Introduction

The acetic acid bacterium (AAB) Gluconobacter oxydans harbors the beneficial ability of regio- and stereoselective incomplete oxidation of a variety of substrates (e.g., sugars and sugar alcohols) in the periplasm by membrane-bound dehydrogenases (mDHs) and release of resulting products into the cultivation medium (Mamlouk and Gullo 2013; Mientus et al. 2017; Pappenberger and Hohmann 2014). Because of this feature, G. oxydans is industrially used for oxidative biotransformations of carbohydrates to produce e.g. the vitamin C precursor l-sorbose, the tanning lotion additive dihydroxyacetone, and 6-amino-l-sorbose used for the production of the antidiabetic drug miglitol (Ameyama et al. 1981; Gupta et al. 2001; Hekmat et al. 2003; Saito et al. 1997; Tkac et al. 2001; Wang et al. 2016).

For the expression of target genes in G. oxydans, so far only constitutive promoters have been used due to the lack of a regulatable promoter demonstrated to be functional and tunable in G. oxydans (reviewed in Fricke et al. 2021). Only recently the first tight system became available for tunable induction of gene expression in G. oxydans. This system is based on AraC-ParaBAD and the induction by l-arabinose binding to the regulator protein AraC (Fricke et al. 2020). AraC typically represses the target promoter ParaBAD by DNA bending in the absence of the inducer l-arabinose and activates it in the presence of the inducer by a modified binding to the promoter DNA and thereby releasing the bending (Schleif 2010). However, in G. oxydans, ParaBAD is almost not active in the absence of araC and thus, repression by AraC is not required in G. oxydans for tightness of ParaBAD in the absence of the inducer (Fricke et al. 2020). In contrast, in Gluconacetobacter and Komagataeibacter plasmid-based AraC-ParaBAD was reported to be very leaky (Teh et al. 2019). While AraC typically is acting both as repressor and activator, the transcriptional regulators TetR and LacI only exert a repressor function and dissociate from their operator DNA when forming a complex with their respective inducer (Hillen et al. 1983; Khoury et al. 1991; Miller 1970; Sellitti et al. 1987; Wray and Reznikoff 1983). Ptet and Plac then enable transcription of the downstream gene by RNA polymerase. In AAB both TetR- and LacI-dependent target gene expression have so far been reported to be very leaky or inhomogeneously induced. In Komagataeibacter rhaeticus iGEM, the TetR-Ptet system from transposon Tn10 exhibited only approximately 1.5-fold induction due to high leakiness in the absence of the inducer anhydrotetracycline (Florea et al. 2016). With a LacI-based expression system in K. xylinus, the induction ratio also appeared to be low and only a fraction of cells showed some induction (Liu et al. 2020). For another LacI-based system, expression was also found to be very leaky in G. oxydans (Condon et al. 1991).

In this study, we aimed to test TetR- and LacI-dependent target gene expression in G. oxydans by applying the same plasmid backbone and construction principles that we have used recently for the construction of the araC-ParaBAD system (Fricke et al. 2020). With pBBR1MCS-5-based plasmid constructs, we found always some leakiness of the LacI-Placuv5 system, yet extremely tight and optimally tunable target gene expression with TetR-Ptet making the latter system highly suitable for applications in G. oxydans and possibly in other AAB.

Materials and methods

Bacterial strains, plasmids, and culture conditions

All strains and plasmids created and used in this study are listed in Table 1. G. oxydans was routinely cultivated in d-mannitol medium (pH 6) containing 4% (w/v) d-mannitol, 5 g L−1 yeast extract, 2.5 g L−1 MgSO4 × 7 H2O, 1 g L−1 (NH4)2SO4, 1 g L−1 KH2PO4 at 30 °C and 180 rpm, and supplemented with 50 µg mL−1 sodium cefoxitin. Besides d-mannitol and cefoxitin which were sterile filtered as stock solutions (20% (w/v) and 50 mg mL−1), all components were autoclaved for sterilization (120 °C, 20 min). Unless stated otherwise, for shake flask cultivations 10 mL or 60 mL of d-mannitol medium was inoculated from overnight precultures to an initial optical density at 600 nm (OD600) of 0.3 or 0.2 (UV-1800, Shimadzu) using 100 mL or 500 mL shaking flasks with three baffles. G. oxydans carrying pBBR1MCS-5- or pBBR1MCS-2-based plasmids were supplemented with 10 µg mL−1 gentamicin or 50 µg mL−1 kanamycin, respectively (Kovach et al. 1995). Escherichia coli strains were routinely grown at 37 °C and 160 rpm in lysogeny broth (LB) medium which was supplemented when appropriate with 10 µg mL−1 gentamicin or 50 µg mL−1 kanamycin. G. oxydans was transformed by conjugal transfer of plasmids from E. coli S17-1 (Kiefler et al. 2017). Competent E. coli cells were prepared by CaCl2 procedure and transformed as described (Hanahan 1983).

Table 1 Strains and plasmids used or constructed in this study
Full size table

Recombinant DNA work

All DNA oligonucleotides used for the construction of plasmids and sequencing are listed in Table S1 and were synthesized by Eurofins MWG. All restriction enzymes were purchased from Thermo Scientific. Polymerase chain reaction (PCR), restriction, and ligation reactions for DNA manipulations were conducted according to standard protocols (Sambrook et al. 1989). DNA fragments were amplified using Q5 polymerases according to the conditions recommended by the manufacturer (New England Biolabs). Unless stated otherwise for the construction of all reporter plasmids, amplified DNA fragments were integrated in the restricted broad-host vector pBBR1MCS-5 in a one-step isothermal Gibson assembly (50 °C, 1 h) (Gibson et al. 2009). All cloning steps to obtain desired plasmids were conducted in E. coli S17-1 and plasmids were isolated using a QIAprep spin miniprep kit (Qiagen). Inserts of all constructed plasmids were checked for correctness by DNA sequencing (Eurofins MWG).

Construction of plasmids

For the construction of plasmids, the empty vector pBBR1MCS-5-TgdhM-MCS-T0028 was generated from pBBR1MCS-5. It carries the terminator sequences of GOX0265 (TgdhM) and GOX0028 (T0028) flanking the multiple cloning site (MCS) to minimize interfering effects between the plasmid backbone and expression of the inserted genes.

Plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 was constructed using the primer pair PF1/PF2 to generate a 763 bp DNA fragment with tetR-Ptet from plasmid pBBR1-tetall-strep_long and primer pair PF3/PF4 to generate a 802 bp DNA fragment with mNG and the terminator BBa_B1002 from the iGEM parts library from pBBR1MCS-5-araC-PBAD-mNG (Fricke et al. 2020). For insertion of the two DNA fragments in pBBR1MCS-5-TgdhM-MCS-T0028, the plasmid was restricted with XbaI and EcoRI.

Plasmid pBBR1MCS-5-tetR-Ptet-mNG-TBBa_B1002-T0028 lacking terminator TgdhM downstream from tetR was constructed by amplification of a 576 bp DNA fragment from pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 with primer pair PF5/PF6 and ligated with Eco81I/MunI-digested pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 replacing TgdhM-tetR by tetR only.

