O‐GlcNAc modification of HSP27 alters its protein interactions and promotes refolding of proteins through the BAG3/HSP70 co‐chaperone

Abstract

Almost all types of cellular stress induce post‐translational O‐GlcNAc modifications of proteins, and this increase promotes cell survival. We previously demonstrated that O‐GlcNAc on certain small heat shock proteins (sHSPs), including HSP27, directly increases their chaperone activity as one potential protective mechanism. Here, we furthered our use of synthetic proteins to prepare biotinylated sHSPs and show that O‐GlcNAc modification of HSP27 also changes how it interacts within the sHSP system and the broader HSP network. Specifically, we show that O‐GlcNAc modified HSP27 binds more strongly to the co‐chaperone protein BAG3, which then promotes refolding of a model substrate by HSP70. We use proteomics to identify other potential HSP27 interactions that are changed by O‐GlcNAc, including one that we confirm with another sHSP, αB‐crystallin. These findings add additional evidence for O‐GlcNAc as a switch for regulating protein–protein interactions and for modifications of chaperones as one mechanism by which O‐GlcNAc protects against protein aggregation.

1. INTRODUCTION

Posttranslational modifications (PTMs) enable cells to dynamically respond to changes in their environment through alterations in protein biochemistry, stability, localization, and so forth. O‐GlcNAc modification is a PTM found on intracellular proteins in the cytosol, mitochondria, and nucleus (Ma et al., 2021; Yang & Qian, ²⁰¹⁷; Zachara, ²⁰¹⁸). This addition of the simple monosaccharide N‐acetylglucosamine is controlled by the action of two enzymes, O‐GlcNAc transferase (OGT) and O‐GlcNAcase (OGA), which add and remove the sugar respectively (King et al., 2019). In mammals, O‐GlcNAc is required for proper cell function and survival and alterations in the overall levels of the modification are associated with human disease. For example, O‐GlcNAc is elevated in almost all types of cancer compared to healthy tissue (Ferrer et al., 2016; Hanover et al., ²⁰¹⁸) but is lost in Alzheimer's diseased brains compared to age‐matched controls (Balana & Pratt, 2021; Pratt & Vocadlo, ²⁰²³). Given these disease associations, it is not surprising that O‐GlcNAc in increased under the induction of various cell stressors, and this increase plays a protective role (Fahie et al., 2022; Groves et al., ²⁰¹³). However, the exact molecular mechanisms by which O‐GlcNAc exerts this protective function remain mysterious.

Another key player in cell stress regulation is the heat shock proteins (Hartl et al., 2011; Hu et al., ²⁰²²; Kregel, ²⁰⁰²). Broadly speaking, this class of proteins can be divided into the ATP‐dependent and ATP‐independent chaperones. The ATP‐dependent chaperones include the HSP70 and 90 families and actively hydrolyze ATP to drive rounds of protein disaggregation and refolding (Krukenberg et al., 2011; Zuiderweg et al., ²⁰¹²). In contrast to these “foldases,” the ATP‐independent chaperones include the small heat‐shock proteins (sHSPs), which function as “holdases” by binding to misfolded protein species and preventing their further aggregation (Haslbeck et al., 2019; Reinle et al., ²⁰²²). We discovered that O‐GlcNAc modification of a subset of sHSPs—HSP27 (gene name HSPB1), αA‐crystallin (HSPB4), and αB‐crystallin (HSPB5)—increases their chaperone activity (Figure 1a) (Balana et al., 2021). These sHSPs function as large heterogeneous oligomers where each monomer contains three domains (Figure S1) (Kappé et al., 2003; Kriehuber et al., ²⁰¹⁰). The N‐terminus is largely unstructured and is important for oligomer formation. The central α‐crystallin domain (ACD) has a cleft formed by two β‐strands that is largely responsible for binding hydrophobic regions of misfolded proteins. Finally, the C‐terminus is also unstructured in solution and contains an IXI‐motif (IPV in HSP27) that can bind‐back to the ACD cleft, thereby regulating the activity of the chaperone (Figure 1a) (Baldwin et al., 2011; Delbecq et al., ²⁰¹²; Freilich et al., ²⁰¹⁸; Jehle et al., ²⁰¹⁰). O‐GlcNAc is found on all three sHSPs at a threonine residue immediately following the IXI (Deracinois et al., 2018; Li et al., ²⁰¹⁹; Roquemore et al., ¹⁹⁹²; Wang et al., ²⁰¹⁷) and functions by inhibiting the interaction between this motif and the ACD cleft (Balana et al., 2021). This frees the cleft to more readily bind client proteins and therefore increased the chaperone activity. It also increases the heterogeneity of the HSP27 oligomer and its average size by ~2‐fold, as determined by size exclusion chromatography, confirming a role for O‐GlcNAc in altering the activity and structure of this key chaperone.

FIGURE 1.

O‐GlcNAc modification and HSP27. (a) O‐GlcNAc modification, the addition on N‐acetylglucosamine to serine/threonine residues, occurs near the IXI‐motif of HSP27 and activates the chaperone. (b) HSP27 can cooperate with HSP70 to refold substrates through interactions with the co‐chaperone BAG3. (c) Synthesis of N‐terminally biotinylated HSP27 and its major O‐GlcNAc modified variant HSP27(gT184).

