In Situ Biomanufacturing of Small Molecules in the Mammalian Gut by Probiotic Saccharomyces boulardii

Abstract

Saccharomyces boulardii is a probiotic yeast that exhibits rapid growth at 37 °C, is easy to transform, and can produce therapeutic proteins in the gut. To establish its ability to produce small molecules encoded by multigene pathways, we measured the amount and variance in protein expression enabled by promoters, terminators, selective markers, and copy number control elements. We next demonstrated efficient (>95%) CRISPR-mediated genome editing in this strain, allowing us to probe engineered gene expression across different genomic sites. We leveraged these strategies to assemble pathways enabling a wide range of vitamin precursor (β-carotene) and drug (violacein) titers. We found that S. boulardii colonizes germ-free mice stably for over 30 days and competes for niche space with commensal microbes, exhibiting short (1–2 day) gut residence times in conventional and antibiotic-treated mice. Using these tools, we enabled β-carotene synthesis (194 μg total) in the germ-free mouse gut over 14 days, estimating that the total mass of additional β-carotene recovered in feces was 56-fold higher than the β-carotene present in the initial probiotic dose. This work quantifies heterologous small molecule production titers by S. boulardii living in the mammalian gut and provides a set of tools for modulating these titers.

Graphical Abstract

Next-generation engineered probiotics offer unique attributes for treatment of disease.¹ By leveraging decades of advances in metabolic engineering and microbial ecology,² these organisms convert unused dietary material to therapeutic molecules directly in the gut.^3–5 Furthermore, integrated sensors enable drug synthesis to be tailored to the severity and location of disease.^6–9 Coupled with the manufacturing and distribution infrastructure in place for traditional probiotics, engineered probiotics promise to reduce drug cost while improving delivery specificity.¹⁰ To date, almost all engineered probiotics have been members of the bacterial domain of life due to their high numerical abundance in the gut^11,12 and ease of engineering.^7,13,14 For example, Escherichia coli Nissle 1917 has been engineered to treat phenylketonuria (PKU),^13,15 hyperammonemia,¹⁶ cancer,¹⁷ and bacterial infections.¹⁸ Unfortunately, bacteria are susceptible to antibiotics and bacteriophage,¹⁹ and have difficulty producing high levels of post-translationally modified proteins.²⁰

In addition to bacteria, a diverse fungal population exists in the human gut.^21–24 While the number of fungal cells in the gut is much lower than bacteria, fungal cells are ~100-fold larger, indicating that their biomass and role in the gut may be larger than metagenomic surveys indicate.^23,25 Indeed, commensal fungi play important roles in the development of inflammatory bowel diseases and also provide protective heterologous immunity to pathogens by training the immune system.^26,27 Furthermore, fungi are not susceptible to bacteriophage and are easily engineered to secrete high titers of post-translationally modified proteins. Thus, in this study, we focused on engineering the only fungal probiotic that is approved by the FDA: Saccharomyces boulardii.²⁸

S. boulardii (Sb) was originally isolated from lychee and mangosteen in 1923 by Henri Boulard and has been used to treat ulcerative colitis, diarrhea, and recurrent Clostridium difficile infection.^29,30 Sb is closely related to Saccharomyces cerevisiae, indicating that it may be similarly amenable to engineering.^31,32 Supporting this, S. cerevisiae (Sc) expression vectors can be transformed and propagated in Sb, and CRISPR/Cas9-mediated genome editing is functional in both strains.³³ However, compared to Sc, Sb better tolerates low pH and grows more rapidly at human body temperature.^31,34,35 These properties position Sb as a unique chassis for the delivery of therapeutics to the gut. Indeed, Sb has been engineered to secrete human synthetic lysozyme³³ and HIV-1 Gag³⁶ in culture, as well as IL-10, atrial natriuretic peptide, and nanobodies targeting the C. difficile toxins in mouse models of disease.^37–39 However, the in situ biomanufacturing of small molecules, an important therapeutic group for which yeast possesses distinct advantages relative to bacteria,⁴⁰ has not yet been demonstrated for Sb. To enable balanced, high metabolic fluxes through small molecule biosynthesis pathways, finely tuned gene expression is critical. Therefore, in this work we sought to elucidate the genetic strategies governing tunable biomolecule production in Sb, compare the activities of genetic parts between Sb and Sc, and compare in vitro and in vivo small molecule production titers.

We began by performing the most comprehensive analysis of engineered gene expression to date in Sb, characterizing a library of Sc genetic parts (promoters, selective markers, origins of replication, and terminators)⁴¹ in plasmid-based and genomic contexts. We then applied these parts to construct for the first time Sb strains exhibiting high levels of vitamin precursor (β-carotene) and drug (violacein) production. Next, we directly compared the residence times of Sb in germ-free, antibiotic treated, and conventional mice. Finally, we investigated the colonization profile and biomolecule production capacities of β-carotene-producing Sb in the monocolonized mouse gut, demonstrating in situ production of a heterologous small molecule for the first time in this strain. This work therefore establishes our ability to engineer Sb to produce small molecules in the mammalian gut.

RESULTS AND DISCUSSION

Selective Marker and Origin of Replication Tune Gene Expression over a Wide Range.

Due to the genetic relatedness of Sb and Sc, we first asked whether synthetic parts that are commonly used in Sc would function in Sb. If so, it would indicate that the large toolbox that has been developed for Sc engineering could be applied to probiotic Sb. We initially focused on plasmids because they enable rapid prototyping. Unlike the genome, gene expression on plasmids is impacted by the origin of replication and selectable marker.⁴² We therefore sought to understand the interplay between origin and marker in determining engineered gene expression in Sb. We used a model fluorescent protein (mRuby2) to facilitate gene expression measurements, as it exhibited a high signal-to-noise ratio (Figure S1). We generated expression cassettes containing pTDH3, mRuby2, and tTDH1 in combination with 7 selective markers (URA3, HIS3, TRP1, KanMX, ZeoR, NatR, and HygR) and two origins of replication (CEN and 2μ). The CEN we used was derived from Sc and differs from the most closely related Sb CEN at 5 positions. We transformed the cassettes containing auxotrophic markers into the corresponding auxotrophic strains (SbΔURA3, SbΔHIS3, and SbΔTRP1), and antifungal markers into wild-type Sb. We achieved maximum fluorescence using URA3-2μ and minimum using KanMX-CEN at 100 μg/mL G418, collectively enabling a 100-fold range in gene expression by varying the plasmid “backbone” alone (Figure 1a, Figure S2a). The ability to tune plasmid-based gene expression without changing the promoter may provide a convenient way to modulate biomolecule output while conserving regulatory logic.

Figure 1.

