In this study we demonstrate that the global MarR-type virulence regulator RovA of Yersinia acts as an intrinsic protein thermometer, that controls its DNA binding activity and regulates its degradation by the ATP-dependent protease Lon, a process which is also subject to growth phase control.
RovA activity is shown to be strongly dependent on temperature, both in vivo and in vitro, in such that it binds with higher affinity and enhanced cooperativity to DNA at lower temperatures. According to our model (), the sensor-regulatory activity is based on a conformational adaptation of RovA in response to temperature for regulation of its DNA-binding function. The loss of structured elements upon a temperature shift from 25°C to 37°C strongly supports partial defolding of RovA in this temperature range. This makes RovA less capable of binding DNA in a cooperative manner, and reduces its ability to stimulate the inv and rovA promoters. Thermo-induced conformational changes are reversible, as α-helicity and DNA-binding capacity of RovA are regained upon cooling to 25°C, and this could be important for the regulatory function of RovA.
Model of temperature-mediated regulation of rovA expression.
To the best of our knowledge only two other bacterial regulators, which belong to a different regulator class are capable of responding directly to temperature: the transcriptional regulator TlpA of Salmonella enterica
serovar Typhimurium which exhibits homology to KfrA and the SMC proteins involved in plasmid partition or chromosome segregation and the heat shock gene repressor protein RheA of Streptomyces albus 
. The dimeric TlpA protein forms a long left-handed supercoiled coiled-coil domain in which two subunit α-helices wind around each other and pack their side chains in a “knobs-into-holes” manner 
. The coiled-coil structure is a versatile and rather flexible motif in mediating protein-protein interactions, and it has been shown that TlpA undergoes a reversible conformational switch in response to temperature changes, leading to alterations between the unfolded monomeric form and the folded DNA-binding coiled-coil oligomeric structure. Also the RheA protein acts as a protein thermometer with shorter coiled-coil domains, and a thermo-induced change in the repressor leads either to an active or inactive form 
. The structure of the MarR-type regulator RovA is significantly different from TlpA and RheA. Its internal region contains the DNA-binding domain that is predicted to adopt a winged-helix fold. The first N- and the last two C-terminal α-helices appear to form an extensive and well-packed dimer interface, similar to that seen in the structure of other MarR-type regulators. These help to stabilize the formation of dimers with properly positioned DNA-binding segments 
. This type of dimer formation makes it very unlikely that thermo-sensing of RovA is based on a monomer-to-dimer conversion. In fact, we found no evidence to suggest dissociation of the dimers into monomers at 37°C; in contrast, RovA dimers are still detectable after heating to 95°C in the presence of SDS (H. Tran-Winkler, unpublished results). However, our previous structural-functional analysis showed that even small changes, i.e. single amino acid substitutions in the N-terminal region (L12A, W16A), which have no detectable effect on RovA dimer formation, cause a severe defect in DNA-binding 
. This suggests that temperature-dependent structural changes within thermo-sensitive elements of the RovA dimer might effect the proper positioning of the DNA-binding segments.
Our studies further demonstrate that the level of RovA in the bacterial cell is not only determined by transcription, but is also subject to growth phase- and temperature-regulated proteolysis. The RovA protein is stable during stationary phase and/or at moderate temperatures (20°C–25°C), but becomes highly unstable during exponential growth at 37°C. Degradation of RovA under these conditions is primarily mediated through the Lon protease, albeit the Clp proteases also appear to participate to a very small extent in the degradation process. Both types of proteases are ATP-dependent and assigned to the AAA+
superfamily of ATPases. Lon and ClpP proteases share two common features: (i) access to the proteolytic chamber of the enzyme is usually prohibited to globular proteins, most likely to prevent unrestrained protein degradation, and (ii) they require ATP hydrolysis to unfold and translocate the substrates into the protease chamber. Degradation within the chamber is processive, including sequential rounds of substrate binding, release and rebinding to the proteolytic site, and generates 10–15 amino acid peptides without generation of partially digested protein intermediates 
. The ClpP and Lon proteases usually degrade improperly folded or damaged proteins. However, undamaged proteins, in particular short-lived regulatory factors which are implicated in developmental processes, stress resistance and bacterial fitness can also serve as substrates for proteolysis 
According to our results, RovA degradation by the Lon protease is clearly temperature-dependent, but alterations of the DNA-binding activity due to thermo-induced conformational changes seems much more critical for thermo-dependent rovA
expression than the change of RovA stability. For instance, little autoactivation of rovA
transcription is evident at 37°C, even when RovA is abundant in the absence of the Lon and Clp proteases (). This raises the question of why is RovA also rapidly degraded at 37°C. Thermo-induced structural alterations led to a strong reduction, but did not cause a complete inactivation of the DNA-binding function of RovA (, ). Accordingly, degradation of the global regulator might be required to prevent binding to higher affinity sites of RovA within the Yersinia
genome. Homologous MarR-type regulators were also shown to interact with small metabolites or signalling molecules 
. In this case, rapid proteolysis of RovA might be essential to prevent sequestration and inactivation of important regulatory components through complex formation with RovA. As RovA proteolysis is also responsible for growth phase regulation of rovA
and RovA-dependent genes, thermo-dependent alterations of the degradation process might be important to link and coordinate thermoregulation and growth phase control ().
