A different scenario was observed in the biosynthesis of NDP-heptoses (Scheme 2), which are building blocks for the assembly of the inner core lipopolysaccharide (LPS). Studies of the formation of ADP-D-glycero-β-D-manno-heptose (ADP-D-β-D-heptose, 20) and GDP-D-glycero-α-D-manno-heptose (GDP-D-α-D-heptose, 23) revealed a kinase/phosphatase cascade to produce heptose 1-phosphate in three steps: (1) isomerization of D-sedoheptulose 7-phosphate to D-glycero-D-manno-heptose 7-phosphate (16 → 17); (2) anomeric phosphorylation of heptose 7-phosphate to heptose 1,7-bisphosphate by a kinase (17 → 18 and 17 → 21); (3) conversion of heptose 1,7-bisphosphate to heptose 1-phosphate via a phosphatase-catalyzed hydrolysis (18 → 19 and 21 → 22).¹⁴

Scheme 2.

NDP-hexose and NDP-heptoses biosynthetic pathways

Sequence analysis of the lmb gene cluster from S. lincolnensis ATCC 25466 found no phosphoryl migration mutase homologue required for installation of the C1 phosphoryl group via a C_N → C1 migration route. However, a comparison of the lmb gene cluster with the NDP-heptose biosynthetic gene clusters from Aneutinibacillus thermoaerophilus DSM 10155/G+^15a and Escherichia coli K-12,^15b showed several genes, including lmbN, lmbP, lmbK and lmbO, whose translated amino acid sequences have low to moderate similarities to those involved in NDP-heptose biosynthesis (Table 1). On the basis of this information, we speculated that MTL biosynthesis might follow an analogous kinase/phosphatase cascade observed in NDP-heptose biosynthetic pathway to make NDP-octose.

Table 1.

Sequence analysis of the genes in lincomycin A (1), ADP-D-β-D-heptose (20) and GDP-D-α-D-heptose (23) biosynthetic clusters.

	LmbN	LmbP	LmbK	LmbO
GDP-D-α-D-heptose (23)	GmhAc	HddA	GmhBc	HddC
identity/similarity (%)	26/43	23/41	25/4	27/47
ADP-D-β-D-heptose (20)	GmhAd	HldEb	GmhBd	HldEb
identity/similarity (%)	31/47	Nonea	32/47a	Nonea

^aNo significant similarity found.

^bHldE from E. coli is a bifunctional enzyme.

^cSequences from A. thermoaerophilus str. DSM10155/G+.

^dSequences from E. coli str. K-12 substr. MG1655.

As depicted in Scheme 1, LmbN catalyzes the C1 → C2 isomerization to produce the corresponding octose 8-phosphate (7 → 8) in a manner similar to the GmhA-catalyzed isomerization (16 → 17).¹² The putative kinase, LmbP, might phosphorylate the C1 hydroxyl group of octose 8-phosphate (8) to afford octose 1,8-bisphosphate (9). The C8 phosphate group of this bisphosphate intermediate might be hydrolyzed by LmbK (annotated as phosphatase) to give octose 1-phosphate (10), which is then converted to the nucleotide-activated octose (11) by LmbO (annotated as nucleotidylyltransferase) (route A). It is also conceivable that the reaction proceeds first with nucleotidylyltransfer by LmbO (9 → 12) followed by LmbK-catalyzed C8 dephosphorylation (12 → 11) (route B). Because LmbP and LmbO exhibit sequence similarities only to their counterparts in the GDP-D-α-D-heptose biosynthetic pathways (Table 1), the C1 activated octose intermediate in lincomycin biosynthesis is likely a GDP-D-α-D-octose. However, it is well known that prediction of substrate specificity for enzymes based solely on sequence alignment can be erroneous, and thus the chemical nature of sugar phosphodinucleotide intermediate in this pathway must be experimentally verified.^2c,d

In order to determine whether the proposed pathways are valid, especially the involvement of a NDP-octose intermediate in lincomycin biosynthesis, the lmbP, lmbK and lmbO genes were amplified from the genomic DNA of S. lincolnensis ATCC 25466 using PCR (polymerase chain reaction) and were individually cloned into pET28 vectors. The recombinant LmbP, LmbK and LmbO with N-terminal His₆-tags were overexpressed in E. coli. LmbK could be isolated and purified to near homogeneity as a soluble protein (Figure S1), but both LmbP and LmbO were expressed only as inclusion bodies. The denatured LmbO could be refolded through a protocol detailed in the Supporting Information, but attempts to obtain the soluble form of LmbP using a similar approach were not successful. To circumvent this obstacle, we chemically synthesized the putative substrate for LmbK (compound 9). The synthetic scheme is shown in Scheme 3.

