Improvement of a synthetic live bacterial therapeutic for phenylketonuria with biosensor-enabled enzyme engineering

Abstract

In phenylketonuria (PKU) patients, a genetic defect in the enzyme phenylalanine hydroxylase (PAH) leads to elevated systemic phenylalanine (Phe), which can result in severe neurological impairment. As a treatment for PKU, Escherichia coli Nissle (EcN) strain SYNB1618 was developed under Synlogic's Synthetic Biotic™ platform to degrade Phe from within the gastrointestinal (GI) tract. This clinical-stage engineered strain expresses the Phe-metabolizing enzyme phenylalanine ammonia lyase (PAL), catalyzing the deamination of Phe to the non-toxic product trans-cinnamate (TCA). In the present work, we generate a more potent EcN-based PKU strain through optimization of whole cell PAL activity, using biosensor-based high-throughput screening of mutant PAL libraries. A lead enzyme candidate from this screen is used in the construction of SYNB1934, a chromosomally integrated strain containing the additional Phe-metabolizing and biosafety features found in SYNB1618. Head-to-head, SYNB1934 demonstrates an approximate two-fold increase in in vivo PAL activity compared to SYNB1618.

Fig. 1. TCA biosensor description and demonstration.

a Schematic representation of biosensor-regulated reporter expression. In the absence of TCA, the allosteric transcription factor (aTF, labeled sensor in figure) binds the DNA operator site and prevents expression of the reporter gene gfp that encodes for a green fluorescent protein (GFP). PAL activity converts l-Phe into TCA; TCA binds to the sensor aTF, causing a conformational change to relieve DNA binding (repression) and allowing the expression of the reporter gene. Symbols and abbreviations: phenylalanine (l-Phe, yellow filled circles), trans-cinnamic acid (TCA, orange empty circles), PAL (phenylalanine ammonia lyase). b Top: whole-cell activity (see Methods, plate-based) of six StlA PAL variants, normalized to wild-type StlA activity (fold-over wild-type, FOWT, wild-type values can be found in the Source Data). n = 2 biological replicates, individual data points shown. Bottom: per cell GFP histograms of variants from the top panel following growth to saturation with simultaneous stlA (gene encoding for StlA) variant expression, sensor response, and gfp (gene encoding for GFP) expression. Colors are consistent between graphs. n = 1 representative culture per strain with 50,000 cells analyzed. c Enrichment demonstration of designed PAL library. Boxplots showing FOWT activity from different sort populations, with accompanying clonal data points. Boxes extend from the first to the third quartile, with a centerline indicating the median. Upper and lower whiskers extend to the maximum and minimum point, respectively, within 1.5 times the interquartile range of the bounds of the box. Data points outside of the whiskers are considered outliers. Sort/control descriptions: wild-type, EcN harboring wild-type StlA, n = 12 biological replicates normalized to the average of the 12 wild-type production values; mock, library sorted based on cell size and doublet exclusion without GFP-based gating, n = 90 independent colonies from the sorted pool; top 1%, selection of top 1% brightest cells also gated based on cell size and doublet exclusion, n = 90 independent colonies from the sorted pool; mock-mock, mock population retransformed into fresh EcN background with sensor and mock-sorted a second time without GFP selection, n = 90 independent colonies from the sorted pool; top 1%-top 1%, top 1% population retransformed into fresh EcN background with sensor and sorted for its top 1% brightest cells, n = 90 independent colonies from the sorted pool. P values were calculated using Welch’s ANOVA test with Dunnett’s T3 multiple comparisons test.

Fig. 2. Depiction of pop ‘n’ sort methodology.

EcN cells harboring the high-copy sensor system plasmid (engineered aTF expression, with aTF-repressed gfp gene encoding for green fluorescent protein GFP) and a library of low-copy inducible stlA variant expression plasmids were grown together within a pool; different cells in this pool contain unique stlA sequences. Following a pre-culture step in a rich medium, the cells are washed and resuspended in an adapted M9 minimal glucose medium for sensor screening, which simultaneously induces stlA expression and provides the substrate Phe. At this stage, the cells are diluted to an OD₆₀₀ that will result in ~1 cell/droplet loading. These diluted, washed cells in medium serve as the aqueous phase for droplet generation in a fluorinated oil with fluorosurfactant. Because cell loading follows a Poisson distribution, some droplets will be empty and some will contain multiple cells/genotypes at time 0, as shown above. The loaded droplets are incubated until saturation, ~16 h, at which point there are many cells per droplet, and these cells have produced TCA and a per cell GFP signal that is correlated with that TCA production. Since the GFP signal is associated with the cell itself, we can break the emulsions using standard techniques, then sort the cells directly on a FACS for the desired phenotype. See Methods for details on recipes, materials, and protocols.

