Transcriptional changes in bone marrow stromal cells of patients with heart failure

Abstract

It is proposed that patients with heart failure may have not only myocardial dysfunction, but also a reduced regenerative capacity of stem cells. However, very little is known about bone marrow stromal cell (BMSC) characteristics in heart failure and its comorbidities (obesity and/or diabetes). We hypothesized that metabolic alterations associated with the latter will be reflected in altered expression of key genes related to angiogenesis, inflammation, and tissue remodeling in patient-derived BMSCs. We found that BMSCs of heart failure patients with lower body mass index have enhanced expression of genes involved in extracellular matrix remodeling. In particular, body mass index <30 was associated with upregulated expression of genes encoding collagen type I, proteases and protease activators (MMP2, MMP14, uPA), and regulatory molecules (CTGF, ITGβ5, SMAD7, SNAIL1). In contrast, these transcript levels did not differ significantly between BMSCs from obese heart failure patients and healthy subjects. Comorbidities (including obesity and diabetes) are known to play role in heart failure progression rate and outcome of the disease. We thus suggest that key molecular targets identified in this study should become the target of the subsequent focused studies. In the future, these targets may find some use in the clinical setting.

Introduction

Cell therapy has recently attracted interest as a new treatment option for cardiovascular diseases. Bone marrow stromal cells (BMSCs) represent one of the possible cell sources for regenerative therapy. BMSCs are multipotent adult stromal cells that express a wide range of molecules that can modulate cell survival, proliferation, differentiation, tissue neovascularization, and remodeling.^1,2 It is proposed that patients with cardiovascular diseases may not only have myocardial dysfunction, but also attenuated function and, thus, a reduced regenerative capacity of stem and progenitor cells.^2,3 In addition, many HF patients have comorbid conditions that may also influence cell functionality. Currently, little is known about BMSC characteristics in HF patients with or without comorbidities, and the impact of these diseases on BMSC gene expression have not yet been examined.

We hypothesized that metabolic alterations associated with heart failure (with and without obesity and/or diabetes) will be reflected in altered expression of key genes related to angiogenesis, inflammation, and tissue remodeling in patient-derived BMSCs. To test this, expression of key genes involved in these processes was assessed by real-time PCR, in a panel of patient-specific primary cell cultures.

Results

Relations of gene expression with demographic and clinical variables

Among 85 gene transcripts analyzed, 48 showed high correlation with the housekeeping genes and were excluded from subsequent analysis. To investigate the relation of the remaining 37 gene transcripts to clinical variables a general linear model (GLM) procedure was used. Since there was a predominance of male population among patients with HF, and patients were older comparing to the control group, GLM was performed with patients’ data only. Predictor variables included DM status, body mass index (BMI), age, and New York Heart Association (NYHA) functional class. Gene expression data (dependent variables) were entered as dCt values. A separate GLM was run for each gene transcript.

Though age was shown to influence gene expression in human BMSCs,⁴ GLM did not reveal any age-related gene transcripts, since age did not vary significantly among individuals with HF. Only 2 gene transcripts were associated with HF functional class, as the majority of patients fit rather homogenous cohort of NYHA class II-III. Similarly, only 2 genes were linked to DM, probably due to a good compensation of patients with comorbid DM, since the mean levels of blood glucose, fructosamine, HbA1c, and insulin were within the normal range and did not differ significantly between study groups (Table 1).

Table 1. Biochemical, clinical, and demographic data of the study population.

