Human haemodynamic frequency harmonics regulate the inflammatory phenotype of vascular endothelial cells

Abstract

Haemodynamic variations are inherent to blood vessel geometries (such as bifurcations) and correlate with regional development of inflammation and atherosclerosis. However, the complex frequency spectrum characteristics from these haemodynamics have never been exploited to test whether frequency variations are critical determinants of endothelial inflammatory phenotype. Here we utilize an experimental Fourier transform analysis to systematically manipulate individual frequency harmonics from human carotid shear stress waveforms applied in vitro to human endothelial cells. The frequency spectrum, specifically the 0 th and 1st harmonics, is a significant regulator of inflammation, including NF-κB activity and downstream inflammatory phenotype. Further, a harmonic-based regression-model predicts eccentric NF-κB activity observed in the human internal carotid artery. Finally, short interfering RNA-knockdown of the mechanosensor PECAM-1 reverses frequency-dependent regulation of NF-κB activity. Thus, PECAM-1 may have a critical role in the endothelium's exquisite sensitivity to complex shear stress frequency harmonics and provide a mechanism for the focal development of vascular inflammation.

Figure 1. Carotid artery atheroprone and atheroprotective shear stress waveforms evoke distinct endothelial phenotypes

(a) Previously blood flow velocities (U_z) from the internal and common carotid arteries were measured. Using this information, the haemodynamic parameters wall shear stress (τ_w), oscillatory shear index (OSI) and the harmonic index (HI) were computed between these different regions circumferentially around the blood vessel. (b) Atheroprone and atheroprotective shear stress waveforms from the human internal carotid sinus (ICS) and common carotid artery (CCA), respectively, were applied in vitro to confluent ECs for 24 h. Pro-inflammatory proteins (c) and gene expression (d) were elevated under atheroprone flow, while the opposite was seen with anti-inflammatory genes (e). Statistics: *P < 0.05 for Student’s one-sample t-test, N = 4–10.

Figure 2. Atheroprone and atheroprotective frequency spectrums regulate NF-κB activity

(a) Fourier transform analysis was performed on the protective and prone waveforms. (b) The corresponding 0th harmonic (time-average shear stress) from the prone (low 0th) and protective (high 0th) was held constant, while exchanging the higher frequency spectrum from the prone or protective waveforms, or compared with steady flow lacking harmonic content. (c) After 24 h of shear stress, NF-κB activity was assessed. Statistics: *P < 0.05 for one-way ANOVA, N = 8–10, ^#P < 0.05 for Student’s t-test compared with equivalent high 0th control. N = 4–10.

Figure 3. Shear stress frequency harmonics contribute to inflammatory phenotype

(a,b) Frequency harmonic content was swapped between control waveforms using Fast Fourier transforms (FFT) to create new ‘mutated’ waveforms for harmonics 1–8. (c,d) The effect of mutating frequency harmonics on NF-κB activity was assessed, along with proinflammatory gene E-Selectin and anti-inflammatory gene, KLF2. Statistics: *P < 0.05 for ANOVA (c,d), N = 4–7.

Figure 4. Regulation of gene and protein expression by 1–2 harmonic mutation in prone and protective shear stresses

Pro- (a,c) and anti-inflammatory (b,d) genes were assessed after 24 h of control or 1–2 harmonic mutation shear stress. (e) VCAM-1 and FN protein was decreased due to frequency exchanges within the prone waveform. Statistics: *P < 0.05 to control paired waveform, Student’s t-test (a,b,e) or one-sample t-test (c,d), N = 4–7.

Figure 5. Amplitude of the 0th and 1st harmonics regulates NF-κB activity

The effect of shear stress harmonic mutations (a) to the prone 1st (M1), 2nd (M2) or both (M1–2) harmonics on NF-κB activity (c) was compared with the prone control (CTR). Dose increases in the 0th or 1st harmonics in the prone or protective waveforms (b) were selectively generated and applied to ECs for 24 h and NF-κB activity was assessed (d). Statistics: *P < 0.05 one-way ANOVA, N = 4.

Figure 6. A Predictive model for NF-κB activation under physiologic flows

Using 20 different shear stress waveforms, a predictive regression model based on haemodynamic frequency content was generated. (a) The final model of NF-κB activity was able to fit the experimental data. (b) From the final model, a generated colour map shows NF-κB activation as a function of the 0th and 1st harmonic amplitude. (c) Shear stress velocity profiles were measured around the circumference of the carotid sinus via MRI, and the velocity profile was used to calculate wall shear stress around the circumference of the vessel over a cardiac cycle (d). (e) These shear stress waveforms were applied to ECs in vitro and NF-κB activity was assessed (coloured bars) along with the amplitude of the 0th or 1st harmonic (solid lines) at each location. (f) To validate the predictive accuracy of the regression model, harmonic content from these five nodes were used as the input to the regression model to get predicted NF-κB, which was compared with the experimental data. Statistics: *P < 0.05 compared with Node 8, one-way ANOVA; ANOVA regression analysis was used to determine P-value for correlation between experimental and predicted NF-κB activity in a and to show that the identity line falls within the 95% confidence interval in f, N = 5.

Figure 7. PECAM-1 mediates shear stress frequency-dependent NF-κB response

NF-κB activity (a) and gene expression (b) were assessed after 24 h of prone control (CTR) or 1–2 harmonic (M1–2) mutated shear stress for control-treated (siControl) or PECAM-1-depleted (siPECAM-1) ECs. (c) PECAM-1 tyrosine phosphorylation levels were measured relative to static conditions using protective, prone or mutated prone shear stresses. Statistics: *P < 0.05 Student’s one-way ANOVA (c), two-way ANOVA (a,b), ^#P < 0.05 one-sample t-test compared with prone control. NS, nonsignificant.