High-Titer Production of the Fungal Anhydrotetracycline, TAN-1612, in Engineered Yeasts

Abstract

Antibiotic resistance is a growing global health threat, demanding urgent responses. Tetracyclines, a widely used antibiotic class, are increasingly succumbing to antibiotic resistance; generating novel analogues is therefore a top priority for public health. Fungal tetracyclines provide structural and enzymatic diversity for novel tetracycline analogue production in tractable heterologous hosts, like yeasts, to combat antibiotic-resistant pathogens. Here, we successfully engineered Saccharomyces cerevisiae (baker’s yeast) and Saccharomyces boulardii (probiotic yeast) to produce the nonantibiotic fungal anhydrotetracycline, TAN-1612, in synthetic defined media–necessary for clean purifications–through heterologously expressing TAN-1612 genes mined from the fungus, Aspergillus niger ATCC 1015. This was accomplished via (i) a promoter library-based combinatorial pathway optimization of the biosynthetic TAN-1612 genes coexpressed with a putative TAN-1612 efflux pump, reducing TAN-1612 toxicity in yeasts while simultaneously increasing supernatant titers and (ii) the development of a medium-throughput UV–visible spectrophotometric assay that facilitates TAN-1612 combinatorial library screening. Through this multipronged approach, we optimized TAN-1612 production, yielding an over 450-fold increase compared to previously reported S. cerevisiae yields. TAN-1612 is an important tetracycline analogue precursor, and we thus present the first step toward generating novel tetracycline analogue therapeutics to combat current and emerging antibiotic resistance. We also report the first heterologous production of a fungal polyketide, like TAN-1612, in the probiotic S. boulardii. This highlights that engineered S. boulardii can biosynthesize complex natural products like tetracyclines, setting the stage to equip probiotic yeasts with synthetic therapeutic functionalities to generate living therapeutics or biocontrol agents for clinical and agricultural applications.

Graphical Abstract

Discovered in the 1940s, tetracyclines have since been prescribed as broad-spectrum antibiotics, treating both Gram-positive and -negative bacterial infections.^1,2 However, with their extensive use, novel and diverse tetracycline-specific resistance mechanisms arose, including multidrug transporter systems,³ ribosomal protection,⁴ rRNA binding mutations,⁵ and enzymatic inactivation,^6–8 diminishing antibacterial efficacy. Innovative technologies for novel tetracycline antibiotics discovery and production are imperative to mitigate the rampant resistance against these previously effective agents.

Tetracyclines are produced by three main methods: (i) isolation from the natural producing bacteria Streptomyces spp.,⁹ (ii) semisynthesis from those isolated compounds,¹⁰ and (iii) total synthesis.^11–13 While these methods enable the commercial production of classical tetracyclines, they present limitations in generating novel analogues; the natural bacterial producers are genetically intractable, and the synthetic routes are very complex. For example, only three tetracycline analogues were FDA-approved within the past decade, and they were developed via semisynthetic modifications (omadacycline and sarecycline) and by the Myers convergent total synthesis (eravacycline).¹¹ This evidences that despite decades of tetracycline research in both academia and industry, new technologies are urgently needed to generate novel analogues. Toward this goal, synthetic biology offers new approaches to tetracycline analogue production by employing heterologous genetically engineerable microbial hosts^14–17 and cell-free strategies.¹⁸

All current synthetic biology approaches that produce tetracyclines rely on heterologously expressing bacterial biosynthetic pathways. Boddy and co-workers produced oxytetracycline in both Myxococcus xanthus (~10 mg/L)¹⁴ and Escherichia coli (~2 mg/L),¹⁶ whereas Tang and co-workers biosynthesized oxytetracycline (~20 mg/L) and dactylocyclinone (~1 mg/L) in Streptomyces lividans K4-114.¹⁵ Moreover, the biosynthetic studies of the bacterial oxytetracycline pathway have led to several new tetracycline analogues/shunt products in different Actinomycetes beyond Streptomyces spp.^17,19–23 For example, Petkovic and co-workers recently generated novel antibiotic analogues of the atypical tetracycline chelocardin in Amycolatopsis sulphurea by expressing tailoring enzymes from the oxytetracycline biosynthesis pathway.^24,25 However, industrial large-scale fermentation of these heterologous bacterial host systems to produce tetracycline analogues is challenging as bacteria are susceptible to phage contamination.²⁶ Alternative nonbacterial heterologous hosts and/or biosynthetic pathways are therefore necessary.

Yeasts can fulfill this need; they grow to high cell densities in fermentation processes and are not prone to phage infection.²⁷ The baker’s yeast, Saccharomyces cerevisiae, is a main host organism that heterologously expresses diverse enzymes ranging from bacteria to eukaryotes. Notably, engineered yeast has been used to biosynthesize many therapeutic natural products, including terpenoids,²⁸ opioids,²⁹ cannabinoids,³⁰ alkaloids,^31,32 statins,³³ and nonribosomal peptide antibiotics.³⁴ In addition to being excellent heterologous hosts, yeasts are also used as probiotics; yeasts are great organisms to add to the microbial living therapeutic world currently dominated by bacterial hosts.³⁵ For example, Saccharomyces boulardii is a probiotic yeast extensively used to treat gastrointestinal diseases;³⁶ its properties could be further improved through genetic engineering to biosynthesize natural products with potential living drug delivery platform applications.³⁵ Because of their close genetic relationship, the synthetic biology toolboxes developed for S. cerevisiae are likely applicable for the genetic engineering of S. boulardii.^37,38 Moreover, S. boulardii has already proven itself as a prominent heterologous host, producing small molecules of therapeutic interest, like β-carotene,³⁷ violacein,³⁷ and leucocin C.³⁸

Engineered S. cerevisiae is a prime candidate for biosynthesizing tetracycline analogues due to its well-established genetic technologies and expression platform for fungal polyketide synthases (PKSs), which are toxic and/or difficult to produce in bacteria.³⁹ Heterologous tetracyclines production involving the bacterial oxytetracycline pathway has been explored in S. cerevisiae.^40,41 Our laboratory successfully achieved the final steps of tetracycline biosynthesis using S. cerevisiae with anhydrotetracycline supplementation, an oxytetracycline biosynthesis intermediate.⁴¹ However, total biosynthesis of tetracyclines in yeasts has yet to be demonstrated.

Fungal anhydrotetracyclines–TAN-1612 (neuropeptide Y antagonist) and its tautomeric form BMS-192548 (neurokinin-1 receptor antagonist)–were first naturally isolated from different filamentous fungi: Penicillium claviforme FL-27337 (~52 mg/L)⁴² and Aspergillus niger WB2346 (~3 mg/L),⁴³ respectively. Due to the absence of key tetracycline antibiotic pharmacophores, like the N-dimethylamino (−NMe₂) and carboxamide (−NH₂−CO−) groups (Scheme 1), TAN-1612 and BMS-192548 do not present therapeutic activity against Gram-positive or -negative bacteria or yeasts like S. cerevisiae or Candida albicans.^43,44

Scheme 1. Structures of the Fungal Anhydrotetracycline, TAN-1612, and Classical Bacterial Tetracyclines^a.

^aKey antibiotic pharmacophore carbon positions are indicated with a gray square.