Scheme 3.

Synthesis of octose 1,8-bisphosphate (9), octose 1-phosphate (10), and GDP-D-α-D-octose (11). ^aBoth 6S- and 6R-isomers of 27 were generated, and the desired 6R-isomer was isolated using silica-gel column chromatography.¹²

The synthetic bisphosphate 9 was first incubated with LmbK and the reaction mixtures were analyzed by HPLC equipped with a Corona charged aerosol detector (CAD) using an anion exchange column (Dionex CarboPac PA-1). A new peak was detected (Figure S2). This new product was collected and characterized by NMR spectroscopy. The ¹H-³¹P heteronuclear multiple quantum correlation (HMQC) spectrum of the substrate, octose 1,8-bisphosphate (9), displayed two ³¹P signals. One is at δ 2.53 and is coupled to the two C-8 proton signals at δ 3.86 and δ 3.83, and the other is at δ 0.49 and is coupled to the anomeric proton at δ 5.35 (Figure S3). In contrast, the NMR spectrum of the isolated product displayed a single ³¹P signal at δ 3.17, which is coupled to the anomeric proton at δ 5.34. The loss of the C8 phosphoryl group from 9 during the LmbK-catalyzed reaction is thus evident. To gain more support for the structural assignment of the isolated product, an octose 1-phosphate (10) standard was also synthesized as depicted in Scheme 3. The LmbK product was found to coelute with the synthetic octose 1-phosphate (10) on HPLC, and the spectral characteristics of the enzymatic product are in good agreement with the synthetic standard. These results fully established the identity of the LmbK product as 10 starting from 9.

After the formation of octose 1-phosphate (10), the next reaction step is likely the nucleotide activation catalyzed by the putative nucleotidylyltransferase, LmbO (route A). To test this hypothesis and verify which nucleotide is utilized in the lincomycin biosynthetic pathway, the synthetic bisphosphate 9 was incubated with LmbK and the refolded LmbO in the presence of ATP, CTP, GTP, TTP or UTP (see Supporting Information section S6 for details). The resulting mixtures were analyzed by HPLC (Figure 2). Although no new peak was detected in the reaction mixture using ATP, CTP, TTP or UTP, a sample derived from the incubation mixture with GTP showed a new peak that eluted in-between those of GMP and GDP with a retention time of 9.45 min (Figure 2, trace e). The same results were obtained when the synthetic octose 1-phosphate (10) was incubated with the refolded LmbO (Figure S4). This new product peak was collected and subjected to ESI-MS analysis. The recorded molecular mass (calcd for C₁₈H₂₈N₅O₁₈P₂⁻ [M−H⁺], 664.0910; obsd, 664.0922) is consistent with the proposed product, GDP-D-α-D-octose (11), which was also chemically prepared from 10, as shown in Scheme 3. As expected, the isolated product from the LmbK/LmbO reaction displayed an identical HPLC retention time as that of the synthetic standard (Figure 3, trace a and b).

Figure 2.

Activity assays for LmbK and refolded LmbO. All reactions contain the synthetic bisphosphate 9, LmbK and refolded LmbO. (a) CTP; (b) TTP; (c) UTP; (d) ATP; (e) GTP.

Figure 3.

Activity assays for the refolded LmbO. (a) The synthetic GDP-octose (11) standard; (b) LmbK and refolded LmbO with bisphosphate 9 and GTP; (c) refolded LmbO with bisphosphate 9 and GTP; (d) refolded LmbO with the synthetic octose 1-phosphate (10) and GTP.

