56 Computational Analysis of the Effects of Nitric Oxide on Mouse Arteriovenous Fistulas

Faculty Mentor: Yan-Ting Shiu (Internal Medicine, University of Utah)

Introduction

Chronic kidney disease (CKD) is a rising concern for Americans, as 37 million Americans are affected by CKD [1]. When CKD advances into its fifth and final stage, it is renamed as end-stage kidney disease (ESKD). ESKD is classified with a glomerular filtration rate below 15 mL/min [2]. At this stage, the patient needs to undergo renal replacement therapy (RRT). A kidney transplant is the most desirable RRT; however, obtaining a suitable donor may take 3-5 years and most patients need hemodialysis while they wait for a suitable kidney donor [3].

Hemodialysis utilizes a dialyzer to filter blood, which requires a vascular access. The arteriovenous fistula (AVF) is the gold standard for vascular access due to its low thrombosis and infection rates compared to other vascular access methods. [4], [5], [6]. The AVF is a surgical connection between a native artery and vein. For the AVF to be usable for dialysis, it must mature, a process where the vein adapts structurally and mechanically to the increased blood flow from the arterial connection. The vein’s lumen expands, and its walls thicken to withstand high-pressure, high- volume blood flow, enabling repeated needle insertion for dialysis. [7]. AVFs are considered mature when the lumen has sufficiently expanded to accommodate the high blood flow rate required for optimal dialysis [6].

The primary cause of morbidity within ESKD patients is failed AVF maturation, as up to 60% of AVFs fail to mature [8]. AVF maturation failure is caused by insufficient lumen expansion and aggressive inward vascular wall remodeling known as neointimal hyperplasia (NH) [9]. Neointimal hyperplasia originates from the migration and proliferation of smooth muscle cells (SMCs) from the tunica media to the tunica intima layer [10]. Luminal stenosis causes a decreased blood flow rate, which leads to an unusable AVF for dialysis. To combat NH, SMC relaxation is necessary to prevent further cellular proliferation into the lumen [11]. SMC relaxation can be achieved by using vasodilators and vessel dilation.

The use of vasodilators is a promising strategy to promote lumen expansion and ensure the high blood flow needed for dialysis. Previous studies have demonstrated the efficacy of vasodilators like phosphodiesterase inhibitors (PDE5) and nitric oxide (NO) in aiding AVF maturation. [8], [9], [12], [13]. These studies show that AVFs treated with vasodilators exhibit improved hemodynamics, with the endothelial nitric oxide synthase (eNOS)-NO system being particularly effective. NO derived from eNOS diffuses from endothelial cells into SMCs, where it activates soluble guanylate cyclase (sGC), increasing cyclic GMP (cGMP) levels. cGMP activates protein kinase G (PKG), which leads to reduced intracellular calcium levels. This reduction in calcium inhibits SMC contraction and allows the vessel to relax and dilate. [14], [15]. This process facilitates lumen expansion and reduces inward cellular proliferation, countering neointimal hyperplasia.

The previous studies mentioned above primarily used computational fluid dynamics (CFD) to analyze the effects of the eNOS-NO system on AVF hemodynamics. CFD has proven to be an effective analytical method to surmise and realize patterns of favorable AVF hemodynamics; however, CFD is incapable of modeling physiologically accurate vessels because of its rigid-wall assumption [16]. However, vascular wall dynamics play a critical role in AVF maturation and must be characterized. Since NO affects the vessel walls and causes vasodilation, CFD alone is insufficient for capturing these effects. This study will use fluid-structure interactions (FSI) to model psychologically accurate vessels. Modeling AVF wall mechanics with FSI can describe the effects of the eNOS-NO system on AVF hemodynamics and wall mechanics. Fluid-structure interaction simulations in the context of vascular hemodynamics analyze the interaction of blood with the AVF vessel walls. By using FSI simulations, new wall parameters such as von Mises Stress and Strain, and wall displacement can be realized that were not possible before. These new parameters can help identify wall mechanics that are favorable to AVF maturation and can help clinicians make informed decisions by improving our database of hemodynamic and vascular wall data.

