Directed evolution of a macrolide-sensing transcription factor biosensor for the detection of macrolactone aglycones via “effector walking” and efflux pump deletion

Abstract

Glycosylated macrolactones (macrolides) often display broad and potent biological activities and are targets for drug development and discovery. The modular genetic organization of macrolide polyketide synthases (PKSs) and various polyketide tailoring enzymes has inspired the combinatorial biosynthesis of new-to-nature macrolides. However, most engineered PKS and macrolide biosynthetic pathways are ineffective and produce reduced or negligible product titers. Directed evolution could improve the activity of engineered PKSs and associated pathways but critically requires a high-throughput screen to identify active variants from large libraries. Transcription factor-based biosensors could be used for this purpose. However, the effector specificity of the only known macrolide-sensing transcription factor MphR is limited to macrolides modified with the sugar desosamine. The potential applications of MphR are subsequently limited, ruling out the possibility of leveraging MphR to screen libraries of pathway variants that make macrolactones that lack sugars (i.e., macrolide aglycones), such as the direct products of PKSs. In this study, we aimed to engineer the effector specificity of the MphR biosensor strain for detecting macrolide aglycones. By developing an “effector walking” strategy, coupled with efflux pump deletion, the effector profile of MphR was dramatically broadened to include several erythronolide macrolactones. This work sets the stage for applying directed evolution and other high-throughput screening approaches to various PKSs. Our results suggest a broadly applicable approach to developing biosensors that detect ligands that are very different in structure from the native effector.

INTRODUCTION

Macrolides are members of the World Health Organization’s list of essential medicines for their antibacterial properties, and their medicinal value continues to rise owing to their impressive antitumor, antifungal, immunosuppressant, and antiviral activities.^1–4 Many macrolides and their intermediates serve as drug leads or as inspiration for semisynthetic analogs.^4–7 In nature, megaenzyme complexes called type I polyketide synthases (PKSs) assemble and offload macrolactones as part of secondary metabolism. These products are then modified by various tailoring enzymes, such as glycosyltransferases and P450s, to afford the mature macrolide. The organizational modularity of PKSs implies that they can be engineered by simply modifying existing and well-studied templates, such as the 6-deoxyerythronolide B (6-dEB) synthase (DEBS), which is responsible for the biosynthesis of the macrolactone core of erythromycin A (ErA, Figure 1).⁸ In practice, however, only a small fraction of engineered PKSs are active.^9,10 Despite the emergence of detailed structural and mechanistic insights,^11–14 the overall failure of combinatorial biosynthesis can be attributed to the sheer size and structural complexity of PKSs, poorly described intact module/polypeptide structures, and broad scope of substrates and other components that must be provided in engineered hosts.^15–17

Figure 1.

MphR-based detection of ErA and other macrolides. Abbreviated depiction of the MphR genetic circuit engineered for macrolide detection. Structures shown are examples of established MphR effectors. Desosamine is highlighted red.

Directed evolution is a successful approach for improving enzyme activity and could be well suited to rescuing the activity of chimeric or engineered PKSs.^18–22 Random mutagenesis methods effectively generate millions of library members, but suitable screens that identify mutant PKSs with improved enzyme activity are not broadly available. To partially address this, we are interested in leveraging genetically encoded biosensors based on transcription factors as potentially generalizable and tunable devices for screening libraries of PKS variants.^18,23,24 Notably, MphR is part of a regulatory gene cassette in pathogenic Escherichia coli (E. coli) strains that controls the expression of the phosphotransferase MphA, which is involved in macrolide resistance²⁵ and forms the basis of the only known transcription factor biosensor for macrolides.^26–28 MphR is a promiscuous repressor protein that responds to various 12-, 14-, 15-, and 16-membered macrolide antibiotics modified with a desosamine sugar, the proposed site of MphA-catalyzed phosphorylation (Figure 1).^26,28 We previously developed a biosensor circuit based on MphR and superfolding GFP (sfGFP) as the sensing module (Figure 1) and engineered its detection capabilities and effector specificity.^26,29

The MphR biosensor provides a potentially powerful platform for guiding the high-throughput engineering of pathways and processes that produce mature macrolides, including screening engineered metabolic pathways, culture conditions, and post-PKS tailoring steps. However, while it also responds to late-stage ErA biosynthetic intermediates, it does not detect macrolides lacking desosamine (Figure 1). As such, biosensors based on MphR cannot distinguish between PKS (macrolactones) and post-PKS products (glycosylated macrolactones, i.e., macrolides) or their associated bottlenecks. A biosensor that can detect and discriminate between ErA pathway intermediates could be leveraged to guide high-throughput engineering of specific post-PKS steps.

Additionally, a biosensor for detecting 6-dEB, the core scaffold of ErA and the direct product of DEBS (Figure 2A), would be useful for directly screening the activity of PKSs without extensive post-PKS tailoring to access a detectable product. Moreover, 6-dEB shares structural similarities with the scaffolds of other polyketides, many of which have potent biological activities.^30–34 A biosensor that can detect 6-dEB would likely provide a screening platform for a broad range of PKSs beyond those that produce ErA.

Figure 2.

