Cryptic cysteines in titin

The amino acid sequence of the Ig domains in the elastic I-band of titin is highly abundant in cysteine residues, all of which are potential targets for S-glutathionylation (Figure 1B) (Kellermayer and Grama, 2002). All I-band Ig domains of titin whose structure has been determined, I1 (PDB code 1g1c), I65–I70 fragment (PDB code 3b43), and I91 (also known as I27, PDB code 1tit, Figure 1A,C), contain at least two cysteine residues that appear to be buried. Some of these cysteines are clustered and have the potential to engage in disulfide bonds (Mayans et al., 2001), while others are far from each other, as in the case of Cys47 and Cys63 in I91 (Figure 1C). To obtain information about the location of cysteine residues in the remaining domains, we carried out homology modeling of all Ig domains in the I-band of the canonical human titin gene (Uniprot code: {"type":"entrez-protein","attrs":{"text":"Q8WZ42","term_id":"384872704","term_text":"Q8WZ42"}}Q8WZ42). Our analysis predicts that 89 out of 93 Ig domains in the I-band of titin contain cryptic cysteines. These cysteines would only become modified by redox-active molecules if the protein domains first unfold (Fig S1B).

We tested experimentally whether the cysteines in I91 are accessible to redox modification by incubation with oxidized glutathione (GSSG). A cysteine residue reacts with GSSG via simple thiol/disulfide exchange (Gilbert, 1990). We incubated I91 in 5 mM GSSG at different temperatures and then detected S-glutathionylation through western blotting and staining using anti-glutathione antibodies. We found that I91 does not react with GSSG at 37°C, implying that the cysteines in native I91 are inaccessible to GSSG under physiological conditions (Figure 1D). However, I91 became S-glutathionylated at temperatures higher than 55°C, which have been shown to unfold I91 (Somkuti et al., 2013) (Figure 1D). S-glutathionylation could be reversed by incubation with 50 mM dithiothreitol (DTT), a strong reducing agent that cleaves the mixed disulfide between cysteine and glutathione (Figure 1D). These results confirm that the cysteines in I91 are buried, and can only be S-glutathionylated when exposed to the solvent through protein unfolding. According to our homology models, similar results are expected for the majority of Ig domains in titin.

Cysteines found in buried positions are typically regarded as inert and not seen as potential targets of redox signaling. However, proteins that experience mechanical forces can be unfolded in cells (del Rio et al., 2009; Johnson et al., 2007; Li et al., 2002; Smith et al., 2007). Hence, the mechanical unfolding of a protein can expose cryptic cysteines that are normally inert, making them reactive (Johnson et al., 2007; Klotzsch et al., 2009). We hypothesized that mechanical unfolding can enable covalent modification of titin’s cryptic cysteines, which would disrupt the folding/unfolding dynamics and cause sustained changes in titin elasticity. We set out to test this hypothesis using force-clamp spectroscopy by atomic force microscopy (AFM), a single-molecule technique that mimics the mechanical perturbations experienced by titin domains in the heart (Fernandez and Li, 2004; Li et al., 2002).

