FAST-mediated fluorescence greatly depends on the choice of the FAST tag. Because protein stability can differ significantly among variants, we first investigated the behavior at higher temperatures of Y-FAST and pFAST, the earliest and latest FAST tag versions, respectively (Plamont et al., 2016; Benaissa et al., 2021). Preliminary in vitro assays showed that the melting temperature for Y-FAST and pFAST in complex with ^TFLime were substantially different (Figure 1B). In the case of pFAST, the observed T_m was close to 70 °C (versus ≈47°C for Y-FAST), underlining its potential compatibility for use as a fluorescent reporter with bacteria from the Thermoanaerobacter genus, which typically display an optimal growth temperature between 60°C and 70°C.

Thermoanaerobacter kivui DSM 2030 was chosen as the model system for this study because of its proven track record for forward genetics approaches. T. kivui has notably been shown to be naturally competent, and can be transformed with pMU131 derivatives, an E. coli–Thermoanaerobacter shuttle plasmid (Shaw et al., 2010; Basen et al., 2018). In addition, gene coding for natural homologs of FAST proteins could not be found in T. kivui genome, indicating FAST could likely be used without prior modification of the strain.

To facilitate the construction of pFAST expression cassettes, we adapted pMU131 to be compatible with a simplified hierarchical Golden Gate assembly system based on GoldenMOCS (Golden Gate-derived Multiple Organism Cloning System), initially established in Aspergillus niger (Sarkari et al., 2017) (Figure 2A). GoldenMOCS systems are modular cloning systems designed for fast and efficient construction of plasmids for a variety of purposes, such as constructing complex metabolic pathways (Novak et al., 2020) or genome editing plasmids (Prielhofer et al., 2017). In our approach, this system allows combining three genetic parts into a pMU131 derivative, which can be used to easily exchange promoters, CDSs and terminators.

The GoldenMOCS system was next exploited to generate shuttle plasmids bearing pFAST expression cassettes. The endogenous pta promoter from T. kivui was first tested to drive the expression of pFAST. The product of the pta gene, phosphotransacetylase, is essential for acetate production from acetyl-CoA. Acetyl-CoA is the terminal product of the WLP, which acetogens such as T. kivui rely on for biomass and energy generation. For homoacetogens (such as T. kivui), excess acetyl-CoA (i.e., not used for biomass generation), is solely used to generate ATP by substrate level phosphorylation, with acetate as the sole product. Because only one pta gene could be found in T. kivui (Hess et al., 2014), we therefore hypothesized its promoter would be strong and constitutive.

After introducing the resulting plasmid pThermoFAST_01 expressing a codon-optimized pFAST under the control of pPta_Tkv, fluorescence was assayed with a plate reader using medium to late exponential phase cells. Green or red fluorescence was observed when cells were suspended in phosphate buffer containing ^TFLime or ^TFCoral (Figure 2B). When cells were grown at their optimal temperature (66°C), a relatively weak fluorescent signal was observed corresponding to appr. 9 (^TFLime) or 7 (^TFCoral) times the signal observed for the negative control (thereafter defined as fold change FC_neg—comparison to cells with the empty backbone plasmid pMU131). We speculated low fluorescence could be due to a potentially high proportion of protein denaturation as growth temperature was close to the melting temperature of pFAST. In addition, the T_m determined for the pFAST-^TFLime complex is potentially not representative of the T_m of pFAST alone, as ligands are susceptible to increase their cognate protein stability (Celej et al., 2003). This potentially beneficial effect of ligand binding on protein thermostability led us to cultivate T. kivui in the presence of 20 µM ^TFLime at 66°C. This approach however proved unsuccessful, presumably because the fluorogen degraded during the time necessary for growth (data not shown). A second strategy consisted in reducing growth temperature to 55°C. Under this condition, the fluorescence signal significantly rose, with an appr. 43 (^TFLime) and 17 (^TFCoral) FC_neg.

