Bacterial viability is ensured within the beads

Objective: We aimed at testing the viability of both strains over time within the beads to estimate the lifespan of our degradation process.

Method: The release and stability of bacteria were evaluated by incubating dried beads in sterile water (Figure 1), encapsulating either Pseudomonas putida with 10% or 20% activated carbon, or Labrys portucalensis with 20% activated carbon. Control beads without bacteria were also prepared using sterile water instead of bacterial culture.

Supernatant samples were collected at different time points to measure bacterial release, and beads from dedicated tubes were dissolved with 1% tricitrate solution to assess bacterial viability within the beads. Bacterial counts were determined by serial dilutions, plating, and colony forming unit (CFU) enumeration.

Results: All strains remained viable and even showed growth for at least 10 days (Figure 2A). Pseudomonas putida in 10% activated carbon beads continued to grow for up to 30 days. Although the other conditions were not tested for the full 30-day period, similar trends are likely. No significant differences in bacterial viability were observed between beads containing 10% or 20% activated carbon, indicating that carbon content does not compromise survival.

Regarding bacterial release, Figure 2B shows that the number of bacteria released into water increases over time, at levels comparable to the bacteria remaining in the beads. This indicates that the current bead formulation is not optimized to retain bacteria. Major adjustments to the formulation will be discussed with YpHen to ensure bacterial confinement and comply with biosafety regulations.

Activated-carbon beads adsorb PFOA from aqueous solutions

Objective: We aimed at testing the efficiency of PFAS removal from aqueous solutions of activated-carbon beads.

Method: The adsorption of PFOA on beads (10% and 20% activated carbon, 10% and 20% talc, and a negative control without beads) was assessed by incubating 1.0 g of beads in a 50 µM PFOA solution.

Results: Figure 3 clearly shows a distinct difference in PFOA concentration over time between beads with or without activated carbon, as well as the negative control without beads. The results indicate that beads containing 20% activated carbon adsorbed 90.2% of PFOA within 120 minutes, while those with 10% activated carbon achieved 96.9% adsorption. In contrast, beads containing talc showed much lower adsorption efficiencies, 36.7% for 20% talc and 33.5% for 10% talc. The control exhibited a 26% decrease in PFOA concentration, likely due to adsorption onto the walls of the tubes used in the experiment.

These findings confirm that the presence of activated carbon in the beads is primarily responsible for PFOA adsorption and highlight the potential synergistic effect between adsorption and bacterial degradation. Interestingly, the 10% activated-carbon beads demonstrated slightly higher adsorption efficiency than the 20% beads. This suggests that similar or even better removal performance can be achieved with a lower activated carbon content, thereby reducing production costs while maintaining high adsorption capacity.

We then attempted to characterize PFOS degradation by L. portucalensis.

L. portucalensis and PFOS degradation

Objective: We aimed at following PFOS degradation by the strain as performed in Wijayahena et al., 2025, but this time by increasing 10 fold the starting concentration of bacteria to measure the impact on degradation efficiency.

Method: Degradation of 10 mg/L of PFOS as the sole carbon source was monitored in M9 minimal medium for 30 days by LC-MS/MS. L. portucalensis was inoculated at two different starting concentrations, OD₆₀₀=0.1 and 1.

Results: Changes in optical density over time suggested different bacterial growth responses depending on the initial inoculum (Figure 4).

At starting OD₆₀₀=1.0, we observed a gradual decrease from 1.00 to 0.50 OD₆₀₀ over time, indicating a decline in the bacterial population.

In contrast, at an initial OD₆₀₀ of 0.1, a slight increase was observed, suggesting that a smaller starting population could better utilize PFOS as a carbon source, likely due to reduced competition.

Further experiments with intermediate inoculum densities may help identify the maximum threshold below OD₆₀₀=1 that still allows growth, thereby clarifying the relationship between initial biomass, resource availability, and bacterial activity under PFOS exposure.

