Gut microbiota in a Saudi population with chronic kidney disease

Abstract

BACKGROUND

The gut microbiota (GM) plays an important role in chronic kidney disease (CKD) progression, and dialysis modalities can differentially impact the GM composition and function. There is also limited information on the GM in Arab populations.

AIM

To investigate the distinct microbial profiles and functional alterations associated with hemodialysis (HD) and peritoneal dialysis (PD) in a Saudi Arabian cohort.

METHODS

We performed whole-genome metagenomic sequencing on fecal samples from 189 participants (controls and CKD, HD, and PD patients).

RESULTS

We detected distinct microbial profiles across all patient groups compared with that of the controls. Microbial risk scores derived from differentially abundant taxa accurately distinguished CKD, PD, and HD patients from controls, with area under the curves exceeding 0.9. Compared with HD patients, PD patients exhibited reduced species richness, an increased abundance of opportunistic pathogens (particularly Proteobacteria), and increased virulence. Functional analysis revealed suppressed energy metabolism and activated proinflammatory pathways in PD patients. Cooccurrence network analysis demonstrated decreased microbial community resilience in PD patients, with increased Proteobacteria interactions. Conversely, the HD group showed partial recovery of microbial balance and beneficial metabolic functions, including increased short-chain fatty acid metabolism and reduced lipopolysaccharide biosynthesis.

CONCLUSION

The findings of this study highlight the potential of the microbial profile as a robust biomarker for CKD classification and underscore the differential impacts of different dialysis modalities.

Key Words: Chronic kidney disease; Hemodialysis; Peritoneal dialysis; Microbiota; Saudi; Microbiota

Core Tip: In a Saudi cohort, the gut microbiota clearly differed between controls, chronic kidney disease patients, and patients on hemodialysis (HD) or peritoneal dialysis (PD). Simple microbial risk scores (area under the curve > 0.9) accurately distinguished chronic kidney disease patients, HD patients, and PD patients from controls, suggesting its potential as a noninvasive biomarker. The PD microbiota was characterized by reduced diversity, increased abundance of Proteobacteria and virulence factors, weaker microbial networks, and a proinflammatory profile. In contrast, partial restoration of a healthier microbiome was observed in HD patients, with increased short-chain fatty acid production and reduced lipopolysaccharide biosynthesis, indicating dialysis modality-specific effects on the gut microbiome and potential targets for microbiome-based interventions.

INTRODUCTION

Chronic kidney disease (CKD) presents a significant global health challenge, affecting approximately 10% of the world’s population and ranking as a leading cause of mortality worldwide[1]. Characterized by the progressive and irreversible loss of nephrons, CKD leads to a decline in renal function and a multitude of complications, including cardiovascular disease and mineral bone disorder (MBD)[2-4]. While traditional research on CKD has focused predominantly on renal and cardiovascular pathophysiology, recent investigations have highlighted the crucial role of the gastrointestinal tract and its resident microbial community, the gut microbiota (GM), in CKD progression[5-8]. Moreover, accumulating evidence underscores the intricate interplay between phosphate, a key uremic toxin, and the GM in the context of CKD and CKD-MBD[9,10].

A growing body of research supports the existence of a bidirectional relationship between CKD and the GM, often referred to as the gut-kidney axis. Uremia, a hallmark of CKD, disrupts the delicate equilibrium of the intestinal microbial ecosystem, leading to a state of dysbiosis characterized by alterations in microbial composition and function. This dysbiosis often involves an increase in proteolytic bacterial abundance and a decrease in beneficial commensal species[6,7], contributing to the accumulation of gut-derived uremic toxins such as indoxyl sulfate and p-cresyl sulfate[8,11,12]. Importantly, the specific modality of renal replacement therapy employed, namely, hemodialysis (HD) or peritoneal dialysis (PD), appears to differentially influence the gut microbiome, with potential ramifications for CKD progression and further CKD-MBD.

Studies have revealed distinct microbial profiles in HD and PD patients. HD patients frequently exhibit more pronounced dysbiosis than their PD counterparts do, characterized by reduced microbial diversity and enrichment of proinflammatory taxa such as Enterobacteriaceae and Ruminococcaceae[13,14]. Conversely, PD patients may demonstrate a relative abundance of beneficial taxa such as Bifidobacterium and Lactobacillus[14-16]. These disparities in microbial communities likely stem from the inherent differences between the two dialysis methods. HD, with its intermittent application and owing to the direct contact between blood and artificial membranes, can induce hemodynamic instability, expose patients to dialyzer materials, and necessitate the use of medications such as phosphate binders, all of which can alter the gut microenvironment and promote inflammation and microbial dysbiosis[17-19]. In contrast, the continuous nature of PD and the lack of direct blood contact with dialyzer membranes may contribute to less severe disruption of the GM, potentially promoting the growth of beneficial bacteria[14-16]. However, PD patients are not immune to alterations in the GM, as factors such as the use of glucose-based dialysates, which can provide a substrate for bacterial fermentation, and the risk of peritonitis, which can introduce pathogenic bacteria into the peritoneal cavity, can also influence microbial composition and function[20,21].

