Abstract Background Arteriovenous fistulae (AVFs) are the gold standard for vascular access in those requiring hemodialysis but may put an extra hemodynamic stress on the cardiovascular system. The complex interactions between the heart, kidney, and AVFs remain incompletely understood. Methods and Results We characterized a novel rat model of five‐sixths partial nephrectomy (NX) and AVFs. NX induced increases in urea, creatinine, and hippuric acid. The addition of an AVF (AVF+NX) further increased urea and a number of uremic toxins such as trimethylamine N‐oxide and led to increases in cardiac index, left and right ventricular volumes, and right ventricular mass. Plasma levels of uremic toxins correlated well with ventricular morphology and function. Heart transcriptomes identified altered expression of 8 genes following NX and 894 genes following AVF+NX, whereas 290 and 1431 genes were altered in the kidney transcriptomes, respectively. Gene ontology and Kyoto Encyclopedia of Genes and Genomes analysis revealed gene expression changes related to cell division and immune activation in both organs, suppression of ribosomes and transcriptional activity in the heart, and altered renin‐angiotensin signaling as well as chronodisruption in the kidney. All except the latter were worsened in AVF+NX compared with NX. Conclusions Inflammation and organ dysfunction in chronic kidney disease are exacerbated following AVF creation. Furthermore, our study provides important information for the discovery of novel biomarkers and therapeutic targets in the management of cardiorenal syndrome. Keywords: animal model, arteriovenous fistula, cardiorenal syndrome, chronic kidney disease, heart failure Subject Categories: Animal Models of Human Disease, Nephrology and Kidney, Pathophysiology, Cardiorenal Syndrome __________________________________________________________________ Nonstandard Abbreviations and Acronyms AVF arteriovenous fistula KEGG Kyoto Encyclopedia of Genes and Genomes NX five‐sixths partial nephrectomy Clinical Perspective. What Is New? * A novel rat model was developed combining five‐sixths partial nephrectomy and an arteriovenous fistula to study the complex interactions between heart failure, chronic kidney disease, and hemodialysis vascular access. * Based on a comprehensive serial physiological, histological, and transcriptomic assessment, we provide the first direct evidence that inflammation and organ dysfunction in chronic kidney disease are exacerbated following application of an arteriovenous fistula. What Are the Clinical Implications? * The profound impact of the arteriovenous fistula on the heart and kidney observed in this study may help explain the increased cardiorenal morbidity and mortality observed in the population undergoing hemodialysis. * This study provides important information for the discovery of novel biomarkers and therapeutic targets in the management of cardiorenal syndrome. Heart failure (HF) is a significant public health problem. Population‐based studies consistently show a prevalence 1.5% to 4% in developed countries[46] ^1 and estimate that this number will increase by 46% in 2030.[47] ^2 Despite improvements in diagnosis and management, costly hospitalizations for decompensated HF are common, and prognosis remains poor, with a 5‐year mortality rate around 40%.[48] ^3 The main barrier to improving outcomes are the comorbidities that many patients have, the most important being kidney dysfunction. According to a recent meta‐analysis, chronic kidney disease (CKD) is present in 49% of patients with HF and is associated with a 2.34‐fold higher risk of all‐cause mortality.[49] ^4 Furthermore, several clinical studies have demonstrated an exponential relationship between decreasing kidney function and cardiovascular mortality, even after adjustment for other established risk factors.[50] ^5 , [51]^6 Based on preclinical observations, there is an emerging understanding that CKD is a chronic, systemic proinflammatory state that mimics accelerated aging of the cardiovascular system through tissue remodeling, vascular and valvular calcification, and cellular senescence.[52] ^7 The finding that cardiovascular and renal diseases are both common, often coexist, and perpetuate each other has led to the recognition of “cardiorenal syndrome” as a distinct clinical entity.[53] ^8 An especially challenging situation is presented when patients with end‐stage kidney disease require hemodialysis. Arteriovenous fistulae (AVFs) are currently the gold standard for vascular access because of demonstrated reductions in infections, lower morbidity and mortality, and overall superior patency compared with other strategies.[54] ^9 However, the application of an AVF can induce a hyperdynamic circulation characterized by increased cardiac output and decreased peripheral vascular resistance.[55] ^10 , [56]^11 In patients with end‐stage kidney disease who already have considerable cardiovascular risk factors, this additional volume overload may contribute to adverse cardiac remodeling and the development of HF. In a study of 137 patients with end‐stage kidney disease undergoing AVF creation, Reddy et al[57] ^12 found that 53% had preexisting HF, while 43% of the remainder developed incident HF within a median of 2.6 years of hemodialysis. Those who developed HF furthermore demonstrated pronounced right ventricular (RV) dilatation and dysfunction and a 3.9‐fold increased risk of mortality. While the mechanisms of cardiorenal syndrome are starting to be untangled, it remains unclear how the AVF modulates the interaction between heart and kidneys. Nonetheless, a better understanding of the intricate relationship between heart, kidneys, and AVFs is essential to develop biomarkers and identify therapeutic targets in the management of this vulnerable patient population. In the present study, we therefore established a novel rat model of AVFs in cardiorenal syndrome and performed a comprehensive serial characterization of its physiological, histological, and transcriptomic features. Methods The data discussed in this article have been deposited in NCBI's Gene Expression Omnibus and are accessible through Gene Expression Omnibus Series accession number [58]GSE211019 ([59]https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE211019). In addition to the raw (FASTQ format) and processed (hit counts) expression data, lists of differentially expressed genes for each of the conditions have been deposited. All other data and supporting materials/methods are available