HSP25 Vaccination Attenuates Atherogenesis via Upregulation of LDLR Expression, Lowering of PCSK9 Levels and Curbing of Inflammation

Abstract

Objective:

Elevated Heat Shock Protein 27 levels predict relative freedom from cardiovascular events. Over-expression or twice daily subcutaneous injections of human HSP27 in ApoE^−/− mice reduces blood and plaque cholesterol levels, as well as inflammation and atherosclerotic plaque burden. Antibodies to HSP27 are present in human blood and the purpose of the current studies is to explore their role.

Approach and Results:

Blood levels of both HSP27 and anti-HSP27 IgG antibodies are elevated in healthy controls compared to patients with cardiovascular disease. ApoE^−/− mice fed a high fat diet and vaccinated with recombinant HSP25 (rHSP25, murine ortholog) show increased levels of anti-HSP25 IgG antibodies and reductions in plasma cholesterol and atherogenesis. Moreover, rHSP25 vaccination markedly lowered serum amyloid A levels as well as hepatic macrophage abundance and inflammatory cytokine expression. The effects of the HPS25 vaccination on cholesterol metabolism are divergent: increased hepatic LDLR mRNA and protein expression and reduced plasma PCSK9 levels - despite no effect on PCSK9 expression. In vitro, the HSP27 immune complex (IC) upregulates hepatocyte LDLR mRNA and protein expression independent of intracellular cholesterol levels and increases LDLR promoter activity. The increase in LDLR expression by the HSP27 IC is dependent upon activation of the NF-κB pathway. Hepatocyte PCSK9 protein levels are reduced after HSP27 IC treatment in vitro despite only minor transient effects on gene expression.

Conclusion:

HSP27 immunotherapy represents a novel means of lowering cholesterol and PCSK9 levels, primarily due to augmentation of LDLR expression and is associated with marked reductions in inflammation.

Graphical Abstract

Introduction

Small heat shock proteins, such as Heat Shock Protein 27 (HSP27), are intracellular chaperones that promote the proper reassembly of misfolded proteins and act as mediators of extracellular cellular signaling.¹ HSP27 effectively preserves cellular homeostasis under various conditions of degenerative or inflammatory stress – including those common to the pathogenesis of atherosclerosis.² Experiments from our group³ and four others using proteomic discovery approaches,^4–7 show that serum HSP27 levels decline as human atherosclerosis develops with its tissue abundance inversely corelated with the degree of coronary artery plaque burden. In atherosclerosis-prone Apolipoprotein E null (ApoE^−/−) mice augmentation of extracellular human HSP27 levels via constitutive over-expression, transplantation of bone marrow from mice that over-express HSP27, twice-daily subcutaneous administration of high dose recombinant HSP27 (rHSP27; 100 μg) or estrogenic therapy post-ovariectomy (that augment HSP27 blood levels) reduce both plasma and plaque cholesterol content, resulting in the formation of more stable plaques that are less inflamed.^8–11 Clinically, elevated HSP27 blood levels are associated with a lower 5-year risk of myocardial infarction, stroke or cardiovascular death.⁹ Interestingly, natural antibodies to HSP27 (AAbs) are detectable in the blood, yet their biological significance is unclear.^12,13

In this study we sought to address three questions. First, what is the correlation between blood HSP27 and AAb abundance in human cardiovascular disease (CVD) patients compared to healthy control subjects (CON)? Second, does augmenting levels of antibodies to HSP25 (the murine ortholog of human HSP27) via vaccination ameliorate the biological effects of rHSP25 and represent a more efficient means of attenuating atherogenesis in ApoE^−/− mice? Third, what are the potential mechanisms by which the HSP27 immune complex (IC) alters the two key drivers of atherosclerosis: cholesterol and inflammation? Our results indicate that blood levels of HSP27 and AAbs are higher in health compared to CVD. Weekly rHSP25 vaccination boosts anti-HSP25 antibodies and is associated with reductions in plasma cholesterol as well as proprotein convertase subtilisin / kexin type 9 (PCSK9) levels,¹⁴ a key negative regulator of low density lipoprotein receptor (LDLR) recycling. These anti-atherogenesis effects are associated with reduced plaque, circulating and hepatic biomarkers of inflammation. Finally, the biological effects of the HSP27 IC are novel, as it markedly upregulates LDLR expression independent of intracellular cholesterol levels yet is reliant on activation of the NF-κB pathway. Taken together, the discovery of HSP27-mediated upregulation of LDLR heralds a unique opportunity to develop HSP27 immuno-therapeutics for the treatment of atherosclerosis and hypercholesterolemia.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request. Please refer to the on-line Supplementary Information document for full methodological details. Briefly, this manuscript incorporates the results of three inter-related studies of: i) HSP27 and AAbs blood levels in healthy control and cardiovascular disease subjects, ii) a murine model of atherogenesis that tests the efficacy of rHSP25 vaccination therapy to reduce plaque formation and lower cholesterol levels, and iii) in vitro studies that dissect out the molecular mechanisms by which the HSP27 IC upregulates the LDLR.

Study Approval

As part of a National Institute of Health sponsored project (#5P20MD002314–08; Sub-Project ID: 6374) designed to identify cardiovascular biomarkers for ischemic heart disease in a population that is medically underserved, subjects were enrolled in a 5-year follow-up study that involved the collection of blood samples at various University of South Alabama-affiliated medical clinics. The study was approved by the University’s Institutional Review Board on June 21, 2012 and complied with the Declaration of Helsinki. Written informed consent was obtained from participants prior to inclusion in the study. Similarly, all experimental murine procedures were approved by the Animal Care Committee of the University of Calgary (protocol: AC17–0015).