Plasmid-based gene expression enabled by S. boulardii (a) markers and origins of replication, (b) promoters, and (c) terminators in vitro. (a) 3 auxotrophic (URA3, HIS3, TRP1) and 4 antifungal (KanMX, ZeoR, NatR, HygR) markers were cloned in plasmids containing the TDH3 promoter, mRuby2 fluorescent protein and TDH1 terminator. Centromeric (CEN6/ARS4) and episomal (2μ) origins were cloned in combination with the 7 selective markers. Plasmids with antifungal markers were tested at 3 different antifungal concentrations (50, 100, and 200 μg/mL). (b) 18 constitutive promoters were cloned 5′ of the mRuby2 fluorescent protein and the TDH1 terminator. (c) 5 terminators were cloned 3′ of 3 different promoters at different strengths (pTDH3-strong, pRNR1-medium and pREV1-weak) and the mRuby2 fluorescent protein. Absolute fluorescence values were obtained via flow cytometry for 3 biological replicates (n = 3). Absolute fluorescence values were normalized to fluorescence values of cells harboring the spacer plasmid. The height of the bars represents the normalized mean fluorescence value of 3 biological replicates and the black dots show the normalized mean of each biological replicate.

The Impact of the Origin on Gene Expression Is Marker-Dependent.

We next asked whether certain markers or origins had consistent effects on gene expression. For auxotrophic markers, 2μ always yielded higher fluorescence than CEN (two-way ANOVA, p < 0.05) (Figure 1a). This is consistent with Sc, in which 2μ plasmids with auxotrophic markers are maintained at 28–58 copies per cell, whereas CEN plasmids are maintained at only 4–8 copies per cell.⁴² However, apart from HygR (two-way ANOVA, p < 0.0001), plasmids containing antifungal markers were not impacted by the origin at the antifungal concentrations tested (two-way ANOVA, p > 0.05). For the CEN origin, using NatR at the maximum antifungal concentration yielded the highest fluorescence (two-way ANOVA, p < 0.05). However, for the 2μ origin, all auxotrophic markers gave higher fluorescence levels than any antifungal marker at the antifungal concentrations tested (two-way ANOVA, p < 0.0001). Taken together, our data indicate that gene expression in plasmids containing auxotrophic markers (but not antifungal markers) is sensitive to the choice of origin, and that selective marker significantly impacts gene expression for both CEN and 2μ origins. These data collectively point to environmental conditions (e.g., antifungal concentration) as strong mediators of plasmid-based gene expression in Sb, indicating that measuring plasmid-based gene expression in vivo (where conditions are very different than in shake flasks) in future work would be very interesting.

Plasmid-based Gene Expression Noise Is Marker- and Origin-Dependent.

Cell-to-cell variability in gene expression can have dramatic impacts on overall production titers.⁴³ Therefore, it is desirable to design engineered probiotics to exhibit minimal gene expression noise. Previous work in Sc showed that cell variability is a function of promoter, terminator, origin, and selective marker.⁴⁴ We thus calculated the population-level coefficient of variance (CV) in the fluorescence values we obtained for Sb. We found that marker, origin, and marker-origin interactions had significant impacts on expression noise (two-way ANOVA, p < 0.05). Specifically, CEN plasmids exhibited less cell–cell variability than 2μ plasmids (two-way ANOVA, p < 0.05) (Figure S2c). Additionally, expression noise in auxotrophic markers (TRP1, HIS3, and URA3) was similar to each other (two-way ANOVA, p > 0.05), while it varied among antifungal markers (Figure S2c). Compared to other selective markers, plasmids with the antifungal marker ZeoR had the highest gene expression noise (two-way ANOVA, p < 0.01). Increasing the concentration of applied antifungal decreased the gene expression noise (two-way ANOVA, p < 0.01) (Figure S2c). These data suggest differences in the ways that the auxotrophic and antifungal markers interact with origins of replication to control gene expression, and offer suggestions for designing genetic circuits that reduce plasmid-based gene expression variance. While our results show that individual Sb cells exhibit stochastic gene expression, population coefficients of variance are similar to Sc harboring similar plasmids.^45–47 Nevertheless, we believe that the tools developed to reduce gene expression noise in Sc such as RNAi, Noise Tuner,⁴⁴ and Transcriptional Regulators⁴⁸ can be applied to Sb. Looking forward, because these measurements were obtained in shake flasks, where conditions are relatively homogeneous, we anticipate that performing future experiments in the gut, with its substantially more numerous environmental niches, would yield higher noise values.

S. cerevisiae Promoters Largely Maintain Their Relative Activities in S. boulardii.

Next, we quantified the ability of 18 constitutive Sc promoters to express mRuby2 in Sb. Each promoter had a homologue in Sb, with percent nucleotide identities ranging from 97% to 100% (Table S1). In these experiments, vectors contained the 2μ origin and the URA3 auxotrophic marker to maximize the range over which changes to fluorescence could be detected (Figure 1a). Each promoter drove the expression of mRuby2 and transcription was terminated by tTDH1. Each promoter yielded different mean fluorescence values (two-way ANOVA, p < 0.0001), and these promoters collectively modulated fluorescence levels over a 40-fold range (Figure 1b, Figure S3). While the relative activities of 16/18 promoters in Sb were similar to reported values in Sc (R² = 0.79), two promoters (pALD6 and pRNR2) showed 10-fold higher activity in Sb than in Sc (Figure S4). While the precise roles of these genes in the probiotic properties of Sb are unknown, Ald6p is involved in the detoxification of aldehyde molecules to their corresponding acid,⁴⁹ and Rnr2p catalyzes the rate-limiting step in dNTP synthesis.⁵⁰ Collectively, these data curate a set of promoters enabling precise tuning of gene expression in Sb, and indicates that constructs developed for Sc will exhibit similar performance in Sb.

Terminators Modulate Gene Expression in a Promoter-Dependent Manner.

Terminators regulate expression in yeast by modulating mRNA stability, translational efficiency and mRNA localization.⁵¹ Furthermore, a terminator’s effect on expression can depend on the strength of the promoter or the sequence of the expressed gene.^52,53 Therefore, we were interested in defining the effect of terminator on gene expression in Sb. We cloned 6 terminators (tENO1, tENO2, tSSA1, tPGK1, tADH1, and tTDH1) into mRuby2 expression constructs driven by strong (pTDH3), medium (pRNR1), or weak (pREV1) promoters. We found that these terminators could collectively modulate gene expression by 1.77, 1.81, and 1.64-fold in constructs containing pTDH3 (tSSA1 vs tTDH1), pRNR1 (tSSA1 vs tTDH1), or pREV1 (tSSA1 vs tTDH1), effectively increasing the expression range beyond that enabled by varying the promoter alone (40-fold, Figure 1b). In particular, using a strong promoter (pTDH3) with tTDH1 yielded 80-fold higher expression than a weak promoter (pREV1) with tSSA1. While the total range of expression observed at the population level varied over 3–4 orders of magnitude (Figure S5), the ability to modulate population expression means over an 80-fold range is meaningful, as the total dose of a delivered biomolecule will be directly proportional to this mean. Relative to tSSA1 and tTDH1, we found that tENO1, tENO2, tADH1, and tPGK1 significantly decreased gene expression in constructs containing the strong pTDH3 promoter (two-way ANOVA, p < 0.005) while for pRNR1 and pREV1, the differences in gene expression we observed among terminators were not statistically significant (two-way ANOVA, p > 0.05) (Figure 1c, Figure S5). Taken together, these data suggest that varying the terminator is a valuable way to fine-tune gene expression, particularly when certain regulatory characteristics of the promoter (e.g., responsiveness to growth phase or gut conditions) must be maintained.

