Alterations of gut microbiota are common in chronic kidney disease (CKD) and contribute to increased uremic toxins including indoxyl sulfate (IS), p-cresyl sulfate (pCS) and trimethylamine N-oxide (TMAO), which are linked to cerebrovascular disease risk. This study examined the kidney–gut–brain axis in CKD mice and in dialysis patients. Male and female mice with adenine-induced CKD were fed a high-amino-acid (HAA) diet to increase precursors of gut-derived uremic toxins. A subgroup of mice received antibiotics in drinking water to suppress gut microbiota and evaluate its role in toxin generation. Behavior tests, gut microbiome composition and brain histology for cerebral microhemorrhages were analyzed. CKD mice had higher serum levels of creatinine, cystatin C and gut-derived toxins, a 2.5-fold increase in brain microhemorrhages, and decreased locomotor activity. The HAA diet significantly increased serum TMAO but not IS and pCS, and all three toxins were reduced by antibiotic therapy. Sex differences were observed; in male animals, higher TMAO was associated with increased brain microhemorrhages, whereas in female mice, pCS was associated with brain microhemorrhage burden. The suppression of toxins with antibiotics improved working memory in male animals. Gut microbiota analysis revealed the expansion of Lactobacillus and Ileibacterium in CKD mice. The HAA diet and antibiotics altered gut microbiota composition without changing alpha diversity. The human study utilized biobanked serum samples and a retrospective review of brain imaging scans in a hemodialysis patient cohort; TMAO levels were associated with increased lobar microbleeds. Our study supports a role for bacterial-derived uremic toxins in the kidney–gut–brain axis and cerebral microhemorrhage formation in CKD.

Mouse Study

Adenine CKD Model and Controls

Male and female C57BL/6J mice from Jackson Laboratories (Bar Harbor, ME, USA) aged 10–12 weeks were randomly assigned to experimental groups. For CKD, a nephrotoxic adenine diet was used to induce tubulointerstitial nephritis as previously described. Mice were fed a diet containing 0.2% adenine for 18 days, placed back on regular chow vs. special diet for 2 weeks, and then re-exposed to the adenine diet for 1 week to maintain CKD. Control mice were fed regular chow. Behavior studies and tissue collection were carried out 5–6 weeks after CKD induction and a comparable time in controls. All animal procedures were approved by the Institutional Animal Care & Use Committee at the University of California-Irvine (UCI), under protocol AUP-19-015. The procedures were conducted in accordance with the American Veterinary Medical Association guidelines.

Experimental Diets and Antibiotic Cocktail

Envigo 2020X was used as the base rodent diet (regular chow). CKD induction utilized a 0.2% adenine diet (Dyet# 611732, Dyets Inc., Bethlehem, PA, USA). The high-amino-acid (HAA) diet (Dyet# 611735) was enriched with precursors of indoxyl sulfate, p-cresyl sulfate and TMAO (1.26% tryptophan, 2.88% phenylalanine and 0.6% choline, respectively), providing 3- to 6-fold higher amounts of these dietary nutrients (usual concentrations are 0.2%, 1.0% and 0.12% respectively). For the group of CKD mice assigned to the HAA diet, a 0.2% adenine HAA chow was given in the 1-week adenine re-exposure period (Dyet# 611736, see Figure 6). To suppress intestinal microbiota, subgroups of animals on the HAA diet were treated with broad-spectrum antibiotics in drinking water ×5 weeks following CKD induction; the antibiotic (Abx) cocktail included ampicillin 1 g/L, vancomycin 500 mg/L, neomycin sulfate 1 g/L and metronidazole 1 g/L. Thus, we had six experimental groups—CTL, CTL + HAA, CTL + HAA + Abx, CKD, CKD + HAA, CKD + HAA + Abx—to which animals were randomly assigned. Control groups included 8 males and 8 females, while CKD groups included 16 males and 16 females.

Behavior Studies

(1) Spontaneous locomotion was assessed using the Open Field Test. The mouse was placed in an empty white box (12” × 12” with 12” high walls) and movements were tracked for 5 min to measure distance traveled and time in the central zone, and the total distance traveled. Data were analyzed using the SMART 3.0 Open Field Exp Module (Data Sciences International, St. Paul, MN, USA). Total distance traveled in the central zone was used as the primary index of spontaneous locomotion.

(2) Recognition (working) memory was assessed using the Novel Object Recognition Test. Habituation phase: On days 1–4, the mouse was allowed to explore the empty white box with no objects for 10 min. Familiarization phase: On days 5–6 the mouse was placed in the box with 2 identical objects in specific locations, and allowed to explore for 5 min. On day 7, the mouse was placed back in the testing box with one copy of the object presented previously along with a new, novel object. The discrimination index (DI) was calculated based on the difference in exploration time for the novel vs. the familiar object, dividing this value by the total amount of exploration of both objects. The DI varies between +1 and −1, where a positive score indicates more time spent with the novel object, a negative score indicates more time spent with the familiar object, and a zero score indicates a null preference.