Statistics

a). Statistical analyses for Clinical Study

For the clinical data, continuous variables are reported as mean ± standard deviation, with potential deviations from normality described as medians with interquartile ranges (25th and 75th percentiles). Categorical variables are reported as number (%) and were compared using χ2 or Fisher’s exact tests. Chi-square tests were performed on nominal data sets: male sex, smoker, diabetes mellitus (defined by use of hypoglycemic agents), hypertension (defined by of conventional anti-hypertensive therapy), statin use (defined by use of any HMG Co-A reductase medication), African American and Caucasian. All comparisons used two-tailed alpha levels of 0.05. In cases of multiple comparisons, one-way analysis of variance (ANOVA) and Tukey’s method were used to correct for the family-wise error rate. The association of health vs. CVD status with blood concentrations of HSP27 and AAb was examined using multivariable linear regression, adjusting for age, sex, ethnicity, body mass index (BMI), diabetes mellitus, and active smoking status. Body mass index was log transformed in the multivariable regression model. Adjusted differential levels in HSP27 and AAb are reported with 95% CIs. All analyses were performed in a blinded fashion using GraphPad Prism 8 (GraphPad Software, La Jolla, CA), except for regression models, which were fit using SAS v9.4 (SAS Institute Inc., Cary, NC).

b). Statistical Analyses for the non-clinical data

Continuous variables are expressed as mean ± standard deviation and the comparison of unpaired treatment groups was performed using a Student’s t-test. Quantitative PCR gene expression data were log-transformed and are reported as log-fold change in order to more normally distribute the data and permit parametric statistical analyses. All analyses were performed in a blinded fashion using GraphPad Prism 8 (GraphPad Software, La Jolla, CA). A two-tailed α level of 0.05 was used to define statistical significance. As indicated, the number of mice per experimental group was typically 8–10, with lesser numbers in some auxiliary experiments (e.g., n=5–6 for the LDLR^−/− mice). All in vitro experiments were typically performed thrice, with representative data sets presented in the manuscript.

Results

Anti-HSP27 IgG antibodies are more abundant in health vs. CVD

To assess HSP27 and AAb levels, blood samples were obtained from 80 CVD patients and 58 CON subjects originally recruited to a National Institutes of Health (NIH) sponsored study designed to look for novel cardiovascular biomarkers in a medically under-serviced population (Supplemental Tables I, II). Similar to our previous report,⁹ CVD was associated with a lower HSP27 blood concentration (median [interquartile range]) compared to CON (538.3 [420.0–653.5] vs. 689.4 [529.8–900.0] pg/mL, p<0.0001; Fig. 1A). This difference persisted after adjusting for age, sex, race, BMI, diabetes mellitus, and active smoking status (adjusted difference: −117.5 pg/mL, 95% confidence interval (CI) −221.3 to −13.7, p=0.027). Diabetes mellitus was associated with lower HSP27 levels in this model (adjusted difference: −127.2 pg/mL, 95% CI −232.1 to −22.2; p=0.018) whereas the remaining parameters were not significant, including sex (p=0.095) and ethnicity (p=0.744 and 0.666 for African American and Other, respectively).

Fig. 1.

Serum HSP27 and anti-HSP27 autoantibody (AAb) levels in healthy human controls (CON) and cardiovascular disease (CVD) patients (median [interquartile range]) compared using non-parametric t-tests.

A) After adjusting for age, sex, race, BMI, diabetes mellitus, and active smoking status, serum HSP27 levels are lower in CVD vs. CON subjects, (adjusted difference: −117.5 pg/mL, 95% CI: −221.3 to −13.7, p=0.027).

B) CVD is associated with lower IgG AAb levels (28.2 [14.4 – 38.7] vs. 44.6 [32.0 – 68.0] a.u.; adjusted difference: −21.7 a.u., 95% CI −33.4 to −10.0, p<0.0001).

C) The ratio of the IgG AAb to HSP27 is higher in CON vs. CVD patients (0.071 [0.044–0.113] vs. 0.043 [0.026–0.082], p=0.001).

D) Components of HSP27 IC detected in pooled human blood samples. Ultracentrifugation of blood was used to obtain a pellet that contained microvesicles that were resolved using SDS-PAGE followed by Western Blotting. Bands show immunodetection of HSP27 and human IgG from both health controls (C1–3) and CVD patients (P1–3), with 6 subjects per group.

E) Photomicrographs of the Duolink® Proximity Ligation Assay (PLA), a highly specific and sensitive means of demonstrating in situ the formation of a HSP27 IC. A red fluorescent signal is generated when rHSP27 and the PAb co-localize within 40 nm of each other. In the top left panel, the red fluorescent signal documents the presence of HSP27 IC in hepatocytes treated with rHSP27 and PAb (200× magnification). All other panels show control experiments and the absence of a specific interaction signal.

A new insight is that CVD is associated with lower IgG AAb levels (28.2 [14.4–38.7] vs. 44.6 [32.0–68.0] absorbance units [a.u.]: crude analysis: p<0.0001, Fig. 1B; adjusted difference: −21.7 a.u., 95% CI −33.4 to −10.0, p<0.001). African American ethnicity was associated with higher IgG AAb levels relative to Caucasians (adjusted difference: +15.4 a.u., 95% CI 5.3 to 25.5; p=0.003). No other parameters were associated with significant differences in IgG AAbs, including sex (p=0.062). The association of CVD with IgM AAb levels was not significant after adjusting for prespecified clinical variables (crude difference: 0.117 [0.090–0.209] vs. 0.161 [0.113–0.274] a.u., p=0.016; adjusted difference: −0.003 a.u., 95% CI −0.059 to 0.053; p=0.910). The ratio of IgG AAb to HSP27 was higher in CON vs. CVD patients (0.071 [0.044–0.113] vs. 0.043 [0.026–0.082], p=0.001) (Fig. 1C). One individual with CVD had an IgG AAb:HSP27 ratio >5.0 standard deviations above the CVD group mean. Excluding this outlier identified a weakly positive association between IgG AAbs and HSP27 (r=0.174, p=0.042) but did not appreciably change the above results.

