Escherichia coli small heat shock protein IbpA is an aggregation-sensor that self-regulates its own expression at a translational level

Aggregation is an inherent characteristic of proteins. Risk management strategies to reduce aggregation are critical for cells to survive upon stresses that induce aggregation. Cells cope with protein aggregation by utilizing a variety of chaperones, as exemplified by heat-shock proteins (Hsps). The heat stress-induced expression of IbpA and IbpB, small Hsps in Escherichia coli, is regulated by the σ32 heat-shock transcriptional regulator and the temperature-dependent translational regulation via mRNA heat fluctuation. We found that, even without heat stress, either the expression of aggregation-prone proteins or the ibpA gene deletion profoundly increases the expression of IbpA. Combined with other evidence, we propose novel mechanisms for the regulation of the small Hsp expression. Oligomeric IbpA self-represses the ibpA/ibpB expression at the translational level, but the self-repression is relieved by the sequestration of IbpA into protein aggregates. Thus, the function of IbpA as a chaperone to form co-aggregates is harnessed as an aggregation sensor to tightly regulate the IbpA level. Since the excessive preemptive supply of IbpA in advance of stress is harmful, the prodigious and rapid expression of IbpA/IbpB on demand is necessary for IbpA to function as a first line of defense against acute protein aggregation. Author summary All organisms have protein quality control systems against stresses disturbing cellular protein homeostasis (proteostasis). The systems have multiple stages: folding, degradation, and sequestration. Sequestration of denatured proteins is the first step to support other maintenance strategies. Small heat shock proteins (sHsps), which are well-conserved chaperones, are representative “sequestrases” that co-aggregate with denatured proteins. We found that IbpA, an Escherichia coli sHsp, is a direct mediator for negative feedback regulation at the translational level. Recruitment of IbpA into the protein aggregates relieves the ibpA expression suppression. This novel mechanism of IbpA as an aggregation-sensor tightly regulates the IbpA level, enabling the sHsp to function as a sequestrase upon aggregation stress.