Stool 16S rRNA Sequencing

Stool pellets collected from the colon at study termination were stored in 300 μL of RNA/DNA Shield (Zymo Research, Irvine, CA, USA). Samples were extracted at the Zymo Research facility (Irvine, CA, USA), and investigators were blinded to group allocation. Extraction of microbial DNA was performed with a bead-beating step using the ZymoBIOMICS^®-96 MagBead DNA Kit with an elution volume of 50 μL. Control samples: The ZymoBIOMICS^® Microbial Community Standard was used as a positive control for each DNA extraction, and a negative control (i.e., blank extraction control) was included to assess the level of bioburden carried by the wet-lab process. A DNA quality control check was made using Nanodrop^TM (Thermo Fisher Scientific, Waltham, WA, USA). The DNA was measured by Qubit and 1.5 ng was used for library construction according to the Illumina 16S Amplicon Library Preparation protocol (Illumina, San Diego, CA, USA) at the UC Irvine Genomics Research and Technology Hub. Libraries were assayed for quality using the Agilent Bioanalyzer 2100 DNA HS chip. The multiplexed library pool with a mock community control sample was sequenced on an Illumina Miseq instrument. Microbiome analysis was carried out at the UCI Microbiome Center. Amplicon sequence variants (ASVs) were inferred in QIIME 2 using DADA2 (qiime dada2 denoise-paired). Demultiplexed paired-end reads were quality filtered, denoised, merged, truncated at 280 (forward) and 202 bp (reverse) based on quality profiles, and screened for chimeras using the consensus method. DADA2 was used to correct sequencing errors and infer exact sequence variants. The resulting ASV table and representative sequences were used for downstream analyses. Taxonomical assignment was achieved using the SILVA database (138.99) trained with the primers used for amplification. Alpha and beta diversity metrics were analyzed within the Vegan package in R (version 2.6-4).

Quantification of Serum Toxins

Serum creatinine was measured using capillary electrophoresis at the O’Brien Kidney Research Core Center (UT Southwestern, Dallas, TX, USA). Cystatin C was measured using the Invitrogen Cystatin C (CST3) Mouse ELISA Kit (EMCST3, Fisher Scientific, Waltham, MA, USA). Gut-derived uremic toxins were quantified using high-performance liquid chromatography–mass spectrometry. A 20 μL aliquot of serum was treated with 200 μL of acetonitrile with 0.1% formic acid and internal standards (hydrochlorothiazide 2 μg/mL and salbutamol 5 ng/mL) for protein precipitation. The mixture was vortexed and centrifuged, and the supernatant was evaporated to dryness. The dried extract was reconstituted with 100 uL of 25% acetonitrile. Standards and prepared samples were injected (10 μL) into the HPLC-MS/MS instrument, a Waters Quattro Premier XE equipped with UPLC. For indoxyl sulfate (catalog# I3875, Sigma-Aldrich, St. Louis, MO, USA) and p-cresyl sulfate (p-tolyl sulfate, catalog# P2091, Fisher Scientific), the internal standard used was hydrochlorothiazide (catalog# 5001437615, Fisher Scientific) and analysis was performed in negative ionization mode using multiple reaction monitoring (MRM) MS/MS. For TMAO (catalog# 317594, Sigma), the internal standard used was salbutamol (catalog# S8260, Sigma) and analysis was performed in positive ionization mode. Uremic toxin standard curves were generated in 25% acetonitrile with hydrochlorothiazide 4 g/mL and salbutamol 10 ng/mL (serial 1/3 dilutions from highest standard 10,000 nM to lowest standard 1.5 nM). The transition (m/z) values are indoxyl sulfate 211.97 → 80.36, p-cresyl sulfate 186.94 → 107.30, TMAO 76 → 59, hydrochlorothiazide 296.96 → 270.08, salbutamol 240 → 148. CKD serum samples required 20× to 80× dilution to ensure that readouts fell within the standard curve. For human dialysis serum TMAO measurements, samples required 40× dilution.

Brain Microhemorrhage Histology

One brain hemisphere was sectioned into 20 μm coronal sections. Sections were collected every 200 µm and stained with Prussian blue to detect hemosiderin (a marker of iron deposition within microhemorrhages). Nuclear fast red was used as a counterstain. Digitized images were analyzed using NIH ImageJ software (version 1.54d) by an observer blinded to the experimental groups, and the number of microhemorrhages was normalized to total brain surface area analyzed per mouse (mm²/cm²). Average microhemorrhage size was also calculated (mm²).

Human Study

The retrospective human study utilized a subgroup of adult hemodialysis patients from the Malnutrition, Diet and Racial Disparities in Chronic Kidney Disease (MADRAD) Study (ClinicalTrials.gov NCT01415570) who had biobanked plasma samples under protocol 2012-9045, approved by the Institutional Review Board at UCI. Medical records at the UCI medical center were reviewed to identify participants who had brain MRI scans performed within 2 years of the archived plasma samples. Gut-derived uremic toxins were quantified using high-performance liquid chromatography–mass spectrometry, as described for the mouse samples. Microbleeds were quantified and localized on brain MRI (1.5T or 3T field strength) from 2 mm axial susceptibility-weighted imaging (SWI), by neuroradiologists who were blinded to patient data. Scores were recorded using the Microbleed Anatomical Rating Scale (MARS) and converted into binomial presence or absence of microbleeds for comparison among groups, as the majority of patients had no microbleeds.