Long-term exposure to air pollution raises circulating levels of proprotein convertase subtilisin/kexin type 9 in obese individuals

Abstract

Aims

Exposure to airborne particulate matter has been consistently associated with early death and increased morbidity, particularly raising the risk of cardiovascular disease. Obesity, one of the leading cardiovascular disease risk factors, increases susceptibility to the adverse effects of particulate matter exposure. Proprotein convertase subtilisin/kexin type 9 has been related to a large number of cardiovascular risk factors, e.g. atherogenic lipoproteins, arterial stiffness and platelet activation. Thus, the present study was aimed at evaluating, in a series of obese individuals, the effects of particulate matter less than 10 µm in diameter (PM₁₀) on proprotein convertase subtilisin/kexin type 9 circulating levels.

Methods and results

In 500 obese subjects, participating in the cross-sectional Susceptibility to Particle Health Effects, miRNAs and Exosomes (SPHERE) study, we evaluated the effects of long- and short-term PM₁₀ exposure on circulating proprotein convertase subtilisin/kexin type 9 levels. In the studied individuals (body mass index: 33.3 ± 5.2 kg/m²) with an annual average PM₁₀ exposure of 40.12 ± 4.71 µg/m³, proprotein convertase subtilisin/kexin type 9 levels were 248.7 ± 78.6 ng/mL. In univariate analysis, PM₁₀ exposure (annual average) was associated with proprotein convertase subtilisin/kexin type 9 levels (β=1.83, standard error = 0.75, p = 0.014). Interestingly, in a multivariable linear regression model, this association was observed only for carriers of lower concentrations of interferon-γ, whereas it was lost in the presence of higher interferon-γ levels. Proprotein convertase subtilisin/kexin type 9 levels were positively associated with the Framingham Risk Score, which was raised by 15.8% for each 100 ng/ml rise of proprotein convertase subtilisin/kexin type 9.

Conclusions

In obese individuals, more sensitive to the damaging effects of environmental air pollution, PM₁₀ exposure positively associates with proprotein convertase subtilisin/kexin type 9 plasma levels especially in those with low levels of interferon-γ. 

Introduction

Cardiovascular disease (CVD) is the leading cause of death worldwide, with over 17 m premature deaths in 2016. Air pollution is responsible for 19% of all cardiovascular (CV) deaths, including 23% of all ischaemic heart disease and 21% of all stroke deaths.¹

Exposure to air pollution, especially to particulate matter (PM), is increasingly rated as one of the main public health issues, with major negative socio-economic consequences.² The Lancet Commission on pollution and health estimated that pollution was responsible for 9 million premature deaths worldwide in 2015.³ Indeed, numerous health studies have shown that acute and chronic PM exposure are associated with early death and increased morbidity, with the highest impact on CVD.^4,5 PM has been linked with endothelial dysfunction and vasoconstriction, increased blood pressure, prothrombotic and coagulant changes, systemic inflammatory and oxidative stress responses, arrhythmias and progression of atherosclerosis.⁶ Among those individuals who are more vulnerable to the damaging effects of PM exposure and the consequent risk of CVD development (e.g. elderlies, diabetics, patients with preexisting coronary heart disease, chronic lung disease or heart failure) the obese ones have arisen as new candidates.⁷ Obesity, a strong risk factor for CVD,⁸ may increase susceptibility to the adverse effects of PM exposure. Indeed, both experimental and epidemiological data have proved obesity to modify the effects of PM exposure on heart rate variability and markers of inflammation, oxidative stress and acute phase response.⁹

Airborne PM is defined as particles suspended in the air, traditionally characterised by their aerodynamic diameter. PM₁₀ represents particles less than 10 µm in diameter: particles of this size can penetrate the airways and have a detrimental impact on health. Although the mechanism of action is still largely unknown, at the cellular level, exposure to PM can be cytotoxic, trigger overproduction of oxygen radicals leading to redox imbalance and oxidative stress¹⁰ and promote the release of pro-inflammatory mediators.¹¹ In addition, PM can cause heritable variations in gene expression via epigenetic effects, by altering DNA methylation patterns.¹²

