Metal-induced sensor mobilization turns on affinity to activate regulator for metal detoxification in live bacteria

SMT Identifies a Mobile and a Clustered Stationary State of CusS in the Membrane.

To visualize CusS in live E. coli cells, we fused the photoconvertible fluorescent protein mEos3.2 (i.e., mE) (30) to its C terminus at its chromosomal locus, creating CusS^mE at a physiological expression level. Using cells that also express CusS^mE from a plasmid to increase cellular protein level, Western blot shows that CusS^mE is largely intact (<11% cleavage of the mE tag; SI Appendix, Fig. S6), and confocal microscopy of these cells further confirms CusS^mE’s membrane localization, where some cells show protein aggregation, likely due to protein overexpression (Fig. 1A and SI Appendix, Fig. S3) (SI Appendix, section 1.6). Cell growth assays [using a ΔcueO base strain to increase cells’ sensitivity to copper stress (9⇓–11)] further show that CusS^mE is as functional as the untagged CusS in conferring cell resistance to copper (SI Appendix, Fig. S5B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

SMT resolves two states of CusS in the membrane. (A) Confocal fluorescence images of cells overexpressing CusS^mE. (B) ITCDD-corrected distribution of displacement length r of CusS^mE from 17,029 molecules and 1,282 cells grown in copper-depleted medium. Lines: overall fit (black) and the resolved quasistationary (red) and mobile (green) diffusion states. The y axis is not the actual counts of displacement lengths; it only represents the relative probabilities of different displacement lengths due to ITCDD correction. (Inset) Typical tracking trajectories in a cell overlaid on its bright-field transmission image. Dashed yellow lines: cell boundaries. (C) Pairwise distance distribution (PWDD) of first locations of CusS^mE tracks. Black, all CusS^mE molecules; red, quasistationary molecules; green, mobile molecules; blue, simulated randomly positioned molecules on cell membrane; magenta, simulated randomly positioned dimers on cell membrane (SI Appendix, section 5). (D) Schematic of copper stress-induced mobilization of CusS in the membrane. (Top) Quasistationary (clustered) and mobile (free) CusS dimers at a dynamic equilibrium. (Bottom) CusS’s periplasmic sensor domains dimerize and CusS protein concentration ([P]) increases upon copper stress; higher [P] and copper binding both shift the equilibrium toward the mobile state. TM, transmembrane domain.

Using sparse photoconversion and time-lapse stroboscopic fluorescence imaging (4-ms laser exposure; 60-ms time lapse) (31, 32), we tracked the motions of single CusS^mE proteins in live cells with tens of nanometer precision until their mE tags photobleached (Fig. 1B, Inset, and SI Appendix, section 1.4). Cells are viable under our imaging conditions (32). This SMT gives the probability distribution function (PDF) of the displacement length r per time lapse, which, although containing rich information about the diffusive behaviors of CusS^mE, is distorted because of the confinement effect from a 2D projection of diffusion on the curved cell membrane surface. We therefore performed ITCDD (inverse transformation of confined displacement distribution) (33⇓–35) to deconvolute this confinement effect (Fig. 1B and SI Appendix, section 4.1).

For cells grown in copper-depleted medium (nutrient-supplemented M9, 0.011 ± 0.002 µM residual Cu; SI Appendix, Table S4), the ITCDD-corrected PDF(r) is sufficiently fitted by two diffusion states (Fig. 1B and SI Appendix, section 4.1): a mobile, faster state with diffusion constant (D_m) of 0.93 ± 0.06 µm²⋅s⁻¹ and fractional population (A_m) of 36 ± 2%; and a much slower state with diffusion constant (D_s) of 0.054 ± 0.003 µm²⋅s⁻¹ and fractional population (A_s) of 64 ± 2%. This slower state is quasistationary because D_s is only slightly larger than the apparent diffusion constant (∼0.01 μm²⋅s⁻¹) for a stationary molecule considering our ∼40-nm localization uncertainties (SI Appendix, section 4.3).

