Association of Autoantibodies to Heat-Shock Protein 60 With Arterial Vascular Events in Patients With Antiphospholipid Antibodies

Abstract

Objective

Anti–heat shock protein 60 autoantibodies (anti-Hsp60) are associated with cardiovascular disease and are known to affect endothelial cells in vitro, and we have recently shown that anti-Hsp60 promote thrombosis in a murine model of arterial injury. Based on those findings, we undertook the present study to investigate the hypothesis that the presence of anti-Hsp60, alone or in combination with other thrombogenic risk factors, is associated with an elevated risk of vascular events.

Methods

The study population was derived from 3 ongoing cohort studies: 2 independent systemic lupus erythematosus (SLE) registries and 1 cohort comprising SLE patients and non-SLE patients. Data from a total of 402 participants were captured; 199 of these participants had had confirmed vascular events (arterial vascular events in 102, venous vascular events in 76, and both arterial and venous vascular events in 21). Anti-Hsp60 were detected by enzyme-linked immunoassay, and association with vascular events was assessed by regression analysis.

Results

Multiple regression analysis revealed that arterial vascular events were associated with male sex, age, and hypertension. Analyses of the vascular events according to their origin showed an association of anti-Hsp60 with arterial vascular events (odds ratio 2.26 [95% confidence interval 1.13–4.52]), but not with venous vascular events. Anti-Hsp60 increased the risk of arterial vascular events (odds ratio 5.54 [95% confidence interval 1.89–16.25]) in antiphospholipid antibody (aPL)–positive, but not aPL-negative, individuals.

Conclusion

We demonstrate that anti-Hsp60 are associated with an increased risk of arterial vascular events, but not venous vascular events, in aPL-positive individuals. These data suggest that anti-Hsp60 may serve as a useful biomarker to distinguish risk of arterial and venous vascular events in patients with aPL.

A growing body of evidence suggests that auto-antibodies to heat-shock protein 60 (anti-Hsp60) constitute an important nontraditional risk factor for cardiovascular disease (1). Immunity to the Hsp60 family has been implicated in endothelial cell stress/activation and the development of atherosclerosis (2). Heat-shock proteins show considerable sequence homology among species, and immune reactions to Hsp60 in infections by microorganisms that express Hsp65 have been widely described (3). Anti-Hsp65 antibodies, induced in response to these pathogens, can cross-react with self Hsp60 expressed on endothelial cells (4–7). Elevated levels of anti-Hsp60 have been associated with progression and severity of atherosclerosis (8–15), with vasculitis in systemic autoimmune diseases (16), and with thrombotic events in the context of systemic lupus erythematosus (SLE) and lupus anticoagulant (LAC) positivity (17). Moreover, anti-Hsp60 from SLE patients were shown to bind to the surface of endothelial cells and induce apoptosis in these cells (17).

We have recently demonstrated that anti-Hsp60 enhance thrombus formation and promote endothelial changes in an in vivo murine model of carotid artery injury (18), supporting the notion that these autoantibodies have a prothrombotic role. Based on these findings and others reported in the current literature, we hypothesized that individuals with circulating anti-Hsp60 might be at risk for vascular events. We further hypothesized that anti-Hsp60 might increase the risk of vascular events in individuals with known thrombovascular risk factors, such as antiphospholipid antibodies (aPL). Antiphospholipid antibodies (LAC, anticardiolipin antibody [aCL], and anti–β₂-glycoprotein I [anti-β₂GPI]) are associated with thrombovascular events (19–21), but it remains unclear why clinical events occur in only some individuals with aPL. Our findings demonstrate that the presence of anti-Hsp60 itself is associated with an increased risk of arterial, but not venous, vascular events. We further show that anti-Hsp60 confer a risk of arterial vascular events in aPL-positive individuals but not in aPL-negative individuals.

PATIENTS AND METHODS

Population

The study population was derived from the databases of 3 ongoing longitudinal observational studies: 1) the University of Toronto Lupus Clinic (UTLC) SLE registry, developed in 1970, 2) the McGill University Lupus Clinic (MULC) SLE registry, developed in 1978, and 3) the Montreal Antiphospholipid Antibody Study (MAPS) registry, developed in 1997. Clinical and laboratory data from a total of 402 participants were captured (140 from the UTLC, 144 from the MULC, 118 from the MAPS), and 4 groups were defined based on aCL and LAC status and vascular event history: 1) aCL and/or LAC positive with vascular events (n = 84), 2) aCL and/or LAC positive without vascular events (n = 83), 3) aCL and LAC negative with vascular events (n = 115), and 4) aCL and LAC negative without vascular events (n = 120).