PFOS degradation assays turned out to be inconclusive. During sampling, the filtering step used to remove bacteria before LC-MS/MS analysis led to adsorption of PFOS to the filter, precluding quantification of PFOS concentration. For future experiments, we recommend substituting the filtration step with cell pelleting by centrifugation and removal of the supernatant containing PFOS for mass spectrometry quantitation.

Quantification of toxicity effects by NaF, TFA and PFPrA

Objective: Degradation of PFOS generates PFPrA, TFA, and fluoride. These products are toxic to cells and might determine the maximal PFOS concentration that our process will be able to treat. We therefore assayed the minimal inhibitory concentration (MIC) of PFPrA, TFA, and NaF on L. portucalensis.

Method: Growth curves were monitored in the presence of NaF, TFA, and PFPrA by monitoring OD₆₀₀ kinetics in a 96-well microplate reader on glucose as the carbon source in M9 minimal medium.

Results: All the collected data are available via our Kinetic Viewer. This allows a complete and easy visual presentation of the kinetic data, and the corresponding R code is available on Contribution page. The most relevant data are presented here. L. portucalensis was able to grow until 40 mM of both TFA and PFPrA, (Figure 5A). However, MIC for NaF was not reached, and lies in a range over 40 mM (Figure 5B). Concentrations above 50 mM NaF were tested but the resulting growth curves could not be analyzed because the M9 medium crystallized at such high concentrations. Interestingly, this effect was not observed in toxicity assays with Pseudomonas putida (as shown later on this page).

As shown in Figure 5A, we observed important variations between biological replicates. This variability is primarily due to the difficulty of synchronizing Labrys portucalensis cultures, as the strain exhibits a long lag phase in pre-cultures (2 days). This issue is evident for all molecules shown in the figure and across all experiments with L. portucalensis (see on the Kinetic Viewer). While this variability does not prevent the determination of MIC values, it makes it challenging to accurately define characteristic growth parameters such as growth rates (μ) and lag time. To address this, we decided to normalize the biological replicates against their control condition (0 mM) in order to highlight trends in the evolution of μ and lag time as a function of molecule concentration (Figure 6).

The maximum growth rate decreases as the concentration increases for the different fluorinated compounds. No clear trend could be detected for the lag time.

These findings indicate that, although growth inhibition by PFOS degradation products can be quantified, variability in the physiological state of the replicates limits the precise determination of kinetic growth parameters.

We next turned to the second axis of the project: directed evolution of DeHa2 for degradation of short-chain PFAS.

The replication operon was successfully assembled and verified

Objective: We aimed to assemble the replication operon, which encodes a double-mutant error-prone DNA polymerase (ODNAP) and associated proteins (TP, DSB, and SSB) required for continuous replication of the linear replicon, with the goal of either integrating it into the chromosome of E. coli or cloning it into a single-copy plasmid.

Method: The replication operon was initially synthesized into three gBlocks (A, B and C) by IDT. Each fragment contained 60 bp overlapping sequences to allow for Gibson assembly as well as homology regions for chromosomal integration in E. coli. The assembled operon was then amplified by PCR using KOD polymerase. To validate the good assembly of the operon, we performed agarose gel electrophoresis on the PCR performed on the assembled product and confirmed these results by linear amplicon sequencing using nanopore technology (Eurofins Genomics).

Results: Gel electrophoresis analysis showed a PCR product of the expected size (Figure 7). In addition, sequencing of the purified fragment confirmed the correct assembly of the replication operon (see part BBa_25HKSDJN). This validates the construct for transformation into a bacterial host, with the goal of generating a strain carrying the ORep system to enable continuous mutagenesis.

The main objective was to integrate the operon into E. coli chromosome to implement the ORep mutagenesis system, a modification made possible using two helper plasmids (pCas and pTarget) enabling CRISPR-assisted recombineering. However, this approach did not yield the expected results (see our Engineering page). We therefore sought for an alternative using a one-copy plasmid.