We assigned the mobile state to CusS^mE dimers freely diffusing in the inner membrane (Fig. 1D, Top Right) because, first, free CusS dimers are expected to exist (CusS is intrinsically dimeric ref. 12), and second, the diffusion constant D_m is comparable to those of mobile membrane proteins (35⇓–37). We attributed the quasistationary state to CusS^mE dimers clustered together (either directly or in association with other proteins) (Fig. 1D, Top Left), considering that membrane proteins can form clusters with reduced mobility (27, 28). To support the assignment of this clustered state, we examined the distribution of pairwise distances between the first locations of each tracked CusS^mE (black curve, Fig. 1C). A small peak is clearly visible at a short distance of ∼70 nm, supporting the presence of CusS^mE protein clusters, whereas simulations of proteins distributed randomly on the membrane do not show such a peak (blue curve, Fig. 1C and SI Appendix, section 5). We further separated these first locations into two categories: those associated with displacements smaller than a threshold displacement r₀ (∼166 nm, SI Appendix, section 5), which would be dominated by CusS^mE in the quasistationary state (Fig. 1B and SI Appendix, Fig. S7A), and those associated with displacements larger than this threshold, which would be dominated by CusS^mE in the mobile free dimer state. Strikingly, the pairwise distance distribution (PWDD) from locations dominantly in the mobile state does not show any discernible peak in the <100-nm distance range (green, Fig. 1C), whereas that for the quasistationary state has a pronounced peak at ∼70 nm (red, Fig. 1C). The PWDD of simulated randomly positioned dimers (where the intermonomer distance was artificially set to 70 nm; SI Appendix, section 5) shows a peak at ∼70 nm (magenta, Fig. 1C), whose amplitude is much smaller than that for the quasistationary molecules in the experiment (red). Therefore, this pronounced peak (red) cannot merely come from the dimeric nature of CusS, further supporting the assignment of the quasistationary state to CusS^mE clusters.

Copper Stress Leads to CusS Concentration Increase and Mobilization in the Membrane.

We performed SCQPC (32) to determine the total CusS^mE protein concentration in each cell ([CusS^mE], in terms of monomers; SI Appendix, section 1.4). In copper-depleted medium, the physiological [CusS^mE] is heterogeneous and ranges from ∼47 to 476 nM, averaging at ∼167 ± 80 nM (Fig. 2A). We then sorted cells into groups, with each group having a similar [CusS^mE], to deconvolute protein concentration heterogeneity in analyzing the displacement distributions. Regardless of [CusS^mE], the same two diffusion states are resolved in each group: the mobile state and the quasistationary state with diffusion constants of 0.97 ± 0.13 µm²⋅s⁻¹ (D_m) and 0.05 ± 0.01 µm²⋅s⁻¹ (D_s), respectively, when globally fitted (SI Appendix, section 4.4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

(A) Distribution of cellular CusS^mE concentration [CusS^mE] (all in terms of monomeric CusS) in the absence or presence of 2 mM Cu²⁺ stress. (B–D) Fractional populations (A’s) of mobile (green) and quasistationary (red) states of CusS^mE in the absence (solid symbols) or presence (open symbols) of 2 mM Cu²⁺ stress vs. [CusS^mE] in WT (B), CusSH42AmE (C), and CusSM133I/M135ImE strains (D). The x error bars are SDs of protein concentration; y error bars are SDs of A from bootstrapping (SI Appendix, section 4.4). (E) Interconversion rate constants between the stationary and mobile states of CusS^mE. k_d: declustering rates from the stationary state, blue; k_c: clustering rates from the mobile state, red. Solid, 0 mM Cu²⁺; dashed, 2 mM Cu²⁺ stress. For calculations on k_d and k_c, see SI Appendix, section 6.

In the presence of steady 2 mM Cu²⁺ stress in the medium, a level that was shown to induce significant CusS protein expression in the cell (38) and inhibit cell growth in strains lacking the cusS gene (12), the distribution of cellular [CusS^mE] broadens significantly toward higher concentrations, reaching up to ∼950 nM for some cells and averaging at ∼312 ± 120 nM, which is ∼2-fold higher than that in the absence of copper stress (Fig. 2A). This protein level increase probably results mainly from transcriptional activation of the cus gene and less from posttranscriptional activation because the mRNA level of cusS gene increases by a similar ∼2.2-fold upon copper stress (SI Appendix, section 1.7 and Fig. S4). Note that this increase in [CusS^mE] happens gradually over a period of ∼2 h after introducing copper stress (SI Appendix, Fig. S22), reflecting that it takes time from transcription to protein synthesis [maturation of the mE tag is faster, at ∼20 min (30)]. Across this larger range of [CusS^mE], the fractional population A_m of the mobile state increases significantly from ∼30% to ∼70% (i.e., more than doubles), with a concurrent decrease of A_s for the quasistationary state (Fig. 2B, open symbols), while both states increase in absolute concentration (e.g., mobile state increases from ∼33 to ∼557 nM; SI Appendix, Fig. S21). (Note that the highest [CusS^mE] data point in Fig. 2B is not reliable because few cells have such high protein concentrations.) Moreover, at any fixed particular [CusS^mE], the 2 mM Cu²⁺ stress increases A_m by a factor of ∼1.3 (e.g., comparing open vs. solid green triangles vertically at ∼200 nM CusS^mE; Fig. 2B). Altogether, these results show that copper stress leads to a significant mobilization of CusS in the cell membrane, which results more from an increase of overall cellular CusS concentration (more than 2-fold) and less from direct copper binding to CusS proteins (only ∼1.3-fold).