A deeper understanding of the specific differences in microbial communities between HD patients and PD patients is crucial for developing targeted interventions aimed at improving gut health and potentially mitigating CKD progression and CKD-MBD. This study builds upon existing knowledge by conducting a comprehensive comparative analysis of the gut microbiome composition and metabolic function in HD and PD patients using microbiota approaches and investigating the influence of different dialysis modalities on specific microbial pathways that selectively modulate the GM to improve clinical outcomes in HD and PD patients. By employing whole-genome metagenomic sequencing, we aim to provide a more nuanced understanding of the gut-kidney axis in CKD patients, with a particular emphasis on elucidating the distinct roles of the GM in HD and PD patients in Saudi populations.

MATERIALS AND METHODS

Setting

This multicenter case-control study was conducted in the Departments of Internal Medicine and Clinical Biochemistry, King Fahd Hospital of the University, Al-Khobar; King Fahd Hospital, Al Ahssa; Qatif Central Hospital, Qatif; and Jubail Central Hospital, Jubail, all of which are in the Eastern Province of Saudi Arabia.

Patient and eligibility criteria

The demographic and clinical data for the patients and controls are shown in Table 1. Consecutive patients were enrolled and evaluated for features of CKD, family history of CKD, and comorbidities.

Table 1 Demographic characteristics of study participants across four groups, chronic kidney disease, hemodialysis, peritoneal dialysis, and healthy controls, n (%).

Characteristics	CKD (n = 48)	HD (n = 49)	PD (n = 43)	Control (n = 49)
Gender
Male	20 (41.7)	21 (42.9)	23 (53.5)	31 (63.3)
Female	28 (58.3)	28 (57.1)	20 (46.5)	18 (36.7)
Age, years
mean (SD)	53.69 (18.98)	58.55 (16.31)	44.86 (19.81)	45.73 (16.74)
Median	54.50	59	39	46
Family history of CKD
Yes	12 (25)	9 (18.4)	2 (4.7)	0 (0)
No	36 (75)	40 (81.6)	41 (95.3)	49 (100)
Comorbidities
CVD	9 (18.8)	15 (30.6)	10 (23.3)	NA
DM	30 (62.5)	30 (61.2)	21 (48.8)	NA
HTN	39 (81.3)	45 (91.8)	36 (83.7)	NA
SCD	1 (2)	0 (0)	0 (0)	NA

CKD: Chronic kidney disease; HD: Hemodialysis; PD: Peritoneal dialysis; CVD: Cardiovascular disease; DM: Diabetes mellites; HTN: Hypertension; SCD: Sickle cell disease; NA: Not available.

Participants were recruited across multiple centers and assigned to CKD, HD, PD, or healthy control cohorts according to clinical status at enrollment. Given the high prevalence of diabetes mellitus in our population and the tendency for CKD to be diagnosed at later stages, we adopted an enrollment strategy that resulted in relatively young patient cohorts overall; the demographic characteristics of each cohort are summarized in Table 1. To minimize acute inflammatory and medication-related effects on the gut microbiome, we included only participants without evidence of intercurrent infection in the preceding four months and without antibiotic exposure during the same period. Dialysis vintage at the time of stool collection was captured and is reported as the mean (SD): PD duration 58 ± 21 months and HD duration 61 ± 21 months across participating sites. In the HD cohort, vascular access was predominantly catheter-based, with non-catheter access more common among participants with diabetes; this distribution is reported in the clinical characteristics of the cohorts. All HD participants were treated with high-flux dialyzers; thus, the dialyzer flux did not vary within the HD group.

The estimated glomerular filtration rate was calculated from the serum creatinine level using the CKD-epidemiology collaboration equation. Total kidney volume was calculated using the ellipsoid formula, and the Mayo Clinic Imaging Classification was obtained using an online tool (Figure 1). Patients with syndromic causes of multiple renal cysts were excluded. Participants were stratified into four groups: CKD (pre-dialysis), HD, PD, and healthy controls.

Figure 1 Ultrasound images of enlarged kidneys. Ultrasound images of the kidneys of one patient on hemodialysis. Both kidneys are small with mildly to moderately increased parenchymal echogenicity and decreased parenchymal sinus differentiation. The right kidney measured 6.9 cm in length and 0.7 cm in parenchymal thickness, showing focal upper calyceal dilatation and irregularity.

