Fractalkine (CX3CL1), GM-CSF and VEGF-a levels are reduced by statins in adult patients

Our research protocol was reviewed and approved by the University at Buffalo Intramural Review Board for Health Sciences research (Approval Number: MED5980509B). Informed consent to undergo the study protocol was obtained in writing from each participant according to the principles expressed in the Declaration of Helsinki. Our study was organized as an observational trial as no direct health outcomes were to be measured.

The study population consisted of 12 adult subjects (7 males and 5 females) with no active medical problems. Study subjects were screened for the absence of chronic health disorders including hypercholesterolemia with additional cardiovascular risk factors, cancer, diabetes, chronic liver or kidney disease[13]. The baseline characteristics of the study population were reported previously[13] and are shown in Table 1. The age of the cohort was 43.4 ± 12.5 years, had a body mass index of 24.9 ± 7.2, and were at low risk for atherosclerotic disease events with a Framingham Risk Score of 1.7 ± 0.5. Prior to statin treatment the mean total cholesterol level was 210.5 ± 27.6 mg/dL, LDL cholesterol 136.2 ± 22.9 mg/dL, HDL cholesterol 53.5 ± 12.9 mg/dL. The cohort had a low index of inflammation as the C-reactive protein level was 1.1 ± 1.3 mg/L. Two of the study subjects had treated hypertension. None of our study subjects had disease states known to increase CX3CL1 levels including common variable immunodeficiency[14], granulomatosis[15], congestive heart failure[16], and did not smoke[17].

Study subjects were randomized to drug regimen groups using a block randomization design. The investigators were blinded to which treatment subjects were receiving. The workflow of the experimental protocol is summarized in Figure 1. Study subjects underwent a baseline blood draw in which a complete blood cell count, lipid panel (total, HDL, and LDL cholesterol, and triglycerides) and C-reactive protein were determined by the Kaleida Health pathology laboratory. The plasma fraction was also retained from each blood draw and frozen at -80°C for chemokine and cytokine assays. Subjects were then treated for two weeks with one of three different HMG-CoA reductase inhibitors (pravastatin 80 mg daily, atorvastatin 80 mg daily, or rosuvastatin 10 mg daily). At the end of the two-week statin treatment, venous blood was sampled again to obtain the lipid panel, C-reactive protein level, and blood samples for cytokine analysis. Subjects were subsequently given a four-week statin free period. At the end of the four week statin-free period, venous blood was sampled again to determine if serum lipids returned to within 5% of their pre-statin levels. In subjects where cholesterol levels did not recover to within 5% of the baseline lipid levels, an additional four-week statin free period was provided before resuming statin therapy to avoid effects of overlap between drugs. Three study subjects required extension of the statin free period for an additional four weeks for serum lipid levels to return to within 5% of their baseline lipid levels. Following this period, the next HMG-CoA reductase inhibitor in the randomization scheme was given for two weeks. The same protocol was repeated for statin drugs two and three until study completion. All twelve subjects completed treatment with the three statin medications.

Plasma levels of fractalkine (CX3CL1), GM-CSF, FLT3 ligand, GRO alpha, Interleukin-3, Interleukin-6, Interleukin-8, macrophage inflammatory protein-1 (MIP-1/CCL3), monocyte chemotactic protein-1 (MCP-1), stromal derived growth factor-1 (SDF-1), and VEGF-A were measured with a Luminex Human Cytokine assay (Millipore) per the manufacturer’s instructions by the Roswell Park Cancer Institute Laboratory of Flow Cytometry. For Luminex based assays, known concentrations of CX3CL1, GM-CSF, and VEGF-A were measured in triplicate to generate standard curves using a non-linear five-parameter curve fit using the nCal package in R[18]. The reliable lower assay limits for each were: CX3CL1-39.5 pg/mL, GM-CSF 3.02 pg/mL, VEGF-A 36.6 pg/mL. The inter-assay %CV for CX3CL1 was 5.13, and the intra-assay %CV for CX3CL1 was 5.85. Plasma cytokine levels from human subjects before and after statin treatment were measured in singulate. Interleukin-17 levels were measured with an ELISA (R&D Systems).

