Protein phosphatase 5 regulates titin phosphorylation and function at a sarcomere-associated mechanosensor complex in cardiomyocytes

Serine/threonine protein phosphatase 5 (PP5) is ubiquitously expressed in eukaryotic cells; however, its function in cardiomyocytes is unknown. Under basal conditions, PP5 is autoinhibited, but enzymatic activity rises upon binding of specific factors, such as the chaperone Hsp90. Here we show that PP5 binds and dephosphorylates the elastic N2B-unique sequence (N2Bus) of titin in cardiomyocytes. Using various binding and phosphorylation tests, cell-culture manipulation, and transgenic mouse hearts, we demonstrate that PP5 associates with N2Bus in vitro and in sarcomeres and is antagonistic to several protein kinases, which phosphorylate N2Bus and lower titin-based passive tension. PP5 is pathologically elevated and likely contributes to hypo-phosphorylation of N2Bus in failing human hearts. Furthermore, Hsp90-activated PP5 interacts with components of a sarcomeric, N2Bus-associated, mechanosensor complex, and blocks mitogen-activated protein-kinase signaling in this complex. Our work establishes PP5 as a compartmentalized, well-controlled phosphatase in cardiomyocytes, which regulates titin properties and kinase signaling at the myofilaments.

D uring the lifetime of a beating heart, the cardiomyocytes must respond dynamically to a multitude of internal and external stresses. Such functional flexibility is supported at the level of the contractile units, the sarcomeres, by the expression of cardiac-specific isoforms of structural, contractile, and regulatory proteins. Some of them, such as cardiac troponin-I, myosin-binding protein-C, or titin, contain unique sequence motifs that can be phosphorylated and dephosphorylated by protein kinases and phosphatases, respectively. These selective biochemical events then help to quickly adjust the mechanical function of the cardiac sarcomere to altered physiological requirements, e.g., during exercise. In the diseased heart this finetuned mechanism can be disrupted. Whereas multiplex kinase signaling has been recognized as an important modifier of cardiac function at the level of sarcomeric proteins 1 , much less is known about how this function is modulated by protein phosphatases 2 .
Titin is a multifunctional protein giant, which determines the 'passive' elasticity of the sarcomere 3,4 and also modulates active contractile properties [5][6][7][8] . Human titin encompasses up to3 6,000 amino acids encoded by the 364 exons of the TTN gene and probably is the protein with the most (potential) phosphorylation sites, but very few have been explored functionally 3 . Only one region in titin, termed N2B (encoded by exon 49 in mouse and human), is unique to the cardiac isoforms 9 . This region is located in the elastic (I-band) segment of the molecule and contains a 572-residue N2B-unique sequence (N2Bus), which is an important spring element 10 . Moreover, N2Bus is a hub for protein-protein interactions 3 and a major site for oxidation 11 and phosphorylation [12][13][14][15] . Several protein kinases (PKs) phosphorylate N2Bus, including PKA 12 , PKG 13 , the mitogen-activated protein kinase (MAPK) extracellular signal-regulated kinase 2 (ERK2, encoded by MAPK1) 14 , and Ca 2+ /calmodulin-dependent protein kinase IIδ (CaMKIIδ) 15 . Functionally, enhanced phosphorylation increases the distensibility of N2Bus and lessens the force needed to stretch this spring element, which lowers cardiomyocyte passive tension 3 . Whether this phosphorylation affects protein-protein interactions at N2Bus is unknown. Interestingly, N2Bus shows a phosphorylation deficit in human and experimental heart failure (HF), which contributes to increased myocardial passive stiffness, notably in HF with preserved ejection fraction (HFpEF) 3 .
In the present study, we identified serine/threonine protein phosphatase 5 (PP5, encoded by PPP5C) as a novel interaction partner of N2Bus and a physiological antagonist to the PKs that phosphorylate N2Bus. PP5 is established as a ubiquitous enzyme in eukaryotic cells involved in multiple signaling pathways important for cell cycle progression, glucocorticoid receptor activation, DNA damage repair, transcriptional activation, or apoptosis 16,17 . PP5 is unique among serine/threonine phosphatases as it contains a tetratricopeptide repeat (TPR) domain at the N-terminus, which binds to the C-terminal catalytic domain and blocks substrate access to the catalytic site, such that 'free' PP5 has low activity 18,19 . However, PP5 becomes activated when the autoinhibitory mechanism is released by interaction of the TPR region with Ca 2+ /S100 proteins 20 or heat-shock protein 90 (Hsp90) 21,22 . Activation is also possible through binding of polyunsaturated fatty acids (e.g., arachidonic acid) or long chain fatty acid-CoA esters (LCACE) 23 . Mice deficient in PP5 24 and those overexpressing PP5 in the heart 25 are viable and grow normally, but do show some (more subtle) functional changes in cell types such as fibroblasts 24 or cardiomyocytes 25 . However, little else is known about PP5 in the heart and nothing in human cardiomyocytes.
Here we show that PP5 dephosphorylates titin specifically at N2Bus and reverses the effect of phosphorylation on titin-based passive tension. Through association with Hsp90 and components of the N2Bus-mechanosensor complex, PP5 operates in a compartmentalized manner at the sarcomeric I-bands. Our work establishes PP5 as a well-controlled enzyme in cardiomyocytes, which regulates titin properties and MAPK signaling to support dynamic heart function.