Notably, the different components of the heat shock protein system do not function as isolated islands, but rather, they physically interact and collaborate to protect cells from protein‐folding stress. This includes an important interaction between sHSPs, including HSP27 and αBC, and HSP70. Hsp70 is well‐established to regulate protein refolding, and we found that sHSPs can direct unfolded proteins to HSP70 for refolding through interactions mediated by the scaffolding protein BAG3 (Rauch et al., 2017). BAG3 is a member of the Bcl‐2‐associated anthanogene (BAG) family of co‐chaperones. BAG3 interacts with the nucleotide‐binding domain of HSP70 and increases the release of ADP, thus assisting in the catalytic refolding cycle (Brehmer et al., 2001; Rampelt et al., ²⁰¹¹). BAG3 also contains two IPV sequences that we reasoned could bind to sHSP ACD‐clefts, localizing them to HSP70, and bridging the ATP‐independent and ATP‐dependent chaperone systems (Carra et al., 2008; Fuchs et al., ²⁰⁰⁹; Hishiya et al., ²⁰¹¹). In our prior work, we tested this possibility in vitro and found that BAG3 could indeed use its IPV sequences to interact with the ACD of various sHSPs and participate in co‐localization, potential hand‐off, and refolding of model substrates like firefly luciferase (Rauch et al., 2017). O‐GlcNAc modification results in a higher level of free ACD clefts in HSP27 and this work showed that these ACD clefts are important for interactions with BAG3. Therefore, we hypothesized that it may also increase the interaction with BAG3 and facilitate communication with HSP70, and we made the HSP27‐BAG3 interaction the initial focus of this study.

Here, we use synthetic proteins to confirm this hypothesis. First, we generated N‐terminally biotinylated versions of either unmodified HSP27 or HSP27 with O‐GlcNAc at threonine 184 (HSP27(gT184)). We chose this modification site because it has been found the most often in proteomics experiments and lies within a primary sequence favored by OGT. We then used the biotinylated proteins to enrich binding partners from cell lysates and found increased interaction between HSP27(gT184) and BAG3. Using in vitro assays, we demonstrated that HSP27 O‐GlcNAc modification does indeed improve the refolding of luciferase by HSP70 in the presence of BAG3. We then used our synthetic proteins to perform an unbiased enrichment and identification experiment by mass spectrometry‐based proteomics, and identified several interactions that were regulated by O‐GlcNAc in either positive or negative directions. Finally, we validated and characterized one of these interactions, the formation of a HSP27/αBC heterooligomer, in vitro. These results further affirm O‐GlcNAc's multifunctional role in preventing and resolving protein misfolding/aggregation, particularly when combined with our previous results on the ability for O‐GlcNAc to directly prevent protein aggregation and activate sHSPs.

2. RESULTS AND DISCUSSION

To facilitate the robust enrichment of HSP27‐binding proteins, we chose to take advantage of biotin/neutravidin. We were concern, however, that non‐selective labeling of lysine residues in HSP27 with NHS‐biotin or similar reagents might disrupt the oligomer structure. Therefore, we took advantage of the Francis 2‐PCA N‐terminal modification chemistry (MacDonald et al., 2015) as the N‐terminus of HSP27 has been tagged in the past and is freely exposed in the one crystal structure of an HSP27 oligomer (Nappi et al., 2020). Our specific synthetic strategy is outlined in Figure 1c. Briefly, HSP27(gT184) was generated from an N‐terminal recombinant thioester and an O‐GlcNAc‐modified peptide through ligation chemistry as previously described (Balana et al., 2021). Unlike other approaches, this method allows creation of uniformly modified HSP27, facilitating insights into the Roels of O‐GlcNAc. This protein, along with fully‐recombinant, unmodified HSP27 from E. coli, were separately subjected to 2‐PCA conditions, purified by RP‐HPLC, and characterized by ESI‐MS (Figure S2). In addition to the major product‐peak in the mass spectra, we also observed a minor peak corresponding to non‐biotinylated starting material but were unable to separate these species by RP‐HPLC. However, because HSP27 functions as a large oligomer, we reasoned that incomplete biotinylation would not prevent us from moving forward. After refolding, the majority of monomers in the oligomer would be modified by biotin, allowing the entire chaperone and its interactors to be captured by streptavidin.

With these proteins in hand, we generated HeLa cell‐lysate and individually added our proteins in triplicate for 1 h at 37°C to initiate any protein–protein interactions. We then either directly performed an enrichment using the biotin‐handle or pretreated the mixed lysates with a disulfide‐linked chemical crosslinker (Lomant's reagent) to trap any transient interactions. Next, we eluted the enriched proteins using denaturing conditions with a reducing agent. We visualized the presence of BAG3 using western blotting and quantified these results by normalizing to inputs (Figure 2). In both experiments, we found that O‐GlcNAc significantly increased the interaction with BAG3, consistent with our hypothesis that O‐GlcNAc‐modification would result in more accessible ACD clefts able to interact with BAG3 IPV‐motifs.

FIGURE 2.

O‐GlcNAc increases the interaction between HSP27 and BAG3. The indicated biotinylated HSP27‐proteins were separately incubated with HeLa cell‐lysate before (a) enrichment or (b) crosslinking with Lomant's reagent followed by enrichment. The interacting proteins were then eluted and BAG3‐levels were measured by western blotting. Each lane represents an individual replicate. Quantification was performed by comparing pulldown to the respective input signals. Results are mean ± SEM of these experimental replicates (n = 3). Statistical significance was determined using a two‐tailed, unpaired t‐test.