The S. boulardii Genome Can Be Efficiently Modified by CRISPR-Cas12a Genome Editing.

Genomic integration of synthetic constructs can sidestep undesirable issues associated with the use of plasmids in the gut. These issues include plasmid instability, spreading to other microbes, collateral damage to the microbiota if antimicrobials are used for plasmid selection, and the increased metabolic burden of maintaining multicopy plasmids. Further, genome integration increases the uniformity of gene expression throughout the engineered cell population.⁵⁴ However, different genomic regions are known to support varying levels of synthetic construct expression, potentially due to nucleosome positioning,⁵⁵ and integrating material into the genome can be difficult for diploid organisms.⁵⁶

To characterize and overcome inefficiencies associated with genome integration, we compared three different editing methods in the diploid Sb: unassisted linear dsDNA integration (dsDNA integration) and genome editing assisted with the CRISPR nucleases SpCas9 or LbCas12a. We tested the ability of these methods to integrate a linear expression cassette containing mTurquoise2 in three different loci (INT1, 2, and 3). In the case of unassisted dsDNA integration, 4/12 transformants did not contain the expected mTurquoise2 insertion, while 7/12 colonies yielded multiple fragments after colony PCR (Figure S6). Sequencing revealed that the larger fragment corresponded to the desired integration, while the shorter fragment corresponded to the native genomic locus. This result is likely due to the fact that the Sb genome is diploid, and that integration of one copy is sufficient to confer uracil prototrophy. Therefore, we tested whether CRISPR/Cas systems could achieve higher genome editing efficiency. We reasoned that since CRISPR/Cas systems enhance the proportion of double-stranded breaks in all copies of the targeted sequence, editing of all chromosomal copies would be a preferred survival strategy. Three sgRNAs for SpCas9 and three gRNAs for LbCas12a were designed for each integration site. Targeted loci were followed by a 5′-NGG-3′ protospacer adjacent motif (PAM) for SpCas9 and preceded by a 5′-TTTV-3′ PAM (V = A, C, G) for LbCas12a. Figure S7 shows the plasmid designs we used for CRISPR-based editing. Table S2 shows integration locations (INT1–INT5) in Sb’s genome. These integration sites were chosen based on high expression levels from previously characterized chromosomal sites in Sc,^57,58 and Table S3 shows the guide RNA sequences we used.

We tested the editing efficiency at each integration site (using one guide RNA per experiment) as above for unassisted linear DNA integration. Table S4 shows the number of fluorescent colonies we observed for each editing method. Genome editing efficiencies were determined by the number of fluorescent colonies divided by the number of all transformants on a transformation plate as previously described.⁵⁹ Colony PCR of random transformants from each experiment revealed a single band corresponding to either the edited or the unedited version of the locus (Figure S8). Altogether, we were able to generate genome integrants in Sb with 82–88% efficiency using SpCas9 and 93–99% efficiency using LbCas12a (Figure 2a, Table S4). This high efficiency indicates that LbCas12a is an excellent candidate for the integration of libraries into the Sb genome in future work. Finally, these experiments indicate that the many CRISPR-Cas tools developed for engineering Sc may be easily extended to Sb.

Figure 2.

S. boulardii genomic location impacts gene expression, and comparison of genome editing techniques. (a) Editing efficiency at 3 different loci using 3 different methods: CRISPR/SpCas9 (SpCas9), CRISPR/LbCas12a (LbCas12a), and unassisted linear dsDNA integration (dsDNA integration). Three different guides were used for both SpCas9 and LbCas12a for each site. Each dot represents one guide. Three replicate cultures were transformed with the DNA repair template to measure the efficiency of genome editing in Sb via dsDNA integration. Table S4 shows the number of colonies screen for each method. (b) Effect of chromosomal locus on the expression of the fluorescent genes mRuby2 and mTurqouise2. Each dot represents one biological replicate. (c) The effects of promoter (strong, medium, or weak) on gene expression at the same chromosomal locus (INT2). Each dot represents one biological replicate.

Integration Locus Has a Minor Effect on Gene Expression Relative to Promoter Choice.

To determine the extent of locus-dependent effects on gene expression in Sb, we integrated expression constructs for mRuby2 or mTurquoise2 into five different locations in the Sb genome (Table S2) using LbCas12a. These genes were regulated by pTDH3 and tTDH1, and constructs contained URA3 to select for edited strains (Figure S7b). The mRuby2 and mTurquoise2 fluorescent proteins are sequence-divergent, with only 30% amino acid identity. These two proteins therefore enabled a cursory view into the generalizability of expression patterns across genes. Across integration sites, the expression of mRuby2 and mTurquoise2 varied 3.4-fold and 2.6-fold, respectively (Figure 2b, Figure S9). INT4 supported the lowest expression for both genes, while INT2 supported the highest expression levels for mRuby2 and INT3 supported the highest expression level for mTurquoise2. We next asked whether differences in promoter strength observed in the plasmid-based system would be similar for genomically integrated constructs. We therefore expressed mRuby2 and mTurquoise2 from two additional promoters in INT2: pRNR1 and pREV1. Intriguingly, genomic mRuby2 expression varied by 5.9-fold between pTDH3 and pRNR1, compared with 2.8-fold for the same constructs on a 2μ plasmid. Conversely, genomic mRuby2 expression varied by 15.3-fold between pTDH3 and pREV1, compared to 31.2-fold for the same constructs on a 2μ plasmid (Figure 2c). This could indicate that gene expression from high-copy plasmids saturates at high promoter strengths, or that promoter strength varies with copy number due to transcription factor dilution. Additionally, we found that mRuby2 and mTurquoise2 fluorescence values were highly correlated (R² = 0.9945, Figure S9c). This correlation, measured for two genes with 30% amino acid identity, indicates that the expression patterns we observed among the promoters we tested may extend to other genes. Together, this work identifies regions of the Sb genome enabling high-level expression of synthetic constructs. Further, these data show that the impact of integration locus on gene expression is relatively small in Sb compared to that of the promoter.

S. boulardii Facilitates High-Throughput Metabolic Engineering through in Vivo Combinatorial Assembly of Multienzyme Pathways.

Having found that Sb could support genome integration via homologous recombination, we asked if recombination efficiencies were sufficient for in vivo assembly of biosynthetic pathways. In Sc, combinatorial in vivo assembly enables rapid optimization of metabolic pathways.^60–62 Large scale assembly of metabolic pathways directly in Sb would simplify strain development and reveal the potential for Sb as a host for small molecule production. For this demonstration, we focused on the β-carotene and violacein synthesis pathways because these compounds provide a direct colorimetric readout of assembly efficacy (Figures S10a and S10b). While β-carotene is a precursor to the essential vitamin A and violacein has been explored as an anticancer drug,⁶³ our focus was on establishing the first in situ production titers from Sb rather than developing a therapeutic. The β-carotene pathway was also interesting because the optimal expression levels for several genes are not obvious from the biosynthetic reaction map.⁶⁴ In particular, HMG1 diverts flux from cellular acetyl-CoA pools into the β-carotene pathway, thereby reducing the amount of acetyl-CoA available for other processes. crtYB can convert lycopene into the desired product β-carotene, but can also convert neurosporene (a precursor to lycopene) into the side product 7,8-dihydro-β-carotene (β-zeacarotene). crtI operates on several pathway intermediates, including phytoene, ζ-carotene, and neurosporene.⁶⁵ We therefore exploited Sb’s homologous recombination machinery to rapidly construct and reveal efficient β-carotene and violacein synthesis pathways.