Introduction
Since proteins tend to form aggregates, cellular maintenance by keeping proteins in their native states and removing denatured proteins is crucial for all organisms. Multi-layered quality control systems are essential to maintain such cellular protein homeostasis (proteostasis) [1,2]. Refolding and degradation of denatured proteins caused by stresses are two primary strategies to prevent the accumulation of protein aggregates. Sequestration of denatured proteins is a third strategy, to keep misfolded proteins in a state that is easy to restore or degrade after stresses [1,2]. Small heat shock proteins (sHsps) participate in the third strategy as "sequestrases", constituting a first line of stress defense against irreversible protein aggregation [3][4][5][6].
sHsps are among the most upregulated Hsps upon stress [4,6,7]. The gene expression of the α-and γ-proteobacterial sHsps is regulated by two mechanisms [7]. One is the heat-shock transcriptional regulator σ 32 , an RNA polymerase subunit, which regulates the transcription of many Hsp genes [8,9]. At normal growth temperatures, σ 32 , which is an extremely unstable protein, is rapidly degraded via a DnaK/DnaJ-mediated pathway. However, σ 32 is stabilized to allow the transcription of many Hsp genes upon heat shock, since DnaK/J is sequestered to rescue the emerging heat-denatured proteins [8,9]. The other is the thermoresponsive mRNA structures in the 5' untranslated region (UTR), called RNA thermometers (RNATs), which mask the Shine-Dalgarno (SD) sequence in their stem loop structures at normal or low temperatures [7,10]. The heat fluctuation by a temperature up-shift melts the stem loops in RNATs and allows the ribosome to initiate translation, using the exposed SD sequence [7,10].
The thermo-responsivity of many bacterial RNATs has been established, and the RNATs of sHsps have conserved shapes harboring two to four stem loops [7,10]. Thus, the expression of sHsps is controlled at both the transcriptional level using heat-shock transcriptional factors, and the translational level using RNATs, in contrast with other Hsps, which are only controlled at transcriptional levels [7,10].
Inclusion body-associated protein A (IbpA) and B (IbpB) are Escherichia coli sHsps, and were originally identified as proteins induced in response to heterologous protein expression in the cell [11,12]. IbpA/IbpB function as a holder of denatured proteins, to facilitate the initiation step of the refolding pathway via DnaK/DnaJ and ClpB [4][5][6]. IbpA/IbpB mediate the efficient transfer of denatured proteins from the sHsp co-aggregation to the DnaK/DnaJ system [13][14][15][16]. IbpA and IbpB are highly homologous proteins with ~50% amino acid sequence identity [11], and form hetero-oligomers in E. coli [14,17,18]. Although most γ-proteobacteria only have a single sHsp (IbpA), a two-sHsp (IbpA and IbpB) system has evolved in a subset of Enterobacterales, including E. coli, from a single gene duplication event [19]. Previous in vitro studies using recombinant IbpA and IbpB proteins revealed the distinct features of the two sHsps: IbpA is more efficient in binding denatured proteins to form the coaggregates [17,19], and even self-forms fibril-like aggregates [18]. However, the coaggregates with IbpA are inefficient substrates for the disaggregation assisted by DnaK/DnaJ and ClpB [17,19]. The additional presence of IbpB in the coaggregates is required for the disaggregation process, suggesting that the interplay between IbpA and IbpB is important to modulate the interactions with substrates in the mixed oligomer states [4,6,16,17,19].
The mRNA encoding the ibpA-ibpB operon has RNATs in the 5' UTRs of both the ibpA and ibpB ORFs, as revealed by RNA structure probing and reporter assays [7,20,21]. Previous analyses of the RNAT in ibpB using a reporter revealed the possible influence of the IbpA protein on ibpB expression [7,21]. In addition to the heat stress, the expression level of IbpA/IbpB was profoundly upregulated, by 10~50-fold, under non-heat stressed conditions such as 30~37 °C in the dnaK-dnaJ deleted strain [22,23] or upon oxidative stress induced by copper [24]. Since the RNAT regulation would not be effective at normal growth temperatures, the mechanism for the massive upregulation under the non-heat stressed conditions remains to be elucidated.
Here we addressed why IbpA is upregulated in non-heat stressed cells. We found that the accumulation of protein aggregates was sufficient for the upregulation. Intriguingly, a reporter assay using the 5' UTR of the ibpA mRNA revealed that the deletion of the ibpA gene increased the reporter translation, which was repressed by the overexpression of oligomeric IbpA. Combined with other evidence, we propose that the IbpA oligomers self-repress the ibpA translation, which is relieved by the sequestration of IbpA by co-aggregation with protein aggregates. The role of the aggregation sensor is specific to IbpA, since the homologous IbpB lacks this self-repression function. The significance of the self-repression by IbpA at the translational level is discussed in relation to the unique role of IbpA in protecting cells from acute heat stress.

