For the adaptation of E. coli towards usage of 6- and 7-fluoroindole a serial transfer regime using batch cultures under aerobic conditions was applied. The cells were propagated in synthetic minimal medium (New Minimal Medium, NMM) (Budisa et al., 1995) which consisted of: 7.5 mM (NH₄)₂SO₄, 8.5 mM NaCl, 22.5 mM KH₂PO₄, 50 mM K₂HPO₄, 1 mM MgSO₄, 20 mM D-glucose, 1 mg/L Ca²⁺ (as CaCl₂), 1 mg/L Fe²⁺ (as FeCl₂), 10 mg/L biotin, 10 mg/mL thiamine and 10 ng/L trace elements (Cu²⁺, Zn²⁺, Mn²⁺, MoO₄ ^2-). In addition, the medium was variably enriched with indole (2.5 µM–0 µM), fluoroindole (70 μM–30 µM) and amino acids (50 mg/L), and was dynamically adjusted during the adaptation process (Supplementary Tables S1–S4).

The cultivation of each experiment (adaptation to 6Fi, 7Fi, W, Ind) was executed in parallel using three biological replicates (assumed to be isogenic) of our metabolic prototype TUB00 (ΔtrpLEDC ΔtnaA), which is a derivative of E. coli MG1655. 10 mL culture were incubated in 100 mL-shaking flasks at 30°C, 180 rpm. When the cells reached their early stationary phase (usually after 50–65 h) fresh medium was inoculated to an OD₆₀₀ of 0.02 with the parent culture. The adjustment of the medium composition (increasing the selective pressure by depleting the indole and cAA concentration) was carried out when the cells proliferated sufficiently at least for two consequential passages under the parent conditions or no further growth improvement could be detected. The transfer steps of this serial dilution approach are indicated as “passages” and samples of each passage were preserved as cryo-stocks (25% glycerol, −80°C) for further analysis.

Beforehand ancestral TUB00 cultures were supplemented with different concentrations of indole (0 μM–30 µM) and fluoroindole (5 μM–1,000 µM) to determine the starting conditions. 2.5 µM indole and 70 µM of 6Fi or 7Fi were found to allow the cells to grow to a sufficient optical density of OD₆₀₀ 0.7–1.0. Within the ALE setup, the concentration of fluoroindole was kept high; then first the concentration of indole was gradually reduced until complete depletion (ALE phase 1) and subsequently the amino acid supply was stepwise ceased (ALE phase 2). For the amino acid removal two different approaches OC (overall concentration) and MB (metabolic blocks) were applied (Figures 1, 2). At this stage the evolving lineages separated into either three (6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3) or two (7TUB165-OC, 7TUB165-MB) independent strains. In case of the 7-fluoroindole ALEs the growth conditions had to be relaxed (reduction of 7Fi concentration and increasing of cAA supply) in order to retain sufficient cell growth.

The same procedure as for the Fi-adaptation was followed with the positive controls, adapted to grow on 70 µM of either indole (Ind-TUBX) or tryptophan (W-TUBX). These cells were cultivated under the same growth regime for 165 passages, except for the amino acid removal process of ALE phase 2 (NMM19-0-70), which was omitted (Supplementary Figure S8).

