Cross-Sectional Correlates of Serum Heat Shock Protein 70 in the Community

Abstract

Background:

Recent studies of referral samples suggest that heat shock proteins play a key role in the pathogenesis of high BP and cardiovascular diseases (CVD) including heart failure. It is unclear whether circulating heat shock protein 70 (HSP70) levels are related to CVD risk factors, echocardiographic indexes of left ventricular (LV) remodeling, and prevalent CVD in the population.

Methods:

We evaluated the cross-sectional relations of serum HSP70 to established CVD risk factors (including hypertension), markers of oxidative stress (urinary 8-epi-PGF_2α) and inflammation (plasma interleukin-6, C-reactive protein, monocyte chemoattractant protein-1 MCP-1, and soluble intercellular adhesion molecule sICAM-1), echocardiographic LV dimensions and prevalent CVD in 456 Framingham Offspring Study participants (mean age 61 years, 42% women).

Results:

In multivariable analyses, serum HSP70 was not associated with age, sex, vascular risk factors (including hypertension), echocardiographic LV mass or prevalent CVD. Also, serum HSP70 was not related to any of the biomarkers evaluated (p≥0.10 for all).

Conclusions:

In our community-based sample, serum HSP70 was similar in men and women, and not significantly related to traditional or novel risk factors, to LV mass or to prevalent CVD. Our data suggest that blood levels may not adequately reflect the important role of heat shock proteins in prevalent CVD.

Heat-shock proteins 70, cardiovascular disease, inflammation, epidemiology, left ventricle

Heat shock proteins (HSP) are constitutively expressed as intracellular cytoprotective chaperones that can increase several-fold within cells in response to stresses that induce protein unfolding or aggregation.¹ More recently it has been recognized that HSP are present in, or can be released into, the extracellular environment in response to tissue damage or under physiologic conditions.² Heat shock protein 70 (HSP70), a key member of this family of proteins, has been studied in relation to many clinical conditions including infections³ and cancers.⁴ Circulating HSP70 serves as an intercellular signaling molecule that mediates a range of immunoregulatory responses via interaction with cell surface receptors to modulate proinflammatory processes.⁵ The relations of HSP70 to inflammation are complex because HSP70 can activate CD14/toll-like receptor 4 and increase cytokine production, which in turn can further induce HSP70 synthesis.^6,7

Lately researchers have investigated the putative role of HSP in the pathogenesis of cardiovascular diseases (CVD).⁸ Such research has focused largely on intracellular HSP70. Experimental evidence suggests that HSP70 is expressed in atherosclerotic plaques and plays a key role in promoting survival of arterial smooth muscle cells.⁹ Vascular HSP expression increases in response to endothelial injury, oxidized LDL accumulation, and rising BP (BP).¹⁰ Expression of HSP70 also increases in vascular and cardiac myocytes in response to stretch¹¹ and elevation of BP.¹²

These experimental observations have motivated investigators to evaluate the importance of extracellular HSP70 by relating circulating HSP70 to high BP¹³ and CVD.¹⁴ Pockley et al reported that serum HSP70 levels in hypertensive patients did not differ from those in nonhypertensive control subjects.¹³ Cross-sectional investigations relating HSP70 to prevalence of CVD have yielded inconsistent results. Some investigators have reported that increased serum HSP70 levels are associated with a higher prevalence of vascular disease¹⁵ and cardiac failure.¹⁶ However, Zhu et al reported inverse relations of serum HSP70 and prevalent CVD.¹⁴ In longitudinal studies, Pockley et al have shown that lower serum HSP70 levels in hypertensives predict the progression of carotid intimal medial thickness.¹⁷ On a parallel note, investigations relating circulating HSP70 to inflammatory markers have yielded conflicting results, with reports of positive,³ null,¹⁶ and inverse associations.¹⁸

In summary the biological and clinical significance of circulating HSP70 levels is not clear. In the present investigation, we studied the cross-sectional relations of serum HSP70 to traditional CVD risk factors, inflammatory and oxidative stress markers, echocardiographic left ventricular (LV) structure and function, and prevalent CVD in a community-based sample. We hypothesized that serum HSP70 would be positively related to CVD risk factors including inflammatory and oxidative stress biomarkers, to subclinical disease as assessed by left ventricular (LV) mass, and to prevalent CVD.