within the article and its [60]supplementary files. Animal Procedures All experiments were performed in accordance with the European Commission Directive 2010/63/EU on the protection of animals used for scientific purposes. All animal experiments were approved by the Animal Ethics Committee at KU Leuven (P060/2018). A total of 29 female Sprague–Dawley rats (9 weeks; 225–249 g; Janvier Labs, Le Genest‐Saint‐Isle, France) were included in this study. The animals were randomly allocated to 3 groups: “AVF+NX” (full model, n=17), in which the animals underwent surgery for the placement of an AVF and a five‐sixths partial nephrectomy (NX); “NX” (partial model, n=7), in which the animals were subjected to NX only; and “Sham” (control, n=5), in which the stress of surgery was mimicked without performing an actual AVF or NX. An “AVF” group, including AVF only, was not included because it is not representative of any clinically relevant scenario (ie, patients will not receive an AVF if they do not have CKD). On the day of surgery, anesthesia was induced using a mix of ketamine (50 mg/kg) and xylazine (5 mg/kg) via intraperitoneal injection and maintained with isoflurane (2%–3% in 1 L/min oxygen). A median neck incision was made, followed by dissection of the right carotid artery and jugular vein. The contralateral blood vessels were also dissected, but no AVF was constructed such that it could serve as an internal control. The right sternocleidomastoid muscle was ligated at its upper and lower ends and subsequently removed. Vascular clamps were placed caudally and cranially on the carotid artery and caudally on the jugular vein. The dorsomedial branch of the vena jugularis was ligated and cut at its cranial end, after which the vein was rinsed with heparinized saline (50 IU/mL) and carefully stretched. A longitudinal incision was then made on the carotid artery with a size matching the diameter of the jugular vein, and the inside was rinsed with heparinized saline (Figure [61]1A). Two staying sutures were placed, followed by construction of the front anastomosis by means of interrupted stitches using a 10/0 suture (Figure [62]1B). The venous and arterial clamps were switched to complete the rear side of the anastomosis (Figure [63]1C and [64]1D). Perfusion was restored by opening the clamp on the vena jugularis first, then the cranial clamp on the carotid artery, and finally the caudal clamp on the carotid artery. Additional stitches were placed if necessary to prevent anastomosis leakage. Once hemostasis was achieved, blood flow through the bilateral jugular veins was measured using a 0.8‐mm perivascular flow probe (Transonic, Ithaca, NY) (Figure [65]1E and [66]1F). These procedures were performed using a surgical microscope (Zeiss, Jena, Germany). The neck incision was closed with resorbable 3/0 sutures. Figure 1. Surgical procedures. Figure 1 [67]Open in a new tab Microsurgical anastomosis of the cranial end of the right jugular vein to the side of the right carotid artery, as a model of brachial AVFs used for hemodialysis (A through F) and NX consisting of full removal of the right kidney and two‐thirds resection of the left kidney, as a model of CKD (G through I). AVF indicates arteriovenous fistula; CKD, chronic kidney disease; and NX, five‐sixths partial nephrectomy. The NX was performed in the same session. On the left side of the back, an incision was made, and the kidney was identified and was pushed externally on a wetted compress (Figure [68]2A). Both the upper and lower one‐third of the kidney were removed, and hemostasis was achieved using absorbable hemostatic sponges (Spongostan, Johnson & Johnson) (Figure [69]2B). The kidney was then returned to its pocket in the retroperitoneal fat (Figure [70]2C). On the right side, the whole kidney was removed after ligation of the hilus. Muscular layers and skin were closed with resorbable 5/0 and 4/0 sutures, respectively. Figure 2. Study design. Figure 2 [71]Open in a new tab AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. Once the surgery was completed, the animals were placed in a warm, dark environment in close proximity to food and water to awaken. Analgesia was administered via subcutaneous injection of buprenorphine (0.01–0.05 mg/kg) in the morning and evening for 2 days after the procedure or longer if needed. Throughout the study, the animals were housed in conditions with constant temperature and a 12/12‐hour day/night cycle. They received a diet of pellets with free access to water. Study Protocol The study followed a longitudinal repeated‐measures design (Figure [72]2). Cardiac magnetic resonance imaging (CMR) was performed at baseline (1 week before surgery) and at 1 and 3 weeks of follow‐up (after surgery). In addition, blood samples were obtained both at baseline and at 3 weeks of follow‐up. After the third week of the experiment, final tissue samples from the AVF, contralateral jugular vein, kidney, and heart were collected. All CMR data were acquired on a Bruker BioSpec 70/30 MRI system (Bruker BioSpin, Ettlingen, Germany) operating at a magnetic field strength of 7 Tesla and equipped with actively shielded gradients (200 mT/m). A dedicated rat quadrature volume coil with 86‐mm internal diameter (Bruker BioSpin) was used. Animals were anesthetized with isoflurane (2% in 1 L/min oxygen). Body temperature, respiration, and cardiac rate were monitored during the measurements and maintained at 36–37°C, 40–60 min^−1 and 310–360 min^−1, respectively. For planning of the CMR scans, localizer scans and 3 orthogonal 2‐dimensional scans (axial, sagittal, and coronal orientation) were acquired. Next, a FLASH self‐gated (IgFLASH, IntraGate Bruker BioSpin) 4‐chamber view scan was acquired with the following parameters: echo time 4.0 ms, repetition time 13.6 ms, flip angle 20 degrees, matrix 256×128, field of view 60×60 mm, 1 single slice of 1.5 mm thickness, oversampling 200. Finally, a multislice short‐axis IgFLASH was acquired with the following parameters: echo time 4.0 ms, repetition time 13.6 ms, flip angle 20 degrees, matrix 256×128, field of view 60×60 mm, 8 contiguous slices of 1.5 mm thickness, oversampling 350. The total acquisition time was ≈60 minutes. Images were reconstructed using the ParaVision software (Bruker BioSpin) version 6.0.1 to 25 frames after verifying retrospective triggering accuracy. To assess global left ventricular (LV) and RV functionality and remodeling, end‐systolic volume, end‐diastolic volume, stroke volume, and ejection fraction, and mass of each chamber, as well as cardiac index were calculated with Segment version 3.0 R8052 ([73]http://segment.heiberg.se).