S-glutathionylation of cryptic cysteines inhibits folding of I91

We first investigated the effects of S-glutathionylation on the refolding of I91. We mechanically unfolded and extended I91 polyproteins at a constant force of 175 pN (exposure pulse) in the presence of 100 mM oxidized glutathione (GSSG; Figure 2). Under force, 25-nm step increases in length mark the mechanical unfolding and extension of one I91 domain in the polyprotein (Figure 2A). Unfolding of multiple domains produces staircase-like changes in length, which we used to fingerprint single-molecule events (Figure 2B). Mechanical unfolding exposes both buried cysteines, which become reactive towards GSSG in the solution, resulting in S-glutathionylation (Figure 2A). We kept the cysteines exposed for variable amounts of time by changing the duration of this exposure pulse (Figure 2B,C). We then ended the exposure by quenching the force to 0 pN for 5 seconds, allowing for the collapse and folding of the I91 polyprotein (quench pulse; Figure 2B). We quantified folding with a probe pulse (175 pN; Figure 2B,C). The presence of a second staircase of 25 nm steps marks the successful refolding of the protein. When the exposure time was brief (2 s; Figure 2B, top trace), folding proceeded successfully. However, when the exposure time was increased (40 s; Figure 2B, bottom trace) allowing for extensive S-glutathionylation of the unfolded protein, unfolding steps were no longer observed during the probe pulse, indicating impaired folding. The near complete inhibition of refolding observed after 40 s exposures to GSSG was not seen in the presence of reduced glutathione or with buffer alone (Figure 2C). To verify that the refolding inhibition was caused by cysteine modification, we repeated the experiment with a cysteine-free I91 mutant. Refolding of this mutant was not affected by exposure to GSSG (ΔCys; Figure 2C). Taken together, these results show that S-glutathionylation of the cryptic cysteines in I91 inhibits protein folding.

Open in a separate window

Figure 2

S-glutathionylation of cryptic cysteines inhibits refolding of I91

(A) I91 contains two cryptic cysteines that become exposed by mechanical unfolding. At a force of 130–175 pN, the unfolding of an I91 domain can be identified as a 25 nm step increase in end-to-end length. Exposure of the unfolded protein to GSSG leads to S-glutathionylation. (B) A single polyprotein is unfolded at 175 pN in 100 mM GSSG and then held extended for a variable length of time (exposure time; 2 s top trace, 40 s bottom trace). The force is then quenched to zero for 5 s to allow the protein to refold. We measure the extent of refolding by applying a probe pulse. (C) Refolded fraction as a function of exposure time for I91₈ polyproteins exposed to buffer only (downward triangles, N>35), 100 mM GSSG (solid circles, N>80), 200 mM GSH (upward triangles, N>80); and refolded fraction in the presence of 100 mM GSSG for a mutant I91₈ polyprotein where all cysteine residues were replaced by alanines (ΔCys, squares, N>120). The refolded fraction was calculated as the ratio between the number of unfolding steps detected during the probe pulse and the number of unfolding steps detected during the exposure pulse. Open symbols represent total refolding of I91 (weak + strong domains) in the presence of 100 mM GSSG, as determined using the double probe pulse protocol in Fig 3. Error bars represent SEM.

S-glutathionylated I91 domains can fold into a weakened state

We noticed that in the presence of GSSG, unfolding events often appeared faster in the probe pulse than in the exposure pulse. This observation led us to believe that some of the glutathionylated domains had refolded into mechanically weakened structures. In order to detect discrete populations of folded domains with different mechanical stabilities, we modified the probe pulse to consist of a force ramp from 0 pN to 200 pN. In the absence of GSSG, most of the refolded I91 domains unfolded at 150–200 pN. Strikingly, when the buffer contained GSSG, we detected a second population of refolded I91 domains that unfolded at much lower forces of around 100 pN (Figure S2).

To quantify the incidence of weak and native I91 domains independently, we modified the probe pulse to include two regimes of constant force (110 pN and 175 pN; Figure 3A). We designed this protocol so that only weak domains would unfold during the initial low force pulse, while natively folded domains would unfold only after application of the high force. Indeed, for short exposure pulses (2 s; Figure 3) or in the absence of GSSG, the majority of the 25 nm steps are observed during the high force regime of the probe pulse, indicating mostly native folding (circles; Figure 3B). Meanwhile, for exposure pulses longer than 10 s, the majority of the 25 nm steps occur during the low force pulse (Figure 3A). Longer exposure pulses led to overall fewer 25 nm steps in the probe pulses (Figure 3B). The total refolding fractions detected using the single and double probe pulse protocols are coincident, indicating that we detected all refolded domains in both experiments (Figure 2C).