Proprotein convertase subtilisin/kexin 9 (PCSK9) is a liver-secreted plasma protein that regulates the number of cell-surface low-density lipoprotein (LDL) receptors, inhibiting LDL uptake.¹³ PCSK9 circulating levels, which vary over an approximately 100-fold range in the general population,¹⁴ have been related to a large number of CV risk factors, i.e. LDL-cholesterolaemia, triglycerides (TGs), atherogenic lipoproteins¹⁵ and platelet activation.¹⁶ Genetic studies confirmed the key role of PCSK9 in atherosclerosis by showing a decrease in coronary artery disease (CAD) events in subjects with certain genetic polymorphisms, e.g. the PCSK9 R46L variant.¹⁷ Results from a meta-analysis from 12,081 subjects showed that elevated PCSK9 levels, compared to low levels, led to a 23% higher risk for total CV events.¹⁸ Interestingly, this predictive ability is higher in patients with a low CV risk and low inflammatory burden,¹⁸ making plasma PCSK9 a possible independent factor beyond the traditional CV risk biomarkers.¹⁹ As a further confirmation of this hypothesis, very recently, a genome wide association study showed a causal effect of PCSK9 levels on CAD, number of coronary vessels with stenosis, and carotid artery plaques. Indeed, a genetic drop of PCSK9 levels by 50% has been associated with a reduction of CAD risk by 50%.²⁰

Based on these premises, since mechanisms linking PM exposure to CV risk have not yet been fully elucidated, we aimed at evaluating, in a cohort of susceptible obese subjects, the effects of PM₁₀ exposure on PCSK9 circulating levels and how these may be associated with the CV Framingham Risk Score (FRS).

Methods

Study design and participants

The baseline study population has been previously described. Briefly, we selected 500 obese subjects among participants of the cross-sectional Susceptibility to Particle Health Effects, miRNAs and Exosomes (SPHERE) study.²¹ Subjects were recruited from the Center for Obesity and Work-Activity (Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico in Milan, Lombardia, Italy). The eligibility criteria of the SPHERE study were: (a) older than 18 years at enrolment; (b) overweight/obese according to body mass index (BMI): overweight, BMI between 25–30 kg/m²; obese: BMI of 30 kg/m² or more; (c) resident in the Lombardy Region at the time of recruitment. Exclusion criteria were: previous diagnosis of cancer, heart diseases, stroke, other chronic diseases or known diagnosis of diabetes. Each participant provided written informed consent which was approved by the Ethics Committee of Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico (approval number 1425). On the day of recruitment, each subject underwent physical and anthropometric evaluations as well cardiovascular and pulmonary function tests. The study was carried out over the period September 2010–April 2014.

PM exposure assessment

We chose to investigate long-term PM₁₀ exposure (six-month- and one-year average) as well as short-term PM₁₀ exposure (two-week lag exposure time window) and their associations with PCSK9 circulating levels. PM₁₀ was chosen for its known detrimental effects on CV health (compared for example to gaseous pollutants) and as monitoring data in Lombardy region are characterised by a better spatial resolution than those available for smaller PM components, e.g. PM_2.5.

Daily PM₁₀ concentrations were collected from fixed monitoring stations of the Regional Environmental Protection Agency (ARPA Lombardy). Using ArcGIS software (Esri), we assigned to each subject the daily PM₁₀ concentration from the nearest monitor to their home address. Meteorological data were obtained from the ARPA monitoring stations, measuring temperature (233 monitors) and relative humidity (163 monitors). Detailed information of the exposure assessment method has been previously described in Pergoli et al.²²

Clinical, laboratory measurements, cytokines evaluation

A detailed description is provided in the Supplementary Material Appendix.

