The strains listed in Table 1 were used in this study. For routine molecular biology studies, growth was on LB-agar plates or in LB-broth at 37°C (Miller, 1972). Anaerobic growths were performed at 37°C as standing liquid cultures and cells were usually grown in M9 minimal medium (47.6 mM Na₂HPO₄ × 2 H₂O, 22 mM KH₂PO₄, 8.4 mM NaCl, 20 mM NH₄Cl, 2 mM MgSO₄, 0.1 mM CaCl₂, 0.1 mM thiamin dichloride, 0.2% w/v casamino acids) containing 0.8% (w/v) glucose, or 0.4% (v/v) glycerol plus 15 mM fumarate, or 0.8% (w/v) glucose plus 1% (w/v) nitrate, where indicated, as described (Sambrook et al., 1989). When growth in rich medium was performed, buffered TGYEP (1% w/v tryptone, 0.5% w/v yeast extract, 0.8% w/v glucose, 100 mM potassium phosphate, pH 6.5) was used (Begg et al., 1977). The growth medium was supplemented with trace element solution SLA (Hormann and Andreesen, 1989). When required, the antibiotic kanamycin or chloramphenicol was added to a final concentration of 50 or 25 μg ml^-1, respectively. Cells were harvested anaerobically by centrifugation at 5,000 g for 15 min at 4°C when cultures had reached an OD_{600 nm} of between 0.8 and 1.2. Cell pellets were used immediately or stored at -20°C until use.

The hupX gene (cbdbA131) was amplified as a 1212 bp DNA fragment from chromosomal DNA isolated from D. mccartyi strain CBDB1 using Pfu DNA polymerase and the oligonucleotides hupX_fw (5′-GGGGCATATGCCTAATGGAATGCTGATTG-3′) and hupX_re (5′-GGGGCTCGAGCTAGTGCTTGCCAGCCTTG-3′) and cloned in plasmid pACYC-Duet-I. Plasmid phupSL was constructed by using pSHH18 (referred to as phupXSL throughout this study) as template in a PCR mutagenesis employing the Q5^® Site-Directed Mutagenesis Kit (New England Biolabs, NEB). Care was taken when deleting the hupX gene to ensure that the ribosome binding site for the downstream hupS gene remained intact by using the oligonucleotides hupSLhoxM_fw (5′-ATGGAGTAGGλAATGTTTAATAC-3′) and hupSLhoxM_re (5′-TCCTGTTGCCCCCCTTGT-3′) and by following the instructions given in the Q5^® Site-Directed Mutagenesis Kit.

E. coli strains were constructed using P1kc-mediated phage transduction (Miller, 1972) to introduce the respective defined deletion mutation from the appropriate donor strain obtained from the Keio collection (Baba et al., 2006) to generate the series of FTD147 mutants lacking the structural genes encoding the three formate dehydrogenases of E. coli. When multiple gene knock-outs were constructed, the plasmid pCP20 was used to remove the kanamycin antibiotic resistance cassette as described (Cherepanov and Wackernagel, 1995).

Unless otherwise stated, all experiments were performed in an anaerobic Coy^TM chamber under an atmosphere of 95% nitrogen/5% hydrogen. For standard hydrogenase enzyme activity determination, E. coli cell paste was re-suspended at a ratio of 1 g cell wet weight to 3 ml 50 mM MOPS buffer, pH 7. Cells were disrupted by sonication (30 W power for 5 min with 0.5 s pulses). Unbroken cells and cell debris were removed by centrifugation for 30 min at 50,000 g and 4°C. The resulting crude extract, unless otherwise stated, was used for all studies reported herein.

In order to perform sub-cellular fractionation, periplasmic, soluble and membrane fractions were isolated as described (Sawers et al., 1985).

Determination of protein concentration was done as described (Lowry et al., 1951).

Unless otherwise specified, non-denaturing polyacrylamide gel electrophoresis (PAGE) was performed anaerobically. Separating gels included 0.1% (w/v) Triton X-100 as described (Ballantine and Boxer, 1985). The crude extracts, or sub-cellular fractions, were incubated with a final concentration of 4% (w/v) Triton X-100 prior to application (usually 50 μg of protein) to the gel, which included 6% (w/v) polyacrylamide. Hydrogenase activity-staining was done in 50 mM MOPS buffer pH 7.0, as described (Sawers et al., 1985; Pinske et al., 2012), and included 0.5 mM BV and 1 mM 2,3,5-triphenyltetrazolium chloride (TTC). Gels were incubated under an atmosphere of 100% hydrogen gas.