Final isolates of the adapted strains 6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3, 7TUB165-OC, 7TUB165-MB and positive controls Ind-TUB165, W-TUB165 as well as TUB00 were grown in the respective NMM supplemented with either 70 µM indole, 6Fi or 7Fi. Exponential growing cells were harvested by centrifugation (18,000 g, 4°C, 20 min) and pellets were stored at −20°C. E. coli extracts were briefly run into an SDS-gel until the dye front had migrated about 5 mm into the gel and only the 3 mm behind the dye front were used for protein digestions. This step removes all positively charged molecules, small highly-mobile negatively charged ions, uncharged molecules and very large agglomerates that did not enter the gel. Following a protocol described by Kublik et. al. (Kublik et al., 2016), the gel slices were washed with ddH₂O, and the proteins were reduced with dithiothreitol to break disulfide bridges. To protect the reduced cysteine residues, carbamidylation using iodoacetamide was performed. Subsequently, an overnight trypsin digestion was initiated to generate peptides. The resulting peptides were extracted from the gel pieces and desalted using C18-tips through zip-tipping. These extracted peptide samples were then analyzed by nLC-MS/MS on a nanoUPLC system (nanoAcquity, Waters) coupled to an Orbitrap Fusion mass spectrometer (Thermo Scientific), as previously described (Seidel et al., 2018). Protein identification was achieved using Proteome Discoverer (v2.4, Thermo Fisher Scientific) with the E. coli fasta genome database and SequestHT as the search engine, setting a false discovery rate threshold of 1% for peptide identification using the Target Decoy PSM Validator node. The abundance of proteins was quantified through label-free quantification based on intensity values in precursor scans using the Minora node in Proteome Discoverer. For the identification of peptides, carbamidomethylation of cysteine residues was taken as a fixed modification, oxidation of methionine residues and fluorination of tryptophan residues were taken as dynamic modifications. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD048225 (Perez-Riverol et al., 2022; Deutsch et al., 2023). Fluorination was calculated as the substitution of a hydrogen atom by fluorine giving a mass change of +17.991 Da (Supplementary Figure S9).

Isolates of the ancestral strain TUB00 and of the final isolates W-TUB165, Ind-TUB165, 6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3, 7TUB165-OC and 7TUB165-MB were subjected to comprehensive growth analysis using different medium compositions along the adaptation course (growth curves see Figure 8, Supplementary Figure S11). These comprised LB (rich, undefined medium), NMM19-0-70 (Ind), NMM19-70-1 (6Fi or 7Fi, Ind), NMM19-70-0 (6Fi or 7Fi) and NMM0-70-0 (6Fi or 7Fi). The lineage nomenclature follows Z-TUB-X-OC/MB pattern where “Z” indicates the Trp-source (Ind, Trp, 6Fi or 7Fi), “X” the passage number and “OC/MB” the cAA removal method; and the growth conditions were abbreviated in all cases as NMM([x]a-b-c), where x is the concentration of cAAs present in mg/L (only indicated when the concentration was below 50 mg/L); a is the number of different cAAs present; b is the concentration of 6Fi or 7Fi in µM; and c is the concentration of indole in µM. Thereby, 6Fi adapted cells could be directly resuscitated in NMM0-70-0 (6Fi) from cryo stock, whereas 7Fi adapted cells had to be revived in nutrient rich LB medium, followed by two-times serial inoculation in their final adapted medium NMM0-70-0 (7Fi supplemented, in order to restore the intracellular Trp-free environment). Subsequently, the cells were washed twice with NMM0-0-0 to remove residual Ind, 6Fi or 7Fi, normalized to OD₆₀₀ = 1, and then 200 µL of the respective medium composition were inoculated to OD₆₀₀ = 0.02. The growth was monitored in 96-well plates (Greiner, flat bottom clear), sealed with breathable membrane (Breathe-Easy sealing membrane, Sigma Aldrich) at 30°C with continuous shaking using a microplate reader (Tecan reader Infinite M200) that measured the absorbance at 600 nm in 10 min intervals. The measurements were performed at least twice and in triplicates (biological replicates).

The cell viability was determined using cell counting kit-8 (CCK-8; Sigma-Aldrich), which is based on dehydrogenase activity detection in viable cells. For the proliferation assay 190 µL of 5.0 × 10⁰⁷ cells/mL suspension were seeded, added with 10 µL of the CCK-8 solution and incubated for 4 h at 30°C without shaking. Then, the absorbance at 450 nm was measured using microplate reader (Tecan reader Infinite M200), which is directly proportional to the number of proliferating cells. As background control medium without cells was used and the viability of TUB00 was set to 100% in comparison to the adapted strains. Measurements were performed in triplicates and repeated at least twice.