For the construction of pBBR1MCS-5-TgdhM-tetR-Ptet-mNG lacking the terminators TBBa_1002 and T0028 downstream from mNG, the primer pairs PF7/PF8 and PF9/PF10 were used to amplify a 1,471 bp DNA fragment comprising tetR-Ptet-mNG and a 643 bp DNA fragment comprising a part of the pBBR1MCS-5 backbone using pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 as a template. Both DNA fragments were ligated with XbaI / Bsp1407I-digested pBBR1MCS-5-TgdhM-MCS-T0028.

The plasmid pBBR1MCS-5-tetR-Ptet-mNG lacking all terminators downstream from tetR and mNG was generated by removing TgdhM from pBBR1MCS-5-TgdhM-tetR-Ptet-mNG with primer pair PF5/PF6 as described above for the removal of TgdhM from pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028.

In another construct, for comparing and enhancing resulting reporter protein level when using the Ptet region including its native ribosome binding site (RBSPtet) in G. oxydans, the RBS AGGAGA (RBSAGGAGA), functional and strong in G. oxydans, was inserted upstream from mNG (Fricke et al. 2020; Hentschel et al. 2013). Therefore, the DNA fragments tetR-Ptet-RBS (761 bp) and RBS-mNG (811 bp) were amplified with the primer pairs PF1/PF12 and PF13/PF4, respectively, and ligated with XbaI/EcoRI-digested pBBR1MCS-5-TgdhM-MCS-T0028 to obtain plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-RBS-mNG-TBBa_B1002-T0028.

For plasmid pBBR1MCS-5-TgdhM-Ptet-mNG-TBBa_B1002-T0028 lacking tetR, fragment Ptet-mNG (901 bp) was amplified from pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 with primer pair PF11/PF4 and ligated with XbaI / EcoRI-digested pBBR1MCS-5-TgdhM-MCS-T0028.

To change the plasmid backbone from pBBR1MCS-5 to pBBR1MCS-2 and create pBBR1MCS-2-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028, the DNA fragment with TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 was excised from pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 and ligated with SacI/XhoI-digested pBBR1MCS-2.

The plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 was constructed by ligating the DNA fragment PlacI-lacI-PlacUV5-lacZα-RBS (1855 bp) and mNG-TBBa_B1002-T0028 (954 bp) with BshTI/SphI-digested pBBR1MCS-5. Here, upstream from mNG the RBS AGGAGA known to be functional in G. oxydans was integrated into the construct (Hentschel et al. 2013). The DNA sequence of PlacI-lacI-PlacUV5-lacZα was derived from E. coli BL21 and obtained with primer pair PF14/PF15 and plasmid pK18mobsacB-DE3 (Kortmann et al. 2015). The DNA fragment mNG-TBBa_B1002-T0028 was amplified with primer pair PF16/PF17 from plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028.

For the construction of plasmid pBBR1MCS-5-PlacUV5-RBS-mNG-TBBa_B1002-T0028 lacking the repressor gene lacI, the DNA fragment PlacUV5-RBS-mNG-TBBa_B1002-T0028 (1575 bp) was amplified with primer pair PF18/PF17 from pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 and ligated with BshTI/SphI-digested pBBR1MCS-5.

The plasmid pBBR1MCS-5-lacI-PlacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 with PlacI-lacI in the opposite orientation to PlacUV5-mNG was constructed by amplification of the DNA fragments PlacI-lacI (1,261 bp) with the primer pair PF19/PF20 and PlacUV5-mNG-TBBa_B1002-T0028 (1556 bp) with the primer pair PF21/PF17. Both fragments were ligated into pBBR1MCS-5 digested with BshTI and SphI.

Measurements of mNG fluorescence

Expression of promoter–reporter constructs in G. oxydans was monitored using the fluorescence protein mNeonGreen (mNG) as a reporter (Shaner et al. 2013). In shake flask experiments, the inducibility of Ptet-derived mNG expression in G. oxydans was tested by the addition of 200 ng mL−1 anhydrotetracycline (ATc) from a 0.2 mg mL−1 stock solution in 50% ethanol. Non-induced reference cultures were supplemented with an equal volume of 50% ethanol. For experiments with PlacUV5-controlled mNG expression, 1 mM of isopropyl-β-d-1-thiogalactopyranoside (IPTG) was supplemented from a 100 mM stock solution in water. An equal volume of water was added to non-induced reference cultures. Throughout the cultivation samples were taken to monitor growth (OD600) by a spectrophotometer (UV-1800, Shimadzu) and fluorescence emission using an Infinite M1000 PRO Tecan reader (λex 504 nm / λem 517 nm, gain 60, ex/em bandwidth 5 nm, infinite M1000 PRO Tecan). Using a BioLector system, 800 µL batches of d-mannitol medium were inoculated from overnight starter cultures to an initial OD600 of 0.3 and incubated at 30 °C (1,200 rpm; 85% humidity) using 48-well Flowerplates® (m2p-labs). Backscattering light intensity (A620 nm) for growth and fluorescence emission (λex 510 nm / λem 532 nm) were monitored online during the cultivation. Backscatter and fluorescence signals were measured using gain 15 or 20 and 50 or 60, respectively, as indicated in the figure legends. Fluorescence values that did not exceed the emission signals from cell-free control samples were set to 1. Specific fluorescence was calculated by taking the quotient of the fluorescence signal per biomass value at a given time point. Specific fluorescence values lower than in the cell-free control samples were set to 0.01. All data presented in the same graphs were obtained in the same BioLector growth experiment using identical gains.

Cell flow cytometer analysis

Reporter gene expression was analyzed by measuring mNG fluorescence on the single cell level with G. oxydans 621H either carrying the plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028, pBBR1MCS-2-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028, or pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 using a FACSAria™ Fusion cell sorter (BD Biosciences) run with 70 psi sheath pressure and equipped with a 70 µm nozzle. Data acquisition and analysis of the flow cytometer was controlled by the FACSDiva 8.0.3 software (BD Biosciences). Using a 488-nm solid blue laser beam, the forward scatter (FSC) and side scatter (SSC) were employed for cell analysis. Particles/events with FSC-H and SSC-H signals below a threshold of 200 a.u. and 300 a.u. were excluded from the analysis. Detection of emitted mNG fluorescence from the SSC signal was performed by combining a 502-nm long-pass and 530/30-nm band-pass filter. The entire cell population was analyzed in a three-step gating strategy. Initially, the assessed cell population was gated in a FSC-H vs. SSC-H plot, to exclude signals originating from electronic noise and cell debris. From the resulting population, the FSC-H signal was plotted against the FSC-W signal. Subsequently, the obtained population was gated in a SSC-H vs. SSC-W plot, to ensure singlet discrimination. The gated singlet population was used for fluorescence acquisition in all experiments (fluorescence intensity vs. cell count). For all samples, 100,000 events were recorded with an event rate below 10,000 events/s. FlowJo 10.7.2 for Windows (FlowJo, LLC) was used for data analysis and visualization of all gated events (n = 100,000).

