Research ArticleStructural Biology

Architecture of the complete oxygen-sensing FixL-FixJ two-component signal transduction system

See allHide authors and affiliations

Sci. Signal.  10 Apr 2018:
Vol. 11, Issue 525, eaaq0825
DOI: 10.1126/scisignal.aaq0825

Signal relay through a two-component system

In two-component systems, a sensor histidine kinase undergoes autophosphorylation and transfers a phosphate group to its cognate response regulator, which then mediates cellular responses by binding to DNA, performing enzymatic reactions, or interacting with other proteins. In the FixL-FixJ two-component system of the plant root nodule symbiont Bradyrhizobium japonicum, the histidine kinase FixL undergoes autophosphorylation only when it is not bound to oxygen. This ensures that its cognate response regulator FixJ stimulates the expression of genes required for nitrogen fixation only under low-oxygen conditions. Wright et al. combined high- and low-resolution structural analyses with modeling techniques and functional analysis to generate a model of signal relay through the FixL-FixJ two-component system. The model shows how the dissociation of oxygen from FixL stimulates FixL autophosphorylation and phosphotransfer from FixL to FixJ.


The symbiotic nitrogen-fixing bacterium Bradyrhizobium japonicum is critical to the agro-industrial production of soybean because it enables the production of high yields of soybeans with little use of nitrogenous fertilizers. The FixL and FixJ two-component system (TCS) of this bacterium ensures that nitrogen fixation is only stimulated under conditions of low oxygen. When it is not bound to oxygen, the histidine kinase FixL undergoes autophosphorylation and transfers phosphate from adenosine triphosphate (ATP) to the response regulator FixJ, which, in turn, stimulates the expression of genes required for nitrogen fixation. We purified full-length B. japonicum FixL and FixJ proteins and defined their structures individually and in complex using small-angle x-ray scattering, crystallographic, and in silico modeling techniques. Comparison of active and inactive forms of FixL suggests that intramolecular signal transduction is driven by local changes in the sensor domain and in the coiled-coil region connecting the sensor and histidine kinase domains. We also found that FixJ exhibits conformational plasticity not only in the monomeric state but also in tetrameric complexes with FixL during phosphotransfer. This structural characterization of a complete TCS contributes both a mechanistic and evolutionary understanding to TCS signal relay, specifically in the context of the control of nitrogen fixation in root nodules.


Two-component systems (TCSs) are widely distributed in bacteria, fungi, and higher plants. They facilitate cellular adaptation in response to environmental change and are considered good targets for the development of novel antibiotics and plant growth modulators because of their conspicuous absence in metazoans (13). TCSs are generally composed of two types of multidomain proteins: sensory histidine kinases (HKs) and response regulators (RRs). The TCSs can be classified into three classes, based on their domain architectures. Class I HKs consist of an N-terminal stimulus-specific sensor domain and a C-terminal HK module. The latter comprises the dimerization and histidine phosphotransfer (DHp) and catalytic adenosine triphosphate (ATP)–binding (CA) domains. Class II HKs, which are specific for chemotaxis (4, 5), contain an N-terminal histidine-containing phosphotransfer (HPt) domain and a C-terminal HK module. Class III HKs have features of both class I and class II HKs, combining the class I HK’s sensor domain and the HPt domain and HK module of the class II HKs (5, 6). RRs contain a conserved N-terminal receiver (REC) domain, which is connected to diverse C-terminal effector domains. In response to an environmental stimulus sensed by the HK sensor domain, the CA domain catalyzes the autophosphorylation of a specific histidine residue in the DHp (class I) or HPt (class II and class III) domain. That phosphoryl group is subsequently transferred to a conserved aspartate residue in the REC domain of the cognate RR. This phosphotransfer activates the RR to promote DNA or RNA binding, enzymatic reactions, or protein interactions that are mediated by the C-terminal effector domain (7). When autophosphorylation activity is turned off, HKs often act as a phosphatase of their cognate phosphorylated RRs (8), thus contributing to shutting down the pathway.

The key questions of TCS signaling are how HKs are activated by stimuli and how they interact with the RRs. The answers to these questions have been hampered by the lack of molecular- and atomic-level structural information on intact, full-length HKs in both the kinase-active and kinase-inactive forms and in complex with RRs. Here, we describe the structural characteristics of the oxygen (O2)–sensing FixL and FixJ (FixL-FixJ) TCS of the rhizobium species Bradyrhizobium japonicum, a root nodule, nitrogen-fixing bacterium that forms symbiotic relationships with leguminous plant such as soybean. B. japonicum FixL is a class I HK that senses the O2 tension in the cytoplasm through a heme-containing Per-Arnt-Sim (PAS) domain and transfers phosphate from ATP by sequential autophosphorylation and phosphotransfer reactions in its C-terminal effector modules to the RR FixJ (Fig. 1). O2 association to and dissociation from the heme-PAS domain of FixL trigger intra- and intermolecular signaling mechanisms such that the deoxy (O2-unbound) form of FixL is active for autophosphorylation and phosphotransfer, whereas the oxy (O2-bound) form does not undergo autophosphorylation and exhibits phosphatase activity toward FixJ (9, 10). As a result, the rhizobial FixL and FixJ system stimulates the expression of genes required for nitrogen fixation only when O2 concentrations in the plant root nodules are low because the active center of nitrogenase is O2-labile (11).

Fig. 1 Schematic representation of the domain structures of full-length and truncated versions of B. japonicum FixL and FixJ.

Full-length Bradyrhizobium japonicum FixL comprises two N-terminal Per-Arnt-Sim (PAS) domains, PAS-A and PAS-B, and C-terminal dimerization and histidine phosphotransfer (DHp) and catalytic adenosine triphosphate (ATP)–binding (CA) domains. His200 in the PAS-B domain is critical for binding to heme; His291 in the DHp domain is the site of autophosphorylation; and Asp431-Val467 of the CA domain constitutes the ATP-binding site. Full-length B. japonicum FixJ contains an N-terminal receiver (REC) domain and a C-terminal effector domain that binds to DNA. FixJ is activated by FixL-mediated phosphorylation at Asp55. Structures of the truncated FixL and FixJ proteins FixLPAS-PAS and FixJN used in this study are indicated. The residues that define the boundaries of these domains are noted. Domain structures were generated using the SMART Tool (

Although most TCS HKs, including FixL homologs from most other species, are integrated into the membrane for sensing extracellular stimuli, B. japonicum FixL is a water-soluble, cytoplasmic sensor. The difficulty of purifying integral membrane TCS HKs in combination with their multidomain configurations and structural flexibility has impeded structure-function studies. Thus, all previous structural studies of TCS HKs, and even those of FixL, were performed with truncated, rather than full-length, proteins. We have isolated full-length B. japonicum FixL and FixJ proteins at high purity and obtained structural information on FixL in both the kinase-active (deoxy) and kinase-inactive (oxy) forms, FixJ, and the FixL-FixJ complex by combining size exclusion chromatography–integrated small-angle x-ray scattering (SEC-SAXS), x-ray crystallography, and molecular modeling techniques. This analysis provides insights into how microorganisms and plants adapt to environmental change at the molecular level and elucidates details of a microbial signaling pathway that facilitates the agro-industrial production of soybeans for human food, livestock feed, and biofuel with only limited need for nitrogenous fertilizers.