Finally, to demonstrate the presence of the HSP27 IC we did the following. First, in pooled human blood samples from CON vs. CVD subjects we document the joint presence of IgG immunoglobulins and HSP27 (Fig. 1D) using Western blotting. Second, using the Duolink Proximity Ligation Assay, we graphically illustrate the interaction between HSP27 and PAb to form ICs in hepatocytes (Fig. 1E).

rHSP25 Treatment Attenuates Atherogenesis in ApoE^−/− mice

With our clinical data showing elevated levels of both HSP27 and AAbs in health vs. CVD, we then tested the hypothesis that enhancing serum levels of these AAbs might be atheroprotective. Atherosclerosis-prone ApoE^−/− mice were started on a high fat diet (HFD) for 2 weeks before receiving four weekly subcutaneous injections of 100 μg of recombinant HSP25 (rHSP25) mixed with the adjuvant Alum (2% aluminum hydroxide; 3:1 v/v, Fig. 2A). Control mice were treated with rC1, the recombinant C-terminal of HSP27 that is biologically inactive (100 μg mixed with Alum). The use of rC1 is particularly relevant as a control, because rC1 and rHSP25 are both generated in E. coli. While the endotoxin contamination concentrations are lower than 2 units/mg of each recombinant protein, any potential confounding endotoxin effects are counterbalanced by comparing the active and control treatment groups as the recombinant proteins are generated in an identical manner. Separately, we validated rC1 as a control treatment, noting that [PBS + Alum] and [rC1 + Alum] were equally ineffective in attenuating atherogenesis and lowering cholesterol levels (Supplemental Fig. IA, IB), even though vaccination with [rC1 + Alum] was similar to [rHSP25 + Alum] in generating AAbs, while [PBS + Alum] was not (Fig. 2B).

Fig. 2.

rHSP25 immunotherapy strategy attenuates atherosclerosis in ApoE^−/− mice.

A) Schematic representation of rHSP25 (mouse ortholog of HSP27) and the biologically inactive truncated C-terminal of HSP27 (rC1 control) proteins. The black box denotes the alpha beta-crystallin domain that is important for protein oligomerization. The IXI box at the C-terminus is a flexible domain involved in the formation of multiple inter-subunit interactions. The timeline for the murine treatment experiments is shown below, highlighting the duration of the high fat diet, and the weekly subcutaneous injections of rHSP25 or rC1 with the adjuvant alum.

B) Time course for the in vivo generation of IgG antibodies in ApoE^−/− mice after weekly treatment with rHSP25 and rC1 but not PBS. (a.u. = arbitrary units for OD @ 450 nm).

C) and D) Aortic en face lesion area visualized with Oil Red O denoting neutral lipid deposits. The aortic lesion area was reduced in male and female mice vaccinated with rHSP25 vs. rC1.

E) and F) Aortic sinus cross-sections stained with hematoxylin and eosin (H&E). Atherosclerotic lesion areas were reduced in male and female mice vaccinated with rHSP25 vs. rC1. Scale bar = 0.5 mm.

G) and H) rHSP25 vaccination reduced the plaque cholesterol cleft content in male and female mice compared to rC1 control treatment. Scale bar = 0.2 mm.

I) and J) Mouse plaque MΦ content was reduced by rHSP25 vaccination in male mice but was not significantly reduced in female mice. The brown color reaction product identifies MΦ immunolabeled with an anti-Mac-2 antibody. Scale bar = 0.2 mm for larger images, and 0.5 mm for insert images.

K) Terminal plasma levels show reduced serum amyloid A (SAA) levels, a biomarker of inflammation in rHSP25 vs. rC1 vaccinated mice.

Atherosclerosis, as reflected by the en face aortic area of oil red O staining (indicative of intracellular lipid deposits), was reduced by 27% with rHSP25 treatment in male (15.2% ± 2.9% vs. 11.1% ± 2.5%, n=10/group; p=0.003) and 32% in female mice (13.3% ± 2.8% vs. 9.0% ± 1.3%, n=8/group; p=0.001; Fig. 2C, 2D). Similarly, the amount of atherosclerotic plaque quantified from aortic sinus cross-sections (Fig. 2E, 2F) was reduced 39% in male (1.17 ± 0.27 vs. 0.71 ± 0.15, n=9/group; p=0.0004) and 32% in female mice (1.34 ± 0.36 vs. 0.91 ± 0.23, n=8/group; p=0.013). These changes in plaque burden were accompanied by alterations in the content of the plaques. The cholesterol cleft area of the plaques (Fig. 2G, 2H) was reduced by 60% in male (0.10 ± 0.04 vs. 0.04 ± 0.02, n= 9/group; p=0.0008) and 69% in female mice (0.13 ± 0.06 vs. 0.04 ± 0.04, n=8/group; p=0.0027). However, there were sex differences with respect to the reduction in plaque macrophage (MΦ) content in response to treatment (Fig. 2I, 2J). rHSP25 treatment reduced plaque MΦ content by 36% in male (0.47 ± 0.08 vs. 0.30 ± 0.08, p=9/group; p=0.0003) but not in female mice (a 11% reduction was not significant: 0.38 ± 0.12 vs. 0.34 ± 0.10, n=8/group; p=0.68). Overall, there was a strong relationship between lesion area and either cholesterol cleft area (r²=0.626; p<0.001) or MΦ area (r²=0.274; p=0.002).

rHSP25 Vaccination Reduces Inflammation

To examine the effect of rHSP25 vaccination on systemic inflammation, circulating levels of serum amyloid A (SAA), a major acute phase protein that is commonly used in murine atherosclerosis studies, were measured.^15,16 Compared to rC1 vaccinated mice, there were marked reductions in SAA plasma levels with rHSP25 vaccination (males: 10.8 ± 5.4 vs. 3.1 ± 2.8; −71%, p=0.004; females: 13.1 ± 4.9 vs. 1.5 ± 0.6; −89%, p=0.0002). In that hepatic inflammation may reflect vessel wall inflammation¹⁷ we also assessed hepatic lipid-laden MΦ content by immunolabeling with an anti-Mac-2 antibody and performing concomitant oil red O staining. Compared to rC1, rHSP25 vaccination dramatically decreased hepatic MΦ content in male (−70%, 10.6 ± 1.3 vs. 3.2 ± 1.3, n=10/group; p<0.0001) and female mice (−60%, 11.2 ± 2.2 vs. 4.5 ± 0.7, n=6/group; p<0.0001) (Fig. 3A, 3B). In addition, the expression of inflammatory MΦ markers, as well as cytokines (as determined using qPCR) was lower in liver tissue from mice treated with rHSP25 vs. rC1 (n=3 F + n=3 M, per group). For male mice (Fig. 3C) the changes were as follows: CD68 (−63%, p=0.006), IL-1β (−15%, p=0.06), MCP1 (−41%, p=0.02) and TNF1α (−55%, p=0.003), with no change in IL-10 expression (p=0.59). Similarly, for female mice (Fig. 3D) the changes were as follows: CD68 (−48%, p=0.008), IL-1β (−57%, p=0.004), MCP1 (−55%, p=0.01) and TNF1α (−52%, p=0.005), with no change in IL-10 expression (p=0.98).

