In situ analysis of osmolyte mechanisms of proteome thermal stabilization

Pepelnjak, Monika; Velten, Britta; Näpflin, Nicolas; von Rosen, Tatjana; Palmiero, Umberto Capasso; Ko, Jeong Hoon; Maynard, Heather D.; Arosio, Paolo; Weber-Ban, Eilika; de Souza, Natalie; Huber, Wolfgang; Picotti, Paola

doi:10.1038/s41589-024-01568-7

In situ analysis of osmolyte mechanisms of proteome thermal stabilization

Article
Open access
Published: 29 February 2024

(2024)
Cite this article

Download PDF

You have full access to this open access article

From

View current issue Submit your manuscript

In situ analysis of osmolyte mechanisms of proteome thermal stabilization

Download PDF

4873 Accesses
17 Altmetric
Explore all metrics

Abstract

Organisms use organic molecules called osmolytes to adapt to environmental conditions. In vitro studies indicate that osmolytes thermally stabilize proteins, but mechanisms are controversial, and systematic studies within the cellular milieu are lacking. We analyzed Escherichia coli and human protein thermal stabilization by osmolytes in situ and across the proteome. Using structural proteomics, we probed osmolyte effects on protein thermal stability, structure and aggregation, revealing common mechanisms but also osmolyte- and protein-specific effects. All tested osmolytes (trimethylamine N-oxide, betaine, glycerol, proline, trehalose and glucose) stabilized many proteins, predominantly via a preferential exclusion mechanism, and caused an upward shift in temperatures at which most proteins aggregated. Thermal profiling of the human proteome provided evidence for intrinsic disorder in situ but also identified potential structure in predicted disordered regions. Our analysis provides mechanistic insight into osmolyte function within a complex biological matrix and sheds light on the in situ prevalence of intrinsically disordered regions.

Protein folding and stability in the presence of osmolytes

Article 01 March 2016

Osmolytes: Key Players in Regulating Protein Aggregation

Proteome-wide solubility and thermal stability profiling reveals distinct regulatory roles for ATP

Article Open access 11 March 2019

Main

Adaptation to acute environmental changes is crucial for organism survival. A key adaptation mechanism under stress conditions is the accumulation of osmolytes, small uncharged organic compounds that regulate osmotic pressure in cells^1,2,3,4. In addition to their ability to control cell water content, osmolytes stabilize lipid membranes⁵ and protect proteins against denaturation^{6,7,8,9,10,11}.

Much remains to be understood about osmolyte mechanisms of action. In vitro studies on monomeric, purified proteins have revealed that osmolytes stabilize proteins against thermal denaturation^1,2,3, but the mechanisms of stabilization are controversial. Models of osmolyte-dependent thermal stabilization include the preferential exclusion model^{8,12,13,14,15,16}, which implies that osmolytes form unfavorable interactions with the protein backbone and render the unfolded state energetically unfavorable. Other models propose effects mediated by an increase in medium viscosity¹⁷ or crowding¹⁸ or protein stabilization via direct osmolyte binding^{Full size image}

The nucleotide-binding domain and substrate-binding domain of DnaK (Fig. 4b) were differentially stabilized by TMAO, betaine, proline and glycerol, with better stabilization of the nucleotide binding domain and the biggest difference for proline and TMAO (Fig. 4c). Notably, the nucleotide binding domain showed a lower thermal stability than the substrate binding domain in the absence of added osmolytes (Fig. 4d), with the difference reduced after the addition of TMAO or proline. We validated these findings using DSF on purified DnaK (Fig. 4e). As in LiP–MS, DSF showed a melting behavior with two transitions in the absence of added osmolytes, and this behavior was lost after the addition of TMAO or proline. Our data show that osmolytes can preferentially stabilize individual domains of multidomain proteins and that this can reduce stability differences between domains.

Direct binding to proteins causes strong stabilization

Protein stabilization could also be affected by direct protein–osmolyte binding, and we observed this effect for known binding partners of betaine, proline and glucose (Supplementary Fig. 4a). The effect was osmolyte specific (Supplementary Fig. 4b), indicating that this set of proteins is not generally better stabilized. For glucose, known monosaccharide-binding proteins were also stabilized (Supplementary Fig. 4c), in line with the monosaccharide binding promiscuity reported for some sugar kinases, for example, fructokinase⁵⁰. Both glucokinase, which phosphorylates glucose, and fructokinase were more strongly stabilized by glucose than by other osmolytes (Supplementary Fig. 4d).

Interestingly, we also observed strong stabilization of RbsK by glucose, which is not known to bind this osmolyte, and to a lesser extent by glycerol (Supplementary Fig. 4d). We therefore asked whether the stabilization of RbsK was due to currently unknown binding of glucose to the ribose-binding site. Melting temperatures of purified RbsK in competition experiments between increasing ribose concentrations and added osmolyte suggest that glucose and glycerol stabilized RbsK through binding to the ribose-binding site because high ribose can compete away their stabilizing effect (Supplementary Fig. 4e). By contrast, stabilization by TMAO rose sharply with increasing ribose, which is known to cause compaction of RbsK⁵¹, and therefore suggests that TMAO acts via a preferential exclusion effect.

We next asked whether osmolytes affected protein thermal stability by altering protein structure more generally, either via currently unknown binding events or other indirect effects such as structure compaction. We focused on the first two temperatures (37 °C and 40.5 °C) in the thermal gradient, at which most proteins are not expected to unfold, and identified proteins with altered proteolytic susceptibility after osmolyte addition. We previously showed that this approach enables the global detection of metabolite–protein binding events (LiP-Smap)⁴¹. More than 50% of stabilized proteins under all conditions (and >90% for trehalose and glycerol) did not show any structural alteration after addition of the osmolyte to the lysate (Supplementary Fig. 4f and Supplementary Data 2). Thus, osmolytes induce structural changes in some proteins, but such changes, including direct osmolyte binding, are not required for stabilization across the proteome.

Role of osmolytes in aggregation

Osmolytes were previously shown to affect aggregation of specific proteins^52,53, but their effects across the proteome are unclear. We therefore investigated the effects of osmolytes on protein aggregation across the proteome. As discussed earlier (Fig. 1d), decreasing protease accessibility in an LiP–MS thermal profiling experiment is likely to indicate protein aggregation. However, LiP data alone cannot distinguish between aggregation and other structural rearrangements that decrease susceptibility to proteolysis. We thus made use of TPP, which measures thermal stability by monitoring protein aggregation.

We performed TPP under the same conditions as LiP–MS and probed the effect of osmolytes on thermal profiles. Insoluble aggregates typically appeared at higher temperatures in the presence of osmolytes than in the absence of osmolytes except for glycerol and proline, which had no effect, but none of the osmolytes prevented aggregation at the highest temperatures (Fig. 5a), when proteins are known to be unfolded^42,47. More than 60% of detected proteins showed agreement between the LiP and TPP experiments for the stabilization effects of all three osmolytes (Fig. 5b, Supplementary Data 3 and Methods). In addition, both TPP and LiP identified TMAO as globally the best stabilizer, followed by glucose (Supplementary Fig. 5a). The good agreement between these two methods indicates that the upward shift in the temperature of aggregation in the presence of osmolytes occurs via stabilization of protein structure.

**Fig. 5: Osmolyte effects on protein aggregation.**

To monitor the effects on aggregation at both near-physiological and high temperatures, we focused on the lowest temperatures (37 °C and 42 °C) and the highest temperatures (69 °C, 72 °C and 76 °C) of the TPP data; for the latter set, most proteins will have lost their native structure. Proline and glucose did not affect protein aggregation. TMAO treatment yielded less aggregation of a few proteins (6/18 proteins at low/high temperatures; Supplementary Fig. 5b), which we confirmed using LiP–MS was via strong stabilization of protein structure (Supplementary Fig. 5c). Surprisingly, TMAO caused more aggregation of 186 and 158 proteins at the tested temperatures (Fig. 5c and Supplementary Data 3). At the lower temperatures, this set was predicted to have low solubility and be aggregation prone (TANGO score; Methods), suggesting that these were proteins on the edge of solubility. By contrast, TMAO promoted the high-temperature aggregation of proteins that were generally more soluble than the proteome (Supplementary Fig. 5d). In addition, these proteins had more predicted short disordered protein-binding regions⁵⁴, higher RASAs and predicted disorder and were shorter than the rest of the measured proteome (Fig. 5d). They also had a lower percentage of histidines (Fig. 5d) and were enriched for periplasmic proteins (Supplementary Fig. 5e), which tend to unfold but not aggregate⁵⁵. Our data thus indicate that TMAO promotes the aggregation of small, partially unfolded proteins that otherwise tend to remain soluble after unfolding.

To understand the effect of TMAO better, we focused on the ribosome recycling factor (Frr), which remained soluble across the entire temperature gradient but precipitated at higher temperatures in the presence of TMAO (Fig. 5e). This effect was also observed by LiP–MS, which further provided peptide-level information on Frr structural changes (Fig. 5f). The LiP–MS unfolding profiles reported that several osmolytes stabilized Frr. For the peptide ASPSLLDGIVVEYYGTPTPLR (amino acids 32–52 (blue region), predicted to be aggregation prone; see Methods), the profile changed from unfolding to nonmonotonous after the addition of TMAO or betaine, indicating that the protein precipitated at high temperatures in the presence of these osmolytes. Purified Frr behaved similarly, suggesting that the effects of TMAO and betaine are direct rather than via chaperones or binding partners (Supplementary Fig. 5f). Further, because TMAO stabilized the Frr structure as shown with both LiP–MS (Fig. 5f) and circular dichroism (Supplementary Fig. 5g), we concluded that TMAO promotes aggregation only after the protein has unfolded.

In conclusion, the addition of osmolytes caused an upward shift in the aggregation temperature of most proteins due to osmolyte stabilization of protein structure. However, TMAO in particular promoted the aggregation of subsets of proteins.

Osmolyte effects on the human proteome

To assess whether our observations are generalizable and to investigate the thermal behavior of intrinsically disordered proteins (IDPs), we analyzed in situ osmolyte effects on the human proteome. IDPs are estimated to comprise 33% of the human proteome versus 4.2% in E. coli⁵⁶ and are in several cases associated with human disease⁵⁷. We thus studied both the in situ thermal unfolding behavior and osmolyte effects of this interesting set of proteins. We focused our analysis on TMAO because it promoted aggregation of proteins with higher predicted disorder in E. coli while strongly stabilizing globular proteins; trehalose was also included for comparison.