Autophosphorylation and phosphotransfer activities of full-length FixL and FixJ

We prepared recombinant full-length B. japonicum FixL and FixJ with very high purity (fig. S1, A and B). B. japonicum FixL is a naturally occurring soluble FixL, in contrast to membrane-anchored FixLs, such as that from Sinorhizobium meliloti. B. japonicum FixL comprises N-terminal tandem PAS domains, PAS-A and PAS-B, followed by C-terminal DHp and CA domains (Fig. 1). The PAS-B domain senses O2 through a heme b cofactor (10, 12). The role of the N-terminal PAS-A domain in B. japonicum FixL is not clear, but some biological studies of the water-soluble FixL from Rhizobium etli suggest that the PAS-A domain influences the oxygen affinity of the heme in the PAS-B domain (13).

We tested the phosphotransfer activity of our highly purified full-length FixL in several heme iron oxidation and ligation states. These included oxy (Fe2+-O2), deoxy (Fe2+), ferric cyanide–bound (cyanomet; Fe3+-CN), and ferric ligand–free (met; Fe3+) forms (fig. S2, A and B). Phosphotransfer from FixL to FixJ was suppressed in the oxy and cyanomet forms, whereas the activities of these reactions were fully restored in the deoxy and met forms (Table 1). These results show that our preparation of full-length FixL and FixJ was successful and that signal transduction by the FixL and FixJ system could be controlled by ligand (O2 or CN) binding to the sensor domain of FixL, irrespective of the ferrous or ferric oxidation state of the heme iron. Because of the low affinity of O2 for FixL (Table 1) and relatively fast autoxidation (τ1/2 ~ 15 min) of oxy FixL (12), we used the met and the cyanomet forms of the full-length FixL as analogs for the deoxy and oxy states, respectively, as samples for the SAXS measurements. We also found that CN binding to the heme iron of the FixL sensor domain in the met form suppressed the autophosphorylation activity (fig. S3A) but promoted the phosphatase activity toward phosphorylated FixJ (fig. S3B). On the basis of these experimental data, we expected that the tertiary and quaternary structures of the full-length FixL are equivalent between ferric CN– and ferrous O2–bound forms.

Table 1 Phosphotransfer activities of full-length FixL to FixJ by ATP-NADH coupled assay (42).
View this table:

Molecular architecture of FixL

Very small amounts of aggregated protein can adversely affect SAXS measurements (14). Despite the high monodispersity of full-length B. japonicum FixL, small amounts of aggregation often resulted in poor SAXS data from static measurements (table S1). To overcome this effect, we performed SEC-SAXS, where the data collection was performed in-line with protein purification, both at the RIKEN beamline BL45XU (15) at the SPring-8 synchrotron in Japan and at the SWING beamline (16) at the SOLEIL synchrotron in France. The SECs of FixL in the met form with SAXS parameters plotted together with absorbance at 280 and 398 nm showed that the aggregated species were eliminated before the SAXS measurements (fig. S4A). Using this method, we obtained high-quality SAXS data from full-length FixL in both the met and the cyanomet forms. The plot of log q versus log I(q) obtained by SEC-SAXS at BL45XU (Fig. 2A) and the SEC-SAXS parameters are summarized in Table 2 and table S2. The data measured at SPring-8 and SOLEIL showed a high degree of reproducibility. Radii of gyration (Rg) of 49.7 ± 0.1 Å and 48.4 ± 0.2 Å for met and cyanomet FixL, respectively, were lower than those collected by static SAXS measurements (Table 2 and table S2) and reflect the ability of SEC-SAXS to isolate scattering from the species of interest. Molecular mass estimations from experimental SEC-SAXS data predict protein masses of 140.1 ± 1.9 kDa and 136.3 ± 2.3 kDa for the met and cyanomet forms, respectively, indicating that FixL is homodimeric in solution. These observations are consistent with the structural characteristics of other HKs, in which two helices of the DHp domain from each monomer interact to form a stable four-helix bundle in the homodimer (17, 18). The distance distribution function, P(r), for each FixL state gave information on the maximum dimension (Dmax) and the average electron distribution (Fig. 2B). The Dmax was calculated at 163 ± 4 Å and 158 ± 3 Å for the met and cyanomet forms, respectively. CN binding to the heme group of FixL suppressed the kinase activity, which was reflected in a slight decrease in Rg (48.4 Å versus 49.7 Å; Table 2) and Dmax (163 Å versus 158 Å; Table 2).

Fig. 2 SEC-SAXS profiles of full-length FixL, FixJ, and FixL-FixJ complexes.

(A) Log-log plots of x-ray scattering for the met and cyanomet forms of full-length FixL, full-length FixJ, the FixL-FixJ complex, and the FixL-FixJN complex, q = 4πsinθ/λ, where 2θ is the scattering angle and λ is the wavelength of incident x-rays. I(q) is the measured scattered intensity at a given value of 2θ. (B) Corresponding pair distance distribution functions for the indicated proteins and protein complexes. P(r) is the frequency of r intramolecular distances. (C) Dimensionless Kratky plots, which compare the compactness of a protein, for full-length FixL (met form), the truncated FixLPAS-PAS (met form), full-length FixJ, and the full-length FixL-FixJ complex. qRg = q, as defined above, multiplied by the radius of gyration of the protein. A compact protein has a peak maximum at √3 and 1.2 on the abscissa and ordinate, respectively. Movement of the peak further into the positive quartile of the graph indicates unfolding and conformational flexibility. Data were collected on the BL45XU beamline at the SPring-8 synchrotron. n > 3 independent protein preparations and data collections.