Fluorescence microscopy

For fluorescence microscopy, cells were placed on agarose-coated microscope slides and covered by a coverslip. Images were taken on a Zeiss AxioImager M2 imaging microscope that was equipped with a Plan-Apochromat 100 × /1.40-numerical aperture phase-contrast oil-immersion objective and an AxioCam MRm camera. Fluorescence was measured using the 46 HE filter set (λex 500/20 nm / λem 535/30 nm). For all images, identical exposure times were applied. Digital images were acquired and analyzed with AxioVision Rel. 4.8 software (Zeiss).

Total DNA extraction, library preparation, Illumina sequencing, and data analysis

Total DNA was purified from a culture aliquot using a NucleoSpin Microbial DNA Mini kit (MACHEREY–NAGEL). DNA concentrations were measured using a Qubit 2.0 fluorometer (Thermo Fisher Scientific). Illumina sequencing libraries of tetR-Ptet and lacI-PlacUV5 samples were prepared from 1 µg of isolated DNA using the NEBNext Ultra™ II DNA Library Prep Kit for Illumina according to the manufacturers’ instructions (NEB). The libraries were evaluated by qPCR using the KAPA library quantification kit (Peqlab) and then normalized for sample pooling. Paired-end sequencing with a read length of 2 × 150 bases was performed in-house on an Illumina MiSeq system. The demultiplexed sequencing output (base calls) was obtained as fastq files and used for trimming and quality filtering, mapping, and coverage calculation using the CLC Genomics Workbench software (Qiagen). For the mappings, the improved genome sequence from G. oxydans 621H and the tetR-Ptet or lacI-PlacUV5 plasmid sequence was used (Kranz et al. 2017).

Computational methods

Homology modeling of TetR, LacI, and AraC structures was performed by YASARA Structure version 19.12.14 (Krieger et al. 2002) using the default settings (PSI-BLAST iterations: 6, E value cutoff: 0.5, templates: 5), and with oligomerization state adjusted to 2 for TetR and AraC, and to 4 for LacI (Altschul et al. 1997). A position-specific scoring matrix (PSSM) was used to score the obtained template structures (Jones 1999; Qiu and Elber 2006). The obtained hybrid models were further evaluated for protein geometry by VERIFY3D and ProSA (Eisenberg et al. 1997; Sippl 1993). Operator models of tetO and araI1 were generated with Avogadro as B-shaped DNA (Hanwell et al. 2012). The initial coordinates for lacO were taken from the X-ray structure (PDB ID 1EFA, resolution 2.6 Å) (Bell and Lewis 2000).

Surface residues were determined with PyMol script (default cutoff of 2.5 Å2). The protonation states of titratable residues were assigned on the basis of pKa values obtained from the PROPKA 3.1 program for the pH values 7, 6, 5, and 4 (Olsson et al. 2011).

Modeling protein–protein and protein-DNA complexes was performed by using the HADDOCK Webserver (van Zundert et al. 2016). To understand the DNA binding behavior, protein-DNA docking simulations of the transcription factors and their respective operator sequence were done. In docking simulations, a dimeric model for TetR and LacI and a monomeric model for AraC were used. The charge states of proteins were adjusted to the respective pH based on the previous pKa calculations and were adjusted to the respective pH either directly within the PDB file according to the HADDOCK specifications (Asp, Glu), or within the HADDOCK interface (His). Oligomerization of proteins was computed by performing protein–protein docking with the transcription factor subunits, i.e., monomer–monomer docking for TetR and AraC, and dimer-dimer docking for LacI. Due to the fact that the LacI homology model was missing the second tetramerization helix, the dimer structure taken from the crystal structure (PDB ID 3EDC, resolution 2.10 Å) was used for the LacI tetramerization modeling (Stenberg and Vihinen 2009).

Results

A pBBR1MCS-5-based TetR-Ptet system was very tight and highly inducible and tunable

The tetR-Ptet region including its native RBS from transposon Tn10 was used to construct a pBBR1MCS-5-based reporter plasmid using the reporter gene mNeonGreen (mNG). The tetR gene under the control of its native promoter PtetR overlapping with the divergently oriented TetR-dependent promoter Ptet followed by mNG were integrated into the MCS of pBBR1MCS-5-TgdhM-MCS-T0028 as described in Material and Methods. DNA sequences for transcription terminators were placed adjacent to the tetR-Ptet-mNG insert to create transcriptional barriers between the genetic elements on the insert and on the plasmid backbone.

The leakiness and inducibility of Ptet were tested in G. oxydans 621H harboring the plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 by omitting and adding anhydrotetracycline (ATc). A pre-culture was split and used to inoculate shake flasks for growth in d-mannitol medium without and with 200 ng mL−1 ATc. The highest mNG fluorescence signals were measured in induced cultures after 10 h of growth at the end of the exponential growth phase, followed by a slight decrease in the stationary phase (Fig. 1a). In the non-induced cultures, the mNG signals barely surpassed background signals of cell-free control samples, suggesting very tight repression of Ptet in the absence of inducer. Based on the absolute and the specific mNG fluorescence, the maximal induction ratios were calculated to be 2284 ± 263-fold and 2661 ± 180-fold, respectively. At the end of the cultivation (24 h), cells of an induced culture were harvested and total DNA was purified for Illumina sequencing. In the read data analysis, 99.5% of the reads mapped to the updated reference sequences of the G. oxydans 621H genome, the 5 endogenous plasmids, and the mNG expression plasmid with tetR-Ptet (Kranz et al. 2017). Thus, the sequencing results excluded undesired contaminations and verified that the Ptet-derived mNG expression was highly induced in G. oxydans 621H carrying the plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028. In microscale BioLector cultivations a similar induction profile of Ptet was observed as in shake flasks. The fluorescence in ATc-supplemented cultures peaked approximately after 8 h also followed by a decreased fluorescence level in the stationary phase (Fig. 1b). Again, the basal expression under non-induced conditions was barely detectable and the maximal induction ratio based on the specific fluorescence was calculated to be 3674 ± 193-fold. The drop in mNG fluorescence signals observed in shake flasks and in BioLector cultivations in the stationary phase was caused by the decreased pH of the medium (pH 4.6) and could be largely recovered at pH 6 as outlined below and described previously (Fricke et al. 2020).