As mentioned above, we previously published that even substoichiometric concentrations BAG3 will increase the release of ADP bound to HSP70 and overcome this rate limiting step in the chaperone cycle (Rauch et al., 2017). Similarly, at these lower concentrations BAG3 facilitates the refolding of the commonly used client protein luciferase by HSP70 and another sHSP, HSP22 (Figure S3). However, we found that BAG3 acts as an inhibitor at higher concentrations, suggesting that the correct cellular ratios of these proteins are important for controlling the rate of protein refolding. Here, we first tested whether HSP27 or HSP27(gT184) would affect HSP70‐mediated refolding of luciferase in the absence of BAG3 (Figure 3a). Interestingly, we found that increasing concentrations of HSP27 reduced luciferase refolding, a phenomenon we attribute to the holdase activity of the sHSP that sequesters unfolded luciferase away from HSP70. Although not statistically significant at all concentrations, we found a slight increase in the inhibitory potency of HSP27(gT184), potentially due to more binding of luciferase to the vacant ACD clefts. We then performed a similar experiment in the presence of different concentrations of BAG3 (Figure 3b). As expected, we found that BAG3 increased refolding at low to moderate concentrations and this activation was lost at the highest concentration tested (1 μM). Again, consistent with our previous data (Rauch et al., 2017), we found that increasing concentrations of HSP27 improved luciferase folding. As a general trend, we discovered that O‐GlcNAc modification further facilitated protein refolding, as HSP27(gT184) had a more pronounced effect. This was particularly evident at moderate equimolar concentrations of HSP27 and BAG3 (0.25 or 0.5 μM, Figure 3c) but could be seen across most of the concentration range. Together, these data indicate that O‐GlcNAc‐modification not only activates the holdase function of HSP27 in accordance with our previous work (Balana et al., 2021), but also facilitates collaboration with the protein refolding system centered around HSP70.

FIGURE 3.

O‐GlcNAc improves luciferase‐refolding by the HSP70/BAG3 chaperone complex. (a) O‐GlcNAc enhances HSP27‐dependent inhibition of luciferase refolding by HSP70. The indicated concentrations of HSP27 or HSP27(gT184) were separately added to HSP70 and unfolded luciferase for 1 h before measuring luciferase activity. (b) O‐GlcNAc increases refolding of luciferase in the presence of BAG3. The indicated combinations of BAG3 and either HSP27 or HSP27(gT184) were incubated with HSP70 and unfolded luciferase for 1 h before measuring luciferase activity. The data from BAG3 concentrations of 0.25 or 0.5 μM are shown as bar graphs for clarity. The y‐axis is all experiments is normalized such that 100% luminescence is signal in the absence of BAG3 for each condition. In all experiments, results are mean ± SEM of experimental replicates (n = 3). Statistical significance was determined using a two‐tailed, unpaired t‐test.

Given the altered interaction between HSP27(gT184) and BAG3, we next wondered whether O‐GlcNAc may globally change the HSP27 interactome. This is supported by recent studies on mutants of HSP27 that cause Charcot–Marie‐Tooth type 2 (CMT2) disease (Alderson et al., 2021; Chalova et al., ²⁰¹⁴). Specifically, these mutants are found within or near the IXI‐motif of HSP27 and can alter the IXI–ACD interaction with deleterious consequences for HSP27 activity. In these studies, it was found that changes to this interaction altered HSP27 binding to several proteins including BAG3. To test this possibility, we again took advantage of our synthetic, biotinylated proteins and performed non‐crosslinked enrichment from HeLa lysates as described above. We then analyzed the full spectrum of enriched proteins by SDS–PAGE, observing qualitative differences in protein enrichment (Figure 4a). Accordingly, we performed this enrichment in triplicate and subjected the proteins to proteomic analysis following separation by SDS–PAGE and in‐gel digestion and extraction. The proteomic data were then curated as follows. First, we required proteins to be identified in all three enrichment experiments from either HSP27 or HSP27(gT184). We then use label‐free quantification to calculate the mean area for proteins from each condition and removed any protein that was not enriched 5‐fold over no‐added HSP27 control. We chose to use no‐added HSP27 as a control instead of adding a random, control proteins, as we were considered that this control might share interactions with HSP27. Finally, we generated a log2 enrichment ratio to compare proteins pulled‐down by HSP27, HSP27(gT184), or both (Figure 4b, Table S1). After this data curation, we found 12 proteins that showed bias towards HSP27(gT184), 164 that showed bias towards HSP27, and 117 that did not discriminate more than 2‐fold between the proteins. Gratifyingly, we found BAG3 to be enriched towards HSP27(gT184) (Figure 4b).

FIGURE 4.

O‐GlcNAc affects other HSP27 protein–protein interactions, including with αBC. (a) The indicated biotinylated HSP27‐proteins were separately incubated with HeLa cell‐lysate before enrichment, separation by SDS–PAGE and Coomassie staining. (b) Enriched proteins were identified by label‐free quantitative proteomics and characterized in their bias towards HSP27 or HSP27(gT184). (c) αBC preferentially forms heterooligomers with HSP27 under competition with HSP27(gT184). The indicated proteins were incubated together before enrichment using biotin and analysis by Coomassie staining. Each lane represents an individual replicate. Quantification was performed by comparing pulldown to the respective input signals. Results are mean ± SEM of experimental replicates (n = 3). Statistical significance was determined using a two‐tailed, unpaired t‐test.