For testing the antibiotic susceptibility, the minimal inhibitory concentration (MIC) of vancomycin was determined (Figure 8, Supplementary Figure S12). Isolates from cryo stocks of the respective strains were resuscitated either in 2–3 mL LB or appropriate NMM, incubated at 30°C, 200 rpm and inoculated into NMM if necessary. Subsequently, cells were used to inoculate 3 mL NMM in culture tubes as follows: NMM0-0-70 (Ind or Trp) for TUB00, W-TUB165, Ind-TUB165 and NMM0-70-0 (6Fi or 7Fi) for adapted strains 6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3 and 7TUB165-OC, 7TUB165-MB. For determination of the MIC a cell density corresponding to 0.5 McFarland standard (OD₆₀₀ = 0.1) was used and cells were treated with serial dilutions of vancomycin (0, 25, 50, 100, 200, 400 μg/mL). Cell growth (incubation at 30°C, 200 rpm) was monitored by measuring the OD₆₀₀ after 24 h and MIC was defined as the lowest concentration at which no increase in cell density could be observed (OD₆₀₀ = 0.1). Measurements were performed using biological triplicates.

Variants of the green fluorescent protein (GFP) were expressed in the final isolates 6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3 and 7TUB165-OC, 7TUB165-MB, as well as TUB00. These variants are the C-terminal his-tagged enhanced green fluorescent protein (EGFP-H6) and the N-terminal his-tagged enhanced cyan fluorescent protein (H6-ECFP). The strains were rendered electro competent (by three times washing in 10% glycerol) and transformed with the respective expression plasmids pQE80L EGFP-H6 or pQE80L H6-ECFP (both endowed with ori ColEI, AmpR).

Inoculated from cryo stocks, precultures of transformed cells were incubated in NMM0-70-0 (6Fi or 7Fi), supplemented with 100 μg/mL ampicillin (Amp) and subsequently 150 mL expression culture (NMM0-70-0 (6Fi or 7Fi), 100 μg/mL Amp) were inoculated to an OD₆₀₀ of 0.05. NMM0-0-70 (Ind) was used accordingly for the TUB00 control. The cultures were incubated at 30°C, 180 rpm until they reach exponential growth phase (OD₆₀₀ 0.4–0.7). Recombinant protein expression was induced by adding 0.5 mM isopropyl ß-D-1-thiogalactopyranoside (IPTG) and performed overnight at 30°C, 180 rpm.

Cells were harvested by centrifugation (20 min, 10,000 g, 4°C) and resuspended in B-PER™ (Bacterial Protein Extraction Reagent, Thermo Fisher Scientific) according to manufacturer’s instructions. The lysis mixture contained B-PER^™, lysozyme, DNAse and was incubated 10–15 min at room temperature. After centrifugation of cell debris (5 min, 15,000 g, RT), the cell lysate was filtered (Rotilabo PVDF Filter, 0.45 µm pore diameter) and target proteins were purified by using affinity chromatography on Protino Ni-NTA column (1 mL Fast Flow, Machery-Nagel). The purification was carried out using a single-channel peristaltic pump P-1 (VWR) and sodium phosphate-based purification buffer (50 mM Na-P, 300 mM NaCl, two-step imidazole gradient: 20 mM and 40 mM, pH 7.8); followed by elution with 500 mM imidazole. EGFP or else ECFP containing fractions were identified and chromophore maturation was proved by detection of fluorescence by irradiation with a UV lamp at 365 nm. Respective elution fractions were collected, and imidazole was removed by dialysis (Microsep^™ Advanced Centrifugal Devices, MWCO 3kDa, Pall Laboratory) against 50 mM Na-P, 100 mM NaCl (pH 7.8). Finally, protein identity and incorporation of 6- and 7-fluorotryptophan was confirmed by LC-ESI-Q-TOF mass spectrometry (Agilent 6530 Accurate-Mass Q-TOF). Protein samples were analyzed in a concentration of 0.1 mg/mL using the following HPLC parameters: linear gradient from 5% to 80% buffer A within 20 min (A: 0.1% formic acid in MQ-H₂O, B: 0.1% formic acid in acetonitrile), flow rate 0.3 mL/min, injection volume 5 µL. For the MS spectrum a range of 27,000–29,000 amu was selected in the total ion current (TIC) plot and the maximum entropy deconvolution algorithm was applied. The MS spectra are shown in Figure 9 and Supplementary Figure S13, the mass values are shown in Supplementary Table S6 and the isolated yields in Supplementary Table S7.

