Disturbed flow in the juxta-anastomotic area of an arteriovenous fistula correlates with endothelial loss, acute thrombus formation, and neointimal hyperplasia

Abstract

Clinical failure of arteriovenous neointimal hyperplasia (NIH) fistulae (AVF) is frequently due to juxta-anastomotic NIH (JANIH). Although the mouse AVF model recapitulates human AVF maturation, previous studies focused on the outflow vein distal to the anastomosis. We hypothesized that the juxta-anastomotic area (JAA) has increased NIH compared with the outflow vein. AVF was created in C57BL/6 mice without or with chronic kidney disease (CKD). Temporal and spatial changes of the JAA were examined using histology and immunofluorescence. Computational techniques were used to model the AVF. RNA-seq and bioinformatic analyses were performed to compare the JAA with the outflow vein. The jugular vein to carotid artery AVF model was created in Wistar rats. The neointima in the JAA shows increased volume compared with the outflow vein. Computational modeling shows an increased volume of disturbed flow at the JAA compared with the outflow vein. Endothelial cells are immediately lost from the wall contralateral to the fistula exit, followed by thrombus formation and JANIH. Gene Ontology (GO) enrichment analysis of the 1,862 differentially expressed genes (DEG) between the JANIH and the outflow vein identified 525 overexpressed genes. The rat jugular vein to carotid artery AVF showed changes similar to the mouse AVF. Disturbed flow through the JAA correlates with rapid endothelial cell loss, thrombus formation, and JANIH; late endothelialization of the JAA channel correlates with late AVF patency. Early thrombus formation in the JAA may influence the later development of JANIH.

NEW & NOTEWORTHY Disturbed flow and focal endothelial cell loss in the juxta-anastomotic area of the mouse AVF colocalizes with acute thrombus formation followed by late neointimal hyperplasia. Differential flow patterns between the juxta-anastomotic area and the outflow vein correlate with differential expression of genes regulating coagulation, proliferation, collagen metabolism, and the immune response. The rat jugular vein to carotid artery AVF model shows changes similar to the mouse AVF model.

INTRODUCTION

The autologous arteriovenous fistula (AVF) is the preferred dialysis access in patients with end-stage kidney disease, but 60% of AVF fail to mature, and only 50% are primarily patent at 1 year (1). Neointimal hyperplasia (NIH) is the leading cause of AVF failure, with multiple areas of NIH and stenoses frequently occurring in the outflow vein from the juxta-anastomotic area (JAA) all the way centrally to the cephalic arch (2, 3). Specimens of failed AVF typically show aggressive NIH (4), and strategies to inhibit NIH via inhibition of proliferation have shown increased patency in selected cases (5). However, multiple clinical trials have failed to improve AVF maturation or prevent NIH in the majority of human patients (6–11), suggesting that current animal models of NIH may be limited for human translation.

The mouse arteriovenous fistula model uses a 25-gauge needle to puncture the aorta and IVC, and recapitulates human AVF maturation (12), showing a consistent pattern of NIH, luminal dilation with outward remodeling of the outflow vein (13–15), as well as reduced patency in female mice (16). In addition, inhibition of NIH using therapeutic drugs shows increased patency in this model (13, 17–19, 20), suggesting the translational application of this model.

These reports consistently analyze the outflow IVC at 100–200 µm distal (cranial) to the AVF, suggesting translational applicability to the outflow vein that is cannulated during hemodialysis or to central stenosis (21); however, since juxta-anastomotic NIH (JANIH) is a major source of early AVF failure to mature (4, 22), it is critical to determine the spatial and temporal evolution of JANIH in this model. We hypothesized that there is relatively increased volume of NIH in the JAA compared with the outflow vein that colocalizes with disturbed flow.

MATERIALS AND METHODS

Arteriovenous Fistula Models

All animal experiments were performed in strict compliance with federal guidelines and with approval from the Institutional Animal Care and Use Committee of Yale University. Mice were 9 to 11 wk of age when the surgeries were performed; microsurgical procedures were performed aseptically in a dedicated facility using a dissecting microscope (Leica MZ 95). Anesthesia was administered using 2 to 2.5% isoflurane, and extended-release buprenorphine (Ethiqa XR; North Brunswick, NJ) was administered subcutaneously (3.25 mg/kg body wt) for intraoperative and postoperative analgesia. After exposing the IVC and aorta, an arteriovenous fistula was created by puncturing the distal aorta into the IVC using a 25-gauge needle, as previously described (23). Visualization of pulsatile arterial blood flow in the IVC was used to assess initial technical success of AVF creation.

Chronic kidney disease (CKD) was induced in mice 3 wk before fistula creation via a surgical 5/6 nephrectomy model. After induction of anesthesia with inhaled isoflurane, the right kidney was removed; the upper and lower poles of the left kidney were excised after decapsulation, and hemostasis was achieved with a gentle manual compression. The abdomen was then closed in two layers using running 6-0 sutures.

The rat jugular-carotid AVF model was created as previously described (24). Briefly, male Wistar rats, ages 6 to 8 wk old, were used for AVF creation. After anesthesia was administered, a midline neck incision was made; blunt and sharp dissection was used to expose the carotid artery and jugular vein. On the animal’s left side, a 1-mm longitudinal arteriotomy was made on the carotid artery, the jugular vein was anastomosed to the carotid artery in an end-to-side configuration. The skin incision was closed using running 4-0 absorbable sutures. Native artery and vein were used as controls. No heparin or antibiotics were used in this model.

Ultrasound Measurements

Doppler ultrasound (Vevo 770 and Vevo F2 High-Resolution Imaging System; Fujifilm Visual Sonics, Inc., Toronto, Canada) using probe RMV704 (40 MHz) was performed at baseline and serially postoperatively to confirm the patency of the AVF, to obtain blood flow velocities, and to measure the inner diameter of the aorta and the AVF.

Macroscopic Photographs

The normal right kidney (day 0) or remnant left 1/6 kidney were harvested (day 7 or 21). Surrounding tissues were carefully dissected away, and photographs were taken using the dissection microscope.