[74] ^13 Blood samples were taken into Li‐heparin tubes on dry ice and centrifuged at 4°C. Plasma had been isolated from the samples and stored at −80°C before assay for urea, creatinine, and a panel of uremic toxins and amino acids (p‐Cresyl glucuronide, phenyl glucuronide, indoxyl sulfate, p‐Cresyl sulfate, hippuric acid, phenyl sulfate, kynurenine, tryptophan, kynurenic acid, tyrosine, indole‐3‐acetic acid, phenylalanine, and trimethylamine N‐oxide). At the end of the experiments, the animals were again anesthetized with a mix of ketamine (50 mg/kg) and xylazine (5 mg/kg). Blood flow through the jugular veins was again measured using a 0.8‐mm perivascular flow probe (Transonic, Ithaca, NY). The animals were then euthanized by means of exsanguination. Before collection of the tissues, the vessels were perfused first with 0.2% adenosine (Sigma‐Aldrich, St. Louis, MO) to induce vasodilation and with 1.5% KCl to arrest the heart in diastole. For each of the tissues, 1 sample was fixed overnight in Zinc Formalin Fixative (Sigma‐Aldrich) for subsequent histological analyses, and another was snap‐frozen in liquid nitrogen and stored at −80°C for subsequent transcriptomic analyses. Histological Analysis Following overnight fixation, the samples for histology were washed with distilled water and transferred to 70% ethanol. They were then run overnight through a tissue processor (TP1020; Leica), embedded in paraffin, and sectioned at 5 μm with a microtome (HM360; Microm). Sections were stained with hematoxylin–eosin or Masson's trichrome and examined using an inverted bright‐field AxioVert 200 microscope (Carl Zeiss, Jena, Germany) and AxioVision software. RNA Sequencing and Bioinformatic Analysis Full details of RNA extraction, library preparation, and bioinformatic analysis are provided in Data [75]S1. Briefly, 4 heart and 4 kidney samples were randomly selected for each of group (AVF+NX, NX, and Sham). RNA was extracted from snap‐frozen samples and purified with RNeasy Plus Mini kit (Qiagen, Venlo, the Netherlands) according to the manufacturer's protocol. RNA sequencing libraries were prepared using poly(A) selection and strand‐specific RNA library prep kits (Illumina, San Diego, CA). RNA sequencing was performed via the Illumina NovaSeq protocol with 2×150 bp configuration. Statistical Analysis Physiological Data Analysis Quantitative data are expressed as mean± SD, and differences between groups were assessed by 1‐way ANOVA for unpaired data and repeated‐measures ANOVA for paired data. Pairwise comparisons were calculated using the Bonferroni post hoc test correction. Results are provided in the figures and text of the main manuscript, while Tables [76]S1 and S2 provide some of the underlying numerical data. P<0.05 was deemed statistically significant. All analyses were completed with R Statistical Software version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). Transcriptome Analysis Normalized read counts for individual genes were calculated from the transcripts per million and corrected for library size using DESeq2 normalization with the R package “DESeq2” (version 3.14).[77] ^14 Gene expression was then compared between groups (AVF+NX, NX, and Sham) within each tissue (heart and kidney) using negative binomial generalized linear models, setting the false discovery rate at α=0.05.[78] ^14 The Wald test was used to generate P values and log[2] fold changes. Genes that differed between AVF+NX and Sham, NX and Sham, and AVF+NX and NX (P<0.05 after adjustment for multiple comparisons, absolute log[2] fold change >1) were considered differentially expressed genes. Transcripts that were identified as differentially expressed genes in any comparison (AVF+NX versus Sham, NX versus Sham, AVF+NX versus NX) were then subjected to unsupervised hierarchical clustering and presented in heatmaps using the R packages “pheatmap” (version 1.0.12) and “clusterProfiler” (version 4.0.5). Transcripts per million was scaled using Z scores to allow for genes expressed at high and low abundance to be included in the same analysis and visualized on the same graph. The same transcripts were also subjected to gene ontology functional annotation (including biological processes, cellular components, and molecular functions) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis using the R package “clusterProfiler” (version 4.0.5).[79] ^15 Results Survival, AVF Failure, and Weight Gain The surgical procedures and study design, as described in the Methods section, are illustrated in Figures [80]1 and [81]2, respectively. Rats were generally healthy and active throughout the study and recovered within 2 to 3 days after surgery. One animal in the AVF+NX group died during anesthesia, while another in the NX group died 15 hours after surgery (Figure [82]3A). Three animals in the AVF+NX group had developed a stenosis of the AVF at the 1‐week follow‐up CMR scans (Figure [83]3B). Body growth was briefly suppressed during the first week after surgery, although without significant loss of body weight and with adequate restoration by the third week follow‐up (Figure [84]3C; Table [85]S1). Figure 3. Survival curves for overall survival (A) and freedom from AVF failure (B), and evolution of body weight during the course of the experiments (C). Figure 3 [86]Open in a new tab AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. ****P≤0.0001. CKD Development The animals that had undergone NX as part of the surgery (groups AVF+NX and NX) developed CKD, as evidenced by >2‐fold increased urea and creatinine levels compared with Sham animals at 3 weeks (all P<0.001; Figures [87]4A and [88]4B; Table [89]S2). Interestingly, urea levels were even higher in AVF+NX compared with NX (P=0.036; Figure [90]4A). Demonstrating a similar pattern, hippuric acid was elevated in both AVF+NX (P<0.001) and NX (P=0.030) compared with Sham at 3 weeks (Figure [91]4C; Table [92]S2). P‐cresyl glucuronide, phenyl glucuronide, indoxyl sulfate, phenyl sulfate, and trimethylamine N‐oxide were also increased compared with Sham, although this was significant only in the AVF+NX group (all P<0.05; Figure [93]4C; Table [94]S2). Figure 4. Evaluation of CKD development from plasma assays of urea (A), creatinine (B), and a panel of uremic toxins and amino acids (C). Figure 4 [95]Open in a new tab AVF indicates arteriovenous fistula; CKD, chronic kidney disease; NX, five‐sixths partial nephrectomy; and TMAO, trimethylamine