We first examined the thermal unfolding behavior of the human proteome in the absence of added osmolytes. Consistent with our previous observations on α-synuclein⁴², protein regions with high predicted disorder tended to show flat rather than thermal unfolding (that is, sigmoidal) profiles. For instance, the CAP-1 protein showed clear thermal unfolding profiles for regions predicted to be folded (peptides 2 and 5), whereas peptides map** to predicted disordered regions (peptides 1, 3 and 4) showed a flat profile not affected by temperature (Fig. 6a). Across the proteome, proteins with a high fraction of predicted disorder had significantly more peptides with a flat profile than folded proteins (Fig. 6b; flat profiles were defined as those with absolute log₂ (fold change) values of <0.5 between the minimum and maximum peptide intensity values). Similarly, peptides originating from regions with high predicted disorder had a significantly flatter thermal profile than peptides from regions with low predicted disorder, which are more likely to be folded (Supplementary Fig. 6a). Overall, 62.6% of peptides predicted to be disordered (pLDDT score of <50) had a flat thermal profile (Supplementary Data 4). Thus, our data enable the definition of regions of the human proteome (n = 1,864 peptides, corresponding to 727 proteins) that have flat thermal unfolding profiles and are predicted to be intrinsically disordered, providing evidence for intrinsic disorder in situ. Interestingly, the remaining 37.4% of peptides predicted to be disordered showed evidence of thermal unfolding in cell lysates (Supplementary Fig. 6b); an example is peptides from the protein PCBP2 (Supplementary Fig. 6c). These data suggest that a substantial fraction of predicted disordered regions may fold, bind other molecules or both in cell lysates (see Discussion).

**Fig. 6: Osmolyte effects on the human proteome.**

We then examined the effects of TMAO and trehalose on human proteins. We first asked how osmolytes affect the structure of folded and disordered regions of human proteins by analyzing LiP–MS data from the first two temperatures (37 °C and 40.5 °C) of our thermal profiling gradient. As in E. coli, only a very small fraction of proteins showed structural effects of TMAO or trehalose, but, interestingly, we observed different effects on folded and disordered regions. TMAO affected around 10% of disordered regions, whereas folded regions were less affected. By contrast, trehalose showed almost no structural change in disordered regions, and its overall effect was smaller (Fig. 6c). Neither TMAO nor trehalose changed the fraction of proteins with flat thermal profiles in a human cell lysate, consistent with these osmolytes having no strong global structural effect (Supplementary Fig. 6d). These patterns could also be exemplarily seen for the oncogenic protein EWS, an IDP associated with Ewing sarcoma and other cancers⁵⁸ (Supplementary Fig. 6c).

We further analyzed how TMAO and trehalose affect aggregation of human proteins. At physiological temperature (37 °C), TMAO, but not trehalose, promoted aggregation of 25% of human proteins (Fig. 6d and Supplementary Data 4). This set was enriched in large disordered proteins with low β-sheet content, low propensity to bind proteins (SCRIBER score) and low AlphaFold prediction score (Fig. 6e), indicating that TMAO promotes the aggregation of large disordered human proteins. This set included the human disease proteins p53, huntingtin and EWS, all of which precipitated in the presence of TMAO but not trehalose (Supplementary Fig. 6e).

Finally, we focused on the subset of around 2,000 human proteins for which we could derive thermal unfolding profiles, which was enriched in globular proteins as expected (Supplementary Fig. 6f). Both TMAO and trehalose stabilized more than 40% of the analyzed proteins (Fig. 6f and Supplementary Data 4), and, as in E. coli, the effect of TMAO was stronger (Fig. 6g). The stabilized proteins were enriched in globular proteins with high AlphaFold prediction score, high β-sheet content, low RASA and low predicted disorder (Fig. 6e), in line with the set of proteins preferentially stabilized by TMAO in E. coli and supporting that human proteins are also stabilized through preferential exclusion. This analysis also highlights the dual effect of TMAO to strongly stabilize globular proteins and promote aggregation of large disordered proteins.

Overall, our analysis of the human proteome shows that the stabilizing effects of osmolytes on proteins are general, supports the existence of IDP regions in situ, highlights interesting instances of potential in situ structure in disordered regions and demonstrates differential effects of TMAO on disordered and globular proteins.

Discussion

We have extended LiP–MS-based thermal profiling into a robust method to study proteome thermal stabilization in situ. This approach allowed us to separately analyze the effects of osmolytes on protein thermal unfolding, native structure and aggregation, thus going beyond the simple identification of osmolyte-stabilized proteins and shedding light on mechanisms of osmolyte action. Our approach further allowed for the assessment of whether direct binding is required for protein stabilization. We report a global analysis of osmolyte effects in a cell-like environment, revealing that osmolyte-dependent protein thermal stabilization is a widespread phenomenon in situ. Our approach can be applied to any small molecule with proposed stabilizing effects. Analysis of the human proteome shows that LiP–MS can be used to study the behavior of intrinsically disordered regions in situ.

We revealed biophysical characteristics of the stabilized proteome as well as mechanisms of osmolyte action. Our data are consistent with a preferential exclusion mechanism for most protein–osmolyte pairs⁸, although observations for several individual proteins did not match the predictions of this theory. This might be because the predictions are based on in vitro models and do not consider protein–protein and protein–metabolite interactions that can influence thermal stabilization⁵⁹. Indeed, our data show an example of such a case, where ribose binding to the protein RbsK increased the stabilization efficiency of TMAO. We identified many other candidate osmolyte binding proteins, although the structure of most proteins did not change after the addition of osmolytes. Most osmolytes caused an upward shift in the temperatures at which proteins aggregate, most probably by stabilizing protein structure. However, the osmolyte TMAO was an exception; it both thermally stabilized many proteins with relatively strong effects and promoted aggregation in both E. coli and human proteomes. Intriguingly, our data suggest that TMAO stabilizes globular proteins while promoting the aggregation of disordered proteins, potentially hel** to explain opposing reports in the literature^32,52,60.

Intriguingly, in nature, TMAO is typically present in contexts in which there are also high levels of urea, a known protein denaturant^61,62. For instance, chondrichthyans (sharks, rays and skates) accumulate high concentrations of urea but also TMAO in a ratio of 2:1 (urea:TMAO)⁶³. It is plausible that TMAO stabilization of most proteins is adaptive in such a context and also that urea prevents the potentially detrimental TMAO-dependent aggregation in some proteins. Such effects have been observed for the IDP α-casein, for which urea prevents TMAO-induced protein aggregation⁶⁴.

Our observation of good overlap between proteins stabilized by different osmolytes helps explain their interchangeability in protecting microorganisms following heat stress. However, use of osmolytes is not uniform; for instance, TMAO is mostly found in marine organisms^65,66, whereas plants largely use proline and betaine⁶⁷. Selection of specific osmolytes by evolution may also have been driven by other factors, for instance, compatibility with enzyme function, availability, cost of uptake or production or other osmolyte functions.

Our analysis of the human proteome provides global in situ evidence for protein intrinsic disorder but also evidence for folding in protein regions corresponding to more than one-third of peptides predicted to be disordered. We note that it is challenging, however, to define appropriate thresholds for foldedness (of a protein region) and for flatness (of a peptide profile), both of which are needed for such analyses. In addition, disorder prediction algorithms are unlikely to be perfect. Nevertheless, our data suggest that a fraction of proteins predicted to be intrinsically disordered may take on unexpected structure in the cellular context, which may be due to protein–protein interactions, post-translational modifications or other consequences of the cellular milieu. LiP–MS thermal profiling could be an exploratory tool both to study such effects and to identify proteins or protein regions that are disordered within cell lysates. Our analysis, with peptide-level resolution, enabled us to probe the behavior of IDRs within a protein sequence and expands our previous protein-level analysis of IDPs⁴².

Finally, our study could help identify new stabilizers. We showed that DSF on cell lysates can accurately predict global stabilization effects of osmolytes, which could form the basis of a new stabilizer screening pipeline. DSF on a complex protein mixture could be used first to select promising molecules that show global effects, followed by LiP–MS on selected hits to profile their effects on specific proteins or protein groups and to study mechanisms. Interestingly, stabilization through direct binding tends to cause a larger shift at the same osmolyte concentrations than stabilization through preferential exclusion. This suggests that, purely for stabilization purposes, small molecules that directly bind to the protein should be favored, although a balance between stabilization and retention of protein activity must be found. Further, differential effects of an osmolyte on different proteins, as we have observed for TMAO, must be considered. Our dataset showed that, although osmolytes are broadly interchangeable, for most proteins, either TMAO or glucose was the best stabilizer; these data can be further used to specifically identify which stabilizer is best for which protein or to study combinatorial effects. Last, our data could drive machine learning approaches to predict protein stabilization by a given osmolyte⁶⁸.

A few caveats must be considered. First, because the experiments were performed in cell lysates, the effects may not entirely reflect the situation within intact cells. Second, the concentrations of osmolytes that we used are close to cellular values^48,49 but may be high for biotechnological applications. This could be circumvented by designing polymers with osmolyte active groups to reach a desired effect at lower concentrations⁶⁹. Third, although LiP–MS provides information-rich datasets of structural changes, mechanistic interpretation of these changes can be challenging; as shown here, these can be tackled by combination with an orthogonal approach such as TPP⁴⁷. Finally, our data do not report on important additional factors, such as osmolyte toxicity, which must be considered when selecting an optimal stabilizer. For example, increased TMAO concentrations have been associated with increased risk of cardiovascular disease⁷⁰.

Increasing the thermal stability of enzymes and protein-based drugs is a major challenge in biotechnology. Our study shows the potential of small molecules as protein stabilizers because we found that specific osmolytes or combinations thereof can stabilize most proteins. Further, we showed that our thermal profiling approach can differentiate between osmolyte mechanisms of action and can help identify molecules that increase thermal stability or counteract irreversible aggregation. This global and in situ analysis of osmolyte action has shown which effects of osmolytes are protein specific and which are global and has allowed us to extrapolate observations for purified proteins into general theory. Our study advances the biophysical and biological understanding of osmolyte mechanism of action and provides practical insight into the development of better protein stabilizers.