Fig. 1
figure 1

Performance of the TetR-Ptet system in G. oxydans 621H. a Growth (OD600) and specific mNG fluorescence in G. oxydans 621H carrying plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 in ATc-induced (200 ng mL−1) and non-induced condition in shake flasks. The mNG fluorescence was measured in a Tecan reader (gain 60). The specific fluorescence was calculated from absolute fluorescence per OD600. Data represent mean values and standard deviation from three biological replicates with three technical replicates each. b Growth according to backscatter (gain 15) and specific mNG fluorescence (with gain 50) of G. oxydans 621H carrying plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 in ATc-induced (200 ng mL−1) and non-induced condition in microscale BioLector cultivations. Data represent mean values and standard deviation from four biological replicates with three technical replicates each. c Graded ATc-dependent mNG expression in G. oxydans 621H carrying plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 in microscale BioLector cultivations. Reporter gene expression measured as fluorescence (gain 50) was induced with increasing concentrations of ATc from 10 to 400 ng mL−1 as indicated. d FACS analysis of G. oxydans 621H carrying plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 or empty vector pBBR1MCS-5 (MCS-5) as a control. Cells were grown in shake flasks with d-mannitol medium without and with 200 ng mL−1 ATc. FACS analysis was performed 7 h after inoculation/induction. Total counts per sample represent 100.000 events

Full size image

To verify that TetR is responsible for the repression of Ptet also in G. oxydans, we constructed the plasmid pBBR1MCS-5-TgdhM-Ptet-mNG-TBBa_B1002-T0028 lacking the tetR gene. In the absence of the plasmid-encoded regulator TetR, in both conditions with ATc supplement and without, the mNG expression from Ptet was even higher than in the induced control cultures containing the tetR gene on the plasmid (Fig. S1). This indicated that in G. oxydans Ptet is also fully repressed by TetR like in E. coli and that non-blocked Ptet is highly active in G. oxydans. Interestingly, even without the tetR gene the addition of ATc resulted in a higher maximal specific fluorescence (28.8 ± 1.3 a.u.) than without ATc (23.5 ± 1.9 a.u). This suggested that in G. oxydans the Ptet activity could additionally be affected by an endogenous factor.

We then tested the tunability of the TetR-Ptet system in G. oxydans by varying the ATc concentration from 10 to 400 ng mL−1. Gradual increase of the inducer concentration led to a gradual increase in mNG fluorescence illustrating the high dynamic range of Ptet-derived gene expression in G. oxydans with the pBBR1MCS-5 backbone (Fig. 1c). Comparing the peaking fluorescence signals, the system appeared to be almost fully induced with 150 to 200 ng mL−1 ATc. According to flow cytometry analysis (FACS) of samples taken 7 h after induction (Fig. 1a), in both ATc-induced and non-supplemented cultures high population homogeneity was observed (Fig. 1d). Ninety-five percent and 94% of the cells of a sample were found to exhibit either very low (~ 400 a.u., non-induced) or very high (~ 80,000 a.u., ATc-induced) mNG fluorescence signals. To visualize this high homogeneity by microscopy, images of induced and non-induced G. oxydans cells harboring the plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 were taken (Fig. S2). In accordance with the data obtained by flow cytometry, the microscopic images confirmed the strong inducibility of pBBR1MCS-5-based TetR-Ptet and the highly homogenous induction response in G. oxydans.

During growth, the G. oxydans d-mannitol medium initially set to pH 6 is acidified to pH 4.7 which causes a loss of intracellular mNG fluorescence suggesting a decreased cytoplasmic pH, at least in the stationary phase (Fricke et al. 2020). We wanted to test experimentally whether an already initially lower medium pH could result in leakiness of Ptet in G. oxydans already during growth as observed in Komagataeibacter (Florea et al. 2016). Therefore, G. oxydans carrying the plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 was grown in d-mannitol medium initially adjusted to pH 6, 5, 4 and 3 both in the absence and presence of ATc. After 23 h cells were centrifuged and resuspended in fresh d-mannitol-free medium set to pH 6 to check for potential pH-dependent recovery of mNG fluorescence above the respective levels monitored before that would indicate a leakiness of Ptet together with a loss of mNG fluorescence during growth. In all non-induced cultures, very low maximal fluorescence signals (0.6 ± 0.1 to 1.3 ± 0.2 a.u.) were measured independent of the initial medium pH (Fig. S3a-b). Cells grown in pH 3 exhibited the lowest signals and a different growth according to backscatter compared to all other cultures. No sufficient increase of mNG fluorescence was observed in any non-induced condition in the exponential growth phase to suggest a leakiness of Ptet in G. oxydans. Moreover, after transfer (23 h) of the cells into fresh d-mannitol-free medium adjusted to pH 6 no recovery of mNG fluorescence above the levels before could be observed. Thus, since the strong induction of Ptet by ATc in the cells grown in each medium of this pH series indicated that the mNG protein could always be produced during the exponential phase of growth and exhibited strong fluorescence property (Fig. S3c-d), the results together indicated that expression from Ptet was really tightly repressed in G. oxydans in all pH conditions, even in medium initially set to pH 3.

Terminators strongly affected mNG expression strength but not repression of Ptet

Since we did not observe and could not show leakiness of the TetR-Ptet system in growing G. oxydans cells per se or in dependence of the medium pH, we wanted to analyze the influence of terminator sequences on the functionality and leakiness of the TetR-Ptet system. Therefore, we constructed plasmids lacking terminators downstream from tetR or mNG and both and compared these constructs in regard to Ptet-derived mNG expression in G. oxydans (Fig. 2). The terminators downstream from mNG had a major influence on the resulting mNG fluorescence level. With plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG not having TBBa_B1002 and T0028 downstream from mNG, the mNG fluorescence signals were reduced by half in ATc-induced G. oxydans cells. Compared to the reference plasmid including all terminators, the maximum in specific fluorescence dropped from 19.4 ± 2.9 a.u. to 10.5 ± 0.2 a.u. (Fig. 2e). The specific fluorescence signals of G. oxydans harboring the plasmid pBBR1MCS-5-tetR-Ptet-mNG-TBBa_B1002-T0028 lacking TgdhM downstream from tetR were significantly (p = 0.0383) higher (22.3 ± 1.1 a.u.) compared to the reference (19.4 ± 2.9 a.u.) suggesting a little positive effect on Ptet-derived induction and expression strength when TgdhM downstream from tetR was absent. The construct pBBR1MCS-5-tetR-Ptet-mNG lacking all terminators performed similar (9.5 ± 0.7 a.u.) as pBBR1MCS-5-TgdhM-tetR-Ptet-mNG lacking only the terminators downstream from mNG (10.5 ± 0.2 a.u.). Hence, in G. oxydans and in case of mNG, termination of the target gene transcription close to the 3’ end might be important to achieve higher expression levels in G. oxydans. None of the terminators appeared to affect the repression of Ptet in the absence of ATc and a Ptet leakiness was not observed.