N‐oxide. *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001. AVF Characterization Intraoperative assessment using a transonic flow probe revealed that blood flow through the AVF immediately after its construction was >10‐fold higher than in the contralateral jugular vein (from 2.25±0.65 to 32.9±4.85 mL/min; P<0.001; Figure [96]5A). By the time of euthanasia, flow through the AVF had increased further by almost 2‐fold (to 58.6±14.8 mL/min; P<0.001; Figure [97]5A). On postmortem histological examination, intimal area (Figure [98]5B) and luminal area (Figure [99]5C) were both increased in the AVF (0.35±0.18 mm^2 and 0.41±0.25 mm^2 at the proximal AVF, and 0.17±0.07 mm^2 and 0.35±0.28 mm^2 at the distal AVF) compared with the contralateral jugular vein (0.04±0.02 mm^2 and 0.07±0.07 mm^2; all P<0.001), suggesting eccentric vascular remodeling. Figure 5. Blood flow measurements (A) and histological analysis of the AVF in the AVF+NX model at the time of euthanasia (B and C). Figure 5 [100]Open in a new tab AVF indicates arteriovenous fistula; JV, jugular vein; and NX, five‐sixths partial nephrectomy. ***P≤0.001; ****P≤0.0001. Serial Measurements of Cardiac Morphology and Function LV end‐diastolic volume, LV end‐systolic volume, and LV stroke volume were significantly increased in AVF+NX compared with both Sham and NX at 3 weeks, with LV end‐diastolic volume already being increased compared with Sham and NX at 1 week and LV end‐systolic volume being increased compared with Sham at 1 week (all P<0.05; Figures [101]6A through [102]6C; Table [103]S1). RV end‐diastolic volume and RV end‐systolic volume were increased in AVF+NX compared with both Sham and NX at 3 weeks, while RV stroke volume was increased only compared with Sham at 3 weeks (all P<0.05; Figures [104]6D through 6F; Table [105]S1). Both LV ejection fraction and RV ejection fraction were preserved throughout the experiments for all groups (Figures [106]6G and [107]6H; Table [108]S1). Nonetheless, cardiac index was increased in AVF+NX compared with Sham at 1 week and compared with both Sham and NX at 3 weeks (all P<0.05; Figure [109]6I; Table [110]S1). While no changes in LVM were observed (Figure [111]6J), RVM was increased at 3 weeks compared with both Sham and NX, suggesting marked RV hypertrophy (all P<0.05; Figure [112]6K; Table [113]S1). RV mass was also increased in NX compared with Sham at 3 weeks (P=0.001; Figure [114]6K; Table [115]S1). Interestingly, several uremic toxins correlated with these markers of cardiac function (Figure [116]S1). Figure 6. Magnetic resonance measurements of cardiac function and morphology. Figure 6 [117]Open in a new tab (A) LVEDV, (B) LVESV, (C) LVSV, (D) RVEDV, (E) RVESV, (F) RVSV, (G) LVEF, (H) RVEF, (I) CI, (J) LVM, (K) RVM. AVF indicates arteriovenous fistula; CI, cardiac index; EDV, end‐diastolic volume; EF, ejection fraction; ESV, end‐systolic volume; LV, left ventricular; M, mass; NX, five‐sixths partial nephrectomy; RV, right ventricular; and SV, stroke volume. *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001. Finally, postmortem analysis confirmed a congestive phenotype for animals in the AVF+NX group, as reflected by elevated “wet” heart and lung weights compared with both Sham and NX (1.24±0.10 g versus 0.88±0.05 g and 0.81±0.12; and 1.60±0.11 versus 1.30±0.09 and 1.26±0.09; all P<0.001; Figures [118]7A and [119]7B). Figure 7. Postmortem measurements of “wet” heart (A) and lung weights (B). Figure 7 [120]Open in a new tab AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. **P≤0.01; ***P≤0.001; ****P≤0.0001. Transcriptomic Results Following quality control and data filtering, RNA sequencing generated 29.4±17.0 × 10^6 high‐quality read pairs per sample. For each sample, 64.7% to 89.3% reads could be mapped to the 23 139 unique coding genes in the rat genome. Gene expression levels were compared between groups (AVF+NX, NX, and Sham) within each tissue (heart and kidney) (all n=3–4). Differential Gene Expression Results from differential gene expression are presented in Figures [121]8, [122]9, [123]10 through [124]8, [125]9, [126]10. Regarding gene expression in the heart, a total of 8 genes differed between NX and Sham (P<0.05 after adjustment for multiple comparisons, absolute log[2] fold change >1; 25% up‐regulated in NX; Figure [127]8A, Figure [128]10), 894 genes differed between AVF+NX and Sham (33.3% upregulated in AVF+NX; Figure [129]8A, Figure [130]10), and 59 genes differed between AVF+NX and NX (40.7% upregulated in AVF+NX; Figure [131]8A, Figure [132]10). Figure 8. Genes altered in the heart of AVF+NX, NX, and Sham animals (n=1487, P<0.05 after adjustment for multiple comparisons, absolute log[2] fold change >1). Figure 8 [133]Open in a new tab (A) A total of 8 genes differed between NX and Sham, 894 genes differed between AVF+NX and Sham, and 59 genes differed between AVF+NX and NX. Percentages indicate the number of genes upregulated in the former of the 2 groups. (B) Heatmap representing mean transcripts per million (TPM) from all samples in the AVF+NX, NX, and Sham groups for genes differentially expressed between these groups. TPM was scaled using Z scores to allow for genes expressed at high and low abundance to be visualized on the same graph. For each gene, the color scale indicates high (red) or low (blue) expression relative to other groups. Green bars indicate significant differences between groups. Unsupervised hierarchical clustering identified 4 clusters with distinct patterns of expression across groups. (C) Box plots summarizing gene expression for all genes in each cluster in (B), based on the mean log TPM values for all samples in each group. AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. Figure 9. Genes altered in the kidney of AVF+NX, NX, and Sham animals (n=1487, P<0.05 after adjustment for multiple comparisons, absolute log[2] fold change >1). Figure 9 [134]Open in a new tab A, A total of 290 genes differed between NX and Sham, 1431 genes differed between AVF+NX and Sham, and 502 genes differed between AVF+NX and NX. Percentages indicate the number of genes upregulated in the former of the 2 groups. B, Heatmap representing mean transcripts per million (TPM) from all samples in the AVF+NX, NX, and Sham groups for genes differentially expressed between these groups. TPM was scaled using Z scores to allow for genes expressed at high and low abundance to be visualized on the same graph. For each gene, the color scale indicates high (red) or low (blue) expression relative to other groups. Green bars