Methods

Escherichia coli

E. coli strain BW25223 K12 (ref. ⁷²) was grown in M9 minimal medium⁷³ containing 2 g l^–1 glucose at 37 °C and 800 rpm. A single colony was picked from a fresh plate to inoculate Luria–Bertani (LB) complex medium and incubated for 6 h at 37 °C with constant shaking. LB cultures were diluted 1:100 with M9 minimal medium and incubated overnight. Overnight cultures were diluted to an optical density at 600 nm (OD₆₀₀) of 0.05 and collected when the OD₆₀₀ reached 0.8.

HEK293T cells

HEK293T cells were cultured in DMEM (Gibco, 41965039) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin and passaged before confluency by detachment with 0.25% trypsin, followed by washing in PBS (pH 7.4; Gibco, 10010015). To store the pellets, cells were centrifuged at 900g for 2 min. PBS was removed, and the pellets were stored at –20 °C until further use.

Cell lysate preparation

E. coli

Frozen cells were thawed on ice and resuspended in lysis buffer (20 mM HEPES, 150 mM KCl and 10 mM MgCl₂, pH 7.5). Acid-washed glass beads were added, and the cells were lysed with three rounds of vortexing for 30 s, followed by a 4-min incubation at 4 °C. The lysates were centrifuged (20,000g, 15 min, 4 °C) to remove cell debris. The supernatants were transferred to a fresh tube, and protein concentration in the lysates was determined by bicinchoninic acid assay (BCA Protein Assay Kit, Thermo Scientific).

HEK293T cells

Cell pellets were resuspended in 300 μl of LiP buffer (20 mM HEPES (pH 7.5), 150 mM KCl and 10 mM MgCl₂) and lysed using a pellet pestle (Argos Technologies) in ten cycles of 10 s of homogenization and a 1-min pause on ice. The lysate was cleared by centrifugation at 15,000g at 4 °C for 10 min. The supernatants were transferred to a fresh tube, and protein concentration in the lysates was determined by bicinchoninic acid assay (BCA Protein Assay Kit, Thermo Scientific).

Preparation of standard control sample of short peptides

E. coli cell lysates were mixed with urea (final concentration of 6 M) and reduced with 1,4-dithiothreitol (DTT; final concentration of 12 mM) for 30 min at 37 °C and subsequently alkylated with iodoacetamide (final concentration of 40 mM) for 45 min at room temperature. Samples were diluted with 0.1 M ammonium bicarbonate to a final urea concentration of 1 M and predigested with lysyl endopeptidase (Wako Chemicals) at an enzyme:substrate ratio of 1:100. After 4 h at 37 °C, sequencing-grade porcine trypsin (Promega) was added at a 1:100 ratio, and samples were incubated overnight at 37 °C and 800 rpm. The digestion was stopped by the addition of formic acid to a final pH of less than 3. The peptide mixtures were loaded onto Sep-Oak tC18 cartridges (Waters), desalted and eluted with 80% acetonitrile. All peptide samples were evaporated in a vacuum centrifuge to dryness and resolubilized with lysis buffer (20 mM HEPES, 150 mM KCl and 10 mM MgCl₂, pH 7.5) to a peptide concentration of 1 mg ml^–1.

Preparation of osmolyte stock solutions

All osmolytes were prepared from ultrapure powders in lysis buffer (20 mM HEPES, 150 mM KCl and 10 mM MgCl₂, pH 7.5) at 2× final concentration. If necessary, the pH was adjusted to 7.5 with 10 M NaOH solution. Stock solutions were frozen at −20 °C until use.

Limited proteolysis

Cell lysates were mixed with osmolyte stock solutions to a final protein concentration of 2 mg ml^–1 and an osmolyte concentration of 1 M (0.5 M for trehalose) or with buffer as a control. Lysates were aliquoted into ten wells (50-μl aliquots, two replicates for each condition) and incubated at a temperature gradient (37–76 °C; exact temperatures: 37, 40.5, 44.4, 49.3, 54.1, 58.9, 63.8, 68.6, 72.5 and 76 °C) for 5 min. Afterward, PK from Tritirachium album (Sigma) was added to the protein extract at an enzyme:substrate ratio of 1:50, mixed well by pipetting with a larger volume (30 μl) and incubated for 1 min at the assigned temperatures. The digestion was stopped by heating the sample for more than 4 min at 99 °C in a thermocycler and adding sodium deoxycholate (DOC) to a final concentration of 5%. The samples were then subjected to complete digestion under denaturing conditions as described below. For the temperature activity assay, the same protocol was used with predigested cell lysate (fully tryptic peptides at a concentration of 0.5 mg ml^–1) instead of native cell lysate. To mimic different PK activities for analysis, the same protocol was used with three different enzyme:substrate ratios for PK (1:20, 1:50 and 1:100).

Thermal proteome profiling

Cell lysates were mixed with osmolyte stock solutions to a final protein concentration of 2 mg ml^–1 and an osmolyte concentration of 1 M (0.5 M for trehalose) or with buffer as a control. Lysates were aliquoted into ten wells (50-μl aliquots, two replicates for each condition) and incubated at a temperature gradient (37–76 °C) for 5 min. For the HEK293T cell precipitation assay, samples in triplicates, with and without osmolytes, were incubated at 37 °C for 5 min. After heating, samples were filtered by centrifugation at 800g for 5 min with a 0.2-μm PVDF membrane filter (Corning FiltrEX 96-well white filter plate). After centrifugation, 40 μl of flowthrough was transferred to a fresh plate and mixed with DOC to a final concentration of 5%. The samples were then subjected to complete digestion under denaturing conditions as described below.

Tryptic digestion

Following LiP or filtration, protein fragments were reduced by incubation with DTT (final concentration of 12 mM) for 30 min at 37 °C and alkylated by incubation with iodoacetamide (final concentration of 40 mM) for 45 min at room temperature in the dark. Samples were diluted with 0.1 M ammonium bicarbonate to a final concentration of DOC of 1%. Proteins were digested overnight with lysyl endopeptidase (Wako Chemicals) and sequencing-grade porcine trypsin (Promega) at an enzyme:substrate ratio of 1:100 at 37 °C with constant shaking (800 rpm). The digestion was stopped by the addition of formic acid to a final concentration of 1% (pH < 3). Precipitated DOC was filtered by centrifugation at 800g for 5 min with a 0.2-μm PVDF membrane filter (Corning FiltrEX 96-well white filter plate). The peptide mixtures were loaded onto 96-well elution plates (Waters), desalted and eluted with 80% acetonitrile and 0.1% formic acid. After elution, peptides were dried in a vacuum centrifuge, resolubilized in 0.1% formic acid and analyzed by MS.

Liquid chromatography–tandem mass spectrometry data acquisition

Peptides were separated on a Waters Acquity M-Class UPLC system on a 40 cm × 0.75 mm inner diameter column (New Objective, PF360-75-10-N-5) packed with 1.9-μm C18 beads (Dr. Maisch, Reprosil-Pur 120). Fractionation was achieved with a gradient of buffer A (0.1% formic acid, Carl Roth) and buffer B (99.9 % acetonitrile and 0.1% formic acid, Carl Roth): a linear gradient from 3% to 35% buffer B over 120 min, followed by 5 min with an isocratic constant concentration of 90% buffer B. The column was heated to 50 °C.

Data-dependent acquisition

Shotgun liquid chromatography–tandem MS data-dependent acquisition on the Orbitrap Fusion Lumos Tribrid mass spectrometer was performed solely for library generation. One replicate of each condition was selected randomly, and 2 μl from the sample was injected. MS¹ spectra were acquired from 350 to 1,500 m/z at an orbitrap resolution of 120,000 with an automated gain control (AGC) target of 150% or 100-ms injection time. Precursors with an intensity exceeding 50,000 and a positively charged state between 2 and 5 were selected for data-dependent MS² scans. A dynamic exclusion after a single occurrence for 30 s with a 10-ppm mass tolerance was applied. Selected precursors were isolated with a quadrupole and an isolation window of 1.2 m/z. Precursors were fragmented with high-energy collision-induced dissociation with a fixed collision energy of 27%. MS² spectra were acquired at an orbitrap resolution of 30,000 and an automatically adapting scan range (minimum – precursor × charge + 10.0) and an AGC target of 200% or a dynamic injection time that automatically calculated the maximal time available.

Data-independent acquisition

For measuring peptide abundances, 2 μl of each sample was injected independently and was measured in data-independent acquisition (DIA) mode. The DIA MS method consisted of a survey MS¹ scan from 350 to 2,000 m/z at a resolution of 120,000 with an AGC target of 50% or 100-ms injection time followed by the acquisition of DIA in 41 variable-width isolation windows⁷⁴. Precursors were isolated by a quadrupole and activated high-energy collision-induced dissociation with a collision energy of 28%. DIA MS² spectra were acquired with a scan range of 200 to 1,800 m/z at an orbitrap resolution of 30,000 with an AGC target of 200% or 54-ms injection time.

Protein expression and purification

E. coli strains carrying plasmids encoding DnaK, Eno, Frr, RbsK and Gnd with an N-terminal His₆ tag from E. coli were commercially obtained⁷⁵. From the strain libraries, precultures were inoculated in LB broth with corresponding antibiotic. LB cultures (1.5 l) were inoculated with 15 ml of preculture and grown to logarithmic phase at 37 °C. Cultures were induced with 1 mM IPTG at an OD₆₀₀ of 0.8 and were allowed to grow at 37 °C for 3 h with vigorous shaking.

Collected cells were suspended in 50 ml l^–1 culture Buffer B (50 mM HEPES-KOH (pH 7.5), 150 mM NaCl and 10% glycerol) supplemented with 1 mM PMSF, 2 mM MgCl₂, 50 U μl^–1 DNase and 1× protease inhibitor cocktail (Sigma). Resuspended cultures were mechanically lysed by cell cracking with five passages at a chamber pressure of 12,000 psi. Cell debris was pelleted by centrifugation of lysed cells at 20,000 rpm from 30 min at 4 °C.