Results
PP5 binds the titin N2Bus domain. In a yeast-2-hybrid screen (Y2H), we used full-length human titin N2Bus as bait and a human adult heart cDNA library as prey. Approximately 150 clones grew on selective media plates and 50 of them were tested positive in a β-galactosidase assay. Among the detected molecules were FHL-2, which was previously identified as a binding partner of N2Bus 31 , and the catalytic domain of PP5 (PP5c), specifically, amino acids 205-266 of human PP5. The interaction with PP5 was verified in a direct Y2H binding assay using N2Bus as bait and PP5c or full-length PP5 as prey (Fig. 1a). Analysis of diploid clones by PCR revealed the expected inserts of 1.7 kb for N2Bus and 1.5 or 1.0 kb for PP5 or PP5c, respectively. The N2Bus-PP5 interaction was also confirmed in vitro by GST-pulldown assays, in which both full-length PP5 and PP5c interacted with N2Bus ( Fig. 1b). Two other I-band titin regions located near N2Bus, PEVK and N2A, did not interact with PP5. To narrow down the interaction site for N2Bus on PP5, we used two PP5 deletion constructs in the GST-pulldown assays, the TPR region at the PP5 N-terminus ('T'-construct, residues 28-129) and the TPR region with additional amino acids reaching into the catalytic domain ('T + '-construct, residues 28 to 211) (Fig. 1a). The Tconstruct showed no interaction with N2Bus, whereas T + did (Fig. 1b), suggesting that the site of interaction involves the Nterminal part of the catalytic domain. Taken together with the Y2H binding data, it appears that amino acids 205-211 of PP5 are critical for the interaction with N2Bus. These residues belong to an α-helix exposed at the outer region of PP5 and are therefore readily accessible 19 . The PP5-N2Bus binding was also probed by co-immunoprecipitation in HEK cells. PP5-HA was coupled to anti-HA agarose beads, N2Bus-myc-containing whole-cell lysate was added and bound protein eluted for detection by western blot. A weak but consistently reproducible interaction was observed (Fig. 1c), confirming that PP5 is a novel interactor of titin at the N2Bus. myofibril and binding detected by indirect immunofluorescence. Consistent with the results of the co-immunoprecipitation assay (Fig. 1c), exogenous PP5c bound only weakly to the myofibrils (Fig. 1d, lower left images). However, the binding was much intensified and clearly localized to the sarcomeric I-bands when the myofibril was phosphorylated (by catalytic subunit of PKA) prior to incubation with PP5c (Fig. 1d, lower right images). In control experiments, neither Cy3-conjugated secondary antibody alone (Fig. 1d, top) nor recombinant green fluorescence protein 32 interacted with the myofibrils, suggesting that PP5 binding was not due to a generally increased stickiness of stretched sarcomeres. Considering the effect of prior phosphorylation of N2Bus on PP5 binding, we performed GST-pulldown assays with recombinant human PP5 and N2Bus, in which the latter was phosphorylated by PKA catalytic subunit or cGMP-activated PKG, or was left unphosphorylated (Fig. 1e). PP5 bound 2-3 times stronger to N2Bus if the titin region was phosphorylated (Fig. 1e, bar graph). Furthermore, in a 162-residue C-terminal fragment of N2Bus we mutated a known PKG-dependent phosphoserine at position 4185 (reference to full-length human titin 13 ) to alanine and used the wildtype (WT) and S4185A N2Bus constructs, phosphorylated by cGMP-activated PKG, in GST-pulldown assays with PP5. A small but significant decrease in binding strength was found for the S4185A mutant (Fig. 1e), providing further evidence that the PP5-N2Bus interaction is enhanced by phosphorylation of N2Bus.

PP5 translocates to sarcomeres if induced in cardiomyocytes.
In testing for PP5 expression in vivo, we found a relatively high level in embryonic (E18) and newborn (P1) rat hearts, which decreased to half in adult rat hearts (Fig. 2a). We thus prepared primary cultures of neonatal rat ventricular myocytes (NRVM) and studied the intracellular localization of PP5 before and after pharmacological manipulation using PP5-activator arachidonic acid (aa; 200 µM) or potent inhibitor okadaic acid (oa; 10 nM) 33 . Interestingly, PP5 activation by aa also increased the expression level of PP5, compared to untreated NRVM, whereas PP5 inhibition by oa decreased it (Fig. 2b) Fig. 1 The elastic N2Bus domain of titin interacts with PP5. a Domain arrangement in the Z-disk/I-band region of cardiac titin N2B/N2BA isoforms and in PP5. Constructs generated for yeast-two-hybrid (Y2H) screens marked in red, those for GST-pulldown assays in blue, and N2Bus-binding amino acids (AA) of PP5 in green. Ig's, immunoglobulin-like domains. Inset: Epitope positions of all phospho-titin antibodies used in this study. (m), anti-mouse; (h), anti-human. b Summary of results of GST-pulldown assays probing interaction of N2Bus with full-length PP5, PP5 catalytic subunit (PP5c), or N-terminal PP5 fragments (T; T+). PP5-binding to PEVK or N2A titin domains was also tested. GST, glutathione S transferase (for negative control). Each test was performed a minimum of two times, mostly three times, with identical results. c Demonstration of PP5-N2Bus association by co-immunoprecipitation assay. PP5 (HA-tag) immunoprecipitations (IP) and whole-cell lysates (WCL) from HEK cells analyzed by western blot for N2Bus (myc-tag) and PP5. PP5 -/+indicates absence/presence of PP5 in the assay. This test was performed three times, with identical results. d Binding of PP5 to sarcomeric I-bands is enhanced by phosphorylation. Top: the experimental design for the stretching of single myofibrils and immunofluorescence image of stretched human cardiac myofibril incubated in relaxing buffer with Cy3-conjugated secondary antibodies alone (control), as well as phase-contrast image (PC). Bottom: representative images of myofibrils incubated with exogenous PP5c and stained against PP5c. The myofibril on the right was incubated with catalytic subunit of PKA before PP5c-treatment (arrowheads, I-band localization of PP5). Binding visualized by anti-PP5c primary and Cy3-conjugated secondary antibodies. Similar results were obtained from four other myofibrils per group. Bars, 2 µm. e Results of GST-pulldown assays probing interaction of PP5 with unphosphorylated