Allostery and protein plasticity: the keystones for bacterial signaling and regulation

Information transmission by repositioning the structure of whole domains

The most classical and easy-to-picture paradigm of allostery involves atomic coordinates shifts of the structure physically connecting the allosteric and orthosteric sites. In other words, according to the occupation of the allosteric site by the signal, structurally distinct ground states of the active functional site result. The entire protein or domains shift via rigid-body motions (Daily and Gray 2009) in the micro- to milliseconds range. Members of the Fur (ferric uptake regulator) superfamily of OCSs are widespread homodimeric regulators that control iron homeostasis in bacteria, as well as other metals (zinc, manganese, nickel, etc.). Fur, among several other OCSs, is an excellent illustration of this allosteric coupling mechanism. Fe²⁺ is bound by the dimerization LBD into a pocket that directly bridges this domain to the output response DBD, allosterically inducing a pronounced rotation. This rigid-body motion aligns the DNA-recognition helices of the DBD helix-turn-helix motif into a competent geometry to bind DNA through its major groove (Fig. 1b). It is in its Fe²⁺-bound, holo form, that Fur is enabled to bind to specific Fur-boxes in the promoters of target genes (Bradley et al. 2020), repressing the transcription of genes coding for proteins involved in iron uptake and assimilation—more infrequently, activatory roles have also been reported (Delany et al. 2004). Apo Fur may be more flexible than initially thought, explaining why relatively few apo structures are available, among which variable inter-domain geometries are observed (Sarvan et al. 2018). Thus, the distinction with dynamic allosteric mechanisms—described next—appears now to be less clear-cut, and might all be specific cases of a more general, common mechanistic scenario.

Information transmission by modulating protein dynamics

The allosteric mechanism might also involve changes of the proteins’ conformational flexibility. Such dynamic features refer to pico- to nanosecond motions of amino acid side chains and/or main chain, around a defined average structure. These motions are not directly observable by ground-state microscopic approaches such as X-ray crystallography or cryo-electron microscopy, although they can be deduced from parameters that are refined with those techniques, e.g., atomic displacement factors (temperature or B factors) or local resolution figures. NMR spectroscopy has become a powerful method to characterize dynamic allosteric transmission with great detail, providing both atomic-level information on chemical environments and quantitative conformational dynamics data. A whole range of illustrative examples comes from another group of OCSs, the TetR family of regulators (TFRs) (Yu et al. 2010). TetR was the first factor describing an OCS architecture that has since been recognized as one of the most numerous transcription factor families in prokaryotes. TetR mediates resistance to tetracycline, by sensing the antibiotic and thereafter inducing the expression of a gene encoding a tetracycline efflux pump in Gram-negative bacteria (Orth et al. 2000). TFRs are homodimeric OCSs, most of which act as repressors, inducing the expression of target genes via ligand-binding-dependent DNA dissociation. Recent work with FasR, a TFR from Mycobacterium tuberculosis, further confirms how these OCSs’ allostery relies on strong dynamic modulation, in the form of order/disorder transitions that link the signal sensing to the output DNA-binding sites (Lara et al. 2020). M. tuberculosis FasR is unusual in that it acts as a transcriptional activator of fas I and acpS genes encoding fatty acid biosynthesis enzymes, key actors in the lipid metabolism and cell wall homeostasis of this pathogen. FasR binds long-chain acyl-CoAs through a deep hydrophobic tunnel open on both ends within the sensor LBD. A physically continuous spine of hydrophobic residues that connects the protein-folding cores of both domains of FasR was identified and seen to be conserved in all TFRs. The hydrophobic spine is interrupted precisely by the ligand-binding cavity, such that only in the ligand-bound condition is the spine completed by the ligand itself (Fig. 1c). The allosteric mechanism at play in TFRs appears to be based on a disordered (flexible, no ligand bound) to ordered transition, with the latter stabilized in a DNA-incompetent conformation, as observed in M. tuberculosis FasR (Lara et al. 2020) and TetR (Kamionka et al. 2004; Reichheld et al. 2009). Another TFR, QacR from Staphylococcus aureus, binds a range of different drugs inducing drug resistance. The integration of crystallographic and NMR data (Takeuchi et al. 2019) in principle upholds a two-state model, which would thus be more similar to Fur (vide supra). However, differently sized drugs trigger a linear range of inducive effects, and NMR relaxation data indicate that apo QacR is in equilibrium between repressive and inducive conformations constituting a conformational selection scenario (Takeuchi et al. 2019) thus ruling out a simple two-state model. Furthermore, revisiting available structures of wild-type QacR and engineered mutants, the hydrophobic spine can be readily identified, with different ligands completing it to different extents. A continuum of dynamic modulation for information transmission, implying different levels of disorder/flexibility, is thus present among the several thousand TFR systems that have evolved in bacteria. Considering OCSs and other signaling systems alike, this idea of a continuum in the way that proteins achieve allosteric transition is a bridging concept between induced fit and conformational selection, mechanisms that are not essentially dissimilar at the molecular level.

