Lumican is increased in experimental and clinical heart failure, and its production by cardiac fibroblasts is induced by mechanical and proinflammatory stimuli

Abstract

During progression to heart failure (HF), myocardial extracellular matrix (ECM) alterations and tissue inflammation are central. Lumican is an ECM-localized proteoglycan associated with inflammatory conditions and known to bind collagens. We hypothesized that lumican plays a role in the dynamic alterations in cardiac ECM during development of HF. Thus, we examined left ventricular cardiac lumican in a mouse model of pressure overload and in HF patients, and investigated expression, regulation and effects of increased lumican in cardiac fibroblasts. After 4 weeks of aortic banding, mice were divided into groups of hypertrophy (AB) and HF (ABHF) based on lung weight and left atrial diameter. Sham-operated mice were used as controls. Accordingly, cardiac lumican mRNA and protein levels were increased in mice with ABHF. Similarly, cardiac biopsies from patients with end-stage HF revealed increased lumican mRNA and protein levels compared with control hearts. In vitro, mechanical stretch and the proinflammatory cytokine interleukin-1β increased lumican mRNA as well as secreted lumican protein from cardiac fibroblasts. Stimulation with recombinant glycosylated lumican increased collagen type I alpha 2, lysyl oxidase and transforming growth factor-β1 mRNA, which was attenuated by costimulation with an inhibitor of the proinflammatory transcription factor NFκB. Furthermore, lumican increased the levels of the dimeric form of collagen type I, decreased the activity of the collagen-degrading enzyme matrix metalloproteinase-9 and increased the phosphorylation of fibrosis-inducing SMAD3. In conclusion, cardiac lumican is increased in experimental and clinical HF. Inflammation and mechanical stimuli induce lumican production by cardiac fibroblasts and increased lumican altered molecules important for cardiac remodeling and fibrosis in cardiac fibroblasts, indicating a role in HF development.

Abbreviations

BNP

brain natriuretic peptide

CFB

cardiac fibroblasts

COL1A2

collagen type I alpha 2

COL3A1

collagen type III alpha 1

DCM

dilated cardiomyopathy

ECM

extracellular matrix

HF

heart failure

ICM

ischemic heart failure

IL-1β

interleukin-1β

IL-6

interleukin-6

IVSTd/s

interventricular septum thickness in diastole/systole

LAD

left atrial diameter

LOX

lysyl oxidase

LPS

lipopolysaccharides

LV

left ventricular

LVIDd/s

LV internal diameter in diastole/systole

MMPs

matrix metalloproteinases

PWTd

diastolic posterior wall thickness

RWT

relative wall thickness

SLRPs

small leucine-rich proteoglycans

TIMP

tissue inhibitor of MMP

TGF-β

transforming growth factor β

Introduction

Pressure overload of the left ventricle, as seen in patients with hypertension and aortic stenosis, leads to increased mechanical wall stress that initiates cardiac remodeling, consisting of cardiomyocyte hypertrophy and alterations of the extracellular matrix (ECM). If not adequately treated, hypertension and aortic stenosis might ultimately lead to heart failure (HF) and death [1]. Despite increasing knowledge and better therapies in recent years, HF is still one of the leading causes of death in the Western world suggesting that essential, mostly unknown, pathogenic mechanisms are responsible for the progression to HF from pressure overload. A deeper understanding of the dynamic changes taking place in the ECM constituents during progression to HF is a prerequisite for increased understanding of the pathogenesis and for yielding novel diagnostic, prognostic and therapeutic targets.

During the progression to HF, myocardial ECM alterations in combination with tissue inflammation are central [2, 3]. Lumican is a small ECM-localized proteoglycan, previously reported to be increased in various inflammatory-like conditions such as keratitis [4], inflammatory colitis [5], pancreatitis [6] and nonalcoholic fatty liver disease [7]. Lumican belongs to a protein family named small leucine-rich proteoglycans (SLRPs) consisting of a core protein with leucine-rich repeats and one or more attached glycosaminoglycan chains [8]. Lumican is also known to bind to fibrillar collagens in the cornea, skin and tendon [9], and might thus serve as an ‘anchor’ tying the collagen fibrils together and thereby affecting the quality of the ECM. Lumican is, however, not well studied in the heart, and the role of lumican in the progression to HF has not been clarified. Based on the described effects lumican might have in the heart [10-12], lumican may play a role in the dynamic alterations in cardiac ECM during development of HF.

In this study, we hypothesized that the collagen-binding SLRP lumican is altered in the myocardium during development of HF in mice. We examined whether there were differences in tissue levels of lumican in failing hearts (ABHF) and hearts at an earlier remodeling stage (AB) in response to pressure overload. To gain further mechanistic insight into the regulation and function of lumican, we investigated the regulation and effects of increased lumican on isolated cardiac fibroblasts (CFB). To confirm our experimental findings in patients, we used left ventricular (LV) cardiac biopsies from patients with end-stage HF due to dilated cardiomyopathy (DCM) and ischemic heart failure (ICM) compared with LV biopsies from control hearts.

Results

Cardiac structure and function in the experimental mouse model of AB and ABHF

To study whether there were alterations in the cardiac expression of lumican during remodeling in response to pressure overload, we assessed lumican mRNA levels in the hearts of mice after aortic banding for 4 weeks. Lung weight and left atrial diameter (LAD) were used to separate mice with hypertrophy (AB; n = 13) from those developing HF (ABHF; n = 13) in response to the aortic banding. Sham-operated animals served as controls (n = 11). LV weight was increased by 38 and 73% in AB and ABHF, respectively, compared with sham (Table 1). Cardiac hypertrophy was confirmed by echocardiographic assessment of wall thickness with increases of 35 and 60% in the diastolic posterior wall thickness (PWTd), and 30 and 45% in the diastolic interventricular septum thickness (IVSTd) in AB and ABHF, respectively, compared with sham (Table 1). LV internal diameter was increased by 12% in diastole and 27% in systole in the ABHF group compared with AB (Fig. 1A and Table 1) and by 11% in diastole and 21% in systole compared with sham. LV internal diameter was not significantly altered in AB compared with sham. Relative wall thickness (RWT) was increased by 34% in AB and 37% in ABHF compared with sham. However, there was no significant difference of RWT between AB and ABHF (Fig. 1A and Table 1) and histological staining with hematoxylin and eosin in representative hearts from sham, AB and ABHF mice illustrated hypertrophy with nuclear variation in AB and ABHF (Fig. 1B). Histological staining with acid fuchsin orange G suggested increased fibrosis in the ABHF group (Fig. 1C). Fractional shortening was decreased in ABHF compared with AB and sham, confirming impaired cardiac systolic function in this group. Moreover, both systolic and diastolic tissue velocities were higher in sham and AB groups compared with ABHF (Table 1), reflecting not only systolic dysfunction, but also impaired relaxation in the ABHF group. In line with this, mitral flow deceleration was 1.6-fold higher in ABHF compared with sham, indicating reduced compliance of the left ventricle (Table 1). The relationship between maximal early mitral inflow and early diastolic mitral annular velocities (E/e′) was increased by 2.2- and 1.9-fold in ABHF compared with AB and sham indicating increased LV diastolic stiffness. Consistent with the increased lung weight and echocardiographic measurements, LV gene expression of markers of HF, atrial natriuretic peptide and brain natriuretic peptide (BNP), were increased in the ABHF group compared with AB, and both groups were increased compared with sham (Table 2). Myosin heavy chain β and actin alpha 1, markers of hypertrophic remodeling, were increased in ABHF and AB compared with sham, however, as seen for RWT, there were no significant differences between ABHF and AB (Table 2).

