Introduction

The importance of the extracellular matrix (ECM) is being recognized increasingly, gradually overturning the concept that the cardiac ECM is inert. ECM proteins not only account for physiological properties of the cardiac tissue, but also contribute to fibrosis and hypertrophy in disease. Fibrosis not only involves structural proteins, but also matricellular proteins, predominantly glycoproteins. Because of the continuous need for dynamic adaptation in the heart, matricellular proteins are of particular importance. Previous studies have demonstrated that ECM deposition facilitates the development of ectopic pacemakers and late potentials as a result of inhomogeneous stimulus conduction, and also can lead to fluctuations in membrane potential.¹ Thus, atrial fibrotic remodeling has been implicated as a potential therapeutic target in atrial fibrillation (AF).²

Proteomics offers the ability to quantify multiple proteins simultaneously without the constraints of antibodies. Most proteomics studies, including the ones on AF, have investigated the cellular proteome.^3,4 We previously have developed a sequential extraction procedure to reduce cellular proteins before studying ECM remodeling in a porcine model of ischemia/reperfusion injury.⁵ No proteomics study has been conducted on the human cardiac ECM.

Therefore, the aims of the present study were 4-fold: (1) to provide a comprehensive characterization of ECM proteins in human atrial appendages, (2) to interrogate protein glycosylation of matricellular proteins, (3) to explore regional differences in the cardiac ECM composition, and (4) to interrogate ECM remodeling in the context of AF. We discovered extensive cleavage in the protein core of the proteoglycan decorin, a key constituent of the cardiac microenvironment. Decorin is a member of the small leucine-rich proteoglycan (SLRP) family^6,7 and regulates collagen fibrillogenesis, and a variety of other ECM molecules involved in cell signaling, as well.⁸ Processed forms of decorin detected in the human cardiac ECM had myostatin and connective tissue growth factor (CTGF) binding properties that might contribute to the pathophysiology of AF by regulating the bioavailability of growth factors.

Methods

An expanded Methods section is available in the online-only Data Supplement. UK, REC number [06/Q0803/69] corresponding to EudraCT number: 2006-005272-40). Atrial tissue was obtained from the atrial appendages during cardiopulmonary bypass and just after cardioplegic arrest of the heart.⁴

ECM Extraction

ECM protein enrichment was performed using our previously published 3-step extraction method, involving sequential incubation with 0.5 mol/L NaCl for 1 hour, 0.1% sodium dodecyl sulfate for 16 hours, and a final incubation for 72 hours with 4 mol/L guanidine hydrochloride.⁵ Matched samples from both atria (n=5) were enriched further in ECM glycoproteins using a lectin binding–based protocol (ie, wheat germ agglutinin and concanavalin A). Guanidine hydrochloride extracts (input), glycoprotein-enriched extracts, and the flowthrough were enzymatically deglycosylated. Gel electrophoresis for deglycosylated proteins was performed using 4% to 12% Bis-Tris gradient gels. Gels were fixed and silver stained. The entire gel lanes were excised and digested using trypsin.⁵ For direct glycopeptide identification, proteins were digested in solution without deglycosylation.

Mass Spectrometry Analyses

Liquid chromatography tandem mass spectrometry (MS/MS) acquisition was performed as previously described.⁹ MS/MS peak lists were matched to a human database using Mascot (version 2.3.01, Matrix Science). Two missed trypsin cleavages were allowed. An additional, no-enzyme search was performed to identify peptides derived from nontryptic cleavages. For the identification of the glycosylation sites, a direct glycopeptide strategy was pursued using a combination of different fragmentation modes on an Orbitrap Elite MS as published elsewhere.¹⁰

P(CAGA)12-Luciferase Reporter Assay

HEK293 cells (ATCC) were transfected with the pGL3(CAGA)12-luciferase reporter plasmid. Recombinant mouse myostatin was added at a final concentration of 2.5 nmol/L ± decorin N-terminal peptides at different concentrations.¹¹ The HEK293 cells were cultured for 24 hours and were then lysed and luciferase activity was assayed.