Results
ibpA translation is upregulated in response to protein aggregation.
Although the thermometer in the mRNA (RNAT) and the transcriptional control by σ 32 are known mechanisms to upregulate the expression of IbpA, previous studies have reported that IbpA expression is also upregulated under non-heat stressed conditions, such as in the dnaKJ deletion strain or upon copper stress [22][23][24]. The absence of DnaK/DnaJ leads to the production of protein aggregates [22,25]. The addition of copper disturbs protein homeostasis in cells with oxidative stress [24,26]. A common consequence in E. coli cells would be the accumulation of protein aggregates [20][21][22][23]. Therefore, we hypothesized that protein aggregation might somehow be involved in the upregulation of IbpA under the non-heat stressed conditions.
To investigate whether the expression of IbpA is upregulated not only by heat shock but also by protein aggregation under non-heat stress conditions, we expressed aggregation-prone proteins in wild-type E. coli. To do so, we overexpressed rhodanese, a bovine mitochondrial protein, which is known to aggregate in E. coli at 37 °C [27] (S1A Fig.). Strikingly, the expression of IbpA increased upon the rhodanese expression (agg ++ , Fig. 1A). Overexpression of another aggregation-prone protein, SerA of E. coli [25] (S1A Fig.), also massively induced the IbpA expression in wild-type E. coli (S1B Fig.). The rhodanese expression did not increase the level of GroEL, one of the representative Hsps in E. coli (Fig. 1A). These results support the idea that the accumulation of aggregated proteins in cells increases the IbpA expression.
We suspected that the aggregates might induce IbpA via upregulated σ 32 -mediated transcriptional control at 37 °C, since the aggregation-prone proteins could sequester DnaK/J, thus protecting σ 32 from DnaK/J-mediated degradation and eventually stabilizing σ 32 to promote the expression of Hsps. If so, then the overexpression of σ 32 would increase the mRNAs to upregulate IbpA as well as other chaperones, such as GroEL. After we confirmed the ~10-fold induction of ibpA mRNAs in the σ 32 -overexpressing cells, using quantitative real-time PCR (qRT-PCR) (Fig. 1B), we compared the protein expression levels. Upon the σ 32 overexpression, the GroEL expression level increased (Fig. 1A), indicating that the excess σ 32 is effective in increasing the Hsp under the non-heat stress. In contrast, the expression level of IbpA did not obviously increase as compared to that of GroEL (Fig. 1A), even though the ibpA mRNA increased, indicating that the IbpA induction in the presence of aggregates is not explained by transcriptional upregulation using σ 32 . Rather, the results suggest a translation suppression mechanism in the IbpA expression.
To demonstrate that the ibpA upregulation by the accumulation of protein aggregates does not occur at the transcriptional level, we investigated the efficiency of the ibpA translational initiation in a reporter assay. Since the 5' UTR in the ibpA mRNA is critical for the translational control using the stem loops in the mRNA, we constructed a plasmid harboring the 5' UTR of ibpA fused with the gfp gene, under the control of an arabinose-inducible promoter (Fig. 1C). The deletion of dnaK/J (ΔKJ) caused a massive increase in the GFP production upon arabinose induction, verifying that the reporter reflected the features of the IbpA upregulation (Fig. 1C). We observed a similar substantial induction of GFP upon the rhodanese overexpression. The effect is specific for the 5' UTR of ibpA, since there was no upregulation of GFP in the ∆KJ cells or upon the rhodanese overexpression when the 5' UTR was substituted with a 5' UTR derived from the parent plasmid. In addition, the overexpression of another aggregation-prone protein, SerA, also induced the GFP reporter (S1B Fig.). The results using the reporter, where the transcriptional levels controlled by the arabinose-inducible promoter were independent of heat-shock, further confirmed that the accumulation of protein aggregates upregulates the ibpA expression at the translational level.

IbpA self-represses ibpA translation
Next, we investigated the connection between protein aggregation and ibpA translation induction. The well-known physiological function of IbpA as a chaperone is the co-aggregation with denatured proteins [1,[3][4][5][6]. The recruitment of IbpA to protein aggregates might cause the entrapment of the free IbpA in the cytosol. Indeed, the localization of IbpA fused with GFP shifted to the cell poles in the DnaK/J-depleted cells and the rhodanese-overexpressing cells ( Fig. 2A), consistent with previous observations that IbpA and denatured proteins accumulate as inclusions at the cell poles [28][29][30]. Taking this into consideration, we hypothesized that the entrapment of free IbpA from the cytosol, due to the sequestration of denatured proteins, would induce the ibpA translation (Fig. 2B). If this model is correct, then the deletion of IbpA, which is nonessential for E. coli growth, would upregulate the translation of the reporter harboring the 5' UTR of ibpA, used in Fig. 1C.
We deleted the operon including ibpA-ibpB. After we confirmed that the growth of the ∆ibpAB cells was similar to that of wild-type E. coli (S2A Fig.), we evaluated the translation initiation of the reporter. Strikingly, the reporter expression was strongly promoted by the ibpAB deletion, even in the absence of induced protein aggregation (Fig. 2C, lane 2), supporting the idea that IbpA suppresses ibpA translation. We then examined the effect of IbpA overexpression, which caused a slower growth rate in wild-type E. coli (S2B Fig.). The Oligomeric IbpA is critical for the self-repression of translation.
What region of IbpA is critical for the self-regulation of translation? At first we deleted the N-or C-terminal domain of IbpA and examined the effect on translation repression (Fig.   3A). The GFP reporter assay revealed that the C-terminal truncation eliminated the ability to suppress the translation in the ∆ibpAB cells (Fig. 3A). In contrast, the translation suppression by the N-terminal truncation was almost the same as that by wild-type IbpA (Fig. 3A), showing that the C-terminal domain is responsible for the self-translation repression.
The C-terminal domain of IbpA contains a universally conserved motif for oligomerization, IX(I/V) [31,32]. We substituted the motif, IEI in IbpA, with AEA, and confirmed the impaired oligomerization of the AEA mutant (S3 Fig.), as reported previously [31]. Co-expression of the IbpA (AEA) mutant with the reporter in the ∆ibpAB cells did not repress the ibpA translation (Fig. 3B), indicating that the oligomeric state of IbpA is critical for the IbpA self-regulation.

