The long Q‐loop of Escherichia coli cytochrome bd oxidase is required for assembly and structural integrity

Cytochrome bd‐I oxidase is a terminal reductase of bacterial respiratory chains produced under low oxygen concentrations, oxidative stress, and during pathogenicity. While the bulk of the protein forms transmembrane helices, a periplasmic domain, the Q‐loop, is expected to be involved in binding and oxidation of (ubi)quinol. According to the length of the Q‐loop, bd oxidases are classified into the S (short)‐ and the L (long)‐subfamilies. Here, we show that either shortening the Q‐loop of the Escherichia coli oxidase from the L‐subfamily or replacing it by one from the S‐subfamily leads to the production of labile and inactive variants, indicating a role for the extended Q‐loop in the stability of the enzyme.

Cytochrome bd quinol oxidases are terminal oxidases of many prokaryotes, including several pathogens [1-3]. The enzyme complex is not related to the well-characterized family of heme-copper oxidases [4,5]. The bd oxidases couple (ubi)quinol oxidation and release of protons to the periplasmic side with proton uptake from the cytoplasmic side to reduce dioxygen to water. In doing so, bd oxidases contribute to the generation of the protonmotive force by vectorial charge transfer [6,7]. Universally, bd-I oxidase is made up of two major subunits CydA and CydB that share the same fold comprising two four-helix bundles and an additional cytoplasmic helix (Fig. 1). CydA harbors the cofactors, namely heme b 558 , b 595 , and d. Heme b 558 accepts electrons from the quinol and transfers them to the heme active site composed of heme b 595 and d where dioxygen is reduced to water [1,8]. In addition, CydA contains a globular domain located on the periplasmic side between transmembranous (TM) helices 6 and 7 (Fig. 1) [9]. This region, termed Q-loop, is expected to be involved in binding and oxidation of the substrate (ubi)quinol (see below) [9][10][11].
It was found that bd oxidases contain a third subunit, called CydX in Escherichia coli and CydS in Geobacillus thermodenitrificans [12][13][14]. CydX consists of 30-40 amino acids, and it is encoded by the cyd operon. In the oxidase, it is located at the interface of TM helices 1 and 6 of CydA probably stabilizing the active heme center (Fig. 1). Surprisingly, E. coli bd oxidase contains a fourth subunit called either CydY or CydH that is not encoded in the cyd operon but derives from the orphan gene ynhF [15,16]. CydY is located in a hydrophobic cleft shaped by TM helices 1 and 9 of CydA and is absent in the G. thermodenitrificans enzyme. Because CydY blocks the dioxygen entry site present in both enzymes, the E. coli enzyme contains an additional unique substrate channel leading from CydB to heme d (Fig. 1). In accordance with the different positions of the dioxygen binding sites, the positions of heme b 595 and d are interchanged in the E. coli enzyme with respect to the G. thermodenitrificans oxidase [15,16].
The Q-loop of CydA is in close proximity to heme b 558 , the primary electron acceptor, implying a function in quinol oxidation [14][15][16]. Due to the presence of a short (S) or a long (L) Q-loop, the family of bd oxidases is classified into the S-and L-subfamilies [1]. The Q-loop of the bd oxidases from the S-subfamily (e.g., G. thermodenitrificans) has a length of about 70 amino acid residues, while the members of the L-subfamily (e.g., E. coli) contain a C-terminal extension of about 60 amino acid residues ( Fig. 1) [1, [14][15][16][17]. The N-terminal region of the Q-loop shows a higher amount of conserved amino acid residues than the C-terminal portion [10,11]. Recent structural data imply that the N-terminal domain of the Q-loop is intrinsically flexible and that binding of the quinone-site inhibitor aurachin D reduces its flexibility, yet still without leading to a defined structural conformation. In contrast, the additional C-terminal extension of the Q-loop present in the members of the L-subfamily is rather rigid and extends all over the periplasmic surface of CydA to the CydAB interface [15,16]. This implies a functional role for the N-terminal domain of the Q-loop and a structural role for its C-terminal extension that is only found in members of the L-subfamily.
Here, we truncated the Q-loop of the E. coli bd-I oxidase as well as changed the enzyme into a bd oxidase of the S-subfamily. Truncations were made based on structure comparison between the E. coli and the G. thermodenitrificans enzyme (Fig. 1). It turned out that all mutations led to the production of variants without the active heme center and consequently to an inactive and unstable oxidase that could not be purified by chromatographic means. Thus, the C-terminal extension of the Q-loop indeed seems to stabilize the entire enzyme complex emphasizing its structural role.