Fig. 3.

rHSP25 Vaccination reduces Hepatic Inflammation

A) and B) Vaccination with rHSP25 vs. rC1 resulted in a marked reduction in hepatic MΦ content (male: −70%, female: −60%; p<0.0001 for both). Mac-2 immunolabeling represented by brown color, Mac-2 immunofluorescence (IF) yields a green color, the lipid deposits are red due to the oil red O staining and the nuclei are blue as a result of the Hoechst stain. Scale bar = 10 μm.

C) and D) Hepatic tissue from ApoE^−/− mice vaccinated with rHSP25 (vs. rC1) showed reduced expression of inflammatory markers / cytokines: CD68, IL-1β, MCP1 and TNF1α, with no change in IL-10 expression in this sex-disaggregated analysis.

rHSP25 Immunotherapy Reduces Plasma Cholesterol and PCSK9 Levels and Markedly Increases LDLR Expression

Total plasma cholesterol levels were distinctly reduced in mice treated with rHSP25 relative to rC1: −59% in male (1,355 ± 356 vs. 552 ± 151 mg/dl, n=10 per group; p<0.0001) and −57% in female mice (1,086 ± 322 vs. 466 ± 102 mg/dl, n=8 per group; p=0.0001) (Fig. 4A, Supplemental Fig. IIA & IIB). Of note, mice maintained on the HFD for an additional 5 weeks after the last rHSP25 injection continued to show a persistent reduction (−54%) in plasma cholesterol levels compared to rC1-treated control mice (Supplemental Fig. IIE). Interestingly, rHSP25 vaccination of LDLR^−/− mice was ineffective in altering total plasma cholesterol (Supplemental Fig. IIF) or aortic atherosclerotic burden – suggesting that the LDLR is necessary for HSP25-mediated cholesterol lowering. Similarly, despite a robust antibody response, rHSP25 vaccination did not reduce plasma cholesterol levels or aortic atherogenesis in ApoE^−/− mice that had HSP25 gene (Hspb1) knocked out (ApoE^−/−Hspb1^−/−) (Supplemental Fig. IIG).¹⁸ Therefore, it appears that at least basal levels of endogenous HSP25 are required for AAbs to lower cholesterol levels.

Fig. 4.

rHSP25 vaccination of ApoE^−/− mice lowers plasma cholesterol and PCSK9 levels.

A) At the completion of the study fasting plasma cholesterol levels were markedly reduced in male (−59%) and female (−57%) mice treated with rHSP25 vs. rC1.

B) rHSP25 treatment reduced plasma PCSK9 levels in male mice (−69%; p=0.0006), however the PCSK9 reduction in female mice was not significant (−42%; p=0.11).

C) Hepatic LDLR mRNA expression increases in the male (+28%) and female (+21%) mice vaccinated with rHSP25 compared to rC1 (p<0.0001 for both).

D) - E) Compared to rC1 vaccination with rHSP25 resulted in higher levels of hepatic LDLR protein expression as assessed by Western blotting in both male and female ApoE^−/− mice. Each band represents protein expression for an individual mouse. The graph symbols represent values for individual mice with the bars indicating the mean and SEM values of the LDLR band intensity (normalized to the housekeeping protein vinculin) for each sex-specific treatment group.

To investigate potential mechanisms by which rHSP25 vaccination reduces atherosclerosis and plasma cholesterol levels, we measured plasma PCSK9 levels. Treatment with rHSP25 produced reductions in measured plasma PCSK9 levels in male mice (−69%; 135 ± 70 vs. 42 ± 11, n=10/group; p=0.0006), with a trend toward significance in female mice (−42%; 96 ± 59 vs. 56 ± 11, n=7/group; p=0.11) (Fig. 4B). When subjected to size exclusion fast protein liquid chromatography (FPLC) separation, the plasma from the rHSP25 vaccinated mice show a clear reduction in the low density lipoprotein (LDL) cholesterol and very low density lipoprotein (VLDL) cholesterol fractions (and in males there was a slight increase in high density lipoprotein cholesterol) compared to rC1 vaccinated mice (Supplemental Fig. IIC). PCSK9 levels in the FPLC elution fractions were lower with rHSP25 vaccination in both male and female mice (Supplemental Fig. IID). Total cholesterol levels were directly related to PCSK9 levels (r² = 0.380; p=0.0008).

Interestingly, hepatic PCSK9 mRNA and protein expression was similar in the rC1 compared to rHSP25 treated mice (Supplemental Fig. IIIA – IIIC), thereby raising the possibility that the reduction in plasma PCSK9 levels may be mediated via non-transcriptional / non-translational mechanisms. Moreover, protein expression for either the Sterol Regulatory Element-Binding Protein 2 (SREBP2) or Hepatocyte Nuclear Factor 1-alpha (HNF1α), two key transcriptional regulators of PCSK9 (Supplemental Fig. IIID, IIIE)^19,20 were unchanged with vaccination. Finally, there was a definite increase in hepatic LDLR expression in the rHSP25- vs. rC1-treated mice: i) mRNA: males: 28%; p<0.001; females: 21%; p<0.001; (Fig. 3C); ii) protein: males: 47%; p=0.002; females: 54%; p=0.004; (Fig. 3D, 3E). Importantly, hepatic lipid content appeared diminished with rHSP25 compared to rC1 vaccination (Supplemental Fig. IIIF). Moreover, as reflected by hepatic histology, plasma aspartate aminotransferase or glucose levels, there was no evidence of off-target toxicity with rHSP25 treatment (Table S3).