Methods

The design and selection criteria of the Framingham Offspring Study have been previously described.¹⁹ Participants who attended the seventh examination cycle (1998 to 2001) were eligible for this investigation. Serum HSP70 levels were measured in a random sample of 456 participants (194 women) that was representative of the entire sample of attendees (data not shown). We chose a subsample to maximize judicious use of limited serologic resources given the novelty of the marker (HSP70) and prior data indicating very low circulating levels in healthy individuals.¹⁴ All attendees underwent routine history and physical examination (including BP measurement and anthropometry) and laboratory assessment of vascular risk factors. Echocardiographic data for these participants were obtained from the sixth examination cycle (about 4 years previously). Cardiovascular disease was determined by a panel of three investigators as described elsewhere.²⁰ All participants gave informed consent, and the study was approved by the Boston Medical Center Institutional Review.

Biochemical Assays

Fasting blood samples drawn from the participants were stored at −70°C without freeze–thaw cycles until assay. Serum HSP70 levels were measured by a standard enzyme-linked immunoassay kit (Stressgen Biotechnologies, Victoria, British Columbia, Canada). The assay has a range of 0.78 to 50 ng/mL, with a sensitivity of ≤0.54 ng/mL. The average intra-assay coefficient of variation (CV) for HSP70 in our sample was 9.5%. We chose a panel of inflammatory markers that were previously measured at the same examination cycle of Framingham Offspring Study as HSP70. This panel included biomarkers representing multiple inflammatory pathways, ie, cytokines such as interleukin-6 (IL-6), chemotactic factors monocyte chemoattractant protein-1 (MCP-1), adhesion molecules soluble intercellular adhesion molecule (sICAM-1), and C-reactive protein (CRP). Inflammatory markers (IL-6, MCP-1, and sICAM-1 levels) were measured in duplicate using enzyme-linked immunosorbent assays (R&D Systems, Minneapolis, MN). Reproducibility of measurements was excellent (mean intra-assay CV for MCP-1, 1.9%; IL-6, 3.1%; sICAM-1, 3.7%). The CRP was measured once using a high-sensitivity assay (Dade Behring BN100 nephelometer, Deerfield, IL) with a mean inter-assay CV 3.2% on 139 phantom replicates. Levels of urinary 8-epi-PGF_2α indexed to urinary creatinine (referred to as urinary isoprostanes) were measured by using a commercially available enzyme-linked immunoassay kit (Cayman, Ann Arbor, MI) with an intra-assay CV of 9.7%, as previously described.²¹

Echocardiographic Variables

The LV measurements were obtained by averaging digital M-mode measurements in three cardiac cycles using a leading edge technique.²² The LV mass was calculated using a standard formula.²³ The reproducibility of echocardiographic measurements has been previously described.²⁴

Statistical Analyses

The HSP70 was modeled as a continuous variable, normalized by using the square root of value (referred to as HSP70 for simplification) and in quartiles. All markers (including CRP, IL-6, MCP-1, sICAM-1, and urinary isoprostanes) were modeled as continuous variables with natural logarithmic transformation. Three sets of analyses were performed, as described below.

Multiple linear regression²⁵ models were used to evaluate the relations of clinical variables (age, sex, vascular risk factors, 3-hydroxy-3-methylglutaryl coenzyme-A reductase inhibitors [statin] use, aspirin use), the echocardiographic measurements and prevalent CVD to serum HSP 70 levels (the latter as the dependent variable).

The relations of serum HSP 70 to biomarkers of interest was evaluated by estimating partial (age- and sex-adjusted) Pearson correlations.

We assessed whether serum HSP 70 levels (independent variable) were associated with prevalent CVD or hypertension (the latter two as dependent variables in separate models) using multiple logistic regression. We modeled HSP70 as quartiles and compared the adjusted prevalence of CVD and hypertension in the top quartile relative to the lowest quartile. We analyzed the quartile-based cut-points because prior investigators have reported a reduced prevalence of cardiovascular disease in individuals with HSP70 above the 75th percentile value.¹⁴

All analyses were done using the REG, CORR and LOGISTIC procedures in the Statistical Analysis System software (SAS Institute, Cary, NC).²⁶ A two-sided P value < .05 was considered to be statistically significant.