Construction of a one-copy plasmid carrying the replication operon

Objective: We aimed at inserting the replication operon into E. coli to generate the ORep strain, prior transformation with the linear replicon. The first strategy was chromosomal integration at the SS9 locus using CRISPR-assisted λ-Red recombination (see our Engineering page). As this was not achieved, an alternative strategy was designed based on assembling the operon into the one-copy plasmid pUD1387, which carries chloramphenicol resistance.

Method: The replication operon was amplified by PCR while excluding pre-existing SS9 sites within its sequence. The pUD1387 plasmid backbone was amplified by PCR to generate a linear fragment. Following Gibson assembly with both fragments, the product was transformed into competent E. coli cells and plated on LB agar supplemented with chloramphenicol for selection. Colonies were screened by colony PCR, and the presence of the operon was verified by PCR on purified plasmid and subsequent sequencing.

Results: PCR amplifications produced bands at the expected sizes: 3 804 bp for the operon and 6 945 bp for the linearized pUD1387 backbone (Figure 8).

Colonies grew on LB-chloramphenicol plates after transformation, confirming uptake of plasmid DNA. However, diagnostic PCR on miniprepped clones did not yield the expected amplicon size for the assembled plasmid; instead, smaller fragments were obtained (Figure 9).

Sequencing confirmed incomplete assembly: only part of gBlock A was integrated, while gBlocks B and C were recovered intact in the plasmid (Figure 10).

Repeated attempts with purified PCR products and optimized assembly conditions yielded the same outcome. The incomplete recovery of the operon suggests instability of the construct in E. coli, possibly linked to the presence of foreign DNA polymerase genes. Future work will involve testing alternative cloning strategies to achieve stable maintenance of the complete operon.

The linear replicon was successfully assembled in pUC19 vector

Objective: We aimed to assemble the linear replicon designed to carry the dehalogenase gene (dehH2), the kanamycin-resistance gene (Kan^R) and flanked by ITRs, then to clone it into the pUC19 backbone, and validate the construction by sequencing.

Method: Linear replicon was first synthesized as two gBlocks (D and E) from IDT. Both fragments harboring dehH2, Kan^R and flanked by ITRs were cloned into the pUC19 vector by Gibson assembly to generate the replicon. The assembled construct was transformed into E. coli DH5α and plated on LB agar supplemented with kanamycin and ampicillin (Kan^R from gBlock D and Amp^R from pUC19). Colonies were screened by restriction digestion. Plasmid DNA was extracted and subsequently sequenced by whole-plasmid sequencing using nanopore technology (Eurofins Genomics) to confirm the correct assembly and integrity of the linear replicon.

Results: Colonies were obtained on LB agar plates containing kanamycin and ampicillin, indicating successful transformation with the pUC19-containing linear replicon construct. Plasmid DNA from five selected clones was digested with BamHI, producing the expected band pattern (Figure 11). This result suggests correct assembly of the linear replicon within the pUC19 backbone.

Sequencing of the miniprepped plasmids confirmed the correct assembly of the biobrick (see pUC19-replicon sequence on Materials page).

The sequence-verified pUC19-replicon was subsequently employed for the expression of DeHa2 in the fluorescence assays using the FluorMango biosensor (see later on this page).

Production of linear replicon with ITRs from pUC19-replicon

Objective: We aimed at amplifying and purifying the linear replicon from the pUC19-replicon biobrick. This linear fragment will later be introduced into the DH10B host carrying the replication operon to enable continuous orthogonal replication of the encoded dehH2 gene.

Method: Two primers targeting the linear replicon were designed to amplify the ITR-flanked fragment from the pUC19-replicon biobrick by PCR using KOD One polymerase. The PCR product was verified by agarose gel electrophoresis, purified, and sent for sequencing by whole-plasmid sequencing using nanopore technology (Eurofins Genomics).

Results: Agarose gel electrophoresis of the PCR product showed a single band at the expected size (Figure 12), confirming successful amplification of the linear replicon from the biobrick.

Sequencing of the purified fragment confirmed the linear replicon had the correct sequence (see part BBa_25794BPU).

We conclude that the linear replicon is ready for transformation into the bacterial host, where it provides the template carrying the DeHa2 enzyme to be subjected to continued mutagenesis by the ORep machinery.

