**1. Introduction**

It is becoming increasingly evident that the gu<sup>t</sup> microbiome plays a role in various diseases such as inflammatory bowel disease, diabetes, autoimmune diseases, and cancer [1]. However, less is known about the role of the gu<sup>t</sup> microbiome in the field of renal transplantation. Renal transplantation is the best available treatment for patients with end-stage renal disease (ESRD). Despite improved prognosis and quality of life (QoL) compared to dialysis treatment, renal transplant recipients (RTRs) su ffer from many problems in the years after transplantation. After transplantation one out of five RTRs su ffers from chronic diarrhea which is associated with a lower QoL, increased abdominal complaints, higher mortality, and gu<sup>t</sup> dysbiosis [2–4]. Furthermore, all RTRs use immunosuppressive drugs and frequently require antibiotics which potentially influence the gu<sup>t</sup> microbiome [5]. Chronic diarrhea and the use of immunosuppressive drugs may change the gu<sup>t</sup> microbiota composition. As a consequence, this can disrupt gu<sup>t</sup> homeostasis leading to a disruption in the balance of the gu<sup>t</sup> microbiome called dysbiosis. This has previously been reported in mice studies. The introduction of prednisolone and tacrolimus to mice resulted in dysbiosis, an overgrowth of *Escherichia coli*, and an increased colonization with opportunistic pathogens [6]. However, the gu<sup>t</sup> microbiome of RTRs has not been studied extensively.

In previous studies among allogenic stem cell transplant recipients and RTRs, a lower diversity of the gu<sup>t</sup> microbiome was observed [7,8]. Furthermore, this lower diversity of the gu<sup>t</sup> microbiome in allogenic stem cell recipients was associated with a higher risk of mortality [9]. In addition, Annavajhala et al. demonstrated that liver transplant recipients with a lower gu<sup>t</sup> microbiome diversity have a higher risk of colonization by multidrug-resistant bacteria [10]. These studies show that the gu<sup>t</sup> microbiome is clinically relevant in the field of transplantation. However, the role of the gu<sup>t</sup> microbiome in renal transplantation has not been adequately studied. Characterization of the gu<sup>t</sup> microbiome in the first three months after renal transplantation showed significant changes in the composition of the gu<sup>t</sup> microbiome and showed that diarrhea was associated with dysbiosis and a loss of diversity [8]. It is currently unknown whether dysbiosis of the gu<sup>t</sup> microbiome remains prevalent more than one year after transplantation and which factors are determinants of the gu<sup>t</sup> microbiota composition in RTRs. The aim of this study was to characterize the gu<sup>t</sup> microbiome of RTRs for at least more than one year post-transplantation. We compared the composition of the gu<sup>t</sup> microbiome between RTRs and healthy controls and identified determinants of the gu<sup>t</sup> microbiome of RTRs.

#### **2. Experimental Section**

#### *2.1. Study Population*

We included 139 RTRs who were at least one year post-transplantation and 105 healthy donors from the TransplantLines Biobank and Cohort Study (ClinicalTrials.gov Identifier NCT03272841). TransplantLines is a prospective observational cohort study in solid transplant recipients [11]. Donors underwent medical screening in the University Medical Center Groningen (UMCG) and can be considered healthy controls. All participants were included during a study visit at the outpatient clinic of the UMCG between September 2015 and April 2018. RTRs were treated with standard antihypertensive and immunosuppressive therapy. The research protocol of the TransplantLines study was approved by the independent medical ethics committee of UMCG (METC 2014/077) and was performed in adherence to the Declaration of Helsinki and the Declaration of Istanbul. All subjects provided a written informed consent.