Nevertheless, because of its chemical structural similarity to the tetracycline core and its fungal origins, TAN-1612 represents an excellent biosynthetic scaffold for novel tetracycline analogue production in heterologous yeast hosts to combat tetracycline-resistant bacteria. For example, heterologous production of TAN-1612 in yeasts does not require the challenging expression or exogenous supplementation of unique bacterial tetracycline biosynthesis cofactors that are non-native to yeasts like F₄₂₀ or F_O.^40,41 Indeed, important steps toward tetracycline analogues production in yeasts were made by Tang and co-workers, who reconstituted the TAN-1612 biosynthetic pathway from A. niger in S. cerevisiae.⁴⁵ Through co-expressing biosynthetic enzymes from the TAN-1612 pathway (AdaA, AdaB, AdaC, and AdaD) and the helper 4′-phosphopantetheinyl transferase, NpgA, in yeast, Tang et al. reported TAN-1612 titers of ~10 mg/L, using complex Yeast Extract–Peptone–Dextrose (YPD) as the production media.⁴⁵

However, using undefined complex media like YPD is nonoptimal for heterologous natural product biosynthesis because the yeast extract within might present end-product metabolites, regulating the host’s biochemical pathway activities and decreasing target molecule production titers.⁴⁶ Its use further complicates extraction and purification because some components would remain unconsumed by the host organism, potentially interfering with downstream processing involving fermentation product recovery.⁴⁷ For example, despite YPD cultures of engineered S. cerevisiae strains producing ~60 mg/L of targeted compounds, like depsipeptides⁴⁸ or polyketides,⁴⁹ only ~1 to 5 mg of the pure product were isolated after the 3–4 step purification process. Contrastingly, complete synthetic media (CSM) facilitates downstream processes, including the chemical characterization and semisynthetic modification of the fermented target products.^28,29,32,34 Thus, achieving high TAN-1612 titers in yeasts using CSM is highly desirable for novel tetracycline analogue development.

Here, we engineered S. cerevisiae to produce high titers of 32.5 ± 5.3 mg/L of the fungal anhydrotetracycline, TAN-1612, in CSM by controlling the expression of the six TAN-1612 biosynthetic enzymes AdaA, AdaB, AdaC, AdaD, a TAN-1612 efflux pump, and NpgA. Furthermore, we demonstrated the S. cerevisiae synthetic biology approaches could be applied to engineer S. boulardii, producing TAN-1612 with titers of 6.1 ± 2.3 mg/L. This is the first engineered S. boulardii strain to produce such a fungal polyketide.

The engineering approaches described herein can be readily applicable to a wide range of therapeutic natural product biosyntheses beyond tetracycline polyketides in S. cerevisiae, S. boulardii, or other therapeutic yeasts of interest, like Kluyveromyces lactis or Hansenula polymorpha.⁵⁰ Our approaches provide scaffolding for more ambitious and transformative applications in synthetic biology, involving engineering yeasts as living therapeutic platforms for the in situ production and/or targeted delivery of therapeutic small molecules, like tetracyclines, to the human body.

RESULTS AND DISCUSSION

Our goal was to engineer yeasts to heterologously produce the fungal anhydrotetracycline, TAN-1612, in complete synthetic media (CSM, referred to as “minimal media” hereafter) to further facilitate tetracycline analogue development in yeasts to combat antibiotic-resistant pathogens. To achieve TAN-1612 production in yeasts from its precursors, acetyl-CoA and malonyl-CoA, it is necessary to co-express five genes: npgA (phosphopantetheinyl transferase) from Aspergillus nidulans; adaA (nonreducing polyketide synthase), adaB (metallo-β-lactamase), adaC (flavin adenine dinucleotide-dependent oxidoreductase), and adaD (O-methyltransferase) all from A. niger ATCC 1015 (Figure 1A). Initial heterologous expression of these five genes biosynthesized TAN-1612 as the only polyketide product in S. cerevisiae at modest titers of ~10 mg/L in complex media (YPD) without strain optimization.⁴⁵

Figure 1.

Overview of TAN-1612 biosynthesis. (A) TAN-1612 biosynthesis involves a five-enzyme pathway where acetyl-CoA and malonyl-CoA are converted into TAN-1612 by npgA (phosphopantetheinyl transferase), adaA (nonreducing PKS), adaB (metallo-β-lactamase), adaC (flavin adenine dinucleotide-dependent oxidoreductase), and adaD (O-methyltransferase). (B) TAN-1612 biosynthetic gene cluster from A. niger ATCC 1015. The putative TAN-1612 efflux pump is shown in purple.

Here, we devised two engineering strategies to achieve TAN-1612 production in both S. cerevisiae and S. boulardii in minimal media. First, we used a promoter-based approach to assess potential TAN-1612 pathway bottlenecks and optimize its production. Second, we successfully explored if efflux pump co-expression could reduce TAN-1612 toxicity to yeasts in minimal media, eventually enhancing TAN-1612 titers. We combined these two approaches to increase production of TAN-1612 in S. cerevisiae. Finally, we showed the applicability of the combinatorial promoter library-based approach to genetically reprogram the probiotic S. boulardii to produce TAN-1612.

Establishment of a Medium-Throughput Screening Assay for TAN-1612.

To analyze the large quantity of strains generated during our engineering cycles to reprogram yeasts for heterologous TAN-1612 biosynthesis, a medium- or high-throughput (>1000 samples/h) screening method was necessary. We sought to establish a rapid, inexpensive, and reliable medium-throughput screening method for straightforward detection of TAN-1612 production in yeast cultures (supernatants and cell pellets).

The state-of-the-art methods for tetracycline detection are high-performance liquid chromatography (HPLC)⁵¹ and liquid chromatography–mass spectrometry (LC-MS).⁵² Because of their high sensitivity and specificity, these chromatography-based methods are generally applicable to diverse tetracycline analogues.^53,54 However, they require expensive instrumentation and time-consuming sample clean-up steps while remaining very low throughput (<10 samples/h).^51–54

Cell-based assays or biosensors using genetically engineered bacteria^55–57 or yeasts,^58,59 like the Y3H assay developed by our laboratory,⁵⁸ enable tetracycline detection with medium/high-throughput reporter gene usage,^55,56,58 fluorescence-activated cell sorting (FACS),⁵⁷ or growth selections^58,59 as readouts. These cell-based assays are highly molecule-specific that, in some cases, must be completely redesigned for each new tetracycline analogue,^56,59 and they also demand long experimentation times (>12 h) associated with sample preparation and cell growth of the genetic reporter systems.^55–59

Commercially available enzyme-linked immunosorbent (ELISA)-based assays, while costly (>$3/sample), are commonly used for the detection of tetracycline analogues like tetracycline, oxytetracycline, and anhydrotetracycline (Scheme 1) in environmental⁶⁰ and food⁶¹ samples with medium/high-throughputs and experimental workflows not requiring labor-intensive sample preparation.^60,61 As TAN-1612 is a fungal anhydrotetracycline analogue, ELISA-based assays could, in principle, detect it. However, contradictory results were reported regarding the ELISA assays performance to detect anhydrotetracycline. For example, while Aga et al. reported anhydrotetracycline detection from aqueous soil extracts using ELISA,⁶¹ Scott more recently not only reported that different commercial tetracycline ELISA assay kits were unsuitable for anhydrotetracycline standard detection but also these kits were incompatible for the detection of tetracyclines in yeasts as the samples (supernatants and cell pellets) interfered with the assay.⁴⁰ Scott’s findings are more analogous to our work because it was performed in yeast,⁴⁰ compared to the other reports using ELISA assays for tetracycline detection in nonyeast biological matrixes.^60,61

The additional aromatic ring in anhydrotetracycline enables both its spectrophotometric detection and differentiation from traditional tetracyclines that only have one aromatic ring (Scheme 1) in the UV–vis range (350–510 nm).⁶² We confirmed this spectrophotometric feature by acquiring UV–vis measurements from different tetracycline analogues solutions, including TAN-1612 (Figure S1). TAN-1612, being a fungal anhydrotetracycline and essentially the only biosynthetic tetracyclic polyketide produced by engineered yeasts co-expressing npgA and adaA–D as reported by Li et al.,⁴⁵ has reduced potential interference from putative tetracyclic shunt products; this encouraged us to develop a UV–vis assay that detects TAN-1612 in producing strains in a rapid manner without any purification.

Accordingly, we developed a straightforward spectrophotometric assay that measures TAN-1612 UV–vis absorbance (445 nm) in yeast liquid cultures that correlates to TAN-1612 production per yeast cell (Supporting Information, Figures S2–S4). This UV–vis assay circumvents conventional chemical extraction methods for quantification, which are low-throughput, tedious, and involve extensive experimental workflows.^{40,45,55–59}

We first confirmed our assay’s functionality by detecting TAN-1612 production from Li et al.’s yeast strain⁴⁵ in YPD (Figure S4). Next, we validated the performance and determined the dynamic range of the UV–vis assay in minimal media by spiking purified TAN-1612 from engineered A. niger ATCC 1015 + pYR311⁴⁵ into yeast cultures of a TAN-1612 nonproducer strain (Figure S5). Moreover, we compared the TAN-1612 quantity extracted from the supernatants of these naive BJ5464 cultures previously spiked with purified TAN-1612 through supercritical fluid chromatography–mass spectrometry (SFC-MS) and the UV–vis assay, and we found that the amounts correlate (Figure S6).