Enzyme-linked immunosorbent assay (ELISA)

A description of PCSK9 measurement is provided in the Supplementary Material Appendix.²³

Statistical analysis

Descriptive statistics were performed on all variables. Continuous data were expressed as the mean ± standard deviation (SD) or as the median and interquartile range (Q1–Q3), as appropriate. Categorical data were presented as frequencies and percentages. We used univariate and multivariable linear regression models to test the relationship between circulating PCSK9 levels and PM₁₀ exposure. Continuous variables were tested for normality and linearity. Multivariable analyses were adjusted for variables significantly related with the outcome in univariate analysis (p < 0.05). Given the existence of multicollinearity among predictor variables (e.g. total, high-density lipoprotein (HDL), LDL and non-HDL cholesterol) the variance inflation factor (VIF) statistic was calculated. To determine the best performing model, we ran several regression equations separately, that included one or more significant explanatory variables used to predict PCSK9 levels. For each model, we separately estimated the respective β coefficient and standard error (SE) for each variable in the model, p-value, VIF statistics, as well as goodness of fit of the model (R²). Finally, the best model selected to predict the association between circulating PCSK9 levels and PM₁₀ exposure was adjusted for age, gender, waist circumference, smoking habit, occupation, use of statin medication, HDL and non-HDL cholesterol, TG, quantitative insulin sensitivity check index (QUICKI) score, lymphocytes % and interferon (IFN)-γ. To examine the potential effect modification of IFN-γ, we added the interaction term PM₁₀ * IFN-γ to the multivariable selected model. We evaluated whether the effect of PM₁₀ exposure on PCSK9 concentrations differs depending on IFN-γ levels, and results are shown using counter plots. A simple linear regression model was applied to verify the association between FRS and circulating PCSK9 levels. The FRS was log (base e) transformed to achieve a normal distribution. All statistical analyses were performed with SAS software (version 9.4; SAS Institute Inc., Cary, North Carolina, USA). A two-sided p value of 0.05 was considered as statistically significant.

Results

Study population

The study population included 500 obese subjects (BMI, 33.3 ± 5.2 kg/m² and waist circumference (WC), 102.8 ± 13.7 cm) aged 50.8 ± 13.8 years, non-diabetic (median glucose, 92 mg/dl and insulin, 12.9 U/ml); 47% were males and 53% females with a mean annual PM₁₀ exposure of 40.12 ± 4.71 µg/m³.

They were moderately hyperlipidaemic with mean total cholesterol (TC) and LDL-C of 213.2 mg/dl and 133.5 mg/dl, respectively. TGs were within normal limits (median 100 mg/dl) (Table 1). These biochemical findings are consistent with a general metabolic normality of the selected population. Mean blood pressure was within normal limits. Out of these patients, 35.6% were on drug treatment, either as single drugs or in combinations: 9.8% were given hypolipidaemic agents, statins being the most prescribed (75.5%); 32.6% were on antihypertensive medications (Supplementary Material Table 1). Application of the FRS prediction model, after correction for lipid-lowering treatments, gave a low median 6.8% 10-year CV risk (3.1% (Q1), 14.2%(Q3)).

Circulating PCSK9 levels were normally distributed, with a mean of 248.7 ± 78.6 ng/ml (Supplementary Material Figure 1). Liver enzymes, i.e. aspartate aminotransferase (AST), alanine aminotransferase (ALT) and Gamma-glutamyl transferase (GGT), and neutrophil, lymphocyte and monocyte blood pattern were all in a normal range, as were C-reactive protein (CRP) levels (median 0.26 mg/l). A pattern of inflammatory cytokines, e.g. tumour necrosis factor (TNF)-α, IFN-γ, interleukin (IL)-18 and C-C Motif Chemokine Ligand 2 (CCL2) known to be affected by PM₁₀ exposure and to be related to the CV risk, was also evaluated (Supplementary Material Table 2).

