Insights into subunit interactions in the Sulfolobus acidocaldarius archaellum cytoplasmic complex



Archaella are the archaeal motility structure that is the functional pendant of the bacterial flagellum but is assembled by a mechanism similar to that for type IV pili. Recently, it was shown by Banerjee et al. that FlaX, a crenarchaeal archaellum subunit from Sulfolobus acidocaldarius, forms a ring-like oligomer, and it was proposed that this ring may act as a static platform for torque generation in archaellum rotation [Banerjee A et al. (2012) J Biol Chem 287, 43322–43330]. Moreover, the hexameric crystal structure of FlaI was solved, and its dual function in the assembly and the rotation of the archaellum was demonstrated [Reindl S et al. (2013) Mol Cell 49, 1069–1082]. In this study, we show by biochemical and biophysical techniques that FlaX from S. acidocaldarius acts as a cytoplasmic scaffold in archaellum assembly, as it interacts with FlaI as well as with the recA family protein FlaH, the only cytoplasmic components of the archaellum. Interaction studies using various truncated versions of FlaI demonstrated that its N- and C-termini interact with FlaX. Moreover, using microscale thermophoresis, we show that FlaI, FlaX and FlaH interact with high affinities in the nanomolar range. Therefore, we propose that these three proteins form the cytoplasmic motor complex of the archaellum.

Structured digital abstract


C-terminal domain


microscale thermophoresis


N-terminal domain


type II secretion system


type IV pilus


Sulfolobus acidocaldarius is a hyper-thermoacidophilic model organism that has been isolated from solfataric hotsprings and is one of the best characterized crenarchaea. In response to changes in the environment, S. acidocaldarius expresses various surface appendages; for example, archaeal adhesive pili, which are type IV pili (T4P), are used for attachment during favorable conditions [1, 2]. In addition, archaella (archaeal flagella) are produced to evade unfavorable conditions by swimming (motility) [3, 4]. The archaellum is the best-characterized surface structure in archaea [4].

The archaellum is structurally similar to T4P of Gram-negative bacteria, but functionally resembles bacterial flagella as it is used for swimming (motility). The homology between the polytopic membrane protein FlaJ in the archaella assembly system and PilC in the T4P assembly systems, the similarity of the P-loop ATPases FlaI and PilT/PilB, the proximal assembly of the subunits FlaB and PilA in the archaellum and T4P filaments, respectively, the narrow diameter of the filament and the lack of a central lumen led to the hypothesis that archaella and T4P are assembled by similar processes [5-7].

Genetic and physiological experiments have shown that deletion of any of the archaella-related proteins leads to non-archaellated cells in all studied phyla [8-11]. In S. acidocaldarius, the archaellum is 12–14 nm in diameter (Fig. 1A) and comprises seven proteins that are encoded by a single gene locus (Fig. 1C). The archaellin FlaB is the filament-forming component of the archaellum cluster. The function of the monotopic membrane proteins FlaF and FlaG is unknown, but they are present in all archaellated archaea. The genetic order of flaHIJ genes is well conserved in all archaella operons [7]. FlaJ is a polytopic membrane protein that forms the membrane core of the archaellum machinery, while FlaH is a RecA family protein containing a classical Walker A motif and an incomplete Walker B motif [7]. Therefore, FlaH may be involved in nucleotide binding and might be involved in regulation of archaellum assembly or rotation. So far, FlaH has only been found in its monomeric form (T.N. and S.-V.A., unpublished results). On the other hand, it has been shown that S. acidocaldarius FlaI forms hexameric species in an ATP-dependent manner [12]. The crystal structure of FlaI was recently solved [13], showing that FlaI has two separate globular domains (the N- and C-terminal domains; NTD and CTD, respectively). The NTD and CTD together form a crown-like structure, in which the CTD is the base and the NTD forms the tip of the crown [13]. The first 29 amino acids on the NTD of FlaI were unstructured in the crystal structure, and deletion of these 29 amino acids of FlaI led to non-motile cells; however, archaella were assembled in this mutant and therefore it was proposed that FlaI is involved in archaellum assembly as well as archaellum rotation [13]. On the other hand, it was shown that FlaX, a Sulfolobales-specific archaellum subunit, forms a ring-shaped oligomeric structure of 30 nm diameter in vitro. The three C-terminal helices were important for ring closure, and it was also demonstrated that the C-terminal half of the protein interacts with the ATPase FlaI [14]. A current model of the archaellum is shown in Fig. 1B.