The UTLC, MULC, and MAPS follow standard protocols. All 3 protocols have been reviewed and approved by the respective institutional research ethics boards, and written informed consent is obtained prior to enrollment in each registry. The UTLC registry (n = 1,255 at the time of the present study) and the MULC registry (n = 494 at the time of the present study) comprise patients with SLE according to the American College of Rheumatology (ACR) criteria (22,23). At the time of recruitment, participants complete questionnaires addressing demographic characteristics, personal health characteristics, and health status assessed by the Medical Outcome Study Short Form 36 (24). The SLE disease activity and damage questionnaires completed by the treating physician include the 2000 update of the Systemic Lupus Disease Activity Index (25) and the Systemic Lupus International Collaborative Clinics/ACR Damage Index (26). All participants are seen at least once annually for recording of intervening medical history, assessment of clinical status, physical examination, questionnaire completion, and blood testing. Routine aPL blood testing, including aCL (IgG/IgM) and LAC assays, is performed annually. Blood samples are collected at entry and annually (since 2005 at the UTLC and since 1993 at the MULC), and stored frozen at −70°C.

The MAPS registry comprises 416 patients being followed up for thrombosis. Participants were recruited between 1997 and 1998 from the McGill University Health Centre and Maisonneuve-Rosemont Hospital. Recruitment was based on the requisition for a blood test. Individuals whose treating physician had requested aCL/LAC testing were age-, sex-, and site-matched with individuals who had been referred for a routine complete blood cell count without aPL testing. Demographic and clinical data were collected at baseline. Blood samples were drawn at entry and annually (since 1997) for an additional 4 years, and stored frozen at −70°C. All participants are contacted semiannually for telephone interview followup. In preparing the data set for the current study, any duplication of patients between the MULC and MAPS registries was verified and patients who were part of both studies were considered only once.

Criteria for study entry

Only individuals who had had routine aCL and LAC testing performed on at least 2 occasions and who had frozen sera available for anti-Hsp60 and anti-β₂GPI testing were included in the study. The earliest serum was used if more than one sample was available. According to the revised classification criteria for antiphospholipid syndrome (APS) (19,20), aCL positivity was defined as an IgG or IgM aCL level of >40 IgG phospholipid units (GPL) or IgM phospholipid units (MPL), respectively, on at least 2 occasions at least 12 weeks apart, and LAC positivity was defined as confirmed LAC presence on at least 2 occasions at least 12 weeks apart. IgG and IgM anti-β₂GPI testing was not routinely available at our clinical laboratories, and this parameter was therefore not included in the criteria for study entry. However, frozen sera from patients were tested for IgG and IgM anti-β₂GPI, as described below.

Clinical variables

Clinical data collected at study entry included the following: demographic parameters (age, sex, and ethnic origin [defined as Caucasian versus other]), comorbidities (diabetes mellitus [DM] [defined by self-report and/or taking DM medications], hypertension [defined by self-report and/or blood pressure >140/90 mm/Hg and/or taking antihypertensive agents], and SLE), family history of cardiovascular disease (CVD) (defined as cerebrovascular accident [CVA], transient ischemic attack [TIA], myocardial infarction [MI], or angina in first-degree relatives), history of smoking, and previous arterial and/or venous vascular events. Followup data collected included new arterial and venous vascular events.

The outcome variables were confirmed vascular events occurring at any time. For our initial analyses, we used all vascular events, which included both arterial and venous vascular events; for later analyses, we evaluated arterial vascular events and venous vascular events separately. Arterial vascular events were defined as CVA, TIA, MI, angina, or arterial thrombosis at other sites. Venous vascular events were defined as deep vein thrombosis (DVT), pulmonary embolism, or venous thrombosis at other sites. All vascular events were confirmed via medical record review by physicians who were blinded with regard to the patient’s aPL status. Criteria for confirmation of a vascular event included a positive result on a diagnostic test and/or a documented clinical diagnosis by the treating physician. Diagnostic tests used to confirm events included Doppler ultrasound scan, venography, and ventilation perfusion scanning for pulmonary embolism and DVT, electrocardiography, angiography, and multiple gated acquisition scanning for MI, angiography and thallium testing for angina, magnetic resonance imaging and computed tomography scanning for CVA, ultrasound scan and venography for other venous vascular events, and angiography for other arterial vascular events. History and neurologic reports were used to confirm TIA.