Soluble supernatant with 10 mM imidazole was applied to Ni-NTA agarose beads (Qiagen) with a column volume of 2.5 ml. Protein bound to the Ni-NTA agarose beads was washed with 10 column volumes of Buffer B supplemented with 20 mM imidazole. Proteins of interest were eluted from beads with 3 column volumes of Buffer B supplemented with 500 mM imidazole. An additional 1 mM EDTA was added to the eluted protein. Protein was further purified by gel filtration chromatography on a HiLoad 16/600 Superdex 75 pg column (Cytiva, formerly GE Healthcare) in Buffer S (50 mM HEPES-KOH (pH 7.5), 150 mM NaCl, 10% glycerol, 1 mM EDTA and 1 mM DTT). Purified DnaK, Eno, Frr, RbsK and Gnd were stored in Buffer S at −20 °C after snap freezing in liquid nitrogen. For various assays, the storage buffer was exchanged for assay buffer using PD MiniTrap G-25 columns (Cytiva, formerly GE Healthcare) by gravity flow.

Differential scanning fluorimetry for whole-cell lysate

E. coli cell lysates were prepared as described above. Lysates were diluted to a final protein concentration of 1 mg ml^–1 together with 10× Sypro Orange Gel stain (Thermo Fischer) and a buffer (control) or 1 M osmolytes (0.5 M for trehalose) in a total volume of 25 μl. Thermal scanning (25–95 °C, 0.5 °C min^–1) was performed in quadruplicates using a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with the scan channel set to FRET. The data were fit using the thermal unfolding equation⁷⁶, assuming a two-state unfolding model with the linear extension of before and after denaturation baseline. Because DSF is typically used with purified proteins, we first examined its performance on an untreated lysate using two different total protein concentrations and two different E. coli lysate preparations. To determine the reproducibility of the method, the experiment was performed on different days; different biological replicates were used and diluted to different protein concentrations (from 0.5 to 2 mg ml^–1). Despite this, the lysate melting temperature was extracted reproducibly.

Differential scanning fluorimetry for purified proteins

Proteins were purified as described above and diluted to a final protein concentration of 0.5 mg ml^–1 together with 10× Sypro Orange Gel stain and a buffer (control) or 1 M osmolytes (0.5 M for trehalose) in a total volume of 25 μl. Thermal scanning (25–95 °C, 0.5 °C min^–1) was performed in quadruplicates using a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with the scan channel set to FRET. For RbsK, the transitions were measured in the presence of increasing concentrations of ribose. The data were fit using the thermal unfolding equation⁷⁶, assuming a two-state unfolding model with the linear extension of before and after denaturation baseline.

Circular dichroism

Protein secondary structure and thermal transitions were measured using a Jasco J-715 spectropolarimeter (Brechbühler). Protein (300 μl) at a concentration of 0.2 mg ml^–1 in buffer C (20 mM NaH₂PO₄.H₂O/K₂HPO₄ (pH 7.5), 150 mM NaCl and 1 mM MgCl₂) was measured in a high-precision cell quartz cuvette with a pathlength of 1 mm (Hellma Analytics). Secondary structure scans were continuously measured from 260 nm to 190 nm with a scanning speed of 20 nm min^–1 and a total of five repetitions. The average ellipticity scans were buffer corrected. Ellipticity (ϴ) was then normalized to mean molar ellipticity in degree cm^–2 dmol^–1 (MRW = (ϴ × 100 × kDa)/(c × d × N)), where N is the number of amino acid residues, d is the cuvette pathlength in centimeters and c is the protein concentration in mg ml^–1. Thermal unfolding of Frr was measured in the presence and absence of 1 M TMAO at 222 nm from 25 °C to 95 °C using a slope of 30 °C h^–1. Changes in ellipticity as a function of temperature were fitted according to the single transition state model.

Precipitation analysis for ribosome recycling factor

Frr was diluted to a final protein concentration of 0.3 mg ml^–1 in lysis buffer or 1 M osmolytes (0.5 M for trehalose) in a total volume of 20 μl. Samples in triplicates were heated to 70 °C or 37 °C (as a control) in a thermoblock for 5 min and subsequently centrifuged at max speed using a table-top centrifuge. Protein concentration in the supernatant was determined in technical duplicates for each replicate using a Bradford assay (Thermo Fischer). Absorbance at 70 °C was divided by the average absorbance at 37 °C to calculate the percentage of remaining soluble protein.

Thermal proteome profiling for whole-cell lysates

E. coli cell lysates were prepared as described above and diluted to a final protein concentration of 1 mg ml^–1 together with lysis buffer or 1 M osmolyte (0.5 M for trehalose) in a total volume of 50 μl. The lysates were incubated at ten temperatures ranging from 37 °C to 76 °C for 5 min. After heating, samples were filtered by centrifugation at 800g for 5 min with a 0.2-μm PVDF membrane filter (Corning FiltrEX 96-well white filter plate). The flowthrough was collected in a 96-well plate, and protein concentration was determined using a Bradford assay (Thermo Fischer).

Extended proteolysis assay

E. coli cell lysates were prepared as described above and diluted to a final protein concentration of 2 mg ml^–1 with lysis buffer or 1 M osmolyte (final concentration) in a total volume of 50 μl. The lysates were incubated at 76 °C for 5 min. The sample was cooled to room temperature, and PK was added at an enzyme:substrate ratio of 1:50. After 5, 10 or 30 min, based on the assigned digestion time, samples were transferred to a cooled centrifuge and centrifuged for 15 min at maximum speed at 4 °C. Immediately after centrifugation, the supernatant was removed, and aggregates were resolubilized in 50 μl of 2% SDS by vortexing and sonication. Protein concentration of resolubilized proteins was determined by bicinchoninic acid assay (BCA Protein Assay Kit, Thermo Scientific). For time point 0 (no PK digestion), PK was not added to the sample.

Detergent resolubilization assay

E. coli cell lysates were prepared as described above and diluted to a final protein concentration of 2 mg ml^–1 with lysis buffer or 1 M osmolyte (final concentration) in a total volume of 50 μl. The lysates were incubated at 76 °C for 5 min, and the sample was cooled to room temperature and mixed (1:1) with increasing concentrations of sarcosyl in 1× PBS. After vortexing and sonication, samples were centrifuged (maximum speed, 15 min at 4 °C), and the concentration of protein in the soluble fraction was determined by bicinchoninic acid assay (BCA Protein Assay Kit, Thermo Scientific).

Viscosity measurements

The viscosity of different 1 M osmolyte solutions was evaluated by measuring the apparent diffusion coefficient of standard 100-nm nanoparticles via dynamic light scattering, following a previously published protocol⁷⁷. Briefly, the viscosity (η) of the solutions was calculated from measurements of the Stokes–Einstein equation,

$$\eta =\frac{{kT}}{6\pi {R}_{{\mathrm{h}}}D},$$

where k is the Boltzmann constant, T is the temperature, and D and R_h are the apparent diffusion coefficient and the hydrodynamic radius of the nanoparticles used as standard tracers, respectively. Dynamic light scattering measurements were performed on a Zetasizer Nano working in backscattering mode at 173°.

Peptide- and protein-level quantification

The data were searched in Spectronaut version 14.5 (E. coli) and version 14.11 (human; Biognosys)⁷⁸. Hybrid libraries for the tryptic control and LiP samples consisting of the corresponding data-dependent acquisition and DIA runs were created based on a Pulsar search using default settings except for the digest type, which was set to ‘semi-specific’, and the minimum peptide length, which was set to six. The data were searched against the UniProt fasta database (for E. coli strain K12, UP000000625, November 2020; for the human proteome, canonical fasta, downloaded March 2020). Targeted data extraction was performed in Spectronaut with default settings except for machine learning, which was set to ‘across experiment’, and data filtering, which was set to ‘Qvalue’. The FDR was set to 1% on the peptide and protein levels. From the LiP search, we exported peptide intensities from the tryptic control search (for TPP data) protein intensities.

Fully tryptic and half tryptic peptides

PK cleavage within the sequence of an fully tryptic peptide often results in the detection of half-tryptic peptides that share either the N or the C terminus with the associated fully tryptic peptide. Our previous study only analyzed fully tryptic peptides that reported an altered accessibility to PK with temperature for the respective protein region⁴². In the current study, both fully tryptic and half-tryptic peptides were included in the analysis.

Correlation analysis for fully tryptic and half tryptic pairs

Fully tryptic and half-tryptic peptides were matched by their assigned position in a protein. Half-tryptic peptides were matched with an fully tryptic peptide if the borders of the half-tryptic peptides were within the borders of the fully tryptic peptide. Pearson correlation was calculated between the peptide intensities of the matched pairs.

Testing and correcting for condition effects on proteinase K activity

Left unadjusted, any effects of experimental conditions on PK activity could bias the analysis. We performed a protease activity assay using a mixture of more than 3,000 short (less than ten amino acids) peptides to evaluate PK activity independent of folding effects. We tested whether we observed significantly different digestion of these short peptides when LiP was performed at different temperatures or in the presence of osmolytes. To eliminate the peptides with potential secondary structure, we only focused on approximately 3,000 fully tryptic peptides that were at most ten amino acids long. Because the fully tryptic peptides were the substrate of our activity assay, decreased abundance of the peptide indicates higher PK activity. To identify significantly changing peptides based on moderated t-tests using the R package protti⁷⁹ followed by a Benjamini–Hochberg correction⁸⁰, we compared peptide abundances at high temperatures (68.6 °C and 72.5 °C) and low temperatures (37 °C and 40.5 °C) for the temperature assay and control and osmolyte (1 M concentration, except for trehalose at 0.5 M) for the osmolyte comparison. The peptides were considered significantly changed if the adjusted P value was <0.05 and the | log₂ (fold change) | value was >1.

We observed significant changes (adjusted P value of <0.05, | log₂ (fold change) | value of >1) in only 6.6% of peptides between the different temperatures applied in our thermal profiling experiment (Supplementary Fig. 1c) and in less than 3.5% of peptides across our tested osmolytes (Supplementary Fig. 1d). To test whether these minor changes in PK activity could influence the thermal profiling readout, we performed independent thermal denaturation experiments using multiple PK concentrations to mimic different PK activities. We observed that rescaling each peptide intensity between 0 and 1 reduced the difference between samples with different PK concentrations while not affecting the expected unfolding effects due to temperature differences.

