PART COLLECTION

I. INTRODUCTION

I.1. Plastic Pollution and its Enzymatic Biodegradation

Over the past century, technological advancements have significantly revolutionized our understanding of the world, leading to remarkable progress in science, healthcare, space exploration, and improvements in lifestyle. However, this golden age of technological progress remains heavily dependent on petroleum extraction. Synthetic plastic is one of the most relevant examples of the many petroleum-derived products found for numerous consumer and industrial products. From packaging to medical devices, plastic shapes our daily lives and society, influencing the way we consume, the food we eat, and the clothes we wear. It invades all our ‘natural’ environment/ecosystems even up to our blood: we may say that we have recently entered into a new era of the Anthropocene, the ‘PLASTICOCENE’ area.
Global plastic production has risen exponentially over recent decades, reaching 367 million metric tons (MMT) in 2020. Due to its low biodegradability, plastic persists in the environment, with an estimated 12000 MMT of global plastic pollution accumulating by 2050 [Zheng & Suh, 2019]. Plastic pollution is now a major global concern, impacting the balance and preservation of ecosystems and posing serious risks to human health, including cardiovascular diseases, reproductive toxicity, and cancer [Chi et al., 2016; Wu et al., 2018].
Since the late 1990s, microbial plastic biodegradation has been observed in aquatic environments, especially by fungal and bacterial strains (mostly Gram-positive), such as Ideonella sakaiensis, Piscinibacter sakaiensis, Rhodococcus sp. 2G, and Gordonia sp. P8219 [Huang et al., 2019; Zhang et al., 2022; Chen et al., 2023]. Microbial plastic degradation occurs through enzymatic processes that break down xenobiotic compounds like polyethylene terephthalate (PET) and phthalate acid esters (PAEs), which are incorporated into plastic to enhance flexibility and durability (Figure 1):

  • PET can be enzymatically degraded by PETase into bis(2-hydroxyethyl) terephthalic acid (BHET), mono(2-hydroxyethyl) terephthalic acid (MHET), terephthalic acid (TPA), and ethylene glycol (EG). MHET hydrolase (MHETase) can further degrade MHET into TPA and EG [Zhang et al., 2022].
  • Phthalate acid esters (PAEs), which are used to increase flexibility in plastic polymers, can be degraded by PAE hydrolases such as MehpH, a monoalkyl phthalate (MBP) hydrolase that converts MBP into phthalic acid (PA) [Huang et al., 2019; Chen et al., 2023].

The detection of these degradation byproducts, such as PA and TPA, is crucial for identifying plastic-polluted areas and better localizing and evaluating risks. Identifying these byproducts can also help for the screening of engineered PET or PAE hydrolases designed for industrial-scale plastic degradation, with the potential to support recycling initiatives and bioremediation efforts.

Figure 1. Enzymatic cascade of the complete degradation pathway of plastic polymer of polyethylene terephthalate (PET) by PETase and MHETase to TPA (A) and of phthalate acid esters (PAEs) by PAE hydrolases to PA (B). (adapted from Zhang et al., 2022 and Chen et al., 2023).

I.2. XylS as an in vivo transcriptional biosensor for plastic degradation products detection

XylS is a promiscuous transcription factor (TF) responsive to a wide range of benzene derivatives, the most potent being the m-toluic acid [Gawin et al., 2017] as detailed under the Contribution page on this wiki. To make XylS an interesting tool for the detection of the plastic degradation products TPA and/or PA, its promiscuity needs to be reduced (Figure 2). Indeed, the design of TF-based biosensors usually requires both a high sensitivity and a high specificity for the target molecule. Maximizing both enables the detection of the analyte at very low concentrations without interference due to other molecules triggering activation of the detection event. Therefore, a promiscuous binder is not ideal for the design of a specific biosensor, as it may lead to unexploitable results in a complex media containing several similar molecules, which is commonly the case when analyzing natural samples.
Transcriptional biosensor activity and efficiency can be modified through protein expression level by modifying its transcription and translation level. In the case of synthetic ligand recognition, the biosensor can be engineered by modifying its ligand recognition and specificity.

Figure 2. XylS induces transcription from the Pm promoter when it binds to the effector, here m-toluic acid (adapted from Ogawa et al., 2019, 2021).