The native aorta, IVC, and AVF were harvested at days 0, 0.25 (6 h), 1, 3, 7, or 21; the IVC lumen was opened or dissected away. For some of the AVF, the contralateral venous wall facing the fistula was kept intact, and photographs were taken using the dissection microscope.

The rat AVF was harvested at 6 and 24 h for macroscopic photographs.

Histology

Animals were euthanized and perfused with normal saline followed by 10% formalin via the left ventricle under physiological pressure, and the AVF was extracted and then embedded in paraffin and cut in 5-μm cross sections. Hematoxylin and eosin (H&E) and Elastin van Gieson (EVG) were used. Some samples were paraffined in a longitudinal direction, cut in 5-μm thickness, and stained with H&E and EVG.

Immunofluorescence

Tissue sections were deparaffined using xylene and rehydrated in a graded series of alcohols. Sections were heated in citric acid buffer (pH 6.0) at 100°C for 10 min for antigen retrieval. The sections were then blocked with 2% bovine serum albumin (BSA) for 1 h at room temperature, before incubation overnight at 4°C with the primary antibodies (Supplemental Table S1). Sections were then treated with secondary antibodies (Supplemental Table S1) at room temperature for 1 h and stained with 4′,6-diamidino-2-phenylindole (DAPI, P36935; Invitrogen).

En Face Staining

After animals were euthanized and perfused with normal saline followed by 10% formalin, the IVC was harvested, and its adventitia was gently removed; the IVC lumen was opened longitudinally. Samples were blocked in 1% BSA for 30 min and then incubated overnight at 4°C with the primary antibodies. IVC samples were carefully placed on the slides with the luminal side facing upward and stained with DAPI.

Three-Dimensional Computational Model

The computational model was based on extracted vessel geometry from computed tomography (CT) scans and boundary conditions (BC) determined based on ultrasound measurements (day 7); the inlet aortic BC was set as −0.8 m/s, and the inlet IVC BC was set at 0.5 m/s. CT was performed on postoperative day 7 following AVF creation (male mice; AVF, n = 3). A nanoparticulate contrast agent (ExiTron nano 12,000; 100 μL; Miltenyi Biotech, Bergisch Gladbach, Germany) was injected via the tail vein. The animals were then placed supine on a horizontal stage under light anesthesia (1.25% isoflurane), and images were obtained from the renal vessels to below the bifurcation of the aorta and IVC using a micro-CT scanner (MicroSPECT4CT; MILabs, Houten, the Netherlands) using retrospective cardiac and respiratory gating, with a cone beam filtered back projection algorithm set to 20-μm effective voxel size. Micro-CT was performed with a 50-kVp X-ray tube voltage, 430-µA tube current, 20 ms per frame, 360° angle, and 0.75° increments per view.

The IVC-aorta complex was extracted from mice (day 21) with care to minimize any disruption of the fistula connection. The IVC-aorta samples then underwent a modified heavy metal-based staining protocol to preserve the global geometry for electron-microscopy high-contrast tissue staining (25). The stained samples were imaged using tomographic microscopy using the 2-BM-B beam line at the Advanced Photon Source, Argonne National Laboratory (26). This parallel-beam micro-CT set-up incorporates an Optique Peter microscope system to allow for sub-micron imaging. Raw data were reconstructed with TomocuPy (27).

The computational model was based on extracted vessel geometry from the in vivo CT scans. Raw Digital Imaging and Communications in Medicine (DICOM) images were imported into Simpleware (Synopsys, Exeter, United Kingdom), and the geometry was segmented by isolating the artery, vein, and fistula from the background noise. The final segmented geometries were then imported into Xflow (Simulia, Dassault Systems), a particle-based Lattice Boltzmann kinetic solver that features a Wall-Modeled Large Eddy Simulation (WMLES) approach to turbulence modeling by providing a consistent local eddy-viscosity and near wall behavior.

Velocity and pressure boundary conditions (BC) were applied to the artery and vein inlets and outlets. First, the BC at the arterial inlet (cranial) and venous inlet (caudal) was prescribed a velocity based on the mean ultrasound measurements of the mice AVF (n = 3). Second, the BC at the arterial outlet (caudal) and venous outlet (cranial) was prescribed using estimated mean arterial physiological blood pressure (100 mmHg) and mean central venous pressure (5 mmHg), respectively (28). Blood was modeled as a Newtonian fluid, with molecular mass 28.996 Da, density 1,060 kg/m³, temperature 288.15 K, viscosity 0.0035 Pa × s, thermal conductivity 0.0243 W × (m × K)⁻¹, and specific heat capacity 1006.43 J × (kg × k)⁻¹.

Characterization of the Flow Field

To characterize the fluid in the model, disturbed flow (turbulence) was analyzed along the length of the system at varying cross-sectional areas within the IVC. To calculate the intensity of disturbed flow, Xflow uses the Wall-Adapting Local Eddy-viscosity (WALE) model that has adequate properties both near to and far from the wall and both for laminar and turbulent flows. WALE considers small-scale eddies near the boundaries in fluid flow simulations and adapts its calculations to account for the changing turbulence characteristics closer to the wall. Specifically, this model accurately captures the asymptotic characteristics of the turbulent boundary layer when the boundary layer is solvable directly, and it does not introduce artificial turbulent viscosity in the shear regions beyond the wake (1). The WALE model is formulated as follows:

$$v_t={\left(c_w\Delta \right)}^2\frac{{(S_{ij}^d{S}_{ij}^d)}^{3/2}}{{\left({\overline{S}}_{ij}{\overline{S}}_{ij}\right)}^{5/2}+{(S_{ij}^d{S}_{ij}^d)}^{5/4}}$$ (1)