Measurement of hydrogenase enzyme activity using BV as electron acceptor was performed as described (Ballantine and Boxer, 1985; Pinske et al., 2011). Briefly, anaerobically prepared cuvettes (1.6 ml) were filled with 0.8 ml of H₂-saturated, anaerobic 50 mM MOPS buffer, pH 7.0, including 4 mM BV and placed under a H₂ atmosphere. After baseline determination, the assay was initiated by adding enzyme sample (approximately 150 μg of protein). All assays were performed at 25°C. The wavelength used was 600 nm and an ε_M value of 7400 M^-1 cm^-1 was assumed for reduced BV. One million unit of enzyme activity corresponded to the reduction of 1 nmol of substrate min^-1. Enzyme assays were performed in triplicate using three biological replicates.

Polypeptides in crude extracts were separated by 12.5% (w/v) sodium dodecyl sulfate (SDS)-PAGE (Laemmli, 1970) and gels were either stained with Coomassie Brilliant Blue R or transferred to nitrocellulose membranes for western blotting, which was performed as described (Towbin et al., 1979). The antibodies used were either anti-Strep-tag (IBA Life Sciences), anti-Hyd-2 (Sargent et al., 1998), anti-HupL or anti-HupX peptide antibodies (Hartwig et al., 2017).

The hupXSLhoxM operon has been shown to be functional in anaerobically grown E. coli (Hartwig et al., 2015b). In order to determine whether all three structural components (HupSL and HupX) are essential for the manifestation of the H₂:BV oxidoreductase activity observed in that study, we constructed two additional plasmid derivatives, one of which carried only the hupX gene, while the other included hupSLhoxM but lacked hupX (Figure 1A). These plasmids, along with pSHH18 (hupXSLhoxM⁺; Hartwig et al., 2015b; here referred to as phupXSL in the aid of clarity), were introduced into FTD147, which lacks the genes encoding the catalytic subunits of Hyd-1, Hyd-2, and Hyd-3 (Redwood et al., 2008). After fermentative growth, crude extracts were separated in native-PAGE and stained for hydrogenase enzyme activity (Figure 1B). As anticipated, the plasmid encoding only HupX showed no hydrogenase enzyme activity in extracts of strain FTD147 (ΔhyaBΔhybCΔhycE), while both of the other plasmids resulted in a fast-migrating activity band corresponding to the HupSL heterodimer (Figure 1B). Notably, although the activity resulting from introduction of the plasmid lacking the hupX gene (phupSL in Figure 1B) was apparently weaker than that resulting from introduction of phupXSL, both enzyme activities showed very similar migration characteristics, indicating that HupX is neither necessary for the ability of the enzyme to reduce BV nor seems to co-migrate with HupSL in this particular activity band. This result correlates well with earlier mass spectrometric analyses of heterologously expressed enzyme, which identified mainly the HupL protein (Hartwig et al., 2015b).

In order to optimize conditions for the analysis of heterologously produced HupSL activity, we tested different anaerobic growth conditions using FTD147 (ΔhyaBΔhybCΔhycE) transformed with either phupXSL or phupSL (Figure 2A). The activity band was slightly more intense when cells were grown with 0.8% w/v glucose compared with half that glucose concentration (0.4% w/v). Suprisingly, however, no HupSL activity could be detected when cells were grown under anaerobic respiratory conditions, with either glycerol and fumarate or glucose and nitrate. Western blot analysis of the extracts derived from anaerobically grown strains after separation by SDS-PAGE using peptide antibodies raised against HupL revealed that the HupL polypeptide could be detected in each extract (Figure 2B). This indicates that a lack of transcription of the hup genes under respiratory conditions was not the reason for absence of HupSL enzyme activity. Surprisingly, it was not possible to restore in vitro HupSL enzyme activity to these extracts, even by incubating the extracts under reducing conditions. This suggests that the HupSL enzyme was irreversibly inhibited under the oxidizing conditions that prevailed within the cells grown under these conditions.

