Uremic toxins levels are associated with miR-223 in chronic kidney disease-associated anemia

Abstract

Chronic kidney disease (CKD) poses a significant threat, with increased rates of cardiovascular and all-cause mortality. Anemia, common in CKD, is associated with accumulation of uremic toxins in the bloodstream. We previously demonstrated that the uremic toxin indoxyl sulfate (IS) impacts the regulation of erythropoiesis in cellular and preclinical CKD models. Here, the role of non-coding RNAs in this toxic effect was evaluated. The effect of IS on microRNA expression was measured in human erythropoietic cell line UT7/EPO, using nanostring. We found a significant increase of miR-223 in cells treated with IS. This finding was further validated in human primary CD34⁺ cells, a more physiological model for human erythropoiesis. Finally, serum levels of miR-223 correlated with representative uremic toxins, including IS, in patients with various stages of CKD, and also with endothelial dysfunction markers, indicating a link with vascular damage. These correlations varied according to erythropoietin treatment and dialysis. These findings suggest that miR-223 may play a role in the development of anemia in CKD. Further investigation into the involvement of miR-223 in erythropoiesis is needed for a better understanding of the mechanisms underlying anemia in CKD and the potential role of uremic toxins. Ultimately, this may open up new therapeutic possibilities for the management of anemia in CKD.

1. Introduction

Chronic kidney disease (CKD) has a rising incidence and prevalence worldwide [1]. It is a progressive disease characterized by the gradual loss of kidney function, ultimately leading to kidney failure [2]. CKD is caused by a spectrum of pathologies, including glomerular diseases, tubulointerstitial injuries, vascular disorders, and may lead to anemia [1]. Despite extensive research efforts, the underlying molecular mechanisms driving CKD progression remain incompletely understood. CKD is often accompanied by high levels of bound solutes, including uremic toxins (UT). It is however not an automatic progress, as CKD does not always cause a significant rise in UT [3]. That said, in a vast number of CKD patients, a rise in bound solutes, among them UT, is believed to contribute to various complications associated with CKD [4].

Anemia is a significant complication in CKD patients, as it is linked to cardiovascular disease (CVD), increased morbidity and mortality, and a decline in kidney function, and treatment with erythropoietin (EPO) entails a significant cost and may cause side effects [5]. To investigate the impact of Indoxyl Sulfate (IS), one of the most well-known UT, on erythropoiesis in CKD, we recently assessed the effects of IS on erythroid cells. We conducted experiments using cellular and pre-clinical models [6]. In cellular and pre-clinical models, IS was shown to increase apoptosis in UT7/EPO cells and in human primary CD34⁺ cells, and blocked the burst-forming unit-erythroid (BFU-E) stage of erythropoiesis in a 5/6 nephrectomy CKD mice model [6,7]. Uremic toxins are also distributed within erythrocytes by active transport mechanisms involving Band 3 proteins, potentially affecting their concentration in the plasma and consequently their toxic effect [8]. This suggests that IS might interfere with the growth and differentiation of erythroid progenitor cells. Additionally, IS was found to deregulate several genes involved in erythropoiesis [7]. These findings propose that IS has the potential to impair cell viability and disrupt the differentiation of erythroid progenitors, thereby affecting erythropoiesis and contributing to the development of anemia in CKD, leading us to search for molecular cues.

In recent years, microRNAs (miRNAs) have emerged as crucial regulators of gene expression and signaling pathways involved in various biological processes [9]. miRNAs are small, non-coding RNA molecules, typically consisting of approximately 21–25 nucleotides (approximately 7–8 kDa), that play a pivotal role in post-transcriptional gene regulation. They act by binding to the 3′ untranslated region (UTR) of target messenger RNA (mRNA), leading to mRNA degradation or translational repression [10]. By modulating the expression of numerous target genes, miRNAs influence various cellular processes, including differentiation, proliferation, apoptosis, and immune responses. miRNAs and other non-coding RNAs have gained considerable attention in the context of CKD due to their potential involvement in disease progression and pathogenesis [11]. To acquire a better understanding of the mechanisms underlying anemia in CKD and highlight the role of IS in EPO producing cells, we decided to explore the effect of miRNAs in this process in erythroleukemic UT7/EPO cells, using the same IS concentration as in our previous work [7]. Recent research has suggested a potential link between miRNA blood levels and uremic toxin levels, raising questions about the role of miRNAs in kidney health and their potential as biomarkers for CKD progression [[12], [13], [14], [15]]. The rationale behind this research lies in the fact that miRNAs are known to be involved in cellular processes that contribute to kidney dysfunction, anemia and fibrosis [16,17]. The primary source of myofibroblasts in fibrotic kidneys is renal EPO-producing cells (REPs) [18]. These authors have published that the transformation of REPs into myofibroblasts has two major consequences: it leads to anemia due to erythropoietin deficiency and contributes to renal fibrosis by activating αSMA-positive matrix-producing cells. This myofibroblastic transition of REPs appears to be regulated independently by NFκB and TGF-β–Smad signaling pathways.