N2Bus or N2Bus phosphorylated by PKA/PKG, as well as wildtype (WT) and S4185A mutant of C-terminal N2Bus fragment, both phosphorylated by cGMP-activated PKG. Left: representative immunoblots using anti-PP5 antibody. Right: relative signal intensities in 'Bound' lane, normalized to the mean intensity of the respective non-phosphorylated/C-Term WT control. Data are mean ± s.e.m., n = 4 assays/condition. *p < 0.05 and **p < 0.01, by two-tailed Student's t-test control NRVM, with only an occasional hint at a more regular striation pattern (Fig. 2c). However, stimulation by aa typically caused some PP5 to translocate to the sarcomeric Z-disk/I-band region, as suggested by co-localization with α-actinin ( Fig. 2c; note that due to the short sarcomere length (SL) of NRVM cultures, the I-band and Z-disk regions are hardly distinguishable by confocal microscopy). Treatment with oa always resulted in PP5 staining patterns with no signs of regular striations. Because PP5 negatively regulates MAPK signaling by dephosphorylating Raf1 27 , we also measured the expression and activity (phosphorylation) of the MAPK effector kinases ERK1/2 in NRVM. Whereas ERK1/2 expression was unaltered by aa and oa treatment, phospho-ERK1/2 was reduced by 65% in aa-treated cells but little affected by oa, as compared to controls ( Supplementary  Fig. 1a). Inhibitor of Raf1 (Raf1-I; 20 µM) lowered phospho-ERK1/2 by~50%. Even larger alterations in ERK1/2 activity were seen in NRVM when the MAPK pathway was first stimulated by angiotensin-2 (AngII) or endothelin-1 (ET-1), prior to treatment with aa, oa, or Raf1-I ( Supplementary Fig. 1b). Following ERK1/2 activation by AngII, treatment with oa now had an additional stimulatory effect on phospho-ERK1/2. Taken together, these findings suggest a link between PP5 and MAPK signaling, for the first time also in cardiomyocytes.
We also studied PP5 localization in primary cultures of adult rat cardiomyocytes (ARC) (Fig. 2d). Before PP5 induction, the phosphatase was diffusely cytosolic in some cells, but appeared in a regular striated (sarcomeric) pattern in others. Quantitation of this distribution showed that the proportion of cells with regular PP5 striations increased substantially in aa-treated (PP5-stimulated) compared with control ARC (Fig. 2d, bar graph). In most aa-stimulated cells, PP5 co-localized with the Z/I-region of the sarcomere marked by α-actinin. Treatment with oa did not alter the proportion of cells with regular PP5 striations, in comparison to controls, and the diffuse PP5 distribution was seen more frequently than in aa-treated cells (Fig. 2d). We conclude that, particularly after induction in cardiomyocytes, PP5 is preferentially translocated to the sarcomeres.
PP5 dephosphorylates titin at the N2Bus element. In order to determine whether PP5 dephosphorylates titin, we first quantified the total titin phosphorylation level in ARC cultures. Induction of  . The merged image also shows staining of nuclei using Hoechst. Bars, 10 µm. d PP5 localization in control, aa-treated, and oa-treated adult rat cardiomyocyte (ARC) cultures by indirect immunofluorescence. The same antibodies as in c were used. Bars, 5 µm. Right bar graph shows proportion of ARC exhibiting clear PP5 striations or no such striated pattern, for each group. Numbers above columns indicate total number of cells included in the analysis. e Total titin phosphorylation in the three ARC groups measured by ProQ Diamond phosphoprotein vs. Sypro Ruby total protein stain. Bar graph shows mean ± s.e.m., n = 9 (from three independent cell preparations); *p < 0.05, by two-tailed Student's t-test. f Localization of phosphoserine P-S3991 (titin N2Bus) in control, aa-treated, and oa-treated ARC cultures by indirect immunofluorescence, using anti-N2Bus P-S3991 antibody (secondary antibody: Cy3-conjugated IgG), counterstained with α-actinin antibody (secondary antibody: FITC-conjugated IgG). Bars, 5 µm. Right bar graph shows proportion of ARC exhibiting clear N2Bus P-S3991 striations or no such striated pattern, for each group. Numbers above columns indicate total number of cells included in the analysis PP5 by aa reduced titin phosphorylation by a small but significant amount, whereas oa treatment did not alter it, in comparison to non-treated control cells (Fig. 2e). Next we performed indirect immunofluorescence on ARC using a phosphospecific antibody against rat/mouse phosphoserine P-S3991 located in titin's N2Bus element (equivalent to P-S4010 in human titin 15 ), which is known to be phosphorylated by ERK2 14 and PKA 34 . In most control cells, the N2Bus P-S3991 epitope appeared in a regular striation pattern, co-localizing with the Z/Iregion of the sarcomere marked by α-actinin (Fig. 2f). In PP5stimulated (aa-treated) ARC, the proportion of cells with regular N2Bus P-S3991 striations was reduced compared to controls, whereas oa treatment did not alter this proportion (Fig. 2f, bar graph). These results suggest that induction and translocation of PP5 to the cardiac sarcomeres may result in dephosphorylation of titin at N2Bus.
PP5 is increased in human and experimental heart failure. In keeping with our earlier findings 34,35 , western blotting using phospho-specific antibodies to human titin (for epitope positions, see Fig. 1a) at residues P-S4010, P-S4099 (a PKG-dependent phosphosite) 34 , and P-S4185 detected a phosphorylation deficit at all three of these N2Bus sites in human end-stage failing hearts vs. non-failing donor hearts, which amounted to~25-50% (Fig. 3e). In contrast, phosphorylation at P-11878 and P-12022 within the titin PEVK element was unaltered in failing vs. donor hearts. Interestingly, PP5 expression was~45% higher in the failing hearts (Fig. 3f). Furthermore, in hearts from elderly hypertensive dogs with diastolic dysfunction, we observed~30% less phosphorylation at N2Bus site P-S4010, compared to healthy canine hearts ( Supplementary Fig. 2a), whereas PP5 expression was 60% higher in the diseased hearts ( Supplementary Fig. 2b). We conclude that hypo-phosphorylation of N2Bus in failing myocardium from humans and animal models, which is known to elevate titin-based passive stiffness 3 , may be attributable, at least in part, to increased PP5-mediated dephosphorylation.