Proton Nuclear Magnetic Resonance Spectroscopy

Metabolites from mouse cardiac tissue were extracted using a methanol-based procedure. Nuclear magnetic resonance (NMR) spectra were acquired on a 700-MHz Bruker spectrometer by using 9.3 kHz spectral width and 32 000 data points with acquisition time of 1.67 s, relaxation delay of 5 s, and 128 scans. The resulting spectra were processed to 65 536 data points and corrected for phasing and zero referencing by using NMRLab software.

Functional Experiments

To test the biological effects of decorin peptides, neonatal rat cardiomyocytes were used for in vitro experiments, and hearts from 10- to 12-week-old C57BL/6J male mice were perfused in a Langendorff system. Details are given in the online-only Data Supplement.

Statistical Analysis

Clinical characteristics are summarized as mean±standard deviation or as percentages, and tested for differences across groups using unpaired t tests for unequal variances or Fisher exact tests. For all significance testing, a 2-tailed P<0.05 was deemed significant.

Results

Glycoproteomic Characterization of the ECM in Human Atria

Atrial appendages were obtained from patients undergoing coronary artery bypass grafting. The clinical characteristics and average atrial sizes of the patient cohort (n=14) are summarized in Table I in the online-only Data Supplement. ECM protein extracts were prepared as described previously⁵ and analyzed by using a high mass accuracy tandem mass spectrometer (Figure 1A). All identified extracellular proteins are listed in Table II in the online-only Data Supplement.

• Download figure
• Download PowerPoint

We also obtained left and right atrial appendages (LAA and RAA) from the same patients undergoing cardiac surgery (n=5 each) and added enrichment steps for glycoproteins (Figure 1A). Peptides belonging to glycoproteins doubled on enrichment for glycoproteins (Figure I in the online-only Data Supplement). The identified ECM-related glycoproteins, and their distribution, as well, in the input (guanidine hydrochloride extracts), glycoprotein-enriched fraction, and unbound flowthrough are shown in Table III and Figure II in the online-only Data Supplement, respectively. In the glycoprotein-enriched fraction, levels of nidogen 2, cadherin 2 and basal cell adhesion molecule were more abundant in LAA, whereas latent transforming growth factor-β–binding protein 4, fibulin 2, lumican, and vitronectin were increased in RAA (Figure 1B). The latter findings were validated by immunoblotting of the original input material (Figure 1C and 1D).

Next, we combined a direct MS/MS method with glycoprotein or glycopeptide enrichment strategies for the identification of the glycosylation sites (Figure 1A). In total, 65 N- and O-linked glycosylation sites were identified in 35 extracellular proteins. Several known glycosylation sites were confirmed. In addition, putative and novel glycosylation sites were detected (Table IV in the online-only Data Supplement). Identified N-glycoforms were mostly complex, some sialylated or core fucosylated. Minor presences of high mannose and hybrid glycans were observed. O-Linked glycans were predominantly core 1 or 2 types with or without sialic acid. A summary of the proteomics findings is given in Table 1.

Decorin Processing in Cardiac Atria

On enrichment for glycoproteins, decorin peptides were consistently identified in the flowthrough (Figure 2A). Decorin has 3 consensus sequences for N-glycosylation at asparagine (Asn) 211, Asn262, and Asn303. Unlike other SLRPs, all N-glycosylation sites within decorin are located at the C-terminal half of the protein (Figure 2B). Glycosylation of decorin was confirmed by sequential deglycosylation and direct confirmation of its glycosylation sites by MS/MS (Table 1; Figure III in the online-only Data Supplement). A comparison of the sequence coverage revealed that only peptides from the nonglycosylated N-terminal half of decorin were identified in the flowthrough. In contrast, peptides spanning the entire sequence were obtained in the input and the glycoprotein-enriched fraction (Figure 2C). Trypsin is commonly used for protein digestion before MS/MS. Identification of cleavage sites produced by enzymes other than trypsin (nontryptic cleavages) can be indicative of proteolysis. Few ECM proteins displayed evidence for nontryptic cleavages (Table V in the online-only Data Supplement). For decorin, however, 18 cleavage sites were detected (Figure 3A), and 11.9% of all spectra were nontryptic (Figure IVA in the online-only Data Supplement). Despite >50% sequence homology to decorin, only 1.2% of spectra for biglycan were nontryptic. A sequence alignment between decorin and biglycan revealed that most cleavage sites are within nonconserved regions (Figure IVB in the online-only Data Supplement), suggesting that sequence variation provides selectivity to the proteolytic processing of cardiac matricellular proteins.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Decorin Processing Is Region Specific