Construction of expression plasmids
Three expression plasmids were generated from plasmid pET28b(+) cydA his BX that encodes the E. coli bd-I oxidase [12]. The newly designed plasmids were designated as Q short , Q helix , and Q Geo . The plasmid Q short encodes an oxidase that lacks amino acid residues 250 to 384, thus, the entire Q-loop. To link TM helices 6 and 7 of CydA, nucleotides encoding a linker sequence 'EDERP' that connects these two helices in the G. thermodenitrificans bd oxidase were inserted. The plasmid Q helix lacks the nucleotides encoding amino acid residues 258-384. Here, only the N-terminal helix of the E. coli Q-loop is encoded in the plasmid and nucleotides coding the linker sequence 'DGDGDP' were inserted that were predicted to form a flexible loop preserving the rest of the structure [18]. The intention behind this approach was that the N-terminal helix of the E. coli Q-loop might be needed to stabilize CydA. The plasmid Q Geo encodes the E. coli oxidase equipped with the Q-loop of the G. thermodenitrificans oxidase. The E. coli nucleotides encoding amino acid residues 258-384 were deleted, and nucleotides encoding the G. thermodenitrificans oxidase from position 258-317 were inserted. The resulting sequences are shown in Table S1. Oligonucleotides Q_short_fwd and Q_short_rev were used to generate plasmid Q short , and oligonucleotides Q_he-lix_fwd and Q_helix_rev were used to generate Q helix . The forward and reverse oligonucleotides share homologous regions to simplify recombination of the linear fragment.
To generate a chimera of the E. coli bd-I oxidase comprising the Q-loop of the G. thermodenitrificans enzyme, the sequence coding the G. thermodenitrificans Q-loop was amplified from the synthetic plasmid pEX-A128 (Table S2; Eurofins Genomics, Ebersberg, Germany) containing the corresponding sequence with the oligonucleotides Q Geo in-sert_fwd and Q Geo insert_rev. The resulting linear fragment was used as primer in a second PCR to amplify the pET28b(+) cydA his BX Q helix plasmid. Hundred and fifty nanogram oligonucleotide and 70 ng pET28b(+) cydA his BX Q helix were used. The sequences of the oligonucleotides (all from Sigma-Aldrich, M€ unchen, Germany) are listed in Table S3. All PCR products were digested with DpnI and purified by agarose gel electrophoresis. The KOD polymerase (Merck Millipore, Darmstadt, Germany) was used according to manufacturer's specifications. Newly generated vectors were checked by sequencing (GATC Biotech, Konstanz, Germany). Restriction enzymes were obtained from Thermo Fisher Scientific (Darmstadt, Germany) or Merck Millipore. All plasmids used in this work are listed in Table S4.
Expression strains and cell growth E. coli strain CBO was used for electroporation [12]. This strain is a derivative of E. coli strain C43(DE3) chromosomally lacking cydABX and appBCX (designated: C43(DE3) recA, cydABX, appBCX). Thus, CBO lacks the genes of both bd-type oxidases leaving cytochrome bo 3 as the only respiratory oxygen reductase. This strain was individually transformed with pET28b(+), pET28b(+) cydA his BX (called WT hereafter), pET28b(+) cydA his BX Q short (called Q short hereafter), pET28b(+) cydA his BX Q helix (called Q helix hereafter), or pET28b(+) cydA his BX Q Geo (called Q Geo hereafter) and grown aerobically in 2 L baffled flasks containing 800 mL LB medium at 37°C. Gene expression was induced at an OD 600 of~2 by an addition of 0.4 mM IPTG. Cells were harvested in the late exponential phase and stored at À80°C.