Laboratory tests

PL assays

Cutoff values for aPL detection were in accordance with the revised APS classification criteria (19,20). Anticardiolipin antibody was measured using the Louisville assay (Louisville APL Diagnostics) (MULC and MAPS) or an assay from Phadia (UTLC) (since 2007). LAC assays were performed on plasma samples mixed 1:1 with normal plasma, according to the guidelines of the International Society on Thrombosis and Haemostasis (27). For MULC and MAPS samples, LAC was detected using a dilute activated partial thromboplastin time (APTT) assay (Automated APTT; BioMérieux Canada), in which the APTT reagent was diluted 1:10 in 20 mM HEPES buffer (pH 7.4) containing 15 mM NaCl, as previously described (28). For UTLC samples, LAC was detected using a dilute Russell’s viper venom time (dRVVT) assay (Precision BioLogic). LAC activity was confirmed by neutralization with hexagonal phase phosphatidylethanolamine, as previously described (28) (MULC and MAPS), or a platelet neutralization procedure using a PTT reagent from Diagnostica Stago and platelets from Precision BioLogic.

Samples were considered to be aCL positive if the GPL or MPL value was >40 as described above, and LAC positive if the APTT was ≥6.0 seconds above the control plasma value (MULC and MAPS) or if the dRVVT ratio was >1.15 (UTLC). The LAC confirmatory test was considered to be positive if the value was >8.0 seconds above control. Participants were tested for IgG and IgM anti-β₂GPI (Corgenix) using serum that had been aliquoted and stored frozen at −70°C. IgG or IgM anti-β₂GPI was considered to be positive if the level was >20 units/ml, a cutoff value greater than the 99th percentile in healthy controls, as recommended in the revised classification criteria for APS (19,20). Details about the assays and their reproducibility (e.g., coefficients of variation) can be found in the documentation accompanying each kit.

Anti-Hsp60 assay

Anti-Hsp60 titers were determined by enzyme-linked immunosorbent assay as described previously (17), with the following modifications: 1) the antigen was recombinant human Hsp60 (StressMarq Biosciences), 2) antibodies were detected using horseradish peroxidase–conjugated goat anti-human IgG (Southern Biotechnology), and 3) plates were developed with the BD OptEIA TMB Substrate Reagent set (BD Biosciences) and quantitated by reading the optical density at 450 nm (OD₄₅₀) (EL800 reader; BioTek Instruments). Anti-Hsp60 titers were also determined in 41 control sera, and the value corresponding to the 75th percentile in the control group (OD₄₅₀ 0.834) was used to define the cutoff forelevated IgG anti-Hsp60. This cutoff value has been shown previously to distinguish between low and high titers of anti-Hsp60 in healthy controls, patients with SLE, and patients with other rheumatic diseases (17). Figure 1 shows anti-Hsp60 levels in 41 healthy controls and in the 402 participants in the current study; the graph clearly demonstrates the selectivity of the OD₄₅₀ 0.834 cutoff value for high-level anti-Hsp60, as anti-Hsp60 levels in the majority of study participants fell below this threshold. The validity of the 75th percentile as an appropriate cutoff value was confirmed by analyzing other cutoff values (50th or 66th percentile in healthy controls) after all statistical analyses had been completed (see Results).

Figure 1.

Anti–heat-shock protein 60 autoantibody (anti-Hsp60) levels in the 41 healthy controls and 402 study participants, as detected by enzyme-linked immunosorbent assay. As anti-Hsp60 are found in both patients and healthy individuals, it is important to select for high levels of anti-Hsp60 IgG in evaluating clinical associations; the dashed line indicates the 75th percentile of anti-Hsp60 levels in the healthy control group (optical density at 450 nm [OD₄₅₀] 0.834), which was used to define the threshold value for elevated anti-Hsp60.

Statistical analysis

Descriptive analyses were created for all study variables using means, standard deviations, medians, and/or ranges, as appropriate. Outcome variables used in the models were arterial vascular events and venous vascular events. These outcomes were defined as binary variables (i.e., whether the patient experienced the outcome or not). The main independent variable for each model was anti-Hsp60 positivity, as defined above. We used covariates that were specific to each outcome. Covariates included in the arterial vascular events model were aPL status (defined as positive if aCL, LAC, and/or anti-β₂GPI were positive as described above and negative if aCL, LAC, and anti-β₂GPI were all negative), age, sex, smoking status, hypertension status, DM, SLE, family history of CVD, and ethnic origin. Covariates included in the venous vascular events model were aPL status, age, sex, and SLE. All covariates were defined as binary variables except for age, which was considered continuous.