XylS mutants, arising from one or several genetic variations of the XylS gene, have been studied for decades for improved properties, including response to an extended range of benzene derivatives or, on the contrary, mutants with a high specificity toward one or few molecules.
Several examples highlight the possibility to successfully engineer XylS to significantly reduce its promiscuity toward one molecule and concomitantly improve the sensitivity. Ogawa et al., 2019 demonstrated the switch of specificity of XylS from m-toluic acid to p-toluic acid with the mutant XylS-N7R-T74P, the latter being the most potent inducer of wild type XylS. XylS-N7R-T74P was 13 times more sensitive to p-toluic acid than m-toluic acid.
Furthermore, the ability of XylS mutants to detect PA or TPA have been studied [Li et al., 2022] and site-directed mutagenesis through error-prone PCR followed by directed evolution was carried out to identify variants more specific to PA or TPA. The results of this study showed that XylS can be engineered to detect PA and TPA at lower concentration than wild type XylS. Notably the two mutants XylS-K38R-L224Q and XylS-W88C-L224Q were found to be particularly beneficial. Compared to XylS wild type, these two mutants had a higher response curve to PA and TPA. The reported limit of detection of PA and TPA was also lowered to 0.1 mM, compared to 5 mM with XylS wild type. The mutant XylS-K38R-L224Q has however a high promiscuity as it is able to detect several other benzoic derivatives tested as XylS wild type does (m-toluic acid, salicylic acid, acetylsalicylic acid, 2-hydroxy-3-methylbenzoic acid, 3-chlorobenzoic acid). For the molecules tested, XylS-W88C-L224Q was only sensitive to PA and TPA, highlighting a higher selectivity toward these two molecules.
The three mutations identified in these two mutants are located in different units of the XylS protein (Figure 3). K38R is located on a beta-loop, hence an unstructured part of the protein, that is not directly involved in the interaction with the ligand. The residue 224, hence the mutation L224Q occurring on both mutants identified by Li et al., 2022, is on the subunit interacting with DNA. This residue may play a role for the switch from the monomeric inactive conformation to the dimeric active conformation of XylS upon binding of the ligand, but we cannot firmly confirm it as the exact structure of XylS with or without ligand has not yet been determined experimentally.
It has been found that residue 111, located in the Effector-Binding Domain (EBD) domain, near the potential/predicted binding pocket of XylS is able, when mutated, to switch XylS substrate specificity from m-toluic acid to PA and TPA [Li et al., 2022]. XylS-A111V, shows a 2.7-fold induction level at 5 mM PA and TPA compared to XylS wild type. However this mutant is limited in terms of sensitivity at lower concentrations: 0.1 to 2.5 mM. Independently, this mutation increases the sensitivity of XylS for m-toluic acid [Vee Aune et al., 2009].
Moreover, recent research published by Ogawa et al., 2019 and 2021 successfully shifted XylS specificity from m-toluic acid to p-toluic acid, a structural analog of TPA that differs by the substitution of a carboxyl (-COOH) group for a methyl (-CH3) group. The best variant candidate, XylS-N7R-T74P, showed a 13-fold change in specificity for p-toluic acid. Both mutations are located in the EBD domain but in different regions: T74P is buried into the predicted binding site, while N7R, which appears to be a mutation essential for p-toluic acid specificity (encounter in other variants), is found on the protein surface in a non-structured region, possibly interacting with the second XylS subunit or with the two αCTDs (C-terminal domain of the α subunits) RNA polymerase complex. The location of T74P confirms the localization of the predicted binding pocket identified in other XylS variants. However, it also shows that mutations enhancing sensitivity or specificity toward p-toluic acid or TPA/PA can occur both near the binding pocket or on distant residues outside the direct binding site.
Moreover, as TPA is a close structural analogue of p-toluic acid, the residues N7 and T74 are highly likely to play a role in switching XylS specificity to TPA and may appear in mutants generated through directed evolution. To better understand their involvement in this specificity switch, we decided to highlight the location of every ligand in the protein's 3D structure we modeled using ChimeraX (Figure 3).

Figure 3. Spatial localization of residues 38, 88, 111, 224 (green) implicated in the switch specificity of XylS for PA and TPA. N7*: at the surface of the EBD region. T74*, W88, A111: proximity to the binding pocket. K38: at the EBD surface, at the entry of the predicted binding pocket. L224: into the DBD. The ligand is colored in red (m-toluic acid). The 3D structure was generated using ChimeraX. * denotes the residues implicated into p-toluic acid specificity switch

The different mutations improving sensitivity or specificity toward PA or TPA occur at different positions on the XylS ternary structure (Figure 3). This highlights the need for powerful directed evolution tools to find new variants of a protein with desired properties, as rational design is still difficult to apply for the residues that are distant from the interaction site with the ligand and have therefore no obvious effect on the interaction between the protein and its ligand. In our PHAGEVO project we combined the latest directed evolution approaches and AI modeling, nature against AI in order to optimize the XylS transcription factor for plastic degradation metabolites recognition and specificity.

II. XYLS VARIANTS DESIGNED WITH AI

II.1. Design

As detailed on the Modeling page on our wiki, we used predictive AI based on the PocketGen model as a baseline [Zhang et al., 2024] and generated iteratively a set of mutants starting from XylS wild type, XylS-K38R-L224Q and XylS-W88C-L224Q.
A back and forth between the wet and the dry lab, resulted in 4 series of mutants that were sequentially generated first using the original PocketGen model (XylS AI PA (BBa_K5061063) and XylS AI TPA (BBa_K5061064)), then the versions 1, 2, 3 of our model and software tool, denoted XylS PHAGEVO-AI v1/2/3 PA/TPA in the summary Table 1.
A sequence comparison of all these mutants along with the wild type XylS and the two mutants described in the literature (Figure 4) shows 2 hotspots where the mutations are grouped. The first hotspot is located on the cysteine 75 that is changed to an aromatic amino acid, tyrosine or phenylalanine, and the closely located histidine 77 (in a few mutants).
The second hotspot covers the residues 126-131, with different combinations of mutations predicted on up to 4 positions in this region.

Table 1. Parts Registry accession numbers and the mutated amino acids of XylS variants generated by AI together with the wild type XylS and the two mutants XylS-K38R-L224Q and XylS-W88C-L224Q described as being able to weakly detect PA and TPA plastic degradation metabolites, respectively [Li et al., 2022].