HSP27 Regulation of PCSK9 and LDLR Expression in HepG2 Cells

As rHSP25 vaccination resulted in striking reductions in cholesterol, with modest changes in PCSK9 levels but a robust increase in LDLR expression, we examined the potential role of the HSP27 IC in controlling the expression of key cholesterol regulatory proteins. Using Stable Isotope Labeling of Amino Acids in Cell culture (SILAC) followed by Mass Spectrometry (MS) the proteomic profile on HepG2 cells was quantified after treatment with the HSP27 IC. First, to form the IC we had to generate and validate a polyclonal anti-HSP27 IgG antibody (PAb). Although the PAb was generated in rabbits, it recognized the same HSP27 epitopes as human AAbs from either normal subjects or CAD patients (Supplemental Fig. IVA – IVD)²¹ and showed a protein immunodetection pattern similar to a commercial anti-HSP27 antibody (Supplemental Fig. IVE). For the SILAC experiment, control HepG2 cells were cultured in ‘light’ media, while cells treated with [rHSP27 + PAb] were cultured in media containing ‘heavy’ isotope-labelled amino acids. The ratio of ‘heavy’ to ‘light’ proteins synthesized with these treatments was normalized to that of GAPDH, and therefore reflects the relative changes in protein expression due to treatment with the HSP27 IC. Treating HepG2 cells with 1 μg/ml rHSP27 combined with 5 μg/ml PAb (~1:1 molar ratio) resulted in a 33% decrease in PCSK9 (p<0.0001) and an 44% increase in LDLR (p=0.0009) levels, with no effect on other known modulators of cholesterol metabolism, such as 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR), SREBP2, or the Inducible Degrader of the LDL receptor (IDOL) (Fig. 5A).

Fig. 5.

HSP27 plus anti-HSP27 polyclonal antibody (PAb) modulate PCSK9 and LDLR in HepG2 cells.

A) SILAC experiment performed after 18 hrs treatment with [rHSP27 (1 μg/mL) + PAb (5 μg/mL)]. The relative abundance of proteins involved in cholesterol metabolism was expressed as the ratio of their abundance with treatment vs. control conditions. Each ratio was then normalized to the ratio for the housekeeping protein GAPDH. Compared to control, the combination of [rHSP27 + PAb] decreased PCSK9 (−33%) and increased LDLR (+44%) expression.

B) rHSP27 plus AAb increase LDLR gene expression. Time course of changes in LDLR gene expression post-treatment with [rHSP27 (1 μg/mL) plus PAb (5 μg/mL)] or control treatments [PBS, PAb alone (5 μg/mL), rC1 (1 μg/mL) plus PAb (5 μg/mL), and rHSP27 (1 μg/mL) plus non-specific polyclonal IgG (5 μg/mL)]. Gene expression was assessed by qPCR and presented as log fold change. Relative to control treatments [rHSP27 + PAb] produced profound and sustained increases in LDLR expression (e.g., +136%, +131% at 36 and 48 hrs).

C) The increase in LDLR expression occurs without a drop in intracellular cholesterol levels. In fact, there is an increase in intracellular LDL-C concentration when HepG2 cells are incubated with LDL-C (10μg/ml) in the media.

D) HepG2 cells transfected with a full length LDLR promoter construct (pLDLR-1192, extending from −989 to +203 relative to the transcription start site of human LDLR gene) and employing firefly luciferase to demonstrate promoter activation. The various controls and treatments are outlined, with the results expressed as log transformed fold increase in promoter activation. Compared to control LDLR promoter activity was markedly increased with [rHSP27 + PAb] treatment (2.4-fold; p<0.0001).

E) Extracellular PCSK9 protein levels are reduced in HepG2 cells treated with small hair pin RNA (shRNA) sequences designed to knockdown the expression of HSP27 (verified in Supplemental Fig. VD & VE). No treatment and treatment with a scrambled shRNA represented control conditions. Extracellular levels of PCSK9 protein in the cell supernatant increased 55% when HSP27 expression was knocked down vs. scrambled shRNA sequence (quantification by ELISA).

F) Overnight treatment of HepG2 cells with AV (10μM) increased secreted PCSK9 protein levels by 66% in the cell supernatant, while [rHSP27 (1 μg/mL) plus PAb (5 μg/mL)] had no effect on its own, but did attenuate the AV-increase by 26% (quantification by ELISA).

Next, the potential role of the HSP27 IC in regulating expression of PCSK9 and its transcriptional factor HNF1α was examined in HepG2 cells treated with: [rHSP27 (1 μg/mL) plus PAb (5 μg/mL)], PBS, PAb alone (5 μg/mL), [rC1 (1 μg/mL) + PAb (5 μg/mL)], [rHSP27 (1 μg/mL) + non-specific rabbit IgG (5 μg/mL)], with the results reported as log-fold change. Relative to PBS, treatment with PAb alone, [rC1 + PAb], and [rHSP27 + control IgG] failed to show a significant and persistent alteration in PCSK9 or HNF1α expression (Supplemental Fig. VA – VB). While [rHSP27 + PAb] treatment resulted in a transient 27% reduction in PCSK9 gene expression compared to PBS treatment, this occurred at 6 hrs only (p<0.001). There were slightly larger and persistent reductions in HNF1α gene expression with [rHSP27 + PAb] treatment – but again these were transient (e.g., −36% at 6 hours, −35% at 12 hours, p<0.0001 for both).

In contrast, LDLR expression in HepG2 cells was intensely upregulated with [rHSP27 + PAb] treatment and for a protracted interval: at 24 hours: +24% (p<0.05), 36 hours: +136% (p<0.0001), and 48 hours: +131% (p<0.0001) (Fig. 5B). Normally, LDLR expression is regulated by intracellular cholesterol concentrations (i.e., via SREBP2); however, SREBP2 protein expression remained unchanged with rHSP25 vaccination (Supplemental Fig. IIID) and [rHSP27 + PAb] treatment (i.e., SILAC experiment; Fig. 5A). Indeed, this increase in HepG2 LDLR expression occurred without a reduction in intracellular cholesterol concentration.²² Treatment with [rHSP27 + PAb] produced an increase in LDL cholesterol uptake and intra-cellular cholesterol concentration (Fig. 5C). Nonetheless, in vivo there was no evidence of excessive intra-hepatic accumulation of lipid (Fig. 3A, Supplemental Fig. IIIF).