Study protocol

A total of 189 stool samples (all those eligible according to our criteria were included) were collected in sterile containers and frozen at -20 °C until DNA isolation. CKD patients were grouped according to the following criteria: (1) Those undergoing HD (n = 49); (2) Those undergoing PD (n = 43); (3) CKD patients who were not receiving HD; and (4) Healthy control individuals (n = 49). DNA was isolated from each stool sample using a QIAamp Fast Stool Mini Kit (Qiagen, Germany) according to the manufacturer’s instructions. DNA was eluted in 150 μL of elution buffer provided with the kit. DNA was quantified using a Qubit^™ dsDNA Quantification broad range assay kit (Invitrogen, United States) on a Qubit 4 fluorometer (Invitrogen, United States). The purity of the DNA was estimated using a Nano Drop 8000 spectrophotometer (Thermo Fisher Scientific, United States). Library preparation was carried out using an Illumina DNA Prep Kit (Illumina, United States) according to Illumina’s protocol. Briefly, the DNA was fragmented with bead-linked transposomes, followed by cleanup with proprietary beads from Illumina. DNA fragments were amplified with different combinations of appropriate index adaptors for each sample, and the polymerase chain reaction products were purified with beads, after which the libraries were quantified using a Qubit^™ dsDNA Quantification High Sensitivity Assay Kit (Invitrogen, United States). The length and quality of the individual libraries were examined with high-sensitivity D1000 screentape (Agilent, United States) on a tapestation 4150 (Agilent, United States). The library concentration was adjusted to 2 nM with resuspension buffer (Illumina, United States). Ten microliters of each diluted library was pooled to create a multiplexed library pool. The final loading concentration (750 pM) was prepared using resuspension buffer with Tween 20 (Illumina, United States). Twenty microliters of the final pooled sample was added to a sample well of the NextSeq P1 300 cycle cartridge (Illumina, United States) of the corresponding flow cell. Sequencing was carried out using a NextSeq 2000 instrument.

Diversity, differential abundance, and microbiome risk score analysis

Microbial alpha diversity was quantified using the Chao1 richness index, which was calculated from species-level taxonomic profiles for each sample. Group differences in the Chao1 index were evaluated using nonparametric Wilcoxon rank-sum tests or Kruskal-Wallis tests, as appropriate, with post hoc pairwise comparisons between CKD, HD, PD, and control groups. Beta diversity was assessed using Bray-Curtis dissimilarity, and principal coordinate analysis was used to visualize differences in overall community structure among groups.

The differential abundance of bacterial and viral taxa between groups was assessed using the Analysis of Compositions of Microbiomes with Bias Correction 2 (ANCOM-BC2) framework. ANCOM-BC2 models log-transformed relative abundances while accounting for compositionality, library size differences, and sampling variability and provides bias-corrected estimates of log-fold changes between groups. Pairwise comparisons were performed among the CKD, HD, PD, and control groups at multiple taxonomic levels. The resulting p values were adjusted for multiple comparisons using the Benjamini-Hochberg false discovery rate (FDR) procedure, and taxa were considered differentially abundant at an FDR q < 0.05.

To summarize group-level microbiome differences in terms of a single metric, we derived microbial risk scores (MRSs) for discriminating patient groups. For each comparison, we first selected taxa that were significantly associated with the outcome using ANCOM-BC2 and resampling-based feature selection (bootstrapping). We then computed the MRS for each individual as a weighted sum of the relative abundances of the selected taxa, where weights corresponded to effect sizes (log-fold changes) from the differential abundance models. The discriminative performance of each MRS was evaluated using receiver operating characteristic curves and the area under the curve (AUC) with 95% confidence intervals. Functional and virulence factor (VF) profiles were analyzed analogously. Differentially abundant VF classes and differentially enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways across groups were identified using the same ANCOM-BC2-based framework with FDR correction. To investigate ecological structure, we constructed co-occurrence networks of bacteria and phages within each group on the basis of pairwise association measures between taxa. Edges were retained for statistically significant positive or negative associations after multiple-testing correction, and network topology was compared qualitatively across CKD, HD, PD, and control groups to highlight shifts in microbial interactions.

Statistical analysis

Statistical analyses were performed to compare the microbial composition, functional capacity, and virulence profiles across the CKD, HD, PD, and healthy control groups.

Ethics approval

This study was approved by the Ethical Committee of Imam Abdulrahman Bin Faisal University, approval No. IRB-2020-01-277 and was conducted in accordance with the 1964 Helsinki Declaration. Informed written consent in English with a verified translation in Arabic was obtained from all participants.