XylS variant Mutated Amino Acids CDS Part Number Expression Cassette Part Number
XylS wild type - BBa_K5061060 BBa_K5061160
XylS-K38R-L224Q K38R L224Q BBa_K5061061 BBa_K5061161
XylS-W88C-L224Q W88C L224Q BBa_K5061062 BBa_K5061162
XylS AI PA C75Y P127K S128K BBa_K5061063 BBa_K5061163
XylS AI TPA C75Y H77Y P127K S128R BBa_K5061064 BBa_K5061164
XylS PHAGEVO-AI v1 PA C75F L126C P127D S128T L131M BBa_K5061065 BBa_K5061165
XylS PHAGEVO-AI v1 TPA C75F H77C L126V P127D S128K BBa_K5061066 BBa_K5061166
XylS PHAGEVO-AI v2 PA C75Y P127K S128K L131I BBa_K5061067 BBa_K5061167
XylS PHAGEVO-AI v2 TPA C75Y H77Y P127E S128R BBa_K5061068 BBa_K5061168
XylS PHAGEVO-AI v3 r3m8 PA C75Y P127D S128R BBa_K5061069 BBa_K5061169
XylS PHAGEVO-AI v3 r5m8 PA C75Y P127D S128R L131M BBa_K5061070 BBa_K5061170
XylS PHAGEVO-AI v3 r5m10 PA C75Y L126F P127D S128R BBa_K5061071 BBa_K5061171
XylS PHAGEVO-AI v3 r5m11 PA C75Y H77L L126V P127D S128R L131M BBa_K5061072 BBa_K5061172
XylS PHAGEVO-AI v3 r6m1 TPA C75Y L126V P127D L131M BBa_K5061073 BBa_K5061173
XylS PHAGEVO-AI v3 r6m5 TPA C75Y L126F P127D S128R L131M BBa_K5061074 BBa_K5061174
XylS PHAGEVO-AI v3 r7m1 PA C75F L126I S128R L131M BBa_K5061075 BBa_K5061175

Figure 4. Sequence comparisons of XylS variants generated by AI together with the wild type XylS and the two mutants XylS-K38R-L224Q and XylS-W88C-L224Q described as being able to detect PA and TPA plastic degradation metabolites, respectively [Li et al., 2022]. The alignment was generated using the MUSCLE algorithm implemented in SnapGene.

To experimentally validate these predicted mutations, we designed whole-cell biosensors following a classical architecture (Figure 5) composed of two parts:

  • the expression plasmid (EP) carrying the XylS variants under the control of the J23104 promoter, a custom made RBS and the B0011 terminator in the BBa_K5061003 backbone composed of the chloramphenicol resistance gene (CmR) and the CloDF13 origin of replication. As plasmids having a CloDF13 ori are medium copy (20-40 / cell), we choose a strong promoter for the XylS gene: the J23104, one of the strongest of the Anderson library promoters according to the data presented on its page in the Registry. We also designed a specific RBS library for XylS using the De Novo DNA’s Library Calculator v2.0 [Reis et al., 2020; Farasat et al., 2014; Ng et al., 2015] and selected the one predicted to have a Translation Initiation Rate (TIR) of about 25000 (BBa_K5061024).
  • the test plasmid (TP) carrying the sfGFP-LVAtag gene under the control of the Pm promoter regulated by XylS in the BBa_K5061002 backbone composed of the ampicillin resistance gene (AmpR) and the pSC101 origin of replication. The choice of sfGFP fused to the LVA degradation tag (BBa_K2675006) was based on the experience of our PI with this reporter gene from previous iGEM projects. Indeed, the LVA degradation tag (AANDENYALVA), which is an ssrA tag that accelerates protein degradation in Escherichia coli at 37°C [Pédelacq et al., 2006], allows a rapid degradation of the reporter gene and thus finely observe the synthetic gene networks dynamics. The backbone carries the pSC101 origin of replication, which is low copy and thus allows reducing the background expression of leaky promoters.

Figure 5. Experimental setup for in vivo transcriptional whole-cell biosensors for characterisation of XylS activity in the presence of various ligands notably PA and TPA with detection via GFP fluorescence (adapted from Alvarez-Gonzalez & Dixon, 2019).

II.2. Build

The construction of our EP (Table 1) and TP (BBa_K5061056) plasmids for XylS variants characterisation was achieved by Golden Gate in two steps using established molecular biology protocols. First we have built EP and TP cloning platforms (BBa_K5061033 and BBa_K5061034, respectively) in which Golden Gates adapters with BsaI sites are present between the J23104 promoter and B0011 terminator in the case of EP and upstream the RBS of sfGFP in the case of TP. Those plasmids were assembled by Golden Gate too, using BsmBI as a restriction enzyme, and DNA fragments obtained either by PCR from plasmid templates purchased from Addgene (MP6 and pJC175e) or provided by the hosting lab or synthesized.
Using these ‘universal’ Golden Gate cloning platforms for EP and TP, we readily built the XylS specific EP (Table 1) and TP (BBa_K5061056) plasmids. For this, XylS variants were synthesized as DNA fragments containing the designed RBS sequence with flanking convenient BsaI type IIS restriction sites. The Pm promoter was synthesized also, but as two long oligonucleotides, one for each DNA strand, that were annealed to form the double stranded DNA that was used in the Golden Gate reaction.