Figure 1

Hypertrophic growth, fibrosis and increased staining of glycosaminoglycans following pressure overload in the experimental aortic banding mouse model. (A) Photographs of representative longitudinal sections of mouse hearts in the four-chamber view. (B) Representative mid-ventricular sections stained with hematoxylin and eosin (HE), which labels cytoplasm and ECM in pink and nuclei in blue. Scale bars 100 μm. (C) Representative mid-ventricular sections stained with acid fuchsin orange G (AFOG), which labels collagen in blue, fibrin in red and cardiomyocytes in orange. (D) Representative mid-ventricular sections stained with alcian blue, which labels alcianophilic polysaccharides, such as glycosaminoglycans, blue. See Table 1 for animal characteristics. The illustrative photomicrography's were taken of mouse hearts following 4 weeks of aortic banding with hypertrophy without heart failure (AB), and heart failure (ABHF), compared with sham.

Table 1. Organ weights and echocardiographic parameters of the experimental mouse model. PWTs, systolic posterior wall thickness; FS, fractional shortening. *P < 0.05 versus sham, **P < 0.01 versus sham, §P < 0.05 versus AB

	Sham (n = 8–11)	AB (n = 11–13)	ABHF (n = 11–13)
Lung weight/tibia length (mg·mm^−1)	8.26 ± 0.12	8.88 ± 0.27	25.6 ± 0.91^**^§
LV weight/tibia length (mg·mm^−1)	5.59 ± 0.15	7.69 ± 0.32^**	9.66 ± 0.16^**^§
LAD (mm)	1.85 ± 0.7	2.01 ± 0.06	3.32 ± 0.09^**^§
PWTd (mm)	0.80 ± 0.03	1.08 ± 0.02^**	1.28 ± 0.03^**^§
LVIDd (mm)	4.00 ± 0.09	3.95 ± 0.08	4.43 ± 0.10^**^§
IVSTd (mm)	0.86 ± 0.05	1.12 ± 0.02^**	1.25 ± 0.03^**^§
PWTs (mm)	1.11 ± 0.06	1.35 ± 0.03^**	1.47 ± 0.04^**
LVIDs (mm)	3.11 ± 0.08	2.96 ± 0.11	3.75 ± 0.11^**^§
IVSTs (mm)	1.13 ± 0.05	1.44 ± 0.03^**	1.52 ± 0.05^**
FS (%)	22.1 ± 1.26	25.11 ± 1.90	15.43 ± 1.11^*^§
RWT (%)	42.0 ± 3.01	56.2 ± 1.86^**	57.7 ± 2.37^**
e′ (early diastolic tissue velocity) (mm·s^−1)	16.5 ± 2.11	23.2 ± 1.86^*	10.6 ± 1.55 ^*^§
a′ (late diastolic tissue velocity (mm·s^−1)	13.5 ± 0.49	12.3 ± 1.08	7.21 ± 1.12^**^§
s’ (systolic tissue velocity) (mm·s^−1)	18.1 ± 0.83	14.8 ± 0.76	9.95 ± 0.59^**^§
E/e′	45.5 ± 5.20	39.9 ± 3.07^**	85.8 ± 10.0^**^§
Mitral flow deceleration (cm·s^−2)	3397 ± 455	4279 ± 308	5499 ± 514^**
Cardiac output (mL·min^−1)	28.6 ± 4.60	24.9 ± 1.92	21.2 ± 2.17
Vmax (m·s^−1)	–	4.09 ± 0.11	3.97 ± 0.10
Heart rate (beat·min^−1)	462 ± 27.8	470 ± 15.0	513 ± 13.4

Table 2. LV mRNA levels in the experimental mouse model. ANP, atrial natriuretic peptide; ACTA, actin alpha 1; MHC-β, myosin heavy chain beta. *P < 0.05 versus sham, **P < 0.01 versus sham, §P < 0.05 versus AB

mRNA	Sham (n = 11)	AB (n = 13)	ABHF (n = 13)
ANP	1.0 ± 0.19	7.83 ± 1.02^**	24.8 ± 1.50^**^§
BNP	1.0 ± 0.14	2.63 ± 0.22^*	6.99 ± 0.66^**^§
ACTA	1.0 ± 0.13	8.48 ± 0.62^**	10.6 ± 1.22^**
MHC-β	1.0 ± 0.10	13.3 ± 3.40^**	21.6 ± 2.62^**
Lumican	1.0 ± 0.06	2.10 ± 0.15^**	2.68 ± 0.20^**^§
Fibromodulin	1.0 ± 0.25	4.73 ± 0.98^*	9.49 ± 0.93^**^§
Decorin	1.0 ± 0.07	1.52 ± 0.12^**	2.14 ± 0.13^**^§
Biglycan	1.0 ± 0.08	2.62 ± 0.23^**	4.88 ± 0.20^**^§
IL-1β	1.0 ± 0.08	1.49 ± 0.28	2.39 ± 0.48^*
IL-6	1.0 ± 0.19	2.22 ± 0.45^**	4.63 ± 0.64^**^§
COL1A2	1.0 ± 0.05	2.01 ± 0.19^**	3.55 ± 0.21^**^§
COL3A1	1.0 ± 0.08	1.62 ± 0.17^*	2.53 ± 0.16^**^§
LOX	1.0 ± 0.06	1.90 ± 0.20^**	3.71 ± 0.23^**^§
MMP-2	1.0 ± 0.07	1.75 ± 0.16^**	3.02 ± 0.16^**^§
MMP-9	1.0 ± 0.14	0.61 ± 0.04^**	0.55 ± 0.06^**
TIMP-1	1.0 ± 0.08	4.15 ± 0.42^**	9.71 ± 0.47^**^§
MMP-9/TIMP-1	1.0 ± 0.11	0.15 ± 0.02^*	0.05 ± 0.00^**^§
TGF-β1	1.0 ± 0.06	1.18 ± 0.04	2.22 ± 0.36^**^§

Interestingly, histology with alcian blue illustrated increased staining in AB and ABHF (Fig. 1D) indicative of increased proteoglycan levels. Lumican mRNA levels were increased 1.3- and 2.1-fold in ABHF compared with AB and sham (Table 2), and the increased cardiac lumican mRNA levels were sustained following aortic banding for 16 weeks (Fig. S1). qPCR confirmed that the increased alcian blue staining unlikely represented increased lumican only, as the mRNA levels of other SLRPs with similar effects in other tissues [8, 13], were increased in ABHF as well (Table 2). In line with this, in LV cardiac biopsies taken from patients with end-stage HF before initiating LV assist device, immunoblotting revealed increased protein levels of fibromodulin, decorin and biglycan compared with controls (Fig. S2).