Fig. 2
figure 2

Variants of pBBR1MCS-5-based expression plasmids with tetR-Ptet and reporter gene expression in G. oxydans 621H. a Map of plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 carrying the fluorescence reporter gene mNeonGreen (mNG) expressed from Ptet with the adjacent tetR gene and terminators TgdhM, TBBa_B1002 and T0028. b Variants of the plasmid insert with reporter gene mNG to test TetR-Ptet-dependent expression in the presence and absence of terminators downstream from tetR and mNG, with RBS AGGAGA (Hentschel et al. 2013) inserted in the 3’ region of Ptet upstream from mNG, and without tetR. c DNA sequence details with Ptet region and TetR binding sites (tetO1 and tetO2) and terminator sequences adjacent to tetR and mNG. d + e) Growth according to backscatter and specific mNG fluorescence in G. oxydans carrying either plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 (1), or the plasmid lacking TgdhM (2) or TBBa_B1002-T0028 (3) or all terminators (4) under non-induced (d) and ATc-induced (e) condition in microscale BioLector cultivations. For induction 200 ng mL−1 ATc was present in the d-mannitol medium. Data represent mean values and standard deviation from two biological replicates with three technical replicates each. TgdhM: terminator sequence of gdhM (GOX0265); T0028: terminator sequence of GOX0028. BioLector settings: backscatter gain 20, fluorescence gain 50

Full size image

Comparison of the pBBR1MCS-2 / -5 backbones and insertion of AGGAGA as RBS

The broad-host-range vectors pBBR1MCS-5 and pBBR1MCS-2 differ in the antibiotics resistance they convey for clone selection. pBBR1MCS-5 gmR encodes for a gentamicin-3-acetyltransferase enabling growth on gentamicin and pBBR1MCS-2 neoR/kanR encodes for a neomycin/kanamycin phosphotransferase enabling growth on kanamycin (Kovach et al. 1995). Furthermore, gmR and neoR/kanR are located on different DNA strands which in the case of neoR/kanR could result in transcripts partially overlapping with and antisense to the mRNA of the target gene present in the cloned insert. To check whether the plasmid backbone and antibiotics resistance affect the inducibility of Ptet in G. oxydans we compared the performance of the TetR-Ptet system with pBBR1MCS-5 and pBBR1MCS-2.

In G. oxydans the pBBR1MCS-2-based TetR-Ptet system performed much worse compared to the pBBR1MCS-5-based system despite the similarities of the two plasmids. With pBBR1MCS-2, the specific mNG fluorescence with induced Ptet was almost halved and reached 19.2 ± 9.7 a.u. ~ 10 h after induction at the end of the exponential phase of growth, while with pBBR1MCS-5 the maximal fluorescence peaked at 36.7 ± 2.2 a.u. after 8 h (Fig. 3a). Nevertheless, the maximal induction ratios were high with both pBBR1MCS-2 and pBBR1MCS-5 and were calculated to be 1915 ± 757 and 3674 ± 193, respectively. The basal expression under non-induced conditions was not affected by the plasmid backbone. Notably, a much higher standard deviation was obtained with pBBR1MCS-2. FACS analysis with gating of cells in regard to their volume and complexity by forward and side scattering (FSC and SSC) of light revealed that a major part of the G. oxydans cells with the pBBR1MCS-2 derivative and kanamycin exhibited a very different non-typical elongated cell morphology that was not observed with the pBBR1MCS-5 derivative and gentamicin (Fig. 3b). Approximately only 42% of the population with the pBBR1MCS-2 derivative passed the FACS gate where almost 100% of G. oxydans cells without plasmid or with the pBBR1MCS-5 derivative passed. This high heterogeneity of G. oxydans carrying the pBBR1MCS-2 derivative was not affected by the inducer ATc. The occurrence of a portion of elongated cells with pBBR1MCS-2 backbone and kanamycin was also demonstrated by fluorescence microscopy (Fig. S4). Several G. oxydans cells carrying pBBR1MCS-2 or pBBR1MCS-2-tetR-Ptet-mNG-TBBa_B1002-T0028 appeared to be 10–15 µm and > 40 µm long, while the 621H reference cells typically were 2–3 µm. Apparently, the altered cell morphology had no direct effect on the inducibility of the TetR-Ptet system in G. oxydans as elongated cells also exhibited strong mNG fluorescence.

Fig. 3
figure 3

Comparison of pBBR1MCS-5 and pBBR1MCS-2 with tetR-Ptet. a Growth according to backscatter (gain 15) and specific mNeonGreen (mNG) fluorescence (gain 50) in G. oxydans 621H carrying either plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 or pBBR1MCS-2-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 in microscale BioLector cultivations. For induction, always 200 ng mL−1 ATc was present in the d-mannitol medium. Data represent mean values and standard deviation from at least three biological replicates with three technical replicates each. b Cell morphology according to FACS analysis of G. oxydans 621H type strain (wt) without plasmid and with either the pBBR1MCS-5- or pBBR1MCS-2-based tetR-Ptet system, all without the inducer ATc

Full size image

To test improved translation and thus resulting reporter protein levels in G. oxydans, we also constructed a plasmid with the RBS AGGAGA that we had also used in the araC-ParaBAD system and inserted it upstream from mNG and downstream from tetO2 overlapping with the native RBS of the Ptet region (Fig. 2c). Thus, in this construct, RBSPtet and RBSAGGAGA were consecutively upstream from mNG. The impact on reporter expression and inducibility was measured in BioLector cultivations. Grown in d-mannitol medium and induced with 200 ng mL−1 ATc, the fluorescence signals of induced cells were further increased and reached > 4000-fold induction, yet at such a high level that the detector gain in the BioLector had to be set from 50 down to 40 to avoid signal saturation. For both constructs with gain 40 setting, the insertion of RBSAGGAGA doubled up the specific fluorescence in induced cells from 8.6 ± 0.4 a.u. for the reference to 20.4 ± 0.5 a.u. (Fig. 4). The basal expression in the absence of ATc was again extremely low, and the calculated maximal induction ratio based on the specific fluorescence with gain 40 was increased from 855 ± 7 for the reference to 2042 ± 1 for the plasmid with RBSAGGAGA.

Fig. 4
figure 4

Influence of the ribosome binding site AGGAGA on mNG reporter expression with the TetR-Ptet system. Growth according to backscatter and specific mNeonGreen (mNG) fluorescence in G. oxydans 621H carrying plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-RBS-mNG-TBBa_B1002-T0028 with the RBS change (a) and with plasmid pBBR1MCS-5-TgdhM-tetR-Ptet-mNG-TBBa_B1002-T0028 as the control (b) in microscale BioLector cultivations. Cells were grown in d-mannitol medium without or with 200 ng mL−1 ATc for induction. Data represent mean values and standard deviation from two biological replicates with three technical replicates each. BioLector settings: backscatter gain 20, fluorescence gain 40

Full size image

pBBR1MCS-5-based LacI-PlacUV5 was leaking and therefore tunable only up to 40-fold

For the lacI-Plac system, the insert containing lacI under control of its native promoter PlacI, promoter PlacUV5 and lacZα was derived from E. coli BL21(DE3). The promotor PlacUV5 differs from the native E. coli Plac promoter by three point mutations leading to enhanced activity and reduced cyclic adenosine monophosphate (cAMP)-dependency for induction in E. coli (Hirschel et al. 1980). Downstream from the PlacUV5-controlled reporter gene mNG, the terminator TBBa_B1002 was placed and RBSAGGAGA upstream from mNG (Fig. 5a, b). The lacI-Plac system includes three different operator sites which typically are bound by LacI in the absence of an inducer molecule to repress Plac-derived expression. Of these operator sites, O1 is located in close proximity to the -10 region of Plac, while O3 and O2 overlap with lacI and lacZα, respectively. In E. coli all three operator sites are bound by tetrameric LacI and are required for maximal repression (Oehler et al. 1990). Therefore, in order to enable full repression by LacI, the α-unit region of lacZ with O2 was included in the insert.