LDLR promoter activity was assessed using the plasmid pLDLR-1192, containing the full length promoter region from −989 to +203 relative to the transcription start site of human LDLR gene, as well as the plasmid pLDLR-234, containing the LDLR core promoter sequence with only the Sterol Response Element-1 (SRE-1) and Sp1 regulatory element sites from −142 to +35.^19,23 Compared to control treatment, the full length LDLR promoter showed a 2.4-fold increase in activity after treatment with low dose [rHSP27 (1μg) + PAb] compared to a 0.5-fold increase with high dose (100μg) rHSP27 alone (Fig. 5D). The results with the LDLR core promoter construct were less impressive (e.g., 0.46-fold increase with [rHSP27 (1μg/ml) + PAb] and 0.23-fold increase with high dose (100μg/ml) rHSP27 alone, thereby suggesting that there may be a novel regulatory mechanism of LDLR transcription that is upstream and absent in the core promoter construct (Supplemental Fig. VC).

Finally, to further assess the role of HSP27 on PCSK9 expression, HepG2 hepatocytes were transfected with a previously described short hairpin ribonucleic acid (shRNA) targeting HSP27 (Supplemental Fig. VD & VE).²⁴ Silencing HSP27 resulted in 55% higher levels of PCSK9 protein in the culture media compared to a scrambled control sequence (p<0.0001; Fig. 5E). To determine if the HSP27 IC can attenuate a statin-induced rise in PCSK9 protein, HepG2 cells were treated with atorvastatin (AV; 10μM), resulting in a 66% increase in PCSK9 protein vs. PBS control (Fig. 5F). On its own [rHSP27 + PAb] was similar to PBS control, while adding [rHSP27 + PAb] dampened the AV-induced increase in PCSK9 protein by 26% (p<0.0001).

LDLR Expression is mediated via the NF-κB Pathway

In the past, we repeatedly demonstrated in MΦ that HSP27 activates the NF-κB pathway to alter the transcription of several key genes that are either pro- or anti-atherogenic.^24–27 We showed in MΦ that activation of NF-κB by HSP27 required Toll Like Receptor 4 (TLR4) and could be blocked with various inhibitors of the NF-κB intracellular pathway (e.g., CLI-095 an inhibitor of TLR4; an inhibitor of IRAK1/4 in the TLR4 pathway; MG132 to block the proteasomal degradation of Iκ-β).²⁴ However, we did not explore if NF-κB activation by the HSP27 IC is integral to LDLR expression, and now present the following data. First, we illustrate the activation of the NF-κB pathway in hepatocytes, showing a strong immunofluorescent nuclear localization signal for the NF-κB p65 subunit in human hepatic cells treated with [rHSP27 + PAb] and much less so after treatment with rHSP27 (alone) or the PAb (alone) (Fig. 6A). Second, to explore the potential role of NF-κB in the upregulation of LDLR expression by HSP27, HepG2 cells were pre-incubated with (or without) the NF-κB pathway inhibitor, BAY 11–7082 for 30–60 minutes before being treated for up to 24 hours with [rHSP27 + PAb]. Compared to control, LDLR mRNA expression was increased by [rHSP27 + PAb] (66%; p<0.0001), and reduced by the BAY 11–7082 alone (35%, p=0.0008; Fig. 6B). Moreover, BAY 11–7082 inhibited the augmentation effect of [rHSP27 + PAb] on LDLR mRNA expression, keeping it at levels similar to control treatment. Third, relative to controls, the HSP27 IC increased LDLR protein expression by 70% at 16 hrs and 40% at 24 hrs (p=0.0008 and p=0.016; respectively) – escalations that were essentially nullified in the presence of BAY 11–7082 alone or combined with [rHSP27 + PAb] (Fig. 6C – 6D). Of note, the NF-κB pathway does not seem to be involved in the regulation of PCSK9 expression (Supplemental Fig. VF), further highlighting how the HSP27 IC shows divergent regulation of LDLR and PCSK9.

Fig. 6.

HSP27 Upregulates LDLR Expression via NF-κB Activation.

A) Human hepatocytes treated with rHSP27 (1μg/ml) with or without PAb (5 μg/ml) for 45 minutes before labelling with a fluorescent rabbit anti-NF-κB antibody and a Hoechst nuclear stain. A fluorescent nuclear localization signal was most abundant when cells were treated with [rHSP27 + PAb] (original photomicrograph: ×400 magnification).

B) In HepG2 cells LDLR gene expression increased with [rHSP27 + PAb] treatment. The addition of NF-κB blocker BAY11–7082 reduced basal LDLR expression by 35% vs. Control, and nullified the increase induced by [rHSP27+PAb].

C) – D) Treatment of HepG2 cells for 16 and 24 hrs with [rHSP27 + PAb] increased LDLR protein expression by 70% and 40% respectively, while co-treatment with BAY 11–7082 impeded these increases.

Discussion

Previously we demonstrated that low blood levels of HSP27 are predictive of cardiovascular events, and in murine models showed that augmenting HSP27 (or HSP25) levels attenuated atherogenesis. More recently we noted the presence of anti-HSP27 antibodies and questioned if they may reduce the salutary biological effects of HSP27. Hence, in the current studies we first examined the prevalence of anti-HPS27 antibodies in clinical cohorts. Next, we conducted rHSP25 vaccination experiments to explore if increasing anti-HSP25 antibodies in mice is atheroprotective. Finally, in vitro we used human hepatocytes to determine how the HSP27 immune complex regulates key components of cholesterol regulatory pathways.

Herein we demonstrate that anti-HSP27 IgG blood levels are elevated in healthy subjects compared to patients with CVD and confirm our previous report that HSP27 blood levels are also higher in health.⁹ To explore the role of these AAbs in atherosclerosis, we subjected ApoE^−/− mice to rHSP25 vaccination, resulting in the generation of anti-HSP25 IgG antibodies and the attenuation of the early inflammatory stages of atherogenesis. Of note, there were reductions in plaque cholesterol accumulation, total plasma cholesterol levels, as well as plaque, circulating and hepatic biomarkers of inflammation. These mice had increased hepatic LDLR mRNA and protein expression without marked alterations in PCSK9 mRNA or protein expression. Nonetheless, in mice vaccinated with rHSP25 compared to rC1, plasma PCSK9 levels appeared lower by FPLC, and quantitatively were reduced in males but not females. In vitro, we confirm that treatment of human hepatocytes with the HSP27 IC upregulates LDLR mRNA and protein expression, increases the activity of the LDLR promoter, is independent of intracellular cholesterol levels and requires the activation of the NF-κB pathway.