RESULTS

In an attempt to understand the microbial landscape and its impact on CKD disease status and in patients undergoing different types of dialysis, we enrolled 189 participants in four sample groups (controls n = 49; CKD, n = 48; HD, n = 49; and PD, n = 43; Figure 2A).

Figure 2 Study design and taxonomic architectures. A: Sample group setting of the study; B: Species richness (Chao 1) across sample groups. The peritoneal dialysis group had the lowest species richness among all the groups (Wilcoxon test, ^aP < 0.05, ^bP < 0.01). Lower species richness is often associated with disease and is generally thought to lead to decreased resilience; C: Principal coordinate analysis of the species abundance matrix showing deviations in the microbial compositions among the four sample groups; D: Microbial species compositions of the microbial community across groups (top 50 species in each sample group); E: Bubble plot showing differentially abundant key bacterial genera across control, chronic kidney disease, hemodialysis, and peritoneal dialysis groups. Bubble size represents relative abundance (%), and bubble color represents the log10 (false discovery rate-adjusted P value) for group differences. The accompanying comparison panel summarizes pairwise between-group differences, with color indicating the direction of change between the indicated groups. CKD: Chronic kidney disease; PD: Peritoneal dialysis; HD: Hemodialysis.

PD patients exhibit more severe taxonomic dysbiosis than HD patients do

We analyzed the gut microbiome, including both bacteria and viruses, at multiple taxonomic levels (kingdom, phylum, genus and species) for each sample. Species richness, characterized by a decreased Chao1 index (Figure 2B), and a greater deviation in the microbial composition from those of the control group (Figure 2C) were observed in the PD group compared with the other groups. As lower species richness and deviation from healthy GM are often associated with disease and are generally thought to cause decreased resilience[22], PD is more detrimental to patients.

Differentially abundant taxa across patient groups

Overall, 2833 bacterial species from 855 genera across 34 phyla were identified in our gut microbiome dataset (Figure 2D and E, Supplementary Figure 1). Differentially abundant taxa between patient groups were identified using ANCOM-BC2[23,24]. Comprehensive profiles of the differentially abundant taxa and their enrichment patterns across all pairwise group comparisons are provided in Supplementary Figure 2. Compared with the healthy controls, all the patient groups (CKD, HD, and PD) had increased abundances of opportunistic pathogens such as Pseudomonas, Klebsiella, Streptococcus, and Brevundimonas (Figure 3A-C and Supplementary Table 1)[25,26].

Figure 3 Differential viral taxa abundance across chronic kidney disease groups. Volcano plots show differential abundance of viral taxa (n = 3193) for six pairwise comparisons; each point is a viral taxon, the X-axis is log2 fold change and the Y-axis is -log10 adjusted P value. Horizontal and vertical dashed lines indicate the thresholds used for significance (adjusted P value and minimum log2 fold change). Color indicates significance: Blue: Non-significant, purple: Significant by fold change only. A: Chronic kidney disease (CKD) vs Control, viral taxa enriched or depleted in conservatively managed CKD patients compared with healthy controls; selected highly significant taxa are labelled on the plot; B: Hemodialysis (HD) vs control, viral taxa enriched or depleted in hemodialysis patients compared with healthy controls; C: Peritoneal dialysis (PD) vs control, viral taxa enriched or depleted in peritoneal dialysis patients compared with healthy controls; D: HD vs CKD, viral taxa differing between hemodialysis and conservatively managed CKD patients (highlights dialysis-associated shifts); E: PD vs CKD, viral taxa differing between peritoneal dialysis and conservatively managed CKD patients; F: PD vs HD, viral taxa differing between peritoneal dialysis and hemodialysis patients (directionality shown by sign of log2 fold change). CKD: Chronic kidney disease; HD: Hemodialysis; PD: Peritoneal dialysis.

Similarly, the abundances of viruses targeting gut bacteria such as Lactobacillus, Streptococcus, Klebsiella, and Escherichia were elevated in all patient groups (Figure 3A-C, Supplementary Table 1), indicating that microbiome stress, likely driven by bacterial overgrowth, infection dynamics, or gut dysfunction, is prevalent in CKD patients. Conversely, the abundances of commensal taxa such as Muribaculum, Riemerella, Prevotella, and Paraprevotella were consistently depleted in CKD, HD, and PD patients compared with healthy individuals (Figure 3A-C, Supplementary Table 1), highlighting persistent gut dysbiosis regardless of treatment modality. The abundances of protective taxa such as Blautia and Ruminococcus increased across all patient groups (Figure 3A-C, Supplementary Table 1), potentially serving as a compensatory mechanism to mitigate dysbiosis-related impacts.

