key: cord-0775439-zsqh8ziq
authors: Chattopadhyay, Gopinath; Bhowmick, Jayantika; Manjunath, Kavyashree; Ahmed, Shahbaz; Goyal, Parveen; Varadarajan, Raghavan
title: Mechanistic insights into global suppressors of protein folding defects
date: 2021-11-19
journal: bioRxiv
DOI: 10.1101/2021.11.18.469098
sha: a1f7e16541b14e1caf2064a2ca0800c0dda685bc
doc_id: 775439
cord_uid: zsqh8ziq

Most amino acid substitutions in a protein either lead to partial loss of function or are near neutral. Several studies have shown the existence of second-site mutations that can rescue defects caused by diverse loss of function mutations. Such global suppressor mutations are key drivers of protein evolution. However, the mechanisms responsible for such suppression remain poorly understood. To address this, we characterized multiple suppressor mutations both in isolation and in combination with inactive mutants. We examined five global suppressors of the bacterial toxin CcdB, the known M182T global suppressor of TEM-1 β-lactamase, the N239Y global suppressor of p53-DBD and three suppressors of the SARS-CoV-2 spike Receptor Binding Domain. The suppressors both alone, and in conjunction with inactive mutants, stabilise the protein both thermodynamically and kinetically in-vitro, predominantly through acceleration of the refolding rate parameters. When coupled to inactive mutants they promote increased in-vivo solubilities as well as regain-of-function phenotypes. Our study also demonstrates that the global suppressor approach can be used to consistently stabilise wild-type proteins, including for downstream translational applications.

Certain specific interactions influencing a protein's structure and function are preserved throughout evolution, either by conservation of the interacting pair, or by mutation of the pair in a correlated fashion (1). Correlated mutational data extracted from multiple sequence alignments, can guide structure prediction of a protein (2) or help in the determination of interfacial residues responsible for binding to various partners (3) . Experimental studies have often led to the identification of second suppressor mutations which can alleviate the defects in folding, function and stability of the protein caused by an initial deleterious mutation. Such compensatory mutations, often occur within the same gene, highlighting the role of intragenic suppression in evolution (4) .

Suppressors can be either spatially proximal or distal to the inactivating mutation. Distal suppressors are typically able to suppress multiple inactivating mutations, not necessarily in proximity to each other. Hence distal suppressors are often referred to as global suppressors. Prior studies on distal suppressors have shown that such suppressors function either by (a) increasing global thermodynamic stability (5), (b) enhancing the activity of the native protein without any effect on stability (6) , or (c) improving the folding of the native protein without substantial enhancement of the thermodynamic stability (7) . In laboratory-based evolution experiments, it has been shown that the evolution of new function is accompanied by second-site compensatory mutations, which compensate for the probable destabilizing function altering mutations (8) . Previous studies have primarily focused on thermodynamic rather than kinetic effects of mutations on protein stability and function, though the latter may be more relevant in vivo (9) .

Here, we endeavour to provide insights into the mechanisms responsible for global suppression. The primary experimental system utilised is a 101-residue homo-dimeric protein CcdB (Controller of Cell Death protein B), which is a part of the toxin-antitoxin (CcdB-CcdA) module, involved in the maintenance of F-plasmid in E. coli (10) . We probe effects of multiple global suppressors of CcdB, on the folding kinetics, stability, in vivo solubility and in vivo activity of the protein. The suppressors are able to rescue folding defects of the inactive mutants, through thermodynamic and kinetic stabilization, with the largest effects on the refolding rate. We also determined the structure of three of the CcdB suppressors mutants, S12G, V46L and S60E using X-ray crystallography to understand the structural basis of stabilization. We probed the effect of the suppressor mutants on aggregation, binding and thermal tolerance. Molecular Dynamics (MD) simulations were employed to investigate changes in flexibility of the CcdB mutants at 293 K and 313 K relative to the WT. We also examined the effects on stability and folding of two known global suppressors: M182T of TEM-1 β-lactamase, an extended spectrum β-lactamase (ESBL) enzyme conferring antibiotic resistance against third generation cephalosporins and N239Y in the DNA binding domain (DBD) of p53, a critical tumour suppressor protein, which is known to suppress the effect of oncogenic inactive mutants. We further examined the effects of additional suppressor mutations that were recently isolated for CcdB and RBD (Receptor Binding Domain of spike glycoprotein of SARS-CoV2), on protein stability and folding. The analysis of data for these diverse systems revealed insights into the mechanism of action of global suppressors.

In order to achieve tuneable expression levels using arabinose (inducer) and glucose (repressor), the ccdB gene was cloned under control of the pBADAraC promoter in the pBAD24 vector (11) . A previously described single site saturation mutagenesis (SSM) library of ~1600 mutants includes several single mutants that lead to partial or complete loss of function (9, 12, 13) . Five such mutants, termed parent inactive mutants (PIMs), (namely V5F, V18W, V20F, L36A and L83S) were chosen for further study. Next, exhaustive second-site saturation mutagenesis libraries were constructed by incorporating each of the PIMs into the original SSM library. PIMs bound GyraseA poorly. In order to identify the suppressors of the PIMs, suppressor libraries were then displayed on the yeast surface and screened for improved binding to GyraseA by FACS (2) . In the study, two residues R10 and E11 on an exposed loop region were also identified as distal suppressors ( Figure 1A ). The global suppressor R10G was able to suppress defects at all the five positions and was also shown to increase the Tm by 8 °C, relative to WT CcdB (2) but E11R was not characterized. Recently, another possible global suppressor S12G was identified (14) . S12, located on an exposed loop, along with R10 and E11, is involved in hydrogen bonding with the cognate antitoxin CcdA ( Figure 1A ). The aim of the present study was to experimentally validate whether these mutations (S12G, E11R) also stabilise WT CcdB and to understand possible mechanisms responsible for the suppressor phenotypes. The study involves characterization of these putative global suppressors both individually and in the context of several known inactive mutants.

In vivo activity and solubility of the inactive mutants in the background of suppressors: The phenotype of the CcdB mutants were studied after transforming the plasmids individually into the Top10pJAT strain which is sensitive to CcdB toxin action. The cells were allowed to grow at different concentrations of glucose and arabinose (9, 13) to allow increasing levels of expression of CcdB from the PBAD promoter. Mutant phenotypes were studied as a function of varying repressor (glucose) and inducer (arabinose) concentration. For WT and the fully active mutants E11R and S12G, the cells fail to grow even at the highest glucose (repressor) concentration (0.2% glucose). On the other hand, the inactive mutants grow even at the highest concentration of arabinose (0.2% arabinose). However, in the background of the E11R or S12G mutations, with the exception of V18W-S12G, the WT phenotype of the remaining (PIM, suppressor) pairs is restored at lower ara concentrations (≥ 0.02%), leading to cell death ( Figure 1B ).

For proper function, proteins need to fold into their native structure avoiding misfolding or formation of aggregates. To examine the relative solubility levels, E. coli strain Top10GyrA was individually transformed with each mutant. The solubilities of the PIMs were significantly lower than that of WT CcdB and suppressors E11R and S12G (Table S1 ). However, the solubilities of the inactive mutants in the background of the suppressors was significantly enhanced ( Figure S1A ). These results reveal that lowered activity and decrease in solubility of the inactive proteins are both improved in the background of the suppressors.

The WT and CcdB mutants were purified by affinity purification against immobilised ligand CcdA.

Yield for all mutants varied from 0.3-12 mg/L depending upon the amount of protein in the soluble fraction. V5F and V5F-S12G could not be purified because of their low expression, solubility and inability to bind to the CcdA column. V18W-E11R and V20F-E11R could not be used for further biophysical studies owing to their high tendency to aggregate.

The purified proteins (4 μM) were subjected to thermal denaturation using nanoDSF and the apparent Tm was calculated ( Figure 1D ). S12G had a 2 °C higher Tm, indicating that the mutation is stabilising (Table S1 ). Further the inactive mutant-suppressor pairs showed increased apparent thermal stabilities (~5-12 °C) relative to the inactive mutants. The ability of the purified proteins to bind to CcdA peptide (8 μM) was also examined by monitoring thermal denaturation using binding of Sypro orange dye (15) , both in the absence and presence of CcdA peptide . Relative to the free proteins, apparent Tm's of the CcdA-bound complexes showed increments (Figure S1B) due to stability-enhancements of the proteins in the presence of peptide (2) .