Univariate data analysis

Circulating PCSK9 levels were positively associated with the average annual (β=1.831, SE = 0.75, p = 0.0144; Figure 1(a)) and six-month (β = 0.734, SE = 0.302, p = 0.0156; Figure 1(b)) PM₁₀ exposure levels. Conversely, short-term exposures were not associated with PCSK9 levels (data not shown). Relative to the main lipid parameters, associated with CV risk, a positive association was found with TC (β=0.47, SE = 0.09, p < 0.0001), LDL-C (β=0.33, SE = 0.10, p = 0.0008), non-HDL-C (β=0.37, SE = 0.09, p < 0.0001), TGs (β=0.11, SE = 0.05, p = 0.0165) and HDL-C (β=0.71, SE = 0.24, p = 0.0034). Inflammatory markers, associated with the initiation and progression of atherosclerosis, were also evaluated. No relationship was found with CRP, inflammatory cytokines, e.g. TNF-α, IL-1β and IL-6, and with the most relevant monocyte recruitment factors, e.g. Macrophage inflammatory protein (MIP)-1α, MIP-1β, and MCP-1. IFN-γ was the only cytokine positively associated with circulating PCSK9 levels (β=1.70, SE = 0.70, p = 0.0155). When glucose metabolism-related variables were examined, only the QUICKI score showed a negative association with PCSK9 (β=–654, SE = 272, p = 0.0167). Finally, the occupational status, categorised into four categories as employee, unemployed, pensioner and housewife, was significantly associated with PCSK9 levels, while no association was observed between the educational status (i.e. primary school or less, middle school, secondary school or university) and PCSK9. The overall results from univariate linear regression models, examining the association between PCSK9 levels and individual characteristics of study subjects, are shown in Supplementary Material Table 3.

Association of PM₁₀ exposure with PCSK9 concentrations at different levels of IFN-γ

The contour plot, reported in Figure 2(a), is the 2D graphical representation of PCSK9 concentration estimates, obtained by a multivariable linear regression model, including PM₁₀, IFN-γ, interaction between PM₁₀ and IFN-γ, age, gender, waist circumference, smoking habit, occupation, use of statins, HDL and non-HDL cholesterol, TGs, QUICKI score, and % lymphocytes. A positive association between PM₁₀ and PCSK9 levels was observed at the lower concentrations of IFN-γ, the only cytokine associated with PCSK9 levels and taken as proxy of the inflammatory status. The strength of this association in terms of PCSK9 observed increments for each µg/m³ rise in PM₁₀, at four fixed levels of IFN-γ (lower limit of quantification (LLOQ)/2: 0.65 pg/ml, mean–SD: 5.02 pg/ml, mean: 10.4 pg/ml and mean ± SD value: 15.78 pg/ml), is depicted in Figure 2(b).

At low concentrations of IFN-γ (0.65 pg/ml), an increase of 3.96 ng/ml of PCSK9 for each µg/m³ of PM₁₀ exposure (β = 3.96, SE = 1.64, p = 0.0160) was observed. A similar association was detected when IFN-γ was equal to 5.02 pg/ml (β = 2.38, SE = 1.09, p = 0.0289). For high levels of IFN-γ (respectively 10.4 and 15.78 pg/ml) the association between PM₁₀ and PCSK9 was instead non significant (β = 0.44, SE = 0.79, p > 0.1; β = –1.50, SE = 1.22, p > 0.1).

We also observed a smaller but still significant effect of a six-month PM₁₀ exposure on PCSK9 when data were adjusted for IFN-γ strata (IFN-γ = 0.65 pg/ml β = 1.44, SE = 0.68, p = 0.0347; IFN-γ = 5.02 pg/ml β = 0.87, SE = 0.46, p = 0.0572; IFN-γ = 10.4 pg/ml β = 0.18, SE = 0.30, p > 0.1; IFN-γ = 15.78 pg/ml β = –0.52, SE = 0.42, p > 0.1). Conversely, short-term exposure lags had a non significant effect (data not shown).

Association between education, occupation and PCSK9 levels

As a secondary analysis, the effect of two major lifestyle variables, i.e. education and occupation, on PCSK9 levels has been evaluated. Multivariable linear regression models, also including age, gender, smoking habits, use of statins, HDL and non-HDL-C, TGs, QUICKI score and %lymphocytes are reported in the Supplementary Material Tables 4 and 5.