Compared with CKD patients, both HD and PD patients exhibited increased abundances of beneficial bacteria such as Ruminococcus, Streptomyces, and Pseudomonas (Figure 3D and E, Supplementary Table 1). The abundances of pathogenic microbes such as Shigella and Salmonella, along with their associated phages (Figure 3D and E, Supplementary Table 1), were reduced in the dialysis groups, likely because of improved toxin clearance and a gut environment characterized by reduced dysbiosis. Dialysis patients also show an altered virome with increased diversity of phages and viruses, potentially enhancing microbial dynamics and reducing pathogen loads. Additionally, the elevated abundances of archaea in both HD and PD patients suggest shifts in gut metabolic activity, such as that associated with hydrogen and methane production, reflecting metabolic adaptations to dialysis.

Compared with HD, PD favors pathogenic taxa such as Shigella, Escherichia, and associated phages, along with reductions in short-chain fatty acids (SCFA)-producing bacteria (Ruminococcus and Eubacterium) (Figure 3F, Supplementary Table 1), indicating increased gut dysbiosis due to systemic stress and intermittent fluid shifts. In contrast, HD promotes increased microbial diversity, with increases in beneficial taxa such as Pseudomonas, Streptomyces, and archaea (Halorubrum and Methanoregula) (Figure 3F, Supplementary Table 1), reflecting greater metabolic stability due to continuous fluid exchange.

The microbiota is a complex community, and analyzing its profile as a whole provides more comprehensive insights than examining individual microbes does. We determined summary MRSs consisting of associated taxa to classify patient status[25,26]. These MRSs effectively distinguished CKD, PD, and HD patients from healthy controls, with AUCs greater than 0.9 (Figure 4). Additionally, MRSs could differentiate HD patients from CKD and PD patients, with AUCs of approximately 0.8 (Figure 4). These results underscore the potential of microbial profiles as robust biomarkers for identifying and differentiating CKD patients.

Figure 4 Receiver operating characteristic analysis of the metabolite risk score for patients with chronic kidney disease. Receiver operating characteristic curves showing the diagnostic performance of the metabolite risk score for discriminating chronic kidney disease patients from healthy controls, with corresponding area under the curve values and 95% confidence intervals. The optimal cutoff points, sensitivity, and specificity are indicated as described in the text. CKD: Chronic kidney disease; HD: Hemodialysis; PD: Peritoneal dialysis; AUC: Area under the curve.

Gut microbial functional profiles reflect increased virulence and suppressed energy metabolism in the PD group

To further characterize the gut microbiome virulence in each sample, we aligned the genes in our microbiome dataset against the VF database. We identified a total of 1285 VFs across 12 classes, involving primarily immune modulation, adherence, the effector delivery system and motility (Figure 5A). Compared with the control group, the CKD group exhibited a notable increase in the levels of VFs associated with immune modulation, exoenzymes, and adherence. Compared with the PD group, the HD group presented a significant reduction in VFs (Figure 5B), particularly those related to motility, nutritional/metabolic factors, and adherence (Figure 2B), potentially suggesting increased virulence in the GM following PD.

Figure 5 Cross-group virulence and metabolic architectures in chronic kidney disease. A: Thirty most prevalent virulence factors (VFs) across sample groups. Each row represents a VF (gene or gene cluster associated with pathogenicity). The first four columns show prevalence within each group (control, chronic kidney disease, hemodialysis, peritoneal dialysis); bubble size and color indicate prevalence/abundance. The six columns on the right display pairwise group comparisons (Fisher’s exact test); red and blue bubbles denote VFs significantly increased or decreased, respectively, in the row group relative to the column group (adjusted P value); B: VF categories across groups. Bubble plot summarizing major VF functional classes (for example, immune modulation, adherence, motility, invasion, biofilm, stress survival). Left columns show category prevalence in each group (bubble size proportional to prevalence; color indicating magnitude); right columns show pairwise comparisons with bubble color indicating direction (purple = higher, blue = lower) and bubble size reflecting -log10 (P value) (adjusted for multiple testing); C: Metabolic pathway architectures across groups. Bubble plot of metagenomically predicted metabolic pathways (for example, glycolysis/gluconeogenesis, carbon fixation, peptidoglycan biosynthesis, citrate cycle, nitrogen metabolism, ubiquinone and other terpenoid-quinone biosynthesis). Left columns show pathway prevalence/abundance per group; right columns show pairwise comparisons with bubble size indicating significance [-log10 (P value)] and color indicating direction and magnitude of log10 (fold change), with purple and blue indicating enriched or depleted pathways after multiple-testing correction.