We started from our characterized promoter, coding sequence (CDS), and terminator parts and assembled pathways directly in yeast with minimal subcloning (Figure 3a). Our assembly strategy involved two steps. First, golden gate assembly was used to assemble a 9-member promoter library for each gene. These gene-level libraries were then PCR-amplified directly from the Golden Gate reaction mixture and contained homology arms to facilitate in vivo assembly. Purified amplicons were transformed into Sb, where flanking sequences directed homology-dependent assembly of the final pathway with one variant of each transcriptional unit. Figures S10c and S10d show spotted cultures of 28 members from the β-carotene library and 24 members from the violacein library. Colony sequencing revealed that each strain harbored a plasmid containing the expected pathway topology, with each gene driven by one of the 9 selected promoters and the desired terminator.

Figure 3.

Combinatorial assembly of the β-carotene pathway in Saccharomyces boulardii. (a) Overview of assembly steps. Genes in the β-carotene pathway were randomly cloned behind 9 promoters (pTDH3, pCCW12, pHHF2, pTEF1, pTEF2, pHHF1, pHTB2, pPAB1, pRNR2). tSSA1, tADH1, tTDH1, and tENO1 terminators were cloned behind crtE, crtI, crttYB-2, and tHMG1, respectively. The backbone plasmid contains Type1 and Type2 YTK connectors to enable in vivo Sb assembly. Promoter-gene-terminator regions between the two connectors were amplified and the PCR products were transformed to Sb for pathway assembly into a linear backbone via connector-assisted homologous recombination, as described in the Methods section. (b) Sequencing and β-carotene quantification of isolates from yeast transformation plates (Figure S10b). Thirteen out of 28 isolates (Figure S10d) were selected to proceed with sequencing and β-carotene quantification based on stability of colony color after 3 serial passages. Promoter sequences were confirmed by Sanger sequencing with reverse primer binding on the gene of interest (backbone for tHMG1) and forward primer binding on the terminator of the previous gene (backbone for crtE). Quantification of β-carotene was done by high-performance liquid chromatography (HPLC) as described in the Methods section. The spot images were taken from Figure S10d. The number above the spots corresponds to the spot number on Figure S10d. The bar heights represent the normalized fluorescence values in linear scale for each promoter according to promoter characterization work (Figure 1a). The numbers in the center of the spots corresponds to average β-carotene produced by 3 biological replicates of each isolate in one liter of saturated yeast culture.

We next identified strains exhibiting qualitatively stable coloration over three rounds of growth of solid media. Thirteen out of 28 β-carotene-producing strains and 8 out of 24 violacein-producing strains exhibited stable production levels. Due to its relevance as a vitamin precursor, we determined the identity of the promoters driving each gene in the β-carotene pathway and quantified the amount of β-carotene produced by each strain to understand how different combinations of promoters regulate biomolecule production (Figure 3b and Figure S11). The highest β-carotene titer we achieved was 7.27 mg/L from SPC554, corresponding to 0.036 pg/cell. Interestingly, we found that strains with the highest qualitative color intensity and β-carotene titer did not harbor strong promoters for all genes. In fact, many high producers contained moderate or low-strength promoters in at least one of their genes, supporting the utility of a diverse panel of gene expression regulators for production of small molecules in high titers. Additionally, we observed cases where multiple genes in a high-producing strain were driven by the same promoter. This suggested that metabolic burden in these strains was not high enough to select for the emergence of nonproducers via homologous recombination. Our results reveal that, as in Sc, large combinatorial pathway libraries can be easily generated in Sb, enabling rapid optimization of small molecule biosynthesis pathways.

The Mouse Microbiota Restricts S. boulardii Colonization.

Host immunity, intestinal peristalsis, intermicrobial competition, and nutrient availability shape the residence time of microbes in the gut. To parse this complexity, gnotobiotic mouse models are attractive to explore the behavior of microbes in defined gut ecologies. While the residence times of Sb has been measured in humans, rats, and antibiotic-treated mice, its residence time in other mouse models was unknown.^37,39,66,67 Therefore, we first measured Sb’s residence time in the germ-free mouse gut by delivering Sb (10⁸, 10⁷, or 10⁶ colony-forming units) to germ-free mice via oral gavage (Figure S12). No deleterious effects on mouse health were apparent during the experiment. While all dosages yielded similar fecal loads of Sb through day 5 (3 × 10⁶ CFU/g feces), we found that when mice were provided at least 10⁸ CFUs, consistent fecal loads of 10⁷ CFU/g Sb were maintained for over 30 days. On the other hand, mice gavaged with 10⁶ and 10⁷ CFUs showed 3-fold lower fecal titers starting on day 7. Thus, the initial dose of Sb impacts its fecal titer and residence time in germ-free mice. Furthermore, in the absence of microbial competitors, Sb is capable of long-term (>30 days) colonization of the gut.

We next asked whether the presence of microbial competitors would modulate the gut residence time of Sb. To answer this, we turned to conventionally raised mice, with and without simultaneous antibiotic treatment to lower colonization resistance.⁶⁸ In these experiments, we were initially unable to track Sb titers via plating because other commensal fungi outgrew Sb on YPD plates containing 0.125 mg/mL penicillin and 0.25 mg/mL streptomycin. Therefore, we hypothesized that integration of an antifungal resistance gene into the Sb genome would allow easier identification via plating on media containing antifungals. On the basis of the observed minimum inhibitory concentrations (MICs) of various antifungals against Sb (Figure S13), we integrated the nourseothricin resistance gene (NatR) into the Sb genome (INT1). We found that NatR-containing Sb was the only microbe from mouse feces able to grow on YPD plates containing 50 μg/mL nourseothricin, 0.125 mg/mL penicillin, and 0.25 mg/mL streptomycin (data not shown). We then provided conventional mice with 1 mg/mL penicillin and 2 mg/mL streptomycin in their drinking water for 4 days, after which 10⁸ CFU of Sb was delivered once per day for 3 days. Antibiotics were maintained over the course of the experiment (Treatment 1, Figure 4b). We observed ~10⁶ CFU Sb per gram feces during and 1 day after the last gavage, after which Sb titers fell to undetectable levels 3 days after the last gavage (Figure 4e). When antibiotic treatment ceased upon Sb delivery (Treatment 2, Figure 4c), its residence time was shortened, dropping to undetectable levels within 48 h after the last gavage. We then provided the same dosage of Sb to conventional mice in the absence of antibiotic treatment (Figure 4d). In this case, the carriage of Sb was much lower—only reaching ~10³ CFU/g 24 h after each gavage. These results are in line with previous Sb colonization studies performed in mice colonized by a human microbiota⁶⁷ and indicate that competitive microbial growth reduces the residence time of Sb in the mouse gut. In these contexts, Sb-mediated drug delivery will only occur during and for several days after probiotic administration, thereby minimizing dosage outside of desired therapeutic time scales.