Creation of a dehalogenase-free linear replicon biobrick

Objective: We aimed at generating a version of the linear replicon biobrick lacking the dehH2 coding the DeHa2 enzyme. This modified biobrick serves as a versatile kit, allowing users to insert their own target gene for continuous orthogonal replication and directed evolution.

Method: Two PCRs were performed from the biobrick: one amplifying the left half of the biobrick and the other the right half, both excluding the dehH2 sequence. PCR products were purified and joined by Gibson assembly to reconstruct the biobrick lacking dehH2. The assembled construct was transformed into E. coli for amplification and extracted by miniprep from multiple clones. Colonies were screened by restriction digestion to verify the absence of the dehH2 sequence. Plasmids from selected clones were sequenced by whole-plasmid sequencing using nanopore technology (Eurofins Genomics) to confirm the correct deletion and integrity of the biobrick.

Results: Agarose gel analysis of the two PCR products showed single bands (an upper band of lower intensity is also observed in one PCR product) of the expected size (~ 2 000 bp each), confirming successful amplification (Figure 13).

Gibson assembly products were amplified in E. coli. Restriction digestion of miniprepped plasmids showed that only clones 3 and 4 had the expected digestion pattern corresponding to the biobrick without dehH2 (Figure 14). Comparison with the original biobrick containing dehH2 corroborated the deletion.

Sequencing results of plasmids from clones 3 and 4 confirmed the removal of the dehH2 gene and the correct sequence of the remaining biobrick (see pReplicon_KIT sequence on Materials page).

We then aimed at developing a screening method to be combined with ORep: fluoride-sensitive biosensor integrated with droplet microfluidics.

FluorMango fluorescence is sensitive to NaF concentration variations

Objective: We aimed at quantifying the responsiveness of FluorMango to varying amounts of fluoride ions.

Method: FluorMango aptamer was incubated for 1 hour with TO1-biotin, 1X refolding buffer and increasing NaF concentrations, and fluorescence was measured in 384-well plates.

Results: Figure 15 clearly demonstrates that the fluorescence intensity increases with the fluoride concentration up to 3 mM NaF. Moreover, the aptamer is sensitive to 0.1 mM NaF.

The next step was to show if this response is specific to fluoride ions.

FluorMango responds specifically to fluoride ions

Objective: We assessed the specificity of FluorMango fluorescence. To this end, we exposed the aptamer to a variety of molecules relevant to our experiments and measured its fluorescence.

Method: FluorMango aptamer was incubated for 1 hour with TO1-biotin (400 nM), 1X refolding buffer and NaF, NaCl, TFA or fluorinated microfluidic oil. Fluorescence was measured in 384-well plates.

Results: A strong fluorescence signal, up to 14-fold, was observed exclusively in the presence of fluoride ions, whereas fluorescence remained low in the presence of NaCl, TFA, or the fluorinated oil used in microfluidics (Figure 16). This remarkable specificity, together with its quantitative responsiveness, validates FluorMango as a reliable biosensor for fluoride ions.

We then tested whether the biosensor could detect fluoride released during bacterial defluorination of a fluorinated substrate.

FluorMango detects fluoride released from enzymatic defluorination

Objective: We investigated if FluorMango is compatible with the detection of fluoride in the presence of bacteria, i.e., under conditions close to the screening of DeHa2 variants with ORep E. coli cells.

Method: FluorMango was incubated with TO1-biotin, 2-fluoropropionate (2-FP) as the monofluorinated substrate for DeHa2, and E. coli cells expressing recombinant DeHa2 (from pUC19-replicon plasmid). Substrate degradation kinetics were monitored over 4 hours in 384-well plates by directly measuring the fluorescence of the FluorMango/TO1-biotin complex.

Results: We found that the defluorination of 2-FP and the accompanying production of F_- by DeHa2 (with IPTG) could successfully be detected by FluorMango (upper curve in Figure 17), while the fluorescence signal remained low in the control samples (no IPTG induction, only bacteria, only 2-FP). Summarizing, we provided a detailed characterization of FluorMango responsiveness, establishing this biosensor as a quantitative tool for the directed evolution of DeHa2.