To confirm these results further, we decided to focus on a known interaction between HSP27 and another sHSP, αB‐crystallin (αBC), which have been demonstrated in multiple studies to form heterooligomeric chaperones (Aquilina et al., 2013; Bova et al., ²⁰⁰⁰; Fontaine et al., ²⁰⁰⁵). In our proteomics data, αBC was enriched towards the HSP27 pulldown (Figure 4b), suggesting that it somewhat prefers to form heterooligomers with the non‐O‐GlcNAc‐modified protein. The formation of sHSP heterooligomers is a complicated process where monomers can be exchanged during and after folding. Therefore, we decided to test the preference of αBC for HSP27 or HSP27(gT184) in a competitive fashion (Figure 4c). The chaperones are shown in the figure as tetramers for simplicity, but are in reality large, polydisperse oligomers. First, we first synthesized biotinylated αBC using the methods described above for HSP27 (Figure S4). We again, found a small amount of non‐biotinylated protein in the mass spectrum, but because αBC acts as an oligomer, we again reasoned would not prevent enrichment by streptavidin. We envisioned an experiment where this modified αBC would be mixed with an equimolar amount of non‐biotinylated HSP27 or HSP27(gT184) to allow for competitive heterooligomer formation. Using streptavidin enrichment, we could enriched the only biotinylated protein in the mixture, αBC, as well as any HSP27 or HSP27(gT184) that was part of the heterooligomer. Finally, we could quantify the enriched HSP27 or HSP27(gT184) relative to their respective inputs to measure the preference of αBC.

Unfortunately, we would not be able to distinguish between HSP27 and HSP27(gT184) using standard SDS–PAGE, as they run at the same molecular weight. However, if we first subject these proteins to chemoenzymatic mass‐tagging (Darabedian et al., 2018), HSP27(gT184) can be selectively modified with a 2 kDa PEG chain, resulting in a mass shift of this protein and separation from HSP27 on a gel. We first tested whether this mass‐shifting could proceed fully on HSP27(gT184) in isolation. As can be seen in Figure S5, chemoenzymatic modification of HSP27(gT184) followed by PEGylation using the strain‐promoted azide‐alkyne cycloaddition (SPAAC) for 16 h at 25°C resulted in a quantitative shift in HSP27(gT184) but did not alter the retention of HSP27 in SDS–PAGE. Therefore, we were confident that we could separate, visualize, and quantify all three proteins in the potential heterooligomers. With this important control completed, we proceeded to heterooligomer mixing followed by streptavidin enrichment. A fraction of the input and the elution were first denatured and subjected to chemoenzymatic mass‐shifting before SDS‐PAGE and Coomassie staining (Figure 4c). Quantification of the enriched HSP27 or HSP27(gT184) versus the corresponding inputs showed a preference of αBC towards HSP27, consistent with our proteomics results. Together, these data suggest that O‐GlcNAc modification may be a switch to dynamically change the HSP27 interactome with downstream consequences not only for the HSP70‐chaperone system, but also potentially on substrates and the sHSPs themselves.

3. CONCLUSIONS

Our results demonstrate that O‐GlcNAc modification of HSP27 may have multiple functions to protect cells from protein misfolding and aggregation. We previously found that O‐GlcNAc near the IXI‐motif of certain sHSPs directly activated their chaperone activity against peptides and proteins that form amyloid aggregates (Balana et al., 2021). The mechanism underlying this activation involves inhibition of an interaction between the IXI and ACD‐cleft of the sHSPs, resulting in more facile binding of this cleft to the client proteins. As mentioned in the introduction, we had previously demonstrated that a similar interaction can occur between IPV sequences in the scaffolding protein BAG3 and the ACD‐cleft of several sHSPs (Rauch et al., 2017). Here, we found that HSP27 O‐GlcNAc modification at threonine 184 also results in an increased physical interaction with the scaffolding protein BAG3, likely due to an increase in unoccupied ACD clefts on the HSP27 oligomer that are free to interact with the IPV sequences in BAG3. However, we did not directly test this mechanism again here. Building upon this result, we then demonstrated that O‐GlcNAc promoted the ability of HSP27 to collaborate with BAG3/HSP70 to refold renilla luciferase as a model substrate. We then went on to use an unbiased proteomics approach to identify other potential HSP27 interactions that might be affected by O‐GlcNAc. From analysis of these data, we found proteins that could be characterized as preferring the modified or unmodified state of the chaperone, as well as proteins that showed no preference one way or the other. We do not know the exact reason for these preferences but speculate that they could be driven by the ACD‐cleft, as is the case for BAG3, and/or differences in direct binding of proteins to HSP27's IXI‐motif in its unmodified versus O‐GlcNAc modified state. Changes in interactions could also be due to more factors beyond opening up the ACD cleft. As mentioned above, our previous analysis showed that the HSP27(gT184) oligomeric chaperones are more heterogeneous and lager (~47 vs. ~28 monomer units) (Balana et al., 2021). It is possible that these different configurations bind with different affinities to different proteins. This is supported by studies by Alderson et al. where small HSP27 dimers are hyperactive chaperones (Alderson et al., 2019) and our own work demonstrating that binding of tau competes with HSP27 oligomer formation (Freilich et al., 2018). Interrogating the biochemistry behind several of these preferences is something that we plan to pursue in the future. Finally, we biochemically characterized the interaction between HSP27 and αBC. As mentioned above, these two sHSPs can form mixed oligomers, and our proteomics data suggested that this interaction may slightly favor the unmodified HSP27. Using enrichments in a competitive format, we found that αBC does indeed seem to prefer forming mixed oligomers with unmodified HSP27, although O‐GlcNAc modified, mixed oligomers can still form. Interestingly αBC can also be O‐GlcNAc modified at the same conserved threonine residue next to its IXI‐motif, making the potential cross‐talk between these two modifications another interesting area for further exploration. We do not necessarily see any direct biological association between these two classes of interactions altered by O‐GlcNAc, but rather they serve as examples where O‐GlcNAc can either increase or decrease associations in a protein‐selective manner.