Equilibrium unfolding experiments for the CcdB mutants were carried out by nanoDSF. The data were fitted to N2↔2D unfolding models for the homo-dimeric CcdB. The fraction unfolded of different CcdB mutants (200 mM HEPES, pH 8.4) in presence of GdnCl is plotted as a function of denaturant concentration ( Figure 1C ). The midpoint of chemical denaturation (Cm), ΔG⁰ and m-values were measured for all the CcdB mutants (Table S5 ). The suppressor S12G alone is 0.8 kcal/mol more stable than the WT, and the double mutants are apparently 2.5-5.0 kcal/mol more stable than the inactive mutants ( Figure 1D ). The significant difference in the apparent effect of the suppressor in the context of PIM relative to WT, is likely due to the high tendency of inactive mutants to aggregate over time thereby reducing the amount of functionally folded form. The E11R is 3.1 kcal/mol less stable than the WT protein. This is surprising because their thermal stabilities are similar and E11R is able to enhance the stability of multiple PIMs by 4-5 kcal/mol. Being a charge reversal mutation, the pI of the protein was changed significantly (6.1 to 8.1) which accounts for the visible aggregation of the E11R mutant at 200 mM HEPES, pH 8.4. Hence, the chemical denaturation and folding kinetics of E11R and WT were repeated in PBS (pH 7.4) where E11R behaves similar to the WT protein ( Figure S1 C-H, Table S4 ).

Refolding and unfolding kinetics for CcdB mutants were also monitored by time-course fluorescence spectroscopy at 25 °C using nanoDSF. Refolding was performed at pH 8.4, at final GdnCl concentrations ranging from 0.5 M-1.5 M. During refolding, the two monomers rapidly associate in a diffusion limited process (16) . Refolding for the WT occurs with a fast and a slow phase as observed earlier (16) . The suppressors E11R and S12G alone refold at a faster rate compared to the WT. Relative to the WT, the inactive mutants refold slowly, whereas their refolding rates were enhanced (both fast and slow phase) in the background of the suppressor (Figure 2A , Table S2 ).The fast phase of refolding for L83S-E11R could not be captured owing to its fast refolding kinetics. The unfolding trace of WT CcdB when fitted to the three-parameter unfolding equation, gives a fitted unfolding rate of 0.05 s -1 . S12G shows a much slower rate of unfolding. For all the PIMs ,we observed very fast unfolding even at low GdnCl concentrations whereas the mutant-suppressor pairs had slower unfolding rates relative to the individual PIMs, though typically faster than the rates for WT and the suppressors E11R and S12G ( Figure 2B , Table S2 ). Refolding and unfolding reactions were carried out at three different GdnCl concentrations and the observed rate constants and m values were plotted as a function of GdnCl concentration ( Figure S2 , Table S3 ).Using this, the refolding rate constants for both fast and slow phases and unfolding rate constants were calculated at 0 M GdnCl as described previously (17) for relative comparison (Table S3) . These experiments correlate with other results and indicate that the suppressors E11R and S12G have similar to marginally higher thermal stabilities than WT and that all the mutant-suppressor pairs are both kinetically and thermodynamically more stable than the corresponding PIMs.

The affinity of fluorescently labeled Gyrase to CcdB mutant proteins (native, native in GdnCl and refolded proteins) was analysed using MST. The assays were carried out with a fixed concentration of the labeled Gyrase (70 nM), which was titrated with different concentrations of the unlabeled CcdB mutants. The obtained data indicates that WT CcdB, S12G and E11R bind GyraseA14 with KD's of about 2.2, 2.3 and 1.2 nM respectively which is consistent with the SPR binding studies ( Figure 2C , Table S1 ). Labeled Gyrase also bound with similar affinity to the native and refolded proteins in 1.5 M GdnCl ( Figure 2C , S3B-C Table S1 ), indicating that refolding was reversible. The Cm of Gyrase was determined to be 4.48 M, confirming that it was folded at 1.5 M GdnCl used in the above refolding assay ( Figure S3D ).

The refolded CcdB proteins and the native proteins in the presence of 1.5 M GdnCl were also subjected to thermal denaturation, and the apparent Tm was calculated ( Figure 2D , Table S5 ). Except for the PIMs V18W and V20F, all the other mutants showed clear thermal transitions confirming that they were in a folded conformation in the presence of GdnCl. The suppressors improved the Tm of the refolded proteins, relative to the PIMs.

The binding of purified CcdB mutants to their target Gyrase, was also probed using SPR. WT CcdB, S12G and E11R bind to Gyrase with K D's of about 1.4, 2.6 and 2.8 nM respectively ( Figure S4 , Table   S1 ). An increased affinity for DNA Gyrase for the inactive mutant-suppressor pairs was observed in all the cases. This explains the ability of these suppressors to rescue the activity of the inactive mutants, which are defective in binding and poisoning DNA Gyrase. The apparent low affinity of these inactive mutants may also arise due to the inability to estimate the fraction of active protein for these purified mutants. The SEC profile of the PIM L36A, shows a significant amount of aggregation as well as degradation as compared to the WT, suppressor S12G and the double mutant L36A-S12G ( Figure 2F ).

For the other mutants, due to poor yields and high tendency to aggregate, SEC was not performed.

Thermal tolerance of the mutants E11R, S12G, L36A, L36A-E11R,L36A-S12G and WT CcdB, was also assessed by determining the binding of CcdB proteins to Gyrase after prolonged heat stress followed by cooling to room temperature ( Figure 2E , Table S5 ). The S12G mutant retained 4% activity after incubation at 80 °C for 1hr, representing a five-fold improvement over WT, whereas the double mutants L36A-E11R and L36A-S12G showed three-fold and ten-fold improvement over WT respectively. Surprisingly, L36A-E11R had residual activity at 40 °C, as compared to E11R alone.

The S12G was experimentally found to be thermally stable compared to the WT protein. To obtain further insights, WT CcdB and the mutants E11R, S12G, L36A, L36A-S12G were simulated as biological dimers for 50 ns at temperatures of 293K and 313K. The root mean square deviation (RMSD) calculations of the Cα atoms of the structures (dimer), from the starting reference structures, indicated greater deviations in WT and L36A, both at 293K and 313K compared to S12G, E11R and L36A-S12G ( Figure 3A) indicating features of instability. The PIM L36A had two distinct clusters at both the temperatures as observed from the RMSD distributions whereas the WT showed two distinct clusters at 313 K, while the S12G, E11R, L36A-S12G exists as a single cluster ( Figure 3B ). The S12G, E11R and L36A-S12G structures cluster at a lower RMSD compared to WT and L36A at both temperatures pointing to the overall dynamic stability of S12G, E11R and L36A-S12G structures. The WT and L36A structures at 313K also showed higher RMSF (root mean square fluctuation) around the residues 9-11, 41-50, 56-61 ( Figure 3C ). The pronounced difference in WT and L36A as compared to S12G, E11R, L36A-S12G (313K), points to the possibility of stabilization of distant regions specifically loop 40-60 which is less flexible in S12G, E11R and L36A-S12G ( Figure 3D -E).

Thermodynamic and kinetic stabilisation mediated by the M182T global suppressor substitution in extended spectrum TEM-1 β lactamases conferring antibiotic resistance TEM-1 β-lactamase, a 263 residue monomeric enzyme confers resistance to β-lactam antibiotics (18) .

There have been several studies on its stability and folding kinetics making it a suitable system to study the effect of global suppressor substitutions (19, 20) . M182T found frequently in clinical isolates in combination with other mutations (R164, E104, M69I, R244) is far from the active site but it restores the stability defects caused by the substitutions near the active site, resulting in increased protein expression and thereby conferring drug-resistance in clinical isolates, resulting in extended spectrum β-lactamase (ESBL) . It was previously shown that M182T was able to rescue the folding defect in a naturally isolated variant M69I, thus conferring resistance to inhibitors such as clavulanate (Inhibitor-resistant TEM β-lactamases, IRTs) as well as in a core engineered substitution L76N, a position which has been shown to be sensitive to substitutions (7, 22) with M182T being distant from the primary mutation ( Figure 4A ).