Figure 1.

(A) Electron micrograph image of Sulfolobus acidocaldarius showing several archaealla. (B) Present model of the archaellum of S. acidocaldarius, in which FlaX, a monotopic membrane protein, forms a ring complex [14]. CM, cell membrane. The S-layer is a proteinaceous layer that enfolds the cell membrane [23]. FlaI forms hexameric species in an ATP-dependent manner, and interacts with the lipids and with FlaX rings [12, 14]. PibD is the membrane-bound class III signal peptidase [24]. (C) S. acidocaldarius archaella gene cluster. The major transcript is flaB, and the second transcript is expressed as flaXGFHIJ, as indicated by the black arrows [8].

In the present study, we characterized the subunit interactions in the S. acidocaldarius motor complex forming the basal body of S. acidocaldarius archaella. We demonstrate that FlaX may be a platform for the motor complex. Using a fluorescence-based biophysical technique, microscale thermophoresis (MST), the interaction of FlaXc (an N terminal 37 amino acid deletion of FlaX) and FlaI was studied in detail. In further experiments, FlaXc interacted with FlaI and FlaH, and FlaI interacted with FlaH. As both the NTD and CTD of FlaI were able to bind to the FlaX ring, we propose that the 14 nm diameter FlaI hexamer fits into the 30 nm diameter FlaX ring. Furthermore, a ternary complex comprising FlaI, FlaX and FlaH was also isolated in vitro, which further supports the hypothesis that the FlaX ring acts as a cytoplasmic scaffold in archaellum assembly.


Archaella subunits have a high affinity towards each other

The stability of FlaX in S. acidocaldarius is dependent on the proteins FlaJ, FlaH, FlaI and also FlaB [8], which may therefore interact with FlaX. FlaX interaction with FlaI has been extensively demonstrated using cell lysate interaction, co-expression pull-down analysis and purified proteins [14]. However, using MST [15], we were able to assess the binding affinities of the two proteins. MST is a fluorescence-based biophysical technique that may be used to determine the affinities of molecular interactions (see Experimental procedures for details). The results confirmed interaction between FlaX and FlaI (Fig. 2A). In this experiment, 11 μm FlaI was labeled with 65 μm of an amine-reactive red N-Hydroxy succinamide fluorescent dye. Upon mixing varying concentrations of unlabeled FlaXc (40 μm–1.22 nm) with 7.5 nm labeled FlaI at a steady-state level under infrared light-mediated local heating, positive thermophoresis of monomeric FlaI was observed that was dependent on changes in the local hydrodynamic radius. The raw data for the capillary scan and the MST profiles of each experiment are shown in Fig. S1. The relative fluorescence was normalized using the mean of Fmax (maximum fluorescence before infrared exposure, cold region). The nanotemper response (hot/cold) [15] after thermophoresis was plotted against the concentration of the unlabeled molecule, the raw data were analyzed using the dose-response sigmoidal fit function, and the dissociation constant (KD) was determined using the half-maximum effective concentration (EC50) or the half-maximum inhibitory concentration (IC50) of the interaction (Fig. 2A–C). In this experiment, FlaXc bound FlaI with very high affinity, with a KD value of 86 ± 32 nm and a Hill coefficient of 0.92 ± 0.33 (Fig. 2A). In order to confirm our hypothesis that FlaX also interacts with FlaH, we performed a further MST experiment. Fluorescent labeling of FlaH was unsuccessful; therefore, purified FlaXc was dialyzed against 25 mm carbonate/bicarbonate pH 9.7, 200 mm NaCl buffer, for better efficiency of amine reactive labeling. The concentration of labeled FlaXc was kept constant at 25 nm, and a variable concentration of FlaH (5 μm–0.15 nm) was titrated at steady-state levels. Figure 2B shows that FlaH bound to FlaXc with a KD value of 42 ± 3.1 nm and a Hill coefficient of 1.36 ± 0.13, indicating positive cooperativity of the interaction. As FlaXc interacts with both cytoplasmic proteins FlaI and FlaH, we investigated whether FlaH and FlaI interact with each other. We therefore performed MST analysis using 35 nm labeled FlaI and variable concentrations of FlaH ranging from 5 μm to 0.31 nm. These proteins showed interaction with the highest affinity measured so far (KD 34.3 ± 2.18 nm, Fig. 2C). The Hill coefficient of 1.86 ± 0.19 also demonstrates positive cooperativity.