Fig. 5
figure 5

pBBR1MCS-5-based expression plasmid with lacI-PlacUV5 and reporter gene expression in G. oxydans 621H. a Map of the pBBR1MCS-5-based lacI-PlacUV5 plasmid. b Sequence information details of the lacI-PlacUV5 plasmid. The promoter PlacUV5 comprises the LacI operator sites O3, O1, and O2. The O3 sites overlaps with the 3’ end of lacI. The RBS AGGAGA was inserted upstream from mNG. The iGEM terminator sequence of BBa_B1002 was located downstream from mNG followed by terminator T0028 from GOX0028. c Shake flasks cultivations of G. oxydans 621H carrying plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 in IPTG-induced (1 mM) and non-induced condition. The mNG fluorescence was measured in a Tecan reader (gain 60. The specific fluorescence was calculated from absolute fluorescence per OD600. d Microscale BioLector cultivations of G. oxydans 621H carrying plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 in IPTG-induced (1 mM) and non-induced condition. All data represent mean values and standard deviation from three biological replicates with three technical replicates each. Backscatter gain 15, fluorescence gain 60. e Microscale BioLector cultivations of G. oxydans 621H carrying plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 with increasing concentrations of IPTG as indicated. Fluorescence gain 60. f FACS analysis of G. oxydans 621H carrying plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028. Cells were grown in shake flasks with d-mannitol medium and 1 mM IPTG. FACS analysis was performed 7 h after inoculation and induction. Total counts per sample represent 100.000 events. As a control G. oxydans 621H carrying the empty vector pBBR1MCS-5 was used (MCS-5)

Full size image

We examined the inducibility of PlacUV5 by IPTG in G. oxydans 621H harboring the plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028. Cells were grown in shake flasks using d-mannitol medium supplemented with and without 1 mM IPTG. After 10 h at the end of the exponential growth phase, the highest fluorescence signal was measured. Due to the basal expression in the absence of IPTG, the maximal induction ratios by the absolute and by the specific fluorescence were calculated to be 18 ± 4 and 18 ± 5, respectively (Fig. 5c). When testing the inducibility of PlacUV5 in microscale BioLector cultivations, a similar trend was observed. Under induced conditions, the fluorescence peaked approximately 6 h after induction and subsequently slightly decreased in the stationary phase while in non-induced cultures a basal fluorescence was measured (Fig. 5d). Here, with 1 mM IPTG an average induction ratio of 41 ± 16 was calculated based on the specific fluorescence of the replicates in the BioLector. The PlacUV5-derived mNG expression exhibited a tunability by IPTG concentrations ranging from 0.1 to 10 mM in G. oxydans (Fig. 5e). The mNG expression from PlacUV5 was gradually increased up to 3 mM IPTG. Above 3 mM IPTG, the mNG expression hardly increased further and thus appeared to be saturated and best-calculated induction ratios were again in the range of 40-fold. Illumina sequencing data obtained with total DNA purified from induced cells showed that 99.5% of the reads mapped to the updated reference sequences of the G. oxydans 621H genome, the 5 endogenous plasmids, and the mNG expression plasmid with lacI (Kranz et al. 2017). Thus, the sequencing results verified that the PlacUV5-derived mNG expression was induced in G. oxydans carrying the intended plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028.

To analyze the induction of PlacUV5 in G. oxydans on the single cell level with regard to the homogeneity of mNG expression, flow cytometer analysis was applied (Fig. 5f). Within the chosen gate, 7 h after induction 95% of all induced cells exhibited high fluorescence (~ 11,000 a.u.) while under non-induced conditions in 96% of all cells, the fluorescence was at the somewhat leaky basal level (~ 1500 a.u.). This indicated that in the exponential growth phase, expression from induced PlacUV5 was very homogenous in G. oxydans. This was also supported by fluorescence microscopy (Fig. S5).

Similarly, as for TetR and Ptet, we tested experimentally if initially lower medium pH values could affect the leakiness of PlacUV5. Therefore, G. oxydans carrying the plasmid pBBR1MCS-5-PlacI-lacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 was grown in d-mannitol medium initially adjusted to pH 6, 5, 4, and 3 in the absence and presence of IPTG. After 23 h, cells were centrifuged and resuspended in fresh d-mannitol-free medium of pH 6 to check for potential pH-dependent recovery of mNG fluorescence above the respective levels monitored before that would indicate a higher leakiness of PlacUV5 together with a previous loss of mNG fluorescence already during growth (Fig. S3e-h). In all non-induced cultures, overall basal fluorescence signals up to 10 a.u. were measured with decreasing maxima in the exponential phase when the initial medium pH decreased. Again, cells grown in pH 3 exhibited the lowest signals and a different growth according to backscatter compared to all other cultures. Compared to the pH 6 condition, no sufficient increase of mNG fluorescence was observed in any non-induced pH condition (6 h) in the exponential growth phase to suggest a higher leakiness of PlacUV5. Additionally, after the transfer (23 h) of the non-induced cells into fresh d-mannitol-free medium adjusted to pH 6 no recovery of mNG fluorescence above the levels with pH 6 could be observed in any condition. Interestingly, only for the non-induced pH 4 and pH 5 conditions, recovery of mNG fluorescence in pH 6 (24 h) above the respective maximal levels observed before in the exponential phase (6 h) were observed. This suggested that in the pH 4 and pH 5 conditions G. oxydans was not able to maintain its typical intracellular pH during growth and mNG fluorescence partially became inactive already in an early phase of growth (6 h).

Strength of PlacUV5 was PlacI-lacI-dependent and leaking was independent of read-through

Unlike most operons, in the E. coli lac operon, the regulatory gene lacI is located immediately upstream from the target operon and is transcribed in the same direction. The repressor LacI forms tetramers binding in the absence of an inducer to up to three operator sites, thereby preventing transcription from the lac promoter directly downstream from lacI. Nevertheless, despite LacI binding and the presence of a lacI terminator region coinciding with the LacI operator site O1 directly downstream from the − 10 region, read-through is possible varying from 10 up to 80% (Horowitz and Platt 1982; Oehler et al. 1990). To check the possibility of read-through from PlacI causing the leakiness observed in G. oxydans and interference by endogenous proteins on the PlacUV5 inducibility, we wanted to check and compare the pBBR1MCS-5-based reporter expression in the absence of PlacI-lacI, as well as with PlacI-lacI in opposite orientation toward PlacUV5.