Table 2 Structural parameters for FixL and FixL-FixJ complexes determined by SEC-SAXS experiments at BL45XU in SPring-8.
View this table:

Using the FixL SEC-SAXS data, we constructed an ab initio model of full-length FixL in the met state. FixL exhibits an extended and club-like form (fig. S5, A and B). Correspondingly, dimensionless Kratky plot analysis of full-length FixL describes a protein that does not adopt either a compact or globular conformation (Fig. 2C). For comparison, we also collected SEC-SAXS data for a truncated form of FixL that contains only the PAS-A and PAS-B domains (FixLPAS-PAS; Fig. 1 and fig. S6, A and B). The Kratky plot analysis of FixLPAS-PAS is much different from that of full-length FixL and shows a high degree of globularity (Fig. 2C). Using this information as a guide, we constructed a pseudoatomic model of full-length FixL, incorporating a combination of previously published crystallographic domain structures [Protein Data Bank (PDB) identifiers (IDs): 3MR0 for PAS-A domain, 1DRM (9) for PAS-B domain, and 4GCZ (19) for HK module], homology modeling, a priori structure prediction, and refinement against the SAXS data, and compared this model with the space-filling model (Fig. 3A). The length of the dimer axis in the model is identical to the experimentally determined Dmax value (Table 2), and its calculated SAXS profile (fig. S7, A and B) has a χ value of 2.57 when compared with the experimental data, indicating the internal consistency of our proposed model and that it is a reasonable snapshot of FixL in solution.

Fig. 3 Space-filling and pseudoatomic models of full-length FixL.

SAXS-based models of the met form of full-length FixL showing the overall shape and domain arrangement.

Our model for FixL displays an elongated structure in which the tandem PAS domains (PAS-A-A and PAS-B-B) and the DHp domains homodimerize. There is no interaction between the PAS-A dimer and PAS-B dimer or between either the PAS-A or PAS-B dimer and the CA domain. We note that the PAS-B sensor domain connects with the kinase module only through a coiled-coil linker region. This architecture of full-length FixL is comparable to an “in-line” model proposed for transmembrane and membrane-associated HKs (17, 18, 20).

Our constructed model has an asymmetric DHp domain (Fig. 3A). To test the possibility that FixL homodimers have a symmetric four-helix bundle, we assembled models of FixL based on the HKs VicK [PDB ID: 4I5S (21)] (fig. S8A), DesKC [PDB ID: 3GIE (22)] (fig. S8B), and CckA [PDB ID: 5IDJ (23)] (fig. S8C). This approach did not yield a better fit to the experimental data than that provided by our proposed structure of intact, asymmetric full-length FixL.

Phosphotransfer activities of FixL mutants

In our proposed model, we found that the sensor PAS-B domains are followed by the coiled-coil linker region and then the four-helix bundle, but there is no direct interaction between the PAS-B and the catalytic kinase domains. From this observation, as one of the possible mechanisms for intermolecular signal transduction of the O2 sensor FixL, we could propose that conformational change in the PAS-B domain is propagated to the DHp domain through the coiled-coil linker region (see Discussion). To examine the importance of the coiled-coil linker region in transducing the signal from the sensor domain to the kinase domain to stimulate autophosphorylation, we prepared 12 proteins bearing substitution mutations of the coiled-coil region by site-directed mutagenesis (R254A, T257A, E258A, E258Q, Q261A, T262A, T262S, Q263A, R265A, L266P, Q267A, and L269P) and measured their autophosphorylation activities in both the met and cyanomet forms (fig. S9, A and B). Except for the E258Q mutation, which was tolerated and responded to the CN binding, each mutation significantly reduced the phosphorylation activity in the met form, indicating that the dimerization of the coiled-coil region might be responsible for activation of the kinase activities of the HK module in FixL. In addition, we also noted that the activity of R254A mutant FixL showed a fivefold lower activity in the met form and moderate impairment in the cyanomet form. The Arg254 residues on the coiled-coil helices near the end of the PAS-B domains might form a hydrogen bonding interaction between the helices to couple CN binding to kinase activity (fig. S9).

Crystal and solution structures of full-length FixJ

One of the most intriguing features of TCSs is the intermolecular communication that facilitates the transfer of a phosphoryl group from the kinase domain to the RR. We analyzed full-length FixJ alone with SEC-SAXS and crystallographic techniques. Full-length FixJ was crystallized in space groups C2221 (one chain in the asymmetric unit) and P212121 (five chains, A to E, in the asymmetric unit) (Fig. 4, A and B, and table S3). Crystallographic analyses revealed that FixJ is a two-domain protein, comprising an N-terminal α/β-type REC domain with a phosphorylation site (Asp55) and a C-terminal all α-type effector domain with a helix-turn-helix DNA binding motif connected by a helical linker (Fig. 4A). The N-terminal REC domain and the C-terminal effector domain each exhibits the same fold as those reported for other DNA binding RRs (24). The linker region contained two α helices (α6A and α6B) in the C2221 data. Chains A to D in the P212121 data represented similar overall conformations with one another, whereas the chain E adopted a different conformation in the linker region compared to chains A to D. Helices α6A and α6B observed in the C2221 data were fused to a single helix, α6, in the chains A to D of the P212121 data, whereas the helix α6A was transformed to a tight curve structure in the chain E of the P212121 data, indicating that the helix α6A is conformationally flexible. Structural comparison of FixJ crystal structures with those of other full-length RRs shows that the α6A helix in the linker region is labile and acts as a hinge enabling TCS RRs, including FixJ, to adopt a variety of overall molecular shapes.

Fig. 4 Crystal and SAXS solution structures of full-length FixJ.

(A) Crystal structure of full-length FixJ (the C2221 data) showing the relative positions of the N-terminal REC (pink), linker (cyan), and C-terminal effector (green) domains. The phosphorylation site, Asp55, is indicated as a stick model with carbon and oxygen atoms indicated in yellow and red, respectively. (B) Comparison of the crystal structures of FixJ in space groups C2221 (blue) and P212121 [yellow (chain A as a representative of chains A to D) and magenta (chain E)]. The view is in the same orientation as (A). (C) Space-filling and pseudoatomic models of full-length FixJ. The ribbon model was colored by temperature factors (b factors). Low and high temperatures are represented in colder and warmer colors, respectively. (D) Comparison of FixJ crystallographic and SAXS models against experimental SAXS data. Experimental SAXS data are shown in black; the calculated scattering curve of the SAXS FixJ model in (C) is shown in red; the calculated scattering curves of the crystal structures of FixJ are shown in other colors as indicated. q = 4πsinθ/λ as defined in Fig. 2.

The SEC-SAXS parameters of B. japonicum FixJ (Table 2), that is, Rg = 22.3 ± 0.2 Å, Dmax = 66 ± 2 Å, and molecular mass = 24.5 ± 0.7 kDa, are comparable with those of the S. meliloti ortholog (25) and indicate that FixJ in the nonphosphorylated state is monomeric in solution. An ab initio model constructed from experimental scattering data showed that FixJ in solution exhibits an ellipsoidal shape (Fig. 4C and Table 2), and dimensionless Kratky plot analysis (Fig. 2C) also indicated that FixJ does not rest in an extended conformation in solution. No crystallographic conformer fitted the experimental SAXS data well, indicating that the average conformation of FixJ in solution is not fully represented by the crystal structures. We were able to refine the FixJ structure against the SAXS data to create a pseudoatomic model that fits the experimental data with a χ value of 2.5 (Fig. 4D). In solution, the N-terminal REC domain and the C-terminal effector domain are able to fold back upon one another because of the kinked linking α helix, thus allowing a more compact structure. The estimated Dmax (66 Å) from the SAXS data is smaller than the longitudinal length of ~78 Å for a dumbbell-like shape of the crystal structures. This difference may arise from the conformational flexibility in the linker region, which confers conformational plasticity in solution.