Previous studies have shown that dimeric CusS can bind Cu⁺/Ag⁺ at two types of sites in its periplasmic domains (12): at two interfacial sites [each with Phe43 and His176 from one monomer and His42 from the other monomer as ligands; K_d ∼8 μM for Ag⁺ (29)] that leads to the dimerization of the two periplasmic domains and the activation of CusS’s downstream signaling; and at two internal sites (each with Met133, Met135, His145, and Ser84 as ligands) that are not conserved across different organisms and appear insignificant for cell viability under copper stress (12). To probe whether copper binding at the interfacial or internal sites is relevant to the observed CusS mobilization in the membrane, we first mutated CusS^mE’s His42 to alanine, which should abolish its copper binding at the interfacial sites and was shown to abolish CusS autophosphorylation and impair cell’s resistance to copper stress (SI Appendix, Fig. S5) (12, 14). The behaviors of this CusSH42AmE mutant are similar to CusS^mE with or without copper stress (Fig. 2C vs. Fig. 2B), including the increase in cellular CusS concentration upon copper stress as well as the significant mobilization with increasing [CusSH42AmE] and at any particular [CusSH42AmE]. This similarity strongly suggests that copper binding at the interfacial sites, although important for CusS activation, has no significant contribution to CusS mobilization in the membrane. The copper stress-induced CusSH42AmE concentration increase is surprising here, as this H42A mutation abolishes the activation of CusS, which autoregulates itself; we attribute this protein concentration increase to cross-activations of cusRS regulon by other regulatory pathways (22).

Alternatively, we performed an M133I/M135I double mutation on CusS^mE to abolish copper binding at the internal sites. As reported previously (12), these mutations do not give phenotypical changes in cells’ resistance to copper stress (SI Appendix, Fig. S5). Without copper stress, this CusSM133I/M135ImE mutant behaves similarly to CusS^mE (Fig. 2D vs. Fig. 2B, solid symbols). However, in the presence of 2 mM Cu²⁺, A_m only increases to ∼50%, significantly smaller than the ∼70% for CusS^mE, even though the protein concentration range reaches high levels (Fig. 2D vs. Fig. 2B, open green triangles). Moreover, at any particular CusS concentration (i.e., comparing data vertically in Fig. 2D), the copper-induced mobilization (i.e., increase in A_m) vanishes. Altogether, these results suggest that copper binding at the internal sites of CusS, albeit having no phenotypic consequences, plays a role in copper-induced CusS mobilization in the cell membrane, perhaps via inducing CusS conformational changes.

We further studied the diffusion dynamics of CusS^mE in a ΔcusR deletion strain. Two-state diffusion behaviors were similarly observed (SI Appendix, Fig. S24), indicating that the protein concentration-induced and copper-induced mobilization of CusS is unrelated to CusR. The copper-induced CusS concentration increase in this ΔcusR strain (as also observed in CusSH42AmE) again indicates that accumulation of CusS can be independent of a functional CusS–CusR regulatory circuit and here likely comes from cross-activation via other regulatory pathways (22).

Mobile and Clustered States of CusS Are in Dynamic Exchange.

To probe whether the mobile free CusS^mE dimers and their quasistationary clusters can exchange dynamically, we examined the interconversion kinetics between these two states (SI Appendix, section 6). Here, we analyzed the displacement length-vs.-time trajectories of single CusS^mE in the cell, obtainable from the tracking trajectories as in Fig. 1B, Inset. Thresholding these displacement trajectories with an upper displacement limit (determined from the diffusion constant of the quasistationary state) selects out those small displacements and provides the estimate of the individual residence times τ of a single CusS^mE dimer within the clusters (SI Appendix, Fig. S17A). By analyzing the distribution of τ and correcting for the photobleaching/blinking kinetics of the mE tag (SI Appendix, Fig. S17B), and using a validated quasiequilibrium two-state kinetic model between the mobile and clustered states (SI Appendix, Fig. S17C), we obtained k_d and k_c, the two apparent first-order rate constants of CusS^mE leaving and entering the clustered state, respectively, as a function of [CusS^mE].