Association between PCKS9 levels and FRS

PCSK9 levels were positively associated with the FRS (Figure 3), this latter being raised by 15.9% for each 100 ng/ml increase in circulating PCSK9 levels, Δ% = 15.88, 95% CI: 2.18–31.43, p = 0.0218.

Impact of IFN-γ on PCSK9 expression

To test the hypothesis of a possible direct effect of IFN-γ on PCSK9 gene and protein expression, HepG2 cells were exposed for 24 h to IFN-γ. IFN-γ raised both PCSK9 protein, determined by Western blot analysis, and mRNA expression (Supplementary Material Figure 2).

Discussion

The present study was carried out in 500 subjects, non-diabetic at the time of recruitment, selected among the 2000 overweight (25 < BMI < 30 kg/m²) or obese (BMI ≥ 30 kg/m²) participants of the SPHERE population. It strongly supports the hypothesis that long-term exposure to PM₁₀, i.e. 12 months, is associated with a significant rise in circulating levels of PCSK9, which were also positively associated with the FRS, a validated CV risk score, showing an increase of about 16% for every 100 ng/ml rise in PCSK9 circulating levels.

PM inhalation is an established trigger of CV events,²⁴ which may occur within hours or days after exposure. Indeed, short-term exposure to PM pollution contributes to acute CV morbidity and mortality; exposure to elevated long-term PM levels reduces life expectancy.²⁵ Because of the ubiquitous and involuntary nature of PM exposure, this may steadily enhance CV risk among millions of susceptible individuals worldwide.⁷ Looking specifically at CVD, it has been calculated that air pollution has been responsible for 19% of all CV deaths worldwide in 2016, including 23% of all ischaemic heart disease and 21% of all stroke deaths.²⁶ Within those yet unclear mechanisms linking PM exposure and CVD, the present data may add PCSK9 to the list of the possible elements connecting air pollution to CV risk.

Among the established mechanisms related to the occurrence of CV events, three main pathways should be considered: (a) a pathway involving proinflammatory mediators, generated in the lung and released into the systemic circulation; (b) an alternative pathway by which soluble or smaller components of environmental compounds enter the circulation directly from lungs; (c) a central pathway characterised by perturbation of the autonomic nervous system and/or sympathoadrenal activation.²⁷ In addition, considering a possible activation of the prothrombotic pathway by PM,²⁸ recent evidence indicates a direct effect of PCSK9 on platelet aggregation.¹⁶

The hypothesis that PCSK9 can be directly involved in a series of pleiotropic effects linked to atherogenesis is well supported by observations reporting the expression of PCSK9 in vascular smooth muscles cells, as well as in human atherosclerotic plaques.²⁹ Absence of PCSK9 is associated with a reduced neointimal formation, further supporting the stimulatory effect of PCSK9 on intimal thickening.³⁰

Studies investigating the association between PCSK9 levels and coronary risk have provided contrasting findings. In the earliest evaluation, in patients with stable CAD, a clear correlation between PCSK9 concentration and CV events was noted, despite well-controlled LDL-C; however, after correction for serum TGs, PCSK9 predictivity lost statistical significance.³¹ A somewhat similar finding was reported for a middle-aged sample of Canadian patients followed for seven years. Despite clear correlations between LDL-C, TGs and insulin with PCSK9, the final findings did not support the hypothesis that PCSK9 may be a marker of atherosclerotic risk or vascular health.³² Similarly, in a sample of women followed for 17 years, baseline levels of PCSK9 did not predict the first CV event.³³ In acute coronary syndrome (ACS) patients undergoing coronary angiography, belonging to a Swiss prospective cohort study, PCSK9 was found to be clearly associated with inflammation and hypercholesterolaemia, but did not predict mortality at one year.³⁴ A different conclusion was reached in a large Swedish sample of men and women, 60 years of age at the time of recruitment. After full adjustment for lipids, diabetes, smoking and other risk factors, CV incidence was predicted by PCSK9 baseline levels, with a hazard ratio of 1.48, when considering the lowest vs the highest quartiles.³⁵ In an Italian population, likewise, the occurrence of ACS in patients with severe carotid artery stenosis was predicted by serum PCSK9, levels above 431.3 ng/ml being best associated with a higher risk of ACS occurrence.³⁶