Learning peptide intensity profiles across temperatures

We extended the LiP–MS thermal profile analysis beyond the analysis of only thermal unfolding by fitting peptide intensity data along the temperature gradient using a nonparametric Gaussian model instead of the thermodynamic model we had previously used. This extended our analysis beyond peptides that follow a typical sigmoidal unfolding profile. To learn the temperature profiles for each peptide under different conditions, we used Gaussian processes (GP)⁸¹, which provide a probabilistic framework to learn a nonparametric relationship of peptide abundance to temperature. In detail, we used gpytorch⁸² version 1.4.2 with an ExactGP model choosing a constant mean function, a squared exponential kernel and a Gaussian likelihood. For each peptide, separate GP models for peptide intensities in the absence (control condition) and presence of an osmolyte (osmolyte condition) and a joint model were defined, and model hyperparameters were found by maximizing the sum marginal log likelihood across all models using Adam optimizer with a learning rate of 0.1 and 1,000 iterations. Based on the resulting posterior of the fit, predicted mean abundance profiles and confidence intervals based on 2 s.d. around the mean were found for each peptide and condition. The residual sum of squares between the observed peptide intensities and the predicted intensities were calculated for each peptide and condition to assess the goodness of the fit.

Clustering of profiles

Fitted values (for 20 temperatures with 2 °C increases) from the learned temperature profiles in the control model were taken for all peptides and clustered into 20 clusters using fuzzy k means clustering (FKM.ent function) from the fclust R package⁸³, setting the fuzzy degree of entropy parameter to 2. After clustering, the median profiles of the clusters were grouped into three groups. First, we separated the clusters based on whether the median profile showed nonmonotonous behavior (orange group; the difference in peptide intensity between the lowest and highest temperature < 0.5). Next, monotonous clusters were grouped into increasing group (purple; peptide intensity at lowest temperature < peptide intensity at highest temperature) and decreasing group (green; peptide intensity at lowest temperature > peptide intensity at highest temperature). To identify whether the peptide belongs to the same cluster shape following osmolyte addition, clustering was repeated by including fitted values from the models for the control and osmolyte conditions. Peptides were clustered and grouped into three groups as before. For each peptide, probabilities of belonging to individual clusters from the same profile group were summed. A peptide was considered as changing when the difference in probability for a certain main group for control or osmolyte was larger than 0.5.

Calculation of peptide- and protein-level scores

After fitting of the GP models, distances between the learnt control and osmolyte curves were calculated for the temperature regions where their confidence intervals do not overlap. In the case of overlap** confidence intervals, the distance was set to 0. To study protein stabilization specifically, temperature intervals with intensity changes were classified as binding (changes at the start of the temperature gradient), stabilization (changes at the middle of the temperature gradient) or aggregation (changes at the end of the gradient). The distances between the curves in temperature intervals were summed as a proxy for the area between the curves. To identify whether the change represents protein stabilization or destabilization, clustering was used as described before to identify whether the peptide has a decreasing or increasing profile, and the sign of the area was corrected accordingly. If two areas of stabilization were identified, only the absolute maximal area was considered the peptide-level stabilization score. To summarize the peptide-level stabilization data into protein-level stabilization data, amino acid-level stabilization was calculated to correct for overlap** peptides and peptides of different lengths. To calculate the amino acid-level score, we first calculated the mean value of all peptide-level scores of overlap** peptides at a specific amino acid position to determine whether that position was generally stabilized (mean > 0) or destabilized (mean < 0). For a ‘stabilized position’, a weighted 0.75 quantile of the peptide-level score for all overlap** peptides at a specific position was calculated, weighted by the goodness of fit (summed for both conditions (control and osmolyte)). For a ‘destabilized position’, a weighted 0.25 quantile was calculated. The mean goodness of fit for each position was also calculated. To summarize the information into a single protein-level stabilization score, we first calculated a mean value of all amino acid-level scores to determine whether the protein was generally stabilized (mean > 0) or destabilized (mean < 0). For stabilized proteins, we calculated the weighted 0.75 quantiles for all positions, weighted by the mean goodness of fit. For destabilized proteins, we calculated the weighted 0.25 quantiles for all positions, weighted by the mean goodness of fit. The 0.75 quantile was chosen to maximize the agreement between TPP and LiP–MS data, while kee** the false-positive rate of analysis below 5%. More details about the analysis can be found in the Rmarkdown script (Supplementary Note). To assess the peptide-level FDR of the proposed approach of identifying stabilized proteins from melting curves, the control experiment (that is, no osmolyte) was performed with four replicates that were grouped into two groups and treated as separate conditions. In such a setup, we would expect no significant changes between groups. We observed that the peptide-level FDR was lower than 0.05. For calculation of the protein-level FDR, different quantiles for summarization were considered, ranging from 0.5 to 1 for stabilized protein and from 0.5 to 0 for destabilized protein. A protein was considered significantly changing when the protein-level score was not equal to 0 (Supplementary Fig. 2c). We observed a protein-level FDR of 0.05 when 0.75 quantile (0.25 quantile for destabilized proteins) was used for summarization.

Comparison of stabilization strength for known binders

Proteins were classified as known binders of glucose, betaine or proline based on UniProt annotations. Analysis was performed on peptide-level stabilization scores, as only few known binders were identified in the literature. Within each dataset, only peptides with significant stabilization scores were considered, and the peptides were classified as originating from a binding protein or not (binding and not binding). Stabilization scores were scaled within a dataset using the following formula to adjust for overall different strengths of different osmolytes, for example, glucose stabilizing the proteome better than glucose:

$$\begin{array}{l}{{\mathrm{Scaled}}}\,{{\mathrm{score}}}\\=\displaystyle\frac{{{\mathrm{Peptide}}}\,{{\mathrm{stabilization}}}\,{{\mathrm{score}}}-{{\mathrm{median}}}\,({{\mathrm{stabilization}}}\,{{\mathrm{scores}}})}{0.7\,{{\mathrm{quantile}}}\,\left({{\mathrm{stabilization}}}\,{{\mathrm{scores}}}\right)-0.25\,{{\mathrm{quantile}}}\,({{\mathrm{stabilization}}}\,{{\mathrm{scores}}})}\end{array}$$

The stabilization scores are all stabilization scores of stabilized peptides of a given osmolyte dataset. Scaled stabilization scores for known binders and nonbinding proteins from glucose, betaine and proline datasets were then combined to produce Fig. 4a. To produce Supplementary Fig. 4a, the same approach was applied, only the binding and nonbinding proteins were mismatched, for example, glucose binding protein was assigned as betaine binding protein to assess whether the subset of binding proteins was generally stabilized better under all osmolyte conditions.

LiP–MS binding analysis

To study changes of native proteins after small-molecule addition, we focused only on the first two temperatures (37 °C and 40.5 °C) where the majority of the proteins should not be unfolded. Peptide differential abundance was calculated based on t-tests, followed by Benjamini–Hochberg multiple testing correction⁸⁰. Because this analysis was performed on nonscaled data (compared to scaled data used for the analysis of thermal stabilization), a more stringent cutoff was used for significance analysis to account for potential changes caused solely due to changes in PK activity that scaling can correct for. The significance cutoff was set to an adjusted P value of <0.01 and a | log₂ (fold change) | value of >1.5. Proteins were identified as significantly changing as soon as one proteotypic significantly changing peptide from that protein was identified.

Calculation of protein characteristics

We assessed 54 protein physicochemical and biochemical characteristics based on protein sequences (for example, charge, pI and hydrophobicity) on AlphaFold-predicted protein structures⁸⁴ (for example, RASA, secondary structure and disorder) or using a variety of additional sequence-based predictors. Protein characteristics based on protein sequence (protein pI, charge, protein length, hydrophobicity, aliphatic index, hydrophobic moment and percentages of individual amino acids or amino acid groups) were calculated using the Peptides package (https://journal.r-project.org/archive/2015/RJ-2015-001/RJ-2015-001.pdf) in R. The predictions SCRIBER score⁸⁵, DRNApredDNAscore⁸⁶, DRNApredRNAscore⁸⁶, MoRFchibi⁵⁴, DFLpred, DisoRNAscore, DisoPROscore and DisoDNAscore were downloaded from DescribeProt⁸⁷. Solubility (CamSol) was calculated using CamSol software⁸⁸. Aggregation prediction was calculated using TANGO⁸⁹. Amino acid-level information was summarized by calculating the percentage of amino acids in the aggregation-prone region (percentage_agg) or the longest consecutive stretch of aggregation-prone amino acids (longest_agg). RASAs and secondary structures were calculated using AlphaFold-predicted structures⁸⁴ with the Bio.PDB package in Python using the DSSP module with default settings. Individual values for RASA were calculated by calculating the average RASA for the whole protein. Individual values for each structural element (helix, sheet or coil) were calculated as a percentage of amino acids of a certain structure to the whole protein. The radius of gyration was calculated using AlphaFold structures with the Bio.PDB package (Rg function) with default settings. To identify whether two groups (stabilized versus nonstabilized, high correlation versus low correlation and increased aggregation versus rest) of proteins significantly differ in certain protein characteristics, t-tests were performed, followed by Benjamini–Hochberg corrections⁸⁰. The difference between mean values for the two groups was calculated, and the control value was always subtracted from the test group to calculate whether the characteristics were higher or lower in the test group.

Identification of nonprecipitators

Proteins were characterized as precipitators or nonprecipitators based on the TPP dataset⁴⁷. Proteins were considered nonprecipitators if at the highest temperature more than 50% of the protein, relative to 37 °C, remained in the soluble fraction.

Analysis of differential scanning fluorimetry data

Raw fluorescence intensities were scaled between 0 and 1 using the following formula: x′ = [x – min(x)]/[max(x) – min(x)]. Scaled intensities as a function of temperature were fit to the following equation: S = ((S_f + m_f × T) + (S_u + m_u × T) × e[(∆H_m/RT) × ((T – T_m)/T_m)])/(1 + e[(∆H_m/RT) × ((T – T_m)/T_m)]). Here, S is the scaled fluorescence signal in arbitrary units, S_f and S_u are the scaled fluorescence signals of the folded and unfolded states at 0 K, respectively, in arbitrary units, m_f and m_u are the linear dependencies of S_f and S_u, respectively, on temperature (that is, slopes of the pre- and post-transition baseline, respectively) in arbitrary units K⁻¹, T is the temperature in K, T_m is the melting temperature in K, ∆H_m is the unfolding enthalpy at the T_m in J mol⁻¹, and R is the gas constant, 8.314 J K⁻¹ mol⁻¹. The fitting was performed as previously described⁴².