PP5 transgenic mouse hearts are hypo-phosphorylated at N2Bus. The increase in PP5 expression observed in failing human hearts prompted us to study titin phosphorylation in a transgenic (TG) mouse model with cardiac-specific overexpression of PP5. The cardiac phenotype of the TG model is mild, but some impairment of cardiac contractility has been noted 25 . The PP5 protein level of the TG hearts was increased, on average by a factor of 7-8, compared to that of matched WT hearts (Fig. 4a). As expected due to the fact that PP5 dephosphorylates Raf1 at S338 27 , phospho-Raf1 S338 was significantly reduced in PP5 TG hearts (Fig. 4a). Immunostaining against endogenous PP5 on myocardial tissue sections demonstrated that the phosphatase was present at low levels in WT cardiomyocytes and was distributed in a diffuse cytoplasmic pattern, with an occasional sarcomeric Iband localization (Fig. 4b, c). In PP5 TG myocytes, the phosphatase localized in a regular striated pattern to the I-bands near the PEVK domain of titin and thus, the N2Bus position, as shown by immunofluorescence (Fig. 4c) and immunoelectron microscopy (Fig. 4b). Quantitation of the distribution of nanogold particles on immunoelectron micrographs indicated the presence of~3 times more particles in PP5 TG than in WT hearts and suggested that~65% of total PP5 protein localized to the I-band springs in PP5 TG cardiomyocytes, compared to~30% in WT cells (Fig. 4b, bar graph).
Total titin phosphorylation was measured by ProQ Diamond/ Sypro Ruby staining and a trend for reduced phosphorylation was observed in PP5 TG vs. WT mouse hearts (Fig. 4d, upper left).
Site-specific phosphorylation was detected by western blot using a panel of phosphoserine-specific antibodies against mouse titin (for epitope positions, see Fig. 1a). Each phospho-specific antibody was used together with a corresponding sequencespecific antibody (Pan), which helped ascertain equal protein loading. All three phosphoserines studied in N2Bus, P-S3991 (ERK2/PKA-dependent), P-S4043 (CaMKIIδ-dependent) and P-S4080 (PKG-dependent), were significantly hypo-phosphorylated (reduction, 45-50%) in PP5 TG vs. WT hearts (Fig. 4d, right  panels). In contrast, a phosphoserine at the Z-disk/I-band junction of titin (P-S2080) showed unaltered phosphorylation, as did P-S12742 in the PEVK domain (which is P-S11878 in humans). We also used the anti-P-S3991 antibody for immunostaining on tissue sections and found regularly spaced doublet lines for N2Bus P-S3991 signals in WT co-localizing with PEVK, but low-intensity N2Bus P-S3991 signals in PP5 TG (Fig. 4e). On immunoelectron micrographs, most nanogold particles indicative of N2Bus P-S3991 were found at the I-band springs in WT, whereas only a small number was at the I-bands in PP5 TG hearts (Fig. 4f). These data led us to conclude that, in vivo, PP5 specifically dephosphorylates phosphoserines of titin located in the N2Bus region.

PP5 transgenic cardiomyocytes have increased passive tension.
Passive force (F passive ) measurements were performed on skinned single cardiomyocytes of WT and PP5 TG mouse hearts in relaxing buffer. Cell stretch over the range 1.8-2.4 µm SL resulted in the typical, quasi-exponential increase in F passive (Fig. 5a). Importantly, the F passive -SL curves were much steeper in PP5 TG than in WT cardiomyocytes, and mean F passive was significantly increased at SL ≥ 2.0 µm (Fig. 5b, c). When PP5 TG and WT cardiomyocytes were incubated with recombinant PP5c, WT cells showed an additional increase in F passive , whereas TG cells were not altered in their stiffness (Fig. 5b). Moreover, F passive could be reduced significantly in both WT and PP5 TG cardiomyocytes upon treatment with ERK2 (Fig. 5c, d), PKA catalytic subunit ( Supplementary Fig. 3a), or cGMP-activated PKG (Supplementary Fig. 3b, c). In the presence of exogenous kinase, F passive was always lower in WT than in TG cells. Finally, incubation of WT myocytes with recombinant PP5c following treatment with ERK2 or PKG (and washout of the kinase) significantly increased F passive to a level well above that measured in untreated WT cells ( Fig. 5d and Supplementary Fig. 3c). We conclude that PP5 is an antagonist to the various PKs that phosphorylate N2Bus, in terms of its effect on cardiomyocyte F passive .
PP5 and N2Bus bind to components of an I-band mechanosensor. N2Bus interacts directly with small heat-shock proteins 32 and FHL proteins, which on their part recruit metabolic enzymes (FHL-2 31 ) and the MAPKs Raf1, MEK1/2 and ERK2 (FHL-1 29 ) to N2Bus, thus forming a putative mechanosensor complex. We studied by GST-pulldown assay whether PP5 also interacts with FHL-1 and found binding with both full-length PP5 and PP5c (Fig. 6a, top). In cardiomyocytes, FHL-1 mainly localized to the sarcomeric I-bands (marked by PEVK antibody) and thus, at the expected N2Bus binding site ( Supplementary  Fig. 4a). There was no difference in FHL-1 localization between PP5 TG and WT hearts. Additional in vitro binding assays revealed that PP5 or PP5c also interacted with ERK2 and Hsp90 (Fig. 6a), both of which have previously been described as binding partners of PP5 28,36 . Since PP1 and PP2a are important phosphatases in cardiomyocytes, we determined their expression levels, as well as that of ERK2, in PP5 TG vs. WT mouse hearts, but observed no differences (Supplementary Fig. 4b).