$v_{\mathrm{t}}$ = eddy viscosity

C_w is the WALE constant, typically 0.2

${\stackrel{-}{g}}_{\mathrm{i}\mathrm{j}}$ is the velocity gradient tensor

Δ is the subgrid characteristic length scale

${\stackrel{-}{S}}_{\mathrm{i}\mathrm{j}}$ is the deformation tensor

δ_ij is the Kronecker symbol

$$S_{ij}^d=\frac{1}{2}\left({\overline{g}}_{ij}^2+{\overline{g}}_{ij}^2\right)-\frac{1}{3}{\text{d}}_{ij}{\overline{g}}_{kk}^2$$ (2)

$${\overline{g}}_{ij}^2={\overline{g}}_{ik}{\overline{g}}_{kj}$$ (3)

Turbulence intensity I (in %) is defined as:

$$I=\frac{u\prime }{U}$$

U = mean velocity

u′ = root-mean-square of the turbulent velocity fluctuations described as

$$u^{\prime }=\sqrt{\frac{1}{3}({u^{\prime }}_x^2+{u^{\prime }}_y^2+{u^{\prime }}_z^2)}=\sqrt{\frac{2}{3}k}$$ (4)

k = turbulent kinetic energy

and the mean velocity U computed from the three mean velocity components:

$$U=\sqrt{U_x^2+U_y^2+U_z^2}$$ (5)

Please note that the eddy viscosity ($v_{\mathrm{t}}$) represents the effect of velocity fluctuations ($u\prime$) on the stress in a fluid. $v_{\mathrm{t}}$ is not a physical property of the fluid, and it depends on the intensity of the turbulent velocity fluctuations ($u\prime$).

Disturbed flow was calculated as the turbulence intensity (I) and imported into Paraview (open-source, Kitware, Sandia National Laboratories, Los Alamos National Laboratory); disturbed flow intensity was averaged at individual cross sections along the IVC, with each cross section ∼1 mm apart. Close to the fistula, disturbed flow intensity was averaged within cross sections ∼0.1 mm apart to capture the dynamic changes in this region.

RNA Extraction and RNA Sequencing

The anastomotic neointimas (combined from four mice) and outflow veins (combined from two mice) were carefully harvested and immediately submerged in RNA protect Tissue Reagent (QIAGEN, Hilden, Germany) and incubated overnight at 4°C, then stored at −80°C without the reagent. Total RNA was extracted using RNeasy Mini kit with DNase I (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. Total RNA quality and concentration were estimated with A260/A280 and A260/A230 ratios measured using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE).

Preparation and sequencing of poly-A mRNA sequencing libraries were performed by the Yale Center for Genome Analysis (YCGA). RNA integrity was estimated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Samples with RNA integrity number (RIN) values of ≥8.0 were used for analysis. Poly-A mRNA was purified from ∼200 ng of total RNA, and the stranded mRNA sequencing libraries were constructed using KAPA mRNA HyperPrep kit for Illumina sequencing (Roche, Basel, Switzerland). Indexed libraries were quantified by qRT-PCR using KAPA Library Quantification kits (Roche, Basal, Switzerland), and insert size distribution was determined using an Agilent 2100 Bioanalyzer. Samples with a yield of ≥0.5 ng/µL and a size distribution of 150–300 bp were sequenced. Sample concentrations were normalized to 1.2 nM and loaded onto an Illumina NovaSeq flow cell (Illumina, San Diego, CA) at a concentration that yielded 25 million passing filter clusters per sample. Samples were sequenced using 101 bp paired-end sequencing on an Illumina NovaSeq 6000 (Illumina, San Diego, CA). Data generated during sequencing runs were simultaneously transferred to the YCGA high-performance computing cluster.

Differential Expression Analysis and Enrichment Analysis

The FASTQ files of raw sequencing reads were imported into a computing cluster based on Red Hat Linux computing system and analyzed. For the purpose of quality control, poly A and poly T sequences, Illumina adaptor sequences, and low-quality sequences in the reads were deleted using PRINSEQv0.20.4 and Trimmomatic v0.39 software (29, 30). The trimmed sequencing data were aligned to the GRCm39 reference genome with the gene annotation information (Ensembl release 106) (31), and expression values for each gene were counted using STAR v2.7.11a and RSEM v1.3.3 software (32).

Further analyses were performed using packages in R v4.2.3 and RStudio v2023.09.1 build 494 on the macOS computing system. Since multiple samples harvested were combined into a single mRNA library, to extract adequate mRNA from the samples, a value of statistical dispersion was estimated from our previous bulk RNA sequencing data. Using the previous data [NCBI Gene Expression Omnibus (GEO) database, Accession No. GSE244317], in which six out flow veins of AVF-created mice had been compared with six inferior vena cava of sham-operated mice, and the dispersion was calculated as 0.1551571. Extrapolating this value, differentially expressed genes (DEGs) were identified by using clusterProfiler v4.6.2 package (33). Gene Ontology (GO) enrichment analysis was also performed using the clusterProfiler package.

Statistical Analysis

Statistical analyses of the sequencing data were performed with the abovementioned packages in R, where an adjusted P value of <0.05 and an absolute fold change of ≥2.0 were considered statistically significant to identify DEG. An adjusted P value of <0.05 was considered statistically significant for GO enrichment analysis. These adjusted P values were determined by applying the Benjamini–Hochberg multiplicity correction method (34).

All data were analyzed using Prism9 software (GraphPad Software, Inc., La Jolla, CA). n represents numbers of animals unless otherwise indicated, and error bars represent means ± SE. The Shapiro–Wilk test was used to test for normality; the F test was performed to evaluate homogeneity of variances. For two-group comparisons, the unpaired Student’s t test was used for normally distributed data. For multiple group comparisons with normally distributed data, the one-way ANOVA test was used. P values of ≤0.05 were considered to indicate statistical significance.