Solution structure of the full-length FixL-FixJ complex

To promote FixL-FixJ complex formation, we performed SEC of FixL in a buffer containing 40 μM FixJ. This concentration is 10-fold greater than the FixL-FixJ dissociation constant (Kd), which has been reported as 0.8 to 4.0 μM in the absence of ATP analogs and Mg2+ (26). Under this condition, the FixL-FixJ complex is expected to predominate over that of homodimeric FixL in solution. When the FixJ concentration in the loading buffer was reduced to 20 μM, the FixL-FixJ complex was not well separated from homodimeric FixL by SEC.

FixJ-saturated FixL elutes earlier than FixL homodimers (fig. S4B). Under these conditions, we collected SEC-SAXS data for the FixL-FixJ complex with the met form of FixL (Fig. 2). Size parameters were found to be 53.1 ± 0.1 Å for Rg, 160 ± 5 Å for Dmax, and 185.4 ± 1.5 kDa for molecular mass (Table 2). The Rg value was larger than that of FixL alone, and the molecular mass was increased by 45.3 kDa, indicative of two molecules of FixJ (24.5 kDa) binding to each FixL homodimer. These parameters indicate that phosphotransfer from FixL to FixJ is facilitated through homodimeric FixL binding to two FixJ molecules to form a heterotetramer. In addition, the SAXS data did not indicate large domain rearrangements upon complex formation, and dimensionless Kratky plot analysis (Fig. 2C) showed that the noncompact FixL conformation is conserved after forming a complex with two molecules of FixJ.

On the basis of this observation, we docked two FixJ monomers onto our FixL pseudoatomic model (Fig. 3B). In this construction, the phosphodonor 3-phospho-His291 in the FixL DHp domain and the phosphoacceptor Asp55 of FixJ are within phosphotransfer distance (~3 Å). The overall fit of this model to the experimental SAXS data is good, with a χ value of 2.71 (fig. S7B), thus validating the proposed model. To our knowledge, this is the first complete model of the complex formed by an intact full-length sensor HK and its full-length cognate RR that mediates TCS transduction.

To compare the kinase-active FixL-FixJ complex with the inactive complex, we also measured the SEC-SAXS of the complex in the cyanomet form, whose parameters are shown in Table 2. Their values, especially Rg and Dmax, were the same as those of the active met FixL-FixJ complex, suggesting that the ligand (CN) binding to the PAS-B sensor domain of FixL, which suppressed the kinase activity, does not strongly affect the overall molecular shape of the FixL-FixJ complex.

In the proposed FixL-FixJ complex, the N-terminal REC domain of FixJ, as determined by our FixJN structure, fits well within the SAXS envelope model, whereas the C-terminal effector domain of FixJ, as determined by our structure of FixJC, does not. We hypothesize that, given the conformational flexibility of monomeric FixJ, the C-terminal domain is also able to adopt multiple conformations while in complex with FixL due to the dynamic nature of the linker region observed in the crystallographic structure of FixJ (Fig. 4B). To test this possibility, we also measured and analyzed the SEC-SAXS of FixL complexed with FixJN. The SAXS parameters of the FixL-FixJN complex showed that the Dmax (170 ± 3 Å) of this complex is similar to that of the FixL-FixJ complex, whereas the molecular mass (153.6 ± 1.4 kDa) and Rg (48.9 ± 0.1 Å) values were smaller, as expected (Table 2). The absence of the FixJ C-terminal domain was reflected in the changes in mass and Rg but did not affect the maximum dimension of the FixL-FixJ complex. In addition, ab initio models of the FixL-FixJ and FixL-FixJN complexes were very similar (fig. S5B). These results suggest that the C-terminal domain of FixJ does not contribute to the FixL-FixJ interaction and is instead free to move in solution.


Intra- and intermolecular signal transduction through TCSs occur in all domains of life except metazoans. These systems facilitate survival by allowing rapid adaptation to environmental changes. To understand the molecular mechanism of TCSs in detail, structural information about TCS HKs, RRs, and the complexes they form have been generated. However, many structural studies on TCSs have relied on breaking the protagonists down into biochemically tractable domains. Here, we have investigated the structure of a complete TCS using the full-length forms of components of the B. japonicum O2-sensing system: FixL, FixJ, and the FixL-FixJ complex. The information gained for the kinase-active and kinase-inactive forms of full-length FixL and of full-length FixL in complex with FixJ is particularly valuable because it enabled us to investigate the modular structures that facilitate signal relay in this agriculturally important TCS. To validate our model of full-length FixL, which used the asymmetric four-helix bundle structure of the FixL DHp domain conjoined to a light sensor domain as a template (19), we created homology models of FixL based on other HK DHp domains. These structures exhibited consistently worse fit to the experimental SAXS data than did the FixL model we proposed.

With respect to the intramolecular signal transduction from the sensor domain to the CA domain of FixL, two possible mechanisms have been proposed so far. Sousa and co-workers proposed a “globular” model based on biochemical studies of the full-length, cytoplasmic R. etli FixL protein (13), which has a similar domain organization as B. japonicum FixL. In the R. etli FixL, interactions between this protein’s cytoplasmic PAS domains (PAS-A and PAS-B) in homodimers or between the PAS homodimers and the CA domain are plausible in the globular form. Therefore, it was proposed that changes to these interactions would be involved in the intramolecular signal transduction stimulated by the association of O2 with or dissociation from the sensor PAS domain. On the other hand, the crystal structure of the transmembrane Thermotoga maritima ThkA, the single PAS domain of which exhibits 24% primary structure similarity to that of FixL PAS-B, displays no interaction between the PAS domains in the homodimeric form, but rather direct interaction of the PAS domains with the CA domains through hydrogen bonding (27). These models, proposing interactions between the PAS domain and the CA domain, were based on the assumption that the mechanism of the intramolecular signal transduction is not necessarily similar between the water-soluble and membrane-integrated TCS HKs. The former HK functions as a sensor of a stimulus in the cytoplasm, whereas the sensor domain of the later HK detects an extracellular stimulus and transfers information into the cytoplasm across the membrane for cellular adaptation. However, the proposed architectures of R. etli FixL and ThkA are incongruent with our present SAXS data of full-length B. japonicum FixL, deduced parameters, and low-resolution models. These structural inconsistencies suggest that neither the globular model nor the direct transfer of the signal from the sensor to the CA domains is possible.