The UV–vis assay we developed exclusively for the straightforward detection of TAN-1612 in liquid cultures of yeasts producing strains is advantageous compared to the tetracycline detection methods mentioned above in terms of throughput (HPLC or LC-MS), cost (ELISA), and sample preparation requirements (HPLC, LC-MS, cell-based, or biosensors).

First-Generation of Strains Using S. cerevisiae BJ5464-NpgA for Potential TAN-1612 Production in Minimal Media.

It is desirable to produce TAN-1612 at mg/L levels in minimal media because this would enable its facile purification and derivatization toward novel tetracycline analogue production to combat antibiotic resistance. One general strategy to enhance heterologous biosynthesis production involves testing promoters of varying transcriptional strength to regulate the corresponding biosynthetic pathway gene expression. This strategy not only enables metabolic pathway balancing but can also help identify potential flux bottlenecks.^63,64

For this first-generation of potential TAN-1612 yeast producer strains, we employed S. cerevisiae BJ5464-NpgA as a production platform because it has been genetically modified to ensure the functional expression and purification of PKSs.³⁹ For example, two vacuolar protease genes (PEP4 and PRB1) were knocked out to significantly increase the expression levels of recombinant fungal PKSs.³⁹ Moreover, BJ5464-NpgA carries two chromosomal copies of A. nidulans npgA (δ sites::pADH2-npgA-tADH2), necessary for efficient heterologous phosphopantetheinylation of fungal PKSs in yeasts.^33,45 BJ5464-NpgA has been used widely in the heterologous production of fungal polyketides like TAN-1612 in YPD,^{33,39,45,48,49} but its performance as a platform for target molecule production in minimal media has yet to be demonstrated.

With BJ564-NpgA readily available, we focused on regulating only TAN-1612 gene expression (adaA–D) with different promoters. Therefore, to drive the expression of adaA–D genes, we chose four promoters whose transcriptional strengths are well documented across different yeast growth phases in minimal media: ADH2 (strong-ethanol-inducible-stationary), HSP26 (strong-stationary), TDH3 (strong-constitutive), and TEF1 (medium-constitutive).⁶⁵

We cloned the TAN-1612 biosynthetic genes on two high-copy-number 2μ plasmids: adaA in a URA3-based plasmid and adaB-adaC-adaD in a TRP1-based plasmid for auxotrophic selection in yeasts with CSM(UT-). Several combinations of these two plasmids harboring the adaA (under pADH2, pHSP26, pTDH3, or pTEF1) and adaB-D (under pADH2 or pTEF1) genes were transformed into S. cerevisiae BJ5464-NpgA (Tables S1–S3).

We tracked TAN-1612 production per cell (expressed as OD₄₄₅/OD₆₀₀) in our primary promoter combination-derived strains every 24 h over 4 days in minimal media. Using our UV–vis assay, we did not detect TAN-1612 production in Li et al.’s⁴⁵ strain (reference strain). Moreover, we did not observe statistically significant TAN-1612 production from any of our first-generation BJ5464-NpgA strains compared to both the reference and respective null strains (previously transformed with empty plasmids) (Figure S7A).

We hypothesized using pADH2 to express npgA might contribute to the unobserved TAN-1612 production in minimal media. pADH2 is a strong stationary phase-inducible promoter during yeast growth in ethanol, a highly abundant byproduct of both yeast extract and peptone, the latter being present in complex media like YPD but absent in minimal media. Therefore, pADH2’s transcriptional strength is stronger in YPD but weaker than those of pHSP26, pTDH3, or pTEF1 in minimal media.^65,66 We indeed confirmed pADH2’s strong transcriptional activity in YPD where all of the yeasts strains—derived from BJ5464-NpgA whose npgA expression was driven by pADH2—showed TAN-1612 production independent of the promoters used to express adaA–D (Figure S7B).

Considering these observations, we tested our reference strain Li et al.’s⁴⁵ for gene expression of npgA and adaA–D and TAN-1612 production in minimal media. We confirmed the expression of the TAN-1612 biosynthetic genes by qRT-PCR (Figure S8) and TAN-1612 production with ultrahigh-performance liquid chromatography–mass spectrometry (UPLC-MS) (Figure S9). We determined Li et al.’s⁴⁵ strain did indeed produce TAN-1612 in minimal media, albeit at low total concentrations of ~7.00 ± 5.71 μg/L (Figure S9).

Li et al.’s⁴⁵ TAN-1612 titers were outside of our UV–vis assay’s dynamic range (64–16 000 μg/L), thereby rendering them undetectable by our spectrophotometric assay. Our BJ5464-NpgA engineered strains showed similar TAN-1612 production per cell values to the reference control Li et al.’s⁴⁵ by our UV–vis assay; we concluded none of these preliminary strains were successful in achieving TAN-1612 production in minimal media at the targeted mg/L levels.

Second-Generation of Strains Using S. cerevisiae BJ5464 for TAN-1612 Production in Minimal Media.

Increasing TAN-1612 titers to mg/L levels in minimal media, rather than in complex media like YPD, allows for an easier recovery process, advancing tetracycline analogue development. We pursued this goal by constructing a second-generation of strains with S. cerevisiae BJ5464 as the background and focusing on regulating only npgA expression, not adaA–D’s (Figure 2A).

Figure 2.

Production of TAN-1612 in minimal media by genetically engineered S. cerevisiae. (A) Engineered S. cerevisiae BJ5464 strains with two copies of genomically integrated npgA under promoters of different transcriptional strengths, pADH2, pHSP26, or pTDH3, were transformed with plasmids expressing adaA–D genes (#1356: non-codon-optimized and #1359: yeast-codon-optimized—^opt). (B) Promoter library in engineered S. cerevisiae BJ5464 strains. TAN-1612 production per cell in engineered yeast strains was determined by UV–vis spectroscopy and expressed as the absorbance of TAN-1612 at 445 nm per cell growth at 600 nm measured directly in yeast cells in a microtiter plate format after 4 days in CSM(T-) or CSM(UT-) at 30 °C, 240 rpm. Li et al. engineered strain as previously reported.⁴⁵ (C) TAN-1612 is partially secreted as evidenced by UPLC-MS TAN-1612 areas for pellets and supernatants from a representative engineered S. cerevisiae strain shown in (B) (#1362: S. cerevisiae BJ5464 + 2x pTDH3-npgA-tENO2 + adaA–D genes). Each symbol represents different replicates. TAN-1612–0.2 mg/L standard (n = 3). Yeast biological replicates (n = 4) were grown in the corresponding synthetic medium, CSM(T-) or CSM(UT-), for 4 days at 30 °C. All data present error bars, though some are not visible at the plotted scale.