Analysis of thermal protein profiling data

Protein-level quantification data were extracted from Spectronaut without data normalization. Protein abundance relative to 37 °C was calculated to determine the amount of protein in the soluble fraction. Scaled abundances were processed, fit and analyzed the same way as peptide-level data from the LiP–MS study without summarization.

Spearman correlation analysis

To calculate whether individual proteins are stabilized in a similar manner as the whole lysate, the Spearman correlation between protein stabilization score and mean stabilization score was calculated for all proteins reproducibly measured under all conditions. To obtain a random distribution, we calculated the Spearman correlation between protein stabilization score and randomized mean stabilization score shuffled between different osmolyte conditions independently for each protein. The calculation of randomized correlations was repeated 1,000 times to assess the variability of random distribution.

Identification of proteins with changed aggregation abundance

To identify the proteins where protein precipitation is affected, we focused on the last three temperatures (68.8, 72.5 and 76 °C) of TPP measurements. To identify whether the protein precipitates significantly more or less after the addition of osmolyte, t-tests were performed on scaled values, followed by Benjamini–Hochberg multiple testing correction⁸⁰. The significance cutoff was set to an adjusted P value of <0.05 and | log₂ (fold change) | value of >1. To identify whether LiP–MS shows significant changes in protein aggregation/structure of protein aggregates, we focused on the last two temperatures (72.5 and 76 °C). We performed significance analysis, as described earlier, at the peptide level. Proteins were considered significantly changing as soon as a single significantly changing peptide was identified.

Gene ontology enrichment analysis

We tested the proteins where TMAO promotes protein aggregation as defined above (Identification of proteins with changed aggregation) for functional enrichments using the topGO package in R⁹⁰. We downloaded current annotation files for E. coli (http://current.geneontology.org/annotations/ecocyc.gaf.gzm, accessed 2 June 2020). To focus on the most informative terms, we tested for enrichment with Fisher’s exact tests using the elim-algorithm in topGO⁹⁰. We performed the test for biological processes, cellular compartments and molecular functions. Only terms with adjusted P values of <0.01 after the Benjamini–Hochberg P value correction⁸⁰ were considered; however, P values and unadjusted P values are displayed.

Domain-level analysis

Peptides were mapped to domains based on Pfam annotation⁹¹. A peptide was considered part of a domain when a single amino acid of a peptide was part of the domain. Proteins with at least two domains with at least three peptides map** to each of the domains were considered for analysis. Differential analysis using t-tests (two domains) or analysis of variance (more than two domains) was performed to identify proteins with domains with significantly different stabilization.

Data visualization and processing

Data visualization was performed using R version 4.0.5 and the ggplot2 (ref. ⁹²) and ComplexHeatmap⁹³ packages for plotting.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All MS proteomics data have been deposited at ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD036186. UniProt fasta databases for E. coli (strain K12, organism ID 83333) were accessed in November 2020 via the UniProt databases download page (https://www.uniprot.org/downloads). Protein structures for DnaK (PDB ID 4JNE) were downloaded in 2022 from the Protein Data Bank website (https://www.rcsb.org/pdb). AlphaFold predictions for the whole E. coli proteome were downloaded in August 2021. The DescribeProt database for the whole E. coli proteome (http://biomine.cs.vcu.edu/servers/DESCRIBEPROT/download.html) was downloaded in February 2021. The TPP dataset from a previously published study⁴⁷ (dataset EV3 of the publication) was downloaded in 2020. All other data needed to evaluate the conclusions described in the paper are present in the Supplementary Information.

Code availability

The custom R script for data analysis is available via GitHub at https://github.com/PicottiGroup/Thermal_unfolding, together with Rmarkdown showing how to perform the analysis, and on Zenodo (https://doi.org/10.5281/zenodo.10514118).