Our results lead us to propose an in-line model for full-length FixL with dimerization through PAS-A–PAS-A, PAS-B–PAS-B, and DHp-DHp interactions, but no direct interaction of the heme-containing PAS-B domain with the CA domain (Fig. 3A). Our model of intact full-length FixL provides us with a structural basis from which to discuss the molecular activation mechanism of TCS HK modules by autophosphorylation in response to stimuli. The full-length FixL met (active) and cyanomet (inactive) forms (Table 2) show that the overall molecular conformation is not largely changed upon activation. Correspondingly, our SAXS envelope models are also unaltered by CN binding (fig. S5A). Thus, it is likely that the conformational changes that control kinase activity after ligand binding are localized, causing subtle changes in the interaction between the sensor domain and the HK domain. We propose that, upon O2 dissociation from the heme iron in the PAS-B domain, a local conformational change in this domain, previously observed in high-resolution structures and spectroscopic characterization of FixL sensor domains (2832), is propagated to the DHp domain through the coiled-coil linker region. These structural changes cause ATP bound to the CA domain and adjust the position of the His291 residue of DHp, resulting in phosphotransfer (Fig. 6).

Our idea was supported by the results of site-directed mutagenesis experiments, in which mutations in the coiled-coil linker region inhibited the phosphotransfer activity of FixL (fig. S9). The results are consistent with our previous work on chimeric sensor proteins, in which the FixL PAS-B sensor domain was fused with the HK domain of T. maritima ThkA at the coiled-coil region (33). The kinase activities of the chimeric proteins were impaired, maintained, or enhanced by ligand binding, depending on where in the coiled-coil region the two functional domains were fused. All these results support our proposal that the coiled-coil linker region between the PAS-B and DHp domains functions as a key modulator of the autophosphorylation activity of FixL.

This proposal, the so-called in-line mechanism, is comparable to those given from previous studies of membrane-integrated or membrane-associated TCS HKs such as VicK and DesKC. The structures of truncated forms of these HKs, in which the transmembrane region or the extracellular domain or both were deleted (18, 2023, 34), were the basis of this proposal. However, there was no structural comparison between the kinase-active and kinase-inactive forms of either. The only full-length structure of an HK in the literature is an engineered, chimeric HK called YF1, which is active in the dark and inactive by stimulation with blue light (19). YF1 was constructed by fusing the FixL HK module (DHp + CA domains) with the light-sensitive light-oxygen-voltage (LOV) domain of Bacillus subtilis YtvA (19). On the basis of the structures of YF1 and the active and inactive forms of the isolated LOV domain, it was suggested that the coiled-coil linker region between the sensor and the DHp domain promotes the symmetry and asymmetry transitions of the four-helix bundle of the DHp domain to adjust the orientation between the DHp and the CA domains. The effects of site-directed mutagenesis of the coiled-coil linker region on the kinase activity of YF1 (19) are comparable to our observations of the effects of similar mutations on the kinase activity of FixL (fig. S9). Therefore, our present study using the intact and full-length FixL establishes the intramolecular signal transducing mechanism of HK.

Autophosporylation of His291 in the FixL DHp domain is followed by the phosphotransfer reaction, in which the phosphoryl group is transferred to Asp55 in the N terminus of FixJ. Formation of a transient FixL-FixJ complex facilitates phosphotransfer. To gain insight into this transient state in solution, we collected SEC-SAXS data under FixJ-saturated conditions and observed SAXS parameters indicating a stoichiometry of 2:2 for the FixL and FixJ complex. This result is consistent with structural data for complexes of truncated HK and truncated RRs: HK853-RR468 (20), ThkA-TrrA (27), DesK-DesR (22), and Spo0B-Spo0F (34). Stranava and co-workers (35) recently reported that AfGcHK and its cognate RR predominantly form a complex with 2:1 stoichiometry, but we did not obtain a 2:1 complex under the condition of excess FixJ in our SEC-SAXS study.

Our SAXS data indicate that FixJ binding induces no large FixL conformational change (Fig. 4, A and B). This observation is consistent with previous crystallographic studies of the truncated HK and RR complexes mentioned above (20, 22, 34), in which a slight rotation of the CA domain was observed, but there were no major changes in the overall structure of the HK (Fig. 6). In addition, the volume of the N-terminal FixJ REC domain is located close to the DHp-CA regions of the space-filling model (Fig. 3B). Because a phosphotransfer reaction from the HK to its cognate RR is common in TCSs, specific recognition of the REC domain by the kinase domain, such as that observed in the full-length FixL-FixJ complex, is almost certainly a general feature of TCSs. After phosphotransfer to the Asp residue in the REC domain of the RR, which activates the RR and releases it from the HK, the C-terminal effector domain of the RR dimerizes and then interacts with its specific target. Phosphorylated FixJ promotes the expression of NifA and FixK by binding to elements upstream of the promoters of these genes, which encode proteins important for nitrogen fixation. Throughout TCSs, the function of the RR is highly variable and may mediate DNA binding (69.4%), RNA binding (1%), protein binding (1.7%), or enzymatic activity (8.1%) (36). Such functional diversity is reflected in the structural diversity of the C-terminal effector domains of RRs. There are at least 35 classes of RR, with each class including anywhere from 1 protein up to roughly 25,000 proteins (3638). It appears to us that a conformationally flexible RR, as described here for FixJ, very effectively limits the role that it can play in forming complexes with the cognate HK (Figs. 5 and 6). This is necessary to ensure that structurally disparate effector domains can be activated by a general TCS phosphotransfer mechanism to the REC domain of the RR. Thus, modularity ensures functionality.

Fig. 5 Space-filling and pseudoatomic models of the FixL-FixJ complex.

SAXS-based models of the FixL-FixJ complex showing the overall shape, domain arrangement, and mode of complex formation. The C-terminal effector domain of FixJ is not present in the model of the FixL-FixJ complex because it does not form part of the complex interface and is allowed conformational freedom by the linker connecting the N-terminal REC domain to the effector domain.

Fig. 6 Schematic representation of the FixL-FixJ TCS.

Our SAXS results suggest that there are no large changes of the overall shape of full-length FixL upon O2 dissociation from the heme group. However, the orientation of the coiled-coil helices between the heme-containing PAS-B (pink) and DHp (green) domains may change. Such localized structural change could alter the distance between the ATP-binding site in the CA domain (orange) and the autophosphorylation site at His291 in the DHp domain. In the full-length FixL-FixJ complex, a phosphorylation site (Asp55) in the FixJ REC domain approaches His291 of the FixL DHp domain, which mediates phosphotransfer. The C-terminal DNA binding domain of FixJ is connected to the REC domain by a flexible linker, allowing the effector domain to exhibit multiple conformations.