We generated three background yeast strains with different promoters to express npgA: pADH2, pHSP26, or pTDH3; their transcriptional activities are all well documented in minimal media⁶⁵ (Tables S1–S5). Moreover, each of our engineered background strains has two copies of npgA chromosomally integrated into the vacuolar protease PEP4 and PRB1 loci, rendering our yeast strains vacuolar protease-deficient and ensuring the functional expression of heterologous PKSs.^{33,39,45,48,49}

Contrastingly to our first-generation promoter combinations where we used two high-copy 2μ plasmids (URA3 and TRP1-based auxotrophic markers) to express adaA–D, we used a single 2μ plasmid (TRP1) to achieve TAN-1612 production in minimal media, freeing the URA3-based plasmid for further co-expression of other relevant enzymes. Furthermore, using only one plasmid could help prevent copy number heterogeneity issues that could arise from potentially losing adaA–D via homologous recombination between plasmids (pADH2-CDS-tADH2, where CDS = adaA, adaB, adaC, or adaD) and the ADH2 locus in the yeast genome (pADH2-ADH2-tADH2).³³

We used an in-house Golden Gate Assembly (GGA) system based on the Yeast Golden Gate (yGG)⁶⁷ and Yeast Toolkit (YTK)⁶⁸ cloning methodologies to express adaA–D in a 2μ TRP1 plasmid (Table S1). Our in-house GGA system acts as a bridge between the yGG⁶⁷ and YTK⁶⁸ methods, facilitating the construction of long biosynthetic pathways like TAN-1612 (~8.3 kb, Figure 2A and Tables S1–S3). To increase TAN-1612 production in minimal media, we also generated codon-optimized versions of adaA–D (referred to as “adaA–D^opt,” Figure 2A and Tables S1–S3) since heterologous gene codon optimization has improved target molecule production in S. cerevisiae.^29–32,46

The adaA–D genes were each expressed from one of four arbitrarily assigned promoters of different strengths and cell growth-phase activity profiles in minimal media—pCCW12, pHSP26, pTEF1, or pTDH3⁶⁵—yielding two different plasmids #1356 (adaA–D) and #1359 (adaA–D^opt) (Table S1). Next, these two plasmids were each transformed into our three engineered S. cerevisiae BJ5464 background strains where npgA expression was driven either by pADH2, pHSP26, or pTDH3, totaling six TAN-1612 engineered strains (Figure 2A and Table S2).

We measured these strains’ TAN-1612 production over 96 h in minimal media with our UV–vis assay and determined four out of six produced TAN-1612 in significant yields compared to their control strains. TAN-1612 productions in these four engineered strains were all comparable to each other. Interestingly, codon optimization of the adaA–D gene sequences proved not to significantly influence TAN-1612 production (Figures 2B and S10).

Importantly, we observed that our engineered strains harboring non-codon-optimized adaA–D gene sequences not only produced TAN-1612 in similar amounts independent of the promoter used to express npgA (pADH2, pHSP26, or pTDH3), but they also produced more TAN-1612 than the Li et al.⁴⁵ reference strain, where all of the TAN-1612 genes (npgA and adaA–D) were expressed using pADH2 (Figure 2B). These results suggested successful TAN-1612 production above ~64 μg/L (the lowest TAN-1612 detectable amount in our UV–vis assay’s dynamic range, Figure S5) in minimal media by our engineered BJ5464 strains.

TAN-1612 production for one representative strain was confirmed by comparing it to a standard, purified from engineered A. niger ATCC 1015 + pYR311,⁴⁵ with UPLC-MS (Figure 2C and Supporting Information). TAN-1612 was detected at 0.49 ± 0.09 mg/L (total), composed of comparable quantities outside (0.20 ± 0.08 mg/L) and inside (0.29 ± 0.04 mg/L) of cells by UPLC-MS. Additionally, UPLC-MS analysis displayed no potential TAN-1612 biosynthetic shunt products in minimal media (Figure S11). These findings correlate with Li et al.’s⁴⁵ observations and confirmed TAN-1612 as the only polyketide biosynthesized in the cultures of engineered yeasts when npgA and adaA–D are coexpressed independent of the culturing conditions used to produce TAN-1612.

These comparable concentrations and absence of TAN-1612 shunt products in minimal media validate our UV–vis-based assay method as a quick, straightforward, and reliable screen for TAN-1612 production in whole yeast cells. With UPLC-MS confirmation of TAN-1612 production in minimal media at low mg/L levels for the first time in hand, we continued with further yeast engineering endeavors using our BJ5464 strains to increase our initial TAN-1612 titers in minimal media.

Testing TAN-1612s Toxicity to Yeasts in Minimal Media.

Achieving titers higher than 0.49 ± 0.09 mg/L of TAN-1612 biosynthesized by engineered S. cerevisiae in minimal media is necessary to better facilitate developing tetracycline analogues production in yeasts. However, TAN-1612 is a fungal anhydrotetracycline, which is known to display toxic effects on both eukaryote and prokaryote cell membranes.¹ We therefore sought to determine TAN-1612’s toxicity to yeasts to better inform our engineering approaches to increase its titers.

Previous antimicrobial studies reported that neither TAN-1612 (MIC >64 mg/L) nor its tautomer, BMS-192548 (MIC >100 mg/L), inhibited cell growth of S. cerevisiae.^43,44 However, these studies were performed in complex-rich growth media where growth rate is favored and not in minimal media. Therefore, we sought to evaluate its toxicity to yeasts in minimal media, toward the goal of achieving high TAN-1612 titers in this media.

We added different concentrations of a panel of four tetracyclic molecules (oxytetracycline, tetracycline, anhydrotetracycline, and TAN-1612—Figure S12A) to liquid minimal media cultures of S. cerevisiae and S. boulardii laboratory strains (BJ5464- and YM5016-derived strains, respectively) and followed cell growth for at least 2 days (Figure S12B,C and Supporting Information). We observed neither oxytetracycline nor tetracycline, FDA-approved antibiotics, inhibited cell growth at the concentrations tested (MIC >1000 mg/L). Contrastingly, both TAN-1612 and a commercially available analogue, anhydrotetracycline, greatly inhibited yeast cell growth (MIC <10 mg/L).

The notable toxicity levels elicited by TAN-1612 and anhydrotetracycline to yeasts are likely caused through their lipophilic unionized forms created by the higher degrees of aromaticity and planarity of their B, C, and D rings present in these two anhydrotetracyclines but absent in oxytetracycline and tetracycline (Scheme 1). The lack of planarity in these two tetracyclines prevents them from perturbing yeast cell membranes, thereby rendering these molecules nontoxic to eukaryotes like yeasts.¹

Our initially achieved TAN-1612 production titers, 0.49 ± 0.09 mg/L (total), were below TAN-1612 MIC in minimal media (<10 mg/L), suggesting toxic levels were not reached. However, we observed growth impairment of our engineered strains expressing the TAN-1612 biosynthetic genes, compared to the corresponding null strains (previously transformed with empty plasmids), during the fermentation process (Figure S10). These findings directed our engineering approaches to enhance yeast TAN-1612 tolerance to further optimize its biosynthesis in this media.

Screening of Putative Efflux Pumps for Optimized TAN-1612 Biosynthesis in Minimal Media.

A potential strategy to reduce TAN-1612 toxicity in yeasts in minimal media is to heterologously express efflux protein pumps. Past studies correlated introducing efflux pumps conferring tolerance into S. cerevisiae strains to producing greater titers of different target molecules like alkanes,⁶⁹ alkenes,⁷⁰ statins,⁷¹ and trichothecenes.⁷² Furthermore, secreting TAN-1612 to its surroundings could facilitate downstream purification processes toward both the development of novel tetracycline analogues and the generation of in situ therapeutic platforms using engineered probiotic yeasts.

To this end, we investigated the function of three different putative efflux pumps bioinformatically mined from A. niger ATCC 1015, a native TAN-1612 producer organism,⁴⁵ via the Joint Genome Institute MycoCosm database:⁷³ (i) an ATP-binding (ASPNIDRAFT_185231), (ii) a trichothecene (ASPNIDRAFT_43349), and (iii) a TAN-1612 (ASPNIDRAFT_48051) pump downstream of its biosynthetic gene cluster (Figures 1B and S13). We chose these three putative efflux pumps because phylogenetic analyses with different multidrug resistance proteins showed the efflux pumps clustered with previously characterized transporter proteins from different families (Figure S13 and Table S6).

We constructed non-codon-optimized and yeast-codon-optimized versions of these putative efflux pumps in both URA3 low- and high-copy plasmids for their further expression in BJ5464-derived strains. We moved forward with growth assays in CSM(U-) to determine if the putative efflux pumps conferred resistance to anhydrotetracycline. In the presence of anhydrotetracycline (10 mg/L), we observed the putative TAN-1612 and trichothecene, but not the ATP binding, efflux pumps allowed for yeast cell growth, compared to the corresponding control strains expressing an empty vector that impaired cell growth. Moreover, the hypothetical TAN-1612 efflux pump enabled higher yeast growth than did the trichothecene pump (Figure 3A).

