To biochemically characterize EnuR from R. pomeroyi DSS-3 further, we expressed a full-length recombinant EnuR-Step-Tag-II protein heterologously in E. coli and purified it to apparent homogeneity via affinity chromatography on a Streptactin column. We also separately expressed and purified the C-terminal ATD of the wild-type protein as a Strep-Tag-II fusion protein. In addition, we carried out similar types of production and affinity purification experiments with a variant of the R. pomeroyi DSS-3 EnuR protein in which the Lys residue to which the PLP cofactor is presumably covalently attached is replaced by a His residue (EnuR^∗; Lys302His). As a result of this amino acid substitution, PLP cannot be covalently bound, leading to the loss of the characteristic PLP-dependent yellow color of the full-length wild-type EnuR protein solutions (Schulz et al., 2017b; Figures 2A,B). Incorporation of the PLP cofactor also occurred during the heterologous production of the ATD from the wild-type protein but not into the ATD derived from the EnuR^∗ protein (Figure 2A).

The GabR protein from B. subtilis is a homodimer where the monomers are arranged in a head to tail configuration (Edayathumangalam et al., 2013) and where the isolated ATD can form homodimers as well (Okuda et al., 2015a). In a similar vein, we found that the purified EnuR protein and its separately produced ATD also assemble into homo-dimers in solution. These proteins eluted from a size exclusion column as corresponding to 109 kDa and 83 kDa molecular mass species, respectively (Figures 2C,D). The calculated molecular mass of the R. pomeroyi DSS-3 monomeric EnuR protein is 51 kDa, while that of its ATD is 40 kDa. The EnuR^∗ protein and its isolated ATD^∗ behaved in these chromatography experiments identical to that of the corresponding wild-type proteins (Figures 2C,D). Taken together, these biochemical experiments show that (i) dimer-formation of EnuR depends on its ATD, regardless whether the ATD carries a covalently attached PLP or not, and (ii) that the PLP cofactor can be attached to the ATD even when the N-terminal DNA-reading head of EnuR is missing.

In order to probe whether the PLP co-purifying with EnuR is covalently attached to Lys302 or just coordinated by this residue, we subjected the EnuR and EnuR-ATD proteins to mass spectrometric (MS) analysis. Due to the lability of the aldimine group that would be formed between the ε-amino group of the lysine side chain and the PLP cofactor, the purified proteins were reduced with NaBH₄ prior to tryptic digestion. In the subsequent MS analysis, we retrieved peptides spanning amino acid residues 294-302 of both the EnuR and EnuR-ATD proteins that exhibited a mass-to-charge (m/z) ratio of 1,318.63305 (theoretical mass-to-charge ratio of M + H⁺ of 1,087.60339), corresponding to a mass shift of 231.02966 Da, a value in excellent agreement with the presence of a covalently attached PLP (theoretical difference of 231.029662 Da). Further MS/MS fragmentation of these peptides corroborates that Lys302 is the site of attachment for PLP (Figures 3A,B). This consolidates the previous suggestion that EnuR binds PLP, similar to type I aminotransferases and MocR/GabR-type regulators (Tramonti et al., 2018), through the formation of an aldimine between an evolutionary conserved lysine residue (Lys302 in EnuR) and the cofactor (Schulz et al., 2017a).

Micro-scale thermophoresis is a sensitive method that traces the movement of fluorescently labelled proteins in a temperature gradient in response to a ligand (Wienken et al., 2010). We used this method to determine the dissociation constant (K_d) (Figure 4) for the binding of the known ectoine-derived inducer α-ADABA (Schulz et al., 2017a) and that of the presumed 5-hydroxyectoine-derived inducer hydroxy-α-ADABA (Figure 1B; Mais et al., 2020). Both compounds were obtained via chemical synthesis through alkaline-mediated hydrolysis of the tetrahydropyrimidine ring of either ectoine (Kunte et al., 1993) or 5-hydroxyectoine. They were purified by repeated chromatography on silica columns to apparent homogeneity as assessed by NMR-spectroscopy (Mais et al., 2020).

Ectoine nutrient utilization regulator bound α-ADABA with a K_d of 2.55 ± 0.5 μM (Figure 4A), a value that fits well with a previous measured K_d-value of 1.7 ± 0.3 μM for this ligand (Schulz et al., 2017a). The full length EnuR protein exhibited a K_d-value of 0.43 ± 0.2 μM for hydroxy-α-ADABA (Figure 4B). We also assessed the binding of α-ADABA to the PLP-bound ATD of the wild-type EnuR protein in a micro-scale thermophoresis ligand-binding experiment. We found that this domain bound this ectoine metabolite with approximately the same K_d-value (1.98 ± 0.45 μM) (Figure 4C) as the full length EnuR protein (K_d of 2.55 ± 0.5 μM).