To investigate regional differences in decorin processing, decorin levels were compared between LAA and left ventricles (LVs). Antibodies raised against 3 different decorin regions were used: the N terminus (Ile36-Leu86), the internal part (Leu136-Ile215), and the C terminus (Phe312-Lys359) of decorin (Figure 3B). All 3 antibodies recognized recombinant human decorin, and decorin in human cardiac tissue, as well (Figure 3C). In agreement with its predicted relative molecular mass (Mr), decorin protein core migrated as a ≈45- to 48-kDa doublet after removal of the glycosaminoglycan chain, but the 3 antibodies provided 3 different results (Figure 3C): according to the antibody recognizing the C terminus, decorin was reduced in LAA in comparison with LV. The antibodies to the internal and the N-terminal part of decorin, however, did not confirm this result, but revealed a Mr difference of ≈3 kDa between decorin in LAA (45 kDa) and LV (48 kDa). C-terminal truncation of decorin may explain this apparent discrepancy. The C-terminal region of decorin inhibits the activity of CTGF. At the protein level, CTGF was more abundant in LV than LAA (Figure 3C, bottom) despite reduced transcript levels (Figure VA in the online-only Data Supplement). A search for putative protease targets on decorin using the PROSPER database returned cathepsins K and G as the most likely effectors (Figure VB in the online-only Data Supplement) for the prominent cleavage at the C terminus observed at Phe330-Ser331 (Figure IVB in the online-only Data Supplement). Transcript levels for cathepsin K were higher in LAA than in LV (Figure VC in the online-only Data Supplement). Taken together, these data suggest region-specific processing of decorin in the human cardiac ECM that may affect the local retention of growth factors.

Decorin Cleavage in AF

AF, the most common arrhythmia encountered in clinical practice, results in significant morbidity and mortality. To provide insight into the consequences of persistent AF on decorin protein abundance, we compared LAA from patients in sinus rhythm (SR) with those with postoperative AF (SR-AF) and in persistent AF (Figure 4A, Figure VIA in the online-only Data Supplement). Tables VI and VII in the online-only Data Supplement summarize their clinical characteristics. At the transcript level, there was no change of decorin or CTGF expression in AF patients and patients who developed postoperative AF in comparison with SR controls (Figure VIB in the online-only Data Supplement). At the protein level, however, the structural damage inflicted by prolonged AF was associated with higher levels of the full-length and the 45-kDa forms of decorin (Figure 4B). The shift in the Mr by 3 kDa was similar to the one previously observed in LVs. Notably, CTGF protein levels were concurrently enhanced (Figure VIC in the online-only Data Supplement). Gene expression data using primers for the different regions of decorin support the notion that the processing of decorin occurs at the protein rather than the transcriptional level (Figure VID in the online-only Data Supplement). In contrast, decorin protein levels were found to be reduced in the SR-AF cohort in comparison with patients who maintained SR postoperatively (SR-SR) (Figure VII in the online-only Data Supplement).

• Download figure
• Download PowerPoint

Next, we interrogated the MS/MS data for potential cleavage sites.¹² Nontryptic peptides of decorin predominantly originated from 4 sites: Val65-Gln66, Leu79-Pro80, Ser121-Pro122, and Phe330-Ser331. The latter cleavage site provides an explanation for the loss of detection with the antibody recognizing C-terminal epitopes. MS/MS spectra are shown in Figure VIIIA in the online-only Data Supplement, alongside elution times for the corresponding tryptic and nontryptic peptides to exclude in-source fragmentation. Notably, an additional cleavage site (Ser49-Leu50) was found in AF (Figure 4C). This cleavage site is next to an N-terminal domain that inhibits myostatin,¹¹ a negative regulator of muscle cell growth. According to the absolute and relative intensities (Figure VIIIB in the online-only Data Supplement), the nontryptic decorin peptide generated by proteolytic cleavage at Ser49-Leu50 was higher in AF patients. Structurally, this part of decorin constitutes a natively disordered region,¹³ which hindered further assessment of putative protease cleavage sites¹⁴ (Figure IX in the online-only Data Supplement).