In the present study, we characterised the global suppressor M182T and observed its effect on stability and folding of known inactive mutants ( Figure 4 ). We determined the MIC and IC90 of the TEM-1 WT, M182T, M69I, M69I-M182T, L76N and L76N-M182T for both ampicillin and cefotaxime, a third-generation cephalosporin ( Figure 4B -C, Table S6 ). In line with a previous study, we found that the M182T suppressor alone and the M69I-M182T double mutant have comparable values to that of the WT (23) and M69I respectively (24) , however M182T rescues the activity of the inactive enzyme mutant L76N (Figure 4B-C Table S6 ) (22, 25) . Further the activity of the purified mutants was monitored in vitro with nitrocefin ( Figure 4D ). The results obtained showed similar effects of the M182T substitution on WT or M69I or L76N background as observed in vivo ( Figure 4D and S5C-E).

The M182T suppressor substitution enhanced both the thermal and chemical stability of the WT protein as well as the inactive mutants M69I and L76N (Figure 4E-F, Table S6 ), in agreement with previously published results (26) . When compared with WT, M69I and L76N are less stable ( Figure   4G ). The suppressor alone is 6 °C, 1.8 kcal/mol more stable than the WT, and M69I-M182T and L76N-M182T are 7 °C, 2 kcal/mol and 8 °C, 1.7 kcal/mol more stable than the inactive mutants M69I and L76N respectively ( Figure 4G ).

Refolding (in 0.5 M GdnCl) and unfolding (in 2.5 M GdnCl) kinetics for all the mutants (5 µM) were also monitored using nanoDSF, in 10 mM HEPES, 300 mM NaCl, 10% glycerol, pH 7.0, at 25 °C.

The unfolding trace of WT TEM-1 when fitted to the three-parameter unfolding equation, gives a fitted unfolding rate of 0.03 s -1 . M182T shows a slightly slower rate of unfolding whereas the inactive mutants M69I and L76N, had two and three fold faster unfolding rates respectively as compared to the WT. The mutant-suppressor pairs however had a slower rate of unfolding as compared to the individual inactive mutants ( Figure 4H and S5B, Table S6 ). The refolding rate constants for each of these mutants, were also calculated for both fast and slow phases. The M182T suppressor alone, refolds at a faster rate compared to the WT. The inactive mutants showed slower refolding kinetics than the WT. In the background of the suppressor, however the inactive mutants refold at a faster rate (both fast and slow phase) ( Figure 4I and S5A, Table S6 ).

The refolded TEM-1 proteins and the native proteins in the presence of 0.5 M GdnCl were also subjected to thermal denaturation, and the apparent Tm was calculated ( Figure 4J , Table S6 ) which was similar in both the cases. Except for the inactive mutant L76N, all the other mutants showed a proper transition indicating that they were in a folded conformation in the presence of GdnCl and the Tm of the refolded suppressors were also higher than the corresponding inactive mutants ( Figure 4J ). Further, lactamase activity of native and refolded proteins, both in 0.5 M GdnCl, checked by nitrocefin hydrolysis yielded results consistent with the other studies ( Figure 4D and S5C-E).

These experiments indicate that the M182T suppressor alone and in the mutant-suppressor pairs (M69I-M182T and L76N-M182T) rescues the folding defects of inactive mutants and confers higher thermodynamic and kinetic stability relative to the inactive mutants. This is in contrast to a previous study (7) which indicated that the L76N-M182T double mutant was destabilized relative to the L76N inactive mutant, despite the fact that the double mutant showed higher activity in vivo. We show that the major contribution to stability mediated by the M182T suppressor substitution is a high refolding rate which allows reversible refolding, even in the background of the inactive mutants.

The transcription factor, p53, acts as a tumor suppressor with multiple anti-proliferative functions upon activation in response to various upstream cellular stress signals (27) Owing to the low thermodynamic stability (28) of its core , DNA Binding Domain (DBD), nearly half of all the human cancers are associated with inactivated p53 proteins, where the mutations primarily located in this domain, destabilise or distort the DBD ( Cañadillas et al., 2006) . The basis of such oncogenic mutations can be structural or functional (30) . Previous studies have identified second-site suppressor mutations in the DBD which could restore the WT p53 functionality (31) . One such suppressor mutation, N239Y in the L3 loop of the DBD, could globally rescue multiple missense mutations located at varying regions of the protein by thermodynamic stabilisation of the inactivated core (28, 30) . The activity of two of the destabilising oncogenic mutations, located in the core of the DBD, V143A and V157F, have been restored by N239Y- (27, 28, 30, 32) .

In the current study, we aimed to obtain mechanistic insight into the N239Y-mediated suppression of the inactivated p53 mutations, V143A and V157F ( Figure 5A ). The double mutants V143A-N239Y and V157F-N239Y showed enhanced expression in the soluble fraction in vivo in E. coli, relative to their corresponding inactive mutants. WT and N239Y single suppressor mutant showed comparable soluble expression and yield ( Figure 5B ). The inactive mutants, V143A and V157F, owing to their low expression levels and aggregation-prone natures, could not be purified or characterised. N239Y, in isolation and in conjunction with the inactive mutants, marginally enhanced the thermal and chemical stabilities of the WT p53 ( Figure 5C -D, Table S7 ). Relative to the WT protein, the suppressor N239Y alone, and the mutants, V143A-N239Y, V157F-N239Y enhance the apparent thermal stabilities by ~1.3 °C, 1 °C and 3 °C respectively. The N239Y, V143A-N239Y and V157F-N239Yhave marginal increments in their thermodynamic stabilities over the WT protein in the range of ~ 0.1 -0.4 kcal/mol ( Figure 5E ).

Refolding (in 2 M urea) and unfolding (in 4.4 M urea) kinetics for the WT and mutants were monitored using nanoDSF at 15 °C. The unfolding traces, fitted to five-parameter unfolding equations, yielded comparable slow-phase unfolding rates for the WT and N239Y proteins (0.003s -1 and 0.0025 s -1 respectively), whereas the unfolding rates for the double mutants were slightly lower than that for the WT (0.0017 s -1 for V143A-N239Y and 0.0019 s -1 for V157F-N239Y) ( Figure 5F and S5G, Table S7 ).

The fast-phase unfolding rates for the single and double mutants were marginally lower than that for the WT (Table S7 ). The refolding traces, also fitted to five-parameter equations, yielded slow-phase refolding rates which were remarkably increased by ~20 and ~10 folds for N239Y and V143A-N239Y respectively, relative to the WT. V157F N239Y refolded with a similar slow-phase rate constant, when compared with the WT. With respect to the fast phase , V157F-N239Y refolds faster than the WT by ~ 3.5 folds, whereas the N239Y and V143A-N239Y refold with similar or marginal increments when compared with the WT ( Figure 5G and S5F, Table S7 ).

The thermal denaturation of the refolded p53 proteins, along with the native proteins in 0.5 M urea as controls were also studied ( Figure 5H ). All the mutants showed a proper transition indicating that they were in a folded conformation in the presence of Urea. ~ 3 °C increment was observed for the apparent Tm of refolded V157F-N239Y relative to that of the refolded WT, whereas the apparent Tm's for the refolded proteins of N239Y and V143A-N239Y were similar to that for the refolded WT ( Figure 5H , Table S7 ).

These observations indicated that the N239Y suppressor mutation likely rescues the inactivated destabilised p53 core by marginal enhancement of the thermodynamic and more importantly the kinetic stability of the proteins containing the suppressor, with the largest effect on the refolding rates.

In order to further, investigate the role of suppressors in the protein stability in the WT background, we performed detailed thermodynamic and kinetic studies of the suppressors alone in CcdB and mRBD proteins ( Figure 6 ).