Directed evolution of MphR effector specificity. (A) Abbreviated ErA biosynthetic pathway. Desosamine is shown orange. Mycarose is shown blue. (B) Residues involved with ErA (yellow sticks) binding in wild-type MphR (monomer A, PDB: 3FRQ). Polar contacts within 4 Å of ErA are shown as dotted lines. Select waters are shown (light blue spheres). Cl anion is shown as a green sphere. Desosamine 2’-OH is circled. (C) Surface view of wild-type MphR showing the ErA binding pocket. MphR is rotated ~90° relative to panel A. Residues with polar sidechains are shown in magenta, hydrophobic orange, and charged in blue. Desosamine 2’-OH is circled. (D) Most transcription factor-directed evolution involves repeated rounds of mutagenesis and screening using the target effector compound. To identify variant MphR biosensors that are derepressed by the aglycone EB or 6-dEB, we propose “effector walking” to walk toward the target effector.

We set out to engineer an MphR variant sensitive to 6-dEB. However, the crystal structure of MphR bound to ErA suggests that the ErA sugars and desosamine 2’-OH phosphate are intimately involved in effector binding (Figure 2B).²⁷ The effector pocket binds the entire ErA structure, including sugars. Given that 6-dEB lacks these moieties, it might be challenging to initially find an MphR variant capable of being derepressed by the aglycone via rational redesign or traditional directed evolution involving iterative rounds of mutagenesis and screening against the target effector (Figure 2C). Dramatic alteration of the substrate specificity of enzymes and transcription factors has been achieved through directed evolution strategies based on substrate or ligand walking. During successive rounds of directed evolution, the native substrate or ligand is substituted with analogs that share increasing structural similarity to the target molecule, which “walks” the protein toward a fitness advantage.^35–38 The availability of ErA pathway intermediates (Figure 2A) conveniently affords a built-in substrate walking progression. Here, we set out to apply “effector walking” to manipulate the effector specificity of an MphR biosensor strain from phospho-ErA to 6-dEB (Figure 2D). Our results support a broadly applicable strategy for modifying the effector specificity of transcription factor biosensors when the target ligand is significantly different from that recognized by known transcription factors.

RESULTS AND DISCUSSION

Evaluating the sensitivity of MphR with nonphosphorylated macrolides

The structure of MphR bound to ErA suggests an electrostatic interaction between Lys21 and the negatively charged phosphate group that, if present, would be attached to the 2’-OH of the ErA desosamine sugar moiety (Figure 2B).²⁷ However, the quantitative impact of 2-phosphate on MphR derepression has not been previously determined, given that the ErA kinase MphA is usually present in MphR biosensor strains to provide macrolide resistance and support growth.²⁹ To address this, we used MphA in vitro to prepare a sample of phosphorylated ErA. We then determined that phospho-ErA displaces wild-type MphR from its cognate operator significantly better than ErA via an electrophoretic mobility shift assay (Supplemental Information, Figure S1). Cumulatively, the available structural information and in vitro studies with phosphor-ErA suggest that phosphor-ErA is the more effective ligand for derepressing wild-type MphR. To begin the effector walking of MphR, we first set out to delineate the ability of MphR to be derepressed in the presence and absence of MphA. Two E. coli TOP10 strains containing a single biosensor plasmid were subsequently constructed. The plasmid pSENSE-MphR-MphA was designed to provide MphA for macrolide phosphorylation (Figure 3A). The plasmid contains genes for MphR and MphA under constitutive expression, a pBR322 origin (100–500 copies per cell), and sfGFP encoding the Superfolder Green Fluorescent Protein (sfGFP) (Supplemental Information, Tables S1–S2). In contrast, pSENSE-MphR-ErmE was designed to provide the ribosomal dimethyltransferase ErmE instead of MphA (Figure 3B and Supplemental Information, Figure S2). ErmE dimethylates adenine 2058 in the E. coli 23S rRNA, preventing macrolides from binding to the ribosomal exit site.³⁹ In this way, resistance to ErA or other antibacterial macrolides could be provided to the biosensor strain without phosphorylation. The biosensor strains were evaluated by quantifying the sfGFP fluorescence signal in the presence of various ErA concentrations. The MphA-based biosensor strain demonstrated high sensitivity to ErA at a half-maximal fluorescence (K_1/2) concentration of 1.09 ± 0.04 μM (Figure 3C, Table 1, and Supplemental Information, Table S3), which is often in the mM range for other published transcription factor-effector pairs.⁴⁰

Figure 3.

Characterization of the wild-type MphR biosensor in the presence of MphA or ErmE. (A) MphA provides resistance to ErA via phosphorylation of the desosamine 2’-OH. Subsequently, in the presence of MphA, phospho-ErA is the likely effector that derepresses MphR. (B). ErmE provides resistance to ErA via demethylation of the ribosome. Subsequently, macrolide effectors are likely not modified when binding to MphR. (C) ErA dose-response curve of the wild-type MphR in the MphA and ErmE E. coli TOP10 strains. (D) MEB/EB dose-response curves of the wild-type MphR in the ErmE E. coli TOP10 strain. Error bars are the standard deviation of the mean (n=3, biological replicates) and are only visible when larger than the data point symbol. Shaded areas (where visible) represent the 95% confidence interval of each fitted curve.