Inhibition of Myostatin Activity

Myostatin gene expression was significantly reduced in AF patients (Figure 5A). Thus, we hypothesized that this processed form of decorin protein core generated during AF could directly interfere with myostatin bioactivity. Myostatin inhibits phosphorylation of AMP-activated protein kinase (AMPK) and activates SMAD2/3 (Figure 5B). Via the SMAD2/3 cascade, myostatin controls cell growth and differentiation. By inhibiting AMPK phosphorylation, myostatin alters cardiac metabolism. We synthesized peptides corresponding to the N-terminal peptides of mouse and human decorin (Figure XA in the online-only Data Supplement) and tested their activity in a reporter gene assay using HEK293 cells transfected with CAGA-luciferase plasmid. CAGA boxes are recognized by SMAD3 on the promoter regions of SMAD3-responsive genes, and therefore this technique screens SMAD3 activity. Incubation of transfected cells with 2.5 nmol/L myostatin triggered the activation of pathways downstream of the myostatin receptor as evidenced by increased luciferase activity. Addition of the synthetic N-terminal peptide (murine decorin amino acids 42–71, cleaved N terminus) dose-dependently reduced this response (ρ=–1.00, P<0.01) (Figure 5C). A longer decorin peptide (mouse amino acids 31–71, full N terminus) was not as effective (ρ=–0.83, P=0.04) (Figure XB in the online-only Data Supplement). Next, cell size was measured on isolated neonatal rat cardiomyocytes (Figure 5D). The shorter N-terminal peptide mediated a strong repression of myostatin activity as evidenced by an increase in cell size on hypertrophy stimuli (ie, isoproterenol and phenylephrine) (Figure 5E and 5F; Figure XC in the online-only Data Supplement). Thus, decorin cleavage generates peptides with antimyostatin activity.

• Download figure
• Download PowerPoint

Decorin Null Mice

The interaction of decorin with myostatin was further confirmed in hearts obtained from decorin null (DCN^–/–) mice. Decorin deficiency resulted in a ≈50% reduction of myostatin expression (Figure 6A). Related ECM proteins showed no changes in abundance, but decorin deficiency was accompanied by a marked increase in SMAD2 phosphorylation (Figure 6B). Metabolic consequences of decorin deficiency were evident in metabolic profiles obtained by ⁺H-NMR spectroscopy (Figure 6C; Figure XIA in the online-only Data Supplement). Levels of glutamine, succinate, aspartate, and nicotinamide adenine dinucleotides (NAD⁺, NADH) were elevated in decorin null hearts, in comparison with littermate controls (Figure 6D). In isolated cardiomyocytes, myostatin reduced AMPK phosphorylation, a master regulator for cardiac energy metabolism (Figure 6E). This myostatin effect was reversed by preincubation with decorin peptides corresponding to the myostatin-binding region in mice and humans (Figure 6F; Figure XIB and XIC in the online-only Data Supplement). A similar inhibitory effect of the N-terminal decorin peptide on myostatin signaling was observed in Langendorff hearts (Figure 6G and 6H; Figure XID in the online-only Data Supplement).

• Download figure
• Download PowerPoint

Discussion

By applying a novel proteomics approach to human atria, fundamental insights were obtained about the composition and the proteolytic processing of the human cardiac ECM, which has not been analyzed by proteomics thus far. Using proteomics in combination with metabolomics, we have previously shown that discordant metabolic alterations are evident in individuals susceptible to postoperative AF.⁴ We now identify endogenous decorin cleavage products in AF that may contribute to atrial remodeling and play a previously unrecognized role in the susceptibility or perpetuation of arrhythmias by altering the local bioavailability of growth factors, such as CTGF and myostatin.

Glycoproteomics of the Cardiac ECM

Thus far, proteomic studies of cardiac disease have focused mainly on evaluating the differences in protein abundance,^4,5 phosphorylation,^9,15 and redox modifications.¹⁶ However, alterations in glycosylation patterns and protein degradation are known to occur in disease.¹⁷ Our study characterizes, for the first time, ECM proteins in human atria by using state-of-the-art glycoproteomics. Previously, Parker et al¹⁸ used glycoproteomics in rat hearts and characterized glycan composition after removal of the sugar moieties from the peptides. In this study, we analyzed intact glycopeptides. Unlike conventional studies, in which ECM proteins are analyzed by antibodies, proteomics does not require a priori decisions on which ECM proteins to study, and MS/MS is not dependent on the recognition of epitopes. Epitopes can become masked, ie, by glycosylation or be rendered undetectable by proteolytic cleavage. These inherent constraints of antibody-based detection can lead to misinterpretations as demonstrated by the 3 immunoblots for decorin presented in Figure 3C.

