Human myocytes are protected from titin aggregation-induced stiffening by small heat shock proteins

HSP27 (HSPB1) and αB-crystallin (HSPB5) are members of the family of small heat shock proteins (sHSPs) ubiquitously expressed in mammalian tissues (Klemenz et al., 1993). sHSPs are important components of the cellular protein quality control machinery as they bind partially unfolded client proteins, hold them in a folding-prone state, and protect them from aggregation (Mymrikov et al., 2011). They are also directly linked to the autophagy and proteasomal degradation pathways (Willis and Patterson, 2010). Typically, sHSPs are up-regulated under diverse stress situations, and their overexpression protects cells from oxidative stress, energy depletion, and other unfavorable conditions (Mymrikov et al., 2011). In the heart, αB-crystallin and HSP27 are induced during ischemic injury, heat stress, or end-stage failure (Martin et al., 1997; Benjamin and McMillan, 1998; Knowlton et al., 1998; Yoshida et al., 1999; Dohke et al., 2006; Li et al., 2012). Induction of sHSPs also occurs in both myopathic skeletal (Kley et al., 2012) and normal muscles after intense exercise (Paulsen et al., 2009) and during aging (Doran et al., 2007).

In response to potentially harmful insults, the myocyte sHSPs, including αB-crystallin and HSP27, preferentially translocate from the cytosol to the myofibrils, where they bind to the sarcomeric Z-disk and/or I-band (Barbato et al., 1996; Lutsch et al., 1997; van de Klundert et al., 1998; Golenhofen et al., 1999; Fischer et al., 2002; Paulsen et al., 2009). Whether or not this translocation depends on the phosphorylation state of sHSPs is controversial (Mymrikov et al., 2011). Sarcomere proteins suggested to be protected by sHSP binding include desmin, α-actinin, actin, troponin-I/-T, titin, and myosin (Bennardini et al., 1992; Liu and Steinacker, 2001; Golenhofen et al., 2002; Wang et al., 2003; Bullard et al., 2004; Golenhofen et al., 2004; Melkani et al., 2006; Singh et al., 2007; Lu et al., 2008; Morton et al., 2009; Paulsen et al., 2009). The protective effect of αB-crystallin on desmin is well supported by genetic evidence (Vicart et al., 1998; Wang et al., 2003). However, in stressed myocytes, HSP27 and αB-crystallin can localize to the I-band region outside the Z-disk (Lutsch et al., 1997; Golenhofen et al., 1999, 2002, 2004; Bullard et al., 2004) where desmin and α-actinin are not present. Even though the two sHSPs interact with actin stress fibers, including cytoskeletal actin fibers in premature cardiomyocytes (Verschuure et al., 2002; Singh et al., 2007), it is not clear whether they associate with sarcomeric actin. Furthermore, the protective effect of sHSPs on isolated muscle myosin (Melkani et al., 2006; Markov et al., 2008) is difficult to reconcile with Z-disk–I-band (but not A-band) localization. The I-band association may be best explained by sHSP binding to titin spring elements (Golenhofen et al., 2002; Bullard et al., 2004), which has been confirmed for mammalian αB-crystallin (Bullard et al., 2004) and tentatively proposed for zebrafish HSP27 (Tucker and Shelden, 2009).

The observation that sHSPs are induced during ischemia suggests that acidic stress is a potential trigger. Acidosis resulting from ischemia-reperfusion injury (along with oxidative stress) can be substantial in the heart, with reductions in intracellular pH by 0.5–1 units (Vaughan-Jones et al., 2009). In skeletal muscle, metabolic acidosis during physical exercise was correlated with a reduction in the intracellular pH by 0.3–0.4 units (Khong et al., 2001). Acidic conditions directly affect sHSPs by promoting the formation and accumulation of large oligomers, thereby increasing chaperone activity (Ehrnsperger et al., 1999; Chernik et al., 2004). Acidosis also promotes the aggregation of many sHSP client proteins, such as actin and desmin, whereas the sHSPs themselves provide increased protection from aggregation at reduced pH (Bennardini et al., 1992; Barbato et al., 1996). Contractility of myocytes is strongly affected by intracellular acidosis (Orchard and Kentish, 1990). Studies also suggested that acidosis increases the passive stiffness of the heart (Berger et al., 1999) and of skeletal muscles (Miyake et al., 2003). Intracellular acidosis induced by ischemic stress caused severe diastolic dysfunction in a double-knockout mouse model deficient for αB-crystallin and HSPB2 (Golenhofen et al., 2006; Pinz et al., 2008). These findings suggest a relationship among acidosis, sHSP induction, and muscle elasticity.

The elasticity and passive stiffness of striated muscle cells are influenced by titin (Linke and Krüger, 2010), an essential component of the sarcomere (da Silva Lopes et al., 2011) and the largest known protein with a molecular mass of 3,000–3,800 kD, depending on the isoform (Bang et al., 2001). The titin isoforms differ primarily in their elastic I-band region (Fig. 1 A), which can be short and relatively stiff (cardiac N2B) or longer and more compliant (cardiac N2BA and skeletal muscle N2A). These titin springs are composed of tandem repeats of folded immunoglobulin-like (Ig) domains grouped into proximal, middle, and distal (to the Z-disk) segments, an N2-A region containing four Ig domains, and two large intrinsically disordered regions: the PEVK domain (titin region rich in proline, glutamate, valine, and lysine) and the N2-B-unique sequence (N2-Bus), which is found exclusively in cardiac titin. The N2-Bus and certain Ig domains were shown previously to interact with αB-crystallin (Bullard et al., 2004), which exerted a stabilizing effect on these domains in vitro (Bullard et al., 2004; Zhu et al., 2009). However, sarcomere stretch unfolds titin domains (Rief et al., 1997; Minajeva et al., 2001), exposing concealed hydrophobic sites, which can lead to aggregation and loss of function. We hypothesized that binding of HSP27/αB-crystallin to unfolded titin domains prevents titin aggregation under stress, thereby maintaining normal myocyte stiffness. We showed that the two sHSPs specifically protect the unfolded Ig regions of titin, but not the intrinsically disordered segments (N2-Bus and PEVK), from aggregating under acidic stress. HSP27 and αB-crystallin were found to translocate preferentially to the titin springs in stretched sarcomeres as well as in diseased human skeletal and cardiac myocytes. This suggests that sHSPs indeed act to prevent abnormally high, titin aggregation-induced, passive tension. This protective mechanism for titin may apply to other “mechanical” proteins in a variety of cell types.