II.3. Test

To evaluated the XylS variants’ capacity to regulate the expression from the Pm promoter in our experimental setup, E. coli DH10B cells transformed with the XylS specific EP (Table 1) and TP (BBa_K5061056) plasmids were grown overnight at 37°C with shaking at 200 rpm in 96-DeepWell plates containing 1 mL of LB medium supplemented with 50 µg/mL ampicillin and 17.5 µg/mL chloramphenicol. The cells were then diluted 1:40 in a fresh LB medium containing the same antibiotics and incubated for 3-4 hours at 37°C with shaking. Following this incubation, the cells were further diluted 1:20 in LB medium containing the antibiotics and varying concentrations of specific inducers: m-toluic acid, 3-chlorobenzoic acid, terephthalic acid (TPA) or phthalic acid (PA), all as sodium or potassium salts. Inducer stock solutions were prepared at 5 mM in LB, and dilutions were performed to achieve final concentrations ranging from 0.01 µM to 5 mM. For each test, 10 µL of the 1:40 diluted cultures were added to 190 µL of the media containing the desired concentration of each inducer in an opaque 96-well polystyrene microplate (COSTAR 3603, Corning).
The microplates were then incubated at 37°C with shaking at 200 rpm. Fluorescence measurements (λexcitation 483 nm, λemission 530 nm) and optical density at 600 nm (OD600) were taken every 15 minutes for 24 hours using a CLARIOstar (BMGLabtech) plate reader. Fluorescence values were normalized by OD600 to account for variations in cell density.

II.4. Learn

After docking experiments performed with different versions of FLINT, thirteen XylS mutants were predicted as having improved binding to PA (BBa_K5061063, BBa_K5061065, BBa_K5061067, BBa_K5061069, BBa_K5061070, BBa_K5061071, BBa_K5061072, BBa_K5061075) or TPA (BBa_K5061064, BBa_K5061066, BBa_K5061068, BBa_K5061073, BBa_K5061074).

To characterize them, we decided to test the effect of 4 different inducers on their activity: m-toluic acid (MTA), 3-chlorobenzoic acid (3CBA), phthalic acid (PA) and terephthalic acid (TPA). MTA and 3CBA are known potent inducers of XylS wild type, while PA and TPA are ineffective effectors of XylS wild type. The results are summarized in Figure 6.

XylS wild type has the highest activity in presence of MTA and 3BCA. WIth the exception of XylS PHAGEVO-AI v3 r7m1 PA (BBa_K5061075), MTA is an ineffective or poorly effective effector of all mutants predicted by AI. XylS PHAGEVO-AI TPA (BBa_K5061064), XylS PHAGEVO-AI v1 PA (BBa_K5061065), XylS PHAGEVO-AI v1 TPA (BBa_K5061066) and XylS PHAGEVO-AI v2 PA (BBa_K5061067) are induced by 3CBA only at a high concentration of 5 mM. XylS PHAGEVO-AI v3 r5m11 PA (BBa_K5061072) and XylS PHAGEVO-AI v3 r6m5 TPA (BBa_K5061074) have induction ratio reaching a maximum of 2 at 1 mM 3CBA before reaching a plateau. The other mutants tested are not sensitive to 3CBA.

With the possible exception of XylS PHAGEVO-AI v3 r5m11 PA (BBa_K5061072) and XylS PHAGEVO-AI v3 r6m5 TPA (BBa_K5061074), none of the mutants tested are induced by TPA. XylS PHAGEVO-AI v3 r7m1 PA (BBa_K5061075) seems on the contrary to be repressed by increasing concentrations of this molecules.

However, XylS PHAGEVO-AI v2 PA (BBa_K5061067) has an induction ratio reaching 2.5 at 5 mM PA. This compares with an induction ratio of 1.4 for XylS wild type in the same conditions. It is also more than the two mutants XylS-K38R-L224Q (BBa_K5061061) and XylS-W88C-L224Q (BBa_K5061062), which are XylS mutants described in the literature and engineered specifically for PA detection, and have a maximal induction ratio of 1.6 and 1.7, respectively. Compared to these two mutants, XylS PHAGEVO-AI v2 PA (BBa_K5061067) is also much more specific. Indeed, though being less sensitive than XylS wild type, XylS-W88C-L224Q (BBa_K5061062) and XylS-K38R-L224Q (BBa_K5061061) are still activated by m-toluic with induction ratio reaching 2.5, while MTA is ineffective on XylS PHAGEVO-AI v2 PA (BBa_K5061067). XylS PHAGEVO-AI v3 r5m11 PA (BBa_K5061072) and XylS PHAGEVO-AI v3 r6m5 TPA (BBa_K5061074) are sensitive to PA at concentrations lower than any other mutant, reaching induction ratio of 1.8 at PA concentration of 10 µM for the former and 1.5 at PA concentration of 100 µM for the latter, remaining stable at higher concentrations. This make them particularly interesting for detection of low concentration of PA compared to other effective mutants that are induced at higher concentrations. Interestingly, the effect of 3CBA on these two mutants is almost identical to the effect of PA, with similar induction ratio for both molecules. In addition, MTA is only poorly effective on these mutants.

These result are very promising and highlight the power of FLINT for predicting new mutants with improved sensitivity and specificity toward PA or TPA.

Figure 6. in vivo characterization of sfGFP-LVAtag expression by E. coli DH10B cells carrying the expressing plasmid of various XylS variants (Table 1) together with the reporter plasmid carrying the sfGFP-LVAtag gene under the control of Pm promoter (BBa_K5061056) in the presence of increasing concentrations of different XylS effectors. (A, B) Fluorescence/OD600 values obtained in the late exponential growth phase as a function of effectors' concentration. (C) Fluorescence/OD600 fold changes in the presence of the inducer compared to the values in the absence of the inducer. The data and error bars are the mean and standard deviation of at least three measurements on independent biological replicates.