Multiple logistic regression analysis was used to investigate the effect of anti-Hsp60 positivity on the probability that a patient would experience a vascular event, with adjustment for the covariates of interest. Interactions between anti-Hsp60 positivity and the covariates of interest were assessed. A separate regression analysis was performed for each outcome, with odds ratios (ORs) and 95% confidence intervals (95% CIs) calculated. For each regression model, the assumption of linearity in the logit for the variable age was assessed by Box-Tidwell test (29). Multicollinearity was assessed using Pearson’s correlation coefficient and by checking the variance inflation factor on a multiple regression model with the same dependent and independent variables (30). The log-likelihood ratio test was used to test the overall significance of the model. The significance of the variables in the model was assessed by Wald’s chi-square test. The fit of the model was assessed by Hosmer-Lemeshow goodness-of-fit chi-square test (31). To assess outliers and detect influential factors, logistic regression diagnostics were performed by plotting several diagnostic statistics against the predicted values, using estimated values and Pearson and deviance residuals (31).

We also ran another set of logistic regression models with the same outcomes and covariates of interest for each separate aPL (i.e., aCL, LAC, and anti-β₂GPI). In each model, we considered one aPL as the main independent factor and evaluated the interaction of that aPL with anti-Hsp60, keeping the other two aPL as covariates. All analyses were performed using SAS version 9.2 (SAS Institute). All hypothesis tests were 2-sided and performed at the 0.05 significance level.

RESULTS

Characteristics of the cohort

The characteristics of the study cohort (n = 402) are shown in Table 1. The study population was predominantly middle-aged Caucasian women, with a 73% prevalence of SLE. Twenty-six percent of the study participants were smokers, and 38% had hypertension. Six percent had DM, and 56% had a family history of CVD. Of the 402 subjects, 199 had confirmed vascular events: 102 (51%) had arterial vascular events, 76 (38%) had venous vascular events, and 21 (11%) had both arterial and venous vascular events. Elevated aCL was present in 127 of the 402 subjects (32%), LAC in 76 (19%), anti-β₂GPI in 116 (29%), and anti-Hsp60 in 43 (11%).

Table 1.

Baseline characteristics of the study cohort (n = 402)^*

Age, mean ± SD years	46.28 ± 13.9
Female sex	341 (84.8)
Caucasian	326 (81.1)
Smoker	105 (26.3)
Hypertension	153 (38.2)
Diabetes mellitus	24 (6.0)
SLE	292 (72.8)
Family history of CVD	223 (55.6)
Vascular events	199 (49.5)
Arterial	102 (51.3)
Venous	76 (38.2)
Arterial and venous	21 (10.6)
Anti-Hsp60	43 (10.8)
aCL	127 (31.6)
LAC	76 (18.9)
Anti-β2GPI	116 (29.2)
aPL†	187 (46.8)

^*Except where indicated otherwise, values are the number (%). Where data were missing for some individuals, values are based on the following numbers: smoking status (n = 400), hypertension (n = 401), diabetes mellitus (n = 401), systemic lupus erythematosus (SLE) (n = 401), anti–heat-shock protein 60 autoantibodies (anti-Hsp60) (n = 400), anti–β₂-glycoprotein I (anti-β₂GPI) (n = 397), and antiphospholipid antibody (aPL) (n = 400). Family history of CVD = history of cardiovascular disease in a first-degree relative.

^† Positivity for at least 1 of the following: IgG or IgM anticardiolipin antibody (aCL) (>40 phospholipid units), lupus anticoagulant (LAC), or anti-β₂GPI.