k_d and k_c are on the order of 10⁰ to 10¹ s⁻¹, corresponding to an interconversion timescale of 10⁻¹ to 10⁰ s between the mobile free CusS^mE and the quasistationary clusters (Fig. 2E). This interconversion timescale is also 1,000 times faster than the typical protein lifetime of ∼1,200 min in bacteria (39⇓⇓–42), indicating that these two states are in dynamic exchange (Fig. 1D) and cannot be from different-aged CusS proteins in the cell (e.g., newly synthesized vs. partially degraded/aggregated). Moreover, under copper stress and with increasing [CusS^mE], k_d, the declustering rate constant increases (Fig. 2E, open blue squares), consistent with the mobilization of CusS in the membrane. An independent analysis using hidden Markov models shows consistent results (SI Appendix, Fig. S18C and section 6). The dynamic exchange between the two diffusion states are also observed for CusSH42AmE and CusSM133I/M135ImE (SI Appendix, Fig. S17 E and F).

Mobilized CusS Interacts with CusR upon Copper Stress.

We postulated that the mobilized, free dimer state of CusS in the membrane should be functionally important for interacting with CusR for signal transduction because of the following: First, significant mobilization of CusS was observed only after copper stress (Fig. 2B). Second, the major contributor to this enhanced mobilization is the increase in [CusS] in the cell, which is again a result of copper stress. Third, the physiological function of CusS is to sense copper (and silver) stress and transduce the signal to CusR. To test this postulate, we imaged in cells the dynamic behaviors of CusR^mE, which carries a C-terminal mEos3.2 tag encoded at CusR’s chromosomal locus. It is worth noting that in making this fusion, we considered the 11 base pair overlap between cusR and cusS genes so that the strain still contains an intact CusS (SI Appendix, section 1.1). Cell growth assays and Western blot show that CusR^mE is as functional as the untagged CusR in conferring cell copper resistance and is largely intact with <9% cleavage of the tag (SI Appendix, Figs. S5 and S6).

We then performed SMT of CusR^mE along with SCQPC to quantify its concentration in each cell ([CusR^mE]). Grown in copper-depleted medium, the cellular [CusR^mE] (in terms of monomers) ranges from ∼33 to 600 nM, averaging at 167 ± 42 nM (Fig. 3A), consistent with a previous immunoblotting study (8). Under 2 mM Cu²⁺ steady stress, [CusR^mE] increases expectedly, averaging at 285 ± 110 nM, but could reach ∼1,100 nM for some cells (Fig. 3A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

SMT reveals CusR interaction with the mobile state CusS upon copper stress. (A) Distribution of cellular CusR^mE concentration (in monomers) in the WT CusR^mE strain in the absence or presence of 2 mM Cu²⁺ stress. (B) ITCDD-corrected distribution of displacement length r at [CusR^mE] = 167 ± 12 nM in WT cells under copper-depleted condition. Black dotted line: overall fit; solid lines: the resolved freely diffusing (FD) (blue), intermediate (IN) (green), and tightly bound (TB) (red) diffusion states. (C and E) Fractional populations of the three diffusion states (C) and diffusion constants of the IN state (E) vs. [CusR^mE] in WT cells under copper-depleted (solid symbols) and copper-stress conditions (open symbols). Blue, green, and red: FD, IN, and TB states, respectively. (D and F) Fractional populations of the TB state (D) and diffusion constants of the IN state (F) at low cellular [CusR^mE] (i.e., ∼0 to 143 nM) in various CusR mE-tagged strains under copper-depleted/stress conditions. Please note the break in the y axis in D. Error bars are SD.