When the functional genes in the gut microbiome were analyzed using the KEGG database, we identified 213184 KEGG orthologs across all sample groups. Notably, the microbial metabolic pathways, including pantothenate and CoA biosynthesis, glycolysis and gluconeogenesis, metabolism of various amino acids, and the one-carbon pool by folate, were suppressed in the CKD group (Figure 5C). As these pathways are related to host energy production, glucose homeostasis, and protein and amino acid metabolism, their suppression suggests overall metabolic disorder in CKD patients, with impaired nutrient absorption and energy and nutritional deficits.

Compared with the control group, the PD group also exhibited a significant decrease in energy production by the GM, characterized by the suppression of metabolic pathways, including glycolysis, the one-carbon pool by folate, pantothenate and CoA biosynthesis, carbon fixation, butanoate metabolism, and the citrate cycle (Figure 5C). Dialysis led to further decreases in butanoate and pyruvate metabolism compared with those in the CKD group, indicating that worsening energy depletion post dialysis may exacerbate kidney dysfunction. Conversely, the HD group exhibited increased microbial glucose metabolism, with significant increases in carbon fixation, pyruvate metabolism, and starch and sucrose metabolism compared with the Control group, and in peptidoglycan biosynthesis compared with the CKD group (Figure 5C). Compared with CKD patients, HD patients also exhibited reduced expression of genes involved in nitrogen metabolism, a metabolic waste product.

When we compared the metabolic functions of the GM between the two dialysis groups, we observed that PD increased the microbial biosynthesis of ubiquinone and terpenoid-quinone, potentially reducing oxidative stress and improving cardiovascular health (Figure 5C). However, more beneficial microbial components were observed in the HD group. After HD, the expression of functional genes related to several energy production pathways increased in the gut microbiome. Additionally, dialysis promoted the metabolism of SCFAs, such as butanoate and propanoate, which are known for their anti-inflammatory effects. HD also inhibited the biosynthesis of proinflammatory and toxic metabolites. Specifically, HD decreased the expression of genes involved in the microbial biosynthesis of lipopolysaccharide (LPS), a potential trigger of inflammation, and increased the microbial degradation of the toxic metabolites nitrotoluene and phenylalanine[27] compared with both the CKD and PD groups (Figure 5C). Overall, the increased LPS biosynthesis, decreased SCFA metabolism, and increases in bacterial pathogen abundance and virulence suggest a greater risk of infection and inflammation after PD than after HD.

Bacteria-phage-function networks across kidney disease states

To characterize the microbial co-occurrence and cooperation landscapes across different sample groups, we analyzed the associations between bacteria, phages, VFs, and KEGG pathways. Our examination of co-occurrence networks among taxa across the bacterial and viral kingdoms revealed varying complexities across the four sample groups (Figure 6, Supplementary Figure 3). The control group exhibited a relatively uniform network with the most microbial co-occurrence links, characterized by the most bacteria-bacteria associations but fewer phage-phage associations (Supplementary Table 2). More robust bacterial cooccurrences were consistently detected in healthy individuals than in patients with kidney disease[28].

Figure 6 Phylum-level co-occurrence networks of the gut microbiome across chronic kidney diseaseand dialysis groups. Circular chord diagrams summarize pairwise co-occurrence relationships among gut microbial phyla in healthy controls (top left), nondialysis chronic kidney diseaseand (top right), hemodialysis (bottom left), and peritoneal dialysis (bottom right). Each colored outer segment represents a microbial phylum (color key at right); segment length reflects the relative abundance of that phylum within the group. Inner ribbons connect phyla with significant co-occurrence, with ribbon width proportional to association strength, enabling comparison of changes in community structure and cross-phyla connectivity across chronic kidney diseaseand progression and dialysis modalities. CKD: Chronic kidney disease; HD: Hemodialysis; PD: Peritoneal dialysis.

In terms of bacteria-bacteria associations, the control group showed the greatest number of significant interactions within or among the most abundant bacterial phyla, Actinobacteria, Bacteroides, Firmicutes, and Proteobacteria (Figure 7, Supplementary Table 3).

Figure 7 Bacteria-phage-function co-occurrence networks across chronic kidney disease-related groups. A: Control: Network of significant co-occurrence associations among bacterial taxa (nodes colored by phylum), a viral/phage taxon (e.g., Moineauvirus), virulence factor (VF) classes, and Kyoto Encyclopedia of Genes and Genomes (KEGG pathways) in healthy controls. Node size reflects relative abundance, and edges indicate significant associations; B: Chronic kidney disease (CKD): Corresponding integrated network for predialysis CKD samples, illustrating group-specific shifts in connectivity between bacterial phyla and linked functional features (VFs and KEGG pathways) and viral/phage signals; C: Hemodialysis (HD): Integrated network for HD patients, showing the pattern of significant associations among bacterial taxa, the viral/phage component, and functional features after initiation of HD; D: Peritoneal dialysis (PD): Integrated network for PD patients, depicting associations among bacterial taxa, viral/phage features, VFs, and KEGG pathways under PD across panels, node shape distinguishes feature type (taxon vs VF vs KEGG pathway vs virus/phage, as indicated in the figure key), node color denotes bacterial phylum for bacterial taxa, and edge thickness reflects association strength. CKD: Chronic kidney disease; HD: Hemodialysis; PD: Peritoneal dialysis.