In a separate study from our lab, along with S12G, several other suppressor substitutions were also identified from yeast surface display (14) . In the present study, we selected three such suppressor substitutions, Y8D, V46L and S60E with ΔTm˃3 °C ( Figure 6A ). The purified proteins were subjected to chemical denaturation ( Figure 6B ). The suppressors Y8D, V46L and the suppressor S60E were ~3 kcal/mol and ~4 kcal/mol respectively more stable than the WT ( Figure 6C , Table S8 ). The suppressors were also subjected to unfolding (in 4.5 M GdnCl) and refolding (in 2 M GdnCl) kinetic studies. The unfolding rates of the suppressors were 2-2.5 times slower than the WT ( Figure 6D and S6B, Table   S9 ), whereas the fast and slow phase refolding rates of the suppressors were 2-5 times and 9-14 times faster than the WT respectively ( Figure 6E and S6A, Table S9 ). Further, the proteins were refolded in 1, 2, 3 and 4 M GdnCl and subjected to thermal denaturation. Native protein at the same GdnCl concentrations was used as control ( Figure 6F and S8). The WT could refold back till 2 M GdnCl, whereas the suppressors Y8D, V46L and suppressor S60E could refold even at 3 and 4 M GdnCl respectively ( Figure S6E-H) . In all cases, the refolded proteins have a broader transition than the native proteins in GdnCl, likely due to the formation of aggregates during refolding.

Next, we investigated the role of the suppressor substitutions in the context of the receptor binding domain (RBD) of SARS-CoV-2 (33) . We recently also identified three suppressors of folding defective mutants in this protein. These suppressors, D389E, L390M and P527I when introduced into WT mRBD have ΔTm≥3 °C ( Figure 6G ). The purified suppressors were subjected to chemical denaturation ( Figure 6H ). In a separate study we have seen that WT mRBD unfolds via a three-state transition during chemical denaturation. In the present study, the suppressors D389E, L390M and P527I were ~0.2-0.4 kcal/mol more stable than the WT for the intermediate to unfolded state (I-U) transition. For the native to intermediate state (N-I) transition, the suppressors D389E, L390M and P527I were ~0.4-0.7 kcal/mol more stable than the WT ( Figure 6I , Table S8 ). The suppressors were also subjected to unfolding (in 3 M GdnCl) and refolding (in 0.5 M GdnCl) kinetic studies. The unfolding rates of the suppressors were marginally slower than the WT for both fast and slow phases ( Figure 6J and S6D, Table S9 ), whereas the refolding rates of the suppressors were 2.5-7 times faster than the WT ( Figure   6K and S6C, Table S9 ). Further, the proteins were refolded in 0.5, 1, and 2 M GdnCl and subjected to thermal denaturation with native proteins in the same GdnCl concentrations as control ( Figure 6L ).

The WT could refold back till 1 M GdnCl, whereas the suppressors D389E, L390M and P527I refolded back till 2 M GdnCl ( Figure S6I -K). The binding of the native proteins, native proteins in 0.5 M GdnCl and refolded proteins in 0.5 M GdnCl with ACE2-hFc neutralizing antibody were also measured using ProteOn ( Figure 6M ). All the refolded proteins showed binding to the ACE2-hFc, indicating that the proteins were properly refolded back to their functional conformation ( Figure S6L -O).

The structure of the S12G, V46L and S60E mutants of CcdB was solved to resolutions of 1.63 Å, 1.35 Å and 1.93 Å respectively. (Figure 7 ). For further validation, the crystals were dissolved in water and the mass was confirmed by ESI-MS.

The structures of S12G, V46L and S60E ( Figure 7A , 7D, 7G) consist of a single chain in the asymmetric unit, with two chloride ions. One of the ions, which is also present in the WT structure . Although there is a water molecule at this position in the wild type structure (3VUB), addition of a water in S12G resulted in an unusually low B-factor whereas a Clion fits well without any negative density and a B-factor of 20 Å 2 . The density for the last residue I101 was not visible in the map for S12G. The electron density map in the region of residues 40-45 for S12G, 43-45 for V46L and 41-42 for S60E was very poor, as a result the side chains could not be fitted. One of the residues in S12G, R40 lies outside but close to the allowed region of the Ramachandran Plot. The mutant structures are very similar to the wild type structure (3VUB) with an RMSD of 0.26, 0.39 and 0.39 Å for S12G ( Figure 7B ), V46L ( Figure 7E ) and S60E ( Figure 7H) respectively.

For S12G, the variation is mainly in the loop region between Y8-Y14 and A39-V46, indicated in Figure   7B by a star. There are two water molecules (254 and 228) in S12G in place of the two conformers of S12[OH] of 3VUB ( Figure 7C ). A cluster of water molecules at a hydrogen bonding distance from G12 stabilizes the loop and anchors it via interactions with the backbone atoms of neighbouring residues ( Figure 7C ) reducing the average B-factor in this region ( Figure 7J ). These two water molecules substitute for the hydroxyl group of both the conformers of serine in the WT structure. Sine the S12G has an additional Clion, an additional comparison was done with the structure 4VUB (WT CcdB) which has the second Clion in the same position as found in S12G. It was found that although the absolute B-factor of S12G and 4VUB were similar in the 8-12 region and lower than that of 3VUB, it was lowest for S12G in the region 39-46 amongst the three structures.

For V46L, the loop region A39-V46, exhibits a major deviation from WT CcdB, as indicated in Figure   7E by a star. L46 is involved in a hydrogen-bond interaction with R62 and hydrophobic interactions with M64 ( Figure 7F ). The hydrogen-bond interactions are formed between the main chain oxygen atom of L46 and side chain nitrogen atoms of R62 (NH1, NH2). The average B-factors of the V46L structure are lower than the WT with the most reductions in the loops 8-14 and 39-46 ( Figure 7J ).

For S60E also the major deviation, is in the starred loop region A39-V46 ( Figure 7h ). E60 is involved in a series of salt-bridge interactions with R48, H55 and R62 ( Figure 7I ). The salt bridge interactions are formed between the side chain oxygen atoms of E60 (OE1 and OE2) and side chain nitrogen atoms of R48 (NE, NH1), H55 (ND1, NE1) and R62 (NE). There is a change in the orientation of the mobile R48 side chain resulting in salt bridge interactions with E60 ( Figure 7I ). The S60E mutation has resulted in reduced B-factor differences between the side chain and main chain in many regions, including R48, resulting in overall stabilisation of the structure. The average B-factors of the S60E structure are also lower than the WT with the most reductions in the loops 8-14 and 39-46 ( Figure 7J ).

In this work we examine the mechanisms by which a second (suppressor) mutation alleviates the protein defects caused by the initial loss of function causing point mutant ( Figure 8 ). The studies have shown physically interacting residues to coevolve (34) . For a pair of residues in contact, the compensation of stability caused by a defect at one position occurs by shape or charge complementarity at a nearby site in the protein core (2) . The location of suppressor mutations may be either spatially proximal or distal from the site of the original mutation. Unlike their proximal counterparts, distal suppressors are generally found on the surface of the protein (35) , and are expected to have a WT like phenotype, when present as single mutants (2) . Global suppressors are versatile mutations that can rescue the protein from damage caused by a number of non-functional point mutants, located in diverse regions of the structure. Some plausible mechanisms responsible for global suppression are: a) improving the foldability of the protein without impacting stability (7) A recent deep sequencing analysis of the SSM library in the background of different PIMs, revealed the presence of several other putative global suppressors, including S12G. In the present study we characterised the mechanisms of suppression by E11R and S12G in considerable detail, and extended these studies to three other global suppressors, Y8D, V46L and S60E. Non active-site, buried mutations typically affect the levels of correctly folded protein (40) . The presence of low levels of active, folded CcdB protein is sufficient to kill the cells and rescue the inactive phenotype caused by the PIMs. This was confirmed in vivo by growth assays and estimation of solubility levels. Equilibrium thermal and chemical denaturation studies reveal a large enhancement in the apparent stability of PIMsuppressor proteins with respect to the PIMs in isolation. These observations might explain a part of the suppression mechanism by the compensation of stability defects, as stated above (5, 41) .

Surprisingly, addition of the suppressor mutation E11R does not improve the stability of WT CcdB, and the S12G mutation shows a marginal improvement in stability. There is however, a surprising nonadditivity of the apparent stabilising effect of E11R in the presence and absence of the PIM. This might be attributed to the fact that the stability of the PIMs are difficult to measure accurately because of their aggregation-prone nature.