indicate significant differences between groups. Unsupervised hierarchical clustering identified 4 clusters with distinct patterns of expression across groups. C, Box plots summarizing gene expression for all genes in each cluster in (B), based on the mean log TPM values for all samples in each group. AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. Figure 10. Volcano plot for each of the comparisons between groups (AVF+NX, NX, and Sham) for heart and kidney tissues. Figure 10 [135]Open in a new tab Each data point in the scatter plot represents a gene. The log[2] fold change of each gene is represented on the x axis, and the log[10] of its adjusted P value is on the y axis. Genes with an adjusted P value <0.05 and a log[2] fold change >1 (upregulated genes) are indicated by red dots, while those with a log[2] fold change <−1 (downregulated genes) are indicated by blue dots. AVF indicates arteriovenous fistula; and NX, five‐sixths partial nephrectomy. Regarding gene expression in the kidney, 290 genes differed between NX and Sham (80.7% upregulated in NX; Figure [136]9A, Figure [137]10), 1431 genes differed between AVF+NX and Sham (94.1% upregulated in AVF+NX; Figure [138]9A, Figure [139]10), and 502 genes differed between AVF+NX and NX (99.2% upregulated in AVF+NX; Figure [140]9A, Figure [141]10). A total of 152 differentially expressed genes overlapped between the heart and the kidney, including 4 that were significant between AVF+NX/NX and Sham (25% upregulated in AVF+NX/NX), 147 that were significant between AVF+NX and Sham (83.7% upregulated in AVF+NX), and 1 that was only significant between AVF+NX and NX (100% upregulated in AVF+NX). The former 4 differentially expressed genes included aryl hydrocarbon receptor nuclear translocator like (Arntl/Bmal1), adrenoceptor alpha 1D (Adra1d), period 2 (Per2), and period 3 (Per3). Of these, Per2 and Per3 were upregulated whereas Arntl was downregulated in both organs of the AVF+NX and NX animals compared with Sham animals (Figure [142]S2). Adra1d was downregulated in the heart but upregulated in the kidney. All of these genes could be mapped, either directly or indirectly, to the circadian rhythm. The lower transcription of the Per genes suggest that oscillatory strength might be decreased in our animal model, a phenomenon known as chronodisruption, which has been linked to dysfunction of various organs, including the heart and kidney.[143] ^16 , [144]^17 This effect might be mediated through interaction of uremic toxins with the aryl hydrocarbon receptor[145] ^18 or effects on adrenergic signaling.[146] ^19 The transcript that overlapped between both organs in AVF+NX versus NX was retinoic acid receptor responder 1 (Rarres1), which encodes a type 1 membrane protein that can be upregulated by tazarotene as well as by retinoic acid receptors and has recently been implicated in the promotion of podocyte injury and glomerular disease.[147] ^20 , [148]^21 This molecule was upregulated in both heart and kidney of AVF+NX animals (Figure [149]S2). Gene Ontology and KEGG Analysis To better understand the functions and pathways involved in the clusters identified in Figures [150]8 and [151]9, we conducted gene ontology functional annotation and KEGG pathway enrichment analysis. Clusters 1 and 2 in the heart, which were downregulated in AVF+NX and NX compared with Sham, consisted of genes that localized to the ribosome and its subunits and were involved in transcriptional activity (Figures [152]S3B through [153]S3D and [154]S4B through [155]S4D). In addition, genes in cluster 1 were involved in the following KEGG pathways: oxidative phosphorylation, Parkinson disease, and thermogenesis (Figures [156]S3D and S4D). Functions and pathways altered in cluster 3, which was upregulated in AVF+NX and NX compared with Sham, included the cell cycle, chromosome segregation, nuclear division, and mitotic cytokinesis, and the extracellular matrix (Figure [157]S3). Cluster 4, which was also upregulated in AVF+NX and NX compared with Sham, linked to various processes suggestive of immune activation such as positive regulation of cytokine production and neutrophil chemotaxis (Figures [158]S3A and [159]S3B, [160]S3D, [161]S4A and [162]S4B, and [163]S4D). Interestingly, periostin (Postn) and several members of the C‐type lectin family (eg, Clec5a, Clec7a) were included in this cluster, the former of which has been linked to myocardial fibrosis[164] ^22 and the latter of which has been strongly linked with activation of the nucleotide‐binding oligomerization domain‐like receptor protein 3 inflammasome and pyroptotic cell death in the pathophysiology of myocardial infarction.[165] ^23 , [166]^24 Regarding the kidney, cluster 1 was downregulated in AVF+NX and NX compared with Sham and included genes (Figures [167]S5D and [168]S6D) that were linked to the renin‐angiotensin system, such as several kallikrein 1–related peptidases (Klk1c2, Klk1c6, Klk1c12) and renin (Ren). This in agreement with clinical studies that found reduced plasma renin activity in patients with an AVF.[169] ^25 Interestingly, a recent study in mice found that deletion of Ren or inhibition of the renin‐angiotensin system resulted in concentric thickening of the intrarenal arteries and arterioles through transformation of renin cells from a classical endocrine phenotype to a matrix‐secretory phenotype.[170] ^26 The present finding might therefore reflect kidney vascular damage in our animal model. Paralleling clusters 3 and 4 in the heart, clusters 2 and 3 in the kidney, which were upregulated in AVF+NX and NX compared with Sham, represented cell division and immune activation, respectively (Figure [171]S5). Interestingly, among the genes included in cluster 3 was apolipoprotein E (Apoe), whose protective roles against atherosclerosis and CKD progression are well known.