Numerous studies have identified specific miRNAs that exhibit altered expression patterns in the blood of CKD patients. These differentially expressed miRNAs can be categorized into those involved in inflammation, fibrosis, oxidative stress, endothelial dysfunction, and other processes relevant to CKD pathogenesis [19]. By quantifying the levels of these miRNAs in the blood, clinicians and researchers can gain insights into the disease status and potentially predict future outcomes. Research has demonstrated that miRNAs are involved in the regulation of erythropoiesis, and specific miRNAs have been identified that regulate genes essential for erythroid differentiation and maturation [20]. For instance, miR-144 and miR-451 are crucial for the terminal differentiation of murine erythroid cells [20]. Dysregulation of these miRNAs can lead to ineffective erythropoiesis and contribute to the development of anemia. One miRNA that has been extensively studied in the context of CKD is miR-223, but its role in the context of CKD-related anemia has not been characterized [21,22]. miR-223 is encoded within the intron of the gene encoding the myeloid-specific transcription factor, Mef2c, and is predominantly expressed in myeloid cells, including granulocytes and monocytes. It has been implicated in multiple physiological and pathological processes, including inflammation, erythropoiesis, and cancer [23]. In the context of kidney disease, miR-223 has demonstrated important regulatory effects on cellular and molecular pathways involved in CKD pathogenesis [21]. Several studies suggest that miR-223 may contribute to the development and progression of CKD through its influence on inflammation and fibrosis. miR-223, alongside other miRNAs, has emerged as a promising blood biomarker in CKD due to stability, specificity, and detectability in various body fluids, including plasma and serum [17,24]. Indeed, the dysregulation of miRNAs in CKD reflects the underlying pathophysiological changes in the kidneys and other affected organs, making them valuable indicators of disease progression, prognosis, and therapeutic response. The role of miR-223 in anemia-related CKD disease has however not been fully investigated to our knowledge.

In this work, the role of miRNAs in CKD-related anemia was explored. In order to identify miRNAs that are disrupted in the context of anemia associated with CKD, UT7/EPO cells were used as an initial screening method to pinpoint the most significantly affected miRNAs, and found miR-223 to be the most relevant. These results were further confirmed using human primary culture cells. Finally, we looked at links between IS, other uremic toxins and miR-223 in a large cohort of CKD patients (process summarized in Fig. 1) (see Fig. 2).

Fig. 1.

Summary of the translational process of the study. To identify miRNAs disrupted in the context of anemia associated with CKD, UT7/EPO cells were used as a screening method to detect the most significantly affected miRNAs. The results were confirmed using human primary culture cells. Finally, links between IS, other uremic toxins and miR-223 were studied in a large cohort of CKD patients.

Fig. 2.

Relative intracellular expression of the 20 most deregulated miRNAs in the presence of IS compared to Control condition in the human erythropoietic cell line UT7/EPO. IS concentrations were 250 μM and 1 mM with an incubation for 48 h (n = 3). (∗P < 0.05). IS: indoxyl sulfate.