Background

Hemodialysis requires a vascular access site, and vascular access dysfunction is the main cause of mortality and morbidity among hemodialysis patients. [11]. Although the AVF is the preferred vascular access method, it suffers from high maturation failure rates. One of the main causes of AVF maturation failure is excessive inward remodeling (i.e., formation of NH) in the venous limb of the fistula. When the AVF is surgically created, the artery exposes the vein to a high pressure, high shear stress, and turbulent flow environment. The aberrant blood flow leads to impaired eNOS and decreased nitric oxide (NO) release, which incites activation of inflammatory, thrombotic, and proliferative genes, resulting in the formation of NH. Our group has explored the possible mechanisms of venous NH. After the creation of the AVF, the high wall shear stress acts upon venous endothelial cells. As a result, the extracellular matrix is exposed to blood flow, causing a cascade of monocyte activation that eventually leads to the proliferation of ECM components into the venous lumen in the AVF. [17].

Successful AVF maturation requires outward vascular wall remodeling of the vein to exceed inward remodeling. Outward remodeling requires the relaxation of SMCs, which dilates the AVF lumen to increase the AVF lumen area, thereby increasing blood flow [9]. To encourage outward remodeling, a vasodilator can be used to relax the SMCs. NO, a natural and potent vasodilator found in the body, is synthesized as a byproduct in the endothelial cells of the intimal layer through the eNOS-catalyzed conversion of L-arginine to L-citrulline. NO diffuses from endothelial cells in the intimal layer to SMCs in the medial layer, leading to SMC relaxation [12]. However, the high wall shear stress and aberrant flow from AVF creation can repress eNOS and decrease NO production [17]. Our group has already shown that overexpressing the eNOS gene leads to favorable hemodynamics in mice (i.e., smoother blood velocity streamlines) [12]. What remains to be investigated are the wall mechanics of these genetically modified mice.

Nitric oxide (NO) is a relatively recent discovery in the biomedical field. Identified in the 1980s as a key component of endothelium-derived relaxing factor (EDRF), NO significantly advanced the understanding of vascular tone regulation [18]. Its role in mediating vasodilation and contributing to blood pressure control marked a major breakthrough in vascular biology. Since then, NO has been widely studied in the context of cardiovascular health due to its critical functions in reducing platelet aggregation, limiting vascular inflammation, and preventing the development of atherosclerosis [19]. The significance of NO in vascular biology and cardiovascular medicine was recognized when Robert F. Furchgott, Louis J. Ignarro, and Ferid Murad won the 1998 Nobel Prize in Physiology or Medicine [20].

Previous studies have used CFD to simulate blood flow and have led to important discoveries of different hemodynamic parameters that predict successful vs. failed AVF maturation [11]. CFD can be explained as the set of techniques that provide a numerical solution to fluid flow simulations [21]. The working principles of CFD involve discretizing the governing equations of fluid motion into a solvable form over a computational domain. The Navier-Stokes system of equations is at the core of fluid mechanics. These equations describe the conservation of mass, momentum, and energy and have the power to realize the effects of viscosity, pressure, and other external forces. [21]. The Navier-Stokes system of equations is are non-linear partial differential equation derived from Newton’s second law of motion. [22].

Because of the numerical techniques required, a common path to solving the Navier-Stokes system of equations is by domain discretization into smaller volumes or elements. These methods are called finite volume and finite element, respectively. This discretization allows the continuous equations to become iterative algebraic equations that can be solved via computational methods. [22].

However, CFD assumes rigid and immobile blood vessel walls. The rigid wall assumption, which simplifies the simulations drastically and thus does not require a lot of computational time, is a reasonable tool as a starting point to investigate hemodynamics in an AVF. However, it is not physiologically accurate because the blood vessel wall is deformable and compliant. Because of this limitation of CFD, fluid-structure interaction (FSI) analysis has received more interest and attention due to its ability to mimic physiological conditions. [23]. With FSI, AVF compliance can be modeled, and thus wall biomechanical parameters can be characterized. My study will focus on comparing eNOS OE mice to WT control mice using hemodynamic parameters (velocity, vorticity, wall shear stress (WSS), oscillatory shear index (OSI)) and the new wall biomechanical parameters (wall displacement, Von Mises stress, and Von Mises strain).

Methods

Mouse model

Mice on the C57BL/6 background were obtained from Jackson Laboratories, Bar Harbor, ME [8]. End-to-side carotid-jugular AVFs were created in 3-4-month-old male mice. [12]. The process of overexpressing the human eNOS gene in the overexpression mice is detailed in previous publications. [24], [25].