To better explain these experimental results, we performed in silico predictive docking analyses using AutoDock Vina [Trott et al., 2010] on XylS wild type and its variants based on their predicted structure (generated by AlphaFold 3) as a receptor and m-toluic acid, PA, and TPA as ligands. Our objective was to compare and validate the experimental data with the predictive outcomes generated through these docking simulations.
With these predicted data we aim to localize each mutated residue and have a better understanding of their contributions to ligand binding, including the interactions and conformational changes that influence the switch in specificity of the engineered XylS variants for PA and TPA.
By localizing the residues that interact with the ligands, we provide insights into the conformational changes that occur as a result of these interactions.
The docking results for each compound were ranked according to their binding free energy values (ΔG).

with :

  • ΔG = Change in Gibbs free energy (kcal/mol)
  • R = Universal gas constant = 1.987 10-3 kcal.mol-1.K-1
  • T = Absolute temperature = 298 K
  • Kd​ = Dissociation constant

Kd reflects the affinity between the ligand and its receptor. In other words it corresponds to the concentration (M, mol.L-1) of the ligands at which 50% of the ligand (L) is binding with receptor (R) forming ligand-receptor complex at equilibrium state.

with :

  • [L] = Concentration of the free ligand (in mol.L-1)
  • [R] = Concentration of the free receptor (in mol.L-1)
  • [LR] = Concentration of the ligand-receptor complex (in mol.L-1)

The lower the Kd​ value is, the higher is the affinity between the ligand and the receptor.

Docking analysis was then realized for XylS wild type, XylS-K38R-L224Q, XylS-W88C-L224Q variants, and XylS PHAGEVO-AI v2, the best candidate screened experimentally (Table 1). The docking simulations were performed with m-toluic acid (MTA), PA and TPA. The best results from ranked 20 simulations for each docking presented in Table 2.

Table 2. Docking modelisation results.

MTA PA TPA
ΔG (kcal/mol) Kd (mol/L) ΔG (kcal/mol) Kd (mol/L) ΔG (kcal/mol) Kd (mol/L)
XylS wild type -5.844 5.12 10-5 -5.166 1.61 10-4 -5.347 1.19 10-4
XylS-K38R-L224Q -5.583 7.96 10-5 -4.829 2.85 10-4 -5.013 2.09 10-4
XylS-W88C-L224Q -5.684 6.71 10-5 -5.149 1.66 10-4 -5.062 1.92 10-4
XylS PHAGEVO-AI v2 PA -5.885 4.78 10-5 -6.233 2.65 10-4 -5.8 5.52 10-5

In all docking simulations, the top-ranking results consistently showed that the ligand docked within the jelly-roll structure, confirming the hypothesis about the binding pocket's location, similar to what was observed in the transcription factor ToxT [Lowden et al., 2010].

With MTA, XylS wild type exhibited the highest affinity (51.2 μM), while the XylS PHAGEVO-AI v2 variant showed the greatest affinity for PA (47.8 μM). These values reflect moderate affinities. XylS-K38R-L224Q and XylS-W88C-L224Q displayed higher values, though they remain within the same magnitude. However, XylS wild type demonstrated a 2- to 3-fold decrease in affinity for PA and TPA, suggesting a reduced or lack of specificity for these two ligands.

In contrast, XylS PHAGEVO-AI v2 showed a similar affinity for TPA (55.2 μM) as it did for MTA, while its specificity for PA was lower. The other variants, XylS-K38R-L224Q and XylS-W88C-L224Q, showed decreased affinity for both PA and TPA compared to MTA, contrary to experimental observations, particularly with XylS-W88C-L224Q (Figure 6).

These results highlight certain limitations in the predictive data. Docking analyses do not account for molecular dynamics, including ligand-induced conformational changes or the contribution of mutations located elsewhere in the protein, which may influence binding and specificity outcomes.

III. XYLS VARIANTS SELECTED BY DIRECTED EVOLUTION

III.1. Design

As detailed on the Engineering Success page on this wiki, we performed directed evolution of the XylS transcription factor to improve its sensitivity and specificity toward two molecules, phtalic acid (PA) and terephtalic acid (TPA), as a proof of concept of the PHAGEVO system. The experiments were performed in different conditions differing in (i) the evolution system (PHAGEVO or PANCE); (ii) the mutagenesis plasmids that express the machinery required for mutagenesis; (iii) the accessory plasmids, which express under the control of the XylS promoter Pm the essential genes (gVI or gVI + gIII) deleted from the phage genome, providing a selection pressure for XylS mutants with improved activity, and (iv) the target molecule (PA or TPA). Among all conditions tested, 30 were retained because phage production increased significantly compared to the condition without mutator, which is expected when the gene of interest (here XylS) evolve toward the target application (here a higher sensitivity and specificity to PA or TPA), and is therefore an indicator of the success of the evolution.

In vivo XylS activity assays based on expression levels of sfGFP under the control of the Pm promoter were performed to screen individually 2880 XylS clones generated from the 30 evolution conditions selected by luminescence monitoring.
Screening of the XylS mutants generated by the evolution experiments required a system that enabled it to easily measure XylS activity depending on the concentration of its inducers. For this purpose, an in vivo system based on fluorescence was designed (Figure 7), based on the same principle of whole-cell biosensors as for testing the AI-designed variants (as described above).