We again sorted cells into groups of similar [CusR^mE] and analyzed ITCDD-corrected displacement length distributions to extract the minimal number of diffusion states, as well as their corresponding diffusion constants and fractional populations. At any cellular [CusR^mE], three states are resolved (e.g., Fig. 3B). Under copper-depleted conditions, the fastest diffusion state has a diffusion constant D_FD of 7.9 ± 0.6 μm²⋅s⁻¹, which we assigned as CusR^mE freely diffusing (FD) in the cytoplasm (either as monomers or dimers), as these species are expected to exist and D_FD is consistent with those for freely diffusing proteins in the bacterial cytoplasm (34, 43, 44). The slowest state has a diffusion constant D_TB of 0.02 ± 0.01 μm²⋅s⁻¹, which we assigned as CusR^mE tightly bound (TB) to the chromosome because CusR, when phosphorylated (i.e., activated), can bind tightly at cognate sites; D_TB is also consistent with those for proteins that bind chromosome specifically (31, 34, 45, 46), and its small magnitude mainly reflects chromosomal dynamics and experimental uncertainties in molecular localization. This TB state may or may not have contributions from CusR^mE interacting with the quasistationary state of CusS, which has a similar diffusion constant (∼0.05 μm²⋅s⁻¹). With increasing cellular [CusR^mE], the fractional population A_FD of the freely diffusing state increases expectedly (Fig. 3C, solid blue squares), while A_TB of the tightly bound state decreases, which is also expected as there are a limited number of sites on chromosome for CusR^mE to bind tightly (Fig. 3C, solid red squares); both trends support the assignments of these two states. The third, intermediate (IN) diffusion state has a diffusion constant D_IN in the range of 0.24 to 0.33 μm²⋅s⁻¹ (Fig. 3E, solid green); we attributed this state to having contributions from CusR^mE nonspecifically bound to the chromosome [which is expected to occur, and the magnitude of D_IN is consistent with reports (31)] and possibly from interactions with the mobile state of CusS, the two of which are not resolved here.

Under copper stress, the diffusion constants D_FD and D_TB of the freely diffusing and tightly bound states remain almost unchanged from those under copper-depleted conditions and subsequently were globally fitted across [CusR^mE] groups (SI Appendix, section 4.4). Importantly, the fractional population A_TB of the tightly bound state increases upon copper stress, especially at low [CusR^mE]—at ∼83 nM, A_TB increases from ∼47% to ∼62% (Fig. 3C, solid vs. open red squares)—consistent with that upon copper stress, there should be more phosphorylated (i.e., activated) CusR^mE, which can bind the chromosome tightly. More interestingly, at the low end of cellular [CusR^mE] range, the value of the intermediate diffusion state D_IN is 0.80 ± 0.02 μm²⋅s⁻¹ (Fig. 3E, open triangles), much larger than under copper-depleted conditions and close to the diffusion constant of the mobile CusS state (0.97 ± 0.13 μm²⋅s⁻¹, Fig. 1B). We attributed this larger D_IN here to a significantly increased interaction of CusR with mobile CusS under copper stress. This attribution is consistent with expectation that, under copper stress, CusS affinity for CusR should increase to recruit CusR for phosphorylation and subsequent signal transduction. Notably, this copper stress-induced increase in D_IN diminishes with increasing cellular [CusR^mE] (Fig. 3E, open triangles), suggesting that at higher [CusR^mE], this intermediate state becomes dominated by nonspecific interactions with the chromosome rather than with mobile CusS (see discussion in SI Appendix, section 6).

To further validate the assignments of the diffusion states of CusR^mE in cells, we examined CusR’s behavior in two control strains. The first is a ΔcusS strain, in which no CusS-to-CusR signal transduction is possible (SI Appendix, Fig. S5). Here, SMT still resolved the three diffusion states of CusR^mE. However, at the low ∼70 nM cellular [CusR^mE], the copper stress-induced increase in A_TB is diminished compared with the wild-type (WT) CusR^mE strain (Fig. 3D, columns 3 and 4 vs. 1 and 2), consistent with that no CusS is present to phosphorylate and activate CusR for subsequent tight binding to chromosome. The residual small increase in A_TB is attributable to CusR phosphorylation via cross-talks by other sensor kinases (22, 47) (Discussion). More important, the copper stress-induced increase in D_IN of the intermediate state in the low protein concentration region, as observed in the WT CusR^mE strain, vanished in the ΔcusS strain (Fig. 3F, columns 3 and 4 vs. 1 and 2), strongly supporting that this increase in D_IN in the WT strain reflects CusR interactions with CusS, more specifically with the mobile state of CusS, where the interactions switch on upon copper stress.

The second control strain harbors a phosphorylation site mutation D51A on CusR^mE (i.e., CusRD51AmE), so that CusR cannot be phosphorylated/activated (SI Appendix, Fig. S5) (14). Past studies (48⇓⇓–51) on other TCS systems showed that this type of mutation on response regulators also alters their conformations and impairs their interactions with sensor kinases. No increase in A_TB was observed upon copper stress (Fig. 3D, columns 5 and 6), consistent with that CusRD51AmE cannot be phosphorylated (not even via cross talks here) to increase its DNA binding affinity. Moreover, the increase in D_IN upon copper stress also vanished (Fig. 3F, columns 5 and 6 vs. 1 and 2), further supporting our assignment of its origin in the WT strain, because CusRD51AmE was expected to not interact with CusS.