To obtain information about the molecular organization of titin spring elements, we performed sedimentation-velocity experiments for recombinant human N2-Bus and PEVK (both intrinsically disordered structures), the N2-B region (N2-Bus plus three Ig domains), and the N2-A segment (four Ig domains). For all constructs, the tests showed a dominant single peak in the continuous sedimentation distribution c(s) with s_20,w values of 2.74 S (N2-Bus), 3.19 S (N2-B), 3.27 S (N2-A), and 1.06 S (PEVK; Fig. 1, B–E). Sedimentation depends, in part, on the size and shape of the molecule and is inversely proportional to the frictional ratio (f/f₀), the latter of which was 2.2 (N2-Bus and N2-B), 1.4 (N2-A), and 2.3 (PEVK). N2-Bus, N2-B, and PEVK were predicted to exist in a slightly extended conformation and N2-A in a more globular-like structure. For each titin construct, the s_20,w value of the dominant peak corresponded to the monomeric form of the respective fragment. A minor s_20,w value peak, suggestive of a small subpopulation of dimers (∼5–15%), was apparent for N2-Bus (4.18 S), N2-B (4.77 S), and N2-A (5.13 S; Fig. 1, B–E). Higher oligomeric states were not detected. The additional, relatively low s value peaks seen for N2-B and PEVK (Fig. 1, D and E) likely resulted from degradation products or contaminants. Thus, the titin spring elements behaved in vitro primarily as monomers.

In addition, size-exclusion chromatography was performed to independently test the molecular organization of two representative human titin spring regions: the four-Ig construct I9-12 (proximal Ig region) and PEVK. I9-12 was predicted to be monomeric and PEVK to be monomeric or dimeric, with no distinction possible because of the low mass of PEVK relative to the peak width (Fig. 1, F and G). These findings are consistent with the results of the sedimentation-velocity tests, suggesting that the elastic titin segment between the Z-disk–I-band junction and the distal Ig region exists primarily in a monomeric state.

To determine which titin spring elements interact with HSP27 and αB-crystallin, we used recombinant human sHSPs and titin constructs in GST pulldown assays (Fig. 2). The two sHSPs associated with the Ig domains I24/25 (part of N2-B), the intrinsically disordered N2-Bus, and the Ig-rich N2-A region, but not with the Z9-I1 construct from the Z-disk–I-band junction (Fig. 2, A and B). Independent binding assays using the yeast two-hybrid system confirmed the interaction between N2-Bus and HSP27 or αB-crystallin (Fig. S1 A). Additional GST pulldown assays demonstrated HSP27 binding to Ig domains I9-12 and I26 (part of N2-B), but not to PEVK (Fig. 2 A). We speculated that the phosphorylation state of HSP27 may be important for the interaction with titin domains, and thus we generated phosphomimicking mutants of human HSP27 to test their binding to N2-Bus in GST pulldown assays. However, relative to the affinity of N2-Bus for wild-type HSP27, none of the mutant HSP27 constructs (single point mutants: S15D and S82D; double mutant: S15/82D; and triple mutant: S15/78/82D) showed significantly altered binding affinity (Fig. S1 B). Therefore, the binding sites for HSP27 on the elastic titin segment were similar to those for αB-crystallin and included Ig-rich segments and the N2-B region, but not PEVK (Fig. 2 C). Interaction between HSP27 and N2-Bus is unlikely to be regulated by HSP27 phosphorylation.

To determine the region of the sarcomere that sHSPs bind, we attached the ends of human cardiomyofibrils coexpressing the N2B (∼65%) and N2BA (∼35%) titin isoforms or the ends of rabbit psoas myofibrils expressing the N2A isoform to glass needle tips to stretch the preparations (Fig. 3 A). Endogenous HSP27 and αB-crystallin were nearly absent from these isolated myofibrils. The myofibrils were incubated with excess amounts of recombinant HSP27 (Fig. 3, B and E) or αB-crystallin (Fig. 3, C and F), the unbound sHSPs were washed away, and myofibril-bound sHSPs were visualized by indirect immunofluorescence. In the control experiments, incubation with GFP or with secondary antibodies alone did not produce any specific immunofluorescence signal (Fig. 3 A). In human cardiomyofibrils, exogenous HSP27 and αB-crystallin both marked the Z-disk–I-band region when the sarcomere length (SL) was 2.4 µm but were clearly observed in the region of the I-band when the SL was 3.0 µm, as indicated by the appearance of double bands (Fig. 3, B and C). Importantly, even after gelsolin-mediated extraction of actin filaments and associated regulatory proteins from the cardiomyofibrils, exogenous HSP27 still localized to the I-bands; the Z-disk (which still contained some nonextractable actin) again did not appear to be the predominant binding site (Fig. 3 D). In stretched rabbit psoas myofibrils (3.4 µm SL), I-band localization of the sHSPs was apparent, although the doublets appeared fuzzier than in the cardiomyofibrils (Fig. 3, E and F). Importantly, when the SL was 2.2 µm, HSP27 did not bind to skeletal myofibrils (Fig. 3 E) and αB-crystallin showed decreased binding to the Z-disk–I-band (Fig. 3 F). Faint staining of the sarcomeric M-band was sometimes observed (Fig. 3 E), but the sHSPs were absent from the remainder of the A-band. These results show that HSP27 and αB-crystallin bound to the extensible segment outside the Z-disk where the titin springs are located, independently of the presence of actin filaments. sHSP binding to the I-band was greatly promoted by sarcomere stretch especially in skeletal myofibrils expressing the N2A isoform lacking the N2-B domain.

We first used the TANGO algorithm (Linding et al., 2004) to predict the aggregation propensity of unfolded Ig-rich and disordered I-band titin segments. The tendency for β aggregation was highest for Ig domains I9-12 (up to nearly 100%), followed by the N2-A segment (>60%) and the N2-Bus (<50%), whereas PEVK was predicted to have no aggregation propensity at all (Fig. S2). Z9-I1 from the Z-disk–I-band junction (which like PEVK did not bind sHSPs; Fig. 2) also appeared unlikely to aggregate. The α-aggregation propensity was small for all segments.