The O2 sensor FixL is categorized as a class I HK, in which the DHp domain is directly adjacent to the CA domain (34, 35). Although all HKs discussed here (VicK, DesK, CckA, EnvZ, HK853, and ThkA) are in this same category, the domain architecture of FixL is the simplest among them. On the other hand, FixJ, which consists of a REC domain and a helix-turn-helix DNA binding domain (effector domain) through which it acts as a transcriptional activator, belongs to the NarL-like superfamily (39), which is the largest family of RRs. For the class I HKs, the REC domain is a common component for receiving a phosphate group. Therefore, the full-length architecture of the FixL-FixJ TCS is expected to represent the general mechanism of intra- and intermolecular signal transduction in TCSs composed of class I HKs. It is possible for the in-line mechanism supported by our findings to operate in all class I HKs, irrespective of whether the HK sensor is water-soluble (cytosolic) or membrane-integrated. The in-line mechanism also does not require that the effector domain of the RR interact with the HK. In addition, the present study provides fundamental knowledge for understanding the physiological, biochemical, and biological importance of TCS, particularly those involving class I HKs, including their molecular evolution (40) or protein engineering for synthetic biological systems for the development of antibiotics and plant growth effectors.

The rhizobia are a group of nitrogen-fixing bacteria that are proficient at establishing symbiotic relationships with leguminous plants, a global staple food group. This relationship provides nitrates to the host and a boost to growth and plant survival with benefits to agricultural productivity. The model organism B. japonicum is of particular interest in this regard because it resides in the root nodules of the soybean plant Glycine max, which provides more protein per hectare cultivated than any other food source. B. japonicum is sprayed onto soybean seed stock on an industrial scale to take advantage of this relationship. Our results open the path for genetic modification of this rhizobial TCS to improve crop yields.


Preparation of recombinant FixL and FixJ proteins

B. japonicum FixL and FixJ genes encoding full-length FixL and FixJ and truncated proteins [FixLPAS-PAS (residues 1 to 275) and FixJN (residues 1 to 124)] were separately amplified by polymerase chain reaction (PCR) with PfuTurbo DNA polymerase (Agilent Technologies). The PCR fragments were cleaved by Bsr GI and Avr II for FixL, Bsr GI and Nhe I for FixJ (New England Biolabs), and cloned into 5′Bsr GI–3′Avr II sites of pET-47b(+) vector (Novagen) for expression with hexa-His tag, followed by the HRV3C protease cleavage site in the N terminus of those proteins. Site-directed mutagenesis of the coiled-coil region in full-length FixL was performed by QuikChange protocol (41) using pET-47b(+) vector inserted in the full-length FixL gene as a template.

Escherichia coli BL21(DE3) cells (Nippon Gene) carrying these plasmids were inoculated in terrific broth (TB) containing kanamycin (50 μg/ml; Wako) and 1% glucose for 4 hours at 37°C with shaking at 150 rpm. One milliliter of the preculture solution was inoculated into 300 ml of TB medium containing kanamycin [50 μg/ml; plus 250 μM 5-aminolevulinic acid (Cosmo Energy Holdings) only for the expression of FixL]. The cultivation was done at 37°C with shaking at 120 rpm. After 4 hours of cultivation, expression of FixL or FixJ was induced with 0.3 or 0.2 mM isopropyl-β-d-thiogalactopyranoside, respectively, and cultivation was allowed to continue for another 15 hours at 23°C with shaking at 80 rpm. The cells were harvested by centrifugation at 4000g for 10 min, and the cells were washed in 30 mM tris-HCl (pH 8.0) twice.

Purification of FixL and FixJ was performed by the following same steps at 4°C. Harvested cells were resuspended in a lysis buffer [50 mM tris-HCl (pH 8.0), 300 mM NaCl, 10% (w/v) glycerol, and one tablet of cOmplete EDTA-free protease inhibitor cocktail (Roche)]. The lysate was mixed with lysozyme (0.1 mg/ml; Sigma-Aldrich), deoxyribonuclease I (0.05 mg/ml; Sigma-Aldrich), and 5 mM MgCl2 for 30 min and disrupted by Microfluidizer M-110Y (Microfluidics). Cell debris was removed by ultracentrifugation at 40 krpm for 1 hour. The supernatant was loaded onto a HisTrap HP (GE Healthcare) column equilibrated with buffer A [50 mM tris-HCl (pH 8.0), 300 mM NaCl, 10% (w/v) glycerol, 10 mM imidazole/HCl (pH 8.0)]. FixL or FixJ was eluted by an imidazole concentration gradient (0 to 300 mM). The eluted fractions were treated with N-terminal 6×His-tagged HRV3C protease (produced in-house) to remove the 6×His-tag from the recombinant FixL and FixJ proteins, and the solution was dialyzed with buffer B [40 mM tris-HCl (pH 8.0), 150 mM NaCl, 10% (w/v) glycerol]. After the dialysis, the protein solution was loaded to HisTrap FF (GE healthcare) equilibrated with buffer B to remove the HRV3C protease and remaining His-tagged proteins, and the flowthrough was collected. The collected solution was concentrated by Amicon Ultra-15 (Merck Millipore) and centrifuged at 15 krpm for 20 min. The supernatant was loaded to HiLoad 16/600 Superdex 200 (GE healthcare) equilibrated with buffer B. The purity of FixL or FixJ was checked by SDS–polyacrylamide gel electrophoresis (PAGE). The purified samples were mixed with 2× SDS-PAGE buffer containing 125 μM tris-HCl (pH 6.8), 4% SDS, 20% (w/v) sucrose, 0.01% (w/v) bromophenol blue (BPB), and 10% (v/v) 2-mercaptoethanol and boiled at 95°C for 10 min before the electrophoresis. NuPAGE Bis-Tris gels (10%; Thermo Fisher Scientific) were used for the electrophoresis with NuPAGE Mops SDS running buffer (Thermo Fisher Scientific) for FixL and NuPAGE MES SDS running buffer (Thermo Fisher Scientific) for FixJ. The gels were stained by EzStain AQua (ATTO). In FixL, the highly purified fractions with an Rz (A398nm/A280nm) value of >1.3 were used for SAXS studies. These spectra were measured in 40 mM tris-HCl (pH 8.0), 150 mM NaCl, and 10% (w/v) glycerol at 20°C by NanoDrop 2000c spectrophotometers (Thermo Fisher Scientific). Because His-tagged full-length FixL showed the kinase inhibition upon cyanide binding to the heme and nearly the same phosphotransfer activity as the His-tag–removed protein, we used the His-tagged full-length FixL proteins in the phosphorylation activity assays of the coiled-coil mutants (fig. S9).