We started the ALE experiments with the metabolic prototype TUB00 (E. coli MG1655 ΔtrpLEDC ΔtnaA) as ancestral strain, which possesses stringent Trp auxotrophy due to genomic deletions in the endogenous pathways of Trp biosynthesis and degradation (Supplementary Figure S4). However, the Trp processing enzymes tryptophan synthase (TrpS; trpBA) (Wilcox, 1974; Goss and Newill, 2006; Watkins-Dulaney et al., 2021) and tryptophanyl-tRNA synthetase (TrpRS; trpS) (Wong, 1983; Bacher and Ellington, 2001; Azim and Budisa, 2008) remained in the genome, exhibit a broad substrate promiscuity and enable TUB00 to produce FTrp from F-indole and to incorporate FTrp in the protein biosynthetic process (Supplementary Figure S5). Since the metabolic sources of indole (tnaA) and Trp (trpLEDC) are eliminated the ALE experiment requires bacteria to accept indole or a fluorinated indole analog and produce the corresponding amino acid via TrpS and L-serine to survive. Thus, depriving the cells of canonical indole while supplying them continually with the fluorinated indole analog exerts selective pressure and brings about the adaptation process.

A conceptual overview of the ALE experiments we performed is shown in Figure 1. Four parallel experiments were designed: 1) a positive “indole control”, 2) a positive “Trp control”, 3) adaptation to 6Fi and 4) adaptation to 7Fi. For a legend explaining the nomenclature for all conditions and lineages obtained please refer to Figure 1. The controls serve to determine fluorine-induced effects and general consequences of long-term cultivation, respectively. We opted to conduct the experiments by allowing liquid cultures to grow under a given set of conditions until they reached early stationary phase; once this occurred, a constant cell number corresponding to an OD₆₀₀ value of 0.02 was used to inoculate the subsequent culture (defined as a “passage”). Unlike the alternative approach of “continuous cultivation”, this kind of serial transfer regime is characterized by alternating “feeding-and-starvation” regimes or “seasonal environments” (Vasi et al., 1994) (i.e., a population bottleneck), but benefits from its flexibility when adjustments to cultivation conditions are needed (Gresham and Dunham, 2014). Furthermore, we chose the serial transfer approach because this type of experimental management is closest to natural selection in that it mimics changing conditions in the real environment of natural populations and enabled cells to undergo all growth phases under each set of new conditions. Consistent inoculation of each new liquid culture with a normalized amount of cells ensures that even poorly growing cultures do not suffer an evolutionary disadvantage (e.g., our primary goal was not to select for rapidly growing populations) (LaCroix et al., 2017).

Three biological replicates of the starting point strain TUB00, which we assumed to be isogenic, were allowed to proliferate in synthetic minimal medium (New Minimal Medium, NMM) containing essential nutrients such as phosphate, ammonium, glucose, and vitamins. NMM was differentially supplemented as follows. Indole control populations grew in NMM with a constant concentration of 70 µM indole and high concentration of 19 of the 20 canonical amino acids (cAA); that is, all except Trp. Trp control populations grew in NMM with no indole but a constant concentration of 70 µM Trp and high concentration of the 19 cAAs (cultivation schemes shown in Supplementary Figure S8). Populations adapting to 6Fi and 7Fi grew in NMM variably enriched (see below) with indole, the 19 cAAs, and the respective fluorinated indole (see Supplementary Tables S1–S4 for medium compositions). To ensure the purity of the purchased 6Fi and 7Fi precursors, we determined the level of parent indole in batches by GC-MS and, indeed, indole was not detected within the limits of the method (see Supplementary Figures S1–S3).

The three phases of adaptation to 6Fi and 7Fi are represented in Figure 1. In phase 1, indole concentration is gradually, stepwise reduced to zero. In phase 2 the 19 cAAs are gradually, stepwise removed from the growth medium according to two distinct strategies (see paragraph below) to exclude possible contamination by traces of indole or tryptophan, which would disturb the adaptation process (Bacher and Ellington, 2001). Phase 3 represents the period of evolution in which only the fluorinated indole analog is present as NMM supplement. As noted above, the adaptation experiments designed here proceeded according to “passages” and it is important to note that the criteria that a population had to meet in order to be subjected to the subsequent growth condition were as follows: the population had to have reached early stationary phase under the given condition and proceeded through at least two full passages under that given condition; in some cases more than 2 passages were required to observe stable growth, or no further growth improvement was observed. This is why the number of days of growth does not correspond 1:1 with passage number. Only after these criteria were met were the conditions made more stringent to increase the evolutionary pressure being placed on the population.