Quantitative assessment of H₂-dependent BV reduction activity in anaerobically prepared, concentrated crude extracts of FTD147 (ΔhyaB, ΔhybC, ΔhycE) transformed with phupXSL measured a low but detectable hydrogenase activity of approximately 60 mU/mg (Table 2), which is in good agreement with previously determined HupSL activity in E. coli extracts (Hartwig et al., 2015b). The phenotypically identical strain CP1170 (ΔhyaB, ΔhybC, ΔhycE) had a background activity of 10 mU/mg. Brief incubation of the extract from FTD147 + phupXSL in the presence of oxygen resulted in a reduction of the HupSL activity by 50% (Table 2).

The lack of HupSL enzyme activity after respiratory growth is reminiscent of the effects of these growth conditions on appearance of E. coli Hyd-3 and Fdh-H enzyme activities (Sawers et al., 1985), with the exception that the effect on synthesis of the E. coli enzymes is at the transcriptional level due to depletion of the regulatory metabolite formate (Rossmann et al., 1991). Due to the fact that HupSL is naturally associated with a formate dehydrogenase-like enzyme, OmeAB (Kublik et al., 2016; Hartwig et al., 2017), we wished to examine the influence of the Fdh-O and Fdh-N enzymes, which are phylogenetically related to OmeAB, on HupSL enzyme activity. Initially, we introduced into strain FTD147 a mutation in the selC gene, which encodes the selenocysteinyl-tRNA_Sec necessary for translation of special UGA codons as selenocyteine, and which, when deleted, renders all Fdhs inactive (Leinfelder et al., 1988). This would provide information on whether HupSL enzyme activity was influenced by defects in formate metabolism. The left panel shown in Figure 3 shows a control for HupSL activity revealing that is was readily detectable in strain FTD150 (ΔhyaBΔhybCΔhycEΔhyfG), which is identical to strain FTD147 with the exception that the gene encoding the catalytic subunit of Hyd-4 (Andrews et al., 1997) is also deleted (Redwood et al., 2008). Thus, both FTD150 and FTD147 yield an identical phenotype with regard to the heterologous HupSL activity (see also below). Analysis of an extract of the FTD147 ΔselC mutant revealed that no HupSL activity was detectable (Figure 3 right panel). The lack of selC was confirmed by the absence of the H₂:BV oxidoreductase activity associated with Fdh-N/O in the strain (Soboh et al., 2011). This result confirms that HupSL enzyme activity is linked to formate metabolism, most likely through one of the formate dehydrogenases (Fdh) the bacterium synthesizes under anaerobic conditions. Introduction of a mutation in selB, which encodes the special translation factor required to decode the UGA codon as selenocysteine (Forchhammer et al., 1989), into FTD150 also revealed a similar lack of HupSL activity (data not shown), confirming that the phenotype was due to a lack of selenocysteine incorporation.

Dependence on the selenocysteine biosynthetic machinery for appearance of HupSL enzyme activity suggests an involvement of one or more of the three Fdhs present in E. coli. To determine which of the three Fdhs is required for the appearance of heterologous HupSL activity, we constructed a series of strains (see Table 1) lacking one or more of the genes encoding the catalytic subunit of FdnG (of Fdh-N), FdoG (of Fdh-O), or FdhF (of Fdh-H) (Pinske and Sawers, 2016; Figure 4). Strain FTD150, which lacked all four hydrogenases and the quadruple and quintuple mutants of FTD147, which lacked Hyd-1, Hyd-2, Hyd-3 as well as either or both respiratory Fdhs (Fdh-N and Fdh-O), retained fully active HupSL (Figure 4). Hydrogenase activity in extracts derived from FTD147ΔfdnGΔfdoG with plasmid phupXSL was approximately 60 mU, while the strain without plasmid had approximately half this activity (Table 2). Exposure of the crude extract from FTD147ΔfdnGΔfdoG transformed with phupXSL to air resulted in a similar 50–60% reduction in hydrogenase activity as was observed with FTD147 containing phupXSL (Table 2). This result confirms that HupSL is oxygen-labile (Hartwig et al., 2015b).