To experimentally test these predictions, recombinant titin constructs were exposed to urea, followed by rapid dilution in pH 7.2 Tris/HCl buffer, to promote their partial denaturation and potential aggregation. These conditions failed to induce aggregation of the mostly disordered titin segments N2-Bus, N2-B, and PEVK (Fig. 4 A). In contrast, I9-12 did aggregate, as indicated by increased light scattering (the absorbance at 320 nm), which plateaued at ∼20 min, whereas N2-A did not aggregate at normal pH (Fig. 4 B). However, when the pH was lowered to 6.7, N2-A aggregated more than did I9-12 (Fig. 4 B). The disordered titin segments still showed no significant aggregation at pH 6.7 (Fig. 4 A). These results were largely consistent with the TANGO predictions.

We then used this assay to test whether aggregation of denatured N2-A can be prevented by sHSPs. The presence of αB-crystallin reduced the aggregation of N2-A partially at an N2A/αB-crystallin molar ratio of 1:5 (∼75% reduction relative to the control at 30 min) and almost completely at a molar ratio of 1:10 (96% reduction; Fig. 4 C). HSP27 did not prevent aggregation of N2-A under these conditions (Fig. 4 D).

Aggregation of unfolded Ig domains of the elastic titin region may affect the passive stiffness of myocytes. We measured the passive force (F_passive) of isolated skinned human cardiomyocytes in response to stepwise stretching, first in relaxing buffer at a normal pH of 7.4 and then after a 10-min prestretch to a SL of 2.6 µm (to increase the probability of titin Ig domain unfolding) in acidic pH 6.6 relaxing buffer, in the absence or presence of recombinant human sHSPs (Fig. 5 A).

The F_passive-SL curve was steeper under the acidic, prestretch conditions than under control conditions (Fig. 5, C and E). However, the increase in F_passive failed to occur in the presence of HSP27 (Fig. 5 C) or αB-crystallin (Fig. 5 E). The protective effect on the F_passive (at 2.2 µm SL) was similar among HSP27 concentrations of 0.01, 0.1, and 1 mg/ml (Fig. 5 C, inset), whereas greater effects were observed with higher compared with lower αB-crystallin concentrations (Fig. 5 E, inset). If the prestretch step was omitted, the F_passive-SL curve was not significantly altered by low pH alone (Fig. 5 B). Likewise, in single isolated cardiac or skeletal myofibrils, the dynamic passive stiffness, measured as the force response to 20-Hz small oscillatory length changes, was independent of pH in the 5.9–7.4 range (Fig. S3). Prestretching of human cardiomyocytes in normal pH buffer again did not alter the F_passive-SL curve (Fig. 5, D and F), and under these conditions, the presence of HSP27 (Fig. 5 D) or αB-crystallin (Fig. 5 F) had no significant effect on the F_passive. Notably, the prevention of an increase in the F_passive by sHSPs under acidic conditions is contrary to what we initially expected according to single-molecule mechanical measurements of recombinant titin constructs (Bullard et al., 2004; Zhu et al., 2009), the results of which suggested that the titin-based force increases in the presence of sHSPs (Fig. S4 A). Instead, the sHSPs protected myocytes from becoming stiffer under acidic stress when titin Ig domains increasingly unfolded, i.e., under conditions that promote Ig aggregation.

If unfolding of Ig domains is critical for titin aggregation and protection by sHSP, then sHSP binding to titin would be unlikely to occur in myocytes expressing extremely compliant titin because the Ig domains would have a very low probability of unfolding, even in stretched sarcomeres. Fetal (day 18) rat cardiomyocytes express a highly compliant (∼3.7 MDa) N2BA titin isoform (Opitz et al., 2004), which the entropic elasticity theory predicts to lack Ig unfolding (Li et al., 2002) within and even beyond the physiological SL range (Fig. 6 A and Fig. S4 B) In contrast, adult rat cardiomyocytes almost exclusively express the stiff 3.0-MD N2B isoform, which is predicted to have already unfolded Ig domains at physiological SLs (Fig. 6 A and Fig. S4 B).

Confirming our expectations, fetal rat cardiomyocytes cultured in standard medium never revealed a sarcomeric (regular spacing) association with HSP27 and αB-crystallin, even though these sHSPs were highly abundant in the cytoplasm (Fig. 6 B). Various stressors applied to these cells, such as acidosis, contractile arrest, hypoxia, and cyclic/phasic stretch (10%), did not affect the distribution of sHSPs, which remained in the cytoplasm. Only upon treatment with the proteasome inhibitor MG132, which is known to boost sHSP levels, was sHSP binding to the Z-disk–I-band observed in some of the sarcomeres (Fig. S5). In contrast, in the adult rat cardiomyocytes, the two sHSPs always localized in the cytosol and at the Z-disk–I-band region (Fig. 6 C). Doublets, which indicate sHSP binding to the I-band, were clearly detectable at higher magnification (Fig. 6 C, insets). Some αB-crystallin was also observed at the M-band. These cells likely experienced mechanical strain and intracellular acidosis during the digestion procedure.

We investigated whether stretch in the isolated adult rat heart under adverse intracellular conditions, as induced by ex vivo perfusion with PBS, could trigger sHSP binding to the elastic I-bands. Hearts perfused for 30 min in the absence of stretch revealed cardiomyocytes with predominantly short sarcomeres (SL of ∼1.9 µm), which showed αB-crystallin exclusively at the Z-disk region (Fig. 7, left). However, a proportion of the myocytes in hearts perfused for 30 min in a high-stretch state showed greatly extended sarcomeres (up to SL of ∼2.6 µm). In those sarcomeres, αB-crystallin usually bound to both the Z-disk and the elastic I-band (Fig. 7, right). As the only difference, theoretically, between the two experimental conditions is the stretch state, the I-band binding of the sHSPs is presumed to directly follow the stretch-induced unfolding of titin Ig domains; the now-exposed hydrophobic sites on the titin springs facilitate protection by αB-crystallin, which is abundant in the cytosol of cardiomyocytes.