	Donors	HF	HF+Ob	HF+DM
	n = 13	n = 14	n = 16	n = 12
Age, years	39.7 ± 1.9	55.8 ± 3.9**	52.0 ± 1.5**	59.5 ± 3.2**
Males, %	23.1	92.8**	81.3**	75.0**
BMI, kg/m2	25.3 ± 1.1	24.5 ± 0.9	36.7 ± 1.8**‡	33.3 ± 1.8**‡
NYHA class	NA	2.75 (2;4)	2.25 (2;3)	2.75 (2;4)
Glucose, mM/L	NA	5.4 ± 0.1	5.5 ± 0.4	5.9 ± 0.3
HbA1c, %	NA	5.7 ± 0.2	5.4 ± 0.3	6.4 ± 0.1
Fructosamine, mkM/L	232.1 ± 7.5	240.4 ± 5.4	232.3 ± 4.0	253.0 ± 12.1
Insulin, pM/L	69.7 ± 19.0	89.3 ± 10.4	74.9 ± 7.0	76.3 ± 10.1
Total cholesterol, mM/L	5.2 ± 0.4	5.1 ± 0.4	4.8 ± 0.3	4.3 ± 0.16
LDL, mM/L	2.5 ± 0.3	3.4 ± 0.3	2.8 ± 0.2	2.3 ± 0.2
Triglycerides, mM/L	1.4 ± 0.3	1.4 ± 0.1	1.8 ± 0.2	2.0 ± 0.4
Leptin, ng/mL	16.5 ± 4.7	18.6 ± 4.6	40.1 ± 7.0*+	22.7 ± 7.1
Adiponectin, mkg/mL	7.0 ± 1.2	10.4 ± 1.8	10.1 ± 1.5	9.3 ± 2.4
NT-proBNP, pM/L	0.0 ± 0.0	83.1 ± 20.0**	67.8 ± 15.2**	140.9 ± 54.0*
proANP, nM/L	1.1 ± 0.3	7.6 ± 0.9**	6.4 ± 0.8**	6.3 ± 1.5**
Ejection fraction, %	NA	31.9 ± 2.7	38.4 ± 2.7	32.2 ± 3.9
Left atrial dimension, cm	NA	5.1 ± 0.3	5.4 ± 0.2	5.4 ± 0.1
Systolic blood pressure, mm Hg	NA	107.9 ± 4.6	116.9 ± 3.0	114.5 ± 4.6
Diastolic blood pressure, mm Hg	NA	68.0 ± 3.0	72.9 ± 1.8	74.6 ± 3.9

Values are presented as mean ± standard error or as median (min; max); gender characteristics are presented as percentage. *P < 0.05; **P < 0.01 compared with healthy donors; ⁺P < 0.05; ^‡P < 0.01 compared with “HF” group. HF, heart failure; Ob, obesity; DM, diabetes mellitus; NA, not available

Among all clinical variables studied, BMI was most commonly associated with changes in gene expression (Table 2). Since BMI-associated transcripts were not related to the presence of DM (except ITGB6), we categorized all individuals with HF into 2 groups according to obesity status (BMI >30). Univariate analysis of these genes showed significant upregulation of COL1A2, CTGF, ITGB5, MMP2, MMP14, SMAD7, and SNAIL1 in BMSCs of non-obese HF patients comparing to obese with a fold-change >1.45 (P < 0.05), while PLAU showed borderline significance (Table 3). The selected transcripts were not associated with cell senescence quantified by β-galactosidase staining (data not shown). Remarkably, these gene transcript levels did not differ significantly between healthy controls and HF patients with obesity.

Table 2. Demographic and clinical variables associated with BMSC gene expression.

		NYHA	BMI	DM
ANG2	Angiopoietin 2	0.002
COL1A2	Collagen type I, α 2		0.045
COL3A1	Collagen type III, α 1			0.040
CTGF	Connective tissue growth factor		0.035
EDN1	Endothelin 1	0.017
ITGB5	Integrin β 5		0.002
ITGB6	Integrin β 6		0.008	0.022
MMP14	Matrix metalloproteinase 14		0.009
MMP2	Matrix metalloproteinase 2		0.005
PLAU	Urokinase plasminogen activator		<0.001
SMAD7	SMAD family member 7		0.006
SNAIL1	Snail homolog 1		0.043
TGFB1	Transforming growth factor β 1		0.022
TGFBR2	TGF β receptor II		0.007

Significant associations (P < 0.05) are in bold. NYHA, New York Heart Association functional class; BMI, body mass index; DM, diabetes mellitus

Table 3. Differences in BMSC gene expression and BMI.