Open in a separate window

Figure 3

Graded mechanical modulation of I91 by S-glutathionylation

(A) Experimental recordings showing unfolding steps during the probe pulse (open arrowheads). Each recording was obtained in 100 mM GSSG after a different exposure time followed by 5 s refolding. Refolded domains with different mechanical stability were detected using a two-level probe pulse (110 pN in blue followed by 175 pN in red). (B) Refolded fraction (circles and squares, left axis) and failures (triangles, right axis) as a function of exposure time (N>40). Squares indicate weak domains detected during the low force regime; circles indicate native domains detected in the high force regime. Data points at zero exposure time were obtained in buffer without GSSG (open symbols, N=24). Solid lines were obtained from the kinetic model shown in Figure 5A. Error bars represent SEM. See also Figure S2

Given that 25 nm steps are fingerprints of folded domains, their absence from the probe pulse can be used to detect domains that failed to fold. We counted the failures (triangles; Figure 3B) as the number of unfolding steps in the exposure pulse minus the number of unfolding steps in the probe pulse. The failures became more frequent with increasing exposure time. Meanwhile, the population of weakened domains peaked at an exposure time of 10 s (Figure 3B). This transient increase in weakened domains followed by complete inhibition of folding indicates the presence of at least three states, which is consistent with the two potential S-glutathionylation sites in each I91 domain. The results can be explained by a scenario in which S-glutathionylation of one cysteine in I91 weakens the domain while S-glutathionylation of the second cysteine abolishes mechanical stability altogether.

The mechanical effects of S-glutathionylation are site specific

We set out to characterize the effects of S-glutathionylation at each of the two cysteines in I91. We conducted refolding experiments using single-cysteine mutants of I91 in which one of the cysteines had been mutated to alanine. We refer to these mutants as I91-Cys47 (C63A) and I91-Cys63 (C47A) respectively, highlighting the remaining cysteine in each case. The probe pulse consisted of a ramp from 0 to 240 pN in 6 seconds, which allowed us to assess the mechanical stability of the refolded domains (Figures 4, S3). Exposure of I91-Cys63 to GSSG for short periods of time led to refolded domains with native mechanical stability, whereas longer exposure times led to weak refolded domains that unfolded at a lower force (Figures 4A,B, S3B). Weak domains were not observed for I91-Cys47, even after exposure times in excess of 40 s (Figure S3A). We conclude that the weak domains seen upon S-glutathionylation of I91 WT are due to modification of Cys63 specifically. If we assume similar exponential force dependence, these weak domains would unfold at a rate two orders of magnitude faster compared to native domains at any given force (Figure S3C).

Open in a separate window

Figure 4

S-glutathionylation of different cysteines produces distinct mechanical phenotypes

(A) The single-cysteine mutant I91-Cys63 was mechanically unfolded and exposed to 100 mM GSSG for a variable amount of time. After a 5 s quench to allow for folding, refolding was probed using a linear force ramp up to 240 pN. (B) Mechanical stability of refolded I91-Cys63 after varying exposure times to 100 mM GSSG (N>25). Solid lines indicate fits of Gaussians with fixed widths and centroids. (C) Refolded fraction of I91-Cys47 (triangles, N>120) and I91-Cys63 (weak domains, squares; strong domains, circles; N>150) at different exposure times for a fixed quench time of 5 s in 100 mM GSSG. Data points at zero exposure time were obtained in the absence of GSSG (open symbols; N>150). Reaction rates of S-glutathionylation were measured from exponential fits (solid lines). Error bars represent SEM. See also Figure S3.

The abundance of weak domains in I91-Cys63 reached a constant value after ~10 s and did not disappear at longer exposure times (Figure 4C). The persistence of these weak domains contrasts with I91 WT, where the appearance of weak domains was transient (Figure 3B). We conclude that disappearance of weak I91 WT domains at longer exposure times was likely caused by a second S-glutathionylation, at Cys47.