As CKD progressed, most of these cross-phylum bacteria-bacteria associations weakened, while positive links between Proteobacteria (harboring many pathogenic genera) and Bacteroidetes strengthened (Figure 7, Supplementary Table 3). Within-phylum co-occurrence associations slightly decreased in Firmicutes but increased in other phyla. Interestingly, after HD, the associations between Proteobacteria and other bacteria weakened, but links among the other major phyla exhibited partial recovery (Figure 7). This transition corresponded with the recovery of species richness in the HD group (Figure 2B), suggesting an improvement in the microbial community balance following dialysis. In contrast, the PD group exhibited a further reduction in cross-phylum associations but an increase in bacterial co-occurrence in the phylum Proteobacteria (Figure 7D), along with a significant reduction in species richness (Figure 2B), indicating decreased resilience in the GM of PD patients.

When taxonomic associations involving phages (either bacteria-phage or phage-phage interactions) were analyzed, we observed increased associations in the patient groups. Compared with the control group, the CKD group had more phage-related associations, which were further enhanced after PD but reversed after HD (Figure 7). Notably, Proteobacteria, which includes a wide variety of pathogens, was consistently the primary phylum linked to phages across all sample groups, with the greatest number of phage associations observed in the PD group (Supplementary Table 4). Further examination of the associations between bacterial pathogens and phages revealed an increase in associations from the control group to the CKD group, which further strengthened in the PD group but reversed in the HD group (Supplementary Table 5).

We then compared the taxon-function associations across sample groups. In general, we observed more microbe-VF links in the CKD and PD groups. The differentially abundant VF classes, especially nutritional/metabolic factors, adherence, and motility, were positively and negatively linked to the abundance of Proteobacteria and Firmicutes, respectively, in these two groups. Moreover, the increased abundance of Proteobacteria and Bacteroidetes and the reduced abundance of Firmicutes likely explain the increased production of proinflammatory LPS and the toxic metabolites nitrotoluene and phenylalanine in the CKD and PD groups (Figure 7). Firmicutes was also the major bacterial phylum and was positively associated with the metabolism of the anti-inflammatory compound propanoate. We inferred that the disrupted abundance of Bacteroidetes (increased), Proteobacteria (increased), and Firmicutes (decreased) in the PD groups compared with the HD group was a potential reason for the increased virulence in the GM of PD patients. These bacteria are therefore potential microbial modulators and treatment targets for improving patient outcomes following PD.

DISCUSSION

In this study, we characterized the GM and associated functional profiles across the continuum of CKD, including patients on conservative (pre-dialysis) management, those receiving HD, those receiving PD, and healthy controls. By integrating taxonomic, functional, and metabolite level information, we provide a comprehensive overview of the GM-kidney axis that is highly relevant to the scope of studies on the kidney and dialysis. Our results confirm that CKD is accompanied by substantial alterations in the gut microbial composition and that dialysis modality is associated with distinct and modality specific microbial and functional signatures[29,30]. These findings extend previous observations of CKD related dysbiosis and highlight the importance of considering treatment context when microbiome changes are interpreted in patients with advanced kidney disease[14].

Consistent with prior reports, we observed a shift in the relative abundance of key bacterial phyla and genera in CKD patients compared with healthy controls, including enrichment of taxa with putative proteolytic and urease activities and depletion of butyrate producing commensals[19]. Such changes are biologically plausible in the context of uremia, constipation, altered gut transit, and dietary modifications that are common among CKD patients. Importantly, by examining both patients managed conservatively and those on HD or PD, we were able to differentiate patterns that appeared to be primarily driven by kidney dysfunction itself from those more strongly associated with dialysis modality.

A central novelty of our work lies in the integration of taxonomic data with functional readouts, including predicted metabolic pathways, VFs, and serum metabolites. Rather than focusing solely on descriptive changes in microbial abundance, we demonstrate that CKD and dialysis modalities are associated with coordinated shifts in the functional potential of the microbiome[31,32]. The enrichment of pathways related to xenobiotic metabolism, amino acid fermentation, and the generation of gut derived uremic toxins, together with the suppression of short chain fatty acid-producing pathways, supports the concept of qualitative reprogramming of the GM in CKD[14]. This functional remodeling may help explain the chronic low-grade inflammation, impaired barrier integrity, and accumulation of protein bound uremic solutes that are characteristic of advanced kidney disease and contribute to cardiovascular and overall mortality[33].