All animal studies and experiments were approved by the University of Alabama at Birmingham (UAB) Institutional Animal Care and Use Committee and by the National Institutes of Health guidelines.

MRI acquisition

MRI acquisition has been previously described. Briefly, a 2D time-of-flight angiography sequence, a 2D T2-weighted fast spin echo sequence, and a 2D gradient echo velocity mapping were performed using a 9.4 Tesla Bruker® Biospec horizontal 20 cm bore instrument (Bruker Biospin, Billerica, MA) [12], [26]. 2D time-of-flight angiography sequence and the 2D T2-weighted fast spin echo sequence were used in the vessel reconstruction. In order to visualize the vessel, black-blood double inversion preparation was used to reduce signal from the blood vessels [26]. In our experiment, two WT and two eNOS OE mice were used.

AVF lumen reconstruction and meshing

Lumen reconstruction was performed in Amira 3D 2021.1 (Thermo Fisher Scientific, Waltham, MA) using the 2D black-blood MRI as previously described. [26]. An STL was exported into VMTK (available at: www.vmtk.org) to add cylindrical flow extensions to stabilize the blood flow into the reconstruction and prevent entrance effects [12]. This new geometry was then exported into ANSYS ICEM CFD 2021 R2 (Ansys, Inc., Canonsburg, PA) and was discretized into 1-1.5 million tetrahedral meshing elements. These meshing elements enabled numerical solutions to be obtained for the governing differential equations.

Velocity mapping

The 2D gradient echo velocity (cine-phase contrast) MRI was used to characterize the velocity in the mouse AVF. Blood flow extraction was performed in the proximal artery and proximal vein using ImageJ (https://imagej.nih.gov/ij/).

FSI simulation

FSI simulation was performed by coupling Ansys Fluent 2023 R1 and Ansys Mechanical 2023 R1 in Ansys System Coupling 2023 R1 (Ansys, Inc., Canonsburg, PA). The mesh from ICEM CFD was exported into Ansys Fluent 2023 R1 for fluid analysis, while an STL of the mesh file was exported into Ansys Mechanical 2023 R1 to generate a computer-aided design (CAD) surface for structural analysis. Fluid analysis has been previously described, assigning blood with a density of 1,050 kg/m³ and a dynamic blood velocity of 0.0035 Pa*s [26]. Convergence criteria were set to 0.001 for x-,y-, and z-residuals and 0.001 for total residual. Simulations were performed for two cardiac cycles. The mechanics portion of the simulation was highly dependent on the fluid portion of the simulation.

Postprocessing

The hemodynamic parameters were extracted as previously described. [12]. Tecplot 360 EX 2022 R1 (Tecplot, Inc., Bellevue, WA) was used to extract the hemodynamic and wall mechanical parameters. The hemodynamic parameters extracted were velocity magnitude and wall shear stress (WSS) magnitude. The equations used to describe these parameters have been previously described and are detailed below. [8], [12].

The wall parameters extracted were von Mises stress, von Mises strain, and wall displacement magnitude. Equations are specified below.

where σ1, σ2, σ3 are the principle stress components.

where ε1, ε2, ε3 are the principle strain components and v’ is the effective Poisson’s ratio.

Statistical Analysis

Statistical analysis was performed in Excel 2023. To test for significance, unpaired t-tests were applied, with significance set at p < 0.05. Data is presented as mean ± SEM and graphs are presented as box and whisker plots.

Results

Data represented as mean ± SEM

Fig. 1 shows that the WT strain has more turbulent velocity streamtraces than its eNOS OE counterpart. The velocity streamtraces in the eNOS strain are smoother and more uniform throughout the whole AVF. Similarly, the WSS contour plots show excessive WSS at the anastomosis in the control mouse. For the wall parameters, the contour plots show an increased Von Mises Stress in the eNOS strain. Notably, Von Mises Stress in the eNOS strain is more prevalent in the anastomosis region than the WT strain. Von Mises Strain follows a similar trend to the Von Mises Stress. Wall Displacement also shows an increase in the eNOS strain, however, most of the Wall Displacement occurs past the 4 mm region.