Figure 4.

Microbial competition reduces S. boulardii’s residence time in mice. (a) Germ-free mice were gavaged with wild-type Sb (pink dots) on day 0. (b) Conventional mice were treated with 1 mg/mL penicillin and 2 mg/mL streptomycin in drinking water for 4 days prior to 3 Sb::NatR (pink dots) treatments every 24 h (antibiotics were replenished every 4 days). (c) Conventional mice were treated with 1 mg/mL penicillin and 2 mg/mL streptomycin in drinking water for 4 days prior to 3 Sb::NatR (pink dots) treatments every 24 h (antibiotic treatment was stopped on day 0). (d) Conventional mice were treated with 3 Sb::NatR (pink dots) gavages every 24 h. (e) Fecal carriage of Sb in the three mouse models. Fecal samples were collected as shown in the timelines and plated on YPD media for the gnotobiotic model or YPD containing antibiotics for the conventional models. Error bars indicate the standard deviation observed among 4 germ-free mice, 3 conventional mice treated with antibiotic treatment #1, and 3 conventional mice treated with antibiotic treatment #2. In all mouse models, feces (brown dots) were collected prior to Sb gavage during days when both occurred.

Studying colonization profiles in the gut provides insights into the behavior of the host and the introduced microbe(s). While further work is necessary to fully elucidate the effects of Sb colonization on both mice and Sb, our work presents an overview for when each mouse model can be used. Specifically, germ-free mice are appropriate for “proof-of-principle” testing of biomolecule production in the gut environment, since colonization levels were the highest in this setting. On the other hand, conventional and antibiotic treated mice are appropriate for testing the impact of transience on biomolecule production, and for testing the effects of Sb on the native microbiota. Elucidating the interactions between Sb and individual microbiota members via gnotobiotic experiments is of significant future interest.

Engineered S. boulardii Can Produce β-Carotene in the Mouse Gut.

Having identified the Sb strains that produce high-titers of β-carotene in culture, we asked whether Sb can produce β-carotene in the gut, how much β-carotene it can produce, and whether β-carotene-producing yeast exhibits a colonization defect. We chose SPC571 (isolate #23), a strong β-carotene producer (Figure 3b), to administer to 7 germ-free mice, since Sb attained its highest and most persistent colonization in this setting (Figure 4), and preliminary experiments in antibiotic-treated mice showed no detectable β-carotene increase (data not shown). As a control, we administered wild-type Sb to 4 germ-free mice. 10⁸ cells were administered to both groups (Figure 5a,b). For SPC571, the average fecal load of Sb was consistent over the course of the 14-day experiment (5 × 10⁶ CFU/g feces) (Figure 5c). For wild-type, the average fecal load was 10⁷ CFU/g (Figure 5c), which is similar to what we observed during the previous colonization experiment (Figure 4d). While colonization levels for wild-type and engineered Sb were different during the first 48 h postgavage (p < 0.05), longer-term colonization profiles were similar (p > 0.05). These data indicate that β-carotene-producing Sb exhibited a similar ability to colonize germ-free mice, as compared to wild-type Sb.

Figure 5.

Engineered S. boulardii can produce β-carotene in the mammalian gut. (a) Germ-free mice were gavaged with wild-type (WT) Sb on day 0 and followed for 7 days. (b) Germ-free mice were gavaged with β-carotene-producing Sb (SPC571) on day 0 and followed for 14 days. (c) Fecal carriage of WT Sb (gray line) and SPC571 (beige line) in feces and β-carotene load (mg) in feces of mice gavaged with WT Sb (black line) and SPC571 (orange line). Smaller circles and triangles represent the data obtained from individual mice (for WT Sb n = 4, for SPC571 n = 7). The mean fecal levels of Sb and β-carotene are presented as large triangles and dots, respectively. One-way ANOVA with Sidak’s multiple comparisons test were conducted between WT and SPC571 treatment groups. (ns P > 0.05, *P < 0.05 **P < 0.005, ***P < 0.0005, ****P < 0.0001)

Knowing that β-carotene-producing Sb could successfully colonize the germ-free mouse gut, we quantified the amount of β-carotene present in feces from both groups. We saw that colonization by wild-type Sb did not change fecal β-carotene levels from baseline dietary levels (~5 μg/g feces) (Figure 5c). However, after colonization by SPC571, fecal β-carotene levels rose to a maximum of 37 μg/g feces on day 4, corresponding to an increase of 27 μg/g feces above day 0 levels. β-carotene levels in these mice remained higher than precolonization levels throughout the experiment (one-way ANOVA with Sidak’s multiple comparisons test, min p-value <0.0001). Additionally, fecal β-carotene from mice colonized by SPC571 was significantly higher than from mice colonized by wild-type on the same day (one-way ANOVA with Sidak’s multiple comparisons test, max p-value <0.05) (Figure 5c).

Integrating production and colonization data, 5.98 pg of β-carotene per viable Sb cell was produced in the large intestine, 174-fold higher than was obtained for the same strain in culture (Supplementary Text S1). While the cause of this discrepancy is unknown, one possibility is that cell death is occurring within the gut. Additionally, this increase could be due to environmental factors (nutrients, pH levels, slow replication rates) which collectively increase the per-cell titers of β-carotene in the gut. Integrating the additional β-carotene produced by SPC571 over the entire experiment, approximately 194.4 μg β-carotene was synthesized in situ by engineered Sb, 56-fold more than the amount of β-carotene present in the Sb cells gavaged on Day 0 (3.5 μg) (Supplementary Text S2). This indicates that in situ biosynthesis of β-carotene occurred during the experiment. Our results reveal a surprising discrepancy between Sb’s in vitro small molecule titers and its performance in vitro, indicating that more realistic in vitro models could enable rapid prototyping of engineered Sb prior to animal experiments.^69–74

On average, 14.14 μg β-carotene was produced per gram of feces. Extrapolating this value to the average fecal mass produced per day in humans (128 g)⁷⁵ yields 1.8 mg of β-carotene, or 150 μg of retinoic acid equivalent.⁷⁶ While the recommended intake of retinoic acid is 900 μg and 700 μg in men and women, respectively,⁷⁶ we expect that the presence of a microbiota and the substantially longer gut residence time will cause β-carotene production in humans to be different from that measured here in germ-free mice. Furthermore, the bioavailability of β-carotene produced by probiotic yeast in the large intestine is likely to be low, as β-carotene absorption occurs in the small intestine. Therefore, release of β-carotene from the yeast cell, and conversion of β-carotene to immunomodulatory carotenoid derivatives,^77,78 are interesting areas of future study.

Interestingly, all yeast isolated from SPC571-colonized mice throughout the experiment exhibited a striking orange color, even though colonies were isolated on nonselective media. We passaged several of these isolates through three rounds of restreaking on solid selective media, and their orange color was maintained. This indicates that Sb successfully maintained the URA3-marked plasmid encoding the β-carotene pathway, presumably in the absence of selective pressure for plasmid maintenance in the gut environment. Taken together, these data show that Sb is metabolically active and can be engineered to synthesize β-carotene, a precursor to an essential vitamin, over 14 days in the mammalian gut.