2. Methods

2.1. Cell culture

The UT7/EPO cell line, which is a type of human leukemic cell line used to study EPO-driven proliferation, was cultured in α-minimum essential medium (α-MEM) from Dominique Dutscher as described in Ref. [7]. The culture medium was supplemented with 10 % fetal calf serum (FCS) from Eurobio, 1 % penicillin-streptomycin from Biowest, and 2 IU/mL of recombinant human EPO (rhEPO, epoetin beta) [14]. To investigate gene expression, UT7/EPO cells were treated with or without IS (250 μM and 1 mM) for 48 h. The cells used in the experiments were derived from passages 7 to 10, and they were regularly subcultured every 3–4 days. All cells were grown under standard cell culture conditions, maintained at 37 °C in an environment with 5 % CO2. Human hematopoietic cells were acquired from mobilized peripheral blood. Primary human CD34⁺ cells were separated using magnetic microbeads (purity >94 ± 3 %) on MACS columns (AutoMACS Separator) [14]. To initiate the process of erythroid differentiation, CD34⁺ cells were cultivated in Iscove's Modified Dulbecco's Medium (IMDM) supplemented with various components: 2 IU/mL unfractionated heparin, 5 % human plasma (provided by Etablissement Français du Sang, EFS), 10 μg/mL human recombinant insulin (Sigma-Aldrich), 1 % penicillin-streptomycin (PAN™-Biotech), and 330 μg/mL human holo-transferrin (Sigma-Aldrich). The expansion method described by Giarratana et al. [25] was employed, involving sequential cytokine stimulation in the media. From day 0 (D0) to day 7 (D7), the media contained rhEPO (3 IU/mL), human stem cell factor (hSCF 100 ng/mL, Miltenyi), recombinant human interleukin-3 (rhIL-3, Miltenyi), and dexamethasone (DXM 10⁻⁶ M, Sigma-Aldrich). DXM and rhIL-3 were removed at D7, hSCF was removed at D10, and EPO was maintained throughout the culture. At D4 and D7, CD34⁺ cells were treated with 250 μM of IS, coinciding with the days when the culture medium was changed. The progression of erythroid differentiation was monitored using flow cytometry and May-Grünwald-Giemsa (MGG) staining, following standard procedures [7].

2.2. Study population

Samples were collected at the outpatient clinic of Ghent University Hospital in Belgium including a total of up to 515 patients who had been diagnosed with CKD stages G1 to G5 (Fig. 1). [26] The sample collection took place from January 2011 till January 2014 and excluded patients under 18 years old, those with active infection, pregnancy, malignancy, or a history of transplantation. The monitoring of outcome parameters lasted for a period of 5.5 years until June 2017. The CKD patients were categorized into subgroups according to their estimated glomerular filtration rate (eGFR) using the 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. The study compared eight subgroups, which included six CKD subgroups not undergoing dialysis (eGFR: over 90, 60–89, 45–59, 30–44, 15–29, and below 15 mL/min/1.73 m²), one CKD subgroup receiving dialysis (stage G5D), and a group of healthy controls. Concerning HD patients, 73 % of them were on post-dilution hemodiafiltration, whereas 13 % were on pre-dilution hemodiafiltration. Most of them (63 %) were filtrated using FX800 membranes. Main treatments were Aranesp, (32 %) and Mircera (37 %). Hypertension was defined as having a systolic blood pressure above 140 mmHg and/or a diastolic blood pressure above 90 mmHg or the use of antihypertensive medication. We considered EPO administration as 1 and the rest as 0. The purpose was to study miR-223 levels in all CKD subgroups. To ensure accurate analysis of serum miR-223, levels uremic toxins, and endothelial markers, samples were collected, divided into smaller portions, and stored at −80 °C until batch analysis. For each analysis a fresh portion was used, and no thaw/freeze cycles were performed.

Ethical approval

This research was granted authorization by the ethical committee of Ghent University Hospital (approval code 2010/033; B67020107926). All participants provided written consent after receiving detailed information about the study. The investigation followed the guidelines outlined in the Declaration of Helsinki and its subsequent revisions. Human hematopoietic cells were acquired from mobilized peripheral blood with the patients' informed consent, following the guidelines of the Helsinki protocol [27]. This research adhered to the regulations of non-interventional studies in France.

2.3. nCounter system (NanoString) gene expression profiling

Concerning the transcriptomics study, 100 ng of RNA was run for each sample, using the nCounter miRNA Expression Panel (NanoString nCounter Human miRNA Expression Assay v.1, NanoString, Seattle, WA, USA), and run on a nCounter Sprint Profiler, as described in Ref. [28]. Dataset was analyzed with nCounter® advanced analysis software. To define statistically significant genes, p-value were adjusted using the Benjamini-Yekutieli (BY) method. The recommendations outlined by NanoString Technologies were all followed regarding mRNA sample preparation, hybridization, detection and scanning, and data normalization.

2.4. RNA extraction and quantification of miRNA and mRNA levels

The techniques have been described in detail previously in Refs. [7,14,17]. Briefly, the RNA extraction and quantification of serum miRNA and cellular mRNA levels were conducted as follows. Blood samples from patients were collected in Venosafe serum tubes and left at room temperature for 30 min. Afterward, the samples were aliquoted at 4 °C and stored at −80 °C until analysis. The extraction of total RNA was performed using the miRNeasy Serum/Plasma kit for serum and the RNeasy Mini kit for the cells following the manufacturer's instructions. To serve as an internal control, a specific quantity of exogenous Caenorhabditis elegans miR-39 was added. The extracted RNA samples were stored at −80 °C until further use. The quality of RNA was evaluated using a NanoDrop spectrophotometer.