Association of anti-Hsp60 with vascular events

Simple regression analyses

The results of the separate simple regression analyses (only 1 predictor variable) are shown in Table 2. The association of anti-Hsp60 with total vascular events approached but did not achieve significance (OR 1.44 [95% CI 0.97–2.15]) (P = 0.07). LAC positivity was significantly associated with vascular events (OR 1.73 [95% CI 1.04–2.88]) (P = 0.034), while anti-β₂GPI positivity also approached but did not achieve significant association (OR 1.46 [95% CI 0.94–2.26]) (P = 0.09). Neither aCL positivity nor aPL positivity (aCL or LAC or anti-β₂GPI) showed an association with vascular events. As expected, age (OR 1.02 [95% CI 1.01–1.04]), hypertension (OR 2.27 [95% CI 1.50–3.43]), and family history of CVD (OR 1.52 [95% CI 1.02–2.26]) were significantly associated with greater odds for vascular events (P = 0.003, P = 0.0001, and P = 0.038, respectively), and female sex with lower odds (OR 0.5 [95% CI 0.28–0.88]) (P = 0.016).

Table 2.

Simple logistic regression analysis of associations with all vascular events^*

Variable	OR (95% CI)
Age (years)	1.02 (1.01–1.04)†
Sex (female)	0.5 (0.28–0.88)†
Smoking	0.92 (0.58–1.45)
Hypertension	2.27 (1.5–3.43)†
Diabetes mellitus	2.12 (0.89–5.07)
SLE	0.96 (0.62–1.48)
Family history of CVD	1.52 (1.02–2.26)†
Ethnic origin	1.35 (0.81–2.24)
Anti-Hsp60	1.44 (0.97–2.15)
aCL	1.10 (0.72–1.68)
LAC	1.73 (1.04–2.88)†
Anti-β2GPI	1.46 (0.94–2.26)
aPL‡	1.06 (0.72–1.57)

^*Analysis was performed on all cohort members with no missing values for any of the variables (n = 398). OR = odds ratio; 95% CI = 95% confidence interval (see Table 1 for other definitions).

^†P < 0.05.

^‡ Positivity for at least 1 of the following: IgG or IgM aCL (>40 phospholipid units), LAC, or anti-β₂GPI.

Appropriateness of the cutoff used for anti-Hsp60 positivity

After completion of the simple regression analyses, we performed analyses to verify that we had selected the appropriate cutoff value for anti-Hsp60 positivity. The association of anti-Hsp60 with arterial vascular events was evaluated by simple regression using 3 different cutoff values (50th, 66th, or 75th percentile). It was not possible to use lower percentiles (e.g., 25th), as the anti-Hsp60 levels for those percentiles were close to zero. We found a statistically significant association of anti-Hsp60 with arterial vascular events when using the 75th percentile in the healthy control group as the cutoff (OR 2.17 [95% CI 1.14–4.13]) but not when using the 66th percentile (OR 1.76 [95% CI 0.991–3.12]) or the 50th percentile (OR 1.47 [95% CI 0.94–2.32]). These data confirm that we selected an appropriate cutoff value for detecting elevated anti-Hsp60 in our study cohort.

Multiple regression analyses of vascular events overall

In a model in which no interaction effects were considered, there was no evidence of a statistically significant association between anti-Hsp60 and total vascular events (OR 1.77 [95% CI 0.90–3.46]) (P = 0.1) (Table 3). The increased risk of vascular events with hypertension (OR 1.84 [95% CI 1.18–2.86]) and decreased risk with female sex (OR 0.51 [95% CI 0.28–0.94]) were also retained in the multiple regression model (P = 0.007 and P = 0.03, respectively), but the association with family history of CVD was lost (OR 1.30 [95% CI 0.84–2.00]) (P = 0.24). Similar to the findings in the separate simple regression analyses, there was no association of aPL positivity with vascular events.

Table 3.

Multiple logistic regression analysis of risk factors associated with all vascular events, arterial events, and venous events^*

Risk factor	Odds ratio (95% confidence interval)
	All vascular events	Arterial events	Venous events
Age (years)	1.01 (0.99–1.03)	1.04 (1.02–1.06)†	0.99 (0.97–1.00)
Sex (female)	0.51 (0.28–0.94)†	0.45 (0.24–0.85)†	1.01 (0.53–1.93)
Smoking	0.88 (0.55–1.42)	0.97 (0.56–1.67)	ND
Hypertension	1.84 (1.18–2.86)†	2.45 (1.51–3.97)†	ND
Diabetes mellitus	1.40 (0.55–3.59)	2.13 (0.82–5.58)	ND
SLE	1.16 (0.71–1.90)	1.42 (0.80–2.54)	0.79 (0.46–1.34)
Family history of CVD	1.30 (0.84–2.00)	1.28 (0.77–2.12)	ND
Ethnic origin	1.23 (0.70–2.10)	1.65 (0.82–3.32)	ND
aPL‡	0.96 (0.63–1.47)	0.88 (0.54–1.43)	1.22 (0.76–1.95)
Anti-Hsp60	1.77 (0.90–3.46)	2.26 (1.13–4.52)†	1.25 (0.61–2.60)

^*Analysis was performed on all cohort members with no missing values for any of the variables (n = 398 for all vascular events [arterial and venous combined] and for arterial vascular events; n = 399 for venous vascular events). ND = not determined (variables that were not entered into the model for venous vascular events) (see Table 1 for other definitions).