Having established a functional fluorescent reporter system compatible with thermoanaerobic growth conditions, we next expanded the T. kivui genetic toolbox by constructing promoter parts compatible with our GoldenMOCS cloning system, which we characterized for their ability to drive pFAST expression. We focused on promoters that were previously used in Thermoanaerobacter (T. kivui S-layer promoter pSlay_Tkv, Thermoanaerobacter sp. X514 gyrase promoter pGyr_X514, pMU131 kanamycin resistance promoter pKan_pMU) (Basen et al., 2018; Jain et al., 2021) and in related bacteria belonging to the Clostridium phylogenetic class (Clostridium acetobutylicum thiolase promoters pThl_Cac and minimal promoter minipThl_Cac, Clostridium sporogenes ferredoxin promoter pFdx_Csp) (Tummala et al., 1999; Dong et al., 2012; Ng et al., 2013). We also added Acetobacterium woodii phosphotransacetylase promoter pPta_Awo, T. kivui glyceraldehyde-3-phosphate dehydrogenase promoter pGap_Tkv as well as potentially inducible promoters (T. kivui fructose repressor and glycine cleavage system promoters pFruR_Tkv and pGcv_Tkv) and a novel synthetic promoter based on pKan_pMU (called pKan_pMU*), for which the 5′-UTR sequence (comprising the ribosome binding site, RBS) was exchanged with the one from pPta_Tkv.

After cloning the genetic parts in BB2_pMU131, we measured ^TFLime-pFAST fluorescence levels for cells grown at 55°C (Figure 2C). The highest fluorescence was recorded for pPta_Tkv, which confirmed our initial hypothesis that this promoter allows for strong and constitutive gene expression in T. kivui. pSlay_Tkv and pGyr_X514 were confirmed to be also relatively strong, with 54% and 20% of the strength of pPta_Tkv, respectively.

Interestingly, heterologous promoters from related mesophiles were also shown to be functional in T. kivui. In particular, pThl_Cac and minipThl_Cac exhibited an interesting behavior. Indeed, minipThl_Cac mainly differs from pThl_Cac by a 84 bp truncation between the transcription start site and the RBS (Dong et al., 2012), which in our case reduced expression by more than half, indicating that the deleted 5′-UTR has a positive effect on expression in T. kivui.

For a few promoters, we could not detect any fluorescence (pGap_Tkv, pPta_Awo, and pKan_pMU), which suggests 1) that these promoters are either relatively weak or inactive and 2) that our reporter system was possibly not sensitive enough to quantify low fluorescence levels (supported by a relatively narrow dynamic range, in terms of maximum FC_neg). By replacing the 5′-UTR sequence from pKan_pMU with the 5′-UTR from the pPta_Tkv promoter (pKan_pMU*), we were able to measure low fluorescence levels, indicating that pKan_pMU likely drives pFAST expression, albeit at a lower level than pKan_pMU* (not detectable with our reporter system). Similarly, we could not detect fluorescence for pFruR_Tkv and pGcv_Tkv, with and without supplementing the potential inducers fructose or glycine, respectively (data not shown).

In the previous experiments, the dynamic range of fluorescence at 55°C and 66°C was relatively narrow (up to 43 and 9 FC_neg, respectively). Plate reader measurements give an output that corresponds to the fluorescence of a cell population, which could potentially be heterogeneous (Mook et al., 2022). Using flow cytometry, we next quantified fluorescence at single-cell resolution and found a significant portion of cells in the T. kivui pThermoFAST_01 (pPta_Tkv) population not to be fluorescent when cultivated at 66°C (43%, Figure 3A), suggesting part of the cell population does not bear or sufficiently express the pFAST gene. Population heterogeneity was also observed at 55°C, in which case both populations exhibited fluorescence (when compared to the empty plasmid population), suggesting both populations expressed pFAST but at different levels. In addition, we compared fluorescence intensity of cells identified as fluorescent (upper panels of Figure 3A) when grown at 55°C and 66°C. We notably observed a stark decrease of fluorescence for the 66°C condition, with a mean of 2,365 relative fluorescence units (rFU) at 55°C compared to 763 rFU at 66°C.

Taken together, the flow cytometry measurements suggested that pMU131 derivatives are not segregationally stable in T. kivui, with stability sharply decreasing with increasing temperatures. To enhance plasmid stability, we serially cultivated wild-type T. kivui with pThermoFAST_01, maintaining antibiotic pressure along the process, thereby favoring growth of cells that stably and/or strongly expressed genes on the plasmid (such as the antibiotic resistance gene, and the pFAST fluorescent marker) (Figure 3B). This adaptive laboratory evolution approach [ALE (Dragosits and Mattanovich, 2013)] allowed us to obtain a strain with a drastic increase in pFAST expression (Figure 3C). At 55°C, fluorescent signal for the evolved strain was appr. 210 FC_neg (≈4.5-fold increase compared to T. kivui pThermoFAST_01). Fluorescence increased the most at 66°C (≈7-fold increase compared to T. kivui pThermoFAST_01), though the dynamic range was not as high as at 55 °C (FC_neg ≈ 63). Flow cytometry analysis further revealed that fluorescence intensity per cell and proportion of fluorescent cells were both increased in the evolved strain at 66°C (mean of 1,230 rFU, 92.7% of the cell population being fluorescent, Figure 3D).