Having validated FluorMango as a sensitive and specific biosensor in microplate assays and even in the presence of living bacteria, the next step was to test whether this technology could be translated into a droplet-based format. Microfluidic encapsulation not only provides the high-throughput capacity required for directed evolution but also allows linking enzymatic activity to individual genetic variants. We therefore explored whether FluorMango retains its functionality inside droplets and whether it can support the sorting of active versus inactive variants in a pooled screening setup.

FluorMango enables fluorescence detection of fluoride in microfluidic droplets

Objective: We aimed at detecting fluoride ions in microfluidic droplets using FluorMango.

Method: FluorMango was incubated with TO1-biotin, 1X refolding buffer and either NaF (1 mM, for positive-fluorescent droplets) or RNAse-free water (for negative droplets). Droplets were generated using two approaches: in the first one, FluorMango activation occurred only after encapsulation, whereas in the second one, FluorMango was pre-activated by fluoride prior encapsulation. Fluorescence of FluorMango/ TO1-biotin in droplets was measured by fluorescence microscopy.

Results: We visualized the positive and negative droplets collected from the microfluidic chip by fluorescence microscopy. With both sample preparation and droplet generation methods, no fluorescence was detected in negative control droplets, whereas positive control droplets consistently exhibited strong fluorescence (Figure 18). These results demonstrate the successful detection of fluoride ions by FluorMango in microfluidic droplets.

We conclude that its implementation in microfluidics does not impair biosensor functionality.

FluorMango allows for fluorescence-based sorting in microfluidics

Objective: We aimed at sorting active droplets based on fluorescence intensity thresholding from a mixed population of positive (with fluoride → fluorescent) and negative (no fluoride → nonfluorescent) droplets. This ‘mock library’ assay would validate the method for the prospective screening and selection of DeHa2 variants.

Method: Positive droplets and negative droplets generated using Method 2 (described above) were mixed at a 2:8 ratio (v/v) to create a mock library of active and inactive variants. Droplets were screened using the integrated fluorescence detector with 𝜆_excitation=488 nm and 𝜆_emission=525 nm and pre-sort and sort population analyzed using fluorescence microscopy.

Figure 19 shows a droplet sorting experiment which was performed under an oil pressure of 500 mbar and a sample pressure of 360 mbar to ensure efficient separation of droplets. When a droplet passing the detector exceeds the fluorescence threshold, the electrode is activated, deflecting the droplet toward the lower outlet for collection of the fluorescent droplets. Droplets with fluorescence below the threshold are directed to the upper outlet as waste.

Video 1: Droplets sorting

Results: Fluorescence-based sorting of the mixed droplet population (‘mock library’) yielded time-resolved events of screened positive droplets (Figure 20). These detection hits activated the sorting of droplets that were collected in an Eppendorf tube. We assessed the sorting efficiency by imaging the collected droplets under a fluorescence microscope. The images (bright field and fluorescence) revealed an enrichment of the positive droplets from 20% (pre-sort) to 80% after sorting (Figure 21).

We conclude that FluorMango, combined with fluorescence-based droplet microfluidic sorting, enables successful enrichment of positive droplets in a mock library containing an excess of inactive droplets. The ‘mock library’ validates this method for the prospective screening and selection of DeHa2 variants.

Successful construction of the pSEVA438-fluC plasmid

Objective: The fluoride-specific transmembrane transporter FluC endogenously expressed under the control of fluoride-responsive mechanisms was overexpressed in P. putida KT2440.

Method: The gene encoding for FluC was cloned into the expression plasmid pSEVA438 under the control of the strong xylS/P_m transcription promoter. P. putida was transformed with the resulting construct, pSEVA438-fluC. Following induction with m-toluic acid, the growth of the recombinant strain KT2440-pSEVA438-fluC was assessed in the presence or absence of 50 mM NaF by monitoring OD₆₀₀ kinetics in 96-well microplates in LB medium. Whole-plasmid sequencing was performed using nanopore technology at Eurofins Genomics.