Table 1.

Performance features of the wild-type and mutant MphR biosensor strains.

entry	mutations	E. coli strain	plasmid	effector	K1/2 (μM)a	dynamic rangeb	Nc
1	Wild type	TOP10	pSENSE-MphR-MphA	ErA	1.09 ± 0.04	107,000	3.0 ± 0.3
2	Wild type	TOP10	pSENSE-MphR-ErmE	ErA	61.7 ± 7.2	79,000	1.9 ± 0.2
				MEB	N.D.	N.D.	N.D.
				EB	N.D.	N.D.	N.D.
3	S106F	TOP10	pSENSE-MphR-MphA	MEB	158.7 ± 5.2	121,000	2.2 ± 0.1
4	S106F	TOP10	pSENSE-MphR-ErmE	ErA	0.45 ± 0.07	120,000	3.1 ± 0.5
				MEB	161.7 ± 9.2	98,000	2.0 ± 0.1
5	S106F	TOP10	pSENSE-MphR	MEB	166.7 ± 4.9	112,000	2.0 ± 0.1
6	N94D/S106F	TOP10	pSENSE-MphR	ErA	1.96 ± 4.82	222,000	1.0 ± 0.4
				MEB	8.49 ± 0.68	111,000	1.2 ± 0.1
				EB	696 ± 204	N.C.	N.C.
				6-dEB	N.C.	N.C.	N.C.
7	N94D	TOP10	pSENSE-MphR	ErA	8.05 ± 0.47	19,000	4.2 ± 0.8
				MEB	N.D.	N.D.	N.D.
8	Wild type	JW5503	pSENSE-MphR-ErmE	ErA	1.65 ± 0.13	91,000	2.6 ± 0.3
				MEB	N.D.	N.D.	N.D.
				EB	N.D.	N.D.	N.D.
				6-dEB	N.D.	N.D.	N.D.
9	S106F	JW5503	pSENSE-MphR	ErA	0.47 ± 0.09	46,000	2.3 ± 0.6
				MEB	0.96 ± 0.10	84,000	2.2 ± 0.4
				EB	79.1 ± 12.6	106,000	1.5 ± 0.1
				6-dEB	88.6 ± 15.1	117,000	1.7 ± 0.1
10	N94D/S106F	JW5503	pSENSE-MphR	ErA	0.31 ± 0.19	64,000	1.1 ± 0.5
				MEB	0.21 ± 0.03	66,000	1.5 ± 0.4
				EB	3.22 ± 0.41	70,000	1.3 ± 0.2
				6-dEB	6.26 ± 0.47	72,000	1.5 ± 0.2
11	N94D/S106F	K207–3 ΔTolC	pSENSE-MphR-ErmE	narbonolide	15.9 ± 5.9	72,000	1.0 ± 0.2
				10-DML	25.1± 13.5	51,000	1.4 ± 0.6
				tylactone	N.D.	N.D.	N.D.

^aConcentration of ligand at half-maximum normalized GFP fluorescence

^bGFP_max−GFP_min

^cHill coefficient, a measure of cooperativity within the MphR biosensor. Values >1 indicate positive cooperativity.

In contrast, dose-response analysis of the ErmE-based biosensor strain revealed a 60-fold lower sensitivity to ErA than with MphA (K_1/2 = 62 ± 7 μM, Figure 3C). These findings suggest that MphR is more sensitive to ErA-P than to ErA, given that in the ErmE biosensor strain, MphR likely encounters nonphosphorylated ErA.

To characterize the effector specificity of the wild-type MphR with ErA biosynthetic intermediates in the absence of MphA-catalyzed phosphorylation, the sfGFP fluorescence of the ErmE-based biosensor strain was determined with various concentrations of 3-O-ɑ-mycarosylerythronolide B (MEB) and erythronolide B (EB) (Figure 1A). Reporter fluorescence was not detected with either polyketide up to a concentration of 100 μM (Figure 3D), indicating that MphR is not derepressed by the monoglycoside or aglycone. However, further engineering is needed to develop MphR variants sensitive to macrolides lacking the desosamine sugar. Furthermore, the phosphorylation dependence must be addressed en route to a biosensor that responds to macrolactones such as MEB and EB.

Detection of MEB via structure-guided mutagenesis of the MphR effector binding site

In previous work, the mutation S106F was found after screening an MphR random mutant library for improved sensitivity to pikromycin.²⁶ Given that pikromycin is a macrolide modified with only one sugar, this mutation could also allow for better accommodation of other monoglycosides, such as MEB. Accordingly, a dose-response analysis of MphR S106F was carried out in all three biosensor strains (E. coli TOP10 pSENSE-MphR-ErmE, pSENSE-MphR-MphA, and pSENSE-MphR) and various concentrations of MEB (Figure 4A). Remarkably, even though the wild-type biosensor displayed no detectable response to MEB in any strain, the single mutation resulted in derepression in the ErmE, MphA, and no-resistance biosensor strains, with indistinguishable sensitivity, as judged by the K_1/2 values of 159 ± 5 μM, 162 ± 9 μM, and 167 ± 5 μM, respectively (Table 1). These findings suggest the MphA kinase does not impact MphR derepression by macrolides without a desosamine sugar. The overall sensitivity of the mutant with MEB was ~2.5-fold lower than that of the wild-type MphR sensitivity with ErA in pSENSE-MphR-ErmE (62 ± 7 μM). Overall, these data imply that the wild-type biosensor strain response depends on the presence of the ErA desosamine and/or the associated phosphorylation for optimal derepression, but the S106F mutant does not. Notably, the viability and fluorescence response of the mutant in the no-resistance biosensor strain (E. coli pSENSE-MphR) revealed that MEB does not impact the growth of the biosensor strain. Consistent with the known negligible antibacterial activity of MEB,⁴¹ this suggests that ErA resistance genes are not required to produce and detect MEB or other biosynthetic intermediates that lack desosamine.