Human umbilical vein endothelial cells (Lonza, passages 3–5) were grown to 80% confluence in EGM-2 medium (Lonza) on type I collagen coated culture dishes. Where indicated, cells were incubated for 18 hours in human LDL cholesterol (100 μg/mL; Biomedical Technologies, Stoughton, MA, USA) and/or atorvastatin (100 nM; Sigma Scientific). TNFα (1 μg/mL; R&D Systems) was then added for an additional 18 hours after LDL cholesterol or atorvastatin treatment where indicated. Cells were then washed in warm PBS and lysed in QIAzol solution (Qiagen), and RNA was extracted following the manufacturer’s protocol. RNA quality was assessed by OD260/280 using a Nanodrop spectrophotometer. Total RNA was reverse transcribed using SuperScript III Reverse Transcriptase with oligo-dT primers (Invitrogen). Real-time PCR analysis was performed using SYBR green and Taq DNA polymerase (Qiagen) on a Bio-Rad CFX96 Connect Real-Time PCR Detection System. PCR primers for CX3CL1, GM-CSF, ICAM, VCAM, CX3CR1, and VEGF-A were obtained from Qiagen. RT-PCR data was normalized to GAPDH expression to identify relative changes in transcript levels.

Regression analyses were tested for significance using Pearson’s correlation. Heat map representations of RT-PCR data were compiled using R. Significant differences in CX3CL1, GM-CSF, and VEGF-A plasma levels between statin therapies were assessed using a non-parametric Friedman’s two-way ANOVA by ranks with a Bonferroni correction for multiple comparisons. Differences in CX3CL1, ICAM, VCAM-1, and VEGF-A transcript levels in response to LDL cholesterol incubation or TNFα versus untreated cells were measured by a one-way ANOVA with a Bonferroni correction for multiple comparisons. RT-PCR data for GM-CSF was analyzed by one-way ANOVA with a Games-Howell post-hoc test due to unequal variances between treatment groups. Numerical data stated in the manuscript text represent mean ± standard error of the mean. Statistical analysis was performed using SPSS software.

To understand whether CX3CL1 levels vary with cholesterol levels or other factors in humans, we assessed CX3CL1 levels in human subjects without known coronary disease and used HMG-CoA reductase inhibitors (statins) of varying potencies to modulate cholesterol levels. The baseline characteristics of the study cohort were published previously[13], and are shown in Table 1. Baseline serum lipids revealed an average total cholesterol of 210.5 ± 27.6 mg/dL (range 168 to 282 mg/dL) and an average LDL cholesterol of 136.2 ± 22.9 (range 114 to 205). The total and LDL cholesterol levels spanned the normal and hypercholesterolemic range. Therapy with statins reduced total and LDL cholesterol levels as expected, demonstrating appropriate drug effect[13]. CX3CL1 levels were significantly reduced from no statin control levels (89.9 ± 18.5 pg/mL) after treatment with atorvastatin (60.0 ± 7.8 pg/mL), pravastatin (54.2 ± 7.0 pg/mL) and rosuvastatin (65.6 ± 7.3 pg/mL) (χ²(2) = 17.4, p ≤ 0.001; see Figure 2), indicating statins downregulate CX3CL1 levels in adult human subjects.

We next sought to elucidate factors regulating CX3CL1 after statin therapy. CX3CL1 is induced by inflammatory cytokines including interferon-γ, interleukin-6, and TNFα[5]. Cholesterol however, is not a known regulator of CX3CL1. We surveyed the levels of inflammatory cytokines in the plasma of our cohort before and after statin therapy. We found FLT3 ligand, Interleukin-3, Interleukin-6, and Macrophage inflammatory protein-1 (MIP-1/CCL3) were undetectable in most subjects, indicating a lack of typical CX3CL1 inducers in our cohort of asymptomatic human subjects. We identified correlations between CX3CL1 plasma levels and granulocyte macrophage-colony stimulating factor (GM-CSF), VEGF-A, stromal cell-derived factor-1/CXCL12 (SDF-1), GROα/CXCL1, monocyte chemotactic protein-1/CCL2 (MCP-1), interleukin-8/CXCL8 (IL-8), neutrophil counts, and cholesterol levels. We performed regression analysis to test significant correlations with CX3CL1 levels in human blood (Table 2). We found that GM-CSF (r² = 0.524; p < 0.005; Figure 3) and VEGF-A (r² = 0.4; p < 0.005; Figure 4) levels were highly and positively correlated with CX3CL1. To explore why GM-CSF and VEGF-A were significantly correlated with CX3CL1 levels, we evaluated if statins decreased the concentrations of GM-CSF and VEGF-A, and if GM-CSF and VEGF-A levels correlated with total, LDL, or HDL cholesterol values. We found that all statin therapies significantly decreased both GM-CSF (Figure 5) and VEGF-A levels (Figure 6), but there was no significant correlation between GM-CSF or VEGF-A levels with total, LDL, or HDL cholesterol. These findings suggest that the effect of statins on GM-CSF and VEGF-A levels are more likely driven by suppression of inflammatory signaling rather than a reduction in cholesterol levels.