Phase 2 was designed according to two different approaches: so-called “overall concentration (OC)” or “metabolic blocks (MB)”. OC and MB are visually represented in Figure 2. In the OC approach, the overall concentration of all 19 amino acids was equally reduced gradually and stepwise from 50 to 0 mg/L. The removal of amino acids in the MB approach was not based on their concentration, but rather on their origin in core metabolism (i.e., their metabolic identity). Canonical amino acids were also assigned to metabolic blocks considering factors such as their metabolic cost in terms of energy and precursor requirements (Akashi and Gojobori, 2002; Tolle et al., 2023). For example, the first block that was removed comprised the aromatic amino acids tyrosine and phenylalanine, both requiring laborious multistep biosynthesis and originating from the same starting metabolites (phosphoenolpyruvate and erythrose 4-phosphate). Whereas we and others successfully applied the MB approach in previous adaption experiments (Hoesl et al., 2015; Agostini et al., 2021), the OC approach presented here represents an innovation. We thought that the gradual reduction of all amino acids is more gentle to cells than the sudden elimination of whole metabolic blocks. This approach gently prompts cells to initiate their amino acid biosynthesis processes, avoiding abrupt ribosome arrest and mistranslation effects that result from a sudden depletion of amino acids (Wong et al., 2018).

Overall, Trp-auxotrophic E. coli tolerated 6Fi remarkably well (Figures 3–5). In phase 1, cells responded sensitively to the challenge of indole depletion, as evidenced by strongly fluctuating OD₆₀₀ values. However, after the populations ceased to be dependent on indole, cell growth remained stable and at a high level.

Surprisingly, the OC and MB approaches give approximately equal results in terms of adaptation speed and success. Adaptation took place in a similar time frame and the different lineages show equally good growth behavior. The mutual phase 1 lasted 53 passages, whereas the cells adapted via the OC approach took 31 passages in phase 2 and those for the MB approach (MB4 and MB3 cells) took 28 passages (Figures 3–5). When the amino acid supply in the MB-lineage was almost depleted (9 cAAs left, at passage 73) we tried the removal of different metabolic blocks, which yielded separate lineages 6TUB165-MB4 (Figure 4) and 6TUB165-MB3 (Figure 5). Although the cultivation schemes of the MB-lineages show large variations in cell densities upon removing the amino acid blocks “AVLI” and “NMT” (MB4) as well as “PEQR” (MB3), even removing different blocks was useful here, and the cells recovered very quickly.

Furthermore, both MB-lineages began to prefer 6Fi and to reject the non-fluorinated counterpart, because it was found that they grew better in 6Fi-supplemented NMM than in indole-supplemented NMM (Supplementary Figure S10). This behavior was not observed in clones of the 6TUBX-OC lineage; thus, its cultivation was terminated after 128 passages (in contrast to the 165 passages of the 6TUBX-MB lineages). In the course of a subpopulation screening (details described in the Supplementary Material), we selected five promising clones and started an additional ALE experiment with increased selection pressure by means of increased 6Fi concentration, from 70 µM to 85 µM to 100 µM, in order to provoke this phenotype. Ultimately, our goal was to create organisms that exhibit inherent adaptation to fluorine, aiming for cells that go beyond being merely facultative FTrp users. Unfortunately, this phenotype did not become dominant.

In contrast to 6Fi, tolerance towards 7Fi appears to be significantly more challenging as indicated by the low population sizes especially during the late adaptation process (Figures 6, 7). Already in phase 1, a high clonal variation in the acceptance to the growth conditions is observed; however, in sum similar high OD₆₀₀ values were sustained as in the same phase of the 6Fi adaptation experiments, indicated by the 6TUBX-OC/MB trace (green line) in Figures 6, 7. In phase 2 cell density decreased continuously in both lineages (OC and MB). We hypothesize that this resulted from the low-grade toxicity of 7Fi that became increasingly difficult for cells to compensate while undergoing cAA starvation. In case of the MB lineage, we alternatively tried the removal of different amino acid as well as performed replay experiments (Blount et al., 2018) with the same setup starting from passages 72 when growth was still acceptable to investigate if the sudden drop in the cell density is reproducible. All such attempts ended with drastic reductions in cell population.