RESULTS

Temporal and Spatial Changes of the Juxta-Anastomotic Area

Since JANIH is the primary etiology of AVF failure in human patients (35, 36), and since previous reports using the mouse AVF model focused on the outflow vein and not the JAA (12, 37), we examined the JAA (Fig. 1A). There was disturbed blood flow at the JAA after AVF creation (Supplemental Fig. S1), and Doppler ultrasound showed increased diameters of both the aorta and the IVC at day 21 compared with baseline (Supplemental Fig. S1B); normal arterial and venous waveforms were present at baseline, whereas disturbed waveforms were present at the JAA, both in the aorta and the IVC at day 21 (Supplemental Fig. S1, A and B), consistent with previous data (38). The IVC and aorta lumens were opened for direct in situ assessment of the JAA; there was thrombus at the anastomosis 3 h after AVF creation (Fig. 1B); in addition, there was protruding anastomotic neointima and contralateral wall neointima opposite the patent fistula exit at day 21, with a 6-0 Prolene suture easily passing through the fistula channel (Fig. 1C). The IVC wall was dissected away (day 21), leaving the anastomotic neointima to be seen as an outgrowth of anastomotic neointima from the aorta; the anastomotic neointima was distinct from the contralateral wall neointima that formed opposite the JAA channel exit (Fig. 1D). We also examined the JAA in mice with CKD induced by 5/6 nephrectomy that mimics human AVF (Fig. 1, E and F, Supplemental Fig. S2, A and B). Anastomotic neointima was clearly observed at days 7 and 21 (Fig. 1, G and H), similar to mice without CKD. Thrombus formed at the anastomosis as early as day 1 (Fig. 1H). In addition, both female and male mice with CKD showed anastomotic neointima and contralateral wall neointima at day 21 (Fig. 1I). These data show that neointima localizes at the JAA in this AVF model, similar to human JANIH.

Figure 1.

Gross view at the juxta-anastomotic area (JAA) in the mouse arteriovenous fistula (AVF) model without and with chronic kidney disease (CKD). A: photograph showing the AVF at day 21, black “*” showing the location of JAA, note the mixed color of blood flow at the JAA; A, aorta; IVC, inferior vena cava. B: in situ photo showing the JAA at 3 h after AVF creation, black “*” showing the anastomosis; A, aorta; IVC, inferior vena cava. C: gross view showing the anastomosis and contralateral wall neointimas, fistula entrance, fistula channel, and fistula exit at the JAA at day 21; note the contralateral wall neointima facing the fistula exit and 6-0 Prolene suture goes through the fistula channel; white rectangle showing the JAA; yellow dashed line area shows the IVC surface. D: front and back gross view of the JAA in the male and female mice at day 21 after AVF creation; note the whole IVC wall was dissected away from the aorta; white square showing the location of the JAA; A, aorta; black “*” showing anastomosis; scale bar, 2 mm; n = 3. E: illustration diagram showing the mouse AVF with CKD induced by 5/6 nephrectomy (red rectangle), and gross view of temporal changes of the remaining 1/6 kidney; n = 3. F: ultrasound measurement and Doppler waveforms of the aorta and IVC at day 21 after AVF creation at the location of JAA in mice with CKD; yellow “*” showing the JAA; A, aorta; IVC, inferior vena cava; n = 3. G: gross view of the JAA at days 0, 7, and 21 in mice with CKD; A, aorta; IVC, inferior vena cava; scale bar, 1 mm; yellow arrows show the fistula exit; yellow rectangle shows the location of the JAA; note the IVC lumen was opened. H: gross view of the JAA at days 1 and 7; note the whole IVC wall was dissected away from the aorta; A, aorta; yellow “*” shows the anastomosis; scale bar, 1 mm; n = 3. I: gross view of the JAA in male and female mice with CKD at day 21, scale bar, 2 mm; zoomed in photographs show the JAA; yellow “*” shows the anastomosis; yellow arrows point to the neointima formed on the contralateral wall; n = 3 animals.

To determine the temporal and spatial changes of the JAA in this model, the entire infrarenal IVC and aorta were examined in both female and male mice (Supplemental Figs. S2C and S3). In mice both without or with CKD, immediately after AVF creation, thrombus formed on the exposed underlying collagen and elastin in the anastomotic channel (black arrows, Supplemental Fig. S3A; pale pink tissue and black arrows, Fig. 2A). At day 1, thrombus was present at the anastomosis (Fig. 2B) similar to observed grossly (Fig. 1H). At days 7 and 21, there was also contralateral wall neointima in both of the female and male mice without or with CKD (Supplemental Fig. S3, B and C; black stars, Fig. 2, B and C). These data show that the anastomotic neointima shows dynamic changes after AVF creation.

Figure 2.

Histological temporal and spatial changes of the JAA in the mouse AVF model with CKD. A: sections stained with Verhoeff–Van Gieson (EVG) at day 0, 10 s and 6 h after AVF creation in female and male mice; A, aorta; V, inferior vena cava; black arrows show the thrombus formed on the fistula channel surface; scale bar, 1 mm or 200 μm; n = 3. B: sections stained with EVG of the JAA from caudal to cranial at days 1, 7, and 21 in female and male mice; A, aorta; V, inferior vena cava; *, JAA; scale bar, 400 μm; n = 3–10; dashed black squares show the contralateral wall neointima; the middle of the fistula was defined as 0 μm, locations to cranial or caudal were defined “+” or “−” as shown in the figure; see also Supplemental Fig. S2C. C: bar graphs showing the anastomosis area in the corresponding locations in the female and male mice at days 0, 1 (female, P < 0.0001; male, P < 0.0001), day 7 (female, P < 0.0001; male, P < 0.0001), and day 21 (female, P < 0.0001; male, P < 0.0001); one-way ANOVA, n = 3. ANOVA, analysis of variance; AVF, arteriovenous fistula; CKD, chronic kidney disease; JAA, juxta-anastomotic area.