CONCLUSION

The gut and the microbes that live within it play a major role in health, and as such are promising targets for small molecule or biologic drugs. Unfortunately, delivering drugs to the gut remains challenging due to the digestive action of the host. Due to its probiotic properties and similarity to the biomanufacturing workhorse Sc, we established a synthetic toolkit for Sb engineering and applied this toolkit to produce a vitamin precursor in the mammalian gut. We first demonstrated tunability of gene expression over 2 orders of magnitude using plasmid-based constructs derived from Sc. We next established >95% efficient genome editing using LbCas12a and revealed up to 3.4-fold variation in gene expression depending on the genomic integration site. We further harnessed Sb’s high rates of homologous recombination to combinatorially assemble pathways for vitamin precursor (β-carotene) and drug (violacein) production, revealing designs that enable up to 7.27 mg/L β-carotene in shake-flask culture. We next determined Sb’s ability to colonize germ-free and antibiotic-treated mice and found that microbial competitors significantly reduce Sb’s gut residence time. Finally, we found that β-carotene-producing yeast can colonize germ-free mice over 14 days and continuously produce β-carotene (194.4 μg total). These results indicate that Sb is a promising probiotic chassis, and point to several areas of future research to further improve its delivery properties, including gut residence time and ability to secrete biomolecules and proteins. We are currently investigating the ability of Sb to release β-carotene to its surroundings via cell lysis or export, as has been demonstrated in related yeast species.^79,80 Taken together, these results expand our ability to tune biomolecule production in the promising probiotic eukaryote Sb.

METHODS

Strains and Culture Media.

Escherichia coli Top10, NEB 5α and NEB 10β were used for plasmid construction and maintenance. E. coli cells were grown in lysogeny broth (LB) (5g/L yeast extract, 10 g/L tryptone, 10 g/L NaCl) at 37 °C supplemented with ampicillin (100 μg/mL), kanamycin (50 μg/mL) or chloramphenicol (50 μg/mL). Saccharomyces boulardii ATCC-MYA796 was used for antifungal marker and origin of replication characterization. Saccharomyces boulardii ΔURA3, provided by Yong-Su Jin, was used for auxotrophic marker, origin of replication, promoter, and terminator characterization, as well as genome editing and metabolic pathway assembly. Saccharomyces boulardii ΔTRP1 and ΔHIS3 strains from the Jin Lab were used for auxotrophic marker and origin of replication characterization. Promoter and terminator characterization experiments were conducted in synthetic media containing 0.67% (w/v) Yeast Nitrogen Base Without Amino Acids (Sigma-Aldrich), 2% Glucose (Fisher Scientific), 1.92g/L Yeast Synthetic Media Drop-Out Mix without Uracil (Sigma-Aldrich). For auxotrophic marker characterization, synthetic media with 0.67% (w/v) Yeast Nitrogen Base Without Amino Acids (Sigma-Aldrich), 2% Glucose (Fisher Scientific), and 1.92% Yeast Synthetic Media Drop-Out Mix without the appropriate nutrient (histidine, tryptophan, or uracil) (Sigma-Aldrich) was used. Antifungal marker characterization experiments were conducted in yeast extract-peptone-dextrose medium (50 g/L YPD broth (Sigma-Aldrich)) supplemented with various concentrations (50 μg/mL, 100 μg/mL and 200 μg/mL) of Geneticin (G418), zeocin, nourseothricin and hygromycin.

Plasmid Construction and Cloning.

A synthetic toolkit (MoClo-YTK) containing yeast parts were gifts from the Dueber Lab (Addgene #1000000061). Characterization plasmids consisted of 8 parts: 2 connectors, promoter, coding sequence, terminator, yeast marker, yeast origin, and E. coli marker and origin. Each of these plasmids were assembled according to Deuber lab YTK protocols via Golden Gate cloning.⁸¹ The Golden Gate reaction mixture, for all plasmids except those containing β-carotene or violacein pathways, was prepared as follows: 0.5 μL of 40 nM of each YTK plasmid (20 fmol), 0.5 μL T7 Ligase (NEB) or 0.5 μL T4 Ligase (NEB), 1 μL T4 Ligase Buffer (NEB) and 0.5 μL BsaI (10 000 U/mL, NEB) and water to bring the final volume to 10 μL. The Golden Gate assembly protocol was performed on a thermocycler with the following program: 30 cycles of digestion (42 °C for 2 min) and ligation (16 °C for 5 min), followed by a final digestion (60 °C for 10 min) and heat inactivation (80 °C for 10 min).

Yeast Competent Cells and Transformations.

We optimized the yeast competent cell preparation and transformation protocol from Gietz et. al for Saccharomyces boulardii.^82,83 For the characterization experiments, we used frozen competent cells, whereas for the genome editing experiments we used unfrozen competent cells prepared on the same day.^82,83 To prepare competent cells, yeast colonies were inoculated into 1 mL YPD and incubated in a shaking incubator overnight at 37 °C, 250 rpm. This culture was diluted into fresh 50 mL YPD (approximately OD₆₀₀ = 0.25), and grown for 6–8 h until OD₆₀₀ = 1. The culture was centrifuged at 3000g for 5 min at room temperature to pellet the cells. After decanting the media, the pellet was resuspended in 25 mL of sterile water, and then pelleted at 3000g for 5 min. This pellet was resuspended in 1 mL of sterile water. The cells were then pelleted at 10 000g for 1 min and water was discarded. The cells were then resuspended in 500 μL of frozen competent cell solution consisting of 2.5% (v/v) glycerol (Sigma-Aldrich) and 5% (v/v) DMSO (Thermo-Fisher). 50 μL of this suspension was aliquoted into 1.5 mL microcentrifuge tubes and stored in −80 °C. For transformations, frozen cells were thawed at 37 °C for 1 min. Then, the cells were pelleted by centrifugation at 3000g for 2 min, after which the supernatant was removed. Then, solutions were added in the following order; 260 μL 50% PEG3350 (Fisher Scientific), 36 μL 1 M Lithium Acetate (Sigma-Aldrich), 50 μL of 2 mg/mL single-stranded salmon sperm DNA (Invitrogen, 15632011), 0.1–10 μg DNA, and water to bring the final volume to 360 μL. Ten mg/mL double-stranded salmon sperm DNA from the supplier was diluted to 2 mg/mL and heat treated at 95 °C for 5 min to denature the double-stranded DNA. The pellet was resuspended in the transformation mix by gently mixing with a pipet tip. This transformation mix was incubated at 42 °C for 1 h. This mixture was then centrifuged at 3000g for 1 min and the supernatant was discarded. The cell pellet was resuspended in 1 mL YPD by gently pipetting up and down and this tube was incubated at 37 °C for 1 h. Then, the cell suspension was centrifuged for 1 min at 3000g, resuspended in 25 μL sterile water, and plated on an appropriate growth media.