Kinetic studies were used to further elucidate the mechanism of action of such distal global suppressors. For a protein to remain functional on biologically-relevant time scales, apart from thermodynamic stability, kinetic stabilization plays a significant role (42) . Several studies have shown the importance of kinetic stability in the evolutionary optimization of the protein function (43) (47, 48) and in detailed studies of the relation between genotype and phenotype (9) . Kinetic destabilisation leads to diseases associated with protein misfolding (49, 50) . Therefore, a mutation which enhances kinetic stability is indeed important in both physiological and pharmaceutical contexts, for example increasing shelf life of monoclonal antibodies (Ionescu et al, 2008 ). The refolding rates for the suppressor pairs are enhanced significantly compared to their PIM counterparts. A dramatic decline in the rates of unfolding is clearly visible for E11R and S12G suppressors in conjunction with the PIMs, when compared to the PIMs individually. In vitro studies confirm the suppressor containing mutants refold back to their native conformations, as is evident from the comparable binding affinities to Gyrase, both in the native and refolded states. It was also observed that the M182T substitution in case of the TEM-1 β-lactamase and the N239Y substitution in case of the p53-DBD were able to rescue the folding defect of inactive mutants, largely through thermodynamic and kinetic stabilisation. In addition, several novel suppressors of the SARS-CoV-2 RBD were characterised and shown to stabilise the protein both thermodynamically and kinetically.

Relative to the WT protein, the individual suppressors have a faster refolding rate whereas unfolding rates differ marginally. The ability to refold back faster is consistently observed across all the systems used in this study. This suggests that additional favourable interactions resulting from the mutation are formed prior to the folding transition state, lowering its energy, relative to the unfolded state. In the case of the CcdB S12G, β-lactamase M182T and p53 N329Y mutants, for which crystal structures are available, additional non-covalent interactions present, relative to corresponding WT structure are shown.

The lower flexibility of the loops in the stabilised CcdB mutants S12G, V46L and S60E as compared to WT structure could be one of the reasons for the mutants being more stable than the WT. The difference seen at the 8-12 region for S12G is largely due the series of hydrogen-bonding interactions formed with the G12 and the water molecules, whereas the difference in structure in the 29-46 region in the V46L is largely due to the additional hydrogen-bond interactions formed with the main chain of L46 and R62 and at the 29-46 region in S60E is largely due to the series of salt-bridges formed with the E60 and R48, H55 and R62.

The exact mechanism responsible for E11R suppressor-mediated structural stabilisation cannot be currently assigned due to the unavailability of structures. Comparative analyses of MD simulations performed on the WT and E11R indicate regions of apparent loop stabilisation from residues 41-50 and 85-101 (involved in the C-terminal Gyrase-binding helix), which can possibly account for the global stabilisation of the E11R-containing mutants. In order to probe into the electrostatics of this charge-reversal stabilising mutation, Tanford-Kirkwood electrostatic calculations have been performed using TKSA-MC server. A significant stabilisation of E11R is observed across a wide pH range from 2 to 12, relative to the WT (data not shown) (52) . Electrostatic potential surface maps generated using the UCSF Chimera package show the appearance of a substantial positive -potential surface patch in the loop from residues 8 -15 harbouring the E11R mutation (Figure S1I-J) (53) . Thus, it can be speculated that the cluster of positive charges in this loop may prevent the formation of misfolded intermediates during folding of the suppressor.

None of the CcdB suppressor mutations are seen in naturally occurring paralogs. The positions 10, 12 and 46 in CcdB are involved in interactions with one of the ligands-the antitoxin CcdA. The loop from residues 9 to 15 ( 9 KRESRYR 15 ) between the two -strands contains four residues with their side chains directed towards the CcdA helix. E11 points away from the CcdA helix and mutating it to arginine will abolish the E11-R15 H-bond within the loop and may alter loop conformation. The presence of R10 and S12 in CcdB, in spite of the enhanced stabilities of R10G and S12G proteins, compared to WT, can be explained by the loss of native H-bonds that R10 and S12 make with CcdA, on mutations to glycine. The S60, though not involved directly in CcdA interaction, along with S12 and V46 has been observed to slower down the rejuvenation process of CcdB poisoned Gyrase by CcdA (data not shown).

In the present work, we focus on protein folding kinetics studied in vitro whereas in vivo, for several proteins, folding occurs co-translationally and/or is assisted by molecular chaperones (54) . How suppressor mutations affect the kinetics and yield of co-translational or chaperone mediated folding/unfolding is beyond the scope of the present work and it is likely that the different proteins studied make use of different chaperone systems and span both cytoplasmic and periplasmic folding compartments. Nevertheless, it is clear from the present studies that acceleration in folding kinetics in vitro is associated with enhanced yield of active protein in vivo whether in the context of E. coli or yeast (14, 55) . A previous study from our laboratory, which looked at the effects of overexpression of a number of different chaperones on ccdB mutant phenotypes reported the rescue of inactive, folding defective CcdB mutants occurred exclusively in the two E.coli strains overexpressing ATPindependent chaperones (SecB and Trigger factor), and not in the strains overexpressing ATPdependent chaperones that act in the later stages of the folding pathway (9) . Till date, there have been reports on the detailed investigation of the Hsp70 and Hsp90-mediated stability and activity of p53 DBD (56) (57) (58) , the GroEL/ES chaperonin system making transient interactions and inhibiting the folding of -lactamase precursor (59, 60) , and of a strongly bound complex of the GroEL chaperone with the receptor-binding domain of the SARS CoV2 spike protein (61) . However, none of these discuss the effects of suppressor mutations on chaperone mediated folding kinetics.

We have summarized the effect of the suppressor mutations characterized in the present study on various kinetic parameters in the background of both WT and inactive mutants (Table 1) . For the PIM/Suppressor pair analysis, we excluded the p53DBD double mutants since we did not have the corresponding inactive mutants for comparison. We observe that except for one suppressor which has a similar rate constant as the WT for the slow phase of refolding and one PIM-suppressor pair which had a similar burst phase of refolding as compared to the inactive mutant, all other suppressors or PIMsuppressor pairs have an effect on both the refolding and unfolding kinetic parameters. To further delineate the parameters which are most effected by the suppressor mutations we calculate the average fold change of each parameter by the suppressor mutations (Table 1) Values of Pi are summarized in Table 1 . and confirm that the effect of the suppressor mutants is statistically significant for the refolding rates in background of both WT and inactive mutations.

In summary, the present study showed that the suppressors are able to achieve high kinetic and thermodynamic stability even in the background of a stable WT protein such as CcdB (Tm= 66 °C) ( Figure 8 ) and glycosylated and secreted proteins such as a potential COVID-19 vaccine candidate.

Further, the effects of suppressors were primarily on the refolding rate parameters. This was observed across multiple suppressors in multiple proteins, suggesting this to be the primary mechanism through which such suppressors function.

The data presented here consistently shows that the global suppressors result in a small amount of stabilization of the wild-type protein but greatly enhance the foldability of folding defective, loss of function mutants, primarily through an increase in the refolding rate constants and burst phase amplitude. Suppressors promote increased in-vivo solubility and yield, as well as regain-of-function phenotypes. These observations hold true across these diverse protein systems with proteins from both prokaryotes and eukaryotes, including both cytoplasmic and secreted proteins, suggesting these are general results. The results also show that such mutations cannot be easily predicted by existing computational predictors of protein stability effects, and thus genetic screens remain essential to isolate such mutants. The study also demonstrates that the global suppressor approach can be used to consistently stabilise wild-type proteins, including ones for therapeutic and vaccine applications and highlights the importance of such suppressor screening to create molecules that are both thermodynamically and kinetically stable to unfolding and aggregation. Results presented here clearly demonstrate that the primary mechanism by which global suppressors function is by increasing refolding rates and also suggest that mutational effects on folding kinetics rather than thermodynamic stability are important in vivo.

CcdB: The ccdB gene was cloned under the control of PBAD promoter in pBAD24 vector (11) . Two Escherichia coli host strains Top10pJAT and Top10GyrA were used. Top10pJAT is a CcdB sensitive strain and was used for screening the phenotypes. The pJAT8araE plasmid which encodes for the arabinose transporter area was introduced into the TOP10 strains to ensure that in all cells there is approximately equal amounts of arabinose uptake (62) . Top10GyrA is resistant to the action of CcdB toxin and was used for monitoring the expression of mutant proteins. The strain contains a GyrA462 mutation in its genome that prevents CcdB from binding to Gyrase (10).