After AVF creation, disturbed flow and exposed collagen lead to the development of JANIH (38, 39), and disturbed flow is observed in this model from immediately after AVF creation to day 21 in the patent AVF (Fig. 1F) that leads to stable remodeling by day 21 (38). To model disturbed flow before neointimal formation, computed tomography (CT) scan images of the IVC and aorta (day 7) showed the JAA, fistula channel, and JANIH (Fig. 3) similar to that seen with histology (Fig. 2). Computational techniques were used to model the blood flow in this mouse model (Supplemental Tables S2 and S3), and they show disturbed flow localized most intensely to the JAA, with increased disturbed flow in the JAA compared with the outflow vein (Fig. 3A). The modeled disturbed flow intensities colocalized with the thickness of the NIH along the remodeling venous outflow (Fig. 3A; black stars, JANIH; black arrows, fistula channel). Quantification of the intensity of the disturbed flow (%intensity) showed increased disturbed flow in the outflow vein compared with that in the vein below the fistula (7.2 ± 4.7 vs. 0.2 ± 0.1%; P < 001; Mann–Whitney U test); in addition, disturbed flow was increased in the JAA compared with the outflow vein (20.4 ± 11.3 vs. 7.2 ± 4.7%; P < 001; Mann–Whitney U test; Fig. 3B). Although we also observed increased percent intensity in the inflow artery (red color in Fig. 3), this may not reflect increased disturbed flow due to the high velocities in the aorta compared with the IVC; similarly, there was neointima formation in the aorta only at the aortic needle puncture site (Fig. 3, yellow arrow). Thus, this computational modeling data show increased disturbed flow in the IVC at the JAA compared with the rest of the outflow vein that precedes venous neointima formation.

Figure 3.

Computational model depicting intensity of disturbed flow throughout the fistula and IVC, day 7. A, left: micro-computed tomography (CT) scan images from cranial to caudal; middle, cross-sectional areas of disturbed flow within the vein and a longitudinal cross-section highlighting the juxta-anastomotic area. A, right: AVF sections stained with Verhoeff–Van Gieson (EVG) from caudal to cranial. Scale bar, 1 mm; *, neointimal hyperplasia; black or white arrows show the fistula channel; A, aorta; V, inferior vena cava; yellow arrow shows the thick neointima at the aortic entry site. B: line graph showing the intensity of disturbed flow along the IVC, n = 884–1,419 turbulence values per cross section. AVF, arteriovenous fistula; IVC, inferior vena cava.

Thrombus Formation Is Followed by Neointimal Hyperplasia

Platelet adhesion and fibrin formation are critical steps of early thrombus formation (40–42); platelets also play a dual role both initiating an immediate repair process and a delayed process to prevent excessive repair after arterial injury (43). Since fibrin and thrombus are rapidly found in the JAA after AVF creation (Fig. 2A, Supplemental Fig. S3A), we examined the temporal changes of platelets and fibrinogen α in the JAA in this AVF model. There were no CD61-positive platelets or fibrinogen α in the JAA at baseline (day 0). After AVF creation, there were a large number of CD61-positive platelets in the anastomotic thrombus in both female and male mice at days 1 and 3, but few CD61-positive platelets at days 7 and 21 (yellow arrow; Fig. 4, A and B). At days 1 and 3, there were neither α-actin nor CD31-positive cells in the anastomotic thrombus; however, at days 7 and 21, there were large number of α-actin-positive cells (Figs. 4A and 5). At day 21, the CD31-positive and α-actin-positive cells covered the anastomotic neointima surface but not the fistula channel (white arrow, neointimal surface; white arrowhead, fistula channel; Fig. 4A). Similarly, at days 1 and 3, both vWF and fibrinogen α were detectable, with no SM-MHC 11-positive cells (Fig. 4A); at days 7 and 21, both vWF- and SM-MHC11-positive cells were detectable in the anastomotic neointima, in both female and male mice (Fig. 4A).

Figure 4.

Disturbed flow induces platelet and fibrin accumulation, and tissue factor expression at the JAA after AVF creation. A: immunofluorescence at the JAA on days 1, 3, 7, and 21; first row, immunofluorescence stained for CD31 (green), CD61 (red), α-actin (cyan), and DAPI (blue); second row, vWF (green), fibrinogen α (red), SM-MHC11 (cyan), and DAPI (blue); third row, tissue factor (TF, red), CD31 (green), and DAPI (blue); yellow arrows show the CD61-positive platelets in the JAA, white arrows show full endothelization on the anastomosis neointimal surface but not in the fistula channel, green arrow shows TF-positive cells in the JAA. A, aorta; V, inferior vena cava; C, fistula channel; scale bar, 200 μm; n = 3. B: bar graph showing CD61-positive platelets in the JAA; P < 0.0001, one-way ANOVA; n = 3. C: bar graph showing the TF-positive cells in the JAA, P < 0.0001, one-way ANOVA; n = 3. ANOVA, analysis of variance; AVF, arteriovenous fistula; DAPI, 4′,6-diamidino-2-phenylindole; JAA, juxta-anastomotic area; TF, tissue factor.

Figure 5.

JAA neointima formation follows early thrombus in this AVF model. A: photograph shows the anastomosis, contralateral wall, and anterior/posterior (A/P) walls in the AVF (day 21) stained with CD31 (green), α-actin (red), and DAPI (blue); scale bar, 1 mm. B: high-power photographs show the temporal changes of anastomosis (first row), contralateral wall (second row), and anterior/posterior (A/P) wall (third row) stained for CD31 (green), α-actin (red), and DAPI (blue) of the baseline (day 0), 6 h, days 1, 3, 7, and 21 in this mouse AVF model; white arrow showing the residue of early thrombus in the anastomosis and contralateral wall neointima, scale bar, 200 μm; n = 3. C: bar graph showing the anastomosis α-actin-positive cells; P < 0.0001, one-way ANOVA; n = 3. D: bar graph showing the contralateral wall endothelial cell coverage, P < 0.0001 one-way ANOVA; n = 3. E: bar graph showing the contralateral wall α-actin-positive cells; P < 0.0001, one-way ANOVA; n = 3. F: bar graph showing the anterior/posterior wall endothelial cell coverage, P = 0.5464, one-way ANOVA; n = 3. ANOVA, analysis of variance; AVF, arteriovenous fistula; DAPI, 4′,6-diamidino-2-phenylindole; JAA, juxta-anastomotic area.