Membrane preparation and detergent extract
Frozen cells (15-20 g wet weight) were suspended in the fivefold volume 50 mM MOPS, 100 mM NaCl, 0.5 mM PMSF, pH 7.0 and disrupted by single pass through a French pressure cell (SLM Alminco) at 17 000 psi. Cell debris was removed by centrifugation at 9500 g for 20 min. Cytoplasmic membranes were sedimented by ultracentrifugation of the cleared lysate at 250 000 g for 75 min and subsequently suspended in 20 mM MOPS, 20 mM NaCl, 0.5 mM PMSF, pH 7.0 to a final protein concentration of 10 mgÁmL À1 . Membrane proteins were solubilized by incubating the membrane suspension 90 min with 1% LMNG (Anatrace, Maumee, OH, USA) under mild stirring. Nonsolubilized material was sedimented by ultracentrifugation at 250 000 g for 15 min.

Redox difference spectra
UV/Vis absorption spectra were recorded using a Tidas II Diode Array Spectrometer (J&M Analytik AG, Essingen, Germany). First, a spectrum of the air-oxidized membranes was recorded. The same sample was then reduced by an addition of a few grains of dithionite, and the spectrum of the reduced sample was recorded using the same set of parameters. The spectrum of the oxidized sample was subtracted from that of the reduced sample resulting in the dithionite-reduced minus air-oxidized difference spectra. The air-oxidized and the ferricyanide-oxidized spectra exhibited no difference.

Oxidase activity
The activity of the oxidases in membranes from the parental strain and the mutants was determined by measuring the NADH:oxygen oxidoreductase activity with an oxygen electrode (oxygraph+, Hansatech) at 30°C. Five microliter membrane suspensions (~50 mgÁmL À1 ) were added to 2 mL buffer (20 mM MOPS pH 7.0, 20 mM NaCl). Strain CBO is lacking both bd-type oxidases. When indicated, the activity of the bo 3 oxidase was inhibited by an addition of 1 mM KCN to the buffer. The reaction was started by an addition of 5 lL NADH (0.5 M). The rates were corrected for the nonenzymatic rate that was < 1% of the rate of the enzymatic reaction. Each data point was assayed in triplicates from three different biological samples.
Western blot analysis SDS/PAGE was performed according to von Jagow and Sch€ agger [19]. Subsequently, proteins were electroblotted onto 0.45 µm pore size PVDF membrane (Schleicher and Sch€ ull, M€ unchen, Germany) according to Ref. [20]. The mouse anti-His-4 antibody was purchased from Merck Millipore and the secondary goat anti-mouse IgG, AP conjugated, from Qiagen (Hilden, Germany).

Attempts to purify bd oxidase variants
Cytoplasmic membranes were prepared as described above. All steps were carried out at 4°C. Membrane proteins were solubilized by incubating the membrane suspension 1 h with 1% LMNG (Anatrace). Nonsolubilized material was sedimented by ultracentrifugation at 250 000 g for 15 min. The supernatant was applied to a 1 mL HisTrap HP (GE Healthcare, Freiburg, Germany) column equilibrated in 50 mM MOPS, 100 mM NaCl, 0.03 mM PMSF, 0.003% LMNG, pH 7.0 (buffer A) containing in addition 20 mM imidazole, with a flow rate of 1 mLÁmin À1 . The column was washed with buffer A containing 68 mM imidazole until the absorption dropped to the original baseline level. Bound proteins were eluted in a 20 mL linear gradient from 68 to 212 mM imidazole in buffer A.