Different from population studies, data from meta-analyses have been fully consistent with an association between CV events and PCSK9 levels. The evaluation of 11 cohort studies, consisting of 13,761 participants, led to the conclusion of a non-linear dose-response relationship between PCSK9 and CV risk, consistent with PCSK9 being an independent risk factor.¹⁹ Similarly, the meta-analysis by Vlachopoulos et al.¹⁸ including nine studies with 12,081 participants, reported a risk increase of 10% for each standard deviation of PCSK9, with a risk increase of 23% in those in the highest tertile. These authors further underlined that the prediction of total CV events was best noted in apparently healthy subjects, rather than in populations with established CV or renal diseases. Finally, a genetic large-scale study of 337,536 individuals of British ancestry reported that the rs11591147 loss-of-function PCSK9 mutation had a protective effect not only on hyperlipidaemia but also on coronary heart disease (–27%) and ischaemic stroke (–39%) risk.³⁷ In agreement with these findings, the largest-to-date genome-wide association study of circulating PCSK9 levels found that, in 3290 patients with CAD, genetic reduction of PCSK9 levels by 50% was associated with a similar percent reduction of CAD risk.²⁰

In the present study, in a similar way, an important role as effect modifier seems to be played by basal inflammation of subjects. The positive association between long-term exposure to PM₁₀ and PCSK9 levels is modulated by IFN-γ concentrations. Indeed, if the association between PM₁₀ and PCSK9 is considered across IFN-γ strata, it becomes clear that the significance of the association is lost when IFN-γ levels are high, i.e. beyond the third quartile of distribution. This loss of PM effect could be well explained by the fact that PCSK9 is upregulated by inflammation, as previously shown by our group,³⁸ and is confirmed by the in vitro experiment, showing that IFN-γ can raise both PCSK9 gene and protein expression (Supplementary Material Figure 2). Thus, when upregulation of the two markers is already maximal, due to the basal level of inflammation, PM₁₀ exposure effects may become negligible. This observation may be of clinical significance, indicating a possible increased sensitivity to PM in subjects with a low-normal inflammatory pattern, e.g. children and out-door athletes.³⁹ Moreover, the lack of association we have found between PCSK9 circulating levels and classical markers of inflammation, e.g. CRP and TNF-α,⁴⁰ is not surprising since, so far, contrasting data have been reported on this topic.^14,33,41 Overall, this possible explanation of the observed results needs to be strengthened in future mechanistic investigations. In the current study a formal statistical assessment of inflammation, i.e. IFN-γ as mediator of PM₁₀ effect on PCSK9 is not feasible, as the assumptions needed to identify the indirect and direct effects as causal effects are not satisfied and therefore conclusions from a mediation analysis would lead to invalid results.

A further feature possibly explaining the loss of association between PCSK9 and the inflammatory pattern may rely on the epigenetic modifications driven by exposure to air pollutants. Hypermethylation of the IFN-γ gene can be found concomitant to parasympathetic activation (manifested by reduced heart rate variability), in PM exposed individuals. This finding may indicate a different activation status of IFN-γ. While PM may increase CV risk over the long term, i.e. months to years of exposure, IFN-γ acutely elicits numerous adverse biological responses (e.g. systemic inflammation) further increasing CV risk. Hypermethylation may be associated with both a reduced and a raised inflammatory status, depending on the level of PM exposure and sensitivity of the exposed population.⁴² On the other hand, we provided new evidence that IFN-γ directly induces and releases PCSK9 from the hepatocyte cell line HepG2. This may explain the lack of a direct association between PM₁₀ and PCSK9 in subjects with higher levels of IFN-γ.

Moreover, in our population sample, as seen from the univariate analysis, circulating levels of PCSK9 are related to a large number of CVD risk factors. All of these observations are consistent with those previously reported in different observational studies^14,43-45 in which a direct association between circulating PCSK9 concentration, age, total cholesterol, LDL-C, non-HDL cholesterol and TGs was found.