Figure 2 shows the statistical analysis of hemodynamic/wall parameters between the WT strain and the eNOS strain via box and whisker plots. The data was averaged 4 mm from the anastomosis where most stenosis is known to occur. For the hemodynamic parameters, there was statistical significance between the velocity in each strain. The velocity in the eNOS strain was significantly lower than the velocity in the WT strain (p < 0.0001). WSS also showed a significant decrease from WT to OE (p < 0.001). The eNOS strain also showed a significant decrease in vorticity from the WT strain (p < 0.001). Surprisingly, the eNOS strain showed an increase in OSI compared to the WT strain (p < 0.0001).

For the wall parameters, the eNOS strain showed a significant increase in Von Mises Stress and Strain, with both p values being less than 0.0001. Wall displacement, on the other hand, showed no significant changes between each strand.

Discussion

Currently, there is no cure to end-stage kidney disease (ESKD) and patients rely on a renal replacement therapy—hemodialysis being the most common—to meet their filtration needs. Hemodialysis needs a vascular access in order to function, but the main cause of morbidity in ESKD patients is vascular access failure. By analyzing hemodynamics and vascular wall function in genetically modified murine arteriovenous fistulas (AVFs) using fluid-structure interaction (FSI) simulations, we realized new parameters that potentially predict favorable vascular remodeling. These novel parameters can potentially be used in clinical applications to help predict vascular access failure and intervention. The fluid-structure simulations yielded similar results to our group’s previous CFD studies. Making a definitive conclusion about the new mechanical wall parameters and their effect on vascular remodeling poses a challenge since there is limited literature that confirm or deny our results. Nonetheless, our data suggests the efficacy of the endothelial nitric oxide synthase (eNOS) – nitric oxide (NO) system in vascular remodeling.

Table 1 shows the decrease in velocity in the eNOS overexpression (OE) compared to the wild- type (WT) control. Table 1 also shows a decrease in wall shear stress (WSS). These hemodynamic differences point to favorable remodeling as the eNOS OE mice showed smoother velocity streamlines and less wall shear stress.

As for the wall mechanical parameters, no definitive conclusion could be made based on the data. The data shows that eNOS OE mice have a higher von Mises stress and strain four millimeters passed the anastomosis; this occurrence was not due to random chance. This result was unexpected because the wall shear stress decreases. The mechanical parameters show a contradiction to the hemodynamic results.

Our hemodynamic results match closely with previous studies. Baltazar et al. and Pike et al. demonstrated that eNOS OE mice 21 days post-AVF creation show decreased velocity and wall shear stress [8], [12]. As aberrant flow contributes to increased neointimal hyperplasia, these previous studies concluded that the smoother flow in the AVF is associated with favorable AVF maturation. These studies used computational fluid dynamics and fluid-structure interactions; our study is an expansion of these studies by adding more mechanical wall parameters. Our study, along with these previous studies, suggest that the longitudinal effects of NO can exhibit therapeutic qualities relating to AVF maturation.

As for the mechanical parameters, no literature analyzes the von Mises stress and strain in murine AVF models. Using FSI simulations in smaller animals is a novel technique that has not been fully addressed. Our group’s previous study involving the use of FSI only analyzed circumferential stretch and radial wall thinning [8]. In this study, increased circumferential stretch and radial wall thinning are suggested to be associated with favorable AVF maturation. Pike et al., however, does not analyze the resultant stresses and strains on the AVF during vascular remodeling [8].

There are limitations of this study. Because of the high computational cost of FSI simulations and the complexity of AVF geometries, the sample size was small (n=2). Another limitation is the material assignment of the model. There is limited literature about the material properties of murine AVFs (Young’s modulus, Poisson’s ratio, etc.). We speculate that the material assignment in the FSI simulations strayed from real physiological values in mouse AVFs. This can explain why our results are contradictory to what we expect. The last limitation is the pressure boundary condition. The use of the RC Windkessel model to convert the extracted velocity from the Cine-Phase contrast MRIs posed a challenge. This is because the temporal resolution of our MRI was insufficient to accurately model the velocity-pressure conversion.

We found that vascular wall mechanics is difficult, but an important parameter to model. Venous wall remodeling after conduit creation is caused by arterial pressure in a venous environment. As a homeostatic response, endothelial cells signal to the smooth muscle cells to proliferate to accommodate this change. This proliferation, if left untreated, can cause stenosis in the AVF, which leads to AVF failure. Our study is an innovation to the field by introducing venous remodeling effects after overexpressing the eNOS gene using FSI simulation. Our findings show that NO can be a potential therapeutic treatment for use in facilitating AVF creation and maintenance.