Since we wanted to further understand the molecular determinants for inducer binding by EnuR, we generated a structural model of its monomer. The EnuR model was fashioned on the crystal structure of the B. subtilis GabR protein (Edayathumangalam et al., 2013), the only GabR/MocR-type regulatory protein whose crystal structure is known. Our EnuR model (Figure 5A) was generated using Phyre² (Kelley et al., 2015) set in the extensive mode¹. An overlay of the GabR protein and of the EnuR model revealed a root mean square deviation (RMSD) of just 0.86 Å (over 390 amino acids), indicating that the overall fold of these two MocR/GabR-type regulatory proteins is most likely closely related.

In addition to the modelling of the putative EnuR structure, we also assessed the phylogenomic conservation of this regulatory protein. For this analysis, we relied on a recently reported manually curated dataset assessing the presence of ectoine/5-hydroxyectoine catabolic gene clusters in 8 850 microbial genome sequences (Mais et al., 2020). The identification of putative ectoine/5-hydroxyectoine catabolic gene clusters was based upon a direct juxtaposition of the eutD and eutE genes. 363 microbial genome sequences contained eutD/eutE pairs. We found that 76% (278 out of 363) of the corresponding ectoine/5-hydroxyectoine catabolic gene clusters contained a juxtapositioned enuR gene (Figure 5B). We then used the information gleaned from this bioinformatic approach (Supplementary Figure 1) to assess the degree of amino acid conservation at each position in the EnuR protein chain (Supplementary Figure 2). Subsequently, we projected these data onto our EnuR model by using the ConSurf algorithm (Berezin et al., 2004; Figure 5A).

In an alignment of the 278 retrieved EnuR-type proteins, we found that the degree of amino acid sequence identity ranged between 82% (for the EnuR protein from Leisingeria sp. NJS201) and 40% (for the EnuR protein from Salipiger pacificus YSBP01) when the R. pomeroyi DSS-3 EnuR protein was used as the search query. As expected, those amino acid residues forming the N-terminally positioned winged-helix-turn-helix DNA reading head are particularly well conserved, as are central segments of the ATD (Figure 5A and Supplementary Figure 2). Notably, the Lys residue in the ATD to which the PLP molecule is covalently attached in EnuR (Lys302 in the R. pomeroyi DSS3 EnuR protein) (Schulz et al., 2017b), is completely conserved among the 278 inspected EnuR-type proteins (Supplementary Figure 2).

Among those 363 genome sequences that contain eutD/eutE pairs, six major microbial orders are represented (Figure 5B), all of which belong to the proteobacteria (Supplementary Table 1). Using computational tools provided via the IMG/M web-server (Chen et al., 2021), a phylogenetic tree was calculated for the 363 EutD-type proteins using aminotransferases from Escherichia coli (Wilce et al., 1998) and Homo sapiens (Bradshaw et al., 1998) as outgroups. The EutD-derived tree was visualized using the iTOL software suit (Letunic and Bork, 2019), and the presence of 278 enuR-type genes was then projected onto this tree (Figure 5B; Supplementary Figure 1). enuR-type genes are dominantly represented among Rhizobiales, Rhodobacterales, Burkholderiales, Pseudomonales, Oceanospirillales, and Vibrionales predicted to use ectoines as nutrients (Figure 5B).

The crystal structure of the dimeric full-length B. subtilis GabR protein contained in the cofactor and inducer binding site of one of its monomers a PLP molecule covalently bound to a Lys residue (Figure 6A). In the second monomer, a free PLP molecule was found that was chemically ligated to γ-ethynyl-GABA, a substrate-mimic of GABA (Figure 6B), thereby revealing the structure of the external aldimine (Edayathumangalam et al., 2013). In addition, a crystal structure of the isolated dimeric ATD of GabR captured the GABA-mediated structural transition catalyzed by the conversion of the internal aldimine to the external aldimine (Park et al., 2017). Collectively, these crystal structures guided our docking experiments of the high affinity inducers α-ADABA and hydroxy-α-ADABA into the presumed ligand-binding site of EnuR. The docking experiments were carried out using AutoDock Vina (Trott and Olson, 2010).

In our EnuR model, hydroxy-α-ADABA is bound in close proximity to the PLP molecule with which it interacts via its free amino group. Nine interactions of the inducer molecule with the EnuR protein are observed (Figure 6C). The oxygen of PLP interacts with the hydroxyl-oxygen of hydroxy-α-ADABA as well as with the primary amino group of this molecule. This nitrogen atom is also bound by the O-1 atom of Asn244. Additionally, hydroxy-α-ADABA is stabilized by interactions with the N-2 atom of Asn244 and through interactions with Thr245, Phe417, Ser431, Ser104 (Figure 6C). Collectively, these nine interactions are the foundation for the high affinity of EnuR for hydroxy-α-ADABA (Figure 4B); they thereby establish the orientation of this 5-hydroxyectoine-derived metabolite in the inducer binding site of EnuR. A more detailed description of the energetics of these interactions are summarized in Supplementary Table 2.