We hypothesized that the lower population heterogeneity and increased intensity obtained through our ALE approach could be linked with a higher stability of the plasmid at high temperatures, resulting in a higher overall fluorescent signal. To investigate DNA modifications arising at the plasmid level in the evolved strain, we extracted DNA from both the initial and evolved T. kivui pThermoFAST_01, amplified the plasmids in E. coli using kanamycin as selective pressure, and submitted the resulting vectors for whole plasmid sequencing. While in the initial strain plasmid sequencing showed a single plasmid corresponding precisely to pThermoFAST_01, in the evolved strain sequencing revealed a mixed population of at least two novel plasmids. These two new plasmids display a distinct enzymatic restriction profile on agarose gel electrophoresis, which was used to isolate them in E. coli. Coined pThermoFAST_01* and pThermoFAST_01**, both plasmids are shorter in size than pThermoFAST_01 (6,387 and 4,488 bp, respectively, compared to 6,984 bp) (Figure 4). Further sequence analysis revealed both plasmids display a truncation of the Gram-positive origin of replication at the same position, resulting in the deletion of three open reading frames (ORFs) coding for short unknown proteins (96, 108, and 57 amino acid residues), but leaving the rep gene intact (likely directly involved in Gram-positive plasmid replication). For pThermoFAST_01*, the 1,029 bp truncation is accompanied by a 432 bp insertion of T. kivui gDNA that maps at a single specific locus and contains part of the transposase gene TKV_RS00225. In contrast, pThermoFAST_01** does not contain this insertion and shows an extended deletion that spans along the ampicillin resistance gene (used for E. coli plasmid construction) up to the start of the Gram-negative origin of replication. However, upon isolation of pThermoFAST_01* and pThermoFAST_01** and reintroduction into wild-type T. kivui, fluorescence levels similar to the original pThermoFAST_01 was observed (data not shown). The increased fluorescence in the evolved strain therefore does not appear to be linked with the deletion of the three ORFs in the Gram-positive origin of replication. Consequently, we further examined whole plasmid sequencing data with respect to identifying additional mutations which might explain the higher fluorescence levels observed for the evolved strain. However, no significant mutation other than the truncation of the Gram-positive origins of replication could be found, indicating that the increased fluorescence displayed by the evolved strain is not a result of plasmid modifications.

Determination of melting temperature was performed using the Real-Time PCR System (QuantStudio 5). Proteins [prepared as previously described (Benaissa et al., 2021)] were diluted to 1 μM in PBS buffer and mixed with 10 μM ^TFLime. Samples were subjected to thermal ramp from 20°C to 95°C with a ramp rate of 1°C/min during which fluorescence was recorded with the following parameters: Excitation: 470/15 nm and emission: 520/15 nm.

QuantStudio Design and Analysis Software v1.5.2 was used to calculate the T_m. Melting curves were characterized by two phases: a first phase during which the fluorescence decreases linearly and a second phase where the decrease is more pronounced. The first phase corresponds to the dissociation of the non-covalent protein:fluorogen complex upon increase of temperature. The second sigmoidal phase corresponds to the actual denaturation of the protein. The reported Tm value corresponds to the peak of the first derivative of the fluorescence data, and thus to the inflection point of the sigmoid curve.

Strains used in this study are described in Table 1.

E. coli was grown in LB medium (10 g L⁻¹ tryptone, 10 g L⁻¹ NaCl, 5 g L⁻¹ yeast extract) at 30°C, with carbenicillin (100 mg L⁻¹) or kanamycin (50 mg L⁻¹) as needed. For solid media, 15 g L⁻¹ agar was added.