Cardiac levels of lumican is increased in experimental and clinical HF

Consistent with the 1.3-fold higher mRNA expression in LV of mice with ABHF compared with AB (Table 2), LV levels of glycosylated lumican were 1.7-fold higher in mice with ABHF compared with AB (Fig. 2A,B). Glycosylated lumican was not significantly altered in AB compared with sham (Fig. 2A,B). Increased lumican protein levels in the ABHF group were confirmed by enzymatic deglycosylation treatment of the LV extracts by keratanase II and PNGase F. This revealed the same increased lumican levels in the ABHF group compared with AB and sham, and confirmed that the 50–75 kDa lumican-positive band represented glycosylated lumican (Fig. 2C,D). As expected, deglycosylated lumican was detected at ~ 35 kDa.

Figure 2

Lumican is increased in experimental and clinical heart failure. (A–D) Protein levels of lumican in the LV of mice following 4 weeks of aortic banding with hypertrophy without heart failure (AB; n = 12), and heart failure (ABHF; n = 12), compared with sham (n = 10). All values are relative to sham. Data are shown as representative western blots and mean with SEM (column bar graphs), without (A, B) and with (C, D) keratanase II and PNGase F treatment. (E) Gene expression of lumican in patients with end-stage HF due to DCM (n = 10) and ICM (n = 10) compared with nonfailing hearts (control: n = 10). Values are presented as relative to controls, mean with SEM (column bar graph) (F,G). Protein levels of lumican in patients with end-stage HF due to DCM (n = 15) and ICM (n = 11) compared with control (n = 14). Data are shown as a bar graph, mean with SEM and representative western blot. ***P < 0.001 versus sham or control, §P < 0.05 versus AB. Vinculin and GAPDH were used for protein loading control.

To confirm the clinical relevance of our findings, LV tissue samples from explanted hearts or LV biopsies taken before initiating LV assist device from patients with end-stage HF due to DCM and ICM were assessed for lumican mRNA and glycosylated protein levels. Nondiseased hearts considered for transplantation but deemed unsuitable served as controls. All patients with DCM had ejection fraction < 25% and all ICM patients had ejection fraction < 35% (Table 3). LV internal diameter was 12% greater in patients with DCM compared with ICM (P = 0.01) and PWTd was 10% less (P < 0.05) in patients with DCM compared with ICM. Thus, RWT was decreased in DCM compared with ICM (P < 0.001) (Table 3). There was a tendency of increased pro-BNP in the DCM group compared with ICM (P = 0.15) (Table 3).

Table 3. Clinical and echocardiographic characteristics of end-stage heart failure patients. Clinical and echocardiographic data (mean ± SEM) of patients with ICM and DCM. Nondiseased hearts considered for transplantation but deemed unsuitable were used as control samples (n = 14). The mean age of control patients was 46.0 ± 4.6 years and included seven males and seven females. NYHA, New York Heart Association; BMI, body mass index; LVEF, LV ejection fraction; LVEDV, LV end diastolic volume; LVESV, LV end systolic volume; SV, stroke volume. *P < 0.05; **P < 0.01

	DCM (n = 15)	ICM (n = 11)
NYHA class	3.7 ± 0.2	3.5 ± 0.2
Pro-BNP (pm)	740 ± 165	525 ± 220
Age	35.0 ± 3.7	55.8 ± 2.5^*
Males	12	10
BMI (kg·m^−2)	24.3 ± 1.1	26.2 ± 0.9
LVEF (%)	16.3 ± 0.95 (all < 25%)	21.1 ± 2.2^* (all < 35%)
IVSTd (cm)	0.75 ± 0.06	0.76 ± 0.06
LVIDd (cm)	7.78 ± 0.17	6.95 ± 0.27^*
PWTd (cm)	0.68 ± 0.02	0.75 ± 0.02^*
LVEDV (mL)	303 ± 22	245 ± 29
LVESV (mL)	254 ± 19	197 ± 30
SV (mL)	48.3 ± 3.8	47.8 ± 5.0
RWT (%)	18.3 ± 0.66	21.8 ± 1.11^**

Confirming that our findings from the murine experimental model of HF were clinically relevant, gene expression levels of lumican were increased both in DCM (2.2 ± 0.32, P < 0.05, n = 10) and ICM (2.6 ± 0.28, P < 0.001, n = 10) compared with control hearts (1.0 ± 0.25, n = 10) (Fig. 2E). Consistently, immunoblotting revealed increased glycosylated lumican levels in DCM (2.3 ± 0.20, P < 0.001, n = 15) and ICM (2.2 ± 0.15, P < 0.001, n = 11) compared with controls (1.0 ± 0.14, n = 14) (Fig. 2F,G). There were no significant differences in cardiac lumican gene expression or glycosylated protein levels in patients with DCM compared with patients with ICM (Fig. 2E–G) suggesting that increased cardiac levels of lumican is etiology independent.

Cyclic mechanical stretch induces release of lumican from CFB

To investigate the mechanisms for increased lumican synthesis and release into the ECM of the failing ventricle, we examined lumican production in an in vitro model simulating in vivo mechanical stress, i.e. whether cyclic mechanical stretch of isolated neonatal rat CFB could induce lumican synthesis. Interestingly, in CFB, cyclic mechanical stretch for 24 h revealed a twofold increase in lumican mRNA expression (Fig. 3A). Lumican protein in the cell culture media was increased after 24 h and still significantly increased following 48 h of mechanical stretch (Fig. 3B,C). Thus, mechanical stretch of CFB is sufficient to induce lumican production and release from the cell, suggesting that this could be one of the mechanisms responsible for the increased lumican levels observed in the ECM of the failing LV in vivo.

Figure 3

Lumican production in CFB increases following cyclic mechanical stretch. (A) Gene expression of lumican in primary cultures of CFB following 24 h of cyclic mechanical stretch (MS; n = 6) compared with controls (n = 4). (B,C) Protein levels of lumican in the cell culture media of CFB following 24 and 48 h, respectively of MS (n = 8) compared with controls (Ctr; n = 7–8) with representative western blots. Data are presented as mean with SEM, relative to Ctr. *P < 0.05, **P < 0.01.