Compared to hydroxy-α-ADABA, α-ADABA appears to be a slightly more linear molecule due to the lack of the hydroxy group. Consequently, it is positioned in the predicted inducer binding site of EnuR in a slightly different orientation (Figure 6D). Like hydroxy-α-ADABA, the primary amino group of α-ADABA interacts with the oxygen atom of PLP. A second interaction of the primary amino group of α-ADABA is found with the side chain of Ser104, a configuration different from that predicted for hydroxy-α-ADABA (Figures 6C,D). As observed for hydroxy-α-ADABA, the carboxyl group of α-ADABA interacts with the nitrogen atom present in the side chain of Asn244. The secondary nitrogen atom of α-ADABA interacts with the oxygen atom of Ser431. In total five interactions of EnuR with α-ADABA are predicted (Figure 6D and Supplementary Table 2), suggesting that this compound will be bound with a somewhat lower affinity by EnuR in comparison with hydroxy-α-ADABA. This is precisely what we observed in our in vitro ligand binding experiments where the affinity of EnuR for hydroxy-α-ADABA was about five times higher than that for α-ADABA (K_d of about 0.43 μM versus 2.55 μM for hydroxy-α-ADABA and α-ADABA, respectively) (Figures 4A,B).

The amino acid sequence alignment of 278 EnuR-type proteins revealed a high degree of conservation of the amino acids predicted to be involved in hydroxy-α-ADABA and α-ADABA binding by our docking studies (Figures 6C,D). Especially Ser104, Asn244, Lys302, Phe417, and Arg429 are either strictly or highly conserved (Supplementary Table 3). Slight deviations can be observed for the position of Ser431; however, this amino acid is mainly exchanged to a Cys residue (135/278). In our model of the EnuR ligand binding cavity, the sulfur atom of Ser431 interacts directly with both inducer molecules and it is thus a reasonable assumption that the sulfur atom of the Cys side chain will adopt the same interaction. The positions Ala412 and Thr245, more peripheral residues in the ligand binding site (Figures 6C,D), seem to be less important for the binding of the hydroxy-α-ADABA and α-ADABA molecules. Ala412 is frequently substituted by a Leu residue (123/278), while a great variety of residues can assume the position of Thr245 in EnuR-type proteins (Supplementary Table 3). A visualization of the EnuR binding site for the PLP cofactor and the inducer α-ADABA is rendered in Figure 6F.

The ectoine metabolite diaminobutyric acid (DABA) (Figure 1B) also serves as an inducer for EnuR (Schulz et al., 2017a; Yu et al., 2017). A K_d-value of about 460 μM has been reported for the EnuR protein from R. pomeroyi DSS-3 (Schulz et al., 2017a). Hence, there is a substantial difference in affinity between DABA on one hand and α-ADABA and hydroxy-α-ADABA on the other hand for EnuR (Figures 4A,B). As a consequence of the low binding affinity of DABA for EnuR, we observed multiple positions for the DABA molecule within the presumed inducer binding site of EnuR in the first round of docking experiments. Optimization and refinement of these positions was difficult and only two interactions of DABA were observed that hinted at a possible binding state (Supplementary Table 2). However, in contrast to α-ADABA and hydroxy-α-ADABA (Figures 6C,D), this would position DABA too far away from the PLP molecule (Figure 6E; Tramonti et al., 2018) in order to serve its function as an inducer for EnuR. While the distances between the primary nitrogen group of α-ADABA and hydroxy-α-ADABA to the PLP cofactor in our EnuR model are 2.6 Å and 3 Å, respectively, the distance of the corresponding nitrogen group of DABA is about 7 Å. Consequently, the actual position of the low-affinity EnuR ligand DABA cannot be reliably predicted by our docking experiments.

The isomer of α-ADABA, γ-ADABA, is the substrate for the ectoine synthase EctC (Czech et al., 2019), the key enzyme for the production of ectoine (Peters et al., 1990; Ono et al., 1999; Hermann et al., 2020). We wondered if interactions of γ-ADABA with EnuR could be found in docking experiments but no stable binding was observed. This was mainly due to steric clashes with amino acid residues whose side chains protrude into the inducer binding site of EnuR. This result is fully consistent with previous experiments in which no binding of γ-ADABA to purified EnuR from R. pomeroyi DSS-3 could be measured (Schulz et al., 2017a). The failure to dock the non-inducer γ-ADABA into the EnuR binding site can be considered as an internal control for the successful docking experiments with its isomer α-ADABA, a high-affinity ligand for EnuR (Figure 4A).