References

Ye, Y. et al. Global metabolomic responses of Escherichia coli to heat stress. J. Proteome Res. 11, 2559–2566 (2012).
Article CAS PubMed Google Scholar
Kim, S. et al. Heat-responsive and time-resolved transcriptome and metabolome analyses of Escherichia coli uncover thermo-tolerant mechanisms. Sci. Rep. 10, 17715 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Jozefczuk, S. et al. Metabolomic and transcriptomic stress response of Escherichia coli. Mol. Syst. Biol. 6, 364 (2010).
Article PubMed PubMed Central Google Scholar
Sévin, D. C., Stählin, J. N., Pollak, G. R., Kuehne, A. & Sauer, U. Global metabolic responses to salt stress in fifteen species. PLoS ONE 11, e0148888 (2016).
Article PubMed PubMed Central Google Scholar
Hincha, D. K. & Hagemann, M. Stabilization of model membranes during drying by compatible solutes involved in the stress tolerance of plants and microorganisms. Biochem. J. 383, 277–283 (2004).
Article CAS PubMed PubMed Central Google Scholar
Levy-Sakin, M. et al. The influence of chemical chaperones on enzymatic activity under thermal and chemical stresses: common features and variation among diverse chemical families. PLoS ONE 9, e88541 (2014).
Article ADS PubMed PubMed Central Google Scholar
Kaushik, J. K. & Bhat, R. Why is trehalose an exceptional protein stabilizer? An analysis of the thermal stability of proteins in the presence of the compatible osmolyte trehalose. J. Biol. Chem. 278, 26458–26465 (2003).
Article CAS PubMed Google Scholar
Street, T. O., Bolen, D. W. & Rose, G. D. A molecular mechanism for osmolyte-induced protein stability. Proc. Natl Acad. Sci. USA 103, 13997–14002 (2006).
Article ADS CAS PubMed PubMed Central Google Scholar
Wang, A. & Bolen, D. W. A naturally occurring protective system in urea-rich cells: mechanism of osmolyte protection of proteins against urea denaturation. Biochemistry 36, 9101–9108 (1997).
Article CAS PubMed Google Scholar
Liu, Y. & Bolen, D. W. The peptide backbone plays a dominant role in protein stabilization by naturally occurring osmolytes. Biochemistry 34, 12884–12891 (1995).
Article CAS PubMed Google Scholar
Baskakov, I., Wang, A. & Bolen, D. W. Trimethylamine-N-oxide counteracts urea effects on rabbit muscle lactate dehydrogenase function: a test of the counteraction hypothesis. Biophys. J. 74, 2666–2673 (1998).
Article ADS CAS PubMed PubMed Central Google Scholar
Lee, J. C. & Timasheff, S. N. The stabilization of proteins by sucrose. J. Biol. Chem. 256, 7193–7201 (1981).
Article CAS PubMed Google Scholar
Arakawa, T. & Timasheff, S. N. Stabilization of protein structure by sugars. Biochemistry 21, 6536–6544 (1982).
Article CAS PubMed Google Scholar
Timasheff, S. N. The control of protein stability and association by weak interactions with water: how do solvents affect these processes? Annu. Rev. Biophys. Biomol. Struct. 22, 67–97 (1993).
Bolen, D. W. & Rose, G. D. Structure and energetics of the hydrogen-bonded backbone in protein folding. Annu. Rev. Biochem. 77, 339–362 (2008).
Article CAS PubMed Google Scholar
Canchi, D. R. & García, A. E. Cosolvent effects on protein stability. Annu. Rev. Phys. Chem. 64, 273–293 (2013).
Article ADS CAS PubMed Google Scholar
Diamant, S., Eliahu, N., Rosenthal, D. & Goloubinoff, P. Chemical chaperones regulate molecular chaperones in vitro and in cells under combined salt and heat stresses. J. Biol. Chem. 276, 39586–39591 (2001).
Article CAS PubMed Google Scholar
Patel, C. N. et al. Effects of molecular crowding by saccharides on α-chymotrypsin dimerization. Protein Sci. 11, 997–1003 (2002).
Article CAS PubMed PubMed Central Google Scholar
**e, G. & Timasheff, S. N. The thermodynamic mechanism of protein stabilization by trehalose. Biophys. Chem. 64, 25–43 (1997).
Article CAS PubMed Google Scholar
Manning, M. C., Liu, J., Li, T. & Holcomb, R. E. Rational design of liquid formulations of proteins. Adv. Protein Chem. Struct. Biol. 112, 1–59 (2018).
Haque, I., Singh, R., Moosavi-Movahedi, A. A. & Ahmad, F. Effect of polyol osmolytes on ΔG_D, the Gibbs energy of stabilisation of proteins at different pH values. Biophys. Chem. 117, 1–12 (2005).
Article CAS PubMed Google Scholar
Singh, R., Haque, I. & Ahmad, F. Counteracting osmolyte trimethylamine N-oxide destabilizes proteins at pH below its pK_a: measurements of thermodynamic parameters of proteins in the presence and absence of trimethylamine N-oxide. J. Biol. Chem. 280, 11035–11042 (2005).
Article CAS PubMed Google Scholar
Kim, Y.-S. et al. Effects of sucrose on conformational equilibria and fluctuations within the native-state ensemble of proteins. Protein Sci. 12, 1252–1261 (2003).
Article CAS PubMed PubMed Central Google Scholar
Kumar, R., Serrette, J. M. & Thompson, E. B. Osmolyte-induced folding enhances tryptic enzyme activity. Arch. Biochem. Biophys. 436, 78–82 (2005).
Article CAS PubMed Google Scholar
Chang, Y. C. & Oas, T. G. Osmolyte-induced folding of an intrinsically disordered protein: folding mechanism in the absence of ligand. Biochemistry 49, 5086–5096 (2010).
Article CAS PubMed Google Scholar
Samuel, D. et al. Proline inhibits aggregation during protein refolding. Protein Sci. 9, 344–352 (2008).
Article Google Scholar
Ignatova, Z. & Gierasch, L. M. Inhibition of protein aggregation in vitro and in vivo by a natural osmoprotectant. Proc. Natl Acad. Sci. USA 103, 13357–13361 (2006).
Article ADS CAS PubMed PubMed Central Google Scholar
Athawale, M. V., Dordick, J. S. & Garde, S. Osmolyte trimethylamine-N-oxide does not affect the strength of hydrophobic interactions: origin of osmolyte compatibility. Biophys. J. 89, 858–866 (2005).
Article ADS CAS PubMed PubMed Central Google Scholar
Liu, R., Barkhordarian, H., Emadi, S., Chan, B. P. & Sierks, M. R. Trehalose differentially inhibits aggregation and neurotoxicity of β-amyloid 40 and 42. Neurobiol. Dis. 20, 74–81 (2005).
Article PubMed Google Scholar
Knowles, T. P. J. et al. Kinetics and thermodynamics of amyloid formation from direct measurements of fluctuations in fibril mass. Proc. Natl Acad. Sci. USA 104, 10016–10021 (2007).
Article ADS CAS PubMed PubMed Central Google Scholar
Scaramozzino, F. et al. TMAO promotes fibrillization and microtubule assembly activity in the C-terminal repeat region of tau. Biochemistry 45, 3684–3691 (2006).
Article CAS PubMed Google Scholar
Bhat, M. Y., Singh, L. R. & Dar, T. A. Trimethylamine N-oxide abolishes the chaperone activity of α-casein: an intrinsically disordered protein. Sci. Rep. 7, 6572(2017).
Article ADS PubMed PubMed Central Google Scholar
Tseng, H. C. & Graves, D. J. Natural methylamine osmolytes, trimethylamine N-oxide and betaine, increase tau-induced polymerization of microtubules. Biochem. Biophys. Res. Commun. 250, 726–730 (1998).
Article CAS PubMed Google Scholar
Nandi, P. K., Bera, A. & Sizaret, P. Y. Osmolyte trimethylamine N-oxide converts recombinant α-helical prion protein to its soluble β-structured form at high temperature. J. Mol. Biol. 362, 810–820 (2006).
Article CAS PubMed Google Scholar
Naik, V., Kardani, J. & Roy, I. Trehalose-induced structural transition accelerates aggregation of α-synuclein. Mol. Biotechnol. 58, 251–255 (2016).
Article CAS PubMed Google Scholar
Levine, Z. A., Larini, L., LaPointe, N. E., Feinstein, S. C. & Shea, J. E. Regulation and aggregation of intrinsically disordered peptides. Proc. Natl Acad. Sci. USA 112, 2758–2763 (2015).
Article ADS CAS PubMed PubMed Central Google Scholar
Macchi, F., Eisenkolb, M., Kiefer, H. & Otzen, D. E. The effect of osmolytes on protein fibrillation. Int. J. Mol. Sci. 13, 3801–3819 (2012).
Article CAS PubMed PubMed Central Google Scholar
Borwankar, T. et al. Natural osmolytes remodel the aggregation pathway of mutant huntingtin exon 1. Biochemistry 50, 2048–2060 (2011).
Article CAS PubMed Google Scholar
Bolen, D. W. & Baskakov, I. V. The osmophobic effect: natural selection of a thermodynamic force in protein folding. J. Mol. Biol. 310, 955–963 (2001).
Article CAS PubMed Google Scholar
Zhou, H. X. Influence of crowded cellular environments on protein folding, binding, and oligomerization: biological consequences and potentials of atomistic modeling. FEBS Lett. 587, 1053–1061 (2013).
Article CAS PubMed PubMed Central Google Scholar
Piazza, I. et al. A map of protein–metabolite interactions reveals principles of chemical communication. Cell 172, 358–372 (2018).
Article CAS PubMed Google Scholar
Leuenberger, P. et al. Cell-wide analysis of protein thermal unfolding reveals determinants of thermostability. Science 355, eaai7825 (2017).
Article PubMed Google Scholar
Feng, Y. et al. Global analysis of protein structural changes in complex proteomes. Nat. Biotechnol. 32, 1036–1044 (2014).
Article CAS PubMed Google Scholar
Strøm, A. R., Falkenberg, P. & Landfald, B. Genetics of osmoregulation in Escherichia coli: uptake and biosynthesis of organic osmolytes. FEMS Microbiol. Lett. 39, 79–86 (1986).
Article Google Scholar
Larsen, P. I., Sydnes, L. K., Landfald, B. & Strøm, A. R. Osmoregulation in Escherichia coli by accumulation of organic osmolytes: betaines, glutamic acid, and trehalose. Arch. Microbiol. 147, 1–7 (1987).
Article CAS PubMed Google Scholar
Schopper, S. et al. Measuring protein structural changes on a proteome-wide scale using limited proteolysis-coupled mass spectrometry. Nat. Protoc. 12, 2391–2410 (2017).
Article CAS PubMed Google Scholar
Mateus, A. et al. Thermal proteome profiling in bacteria: probing protein state in vivo. Mol. Syst. Biol. 14, e8242 (2018).
Article PubMed PubMed Central Google Scholar
De Virgilio, C., Hottiger, T., Dominguez, J., Boller, T. & Wiemken, A. The role of trehalose synthesis for the acquisition of thermotolerance in yeast: I. Genetic evidence that trehalose is a thermoprotectant. Eur. J. Biochem. 219, 179–186 (1994).
Article PubMed Google Scholar
Hottiger, T., De Virgilio, C., Hall, M. N., Boller, T. & Wiemken, A. The role of trehalose synthesis for the acquisition of thermotolerance in yeast: II. Physiological concentrations of trehalose increase the thermal stability of proteins in vitro. Eur. J. Biochem. 219, 187–193 (1994).
Article CAS PubMed Google Scholar
Miller, B. G. & Raines, R. T. Identifying latent enzyme activities: substrate ambiguity within modern bacterial sugar kinases. Biochemistry 43, 6387–6392 (2004).
Article CAS PubMed Google Scholar
Sigrell, J. A., Cameron, A. D. & Mowbray, S. L. Induced fit on sugar binding activates ribokinase. J. Mol. Biol. 290, 1009–1018 (1999).
Article CAS PubMed Google Scholar
Bhat, M. Y., Malik, M. A., Singh, L. R. & Dar, T. A. Trimethylamine N-oxide alters structure-function integrity of β-casein: structural disorder co-regulates the aggregation propensity and chaperone activity. Int. J. Biol. Macromol. 182, 921–930 (2021).
Article CAS PubMed Google Scholar
Verma, G., Singh, P. & Bhat, R. Disorder under stress: role of polyol osmolytes in modulating fibrillation and aggregation of intrinsically disordered proteins. Biophys. Chem. 264, 106422 (2020).
Article CAS PubMed Google Scholar
Disfani, F. M. et al. MoRFpred, a computational tool for sequence-based prediction and characterization of short disorder-to-order transitioning binding regions in proteins. Bioinformatics 28, i75–i83 (2012).
Article CAS PubMed PubMed Central Google Scholar
Liu, Y. et al. Periplasmic proteins of Escherichia coli are highly resistant to aggregation: reappraisal for roles of molecular chaperones in periplasm. Biochem. Biophys. Res. Commun. 316, 795–801 (2004).
Article CAS PubMed Google Scholar
Ward, J. J., Sodhi, J. S., McGuffin, L. J., Buxton, B. F. & Jones, D. T. Prediction and functional analysis of native disorder in proteins from the three kingdoms of life. J. Mol. Biol. 337, 635–645 (2004).
Article CAS PubMed Google Scholar
Uversky, V. N. Wrecked regulation of intrinsically disordered proteins in diseases: pathogenicity of deregulated regulators. Front. Mol. Biosci. 1, 6 (2014).
Paronetto, M. P. Ewing sarcoma protein: a key player in human cancer. Int. J. Cell Biol. 2013, 642853 (2013).
Auton, M. & Bolen, D. W. Prediction the energetics of osmolyte-induced protein folding/unfolding. Proc. Natl Acad. Sci. USA 102, 15065–15068 (2005).
Article ADS CAS PubMed PubMed Central Google Scholar
Uversky, V. N., Li, J. & Fink, A. L. Trimethylamine-N-oxide-induced folding of α-synuclein. FEBS Lett. 509, 31–35 (2001).
Article CAS PubMed Google Scholar
Lin, T. Y. & Timasheff, S. N. Why do some organisms use a urea–methylamine mixture as osmolyte? Thermodynamic compensation of urea and trimethylamine N-oxide interactions with protein. Biochemistry 33, 12695–12701 (1994).
Article CAS PubMed Google Scholar
Ganguly, P., Boserman, P., Van Der Vegt, N. F. A. & Shea, J. E. Trimethylamine N-oxide counteracts urea denaturation by inhibiting protein–urea preferential interaction. J. Am. Chem. Soc. 140, 483–492 (2018).
Article CAS PubMed Google Scholar
Laxson, C. J., Condon, N. E., Drazen, J. C. & Yancey, P. H. Decreasing urea: trimethylamine N-oxide ratios with depth in chondrichthyes: a physiological depth limit? Physiol. Biochem. Zool. 84, 494–505 (2011).
Article CAS PubMed Google Scholar
Bhat, M. Y., Mir, I. A., Ul Hussain, M., Singh, L. R. & Dar, T. A. Urea ameliorates trimethylamine N-oxide-induced aggregation of intrinsically disordered α-casein protein: the other side of the urea–methylamine counteraction. J. Biomol. Struct. Dyn. 41, 3659–3666 (2022).
Article PubMed Google Scholar
Yancey, P. H., Fyfe-Johnson, A. L., Kelly, R. H., Walker, V. P. & Aun, M. T. Trimethylamine oxide counteracts effects of hydrostatic pressure on proteins of deep-sea teleosts. J. Exp. Zool. 289, 172–176 (2001).
Article CAS PubMed Google Scholar
Yancey, P. H. Cellular responses in marine animals to hydrostatic pressure. J. Exp. Zool. A Ecol. Integr. Physiol. 333, 398–420 (2020).
Holmström, K. O., Somersalo, S., Mandal, A., Palva, T. E. & Welin, B. Improved tolerance to salinity and low temperature in transgenic tobacco producing glycine betaine. J. Exp. Bot. 51, 177–185 (2000).
Article PubMed Google Scholar
Narayanan, H. et al. Design of biopharmaceutical formulations accelerated by machine learning. Mol. Pharm. 18, 3843–3853 (2021).
Article CAS PubMed Google Scholar
Lee, J. et al. Trehalose glycopolymers as excipients for protein stabilization. Biomacromolecules 14, 2561–2569 (2013).
Article CAS PubMed PubMed Central Google Scholar
Koeth, R. A. et al. Intestinal microbiota metabolism of l-carnitine, a nutrient in red meat, promotes atherosclerosis. Nat. Med. 19, 576–585 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Wilson, C. J., Choy, W. Y. & Karttunen, M. AlphaFold2: a role for disordered protein/region prediction? Int. J. Mol. Sci. 23, 4591 (2022).
Article CAS PubMed PubMed Central Google Scholar
Baba, T. et al. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2, 2006.0008 (2006).
Fischer, E., Zamboni, N. & Sauer, U. High-throughput metabolic flux analysis based on gas chromatography–mass spectrometry derived ¹³C constraints. Anal. Biochem. 325, 308–316 (2004).
Article CAS PubMed Google Scholar
Malinovska, L. et al. Proteome-wide structural changes measured with limited proteolysis–mass spectrometry: an advanced protocol for high-throughput applications. Nat. Protoc. 18, 659–682 (2022).
Kitagawa, M. et al. Complete set of ORF clones of Escherichia coli ASKA library (a complete set of E. coli K-12 ORF archive): unique resources for biological research. DNA Res. 12, 291–299 (2005).
Article CAS PubMed Google Scholar
Bolen, D. W. & Santoro, M. M. Unfolding free energy changes determined by the linear extrapolation method. Biochemistry 27, 8063–8068 (1988).
Article PubMed Google Scholar
Kopp, M. R. G., Villois, A., Capasso Palmiero, U. & Arosio, P. Microfluidic diffusion analysis of the size distribution and microrheological properties of antibody solutions at high concentrations. Ind. Eng. Chem. Res. 57, 7112–7120 (2018).
Article CAS Google Scholar
Bruderer, R. et al. Extending the limits of quantitative proteome profiling with data-independent acquisition and application to acetaminophen-treated three-dimensional liver microtissues. Mol. Cell. Proteom. 14, 1400–1410 (2015).
Article CAS Google Scholar
Quast, J.-P., Schuster, D. & Picotti, P. protti: an R package for comprehensive data analysis of peptide- and protein-centric bottom-up proteomics data. Bioinform. Adv. 2, vbab041 (2022).
Article PubMed Google Scholar
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300 (1995).
MathSciNet Google Scholar
Rasmussen, C. E. & Williams, C. K. I. Gaussian Processes for Machine Learning (MIT Press, 2005).
Gardner, J. R., Pleiss, G., Bindel, D., Weinberger, K. Q. & Wilson, A. G. In Advances in Neural Information Processing Systems (eds Bengio, S. et al.) 7587–7597 (Curring Associates, 2018).
Ferraro, M. B., Giordani, P. & Serafini, A. Fclust: an R package for fuzzy clustering. R Journal 11, 1–18 (2019).
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhang, J. & Kurgan, L. SCRIBER: accurate and partner type-specific prediction of protein-binding residues from proteins sequences. Bioinformatics 35, i343–i353 (2019).
Article CAS PubMed PubMed Central Google Scholar
Yan, J. & Kurgan, L. DRNApred, fast sequence-based method that accurately predicts and discriminates DNA- and RNA-binding residues. Nucleic Acids Res. 45, e84 (2017).
PubMed PubMed Central Google Scholar
Zhao, B. et al. DescribePROT: database of amino acid-level protein structure and function predictions. Nucleic Acids Res. 49, D298–D308 (2021).
Article ADS CAS PubMed Google Scholar
Sormanni, P., Aprile, F. A. & Vendruscolo, M. The CamSol method of rational design of protein mutants with enhanced solubility. J. Mol. Biol. 427, 478–490 (2015).
Article CAS PubMed Google Scholar
Fernandez-Escamilla, A. M., Rousseau, F., Schymkowitz, J. & Serrano, L. Prediction of sequence-dependent and mutational effects on the aggregation of peptides and proteins. Nat. Biotechnol. 22, 1302–1306 (2004).
Article CAS PubMed Google Scholar
Alexa, A. & Rahnenführer, J. Gene set enrichment analysis with topGO. Bioconductor https://bioconductor.org/packages/release/bioc/vignettes/topGO/inst/doc/topGO.pdf (2007).
Mistry, J. et al. Pfam: the protein families database in 2021. Nucleic Acids Res. 49, D412–D419 (2021).
Article CAS PubMed Google Scholar
Wickham, H. ggplot2. WIRES Comput. Stat. 3, 180–185 (2011).
Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32, 2847–2849 (2016).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