	HF patients with BMI < 30 (n = 18)			Healthy controls (n = 13)
Symbol	Fold-change			Fold-change
		95% CI	P value		95% CI	P value
COL1A2	1.60	(1.21, 1.99)	<0.001	1.16	(0.90, 1.42)	0.385
CTGF	1.49	(1.05, 1.93)	0.009	0.93	(0.63, 1.22)	0.488
ITGB5	1.55	(1.14, 1.95)	0.001	0.95	(0.70, 1.21)	0.468
MMP2	1.77	(1.33, 2.21)	<0.001	0.91	(0.61, 1.21)	0.682
MMP14	1.54	(1.02, 2.06)	0.032	0.91	(0.48, 1.33)	0.843
PLAU	1.91	(1.01, 2.81)	0.076	1.15	(0.38, 1.91)	0.632
SMAD7	1.47	(1.07, 1.88)	0.008	0.97	(0.65, 1.30)	0.986
SNAIL1	1.74	(0.95, 2.52)	0.022	0.94	(0.45, 1.44)	0.868
TGFB1	1.14	(0.89, 1.40)	0.543	0.97	(0.62, 1.31)	0.921
TGFBR2	1.30	(1.02, 1.58)	0.026	0.88	(0.64, 1.12)	0.379

Fold-change is indicated in comparison to heart failure (HF) patients with body mass index (BMI) > 30 (n = 24).

Discussion

It is assumed that HF encompasses cellular and molecular modifications that are aimed at maintaining heart function, but in the long-term would contribute to HF deterioration.⁵ BMSCs express a wide range of molecules that can influence the microenvironment directly or through paracrine action.¹ BMSC-derived transcripts demonstrating association with BMI in the present study encoded ECM proteins, remodeling enzymes, and molecules regulating ECM deposition. This finding is noteworthy, since the changes in ECM structure and function directly contribute to the adverse myocardial remodeling in HF of various etiology.⁶

ECM proteins and remodeling enzymes

Cardiac ECM is mainly composed of collagens I and III, and changes in collagen content or isoform distribution may affect the properties of myocardium.^5,7 Increased synthesis of collagen type I with its large-diameter fibers, and/or a high degree of cross-linking may increase myocardial stiffness and alter left ventricular function.^6,8 The degradation of collagen fibers is mediated by the family of matrix metalloproteinases (MMP).⁹ The product of PLAU gene, urokinase plasminogen activator (uPA) may be involved in the degradation of myocardial ECM via plasmin generation and activation of pro-MMPs.^10,11 However, MMP2, MMP14, and uPA can play a dual role in tissue remodeling, being responsible for a breakdown of ECM proteins as well as for matrix deposition through proteolytic activation of TGFβ.^6,10,11 So far, elevated protease levels were mostly associated with cardiac fibrosis. Animal models clearly demonstrated a contribution of MMP2, MMP14, and uPA activity to collagen accumulation and adverse myocardial remodeling.^10,12,13 In humans, MMP2 and MMP14 are increased in myocardium and circulation in all types of myocardial remodeling,⁶ and cardiac fibrosis with increased collagen deposition is found in end-stage HF independent of etiology.^8,9,14 It was hypothesized that proteases degrade normal matrix and facilitate the deposition of a more “fibrotic” myocardial ECM.⁹ All these observations suggest that enhanced COL1A2, MMP2, MMP14, and uPA expression may contribute to profibrotic properties of BMSCs isolated from non-obese HF patients.