T. kivui LKT-1 DSM 2030 was handled and maintained anaerobically as previously described (Hess et al., 2014), with some modifications. All experiments were conducted in a vinyl anaerobic chamber (Coy lab, Grass Lake, United States) or in an anaerobic workstation (Concept 500, Baker Ruskinn, Bridgend, UK). Mineral medium was used throughout the study with the following composition, per liter: 5.5 g glucose monohydrate, 7.8 g Na₂HPO₄*2H₂O, 6.9 g NaH₂PO₄*H₂O, 210 mg K₂HPO₄, 160 mg KH₂PO₄, 250 mg NH₄Cl, 225 mg (NH4)₂SO₄, 438 mg NaCl, 91 mg MgSO₄*7H₂O, 6 mg CaCl₂*2H₂O, 2 mg FeSO₄*7H₂O, 5.41 g KHCO₃, 500 mg Cysteine-HCl*H₂O and 10 mL DSM141 modified Wolin’s mineral solution. 2 g L⁻¹ yeast extract was only added for transformation experiments and plates. Plates were made using the same medium supplemented with either 8 g L⁻¹ Noble agar (Carl Roth, Karlsruhe, Germany) or 3 g L⁻¹ Gelrite (Carl Roth). 200 mg L⁻¹ kanamycin or neomycin were added when required. Hungate tubes or 125 mL serum bottles were used for anaerobic cultivation. Prior to cultivation, the headspace was flushed with a gas mixture containing 20% CO₂ and 80% N₂, and the pressure set to 2 bar. A rotary shaking incubator (55°C) or a shaking water bath (66°C) were used for cultivation.

Plasmids used in this study are described in Table 1. DNA parts used in this study are further described in Supplementary File S1.

Site directed mutagenesis was first performed to remove a BbsI restriction site in the plasmid. A Golden Gate cassette containing BbsI recipient sites 1 and 4 was next inserted by enzymatic restriction between SacI and KpnI sites, yielding BB2_pMU131.

Next, Golden Gate assembly (GGA) was used to create pFAST expression plasmids. Promoters were first synthesized (Twist Bioscience) or cloned (PCR and BsaI GGA in BB1_12) in plasmids containing BbsI donor sites 1 and 2. pFAST was codon optimized for T. kivui and synthesized (Twist Bioscience) in a plasmid containing BbsI donor sites 2 and 3. The Kan terminator from pMU131 thermoKanR gene was synthesized (Twist Bioscience) in a plasmid containing BbsI donor sites 3 and 4. Promoters, pFAST and tKan parts were subsequently combined by GGA in BB2_pMU131 using BbsI-HF (NEB), yielding the pThermoFAST plasmid suite.

Plasmids were introduced in T. kivui by natural competency as previously described (Basen et al., 2018). Briefly, cells were cultivated overnight at 66°C in 2 mL mineral medium supplemented with 2 g L⁻¹ yeast extract and 1 µg of plasmid. Cells were then embedded in hot selective medium at various dilutions and left under anoxic conditions until plates solidified. Neomycin was preferred over kanamycin for initial selection. Plates were incubated until colony formation at 66°C in a custom anaerobic jar under 2 bar N₂:CO₂ (80:20). Plasmid presence was assessed by PCR targeting the pFAST expression cassettes using colonies grown in liquid cultures as DNA templates.

To measure expression using pFAST as a reporter, a plate reader photometer was routinely used (Infinite M200, Tecan, Männedorf, Switzerland). Cells bearing the pFAST plasmids were cultivated in 3 mL mineral medium at 55°C or 66°C until mid-late log phase (0.8–2.0 OD₆₀₀ units). Cells were concentrated by centrifugation to 5 OD₆₀₀ units in 70 mM sodium phosphate buffer pH 7.4 with 5 µM fluorogen (^TFLime or ^TFCoral, Twinkle Factory, Paris, France). Fluorescence of 100 µL of cell suspension was quantified at 30°C in relative fluorescence units (^TFLime: excitation 485/20 nm, emission 535/25 nm, gain 50 and 60; ^TFCoral: excitation 485/20 nm, emission 595/35 nm, gain 75). Fluorescence of a well containing 100 µL of the fluorogen solution without cells was subtracted to the values obtained. Results were normalized using the fluorescence value obtained for cells bearing pMU131 (no insert control).

Single-cell fluorescence was measured using flow cytometry (Cube 8, Sysmex, Kobe, Japan). ^TFLime-pFAST fluorescence was monitored using a blue solid-state laser (488 nm) for excitation and a green filter (536/40 nm) for emission. Cells were grown to ≈1 OD₆₀₀ unit and 5–10 µL were mixed with 1 mL H₂O containing 5 µM ^TFLime. At least 50,000 events were analyzed for each sample. Analysis was performed using FlowJo (v10.8.2).

DNA was extracted from overnight cultures of T. kivui (Hi Yield Mini Plasmid DNA extraction kit, SLG) and directly introduced into E. coli. Cells were selected with kanamycin in liquid and plasmid DNA was purified, and sent for whole plasmid sequencing (Plasmidsaurus). The resulting files were analyzed using Geneious Prime 2023.1.1 (https://www.geneious.com).