Figure 2.

Archaellum subunit interactions. (A) FlaXc interacted with the NT-647-labeled FlaI monomer in the MST experiment. Unlabeled FlaXc (40 μm–1.22 nm) was titrated in 7.5 nm labeled FlaI, and thermophoresis was performed. The data were analyzed by plotting the NT response (hot/cold) of thermophoresis against the FlaXc concentration using a non-linear fit (sigmoidal dose-response). FlaXc interacts with FlaI with a KD of 86 ± 32 nm and a Hill coefficient of 0.92 ± 0.33. (B) Binding of FlaH and FlaXc. Labeled FlaXc (25 nm) was used in a thermophoresis experiment with 5 μm–0.15 nm unlabeled FlaH. The KD was 42 ± 3.1 nm and the Hill coefficient was 1.36 ± 0.13. (C) MST experiment using 35 nm labeled FlaI monomer with 5 μm–0.31 nm unlabeled FlaH, showing that FlaI interacts with FlaH with an affinity (KD) of 34.3 ± 2.18 nm and a Hill coefficient of 1.86 ± 0.19.

Molecular architecture of the FlaI hexamer and its position in the FlaX ring complex

The crown-shaped FlaI hexamer has a diameter of 14 nm. The base of the crown mainly comprises the CTD of FlaI, which contains the active sites for ATP hydrolysis, and the tip of the crown comprises the NTD [13]. FlaI contains several domains that may play important roles in the dual functionality of FlaI. The extreme N-terminal region was unstructured, and the first 29 amino acids were shown to be involved in the switch for assembly to rotation [13]. The unstructured region is followed by a triple helix domain (amino acids 61–127), which is believed to play a key role in interaction with the polytopic membrane protein FlaJ (Fig. 3A).

Figure 3.

Molecular architecture of the FlaI hexamer, and its position in the FlaX core complex. (A) Molecular architecture of FlaI monomer as determined previously [13] and of the truncated versions constructed in present study, FlaI∆C and FlaI∆N. (B) Binary complex formation using pure protein interaction between FlaXc (C-terminal His6-tagged) and FlaI∆C (deletion of amino acids 227–513 of FlaI) with an N-terminal Strep II tag. The co-elution of the proteins suggests direct interaction. (C) Refolded inclusion body of the N-terminal Strep II-tagged FlaXc and FlaI∆N (His6-tagged) interaction was performed using Ni-NTA affinity beads. Co-elution of FlaXc and FlaI was confirmed using Coomassie-stained SDS/PAGE and immunoblotting with α-His and α-strep antibodies. (D) Control experiment for Strep II-tagged FlaXc (refolded inclusion bodies) using Ni-NTA beads. (E) In vitro interaction analysis of His6-tagged FlaI hexamer and strep II-tagged FlaXc. His6-tagged FlaI was hexamerized using the non-hydrolyzable ATP analog adenosine 5′- (β,γ-imido) triphosphate (25 μm). After mixing the proteins, a binary complex eluted through the Ni-NTA column. S, starting material; U, unbound fraction; W, pooled wash fractions; E, elution; M, marker. (F) Hypothetical spatial localization of the FlaI hexamer in the FlaXc ring.

To identify the region of FlaI interacting with the FlaXc ring, we generated two truncated variants of FlaI, FlaI∆N and FlaI∆C, i.e. CTD and NTD deletions of FlaI, respectively (Fig. 3A). The proteins were purified using either Ni2+ affinity or streptavidin-based affinity chromatography (see Experimental procedures for details). In the first sets of experiments, we used purified protein-mediated interaction analysis as described in Experimental procedures. Equal amounts of pure His6-tagged FlaXc were mixed with Strep II-tagged FlaI∆C and loaded onto streptavidin beads, followed by washing and elution steps. The results showed that the NTD of FlaI interacts with the FlaXc ring (Fig. 3B). Interestingly, when we performed a cell lysate interaction analysis using His6-tagged FlaI∆N and strep II-tagged FlaXc on Ni-NTA beads, we observed co-elution of both proteins (Fig. 3C), which indicates that FlaI also binds the FlaXc ring through its CTD. A control experiment using the same protocol was also performed for strep II-tagged FlaXc on Ni-NTA beads (Fig. 3D), showing that the protein did not bind to the column material.