Figure 7. In vivo transcriptional whole-cell biosensors for high-throughput screening of engineered XylS mutants with altered and enhanced specificity for PA and TPA and their fluorimetric detection through Green Fluorescent Protein (GFP). (adapted from Alvarez-Gonzalez & Dixon, 2019)

III.2. Build

The construction of the XylS specific expression plasmids carrying the XylS variants generated by evolution experiments and TP (BBa_K5061056) plasmids was performed following the same protocols as detailed above in the chapter “XylS variants generated by AI”. For this, after one round of overnight evolution of XylS with either PANCE or PHAGEVO, 30 conditions were selected (as described on the Engineering Success page on this wiki). The 200 µL of culture medium from these evolution experiments containing bacteria and evolved phages were collected and centrifuged 3 min at 7000 rpm. The supernatant filtered using a MultiScreen-GV Sterile, clear 96-well filter plate with 0.22 um pore size Hydrophilic PVDF membrane (Millipore) fixed on a DeepWell plate. Filtered phages were boiled at 80°C for 30 min to disrupt them, then the XylS sequences of the polyclonal phage solution were amplified by PCR with the high fidelity Q5 DNA polymerase and cloned into the EP cloning platform (BBa_K5061033) by Golden Gate. The assembly product was transformed into E. coli DH10B cells transformed with the XylS specific TP (BBa_K5061056) plasmids.

III.3. Test

96 mutants of each evolution experiment were randomly picked on the corresponding Petri dish to inoculate 1 mL of LB medium supplemented with 50 µg/mL ampicillin (to select for EP plasmid containing cells) and 17.5 µg/mL chloramphenicol (to select for EP plasmid containing cells) in a 96-DeepWell. After 4 hours of incubation at 37°C with shaking at 200 rpm, 10 µL of each well were diluted on a transparent 96-well plate in 190 µL of LB medium containing the same antibiotics and either 0.5 mM PA or TPA depending on the experimental conditions of the corresponding evolution experiment (to activate XylS and trigger sfGFP expression). Both the DeepWell and the 96-well plate were incubated overnight at 37°C, 200 rpm. The next day, optical density at 600 nm (OD600) and sfGFP fluorescence (λexcitation 483 nm, λemission 530 nm) of the 96 well plate was analyzed using a CLARIOstar (BMGLabtech) plate reader. Fluorescence values were normalized by OD600 to account for variations in cell density.
For the clones displaying a Fluorescence/OD600 ratio above average, 200 µL from the Deepwell were transferred on a new transparent 96 well plate and readily analyzed for OD600 and sfGFP fluorescence as described above to serve as a control without PA or TPA induction.
After this first screening step, the clones with an induction ratio (i.e: Fluorescence/OD600 values with the inducer compared to the control without inducer) were further investigated.
Overnight cultures of the selected clones were diluted 40 fold in LB media supplemented with 50 µg/mL ampicillin and 17.5 µg/mL chloramphenicol and incubated 4h at 37°C. An opaque 96-well polystyrene microplate (COSTAR 3603, Corning) with 190 µL LB with antibiotics and an inducer concentration ranging from 0 to 5 mM (no inducer, 10 µM, 100 µM, 500 µM, 1 mM, 5 mM) was prepared and inoculated with 10 µL of the XylS clone culture. Each condition was performed in quadruplicate. The plate was placed overnight in a CLARIOstar (BMGLabtech) plate reader and incubated at 37°C with shaking. Each 15 minutes, OD600 and sfGFP fluorescence were measured with the same parameters as previously described.

III.4. Learn

The screening of XylS mutants from the 30 evolution conditions considered as promising was carried out by in vivo characterization of sfGFP expression controlled by the XylS promoter Pm.

In a first screening step, fluorescence of all randomly picked XylS clones in presence of the PA or TPA inducer was analyzed (Figure 8). This enabled us to eliminate the majority of the clones which did not have a fluorescence signal above background levels. A second analysis was then performed in absence of the inducer, to establish an induction ratio comparing expression with and without the inducer (Figure 9). Clones with high fluorescence signals in presence of the inducer did not necessarily have a high induction ratio. This means that some clones may have evolved higher constitutive induction levels even in the absence of an activator molecule. In the design of our directed evolution experiment, this is one of the main risks of escape from the desired target property.

Figure 8. Results of the 1st step of the screening experiments through in vivo characterization of sfGFP-LVAtag expression by E. coli DH10B cells carrying randomly selected colonies expressing XylS together with the reporter plasmid carrying the sfGFP-LVAtag gene under the control of Pm promoter (BBa_K5061056) in the presence of 0.5 mM of either PA or TPA. The data are the Fluorescence/OD600 values obtained after the overnight culture of 2880 colonies from 30 independent evolution experiments. 527 colonies displaying a Fluorescence/OD600 value above average (in pink) were selected for further analysis, the results of each are presented in Figure 9.

Figure 9. Results of the 2nd step of the screening experiments through in vivo characterization of sfGFP-LVAtag expression by E. coli DH10B cells carrying the 527 pre-selected colonies (Figure 8) expressing XylS variants together with the reporter plasmid carrying the sfGFP-LVAtag gene under the control of Pm promoter (BBa_K5061056). The data are the ratio between the Fluorescence/OD600 values obtained after the overnight culture in the presence of 0.5 mM of either PA or TPA compared to the values in the absence of the inducer. Colonies displaying a Fluorescence/OD600 fold change value above average (in pink) were selected for further analysis, the results of each are presented in Figure 10.

The clones with highest induction ratio were selected for further characterisation and comparison with XylS wild type and mutants from literature (Figure 10). Surprisingly, most clones have a low induction ratio that still remains in the range of observations with XylS wild type, which contrast with previous screening steps results. A few clones clearly surpass both XylS wild type and the two mutants from literature XylS-K38R-L224Q (BBa_K5061061) and XylS-W88C-L224Q (BBa_K5061162).