Figure 3.

TAN-1612 efflux pump improves heterologous production of TAN-1612 in S. cerevisiae. (A) Screening of heterologous efflux pumps in S. cerevisiae to alleviate toxicity from 10 mg/L anhydrotetracycline. The heterologous efflux pumps were expressed in engineered BJ5464 strains from low-copy (CEN/ARS) and high-copy (2μ) plasmids using different DNA recodes: non-codon-optimized and yeast-codon-optimized. Growth assays were performed in CSM(U-). Cell growth (starting OD₆₀₀ = ~0.05) was recorded after 2 days at 30 °C, 800 rpm, and are depicted as arbitrary units (a. u.). Each symbol represents a biologically independent replicate (n = 5). (B) Representative TAN-1612 yeast producer strains identified from a TAN-1612 efflux pump library. TAN-1612 production per cell calculated as TAN-1612 OD₄₄₅/OD₆₀₀ values was measured for different engineered BJ5464-derived strains expressing the TAN-1612^opt efflux pump in a low-copy URA3 plasmid and the adaA–D (non-codon-optimized and yeast-codon-optimized) genes in a high-copy TRP1 plasmid after 4 days at 30 °C. Each symbol represents a biologically independent replicate (n = 4). In some cases, no symbol(s) is shown because its value is below the quantifiable threshold. All data present error bars, though some are not visible at the plotted scale.

By identifying a putative TAN-1612 efflux pump that reduces anhydrotetracycline toxicity to yeasts, a baseline for the biosynthesis optimization of TAN-1612, a fungal anhydrotetracycline, in yeasts was established. Therefore, we constructed a TAN-1612 efflux pump library by transforming a CEN/ARS URA3 plasmid, expressing the putative codon-optimized TAN-1612 efflux pump gene (TAN-1612^opf), into our engineered BJ5464 strains from the second-generation promoter library, expressing the biosynthetic genes either adaA–D or adaA–D^opt in 2μ TRP1 plasmids (Tables S1 and S2).

After 96 h of growth in CSM(UT-), we measured TAN-1612 production per cell in this library against strains without the efflux pump. We identified #1379 as the highest TAN-1612 producer strain from this TAN-1612 efflux pump library (Figure 3B). Moreover, all strains co-expressing the efflux pump enhanced TAN-1612 production independent of adaA–D codon optimization (Figure S14).

Based on the results from this library, we concluded TAN-1612 production in S. cerevisiae was significantly improved when all of the genes from the TAN-1612 biosynthetic cluster (adaA, adaB, adaC, adaD, and the putative efflux pump) were coexpressed (Figure 1B), highlighting the use of the TAN-1612 efflux pump in further rounds of TAN-1612 biosynthesis optimization in minimal media.

Combinatorial Optimization of TAN-1612 Biosynthesis in Minimal Media for S. cerevisiae BJ5464 Strains.

We continued optimizing our engineered BJ5464-derived strains for higher TAN-1612 production in minimal media because of its potential to facilitate novel tetracycline antibiotic development. To this end, we pursued two different, yet complementary, approaches: (i) a general combinatorial promoter library-based strategy to regulate TAN-1612 biosynthetic gene expression coupled with (ii) the coexpression of TAN-1612 efflux pump to reduce TAN-1612 toxicity in minimal media.

The promoter library-based approach consisted of testing promoters of different transcriptional strengths to balance the expression levels of adaA–D. It has been demonstrated that screening promoters spanning a wide range of transcriptional strengths are essential to achieve high production levels of heterologous target molecules in yeasts.^34,64,74 Consequently, we used five promoters to drive the expression of adaA–D. These five promoters can be categorized into three different types of transcriptional activities across yeast cell growth: (i) strong-constitutive, logarithmic phase (pCCW12 and pTDH3); (ii) medium-constitutive, logarithmic phase (pPGK1 and pTEF1); and (iii) strong-constitutive, stationary phase (pHSP26).⁶⁵ Terminators with similar transcriptional activities^68,75 were also used for each TAN-1612 gene (Figure 4A, Table S1, and Supporting Information).

Figure 4.

Optimization of TAN-1612 production in engineered S. cerevisiae and S. boulardii strains. (A) Construction of adaA–D biosynthetic libraries by Golden Gate Assembly (GGA). PRO = Promoter, CDS = Coding DNA Sequence, and TER = Terminator. P1–P5 represent the following promoters with different transcriptional strengths: P1–pCCW12, P2–pTDH3, P3–pTEF1, P4–pPGK1, and P5–pHSP26. The following terminator sequences were assigned to each of the ada genes: tPGK1–adaA, tADH1–adaB, tENO1–adaC, and tRPL15A–adaD. Non-codon-optimized and yeast-codon-optimized DNA recodes of adaA–D were used to yield a total theoretical library of 1250 (2 × 5⁴) different biosynthetic pathways. (B) Screening workflow to identify TAN-1612 producer strains. TAN-1612 promoter library from (A) was transformed into engineered S. cerevisiae BJ5464 and S. boulardii YM5016 strains for further screening in 24 or 96 deep-well microtiter plates using UV–vis spectroscopy. (C) Top TAN-1612 engineered yeast producer strains of S. cerevisiae and S. boulardii identified from 14 combinatorial libraries resulting from expressing the adaA–D biosynthetic libraries in a high-copy TRP1 plasmid with (○) or without (◇) the TAN-1612^opt efflux pump in a low-copy URA3 plasmid from (B). TAN-1612 production per cell values were measured for different yeast strains after 4 days in CSM(T-) or CSM(UT-) at 30 °C, 240 rpm. Data represents the mean of n = 4 biologically independent replicates, and error bars show standard deviation. All data present error bars, though some are not visible at the plotted scale.

Next, we used our GGA system to construct a diverse library of 2μ plasmids, where the adaA–D DNA sequences (both non-codon-optimized and codon-optimized for yeast expression) were each expressed using one of these five promoters (Figure 4A). The resultant plasmid libraries, GGL-174 and GGL-175 (Table S1), each with 625 different theoretical combinations, were transformed into six engineered S. cerevisiae BJ5464 strains with two chromosomal copies of npgA expressed under pTDH3, pHSP26, or pADH2, crossed either with the coexpressed TAN-1612^opt efflux pump or without. These S. cerevisiae strains totaled 12 TAN-1612 combinatorial libraries (Tables S2 and S7).

We then screened these libraries with our TAN-1612 UV–vis assay and observed a broad distribution of TAN-1612 production levels, as expected (Figures 4B, S15, and S16). From this, we identified 11 engineered S. cerevisiae strains whose TAN-1612 production values were not only higher than those of their corresponding strain controls but also than that of the Li et al. strain⁴⁵ (Figures 4C and S17).

We characterized the plasmids from these 11 top TAN-1612 producing strains by colony PCR followed by Sanger sequencing to determine which promoter combinations favored heterologous biosynthesis of TAN-1612 in yeasts. Sequencing revealed an apparent over-representation of the strong stationary promoter, HSP26, at the adaA and adaD genes (6 and 5 out of 11, each, respectively) and of the medium-constitutive promoters, PGK1 or TEF1, at the adaC gene (6 out 11, total) in TAN-1612 yeast producer strains (Tables S8–S12).

These results suggested strong expression levels of adaA or adaD during the stationary phase paired with medium levels of adaC expression are required to achieve top TAN-1612 production in yeasts. However, this could also suggest constructing TAN-1612 libraries through GGA had an unintended bias to incorporate adaA/adaD/adaC genes with these promoters into the final TAN-1612 plasmid libraries (Figure 4A).

To rule out any potential bias during the library assembly, we sequenced plasmid libraries GGL-174 and GGL-175. We detected a clear bias for medium-constitutive promoters in adaC during the multigene assembly of the plasmid libraries compared to the theoretical normal distribution expected by chance (Tables S13–S19, χ² test). This observed bias could be attributed to technical error during the manual setup of the intended equimolar mix of all 40 “Level 1” sequenced-verified single transcription units (Figure 4A and Supporting Information).