To address the relevance of the sHSP–titin interaction in disease, we investigated the localization of HSP27 and αB-crystallin in heart samples obtained from healthy donors and patients with dilated cardiomyopathy (DCM) as well as in skeletal muscle biopsies (Vastus lateralis) from healthy subjects and patients with limb girdle muscular dystrophy type 2A (LGMD2A). In LGMD2A myocytes, endogenous HSP27 and αB-crystallin associated mainly with the elastic I-band region and rarely with the Z-disk (Fig. 8, A and B). Anti-HSP27 (Fig. 8 A) and anti–αB-crystallin antibodies (Fig. 8 B) showed marked staining on either side of the Z-disk; this staining pattern was always closer in distance to the Z-disk relative to the anti-PEVK staining representing titin sites (Fig. 8 C). The greater the degree of double band separation the greater that of sarcomere stretching, confirming the elastic nature of the sHSP binding region (Fig. 8 C). In contrast, healthy skeletal myocytes showed binding of HSP27 to the Z-disk (Fig. 8 A) and fuzzy cytosolic localization of αB-crystallin with faint Z-disk and M-band staining (Fig. 8 B). I-band staining was almost absent. In DCM hearts, many cardiomyocytes showed a double-banded pattern for sHSPs, indicative of I-band localization (Fig. 9, A and B). In these cases, HSP27 and αB-crystallin again localized between the Z-disk and the PEVK epitope, migrating away from the Z-disk during sarcomere stretch (Fig. 9 C). In control donor cardiomyocytes, HSP27 and αB-crystallin were abundant in the cytosol, especially at the Z-disk but not the I-band; anti-sHSP and anti-PEVK antibodies showed hardly any colocalization (Fig. 9, A and B).

HSP27 and αB-crystallin have been investigated extensively in striated muscle, especially in regard to their induction during cardiac ischemia-reperfusion injury and their potential cardioprotective role. Here we provide new insight into the functional relevance of the translocation of these sHSPs to the sarcomeres, their primary binding sites in stressed myocytes. We demonstrated that HSP27 and αB-crystallin bound to titin spring elements and colocalized on I-band titin. The I-band association in sarcomeres was independent of actin filaments and enhanced by stretch. With increasing stretch, more titin Ig domains unfold, expose previously buried hydrophobic sites, potentially aggregate, and lose function. Our findings suggest that the unfolded Ig regions of the I-band titin, but not the intrinsically disordered segments, can aggregate to cause abnormally high titin-dependent myocyte stiffness. However, HSP27 and αB-crystallin translocated—likely promoted by acidic conditions—from the abundant pool of cytoplasmic sHSPs to the unfolded Ig regions, preventing their aggregation and maintaining functional (low-stiffness) titin. In line with this view, the sHSPs did not bind to cardiac sarcomeres that express the very compliant titin isoforms because the probability of Ig unfolding was very low even with stretching. In contrast, both sHSPs readily translocated to the I-bands of stretched sarcomeres that express relatively stiff titin. The two sHSPs were enriched at the elastic I-bands of human dystrophic muscles and failing hearts, suggesting a protective effect of sHSPs on titin springs in skeletal and cardiac muscle disease.

Previous studies of titin elasticity in situ and at the level of the single molecule demonstrated that the combined mechanical properties of individual titin spring elements can explain the elasticity of titin in myocytes (Li et al., 2002). We supported this finding with evidence that the elastic elements traverse the I-band in a primarily monomeric state, whereas the dimeric state seldom existed. The monomeric behavior observed for the proximal Ig, N2-B, N2-A, and PEVK regions contrasted with the established hexameric state of the distal Ig region of titin (Fig. 10; Houmeida et al., 2008). Although distal Ig domains can unfold and aggregate in vitro (Wright et al., 2005; Marchetti et al., 2008; Somkuti et al., 2013), these domains are stabilized through their self-association in vivo (Houmeida et al., 2008). In contrast, the monomeric spring elements are more vulnerable to unfolding/aggregation and may bind preferentially to chaperones. Indeed, HSP27 and αB-crystallin did not interact with the self-associating distal Ig region near the A-band, as shown here and in Bullard et al. (2004), but rather with the elastic titin elements in between the Z-disk and the PEVK domain.

Only those titin spring segments predominantly or exclusively composed of Ig domains (proximal Ig/N2-A), but not the intrinsically disordered PEVK and N2-Bus (including the N2-Bus flanked by Ig domains, as in the N2-B construct), aggregated under conditions of partial denaturation. The primary strategy by which proteins escape aggregation is through folding, and Ig domains are protected from aggregation by their β-barrel folds. However, titin Ig domains mechanically unfold in stretched myofibrils (Minajeva et al., 2001). Overstretch is common under both physiological and pathological conditions, such as in skeletal muscle during exercise and in ischemic heart failure when myocyte death can lead to overstretch of adjacent myocytes. With overstretch, the probability of titin Ig unfolding increases greatly (Li et al., 2002). Although the unfolded Ig domains readily refold in vitro in the absence of chaperones (Rief et al., 1997; Linke and Grützner, 2008), they are at risk of aggregation under conditions of macromolecular crowding in the sarcomeres, requiring protection by sHSPs.

Intrinsically disordered proteins lack a well-defined secondary structure. They are resistant to aggregation through a variety of mechanisms, including a relatively high net charge and low hydrophobicity, low number of aggregating sequences, and evolutionary conservation of proline residues, which introduce constraints because of their distinctive cyclic structure (Wright et al., 2005; Monsellier and Chiti, 2007). The titin-PEVK domain is proline-rich with a higher net charge than other titin regions (Tskhovrebova and Trinick, 1997). The N2-Bus also contains more proline residues and an above average net charge than a typical titin Ig domain. Supporting our experimental results, the TANGO algorithm (Linding et al., 2004) predicted lack of aggregation propensity for the PEVK domain, and the aggregation tendency of N2-Bus, albeit much higher compared with PEVK, was predicted to be lower than that of unfolded Ig segments. Consistent with our findings, aggregation of thermally unfolded (distal) titin Ig domains, excluding PEVK, was recently demonstrated in vitro (Somkuti et al., 2013). In summary, the disordered N2-Bus and PEVK segments have intrinsic properties that help prevent aggregation, potentially precluding the protection provided by sHSPs.