In the clinical cohorts that we studied, the origin of the AAbs is unclear, but presumably occurs in the absence of immunization or infection by an organism with cross-reactive epitopes (i.e., molecular mimicry). Perhaps these AAbs are akin to natural antibodies, and represent an immunological backup defense mechanism (e.g., to provide primary protection against infection during the gap prior to germinal center formation and adaptive antibody production)?²⁸ Recently, it has come to light that elevated antibody titers of any nature may represent some form of health advantage. For example, in the large Anglo-Scandinavian Cardiac Outcome Trial (ASCOT), elevated levels of IgG (regardless of the recognized antigen) are associated with a reduced risk of coronary heart disease.²⁹ Indeed, we recently showed how these anti-HSP27 IgG AAbs potentiate HSP27 signaling, facilitating its binding at the cell surface where it competes (e.g., with LPS) to interact with TLR4 and modestly activate the NF-κB pathway.³⁰ Moreover, the HSP27 IC can bind with scavenger receptors (e.g., SR-AI, CD-36) and competitively reduce the uptake of oxidized LDL. Hence, we coined the term ICAST (Immune Complex Activated Signaling and Transport) to describe these properties of the HSP27 IC. In the current study we show how HSP27 IgG antibodies co-localizes with HSP27 in human blood samples, and in vitro demonstrate the formation of the HSP27 IC using a Proximity Ligation Assay (Fig. 1D, 1E).

To study the mechanisms by which the HSP27 IC reduces plasma cholesterol levels, we examined the expression of two key regulators of LDL cholesterol, PCSK9 and the LDLR. rHSP25 vaccination did not alter hepatic PCSK9 mRNA and protein expression (Supplemental Fig. IIIA – IIIC). Yet, vaccination with rHSP25 compared to rC1 results in lower plasma PCSK9 levels (using FPLC, Supplemental Fig. IID), and reductions in measured PCSK9 plasma levels (using ELISA for males: −69%, p=0.0006; females: −42%, but non-significant p=0.11; Fig. 4B). This sex-difference in PCSK9 levels may be important,³¹ as recently we showed that estrogens augment PCSK9 expression in vivo, as well as increase PCSK9 expression and promoter activity in vitro.³² Interestingly, treatment of human hepatocytes in vitro with [rHSP27 + PAb] reduced PCSK9 protein expression by 33% (SILAC experiment; Fig. 5A). However, in the absence of any impressive or persistent changes in PCSK9 mRNA expression in vitro (Supplemental Fig. VA), the decrease in PCSK9 protein expression in the SILAC experiment could be due to the corresponding 44% increases in LDLR expression (that may increase PCSK9 protein disposal). Moreover, rHSP25 vaccination or the HSP27 IC in vitro had only minor transient effects on the expression of two transcriptional regulators of PCSK9, SREBP2 and HNF1α (Supplemental Fig. IIID, IIIE, VB). Hence, these data clearly point to an important and reproducible increase in LDLR expression with rHSP25 vaccination or treatment of hepatocytes in vitro with the HSP27 IC. Future experiments will take aim at sorting out whether the reduction in PCSK9 by HSP25/27 treatment is a direct transcriptional effect or related to enhanced clearance. Until then, our working hypothesis for these changes in PCSK9 levels is that rHSP25 vaccination or treatment with the HSP27 IC has only a modest and transient negative effect on PCSK9 expression, and it is the marked increase in LDLR levels that regulates PCSK9 levels (Graphical Abstract).

LDLR is the primary determinant of not only LDL cholesterol³³ but also PCSK9 levels – as the LDLR is the principal clearance mechanism for plasma PCSK9.³⁴ Considering that the LDLR recycles every 10 minutes, internalizing hundreds of LDL cholesterol particles during its ~20 hr lifespan,³⁵ small increases in LDLR expression translate into large decreases in LDL cholesterol as well as PCSK9 removal from the plasma. Moreover, knocking down HSP27 expression increases extracellular PCSK9, while adding [rHPS27 + PAb] to HepG2 cells attenuates statin-induced increases in extracellular PCSK9 levels (Fig. 5E, 5F). Finally, it is interesting to note that total plasma cholesterol levels did not change in LDLR^−/− mice vaccinated with rHSP25 – thereby supporting the concept that the central atheroprotective effect of rHSP25 immunotherapy is its upregulation of LDLR expression (Supplemental Fig. IIF).

Hepatocyte LDLR expression is thought to be regulated at the transcriptional level by a SRE in the LDLR promoter. In response to low intracellular cholesterol levels (e.g., with statin therapy), SREBP2 activity increases and binds to this SRE. However, other mechanism may also alter LDLR expression (e.g., ubiquitin-induced proteasomal degradation of HNF1α by berberine, or a PPAR-response element in the LDLR promoter).^23,36 In vitro, we note that intracellular cholesterol levels actually rise with rHSP27 IC treatment (with no corresponding hepatic steatosis or toxicity in vivo) (Fig. 3A, 5C; Supplemental Fig. IIIF, Supplemental Table III) and do not alter SREBP2 levels (Supplemental Fig. IIID). Hence, unlike statins, which result in increased expression of both LDLR and PCSK9, the effect of [rHSP27 + PAb] appears to be divergent, reducing PCSK9 levels yet promoting marked increases in LDLR promoter activity and mRNA/protein expression (Table 1).