It was previously shown that FlaI maintains a hexameric conformation upon binding to the non-hydrolyzable ATP analog adenosine 5′-(β,γ-imido) triphosphate [12]. Therefore, monomeric FlaI (N-terminal His6-tagged) was incubated with 25 μm adenosine 5′- (β,γ-imido) triphosphate and then incubated with strep II-tagged FlaXc. Ni-NTA beads were used to bind hexameric FlaI, and Fig. 3E shows that FlaXc and hexameric FlaI co-elute.

The FlaXc ring is the scaffold for archaellum assembly

Archaellum rotation depends on the local ATP concentration. In S. acidocaldarius, the dual-function ATPase FlaI provides the energy for both archaellum assembly and swimming motility by hydrolyzing ATP [13]. Above, we have shown that both FlaH and FlaI interact with the FlaXc ring. Moreover, we showed that FlaI interacts with FlaXc via both the NTD and the CTD. We then wished to determine whether these three proteins can interact together. To isolate the ternary complex, we used strep II-tagged FlaXc, His6-tagged FlaH and strep II-tagged FlaI. The three proteins were mixed and incubated with Ni2+ affinity beads (see Experimental procedures for details). After extensive washing, the bound protein fraction was eluted and further characterized using Coomassie-stained SDS/PAGE and western blot analysis using α-His and α-Strep antibodies. The results indicate that the three proteins FlaI, FlaX and FlaH, also forms a ternary complex in vitro (Fig. 4).

Figure 4.

FlaXc rings acts as a platform for archaella assembly. To form a ternary complex of FlaXc, FlaH and FlaI, His-tagged FlaH, Strep II-tagged FlaI and Strep II-tagged FlaXc were mixed and incubated with Ni-NTA affinity beads. The complex was eluted using 50 mm Tris/HCl pH 8, 150 mm NaCl, 500 mM imidazole buffer. The ternary complex was visualized by immunoblotting using α-His and α-strep antibodies. Individual proteins were loaded on the same SDS/PAGE to ensure the migration pattern of pure Strep II-tagged FlaXc, His6-tagged FlaH and Strep II-tagged FlaI. S, starting material; U, unbound fraction; W, wash; E, elution; M, marker. ‘O’ represents the oligomer of FlaXc [14].


Type II secretion systems (T2SS), type IV pili (T4P) and archaella assembly apparatuses are membrane-associated complexes that require ATP hydrolysis for their assembly and/or function. The energy required for all of these membrane-associated nanomachines is provided by soluble ATPases [4, 7, 16]. Association of soluble ATPases with membrane proteins is an essential step for proper assembly or secretion of proteins in T4P assembly or T2SS, respectively. For T2SS, it was shown that one of the membrane-integrated proteins e.g. Vibrio cholerae EpsL, recruits the ATPase, e.g. EpsE, to the membrane by interaction with its C-terminal domain [17]. Monotopic EpsL stimulated the activity of EpsE by 10 000-fold when EpsL and phospoholipid membrane were mixed with EpsE [18]. Similar results were observed for the archaella assembly system, for which the activity of the ATPase FlaI was enhanced by S. acidocaldarius tetraether-linked lipids [12].

The data presented in this study show that FlaX, FlaH and FlaI interact with each other with high affinities (Fig. 5A), and we propose that the FlaXc ring may act as a platform during S. acidocaldarius archaellum assembly. Figure 5C shows our present model, in which the FlaX ring acts as a membrane-bound cytoplasmic assembly platform for subsequent assembly of the FlaI hexamer and FlaH. The affinities of the subunit interactions were in the nanomolar range, which strongly suggests that archaella-associated proteins bind their interacting partners immediately after expression. Interestingly, the FlaXc/FlaH and FlaI/FlaH interactions revealed significant evidence for cooperativity. Given that the apparent affinity of the FlaI/FlaX complex is approximately twofold lower than that for formation of the other two complexes, this may represent a rate-limiting step for basal body assembly. In the pull-down experiment (Fig. 4), FlaX and FlaH interacted with the same stoichiometry, whereas FlaI was present in lower amounts in the elution fraction. This indicates that FlaI and FlaH compete for similar binding sites in FlaX, as, in the FlaI/FlaX interaction assay, FlaI and FlaX eluted with a 1 : 1 stoichiometry (Fig. 3). FlaX may therefore serve as the priming protein for proper archaella assembly in S. acidocaldarius, but is also dependent on the presence of FlaI, FlaH and FlaJ, as FlaX is instable in ∆flaI and ∆flaH deletion strains [8]. Together, these proteins probably form the motor complex of the S. acidocaldarius archaellum.