S. boulardii Colony PCR.

Since Sb has a thicker cell wall than Sc, we developed a new method to perform colony PCR on Sb. Colonies were picked into 25 μL 20 mM sodium hydroxide and incubated at 98 °C for 80 min. Then we centrifuged these reactions at max speed (Mini Microcentrifuge Sargent-Welch) for 10 min and started a 25 μL PCR reaction with 2 μL of this reaction as a template. Otherwise, PCR reactions were performed according to the supplier’s instructions.

Flow Cytometry.

Constructs were inoculated from −80 °C freezer stocks into 750 μL of appropriate media in 96-deep-well plates (VWR International) and incubated at 37 °C for 36 h with shaking at 250 rpm. After 36 h, the cultures were diluted 1:100 in 200 μL of fresh media in 96-well-plates (Costar) and were grown for ~8–10 h at 37 °C. Once the cultures reached OD₆₀₀ values between 0.1 and 0.4, flow cytometry was performed using a MACSQuant VYB (Miltenyi Biotec). Cells containing empty vector were used as a control. The voltage of the channels was adjusted in order to tune the gain of the detectors. The voltage values applied to FSC, SSC, Y1 (PE-A, excitation:561, emission: 586/15 nm), V1 (CFP-VioBlue, excitation: 405 nm, emission: 450/50 nm) and B1 (GFP_FITC, excitation: 488 nm, emission: 425/50 nm) channels/filters were 326V, 290V, 396V, 324V, and 265V respectively. No gating was performed. Flow cytometry data analysis was conducted on the FlowJo program (FlowJo LLC). The mean fluorescence values obtained from the flow cytometry for each transcriptional construct were then normalized to background fluorescence (i.e., the cells with an empty vector).

Mouse Experiments.

All mouse experiments were approved by the NC State University Institutional Animal Care and Use Committee (IACUC).

Germ-Free Mice.

Colonization Experiments (Figure 4).

6–10 week old male and female germ-free C57BL/6 mice were monocolonized by wild type S. boulardii. Two mice (males) were gavaged with 10⁶ colony forming units in 100 μL PBS, 8 mice (6 females and 2 males) were gavaged with 10⁷ and 4 mice (2 females and 2 males) were gavaged with 10⁸ Sb cells on day 1 and fecal samples were collected on days 2, 5, and 7. Fecal samples from mice housing the 10⁸ inoculum were also collected on day 30.

β-Carotene Production Experiment Negative Control (Figure 5).

20–28 week old male germ-free C57BL/6 mice were monocolonized by wild type Sb. All 4 mice were gavaged with 10⁸ Sb cells on day 0 and fecal samples were collected on days 0 (prior to gavaging with Sb) 2, 4 and 7.

β-Carotene Production Experiment (Figure 5).

12–14 week old male germ-free C57BL/6 mice were monocolonized by recombinant Sb capable of synthesizing β-carotene. All 7 mice were gavaged with 10⁸ β-carotene producing Sb cells on day 0. Prior to gavaging on day 0, fecal samples were collected from all mice to capture baseline β-carotene levels. After the gavage took place, fecal samples were collected on days 2, 4, 7, 9, 11, and 14. On day 14 after the fecal sample collection, the mice were sacrificed and the cecum and the colon contents were collected.

Stool Cultures.

Two pieces of stool were added to preweighed 1.5 mL microcentrifuge tubes and then weighed again to determine the fecal mass. Cecum and colon contents were collected in preweighed 5 mL reagent tubes and then weighed again to determine the intestinal content mass. Fecal and intestinal matter was then resuspended in 1 mL PBS per 100 mg feces/intestinal matter. Fecal and intestinal matter suspensions from monocolonized mice were placed on solid YPD media. Plates were incubated in 37 °C for 2–3 days until the colonies appeared.

Conventional Mice.

Six-week old female C57BL/6J mice were obtained from Jackson Laboratories and hosted at the NCSU Biological Resources Facility (BRF) for 3–4 days before experiments. Mice were housed in groups of three and their cages were changed before treatment with antibiotics (if necessary) and before treatment with Sb::NatR or water. Antibiotic-treated mice were provided 1 mg/mL penicillin and 2 mg/mL streptomycin in their drinking water for 4 days starting on day −3. On day 1, drinking water was either replenished with antibiotic solution or changed to fresh water. Mice were then gavaged with 10⁸ CFU Sb or water once every 24 h for 3 days. Fecal samples were collected every 24 h from day 1 to day 5.

Stool Cultures.

1–2 pieces of stool were collected in preweighed 1.5 mL centrifuge tubes and then weighed again to determine fecal mass. Fecal matter was then resuspended in 1 mL PBS per 100 mg feces. Fecal suspensions from monocolonized mice were placed on solid YPD media, while fecal suspensions from conventional mice treated with antibiotics were plated on YPD media containing 50 μg/mL nourseothricin and 0.25 μg/mL streptomycin. Plates were sealed with parafilm and incubated at 37 °C for 2–3 days.

Genome Editing.

Integration plasmids contained URA3 to aid selection and contained homology arms between 350 to 900 base pairs in length as shown in Figure S7. For DNA homologous recombination without DSB (dsDNA integration), 2 μg of linear donor DNA was transformed to 10⁷ Sb (Sb ΔURA3) competent cells prepared on the same day. For SpCas9, 2 μg of linear donor DNA and 1.5 μg of gRNA plasmids were transformed to 10⁷ freshly prepared chemical competent Sb ΔURA3 cells containing SpCas9 plasmid. For LbCas12a, 2 μg of linear donor DNA and 1.5 μg of gRNA plasmids were transformed to 10⁷ freshly prepared chemical competent Sb ΔURA3 cells containing LbCas12a plasmid.

Overnight cultures of Sb ΔURA3 were grown in YPD only for Homologous Recombination, Sb ΔURA3 containing the SpCas9 plasmid were grown in YPD + 100 μg/mL nourseothricin, and Sb ΔURA3 containing the LbCas12a plasmid were grown in YPD + 200 μg/mL G418. Cultures were subinoculated to prewarmed YPD media with or without antifungals at an OD₆₀₀ of 0.25 and incubated at 37 °C with shaking at 250 rpm until cultures reached an OD₆₀₀ of 0.9–1.0. Cultures were then centrifuged at 3000g for 5 min. Cells were then washed twice with 0.5 and 0.1 volumes of sterile DI water at room temperature at 3000g for 5 min. The cells then were transformed according to the protocol described above, with the exception that heat shocked cells were spun down, resuspended in YPD, and incubated for 3 h at 37 °C. Transformed cells were then spread on Yeast Complete Synthetic Media (CSM) plates without uracil and with or without antibiotics. Plates were incubated at 37 °C for 2–4 days. Integration was verified with colony PCR.

Combinatorial Pathway Assembly.