The above results clearly indicate that the refolded LmbO is catalytically active and suggest a pathway in which C8 dephosphorylation by LmbK preceeds prior to nucleotidylyltransfer catalyzed by LmbO (route A in Scheme 1). However, a reversal of the sequence of these two reactions is also conceivable for the biosynthesis of GDP-octose 11 (route B). To check this possibility, the bisphosphate 9 was incubated with the refolded LmbO in the presence of GTP. As shown in Figure 3 (trace c), no new peak in the range of GDP-activated sugars is observed. In contrast, formation of GDP-D-α-D-octose (11, Figure 3, trace d) is clearly visible when the synthetic octose 1-phosphate (10) was incubated with the refolded LmbO and GTP. Evidently, C8 dephosphorylation is a prerequisite for the subsequent nucleotidyl activation. These results unequivocally establish that GDP-octose (11) is indeed an intermediate in MTL biosynthesis and formation of GDP-octose follows route A, not route B.

In the heptose biosynthetic pathways, the bisphosphate products (e.g., 18 and 21) generated in the kinase reactions have defined anomeric configurations, which directly correlate with the anomeric specificities of the subsequent nucleotidylyltransferase (e.g. HldE and HddC) catalyzed transformations (e.g., 19 → 20 and 22 → 23, Scheme 2). An analogous anomeric specificity also appears to be conserved in LmbK- and LmbO-catalyzed reactions because their products are all α-anomers. To investigate whether LmbO tolerates flexibility in its anomeric specificity, the refolded LmbO was incubated separately with GTP and each of the two anomers of glucose 1-phosphate, which were used as substrate analogues. Since formation of GDP-α-glucose was found only in the sample of α-glucose 1-phosphate (Figure 4, trace a and b), LmbO shows stringent α-anomeric stereospecificity for both octose and hexose substrates. Interestingly, although LmbO is capable of processing six-carbon sugar 1-phosphate, LmbK could not hydrolyze α-D-glucose 1,6-bisphosphate (Figure 4, trace c). A similar observation was reported for heptose 1,7-bisphosphate phosphatase GmhB, which could not recognize α-D-glucose 1,6-bisphosphate as a substrate.¹⁵ Apparently, the location of the phosphoryl group being removed on the substrate is essential for these phosphatases.

Figure 4.

Activity assays for LmbK and refolded LmbO. (a) Refolded LmbO with α-glucose 1-phosphate and GTP; (b) refolded LmbO with β-glucose 1-phosphate and GTP; (c) LmbK and refolded LmbO with α-glucose 1,6-bisphosphate and GTP; (d) coinjection of the sample from LmbO reaction with α-glucose 1-phosphate (trace a), and the GDP-D-α-glucose standard; (e) GDP-D-α-glucose standard.

Taken together, these results provide significant insight into the lincomycin biosynthetic pathway, part of which is reminiscent of the NDP-heptose pathway. The transformation of octose 8-phosphate (8) to GDP-D-α-D-octose (11) is shown to involve kinase and phosphatase reactions as intermediary steps. Our data also reveal that the dephosphorylation step, catalyzed by LmbK, is critical for the nucleotide activation reaction. However, direct demonstration of the predicted kinase activity of LmbP was unsuccessful due to the difficulty in refolding insoluble LmbP. Nevertheless, the involvement of a kinase-catalyzed step and the α-stereospecificity of LmbP reaction are supported by the effective reconstitution of the biosynthesis of GDP-D-α-D-octose using the synthetic bisphosphate (9) as substrate.

The formation of a nucleotide-activated octose intermediate in the lincomycin biosynthetic pathway likely serves two purposes: (1) the phosphonucleotidyl group might be an important recognition/binding element for the later enzymes in the pathway, and (2) it can function as a good leaving group in a nucleophilic substitution reaction, which may eventually allow the installation of the C1 thiol group. Furthermore, the identification of the GDP-octose intermediate has settled a long-standing dispute concerning whether the proteins encoded in the lmb gene cluster with sequence similarity to some NDP-deoxyhexose modifying enzymes play any roles in lincomycin biosynthesis.¹² It should also be noted that most GDP-activated sugars are used in the biosynthesis of bacterial cell-wall polysaccharides and eukaryotic glycans.^2,16 Only GDP-mannose has been demonstrated or suggested to be a biosynthetic precursor of the sugar subunit in some secondary metabolites including the polyene macrolide nystatin, amphotericin and candicidin,¹⁷ the aminoglycoside hygromycin,¹⁸ and the antitumor drug bleomycin.¹⁹ Thus, these results identify several key intermediates in the lincomycin pathway and expand our knowledge of the roles of nucleotide-activated sugars in natural product biosynthesis.