Firstly, we constructed the plasmid pBBR1MCS-5-PlacUV5-RBS-mNG-TBBa_B1002-T0028 lacking PlacI and most of the lacI gene, while keeping a short 3’ region of lacI to maintain all three operator sites of PlacUV5 including O3 partly overlapping with the end of the lacI gene. The analysis of the IPTG-dependent inducibility of PlacUV5 in G. oxydans in the absence of PlacI-lacI revealed no difference in fluorescence signals supplemented with IPTG or not (Fig. S6). This indicated that the inducibility observed in G. oxydans solely depended on derepression by LacI. Interestingly, compared to the original system with PlacI-lacI, the maximal reporter expression from PlacUV5 was strongly reduced from 8.4 ± 0.4 to 2.3 ± 0.2 according to the specific fluorescence (6–7 h) when PlacI-lacI was lacking. This suggested that a significant read-through from PlacI appeared to contribute to the expression of the gene downstream from PlacUV5, thus increasing the apparent PlacUV5 strength in G. oxydans almost fourfold when induced.

Secondly, we constructed plasmid pBBR1MCS-5-lacI-PlacI-PlacUV5-RBS-mNG-TBBa_B1002-T0028 with PlacI-lacI in opposite direction to exclude potential read-through toward PlacUV5-mNG. Despite the opposite directions of PlacI and PlacUV5, an even somewhat higher leakiness of PlacUV5 was observed when not induced (Fig. S7). Therefore, with 1 mM IPTG the maximal induction fold-change (6.4 h) calculated was only 15.4 ± 0.9 based on the specific fluorescence, while the original PlacI-lacI-PlacUV5 construct reached a maximal induction fold-change (6.7 h) of 39.7 ± 13.9. Surprisingly, the apparent induced PlacUV5 strength was restored according to the reporter fluorescence and reached the same levels as when PlacI-lacI were in the same direction as PlacUV5. Together, the results indicated that in G. oxydans the promoter strength of plasmid-based PlacUV5 appeared to depend on sequences further upstream and that the PlacUV5 leakiness was independent of potential read-through from PlacI.

Modeling LacI and TetR binding to DNA at acidic pH values

Our results suggested that the leakiness of PlacUV5 in G. oxydans was independent of potential read-through from PlacI and thus should result from transcription initiation from PlacUV5 and insufficient repression by LacI. Since the cytoplasmic pH of G. oxydans may readily be acidified to some extent already during growth when medium pH decreases, we modeled LacI- and TetR-DNA structures at different pH values using YASARA to predict if and how a potentially lower cytoplasmic hydrogen ion activity might affect the binding of LacI and TetR to DNA (Krieger et al. 2002). As a control, the AraC structure was also modeled since an l-arabinose-inducible AraC-ParaBAD system was shown to be fully functional in G. oxydans with plasmid pBBR1MSC-5, yet was very leaky in Gluconacetobacter and Komagataeibacter using another expression plasmid (Fricke et al. 2020; Teh et al. 2019). Protein-DNA bindings were simulated in two different steps for pH 7, 6, 5, and 4 (see supplementary Text S1 and S2).

Analysis of the HADDOCK simulations in different pH conditions showed that the HADDOCK scores of all five best-scoring clusters obtained in the TetR-DNA docking simulation for pH 7, 6, 5, and 4 lie within the standard deviation of each other. To make a prediction of the DNA binding behavior in dependence of the pH value, similar docking positions were identified for each pH value and the cluster scores were compared. It can be observed that with decreasing pH values the scores decrease. The smaller the score, i.e., the more negative, the better is the predicted binding to DNA, suggesting an increased binding of TetR to DNA (Fig. 6). Similar results were obtained for AraC. Thus, when comparing the scores of clusters with corresponding docking positions, the values generally decreased with decreasing pH values, suggesting a generally increased binding to DNA for both TetR and AraC (Fig. 6). On the contrary, for LacI unique clusters with correlating docking poses identified for pH 7, 6, 5, and 4 exhibited increasing HADDOCK scores for decreasing pH values, suggesting a decreased DNA binding of LacI when pH decreases (Fig. 6 and Supplementary Text S1 and S2).

Fig. 6
figure 6

Prediction of DNA-binding behavior of LacI, TetR, and AraC at different pH values computed by the HADDOCK web server. Each data point corresponds to a cluster of docking poses. The lower (more negative) the HADDOCK score the energetically more favorable the docking pose. Trend lines plotted are based on clusters containing similar docking poses. For details see supplementary Text S1 and Text S2

Full size image

Discussion

In this study, we found with a pBBR1MCS-5-based TetR-Ptet system that inducible target gene expression based solely on de-repression of the heterologous target promoter can also perform extremely well in G. oxydans. With the pBBR1MCS-5-based plasmid constructed here, the anhydrotetracycline (ATc)-inducible promoter Ptet derived from the E. coli transposon Tn10 exhibited excellently tunable expression performance in an inducer concentration-dependent manner with maximal induction ratios up to more than 3500-fold. This was due to extremely low basal reporter expression in the absence of an inducer, thus well repressed Ptet, and Ptet being very strong in G. oxydans.

In contrast, expression from PlacUV5 was always leaky when not induced and this appeared to be independent of potential read-through from PlacI which was located with lacI immediately upstream from PlacUV5 and, unlike most regulator genes and target operons, coupled very close and transcribed in the same direction. Therefore, the transcription termination by the native lacI terminator is expected to be very effective in G. oxydans and the PlacUV5 leakiness is rather related to the early events after transcription initiation from PlacUV5. In E. coli, the LacI repressor is not preventing per se the binding of RNA polymerase (RNAP), but is blocking the progress of RNAP which has bound to the DNA and has begun transcribing (Reznikoff et al. 1969). Transcription through roadblocks has been shown to depend on the cooperation of the leading and trailing RNAP molecules and thus depends on how many RNAP molecules are allowed to initiate from the same promoter upstream from the roadblock (Epshtein et al. 2003; Hao et al. 2014). In the case of Plac or PlacUV5, the distance from the transcriptional start (TSS) to the LacI binding site O2 is rather short and site O1 is even overlapping with the TSS; thus, only one RNAP molecule is expected to initiate upstream from the roadblock. Additionally, the E. coli σ70 subunit of RNAP mediates a pause at the lac promoter closely downstream from the transcriptional start which is proposed to function in limiting the downstream gene expression (Nickels et al. 2004). If the G. oxydans σ70 subunit of RNAP also mediates this pause at the lac promoter is unknown. Nevertheless, RNAP cooperation should not be possible and thus not contribute to the leakiness of PlacUV5 in G. oxydans, yet the extent of read-through can also be affected by accessory factors. For example, in E. coli GreA and GreB rescue backtracked (roadblocked) RNAP by cleavage of the RNA, regenerating a new 3’ end at the catalytic site, and can aid passage of RNAP through a LacI roadblock in cells (Toulme et al. 2000). Furthermore, in E. coli the transcription-coupled repair protein Mfd binds to the DNA behind RNAP and uses ATP to push backtracked RNAP forward until the 3’ end of RNA is back at the catalytic center, yet the forces generated by Mfd may also result in RNAP termination (Park et al. 2002). E. coli GreA (158 aa, b3181) and G. oxydans GreA (157 aa, GOX0324) share 54% sequence identity, E. coli GreB (158 aa, b3406) and G. oxydans GreB (168 aa, GOX1860) share 47%, and E. coli Mfd (1148 aa, b1114) and G. oxydans Mfd (1173 aa, GOX0055) share 39%. It is unknown how these factors support roadblocked RNAP and read-through behind roadblocks in G. oxydans and AAB in general, yet the leakiness of Plac or PlacUV5 is likely rather related to weaker binding of LacI to operator DNA and thus weaker promoter repression in G. oxydans.