The proinflammatory cytokine IL-1β induces lumican synthesis in CFB

Interestingly, glycosylated lumican protein correlated positively to the in vivo LV gene expression of the proinflammatory cytokines interleukin (IL)-1β (R² = 0.29, P < 0.001) and IL-6 (R² = 0.34, P < 0.001), respectively. Similar to lumican, IL-1β mRNA levels were 2.4-fold higher in mice with ABHF compared with sham, however, IL-1β was not significantly different in AB compared with ABHF or to sham (Table 2). IL-6 mRNA levels were increased 2.1- and 4.6-fold in ABHF compared with AB and sham, respectively, and increased 2.2-fold in AB compared with sham (Table 2). To examine whether proinflammatory cytokines associated with HF could directly increase the synthesis of lumican, we stimulated CFB with IL-1β (10 ng·mL⁻¹) and IL-6 (10 ng·mL⁻¹) for 24 h, and analyzed the gene expression of lumican. In CFB, IL-1β and IL-6 directly increased the lumican gene expression 2.9- and 2.1-fold (Fig. 4A). Vehicle (sterile NaCl/P_i) served as a negative control and lipopolysaccharides (LPS) as a positive control [14]. Accordingly, stimulating CFB with 1 μg·mL⁻¹ LPS increased lumican mRNA 3.4-fold and secreted lumican protein 5.7-fold (Fig. 4). Secreted glycosylated lumican protein in the cell culture media of CFB increased 4.7-fold following stimulation with IL-1β (Fig. 4B). However, IL-6 had no significant effect on levels of secreted lumican protein from CFB (Fig. 4B). Thus, IL-1β stimulation of CFB seems to be sufficient to cause production of lumican and release from the cells, suggesting that the proinflammatory IL-1β pathway may constitute another mechanism to increase lumican levels in the ECM of the failing LV in vivo.

Figure 4

Increased lumican synthesis in CFB following stimulation with the proinflammatory cytokine IL-1β and LPS. (A) Gene expression of lumican in primary cultures of CFB following 24 h of stimulation with IL-1β (n = 9) and IL-6 (n = 6) compared with vehicle (Ctr; n = 9). LPS (n = 6) was used as positive control. (B) Protein levels of lumican in the cell culture media of CFB following stimulation with IL-1β (n = 6), IL-6 (n = 6) and LPS (n = 6) compared with Ctr (n = 6). Data are presented as relative to Ctr, mean with SEM. ***P < 0.001 versus Ctr.

Lumican may induce collagen cross-linking through increased production of collagen type I alpha 2 and lysyl oxidase in CFB in vitro

Increased lumican levels in the ECM of the LV in response to increased mechanical stretch and proinflammatory stimuli might affect CFB function and fibrosis. In the experimental mouse model, we revealed a 3.6-fold increase in collagen type I alpha 2 (COL1A2) mRNA in ABHF, and a 2.0-fold increase in AB, compared with sham (Table 2). COL1A2 was 1.8-fold increased in ABHF compared with AB (Table 2) and positively correlated with glycosylated lumican protein (R² = 0.37, P < 0.01). COL3A1 was also positively correlated with glycosylated lumican (R² = 0.56, P < 0.001) and was increased in ABHF compared with AB. ABHF and AB were 2.5- and 1.6-fold increased compared with sham (Table 2). Lysyl oxidase (LOX) is an ECM enzyme that plays a critical role in collagen cross-linking, resulting in the deposition of insoluble collagen fibers and myocardial stiffness [15, 16]. In the ABHF group, LOX mRNA was increased 2.0- and 3.7-fold compared with AB and sham (Table 2), increased in AB compared with sham, and positively correlated to glycosylated lumican protein (R² = 0.40, P < 0.001). Thus, we examined whether increased lumican levels could directly influence collagen synthesis and expression of LOX in CFB in vitro.

When stimulating neonatal rat CFB with recombinant glycosylated lumican (55–65 kDa) we found a 1.9-fold increased mRNA expression of COL1A2 (Fig. 5A), but no significant differences in COL3A1 (data not shown). Importantly, increased lumican levels increased LOX mRNA expression 1.8-fold in CFB (Fig. 5B), suggesting that lumican may increase the amount of cross-linking of collagens. In support of this, immunoblotting showed that stimulation with lumican increased the formation of secreted collagen type I β at ~ 250 kDa in the cell culture media of CFB, probably reflecting increased levels of the dimeric and cross-linked form of collagen type I [17, 18] (Fig. 5C,E), whereas the levels of secreted collagen type I α was decreased (Fig. 5D,E).

Figure 5

Increased collagen type I expression, collagen cross-linking and reduced MMP-9 activity following stimulation of CFB with lumican. (A, B, F, G) Gene expression of COL1A2, LOX, MMP-2 and -9 in primary cultures of CFB following 24 h stimulation with 1.7 nm (n = 24) and 17 nm (n = 6) recombinant N-glycosylated lumican (lum; 55-65 kDa) compared with vehicle (NaCl/P_i; n = 24). (C,D) Protein levels of dimeric collagen type I β and collagen type I α in the cell culture media of CFB stimulated with recombinant N-glycosylated lumican (55–65 kDa) and ascoribic acid (AA) (n = 7) compared with control (vehicle and AA, n = 8). (E) Representative western blots of collagen type I. (H) MMP-9 activity in cell culture media of CFB stimulated with lumican (n = 7) compared with vehicle (n = 9). (I) Representative gelatin zymography. All values are relative to vehicle in their respective cell isolation. *P < 0.05, **P < 0.01, ***P < 0.001.

In the ECM, there is an important balance between collagen deposition and degradation by metalloproteinases (MMPs) and MMP-2 and -9 have been shown to be most altered in the pressure-overloaded heart [19]. In our in vivo model, we revealed decreased MMP-9 and increased MMP-2 mRNA in ABHF and AB compared with sham (Table 2). Stimulation with recombinant glycosylated lumican increased MMP-9 and -2 mRNA in CFB (Fig. 5E,F). Because mRNA levels do not distinguish between the inactive and active form of MMPs, we performed gelatin zymography in the cell culture media of CFB. Interestingly, we found that glycosylated lumican decreased the MMP-9 activity in the cell culture media of CFB (Fig. 5H,I), suggesting that increased lumican might stimulate cardiac ECM accumulation through decreased MMP-9 activity [20]. These data also suggested that lumican had no direct effects on the activity of MMP-2 (Fig. 5I).

Lumican might stimulate cardiac fibrosis through increased TGF-β production and phosphorylation of SMAD3 in CFB in vitro

Although speculative, based on our results indicating that lumican increased the amount of cross-linked collagen type I and decreased MMP-9 activity, we suggest that lumican might be involved in development of cardiac fibrosis during HF progression. In the experimental mouse model, transforming growth factor (TGF)-β1, a central regulator of cardiac fibrosis, was increased 2.2- and 1.9- fold in ABHF compared with sham and AB (Table 2), with a positive correlation to glycosylated lumican (R² = 0.20, P = 0.008). TGF-β1 was not increased in AB compared with sham (Table 2). A connection between lumican and TGF-β has been postulated [21-23], and we chose to examine this further in vitro. Interestingly, we found that lumican increased the mRNA expression of TGF-β1 in CFB 2.2-fold (Fig. 6A). SMAD 2/3 is a downstream pathway of TGF-β, and lumican increased the phosphorylation of SMAD3 in CFB (Fig. 6B). In contrast, we found no significant alterations of phosphorylated SMAD2 (Fig. 6C) in CFB following stimulation with lumican.