Since α-ADABA, hydroxy-α-ADABA, DABA, and possibly also hydroxy-DABA can interact with EnuR and serve as inducers, we wondered if these compounds can be found in cells of R. pomeroyi DSS-3 actively catabolizing ectoines. We therefore performed targeted metabolic analysis of ectoine- and 5-hydroxyectoine-derived metabolites in cells that were grown with either ectoine or 5-hydroxyectoine as sole carbon, energy and nitrogen sources. The metabolic profile of these cultures was compared with that of cells using glucose as carbon and energy source and NH₄Cl as nitrogen source in a chemically fully defined minimal medium. The analyzed samples contained substantial amounts of either ectoine or 5-hydroxyectoine, but these values represent in all likelihood not only intracellular pools of these compounds but probably also reflect incomplete removal during the harvesting and washing of the cells (Figure 7A).

Substantial amounts of α-ADABA were found in the extracts of cells grown in the presence of ectoine, while hydroxy-α-ADABA was found in cells grown in the presence of 5-hydroxyectoine (Figure 7B). Interestingly, DABA was found under both cultivation conditions, regardless whether the cells were grown in the presence of ectoine or of 5-hydroxyectoine. The pool of hydroxy-DABA in cells that received 5-hydroxyectoine as their sole carbon and nitrogen sources was very low and, as expected, not detectable in cells that were exposed to ectoine (Figure 7B). A rather surprising finding was the detection of substantial amounts of γ-ADABA in cells that were grown in the presence of ectoine, while γ-ADABA was present only in very low amounts in cells grown in the presence of 5-hydroxyectoine (Figure 7B).

Using comparative genomics and metabolic reconstruction, Suvorova and Rodionov (2016) have previously analyzed putative EnuR (EutR) operator binding sites in 69 microbial genomes. This analysis suggested a consensus operator sequence for EnuR (EutR)-type proteins that consists of an inverted repeat of five base pairs separated by six base pairs [ATTGTnnnnnnACAAT] (Suvorova and Rodionov, 2016). However, depending on the microbial species under study, variations on this theme exist (Schulz et al., 2017a; Yu et al., 2017).

In R. pomeroyi DSS-3, two closely spaced potential EnuR binding sites in the intergenic region between the 3′-end of enuR and the beginning of the ectoine/5-hydroxyectoine catabolic operon (Figure 1A) can be observed. They overlap with core elements (−10 and −35 sequences separated by 17 bp) of the predicted sigma-70 type promoter (Feklistov et al., 2014; Figure 8A) for the ectoine/5-hydroxyectoine catabolic gene cluster. DNA band-shift assays have previously shown that EnuR can specifically bind in vitro to a 278 bp DNA fragment carrying both predicted operators (Schulz et al., 2017a). The putative binding site one is a perfect repeat with eight bp in each of the half-sites, while the second putative binding site has an overall length of 18 bp and only the outermost four bp are a perfect inverted repeat (Figure 8A). It is currently unknown if both of the in silico predicted EnuR binding sites (Suvorova and Rodionov, 2016) are biological relevant to control the ectoine/5-hydroxyectoine importer and catabolic gene cluster from R. pomeroyi DSS-3.

Since no quantitative data for the interaction of the R. pomeroyi DSS-3 EnuR protein with its presumed operator(s) have been reported, we carried out DNA-binding assays with the purified full-length and PLP-containing EnuR protein with a labeled 48 bp DNA fragment containing both presumed EnuR binding sites using microscale thermophoresis. We measured a K_d-value of 1.9 μM (Figure 8B). This is a higher yet physiologically relevant K_d-value than the K_d-value (9.14 nM) reported for the promoter/operator interaction of EnuR with its two established operators for the S. meliloti ectoine/5-hydroxyectoine catabolic gene cluster (Yu et al., 2017).

The antibiotics gentamycin, rifampicin, and kanamycin were obtained from Serva (Heidelberg, Germany); ampicillin was purchased from Carl Roth GmbH (Karlsruhe, Germany). Anhydrotetracycline hydrochloride, desthiobiotin, and Strep-Tactin Superflow chromatography material were obtained from IBA GmbH (Göttingen, Germany). Marker proteins for size exclusion chromatography experiments were purchased from Sigma-Aldrich (Taufkirchen, Germany). Restriction endonucleases and DNA ligase were obtained from ThermoScientific (St. Leon-Rot, Germany) and used as suggested by the manufacturer. Ectoine was a kind gift from the bitop AG (Witten, Germany) and 5-hydroxyectoine was purchased from Merck (Darmstadt, Germany). γ-ADABA was purchased from abcr GmbH (Karlsruhe, Germany).