Since tissue factor (TF) regulates hemostasis (44, 45), we examined TF changes after AVF creation. There were no TF-positive cells at baseline in the IVC endothelium (Fig. 4A); however, after AVF creation, TF-positive cells were detectable in the anastomosis as early as 6 h (Fig. 4A), and there were increased numbers of TF-positive cells in the JAA at days 1 and 3 (green arrow; Fig. 4, A and C); TF-positive cells increased at day 7 and then decreased at day 21 (Fig. 4, A and C). These data suggest that TF plays a role in the early thrombus formation that occurs after AVF creation.

Since endothelial cells (EC) and smooth muscle cells (SMC) are major components of the neointima, we determined the temporal localization of EC and SMC in the JAA. En face staining showed that the endothelium was fully covered by a fishnet pattern layer composed of CD31-positive EC in the IVC at day 0, that was not present at day 7 (Supplemental Fig. S4A). The locations of anastomosis or juxta-anastomosis, contralateral wall, and anterior/posterior (A/P) wall are shown in a day 7 and day 21 AVF (Fig. 5A, Supplemental Fig. S4B). At baseline, the IVC showed continuous CD31- and vWF-positive cells, and orderly arranged α-actin- and SM-MHC11-positive cells (Fig. 5B, Supplemental Fig. S4C). At the anastomosis, there were neither CD31- nor α-actin-positive cells before day 3; and significantly more α-actin-positive cells were in the anastomosis at days 7 and 21 (Fig. 5, B and C). Six hours after AVF creation, the EC was lost and there was mural thrombus formation on the contralateral wall facing the fistula exit; there were no CD31-positive cells before day 7 in the contralateral wall facing the fistula exit, but part of the contralateral wall gained CD31-positive cell coverage at day 21 (Fig. 5, B, D, and E). In the A/P wall not directly facing the fistula exit, the EC was intact from 6 h to day 21, with only slightly thickening of the SMC layer (Fig. 5, B and F). Similar results were found when the sections were stained with vWF and SM-MH11 (Supplemental Fig. S4C). These data confirm that the anastomotic and contralateral walls contain thrombus immediately after AVF creation, in areas of focal EC loss, and then also contain NIH in the identical area at later time points, similar to the direct observations (Fig. 1).

Patency of the AVF Correlates with Channel Endothelialization

In this AVF model, failure occurs between days 28 and 42 in ∼1/3 of the male mice (12), and there is increased failure in female mice (16, 37). Since mice with CKD typically have a high (20–60%) mortality rate (24, 46, 47), we used mice without CKD to assess later time points. At day 7, there were several layers of SM-MHC11-positive cells with few CD31-positive cells on the anastomotic neointima surface, and there were no CD31-positive cells in the fistula channel; at day 42, there were fewer SM-MHC11-positive cells but a continuous CD31-positive cell layer on the anastomotic neointima surface and fistula channel (white arrows, neointima surface; white box, fistula channel; Fig. 6, A and B). There were increased numbers of SM22α-positive cells in the fistula channel at day 42 compared with day 7, but a decreased number of SM22α-positive cells on the anastomotic neointima surface at day 42 compared with day 7; some of these SM22α-positive cells also colocalized with collagen-1 and the synthetic marker vimentin (Fig. 6, A and D). There were a large number of PCNA- and SM22α-dual positive cells in both of the anastomotic and contralateral wall neointimas at day 7, but there were significantly decreased numbers of these cells at day 42 (Fig. 6, A and E). The TGFβ pathway mediates collagen synthesis and extracellular matrix deposition (14); there was a similar percentage of p-smad2- and SM22α-dual positive cells in both of the anastomotic and contralateral wall neointimas at days 7, 21, and 42 (Supplemental Fig. S5, A–C).

Figure 6.

Histological changes at the JAA after maturation in this AVF model. A: merged immunofluorescence showing the JAA at days 0, 7, and 42; first row, merge of CD31 (green), SM-MHC11(red), and DAPI (blue); second row, merge of SM22α (green), collagen-1 (red), and DAPI (blue); third row, merge of SM22α (green), vimentin (red), and DAPI (blue); fourth row, merge of SM22α (red), PCNA (green), and DAPI (blue); A, aorta; V, inferior vena cava; C, fistula channel; *, anastomosis; scale bar, 1 mm or 100 μm; white arrows show the anastomosis neointimal surface; white box shows the fistula channel; n = 3. B: bar graph showing endothelial cell coverage in the fistula channel (P < 0.0001) and on the anastomosis neointima surface (P = 0.0003) at day 7 and day 42; t test; n = 3. C: bar graph showing the SM-MHC11-positive cells in the fistula channel (P = 0.0004) and on the anastomosis neointima surface (P = 0.0008) at day 7 and day 42; t test; n = 3. D: bar graph showing the SM22α-positive cells in the fistula channel (P = 0.0017) and on the anastomosis neointima surface (P = 0.0177) at day 7 and day 42; t test; n = 3. E: bar graph showing the PCNA- and SM22α-dual-positive cells in the fistula channel (P = 0.0074) and anastomosis neointima surface (P = 0.0005) at day 7 and day 42; t test; n = 3. AVF, arteriovenous fistula; DAPI, 4′,6-diamidino-2-phenylindole; JAA, juxta-anastomotic area.