TEM-1 and p53-DBD: WT and mutant TEM-1 β lactamase with a C-terminal 6xHistidine tag were cloned and expressed under the control of the T7 promoter in the pET-24a vector. The native signal sequence was used for efficient secretion in the Escherichia coli host strain BL21 (DE3) pLysE. WT and mutant p53-DBD genes with N-terminal 6xHistidine tag were cloned and expressed under the control of the T7 promoter in the pET-15b vector. Escherichia coli host strain BL21 Rosetta (DE3) was used for expressing the p53-DBD proteins. mRBD: mRBD WT and mutants were expressed from mammalian cell culture as described previously (33) under the control of the CMV promoter along with a tPA signal sequence for efficient secretion.

CcdB: For single mutants V5F, Y8D, E11R, S12G, V18W, V20F, L36A, V46L, S60E and L83S, as well as for double mutants, V18W-E11R, V20F-E11R, L36A-E11R and L83S-E11R, the ccdB gene was amplified in two fragments with the desired point mutations. The fragments had overlapping regions (introduced during PCR) of 15-20 nucleotides, which were then recombined using Gibson assembly. Amplification was done using Phusion Polymerase from NEB as per the manufacturer's protocol. The double mutants V5F-S12G, V18W-S12G, V20F-S12G, L36A-S12G and L83S-S12G were synthesized by GeneArt (Germany). 

CcdB: WT CcdB and all mutants were expressed from the arabinose promoter PBAD in the pBAD24 vector in the CcdB resistant Top10Gyrase strain of E. coli. The purification of the CcdB mutants were carried out as described previously (17) . Briefly, 500 mL of LB medium (HiMedia) was inoculated with 1% of the primary inoculum and grown at 37 °C until the OD600 reached 0.6. Cells were then induced with 0.2% (w/v) arabinose and grown at 37 °C for 5 hours for WT CcdB, Y8D, E11R, S12G, V46L, S60E and the double mutants V18W-S12G, V20F-S12G, L36A-S12G and L83S-S12G, at 25 and P527I was carried out as described previously (33) . Briefly, proteins were expressed by transient transfection of Expi293 cells and purified by Ni-NTA chromatography. The eluted fractions were pooled and dialysed thrice using a 3-5 kDa (MWCO) dialysis membrane (40mm flat width) (Spectrum Labs) against 1XPBS, pH 7.4 (storage buffer). The eluted fractions were subjected to 15% Tricine SDS-PAGE and the protein concentration was determine by measuring the A280 and using an extinction coefficient of 33850 M -1 cm -1 .

Wild type and mutant CcdB were transformed into E.coli Top10pJAT, grown for 1 hour in 1 ml LB media containing 0.2% glucose (highest repressor level to avoid leaky expression). After 1 hour, the cells were pelleted and glucose was removed by subjecting cells to three washes with 1 ml LB. Finally, equal amounts of cells resuspended in 1 ml of LB media and serially diluted were spotted on seven agar plates (LB agar plates containing 100 µg/mL ampicillin, 20 µg/mL gentamycin) containing various amounts (%) of glucose (repressor) and arabinose (inducer) concentrations (i.e. 2×10 -1 % glu, 4×10 -2 %glu, 7×10 -3 % glu, 0% glu/ara, 2×10 -5 % ara, 7×10 -5 % ara and 2×10 -2 % ara) at 37 °C . Since active CcdB protein kills the cells, colonies are obtained only for mutants that show an inactive phenotype under the above conditions (9) .

CcdB: Solubility levels were monitored for all the single and double mutants in E. coli Top10Gyrase in the presence of 0.2% arabinose as described previously (9) . Cultures were grown in LB media, induced with 0.2% arabinose at an OD600 of 0.6 and grown for 5 hours at 37 °C. Sigmaplot TM for Windows TM scientific graphing software, and the plots were fitted to a two-state unfolding model (N2↔2U). The fraction unfolded for all CcdB mutants was calculated as described (11, 17) and is summarized below. native and denatured baseline, as described previously (9, 16, 17) . Unfolding kinetic traces of fluorescence intensity from 2 to 4 M GdnCl as a function of time for different CcdB mutants in 200 mM HEPES, pH 8.4 were normalized from 0 to 1 between native and denatured baseline, as described previously (16, 17) . For the stable CcdB mutants (Y8D, V46L and S60E as well as WT CcdB) refolding and unfolding was carried out in 2 M and 4.5 M GdnCl respectively in 200 mM HEPES, pH 8.4 at 25

°C and normalisation was done as described above. The data was analyzed using Sigmaplot TM for Windows TM scientific graphing software and plots were fitted to a 5 parameter equation for exponential decay for refolding (y=a0+a1*exp(-kf1*x)+a2*exp(-kf2*x)), and a 3 parameter exponential rise for unfolding (y=A0+A1*(1-exp(-ku1*x))) as described previously (16) Refolding and unfolding kinetic traces were normalized from 0 to 1 between native and denatured baselines. The data for the TEM-1 β lactamase mutants was analyzed using Sigmaplot TM for Windows TM scientific graphing software and plots were fitted to a 5 parameter equation for exponential decay for refolding (y=a0+a1*exp(-kf1*x)+a2*exp(-kf2*x)), yielding slow and fast phase rate constants and a 3 parameter exponential rise for unfolding (y=A0+A1*(1-exp(-ku1*x))) as described above, where x is the time of refolding/unfolding. The data for the mRBD mutants was analyzed using Sigmaplot TM for Windows TM scientific graphing software and plots were fitted to a 3 parameter equation for exponential decay for refolding (y=a0+a1*exp(-kf1*x)), and a 5 parameter exponential rise for unfolding (y=A0+A1*(1-exp(-ku1*x))+a2*(1-exp(-ku2*x))), yielding slow and fast phase rate constants as described above, where x is the time of refolding/unfolding. The data for the p53 DBD mutants was analyzed using Sigmaplot TM for Windows TM scientific graphing software and plots were fitted to a 5 parameter equation for exponential decay for refolding (y=a0+a1*exp(-kf1*x)+a2*exp(-kf2*x)), yielding slow and fast phase rate constants and a 5 parameter equation for exponential rise for unfolding (y=A0+A1*(1-exp(-ku1*x))+a2*(1-exp(-ku2*x))), yielding slow and fast phase rate constants as described above, where x is the time of refolding or unfolding. Gyrase to refolded CcdB mutants in 1.5 M GdnCl (also in 0.1 M GdnCl for V18W, V18W-S12G, V20F and V20F-S12G) as well as native CcdB proteins, not subjected to refolding, in presence of 1.5 M GdnCl (also in 0.1 M GdnCl for V18W, V18W-S12G, V20F and V20F-S12G) was measured. parameters were obtained by fitting the data to a simple 1:1 Langmuir interaction model by using BIA EVALUATION 3.1 software as described previously (9, 65) . Thermal tolerance of the CcdB mutants E11R, S12G, L36A, L36A-E11R and L36A-S12G and WT CcdB using 500 nM of protein was assessed by their ability to bind GyraseA-14 after heat stress. The protein sample was incubated at 40 °C and 80 °C respectively for 1 h in a PCR cycler (BioRad) with a heated lid to prevent evaporation. The samples were cooled to 25 °C and binding affinity to Gyrase was determined by SPR experiments as described above The fraction of active protein following thermal stress was quantitated by measuring the RUs at the end of the association time period relative to those for the same protein incubated throughout at 25 °C. This was designated as Residual activity.

The starting structures for the simulation were 3VUB for WT and the solved structure S12G for the mutant. For the other three mutants E11R, L36A, L36A-S12G, the structure was modelled and then simulated. Before proceeding for simulation, waters were removed from the structure but chloride ions were retained, residues with multiple occupancies were edited to include the conformation with the highest occupancy and any missing side chains were modelled using Coot. In all cases, the dimeric structures were used and were generated by selecting the suitable symmetry equivalent molecule.

Simulations were carried out using Gromacs version 2019.2 (66) with OPLSAA forcefield (67) .