^†P < 0.05.

^‡Positivity for at least 1 of the following: IgG or IgM aCL (>40 phospholipid units), LAC, or anti-β₂GPI.

Individual models for arterial vascular events and venous vascular events

As the origin of a vascular event (i.e., arterial or venous) can be important in associating risk factors with clinical events (32), we determined whether anti-Hsp60 were associated specifically with arterial vascular events or venous vascular events (Table 3). Arterial vascular events modeling was done for 398 individuals and venous vascular events modeling for 399 individuals with complete data. When we applied the multiple regression model with no interactions using arterial vascular events as the outcome, anti-Hsp60 showed a significant association with arterial vascular events (OR 2.26 [95% CI 1.13–4.52]) (P < 0.0001). As expected, increased age (OR 1.04 [95% CI 1.02–1.06]) and hypertension (OR 2.45 [95% CI 1.51–3.97]) were associated with greater odds of arterial vascular events (P < 0.0001 and P < 0.0003, respectively), and female sex with a lower odds of arterial vascular events (OR 0.45 (95% CI 0.24–0.85]) (P = 0.014). None of the variables studied in the model were significantly associated with venous vascular events.

Effect of the presence of both anti-Hsp60 and aPL on arterial vascular events

As both aPL and anti-Hsp60 are potential risk factors for vascular events, we investigated whether individuals with both autoantibodies were at greater risk of having a vascular event, by including an interaction effect between aPL and anti-Hsp60 in the multiple logistic regression models. The results showing the interaction between aPL and anti-Hsp60 are summarized in Table 4. The presence of anti-Hsp60 was strongly associated with total vascular events when aPL was also present (OR 3.17 [95% CI 1.06–9.46]) (P = 0.04). Analyses taking into account the origin of the vascular event (arterial or venous) revealed that aPL and anti-Hsp60 were associated with arterial vascular events (OR 5.54 [95% CI 1.89–16.25]) (P = 0.002) but not with venous vascular events (OR 1.06 [95% CI 0.36–3.14]) (P = 0.92). Notably, the effect of anti-Hsp60 in the presence of aPL on increasing the risk for arterial vascular events, but not venous vascular events, was replicated for each individual aPL, i.e., aCL (OR 6.58 [95% CI 1.68–25.74]) (P = 0.01), LAC (OR 5.81 [95% CI 1.21–27.83]) (P = 0.03), and anti-β₂GPI (OR 7.44 [95% CI 1.92–28.78]) (P = 0.004). The association of aPL and anti-Hsp60 with arterial vascular events was also replicated for each individual study center: for the Montreal cohorts (MULC and MAPS) combined, OR 4.17 (95% CI 1.08–16.17]) (P = 0.03), and for the Toronto cohort (UTCL), OR 11.04 (95% CI 1.62–75.03]) (P = 0.02). Similar to the findings in the entire cohort, there was no association of aPL and anti-Hsp60 with venous vascular events in the individual centers.

Table 4.

Multiple logistic regression analysis of risk factors, including an interaction effect between anti-Hsp60 and aPL, associated with all vascular events, arterial events, and venous events^*

	Odds ratio (95% confidence interval)
	All vascular events	Arterial events	Venous events
Age (years)	1.01 (0.99–1.03)	1.04 (1.02–1.06)†	0.99 (0.97–1.00)
Sex (female)	0.51 (0.28–0.94)†	0.44 (0.23–0.85)†	1.01 (0.52–1.93)
Smoking	0.89 (0.55–1.43)	0.98 (0.56–1.69)	ND
Hypertension	1.87 (1.20–2.91)†	2.55 (1.56–4.16)†	ND
Diabetes mellitus	1.36 (0.53–3.50)	2.02 (0.76–5.37)	ND
SLE	1.17 (0.72–1.93)	1.48 (0.82–2.66)	0.78 (0.46–1.33)
Family history of CVD	1.32 (0.85–2.04)	1.31 (0.79–2.17)	ND
Ethnic origin	1.26 (0.72–2.21)	1.77 (0.87–3.60)	ND
Anti-Hsp60+ when aPL−	1.16 (0.48–2.79)	1.09 (0.41–2.87)	1.43 (0.55–3.68)
Anti-Hsp60+ when aPL+‡	3.17 (1.06–9.46)†	5.54 (1.89–16.25)†	1.06 (0.36–3.14)