In order to rescue the cells, we reduced the concentration of 7Fi while temporarily increasing the cAA supply both starting with passage 84 (displayed in the legends of Figures 6, 7). Finally, the cells slowly recovered upon the relaxed selection pressure and the concentration of 7Fi could again be raised to 70 µM. Still, this process demonstrated that cAAs had to be removed extremely conservatively (compared to 6Fi): in smaller steps in the OC approach, and not at all as metabolic blocks in the MB approach, but rather removed on an individual basis. Even then not all clones survived the treatment, and the population sizes were considerably lower compared to the 6Fi adaptations. Taken together, this underlines the great challenge for E. coli to accept and adapt to 7Fi. Based on the described hurdles, a noteworthy difference between the OC and the MB approach is difficult to be determined. The OC and MB approaches both produced functionally adapted cells, however, from an experimental point of view the OC approach was easier to implement.

We conducted extensive growth experiments of all strains in different media that reflect the conditions along the adaptation trajectory (Figure 8B, upper panel). Overall, the impression about the adaptation success gained from the cultivation schemes discussed above (Figures 3–7) is reflected in the growth curves (Figure 8A). 6Fi adapted cells show better growth performance than 7Fi adapted cells. The final isolates of the 6Fi adapted cells exhibit similar growth behaviors to one another in their final growth medium (NMM0-70-0), although 6TUB165-MB4 appears to be the best adapted lineage. In contrast, the 7Fi cells exhibit a significant increased lag time and decreased growth rate, which is more pronounced in the 7TUB165-MB strain, which also shows a lower maximum OD₆₀₀ value.

Furthermore, Figure 8B provides an overview of the growth of all investigated strains along the adaptation course given by calculation of the specific growth rate (Hall et al., 2013) (for comprehensive view see detailed growth curves in Supplementary Figures S11A–H). Generally, the growth correlates with the concentration of provided nutrients. Interestingly, the ancestral TUB00 tolerates 7Fi better than 6Fi, although based on the cultivation schemes (Figures 6, 7) we had considered 7Fi to be more toxic. The growth performance of the positive controls in NMM19-0-70 approaches that in LB (Supplementary Figures S11B, C), which is comprehensible because these strains were adapted to grow on minimal medium.

Among the 6Fi adapted cells, the OC lineage exhibits the highest specific growth rate in all tested media. However, as also seen in the MB3 lineage, there is no further growth improvement observable after the transition from NMM19-70-0 to NMM0-70-0. Remarkably, the MB4 lineage displays such enhanced growth behavior. Although its growth rates are lower compared to both other lineages, in 6Fi supplemented NMM, the highest growth rate is observed in the final medium NMM0-70-0, which is only surpassed by growth in the ancestral medium. This finding is even more evident in the entire growth curves (Supplementary Figures S11D–F). This observation strongly indicates a fluoroindole preferring phenotype, which is also supported by the subpopulation screening results.

Both 7Fi-adapted strains show decreased growth rates along with reducing nutrient supply (cAAs, Ind). Indole supplemented media enables enhanced growth in the absence of 7Fi (NMM19-0-70 > NMM19-70-1), which is a clear signal for an incomplete adaptation, as these cells still strongly prefer indole over 7Fi. However, the growth performance in 7Fi supplemented medium with and without cAAs addition (NMM19-70-0, NMM0-70-0) is equal, suggesting that removal of the canonical amino acids is well accepted.

We completed the picture of bacterial fitness by determination of the cell viability using a CCK-8 assay (Figure 8C), which measures the bio-reduction of tetrazolium salts (WST-8) (Yang et al., 2021). The viability of all 6Fi adapted lineages correlates very well with the growth parameters, whereas among the 7Fi adapted strains there is a discrepancy found for 7TUB165-MB. This strain shows a cell viability of 150% compared to TUB00, although this seems inconsistent with growth curves and parameters.