Figure 6

Increased TGF-β1 and phosphorylation of SMAD3 in CFB following stimulation with lumican in vitro. (A) Gene expression of TGF-β1 in primary cultures of CFB stimulated for 24 h with 1.7 nm (n = 24) and 17 nm (n = 6) recombinant N-glycosylated lumican (lum; 55-65 kDa) compared with vehicle (NaCl/P_i, n = 24). (B) Protein levels and representative western blot of phosphorylated SMAD3 (pSMAD3) in CFB stimulated with 1.7 nm lumican (n = 8) compared with vehicle (n = 8). (C) Protein levels and representative western blot of phosphorylated SMAD2 (pSMAD2) in CFB stimulated with 1.7 nm lumican (n = 8) compared with vehicle (n = 8). All values are relative to vehicle in their respective cell isolation. *P < 0.05, **P < 0.01, ***P < 0.001.

Lumican might act through the proinflammatory NFκB signaling pathway in CFB in vitro

Lumican has previously been suggested to act on colonic cells and macrophages through the intracellular proinflammatory NFκB pathway [14, 24]. Interestingly, we found that costimulation with lumican and SM7368, an inhibitor of NFκB [25], attenuated the increased expression of COL1A2, LOX, TGF-β1 and MMP-9 (Fig. 7A–D). Thus, lumican acts, at least in part, through the NFκB pathway in CFB. This was further supported by increased levels of the P65 subunit of the NFκB protein in CFB following stimulation with lumican (Fig. 7E). The proposed mechanisms for lumican in HF are illustrated in Fig. 8.

Figure 7

In vitro inhibition of NFĸB attenuated the lumican-induced increased gene expression of COL1A2, LOX, TGF-β1and MMP-9 in cardiac fibroblasts. (A–D) Gene expression in primary cultures of cardiac fibroblasts of COL1A2, LOX, TGF-β1 and MMP-9 following 24 h stimulation with vehicle (n = 24), 1.7 nm lumican (lum; n = 24), 17 nm lumican (n = 6), lum + SM7368 (n = 8) or SM7368 (n = 6). Recombinant N-glycosylated lumican (55–65 kDa) was used. SM7368 was used to inhibit NFĸB activation. Data are presented as relative to vehicle in their respective cell isolation, mean ± SEM. **P < 0.01 versus vehicle, ***P < 0.001 versus vehicle, §P < 0.05 versus lum. (E) Protein levels of P65 NFĸB in CFB following 24 h stimulation with vehicle (n = 8) or 1.7 nm lumican (n = 9). Data are presented as representative western blots and mean ± SEM (column graph). *P < 0.05.

Figure 8

Proposed mechanisms for lumican in heart failure. (A) Increased lumican in the pressure-overloaded failing LV that has undergone the transition from hypertrophy (AB) to heart failure (ABHF). (B) Mechanical stretch and IL-1β lead to increased lumican transcription and secretion in cardiac fibroblasts in vitro. (C) Stimulating cardiac fibroblasts with recombinant N-glycosylated lumican leads to increased transcription of COL1A2, LOX and TGF-β1 in vitro, possibly through the NFκB pathway. Furthermore, recombinant N-glycosylated lumican induced phosphorylation of SMAD3 in cardiac fibroblasts in vitro. Thus, lumican might affect cardiac fibrosis in the failing heart.

Discussion

In this study, mRNA and cardiac protein levels of the ECM SLRP lumican were increased in experimental and clinical HF. Secreted lumican protein from CFB in vitro was increased following mechanical stretch and stimulation with the proinflammatory cytokine IL-1β, two important stimuli during HF progression. Stimulation of CFB with glycosylated lumican induced the expression of molecules important for cardiac remodeling and fibrosis such as the collagen cross-linking enzyme LOX, the dimeric and cross-linked form of collagen type I, TGF-β1, SMAD3 phosphorylation and decreased MMP-9 activity. Thus, our results indicate that increased lumican affects the quality of the ECM during HF progression.

In the in vivo experimental mouse model of pressure overload, lumican mRNA and glycosylated protein levels were increased in ABHF compared with AB and sham. An important finding was that the glycosylated lumican protein was increased only in the hearts of mice subjected to LV pressure overload that had undergone the transition from AB to ABHF, characterized by increased lung weight and LAD. Enzymatic deglycosylation treatment confirmed that the 50–75 kDa lumican-positive bands in ABHF represented glycosylated lumican. As seen in the mouse model, lumican mRNA and glycosylated protein levels were increased in patients with end-stage HF due to DCM and ICM, compared with controls. A microarray study has previously identified increased gene expression of lumican in patients with DCM compared with hypertrophic CM [26]. Hwang et al. [26] did not present any protein data on patients with HF and to our knowledge, our study is the first to reveal increased glycosylated protein levels of lumican in patients with DCM and ICM.

In the in vitro experimental model of mechanical stretch in CFB, we revealed increased lumican mRNA and secreted glycosylated protein levels. Hence, both the in vivo mouse model of pressure overload as well as the in vitro model of mechanical stress using neonatal rat CFB obtained the same results with increased lumican. Although there might be differences in behavior and function of adult mouse CFB and neonatal rat CFB [27, 28], lumican was increased by mechanical stress in mice in vivo as well as in rat CFB in vitro. A recent study reported that mechanical ventilation increased lumican levels in the diaphragm, suggesting that tissue stretch is a mechanism leading to increased lumican in both cardiac tissue and the diaphragm [22]. Our experimental ABHF group had increased LV dilatation compared with AB. Stretch of the cardiac tissue in HF in vivo might induce production of lumican protein, and our in vitro experiments showed that mechanical stretch increased lumican gene expression and lumican protein secretion from CFB. However, although the DCM patients in our study had increased LVID diameter compared with ICM, lumican was not increased in DCM compared with ICM. This suggests additional mechanisms beyond an increase in LV dilatation leading to increased lumican.