We thank members of the Picotti lab and M. Savitski (EMBL Heidelberg) for useful discussions. P.P. was supported by funding from the European Research Council (grant number 866004) under the EU Horizon 2020 research and innovation program, by a Personalized Health and Related Technologies grant (PHRT-506) and by a National Center of Competence in Research AntiResist grant from the SNSF (grant number 51NF40_180541).

Funding

Open access funding provided by Swiss Federal Institute of Technology Zurich.

Author information

Authors and Affiliations

Department of Biology, Institute of Molecular Systems Biology, ETH Zurich, Zurich, Switzerland
Monika Pepelnjak, Nicolas Näpflin, Natalie de Souza & Paola Picotti
Division of Computational Genomics and Systems Genetics, German Cancer Research Center (DKFZ), Heidelberg, Germany
Britta Velten
Centre for Organismal Studies (COS) & Center for Scientific Computing (IWR), Heidelberg University, Heidelberg, Germany
Britta Velten
Department of Biology, Institute of Molecular Biology & Biophysics, ETH Zurich, Zurich, Switzerland
Tatjana von Rosen & Eilika Weber-Ban
Department of Chemistry and Applied Biosciences, Institute of Chemical and Bioengineering, ETH Zurich, Zurich, Switzerland
Umberto Capasso Palmiero & Paolo Arosio
Department of Chemistry and Biochemistry, University of California, Los Angeles, CA, USA
Jeong Hoon Ko & Heather D. Maynard
Department of Quantitative Biomedicine, University of Zurich, Zurich, Switzerland
Natalie de Souza
Genome Biology Unit, European Molecular Biological Laboratory, Heidelberg, Germany
Wolfgang Huber

Authors

Monika Pepelnjak
View author publications
You can also search for this author in PubMed Google Scholar
Britta Velten
View author publications
You can also search for this author in PubMed Google Scholar
Nicolas Näpflin
View author publications
You can also search for this author in PubMed Google Scholar
Tatjana von Rosen
View author publications
You can also search for this author in PubMed Google Scholar
Umberto Capasso Palmiero
View author publications
You can also search for this author in PubMed Google Scholar
Jeong Hoon Ko
View author publications
You can also search for this author in PubMed Google Scholar
Heather D. Maynard
View author publications
You can also search for this author in PubMed Google Scholar
Paolo Arosio
View author publications
You can also search for this author in PubMed Google Scholar
Eilika Weber-Ban
View author publications
You can also search for this author in PubMed Google Scholar
Natalie de Souza
View author publications
You can also search for this author in PubMed Google Scholar
Wolfgang Huber
View author publications
You can also search for this author in PubMed Google Scholar
Paola Picotti
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

M.P. designed and performed LiP experiments, designed the computational pipeline and performed bioinformatics analysis. B.V. designed part of the computational pipeline and statistical analysis. N.N. performed bioinformatics analyses. T.v.R. performed validations with purified proteins. U.C.P. performed viscosity measurements. J.H.K. and H.D.M. contributed to experimental design. N.d.S., M.P. and P.P. wrote the manuscript with contributions from all authors. N.d.S. and P.P. supervised the revision process. E.W.-B., W.H. and P.A. supervised parts of the project. P.P. conceived and supervised the project.

Corresponding author

Correspondence to Paola Picotti.

Ethics declarations

Competing interests

P.P. is an inventor of a patent licensed by Biognosys AG that covers the LiP–MS method used in this manuscript (patent number WO-A-2014082733). P.P. is also a scientific advisor of Biognosys AG. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Chemical Biology thanks Markus Schirle and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Figs. 1–6 and Note.

Reporting Summary

Supplementary Data 1

Analysis of thermal stabilization by osmolytes.

Supplementary Data 2

Binding analysis following the addition of osmolytes.

Supplementary Data 3

Analysis of osmolyte effects on protein aggregation.

Supplementary Data 4

Analysis of osmolyte effects on the human proteome.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Pepelnjak, M., Velten, B., Näpflin, N. et al. In situ analysis of osmolyte mechanisms of proteome thermal stabilization. Nat Chem Biol (2024). https://doi.org/10.1038/s41589-024-01568-7

Download citation

Received: 19 August 2022
Accepted: 03 February 2024
Published: 29 February 2024
DOI: https://doi.org/10.1038/s41589-024-01568-7
Springer Nature America, Inc.

In situ analysis of osmolyte mechanisms of proteome thermal stabilization

Abstract

Similar content being viewed by others

Main

Direct binding to proteins causes strong stabilization

Role of osmolytes in aggregation

Osmolyte effects on the human proteome

Discussion

Methods

Escherichia coli

HEK293T cells

Cell lysate preparation

E. coli

HEK293T cells

Preparation of standard control sample of short peptides

Preparation of osmolyte stock solutions

Limited proteolysis

Thermal proteome profiling

Tryptic digestion

Liquid chromatography–tandem mass spectrometry data acquisition

Data-dependent acquisition

Data-independent acquisition

Protein expression and purification

Differential scanning fluorimetry for whole-cell lysate

Differential scanning fluorimetry for purified proteins

Circular dichroism

Precipitation analysis for ribosome recycling factor

Thermal proteome profiling for whole-cell lysates

Extended proteolysis assay

Detergent resolubilization assay

Viscosity measurements

Peptide- and protein-level quantification

Fully tryptic and half tryptic peptides

Correlation analysis for fully tryptic and half tryptic pairs

Testing and correcting for condition effects on proteinase K activity

Learning peptide intensity profiles across temperatures

Clustering of profiles

Calculation of peptide- and protein-level scores

Comparison of stabilization strength for known binders

LiP–MS binding analysis

Calculation of protein characteristics

Identification of nonprecipitators

Analysis of differential scanning fluorimetry data

Analysis of thermal protein profiling data

Spearman correlation analysis

Identification of proteins with changed aggregation abundance

Gene ontology enrichment analysis

Domain-level analysis

Data visualization and processing

Reporting summary

Data availability

Code availability

References

Acknowledgements

Funding

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Rights and permissions

About this article

Cite this article

Share this article

Search

Navigation