#### *2.2. Patient Characteristics*

All measurements were performed during a study visit at the outpatient clinic. Weight, length, and waist and hip circumference were measured in duplicate. Body fat percentage was measured using the multifrequency bioelectrical impedance device (BIA, Quadscan 4000, Bodystat, Douglas, British Isles). Blood pressure was measured by qualified nurses according to a standard clinical protocol as described previously [11]. Hypertension was classified as a mean systolic pressure >140 mm Hg, and/or a mean diastolic pressure >90 mm Hg and/or use of antihypertensive medication. Diabetes mellitus was defined according to the guidelines of the American Diabetes Association [12]. Estimated glomerular filtration rate (eGFR) was calculated using the serum creatinine-based chronic kidney disease epidemiology collaboration (CKD-EPI) formula. Proteinuria was defined as urinary protein excretion >0.5 g per 24 h. Glucose and hemoglobin A1c (HbA1c) were determined using standard laboratory methods. Smoking status was recorded using a questionnaire. Medication use was retrieved from medical records and verified with patients during study visits. The study design is described in detail in the TransplantLines design paper [11].

## *2.3. Sample Collection*

Blood samples were collected after an overnight fasting period of 8–12 h and stored at −80 ◦C. Participants were instructed to collect a fecal sample the day prior to the study visit at home and store the sample on ice. Upon arrival at the UMCG the fecal samples were immediately stored at −80 ◦C. Participants also collected 24-hour urine samples the day prior to the study visit.

#### *2.4. DNA Extraction and 16S rRNA Sequencing*

Deoxyribonucleic acid (DNA) was extracted from 0.25 g feces [13]. The genes for the 16S rRNA V4 and V5 region were amplified by polymerase chain reaction (PCR) using the TaKaRa Taq Hot start version kit (TaKaRa Bio Inc., Kusatsu, Japan). We used the 341F and 806R primers containing a 6-nucleotide Illumina-MiSeq adapter sequence. The PCR product was purified with AMPure XP beads (Beckman Coulter, USA). DNA concentrations were measured with Qubit 2.0 Fluorometer to ensure equal library presentation for each sample, dilutions were made accordingly [14]. The normalized DNA library was sequenced using the MiSeq Benchtop Sequencer.

#### *2.5. Microbiome Profiling*

Bacterial taxonomy was assigned using PAired-eND Assembler for DNA sequences (PANDAseq), Quantitative Insights Into Microbial Ecology (QIIME), and ARB [15–17]. QIIME was used to assign taxonomy to the phylum, class, order, family, and genus level. ARB was used to assign taxonomy to the species level. As previously described, PANDAseq was used to increase the quality of sequence reads. Readouts with at quality score lower than 0.9 were discarded according to the protocol followed by Heida et al. [14].

#### *2.6. Statistical Analyses*

Data are presented as mean ± standard deviation (SD) for normally distributed data and median with interquartile range (IQR) for non-normally distributed data. Differences between baseline characteristics of RTRs and healthy controls were tested using a t-test or a Mann–Whitney u-test.

Sample richness/evenness was estimated using the Shannon index using QIIME. The microbial dissimilarities matrix (Bray–Curtis) was obtained using *vegdist* from the *vegan* R-package [18]. Principal coordinates were constructed and plotted with the *cmdscale* function. We used permutational multivariate analysis of variance using distance matrices (ADONIS) to analyze the variance in the Bray–Curtis matrix that could be explained by metadata such as age, sex, body mass index (BMI), fat percentage, smoking, eGFR, and medication. Pearson correlation was used to correlate metadata to the Shannon diversity index. *p*-values <0.01 were considered statistically significant.

Multivariate analysis by linear models (MaAsLin) is a tool to find associations between clinical metadata and bacterial abundance. We used MaAsLin to find associations between microbiome data and clinical phenotype. MaAsLin performs a boosted, additive general linear model between metadata and microbial abundance [19]. Covariates including sex, body mass index (BMI), smoking, use of antihypertensive medication, use of antibiotics, use of statins, use of proton-pump inhibitor (PPI), and read depth were forced into the model. These covariates are known to influence the gu<sup>t</sup> microbiome [20]. All *p*-values were corrected for multiple testing using false discovery rate (FDR). *p*FDR < 0.10 was considered statistically significant for taxonomic analysis.