Cardiac inflammation has previously been shown to be of importance in cardiac remodeling and HF [10, 29, 30], and has been associated with worsening of symptoms, hospital readmission and mortality [31]. In the in vivo model, we show increased LV gene expression of the proinflammatory cytokines IL-1β and IL-6 in ABHF mice. Furthermore, we found that IL-1β induced lumican mRNA release of lumican from CFB into the cell culture media. As for IL-1β, stimulating CFB with IL-6 increased lumican mRNA expression, although we found no significant alterations in lumican protein secretion. This may suggest that IL-6 plays a role in lumican protein turnover, possibly contributing to increased degradation of the lumican protein that might lead to increased gene transcription. As expected [14], LPS increased lumican mRNA and induced lumican secretion from CFB. Interestingly, it has previously been speculated that intestinal leakage of LPS in HF due to compromised gut function might be the driving factor provoking the increased synthesis of proinflammatory cytokines in HF patients [32]. Furthermore, we have previously reported that several chemokines [10, 12] lead to increased synthesis of lumican from CFB. Thus, because the proinflammatory cytokine IL-1β and chemokines, as well as cyclic mechanical stretch, lead to increased synthesis of lumican in CFB in vitro, this may suggest that cardiac inflammation in HF as well as cardiac tissue stretch induce lumican in HF. Although we have focused on the production of lumican by CFB, other cell types (i.e. endothelial cells, cardiomyocytes and macrophages) might have contributed to the production of lumican in the failing heart in vivo. Furthermore, decorin mRNA has previously been shown to be induced by IL-1β in arterial smooth muscle cells [33] and as seen for lumican, the synthesis of decorin from CFB was increased following stimulation with several chemokines [12]. Thus, in addition to lumican, alterations of other SLRPs as well as other ECM molecules are likely to contribute to ECM remodeling and development of HF.

Activation of CFB and their differentiation into myofibroblasts is a key event in the progression of cardiac fibrosis and HF [34]. We have previously shown that aortic banding induces a transdifferentiation from CFB to myofibroblasts (i.e. increased expression of the myofibroblasts markers αSMA and SM22) following AB [35]. CFB kept under standard 2D cell culture conditions on rigid substrates, such as in the present study, undergo a phenotype switch to myofibroblasts which makes it difficult to investigate functional differences between fibroblasts and myofibroblasts in culture [36]. Thus, in vitro cultures of CFB are likely to contain proto- or fully differentiated myofibroblasts similar to those seen in the pressure-overloaded heart [35-37], and therefore it is likely that lumican produced in our CFB in cultures come from myofibroblasts.

From knockout studies, lumican is known to be important for collagen organization in the skin and cornea [38], and the quality and amount of collagens present in the myocardium are known to have important influence on the mechanical properties of the heart [16, 39]. Stimulation of CFB with recombinant lumican induced COL1A2 mRNA, and increased the dimeric cross-linked form of collagen type I. Furthermore, lumican induced increased expression of LOX which is an important enzyme in regulating collagen cross-linking [40]. Although increased cross-linking might counteract cardiac dilatation, increased collagen and collagen cross-linking will also increase the stiffness of the heart, which may impair both systolic and diastolic function, thus contributing to HF [16, 39]. This is supported by our echocardiographic findings, revealing impaired systolic and diastolic function in the ABHF group, with decreased diastolic and systolic tissue velocities in ABHF compared with AB and sham, as well as increased mitral flow deceleration in ABHF. Increased LOX and collagen cross-linking have been associated with myocardial fibrosis [16]. Histology with acid fuchsin orange G indicated increased fibrosis in ABHF mice, and myocardial fibrosis is a characteristic feature in response to pressure overload [41, 42].

One mechanism contributing to increased fibrosis in HF might be related to lumican-induced TGF-β1. TGF-β has been shown to induce its fibrotic effects through phosphorylation of SMAD3 via the downstream signaling pathway [43-45]. In our in vivo mouse model of pressure overload, we found that LV TGF-β1 was increased in the ABHF group compared with AB and sham. Interestingly, lumican increased both gene expression of TGF-β1 and SMAD3 phosphorylation in CFB, suggesting that lumican may influence fibrosis, at least partly through this pathway. Interestingly, lack of lumican has previously been shown to attenuate the increase of TGF-β1 in the skin and diaphragm [21, 22], and a recent report showed that mice lacking lumican are protected against fibrosis when subjected to hepatic injury [46]. This supports our hypothesis of lumican playing a role in cardiac fibrosis. Thus, lumican might have the opposite effect on fibrosis as seen for decorin, a SLRP that has received attention due to its ability to inhibit fibrosis, possibly due to sequestration of TGF-β1 [47]. We did not find any induction of phosphorylation of SMAD2, another well characterized downstream pathway of TGF-β [48] in response to lumican, although lumican has previously been shown to inhibit SMAD2 and TGF-β2 in osteosarcoma cells [23]. The absence of altered SMAD2 phosphorylation in CFB, as opposed to the inhibitory effect seen in osteosarcoma cells, may be due to different effects on different cell types such as previously demonstrated for the effect of lumican on proliferation in various cancer cell types [49].

Interestingly, stimulation with lumican in CFB increased the P65 NFκB protein and we found that the increased COL1A2, LOX, TGF-β1 and MMP-9 mRNA expression following stimulation with lumican could be attenuated with inhibition of NFκB, suggesting that lumican also acts through the NFκB pathway in CFB, which is supported by previous findings in other cell types [5].

Activation of MMP-9 via the NFκB pathway has previously been shown by others [50, 51]. MMPs are known modulators of myocardial remodeling [52] and the balance between MMPs and their tissue inhibitors determines the maintenance of interstitial tissue homeostasis [53]. Proteoglycans constitute an important class of ECM molecules regulating activation and activity of MMPs [54]. In our in vivo model we found decreased MMP-9 gene expression as well as MMP-9 to tissue inhibitor of metalloproteinase 1 ratio in both the ABHF and the AB group, compared with sham. MMP-9 has previously been shown to be both increased and decreased in patients with HF [20, 55]. Interestingly, gelatin zymography revealed decreased activity of MMP-9 in cell culture media of CFB following stimulation with lumican. This is consistent with the findings of a previous study performed on endothelial cells [56]. Targeted deletion of MMP-9 has been shown to lead to LV enlargement and collagen accumulation [57]. This might indicate that increased lumican could contribute to ECM accumulation. Divergence between increased MMP-9 mRNA expression and decreased activity as shown in this study has also been shown by others [58], and might be due to a negative feedback mechanism. Hence, lumican may bind secreted MMP-9 in the ECM as seen for other proteoglycans [59], and such a decrease in activity might lead to increased mRNA transcription of MMP-9. However, based on our results and previous studies, the connection between lumican and MMPs is not clear and requires further studies.

SLRPs were initially thought to act as structural components, but are now also recognized as signaling molecules, with an expanding repertoire of molecular interactions capable of influencing proliferation, differentiation, survival, adhesion and inflammation [60]. This study supports the idea that lumican might have important signaling functions mediating communication between the ECM and cells regulating cellular behavior, and thus acting as a matrikine [49].

In conclusion, we report increased lumican mRNA expression and protein levels in ABHF mice following pressure overload. This experimental finding was confirmed in the myocardium of patients with HF. In vitro, cyclic mechanical stretch and the proinflammatory cytokine IL-1β increased lumican mRNA expression and secretion of glycosylated lumican from CFB. We suggest that increased lumican in the failing heart might play a pathophysiological role affecting the ECM, initiating increased ECM collagen cross-linking and cardiac fibrosis. This may be due to increased TGF-β1, increased SMAD3 phosphorylation and decreased MMP-9 activity, possibly partly regulated by the NFκB pathway. Hence, in LV pressure overload, lumican may be an important player in HF development, both as a component, and in the regulation of the ECM.