[172] ^27 , [173]^28 Cluster 4 in the kidney was downregulated in NX compared with Sham and tended to be partially “rescued” in AVF+NX (Figure [174]9C). This cluster represented genes involved in the circadian rhythm (eg, Per1, Per2, Per3), which was a hallmark pathway that overlapped between the kidney and the heart, as described above (Figures [175]S5A, S5C, S5D, S6A, S6C and [176]S6D). In line with this, ubiquitin‐specific peptidase 2 (Usp2), which regulates the intracellular location of the Per genes,[177] ^29 and nocturin (Noct), which regulates metabolism under the control of the circadian clock,[178] ^30 were also included in this cluster. Finally, gene ontology and KEGG analyses were conducted for the set of genes that overlapped between both organs (Figures [179]S7 and [180]S8). These analyses recapitulated the processes of cell division, immune activation, transcriptional suppression, and the circadian rhythm. Discussion In the present study, we harnessed a novel rat model combining NX and AVF to study the intricate relationship between the heart, kidneys, and AVF. The key findings are summarized in Figure [181]11. Serial measurements before and after the operation showed that NX induced increases in urea, creatinine, and hippuric acid. The addition of AVF further increased the levels of uremic toxins and led to a hyperdynamic circulation characterized by increased cardiac index, left and right ventricular volumes, and RV mass. Transcriptomic analyses of the kidney and heart revealed enrichment in genes related to cell division, immune activation, transcriptional suppression, altered renin‐angiotensin signaling, and chronodisruption. All of these processes except for the latter were worsened in AVF+NX compared with NX. These data suggest that inflammation and organ dysfunction in CKD are exacerbated following AVF creation, thereby potentially accounting for the increased cardiorenal morbidity and mortality observed in the population undergoing hemodialysis. Figure 11. Summary of the key study findings. Figure 11 [182]Open in a new tab AVF indicates arteriovenous fistula; CI, cardiac index; CKD, chronic kidney disease; EDV, end‐diastolic volume; EF, ejection fraction; ESV, end‐systolic volume; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; LV, left ventricular; M, mass; NX, five‐sixths partial nephrectomy; RV, right ventricular; SV, stroke volume; and TMAO, trimethylamine N‐oxide. Hallmarks of Cardiorenal Syndrome Our NX model recapitulated many of the features observed in human CKD. Subtotal NX is the most frequently used model of CKD in rodents and has been shown to exhibit hypercreatinemia, albuminuria, glomerulosclerosis, tubular injury, interstitial fibrosis and inflammation, hypertension, anemia, endothelial dysfunction, cardiac hypertrophy, cardiac fibrosis, and decreased systolic and diastolic function in multiple studies.[183] ^31 , [184]^32 , [185]^33 Apart from confirming hypercreatinemia, our present study also found elevated urea and hippuric acid in the NX model. While traditionally considered to be biologically inert, recent experimental data support roles for urea in driving insulin resistance, free radical production, apoptosis, and disruption of the protective intestinal barrier.[186] ^34 Furthermore, urea is the main source of cyanate, ammonia, and carbamylated compounds, which have been linked to atherogenesis and the pathophysiology of cardiovascular disease.[187] ^35 Accumulation of hippuric acid has also been observed in human CKD, showing almost 24‐fold higher plasma levels compared with healthy people.[188] ^36 This toxin can lead to kidney fibrosis mainly through disruption of antioxidant systems and promotion of reactive oxygen species.[189] ^37 It has now been well recognized that uremic toxins constitute major drivers of cardiovascular disease and CKD progression.[190] ^38 Consistent with previous transcriptomic studies,[191] ^39 , [192]^40 , [193]^41 , [194]^42 , [195]^43 our investigation of the NX model identified changes in gene expression that were consistent with immune activation and inflammation. These data add to work by Fan et al[196] ^40 in kidney biopsies of patients with advanced diabetic nephropathy. They found upregulation of genes associated with the immune response, which correlated inversely with estimated glomerular filtration rate. Several of the differentially expressed genes in their study were also identified in our analysis, most notably including CC‐chemokine ligand 19 (Ccl19) and its receptor 7 (Ccr7), which are essential in establishing dendritic cell and T cell recruitment[197] ^44 ; several members of the C‐type lectin family (eg, Clec5a, Clec7a), which have been strongly linked with activation of the nucleotide‐binding oligomerization domain‐like receptor protein 3 inflammasome and pyroptotic cell death[198] ^23 , [199]^24 ; SLAM family member 6 (Slamf6) and glycoprotein CD2 (Cd2), both of which belong to the immunoglobulin superfamily and are expressed on natural killer, T, and B lymphocytes[200] ^45 ; interferon regulatory factor 4 (Irf4), which has been implied in kidney fibrosis as it may regulate the migration and activation of macrophages[201] ^46 ; surface receptor sialic acid binding Ig‐like lectin 1 (Siglec1), which regulates monocyte–macrophage activation and has been linked to increased frequency of kidney complications in patients with systemic lupus erythematosus[202] ^47 ; and members of the matrix metalloproteinase family (eg, Mmp2, Mmp7, MMp12), which play and important role in the degradation of the extracellular matrix and have been linked to CKD progression.[203] ^48 These findings are also well in accordance with the study by Tajti et al,[204] ^39 who confirmed the role of inflammation across a wide spectrum of CKD entities. Chronic inflammation and fibrosis are indeed key features of dysfunctional tissue repair culminating in irreversible organ dysfunction, as occurs in CKD progression and cardiovascular disease. A striking finding was that those genes that were differentially expressed in NX compared with Sham in both the heart and the kidney (Per2, Per3, Arntl/Bmal1, Adra1d) could all be mapped, either directly or indirectly, to the circadian rhythm. In normal physiology, almost all cell types within the human body possess their own peripheral clock consisting of complex autoregulatory transcription‐translation feedback loops.[205] ^16 , [206]^17 The rhythmic interplay of these components (which include Per2, Per3, and Arntl/Bmal1) leads to the rhythmic activation of clock‐controlled genes the govern divergent functions in tissue homeostasis. Any disruptions of these oscillations, referred to as chronodisruption, can result in disease. Numerous studies have demonstrated the association of chronodisruption with the development of atherosclerosis, cardiac dysfunction, hypertension, CKD, and kidney fibrosis.[207] ^16 , [208]^17 The expression pattern in our study (downregulation of Per2 and Per3) was consistent with chronodisruption and might be mediated through interaction of uremic toxins with the aryl hydrocarbon receptor[209] ^18 or effects on adrenergic signaling.