Figure 4.

Characterization of the wild-type and S106F MphR biosensor strains in the presence of MphA or ErmE with various polyketides. (A) MEB or EB dose-response curves of MphR S106F in the MphA, ErmE, and no resistance E. coli TOP10 strains. Shaded areas (where visible) represent the 95% confidence interval of each fitted curve. (B) Normalized and background subtracted (0 μM macrolide) fluorescence of the wild-type and S106F MphR in MphA and ErmE E. coli TOP10 strains in response to 100 μM various macrolides. Error bars are the standard deviation of the mean (n=3, biological replicates) and are only visible when larger than the data point symbol. See Figure 1B for structures of the macrolides tested.

To further confirm that MphR S106F is not dependent on phosphorylation, the mutant was assayed with various concentrations of ErA and clarithromycin, pikromycin, and azithromycin in the MphA (E. coli pSENSE-MphR-MphA) and ErmE biosensor strains (E. coli pSENSE-MphR-ErmE). These macrolides contain desosamine and are expected to be phosphorylated by MphA. The S106F mutant was derepressed by all the tested macrolides in the MphA biosensor strain (Figure 4B). Compared with the ErmE strain, the MphA-based S106F biosensor strain presented greater fluorescence with all the macrolides tested (Figure 4B), likely reflecting enhanced derepression by the macrolides phosphorylated with MphA and/or improved sequestration via macrolide phosphorylation. In the S106F mutant, pikromycin was not a detectable effector in the ErmE strain. Notably, in the presence of ErmE and the absence of macrolide phosphorylation, only ErA was a detectable effector of wild-type MphR. This result suggests that the S106F mutation is insufficient to drive derepression by pikromycin in the absence of phosphorylation of the macrolide by MphA, perhaps due to differences in structure compared with the other macrolides tested. Together, these results highlight the importance of macrolide phosphorylation for maximal derepression with wild-type MphR and the impact of S106F on the effector specificity.

To evaluate whether Phe at position 106 was the optimal substitution for further effector walking, a saturation library at Ser106 was constructed by site-directed mutagenesis with a partially degenerate NNK codon and ~400 library members screened with MEB (data not shown). Mutants with an improved response to MEB (>1 standard deviation of the S106F mean) were not identified, indicating that Phe is likely the best substitution at Ser106 for derepression by MEB. Notably, however, S106F displays no detectable derepression by EB (Figure 4A), indicating that further engineering is needed to detect the ErA aglycone.

Directed evolution of MphR S106F for increased sensitivity to MEB

The detection range of MphR S106F likely needs to be improved to be useful in screening libraries of pathway variants, given that MEB titers in engineered E. coli are ~73 μM.⁴¹ To this end, an error-prone PCR library was generated using pSENSE-MphR-S106F as the template. The mutant library was transformed into E. coli TOP10 and subjected to an initial round of negative sorting in the presence of MEB via fluorescence activated cell sorting (FACS) to remove variants that were not responsive to MEB (Supplemental Information, Figure S3). The remaining library members were screened with a fixed concentration of MEB in microtiter plates. One library member showed an increase in sensitivity toward MEB compared with S106F and was selected for dose-response analysis, which revealed a K_1/2 of 5.3 ± 0.4 μM, a 30-fold improvement compared with that of S106F with MEB (Figure 5). The DNA sequence of the mutant gene revealed an additional N94D mutation. Asn94 is a solvent-facing residue positioned at the entrance to the effector binding site and is noted for being close to the desosamine sugar amine and likely the phosphate binding site (Figure 2A/B).²⁷ To delineate the impact of each mutation in this double mutant, the single mutant N94D was constructed, and its response to various concentrations of MEB was determined. Notably, the fluorescence response of N94D was not detected, indicating that N94D and S106F synergistically affect derepression by MEB. The MphR N94D/S106F mutant was subjected to saturation mutagenesis at position 94 by site-directed mutagenesis with a partially degenerate NNK codon. Then, library members were screened with a fixed concentration of MEB. However, mutants with further improvements in the fluorescence response were not identified (data not shown).

Figure 5.

Characterization of the MphR N94D/S106F and S106F biosensor with MEB, EB, and 6-dEB. Error bars are the standard deviation of the mean (n=3, biological replicates) and are only visible when larger than the data point symbol. Shaded areas represent the 95% confidence limit of each fitted curve.