Results: After cloning and transformation of P. putida, plasmid extraction, and purification, pSEVA438-fluC was sent for sequencing. The alignment shows that fluC has successfully been cloned in pSEVA438. See the complete sequencing of pSEVA438-fluC in the Materials page.

We then aimed to verify that FluC was successfully overexpressed by comparing the growth curve of the recombinant strain to that of wild-type P. putida in the presence of NaF, both with or without induction with m-toluic acid (Figure 22).

Figure 22A shows that for the WT strain, there is no significant difference in fluoride resistance with or without the inducer. This suggests that the presence of m-toluic does not affect bacterial growth. For P. putida KT2440-pSEVA438-fluC, the lag time is reduced by 2-fold compared to the WT strain in the presence of the inducer, while the growth rate is unchanged (Figure 22 B and C). These results suggest a positive effect on fluoride stress adaptation. Further investigation to characterize these effects were pursued.

Overexpression of fluC accelerates cellular adaptation to fluoride stress

Objective: We further investigated the effect of fluC overexpression on P. putida’s response to fluoride stress. Additional experiments were conducted in M9 minimal medium to better mimic the conditions encountered during water treatment.

Method: Growth of the recombinant strain P. putida KT2440-pSEVA438-fluC in M9 minimal medium was monitored in the presence of m-toluic acid inducer, across different NaF concentrations. Two individual carbon sources were used, acetate and propionate, as they are the expected degradation products of TFA and PFPrA. Lag time, specific growth rate, and Minimal Inhibitory Concentrations (MIC) were evaluated.

Results: The complete dataset and results are provided on the Kinetic Viewer. The growth kinetics of the recombinant strain on both carbon sources were compared, as shown in Figure 23.

In the absence of fluoride ions, P. putida pSEVA438-fluC grows similarly on propionate or acetate (Figure 23A). In fact, specific growth rates are equivalent with values of 0.58 h_-1 on acetate and 0.63 h^-1 on propionate (Figure 23B). When increasing the NaF concentration, the specific growth rates follow the same trend, and carbon source does show any influence. However, lag times for acetate are clearly reduced compared to that on propionate (Figure 23C). Overall, the MIC for NaF on acetate is clearly higher (over 100 mM) than the MIC on propionate (75 mM) (Figure 23D).

To further characterize the effect of the overexpression of fluC on growth, we then compared growth of the wild-type strain and the recombinant strain on acetate in the presence of different NaF concentrations (Figure 24).

We found that higher concentrations of NaF led to reduced growth rates (µ) and increased lag times for both strains (Figure 24). This confirms that the toxic effect of fluoride ions on bacterial growth is not confined to a threshold concentration leading to cell death, but encompasses a progressive impairment of metabolic activity (Qin et al., 2006).

Interestingly, the recombinant strain was less affected than the WT strain at a given NaF concentration. P. putida KT2440-pSEVA438-fluC was able to grow at NaF concentrations up to 100 mM, whereas the WT strain only grew up to 60 mM (Figure 24A). The resulting MIC for P. putida WT is congruent with the MIC found in literature in M9 minimal medium supplemented with glucose, which lies in the range of 75 mM (Calero et al., 2022) (Figure 24D). A difference of 25 mM is then observed in the MIC analysis, showing that the overexpression of the FluC transporter enables growth in the presence of higher fluoride concentrations.

The overexpression of fluC also accelerated the adaptation of P. putida to high fluoride concentrations. Specific growth rates were less affected by NaF for the recombinant strain, and lag times were significantly reduced compared to that of the WT strain (Figure 24B and C). For example at 40 mM NaF, the lag time of KT2440-pSEVA438-fluC was 12 h compared to 25 h for the WT, and the growth rate was 0.32 h^-1 versus 0.25 h^-1, respectively. Similar trends were observed in M9 medium supplemented with propionate: data are provided in the supplementary Figure S1. The complete dataset and results are provided on the Kinetic Viewer.