Ruegeria pomeroyi strains (Supplementary Table 4) were maintained on half-strength YTSS agar. For all growth experiments, the strains were cultivated in defined basal minimal medium (Baumann et al., 1971). Both media were prepared as described previously (Schulz et al., 2017b). When applicable, the antibiotics gentamycin and rifampicin, were added to liquid and solid media at a concentration of 20 μg ml⁻¹. When the use of ectoine or 5-hydroxyectoine as combined carbon and nitrogen sources by R. pomeroyi strains was tested on basal minimal medium agar plates, the plates were supplemented with 28 mM ectoine or 28 mM 5-hydroxyectoine as sole carbon, energy and nitrogen sources. Agar plates with streaked R. pomeroyi strains were typically incubated at 30°C for five days.

The IBA-Stargate plasmids containing either the enuR (pBAS3), the enuR^∗ (pBAS17), enuR wild-type C-terminal aminotransferase (ATD) domain (pLH17), or the enuR^∗ C-terminal aminotransferase domain (pLH17) genes, were routinely maintained in the E. coli K-12 DH5α (Invitrogen, Karlsruhe, Germany) on LB agar plates containing ampicillin (100 μg mL^–1). Minimal Medium A (MMA) (Miller, 1972) containing 0.5% (w/v) glucose as the carbon source, 0.5% (w/v) casamino acids (0.5%), 1 mM MgSO4, and 3 mM thiamine was used for cultivation of the E. coli B strain BL21 carrying plasmids pBAS3 (enuR⁺), pBAS17 (enuR^∗), pLH17 (enuR-ATD) or pLH26 (enuR^∗-ATD) (Supplementary Table 5) for the overproduction of the EnuR protein and its mutant derivatives (Schulz et al., 2017a,b).

The cyclic ectoine and 5-hydroxyectoine molecules were linearized through alkaline hydrolysis as previously described (Kunte et al., 1993; Mais et al., 2020). Subsequently, the α- and γ-isomers of N-acetyl-L-2,4-diaminobutyric acid (α-ADABA and γ-ADABA) and hydroxy-α-ADABA were separated from other hydrolysis products of ectoine and 5-hydroxyectoine by repeated chromatography on a silica gel column (Merck silica gel 60) (Mais et al., 2020). The purity of the isolated α-ADABA and hydroxy-α-ADABA samples was at least 90% as determined by HPLC analysis. The identity and purity of these compounds was assessed by NMR spectroscopy as described by Mais et al. (2020), although we cannot exclude that minor impurities resulting from the alkaline hydrolysis of ectoines are still present in the preparations that we used for our experiments (Mais et al., 2020).

To identify metabolites derived from ectoines in R. pomeroyi DSS-3 cells growing in the presence if either ectoine or 5-hydroxyectoine, we carried out targeted metabolic analysis. In one set of experiments, we grew the cells in basal minimal medium with glucose (28 mM) as a carbon source and NH₄Cl (56 mM) as a nitrogen source in the absence of ectoines (control culture). In the second set of experiments, we grew the cells in an basal medium in the absence of glucose and NH₄Cl and provided either ectoine (56 mM) or 5-hydroxyectoine (56 mM) as sole and combined source of carbon, energy and nitrogen to the cells. In both sets of experiments, the cells were grown at 30°C in orbital shaker (20 ml culture volume in a 100 ml Erlenmeyer Flask) until the R. pomeroyi DSS-3 cultures reached an OD₅₇₈ of about 1. The cells were pelleted by centrifugation, resuspended in basal medium and were then re-centrifuged. For the extraction of ectoine/5-hydroxyectoine-derived metabolites, one ml of 20% ethanol was added to the cells and they were vigorously shaken at room temperature for 30 min; cellular debris was then removed by centrifugation in table top Eppendorf centrifuge (13 000 rpm for 30 min at 4°C). The supernatant was evaporated at 50°C for at least 24 hours and the formed dry residue was re-suspended in 500 μl of double distilled water. After another centrifugation step, the supernatant was analyzed, and intracellular concentrations were estimated by assuming a volume of 0.5 μl of the cytoplasm of 1 ml R. pomeroyi DSS-3 cells at an OD₅₇₈ of 1.

Separation and quantification of ectoine, 5-hydroxyectoine and its metabolites DABA, α-ADABA, γ-ADABA, hydroxy-DABA, hydroxy-α-ADABA and hydroxy-γ-ADABA in ethanolic cell extracts were conducted on a HPLC-ESI-MS system (Agilent 1,100 system with MSD1946D) using 100 mM NH₄HCO₃ in 90% H₂0/10% acetonitrile as eluent. The separation column was a 250 × 2 mm i.d. Metrohm Carb 2 strong anion exchanger operated at 0.2 ml/min. The analytes were detected in selected ion modus as their positively charged H⁺ adducts. Possible interference of aspartic acid on hydroxy-DABA were checked and discarded. Calibration was performed using commercially available γ-ADABA samples. For all measurements, four independently grown R. pomeroyi DSS-3 cultures were used and form each of them two ethanolic extracts were prepared.