SDF-1/CXCL12, LDL cholesterol, MCP-1/CCL2, total cholesterol, and IL-8/CXCL8 levels all showed statistically significant associations with CX3CL1 levels. GROα/CXCL1, circulating neutrophil levels, and HDL cholesterol levels (Figure 7) were negatively correlated with CX3CL1 levels. We did not find significant correlation between CX3CL1 levels and C-reactive protein, IL-17, CD34+ HSPC levels, VEGF-C or G-CSF. Interestingly, the shared biological function of GM-CSF, VEGF-A, SDF-1, GROα, and IL-8 lies in regulation of chemotaxis and response to inflammation. These findings suggest that CX3CL1, along with several other factors involved in response to inflammation and chemotaxis of inflammatory cell types are modulated by statin therapy.

To determine whether CX3CL1, GM-CSF and VEGF-A levels are modulated by LDL cholesterol itself or by inflammatory cytokines, and how statins may affect CX3CL1, GM-CSF and VEGF-A expression, we incubated human umbilical vein endothelial cells with loaded with LDL cholesterol with or without atorvastatin (100 nM) for 18 hours. Then TNFα was added for an additional 18 hours. Real-time PCR analysis was performed to measure changes in the levels of CX3CL1, GM-CSF, and VEGF-A. ICAM and VCAM-1 are induced by TNFα treatment of endothelium via an NFκB dependent mechanism[19], and were positive experimental controls. Experimental results are summarized in Figure 8 showing log 2 fold changes in gene expression relative to untreated endothelial cells. We found that LDL cholesterol alone resulted in small increases in CX3CL1, ICAM1, and VCAM1, but were not significant due to a high degree of variability between experimental replicates. GM-CSF expression was modestly decreased by incubation with LDL cholesterol. LDL cholesterol incubation alone caused modest but significant increase in VEGF-A (0.1 fold increase versus untreated cells; p ≤ 0.05). TNFα treatment resulted in > 10-fold increases in transcripts of CX3CL1 (99.0 ± 0.11 fold vs. untreated; p < 0.0005), and GM-CSF (106.1 ± 1.3 fold vs. untreated; p < 0.0005). Positive controls for TNFα treatment, ICAM1 (39.7 ± 0.06 vs. untreated; p < 0.0005) and VCAM1 (82.8 ± 0.53 vs. untreated; p < 0.0005), also showed > 10-fold induction as expected. In contrast, TNFα treatment resulted in minor increases in VEGF-A (0.29 ± 0.01 vs. untreated; p < 0.006) transcript levels. Incubation with TNFα and LDL cholesterol together resulted in modest, but not significant, augmentation of expression of CX3CL1, GM-CSF, VCAM1, ICAM1, and VEGF-A compared with TNFα treatment alone. Finally, addition of atorvastatin (100 nM) significantly reduced induction of ICAM1, VCAM1, CX3CL1 and GM-CSF by both LDL cholesterol and TNFα. In conclusion, LDL cholesterol alone did not substantially augment CX3CL1 or GM-CSF transcript levels in human endothelial cells in vitro. TNFα coordinately increased transcript levels of CX3CL1 and GM-CSF, and the addition of LDL cholesterol modestly increased both transcripts, while atorvastatin significantly reduced CX3CL1 and GM-CSF expression. The findings support a model where inflammatory cytokines intersect with cholesterol levels to augment expression of a module of inflammatory response factors including CX3CL1 and GM-CSF, and statins inhibit their induction.