Results and Discussion
Cells were grown under aerobic conditions in LB medium, and expression of the oxidase genes was induced by an addition of IPTG. Strain CBO/pET28b(+) grew to an OD of about 4.0 resulting in 6-9 g cells (wet weight) per L medium (Fig. 2). The strains with an insert in pET28b(+) grew significantly slower to an OD between 1.2 and 2.2 and yielded~3-5 g cells (wet weight) per L medium (Fig. 2). This indicates that gene expression resulted in protein production. Cytoplasmic membranes from the various strains differed in color. Membranes from strain CBO/pET28b(+) cydA his BX had a brownish-reddish color, while those from the strains with pET28b(+) cydA his BX Q short and pET28b(+) cydA his BX Q helix showed a light reddish color. Membranes from strains CBO/pET28b(+) cydA his BX Q Geo and CBO/pET28b(+) were greenishbrown suggesting a higher amount of bo 3 oxidase in the membranes.
The heme content of the cytoplasmic membranes was determined by UV/Vis spectroscopy (Fig. 3). The reduced-minus-oxidized difference spectrum of membranes from strain CBO/pET28b(+)cydA his BX (WT strain) showed the signals of heme b 558 at 430 nm and of heme b 595 at about 440 nm in the c-Soret region. The signal of heme d is superimposed by that of heme b 558 at 430 nm. Additional absorbance of heme b 558 was detected at 531 and 562 nm, of heme b 595 /d at 594 nm and of heme d at 629 and at 655 nm (Fig. 3). Membranes from the strain with the empty plasmid and the strains with the plasmids coding for the bd variants just showed the signals of heme b 558 , but lacked absorbance of heme b 595 and heme d (Fig. 3). The signals at 430, 531, and 562 nm contained contributions from the b-type hemes of bo 3 oxidase [8] and succinate dehydrogenase [22].
The activity of the bd oxidase variants in the membrane was measured as NADH oxidase activity in the presence and absence of 1 mM KCN to inhibit bo 3 oxidase [1,7] (Table 1). Strain CBO with the empty plasmid features the bo 3 oxidase as sole respiratory oxidase with an IC 50 of 10 µM toward KCN [8]. Thus, the oxidase activity is nearly fully inhibited by an addition of 1 mM KCN. The bd oxidase has an IC 50 of 2 mM toward KCN [21] leading to an only minor inhibition under the applied conditions. Due to this, the activity of the strain producing the WT bd oxidase was only inhibited to 38% by KCN demonstrating the presence of a functional bd oxidase in the membrane. The oxidase activity of strains expressing Q short , Q helix , and Q Geo was also inhibited by 94% and 93% by KCN, respectively, indicating a complete lack of a functional bd oxidase. Noteworthy, the oxidase activity of Q Geo in the absence of KCN is in the same range as the one obtained with the strain producing the WT bd oxidase (Table 1). Since the activity of bd oxidase is lower compared to that of bo 3 oxidase [1,8] and as the susceptibility to KCN inhibition indicates enzymatic activity mainly of the bo 3 oxidase, we conclude that the production of a bd oxidase assembly apparently suppresses the production of the bo 3 oxidase. Thus, the cellular levels of bo 3 oxidase are higher in the strain carrying the empty plasmid than in the strain producing the WT bd oxidase. Conclusively, the cellular levels of bo 3 oxidase are substantially higher in the strains producing the Q short and Q helix variants. However, the amount of bo 3 oxidase is significantly reduced in the strain producing the Q Geo variant to approximately the same level as in the strain producing the WT bd oxidase (Table 1). This suggests that the Q Geo variant is produced most efficiently among the variants, although it is inactive (Table 1).
The amount of the bd oxidase variants in the mutant membranes was detected by western blot analysis using an antibody raised against the His-tag on CydA (Fig. 4). A band with an apparent molecular mass of about 53 kDa was detected in membranes from wild-type cells (Fig. 4). The molecular mass of CydA as deduced from its DNA sequence is 59 kDa. However, it is well known that due to its hydrophobic nature, CydA binds more SDS than a standard protein resulting in significantly lower apparent molecular masses after SDS/PAGE [12]. The host strain CBO does not encode the bd oxidase, and, consequently, no signal was seen in this lane. Surprisingly, no signal could be detected in membranes from the strain producing the Q short variant (Fig. 4), indicating that the variant protein is rapidly degraded within this strain. A faint band at around 35 kDa was detectable in the lane loaded with membranes from the strain containing the Q helix variant in agreement with its predicted molecular mass of 46 kDa. Thus, the N-terminal helix of the Q-loop indeed stabilizes CydA. A more intense band at 40 kDa was detected in membranes of strain Q Geo , which contains CydA with a molecular mass of 51.5 kDa. As judged from the intensities, the concentration of CydA in the membrane is about one-third in the Q Geo mutant and one-tenth in the Q helix mutant, respectively, in comparison with the wild-type strain. Thus, as expected from activity measurements, the amount of bd oxidase is somewhat reduced in strain Q Geo , drastically reduced in strain Q helix , and not detectable in strain Q short .
Attempts to purify the bd oxidase variants from the mutant strains failed. Membrane proteins were extracted with 1% LMNG, and UV/Vis difference spectra were taken from the cleared extract (Fig. 3). The typical absorbance of the heme groups from bd oxidase was detectable in the extract from the strain producing WT bd oxidase. However, the detergent extract from membranes of the strains expressing the bd oxidase variants showed just the signals from heme b 558 at 430, 531, and 562 nm, while signals from heme b 595 and heme d were not detectable (Fig. 3). Thus, none of the mutants coding variants of bd oxidase produced a stable and fully assembled bd oxidase. The heme b signals in the extract from the strain containing the empty plasmid most likely derived from bo 3 oxidase [8] and succinate dehydrogenase [22]. In an attempt to purify bd oxidase from strain Q Geo exhibiting the highest amount of the variant oxidase among all mutant strains (Fig. 4), the detergent extract was loaded onto the affinity chromatography column. However, no protein could be eluted from the column, while the flow-through exhibited a UV/Vis spectrum reminiscent to the one obtained with the extract from this strain (Fig. S1). A minor peak within the elution profile most likely derives from bound bo 3 oxidase that contains a natural His-tag.