2.5. qRT-qPCR

RNA was reverse-transcribed into cDNA using TaqMan miRNA-specific primers and the dedicated TaqMan microRNA reverse transcription kit (Thermofisher), as described in Refs. [14,17]. For each patient and cell culture sample, the relative expression level of miR-223 was determined using the 2^−ΔCq method, with ΔCq calculated as the difference between the quantification cycle (Cq) values of the target miRNA and cel-miR-39 (control gene). Concerning LIM domain only2 (LMO2) levels, SYBR was used, and the reference gene used was GAPDH. To avoid bias, all experiments were conducted by a single individual who was blinded to patient information. Exogenous Caenorhabditis elegans miR-39 and GAPHD (internal controls) levels were measured as described previously [7] (Supplementary Table 1).

2.6. Quantification of endothelial dysfunction markers

We simultaneously gauged the levels of various markers related to endothelial dysfunction in plasma samples. The specifics of these measurements were extensively explained in Ref. [26]. Magnetic Luminex Assays from R&D Systems (Minneapolis, MN, USA) with a premix multiplex was employed to measure the endothelial markers, and all patient samples were analyzed using the same lot number (L123484). These markers included matrix metalloproteinase (MMP-7), vascular cell adhesion molecule-1 (VCAM-1), angiopoietin-2 (ANGPT2), and syndecan-1 (Sdc-1). The magnetic luminex assay premix multiplex was categorized as LXSAHM. The experimental data underwent analysis by fitting a 5-parameter logistic curve to the standard analyte curves. The calibration curves had the following lowest standards: 115.08 pg/mL for SDC-1, 8264.8 pg/mL for VCAM-1, 91.3 pg/mL for ANGPT-2, and 320.61 pg/mL for MMP-7.

2.7. Statistical analysis

Concerning the cohort, we examined the relationships between patient characteristics and levels of various biomarkers [as previously described in 12,15]. Statistical analyses were performed with JASP (version 0.18.3 for Windows, JASP Team, VU Amsterdam, Amsterdam, The Netherlands). All data are expressed as mean and standard deviation (±SD). According to the normality of the data evaluated by Shapiro-Wilk and equality of variance (Levene's test), Kruskal-Wallis and Dunn's post hoc tests were performed to compare difference in quantitative variables between all groups. Pearson (r) and/or Spearman's rank-order correlation (ρ) were used to evaluate the strength and direction of the monotonic association between miRNA-223 expression and other physiological variables. Spearman test was considered relevant when the r value was between 0.5 and 1 or between −0.5 and −1. Concerning cell cultures, experiments were repeated at least three times and all results were expressed as mean ± standard error of the mean (SEM). Comparisons between groups were performed using the Kruskal-Wallis test and a comparison of two groups was done by Mann–Whitney U test [7]. The threshold for statistical significance was set at p < 0.05 for all tests.

3. Results

3.1. Transcriptomic analysis by nanostring

A transcriptomic analysis was performed using a Nanostring setup dedicated to miRNA expression. Among 804 miRNAs, we selected the 20 most deregulated miRNAs in the presence of IS compared to Control condition. Two IS concentrations were used, 250 μM and 1 mM, found in later stages CKD patients [29]. 250 μM corresponds to the average concentration found in CKD patients whereas 1 mM corresponds to the highest concentrations found [7]. Also, we have defined 250 μM as the condition induce apoptosis in erythropoieic models [7]. Among all the miRNAs analyzed, 2 miRNAs were significantly overexpressed at both IS concentrations (250 μM and 1 mM) compared to the control condition: miR-1246 and miR-223-3p. At 1 mM IS, the miR-1246 appears to be the most deregulated miRNA, as it was three times more expressed (p < 0.05) compared to control. As for miR-223, it was 1.8 times more expressed (p < 0.05) in the presence of 1 mM IS versus the control condition.