Next, inspired by enzymes that often display broad substrate specificity via directed evolution,^42,43 the effector specificity of MphR N94D/S106F was tested with EB and 6-dEB. Remarkably, dose-response analysis of the double mutant revealed derepression in the presence of EB or 6-dEB (Figure 5). The response with EB and 6-dEB was poor, and only the sensitivity with EB was fitted by the Hill equation (696 ± 204 μM, Table 1). Nevertheless, this MphR double mutant is the only reported transcription factor biosensor for detecting the polyketide macrolactone.

Improving the biosensor strain sensitivity to EB and 6-dEB via efflux pump knockout

The emergence of derepression with EB and 6-dEB by the MphR N94D/S106F double mutant provides a platform for further engineering biosensor sensitivity. As an alternative to further mutation of the transcription factor, we hypothesized that trapping the macrolactone intermediates in the cell may increase the likelihood of MphR encountering them, thus decreasing the K_1/2 of the biosensor. Although the action of MphA is expected to trap macrolides within E. coli via phosphorylation, the target macrolactones (6d-ED, EB) lack the desosamine sugar needed. Instead, we explored whether transporter manipulation could achieve a similar effect. The transmembrane transporter TolC is known to export macrolides from E. coli,⁴⁴ and its deletion increases MphR sensitivity to macrolides.⁴⁵ Accordingly, the E. coli mutant strain JW5503 (ΔtolC)⁴⁶ was used to house the wild-type MphR, S106F, and N94D/S106F variant ErmE-based plasmids. The biosensor strains were then characterized by measuring the fluorescence response in the presence of ErA, MEB, EB, and 6-dEB (Figure 6).

Figure 6.

Characterization of the wild-type and variant MphR biosensors in the TolC deletion strain, JW5503. (A) Scheme showing abbreviated MphR biosensor genetic circuit, ErmE resistance, and deletion of the TolC efflux pump. (B) Dose-response curves of the wild-type MphR. (C). Dose-response curves of MphR N94D/S106F. (D) Dose-response curves of MphR S106F. Error bars are the standard deviation of the mean (n=3, biological replicates) and are only visible when larger than the data point symbol. Shaded areas represent the 95% confidence limit of each fitted curve.

As expected, of the polyketides tested, only ErA derepressed the wild-type MphR in the tolC deletion strain. Its sensitivity, as judged by the K_1/2, was improved ~37-fold compared with that of the wild-type biosensor in E. coli Top10. In contrast, as anticipated, the N94D/S106F double mutant strain was derepressed with all four compounds tested. The sensitivity with EB, as judged by the K_1/2 value (3.22 ± 0.41 μM), was improved 216-fold compared with that of the same mutant in E. coli Top10. The increase in sensitivity of the double mutant strain with 6-dEB is likely greater than that of EB, but the K_1/2 with 6-dEB could not be determined in Top10. Notably, the fold-induction of the biosensor strains that include the N94D mutation is lower than those without N94D (Supplemental Information, Table S3). This is attributed to a higher level of leakiness in the absence of an effector. Regardless, the dramatic increase in the sensitivity of the MphR variant strains with 6-dEB, EB, and MEB indicates that the tolC deletion likely increases the intracellular concentration of the macrolactones, as designed. Interestingly, all four compounds tested also derepressed the S106F mutant in the tolC deletion strain, although not as well as the double mutant. MEB was the best effector, as judged by the K_1/2 (0.94 ± 0.10 μM), followed by 6-dEB (88.62 ± 15.13 μM, ~14-fold greater than N94D/S106F) and EB (79.11 ± 12.62 μM, ~25-fold greater than N94D/S106F). The sugar moiety of MEB might enable additional interactions within the binding pocket of S106F that make derepression more favorable. These results indicate that macrolactone colocalization with MphR is insufficient to explain the de-repression of N94D/S106F in JW5503 and that the effector walking mutations S106F and N94D are required for optimal derepression to occur by EB and 6-dEB.

Effector scope of MphR with other macrolactones

Part of the rationale for engineering a 6-dEB biosensor was that the MphR biosensor strain might also detect macrolactones unrelated to the ErA biosynthetic pathway and could be used as a biosensor platform for reporting the activity of other macrolide biosynthetic systems. However, this is only possible if MphR binds macrolactones independently of any one feature on the effector scaffold. Accordingly, a small panel of macrolactones was tested with pSENSE-ErmE-MphR N94D/S106F (Figure 7A) in the E. coli strain K207–3, which was engineered as a host for heterologous polyketide production.⁴⁷ Tylactone did not elicit a response (data not shown), likely because the 16-membered ring is too large for the MphR effector pocket.²⁸ In contrast, both narbonolide and 10-deoxymethynolide (10-DML) were able to derepress the double mutant MphR biosensor strain (Figure 7B), with K_1/2 values 4.9- (15.9 ± 5.9 μM) and 7.7-fold (25.1 ± 13.5 μM) greater than those of the double mutant MphR in JW5503 with EB, respectively. These findings suggest that the effector specificity of the evolved MphR extends beyond those originally screened for and that the engineered biosensor strain is relevant to other polyketide biosynthetic pathways.

Figure 7.

Characterization of the MphR N94D/S106F biosensor in the K207–3 DTolC strain with various macrolactones. (A) Structures of macrolactones tested with MphR. (B) Dose-response curves of MphR N94D/S106F with narbonolide and 10-DML. Error bars are the standard deviation of the mean (n=3, biological replicates) and are only visible when larger than the data point symbol. Shaded areas represent the 95% confidence limit of each fitted curve.