^*Analysis was performed on all cohort members with no missing values for any of the variables (n = 398 for all vascular events [arterial and venous combined] and for arterial vascular events; n = 399 for venous vascular events). ND = not determined (variables that were not entered into the model for venous vascular events) (see Table 1 for other definitions).

^†P < 0.05.

^‡Antiphospholipid antibody positivity was defined as positivity for at least 1 of the following: IgG or IgM aCL (>40 phospholipid units), LAC, or anti-β₂GPI.

In contrast, anti-Hsp60 in the absence of aPL were not associated with arterial vascular events (OR 1.09 [95% CI 0.41–2.87]) (P = 0.86) (Table 4), nor was there an association of aPL (aCL, LAC, and/or anti-β₂GPI) (OR 0.88 [95% CI 0.54–1.43]) (P = 0.60) or individual aPL specificities with arterial vascular events. Of note, anti-β₂GPI was the only individual aPL to show an association with total vascular events, when present with anti-Hsp60 (OR 5.65 [95% CI 1.15–27.75]) (P = 0.03).

DISCUSSION

Conventional cardiovascular risk factors only partially predict the occurrence of vascular events. Chronic infection and autoimmune reactions, especially to heat-shock proteins, may constitute another important variable in the pathogenesis of these events (1). We have recently shown that anti-Hsp60 autoantibodies can arise in response to mycobacterial Hsp65 and that these autoantibodies can potentiate thrombus formation in an in vivo murine model of arterial thrombosis (18). Based on these findings, we hypothesized that individuals with circulating anti-Hsp60 might be at risk for vascular events and that anti-Hsp60 might increase the risk of vascular events in individuals with known thrombovascular risk factors, such as aPL. Our findings demonstrate that anti-Hsp60 are associated with an increased risk of arterial vascular events, but not venous vascular events. We further show that anti-Hsp60 increase the risk of arterial vascular events in aPL-positive, but not aPL-negative, individuals.

The selective association of anti-Hsp60 with arterial vascular events and not venous vascular events is potentially important in understanding the pathogenicity of anti-Hsp60. Arterial thrombi tend to occur at sites of arterial plaque rupture where shear stress is high, while venous thrombi tend to occur at sites where the vessel wall is often normal and blood flow and shear stress are low (33). Elevated shear stress, which occurs more frequently in arterial than in venous beds, can result in endothelial surface expression of Hsp60 both in vitro and in vivo (34). Increased expression of Hsp60 in arterial vessels may explain the association of anti-Hsp60 with arterial but not venous thrombovascular events.

Given the association of anti-Hsp60 with arterial vascular events, we were interested in its potential effect in the presence of other risk factors, in particular aPL. We and others have shown that aPL is associated with thrombovascular events (21). In murine models of APS, aPL promotes thrombus formation in the presence of vascular injury and/or endothelial activation, but does not itself initiate formation of the thrombus (35,36). Indeed, in the absence of vascular injury or Toll-like receptor 4 activation, aPL alone has little or no effect (36,37). These data suggest that aPL alone may not be sufficient to initiate in vivo thrombus formation, and that more than one factor or “hit” is necessary to cause thrombovascular events in vivo. In a recent prospective study, we found that aPL predicted new arterial vascular events but not venous vascular events, and that in individuals who had experienced a previous event, there was a tendency for the same type of event (arterial vascular event or venous vascular event) to recur (32). These findings suggest that the predisposing features for arterial vascular events and venous vascular events are different.