Focusing on interactors of N2Bus, we confirmed binding of this titin region to FHL-1 29 and FHL-2 31 by GST-pulldown assay (Fig. 6b). FHL-2 also binds ERK2 37 and could thus play a role in the N2Bus-associated mechanosensor. Interestingly, N2Bus interacted in vitro with the α and β isoforms of Hsp90 (Fig. 6b). Since Hsp90 activates PP5 through binding to the TPR region of the phosphatase, we performed in vitro 'competition' assays to test the impact of Hsp90 on the N2Bus-PP5 association. In these assays, pre-incubation of N2Bus-bound sepharose beads with PP5, followed by incubation with Hsp90, resulted in relatively weak binding of PP5 to N2Bus, whereas pre-incubation of N2Bus-bound sepharose beads with Hsp90, followed by incubation with PP5, resulted in significantly stronger binding of PP5 to N2Bus (Fig. 6c). Thus, binding of Hsp90 to N2Bus could promote subsequent PP5 binding to N2Bus. In this context, we immunostained against Hsp90 on myocardial tissue sections from PP5 TG and WT mice and found a diffuse cytosolic distribution of the chaperone in WT, with only an occasional hint at a more regular (sarcomeric) striation pattern (Fig. 6d). However, in PP5 TG hearts, Hsp90 consistently showed a regular striation pattern, in addition to the cytosolic localization, and co-localization with PEVK suggested binding to I-band titin, presumably N2Bus (Fig. 6d). Thus, Hsp90 and PP5 (Fig. 4c) may translocate to N2Bus in a coordinated manner. In conclusion, we confirmed known binary interactions within the N2Bus-associated mechanosensor and provided evidence for additional interactions involving PP5 and some of its binding partners. Data are mean ± s.e.m., n = 5 ROIs from 2 hearts/ group. c PP5 localization in cardiomyocytes from PP5 TG and WT mouse hearts by indirect immunofluorescence. PP5 antibody (secondary antibody: Cy3conjugated IgG), counterstained with anti-PEVK (titin) antibody (secondary antibody: FITC-conjugated IgG). Bars, 2 µm (main) and 1 µm (insets). d PP5 overexpression specifically decreases titin phosphorylation at N2Bus in PP5 TG vs. WT hearts. Total titin phosphorylation measured by ProQ Diamond/ Sypro Ruby staining (upper left), site-specific titin phosphorylation detected by western blot using antibodies to P-S3991, P-S4043, P-S4080 (all N2Bus; right panels), P-S2080 (titin Z/I junction), and P-S12742 (PEVK region). Phospho-titin signals were normalized to total titin signals detected by WB using a panel of sequence-specific antibodies (Pan). Means were indexed to those of control (WT) groups. Data are mean ± s.e.m., n = 4 hearts/group, samples analyzed in triplicate. e Localization of phospho-N2Bus P-S3991 in cardiomyocytes from PP5 TG and WT hearts by indirect immunofluorescence. Anti-N2Bus P-S3991 antibody (secondary antibody: Cy3-conjugated IgG), counterstained with anti-PEVK antibody (secondary antibody: FITC-conjugated IgG). Bars, 2 µm (main) and 1 µm (insets). f Sarcomeric localization of phospho-N2Bus P-S3991 in PP5 WT and TG hearts by immunogold electron microscopy. In vitro data suggested that FHL-1 blocks ERK2-mediated phosphorylation of specific N2Bus residues, including S3991 14 .
However, FHL-1 may also promote N2Bus phosphorylation, because FHL-1 is needed to anchor MAPKs at N2Bus 29 . To address this conundrum, we quantified cardiac titin phosphorylation in a mouse model deficient in FHL-1 29 . Total phosphotitin was significantly lowered by~30%, whereas N2Bus phosphorylation at S3991 was reduced by~50% in FHL-1 knockout (KO) vs. WT hearts (Fig. 7a). As a control, we measured phosphorylation of S2080 at the Z-/I-band junction of titin and found no difference between WT and FHL-1 KO (Fig. 7a). On immunofluorescently stained tissue sections of both WT and FHL-1 KO hearts, PP5 appeared mainly in the cytosolic space and sometimes at the sarcomeres; a difference in PP5-staining intensity and pattern was not consistently observed (Fig. 7b). Anti-P-S3991 antibody marked WT cardiomyocytes in a regular striated (doublet) pattern and at relatively high intensities, overlaying with PEVK, whereas FHL-1 KO cardiomyocytes consistently showed very low anti-N2Bus P-S3991 staining intensity (Fig. 7c). In contrast, anti-P-S2080 antibody gave intense striation patterns in both WT and FHL-1 KO cardiomyocytes (Fig. 7d). On immunoelectron micrographs, N2Bus P-S3991 nanogold particles localized abundantly to the sarcomeric I-band springs in WT but were nearly absent from FHL-1 KO cardiomyocytes (Fig. 7e), whereas anti-P-S2080 antibody again marked the Z/I titin region similarly in both WT and FHL-1 KO (Fig. 7f). These findings show that FHL-1 does not block phosphorylation of N2Bus in vivo (at least at ERK2/PKA-dependent P-S3991) and suggest that FHL-1 supports phosphorylation specifically at N2Bus sites (but not I-band titin in general), presumably by targeting PKs to this spring element.