DISCUSSION

The streamlined engineering of macrolide PKSs has not been fully realized, partly owing to the lack of high-throughput screens that can identify active or improved PKS variants from large libraries. MphR is the only known transcription factor biosensor for macrolide detection but, critically, is thought to depend on macrolides phosphorylated by MphA and is not derepressed by polyketides that lack the desosamine needed. This narrow effector specificity limits the detection scope of MphR and restricts its use for screening the production of heavily modified macrolides, excluding the possibility of screening PKSs directly.

To address this, the effector specificity of the MphR biosensor strain was engineered via an effector walking approach, resulting in the development of the double mutant N94D/S106F strain, which can detect 6-dEB when coupled with the deletion of the tolC efflux pump gene. As the starting point for effector walking and directed evolution, the wild-type MphR could not detect MEB or EB and was highly dependent on the presence of the MphA kinase. These findings suggested that MphR is better derepressed by phosphorylated macrolides, possibly because of better binding to MphR and/or sequestration inside the cell. Remarkably, the single mutation S106F, identified from a previous directed evolution campaign,²⁶ was sufficient for MphR to be derepressed by MEB in the absence of MphA. Although this mutant was only ~2.5-fold less sensitive to MEB than the wild-type biosensor with ErA, significant further improvement was needed to provide a viable system for reporting MEB production in engineered hosts and to effectively “walk” towards derepression by the aglycones EB and 6-dEB. One round of random mutagenesis and screening provided the N94D/S106F strain with a 30-fold improved sensitivity with MEB. The double mutant was also derepressed by EB and 6-dEB. Deletion of tolC further enhanced the detection of all the polyketides tested, with the sensitivity for EB improving 216-fold. Notably, the evolved N94D/S106F double mutant was also derepressed by narbonolide and 10-DML.

Given that the wild-type MphR strain does not detect EB and 6-dEB, and the sugars and phosphate of the native effector phospho-ErA are intimately involved in MphR binding according to the MphR structure, we anticipated that extensive mutagenesis of MphR might have been required to shift specificity toward the erythronolide aglycones. Notably, only two amino acid changes to wild-type MphR in combination with tolC deletion were needed to develop a sensor strain capable of detecting EB and 6-dEB. Interestingly, developing the N94D/S106F biosensor strain required two rounds of random mutagenesis (one originally used to generate S106F, the other to identify N94D)²⁶ and the genomic modification to delete tolC. In hindsight, if the effector walking approach had started in the tolC deletion strain, the double mutant could have also been discovered stepwise via S106F or by screening a single MphR library with a high mutation rate. The enormous impact of S106F and N94D combined with the tolC deletion suggests that MphR and perhaps other transcription factor biosensors could be evolved to respond to effectors very different in structure from their native effectors via effector walking and/or if subjected to high mutation rates and extensive screening, providing a viable alternative to computational (re)design.^48–50

Although tolC deletion has previously been shown to improve MphR sensitivity to mature macrolides such as ErA,⁴⁵ the results described here indicate for the first time that this deletion can improve sensitivity to the precursor MEB and the aglycones EB and 6-dEB. This result also suggests that tolC deletion is a viable alternative to MphA-based sequestration of macrolides as a strategy to improve biosensor sensitivity.⁵¹ Indeed, the broad specificity of TolC, including antibiotics, detergents, dyes, and other toxic compounds,⁴⁴ suggests that its deletion could also improve the sensitivity of other transcription factor biosensors unrelated to MphR.

The MphR N94D/S106F biosensor strain is the first known transcription factor for detecting the aglycone scaffolds of ErA or other macrolides. The titers of macrolide aglycones in engineered microbial strains typically approach ~100 μM, so our engineered MphR biosensor strains should be relevant for screening their heterologous production,^41,52,53 perhaps with additional optimization of the fold-induction.^54,55 Interestingly, the MphR N94D/S106F biosensor strain, although dramatically improved with 6d-EB and EB, remains an efficient ErA/MEB biosensor. Given that typically, only the minimally required pathway components are provided in polyketide-producing strains, this effector promiscuity is not expected to negatively impact the suitability of the biosensor strain for applications related to screening libraries of pathway or PKS variants. Nevertheless, additional directed evolution of the biosensor strain could be performed to tailor the specificity if desired.²⁹

Cumulatively, our results demonstrate that effector walking and tolC deletion is a powerful strategy to dramatically alter the effector scope of MphR biosensor strains. These results further suggest that the effector plasticity of MphR is much greater than previously recognized. The sensitivity and effector scope of the MphR biosensor strains described here set the stage for applying directed evolution and other high-throughput approaches to screening PKS libraries and improving the activity of engineered PKSs.^56–60 The MphR biosensor strains significantly increase the potential to expedite PKS engineering and provide a basis for expanding biosensor-guided engineering of PKSs beyond ErA.