Intrinsically disordered proteins generally exhibit no preference for chaperone binding in vivo (Hegyi and Tompa, 2008). We showed here that the PEVK domain did not interact with HSP27, and others have shown little to no interaction with αB-crystallin (Bullard et al., 2004). The fact that PEVK is not recognized by sHSPs as “unfolded” underscores the specialized nature of this sequence, which is not merely a canonical unfolded peptide (Linke et al., 1998; Leake et al., 2004). N2-Bus, the other large sequence insertion in I-band titin, also demonstrates unique structural organization (Leake et al., 2006), but this domain, although not undergoing aggregation, did associate with HSP27 and αB-crystallin. Because the N2-Bus is intrinsically disordered, mechanical unfolding of this region is probably not required for sHSP binding. Consistent with this view was our observation that exogenous sHSPs bound to unstretched isolated cardiac myofibrils, whereas they showed little (αB-crystallin) to no (HSP27) binding to unstretched skeletal myofibrils, which do not express N2-Bus. In an attempt to find a possible regulatory mechanism for the interaction between N2-Bus and HSP27, we generated HSP27 phosphomimicking mutants because the phosphorylation of sHSPs has occasionally (though not consistently) been reported to promote client binding (Mymrikov et al., 2011). However, we did not observe altered binding of these mutants to N2-Bus compared with wild-type HSP27, suggesting this interaction may be phosphorylation independent. We speculate that sHSP binding to the N2-Bus, regardless of the stimulus, plays a role in promoting the formation of protein complexes, a function proposed for several other chaperones (Hegyi and Tompa, 2008; Donlin et al., 2012). The N2-Bus interacts with the four-and-a-half-LIM domain family of proteins, which target protein kinases to the N2-Bus (Lange et al., 2002; Sheikh et al., 2008). The sHSPs may help regulate the formation of these macromolecular complexes, perhaps in conjunction with other chaperones/cochaperones. A similar scenario was recently suggested for complexes involving the Z-disk protein filamin (Ulbricht et al., 2013).

Aggregation of unfolded titin Ig regions in the sarcomere and sHSP-mediated protection from aggregation have important consequences for the mechanical properties of cardiac and skeletal myocytes. Although αB-crystallin increased the mechanical stability of proximal and distal Ig domains as well as the N2-Bus, according to single-molecule atomic force spectroscopy experiments (Bullard et al., 2004; Zhu et al., 2009), isolated human cardiomyocytes did not become stiffer in the presence of sHSPs. Instead, under conditions promoting titin aggregation (prestretch and acidic pH), the passive tension was abnormally high and αB-crystallin/HSP27 suppressed this increase in stiffness. Thus, the prevention of titin aggregation by sHSPs more than compensated for their domain-stabilization effect: the overall mechanical effect of sHSP binding to I-band titin in stressed sarcomeres decreased the stiffness.

Acidic buffer together with prestretch were prerequisites for the development of abnormally high cardiomyocyte stiffness and sHSP mechanical effects. Acidosis is known to promote the aggregation of unfolded proteins and to boost sHSP binding and client protection (Bennardini et al., 1992; Barbato et al., 1996; Ehrnsperger et al., 1999; Chernik et al., 2004). Acidosis alone is sufficient to trigger αB-crystallin binding to cardiomyofibrils (Barbato et al., 1996) and to increase the chaperone activity of HSP27 (Chernik et al., 2004). In our in vitro assays, acidic conditions marginally promoted aggregation of the proximal Ig domains I9-12 and strongly promoted that of N2-A; αB-crystallin fully protected N2-A from aggregation when present in 10-fold molar excess. Somewhat surprisingly, HSP27 did not prevent N2-A aggregation, even though it protected from abnormally high cardiomyocyte stiffness. However, HSP27 (as well as αB-crystallin) presumably protects titin domains that were not tested here against aggregation, good candidates being the proximal and middle Ig domains according to the I-band position of the sHSPs in stressed myocytes. Moreover, HSP27 can exchange subunits with αB-crystallin and form both homo- and heterooligomers (Sun and MacRae, 2005). Because the two sHSPs were detected at the same I-band titin positions, heterooligomerization cannot be excluded. HSP27 is known to stabilize αB-crystallin (Fu and Liang, 2003), such that the presence of HSP27 could enhance binding of αB-crystallin to titin. In any case, the 6.6–6.7 pH range at which titin Ig domains aggregated and sHSPs exerted their protective effect has been observed in myocytes of metabolically stressed muscle (Khong et al., 2001) and in cardiomyocytes of acute ischemic hearts (Vaughan-Jones et al., 2009), together with the induction of sHSPs. Acidic conditions impair the generation of active tension in skeletal muscle and myocardium (Orchard and Kentish, 1990) and can also increase passive tension (Berger et al., 1999; Miyake et al., 2003). In principle, active contraction may be enhanced by sHSP protective effects on actin, troponin, or myosin. However, the preferential association of sHSPs with the elastic I-band in stressed myocytes indicates that titin is the primary sHSP client. The main role played by sHSPs may thus be the maintenance of low passive stiffness. In line with this view, contractile parameters were unaltered but passive stiffness pathologically elevated during ischemia in mouse hearts deficient in αB-crystallin and HSPB2, compared with ischemic wild-type mouse hearts (Golenhofen et al., 2006). Thus, sHSPs have an unappreciated role in protecting the myocardium from diastolic mechanical dysfunction, such as that during ischemia-reperfusion damage.

The protective effect of sHSPs on titin-based passive stiffness may also be present in other types of heart and skeletal muscle diseases. During heart failure in patients and in animal models, and in subjects with dystrophic muscles, sHSPs tended to be up-regulated in myocytes and translocated to the myofilaments (Knowlton et al., 1998; Fischer et al., 2002; Dohke et al., 2006; Li et al., 2012; Kley et al., 2012). We showed that αB-crystallin and HSP27 were targeted to the cytosol and/or sarcomeric Z-disks in the normal human heart and skeletal muscles. However, in skeletal myocytes of LGMD2A patients and in cardiomyocytes of failing human DCM hearts, a significant portion of the sHSP pool was found in the elastic I-band region colocalizing with the titin springs. Failing hearts often show pathologically altered cardiomyocyte passive stiffness caused by isoform switching and aberrant phosphorylation of titin (Linke and Krüger, 2010); degenerating skeletal myofibers also display abnormal stiffness (Olsson et al., 2006). Our study has revealed additional stiffness-modulating processes common to diseased myocytes involving titin aggregation and its protection by sHSPs.