Our study can be improved by verifying our material parameters using pressure myography to calculate mouse-specific viscoelastic properties. AVFs can have vastly different material properties as the amount of remodeling and hyperplasia is dependent on the blood flow through the AVF. Another expansion on this study can look at the effects of NO in human patients to access its efficacy in AVF creation and maintenance.

AVF abandonment and stenosis are the primary causes of mortality within ESKD patients on hemodialysis. There is no cure for ESKD, and there are limited treatments to prevent AVF failure. Our computational study showed the efficacy of the eNOS-NO system in murine AVFs, which can be applied to the human model. AVF treatments can help patients remain on hemodialysis longer while they wait to receive a transplant. Also, clinicians can use our hemodynamic data to predict when intervention is needed to assist AVF maturation and make informed decisions on AVF maintenance.

References

[1] “Kidney Disease Statistics for the United States – NIDDK,” National Institute of Diabetes and Digestive and Kidney Diseases. Accessed: Sep. 08, 2024. [Online]. Available: https://www.niddk.nih.gov/health-information/health-statistics/kidney-disease

[2] M. F. Hashmi, O. Benjamin, and S. L. Lappin, “End-Stage Renal Disease,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2024. Accessed: Sep. 08, 2024. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK499861/

[3] “Survival for waitlisted kidney failure patients receiving transplantation versus remaining on waiting list: systematic review and meta-analysis | The BMJ.” Accessed: Nov. 10, 2024. [Online]. Available: https://www.bmj.com/content/376/bmj-2021-068769

[4] “KDOQI Clinical Practice Guideline for Vascular Access: 2019 Update – ScienceDirect.” Accessed: Nov. 10, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0272638619311370?via%3Dihub

[5] “Editor’s Choice – Vascular Access: 2018 Clinical Practice Guidelines of the European Society for Vascular Surgery (ESVS) – ScienceDirect.” Accessed: Nov. 10, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1078588418300807?via%3Dihub

[6] A. M. Marsh, R. Genova, and J. L. Buicko Lopez, “Dialysis Fistula,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2024. Accessed: Sep. 08, 2024. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK559085/

[7] Y.-T. Shiu, Y. He, J. C. S. Tey, M. Knysheva, B. Anderson, and K. Kauser, “Natural Vascular Scaffolding Treatment Promotes Outward Remodeling During Arteriovenous Fistula Development in Rats,” Front. Bioeng. Biotechnol., vol. 9, p. 622617, Feb. 2021, doi: 10.3389/fbioe.2021.622617.

[8] D. Pike et al., “The effect of endothelial nitric oxide synthase on the hemodynamics and wall mechanics in murine arteriovenous fistulas,” Sci. Rep., vol. 9, no. 1, p. 4299, Mar. 2019, doi: 10.1038/s41598-019-40683-7.

[9] H. Northrup et al., “Analysis of Geometric and Hemodynamic Profiles in Rat Arteriovenous Fistula Following PDE5A Inhibition,” Front. Bioeng. Biotechnol., vol. 9, p. 779043, Dec. 2021, doi: 10.3389/fbioe.2021.779043.

[10] M. A. Zain, R. T. Jamil, and W. J. Siddiqui, “Neointimal Hyperplasia,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2023. Accessed: Jun. 12, 2023. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK499893/ RANGE: Undergraduate Research Journal Template

[11] M. Somarathna et al., “Nitric oxide releasing nanomatrix gel treatment inhibits venous intimal hyperplasia and improves vascular remodeling in a rodent arteriovenous fistula,” Biomaterials, vol. 280, p. 121254, Jan. 2022, doi: 10.1016/j.biomaterials.2021.121254.

[12] S. Baltazar et al., “Effects of endothelial nitric oxide synthase on mouse arteriovenous fistula hemodynamics,” Sci. Rep., vol. 13, no. 1, p. 22786, Dec. 2023, doi: 10.1038/s41598-023-49573-5.

[13] G. E. Havelka et al., “Nitric oxide delivery via a permeable balloon catheter inhibits neointimal growth after arterial injury,” J. Surg. Res., vol. 180, no. 1, pp. 35–42, Mar. 2013, doi: 10.1016/j.jss.2012.10.048.