S-glutathionylation at either Cys47 or Cys63 leads to a decrease in total refolding (Figure 4C). To investigate if this decrease could be explained by a decrease in folding kinetics, we measured the folding rate of S-glutathionylated I91-Cys63 (Figure S3D). First, we unfolded the protein for 20 s in the presence of GSSG to allow for ample S-glutathionylation. We then quenched the force for a variable amount of time, and probed refolding after a variety of quenching times. The folding rate of S-glutathionylated I91-Cys63 appears indistinguishable from the folding rate of the unmodified protein (Figure S3D). Hence, the decrease in refolding was not caused by altered folding kinetics. Instead, it appeared that a fraction of the glutathionylated domains simply fail to fold at all. One possibility is that during refolding, S-glutathionylated domains have a stronger tendency to become trapped in off-pathway misfolded states that do not mature into folded, mechanically stable structures.

Since folding rates appeared unchanged by S-glutathionylation, we took advantage of the decrease in overall folding to measure the site-specific rates of S-glutathionylation. We found that the first-order rate of glutathionylation at Cys63 was nearly 5 times higher than the glutathionylation at Cys47 (Figure 4C, ​,5A).5A). The lower reactivity of Cys47 is likely caused by the presence of negatively charged neighboring residues (Figure 5B) (Wu et al., 2011). Refolding measurements of I91-Cys63 in the presence of 10 mM GSSG confirmed that the glutathionylation rate changes linearly with the concentration of GSSG, as expected for a first-order chemical reaction (Figures 5C, S4).

Open in a separate window

Figure 5

Kinetic model

(A) Scheme that includes all possible reaction pathways for S-glutathionylation and mechanical outcomes. Numbers in parentheses are the asymptotic folding fractions. (B) Sequence flanking the two cysteines in I91. Negatively charged residues are labeled red, positively charged residues are labeled blue. (C) The rate of S-glutathionylation of Cys63 depends linearly on the concentration of GSSG. Error bars represent SEM. See also Figure S4.

Based on our results with the single-cysteine mutants, we constructed a kinetic model that predicts the mechanical outcomes of S-glutathionylation in I91 WT (Figure 5A). The model contains two classes of parameters: amplitude of folding of each of the four states of glutathionylation, and site-specific glutathionylation rates k₄₇ and k₆₃. The only free parameter in the model is the extent of refolding of the doubly glutathionylated I91 WT. The value of this parameter was determined through numerical optimization. We found that a value of 0.054 best fits the I91 WT data (Figure 3B), confirming that folding is severely compromised in the doubly glutathionylated I91.

The mechanical effects of S-glutathionylation are reversed by the enzyme glutaredoxin

Reversible S-glutathionylation is emerging as an important downstream regulatory posttranslational modification in various redox signaling pathways (Dalle-Donne et al., 2009). Although several chemical and enzymatic pathways can lead to S-glutathionylation (Gallogly and Mieyal, 2007), deglutathionylation is mainly the result of the activity of specific enzymes, primarily glutaredoxin (GRX) (Lillig et al., 2008). Hence, we reasoned that GRX could rescue refolding following S-glutathionylation of I91. To test this hypothesis, we used an alternative pathway of S-glutathionylation that is compatible with the activity of GRX. A protein-glutathione complex can be obtained through the attack of a free cysteine on GSSG, but also through the cleavage of a protein disulfide bond by a reduced glutathione molecule (GSH). We took advantage of this latter route because it allows obtaining S-glutathionylation in a reducing buffer, necessary to preserve GRX activity. Since the cysteines in I91 WT are far apart and do not establish a disulfide bond, we used the mutant I91_32–75. This I91 mutant readily forms a buried disulfide bond between engineered cysteines at positions 32 and 75. The disulfide in I91_32–75 becomes exposed and reactive only upon mechanical unfolding (Wiita et al., 2006) (Figure 6).