Regulatory molecules implicated in ECM deposition

While integrin αVβ5 supports non-proteolytic activation of TGFβ via a conformational change,^15,16 the induction of SMAD7 constitutes an inhibitory feedback mechanism to TGFβ signaling.¹⁷ In vivo and in vitro studies revealed that overexpression of Smad7 inhibited collagen deposition induced by angiotensin II and TGFβ1 in cardiac fibroblasts and myofibroblasts.^18,19 In contrast, increased expression of β5 integrin subunit was associated with elevated collagen production in dermal fibroblasts.^20,21 Therefore, increased expression of ITGB5 and SMAD7, which was associated with BMI <30 in our study, may contribute to pro- and antifibrotic properties of BMSCs, respectively.

Another transcript inversely related to BMI in our study was CTGF. This matricellular protein can impact ECM deposition and remodeling either independently or as a downstream mediator of TGFβ, and can also inhibit its antagonist BMP7.^22,23 CTGF is induced in many fibrotic disorders, and its level correlates with the degree and severity of fibrosis.²² In HF, elevated CTGF protein levels were found in myocardial biopsies and circulation.^24,25 It was proposed that CTGF can play a physiological role in early tissue repair and contribute to replacement of normal tissue with scar tissue when its threshold level is reached and sustained.²⁶

CTGF can also affect ECM turnover in the heart through induction of myofibroblast formation from fibroblasts and epithelial cells (via epithelial-to-mesenchymal transition, EMT).²³ EMT is mediated by the activation of a transcriptional factor Snail1.^27,28 It was demonstrated that Snail1 not only can induce EMT, but also controls the transcriptional levels of MMP2 and MMP14.^29,30 These data suggest that increased expression of Snail1 and CTGF in BMSCs from non-obese HF patients may provide autocrine and paracrine mechanisms for enhanced ECM remodeling.

The relation of gene expression to BMI

The association of transcriptional changes with BMI was shown for several cell types. A community-based cohort study demonstrated that for platelets and leukocytes, gene expression was associated with both increased and decreased BMI.³¹ Whole-genome expression profiles of cultured mononuclear progenitor cells identified both downregulated and upregulated genes in obese subjects.³² Obesity itself significantly influenced gene expression in myocardial biopsies.³³ However, the present study is the first to demonstrate the relation between BMI and gene expression in BMSCs.

Conclusions

Our study indicates that BMSCs of HF patients with lower BMI have enhanced expression of genes encoding collagen type I, proteases and protease activators, and regulatory molecules. Of particular interest is that modifications are found not in the myocardium, but in the bone marrow stromal cells. The changes in certain BMSC-derived transcripts (e.g., COL1A2, MMP2, MMP14, CTGF) resemble those described in myocardium of HF patients, suggesting that in HF certain common mechanisms are induced in various organs and tissues. In contrast, the present study did not show significant differences in expression of genes involved in ECM remodeling between BMSCs from obese HF patients and healthy donors.

Several clinical trials were launched to evaluate the safety and efficacy of BMSCs administration in HF (for a review, see ref. 2). Most of these studies perform BMSCs delivery by intramyocardial injections; however, the optimal cell type—autologous or allogenic—has not yet been determined. Our data demonstrate that lower BMI was associated with upregulated expression of proteins that constitute ECM and mediate its turnover, or modulate ECM remodeling through paracrine action. These alterations in gene expression presumably may impact the myocardial microenvironment in case of autologous BMSCs injections.

Comorbidities (including obesity and diabetes) are known to play a role in heart failure progression rate and outcome of the disease. We thus suggest that key molecular targets identified in our study should become the target of the subsequent focused studies. In particular, such studies should investigate whether the observed changes are realized at protein level, and whether BMSCs of obese HF patients have a better capacity to enhance endogenous repair compared with non-obese. We believe that in the future, these targets may find some use in the clinical setting: in particular, in designing cell-based clinical trials and in evaluating the effect of BMSC-based cell therapy.