Taken together, we believe that these results continue to build a strong foundation for understanding the role of O‐GlcNAc in cellular stress or human diseases that involve protein misfolding and/or aggregation. An increasing amount of evidence suggests that O‐GlcNAc directly prevents protein/protein interactions and misfolding. For example, O‐GlcNAc is found co‐translationally on certain proteins to prevent their premature degradation (Zhu et al., 2015). It also slows the kinetics of aggregation for the amyloid forming proteins tau and α‐synuclein (Levine et al., 2019; Yuzwa et al., ²⁰¹²), and antagonizes the phase‐separation of SynGAP/PSD‐95 (Lv et al., 2022). Our sHSP data indicate that O‐GlcNAc may also play a role in activation of chaperone proteins to block aggregation and now promote refolding. We believe that this may have important implications in human diseases, like cancer, that are associated with increased basal levels of cell stress. Alternatively, diseases caused by protein misfolding might be more generally targeted by an increasing range of OGA inhibitors that are entering the clinic (Bartolomé‐Nebreda et al., 2021).

4. MATERIALS AND METHODS

4.1. General

Solvents and reagents were purchased from commercial sources. Aqueous solutions were prepared using ultrapure water. Reversed‐phase high‐performance liquid chromatography was performed on an Agilent 1260 Series instrument with the following solvents: solvent A, 0.1% trifluoroacetic acid in H₂O and solvent B, 0.1% trifluoroacetic acid in 90% acetonitrile and 10% H₂O. Semi‐preparative and analytical columns (Luna Omega 5 μm Polar C18 100Å and Jupiter 5 μm C4 300Å) were purchased from Phenomenex.

4.2. Peptide synthesis

HSP27 173–205 gT184 peptide was synthesized using Fmoc solid‐phase chemistry on a lysine‐preloaded Wang resin. For standard residues, the couplings were 1 h long involving Fmoc‐protected amino acids (5 equiv.), HBTU (4.5 equiv.), and DIEA (10 equiv.) in DMF. For glycosylated threonine, 2 equiv. of Pfp‐activated residue was coupled overnight in DMF. Upon the final deprotection of peptide synthesis, the peptide was globally deprotected and cleaved from the resin for 4 h at RT in 95:2.5:2.5 TFA/H₂O/tri‐isopropylsilane. Cleaved peptide was precipitated out of cold diethyl ether. To deacetylate the glycosylated residue, the peptide was deprotected with 5% hydrazine monohydrate in MeOH. The peptide was then purified by reverse‐phase HPLC on a Phenomenex Polar C18 column and characterized by ESI‐MS. Purity was assessed by analytical HPLC.

4.3. Protein expression and purification

Lysis buffer—50 mM phosphate, 300 mM NaCl, 5 mM imidazole, 2 mM PMSF, 2 mM TCEP, and 6 M GnHCl, pH 7.4; wash buffer—50 mM phosphate, 300 mM NaCl, 20 mM imidazole, and 4 M urea, pH 7.4; elution buffer—50 mM phosphate, 300 mM NaCl, 250 mM imidazole, 2 mM TCEP, and 4 M urea, pH 7.

HSP27—BL21(DE3) E. coli cells were transformed with HSP27 pTXB1 (intein‐fusion) plasmid DNA and plated on LB agar plates containing 100 μg/mL ampicillin. Single colonies were picked and grown in LB overnight at 37°C at 225 rpm to generate a starter culture. The next day, the starter culture was used to inoculate expression flasks containing Terrific Broth, which were grown at 37°C with shaking at 225 rpm to an OD₆₀₀ of 0.6 prior to induction. Protein expression was induced by the addition of IPTG at a final concentration of 1 mM with shaking for 6 h at 37°C. After harvesting by centrifugation at 6000 × g, cells were resuspended in lysis buffer. To lyse, cell suspension was tip sonicated for 30 s on/30 s off at 70% Amp for 5 min prior to clearing the lysate by centrifugation. Lysate was loaded onto a Cobalt IMAC column pre‐equilibrated with binding buffer to rock at 4°C and then eluted with elution buffer. After washes, protein was eluted and concentrated using 10K MWCO Amicon‐Ultra 15 mL spin filters (Millipore Sigma) prior to buffer exchange into DPBS with 4 M urea.