In summary, this work established a novel role for αB-crystallin and HSP27 in myocytes involving the protection of unfolded titin Ig domains from aggregation in the presence of acidosis. Thus, sHSPs may help prevent titin spring loss of function and the development of abnormally high passive stiffness in myocytes. This protective mechanism is operative in cardiomyocytes of human failing hearts and in skeletal myocytes of muscular dystrophy, in which it can be induced by overstretch and acidic stress. Even in healthy skeletal myocytes during exercise, such could be an early repair mechanism for titin to preserve the protein’s elasticity and additional roles in mechanotransduction and mechanosensation. More generally, the concept of aggregation-induced mechanical stiffening under adverse intracellular conditions and its protection by sHSPs may apply to other mechanical proteins in several cell types. We conclude that in muscle/myocardium, chaperone-mediated protection of titin springs is an effective stress mechanism that maintains low passive myocyte stiffness and preserves diastolic heart function.

Fragments of human I-band titin were expressed in Escherichia coli BL21De3 pLys cells (New England Biolabs, Inc.) and purified using the GST-Fusion System (GE Healthcare). Each construct was expressed twice at minimum. The generated constructs were as follows: Z7-I1 (Z-disk–I-band junction), I9-12 (proximal Ig domains), I24/25 (Ig domains N-terminal to N2-Bus), N2-Bus (large unique sequence of N2-B region), I26 (Ig domain C-terminal to N2-Bus), entire N2-B region (I24/25 + N2-Bus + I26 of exon 49), N2A (four Ig domains and intervening sequences of exons 102–109), and the constitutively expressed PEVK segment (exons 219–225). The human titin sequence was derived from the NCBI accession numbers nm_003319 (N2B isoform) and nm_133378 (N2A isoform). Titin domains were numbered according to Bang et al. (2001). Recombinant human αB-crystallin (NCBI accession no. nm_001885) and HSP27 (NCBI accession no. nm_001540) proteins each were generated using the same expression and purification system. In the cardiomyocyte mechanical studies, we also used commercial recombinant human HSP27 (ab48740; Abcam) and αB-crystallin (ab48779; Abcam). All constructs were verified by sequencing. Primer sequences are listed in Table S1.

To mimic constitutive phosphorylation of HSP27, serine residues 15, 78, and 82 were mutated to aspartate. Site-directed mutagenesis was performed using a two-step PCR reaction (Hamdani et al., 2013), as described in Table S1. Titin fragments were ligated into the pGEX4T2 expression vector and transformed into competent E. coli XL1-blue (Agilent Technologies) or BL21De3 pLys cells (New England Biolabs, Inc.). All constructs were validated by sequencing. Fragments were expressed as GST fusion proteins using 0.2 mmol/l isopropyl-β-d-thiogalactopyranoside and purified using glutathione affinity chromatography. The GST tag was cleaved with thrombin (10 U) and removed from solution using paraaminobenzamidine sepharose beads.

This technique was used to further purify constructs after the GST affinity purification protocol using 1-ml RESOURCE Q columns (GE Healthcare).

Sedimentation-velocity centrifugation was performed in an XL-A centrifuge (Beckman Coulter) equipped with absorption optics and a four-hole titanium rotor. Samples (300–400 µl) were loaded into standard aluminum double-sector centerpieces with quartz windows and spun at 45,000–52,000 rpm at 20°C. Purified titin constructs were suspended in 30 mmol/l Tris-HCl, pH 7.2, and 50 mmol/l KCl buffer in the following concentrations: N2-Bus, 9.5 or 13 µmol/l; N2-B, 3.5 or 5 µmol/l; N2-A, 9.5 µmol/l; and PEVK, 63 µmol/l. Radial scans were recorded at a 0.003-cm resolution every ∼1.5 min. The detection wavelength was 280 nm (N2-Bus/N2-B/N2-A) or 220 nm (PEVK). Data analysis was performed with SedFit, version 11.9 (National Institutes of Health), which estimates the mixture of macromolecular species as a system of noninteracting species with the weight-averaged frictional coefficient ratio (f/f₀). Parameters were set as follows: s-value resolution, 200; s-value range, 0–20 S; confidence interval, 0.95. Time- and radial-invariant noise was calculated; the bottom of the cell and the baseline were fitted. The frictional ratio was floated for the calculation of c(s) and c(M) distributions. Hydrodynamic parameters were corrected for buffer density, viscosity, and partial specific volume, as implemented in SEDNTERP, version 1.09. The partial specific volumes at 20°C were calculated as 0.7347 cm³ g⁻¹ for N2-B and N2-Bus, 0.7323 cm³ g⁻¹ for N2-A, and 0.7589 cm³ g⁻¹ for PEVK, based on the amino acid composition. The mass density of the buffer was calculated as 1.0015 g cm⁻³ and the viscosity as 1.0059 cp at 20°C. Experiments were run at least twice per construct with similar results (only one dataset shown).

Gel-filtration assays were performed using the ÄKTAFPLC on a Superose 12 column (GE Healthcare) in 20 mmol/l Tris/HCl, pH 7.2, and 50 mmol/l NaCl. Proteins were added in a 500-µl super loop and separated by a flow rate of 500 µl/min. The eluate was collected in 500-µl fractions. Eluted proteins were detected at 280 nm (I9-12 construct) or 220 nm (PEVK construct), and the elution profiles (absorption versus elution volume) were recorded. For size determination, results were compared with those obtained with a reference protein of known size (66 kD BSA). Experiments were run at least three times for each construct with similar results (only one dataset shown).

Binding assays were performed at 4°C using recombinant fragments of human I-band titin and human αB-crystallin or HSP27 protein according to standard protocols (Donlin et al., 2012). In brief, the titin fragment or sHSP protein was immobilized on GSH-sepharose beads and incubated with supernatant containing the potential interaction partner. The beads were washed four times, and the interaction was detected by Western blotting with 1-h incubations of anti-titin or anti-sHSP antibodies (dilution 1:1,000) and 1-h incubations of horseradish peroxidase–conjugated anti-rabbit or anti-mouse secondary antibodies (dilution 1:10,000). Detection was performed using the LS 4000 imager (Fujifilm). Each binding test was repeated at least twice.

Using an established protocol for direct binding assays in yeast (Donlin et al., 2012), human titin N2-Bus (bait) was cloned into a pGBKT7 vector and transformed into the yeast strain Y187 (Agilent Techonologies). HSP27 or αB-crystallin (prey) was cloned into a pACT2 vector and transformed into the yeast strain AH109 (Agilent Techonologies) using electroporation. Strains were crossbred and plated on triple drop-out plates. To eliminate false-positive interactions, yeast colonies were tested for active β-galactosidase using an X-Gal filter lift assay. In the presence of X-Gal, positive clones turned blue and were analyzed using PCR.