Materials and methods

Mouse model of pressure overload

All animal experiments were approved by The Norwegian Animal Research Committee, which conforms to Guide for the Care and use of Laboratory Animals published by US National Institute of Health. C57Bl/6 wild-type mice were subjected to a banding operation of the ascending aorta. Anesthesia was induced with isoflurane gas in a chamber, the mice were intubated (BioLite Intubation Illuminating System, Braintree Scientific Inc., Braintree, MA, USA) and ventilated on Mini-Vent ventilator (Harvard Apparatus, Holliston, MA, USA). Then aortic banding was performed through a left-sided, muscle-saving thoracotomy under a dissecting microscope (Carl Zeiss Microsopy GmbH, Jena, Germany), as previously described [48, 61, 62]. Within a week after the operation, echocardiography was performed to measure the gradient across the stenosis. Animals with a V_max of 3.5–4.5 m·s⁻¹ were included. Four weeks after the operation, mice were fully characterized by echocardiography before sacrifice. Lung and LV weight were measured before the material was snap-frozen in liquid nitrogen and kept in a −80 °C freezer until molecular analyses. The right tibia bone was carefully removed and the length measured. In the presented data lung and LV weight were normalized to tibia length. Mice were divided into two groups of AB and ABHF based on lung weight and echocardiographic measurements of LAD. Mice with lung weight/tibia length ≤ 11 mg·mm⁻¹ (8.9 ± 0.28) and LAD ≤ 2.3 mm (2.0 ± 0.06) were included in the AB group. Mice with lung weight/tibia length ≥ 20 mg·mm⁻¹ (25.6 ± 0.91) and LAD ≥ 3.0 mm (3.3 ± 0.09) were included in the ABHF group. Sham-operated animals were used as controls. There were no significant differences in maximum flow velocity across the banding-induced stenosis between AB (4.08 ± 0.12) and ABHF (3.97 ± 0.11, P = 0.5). Lung weight was increased approximately threefold in ABHF compared with AB and sham, reflecting pulmonary congestion. LAD was increased ~ 1.7-fold in ABHF compared with AB and sham, reflecting increased LV preload. Eleven sham-operated animals and 13 animals in the ABHF and AB groups were included. There were no correlations between maximum flow velocity and lung weight in the ABHF (R² = 0.01, P = 0.7) or AB (R² = 0.04, P = 0.5) group, supporting a different mechanism from the degree of stenosis responsible for mice developing ABHF or not. RWT was defined in percentage of IVSTd and PWTd divided by LV internal diameter (RWT = (IVSTd + PWTd)/LVIDd).

Sampling of biopsies from HF patients

Biopsies were taken with the understanding and written consent of each patient with HF and from next of kind from controls, and conformed to the Declaration of Helsinki. Permission to obtain myocardial biopsies was granted from The Regional Committee for Medical Research Ethics. Biopsies were taken from explanted hearts or before initiating left ventricular assist device from patients with end-stage HF due to DCM (n = 15) and ICM (n = 11) as previously described [63]. For the failing hearts, in patients undergoing transplantation, LV tissue samples were removed immediately from the still beating explanted hearts. For the failing hearts in patients receiving LV assist device, biopsies were taken immediately before the device was inserted at the same place in the apex of the LV. Nondiseased hearts considered for transplantation but deemed unsuitable because of surgical reasons were used as control samples (n = 14). The control hearts were kept on ice for 1–4 h before tissue sampling. All LV tissue samples were snap-frozen in liquid nitrogen and stored at −80 °C until use. The biopsies were taken in the period 2007–2011. The cause of death of the nonfailing control hearts were cerebrovascular accidents and none had a history of heart disease. Mean age of the control patients was 46.0 ± 4.6 years and seven males and seven females were included. All the nonfailing control hearts had normal echocardiography and ejection fraction as well as normal angiography of the coronary arteries. None of the patients (nonfailing or failing) had significant concomitant disease such as infection, malignancy, or autoimmune disorder. Cardioplegia was only used in nondiseased control hearts immediately before explantation which received one liter modified St. Thomas crystalloid cardioplegia with a temperature of 4 °C.

Isolation and stimulation of neonatal rat CFB

Primary neonatal CFB were isolated from the LV of 1–3-day-old Wistar rats, isolated and digested both mechanically and by a collagenase solution. The cell suspension was transferred to uncoated culture flasks for 20 min, allowing CFB to attach. Unattached cells were removed. The fibroblasts were kept in a Dulbecco's modified Eagle's medium (41965; Gibco, Camarillo, CA, USA) supplemented with fetal calf serum (14-701E; BioWhittaker, Lonza Group Ltd, Basel, Switzerland) and penicillin/streptomyocin (P0781; Sigma-Aldrich, St Louis, MO, USA) and maintained in culture for up to 1 week, before being passaged and seeded onto standard six-well plates or collagen type I-coated six-well Bioflex culture plates (Flexcell; Dunn Labortechnik GmbH, Asbach, Germany) at a density of 1.8 × 10⁵ mL. Cells were kept in a 37 °C, 5% CO₂, humidified incubator. Following 24 h starvation from serum, the cells were stimulated for 24 h with vehicle (sterile NaCl/P_i), 10 ng·mL⁻¹ IL-1β (PRC0814; Gibco), 25 ng·mL⁻¹ IL-6 (406-ML; R&D Systems, Minneapolis, MN, USA), 1 μg·mL⁻¹ LPS (201; List Biological Laboratories, Campell, CA, USA) or cyclic mechanical stretch (1 Hz, 10% stretch), using the FlexCell Tension System (FX-4000; Dunn Labortechnik) as previously described [64]. To study the effect of lumican on CFB recombinant lumican from a mouse myeloma cell line (NS0 cells), GLn19-Asn338 with a C-terminal 6-His tag, was purchased from R&D Systems. Immunoblotting of the recombinant lumican revealed a molecular mass of 50–65 kDa, similar to that seen following immunoblotting of the LV and almost identical to what was seen in the cell culture media of CFB following stimulation with IL-1β (Fig. S3), likely representing N-glycosylated lumican [11]. Concentrations of 1.7 and 17 nm recombinant N-glycosylated lumican were used for the experiments in the present study and represents estimations as precise local in vivo myocardial ECM concentrations are difficult to obtain. Costimulation with 10 μm of the NFκB inhibitor SM7368 (Sigma-Aldrich) was performed. L-Ascorbic acid 2-phosphate has previously been shown to be necessary for decorin induced collagen synthesis [65], thus when studying the effect of secreted collagen type I protein levels following stimulation with lumican, 1 mm of l-ascorbic acid 2-phosphate (Sigma-Aldrich) was added to the cell culture media of both vehicle and lumican stimulated fibroblasts.