Discussion
Unlike the phosphatases PP1, PP2a, and PP2b (calcineurin) 2,38 , PP5 has gained little attention in the cardiac field, although it is readily expressed in cardiomyocytes. A reason may be the low basal activity of PP5 brought about by autoinhibition 18,19 . However, PP5 is activated when Hsp90, Ca 2+ /S100 proteins, arachidonic acid or LCACE bind to the enzyme (Fig. 8)   the nucleus of the cardiomyocyte where it normally acts as a transcriptional co-factor 39 and mechanosensor function will be compromised (Fig. 8). In support of this scenario, Raf1 activity was reduced in the cardiomyocytes of PP5-overexpressing TG mice (Fig. 4a). Another effect of PP5 on the mechanosensor is triggered by the dephosphorylation of N2Bus. In PP5 TG hearts, we found lowered site-specific N2Bus phosphorylation and increased cardiomyocyte passive tension, compared to WT. Since the rise in passive tension follows from reduced distensibility of N2Bus due to the hypo-phosphorylation 13 , and considering that mechanosensor function will depend on N2Bus distensibility 29 , PP5 may inactivate the mechanosensor by stiffening N2Bus (Fig. 8). Taken together, PP5-mediated regulation of mechanosensor activity is possible, in principle, by dephosphorylation of Raf1 or N2Bus, or both. MAPK/ERK signaling via Gαq is promoted by G-protein coupled receptor (GPCR) agonists, such as ET-1 or AngII, in cardiomyocyte cultures and intact hearts (Fig. 8) [39][40][41][42] . Our results suggest that PP5 interferes with this pathway under physiological conditions in cardiomyocytes and regulates it in a compartmentalized manner. In cultured NRVM, we observed ERK1/2 activation by AngII or ET-1, hyper-activation of ERK1/2 in the presence of PP5-inhibitor okadaic acid, and very effective suppression of ERK activity by PP5-activator arachidonic acid Reduced N2Bus phosphorylation in FHL-1-deficient cardiomyocytes. a Total titin phosphorylation measured by ProQ Diamond/Sypro Ruby staining (left) and site-specific titin phosphorylation detected by western blot using antibodies to P-S3991 (N2Bus; middle) or P-S2080 (Z/I junction; right). Sitespecific titin phosphorylation levels were normalized to total titin levels detected by WB using sequence-specific antibodies (Pan). Means were indexed to those of control (WT) groups. Data are mean ± s.e.m., n = 3 hearts/group, samples analyzed in triplicate; *p < 0.05, by two-tailed Student's t-test. b, c, d Localization of PP5 (b), phospho-N2Bus P-S3991 (c), and phospho-Z/I-junction P-S2080 (d) in cardiomyocytes from FHL-1 WT and KO hearts by indirect immunofluorescence. Anti-PP5, anti-N2Bus P-S3991 or anti-Z/I-junction P-S2080 (secondary antibody: Cy3-conjugated IgG), counterstained with anti-PEVK antibody (secondary antibody: FITC-conjugated IgG). Bars, 5 µm (main) and 1 µm (insets). e, f Sarcomeric localization of phospho-N2Bus P-S3991 (e) and phospho-Z/I-junction P-S2080 (f) in FHL-1 WT and KO hearts by immunogold electron microscopy. Bars, 500 nm (main) and 100 nm (insets). Bar graph in e and f shows average number of gold particles counted in 50-µm 2 -sized regions-of-interest (ROI), either on the sarcomeric I-band or elsewhere in the cardiomyocyte ('Not on I-band'). Data in e and f are mean ± s.e.m., n = 5 ROIs from two hearts/group. In a and e, *p < 0.05 and ***p < 0.001, by two-tailed Student's t-test  Fig. 1). Cardiomyocytes treated with aa revealed enhanced sarcomeric I-band association of PP5 (Fig. 2), which could be the reason why the phosphatase appeared to have an increased half-life when activated (as indicated by elevated protein expression). Cardiomyocyte-restricted induction of PP5 in a TG mouse model also caused translocation of PP5 to the sarcomeres, specifically to the position of N2Bus (Fig. 4). Although the exact mechanism behind this translocation remains to be elucidated, our work suggested potential triggers. First, we found that phosphorylation of N2Bus enhanced PP5 binding to N2Bus in vitro, and the effect was partially suppressed by mutating a phosphoserine in N2Bus to alanine (Fig. 1d, e), suggesting the added negative charge at N2Bus was the reason for the increased PP5 affinity. Second, we found that PP5-activator Hsp90 translocated to the N2Bus region in PP5-overexpressing cardiomyocytes (Fig. 6d), just as did PP5. Because Hsp90 promoted PP5 binding to N2Bus in vitro (Fig. 6c) and itself interacted with N2Bus (Fig. 6b), the translocation of PP5 to N2Bus might be guided by Hsp90. In this context, it may be of relevance that the N2Bus binding site on PP5 is located not in the TPR region, but at the beginning of the catalytic subunit (Fig. 1a, b). Hsp90 can thus dock to the TPR region and activate PP5, while the phosphatase interacts with N2Bus. In summary, activated PP5 likely performs its physiological functions in cardiomyocytes, including the regulation of Gαq-mediated MAPK/ERK signaling, within a defined compartment at the sarcomeric I-bands (Fig. 8).