BAG3—BL21(DE3) Rosetta E. coli cells (EMDMillipore) were transformed with BAG3 pMCSG7 plasmid DNA by heat shock and plated on LB agar plates containing 100 μg/mL ampicillin. Starter cultures and expression flasks were prepared in the same manner described above. Protein expression was induced by the addition of IPTG at a final concentration of 0.5 mM with shaking at 225 rpm for 16 h at 18°C. After harvesting by centrifugation at 6000 × g, cells were resuspended in BAG3 lysis buffer (50 mM Tris, 100 mM NaCl, 1 mM EDTA, and 15 mM BME, pH 8.0) with two protease inhibitor tablets (Roche cOmplete mini) per liter of culture added and lysed by tip sonication. Lysate was cleared by centrifugation at 20,000 × g and then proteins were precipitated with ammonium sulfate until a final concentration of 35%. The resulting protein pellet was obtained by centrifugation at 7000 × g and then resuspended in binding buffer (50 mM Tris, 1 mM imidazole, and 300 mM NaCl, pH 8).

4.4. Synthesis of HSP27(gT184)

HSP27(gT184) was synthesized and purified as previously described (Balana et al., 2021). Briefly, the C‐terminal, O‐GlcNAc modified peptide was prepared using standard Fmoc solid‐phase chemistry with pre‐loaded Wang resin as a support. Coupling reactions for standard amino acids utilized Fmoc‐protected amino acid (5 equiv.), HBTU (5 equiv.), and DIEA (10 equiv.) in DMF for 1 h. For the glycosylated threonine residue, 2 equiv. of the Pfp‐activated monomer (Leon et al., 2018) was coupled overnight in 3 mL of DMF. Upon completion of peptide syntheses, the acetyl groups on the O‐GlcNAc were removed on resin by treatment with hydrazine monohydrate (80% v/v in MeOH) twice for 45 min with mixing. The peptide was then fully deprotected and cleaved (95:2.5:2.5 TFA/H₂O/tri‐isopropylsilane) for 4 h at room temperature, followed by precipitation out of cold ether. The resulting pellet was resuspended, purified by reverse‐phase chromatography, and characterized by ESI‐MS. Purity was assessed by analytical HPLC and the peptide was lyophilized and stored at −20°C until use.

BL21(DE3) Escherichia coli (EMD Millipore) cells were transformed with and HSP27‐intein‐fusion plasmid DNA by heat shock and plated on selective LB agar plates containing 100 μg/mL ampicillin. Single colonies were selected and grown in LB to an OD₆₀₀ of 0.6–0.7 at 37°C while being shaken at 250 rpm. Expression was then induced with IPTG at a final concentration of 1 mM at 37°C with shaking at 250 rpm for 6 h. The bacterial pellet was harvested at 6000 × g, resuspended in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 1 mM TCEP, 5 mM imidazole, 6 M GuHCl, and 2 mM PMSF, pH 7.5), tip sonicated, and clarified by centrifugation (7000 × g for 30 min at 4°C). The resulting protein lysate was incubated with Co‐NTA agarose beads (Genessee Scientific) and then extensively washed (50 mM NaH₂PO₄, 300 mM NaCl, 2 mM TCEP, 20 mM imidazole, and 4 M urea, pH 7.5) to remove impurities. The HSP27–intein fusion was eluted (50 mM NaH₂PO₄, 300 mM NaCl, 1 mM TCEP, 250 mM imidazole, and 4 M urea, pH 7.5) and dialyzed into PBS. The corresponding HSP27‐thioester was generated by addition of sodium mercaptoethanesulfonate (MESNa) at final concentration of 200 mM and adjustment of the pH to 7 before overnight incubation at room temperature. The resulting HSP27 thioester was purified by reversed phase liquid chromatography and characterized by analytical RP‐HPLC and mass spectrometry. The thioester was then freeze‐dried to give a lyophilized powder.

The glycopeptide fragment (1.1 equiv.) and 1 equiv. of the HSP27 thioester 1 (3 mM final concentration) were dissolved in ligation buffer (300 mM NaH₂PO₄, 6 M guanidine HCl, 100 mM MESNa, and 1 mM TCEP, pH 7.4) allowed to react at 37°C until the reaction was complete as determined by HPLC. The reaction was then diluted 4‐fold into desulfurization buffer (200 mM NaH₂PO₄, 3 M guanidine HCl, and 300 mM TCEP, pH 7.0) containing 2% (v/v) ethanethiol, 10% (v/v) tert‐butyl‐thiol, and the radical initiator VA‐061 (as a 0.2 M stock in MeOH). The reaction mixture was stirred at 37°C for 15 h to remove the cysteine required for ligation. The resulting protein was purified by RP‐HPLC to yield HSP27(gT184). Proteins were analyzed by analytical HPLC for purity, and masses were confirmed by MALDI‐TOF MS. Purified HSP27 variants were lyophilized and stored at −20°C until use.

4.5. N‐terminal biotinylation of HSP27 and CRYAB

To an eppendorf tube, 500 μg (20 nmol) of lyophilized protein was dissolved in 360 μL 6 M GnHCl, 50 mM sodium phosphate, pH 8 buffer. 100 mM PCA–biotin stock was prepared in 1:1 acetonitrile/water and added to a final concentration of 10 mM. Biotinylation occurred over a period of 24 h at 37°C with agitation at 300 rpm. Biotinylated protein was purified by reverse‐phase HPLC and characterized by ESI‐MS. Purity was assessed by analytical HPLC. Biotinylated protein was lyophilized and stored at −20°C.

4.6. sHSP refolding

Lyophilized sHSPs were denatured in 6 M GnHCl, 50 mM sodium phosphate, pH 8 buffer at a concentration of 0.5 mg/mL and dialyzed against DPBS for slow refolding overnight at 4°C. Refolded proteins were concentrated using 10K MWCO Amicon‐Ultra 0.5 mL spin filters (Millipore Sigma) at 4°C.