Decorin and Other SLRPs

Decorin is a member of SLRPs, a family of 18 structurally related ECM proteins. After synthesis, SLRPs are secreted into the extracellular space and exert diverse roles as biologically active regulators of the ECM.¹⁹ Similar to biglycan, decorin has a horseshoe shape and binds to collagen fibers to control their shape, size, and distribution. Decorin and biglycan share >50% sequence identity and have chemically similar amino acid substitutions at most other positions.²⁰ Yet, decorin, but not biglycan, was subject to extensive proteolytic processing in human atria. Despite the structural and functional similarities of SLRPs, decorin is located in the interstitial ECM, whereas biglycan is mainly associated to collagens in the pericellular regions.²¹ A sequence alignment between decorin and biglycan revealed that most cleavage sites in decorin mapped to nonconserved regions. Thus, specific cleavage sites might endow decorin with additional functional properties in the heart.

Biological Activity of ECM Cleavage Products

Besides its role in ECM assembly, decorin has structural moieties that recognize not only collagens, but also growth factors and cell surface receptors, affecting their activity and bioavailability.^19,22 The regulatory mechanisms involve interactions with sugar residues or specific domains within the core protein.²³ It is noteworthy that, among all SLRPs, decorin had the greatest number of nontryptic cleavage sites, which were predominantly located on the N-terminal half of the protein. In general, the N-terminus and the C-terminus of SLRPs offer the greatest variability for interaction partners.²² The decorin peptides within the N-terminal half have antimyostatin properties.¹¹ An abundant cleavage site at the C-terminus (Phe330-Ser331) gives rise to a peptide with CTGF-binding properties.²⁴ The known cleavage sites of decorin, and the nontryptic cleavages identified in atrial appendages and their relation to binding sequences for collagen and growth factors, as well, are summarized in Figure XII in the online-only Data Supplement.

C-Terminal Decorin Peptides Bind CTGF

The C-terminal tail of decorin, also known as the cysteine-rich ear, is involved in ligand-binding regulation and conformational stability via 2 cysteines at positions 313 and 346. A point mutation within this region is the genetic cause for congenital hereditary stromal dystrophy of the cornea and linked to abnormal fibrillogenesis.²⁵ The abundant cleavage site at position Phe330-Ser331 results in a truncated form of decorin (loss of 3 kDa). C-Terminal truncation of decorin was predominantly observed in LAA and to a lesser extent in LV. Importantly, a synthetic decorin peptide within LRR12 (ie, Gln335-Lys359) has been shown to bind CTGF and attenuate fibrosis.²⁴ Thus, C-terminal truncation may result in diminished collagen-binding ability and repression of CTGF activity in human atria with their lower mechanical load and decreased stiffness compared to LV (Figure 7).

• Download figure
• Download PowerPoint

N-Terminal Decorin Peptides Regulate Myostatin

At the N terminus, 4 cysteine residues form a disulfide knot (Cys54-Cys60 and Cys58-Cys67) that buries the hydrophobic core of the first leucine-rich repeat (LRR1). The binding of Zn²⁺ to this region results in a conformational change required for myostatin binding.²⁶ Equivalent biglycan moieties do not bind myostatin.^26,27 Myostatin, a potent antagonist to hypertrophic stimuli,²⁸ is synthesized as inactive protein with a N-terminal propeptide. After activation, myostatin is regulated via noncovalent binding to its propeptide, forming a latent complex.²⁹ Our experiments demonstrated a markedly lower expression of myostatin in hearts of decorin null mice, supporting the notion of a myostatin-decorin feedback loop. In cardiomyocytes, myostatin regulates at least 2 pathways downstream of the activin receptor (Figure 5B): (1) Coincubation of myostatin with the N-terminal decorin peptide led to inhibition of SMAD2/3 activity in a dose-dependent manner. SMAD3 activation via myostatin represses muscle cell growth.³⁰ When cardiomyocytes were treated with increasing concentrations of the N-terminal decorin peptide, the effect of myostatin on cell size was reversed. In line with these in vitro findings, SMAD2 phosphorylation was increased in hearts of decorin null mice. (2) By inhibiting AMPK phosphorylation via activin receptor and TAK1, myostatin regulates cardiac energy homeostasis.³¹ AMPK is a crucial energy sensor with a phosphorylation site in its activation loop at threonine 172 (Thr₁₇₂).³² In in vitro and ex vivo experiments, both the murine and the human N-terminal peptides of decorin maintained phosphorylation of AMPK at Thr₁₇₂ in the presence of myostatin. In hearts of decorin null mice, with reduced myostatin levels, NMR spectroscopy profiling revealed changes in metabolites linked to mitochondrial metabolism and an accumulation of nicotinamide adenine dinucleotides. Activation of AMPK increases ATP production mostly by enhancing oxidative metabolism leading to a gradual increase in NAD⁺ levels as a consequence of fatty acid oxidation.³³