Conclusion
The role of the different lengths of the Q-loop of bd-I oxidases from various species is not understood [1,8,22,23]. It might play a functional role in quinol binding and reduction, a structural role in conferring rigidity to the protein, or a combination of both. All bd-I oxidases contain the N-terminal part of the Qloop that was shown to participate in quinone binding [9,10,24-27]. Furthermore, it builds a hydrophilic loop in the E. coli enzyme that covers heme b 558 to protect it from being solvent-exposed [15,16]. Hydrogen/deuterium exchange mass spectrometry (HDX-MS) showed that binding of a quinone-competitive inhibitor just affects the N-terminal part of the E. coli long Q-loop [15]. Members of the L-subfamily of bd-I oxidases feature a C-terminal extension of the Q-loop that is rigid and covers the periplasmic surface of CydA to the interface toward CydB [15,16]. This part  of the Q-loop was not influenced by binding of a quinone and a quinone-competitive inhibitor as determined by HDX-MS [15]. The structures of bd-I oxidase from G. thermodenitrificans and E. coli thus imply that the N-terminal part of the Q-loop is involved in quinone binding, while the C-terminal extension confers stability [14][15][16]. Here, we show that the deletion of the entire E. coli bd-I oxidase Q-loop either led to a perturbed production or a rapid degradation of the oxidase so that no signal from CydA was obtained in western blot analysis. Retaining solely the N-terminal helix of the Q-loop led to the production of an inactive enzyme that is not fully assembled (Figs 3 and 4, Table 1). The same phenotype was observed when the long Q-loop was replaced by the short one from the G. thermodenitrificans enzyme in the Q Geo variant or, in other words, when the rigid Cterminal part of the Q-loop was lacking (Fig. 1). Noteworthy, this protein accumulated in considerable amounts in the mutant membranes (Fig. 4). However, the Q Geo variant seems to be too fragile to withstand protein purification (Figs 3 and S1). These findings support the proposed structural role of the C-terminal extension that might be caused by the special mechanistic needs of the E. coli enzyme. Because the long Q-loop extends to the CydAB interface [15,16], it might strengthen the interactions between the two subunits. The oxygen entry channel of E. coli bd-I oxidase starts on CydB and extends further to heme d in CydA. A close connection between the two subunits might be necessary to maintain this substrate channel. This is in agreement with the fact that homologues of CydY have so far only been detected in bd-I oxidases from the L-subfamily [15] and suggests that all enzymes belonging to this subfamily employ CydY to seal the common oxygen entry site, and thus, have to use the alternate oxygen access channel. It is most likely that the acquisition of CydY led to the evolution of the extended oxygen entry channel, a necessity to maintain a functional enzyme. Accordingly, all oxidases featuring a homologue of CydY would share these properties. Following this line of arguments, the presence of CydY should be concomitant with a long Q-loop and both indicators might be used as a probe to classify oxidases as functioning mainly as terminal oxidases exemplified by the E. coli bd oxidase or detoxifying enzymes exemplified by the G. thermodenitrificans enzyme [16].

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Attempts to purify the Q Geo variant by affinity chromatography. Table S1. Amino acid sequence of the Q-loop present in Q short , Q helix and Q Geo. Table S2. Sequence of pEX-A128. Table S3. Oligonucleotides used in this work. Table S4. Plasmids used in this work.