Recruitment of PP5 to this compartment is promoted by stimulation via Hsp90 or aa (perhaps also Ca 2+ /S100 and LCACE) and through phosphorylation of N2Bus. Elevated PP5 expression was observed in end-stage failing human hearts and hypertensive dog hearts with diastolic dysfunction (Fig. 3f and Supplementary Fig. 2b). Pathological triggers could thus activate and recruit PP5 to the sarcomeric N2Bus-FHL-MAPK complex in heart disease. In the failing human and dog hearts, we confirmed the reduced site-specific titin phosphorylation at N2Bus, but not PEVK ( Fig. 3e and Supplementary Fig. 2a), reported earlier 34,35,43 . Such phosphorylation deficit at N2Bus is discussed as a main reason for the pathological increase in titin-based myocardial passive stiffness in HF, especially in HFpEF 3,13,44 . Whereas hypo-phosphorylation of N2Bus in HFpEF has so far been explained mainly in terms of pathologically downregulated cGMP-PKG signaling 3,45 , our findings suggest another mechanism, which involves PP5. The phosphatase most efficiently dephosphorylated ERK2phosphorylated human N2Bus in vitro, but also dephosphorylated PKA-or PKG-phosphorylated N2Bus (Fig. 3). Moreover, recombinant PP5c reversed the softening effect of ERK2 or cGMP-activated PKG on isolated skinned cardiomyocytes ( Fig. 5d and Supplementary Fig. 3c). In PP5 TG mouse cardiomyocytes, ERK2/PKA-dependent N2Bus phosphosite P-S3991 was much less phosphorylated than in WT, as was CaMKIIδ-dependent P-S4043 and PKG-dependent P-S4080, but not PKCα-dependent P-S12742 in the PEVK domain or P-S2080 at the Z/I junction of titin (Fig. 4). Thus, the induction of PP5 in failing hearts and the activity of PP5 towards N2Bus (Fig. 3e) can explain why there is a phosphorylation deficit specifically at N2Bus but not at the PEVK domain 34,35,46,47 .
These findings implicate several potential treatment strategies to reduce the high myocardial passive stiffness of HFpEF patients 44 , which are worthwhile to be tested. First, the inhibition of PP5 could increase N2Bus phosphorylation and lower titinbased myocardial passive stiffness. Second, PP5 can associate with  The strain-dependent mechanosensor connecting MAPKs to N2Bus via FHL-1 functions normally, as downstream signaling from Raf1 to ERK2 is enabled. When PP5 expression is increased (as in failing hearts) and PP5 becomes activated through interaction with Hsp90, Ca 2+ /S100 protein, arachidonic acid (aa), or long chain fatty acid-CoA esters (LCACE), the phosphatase translocates to the I-band mechanosensor at N2Bus (right side). Thus, N2Bus (previously phosphorylated by ERK2, PKA, PKG, or CaMKII) is dephosphorylated, which reduces its distensibility and increases titin-based passive tension; the mechanosensor is now less sensitive. Raf-1 is also dephosphorylated and signaling to ERK2 is disabled, such that the mechanosensor function is additionally compromised. The process is embedded in signaling pathways activated via G-protein coupled receptor (GPCR) and Ras, and it can be reversed when PP5 is deactivated. (Molecules that have a color code were studied here, those with no color/white background were inferred from the literature) PP2a, either directly 48 or indirectly through a regulatory subunit that binds both phosphatases 49 . PP2a is known to dephosphorylate multiple cardiomyocyte proteins 38,50 and it has been used experimentally to dephosphorylate titin in vitro 13 . Thus, one can speculate that PP5 may recruit PP2a to the sarcomeric I-band complex at N2Bus (Fig. 8), where these phosphatases could work in concert to dephosphorylate myofilament proteins like titin. If so, it will be interesting to test whether inhibition of PP2a benefits titin properties in normal and failing hearts. Moreover, PP5 could be linked structurally or functionally to other important phosphatases in cardiomyocytes, such as PP1 and calcineurin 2,38,51,52 , which is a possibility worth studying in the future. Finally, PP5 is strongly inhibited by spermine 17 , a natural polyamine closely related to spermidine, which was recently shown to protect aging rodent hearts from diastolic dysfunction by softening the cardiac walls 53 . Importantly, the spermidine effect appeared to be related, in part, to increased titin phosphorylation at N2Bus. If spermine was found to have the same beneficial effects on the heart, an interesting potential therapeutic option for HFpEF patients would be to increase N2Bus phosphorylation and reduce titin-based stiffness by oral supplementation with this polyamine. Conversely, one can test spermidine for possible inhibitory effects on PP5. These novel approaches targeting PP5 should complement other efforts in HFpEF research directed at promoting the PKs (such as PKG 54 ) that phosphorylate N2Bus and reduce cardiomyocyte passive stiffness.
In summary, we demonstrated an important signaling role for PP5 in cardiomyocytes, as the phosphatase binds and dephosphorylates the N2Bus region of titin, which affects cardiomyocyte passive stiffness. PP5 also interacts with components of the N2Bus-associated mechanosensor and blocks MAPK/ERK signaling in this complex. This way, PP5 functions in confined compartments at the sarcomeric I-bands. Reversing the induction of PP5 in heart failure would promote titin phosphorylation and help reduce pathologically increased diastolic stiffness.

Methods
Yeast-two-hybrid screens. To search for interactors of N2Bus we first used an unbiased yeast-two hybrid binding assay, as described 55 . Human N2Bus titin region serving as bait was cloned into the pGBKT7 vector and transferred into yeast strain AH109 (Clontech). Human cardiac cDNA library (Clontech) serving as prey resided in a pACT2 vector, which was transformed into yeast strain Y187 (Clontech) using electroporation. Mating and screening procedures were carried out as described by the manufacturer (Clontech); 3-amino-1.2.4-triazol (70 mM) was used to eliminate background growth on selective media. Diploid yeast clones were let to grow on triple dropout plates. To exclude false-positive interactions, colonies were further analyzed for active β-galactosidase using an X-Gal filter-lift assay. Plasmids from blue-turned, positive clones were isolated and sequenced. A BLAST analysis revealed possible binding partners of N2Bus. To verify interactions, direct binding assays were performed in yeast with the potential binding partner (in pACT2 and Y187) and N2Bus (in pGBKT7 and AH109). Resulting diploid yeast clones from the small scale mating were analyzed by PCR for the presence of both fragments.