No significant bias was observed for pHSP26 in adaA or adaD genes within the plasmid library assembly; however, we found there was a significant bias for this promoter once transformed into our TAN-1612 producing yeast strains (Table S19, Fisher’s exact test). The exact relationship between the heterologous production of chemicals and the promoter strengths of the genes involved has been reported as complex³⁴ and could be further studied using various metabolic flux analysis approaches.^63,76–78

Critically, our findings reinforce the importance of using well-characterized biological parts, e.g., promoters, terminators, and DNA sequences (non-codon- or codon-optimized), for metabolic engineering in yeast. As previously reported for the heterologous secondary metabolites production—like of naringenin, β-carotene, or violacein—in yeast, balancing heterologous or endogenous gene expression is crucial, sometimes yielding counterintuitive results, where moderate or low-strength promoters result in higher production titers.^37,64

Our engineered BJ5464 strains—where npgA expression is driven by three different, well-characterized, native yeast promoters pADH2, pHSP26, and pTDH3^65,68,79—enabled the generation of a diverse set of TAN-1612 producers (with or without co-expression of the efflux pump) as demonstrated in our combinatorial library (Figure 4C). This wide variety of engineered strains producing TAN-1612 at different levels is advantageous, especially for potential synthetic biology or therapeutic applications where higher TAN-1612 concentrations would not prove beneficial. This TAN-1612 producer strain diversity highlights the potential use of our engineered BJ5464 strains for the heterologous expression and purification of functional PKSs in minimal media. Furthermore, the engineering approaches used to improve TAN-1612 production in minimal media for engineered S. cerevisiae strains represent a feasible route toward achieving TAN-1612 biosynthesis in other yeast species like in the probiotic S. boulardii.

Engineering the Probiotic Yeast S. boulardii to Produce TAN-1612 in Minimal Media.

As a proof of principle, we sought to engineer S. boulardii to produce TAN-1612 in minimal media because of its potential for therapeutic applications. To this end, we constructed the S. boulardii YM5016 strain with TAN-1612 engineering approaches learned from S. cerevisiae (Figures 2–4). We reprogrammed this probiotic strain to drive the expression of two chromosomally integrated copies of npgA with pTDH3 (Table S2). We chose pTDH3 as the only promoter to express npgA because of the promising TAN-1612 production per cell observed using this promoter across our top S. cerevisiae producer strains (Figures 2–4).

Our TAN-1612 plasmid libraries, GGL-174 and GGL-175 (Table S1), successfully achieved TAN-1612 production in minimal media in our engineered S. cerevisiae BJ5464 strains (Figure 4). Thus, to test S. boulardii’s capabilities as a heterologous host to produce tetracycline analogues like TAN-1612, we transformed these two TAN-1612 plasmid libraries into our engineered S. boulardii strain, totaling two TAN-1612 S. boulardii combinatorial libraries (Tables S2 and S7).

Analogous to our work in engineered S. cerevisiae BJ5464-derived strains, we attempted to co-express the TAN-1612 efflux pump in S. boulardii to reduce any potential TAN-1612 toxicity observed in minimal media (Figure S12C). Interestingly, we did not obtain any efflux pump transformants with the probiotic strain (Tables S1 and S2). Attempting to heterologously express any efflux pump may have induced stress on S. boulardii, causing cell death.

We then screened the two probiotic libraries using our TAN-1612 UV–vis assay and identified one engineered S. boulardii strain (#1407) with TAN-1612 production per cell values higher than both those of its corresponding strain control and the reference strain Li et al.’s.⁴⁵ Furthermore, this TAN-1612 engineered probiotic strain (#1407) showed higher (to #1337 and #1401) or comparable (to #1333, #1350, #1354, #1355, and #1402) TAN-1612 production per cell titers than some of our top TAN-1612 S. cerevisiae producers (Figures 4C, S16, and S17).

Taken together, we demonstrate that large TAN-1612 combinatorial pathway libraries can be generated and functional in S. boulardii, enabling rapid optimization and further development of probiotic yeasts equipped to biosynthesize tetracycline analogues for their potential use as engineered live therapeutics.

UPLC-MS Quantification of TAN-1612 Production in Minimal Media by Engineered S. cerevisiae and S. boulardii.

We identified four strains—in S. cerevisiae (#1333, #1336, and #1474) and S. boulardii (#1407)—as the top TAN-1612 producers from our promoter libraries grown in microtiter plates and screened using our UV–vis assay (Figure 4).

We scaled production up of these top four strains in shake flasks to determine the TAN-1612 titers achieved in minimal media via UPLC-MS. We detected TAN-1612 in comparable amounts in both pellet and supernatants across all of the producer strains, confirming previous results (Figure 2C). The state-of-the-art UPLC-MS method directly supports TAN-1612 production per cell and importantly corroborates with the results acquired through our UV–vis assay that indirectly measures TAN-1612 production, highlighting the UV–vis assay’s use as a straightforward medium-throughput assay screening for TAN-1612 production in whole yeast cultures (Figure 4).

We report here for the first time the production of the fungal anhydrotetracycline, TAN-1612, in the probiotic yeast, S. boulardii, with titers of 6.1 ± 2.3 mg/L through the heterologous expression of the adaA–D genes (Figure 5). These achieved titers correlate with TAN-1612s MIC to S. boulardii (Figure S12C).

Figure 5.

TAN-1612 production of the top engineered S. cerevisiae and S. boulardii strains in shake flasks. TAN-1612 was quantified for the top producer strains from Figure 4C by UPLC-MS and compared to the strain engineered by Li et al. TAN-1612 was extracted from the cell pellets and supernatants of engineered yeast strains grown in the corresponding medium, CSM(T-) or CSM(UT-), after 4 days at 30 °C, 240 rpm. Each symbol represents a biologically independent replicate (n = 4). Black horizontal lines indicate mean values. In some cases, no symbol(s) is shown because its value is below the quantifiable threshold. All data present error bars, though some are not visible at the plotted scale.

Moreover, this engineered S. boulardii strain (#1407) achieved TAN-1612 production in minimal media with comparable amounts to two of three top TAN-1612 S. cerevisiae producers, #1333 (3.7 ± 0.6 mg/L) and #1336 (5.2 ± 4.2 mg/L), highlighting our TAN-1612 approaches designed for S. cerevisiae are readily applicable toward the reprogramming of this probiotic yeast to heterologously biosynthesize tetracycline polyketides (Figure 5).

Further optimization of TAN-1612 biosynthesis in nontraditional yeasts, like S. boulardii, is yet to be explored. For example, high production of TAN-1612 in these yeasts could be achieved by improving the synthetic biology methods explored herein—identifying and co-expressing a viable helper protein (like an efflux pump) to reduce any potential TAN-1612 cell burden or toxicity—coupled with growth optimization approaches involving different media and temperature conditions.³¹

We also engineered S. cerevisiae to produce TAN-1612 at 32.5 ± 5.3 mg/L, representing an ~465-fold improvement over the strain reported by Li et al.⁴⁵ (Figure 5). We achieved these high TAN-1612 titers in S. cerevisiae (strain #1474) by co-expressing the biosynthetic adaA–D genes with the TAN-1612 efflux pump (Figure 1B) that alleviated TAN-1612 toxicity to S. cerevisiae in minimal media (Figure S10). Furthermore, co-expressing the TAN-1612 efflux pump in S. cerevisiae not only achieved high TAN-1612 titers but also reduced its highly variable production observed for engineered strains like #1336, which does not express the TAN-1612 efflux pump (Figure 5).

Heterologous production of TAN-1612 in yeasts at high titers, like those reported herein, has the potential to advance tetracycline analogue total biosynthesis and semisynthesis with TAN-1612 as the starting material. Some highlighted advantages of high TAN-1612 production in yeasts include faster isolation time in fewer steps toward purification from engineered yeasts than from filamentous fungi, like A. niger, which also naturally produces other fungal polyketides,⁸⁰ potentially complicating TAN-1612 isolation.