Autophosphorylation, phosphotransfer, and phosphatase activity assay of purified FixL and FixJ proteins

Autophosphorylation of the FixL protein was monitored by radioactivity of phosphorylated FixL by 32P. Reaction mixtures contained 2 or 6 μg of FixL in 7.5 μl of 50 mM tris-HCl (pH 8.0), 50 mM KCl, 1 mM MgCl2, 50 μM MnCl2, and 5% (w/v) glycerol with or without 5 mM KCN. The reactions were started with the addition of the mixture of ATP (10 or 25 nM) and γ-32P-ATP (4 or 10 μCi) (PerkinElmer), incubated the reaction mixture at 23°C for 10 min, and stopped with one-fourth volume of SDS-PAGE sample buffer. After SDS-PAGE, phosphorylated proteins were visualized with an imaging analyzer BAS-1800 on an imaging plate (Fujifilm).

Phosphotransfer activities from FixL to FixJ were determined using an enzymatic assay in which ATP hydrolysis is coupled to NADH (reduced form of NAD+) oxidation using lactate dehydrogenase and pyruvate kinase (42). To measure basal activity, the reaction buffer [50 mM tris-HCl (pH 8.0), 50 mM KCl, 3 mM phosphoenolpyruvic acid (Wako), 0.3 mM NADH (Sigma-Aldrich), 8 U of lactate dehydrogenase (Toyobo), 25 U of pyruvate kinase (Sigma-Aldrich), 5 mM MgCl2, and 5 mM ATP] was equilibrated at 20°C for 5 min. The reaction was started by the addition of 1 μM purified FixL and 10 μM purified FixJ, and the time course of A340nm was monitored for 10 min at 20°C. An NADH standard curve with a range of NADH concentrations between 0 and 150 μM was measured in the same buffer. To measure the activity of the cyanomet form, we added 5 mM KCN to the reaction buffer before the addition of FixL protein.

Phosphatase activity of FixL was monitored the dephosphorylation of acetylphosphorylated FixJ by unphosphorylated FixL. The acetylphosphate-dependent FixJ autophosphorylation was performed in 50 mM tris-HCl (pH 8.0), 50 mM KCl, 1 mM MgCl2, 50 μM MnCl2, 20 μM FixJ, and 50 mM acetyl phosphate lithium potassium salt (Sigma-Aldrich) at 23°C. After 2 hours, unphosphorylated FixL (10 μM, wild type) was added to the reaction mixture. The incubation times for the reactions were 0 (before FixL addition), 5, 15, 30, 60, and 120 min after the addition of the FixL. The reactions were stopped with two-third volume of SDS-PAGE sample buffer. The reaction products were subjected to 15% Zn2+-Phos-tag SDS-PAGE containing 50 μM Phos-tag acrylamide (Wako) and 100 μM ZnCl2. Tris-glycine buffer was used for the electrophoresis. The gel was stained by EzStain AQua (ATTO).

SEC-SAXS data collection and analysis

SEC-SAXS data collection was performed at RIKEN beamline BL45XU (15) in SPring-8 and at beamline SWING (16) in the French national synchrotron SOLEIL. At BL45XU in SPring-8, the purified protein was loaded onto a Superdex 200 Increase 3.2/300 column (GE Healthcare) in a 20-μl volume at 50 μl/min flow rate with a SEC buffer [40 mM tris-HCl (pH 8.0), 10% (w/v) glycerol, and 5 mM MgCl2] at 20°C. For the data collection of cyanomet FixL, 5 mM KCN was added to the SEC buffer, and the loading FixL sample was mixed with KCN at the final concentration of 5 mM. For the detection of the FixL-FixJ complex or the FixL-FixJN complex, SEC buffer containing saturating amounts of FixJ or FixJN (more than 40 μM, 10 times of the Kd value) was used to improve the affinity of FixL and FixJ. X-ray exposure was 1 s every 4 s with the incident beam energy 12.4 keV. Buffer frames (30 × 1 s) were averaged and subtracted from 30 × 1 s frames taken over the course of protein elution. The sample detector distance was 2 m, giving an angular momentum transfer range of qmin = 0.009 Å−1 to qmax = 0.5 Å−1. The flux density was about 2 × 1012 photons/s/mm. Scattering was collected on a PILATUS3 X 2M detector (Dectris). Data averaging and reduction were calculated by the program DataProcess installed in BL45XU. Measurements at SWING were performed as above but using an Agilent Bio SEC-3 4.6 × 300–mm column (Agilent) with 3-μm bead size and 300 Å pore size at 15°C after double purification on a Superdex 200 Increase 10/300 column (GE Healthcare) at 4°C. Beam energy was 12 keV, and sample detector distance was 1.8 m.

Rg and I0 calculations on the SEC-SAXS data were performed with AutoRg (43). Data of interest was averaged, and the Guinier estimation was performed in MATLAB and PRIMUS (fig. S10) (44). Distance distribution functions P(r) were calculated with GNOM (45) and ScÅtter based on agreement between real and reciprocal space Rg values (<3% difference) and fit to the experimental data. This was performed independent of crystallographic and homology model building. Bead models were generated with DAMMIN (46).

Crystallization, data collection, and refinement for the full-length FixJ structure

Crystals of FixJ in space group C2221 (form 1) were obtained at 20°C by vapor diffusion using a mother liquor containing 10% (w/v) PEG 8000, 10% (w/v) PEG 1000, 0.8 M sodium formate, 20% (v/v) glycerol, and 0.1 M tris-HCl (pH 7.5). The asymmetric unit contained one polypeptide chain. Crystals of FixJ in space group P212121 (form 2) were obtained at 20°C by vapor diffusion using a mother liquor containing 20% (w/v) pentaerythritol ethoxylate (15/4 EO/OH), 0.1 M magnesium formate, 20% (v/v) glycerol, and 0.1 M tris-HCl (pH 8.5). The asymmetric unit contained five polypeptide chains. Crystals of both forms were grown in 3 days, frozen, and stored in liquid nitrogen. Data collection was carried out at BL26B2 in SPring-8, Harima, Japan (47, 48), equipped with an automated sample mounting system (49). Crystals were cryocooled in a nitrogen gas stream at 100 K during data collection. The data were integrated and scaled using HKL2000 (50). Initial phases for the data set of the crystal form 1 were obtained by molecular replacement using coordinates of StyR (PDB ID: 1YIO) (51) as a search model in PHENIX (52). The coordinates of the N- and C-terminal domains of StyR were extracted and used for the molecular replacement calculation because the structure of the linker region may vary from protein to protein in the RR family. Initial phases were used for model building and improved by refinement of the model coordinates, including model building of the linker domain, in PHENIX (52) and COOT (53). The final model included residues from 2 to 203, and 4 glycerols, 6 formic acids, and 257 water molecules. Initial phases for the data set of the crystal form 2 were obtained by molecular replacement using coordinates of the form 1 structure in PHENIX, with the N- and C-terminal domains separated. Four solutions for each domain were obtained and used for model rebuilding and refinement including the linker regions (chains A to D) in PHENIX and COOT. During refinement, weak but continuous densities appeared in the solvent region. Lowering the contour level revealed the fifth molecule for the densities. Each separated copy of the models of the three domains was manually fitted in the electron densities (chain E). Coordinates of the five polypeptide chains were further refined including magnesium ions and water molecules. The final model included 1015 residues, 4 magnesium ions, and 8 glycerols, 7 formic acids, and 163 water molecules. Model quality was checked by MolProbity (54) in PHENIX.