Since early AVF failure is typically the result of thrombosis that occurs in the fistula channel, we examined the temporal changes of EC lining the fistula channel. There were no luminal EC in the fistula channel at days 1, 3, or 7, and rare luminal EC at day 21 (Fig. 7, A–C). The JAA at day 21 was also sectioned longitudinally, and EVG staining showed a patent fistula channel without thrombus formation (Fig. 7D); immunofluorescence confirmed endothelial coverage on the anastomotic neointima surface but not in the fistula channel (Fig. 7D). Immunofluorescence also showed neointima in the aortic entry site (Fig. 7D, yellow arrows). In patent AVF (day 42), the fistula channel was patent, whereas the fistula channel of occluded AVF contained significant thrombus without a lumen (Fig. 7E). Both patent and occluded AVF contained luminal EC at the fistula entrance (Fig. 7F). Compared with the patent fistula channel that contained a monolayer of luminal EC (Fig. 7F, yellow arrows), α-actin- and CD31-positive cells filled the channel in the occluded AVF; contralateral wall neointima was found in both patent and occluded AVF (Fig. 7F). These data show that AVF patency correlates with patency of the fistula channel.

Figure 7.

Endothelization of the fistula channel correlates with AVF patency in this AVF model. A: low power photographs showing the temporal changes of fistula channel stained with CD31 (green), α-actin (red), and DAPI (blue) from day 1 to day 21 in the female (first row) and male (second row) mice; scale bar, 1 mm; C, fistula channel; n = 3. B: bar graphs showing endothelial cell (EC) coverage (day 7) in the fistula channel and on the anastomosis surface (P = 0.3529), t test; n = 3. C: bar graph showing EC coverage (day 21) in the fistula channel and on the anastomosis surface (P = 0.5734), t test; n = 3. D: longitudinal sections stained with EVG; CD31 (green), α-actin (red), and DAPI (blue) of the JAA at day 21, A, aorta; IVC, inferior vena cava; black arrow shows the fistula channel; note EC coverage on the anastomosis surface but not in the fistula channel; yellow arrow shows thick neointima at the aortic entry site; scale bar, 1 mm. E: photographs show the spatial changes of patent and occluded AVF at day 42, black arrows show the patent or occluded fistula channel, scale bar, 1 mm; n = 3–6. F: immunofluorescence of the patent and occluded fistulae at day 42 stained with CD31 (green), α-actin (red), and DAPI (blue); scale bar, 1 mm or 200 μm; dashed white circles show the fistula channel; yellow arrows show luminal CD31-positive cells; n = 3. AVF, arteriovenous fistula; DAPI, 4′,6-diamidino-2-phenylindole.

Transcriptomic Differences between the Anastomotic Neointima and the Outflow Vein

Since these data suggest increased NIH in the JAA compared with the outflow vein, we compared the global RNA transcript differences between the anastomotic neointima and the outflow vein with an unbiased approach using bulk RNA-seq; due to the small amount of RNA, we combined four anastomotic neointimas and two outflow veins of AVF (male mice, day 7) for analysis (Fig. 8A). GO enrichment analysis of the 1,862 DEG obtained from the comparison between the anastomotic neointima and outflow vein identified 525 overexpressed genes and 483 overrepresented GO terms. Compared with the outflow vein, multiple genes were differentially regulated in the JAA, including genes related to tissue remodeling, regulation of response to wounding, coagulation and regulation of coagulation, fibroblast proliferation, collagen biosynthetic, and catabolic process (Fig. 8, B and C; Fig. 9; Supplemental Tables S4–S10). There were also differences in the inflammatory response to wounding, positive regulation of T cell-mediated immunity, macrophage activation, T-cell activation, and macrophage cytokine production between the anastomotic neointima and the outflow vein (Supplemental Tables S11–S15). These data suggest that disturbed flow at the JAA differentially influences genes that regulate coagulation, proliferation, collagen metabolism, and the immune response, validating the data suggesting differences between the JAA and the outflow vein.

Figure 8.

RNA-seq analysis showing the difference between the JAA anastomotic neointima and the outflow vein. A: illustration showing the JAA anastomotic neointima and outflow vein harvested for bulk RNA-seq, four JAA anastomotic neointimas were combined. B: GO enrichment analysis of the anastomotic neointima and outflow vein. C: heatmap showing the fold changes of different genes. GO, gene ontology; JAA, juxta-anastomotic area.

Figure 9.

Gene expression in the juxta-anastomotic area compared with the outflow vein.

Validation Using the Rat Jugular Vein to Carotid Artery AVF Model

We validated the data in the mouse AVF model using the rat jugular vein to carotid artery AVF model, since the sutured anastomosis is still the most commonly used method to create AVF in humans. We previously showed the presence of JANIH in the rat AVF model after day 7 (24, 48); therefore, the rat JAA was harvested at 6 or 24 h to examine potential endothelial cell loss and acute thrombus that might precede JANIH. Gross examination showed dilation of the JV at 6 h and 24 h (Supplemental Fig. S6A), and EVG staining showed mural thrombus on surface of the transected CA and JV (Supplemental Fig. S6A), which is similar to the mouse needle puncture AVF (Fig. 2, Supplemental Fig. S3). Immunofluorescence showed CD31- and α-actin-positive cells in the baseline CA and JV (Supplemental Fig. S6B); however, CD31-positive cells were lost, and mural thrombus formed at 6 h at the JAA (Supplemental Fig. S6B), which is also similar to the mouse needle puncture AVF. These data show disturbed flow at the JAA with focal EC loss followed by mural thrombus formation in a sutured anastomotic AVF model, similar to the needle puncture AVF model.

DISCUSSION

We show that the JAA of the mouse AVF, immediately adjacent to the fistula, is actively regulated both temporally and spatially during venous remodeling, in a different pattern compared with the more cranial outflow vein. In addition, rapid endothelial loss with disturbed flow at the JAA, in the wall contralateral to the fistula channel, is followed by thrombus deposition and then NIH. The JAA also corresponds to areas of disturbed flow predicted by computational modeling and unbiased transcriptomics. As such, the JAA may be even more relevant for translational studies to improve AVF outcomes compared with the rest of the outflow vein.