We routinely purify ~3 to 4 mg/L TAN-1612 (>90% purity as judged by ¹H NMR) from ~1 L of S. cerevisiae #1474 (Figure S19) spanning ~2 weeks, starting from glycerol stock restreaking to NMR characterization, compared to the ca. 3–4 weeks needed when isolating TAN-1612 from A. niger pYR311 (Table S20 and Supporting Information).

CONCLUSIONS

In this study, we engineered different components of the biosynthetic pathway of TAN-1612 leading to an improvement of its production of over 450-fold in yeasts in minimal media, a prerequisite for clean purification processes. Through our optimization approach, we discovered and characterized, for the first time, a putative TAN-1612 efflux pump mined from the genome of A. niger ATCC 1015. This work presents the yeasts S. cerevisiae and S. boulardii as viable heterologous hosts for producing complex fungal anhydrotetracyclines, like TAN-1612, paving the road for tetracycline analogue development to combat antibiotic threats.

The engineering approaches used here have general implications for the heterologous production of other fungal polyketides and natural products-derived compounds⁸¹ in yeasts in minimal media. Synthetic biology is a thriving field, and the approaches presented in this work set the stage for more ambitious, yet-to-be imagined applications for engineered yeasts. For example, we envision engineered yeast strain producers of therapeutic molecules can be part of a sense-response community^82,83 capable of sensing a trigger, like a pathogen, and the yeast responds by killing the trigger by producing and secreting an antimicrobial, like a tetracycline. Further, the engineered yeasts described in this work act as a steppingstone toward clinical applications of yeast synthetic biology for the in situ production and delivery of therapeutics, like tetracyclines, in the human body.

MATERIALS AND METHODS

Strains and Growing Conditions.

S. cerevisiae and S. boulardii strains used in this study are listed in Table S2. For maintenance and propagation, yeasts were grown in YPD: 10 g/L Yeast Extract (Catalogue No. 212750, BD), 20 g/L Peptone (Catalogue No. 211677, BD), and 50 mL of 40% w/v of glucose (Catalogue No. MK-4912–212, Fisher Scientific). For selection purposes and plasmid propagation, Complete Synthetic Medium (CSM) drop-out dextrose (glucose) media was used and prepared as described previously.⁸⁴ For solid media, 20 g/L agar (Catalogue No. 214010, BD) was added to the other components. The fungus A. niger ATCC 1015 transformed with pYR311, producing TAN-1612, was cultured as previously reported.⁴⁵ NEB Turbo Competent E. coli (Catalogue No. C2984, NEB) and/or NEB Stable Competent E. coli (Catalogue No. C3040, NEB) were used as cloning strains and/or for plasmid propagation. Both strains were cultured in liquid LB broth Miller (Catalogue No. 1.10285.0500, Merck Millipore) and grown on solid LB agar plates (Catalogue No. 1.10283.0500, Merck Millipore), both prepared according to the manufacturer’s specifications. If necessary, antibiotics were added to the medium (either liquid or solid). These were used at the following final concentrations: Ampicillin or Carbenicillin (100 μg/mL), Kanamycin (30 μg/mL), Spectinomycin (50 μg/mL), and Chloramphenicol (33 μg/mL). Strains harboring desired plasmids were typically grown overnight from picked single colonies in 6 mL of media in 14 mL Falcon round-bottom polypropylene tubes (Catalogue No. 352059, Corning) at 37 °C at 240–300 rpm.

DNA Manipulation.

Genetic part sequences used in this study are described in Table S3. DNA parts were amplified by polymerase chain reaction (PCR) using Q5 High-Fidelity DNA polymerase (Catalogue No. M0491L, New England BioLabs), resolved using 0.5–0.8% (w/v) agarose gel electrophoresis, and purified using the QIAquick Gel Extraction kit (Catalogue No. 28706, Qiagen). Plasmids were constructed by Gibson Assembly of PCR fragments and/or gBlocks with digested plasmid backbones from Table S1. Gibson isothermal reactions were performed using Gibson Assembly Master Mix from NEB (Catalogue No. E2611L, NEB). A 20–40 bp sequence overlap or an equivalent overlap with a melting temperature (T_M, Nearest Neighbor) of ~65 to 67 °C according to the Oligo Calc web tool (http://biotools.nubic.northwestern.edu/OligoCalc.html) was used for Gibson Assembly. Genetic parts were ordered as gBlocks and/or plasmids synthesized by Integrated DNA Technologies (IDT), Genscript, Genewiz, and/or Twist Bioscience. If needed, CDSs were codon-optimized for S. cerevisiae expression using the JCAT portal (http://www.jcat.de).⁸⁵ “^Opt” indicates yeast-codon-optimized DNA sequences. If present, the following restriction enzyme sites: BsaI, BsmBI, BsbI, and NotI were removed from DNA sequences to be compatible with Golden Gate assembly.⁶⁷ TAN-1612 transcription units (TUs) used in this study were constructed using Golden Gate assembly as described in the Supporting Information. Plasmids were purified from E. coli stocks using the QIAprep Spin Miniprep kit (Catalogue No. 27106, Qiagen). After E. coli minipreps, plasmids were verified by Sanger sequencing provided by Genewiz.

qRT-PCR.

Total RNA from yeast cultures grown in minimal media for 96 h at 30 °C, 240 rpm, was isolated using YeaStar RNA Kit (Catalogue No. R1002, Zymo Research) according to the manufacturer’s instructions. RQ1 RNase-Free DNase kit (Catalogue No. M6101, Promega) was used to remove any potential DNA contaminant. DNA-free RNA was quantified by an Infinite M200 plate reader (TECAN) and cDNA was generated from each DNA-free RNA prep using a High-Capacity cDNA Reverse Transcription Kit (Catalogue No. 4368814, ThermoFisher Scientific). All quantitative PCR (qPCR) reactions were performed in a ViiA 7 Real-Time PCR System (ThermoFisher Scientific) using Power SYBR Green PCR Master Mix (Catalogue No. 4367659, Applied Biosystems) according to the manufacturer’s instructions. Each qPCR reaction contained 20 ng of cDNA. qPCR results were normalized to the housekeeping gene ACT1. Results were quantified by the relative 2^−ΔCt method.⁸⁶ qPCR primers (Table S21) were designed using the PrimerQuest Tool, selecting an amplicon length between 90 and 100 bp and a T_M of 62 °C.

Genome Mining, Bioinformatic Analysis, and Experimental Characterization of Putative Efflux Pumps.

DNA sequences of putative efflux pumps were mined from the genome of A. niger ATCC 1015 via the Joint Genome Institute MycoCosm database (https://mycocosm.jgi.doe.gov/mycocosm/home)⁷³ and ordered as gBlocks/plasmids from IDT or Twist Bioscience. Protein sequences of the multidrug resistance transporters used for the construction of the phylogenetic tree were obtained from UniProtKB.⁸⁷ The phylogenetic tree (Figure S6) was constructed with the multiple sequence alignment of the proteins of interest obtained with Clustal Omega and Simple Phylogeny tools from the European Bioinformatics Institute.⁸⁸ Both tools were used with their default settings. The software FigTree v1.4.4⁸⁹ was used to display the tree. Efflux pump TU plasmids were assembled, transformed into S. cerevisiae BJ5464-derived strains, and screened for their capability to recover yeast cell growth in the presence of the toxic anhydrotetracycline (a commercially available, TAN-1612 analogue) and/or TAN-1612 as described in the Supporting Information.

Yeast Transformation and CRISPR/Cas9-Mediated Genome Integration.