[14] “28.8: Receptor Guanylyl Cyclases, cGMP, and Protein Kinase G,” Biology LibreTexts. Accessed: Nov. 10, 2024. [Online]. Available: https://bio.libretexts.org/Bookshelves/Biochemistry/Fundamentals_of_Biochemistry_(Jakubowski_and_Flatt)/Unit_IV_-_Special_Topics/28%3A_Biosignaling_-_Capstone_Volume_I/28.08%3A_Receptor_Guanylyl_Cyclases_cGMP_and_Protein_Kinase_G

[15] “NO-Independent Activators of Soluble Guanylate Cyclase: Therapeutic Potential | American Journal of Hypertension | Oxford Academic.” Accessed: Nov. 10, 2024. [Online]. Available: https://academic.oup.com/ajh/article/22/5/470/330511?login=true

[16] Y. Zhu, S. Mirsadraee, U. Rosendahl, J. Pepper, and X. Y. Xu, “Fluid-Structure Interaction Simulations of Repaired Type A Aortic Dissection: a Comprehensive Comparison With Rigid Wall Models,” Front. Physiol., vol. 13, Jun. 2022, doi: 10.3389/fphys.2022.913457.

[17] Y.-T. Shiu, J. I. Rotmans, W. J. Geelhoed, D. B. Pike, and T. Lee, “Arteriovenous conduits for hemodialysis: how to better modulate the pathophysiological vascular response to optimize vascular access durability,” Am. J. Physiol. – Ren. Physiol., vol. 316, no. 5, pp. F794–F806, May 2019, doi: 10.1152/ajprenal.00440.2018.

[18] R. F. Furchgott and J. V. Zawadzki, “The obligatory role of endothelial cells in the relaxation of arterial smooth muscle by acetylcholine,” Nature, vol. 288, no. 5789, pp. 373–376, Nov. 1980, doi: 10.1038/288373a0.

[19] U. Förstermann and W. C. Sessa, “Nitric oxide synthases: regulation and function,” Eur. Heart J., vol. 33, no. 7, pp. 829–837, Apr. 2012, doi: 10.1093/eurheartj/ehr304.

[20] P. Derentowicz, K. Markiewicz, M. Wawrzyniak, I. Czerwińska-Kartowicz, E. Buława, and H. Siwińska-Gołebiowska, “[Nitric oxide (NO)–Nobel prize in medicine and physiology for 1998],” Med. Wieku Rozwoj., vol. 4, no. 2, pp. 209–217, 2000.

[21] M. M. Bhatti, M. Marin, A. Zeeshan, and S. I. Abdelsalam, “Editorial: Recent Trends in Computational Fluid Dynamics,” Front. Phys., vol. 8, Oct. 2020, doi: 10.3389/fphy.2020.593111.

[22] T. J. Chung, Computational Fluid Dynamics. Cambridge University Press, 2002.

[23] K. Khanafer and K. Vafai, “Chapter 1 – Applications of porous media in biological transport modeling,” in Modeling of Mass Transport Processes in Biological Media, S. Becker, A. V. Kuznetsov, F. de Monte, G. Pontrelli, and D. Zhao, Eds., Academic Press, 2022, pp. 1–15. doi: 10.1016/B978-0-323-85740-6.00014-5.

[24] S. P. Jones, J. J. M. Greer, R. van Haperen, D. J. Duncker, R. de Crom, and D. J. Lefer, “Endothelial nitric oxide synthase overexpression attenuates congestive heart failure in mice,” Proc. Natl. Acad. Sci. U. S. A., vol. 100, no. 8, p. 4891, Apr. 2003, doi: 10.1073/pnas.0837428100.

[25] S. P. Jones et al., “Endothelial nitric oxide synthase overexpression attenuates myocardial reperfusion injury,” Am. J. Physiol.-Heart Circ. Physiol., vol. 286, no. 1, pp. H276–H282, Jan. 2004, doi: 10.1152/ajpheart.00129.2003.

[26] D. Pike et al., “High resolution hemodynamic profiling of murine arteriovenous fistula using magnetic resonance imaging and computational fluid dynamics,” Theor. Biol. Med. Model., vol. 14, p. 5, Mar. 2017, doi: 10.1186/s12976-017-0053-x.