Application of the FRS has, at times, been fraught with perplexities when applied to populations outside of the USA. The FRS is based on risk factor categories for CAD, as retrieved from the original cohort of Framingham, examined from 1971–1974 and attending the 11th examination at the 12-year follow-up. This multivariate CAD risk has been used on millions of patients, being available in an appropriate software.⁴⁶ Subsequent evaluations of the FRS allowed updating, thus expanding the profile of risk up to older ages, i.e. >45 years for men and >55 years for women.⁴⁷ For the same risk function, an Italian Score predicted a far lower event rate in 10 years, indicating that, most likely, the application of the FRS in an Italian population samples with a low CV risk should be used with caution.⁴⁸ More recently, the CUORE prediction algorithm, also developed in the Italian population⁴⁹ indicated a higher predictivity vs the FRS, but successful application is best achieved in a high-risk population. Conversely, for a low-risk population, such as that of the present study, higher reliability of the FRS vs the CUORE algorithm has been reported.⁵⁰ In addition, the positive association between PCSK9 and the FRS is in line with findings describing how, in a cohort of asymptomatic subjects without a history of CVD and a low value of FRS, this latter was related to PCSK9.⁵¹

The present study conducted on a large number of subjects has, however, some limitations. First, the study population itself may be considered as unusual. If the choice of investigating obese subjects offers the possibility to a focus on a high CVD risk population, it poses a challenge due to the many comorbidities that need to be excluded. In order to limit this possible problem, we decided to select a priori the healthiest subjects (n = 500) and we further corrected for possible confounders.

Moreover, since the kinetics of association between PM₁₀ and PCSK9 and between PM₁₀ and cardiovascular parameters were different (i.e. the effects were respectively observed for long-term and short-term exposures), it was not possible to statistically evaluate if PCSK9 can be an intermediate variable for the association between PM exposure and CV risk.

Moreover, the SPHERE population is characterised by a well-defined PM₁₀ exposure. We used PM₁₀ instead of PM_2.5 as the air pollutant of choice, because the PM₁₀ dataset was more complete and can be identified by better spatial resolution. However, in the study area, PM₁₀ is mainly constituted by fine particles, and PM_2.5 represents 58–94% of PM₁₀.⁵² A third limitation is the lack of personal exposure monitoring, due to the large study sample, that hampered the possibility of accounting for indoor air pollution.⁵³ This type of approach, however, is recognised for large epidemiological studies, having been previously used in a number of similar investigations.²²

Conclusions

The evaluation of CV risk, occurring in association with the environmental exposure to PM is accompanied by changes in risk biomarkers.⁵⁴ The present investigation examined 500 obese individuals, known to be more sensitive to the damaging effects of environmental PM exposure. The major finding was that individuals with the lowest expression of inflammatory markers, particularly IFN-γ, appeared to be most sensitive to PM₁₀ exposure, as expressed by raised PCSK9 concentrations. These appeared to be associated with a significant rise of the FRS, a well-established predictor of CV risk. This conclusion indicates that raised PCSK9 levels are associated to PM₁₀ exposure and that individuals with a low expression of inflammatory markers may be most sensitive to the potential CV damaging effects of environmental PM exposure.

Author contribution

CM performed the ELISA assays and wrote the manuscript; NF conceived the results reported in the Supplementary Material Figure 3 and critically revised the manuscript; CF performed all the statistical analyses; LC performed the determination of particulate matter exposure; ACP, CRS and AC critically revised the manuscript; MGL performed the experiments presented in the Supplementary Material Figure 3; LV visited patients; VB and MR conceived the study and wrote the manuscript. All gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: CM, NF, CF, LC, LV, ACP, MGL, CRS, VB and MR declare that there is no conflict of interest. AC has received honoraria from AstraZeneca, AMGEN, Sanofi, Recordati, Novartis, MSD, Mediolanum, DOC, Mylan and Pfizer.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was supported by Fondazione Cariplo 2015-0552 (MR) and EU Programme ‘Ideas’, European Research Council (ERC-2011-StG 282,413 to VB).

References

Author notes