Our network analyses further advance the current understanding by mapping significant correlations among differential microbial taxa, viral genera, VF classes, and KEGG metabolic pathways across clinical groups. These integrated microbe-function networks reveal tightly connected modules that are differentially represented in controls and CKD, HD, and PD patients[34,35]. The identification of specific genera and functional nodes that occupy central positions within these networks suggests that these factors may act as key regulators of microbe-host interactions in CKD. From a translational perspective, such hubs represent attractive candidates for biomarker development and microbiota directed interventions, including dietary strategies, prebiotics, probiotics, or approaches designed to reduce the generation or increase the clearance of gut derived toxins[30,36-38].

The present study also contributes novel insights into how dialysis modality may shape the GM-metabolite interface. HD and PD differ markedly in terms of solute and fluid removal, exposure of blood and peritoneal surfaces to dialysis membranes and solutions, and nutritional and pharmacologic profiles. Our observation that each modality is associated with a partially distinct microbial and functional configuration suggests that therapy specific perturbations of the gut environment may influence the production and systemic accumulation of microbial metabolites. This might, in part, explain the differences in inflammatory burden, nutritional status, and cardiovascular risk reported between HD and PD patients, although causality cannot be established from our data[33,37]. Future work integrating longitudinal microbiome profiling with detailed metabolomics in incident dialysis patients could clarify how the initiation and modification of dialysis therapy reshape the gut ecosystem and its metabolic output over time.

Our findings should be interpreted in terms of several strengths and limitations. Strengths include the inclusion of well characterized CKD cohorts spanning conservative management and two major dialysis modalities, the parallel comparison with healthy controls and the use of high resolution sequencing with curated functional annotation. We also applied stringent statistical thresholds with correction for multiple testing to reduce the risk of spurious associations in the correlation networks. Conversely, the observational and predominantly cross sectional design precludes firm conclusions about causality, and residual confounding by unmeasured dietary, pharmacological, and lifestyle factors is possible. The sample size within specific subgroups may limit the ability to detect more subtle associations, and we did not include direct measures of intestinal permeability, systemic inflammation, or hard clinical outcomes such as hospitalization and mortality, which would strengthen our mechanistic inferences[39].

Despite these limitations, our study has several important implications for clinical nephrology and dialysis practice. First, it reinforces the concept that the GM in CKD patients is not merely an epiphenomenon of advanced disease but also a potentially modifiable contributor to the uremic milieu. Second, by providing an integrated, modality specific map of microbial and functional alterations across the CKD spectrum, our results offer a framework for designing targeted interventions that are tailored to distinct CKD phenotypes and treatment modalities. Third, the identification of reproducible microbial and functional signatures associated with kidney function and dialysis status suggests that microbiome based biomarkers could complement existing tools for risk stratification and therapeutic monitoring in patients with kidney disease and those undergoing dialysis. Future interventional studies, including randomized trials of microbiota modifying strategies, will be needed to determine whether altering these microbial configurations can translate into meaningful improvements in renal, cardiovascular, and patient reported outcomes. In this context, the present work provides a foundation and clear set of hypotheses for studies that aim to move from association to intervention in the GM-kidney axis.

Some clinical factors that can shape the gut microbiome were not captured with the same level of detail across all centers, which limited how fully we could adjust for them. We recorded dialysis vintage for both PD and HD patients but did not consistently collect more granular PD exposure measures (such as cumulative glucose-based dialysate exposure) or dialysis adequacy metrics. In the HD cohort, vascular access type may still contribute to residual confounding, and because all HD participants used high-flux dialyzers, we could not assess the impact of dialyzer flux within that group. These limitations highlight clear priorities for future multicenter, longitudinal studies with standardized clinical data collection and closer matching of key demographic variables to better isolate modality-specific microbiome effects.

CONCLUSION

This study provides a detailed characterization of the gut microbiome in CKD patients undergoing HD and PD in a Saudi population. Our findings highlight the distinct microbial profiles and functional alterations associated with each dialysis modality, emphasizing the importance of considering the gut-kidney axis in CKD management. Further research is needed to translate these findings into targeted interventions that can improve gut health and ultimately benefit CKD patients.

ACKNOWLEDGEMENTS

We are grateful to the nurses, technical staff, and students for their efforts, commitment, and dedication in supporting this work.