Results

The characteristics of our sample are summarized in Table 1. There were 57 participants (12.5%) with prevalent CVD. Compared with women, men were heavier and had higher BP levels and ratios of total to HDL cholesterol. Men also had a higher prevalence of myocardial infarction and CVD and were more often receiving treatment with aspirin or statins or both. There were no significant differences in levels of inflammatory markers for men versus the women in our sample.

Table 1

Baseline characteristics of study sample

Characteristic	Men (n = 262)	Women (n = 194)	P value*
Age (y)	62 (10)	61 (9)	.75
Body mass index (kg/m2)	29.1 (4.7)	27.0 (5.2)	<.0001
Total:HDL cholesterol (ratio)	4.6 (1.3)	3.7 (1.2)	<.0001
Systolic blood pressure (mm Hg)	127 (17)	123 (17)	.04
Diastolic blood pressure (mmHg)	75 (9)	72 (9)	<.0001
Hypertension (%)	44	38	.22
Hypertension Rx (%)	33	31	.61
Diabetes (%)	16.0	15.5	.87
Smoking (%)	8.4	11.3	.29
Aspirin use (%)	40.1	27.3	.005
Statin use (%)	19.5	11.9	.03
Myocardial infarction (%)	7.7	1.03	.001
Congestive heart failure (%)	1.2	1.03	.91
Cardiovascular disease (%)	16.4	7.2	.003
Serum heat shock protein 70 (ng/mL)
All subjects, median, (25th, 75th percentile)	1.76 (1.05, 2.94)	1.59 (0.92, 2.66)	.37
Subjects 30–49 years, median, (25th, 75th percentile) n = 28/12	1.32 (0.78, 3.22)	1.59 (0.74, 2.97)	.72
Subjects 50–59 years, median, (25th, 75th percentile) n = 86/77	1.47 (0.87, 3.23)	1.57 (0.97, 2.43)	.35
Subjects 60–69 years, median, (25th, 75th percentile) n = 82/62	1.84 (1.18, 2.79)	1.43 (0.79, 2.56)	.45
Subjects ≥ 70 years (25th, 75th percentile) n = 66/43	1.98 (1.23, 2.70)	2.02 (1.09, 2.94)	.69
Inflammatory markers
C-reactive protein, mg/L, median, (25th, 75th percentile)	2.1 (1.2, 5.3)	2.2 (1.0, 5.5)	.60
Interleukin-6 (pg/mL), median (25th, 75th percentile)	3.0 (1.9, 4.5)	2.7 (1.7, 4.0)	.10
MCP-1 (pg/mL), median (25th, 75th percentile)	301 (239, 393)	302 (243, 375)	.53
Urinary isoprostanes (ng/mmol), median (25th, 75th percentile)	114 (87, 175)	132 (83, 191)	.07
sICAM-1 (ng/mL), median (25th, 75th percentile)	245 (212, 284)	243 (210, 284)	.74
Echocardiographic variables
LV mass (g)	192 (42)	140 (31)	<.0001
LV diastolic dimension (cm)	5.1 (0.5)	4.6 (0.4)	<.0001
LV wall thickness (cm)	2.0 (0.2)	1.8 (0.2)	<.0001
Fractional shortening	0.4 (0.1)	0.4 (0.1)	<.0001

CRP = C-reactive protein; IL-6 = interleukin-6; LV = left ventricular; MCP-1 = monocyte chemoattractant protein–1; sICAM-1 is soluble intercellular adhesion molecule–1; urinary isoprostanes = 8-epi-PGF_2α indexed to urinary creatinine.

Values are means unless indicated and values in parentheses are standard deviations.

*

P values are for comparison between men and women.