Figure 5.

FlaX is a membrane-bound cytoplasmic assembly platform. (A) Schematic overview of archaellum subunit interactions for building the motor complex. (B) The polytopic membrane protein FlaJ in S. acidocaldarius consists of seven transmembrane segments (TMS). The cytoplasmic loops cytoI and cytoII are highly positively charged (pI 9.75 and 10.3, respectively). The crystal structure of FlaI revealed that the crown grooves are negatively charged (pI 5.75) and constitute a triple helix domain. Reindl et al. [13] proposed that the positively charged cytoplasmic loops of FlaJ and the negatively charged crown grooves of FlaI interact with each other. CM, cell membrane. (C) Current working model of the archaellum motor complex. FlaX is a monotopic membrane protein, and Banerjee et al. [14] showed that the cytoplasmic domain (FlaXc) forms a 30 nm ring-like oligomer. FlaX binds to FlaI and FlaH with high affinities; FlaI and FlaH also interact with each other. The hexameric FlaI crystal exhibited a 14 nm diameter crown shape. As both the NTD and CTD of FlaI and also the hexameric FlaI interact with FlaXc, we propose that the FlaI hexamer is inside the FlaX ring.

FlaX is a unique crenarchaeal archaellum component, and is not present in euryarchaeal genomes. However, euryarchaea possess three proteins (FlaC/D/E) that are not present in crenarchaeal genomes [4]. Therefore, we hypothesize that, in euryarchaea, these proteins may take over the function of FlaX in stabilizing the two conserved proteins FlaH and FlaI in the archaellum, but this remains an open question.

In T2SS and T4P assembly systems, it was predicted that the polytopic membrane protein GspF or PilC, respectively, interacts with the assembly and/or retraction ATPases through their positively charged cytoplasmic loops, termed cytoI or cytoII, which in turn facilitate membrane association of the ATPases [19, 20]. FlaJ, the polytopic membrane protein is thought to be essential in providing the assembly platform for the archaellum in the membrane [7, 21]. The polytopic membrane protein FlaJ shares homology to the T4P assembly platform-forming inner membrane protein, PilC. FlaJ of S. acidocaldarius has seven TMDs. The N-terminal cytoplasmic domain (cyto1) (amino acids 1–130) and the cyto2 domain (amino acids 233–359) have several stretches of positively charged amino acids, giving high theoretical pI values of 9.75 and 10.3 [13]. In contrast to crenarchaea, there is a putative Nterminal TMD present in euryarchaea that is followed by the cyto1 loop, a stretch of positively charged amino acids on the cytoplasmic side. Interestingly, the cytoplasmic loops show homology to each other [7], and the loops of FlaJ, PilC and GspF are similar [6]. Recently, it was proposed that these charged loops interact with the crown grooves of FlaI, which is partially negatively charged and has a theoretical pI of 5.75 (Fig. 5B). Previously, a model based on FlaI ATP hydrolysis and interaction with the cyto loops of FlaJ suggested generation of a possible 8–10 Å space in the membrane, which may lead to proximal incorporation of FlaB to a newly growing archaellum filament. [13]. This remains to be proven, but the in vivo stability of FlaX is also dependent on FlaJ and FlaB subunits. Therefore, we suggest that, in S. acidocaldarius, the FlaX ring may assemble around FlaJ, and that FlaI confers conformational change of FlaJ transfers to the membrane domain of FlaX ring, which leads to correct incorporation of archaellin (FlaB) into the newly growing archaellum filament.

This study provides an insight into how the cytoplasmic proteins FlaH and FlaI are assembled at the membrane in the archaella assembly complex, and will help to understand their function in assembly and rotation of the archaellum.

Experimental procedures

Plasmids and strains used in present study

The pET-based dual over-expression vector pET Duet1 (Novagen) was used for expression of the FlaXc, FlaI and FlaH proteins in Escherichia coli BL21-CodonPlus-RIL strain. Tables S1 and S2 list plasmid and primer details.