MATERIALS AND METHODS

Bacterial Strains, Plasmids, and Materials

The strains and plasmids used in this study are listed in the Supporting Information, Table S2. E. coli strains 10G (Lucigen) and TOP10 (Invitrogen) were used for cloning and biosensor expression, respectively. E. coli JW5503 was obtained from the Yale E. coli Genetic Stock Center. Bacteria were grown in Luria broth (Fisher Scientific) supplemented with ampicillin (Fisher Scientific) as appropriate. ErA, clarithromycin, and azithromycin were obtained from Sigma-Aldrich. Pikromycin was purchased from Abcam. MEB, EB, and 6-dEB were obtained from Fermalogic (Chicago, IL), and narbonolide and 10-deoxymethynolide were kind gifts from Prof. David Sherman (University of Michigan). Each macrolide was prepared in dimethyl sulfoxide (DMSO) at stock concentrations of 50 mM, 5 mM, 500 μM, or 50 μM. DMSO and 96-deepwell plates were purchased from Fisher Scientific. All other chemicals were purchased from Sigma-Aldrich unless otherwise stated. The PCR products were extracted with a Bio Basic Gel Extraction Kit. All the enzymes used for DNA manipulation were purchased from New England Biolabs. The plasmids were isolated via a plasmid miniprep kit from Bio Basic. All oligonucleotides were purchased from Integrated DNA Technologies. Clear and opaque flat-bottom 96-well plates were purchased from Greiner Bio-One. All other chemicals were of reagent grade or better.

Construction of pSENSE-MphR Biosensor Plasmids

The plasmid pSENSE-MphR was constructed by inserting MphR into multiple cloning site 1 of pSENSE2^29,61 via EcoRI and XbaI restriction sites. Briefly, the pSENSE2 plasmid was amplified via primers 1 and 2, and the MphR gene was amplified from pMLGFP via primers 3 and 4 (Supporting Information, Table S4). The amplified products were subjected to restriction digestion with EcoRI and XbaI, purified on a 1% agarose gel, and extracted from the gel via a Monarch gel extraction kit (NEB, Inc.). The purified fragments were digested with EcoRI and XbaI and then ligated overnight. The mixture was subsequently transformed into E. coli DH5α chemically competent cells. The transformation mixture was plated on ampicillin-containing (67 μg/mL) LB agar and incubated overnight at 37 °C. A colony was picked and cultured in 3 mL of LB plus ampicillin overnight at 37 °C, with shaking at 250 rpm. Then, 500 μL of 20% glycerol (v/v%) and 500 μL of culture were placed in a sterile 1.5 mL cryovial and stored at −80 °C.

The pSENSE-MphR-MphA plasmid was constructed via Gibson Assembly. The pSENSE-MphR plasmid was PCR amplified via primers 5 and 6 (Supporting Information, Table S4) to produce the vector fragment. The MphA gene was PCR amplified from pJZ12 via primers 7 and 8 (Supporting Information, Table S4). The PCR products were evaluated via agarose gel electrophoresis, and the resulting products were extracted via an NEB Monarch Gel Extraction Kit. The two fragments were assembled via Gibson Assembly, which leverages the NEBuilder Hifi DNA Assembly Kit following the manufacturer’s instructions. 2μL of the Gibson assembly was then transformed into E. coli TOP10 competent cells and the construct was sequenced to confirm successful assembly.

Similarly, the pSENSE-MphR-ErmE plasmid was constructed via Gibson Assembly. The pSENSE-MphR-MphA plasmid was PCR amplified via primers 9 and 10 (Supporting Information, Table S4) to produce the vector fragment. The ErmE gene was PCR amplified from pJZ12²⁷ via primers 11 and 12 (Supporting Information, Table S4). The PCR products were evaluated via agarose gel electrophoresis, and the resulting products were extracted via an NEB Monarch Gel Extraction Kit. The two fragments were assembled via Gibson Assembly, which leverages the NEBuilder Hifi DNA Assembly Kit following the manufacturer’s instructions. 2 μL of the Gibson assembly was then transformed into E. coli TOP10 competent cells and the construct was sequenced to confirm successful assembly.

Site-Directed Mutagenesis and Saturation Mutagenesis of MphR

Site-directed mutagenesis for the construction of MphR variants and saturation libraries was carried out via a modified “round the horn” procedure in a total volume of 16 μL containing 5X GC Phusion DNA polymerase buffer (4 μL), DNase-free water (11.8 μL), forward/reverse primer mixture (2 μL, 10 μM) (Supporting Information, Table S4), template DNA (1 μL, 50 ng/μL), dNTPs (0.4 μL, 2.5 mM each dNTP), DMSO (0.6 μL), and Phusion High Fidelity DNA Polymerase (0.2 μL) (New England Biolabs, NEB). The PCR cycling parameters were as follows: step 1) 98 °C, 30 s; step 2) 29x [a) 98 °C, 10 s; b) 55 °C, 1 min; c)72 °C, 5 min]; and step 3) 72 °C, 5 min. The PCR product was digested with DpnI at 37 °C for 1 h to remove any remaining template DNA. The DpnI reaction mixture contained 10X CutSmart Buffer (2 μL), the PCR mixture (17 μL), and DpnI (1 μL). The mixture was vortexed, centrifuged, and incubated for 3 h at 37 °C. The digested DNA was then purified by agarose gel electrophoresis on a 1% agarose gel, ligated overnight, and transformed into E. coli Top10 competent cells. An aliquot of the overnight culture was plated onto LB agar plates supplemented with 100 μg/mL ampicillin. The DNA sequences of the plasmids from individual transformants confirmed that the correct mutation was obtained and that there were no spurious mutations. Purified plasmids were transformed into E. cloni 10G electrocompetent cells, which were subsequently grown overnight in LB media supplemented with 67 μg/mL ampicillin. Glycerol stocks were prepared.