Serum HSP70 levels were positively skewed, with a median value of 1.69 ng/mL and an interquartile range of 0.98 to 2.84 ng/mL. The HSP70 levels were similar in the two sexes (Table 1). In regression models, HSP70 was not associated with any of the CVD risk factors including age, sex, hypertension, and diabetes mellitus (P > .20). In addition HSP70 was not related to prevalent CVD (P > .20). In both age- and sex-adjusted models and multivariable models, HSP70 was not related to statin or aspirin use (P > .20 for all models). After adjustment for age, sex, and other vascular risk factors, HSP70 levels were not related to any of the measured inflammatory markers or to urinary isoprostanes. All age- and sex-adjusted correlation coefficients ranged between 0.03 (for log sICAM-1) to 0.08 (for log CRP and log IL-6; all P values ≥ 0.10). On further analyses HSP70 was not associated with LV mass, LV diastolic dimensions, LV wall thickness, or fractional shortening (all P values > .20).

The prevalence of CVD and hypertension in the top quartile of serum HSP70 were not different from that in the lowest quartile (multivariable-adjusted odds ratio: for prevalence of CVD, 1.71, 95% CI 0.63 to 4.67, P = .30; for hypertension, 1.14, 95% CI 0.59 to 2.21, P = .70).

Because other investigators have studied associations between serum HSP70 and hypertension¹³ and CVD,¹⁴ we estimated the statistical power of our sample. In our sample we had >80% power (at α = 0.05) to detect very small increments in R² (of 0.017 (1.7%) for HSP70 in linear regression models incorporating explanatory variables such as vascular risk factors. Furthermore, in logistic regression models we had >80% power to detect adjusted-odds ratios of 1.56 for CVD and 1.35 for hypertension per 1-SD increment in HSP70.

Discussion

In our community-based sample, levels of serum HSP70 were similar in men and women, unlike a previous report² in which women were reported to have higher levels compared with men. We found no significant relation between serum HSP70 and traditional or novel CVD risk factors (including hypertension), echocardiographic LV dimensions or prevalent CVD. We had adequate statistical power to detect very modest associations of prevalent CVD and hypertension with circulating HSP70 levels.

There are several reasons why our findings differ from those of prior reports. First, our assay results yielded levels of HSP70 that are considerably lower than those reported by some investigators. However our values correspond closely with those reported by others using an assay similar to the one that we used.¹⁴ Second, our sample was community-based and the clinical correlates in community-dwelling individuals may differ from those observed in referral samples that may have a greater proportion of diseased individuals.¹⁷ Third, it is important to note that a lack of correlation of serum levels of HSP70 with risk factors or subclinical disease measures does not negate a role for HSP70 in vascular disease. It is conceivable that serum levels do not adequately reflect tissue levels or tissue responses. In addition HSP70 responses to stress may be transient and may not be adequately reflected in steady-state circulating levels.

The use of a community-based sample well characterized for established and newer risk factors, subclinical disease, and clinical CVD, strengthens our investigation. Nonetheless there are several limitations of our study that merit comment. We did not measure serum levels of other heat shock proteins besides HSP70 (such as HSP 65, 60, or 20), nor did we measure anti-HSP70 levels. We chose to measure HSP70 in our sample because it has been studied in relation to atherosclerosis and inflammation, and data relating this marker to protection against ischemic injury⁸ are stronger than for other HSP types. We did not assess associations of HSP70 levels with antibodies against microbial antigens (such as Chlamydia pneumonia titers²⁷) because antimicrobial antibody levels were not obtained at the seventh examination cycle of the offspring cohort. Also our sample was predominantly middle-aged to elderly and of white ethnicity; the generalizability of our results to younger individuals and other ethnic groups remains untested. Furthermore because of the modest sample size our ability to assess effect modification by age and sex was limited. Although our study suggests that serum HSP70 levels may have limited diagnostic value in community-based samples, we cannot exclude the possibility that HSP70 levels may have value in samples with a higher burden of CVD. Because our study was cross-sectional, we cannot evaluate the relations of HSP70 to long-term risk of CVD morbidity and mortality.

In conclusion, in our community-based sample of individuals, serum HSP70 was not associated with known or novel CVD risk factors, LV dimensions and systolic function, or prevalent CVD. Our findings suggest correlates of circulating HSP70 in the general population may differ from that reported in high-risk referral samples.