IbpA also suppresses the translation of ibpB.
E. coli also possesses IbpB, a paralog of IbpA, and its expression is also regulated by RNAT in the 5' UTR of the ibpB mRNA [7,20,21]. Gaubig et al. observed that the presence of the ibpA ORF suppresses the ibpB translation [21], but the reporter system used in the study could not reveal the influence of IbpB on the ibpB translation. Therefore, we tested the effect of IbpB on the ibpB translation, after we replaced the 5' UTR of ibpA with that of ibpB in the GFP reporter system. IbpA, but not IbpB, suppressed the reporter for the ibpB translation in the ∆ibpAB strain (Fig. 4A). This result shows the specific function of IbpA as a translation repressor of the small Hsps, IbpA and IbpB, in E. coli.
In contrast, the overexpression of IbpB did not change the ibpA reporter upregulation in the ∆ibpAB strain (Fig. 4B, lane 4), indicating that IbpB cannot substitute for IbpA in the self-repression of the ibpA translation. In the wild-type strain, the IbpB overexpression increased the amount of the GFP reporter (Fig. 4B, lane 3), probably reflecting the hetero-oligomerization of IbpB with endogenous IbpA to reduce the amount of free IbpA for the self-repression. Previous studies revealed that the secondary mRNA structures of the ibpA 5' UTR regulate the ibpA translation [20]. The ibpA 5' UTR contains three stem loops, and the two upstream stem loops (SL1 and SL2) are thought to stabilize the downstream thermo-responsive stem loop (SL3) to mask the SD sequence ( Fig. 5) or protect the mRNA from degradation [10,20]. Since the contribution of these stem loops to translational regulation remains unclear, we constructed a series of stem loop variants in which the stem loop stabilities were weakened or strengthened (S4 Fig.), and evaluated the effects on the GFP reporter. For all three weakened stem loops, the GFP expressions were upregulated in wild-type E. coli as compared to those of the unchanged stem loops ( The expression patterns of the reporters using the strong SL1 and SL2 stem loops were almost the same as those using the weak variants ( Fig. 5 A, B). In contrast, the expression of the reporter using the strong SL3 was not observed under all conditions examined (Fig. 5C), probably because the strong stem loop containing the SD motif is too tight to expose the SD motif for the translation initiation.