Myofibrils were isolated from thawed human donor heart, rabbit psoas, or rabbit cardiac muscle tissues following our established protocols (Linke et al., 1999; Neagoe et al., 2002). In brief, single myofibrils were stretched in relaxing buffer (1 mmol/l of free Mg, 100 mmol/l KCl, 2 mmol/l EGTA, 4 mmol/l Mg-ATP, and 10 mmol/l imidazole, pH 7.0) at room temperature using an inverted microscope (Axiovert 135; Carl Zeiss) with a pair of micromanipulators. Non-stretched and stretched samples were incubated for 30 min with recombinant αB-crystallin or HSP27 at a concentration of 0.2 mg/ml in relaxing buffer; this concentration is within the range reported for cardiac and skeletal muscle cells (Inaguma et al., 1995; Larkins et al., 2012). After washing away unbound sHSP, myofibrils were incubated with anti–αB-crystallin or anti-HSP27 antibodies (1:250 dilution) for 30 min, followed by Cy3-conjugated secondary antibodies (1:500) for 30 min. In control experiments, myofibrils were incubated with secondary antibodies alone or with recombinant GFP (0.2 mg/ml). Human cardiomyofibrils were treated before antibody incubation with sHSPs and antibodies with a Ca²⁺-independent gelsolin fragment to extract sarcomeric actin filaments (and their associated regulatory proteins; e.g., troponin). Extraction was verified by actin staining using rhodamine-phalloidin. Fluorescence measurements and the corresponding phase-contrast images were obtained using a Plan-Neofluar 100× oil-immersion objective (NA 1.3; Carl Zeiss), a color 3CCD camera (DXC 930P; Sony), and ImageJ acquisition software (National Institutes of Health). Images were filtered using the built-in “Soften” command of ImageJ. The positions of the fluorescence signals were determined using the measurement tool of ImageJ. The staining results for each experimental condition were verified at least once.

Single rabbit psoas and cardiac myofibrils were used for measurements of dynamic stiffness in relaxing buffer at room temperature. Slightly stretched samples were oscillated using a piezo motor at an amplitude of 20 nm per half-sarcomere and a frequency of 20 Hz, and the oscillatory force response was recorded over a 1-s period using a custom-designed force transducer (Linke et al., 1998, 1999). The mean force-oscillation amplitude was used as a measure of the dynamic stiffness. Measurements were performed in relaxing solutions of varying pH (7.4, 7.0, 6.6, and 5.9), and in the case of the rabbit psoas myofibrils, after treatment with the actin-severing, Ca²⁺-independent gelsolin fragment for 20 min.

TANGO (http://tango.crg.es/protected/academic/calculation.jsp), an algorithm that predicts the regions subjected to aggregation in unfolded polypeptide chains (Linding et al., 2004), was used to predict the aggregation tendency of the I-band titin unfolded regions. The following parameters and conditions were selected: no N- and C-terminal protection, pH 7.0, 300°K temperature, and 0.2 mol/l ionic strength. The relative propensity for β aggregation was plotted for each amino acid of the respective sequence.

Recombinant titin fragments (1 µmol/l final concentration) were exposed to conditions promoting partial denaturation (8 mol/l urea followed by dilution in 30 mmol/l Tris/HCl, pH 7.2 or 6.7, and 50 mmol/l KCl buffer). Many proteins do not refold efficiently after dilution with denaturants but rather form misfolded aggregates. As a measure of aggregation, the absorbance was recorded at 320 nm for up to 30 min. Rhodanese protein aggregation, a known positive control (Martin et al., 1991), was measured under the same conditions and found to be similar to N2-A-titin aggregation in terms of absolute absorbance values. In experiments with the aggregating N2-A construct, we added recombinant human αB-crystallin or HSP27 to Tris buffer at a final concentration of 5 or 10 µmol/l. Experiments were performed at room temperature and run multiple times per condition.

Demembranated single human cardiomyocytes were prepared from donor heart tissue as described previously (Hamdani et al., 2013). In brief, samples were thawed in relaxing buffer (1 mmol/l of free Mg, 100 mmol/l KCl, 2 mmol/l EGTA, 4 mmol/l Mg-ATP, and 10 mmol/l imidazole, pH 7.4), mechanically disrupted, and incubated for 5 min in relaxing solution supplemented with 0.5% Triton X-100. The cell suspension was washed five times in relaxing solution. A single cardiomyocyte was attached in relaxing buffer between a force transducer and a micromotor on the stage of an Axiovert 135 microscope (Plan-Neofluar 40× objective, NA 0.75) as part of the Permeabilized Myocyte Test System (1600A; force transducer 403A; Aurora Scientific). The force response to the stepwise cell stretches (F_passive) was recorded at room temperature in relaxing buffer for a range of SLs (1.8–2.4 µm). The experimental protocol consisted of recording the F_passive, first in normal pH 7.4 buffer (control) and again after the cell was stretched to ∼2.6-µm SL, incubated in pH 6.6 relaxing buffer, and held in the stretched state for 10–15 min (prestretch). The myocyte was returned to the slack length and rested for several minutes before recording the F_passive in the low pH buffer. Alternatively, the pH 6.6 buffer was supplemented with recombinant human αB-crystallin or HSP27 at three different concentrations (0.01, 0.1, or 1 mg/ml), which cover their reported expression range in cardiac and skeletal myocytes (Inaguma et al., 1995; Larkins et al., 2012), and the F_passive was measured in the presence of the sHSPs. In other experiments, the protocol was modified to either omit the prestretch step or to use normal pH 7.4 relaxing buffer throughout. Force values were normalized to the myocyte cross-sectional area calculated from the major and minor diameters of the cell, assuming an oval shape. SL was measured in the experiments using the built-in features of the setup (Aurora Scientific). To test cell viability, each cardiomyocyte was additionally transferred from a relaxing to maximally activating solution (pCa 4.5). Once the steady-state force was reached, the cell was quickly released to 80% of its original length to determine the baseline force. Only cells that developed active forces >20 kN/m² were included in the analysis.

Adult rat hearts were quickly excised and the aorta was tied with a thread to a cannula connected to a perfusion system. Retrograde perfusion with nonoxygenated PBS (20°C) was performed using a peristaltic pump generating pressure sufficient to maintain the left ventricle in a well-stretched state. Alternatively, the aorta was connected to the cannula only loosely to allow ample perfusion of the left ventricle at low pressure and no stretch. Total perfusion time was 30 min. Subsequently, the heart was immediately fixed using a solution containing 4% paraformaldehyde and 15% picric acid. Two left ventricles/experimental condition were prepared for immunohistochemistry.