[210] ^19 Taken together, these data suggest that accumulation of uremic toxins, immune activation, and chronodisruption are major drivers of CKD progression and cardiac dysfunction alike and can therefore be regarded as hallmarks of the cardiorenal syndrome. Influence of AVF on Cardiorenal Interactions Although the effect of NX and the general mechanisms of cardiorenal syndrome have been relatively well examined, conclusions regarding the impact of the AVF on the interaction between both organs are far less definitive. However, several clinical observations have suggested a detrimental effect. Even in their report of the first surgically created AVF for hemodialysis in 1966, Brescia and Cimino acknowledged that it led to significant cardiac strain in the long term and could potentially cause HF and arterial “steal.”[211] ^49 Subsequent studies estimated that AVFs with a flow between 1 and 1.5 L/min or a flow that is >20% of the CO result in the greatest predisposition to HF.[212] ^50 Of note, this degree of flow corresponds to AVFs in the upper arm (flow=0.9–1.5 L/min, which corresponds to flow/CO=18–30% assuming a CO of 5 L/min) of clinical patients[213] ^51 and the AVFs created in our animal model (flow/CO=30%–50%). The creation of an AVF thus induces a hyperdynamic circulation, which may initially be tolerated by eccentric hypertrophy and a decrease in peripheral vascular resistance.[214] ^10 , [215]^11 Several studies have demonstrated enlarged right and left ventricular volumes,[216] ^25 , [217]^52 , [218]^53 , [219]^54 , [220]^55 as has been recapitulated in our animal model. However, the above studies have also reported progressive increases in central hemodynamic pressures and markers suggestive of diastolic dysfunction. Of note, increased filling pressures significantly increase myocardial workload and energy consumption. Using pulse wave analysis, Savage et al[221] ^53 studied subendocardial viability ratio—a surrogate marker of oxygen supply and demand, with low values indicating ischemia—was decreased following AVF creation and improved immediately following AVF ligation. Using the ratio of diastolic pressure time index over systolic pressure time index as a marker of oxygen supply and demand, Bos et al[222] ^52 similarly found that coronary perfusion was negatively affected by AVF creation and improved when the AVF was compressed. Importantly, these relationships were found to be relevant even for AVF with low flow. The histopathological finding of eccentric remodeling of the AVFs in our present study, in conjunction with increases in AVF flow over time, progressive ventricular dilatation, and RV hypertrophy, suggest that maturation of the AVF may eventually shift hemodynamics toward a situation that is no longer sustainable for the heart. Moreover, associated factors such as arterial “steal,” anemia, and uncontrolled hypertension may eventually precipitate overt HF in these patients. In the study by Reddy et al,[223] ^12 this turned out to be the case in 43% of those who did not have preexisting HF, within a median of 2.6 years of hemodialysis. Whereas all of these reports have provided anecdotal evidence that AVF creation may be associated with progressive cardiac and kidney dysfunction, the current study presents direct serial functional, histological, and transcriptomic data showing that (1) a hyperdynamic circulation was established by the AVF that was furthermore associated with signs of congestion; (2) this occurred in accordance with increases in the plasma levels of several uremic toxins beyond the levels observed in NX only; and (3) inflammation in the animal model was further exacerbated in the AVF+NX model, with many genes having a 2‐fold higher expression than in NX only, and this time even having a profound impact on gene expression in the heart. This begs the question how the cardiac phenotype, the uremic milieu, and inflammation in our AVF+NX model all relate to one another. The association of congestion with worsening kidney function has been described well in the setting of acute decompensated HF. In a foundational experiment in 1988, Firth et al[224] ^56 demonstrated that increases in kidney venous pressure proportionately reduced kidney perfusion pressure, fractional sodium excretion, and glomerular filtration rate in isolated perfused rat kidneys. Subsequent human studies have confirmed this finding. Damman et al[225] ^57 found that both kidney blood flow and right atrial pressure were independently associated with glomerular filtration rate in 51 patients with cardiac dysfunction, whereas cardiac index was not. In an analysis of 145 patients with acute decompensated HF included in the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial, Mullens et al[226] ^58 found that greater central venous pressure, but not cardiac index, was the most important hemodynamic predictor of worsening kidney function, defined as a serum creatinine increase ≥0.3 mg/dL during hospitalization. Worsening kidney function has furthermore been linked to a 1.62‐fold increased risk of death and hospitalization in patients with HF.[227] ^59 It thus stands to reason that congestion, as induced by the AVF in our rat model, may have accelerated CKD progression. Importantly, the levels of uremic toxins in our study correlated with markers of cardiac remodeling and congestion, further supporting this notion. Congestion of the kidneys alone does not, however, explain the large extent to which uremic toxins such as p‐cresyl glucuronide, phenyl glucuronide, indoxyl sulfate, phenyl sulfate, and trimethylamine N‐oxide were increased in AVF+NX animals compared with NX animals. Many of these uremic toxins, most notably indoxyl sulfate and trimethylamine N‐oxide, are end products of bacterial protein fermentation in the gut.[228] ^60 Impaired filtration capacity in CKD initially leads to accumulation of small water‐soluble molecules such as urea and creatinine, which establish a uremic milieu.[229] ^61 These changes subsequently seem to impair intestinal barrier function and promote inflammation throughout the gastrointestinal tract, allowing for various toxins originating from the intestinal microflora to enter the bloodstream, a phenomenon referred to as the “gut‐kidney axis” or the “leaky guts hypothesis.”