Coiled-coils can be long-range carriers of information…if they are not too stable

The active site in HKs is under allosteric regulation from the distant sensor domains. Input signals in the form of chemical ligands or physical modifications perceived by the HKs’ sensor domains act as the allosteric effectors. Connecting the sensor domains with the HK’s catalytically active region, a coiled-coil works as an effective allosteric route, exploiting its structural features. The central domain, including the DHp in all HKs, is α-helical, and α-helices are known to span approximately seven residues (positions a through g) every two turns. When positions a and d are both occupied by hydrophobic residues, their side chains stick out from the same side of the helix and tend to interact with neighbor helices comprising similar heptad-repeating hydrophobic patterns. This side-by-side packing results in a left-handed superhelix or coiled-coil. Shifts at one end of a coiled-coil can be mechanically transmitted into substantial positional shifts at distantly located residues, exploiting the stiffness of α-helices and sequence-encoded heptad-repeat deviations. Non-ideal heptad repeats, bearing polar residues at a/d positions and/or insertions, reduce the coiled-coil’s energetic stability and enable conformational switching for allosteric regulation. HK sequences comprise a coiled-coil structure along the DHp four-helix bundle where the reactive histidine is located and through an N-terminal two-helix extension that connects to the sensor domain(s). Such coiled-coil region was presciently identified as a key functional element on the basis of HKs sequence analyses (Singh et al. 1998). This two-helix coiled-coil segment has also been named helix Jα (Möglich et al. 2009) or S-helix (Anantharaman et al. 2006), indeed exhibiting non-ideal heptad-repeat motifs. Additional S-helices, when more than one sensory or signal-transmission domains are present, connect the intervening domains (Lesne et al. 2018).

The S-helix is the means to transmit positional information downstream, by helical rotation and helix axis translation following signal-effected coiled-coil rearrangements. Rotational rearrangements of S-helices had been proposed to underlie HK activation in quorum sensing (Neiditch et al. 2006) and light perception (Möglich et al. 2009) pathways, based on functional experiments. A first direct observation of this allosteric route involving S-helix coiled-coil rearrangements was obtained with X-ray structures of Bacillus subtilis DesK, a thermosensor, in different conformational/signaling states (Albanesi et al. 2009; Saita et al. 2015; Trajtenberg et al. 2016; Imelio et al. 2017) and then confirmed in several other cases including NarX (Huynh et al. 2013), NarQ (Gushchin et al. 2017), and NsaS (Bhate et al. 2018). Further, supporting evidence for a universal coiled-coil-borne allosteric mechanism has been provided with elegant experiments on light-sensitive (Berntsson et al. 2017b, 2017a; Engelhard et al. 2017) and virulence-regulating (Wang et al. 2014; Lesne et al. 2016, 2017, 2018) HKs. Coiled-coil formation/stabilization typically drives HKs to kinase-off/phosphatase-on conformations (Fig. 3a), whereas coiled-coil disruption/destabilization leads to kinase-on/phosphatase-off states (Trajtenberg et al. 2016; Gushchin et al. 2017). Transmission can also travel through such a coiled-coil in the opposite direction. The phosphorylated RR P ~ DivK—member of the Caulobacter crescentus DivJ-DivK-PleC phosphorelay that controls cell identity during asymmetric cell division—binds to the DHp domain of the DivL HK and, via DivL’s S-helix, triggers the allosteric upstream reorganization of its own sensory PAS domains (Childers and Shapiro 2015), a key event for subsequent DivL-dependent inhibition of a second phosphorelay—CckA-ChpT-CtrA—involved in differentiation and cell-cycle progression.

Information transmission via coiled-coil segments in signaling proteins goes beyond HKs. In the TCS-related realm of chemotaxis, analogous coiled-coil destabilization/stabilization switching has been put forward as the molecular basis of allostery connecting MCP perception of attractant/repellent molecules and downstream control of the CheA HK activity (Bartelli and Hazelbauer 2016). RRs provide with additional examples, like in DgcR from Leptospira biflexa, which possesses a coiled-coil segment connecting the phosphorylatable REC domain to the output diguanylate cyclase (DGC) domain. REC phosphorylation modulates allosterically the DGC domains in the dimer via a register shift and extension of the coiled-coil (Fig. 3b), which repositions the inter-domain hinge, eventually triggering c-di-GMP production (Teixeira et al. 2021).