4.7. Mass shift

Refolded protein amounts were quantified using Pierce BCA Protein Assay kit. In 500 μL of DPBS, 10 μg each of refolded of CRYAB‐biotin, HSP27(unmod) and HSP27(gT184) were incubated at 37°C for 1 h to form heterooligomers, and then pulled down with 30 μL of pre‐washed neutravidin beads for 1 h at room temperature. Beads were washed with DPBS prior to elution with 50 μL 2% SDS in 50 mM TEA buffer at 60°C for 10 min. Elutions were diluted with 50 μL of 50 mM TEA buffer prior to MeOH/CHCl₃ precipitation, chemoenzymatic transfer, and mass shift with DBCO‐PEG‐2 kDa (Darabedian et al., 2018). Gel samples were run on 12% Criterion™ XT Bis‐Tris Protein Gel for 75 min at 150 V for separation and proteins were visualized using Coomassie brilliant blue.

4.8. Enrichment

HeLa cells were cultured in DMEM supplemented with 10% fetal bovine serum. Cells were grown at 37°C in a humidified atmosphere with 5% CO₂. After harvesting, cell pellets were stored at −80°C. Pellets were resuspended in native lysis buffer (20 mM HEPES, 2.5 mM MgCl₂, 100 mM KCl, 0.5 NP‐40, pH 7.4 with 5 mg/mL Roche cOmplete mini protease tablets) and lysed using a Dounce homogenizer. After clearing the lysate, protein concentration was measured by BCA protein assay and adjusted to 1 mg/mL. High capacity neutravidin beads (50 μg/mg protein) were washed with 0.2% SDS buffer prior to incubation with HeLa lysate for 1 h at room temperature to remove any endogenously biotinylated proteins.

Native enrichment—After filtering out the beads, 10 μg (quantified by BCA assay) of refolded, biotinylated HSP27 and HSP27(gT184) were then spiked into separate aliquots of HeLa lysate prior to incubation with fresh beads at 37°C for 1 h in fritted columns. Non‐interactors were then removed and beads were subjected to a series of washes (3× with native lysis buffer and 6× with 25 mM ammonium bicarbonate) prior to elution with 4% SDS for 10 min. Biotinylated HSP27s and any interacting proteins were precipitated in cold acetone. SDS–PAGE and subsequent trypsinolysis enabled analysis and identification of protein interactors by TIMS‐TOF mass spectrometry.

Crosslinked enrichment—After filtering out the beads, 10 μg (quantified by BCA assay) of refolded, biotinylated HSP27 and HSP27(gT184) were then spiked into separate aliquots of HeLa lysate prior to incubation at 37°C for 1 h. To crosslink interacting proteins, each aliquot was slightly agitated at RT for 30 min with Lomant's reagent (1 mM final concentration), prior to quenching with Tris (20 mM final concentration) at RT for 15 min with gentle rocking. Proteins were then incubated with fresh neutravidin beads for 1 h at RT in a fritted column prior to washing with 1% SDS in DPBS and elution with 50 mM DTT, 1% SDS in DPBS (30 min at 37°C). Gel samples were prepared for separation on a Criterion 4–20% Tris‐Glycine Gel (Bio‐Rad).

4.9. Western blot

After running the gel, proteins were transferred onto PVDF membrane via semi‐dry transfer. Membrane was blocked for 1 h at RT with WesternCL blocking buffer (Bio‐Rad) and incubated with 1:5000 anti‐BAG3 (ThermoFisher MA5‐35563) ON at 4°C. After washing 3× with TBS‐Tween 20, the membrane was incubated with HRP‐conjugated goat anti‐rabbit for 1 h at RT prior washing and chemiluminescence imaging.

4.10. Proteomics

A nanoElute was attached in line to a timsTOF Pro equipped with a CaptiveSpray Source (Bruker, Hamburg, Germany). Chromatography was conducted at 40°C through a 25 cm reversed‐phase C18 column (PepSep) at a constant flow‐rate of 0.5 μL/min. Mobile phase A was 98/2/0.1% water/ACN/formic acid (v/v/v) and phase B was ACN with 0.1% formic acid (v/v). During a 108 min method, peptides were separated by a three‐step linear gradient (5%–30% B over 90 min, 30%–35% B over 10 min, and 35%–95% B over 4 min) followed by a 4 min isocratic flush at 95% for 4 min before washing and a return to low organic conditions. Experiments were run as data‐dependent acquisitions with ion mobility activated in PASEF mode. MS and MS/MS spectra were collected with m/z 100 to 1700 and ions with z = +1 were excluded.

Raw data files were searched using PEAKS Online Xpro 1.6 (Bioinformatics Solutions Inc.). The precursor mass error tolerance and fragment mass error tolerance were set to 20 ppm and 0.05, respectively. The trypsin digest mode was set to semi‐specific and missed cleavages were set to 3. The human Swiss‐Prot reviewed (canonical) database (downloaded from UniProt) and the common repository of adventitious proteins (cRAP, downloaded from The Global Proteome Machine Organization) totaling 20,487 entries were used. Carbamidomethylation was selected as a fixed modification. Acetylation (N‐term), deamidation (NQ), and oxidation (M) were selected as variable modifications. The maximum number of variable PTMs per peptide was to 3. All experiments were repeated in triplicate and combined datasets subjected to the filtration criteria explained in the manuscript text. Raw data files and searched datasets are available on the Mass Spectrometry Interactive Virtual Environment, a member of the Proteome Xchange consortium, under the identifier: MassIVE MSV000091378. The filtered dataset is also available in Table S1.