The R. pomeroyi strain DSS-3 (Moran et al., 2004) was obtained from the German Collection of Microorganisms (DSMZ; Braunschweig, Germany), and a rifampicin-resistant [Rif^R] derivative of this isolate (strain J470) (Todd et al., 2012) was kindly provided by J. Todd and A. Johnston (University of East Anglia, United Kingdom). E. coli K-12 DH5α carrying the helper plasmid pRK2013 [Kan^R] (Figurski and Helinski, 1979) for conjugation experiments between E. coli and R. pomeroyi were also provided by these colleagues. The construction of the R. pomeroyi eutD mutant (ASR8) and the complete operon deletion strain ASR6 [Δ(enuR-atf:Gm^R)] were described previously (Schulz et al., 2017a,b), as were plasmids pBAS3 (enuR⁺; wild-type) and pBAS17 (enuR^∗; Lys302His) (Schulz et al., 2017a). These plasmids are derivatives of the expression vector pASG-IBA3 (IBA GmbH, Göttingen, Germany) and express enuR genes under the control of the anhydrotetracycline hydrochloride (AHT) responsive TetR controlled tet promoter carried by pASG-IBA3 and its recombinant derivatives. Both the wild-type EnuR protein and its Lys302His mutant (EnuR^∗) carry a Strep-TAG-II peptide fused to their C-termini to allow affinity purification of the recombinant EnuR and EnuR^∗ proteins from cell extracts of the E. coli B strain BL21 (DE3) (Schulz et al., 2017a,b).

To construct a deletion of the R. pomeroyi chromosomal eutE gene, 500 bp fragments located upstream and downstream of the respective genomic region (Moran et al., 2004) were amplified by PCR using custom synthesized primers (Supplementary Table 6). A DNA fragment encompassing a gentamycin resistance cassette (Gm^R) was amplified from plasmid p34S_Gm (Dennis and Zylstra, 1998). Using the Gibson assembly procedure (Gibson et al., 2009), the three DNA fragments were cloned into the linearized (with EcoRI) and dephosphorylated suicide vector pK18mobsacB (Kvitko and Collmer, 2011), which confers resistance to kanamycin. The resulting plasmid was pLH72 and carries the Δ(eutE:Gm^R)1 deletion mutation (Supplementary Table 5).

Plasmids for the overproduction of the aminotransferase domains of EnuR and its Lys302His mutant derivative EnuR^∗ were constructed via the IBA-Stargate cloning procedure as described by the manufacturer (IBA GmbH, Göttingen, Germany). Custom designed primers (Supplementary Table 6) (Microsynth AG, Balgach, Switzerland) were used to amplify the 1,110 bp aminotransferase domains (ATD) for the enuR and enuR^∗ genes from the respective plasmids pBAS3 (enuR⁺) and pBAS17 (enuR^∗), and were then inserted into the expression plasmid pASG-IBA3 so that recombinant proteins with a Strep-TAG-II affinity peptide at their carboxy-termini were produced. The resulting plasmids were pLH17 (enuR-CTD) and pLH26 (enuR^∗-CTD), respectively (Supplementary Table 5).

Chromosomal DNA of R. pomeroyi strain DSS-3 was isolated as described (Marmur, 1961). The High Pure Plasmid Isolation Kit (Roche, Mannheim, Germany) was used to isolate plasmid DNA from E. coli strains. Chemically competent E. coli cells were prepared and transformed with plasmid DNA as reported (Sambrook et al., 1989). All recombinant DNA methods were carried out via routine procedures (Sambrook et al., 1989).

Plasmid pLH73 [Δ(eutE:Gm^R)1] (Supplementary Table 5) was conjugated by tri-parental mating into R. pomeroyi by mixing the E. coli strain PRK2015 (pRK2013 [Kan^R]) (Figurski and Helinski, 1979), DH5α (pLH73) [Kan^R and Gm^R] and the Rif^R R. pomeroyi recipient strain J470. R. pomeroyi J470 trans-conjugants that had received plasmid pLH73 were selected on 1/2 YTSS agar plates containing the antibiotics rifampicin and gentamycin and 10% saccharose as described (Schulz et al., 2017b). The resulting colonies were tested for their antibiotic resistance profile and Kan^S Gm^R strains were then evaluated for the presence of the chromosomal Δ(eutE:Gm) deletion/insertion mutation via PCR using chromosomal DNA as the template and DNA primers listed in Supplementary Table 6 that hybridize to genomic regions flanking the eutE gene. The resulting R. pomeroyi J470-derived strain was named LHR7 [Δ(eutE:Gm^R)1] (Supplementary Table 4).