RNA extraction

RNA was extracted from LV tissue samples and CFB with the use of RNeasy fibrous tissue mini protocol (RNeasy Mini Kat; #74104; Qiagen GmbH, Hilden, Germany). Concentrations of RNA was measured with NanoDrop (ND-1000; Saveen Werner, Malmö, Sweden). RNA integrity number (RIN) was measured on the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) to characterize the RNA quality. Reverse transcription, was performed with a concentration of 25 ng·μL⁻¹ RNA template using iScript cDNA Synthesis Kit (BioRad, Hercules, CA, USA).

qPCR

All the samples were run in a final reaction volume of 25 μL in duplicates using TaqMan (R) Gene Expression Assays (Applied Biosystems, Foster City, CA, USA) on a 7900 HT Fast Real-Time PCR System (Applied Biosystems). See Table S1 for Inventoried assay ID and inputs. In order to quantify the amount of gene specific RNA, sds 2.3 software (Applied Biosystems) was used. The results were normalized against the reference genes RPL4 or RPL23 for the murine and human samples, respectively.

Protein extraction

Protein was extracted from snap-frozen LV myocardium tissue samples, and homogenates were prepared in homogenization buffer with a polytron 1200 knife. From the cell culture, cell lysate was extracted with a protein lysis buffer (1% Triton X-100, 0.1% Tween 20 and protease inhibitors (Complete; Roche Diagnostics, Basel, Switzerland). The cell culture media was concentrated and purified using Amicon Ultra centrifugal filter units (Millipore Corporation, Billerica, MA, USA). Protein concentrations were measured using Micro BCA protein assay kit (Pierce, #23235; Thermo Scientific, Rockford. IL, USA) and concentrations were read on the Victor 31420 Multi Laber Counter (Perkin-Elmer, Waltham, MA, USA). The absorbance in a serum-free medium sample (Dulbecco's modified Eagle's medium) was subtracted from the absorbance of each of the samples from the cell culture media of stimulated CFB before calculating the protein concentration.

Western blots

Protein lysates were denatured for 5 min at 95 °C in sample buffer containing: 50% sucrose, 7.5% SDS, 0.0625 m Tris/HCl (pH 6.8), 2 mm EDTA (pH 7.5), 3.1% dithiothreitol, 0.01% bromophenol blue, prior to electrophoresis using a Criterion TM Precast Gel 4–15%, Tris/HCL, 1.0 mm 18-well Comb. Catalog#345-0028 (BioRad) and blotted on to polyvinylidene difluoride membranes (GE Health Care Life Science, Uppsala, Sweden). The membranes were blocked for 1 h at room temperature before probing with antibodies (Table S2) at 4 °C overnight. Then the membranes were washed with TBS-T before incubating with species specific horseradish peroxidase secondary antibodies 1 : 4000. To visualize the proteins, enhanced chemiluminescence were used according to the manufacturer's guidelines (GE Healthcare Life Science). Enzymatic deglycosylation was performed to confirm that the 50–75 kDa band of lumican in the in vivo LV tissue samples represented glycosylated lumican. Forty micrograms of protein was digested with 8 μL keratanase II (Seikagaku Biobusiness Corp., Tokyo, Japan) at 37 °C overnight, to remove KS side chains from the protein core of lumican. This sample was further digested with 4 μL of the deglycosylation enzyme mix containing PNGase F. The reaction was terminated by boiling the samples for 10 min at 100 °C. Relative intensities of the bands were quantified with image-quant-tl software (GE Healthcare Life Sciences). The membranes were re-probed with vinculin or GAPDH for protein loading control of the murine and human samples, respectively.

Zymography

Cells were treated for 24 h with vehicle (sterile NaCl/P_i) or recombinant lumican 1 μg·mL⁻¹ (R&D Systems). The samples were not boiled. Mixed sample 1:2 with cell culture medium and 2 × Tris glycine SDS sample buffer were loaded on BioRad 10% zymogram (0.1% gelatin) gels and electrophoresis were run at 125 V for 1.5 h. Following electrophoresis, the gels were incubated with100 mL 1 × Novex zymogram renaturing buffer for 1 h at room temperature before incubated with 1 × Novex zymogram developing buffer at room temperature for 0.5 h and at 37 °C overnight at gentle agitation. Gels were stained with Comassie Brilliant Blue R250 for 45 min and destained (with methanol 40%, acetic acid 10%, dH₂O 50%) for 45 min. The destain solution was changed several times. Gels were then photographed and quantified using image quant-tl software (GE Healthcare Life Sciences). The representative picture of the blue gel has been adjusted to gray scale mode and the picture colors have been inverted.

Illustrative photomicrography

Representative hearts from sham, AB and ABHF were fixed in 4% formaldehyde, cut in longitudinal axis and photographed to demonstrate the general morphology in the four chambered view. They were embedded in paraffin and mid-ventricular 3-μm sections were stained with hematoxylin and eosin and photographed using × 20 objectives to demonstrate the tissue morphology. Alcian blue was used to display staining of mucin, likely reflecting proteoglycans and GAG-chains, while acid fuchsin orange G was used to distinguish the connective tissue from the cells.

Statistical analyses

Data are presented as mean ± SEM. Relative values were normalized to sham/controls. Differences between two groups were analyzed using Student's t-test or Mann–Whitney rank sum test when appropriate and a two-tailed significance level of α = 0.05 were used. Differences among groups were analyzed using one-way analysis of variance (ANOVA) or one-way ANOVA on ranks, and the Holm Sidak or Dunn's method were used to correct for post-hoc multiple testing. Comparisons among three groups were made for all experiments regarding the mouse model (sham, AB and ABHF) and patient data (control, DCM and ICM). For the cell culture experiments with more than two groups, a comparison was made to vehicle/control only, unless otherwise stated in the figure legend (i.e. Fig. 7). Data not normally distributed were transformed by natural logarithm. sigma plot 12.0 (Systate Software Inc, San Jose, CA, USA) was used for the analyses.

Acknowledgements

The authors would like to thank scientists and staff at the Institute for Experimental Medical Research for helpfulness and constructive criticism. We are also grateful for skilful pathological assistance for illustrative photomicrograpy of the mouse model by Sigrid Bjørnstad and technical assistance by Marina Pavlovna Novikova, Marita Martinsen, Heidi Kvaløy, Dina Behmen and Bjørg Austbø. We are deeply thankful to the patients who donated pieces of their heart to science, and to the surgical teams at Oslo University Hospital. This work was funded by The University of Oslo, Oslo University Hospital, Norwegian council on cardiovascular disease, KG Jebsen Cardiac Research Center and Center for Heart Failure Research, Blix Family thrust, Rakel and Otto Bruun thrust, Simon Fougner Hartmanns Family Fund, Anders Jahre's Fund for the Promotion of Science and South-Eastern Norway Regional Health Authority.