4.11. Luciferase refolding assays

Luciferase refolding assays were performed as described previously (Johnson et al., 2022; Rauch & Gestwicki, ²⁰¹⁴; Wisén & Gestwicki, ²⁰⁰⁸). Briefly, Renilla luciferase (Promega) was chemically denatured in 6 M GnHCl for 1 h at room temperature. Hsc70, DnaJB4, denatured luciferase, and HSP27 were then diluted in buffer containing an ATP regenerating system (23 mM HEPES, 120 mM K(C₂H₃O₂), 1.2 mM Mg(C₂H₃O₂)₂, 15 mM DTT, 61 mM creatine phosphate, 35 units/mL creatine kinase, and 5 ng/μL BSA, pH 7.4) such that the final reaction concentrations were 1 mM Hsc70, 300 nM DnaJB4, 100 nM Renilla luciferase, and the indicated concentration of each Hsp27 construct. All proteins were quantified by BCA assay. A two‐fold dilution series of Bag3 was added, followed by 2.5 mM ATP. The refolding reactions were incubated in a white, 96‐well plate for 1 h at 37°C, SteadyGlo luminescence reagent (Promega) was added to each well, and luminescence was measured on a SpectraMax M5 plate reader. Each condition was run in triplicate, and background (100 nM denatured Renilla luciferase without chaperones) was subtracted from each replicate.

4.12. Synthesis of PCA–biotin

All reagents used for chemical synthesis were purchased from Sigma‐Aldrich. All anhydrous reactions were performed under nitrogen atmosphere. Analytical thin‐layer chromatography (TLC) was performed on glass‐backed plates with detection by cerium ammonium molybdate (CAM), ninhydrin, or UV. PCA–biotin was synthesized according to published procedures (MacDonald et al., 2015). Characterization of the final product was consistent with the published data: ¹H NMR (400 MHz, CDCl3) 10.03 (s, 1H), 7.83 (d, J = 4.3 Hz, 2H), 7.65 (m, J = 4.4 Hz, 1H), 6.02 (s, 1H), 5.26 (s, 1H), 4.46 (dd, J = 8.0, 4.7 Hz, 1H), 4.27 (dd, J = 8.0, 5.0 Hz, 1H), 3.75 (s, 2H), 3.62 (t, J = 4.9 Hz, 2H), 3.47 (t, J = 4.9 Hz, 2H), 3.12 (q, J = 2.9 Hz, 1H), 2.86 (dd, J = 12.4, 4.9 Hz, 1H), 2.69 (d, J = 12.4 Hz, 1H), 2.49 (m, 4H), 2.32 (t, J = 7.5 Hz, 2H), 1.75–1.56 (m, 4H), 1.5–1.4 (m, 2H).

AUTHOR CONTRIBUTIONS

Afraah Javed: Conceptualization; investigation; writing – original draft; methodology; validation; visualization; writing – review and editing; formal analysis. Oleta T. Johnson: Investigation; methodology; validation; visualization; writing – review and editing; formal analysis. Aaron T. Balana: Conceptualization; investigation; methodology; validation; visualization; formal analysis. Regan F. Volk: Investigation; methodology; validation; visualization; formal analysis. Andreas Langen: Investigation; methodology; validation; visualization; formal analysis. Benjamin S. Ahn: Investigation; methodology; validation; visualization; formal analysis. Balyn W. Zaro: Investigation; funding acquisition; methodology; validation; visualization; writing – review and editing; formal analysis; supervision. Jason E. Gestwicki: Investigation; funding acquisition; writing – review and editing; formal analysis; supervision. Matthew R. Pratt: Conceptualization; investigation; funding acquisition; writing – original draft; writing – review and editing; project administration; supervision.

Supporting information

FIGURE S1: Schematic of HSP27 primary sequence. HSP27 is comprised of an N‐terminal unstructured domain that is important for oligomer formation, a central α‐crystallin domain (ACD) that contains a cleft important for substrate binding, and a C‐terminal unstructured domain with an IXI sequence and an adjacent O‐GlcNAc site.

FIGURE S2: Characterization of HSP27 proteins. RP‐HPLC traces and ESI‐MS characterization of the indicated HSP27 proteins. The smaller set of peaks represents non‐biotinylated protein.

FIGURE S3: Schematic of the HSP27/BAG3/HSP70 refolding assay.

FIGURE S4: Characterization of Biotin‐αBC. RP‐HPLC traces and ESI‐MS characterization of biotinylated αBC. The smaller set of peaks represents non‐biotinylated protein.

FIGURE S5: Characterization of the mass shifting experiment. The indicated proteins (10 μg) were mixed with 90 μg of BSA and then subjected to chemoenzymatic modification with GalNAz followed by the indicated strain‐promoted azide‐alkyne cycloaddition (SPAAC) conditions with DBCO‐PEG (2 kDa) before analysis by Coomassie blue staining.

TABLE S1:

Javed A, Johnson OT, Balana AT, Volk RF, Langen A, Ahn BS, et al. O‐GlcNAc modification of HSP27 alters its protein interactions and promotes refolding of proteins through the BAG3/HSP70 co‐chaperone. Protein Science. 2024;33(10):e5173. 10.1002/pro.5173