The clones PHAGEVO-DE10-D12, PHAGEVO-DE18-D7, PHAGEVO-DE20-C3 and PHAGEVO-DE20-F12 reach induction ratio surpassing 2 at PA concentration of 5mM. This compares with an induction ratio of 1.4 for XylS wild type at the same concentration, and and 1.6 to 1.7 for XylS-K38R-L224Q (BBa_K5061161) and XylS-W88C-L224Q (BBa_K5061162). As for the experiments with TPA, there is no major differences in terms of induction ratios between wild type XylS (BBa_K5061160), XylS-K38R-L224Q (BBa_K5061161), XylS-W88C-L224Q (BBa_K5061162) and the 15 clones from directed evolution. As for the experiments with m-toluic acid and 3-chlorobenzoic acid, the clones PHAGEVO-DE10-D12, PHAGEVO-DE20-F12 and PHAGEVO-DE22-H9 behaves similarly to XylS-K38R-L224Q (BBa_K5061161) and XylS-W88C-L224Q (BBa_K5061162).

Figure 10. in vivo characterization of sfGFP-LVAtag expression by E. coli DH10B cells carrying selected colonies (Figure 9) expressing XylS variants together with the reporter plasmid carrying the sfGFP-LVAtag gene under the control of Pm promoter (BBa_K5061056) in the presence of increasing concentrations of phthalic acid (A) and terephthalic acid (B). The controls were performed with the wild type XylS (BBa_K5061160) and the two mutants XylS-K38R-L224Q (BBa_K5061161) and XylS-W88C-L224Q (BBa_K5061162) described as being able to detect PA and TPA plastic degradation metabolites, respectively [Li et al., 2022]. The data are the ratio between the mean Fluorescence/OD600 values of at least three measurements on independent biological replicates obtained in the late exponential growth phase and the values in the absence of the inducer. Same data in line charts representation is available here. Some clones were tested with only PA (A) or only TPA (C) and others were tested with both molecules (B & D).

Three rounds of screening enabled the reduction of the number of selected XylS clones from 2880 to 527 and then 15. However, for the moment, we identified only one XylS mutant after sequencing of the selected clones (PHAGEVO-DE19-B8: XylS-K38R-G198S-L224Q), that has one additional mutation G198S compared to the initial sequence XylS-K38R-L224Q (BBa_K5061161) after evolution experiments using PHAGEVO. Unfortunately, this mutant does not show improved properties compared to XylS wild type or the two mutants from literature XylS-K38R-L224Q (BBa_K5061161) and XylS-W88C-L224Q (BBa_K5061162), and is therefore not considered as interesting for future works.

All other clone sequences were either identical to the initial sequence of the evolution experiment, or were empty plasmids without incorporation of the XylS sequence into the expression plasmid (EP).

Differences of GFP expression observed during screening between the different clones selected after directed evolution experiments may be the result of experimental variability instead of actual differences in terms of induction by PA or TPA.

CONCLUSIONS

The transcription factor XylS, which is a promiscuous transcription factor sensitive to a vast array of benzoic acid derivatives, has been extensively studied and engineered for mutants with improved specificity or sensitivity toward one target molecule in order to serve as an efficient biosensor.

Recently, directed evolution methodologies enable to engineer it for detection of phtalic acid (PA) and terephtalic acid (TPA), with both mutants XylS-K38R-L224Q (BBa_K5061161) and XylS-W88C-L224Q (BBa_K5061162) described as being able to weakly detect PA and TPA, two plastic degradation metabolites [Li et al., 2022].

Our team aimed at extending the knowledge about XylS mutants with improved specificity and sensitivity toward PA and TPA. For this purpose, we have tested two radically different approaches: rational design by artificial intelligence using FLINT, a model developed by our dry lab team, and directed evolution using the PHAGEVO technology developed by our wet lab team.

13 mutants generated by the AI model FLINT were characterized. Among them, the mutants XylS PHAGEVO-AI v2 PA (BBa_K5061067), XylS PHAGEVO-AI v3 r5m11 PA (BBa_K5061072) and XylS PHAGEVO-AI v3 r6m5 TPA (BBa_K5061074) are particularly significant as they show improved specificity and sensitivity to PA compared to the two mutants previously described in the literature, highlighting the power of the FLINT model for prediction of new XylS mutants. The former seems to reach higher induction ratio at high PA concentration while the two latter are comparable to XylS-K38R-L224Q and XylS-W88C-L224Q in terms of intensity of the signal but reach their maximum at PA concentrations 50 to 500 lower.

After one round of overnight directed evolution using PHAGEVO or PANCE in different conditions, we screened a total of 2880 clones. After two steps of screening, this number was reduced to 15 clones. However, sequencing results revealed only one XylS mutant, while the others were either identical to the initial sequence of the evolution experiment, or empty backbones.

From these results, we cannot conclude whether PHAGEVO works as intended or not as we lack important data. Indeed, we didn’t manage to perform Next Generation Sequencing (NGS) on the phage metapopulation after evolution. These data would have been important to understand where mutations occur on the selection phage and what is the rate of mutation in PHAGEVO compared to PANCE.

We also only performed one round of overnight evolution with PHAGEVO and PANCE, which may not be sufficient for emergence of a broad diversity of sequences and the subsequent selection of the best variants. Multiple rounds of evolution are probably required before we can identify new interesting XylS mutants. Indeed, the phage population is initially composed of only one single sequence. In our case, this sequence was either XylS-K38R-L224Q (BBa_K5061061) or XylS-W88C-L224Q (BBa_K5061062), two variants that are already described in literature as being activated by PA and TPA. Therefore, the initial selection pressure may not have been important enough to trigger high amplification of eventual mutants occurring during our evolution experiment. In this situation, a majority of phage containing the initial, non-modified, sequence of the gene of interest will be produced before mutagenesis happen.