For overproduction of the EnuR-Strep-tag-II and EnuR^∗-Strep-tag-II recombinant proteins, cells of the E. coli B strain BL21 (DE3) were transformed with the appropriate overproduction plasmids pBAS3 (enuR⁺) or pBAS17 (enuR^∗) (Supplementary Table 5). These plasmids allow the expression of the enuR⁺ and enuR^∗ genes under the control of the tet promoter, a system that is controlled by the anhydrotetracycline (AHT) responsive TetR repressor whose structural gene is present on the expression plasmids (Schulz et al., 2017b). The same type of overproduction system was used to produce either the ATD from the wild-type EnuR protein (plasmid pLH17), or of the ATD from the mutant EnuR^∗ protein (plasmid pLH26) (Supplementary Table 5). The plasmid-containing E. coli cells were grown at 37°C in MMA containing 0.5% casamino acids until the cultures reached an OD₅₇₈ of about 0.5. tet-promoter/TetR-mediated overexpression of the various plasmid-encoded genes was triggered by adding the inducer AHT (final concentration: 0.2 mg l^–1) to the cultures. The growth temperature of the cultures was then reduced to room temperature (about 25°C) and the cultures were subsequently incubated for additional two hours to allow overproduction of the recombinant proteins. Cells were harvested by centrifugation, resuspended in purification buffer (100 mM Tris-HCl (pH 7.5), 150 mM NaCl), lysed by passing them three to five-times through a French Pressure Cell (Aminco, Urbana, Il, United States) at 900 psi, and a cleared cell extract was obtained by centrifugation at 35,000 × g for 1 h at 4°C. The recombinant proteins marked with a Strep-TAG-II peptide were purified from the cleared cell extracts via affinity chromatography on a Strep-Tactin Superflow column as described (Schulz et al., 2017b). Strep-Tactin purified proteins were analyzed and further purified via Size-Exclusion-Chromatography (SEC) on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare Europe, Freiburg, Germany), using either a buffer containing 10 mM Tris-HCl (pH 7.5) and 150 mM NaCl when the proteins were subsequently used in ligand-binding assays. The purity of all isolated proteins was assessed by sodium-dodecylsulfate (SDS) polyacrylamide gel electrophoresis (12% acrylamide). Proteins were stained and visualized with InstantBlue (Expedion, Cambridgeshire, United Kingdom).

Ectoine Nutrient Utilization Regulator and EnuR-ATD proteins were purified as described above. The buffer used for these preparations was 20 mM HEPES-Na pH 7.5, 115 mM NaCl, 1.2 mM CaCl₂, 1.2 mM MgCl₂, 2.4 mM K₂HPO₄. 25 μl (10 μM) of affinity purified full-length EnuR, or EnuR-ATD were treated with 10 mM NaBH₄ (1 μl of 250 mM stock prepared freshly in 0.1 M NaOH) and incubated at room temperature for 30 min to reduce the aldimine. The NaBH₄ reduction was quenched by acidification of the solution to pH of 5-6 with HCl and neutralized to approximately pH 7 with NaOH (Hoegl et al., 2018). These samples were immediately supplemented with 6 μl of SDS loading dye (300 mM Tris-Cl pH 6.8, 10% (w/v) SDS, 25% (v/v) β-mercaptoethanol, 25% (v/v) glycerol, 0.05% (w/v) bromo phenol blue) followed by mixing and heat treatment at 95°C for 5 min. Samples were loaded and separated on 15% polyacrylamide SDS-PAGE gels at 200 V. Gels were stained with Coomassie brilliant blue R250 [0.36% (w/v) Coomassie R250 dissolved in 46% (v/v) ethanol supplemented with 9% (v/v) glacial acetic acid] and destained with 30% (v/v) ethanol supplemented with 10% (v/v) glacial acetic acid. After destaining the protein bands corresponding to full-length EnuR, or EnuR-ATD were excised out of the gel and digested in gel by the addition of Sequencing Grade Modified Trypsin (Serva) at 37°C for 45 min, after which the supernatant was removed and further incubated at 37°C overnight. Peptides were desalted and concentrated using Chromabond C18WP spin columns (Macherey-Nagel). Finally, Peptides were dissolved in 5% (v/v) acetonitrile supplemented with 0.1% (v/v) formic acid.