[230] ^62 When an AVF is created, venous congestion not only of the kidneys but also of the intestines ensues along with reduced intestinal perfusion attributable to arterial “steal,” both of which can exacerbate intestinal barrier dysfunction. In fact, a recent line of research has provided evidence of intestinal barrier dysfunction in a wide range of cardiovascular pathologies characterized by congestion. Direct experimental evidence has been provided by Huo et al,[231] ^63 who demonstrated that mice with HF had intestinal villus damage, decreased expression of the tight junction proteins occludin and zonula occludens 1, and increased permeability for dextran, d‐lactate, and lipopolysaccharide. Interestingly, the latter was associated with increased sympathetic activity, fibrosis, and inflammation. In 29 patients with advanced decompensated HF, Kitai et al[232] ^64 found significantly impaired intestinal barrier function, which was associated with elevated right atrial pressure, an important marker of congestion. Similar findings have been reported by Sandek et al[233] ^65 in 22 patients with chronic HF. Although no intestinal samples were evaluated as part of our present study, collectively these findings suggest that increased intestinal permeability in the setting of CKD may be exacerbated by venous congestion attributable to AVF creation and may contribute to the origin of inflammation. While the AVF is created with the intention to facilitate dialysis, our findings suggest that its hemodynamic sequelae may in fact accelerate the accumulation of uremic toxins and thereby contribute to cardiorenal deterioration. This is especially alarming because many of these toxins are protein bound and thus not easily removable via dialysis.[234] ^66 Once translocated in the bloodstream, uremic toxins may (1) further increase damage to the intestinal barrier, (2) lead to CKD progression, and (3) induce cardiac dysfunction, which may eventually culminate in HF. The putative mechanisms through which uremic toxins effect these changes have been linked to the activation of the nuclear factor κ‐light‐chain enhancer of activated B cellsB, transforming growth factor‐β, and nicotinamide adenine dinucleotide phosphate oxidase signaling pathways—all of which were also upregulated in our study—converging in the generation of reactive oxygen species, inflammation, endothelial dysfunction, and cardiotoxicity.[235] ^38 Furthermore, meta‐analyses of prospective clinical cohorts have linked elevated plasma levels of several uremic toxins to cardiovascular events and all‐cause mortality.[236] ^67 , [237]^68 The mechanisms established in our present study may thus help explain the clinical observation that patients with an AVF are at an elevated risk of adverse ventricular remodeling and eventually HF. Future Perspectives The data generated in this study provide a valuable tool for future investigations to establish novel biomarkers to diagnose, monitor, and predict the course of cardiorenal syndrome. Future studies may wish to evaluate the effect of AVF ligation or surgical strategies to reduce flow on the transcriptomes of the heart and kidney. Furthermore, we have generated insights that could inspire future pharmacological modulation in cardiorenal syndrome. In this regard, various randomized controlled trials have now demonstrated the benefits of sodium glucose cotransporter 2 inhibitors in terms of mortality, HF hospitalizations, and kidney events across a wide range of cardiovascular diseases and CKD etiologies.[238] ^69 , [239]^70 Interestingly, experimental evidence has shown that their benefits are not limited to a diuretic and glucosuric effect but extend to increased hematocrit with enhanced myocardial oxygen delivery, improved cardiac energy efficiency, reduced kidney metabolic stress and fibrosis, and attenuation of inflammation, autophagy, and hypertrophy.[240] ^69 These mechanisms were also shown to be impaired in our NX model and further worsened in our AVF+NX model. Given the role of the intestinal microbiota in AVF‐induced exacerbation of uremia, therapies such as prebiotics and synbiotics may deserve further investigations, with early evidence demonstrating that these may reduce the levels of p‐cresyl sulfate in patients with CKD.[241] ^71 Limitations Our study is limited by a number of factors that merit consideration. First, the effects of repeated puncture of the venous segment of the AVF as well as fluid shifts and electrolyte replacements that would be observed during the course of dialysis could not be assessed in our animal model. However, patients usually undergo a 3‐month period where the AVF is allowed to mature before the initiation of dialysis; this was the clinical setting that we aimed to simulate in our animal model. Second, although NX is the most frequently used model of CKD in rodents, reductions in kidney function in this model are established relatively acutely as opposed to the usual progressive onset of the disease in human patients. Third, only a subset of samples was submitted for RNA sequencing such that correlation of gene expression with functional data was not possible. Fourth, analyses of intestinal permeability were not performed. We could thus not provide direct evidence for the role of the “gut‐brain axis” in our animal model; however, substantial evidence supporting this notion has been reported by other groups, and future work may further investigate this. Fifth, for a limited number of animals, not all serial CMR scans could be evaluated because of poor imaging quality. Finally, while the present experiments identified molecular pathways and processes that may play an important role in the cardiorenal syndrome, further validation is warranted. Conclusions In conclusion (see Figure [242]11), we have characterized the complex interaction between the heart and kidney as well as the modulating effect of AVFs in cardiorenal syndrome. Based on a comprehensive serial physiological, histological, and transcriptomic assessment, we provide the first direct evidence that inflammation and organ dysfunction in CKD are exacerbated following application of an AVF. These factors may drive the increased cardiorenal morbidity and mortality observed in the population undergoing hemodialysis. Furthermore, our study provides important information for the discovery of novel biomarkers and therapeutic targets in the management of cardiorenal syndrome. Sources of Funding The research presented in this work was supported by a grant from the Fund for Cardiac Surgery (project “New strategies in the management of heart failure: metabolic imaging and gene targeting for a volume‐overload rat model”) and a KU Leuven financing grant (C14/20/095). Disclosures None. Supporting information Data S1 Tables S1–S2 Figures S1–S8 [243]Click here for additional data file.^ (2.4MB, pdf) Acknowledgments