Structures were enclosed in a dodecahedral simulation box in which the distance between the protein and the box was set to 1.2 nm. The box was solvated using a TIP4P (68) water model, protein inside the box was used with standard protonation states and charges of the system were neutralized using Na + or Clions. The PME or particle mesh Ewald (69) was used for treating electrostatics. A distance cut off for Coulombic interaction was 1.0 nm, a Fourier spacing of 0.16 nm with a fourth degree interpolation, while van der waal's cut off was set to 1 nm. In order to remove any short contacts, energy minimization was carried our using a conjugate gradient method and with 1 kJ mol -1 nm -1 as the maximum force before the minimization converges. The minimized structure was subjected to temperature equilibration (NVT) at temperatures of either 293 K or 313 K using Nośe-Hoover (70, 71) thermostat with a coupling time constant τT of 0.4 ps. The protein and non-protein atoms were separately coupled to the temperature baths. Initial velocities at appropriate temperatures were generated using a Maxwell-Boltzmann distribution. The structure after temperature equilibrium was further pressure (NPT) equilibrated for another 100 ps using the Parinello-Rahman barostat (72) µg/mL. The cultures were then incubated at 37 °C with shaking for 24 hrs following which OD600 measurements were carried out on Varioskan Flash (ThermoScientific) using Nunclon delta surface plates (ThermoScientific). and the MIC was determined by recording the lowest concentration of ampicillin or cefotaxime on which no growth was observed. In practice, this was an all or none phenomenon (24) .

IC90 was derived directly from the plate data measurements. Briefly, E. coli plysE) containing the pET24a plasmid that encodes TEM-1 β-lactamase mutants was grown overnight in LB broth with 50 µg/mL kanamycin. Overnight cultures were diluted 1:100 into LB broth with 50 µg/mL kanamycin and incubated for 4 hrs at 37 °C to mid-log phase (OD600 ~0.6). Ten-fold serial dilutions of each culture were made, and 100 µL of each dilution was spread onto LB agar plates containing 100 µM IPTG inducer and 0-4000 µg/mL ampicillin and 0-50 µg/mL of cefotaxime, in a series of twofold increases. After incubation for 24 hrs at 37 °C, colony forming units (cfu) on each plate were counted to calculate the cfu/mL of culture, and IC90 was defined as the concentration of ampicillin or cefotaxime that reduces cfu/mL of culture by ≥90% (22) .

Nitrocefin assay of TEM-1 β lactamase mutants: TEM-1 β -lactamase activity was assayed as described previously (7). Briefly the rate of nitrocefin (50 µM) hydrolysis was observed at 486 nm at 25 °C for 60 minutes in 10 mM HEPES, 300 mM NaCl, 10% glycerol (pH 7.0) using10 nM native protein, native protein in 0.5 M GdnCl and refolded protein in 0.5 M GdnCl. Activity measurements were carried out on Varioskan Flash (ThermoScientific) using Nunclon delta surface plates (ThermoScientific).

immobilized ACE2-hFc: Binding studies of various mRBD mutants with ACE2-hFc neutralizing antibody were carried out using the ProteOn XPR36 Protein Interaction Array V.3.1 from Bio-Rad as described previously (33) . Briefly, following activation of the GLM sensor chip with EDC and sulfo-NHS (Sigma), Protein G (Sigma) was coupled at 10 µg/mL in the presence of 10 mM sodium acetate buffer pH 4.5 at 30 µL/min for 300 seconds until ~3500-4000 RU was immobilized. After quenching the excess sulfo-NHS esters using 1 M ethanolamine, ~1000 RU of ACE2-hFc was immobilized at a flow rate of 5 µg/mL for 100 seconds on various channels except one blank channel that served as the For the studies carried out in GdnCl, the running buffer did not have any GdnCl and the jumps obtained in all the channels were removed after reference subtraction.

The purified mutants of CcdB, S12G, V46L and S60E CcdB, in 1xPBS pH 7. The initial crystals diffracted to ~2.5 Å. Conditions were further optimized to obtain single crystals for better diffraction quality. The best crystals were obtained in the condition 0.15 M calcium chloride dihydrate, 20% w/v PEG 3350 for S12G; 0.20 M calcium chloride dihydrate, 25% w/v PEG 3350, pH 5.1 with 20% glycerol (cryo) for V46L; 0.20 M calcium chloride dihydrate, 10% w/v PEG 3350; pH 5.1 with 20% glycerol (cryo) for S60E, with a protein:precipitant ratio of 1:2 at 18 °C using the hanging drop method.

Diffraction data was collected at 100K using Rigaku FR-X with R-AXIS IV++ detector facility at NCBS/Instem for S12G. For V46L, diffraction data was collected at 100K using XRD2 beamline with Dectris Pilatus-6M detector facility at Elettra synchrotron, Trieste, Italy. For S60E, diffraction data was collected at 100K using Rigaku MicroMax-007HF with mar345dtb detector facility at homesource MBU, IISc. The crystals diffracted to 1.63Å, 1.35 Å and 1.93 Å for S12G, V46L and S60E respectively. Data was processed using iMosflm (73, 74) with an overall completeness of 90.6% for S12G, 100% for V46L and 97.5% for S60E. For S12G, the crystal belonged to the C2 space group, with the unit cell parameters, while the pointless predicted I2. Thus, the data was re-scaled in I2 with respectively. Composite omit maps were calculated around the 12 th , 46 th and 60 th residue for S12G and S60E respectively. The final model was validated using the validation server (https://validate.wwpdb.org). Data processing and refinement statistics are given in Supplementary   Table S10 . The server "https://swift.cmbi.umcn.nl/servers/html/listavb.html" was used for calculating the average B-factors.

Ethics Approval and consent to participate: Not Applicable.

Data availability: The crystal structures of the CcdB mutants have been deposited in PDB with PDB IDs: 7EPG (CcdB S12G), 7EPJ (CcdB V46L) and 7EPI (CcdB S60E). The validation reports of the structures are submitted along with the manuscript. The structures will be available after publication.

The data relevant to the figures in the paper have been made available within the article and in the supplementary information section. Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Prof. Raghavan Varadarajan (varadar@iisc.ac.in). Lead contact for the material availability: Prof. Raghavan Varadarajan (varadar@iisc.ac.in). All unique/stable reagents generated in this study are available from the Lead Contact without restriction.

Competing interests: None. Regenerative Medicine is acknowledged for the use the X-ray crystallography facility for the diffraction data collection for the S12G mutant. For the V46L mutant, data was collected at the Elettra synchrotron facility, for S60E data was collected at home source, MBU, IISc. We acknowledge them all. We thank the beamline staff at the Elettra XRD2 particularly Dr. Raghurama Hegde for beamline support. Access to the XRD2 beamline at the Elettra synchrotron, Trieste was made possible through a grant-in-aid from the Department of Science and Technology, India, vide grant number DSTO-1668.

We thank Dr. Mahavir Singh, (Indian Institute of Science) for providing the BL21-Rosetta (DE3) strain. Transfection of the mRBD was carried out by Mohammad Suhail Khan and Kawkab Kanjo assisted with ProteOn Studies. We thank them both. The assistance of Nonavinakere Seetharam Srilatha is duly acknowledged for the SPR experiments. We thank Arunabh Athreya for helping with the data collection for V46L mutant at the synchrotron facility. We also thank all the members of the RV lab for their valuable suggestions.

All the experiments are carried out in biological replicates (n=2), and the listed errors are the standard error derived from the values obtained for individual replicates. For the nanoDSF and MST measurements, each experiment has been carried out twice, each time with two sets of capillaries (n=4) and the listed errors are the standard error derived from the values obtained for individual replicates. The SPR experiments in Figure S4 have been performed once at each concentration with four different concentrations of the protein and the listed error is the standard error derived from the values at multiple concentrations. The P values for comparing the kinetic parameters, were analysed with a two-tailed Mann Whitney test using the GraphPad Prism software 8.0.0 (* indicates P < 0.05, ** indicates P < 0.005, *** indicates P < 0.0005). Structural superposition of WT (3VUB) and S12G monomers, regions displaying deviation are indicated by ⋆. (C) Network of interactions at the 12th position in S12G. WT structure is shown with a grey backbone. S12 in 3VUB adopts two conformations with partial occupancy, the position of the corresponding hydroxyl group in each conformation is taken up by two water molecules in S12G.

Water molecules directly interacting with G12 is shown in red and the corresponding water in 3VUB parameters. We observed that the mean of the distributions of the values for each of the parameters are significantly higher than 1 for refolding, and lower than 1 for unfolding. P value indicated with *, ** and *** indicates < 0.05, < 0.005 and < 0.0005 respectively. Table S1 . Table S4 ). The error bars wherever shown represent the standard deviation from two independent experiments, each performed in duplicates.