Additionally, we investigated the membrane integrity of the adapted cells through determination of the susceptibility towards the antibiotic vancomycin, which is inactive against Gram-negative E. coli because the intrusion through the (intact) outer membrane is blocked (Figure 8D, Supplementary Figure S12). Compared to TUB00 (MIC 200 μg/mL) the minimal inhibitory concentration of the positive controls (MIC 100 μg/mL) shows increased vancomycin sensitivity. This may be due to temperature: on the one hand, membrane permeability in general is influenced by temperature, and on the other hand, cold stress is known to increase E. coli’s susceptibility to vancomycin (Stokes et al., 2016). The 7Fi adapted cells show a strongly increased vancomycin sensitivity (MIC 25 μg/mL), which could be attributed to a membrane rearrangement (Agostini et al., 2021). Interestingly, the strains adapted to 6Fi exhibit the opposite effect being highly resistant to vancomycin treatment (MIC 400 μg/mL). A potential explanation could be the accumulation of phosphatidic acid, a standard membrane component. However, it is assumed that an increased phosphatidic acid level hinders the penetration of vancomycin and thus causes resistance (Sutterlin et al., 2014). Certainly, additional research is needed to explore this hypothesis in the context of the adapted strains.

Global substitution of Trp for FTrp was shown by protein nano-electrospray-ionization-mass-spectrometry coupled with liquid chromatography (nLC-ESI-MS/MS) within the limits of the method. Trp-containing peptides were exclusively detected in TUB00 and both positive controls, while FTrp-containing peptides were only present in the isolates of adapted strains (Supplementary Figure S9). Despite of our careful experimental setup (strict prevention of Trp import in the growth medium and Fi substrates), the following background Trp incorporation was detected. Occurrence of FTrp-containing peptides in positive controls: TUB00 (0.0225% ± 0.0318), Ind-TUB165 (0.0878% ± 0.0098), W-TUB165 (0.0485% ± 0.0485), and occurrence of Trp-containing peptides in adapted strains: 6TUB128-OC (0.2309% ± 0.1315), 6TUB165-MB4 (0.1702% ± 0.0555), 6TUB165-MB3 (0.1269% ± 0.1269), 7TUB165-OC (0.3984% ± 0.0822) and 7TUB165-MB (0.5154% ± 0.1173); the mean of three biological replicates is reported, respectively. These values are within the false discovery rate threshold of 1% for peptide identification of the method.

Furthermore, the 6Fi and 7Fi adapted strains were tested for their robustness in driving recombinant protein expression (Kuthning et al., 2016). Two different variants of the green fluorescent protein (GFP) model protein were used for heterologous expression: 1) the enhanced green fluorescent protein EGFP, which harbors one single Trp residue at position 57 in its primary sequence, and 2) the enhanced cyan fluorescent protein ECFP, which additionally contains a Trp in the chromophore at position 66 (cartoons in Figure 9).

Cells of the final isolates of all adapted strains (6TUB128-OC, 6TUB165-MB4, 6TUB165-MB3 and 7TUB165-OC, 7TUB165-MB) and TUB00 were transformed with a plasmid bearing the sequence of either His-tagged EGFP (C-terminal EGFP-H6) or ECFP (N-terminal H6-ECFP). The recombinant protein expression was performed under the same conditions as used for adaptation, meaning cultivation in NMM0-70-0 containing either 6Fi or 7Fi at 30 °C; followed by isolation and purification using Ni-NTA affinity chromatography and analysis via LC-ESI-MS (Supplementary Figure S13 and Supplementary Table S6).

Almost all adapted strains support recombinant expression of GFP variants in which all Trp residues are quantitatively replaced by fluorinated counterparts. Mass spectrometric analyses of 6FTrp- and 7FTrp-substituted EGFP and ECFP show homogeneous mass profiles associated with an expected mass shift of 18 Da (which correlates with a H-to-F exchange on the aromatic ring of the indole). The expected theoretical values for translated EGFP and ECFP after chromophore maturation (Lundqvist et al., 2019) are consistent with the observed masses (Figure 9). However, the amount of 6FTrp-tagged EGFP expressed with 6TUB165-MB4 was very low; and the 7TUB165-OC strain was either able to produce only small amounts of 7FTrp-EGFP or was even unable to form 7FTrp-ECFP at all. The successful expression confirmed that in the adapted strains protein synthesis was also reprogrammed as Trp was completely replaced by 6FTrp or 7FTrp, respectively.