Decorin in Patients With AF

In human AF, we observed an additional decorin cleavage site next to the myostatin-binding region (Figure 7, position Ser49-Leu50).¹¹ This cleavage site generates a decorin form devoid of the glycosaminoglycan chain at Ser34.¹⁴ We demonstrated that the resulting peptide suppresses myostatin activity and increases cardiomyocyte size. Notably, in a goat model of pacing-induced AF, changes in cellular substructures were accompanied by an increase in myocyte size (up to 195%).³⁴ The additional cleavage site in AF releases decorin peptides that only contain the myostatin-binding region and are therefore less likely to interact with other binding partners. In comparison with other regulators of myostatin activity, ie, myostatin propeptide and follistatin,³⁵ decorin is abundant and readily available within the cardiac ECM. Thus, its cleavage products may constitute an important local regulatory mechanism of myostatin activity.

Evidence in Preclinical Models

Inhibition of myostatin by cardiac-specific overexpression of the myostatin propeptide induced enlarged atria and AF in the transgenic mice.³⁶ Similarly, mice overexpressing microRNA-208a, a myomiR enriched in hearts, developed spontaneous AF. Besides thyroid hormone receptor–associated protein 1, myostatin was identified as a regulatory target of microRNA-208a.³⁷ Similarly, AMPK signaling, one of the key downstream signaling pathways of myostatin, has been implicated as a mechanism for AF in mice with cardiac-specific deletion of liver kinase B1.³⁸ Taken together, these data provide in vivo evidence that loss of myostatin activity and dysregulation of AMPK signaling can act as a substrate for AF, at least in mice.

Study Limitations

Our proteomics findings in clinical samples do not directly prove a decorin-mediated effect on atrial myostatin signaling. Although local injection of decorin peptides in skeletal muscle increased muscle mass,¹¹ in vivo studies are required to further explore the role of decorin cleavage products in cardiac disease, ie, the specificity of the interaction and the location of the effects. Decorin has many binding partners.⁸ Thus, the genetic deletion of decorin in mice may have pleiotropic effects. Lectin enrichment of glycoproteins introduces a selectivity bias toward specific glycoforms, and high Mr glycoproteins in basement membranes, ie, collagens, laminins, and perlecan, are barely resolubilized by buffers compatible with the affinity enrichment of glycoproteins. The direct glycopeptide method used in this study identifies the peptide sequence, the glycosite, and the glycan mass, but the glycan composition is only calculated based on accurate mass and the glycan structure remains unresolved. Emerging technologies combining MS with glycan structural data are yet to be applied to human cardiac tissue.³⁹ Moreover, nontryptic cleavage sites at positions containing lysine or arginine will be missed by digestion with trypsin. Advances in positional proteomics will help to better characterize proteolytic cleavage in cardiac disease.⁴⁰

Conclusions

The glycoproteomic analysis of atrial appendages allowed, for the first time, a detailed analysis of human cardiac ECM proteins. It examined differences in the ECM composition between RAA and LAA, confirmed putative and revealed new glycosylation sites in the cardiac ECM, and identified decorin-derived cleavage products that can act as local regulators of growth factor activity. Proteomic workflows, as outlined in this study, will be essential to explore the different aspects of ECM remodeling in cardiovascular disease, as exemplified by the observed changes in decorin processing during structural remodeling in persistent AF.

Footnotes