Fig. 3. Designed and enriched mutation positions.

Positions of interest are highlighted on a homology model generated for StlA using RosettaCM. Active site residues are highlighted in red. a Positions targeted for combinatorial mutagenesis during library construction. Any positions mutated in the library templates (see Methods) are highlighted in green, and additional mutations targeted during combinatorial mutagenesis are shown in pink. b Positions mutated in top hits. Positions mutated in the final SYNB1934 PAL variant are highlighted in yellow, and those mutated in other top hits (indicated with special markers in c, further detail in Supplement) are shown in blue. c A phylogenetic tree was generated by computing the distance between all the sequences using the Blosum62 substitution matrix and constructing the tree by neighbor-joining in Biopython^,. Colors indicate activity level normalized to wild-type (FOWT). Markers: square = S92G_H133F_A433S_V470A, diamond = S92G_H133M_I167K_L432I_V470A, x = A93C_H133F_T322W_Y437N, star-square = I28R_S92G_H133F_V470A, large circle = S92G_H133F_R185E, hexagram = S92G_F109A_H133M_T503E, wild-type = hourglass; all other unique variants are shown as a small circle.

Fig. 4. Activity at different pH or after exposure to low pH.

a Whole-cell TCA production for EcN with wild-type StlA after 0.5 h, 1 h, 1.5 h, 2 h, or 4 h at pH 5, 6, 7, or 8. n = 3 biological replicates with individual data points shown. b Whole-cell PAL activity after 4 h at pH 5, 6, 7, or 8, normalized to wild-type StlA activity (fold over wild-type, FOWT). n = 3 biological replicates ± s.d. c Whole-cell PAL activity after 4 h at pH 7, assayed without pH treatment (control) or after incubation for 1 h at pH 5 (pH treated), normalized to wild-type StlA activity (FOWT). n = 3 biological replicates ± s.d. X axis labels refer to strain IDs, EcN with PAL variants. EP2315: wild-type StlA, EP2516: S92G_H133F_A433S_V470A, EP2525: S92G_H133M_I167K_L432I_V470A, EP2495: A93C_H133F_T322W_Y437N, EP2502: I28R_S92G_H133F_V470A, EP2528: S92G_H133F_R185E, EP2526: S92G_F109A_H133M_T503E. For additional time points and non-normalized TCA production data, refer to Supplementary Figure 4 and 5.

Fig. 5. In vitro and in vivo activity of SYNB1934.

Optimized PKU strain activity is compared to SYNB1618. a Schematic representation of the SYNB1934 bacterium shows key engineered elements, including the genes encoding mPAL, PheP, LAAD, and deletion of the dapA gene. b TCA production of 2.5 × 10⁹ resuspended lyophilized SYNB1618 vs SYNB1934 cells was measured in an in vitro simulation of the gut environment. n = 3 independent biological triplicates ± s.d. c and d NHP subjects were dosed orally with a 5 g peptide and 0.25 g d₅-Phe bolus followed by dosing with 1 × 10¹¹ resuspended lyophilized SYNB1618 or SYNB1934 cells. Plasma areas under the curve (AUCs) for strain-specific biomarkers TCA and d₅-TCA are displayed. For c, n = 18 biologically independent NHP subjects per group ± s.e. For each comparison, data were analyzed using a two-tailed unpaired t test (p < 0.0001). For urinary d₅-HA concentration normalized to creatinine (d) n = 18 for SYNB1618-treated and 17 for SYNB1934-treated subjects ± s.e (one NHP in the SYNB1934-treated group failed to produce a urine sample over the course of experimentation). Data were analyzed using a two-tailed unpaired t test with Welch’s correction (p = 0.0184).