Pseudoatomic model building and refinement against SAXS data

JPred (55) and Coils (56) were used to ascertain which parts of the FixL sequence form α helices and coiled coils. The structure of the blue light receptor (PDB ID: 4GCZ) was used as a model for the HK domain and the α helices beyond Thr257. This was linked to the structure of FixL heme-PAS domain (PDB ID: 1DRM). The coiled-coil helix N terminus of the heme-PAS to the PAS-A domain was created with PEP-FOLD (57). This was linked to a homology model of the PAS-A domain created from a sensory HK from Burkholderia thailandensis (PDB ID: 3MR0). The cloning fragment from the pET-47b(+) vector (10 amino acid residues, GPGYQDPNSV) was also constructed using PEP-FOLD. This initial structure had χ2 value of 3.93 against experimental data calculated using FoXS (58). Torsion angle molecular dynamics (MD) in CNS (59) was used to refine the positions of domains and loops in this structure against SAXS data as described by Wright et al. (60). To validate the region of the model encompassing amino acids 1 to 256, which was assembled by combining a crystal structure, a homology model, and several ab initio structure predictions, we measured SAXS of a C-terminally truncated form of FixL consisting of amino acids 1 to 275 and including the cloning fragment. These data were exceptionally congruent with the PAS-A–PAS-B part of our FixL model, indicating the proposed domain arrangement and interfaces are correct. The chimeric sensor protein structure 4GCZ shows structural asymmetry possibly resulting from crystal packing or partial adenosine diphosphate (ADP) binding. We created three models of FixL based on VicK, DesKC, and CckA HKs, each of which has symmetric four helix bundles, based on their coverage of and identity to FixL amino acids 258 to 505. Homology models were generated with SWISS-MODEL in conjunction with the method above and HADDOCK (61) to refine and dock the ADP-free CA domains to find an optimum ADP-free structure. This pool of structures has consistently worse fit to our experimental data than that based on 4GCZ, indicating that FixL adopts an asymmetric conformation in solution in the ADP-free state.

The structures of FixJ were refined against SAXS data using the torsion angle MD process described above. We initially defined the location of FixJ binding to FixL using pyDockSAXS (62) and FoXSDock (63). These programs use a combination of FTDock/Crysol and PatchDock/FoXS to assign interactions. Using this approach, and without defining the FixL His291–FixJ Asp55 interaction site, FixJ was consistently positioned on the FixL four-helix bundle. FixJ REC domains were then directed to FixL His291 on both chains using HADDOCK. The FixJ linkers and DNA binding domains were then added based on the SAXS refined structure of the FixJ monomer. The positions of these latter domains were then determined using CNS torsion angle MD. Trimeric complexes of dimeric FixL with one FixJ monomer were also constructed and optimized by torsion angle MD/rigid body refinement but were found to consistently produce models that fit the data poorly in comparison with the tetrameric architecture.


Fig. S1. Purification of FixL and FixJ.

Fig. S2. Optical absorption spectra of full-length FixL.

Fig. S3. Autophosphorylation and phosphatase activities of FixL.

Fig. S4. SEC profiles.

Fig. S5. Space-filling models of FixL alone and in complex with FixJ or FixJN.

Fig. S6. Pseudoatomic model and SAXS curve of truncated FixL comprising PAS-A, PAS-B, and the coiled-coil region (FixLPAS-PAS).

Fig. S7. Experimental and simulated SAXS curves of met FixL and the FixL-FixJ complex.

Fig. S8. SAXS curves and pseudoatomic models of full-length FixL based on comparisons with other HKs.

Fig. S9. Phosphorylation activities of FixL mutants bearing mutations in the coiled-coil linker.

Fig. S10. Guinier plots with Pearson residuals for full-length FixL, FixLPAS-PAS, FixL-FixJ, full-length FixJ, and FixJN.

Table S1. Structural parameters for FixL and FixJ determined by static SAXS experiment at the BL45XU beamline at the SPring-8 synchrotron.

Table S2. Structural parameters for FixL and FixJ determined by SEC-SAXS at the SWING beamline at the SOLEIL synchrotron.

Table S3. Crystallographic statistics for full-length FixJ.


Acknowledgments: Thanks to synchrotron SOLEIL and SPring-8 for the provision of SAXS facilities. We acknowledge the support and the use of resource of instruct, a landmark ESFRI project (iNEXT 2822). Funding: This work was supported by the Fumi Yamamura Memorial Foundation for Female Natural Scientists from Chuo Mitsui Trust and Banking (to H.S.), Hyogo Science and Technology Association (to H.S.), RIKEN Pioneering Project “Integrated Lipidology” (to H.S.), and “Molecular System” (to Y.S.), and the Japan Society for the Promotion of Science (JSPS) KAKENHI grant numbers JP26220807 (to Y.S. and H.S.) and JP25871213 (to H.S.). Author contributions: G.S.A.W., S.V.A., S.S.H., Y.S., and H.S. designed this study. A.S., H. Nakamura, and H.S. created the systems for expressing recombinant FixL and FixJ in E. coli. T. Hikima and M.Y. installed the SEC-SAXS system at BL45XU in SPring-8. G.S.A.W., A.S., Y.N., and H.S. purified the FixL and FixJ samples. H.S. prepared the expression systems for FixL mutants. M.K. purified the FixL mutant proteins and measured their phosphotransfer activities. K.N. and H. Nishitani performed the autophosphorylation activity assay using γ-32P-ATP. H.S. performed the phosphatase activity assay by Phos-tag SDS-PAGE. G.S.A.W., A.S., T. Hikima, and H.S. measured and analyzed the SAXS data. G.S.A.W. modeled the pseudoatomic structures. Y.N. crystallized FixJ. Y.N. and T. Hisano collected, processed, and refined the crystal data. G.S.A.W., T. Hisano, S.V.A., S.S.H., Y.S., and H.S. wrote the manuscript. All authors analyzed data and discussed the results. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The atomic coordinates and structure factors for FixJ (PDB IDs: 5XSO and 5XT2) have been deposited in the PDB ( The SAXS measurements at SPring-8 BL45XU were performed under proposals 20140099, 20150017, 20160015, and 20170092. The x-ray diffraction measurements were performed at SPring-8 BL26B2 (proposal 20160015).
View Abstract

Navigate This Article