In sum, marginally stable coiled-coils act as information carriers, and their shifts provide mechanical energy to switch among conformational/functional states.

Covalent stabilization of protein conformations by phosphorylation: a particular case of allostery?

The classic definition of allosteric regulation excludes covalent modification of proteins as allosteric effectors (Fenton 2008), considering that such effectors must bind reversibly onto their regulatory site. However, obvious parallels between covalent and non-covalent modifications can be drawn. Protein phosphorylation, even though covalent, is eventually reversible and can thus be considered a special case of binding with very slow dissociation rates (enhanced by dedicated phosphatases in some scenarios). Indeed, the phosphorylation of the reactive aspartate on the REC domain of RRs allosterically regulates the conformation of remote functional sites within the same domain and eventually also the configuration of additional output effector domains when present (Bourret 2010; Gao et al. 2019).

Analogously to the picture described earlier for OCSs, according to which a continuum of protein dynamics underlines allosteric regulation mechanisms, the REC domain in TCS RRs is again a fascinating example of encoded allostery. REC allosteric modulations range from discrete structural switching of key residues’ positions all the way to essentially dynamic order/disorder transitions. Both case scenarios are controlled by the phosphorylatable aspartate reaction center that acts as the allosteric site. This site is phosphorylated by the cognate HK upstream, hence connecting the HK-sensed input signal to the RR-mediated response. Examples of positional shifts in RECs triggered by Asp phosphorylation were observed early on (Zhu et al. 1997; Hastings et al. 2003) and designated as the Y–T coupling mechanism. A conserved threonine that is part of the P ~ Asp environment propagates a phosphorylation-triggered conformational rearrangement to a nearby tyrosine, ultimately channeling an allosteric modulation to distant, solvent-exposed areas. The latter are typically defined by the β4-α4 and β5-α5 loops and part of the α4-β5-α5 surface (for the standard nomenclature of RR secondary structure elements see Gao et al. (2019)), which act as the orthosteric functional sites of RECs: their allosteric alteration typically drives the domain to interact with other proteins exerting response functions, or with itself, forming homo-oligomeric species (dimers or tetramers most often) critical to gain output function attributes. Correlated to REC-mediated dimerization, phosphorylation-dependent allostery also triggers the reconfiguration of output effector domains. For example, RRs that comprise DBDs often exhibit these attached to the REC domain in the inactive state and then unleashed ready to bind DNA upon REC phosphorylation. The antibiotic resistance regulator VraR from S. aureus is an example of this (Leonard et al. 2013). VraR in its inactive form is monomeric and exhibits a “closed” configuration, with the output response DNA-binding domains tightly associated to the N-terminal REC and thus impeded to bind DNA. Upon phosphorylation by VraS, coupling is observed between the P ~ Asp acting as allosteric site and two different areas of the REC domain: while one promotes homo-dimerization, the other one liberates the DBD to trigger the response (Fig. 3c). Further confirming this, engineered perturbations of the DBD:REC interface result in enhanced RR phosphorylation (Barbieri et al. 2010), consistent with the fact that allosteric coupling drives reciprocal effects between allosteric and orthosteric sites.

NMR relaxation analyses of the REC showcase other examples where phosphorylation-triggered allostery in RRs occurs through more dynamic pathways as compared to the paradigmatic Y–T coupling mechanism. The Escherichia coli NtrC is a TCS RR that plays a key role in nitrogen metabolism regulation. The non-phosphorylated inactive state of NtrC REC comprises not one but a heterogeneous collection of conformers that sample the active conformation as well, even though barely populated (Volkman et al. 2001). Phosphorylation stabilizes this single, well-defined active conformation from among the inactive mixture, without the requirement of positional shifting of connected residues, and in this way reinterpreting the Y–T coupling paradigm (Villali et al. 2014). This more general perspective, of a flexible non-phosphorylated REC, inactive for downstream signaling functions, allows for a better understanding of alternative activatory allosteric effectors beyond Asp phosphorylation. In the hypoxia sensor NtrYX TCS from Brucella abortus, the RR NtrX exhibits meta-active conformations independent of phosphorylation, including a homodimeric species with a similar-to-active configuration of the β4-α4 loop (Fernandez et al. 2015). The active form of the HK NtrY selects the meta-active conformation of NtrX, phosphorylating it and stabilizing its active conformation. A similar picture had been reported in B. subtilis DesKR, where the HK-RR interaction stabilizes a pre-existent active state of DesR promoting its dimerization and transcriptional activity (Trajtenberg et al. 2014).

Conflict of interest

The authors declare no competing interests.