The only strains that lacked a detectable HupSL enzyme activity band were those that lacked the fdhF gene, which encodes the Fdh-H component of the FHL complex (Figure 4). Assay of hydrogenase activity in extracts derived from FTD147ΔfdhF + phupXSL failed to show HupSL-dependent hydrogenase activity (Table 2). These findings indicate that for full HupSL activity to be manifested, an active Fdh-H enzyme is required.

In order to determine what the link between the appearance of HupSL enzyme activity and Fdh-H might be, we first performed a western blot using anti-HupX antibodies and with crude extracts derived from some of the strains shown in Figure 4. Surprisingly, HupX was only detectable in extracts of strains lacking fdhF, which encodes the Fdh-H enzyme, and, as expected, only in those strains carrying the phupXSL plasmid (Figure 5). This suggests that when Fdh-H was absent, HupX was stably synthesized and when Fdh-H was present in the cells, HupX became unstable and was presumably degraded.

To examine whether HupX might influence HupSL activity, plasmids phupSL and phupXSL, both encoding HupL, HupS and the endoprotease HoxM, but only the latter also encoding HupX (Figure 1A), were introduced into strains FTD147 and FTD147ΔfdhF and enzyme activity was compared after anaerobic growth with glucose (Figure 6A). The products of both plasmids in strain FTD147ΔfdhF showed a strongly reduced activity of HupSL compared with the fdhF⁺ strain FTD147. This result indicates that the dependence on Fdh-H for HupSL activity was retained in the absence of HupX.

A recent study in E. coli identified a flexible interaction network of ferredoxin-like proteins with Fdh-H, including its small, electron-transferring subunit, HycB (Pinske, 2018). HycB and HupX belong to this family but share only 24% amino acid identity (38% similarity) in a MuscleWS alignment. The HupX protein, however, cannot functionally replace HycB in formate-dependent BV reduction (data not shown). Due to the link between HupSL activity and Fdh-H demonstrated above, we therefore examined whether the presence of HycB influenced HupSL’s ability to reduce BV. To do this, we analyzed HupSL activity in strain CP1170 (ΔhyaB, ΔhybC, ΔhycA-I), which is similar to FTD147 (ΔhyaB, ΔhybC, ΔhycE) with the exception that the complete hyc operon is deleted in CP1170, rather than only the hycE gene (Table 1). Introduction of plasmid phupXSL into CP1170 and its ΔfdhF derivative CPH020 (ΔhyaBΔhybCΔhycA-IΔfdhF) (Table 1), revealed that the dependence on Fdh-H for HupSL activity was retained (Figure 4, right side of right panel). However, introduction of plasmid phupSL (lacking hupX) into CPH020 (CP1170 ΔfdhF) revealed that HupSL activity in native PAGE was no longer reduced in the absence of Fdh-H (Figure 6B, lane 1, right panel). Introduction of phupSL into strain CPH008 (ΔhycA-I, ΔfdhF) in which Hyd-1 and Hyd-2 are still active, but all structural components of the FHL complex are missing, demonstrated that HupSL activity was retained, and even slightly more intense (Figure 6B, lane 2 left panel). As a final control, we analyzed HupSL activity after introduction of phupSL into strain RT2 (ΔhyaBΔhybCΔhycA-IΔfdhEΔpflA), which lacks Hyd-1, Hyd-2 and Hyd-3, as well as Fdh-N/O (through the fdhE mutation; Mandrand-Berthelot et al., 1988; Lüke et al., 2008) and the formate-inducible Fdh-H, due to the lack of active PflB (due to the pflA mutation), which is required for formate production (Sawers and Böck, 1988; Rossmann et al., 1991). This strain was chosen because it uses a different combination of mutations to generate the same phenotype, i.e., no hydrogenase or formate dehydrogenase enzymes, and no HycB protein. HupSL activity was also retained in this genetic background confirming that strains devoid of HupX and HycB exhibit HupSL- and H₂-dependent reduction of BV (Figure 6B, lane 1).

Surprisingly, however, when phupSL was introduced into strain RM220 (ΔpflAB), which generates considerably reduced levels of formate under respiratory conditions (Suppmann and Sawers, 1994), and thus expresses only low levels of the formate-inducible hyc operon (Rossmann et al., 1991), the anticipated high activity of HupSL was not observed (Figure 6B, lane 4). This result indicates that even the low levels of HycB produced in this strain (Rossmann et al., 1991) are likely sufficient to inhibit HupSL activity.