Protein expression and membrane isolation

The E. coli Codon Plus strain was freshly transformed with the pET-based vector and grown overnight in Luria-Bertani medium containing 50 μg·mL−1 ampicillin and 34 μg·mL−1 chloramphenicol. Then 1 mL of pre-culture was used to inoculate 1 L of Luria-Bertani medium supplemented with ampicillin (50 μg·mL−1) and chloramphenicol (34 μg·mL−1), and grown at 37 °C until an attenuance at 600 nm of 0.5–0.6 was reached. Then 1 mm isopropyl thio-β-d-galactoside was used to induce protein over-expression. All inductions were performed at 16 °C overnight for controlled protein expression. The cells were collected by centrifugation at 10 000 g for 25 min at 4 °C, and were resuspended in 50 mm Tris/HCl pH 8, 150 mm NaCl, 1 x protease inhibitor cocktail mix. The resuspended cells were lysed using a Sonoplus HD3100 sonicator (Bandelin Sonorex Biotechnique, Berlin, Germany) with probe HD3100. Cell debris was removed by centrifugation at 10 000 g for 25 min at 4 °C, and the cell-free lysate was centrifuged at 100 000 g for 45 min at 4 °C to collect the membrane and the soluble fractions. The membrane fraction was solubilized using 50 mm Tris/HCl buffer, pH 8, 150 mm NaCl, 5% glycerol, 20 mm imidazole and 2% w/v n-dodecyl-β-maltoside, followed by incubation at 50 °C for 1–2 h for solubilization of the membrane proteins [14]. The fractions were analyzed by Coomassie-stained SDS/PAGE and immunoblot analysis using α-His, α-FlaX or α-FlaI antibodies.

Purification of recombinant proteins

His6-tagged FlaH was purified using Ni-NTA beads. Purification was performed in 50 mm MES pH 6.2, 500 mm NaCl buffer, using a 20–500 mm imidazole gradient on a gravity column. The elution fraction was desalted in 50 mm MES pH 6.2, 500 mm NaCl buffer, concentrated and kept at −80 °C for future experiments. Strep II-tagged proteins (FlaI or FlaI∆C) were purified using streptavidin beads. A concentration gradient of 2.5–5 mm desthiobiotin in 50 mm Tris/HCl, pH 8, 150 mm NaCl buffer, was used to elute Strep II-tagged FlaI from the column, and the pure sample was desalted and kept for future use. His6-tagged FlaI and His6-tagged FlaI∆N were purified as described previously [13]. Moreover, His6-tagged FlaXc was purified using the protocol described previously [14] with a slight modification. His-tagged FlaXc was purified using an IMAC (immobilized metal ion affinity chromatography) column on a Bio-Rad (Berkeley, CA, USA) Profinia purification system, and the purified protein was desalted using a Profinia desalting column with 25 mm glycine/NaOH pH 10, 200 mm NaCl buffer and purified FlaXc was concentrated to 10 mg·mL−1.

Isolation and refolding of inclusion bodies

Upon heterologous over-expression, strep II-tagged FlaXc was present solely in inclusion bodies. To isolate strep II-tagged FlaXc, inclusion bodies were resuspended in buffer I (50 mm Tris/HCl pH 8, 150 mm NaCl, 3 m guanidium hydrochloride, 1% Triton X-100, 3% glycerol). Resuspended inclusion bodies were incubated on ice for 20–30 min. To ensure proper solubilization, resuspended inclusion bodies were further sonicated and incubated under constant stirring on ice for 3 h or overnight. Solubilized inclusion bodies were separated from insoluble particles by centrifugation at 10 000 g for 30–60 min. Refolding of proteins from solubilized inclusion bodies was performed by three-step dialysis in buffer R (100 mm Tris/HCl pH 8, 500 mm l-arginine) by reducing the guanidium hydrochloride concentration (1, 0.25 and 0.125 M). To ensure proper refolding, dialysis was performed in buffer R only. To remove l-arginine, a final last dialysis step was performed using buffer A (100 mm Tris/HCl pH 8, 150 mm NaCl). Centrifugation at 13 000 g for 10 min was performed to ensure protein solubility. The soluble fraction was further dialyzed against 25 mm glycine/NaOH pH 10, 200 mm NaCl buffer, and concentrated to 5 mg·mL−1. Refolded FlaXc was flash-frozen in liquid N2 and kept at −80 °C for further use. Correct folding of FlaXc was tested by CD spectroscopy (Fig. S2).