Remarkably, despite AVF being the best vascular access for human patients requiring hemodialysis, there is only a 50% patency rate at 1 yr; this low patency rate may be the sum of three distinct failure modes with early AVF failure due to acute thrombosis, intermediate AVF failure due to failure to mature, and late AVF failure due to NIH (1, 49). After AVF creation, collagen and elastin are exposed by the arteriotomy, vein transection, and anastomosis; in addition, EC injury and detachment are potentiated by the sudden exposure to increased magnitudes of disturbed blood flow, all of which can cause acute thrombus formation in the JAA (50). Our data show that early thrombus in the JAA was uniformly present among all AVF and colocalized with later NIH, with thrombus gradually replaced by SMC with high proliferative capability (Figs. 4 and 6). However, there does not appear to be a clear separation of timing between the early thrombus and late NIH in this mouse model; both anastomotic and contralateral wall thrombus appear to be gradually replaced predominately by SMC and EC in this model (Figs. 4–6). We used day 7 to divide the thrombus and neointima phases in this model (Supplemental Fig. S7) based on the temporal staining of different cell markers (Figs. 4–6). The EC loss on the contralateral wall occurs as early as 6 h after AVF creation (Fig. 5), which is consistent with cell loss due to the large mechanical force of the sudden increased disturbed flow. Needle puncture is not likely to induce the cell loss on the contralateral wall because the needle tip does not contact the contralateral wall. We do not have data regarding apoptosis, but since this process typically takes at least several hours (51), we cannot measure conventional markers of apoptosis without the cells in the face of such rapid loss. As such, this mouse model of AVF recapitulates both failure modes of human AVF, acute thrombus formation, and late JANIH.

The complex temporal and spatial venous remodeling at the JAA reflects changes both at the anastomosis as well as in the contralateral wall and may be due to changes in resident vascular cells, infiltrating cells, as well as changes in gene expression, driven by the sudden exposure to high-pressure disturbed flow, endothelial loss, as well as formation of thrombus. Although other models of neointimal hyperplasia show more homogeneous lesions (24, 48, 52–57), these models may have similar complexity that is not yet documented. A feature of the mouse AVF model is the large increased magnitude and frequency of the arterial flow into the venous system; although altered hemodynamics are a major mechanism of the adaptive venous remodeling (38), it is likely that a focal response to venous injury is also present that contributes to the remodeling. As such, other mechanisms, such as regulation of proliferation and apoptosis (58), cell migration (59), deposition of extracellular matrix (60), and others, are likely to be important regulators of adaptive venous remodeling as well as JANIH.

We used a computational model of disturbed flow to validate the histological data; this model showed that disturbed flow is mainly located at the JAA and the channel, which correlates with the neointimal area (Figs. 2 and 3). Quantification of the disturbed flow and turbulence intensity using the Wall-Adapting Local Eddy-viscosity model is not well documented in this mouse model; although comparison with other methods, such as oscillatory shear index or the spectral broadening index (21), would be useful, documentation of shear stress within the channel is not reproducible with the ultrasound hardware to allow validation. Similarly, a limitation of this computational model is that the boundary conditions are based upon ultrasound measurements of the velocities in the outflow tract, without consideration of the flow contribution below the fistula, as this was not able to be accurately measured with our ultrasound hardware. In addition, the model segmentations are unable to capture microscopic changes in JANIH that are observed at the histological level. Rather, the purpose of segmentation is to establish global geometry, and to create a computational framework that can be used to study hemodynamic changes; these hemodynamic changes can then be compared with biological changes as potential explanations for biological responses. Interestingly, although our model showed differences between the inflow artery and the outflow vein, the inflow artery showed little neointimal hyperplasia compared with that seen in the outflow vein (Fig. 2); we believe that the turbulence intensity of the aorta, a function of velocity, appears high (red color in Fig. 3) due to the higher velocities of flow in the aorta. However, the lack of gradient in turbulence suggests that there is less disturbed flow in the aorta compared with the IVC. In addition, the increased capacity of venous cells to form neointimal lesions, compared with arterial cells, has been previously shown (61–64). Nevertheless, the potential of computational modeling as a noninvasive method to show development of altered flow patterns may predict JANIH and the risk of AVF failure for translational applications.

Although the sutured anastomosis is still the most commonly performed surgical AVF, endovascular AVF creation, such using the WavelinQ and Ellipsys devices, has showed promising clinical results (65–67). Interestingly, endovascular AVF creation uses a focal puncture that avoids the general vessel injury that is associated with surgical dissection and sutures (68, 69); thus, the mouse needle puncture model mimics endovascular AVF creation to a greater degree than it does traditional open surgical models. As such, this mouse AVF model may help understand how the endovascular AVF matures and remodels.

In conclusion, disturbed flow at the JAA correlates with early EC loss, early thrombus formation, and later NIH, with increased NIH at the JAA compared with the outflow vein during venous remodeling. This suggests the increased importance of using the JAA, rather than the outflow vein, to assess potential data for translational applications.

DATA AVAILABILITY

The raw sequencing data and the results of the DEG analyses generated in this study have been submitted to the NCBI Gene Expression Omnibus (GEO) database under Accession No. GSE249267.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S7: https://doi.org/10.6084/m9.figshare.25293877.v1.

Supplemental Tables S1–S15: https://doi.org/10.6084/m9.figshare.25293877.v1.

GRANTS

This work was supported by National Health, Lung, and Blood Institute Grants R01-HL162580 and R01-HL144476 (to A. Dardik), and the resources and the use of facilities at Veterans Affairs Connecticut Healthcare System, West Haven, CT.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

H.B. and A.D. conceived and designed research; H.B., M.A.V., and C.T. performed experiments; H.B., M.A.V., C.T., Y.O., and A.D. analyzed data; H.B., M.A.V., C.T., Y.O., and A.D. interpreted results of experiments; H.B., M.A.V., C.T., Y.O., and A.D. prepared figures; H.B., M.A.V., C.T., Y.O., and A.D. drafted manuscript; H.B., M.A.V., C.T., Y.O., L.G., W.Z., Z.L., L.P., and A.D. edited and revised manuscript; H.B., M.A.V., C.T., Y.O., L.G., W.Z., Y.A., M.K., B.Y., Z.L., L.P., and A.D. approved final version of manuscript.