Interestingly, CusRD51AmE always has a significant fractional population A_TB of the tightly bound state to the chromosome (e.g., ∼50% at low CusR concentration), comparable to that for WT CusR^mE under copper-depleted conditions, in which the unphosphorylated form should be a significant component of cellular CusR (Fig. 3D, columns 5 and 6 vs. 1). This comparability directly suggests that unphosphorylated CusR can also bind DNA tightly (although not as strongly as the phosphorylated form), consistent with reports on unphosphorylated forms of other response regulators (52⇓–54) (this binding does not lead to gene activation, however, possibly due to wrong interaction conformations).

CusS’s Affinity toward CusR Switches on upon Sensing Copper at Its Sensor Domain Prior to ATP Binding and Autophosphorylation at Its Kinase Domain.

The pathway from CusS sensing copper (or silver) to the eventual phosphorylation and activation of CusR comprises a sequence of events (14) (Fig. 4): i) copper binding at the interfacial sites in CusS’s periplasmic sensor domains, leading to sensor domain dimerization and allosteric conformational changes in the cytoplasmic kinase domains; ii) ATP binding at the kinase domains; iii) cis-autophosphorylation of the kinase domain H271 residue (14); and iv) phosphoryl group transfer to the D51 residue on CusR, activating CusR for subsequent binding to chromosomal operator site.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Schematic of CusS-to-CusR signal transduction in cells upon sensing copper, showing (A) the two-state equilibrium of CusS in the inner membrane (IM); (B) CusS’s mobilization upon binding copper at its periplasmic sensor domain, and the recruitment of CusR; (C) ATP binding to CusS’s cytoplasmic kinase domain and its autophosphorylation; (D) phosphoryl group transfer to the bound CusR; and (E) the eventual binding of phosphorylated CusR to chromosomal operator site for transcription activation.

We have shown that, upon copper stress, CusR interacts significantly with the mobile state of CusS. To probe where among the sequence of events this interaction affinity switches on, we first mutated the autophosphorylation site H271 at CusS’s kinase domain to alanine (i.e., CusS_H271A), eliminating step iii, the last step before phosphoryl group transfer to CusR, and imaged the behaviors of CusR^mE in cells. SMT shows that upon copper stress, the increase in the fractional population A_TB of tightly bound state to chromosome is much less compared with the WT CusR^mE (Fig. 3D, columns 7 and 8 vs. 1 and 2), consistent with that CusS_H271A cannot phosphorylate CusR, leaving only CusR phosphorylation via cross-talk mechanisms for subsequent chromosomal binding, similar to the behavior of the ΔcusS strain. However, unlike the ΔcusS strain, D_IN, the diffusion constant of the intermediate state that contains CusR interaction with the mobile CusS, still increases (as in the WT CusR^mE strain) upon copper stress (Fig. 3F, columns 7 and 8 vs. 1 and 2), indicating that mobile CusS can interact with CusR even without autophosphorylation.

We then made N386A/N414A double mutations in CusS’s kinase domain to abolish its ATP binding (14) (i.e., step ii; Fig. 4). Upon copper stress, similar behaviors of CusR^mE were observed as in the above CusS_H271A strain especially the increase in D_IN (Fig. 3F, columns 9 and 10 vs. 1 and 2). These results indicate that copper-induced CusR–CusS interactions are still feasible even without ATP binding to CusS.

Finally, we made an H42A mutation at CusS’s interfacial metal-binding sites in its periplasmic sensor domains, abolishing CusS’s copper sensing (step i in Fig. 4) and examined CusR^mE. Upon copper stress, there is a small increase in the tight DNA binding population (A_TB) (Fig. 3D, columns 11 and 12), as expected for the same cross-talk mechanism. Strikingly, there is no increase in D_IN (Fig. 3F, columns 11 and 12), as in the ΔcusS strain, reflecting no significant interactions between CusR and mobile CusS_H42A. Altogether, these results show that within the CusS-to-CusR signal transduction pathway, significant CusS–CusR interaction switches on as soon as CusS senses copper and before CusS undergoes subsequent ATP binding and autophosphorylation.