Prior to library assembly, promoter, gene, and terminator part plasmids were assembled into “entry vector” plasmids by BsmBI assembly. These plasmids had BsaI overhangs and follow the “Yeast golden gate toolkit” standard.⁴¹ Backbones were assembled for subcloning of the transcriptional units. These backbones housed the appropriate homology arms for later final assembly of transcriptional units into a full pathway. BsaI assembly of transcriptional units into the matching backbone was completed by combining 10 fmol of each promoter part plasmid (to a total of 90 fmol) with 100 fmol of gene, terminator, and backbone plasmids. Reaction mixtures included 0.5 μL of BsaI-HF v2, 0.5 μL of T7 DNA Ligase, and 1 μL of T4 DNA ligase buffer (NEB), and underwent 50 cycles of golden gate assembly (2 min at 42 °C, 5 min at 16 °C), before a single 30 min final digestion at 60 °C, and 10 min of heat inactivation at 80 °C. Subsequently, we have noticed that final digestion steps of 5 min at 42 °C are more optimal for BsaI assembly.⁸¹ Transcriptional units were amplified directly from the golden gate mixture using 1 μL of reaction mixture as template in 50 μL Q5 Hot-Start Master Mix reactions (NEB). PCR reactions were performed according to manufacturer’s instructions and cleaned up using clean and concentrate kits (D4004-Zymo Research). 600 nM of each amplified transcriptional unit library was transformed using the lithium acetate protocol described above. Cells were recovered for 1 h in YPD prior to dividing the culture into 5 aliquots of 200 μL and plating on yeast nitrogen base plus amino acids without uracil (Sigma).

Pathway Sequencing.

Random colonies were collected in 1.5 centrifuge tubes and resuspended in 1 mL of 10 mg/mL Lysing Enzymes from Trichoderma harzianum (L1412 Sigma). The cells were incubated in the lysing enzyme at 37 °C for 3 h and then the ZymoPURE Plasmid Miniprep (D4208T) was used to miniprep the plasmids. The Lysing Enzymes from Trichoderma harzianum help digest Sb’s cell wall and produce protoplasts that are easy to miniprep. 2 μL of the miniprep product was then used as a template to amplify the assembled region with primers complementary to the pathway genes or terminators, as the ordering of these was common to all constructs. The PCR product was then sent for Sanger sequencing (Genewiz, NJ, USA) to verify the promoters assigned for each gene.

Antifungal Minimum Inhibitory Concentration Tests in S. boulardii.

Three biological replicates of Sb (ATCC-MYA796) were grown overnight in YPD media. Overnight cultures were subinoculated into YPD media with different antifungal concentrations ranging from 0 μg/mL to 200 μg/mL in a 96 well plate at starting OD₆₀₀ 0.02 and grown for 36 h in a plate reader (BioTek Synergy H1, Shake Mode: Double Orbital, Orbital Frequency: continuous shake 365 cpm, Interval: 10 min). Geneticin (G418), Nourseothricin, Zeocin, Hygromycin B were used in this study.

Extraction of β-Carotene from S. boulardii Cultures.

Three replicates of each β-carotene producing Sb strain were inoculated into 1 mL CSM without uracil (2.5% glucose) and grown overnight in a shaking incubator at 250 rpm at 37 °C. Overnight cultures were then subinoculated (1:100) into 30 mL CSM without uracil (2.5% glucose) and grown for 3 days in a rotary incubator at 250 rpm at 37 °C. After 3 days, 2 × 10⁹ cells were collected by centrifuging at 4 °C at 5000g for 5 min. The media was discarded and the cells were washed with 10 mL water and were centrifuged at 4 °C at 5000g for another 5 min. After the water was discarded, the cells were resuspended in 500 μL water and transferred into ZR Bashing Bead Tubes (Zymo Research, CA). The ZR Bashing tubes were centrifuged for 1 min at 8000g to discard the water. 500 μL acetone was then added into the bashing tubes. The tubes containing the cells, acetone, and lysis beads then transferred to a homogenizer (TissueLyser, Qiagen, Germany) for physical lysis (Lysis duration: 10 min, frequency: 30/s). The cells and the lysates then centrifuged at 16 000g for 5 min at 4 °C. The supernatants (acetone lysate) were transferred to 1.5 mL black microcentrifuge tubes. Fresh acetone was then added to the bashing tubes, and lysis steps were repeated 2 more times, at which point the cell pellets were completely white. β-carotene extracts were dried by speed vac (aqueous solvent, 30 °C for 2 h). Dried β-carotene pellets were reconstituted in 500 μL acetone and stored at −20 °C in the dark.

Extraction of β-Carotene from Stool and Intestinal Samples.

Fecal and intestinal matter suspensions in PBS were centrifuged at 4500g for 20 min. PBS was removed from the fecal and intestinal matter pellets containing recombinant Sb cells. After the PBS was discarded, the pellets were resuspended in 1 mL 1× PBS and were transferred to ZR Bashing Bead Tubes (Zymo Research, CA). The ZR Bashing tubes were centrifuged for 1 min at 8000g to discard the water. 500 μL acetone was then added into the bashing tubes. The tubes containing the cells, acetone, and lysis beads then transferred to a homogenizer (TissueLyser, Qiagen, Germany) for physical lysis (Lysis duration: 10 min, frequency: 30/s). The cells and the lysates then centrifuged at 16 000g for 5 min at 4 °C. The supernatants (acetone lysate) were transferred to 1.5 mL black microcentrifuge tubes. Fresh acetone was then added to the bashing tubes, and lysis steps were repeated 2 more times, at which point the extract solution was clear. Approximately 1.5 mL β-carotene extract solutions were obtained from each sample and the extract solutions were stored at −20 °C in the dark until the analysis.

Detection of β-Carotene by HPLC.

Cell Cultures.

5 μL of each sample was processed in an Agilent 1260 Infinity high performance liquid-chromatography (HPLC) system (Agilent Technologies, CA, USA). 0.002 mM naphthalene solution was used as internal control and pure β-carotene (Thermo Fisher, Waltham, MA) at varying concentrations (mg/mL) 0.00625, 0.0125, 0.025, 0.05, 0.0625, 0.125, 0.25, 0.5, 0.625, 1.25, 2.5, 5) as external control. The channels used for internal control and β-carotene detection were 275 and 453 nm, respectively. The column used in the system was an Agilent HC-C28 Reversed Phase column (4.6 × 250 mm, 50 μL) (Agilent Technologies, CA, USA). The composition of the mobile phase was acetonitrile:methanol:isopropanol (v/v/v: 85/10/5) The flow rate for the mobile phase was 3 mL/min.

Fecal and Intestinal Samples.

After all extractions, the samples were dried completely by nitrogen gas and then reconstituted with 400 μL high performance liquid-chromatography (HPLC)-grade acetone. A 250 μL sample of final extract was injected into the HPLC system (Ultimate 3000, Thermo Fisher, Waltham, MA). Pure β-carotene (Sigma-Aldrich, St. Louis, MO) at various concentrations (μM) 1, 5, 100, 500, 1000, 5000 was used as external control. The channel used for β-carotene detection was 475 nm. The column used in the system was Acclaim C30, 5 μm, 4.6 × 250 mm. The composition of the mobile phase was acetonitrile: methanol: methyl tert-butyl ether (v/v/v: 25:75:0 at time 0–20 min, 15:35:50 at time 20–25.5 min, 25:75:0 at time 25.5–30 min). The flow rate for the mobile phase was 1.0 mL/min.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssynbio.0c00562.

Supplementary figures, tables, and texts described throughout the manuscript (PDF)

Plasmid maps (ZIP)

Strains, plasmids, primers used in each figure (XLSX)