Our docking simulations predicted that below pH 6 the binding of LacI to DNA is decreased which could cause or contribute to promoter leakiness. Neutralophilic bacteria maintain an intracellular pH between 7 and 8, while in acidophilic bacteria intracellular pH is considered to be maintained between 6 and 7 (Baker-Austin and Dopson 2007; Krulwich et al. 2011). The mannitol LB medium typically used for G. oxydans is initially set to pH 6 and can be acidified during growth in dependence of the carbon and energy sources, for example, to approximately pH 4.5 or 3.3 as well (Fricke et al. 2020). In such acidic pH conditions of the growth medium strongly decreased, intracellular fluorescence reporter activities of mNeonGreen (mNG) were observed in G. oxydans in the stationary phase (Fricke et al. 2020). Studies with isolated mNG revealed a strong decrease in fluorescence at pH 4.4 and below (Steiert et al. 2018). Therefore, the decrease in mNG fluorescence in G. oxydans cells was attributed to an acidified intracellular pH, at least in the stationary phase. The loss in mNG fluorescence in stationary G. oxydans cells can quickly be restored without protein synthesis just by short incubation in pH 6 condition as shown in this study and before (Fricke et al. 2020). Moreover, for the AAB Acetobacter aceti, the intracellular pH was reported to change from 5.8 to 3.9 already during growth when the medium pH acidified from 6.2 to 3.5 (Menzel and Gottschalk 1985). Despite intracellular pH maintenance mechanisms, it is conceivable that depending on the growth conditions in some or many AAB the cytoplasmic pH can decrease already during growth below the range of pH 6 to 7 considered to be typically maintained in acidophilic bacteria. According to our data, the mNG reporter fluorescence in G. oxydans cells with LacI-PlacUV5 grown in non-induced pH 4 and pH 5 conditions suggested that G. oxydans was not able to maintain its typical intracellular pH during growth and the mNG fluorescence became partially inactive already in the middle phase of growth (6 h). To get LacI-dependent systems tight and suitable for use in G. oxydans or AAB in general, the LacI protein may require engineering for increased binding to DNA in acidic pH condition. However, more intracellular pH data of AAB strains in various growth conditions are required. Besides leakiness, in G. oxydans, the strength of fully induced PlacUV5 was much lower compared to that of the fully induced Ptet (compare absolute fluorescence Fig. S3d and S3h). It was already reported before that Plac is rather weak in G. oxydans 621H and the PlacUV5 mutant promoter exhibiting three base pair changes including two changes at positions − 9 and − 8 apparently also exhibits a similar low expression strength in G. oxydans (Hirschel et al. 1980; Kallnik et al. 2010). Thus, in G. oxydans the TetR-Ptet system will probably always be the choice for applications because of its extremely low basal expression and very high dynamic range of the tunable expression strength from Ptet.

Interestingly, in K. rhaeticus iGEM, the TetR-Ptet system tested exhibited high leakiness in the absence of the inducer (Florea et al. 2016). In G. oxydans, we could neither observe nor demonstrate relevant leakiness of TetR-Ptet, not with the pBBR1MCS-5-based construct variants (± terminators), not with the pBBR1MCS-2 backbone and kanamycin causing abnormal cell morphology, and not in the media of the pH series. Without terminators downstream from mNG, the length of the mNG transcript is assumed to exhibit an extended 3’ UTR. This appeared to only decrease transcript stability or translation, since the mNG signals were reduced to half, yet repression of Ptet was not affected. For the tetR transcript, it is unknown how stability and translation are affected in G. oxydans when the 3’ UTR is longer or shorter, yet altered repression of Ptet was also not observed with either of the constructs. A major difference of the two tetR-Ptet constructs tested in K. rhaeticus iGEM and in G. oxydans is that we maintained the opposite orientation of tetR-PtetR and Ptet as originally present on Tn10, while in K. rhaeticus iGEM the construct tested contained the promoter J23118 or PlacI for expression of tetR followed by a terminator sequence and all directly upstream from the target promoter Ptet and in the same direction as Ptet (Florea et al. 2016). It was assumed that the terminator is effective, yet it seemed to be unknown to what extend the terminator allowed potential read-through in these constructs. This could possibly cause the high leakiness of Ptet in K. rhaeticus iGEM, especially when cooperative RNAP molecules read through the TetR roadblock. On the other hand, our computational simulation also predicted that with decreasing pH value TetR dimerization decreases (Text S1 and Text S2). That could finally affect binding to DNA and thus promoter repression since TetR binds as a dimer (Hillen and Berens 1994). Thus, if the prediction holds true, in G. oxydans TetR dimerization is not sufficiently negatively affected by mild acidic intracellular pH. Komagataeibacter is typically grown in Hestrin − Schramm (HS) medium with glucose or in LGI medium with sucrose, both initially set to pH 6 or pH 4.5 and acidified further during growth (Florea et al. 2016; Hestrin and Schramm 1954). The intracellular pH of A. aceti was reported to change from 5.8 to 3.9 already during growth when the medium pH acidified from 6.2 to 3.5 (Menzel and Gottschalk 1985). The extent of being able to maintain the intracellular pH not only in the stationary phase but also during exponential growth could differ between AAB species. This could possibly be seen in the results with the LuxR-Plux system in Komagataeibacter (Florea et al. 2016). The system exhibited condition-dependent induction only up to fivefold due to high leakiness, yet it also exhibited extremely low leakiness and much better induction performance according to fluorescence microscopic analysis in cells inside cellulose pellicles. Cellulose pellicles generate a micro-environment with the advantage of potentially protecting the cells from harsh conditions which could also include protection from acidic environmental pH. Again, to better understand the pH-dependent binding of regulators to DNA and to improve the expression performance of heterologous regulatable expressions systems so far leaky in AAB and their growth conditions, more intracellular pH data of AAB strains are required for various growth conditions. Additionally, experimental pH-dependent protein–protein and protein-DNA interaction data are needed.

In summary, we could show that in G. oxydans the TetR-Ptet system tested was extremely well performing with the pBBR1MCS-5 backbone and exhibited very strong reporter expression when induced, while the LacI-PlacUV5 system was always leaking and thus resulting in much lower induction ratios. Therefore, the pBBR1MCS-5-based TetR-Ptet system with the low inducer concentrations required for gradually tuning target gene expression appeared to be highly suitable for applications in G. oxydans and possibly in other AAB.