Table 1:

HSP27 Immunotherapy and the Divergent Regulation of LDLR and PCSK9 Expression

	LDLR	PCSK9
Hepatic (in vivo): rHSP25 vaccination
• mRNA	Increase in LDLR mRNA (qPCR; Fig. 4C)	No change in mRNA (Suppl. Fig. IIIA)
• protein	Increased protein (Western blot; Fig. 4D,4E)	No change in hepatic protein levels (Suppl. Fig. IIIB,IIIC) but reduced plasma levels in male but not female mice (Fig. 4B)
• transcriptional regulation	No change in upstream in upstream transcriptional factors SREBP2 & HNF1-α (Suppl. Fig. IIID, IIIE)	No change in upstream in upstream transcriptional factors SREBP2 & HNF1-α (Suppl. Fig. IIID, IIIE)
HepG2 Hepatocytes (in vitro): [rHSP27 + PAb] treatment	LDLR	PCSK9
• mRNA	Increased mRNA (qPCR; Fig. 5B)	Minor, transient decrease in mRNA (qPCR; Suppl Fig. VA)
• protein	Increased protein (SILAC; Fig. 5A)	Decreased protein (SILAC; Fig. 5A)
• dependence on intracellular cholesterol	Increased expression is independent of intracellular cholesterol (Fig. 5C)	Decreased expression is independent of intracellular cholesterol (Fig. 5C)
• dependence on NF-κB activation	Decreased expression with NF-kB blockade (Fig. 6B–6D)	No change in expression with NF-kB blockade (Suppl. Fig. VF)

How HSP27 upregulates LDLR expression is also becoming better understood. Previously we demonstrated that extracellular HSP27 (alone) can activate intracellular signaling cascades such as NF-κB via TLR4 in MΦ.^24–26 While NF-κB signaling is commonly associated with pro-inflammatory responses,^37,38 it also regulates anti-inflammatory processes³⁹ and deletions of components of this pathway aggravate inflammation.^40–42 In the current murine experiments we see a marked reduction in not only SAA (−71%, −89%; Fig. 2K), a circulating biomarker of inflammation, but also hepatic MΦ abundance and expression levels of both IL1β and TNFα (Fig. 3C, 3D). Intriguingly, we now show for the first time that [rHSP27 + PAb] activates NF-κB in hepatocytes (Fig. 5A) and that this is a required step for the upregulated expression of LDLR mRNA (66%) and protein (70%) (Fig. 6B – 6D). This newly recognized HSP27 immuno-regulation pathway for LDLR may be druggable and superior to statin therapy which unfortunately upregulates both LDLR and PCSK9.

This study has limitations. First, because of the prevalence of statin use in our clinical cohort from Alabama (CVD: 96% and CON 17%; Table S1) we are unable to determine a meaningful relationship between the LDL cholesterol (or PCSK9) and HSP27 or AAbs. Hence, going forward we will look for additional cohorts to explore these relationships, as well as confirm that AAbs are higher in healthy subjects compared to CVD patients, paying specific attention to potential differences (e.g., race). Second, while we observed upregulated LDLR expression to be independent of a reduction in total intracellular cholesterol levels, future studies should focus on measuring endoplasmic reticulum cholesterol levels which are more relevant to regulation of SREBP2. Similarly, we measured SREBP2 protein abundance (Supplemental Fig. IIID) but not transcriptional activity, which might be more insightful. Third, we made important discoveries on the role of the NF-κB pathway in the regulation of LDLR transcriptional regulation, and how rHSP25 vaccination reduced several markers of inflammation, but additional studies are required to more completely understand the mechanisms for these effects (e.g., the regulation of the LDLR promoter and the specifics of the pathways for modulating inflammation).

In summary, HSP27 immunotherapy is a promising new strategy to reduce atherogenesis by lowering cholesterol levels through the upregulation of LDLR expression that promotes enhanced clearance of LDL cholesterol. Given the rapid recycling of the LDLR, we postulate that this increase in LDLR synthesis augments the intracellular disposal of PCSK9. In addition, the observation that [HSP27 + PAb] increases LDLR expression in an NF-κB dependent manner that is independent of intracellular cholesterol levels is novel and uncovers an attractive opportunity for the development of a new therapeutic avenue for managing cholesterol disorders. Going forward, we know that low HSP27 and AAb levels are associated with CVD, and are focusing our efforts on developing therapies to boost AAb levels in order to potentiate existing (low) levels of HSP27. Having mapped the HSP27 epitopes (Supplemental Fig. IVA – IVC) we are striving to optimize the formulation of a HSP27 vaccine that will provide a prolonged therapeutic benefit, as well as develop other HSP27 immunological approaches (e.g., the direct administration of anti-HSP27 monoclonal antibodies), for treating both dyslipidemia and inflammation, thereby providing a one-two punch targeting atherogenesis.

Highlights.

• HSP27 and IgG antibodies to HSP27 are higher in healthy control subjects compared to cardiovascular disease patients

• ApoE^−/− mice vaccinated with rHSP25, the mouse ortholog of the human HSP27 protein, generated anti-HSP25 IgG antibodies and reduced atherogenesis

• Plasma cholesterol levels were lowered with rHSP25 vaccination, which was associated with upregulated expression of the LDLR without major perturbations in that of PCSK9

• rHSP25 vaccination markedly reduced systemic (Serum Amyloid A) and hepatic (IL1β, TNF1α) indices of inflammation

• In vitro, the HSP27 immune complex upregulation of hepatocyte LDLR expression requires activation of the NF-κB pathway

Abbreviations

AAb

anti-HSP27 (or anti-HSP25) antibodies

ApoE^−/−

knockout of apolipoprotein E gene

a.u.

absorbance units

CI

confidence interval

CON

healthy control subjects

CVD

cardiovascular disease

FPLC

fast protein liquid chromatography

HFD

high fat diet

HNF1α

hepatic nuclear factor 1-alpha

HSP25

murine ortholog of human HSP27

HSP27

Heat Shock Protein 27

HSP27^o/e

over-expressing (human) HSP27

IC

immune complex

IgA, IgG, IgM

immunoglobulin A, G, M

LDL-C

low density lipoprotein cholesterol

LDLR

low density lipoprotein receptor

MΦ

macrophage

mRNA

messenger RNA

NF-κB

nuclear factor kappa light chain enhancer of activated B cells

PAb

polyclonal anti-HSP27 IgG antibody

PCSK9

Proprotein Convertase Subtilisin/Kexin type 9

rC1

recombinant protein representing the C-terminal of HSP27, spanning amino acids 90–205

rHSP27

recombinant HSP27

SAA

serum amyloid A