Discussion
IbpA and IbpB, small Hsps in E. coli, are chaperones that sequester aggregation-prone proteins by co-aggregation during stress. This study unveiled a previously unknown function of IbpA, as a mediator for negative feedback regulation at the translational level. Our experiments revealed that the IbpA oligomers serve as their own translation suppressor. The titration of IbpA by co-aggregation with denatured proteins relieves the translation suppression.
We propose a novel mechanism of the ibpA regulation, where IbpA-mediated aggregation sensing self-regulates the ibpA translation (Fig. 6). Without protein aggregation, free IbpA oligomers suppress ibpA translation initiation. In contrast, under stress conditions involving protein aggregate accumulation, IbpA recruits aggregation-prone proteins to form co-aggregates. This titrates the free IbpA away to relieve the translation suppression, leading to the massive increase in IbpA expression to maintain cellular proteostasis.
This model of IbpA expression regulation resembles the σ 32 -mediated transcriptional regulation of Hsps [8,9], since the chaperones are titrated away by denatured proteins under stress conditions in both mechanisms. One of the main differences between the two mechanisms is that the IbpA-mediated self-regulation takes place at the translational level, providing an advantage for a rapid response to the emergence of aggregation-prone proteins.
Previous studies revealed the layered regulation of IbpA expression: σ 32 -mediated transcriptional control and RNA thermometer (RNAT) translational control [7,21]. Since the stem loops in the RNAT system influence the IbpA-mediated translation repression (Fig. 5), RNAT and the self-repression control are not independent. In the RNAT mechanism, the stem loops fluctuate and melt to expose the SD region, depending on the temperature. As reported previously, higher temperatures cause more melting of the stem loops, like a "thermometer".
In other words, the degree of melting gradually changes, and is not all or none, over a wide range of temperatures [7,10]. This inherent property would allow the stem loops to partly open even under mild conditions such as 37 °C, leading to the leaky expression of certain amounts of IbpA. Thus, the IbpA-mediated self-repression would function to tightly shut off the IbpA expression as a "safety catch" in the leaky RNAT system. Taken together, the stringent repression mechanism, combining RNAT and the IbpA-mediated negative feedback control, has evolved to fulfill the following requirements: tight repression under unstressed conditions, and acute upregulation upon aggregation-stress. This mechanism enables IbpA to be one of the most upregulated chaperones upon aggregation inducing stresses.
IbpA serves as a first line of defense against protein aggregation, where oligomerized IbpA co-aggregates with aggregation-prone proteins for sequestration. Why does IbpA employ such a feedback control mechanism in addition to the known regulation controls including σ 32 and RNATs? Considering that IbpA is an ATP-independent oligomeric chaperone, a greater than stoichiometric amount of IbpA would be necessary to sequester the aggregation-prone proteins. One strategy for risk management is to prepare an abundance of IbpA protein even under unstressed conditions. However, this might not be appropriate, since IbpA overexpression had detrimental effects under the normal conditions (S2B Fig.), probably due to the self-formation of fibril aggregates [18], which could perturb proteostasis and compromise the sequestration activity. Thus, the expression of IbpA should be tightly repressed under normal conditions, since IbpA can be regarded as a "double-edged sword".
The self-regulation mechanism proposed here can overcome the dilemma that the high abundance of IbpA is necessary in cases of aggregation stress, but an excessive preemptive supply could be detrimental to the cell.
Our analysis of the stem loops in the 5' UTR of the ibpA mRNA revealed that the secondary structures of the mRNA are critical to regulate the translation. How does IbpA couple to the mRNA structure in RNAT for the translation suppression? One possibility is that the oligomeric states of IbpA bind to their own mRNA to suppress the translation, although no RNA-binding motif has been detected in IbpA. Our stem loop variants revealed that SL1/SL2 affect SL3 for the translation initiation, suggesting that SL1/SL2 interact with SL3 to control the translation. In this case, the IbpA oligomers might associate with the RNAT system to suppress the translation. Alternatively, a trans-acting modulator might contribute to the link between RNAT and IbpA.
E. coli has two small Hsps, IbpA and its highly homologous paralog IbpB, which are encoded in the ibpAB operon [11]. Several lines of evidence have shown the distinct roles of IbpA and IbpB, where IbpA and IbpB function as a canonical binder and its noncanonical paralog that enhances the dissociation of sHsps from the co-aggregates, respectively [17,19].
In addition to this distinction, our findings demonstrated another aspect of the difference between the two sHsps in the expression regulation. IbpA suppressed the ibpB reporter translation in the ∆ibpAB strain (Fig. 4, [21]). In contrast, IbpB could not suppress the ibpA translation. Thus, IbpA plays a pivotal role as a master regulator of the expression of sHsps at the translational level, ensuring that IbpA and IbpB cooperate to cope with protein aggregation.
The overexpression of IbpB in the wild-type strain increased the IbpA level (Fig. 4). This IbpA induction is interpreted to be due to the IbpA deprivation by hetero-oligomer formation between IbpA and IbpB, implying that, in addition to the aggregation-prone proteins, the factors that can associate with IbpA could trigger its upregulation. Therefore, we suggest the possibility that IbpA plays a pivotal role as a trans-regulator for the expression of other proteins. Indeed, the translation level of ibpB is decreased upon IbpA co-expression [21]. The fact that the ibpB mRNA also has an RNAT in the 5' UTR [20,21] implies that IbpA recognizes a series of stem loops in 5' UTRs, such as RNAT structures.

E. coli strains.
The E. coli DH5α strain was used for cloning. The BW25113 strain was used for each assay.
Primers used for cloning are listed in S1 Table. SDS-PAGE and western blotting. and HRP conjugated anti-rabbit IgG (Sigma-Aldrich) were used as secondary antibodies.
Blotted membranes were detected with an LAS 4000 mini imager (Fujifilm).