Far-UV CD spectroscopy

The structural integrity of refolded strep II-tagged FlaXc was determined using CD spectroscopy, and CD spectra were recorded at room temperature using a Jasco (Easton, MD, USA) J-810 spectropolarimeter. Experiments were performed in 1 mm path length cells. Pure refolded inclusion body was dialyzed against 25 mm carbonate/bicarbonate pH 9.7, 200 mm NaCl buffer, to ensure no interference arising from the ‘buffer only’ spectrum. Confirmation of the structural integrity of the refolded inclusion body was obtained by acquisition of far-UV (200–280 nm) CD spectra. Far-UV spectra of the tested proteins are presented as the smoothed mean of three accumulations. Secondary structure prediction analysis was performed using the supplied J-810 software, using the Yang statistical algorithm [22]. All recorded CD spectra were baseline-corrected by subtraction of the ‘buffer only’ spectrum.

Binary or ternary complex formation

In vitro protein-protein complex formation was performed as described previously [14]. To form the binary or ternary complexes, pure proteins tagged with either His6 or Strep II were incubated together in a 50 mm Tris/HCl pH 8, 150 mm NaCl buffer, and then applied to affinity beads (Ni-NTA; Sigma Aldrich, Seelze, Germany or streptavidin; IBA GmbH, Göttingen, Germany, respectively). Complexes were eluted with either 250–500 mm imidazole or 2.5 mm desthiobiotin, visualized by Coomassie-stained SDS/PAGE, and α-His or α-Strep antibodies.

Affinity determination using thermophoresis

Microscale thermophoresis (NanoTemper Technologies GmbH, Munich, Germany) is a fluorescence-based biophysical technique used to determine the binding affinities of proteins in solution by using infrared laser to induce a local temperature gradient in glass capillaries and thus to monitor thermophoretic movement. The detailed procedure has been described previously [15]. Purified His6- tagged FlaI was precipitated using 80% ammonium sulfate to eliminate all bound nucleotides and to ensure that only monomeric FlaI was present. The precipitated protein was refolded in 50 mm HEPES pH 7.5, with 150 mm NaCl buffer. The ATPase activity of the monomeric FlaI was analyzed in order to confirm proper folding (data not shown). The FlaI monomer was fluorescently labeled using the amine-reactive Monolith NT™ protein labeling kit NT-647 (NanoTemper Technologies GmbH) according to the manufacturer's instructions. FlaXc was purified in glycine/NaOH buffer, and, as glycine interferes with the labeling efficiency of the dye, pure FlaXc was dialyzed against 25 mm carbonate/bicarbonate, pH 9.7, 200 mm NaCl buffer. FlaXc was labeled using the same protocol as FlaI. To determine the binding affinity of the FlaI/FlaH binary complex, 35 nm labeled FlaI was used and FlaH was titrated in a 1 : 2 serial dilution. For the FlaXc/FlaI interaction, 7.5 nm labeled FlaI was used, and for the FlaXc/FlaH interaction, 25 nm labeled FlaXc was used. The measurements were performed at 298 K, 80% LED power and 20% infrared-laser power, which induces a temperature increase of ~ 2 K. Samples were incubated for 10–20 min before measurements. The laser on and off times were adjusted to 30 and 5 s, respectively. The FlaI/FlaH interaction experiment was performed in 50 mm Tris/HCl pH 8 buffer, containing 150 mm NaCl and 0.05% Tween-20, the FlaXc/FlaI interaction experiment was performed in 25 mm glycine/NaOH pH 10 buffer, containing 200 mm NaCl and 0.05% Tween-20, and the FlaXc/FlaH interaction experiment was performed in 25 mm carbonate/bicarbonate pH 9.7 buffer, containing 200 mm NaCl and 0.05% Tween-20. The measurements were performed using a NanoTemper Monolith NT.115 instrument in standard treated capillaries, and analyzed using nanotemper analysis software version 1.4.27 (NanoTemper Technologies GmbH). To generate representative curves, the raw data were extracted and plotted using origin 6.1 (OriginLabs, Northampton, MA, USA), and the KD value and Hill coefficient were calculated using a non-linear sigmoidal (dose-response) fitting equation provided in the software.


A.B., P.T. and S.-V.A. were supported by intramural funds from the Max Planck Society. T.N. received support from European Research Council starting grant ARCHAELLUM/311523. We thank Anna-Lena Henche, Max Planck Institute for Terrestrial Microbiology, for providing the electron micrograph image of Sulfolobus acidocaldarius.