In addition, one should not overlook the residual leaky expression of the Pm promoter, the basal level of activity of XylS even in absence of inducer. In this case, the selection pressure for the selection phage may also be considerably lowered.

Currently, artificial intelligence with FLINT revealed more interesting mutants of XylS than directed evolution. Indeed, artificial intelligence was easier to implement and enabled to quickly propose new mutants that can be readily tested in the lab. On the contrary, our new directed evolution tool PHAGEVO is still in its early phase of development, experimental design and troubleshooting. Once research on PHAGEVO will progress and more robust protocols for directed evolution with this tool established, we anticipate a significant acceleration in the generation of beneficial mutants.

Using artificial intelligence or directed evolution, we successfully managed to generate, identify and characterize several new mutants of the XylS transcription factor. This collection of new parts represent a valuable set of data for future development of biosensors for PA or TPA based on XylS. It also helps to better understand which residues of XylS are important for interaction with its effectors. At least three mutants (BBa_K5061067, BBa_K5061072 and BBa_K5061074) are promising PA biosensors and require deeper characterization.

REFERENCES

  • Alvarez-Gonzalez G, Dixon N. (2019) Genetically encoded biosensors for lignocellulose valorization. Biotechnol Biofuels 12, 246.
  • Chen Y, Wang Y, Xu Y, Sun J, Yang L, Feng C, Wang J, Zhou Y, Zhang ZM, Wang Y. (2023) Molecular insights into the catalytic mechanism of plasticizer degradation by a monoalkyl phthalate hydrolase. Commun Chem 6, 45.
  • Chi Z, Zhao J, You H, Wang M. (2016) Study on the mechanism of interaction between phthalate acid esters and bovine hemoglobin. J Agric Food Chem 64, 6035-6041.
  • Farasat I, Kushwaha M, Collens J, Easterbrook M, Guido M, Salis HM. (2014) Efficient search, mapping, and optimization of multi-protein genetic systems in diverse bacteria. Mol Syst Biol 10, 731.
  • Gawin A, Valla S, Brautaset T. (2017) The XylS/Pm regulator/promoter system and its use in fundamental studies of bacterial gene expression, recombinant protein production and metabolic engineering. Microb Biotechnol 10, 702-718.
  • Huang H, Zhang XY, Chen TL, Zhao YL, Xu DS, Bai YP. (2019) Biodegradation of structurally diverse phthalate esters by a newly identified esterase with catalytic activity toward di(2-ethylhexyl) phthalate. J Agric Food Chem 67, 8548-8558.
  • Li J, Nina MRH, Zhang X, Bai Y. (2022) Engineering transcription factor XylS for sensing phthalic acid and terephthalic acid: an application for enzyme evolution. ACS Synth Biol 11, 1106-1113.
  • Lowden MJ, Skorupski K, Pellegrini M, Chiorazzo MG, Taylor RK, Kull FJ. (2010) Structure of Vibrio cholerae ToxT reveals a mechanism for fatty acid regulation of virulence genes. Proc Natl Acad Sci USA 107, 2860‑2865.
  • Ng CY, Farasat I, Maranas CD, Salis HM. (2015) Rational design of a synthetic Entner-Doudoroff pathway for improved and controllable NADPH regeneration. Metab Eng 29, 86–96.
  • Ogawa Y, Katsuyama Y, Ueno K, Ohnishi Y. (2019) Switching the ligand specificity of the biosensor XylS from meta to para-toluic acid through directed evolution exploiting a dual selection system. ACS Synth Biol 8, 2679-2689.
  • Ogawa Y, Katsuyama Y, Ohnishi Y. (2021) Engineering the ligand specificity of the transcriptional regulator XylS using deep mutational scanning. ACS Synth Biol 11, 473-485.
  • Pédelacq JD, Cabantous S, Tran T, Terwilliger TC, Waldo GS. (2006) Engineering and characterization of a superfolder green fluorescent protein. Nat Biotechnol 24, 79-88.
  • Reis AC, Salis HM. (2020) An automated model test system for systematic development and improvement of gene expression models. ACS Synth Biol 9, 3145–3156.
  • Trott O, Olson AJ. (2010) AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31, 455-461.
  • Vee Aune TE, Bakke I, Drabløs F, Lale R, Brautaset T, Valla S. (2009) Directed evolution of the transcription factor XylS for development of improved expression systems. Microb Biotechnol 3, 38-47.
  • Wu C, Zhang K, Xiong X. (2018) Microplastic pollution in inland waters focusing on Asia. In: Wagner M, Lambert S (eds) Freshwater Microplastics. The Handbook of Environmental Chemistry Springer, Cham, 58, 85-99.
  • Zhang Y, Pedersen JN, Eser BE, Guo Z. (2022) Biodegradation of polyethylene and polystyrene: From microbial deterioration to enzyme discovery. Biotechnol Adv 60, 107991.
  • Zheng J, Suh S. (2019) Strategies to reduce the global carbon footprint of plastics. Nat Clim Chang 9, 374-378.
  • Zhang Z, Shen WX, Liu Q, Zitnik M. (2024) Efficient generation of protein pockets with PocketGen. Nature Machine Intelligence 6, 1382–1395.