Mass spectrometric analysis of the tryptic digests was performed using a timsTOF Pro mass spectrometer (Bruker Daltonic). A nanoElute HPLC system (Bruker Daltonics), equipped with an Aurora column (25 cm × 75 μm) C18 RP column filled with 1.7 μm beads (IonOpticks), was connected online to the mass spectrometer. Sample loading was performed at a constant pressure of 800 bar, and 2 μl of a 1:3 dilution of the tryptic digests in double-distilled water injected directly on the separation column. Separation was conducted at 50°C column temperature with the following gradient of water + 0.1% (v/v) formic acid (solvent A) and acetonitrile + 0.1%(v/v) formic acid (solvent B) at a flow rate of 400 nl/min: A linear increase from 2% solvent B to 17% solvent B within 60 min was followed by a linear gradient to 25% solvent B within 30 min and a linear increase to 37% solvent B in additional 10 min. Finally, solvent B was increased to 95% within 10 min and held for additional 10 min. The built-in “DDA PASEF-standard_1.1sec_cycletime” method developed by Bruker Daltonics was used for mass spectrometric measurement. Data analysis was performed using Proteome Discoverer 2.4 (ThermoScientific) with SEQUEST search engine and Byonic version 3.7.4 (Protein Metrics) using the amino acid sequences of full-length EnuR, trypsin, and keratin, as database.

Ligand binding assays with the purified EnuR and EnuR-ATD proteins were carried out by microscale thermophoresis (MST) (Wienken et al., 2010). All experiments were performed on a Monolith NT.115 (NanoTemper Technologies GmbH, Munich, Germany) at 21°C (red LED power was set to 80% and infrared laser power to 70%). The buffer of the purified EnuR and EnuR-ATD [in 10mM Tris- HCl (pH 7.5), 150 mM NaCl] was first exchanged with the labeling buffer of the Monolith NTTM Protein Labeling Kit RED (NanoTemper) to avoid interference of the labeling reactions with free amines in the buffer solution. Subsequent to the labeling of either EnuR, or EnuR-ATD (20 μM each) with the NT 647 dye (according to the supplier’s reaction scheme), the proteins were re-buffered into a solution buffer containing 10 mM Tris-HCl (pH 7.5), 150 mM NaCl and 0.07% Tween20. EnuR (200 nM) was titrated with α-ADABA and hydroxy-α-ADABA (starting from a ligand concentration of 1 mM). Likewise, the EnuR-ATD protein was also titrated with α-ADABA (starting from a ligand concentration of 1 mM). To determine the DNA-binding properties of EnuR, the protein was treated in the same manner and titrated with buffer containing a DNA-fragment (48 bp) harboring the presumed EnuR operator site(s) and the promoter region of the R. pomeroyi DSS-3 ectoine/5-hydroxyectoine importer and catabolic gene cluster (Supplementary Table 6 and Figure 8A). At least six independent MST experiments per ligand of the EnuR protein were recorded at 680 nm and analyzed using NanoTemper Analysis 1.2.009 and Origin8G software suits.

To analyze the phylogenomic distribution of enuR-type genes, a recently compiled and manually curated dataset of 363 bacterial ectoine/5-hydroxyectoine catabolic gene clusters was used as a starting point (Mais et al., 2020). In this dataset, only microorganisms harboring eutD/eutE-genes in direct genetic neighborhood were included, as both proteins are needed to degrade ectoines (Mais et al., 2020). Accordingly, this analysis does not include ectoine/5-hydroxyectoine degradation gene clusters in which the eutD and eutE catabolic genes are not juxtapositioned [e.g., from M. alcaliphilum (Reshetnikov et al., 2020)]. The 363 EutD-protein sequences represented in the dataset reported by Mais et al. (2020) was retrieved from the IMG/M database (Chen et al., 2021) and represented in a tree-format visualized using the iTOL software (Letunic and Bork, 2019). Onto this EutD-protein based tree, we projected the presence of enuR-type genes (278 representatives) that were positioned in the immediate vicinity of ectoine/5-hydroxyectoine degradation gene clusters. Alignments of EnuR-type proteins that were obtained through IMG/JGI Web resources, were visualized with Jalview (Waterhouse et al., 2009).

A model of the presumed EnuR structure was created using the crystal structure of the B. subtilis GabR protein as the template (Edayathumangalam et al., 2013) and by employing the Phyre² software (Kelley et al., 2015) set in the extensive mode (see Text Footnote 1). An overlay of the GabR protein and the EnuR model revealed a root mean square deviation (RMSD) of 0.86 Å (over 390 amino acids). In silico modelling and docking experiments for EnuR and its various ligands were carried out using Chimera (Pettersen et al., 2004) and AutoDock Vina (Trott and Olson, 2010). The definition files for the ligands α-ADABA, hydroxy-α-ADABA, γ-ADABA and DABA were created using the Schrödinger Meastro package (Release, 2017). Initial docking was performed using a wide grid setting and by allowing the positioning of the ligand all around the EnuR protein. The best solution was further optimized by multiple cycles of AutoDock Vina (Trott and Olson, 2010). Final assessment was performed by manual inspection of the interactions of each ligand within the predicted EnuR ligand binding site.