Quantitative RT-PCR.
Total mRNA was extracted using Tripure Isolation Reagent (Merck) and treated with recombinant DNase I (Takara). The treated RNA was purified with an RNeasy Mini kit (Qiagen). The samples were prepared using a Luna Universal One-Step RT-qPCR Kit (New England Biolabs). Quantitative RT-PCR was performed with an Mx3000P qPCR system (Agilent) and analyzed by the MxPro QPCR software (Agilent). The amount of target mRNA was normalized with the ftsZ mRNA by the ∆∆Ct method [38]. Primers used for PCR are listed in SI appendix Table S1.

Microscopy.
To observe the IbpA localization in cells, we used the E. coli BW25113 wild-type strain carrying pCA24N-rhodanese and the dnaKJ deletion strain. Each strain carrying pBAD30-ibpA 5' UTR-ibpA-gfp was grown to an OD660 of ~0.4 at 37 °C in LB medium. Cells were observed with an inverted microscope IX71 (Olympus) and a mercury lamp with a GFP filter. Fluorescent images were recorded with an iXon DV897 electron multiplying CCD camera (Andor) and the Andor SOLIS software (Andor).

Protein purification.
To purify IbpA, we used the BL21 (DE3) strain carrying pCA24N-ibpA or pCA24N-ibpA_AEA. Cells were grown in LB media at 37 °C to an OD660 of 1.0, and IbpA production was induced with 1 mM IPTG for 2 h. The cells overexpressing IbpA were lysed by sonication (Branson) in buffer A (50 mM Tris-HCl, pH 7.4, 10% glycerol, 1 mM dithiothreitol, 100 mM KCl). After the sonication, we followed the established methods for the purification of the wild-type IbpA [14] and the AEA mutant [31].
Sucrose density gradient assay.
The purified IbpA (10 µM) was incubated for 30 min at 48 °C in buffer B (50 mM Tris-HCl, pH 7.4, 150 mM KCl, 20 mM magnesium acetate, 5 mM dithiothreitol). After the incubation, the samples were applied onto an 11 ml gradient of 10-50% (w/v) sucrose in buffer B and centrifuged, using a Beckman SW41Ti rotor at 35,000 rpm at 4 °C for 80 min. The samples were collected as 20 separate fractions, using a fractionator (BioComp). The fractions were separated by SDS-PAGE and visualized by Coomassie Brilliant Blue staining.

Growth assay.
The E. coli wild-type strain, the ibpAB deleted strain, and the wild-type strain harboring pCA24N-ibpA or pCA24N-gfp were precultured at 30 °C for 16 h in LB medium. The precultured cells were incubated at 37 °C with shaking at 70 rpm, using a TVS062CA incubator (Advantec).

Statistical Analysis.
Student's t test was used for calculating statistical significance, with a two-tailed distribution  coli wild-type (WT) cells under various conditions. E. coli cells were grown at 37 °C or shifted to 42 °C for 10 min. ∆KJ, the dnaKJ deletion strain; agg ++ , E. coli wild-type expressing rhodanese; σ 32++ , E. coli wild-type strain expressing σ 32 . Unless otherwise indicated, E. coli cells were grown at 37 °C. Expressions of GroEL and FtsZ were also examined as controls for a typical Hsp and a constitutively expressed protein, respectively. (B) Ratios of the ibpA mRNA amounts in cells under conditions corresponding to (A). Error bars represent SD; n =3 biological replicates. Student's t test was used to assess the statistical significance of differences (*: p < 0.01). (C) Evaluation of ibpA translation by GFP reporters. The reporters harboring the 5' UTR of ibpA or the 5' UTR from a plasmid without the ibpA sequence were expressed under various conditions. Western blotting was performed using anti-GFP and anti-FtsZ antibodies.         WT, E. coli wild-type; SerA++, E. coli wild-type expressing SerA; σ32++, E. coli wild-type expressing σ 32 . Expressions of GroEL and FtsZ were examined as controls for a representative Hsp and a constitutively expressed protein, respectively. Western blotting using anti-IbpA, anti-GroEL, and anti-FtsZ antibodies is shown. (C) Evaluation of the ibpA translation initiation by the GFP reporter used in Fig. 1C. E. coli lysates from wild type with (++) or without (-) the SerA induction by 0.1 mM IPTG for 1 h were analyzed. Western blotting using anti-GFP and anti-FtsZ antibodies is shown. S1