Open in a separate window

Figure 6

Glutaredoxin reverses the mechanical effects of S-glutathionylation

(A) I91_32–75 is S-glutathionylated through cleavage of its disulfide by GSH. Unfolding of I91_32–75 increases the protein length by 11 nm, while S-glutathionylation increases it further by 14 nm. Exposure-quench-probe pulse protocol probes the refolding of an I91_32–75 polyprotein in the presence of 100 mM GSH. During the exposure pulse, three unfolding events are followed by three S-glutathionylation events (solid arrowheads). The force is then quenched to zero for 5 s. No refolding events are observed during the probe pulse. (B) Recording showing the unfolding and S-glutathionylation of four I91_32–75 domains in the presence of 100 mM GSH and 45 µM GRX. After a 5 s quench, refolding is apparent during the probe pulse (25 nm steps, open arrowheads). (C) Diagram of unfolding and S-glutathionylation of I91_32–75. Numbers indicate the associated step sizes detected in force-clamp. (D) GRX-mediated deglutathionylation. (E) Refolded fraction after a 5 s quench in 100 mM GSH with or without 45 µM GRX. A baseline for refolding was measured using pre-reduced I91_32–75 polyprotein in the presence of 100 mM GSH (N>90). Error bars represent SEM. See also Figure S5.

When an I91_32–75 protein domain is placed under a constant stretching force of 150 pN, unfolding yields an 11 nm stepwise increase in length, resulting in the exposure of the buried disulfide (Wiita et al., 2006). Cleavage of the exposed disulfide by GSH further increases the length of the protein by 14 nm, providing an unambiguous fingerprint for S-glutathionylation (Figure 6A,C). Figure 6A shows a recording where an I91_32–75 polyprotein is first unfolded and S-glutathionylated, and then allowed to collapse and refold. 25 nm steps in the probe pulse indicate a refolded protein without disulfide. S-glutathionylated domains did not refold well, as 25 nm steps were only rarely seen in the probe pulse (Figure 6 A,E). Those few domains that refolded showed native mechanical stability (Figure S5A–C). We repeated the glutathionylation-refolding experiment, but now in the presence of GRX in conditions where the enzyme does not effectively cleave disulfide 32–75 (Figure 6 B,D, S5D). S-glutathionylated proteins exposed to 45 µM GRX showed robust refolding similar to controls using reduced I91_32–75 in the presence of GSH, a condition where S-glutathionylation is not possible (Figure 6E). The rescue of protein folding by GRX shows that the mechanical effects of I91 S-glutathionylation are reversible.

Measurement of I91 cysteine accessibility to GSSG

A fusion protein SUMO-I91 (0.6 mg/mL) was incubated with 5 mM GSSG at different temperatures for 1 hour. Then, Laemli sample buffer including 2 mM maleimide was added. Samples were incubated at 90°C for 15 minutes before SDS-PAGE. The inclusion of 2 mM maleimide assures that all cysteines that are exposed during denaturation of the sample for SDS-PAGE are blocked and do not react with GSSG. S- alkylation by maleimide is much faster than S-glutathionylation. Control experiments showed that no S-glutathionylation is detected when SUMO-I91 is boiled in the presence of 5 mM GSSG and 2 mM maleimide. To show reversibility by DTT, we incubated the samples with 50 mM DTT at 90°C for 15 minutes before loading in the SDS-gel. After electrophoresis and western blotting, S-glutathionylation was detected using anti-GSH monoclonal antibodies (Virogen).