(I-J) Coulombic potential surfaces generated for CcdB WT and E11R using the UCSF Chimera. The coulombic potential surfaces generated were coloured using the colour gradation range from blue (-10 kcal mol -1 ε -1 ) to white ( 0 kcal mol -1 ε -1 ) to red (+10 kcal mol -1 ε -1 ), where ε is the dielectric constant (Table S1 ). Table S7 .

unfolding of mRBD proteins follows biphasic exponential kinetics with burst, fast and slow phases.

Representative Figure 4) . Top: MIC and IC90 for ampicillin and cefotaxime from OD600 and plate data measurements, thermodynamic stability parameters (Cm, ΔG⁰, m), apparent thermal stability (Tm), thermal stabilities of refolded proteins and native proteins in presence of 0.5 M GdnCl (TmRefold,, TmGdnCl), of different TEM-1 β-lactamase mutants 1 . Bottom: Kinetic Parameters for refolding and unfolding of TEM-1 β-lactamase mutants measured in 0.5 M and 2.5 M GdnCl respectively carried out in 10 mM HEPES, 300 mM NaCl, 10% glycerol, pH 7.0, at 25 °C 1 . Figure 7) . Data collection and refinement statistics for CcdB mutants. 

Quantification of the transferability of a designed protein specificity switch reveals extensive epistasis in molecular recognition

Residue proximity information and protein model discrimination using saturation-suppressor mutagenesis

Combining Natural Sequence Variation with High Throughput Mutational Data to Reveal Protein Interaction Sites

The coupon collector and the suppressor mutation: Estimating the number of compensatory mutations by maximum likelihood

A fundamental protein property, thermodynamic stability, revealed solely from large-scale measurements of protein function

Phage lambda repressor revertants. Amino acid substitutions that restore activity to mutant proteins

A secondary drug resistance mutation of TEM-1

β-lactamase that suppresses misfolding and aggregation

Intense Neutral Drifts Yield Robust and Evolvable Consensus Proteins

Molecular Determinants of Mutant Phenotypes, Inferred from Saturation Mutagenesis Data

Cell killing by the F plasmid CcdB protein involves poisoning of DNAtopoisomerase II complexes

Thermodynamic Characterization of Monomeric and Dimeric Forms of CcdB

A rapid, efficient, and economical inverse polymerase chain reactionbased method for generating a site saturation mutant library

Protein model 42 discrimination using mutational sensitivity derived from deep sequencing

Stabilizing proteins through saturation suppressor mutagenesis. bioRxiv

The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability

Homodimeric Escherichia coli Toxin CcdB (Controller of Cell Division or Death B Protein) Folds via Parallel Pathways

Facile measurement of protein stability and folding kinetics using a nano differential scanning fluorimeter

Amino acid sequence determinants of β-lactamase structure and activity

Investigation of the folding pathway of the TEM-1 βlactamase

Βeta-Lactamases As Models for Protein-Folding Studies

Evolution of an antibiotic resistance enzyme constrained by stability and activity trade-offs

Genetic and Structural Characterization of an L201P Global Suppressor Substitution in TEM-1 β -Lactamase

Enzyme Efficiency but Not Thermostability Drives Cefotaxime Resistance Evolution in TEM-1 β-Lactamase

A natural polymorphism in β-lactamase is a global suppressor

Multiple Global Suppressors of Protein Stability Defects Facilitate the Evolution of Extended-Spectrum TEM β-Lactamases

The structural bases of antibiotic resistance in the clinically derived mutant β-lactamases TEM-30, TEM-32, and TEM-34

A global suppressor motif for p53 cancer mutants

Semirational design of active tumor suppressor p53 DNA binding

Solution structure of p53 core domain: Structural basis for its instability

Mechanism of rescue of common p53 cancer mutations by second-site suppressor mutations

Genetic selection of intragenic suppressor mutations that reverse the effect of common p53 cancer mutations

Structures of oncogenic, suppressor and rescued p53 core-domain variants: Mechanisms of mutant p53 rescue

Design of a highly thermotolerant, immunogenic SARS-CoV-2 spike fragment

Can three-dimensional contacts in proteins structures be predicted by analysis of correlated mutations?

A Systematic Survey of an Intragenic Epistatic Landscape

Guanidine hydrochloride denaturation studies of mutant forms of staphylococcal nuclease

Amino Acid Substitutions That Increase the Thermal Stability of the lambda Cro Protein

Multiple Global Suppressors of Protein Stability Defects Facilitate the Evolution of Extended-Spectrum TEM β-Lactamases

Protein dynamism and evolvability

Structural correlates of the temperature sensitive phenotype derived from saturation mutagenesis studies of CcdB

Robustness-epistasis link shapes the fitness landscape of a randomly drifting protein

Protein kinetic stability

Energetic landscape of α-lytic protease optimizes longevity through kinetic stability

Regulation of protein function by native metastability

Core structure of gp41 from the HIV envelope glycoprotein

Structure of influenza haemagglutinin at the pH of membrane fusion

Type I collagen is thermally unstable at body temperature

Unstable molecules form stable tissues

Kinetic stability of Cu/Zn superoxide dismutase is dependent on its metal ligands: Implications for ALS

Dominant role of copper in the kinetic stability of Cu/Zn superoxide dismutase

Contribution of variable domains to the stability of humanized IgG1 monoclonal antibodies

TKSA-MC: A web server for rational mutation through the optimization of protein charge interactions

UCSF Chimera -A visualization system for exploratory research and analysis

Cotranslational protein folding

Prediction of residue-specific contributions to binding and thermal stability using yeast surface display. bioRxiv

Hsp70-and Hsp90-Mediated Regulation of the Conformation of p53 DNA Binding Domain and p53 Cancer Variants

Kinetic partitioning during folding of the p53 DNA binding domain

Hsp70 molecular chaperones are required to support p53 tumor suppressor activity under stress conditions

Transient association of newly synthesized unfolded proteins with the heat-shock GroEL protein

The Escherichia coli heat shock proteins GroEL and GroES modulate the folding of the β-lactamase precursor

Structural and functional comparison of SARS-CoV-2-spike receptor binding domain produced in Pichia pastoris and mammalian cells

Homogeneous expression of the PBADpromoter in Escherichia coli by constitutive expression of the low-affinity highcapacity araE transporter

Quantitative analysis of residual folding and DNA binding in mutant p53 core domain: Definition of mutant states for rescue in cancer therapy

Molecular Determinants of Temperature-Sensitive Phenotypes

Glycosylation of the core of the HIV-1 envelope subunit protein gp120 is not required for native trimer formation or viral infectivity

Evaluation and reparametrization of the OPLS-AA force field for proteins via comparison with accurate quantum chemical calculations on peptides

Comparison of simple potential functions for simulating liquid water

Particle mesh Ewald: An N·log(N) method for Ewald sums in large systems

A unified formulation of the constant temperature molecular dynamics methods

Canonical dynamics: Equilibrium phase-space distributions

Polymorphic transitions in single crystals: A new molecular dynamics method

Overview of the CCP4 suite and current developments

iMOSFLM: A new graphical interface for diffraction-image processing with MOSFLM

Phaser crystallographic software

Crystal structure of CcdB, a topoisomerase poison from E.coli

Refinement of macromolecular structures by the maximum-likelihood method

REFMAC5 for the refinement of macromolecular crystal structures

Coot: Model-building tools for molecular graphics

Features and development of Coot

TEM1 beta-lactamase structure solved by molecular replacement and refined structure of the S235A mutant

Structure of the human p53 core domain in the absence of DNA

Structural basis for the neutralization of SARS-CoV-2 by an antibody from a convalescent patient

Protein-binding assays in biological liquids using microscale thermophoresis

The error bars represent the standard deviation from two independent experiments, each performed in duplicates. (E) Binding of 500 nM WT and mutant CcdB to immobilised GyraseA-14.measured by passing the same concentrations of the analyte (CcdB proteins), after heat stress at two different temperature (40 and 80 °C), followed by cooling back to 25 °C. A room temperature control (25 °C) was also used. The residual active fraction was calculated as described in the materials section. (see Table S5). (F) The SEC profiles of a few of the CcdB mutants are shown. The PIM L36A shows aggregation as well as degradation as compared to the WT and E11R