We obtained nonfailing donor heart tissue and end-stage failing left ventricular tissue from patients with ischemic DCM, as well as Vastus lateralis muscle biopsies from healthy human subjects and patients with LGMD2A, who have a mutation in the CAPN3 gene encoding the calpain 3 protease. Tissue was obtained from two different subjects per condition. Human heart and muscle biopsy collections were in accordance with national guidelines. This study conformed to the principles outlined in the Declaration of Helsinki and was approved by the Ethics Committee at Ruhr University Bochum (entries 3447–09 and 3483–09, respectively).

Heart and muscle tissues were prepared for indirect immunofluorescence using standard protocols. In brief, stretched or unstretched samples were fixed with 4% formaldehyde and embedded in paraffin, and 5-µm-thick sections were prepared. Sections were washed twice with PBS and incubated in blocking buffer (0.5% Triton X-100 and 5% goat serum in PBS) for 1 h. Antibodies were diluted in staining buffer (0.5% goat serum in PBS), the primary antibodies were applied overnight at 4°C, and the secondary antibodies were applied for 2 h at room temperature, with three washes between incubations. Antibody incubations were performed in the dark under agitation. Sections were embedded in Mowiol, and microscopic images were recorded at room temperature using a confocal laser scanning microscope (A1; Nikon) equipped with a CFI Plan Apo 100× oil-immersion objective (NA 1.4) and NIS-Elements software (Nikon). The primary antibodies used were anti-HSP27 (1:250 dilution in PBS; SMC161 5D12; StressMarq), anti-HSP27 (1:200; SR B800; MBL), anti–αB-crystallin (1:250; SMC 165 3A10; StressMarq), anti–αB-crystallin (1:400; SR 223F; MBL), anti-titin PEVK (1:200; 9D10; Hybridoma Bank), and anti-titin PEVK (1:500; custom-made, affinity-purified; Eurogentec; Hamdani et al., 2013). Secondary antibodies were Cy3- or FITC-conjugated IgG (1:400; Rockland). Recorded myocyte images were filtered using the “Soften” command in ImageJ. For some images, we measured the nearest distance across the sarcomeric Z-disk between the mean epitope positions of αB-crystallin, HSP27, and PEVK using ImageJ, as described previously (Linke et al., 1999). The distance between epitopes was plotted against the SL, and data points for each antibody type were fitted by a regression model of order 2.

Hearts were obtained from day 18 embryos of pregnant adult Sprague-Dawley rats. Animals were anaesthetized by isoflurane and euthanized by decapitation. Cells were isolated using enzymatic digestion, the fibroblasts removed by a 60-min preplating step, and cardiomyocytes cultured at 37°C on gelatin-coated coverslips (Krüger et al., 2008). The day after plating, a portion of the fetal cells was treated with the 26S proteasome inhibitor MG132 (5 µM) for 2 h. Other cell groups were perturbed by various stressors, including hypoxia, lactic acid (100 mmol/l; 2 h), contractile arrest (high KCl; Krüger et al., 2008), or 10% cyclic or static stretch using the Flexcell stretcher apparatus (Dunn Labortechnik). Cells were fixed in 3% paraformaldehyde (20 min) and stained with the respective antibodies using standard protocols. Cardiomyocytes were also isolated from adult (3-mo-old) rat hearts by enzymatic digestion as described previously (Kienitz et al., 2011). In brief, Wistar Kyoto rats (∼200 g) were anesthetized by intravenous injection of urethane (1 g/kg) and killed following protocols approved by the animal welfare regulation local authorities in accordance with the guidelines of the European Community (86/609/EEC). Hearts were perfused in a modified Langendorff apparatus under constant flow and sterile conditions; left ventricular myocytes were prepared by enzymatic extraction. Cells were cultured overnight on laminin-coated cover glasses at 37°C in M199 medium (Life Technologies) containing 5% fetal calf serum. Before plating, the serum was removed with several washes of serum-free culture medium (Hepes-buffered M199). Cells were examined between the second and fifth days of culture. Cells were fixed in 3% paraformaldehyde for 20 min at 20°C and stained with antibodies.

For indirect immunofluorescence, cultured cardiomyocytes were incubated for 1 h at room temperature with primary antibodies against HSP25/27 (ProteinTech; dilution 1:250 in PBS) or αB-crystallin (1:500; MLB Antibodies) and counterstained with antibodies against α-actinin (1:500; Sigma-Aldrich) or myosin-binding protein C (Witt et al., 2001), followed by staining for 1 h with FITC- (Sigma-Aldrich) or Cy3-conjugated (Rockland) secondary antibodies (1:500). Images were recorded using an Axiovert 200 microscope (Plan-Neofluar 100× oil-immersion objective, NA 1.3; Carl Zeiss) in epifluorescence mode using a color 3CCD camera (DXC 930P) and ImageJ acquisition software. Images were filtered using the built-in “Soften” command of ImageJ. The epitope positions of endogenous HSP27 and αB-crystallin in sarcomeres were determined in relation to that of the Z-disk or A-band markers. At least 10 different cells per experimental condition were included in the analysis.

The statistical significance of the data were determined using the paired Student’s t test (Prism 5; GraphPad). A value of P < 0.05 was taken to indicate statistical significance.

Fig. S1 shows additional tests for interaction between N2-Bus and sHSPs. Fig. S2 shows the β aggregation tendency of titin regions predicted by TANGO. Fig. S3 shows control measurements of dynamic myofibril stiffness at varying pH values. Fig. S4 shows predictions of force/titin based on entropic elasticity theory. Fig. S5 shows the partial translocation of sHSPs to the Z-disk–I-band region of cultured fetal rat cardiomyocytes after proteasome inhibition. Table S1 lists the human titin primer sequences used in this study.

We thank Anna Eliane Müller for assistance with the adult cell cultures, Judith Krysiak for the yeast two-hybrid data, and Wolfgang Obermann for advice on the protein aggregation assays.

The study was funded by grants from the Deutsche Forschungsgemeinschaft to W.A. Linke (Li 690/7-2; SFB1002, TPB3).

The authors have no conflicting financial interests.