In this study, we examined the effects of the presence of aPL (aCL, LAC, and/or anti-β₂GPI) and anti-Hsp60 on the occurrence of vascular events. Indeed, anti-Hsp60 showed a strong association with vascular events when aPL was present (OR 3.17 [95% CI 1.06–9.46]), and an even greater association with arterial vascular events (OR 5.54 [95% CI 1.89–16.25]), but not with venous vascular events. Importantly, this was replicated for each individual aPL (i.e., aCL, LAC, and anti-β₂GPI). Of note, anti-β₂GPI not only showed the highest OR for this association with arterial vascular events, but was also the only individual aPL to show an association with total vascular events when present with anti-Hsp60. In contrast, in the absence of anti-Hsp60, there was no association of aPL with arterial vascular events. Thus, aPL was associated with vascular events, but its association depended on the concomitant presence of anti-Hsp60. Our data indicate that anti-Hsp60 may serve as a biomarker to distinguish risk for arterial vascular events and venous vascular events, especially in patients with aPL.

Anti-Hsp60 are induced by exposure to a pathogen expressing Hsp65, and likely result from cross-reactivity of the anti-Hsp65 antibodies with autologous Hsp60 (3). Based on our findings, we propose a “double-hit model” (Figure 2), which provides a possible explanation for the association of anti-Hsp60 and aPL with arterial vascular events. The association of anti-Hsp60 with arterial vascular events but not venous vascular events suggests that Hsp60 is expressed in the arterial bed but not the venous bed. Elevated shear stress, which occurs more frequently in arterial than in venous beds, can result in endothelial surface expression of Hsp60 both in vitro and in vivo (34). Differential endothelial cell expression of Hsp60 may at least partially explain the greater susceptibility of the arterial bed to anti-Hsp60. We and others have provided both in vitro and in vivo evidence that the presence of anti-Hsp60 is itself sufficient to cause endothelial cell dysfunction (apoptosis, activation, and/or injury) (2,16–18). Such anti-Hsp60–induced endothelial damage would constitute the “first hit” and would favor the recruitment of aPL by exposing phosphatidylserine, a negatively charged phospholipid that interacts with phospholipid-binding proteins. Antiphospholipid antibody binding to phospholipid-binding protein on the endothelial cell membrane (the “second hit”) could promote a prothrombotic environment and predispose to arterial vascular events (38).

Figure 2.

Proposed “double-hit model” for arterial vascular events in the presence of anti–heat-shock protein 60 autoantibodies (anti-Hsp60) and antiphospholipid antibody (aPL). The illustration outlines a minimal, but sufficient, model to explain the association of anti-Hsp60 and aPL, but not aPL alone, with arterial vascular events. In this double-hit model, the association of anti-Hsp60 with arterial vascular events is due to selective expression of Hsp60 in the arterial bed. Elevated shear stress, which occurs more frequently in arterial than in venous beds, results in endothelial surface expression of Hsp60, providing a target for circulating anti-Hsp60. We propose that binding of anti-Hsp60 serves as the “first hit” for arterial vascular events, and that anti-Hsp60 binding leads to the exposure of negatively charged phospholipid (anionic PL) that is bound by phospholipid-binding proteins (PL-BP). In aPL-positive individuals, cell-bound phospholipid-binding protein (e.g., β₂-glycoprotein I) is targeted by circulating aPL, constituting the “second hit.” Antiphospholipid antibody binding would promote a prothrombotic environment and predispose to arterial vascular events. This model provides one possible explanation for the increased risk of arterial vascular events observed in individuals who are positive for both anti-Hsp60 and aPL, but does not preclude the possibility that anti-Hsp60–induced endothelial changes may require other factors that predispose to thrombosis.

This double-hit model does not preclude the possibility that anti-Hsp60–induced endothelial damage may require other factors besides aPL that predispose to endothelial dysfunction, or the possibility that other risk factors for endothelial dysfunction or thrombophilia may result in a prothrombotic state similar to that seen with anti-Hsp60 and aPL. In this regard, it is worth noting that, in our logistic regression models, interactions of anti-Hsp60 with other known cardiovascular risk factors (e.g., hypertension, smoking, or DM) were not statistically significant (data not shown).

In conclusion, we have presented evidence that anti-Hsp60 are associated with an increased risk of arterial vascular events, but not venous vascular events. We further show that anti-Hsp60 may increase the risk of arterial vascular events in individuals with aPL. In our study population, the frequency of anti-Hsp60 positivity was similar in patients with primary APS (4 of 21 [19.0%]) and those with secondary APS (10 of 72 [13.9%]), suggesting that anti-Hsp60 may play a role in both conditions. However, given the limited number of individuals in these subgroups, large prospective studies are needed to replicate these findings and to evaluate anti-Hsp60 as a clinical predictor of arterial vascular events.