GST-pulldown assays were conducted as described previously 32 . Briefly, a specific purified protein was incubated with a GST-fusion protein immobilized on glutathione sepharose beads at 4°C for 1.5 h. Beads were then washed in high and low salt buffer three to four times. Samples were collected from each washing step including a sample of the beads. Analysis was performed by SDS-PAGE and western blot. Mostly, anti-PP5 antibody (target, N-terminal of human PP5; Cell Signaling, #2289; polyclonal, rabbit; 1:2000) or anti-PP5c (catalytic subunit) antibody (target, rat PP5 amino acids 36-238; 3/PP5; BD Biosciences, 611021; monoclonal, mouse; 1:2000) was used for detection, sometimes also anti-N2Bus (titin) antibody (custom-made by Eurogentec; affinity-purified polyclonal, rabbit; 1:500) 15,34 or anti-Hsp90(pan) antibody (target, peptide surrounding Asn300 of human Hsp90; C45G5; Cell Signaling, #4877 S; monoclonal, rabbit; 1:1000). A given interaction test was performed at least three times. Uncropped images of gels and western blots are shown in Supplementary Fig. 5 and Supplementary Fig. 6.
Healthy and failing heart muscle tissues. Left ventricular (LV) samples of nonfailing human hearts were from brain-dead donors with normal LV function. Failing heart samples were from patients with dilated cardiomyopathy (DCM) who underwent heart transplantation due to severe systolic dysfunction (NYHA class III or IV). Sample collection was done in full accordance with Australian National Health Medical Research guidelines and approved by the Human Research Ethics Committee of the University of Sydney (HREC approval: 2012/2814). Informed consent was obtained from the next of kin of all organ donors. It is a requirement of the University of Sydney Human Research Ethics Committee that the consents not only must be sighted before the heart can be included in the Sydney Heart Bank 54 , but any identifying details be redacted. Healthy dog heart tissue was obtained from adult mongrels (8-12 years) and failing heart samples from adult mongrel dogs made hypertensive by bilateral renal wrapping 55 . All dog samples were collected at the Mayo Clinic (Rochester, Minnesota, USA) in full accordance with the institutional guidelines and approved by the Mayo Clinic Ethics Committee. Fetal (E18), newborn (P1) and adult heart tissue was obtained from female Wistar-Kyoto rats, following the guidelines and under the approval of the Animal Care and Use Committee at Ruhr University Bochum, Germany.
PP5-overexpressing transgenic mouse hearts. PP5 TG mice (CD1 strain) with cardiomyocyte-specific overexpression of the phosphatase under the α-myosin heavy-chain promoter were generated as described previously 25 . The transgene consisted of the α-myosin heavy-chain promoter, the entire protein coding region for rat PP5 (plus 483 base pairs of 3′ untranslated sequence), and the SV40 polyadenylation signal sequence. The transgene, isolated from the parent plasmid, was microinjected in fertilized mouse eggs. Mice positive for the transgene were identified via Southern blot and PCR of tail genomic DNA. In total, we studied the hearts of four TG and four litter-matched WT male mice aged 5-6 months, using a TG line with 7-8-fold overexpression of the phosphatase. These PP5 TG hearts showed a slightly compromised contractility in a previous study that provided phenotypic characterization of the model 25 . Hearts from TG and WT animals were collected at University of Muenster in full accordance with the institutional guidelines. Approval for the study was granted by the State Office for Nature, Environment and Consumer Protection North Rhine-Westphalia (LANUV NRW; reference number 8.87-50.10.36.09.006). Samples were prepared for immediate mechanical measurements of isolated cardiomyocytes, immunofluorescence, and immunoelectron microscopy, or deep-frozen for later biochemical analysis.
FHL-1 knockout mouse hearts. A mouse model deficient in FHL-1 (kind gift from Dr. Ju Chen, University of California-San Diego, La Jolla, CA, USA) was generated previously 29 . From this model we obtained adult heart tissue (6-8-month-old male mice) and analyzed three KO and three litter-matched WT hearts. All animal procedures were in full compliance with the guidelines approved by the UCSD Animal Care and Use Committee. The committee approved the procedure of heart extraction from the deceased animals for later biochemical and histochemical analysis.
Adult rat cardiomyocyte cultures. ARC were prepared as described 32 . Briefly, Wistar-Kyoto rats were euthanized with isoflurane, following the guidelines and under the approval of the local Animal Care and Use Committee. The abdomen was opened and ice-cold perfusion buffer (PB) with heparin was injected. The heart was excised and mounted on the cannula of a Langendorff perfusion system. Enzymatic digestion and rinsing were performed with different buffers in the cell isolation process, as described 32 . Cells were incubated in medium supplemented with 4% FCS, ITS (insulin/transferrin/selenium) and butanedione monoxime (BDM; 10 mM). Shortly after plating, this medium was exchanged for serum-free medium (M199, Invitrogen; +ITS, +10 mM BDM). Ventricular cardiomyocytes were used for experiments from day 2 onwards. Cells were treated with arachidonic acid (aa, 200 µM; 2 h), or okadaic acid (oa, 10 nM; 1 h), while other cells were left untreated ('control'). Cells were then harvested as described above for NRVM. Total titin phosphorylation was detected on titin gels stained with the ProQ Diamond (phosphoprotein) vs. Sypro Ruby (total protein) system. Other cells were immunostained using antibodies against PP5 (target, N-terminal of human PP5; Cell Signaling, #2289; polyclonal, rabbit; 1:50) and α-actinin, (target, rabbit skeletal α-actinin; EA-53; Sigma-Aldrich, #A7811; monoclonal, mouse; 1:300) or phospho-N2Bus S3991 (custom-made by Eurogentec; affinity-purified polyclonal, rabbit; 1:400) 15,34 and α-actinin, and indirect immunofluorescence was again performed using confocal laser scanning microscopy. On those samples, we counted the proportion of cells showing a clear striation pattern for either PP5 or titin P-S3991, relative to the total number of cells. Counting was done by a person blinded to the identity of the cells.