S. cerevisiae and S. boulardii transformation and selection of correct transformants were performed as previously described.⁹⁰ All genomic edits were performed using CRISPR/Cas9 via a two-plasmid system consisting of CRISPR/Cas9 and gRNA expression plasmids. Genome editing in S. cerevisiae BJ5464 (ATCC 208288)⁹¹ and BJ5464-NpgA⁹² was performed according to the “Quick and easy CRISPR engineering in S. cerevisiae” protocol.⁸³ gRNA expression plasmids were generated by Gibson assembly of corresponding gRNA sequences (Table S4) into pWS082 (Table S1). pWS158, pWS172, and/or pWS173 were used as Cas9 expression plasmids (Table S1). For genome editing in S. cerevisiae, the Cas9 and gRNA expression plasmids were first digested with BsmBI. Then, ca. 100–150 ng of the linearized Cas9 plasmid and ca. 200–250 ng of each linearized gRNA plasmid were co-transformed with linearized donor DNA fragment (≥2 μg) into yeast cells. Genome engineering in S. cerevisiae var boulardii YM5016 was carried out using the CRISPR/Cas9 system developed by DiCarlo et al.⁹³ gRNA sequences (Table S4) were Gibson assembled into p426-pSNR52-gRNA.CAN1.Y-tSUP4 (Table S1). The Cas9 expression plasmid used was pAV115-pTEF1-Cas9-tCYC1⁸² (Table S1). For engineering S. boulardii using the CRISPR/Cas9 system, yeast cells were first transformed with the Cas9 expression plasmid (ca. 1.5–2 μg), followed by a co-transformation of the gRNA expression plasmid (500–700 ng) and linearized donor DNA fragment (≥2 μg). Correct integration into the yeast genome was verified by colony PCR (cPCR) using Thermo Scientific Phire Plant Direct PCR Kit (Catalogue No. F130WH, Thermo Scientific) followed by enzymatic purification with ExoSAP-IT (Catalogue No. 78-201-1ML, Fisher Scientific) for further Sanger sequencing.

Determination of TAN-1612 Production Per Cell in Microtiter Plate Format.

The TAN-1612 production per yeast cell absorbance assay was developed as described previously.^84,94 Briefly, to apply this method on TAN-1612 produced by yeast in both YPD- and CSM-derived media, we determined the appropriate sensitive and robust wavelengths⁸⁴ by obtaining the absorbance spectrum of TAN-1612 directly in yeast cells (Figure S1). TAN-1612’s spectrum was obtained by subtracting the optical density (OD) spectrum of TAN-1612 nonproducer yeast strain, BJ5464-NpgA, from a TAN-1612 yeast producer strain, EH-3-54-4 (Figure S2). Based on this spectrum (Figure S2), we chose 445 nm as the sensitive wavelength and 365 and 600 nm as the robust wavelengths. Finally, we adopted the same technical considerations as previously described:⁸⁴ (i) OD measurements recorded at the same time for all of the wavelengths studied, (ii) control wells for each medium, and (iii) correction factor for high cell densities to determine TAN-1612 production per cell (expressed as absorbance of TAN-1612 at 445 nm per cell growth at 600 nm) directly in yeast cells in a microtiter plate format (Figure S3). High OD values that were outside the linear range of the photodetector were corrected as previously reported.⁸⁴ Briefly, we corrected OD values using the following formula to calculate true OD values: ${\text{OD}}_{\text{true}}=\frac{k\times \text{O}{\text{D}}_{\text{measured}}}{{\text{OD}}_{\text{saturation}}-{\text{OD}}_{\text{measured}}}$, where ${\text{OD}}_{\text{measured}}=\text{measured OD},{\text{OD}}_{\text{saturation}}$ = saturation value of the photodetector (3.57 for our instrument, as experimentally determined), and k = true OD at which the detector reaches half-saturation of a given OD (3.16 for our instrument, as experimentally determined).

TAN-1612 Sample Preparation, Isolation, and Quantification.

For sample preparation, yeast strain patches were obtained from glycerol stocks (yeast strains were kept at −80 °C with 20% glycerol final concentration) by streaking on an agar plate of corresponding medium and incubating at 30 °C until colonies appeared (typically 3–5 days). Then, single colonies were patched onto a new agar plate of respective medium and incubated at 30 °C until the patches were fully grown (usually after 2–4 days). Fresh patches of each strain were inoculated into 1–1.5 mL of CSM(T-) or CSM(UT-) and grown until saturation (usually 1.5–2 days) at 30 °C, 240 rpm. OD₆₀₀ was then measured, and cells were diluted accordingly into CSM(T-), CSM(UT-), or YPD to OD₆₀₀ = ~0.05. Samples were grown in 15 mL Falcon round-bottom polypropylene tubes (Catalogue No. 352059, Fisher Scientific), 24-well deep-well plates (Catalogue No. WHA77015110, Sigma-Aldrich), or 96-well deep-well plates (Catalogue No. 89237–526, VWR) at 30 °C, 240 or 1000 rpm for 4 days. TAN-1612 production was reported as the ratio of TAN-1612 absorbance at 445 nm and cell growth (OD₆₀₀), which was recorded every 24 h by transferring 200 μL of resuspended cells onto an assay plate, 96-well polystyrene black plate, clear bottom with lid (Catalogue No. 3603, Corning). Absorption spectra (OD₄₄₅ and OD₆₀₀) were recorded on an Infinite M200 microplate reader (TECAN) and analyzed as described above.

TAN-1612 was isolated from large-scale cultures (ca. 0.5–1 L) of engineered A. niger ATCC 1015 or engineered S. cerevisiae strains as described in the Supporting Information. NMR ¹H spectra of TAN-1612 (Figures S18 and S19) were acquired on a Bruker AV III 500 MHz spectrometer.

Experiments involving the measurements of TAN-1612 extracted from the supernatants of naïve BJ5464 strains previously spiked with different concentrations (0–16 mg/L) of purified TAN-1612 were performed using SFC-MS equipped with a photodiode array (PDA) UV–vis detector to identify TAN-1612 peak (MS(ESI+): m/z calculated for TAN-1612: 414.10; m/z found for TAN-1612: 415.10 [M + H]⁺).

TAN-1612 was quantified via UPLC-MS using purified TAN-1612 to prepare standard calibration curves to correlate the area under TAN-1612 peak (MS (ESI–): m/z calculated for TAN-1612: 414.10; m/z found for TAN-1612: 413.09 [M − H]⁻) to TAN-1612 standards. Standard curves (Figure S20) were made as follows: Different concentrations (0–200 μg/L) of pure TAN-1612 (>90% as indicated by NMR) were mixed in a serial dilution manner, using MeOH as the diluent. For TAN-1612 quantification in engineered yeast strains, TAN-1612 was extracted from yeast cultures’ supernatants and cell pellets using a method modified from Li et al.⁴⁵ Briefly, yeast cultures (~5 mL) previously grown in shake flasks for 96 h at 30 °C, 240 rpm, were spun down at 3300 rpm at 25 °C for 10 min to separate the yeast cell pellets and aqueous supernatants. Then, the supernatants were transferred to a Falcon 50 mL conical centrifuge tube (Catalogue No. 14-959-49A, Fisher Scientific) and mixed with ethyl acetate (EtOAc) (1:1) by placing the Falcon tube on an orbital shaker at maximum speed at 25 °C for 90 min. After that time, the mixture was spun down at 3300 rpm, 25 °C, for 10 min to separate the organic (EtOAc) and aqueous layers. The EtOAc layer was concentrated in vacuum until dry. Yeast cell pellets were mixed with glass beads, acid-washed, 425–600 μm (Catalogue No. G8772, Sigma), and extracted with 1 mL of acetone in a vortex mixer at maximum speed for 90 min at 25 °C. The organic phase (acetone) was separated and evaporated to dryness. EtOAc and acetone extracts were resuspended in methanol (MeOH) and filtered using Acrodisc syringe filters with wwPTFE membrane (Catalogue No. 28143-982, VWR) before UPLC-MS analysis. UPLC-MS measurements were performed at the Columbia University Chemistry Department Mass Spectrometry Core on a Waters Xevo G2-XS QToF mass spectrometer equipped with an H-Class Plus UPLC and a LockSpray source electrospray ionization (ESI) probe using negative-mode electrospray ionization. Samples were analyzed with an Eclipse Plus C8 4.6 × 50 mm², 1.8 μm column (Agilent Technologies) with injection volumes of 3 μL using a gradient of 70% acetonitrile (v/v in water, 0.2% formic acid) to 100% acetonitrile (v/v in water, 0.2% formic acid) over 4 min with a flow rate of 0.5 mL/min.