Single-molecule Atomic Force Microscopy

We used custom-built force-clamp AFM setups that have been described in previous publications (Fernandez and Li, 2004) or a commercial force-clamp AFM (Luigs and Neumann, Germany) (Popa et al., 2013). All AFM experiments were conducted in degassed solution consisting of 10 mM HEPES pH 7.2, 150 mM NaCl, 1 mM EDTA. We coated freshly evaporated gold coverslips with the polyprotein of interest by adding a small aliquot (less than 10 µL) of polyprotein solution (0.2 – 1 mg/mL). Polyproteins were adsorbed for several minutes on the gold surface. Then buffer was added to the AFM fluid cell and the experiment was started after in situ calibration of the cantilever using the equipartition theorem (Popa et al., 2013). Single molecules were picked up by pushing the cantilever against the gold surface at a contact force of 0.5–2 nN for 1 second. Then, the piezoelectric actuator was retracted by applying a pulling force. If a tether had been formed, the force rapidly stabilized at the set-point. The tether was maintained until the force protocol ended or spontaneous detachment occurred. Refolded fractions were calculated as the ratio between the number of unfolding events in the probe pulse and in the exposure pulse. Once set up, AFM experiments run automatically in the same buffer for hours or even days. We did not observe any changes in the results with time since the start of the experiment, in agreement with I91 cysteines being buried in the fold of the protein (see also Figure 1D).

Data analysis

Data were analyzed using custom-written software in IGOR Pro (Wavemetrics). We included in the analysis traces that showed a fingerprint of 2 or more unfolding events in the exposure pulse. Only traces that reached the same length at the end of exposure and probe pulses were selected. This criterion ensured that the same molecule was attached in both pulses. Traces showing steps that could not be identified neither as unfolding nor reduction events were excluded from the analysis. For the I91-Cys63 refolding data, weak and strong refolded domains were quantified from double Gaussian fits to the histograms shown in Figure 4B. In these fits, the widths and centroids were held constant across the datasets, leaving only two free parameters (i.e. the amplitudes of the Gaussians). The numbers of weak and strong domains in each condition were then determined given the total number of folded domains and the relative amplitude of the two Gaussians. In refolding experiments, the number of independent observations N for a particular condition was counted as the total number of protein domains (as observed in the exposure pulse). SEM was estimated using the bootstrap method (Efron, 1982). In the force histograms, N is the number of observations. S-glutathionylation rates were determined from refolding experiments in 100 mM GSSG using the single-cysteine mutants (Figure 4). k₄₇ = 0.33 M⁻¹ s⁻¹ was measured from an exponential fit to the time course of folding inhibition upon exposure to GSSG (grey line in Figure 4C). k₆₃ = 1.6 M⁻¹ s⁻¹ was measured as the average of the rate of disappearance of the strong domains (red line in Figure 4C) and the rate of appearance of the weak domains (blue line in Figure 4C). To examine the GSSG concentration dependency of the reaction, k₆₃ was also measured in 10 mM GSSG (Figure 5C). To obtain the refolding fraction of the doubly glutathionylated I91, we fit the data in Figure 3B to the kinetic model in Figure 5A (see Extended Experimental Procedures).

We thank Sergi García-Manyes for his original observation that mixed disulfides prevent refolding of Ig domains. We thank Thomas Kahn for the SUMO-191 fusion protein. GRX from E. coli was a kind gift from Arne Holmgren. Monoclonal antibodies MF20 (Fischman, D. A) and 9D10 (Greaser M. L) were obtained from the Developmental Studies Hybridoma Bank. This work was supported by the NIH grants HL66030, HL61228 (J.M.F.), P01 HL062426 (R.J.S., C.M.W.), and grants FP7-MEDIA from the European Union and SFB 1002/TP-B03 from the German Research Foundation (to W.A.L.). J.A.-C. was the recipient of a fellowship from Fundación Ibercaja (Zaragoza, Spain) and a K99 career development award ({"type":"entrez-nucleotide","attrs":{"text":"AI106072","term_id":"3461175","term_text":"AI106072"}}AI106072) from NIH. J.A.-C., P.K, D.G., and J.M.F. conceived the project. J.A.-C., P.K., D.G., E.E., and J.A.R.-P. did the single-molecule experiments. D.G. developed the homology models. J.A.-C and J.A.R.-P examined the solvent accessibility of the cysteines of I91. P.K. implemented the Monte Carlo simulations. N.H. and W.A.L. did the cardiomyocyte stretching experiments. C.M.W. and R.J.S. examined S-glutathionylation of murine titin. J.A.-C., P.K., and J.M.F. wrote the paper.