Error-Prone Mutagenesis of MphR

Error-prone PCR was performed via the GeneMorph II Random Mutagenesis Kit (Agilent Technologies) in a total volume of 50 μL with 20 ng of DNA template (pSENSE-MphR-S106F-ErmE), 0.8 mM dNTPs, 1.0 μL of Mutazyme II DNA polymerase, 4.0 μL of 10X Mutazyme buffer, 0.5 μL of DMSO, and 1.5 μL of the 10 mM forward primer, and 2.1 μL of the 10 mM reverse primer (entries 5−6, Supporting Information, Table S4). The thermocycling parameters were as follows: 95 °C, 2 min; [95 °C, 30 s; gradient 72 °C - 67 °C, 30 s; 72 °C, 1 min] cycle step 2–4 30X; 72 °C, 10 min, and hold at 12 °C. The PCR products were purified via agarose gel electrophoresis and double digested via EcoRI and HindIII following standard protocols. The double-digested product was ligated at 16 °C for 18 h into similarly treated pSENSE-MphR-S106F-ErmE. The ligation product was ethanol precipitated and electroporated into E. cloni 10G electrocompetent cells and incubated overnight according to the manufacturer’s instructions. An aliquot of the overnight culture was plated onto LB agar plates supplemented with 100 μg/mL ampicillin. Randomly selected colonies were used to prepare plasmids for DNA sequencing, revealing, on average, four nucleotide mutations per MphR gene. The remaining overnight culture was grown as an overnight liquid culture in LB supplemented with 100 μg/mL ampicillin. A copy of the library was made by the purification and storing of the plasmid DNA from an aliquot of the overnight culture. The remainder of the overnight culture was stored as glycerol stocks.

General Procedure for Microplate Screening of MphR Variants

Single colonies from mutants or mutant libraries in E. coli TOP10 cells were picked from LB agar plates and used to inoculate 300 μL of LB media containing 67 μg/mL ampicillin in the wells of a deep-well 96-well microplate. The cultures were incubated for 5 h at 37 °C and with shaking at 350 rpm. Then, 10 μL of each culture was added to 290 μL of LB media containing 67 μg/mL ampicillin and DMSO or polyketide at the indicated concentration. The plates were incubated overnight at 37 °C and with shaking at 350 rpm. The cultures were subsequently centrifuged at 1,509 × g for 5 min, after which the cell pellet was resuspended in 600 μL of phosphate-buffered saline (PBS). Next, 100 μL of each cell suspension was each transferred into a clear Greiner 96-well microplate and a Greiner 96-well black microplate. The optical density of the cells was analyzed at 600 nm, and the fluorescence was measured at 485 nm excitation and 510 nm emission. The fluorescence intensity was divided by the OD₆₀₀ to yield a normalized GFP fluorescence value.

Dose-Response Analysis of Wild-Type and Variant MphR Biosensor Strains

Single colonies of the wild-type or variant MphR biosensor strain in E. coli TOP10 cells were picked from LB agar plates and used to inoculate 3 mL of LB containing 67 μg/mL ampicillin. The cultures were incubated for 5 h at 37 °C and with shaking at 350 rpm. Then, 10 μL of each culture was added to 290 μL of LB media containing 67 μg/mL ampicillin and DMSO or polyketide at the appropriate concentration. The microplate was incubated overnight at 37 °C with shaking at 350 rpm. The plate was centrifuged at 1,509 × g for 5 min, and the cell pellet was resuspended in 600 μL of phosphate-buffered saline (PBS). Next, 100 μL of each cell suspension was transferred into a clear Greiner 96-well microplate and a Greiner 96-well black microplate. The optical density of the cells was analyzed at 600 nm, and the fluorescence was measured at 485 nm excitation and 510 nm emission. The fluorescence intensity was divided by the OD₆₀₀ to yield a normalized GFP fluorescence value. The concentrations and relative fluorescence intensities were plotted and analyzed via the Hill equation (equation I) via GraphPad Prism 7 software.

$$\text{GF}{\text{P}}_0=1+\frac{\text{GF}{\text{P}}_{\text{max}}{\left[\text{I}\right]}^{\text{n}}}{{\text{K}}_{1/2}^{\text{n}}+{\left[\text{I}\right]}^{\text{n}}}$$ equation I:

In this equation, GFP₀ is the normalized GFP expression when no inducer is present, GFP_max is the maximum normalized GFP expression observed, [I] is the inducer concentration, n is the Hill coefficient that quantifies the cooperativity of the protein, and K_1/2 is the inducer concentration at half-maximal normalized fluorescence.

Supplementary Material

Includes supplementary tables, supplemental figures, and detailed methods that describe DTolC strain construction, in vitro preparation of phosphor-ErA, and flow cytometry.