To date very little is known as to what feature identifies transcriptional virulence regulators as substrates of Lon or Clp proteases. Certain peptide motifs of an exposed or unstructured region of a protein, including a few non-polar, aromatic amino acid residues can serve as recognition signal for proteases 
. Although the major determinants for Lon-mediated proteolysis are frequently found at the N- and C-termini of target proteins 
, it has recently been reported that Lon can obviously also recognize internal tags 
. In MarR-type regulator proteins, such as RovA, both termini form α-helical structures, which contribute to the formation of the dimer interface 
. Here we report that transplantation of the N-terminal 96 amino acids of RovA to a normally stable protein confers instability to Lon, but a transfer of the first 42 amino acids does not. Also introduction of a mutation abolishing the DNA-binding function of RovA rendered the protein more susceptible to Lon, whereas presence of a multi-copy plasmid harbouring RovA-binding sites reduced degradation of the protein. This suggests that amino acid residues in the vicinity of the central winged-helix DNA-binding domain comprise the information necessary for Lon binding and/or provides the foundation for degradation by Lon. This finding raises the intriguing possibility that Lon recognition and degradation of RovA is in direct competition with the DNA-binding function of the regulator. In fact, two other substrates, SoxS and the N protein of bacteriophage λ are known to be protected from Lon-mediated degradation when bound to DNA, although their instability is an intrinsic property and does not require an external signal to trigger degradation 
. From results in this study we know that temperature strongly affects the susceptibility of the RovA protein to Lon-mediated proteolysis. In this regard, it is very likely that the protein degradation signals are normally buried in active RovA and less accessible in RovA-DNA complexes, but become more accessible to the AAA+
protease in the non-bound state as a consequence of thermo-induced defolding events ().
Lon-mediated turnover of RovA in vivo
is likely modulated or regulated by accessory factors that affect either the enzymatic activity of the protease or the conformational state of the target protein. Experiments with an in vitro
degradation system demonstrated rapid degradation of α-casein by the Yersinia
Lon protease, but RovA proteolysis was significantly less efficient than proteolysis in vivo
. In contrast, rapid degradation of RovA could be observed when crude extract of an exponentially grown lon
mutant strain was added to the in vitro
system, indicating that an additional component is required for RovA-mediated proteolysis. As RovA degradation is significantly reduced during stationary phase, it seems likely that the postulated accessory component is only present during exponential growth (). Lon-mediated proteolysis might be influenced by interactions with partner proteins, such as adaptors that tether substrates, or chaperones and proteases, which might create or help to expose the degradation signals of RovA. For example, the DnaJ/DnaK/GrpE molecular chaperone system is required to promote formation of certain AAA+
protease-substrate complexes 
, and only endoproteolytic cleavage and interaction with the SspB adaptor protein render the transmembrane RseA protein susceptible for ClpXP-mediated proteolysis 
. Activity of AAA+
proteases was further shown to be subject to modulation by cellular compounds, ions and metabolites 
. Since MarR-type regulators are frequently modulated by small effector molecules 
, it will also be interesting whether RovA also undergoes effector-induced conformational changes, which could be crucial for growth phase-dependent RovA degradation.