Materials and Methods

Ethics statement

The research was conducted according to the principles expressed in the Declaration of Helsinki. The samples were collected under agreement of the Ethics Committee at Federal Almazov Medical Research Centre with written informed consent of each patient and donor.

Study population

BMSCs were isolated from bone marrow aspirates of patients enrolled in the Studies Investigating Co-morbidities Aggravating Heart Failure³⁴ and healthy donors for allogenic bone marrow transplantation. A study population of 55 subjects enrolled included 13 healthy donors and 42 HF patients: 14 with isolated HF, 16 with comorbid obesity, and 12 with comorbid DM. The clinical diagnosis of heart failure was evidenced by at least one of the following: (1) left ventricular ejection fraction ≤40%; (2) left atrial dimension >4.0 cm; or (3) N-terminal pro-B-type natriuretic peptide >47.3 pM/L. The diagnosis of DM was verified by fasting or random blood glucose test or oral glucose tolerance test. Patients were categorized as obese according to BMI >30 kg/m2. The detailed characteristics of a study population are provided in Table 1.

BMSCs isolation and characterization

Bone marrow aspirates were obtained from the iliac crest or sternum. BMSCs were isolated and cultured as described previously.³⁵ Briefly, a density gradient was used in the isolation procedure to eliminate unwanted cell types that were present in the marrow aspirate. Isolated bone marrow mononuclear cells were plated in T75 flasks at a density of 3 × 10⁵ cells/cm² for expansion. This initial passage of the primary cell culture was referred to as passage 0. Cells were cultured in α-MEM (PanEco, cat #C180) supplemented with 10% FBS (HyClone, cat #SH30070.03), 1% penicillin-streptomycin (Invitrogen, cat #15140-122), 1% L-glutamine (Invitrogen, cat #25030-024) with medium changes twice a week.

BMSCs were evaluated for gene expression after the passage 2, which corresponded to 15–16 cumulative population doublings.³⁵ At this time point all BMSC samples were positive for stromal cell-associated markers CD105, CD90, CD73, lacked expression of CD34, CD19, CD14, CD45, and were able to differentiate into osteocytes and adipocytes (data not shown).

RNA isolation and real-time PCR analysis

Total RNA was isolated using Aurum Total RNA Mini Kit (Bio-Rad, cat #732-6820). RNA quality was verified using RNA LabChip Kit (Agilent, RNA 6000 Nano) and concentration measured with a NanoDrop ND-1000 spectrophotometer (Thermo Scientific). One microgram of total RNA was converted to cDNA using RevertAid First Strand Synthesis Kit (Thermo Scientific, K1621) and RT2 First Strand Kit (SABiosciences, cat #330401).

The level of specific transcripts was assessed using RT2 Profiler PCR Array Human Fibrosis (SABiosciences, cat #PAHS-120) according to the manufacturer’s protocols. Several genes associated with angiogenesis or inflammation that were absent in the array were quantified using TaqMan Gene Expression Assays (Applied Biosystems, cat #4331182). All PCR reactions were performed using 7500 Real-Time PCR System (Applied Biosystems).

Statistical analysis

Continuous variables are displayed as mean ± standard error, order variables are shown as median (min, max), and categorical variables are displayed as percentages. Clinical characteristics of study groups were analyzed by Student t test. Differences in gender among groups were compared using Pearson Chi-square test.

To study the relation of clinical characteristics to gene expression a GLM procedure was used. Predictor variables included DM status as a factor, and BMI, age, and NYHA functional class as covariates. Gene expression data (dependent variables) were entered as dCt values. A separate GLM was run for each gene transcript.

Statistical analysis was performed using Statistica 6 software (StatSoft); P values < 0.05 were considered statistically significant.

Glossary

Abbreviations:

BMI

body mass index

BMSCs

bone marrow stromal cells

DM

diabetes mellitus

GLM

general linear models

ECM

extracellular matrix

HF

heart failure

NYHA

New York Heart Association