3.2. IS induces intracellular expression of miR-223 in two cellular models

Subsequently, we tried to confirm by RT-qPCR the results with the two most deregulated miRNAs obtained by the Nanostring analysis, ie miR-223 and miR-1246. Only miR-223 was significantly more expressed in the presence of IS compared to the control condition in UT7/EPO cells (Fig. 3A). We next quantified the expression of LIM-only 2 (LMO2), a documented target of miR-223 involved in erythropoiesis [30]. As expected, as LMO2 is a target of miR-223, its levels were decreased in the presence of a rise of miR-223 (Fig. 3A). In addition, we checked overexpression of miR-223 in primary human CD34⁺ cells, directed towards the erythropoietic lineage, on D7 of culture, in the presence of IS 250 μM with respect to the control condition (Fig. 3B). Since IS alters early erythropoiesis, we focused on the 7th day in culture as this condition reflects this particular stage. This overexpression of approximately 2.2 was confirmed, and found to be more pronounced in this physiological model of erythropoiesis than in UT7 cell line. We however could not confirm by RT-qPCR the results of the Nanostring for miR-1246 as we found no significant difference in either experimental models (data not shown), therefore we did not further investigate this miRNA.

Fig. 3.

Relative expression of miR-223 and its target LIM-only 2 (LMO2) after 48h of incubation with IS (250 μM) in the human erythropoietic cell line UT7/EPO (A) and of miR-223 in CD34⁺ primary culture cells (after 7 days of culture) (B) (n = 4). (∗P < 0.05 or ∗∗ < 0.01). IS: indoxyl sulfate.

3.3. Cohort study: correlation between serum levels of various biological parameters

To explore the association between circulating levels of miR-223 and uremic toxins, correlation analyses were conducted on major uremic toxins: IS, PCS, PCG, IAA., etc. When relevant, both total and free plasma concentrations of these toxins were considered. Previously, all uremic toxin levels were shown to be increased with advancing stages of CKD [14,17].

When subdividing the total population on the basis of EPO and/or dialysis treatment, the expression of miR-223 varied between groups, with higher levels in patients on dialysis and/or treated with EPO (Table 1). As IS levels are highest in dialysis patients, these results are in line with our in vitro results on CD34⁺ cells in which a difference in miR-223 expression was observed according to the presence or absence of IS (see Table 2).

Table 1.

Expression of miR-223 according to different subgroups based on EPO and/or hemodialysis treatment - Dunn's comparison test.

subgroups	n	miR-223 (Mean ± SD)
Controls	31	3.8 ± 2.3
With EPO and on dialysis	14	1.0 ± 0.9a
With EPO and without dialysis	71	2.1 ± 1.9a,b
Without EPO and on dialysis	19	1.3 ± 0.8a
Without EPO and without dialysis	493	3.4 ± 2.9b,c,d

^ap<0.05 vs controls; ^bp < 0.05 vs “with EPO and on dialysis”; ^cp < 0.05 vs “with erythropoietin (EPO) and without dialysis”; ^dp < 0.05 vs “without EPO and on dialysis”.

Table 2.

Expression of miR-223 as a function of different variables in the general population. Spearman test.

Parameter	Spearman r (related to miR-223)	p-value
Serum creatinin	−0.27	<0.001∗∗∗
IS tot	−0.25	<0.001∗∗∗
IS Free	−0.30	<0.001∗∗∗
pCS Tot	−0.16	<0.001∗∗∗
pCS Free	−0.18	<0.001∗∗∗
pCG Tot	−0.19	<0.001∗∗∗
pCG Free	−0.20	<0.001∗∗∗
IAA Tot	−0.15	<0.001∗∗∗
IAA Free	−0.19	<0.001∗∗∗
CMPF	−0.06	0.09
HA Tot	−0.10	0.013∗
HA Free	−0.10	0.007∗
Uric Acid	−0.06	0.089
CRP	0.01	0.756
Hemoglobin	0.31	<0.001∗∗∗
MMP-7	−0.12	0.005∗∗
VCAM-1	−0.12	0.004∗∗
Sdc-1	−0.098	0.023∗
ANGPT2	−0.10	0.018∗
ACE-ARB	−0.04	0.333
ADPKD	0.01	0.682
EPO	−0.22	<0.001∗∗∗

Abbreviations: Angiopoietin-2 (ANGPT2); angiotensin-converting–enzyme (ACE) inhibitors; angiotensin-receptor blockers (ARBs); Autosomal dominant polycystic kidney disease (ADPKD); Chronic Kidney Disease (CKD); C-reactive protein (CRP); Erythropoietin (EPO); Indoxyl sulfate (IS); p-cresyl sulfate (pCS); P-cresol (pCG); indole-3-acetic acid (IAA); 3-carboxy-4-methyl-5-propyl-2-furan propionic acid (CMPF); Hippuric acid (HA); matrix metalloproteinase-7 (MMP-7); Syndecan-1 (Sdc-1); Vascular cell adhesion protein 1 (VCAM-1).

In the total population, there was no significant correlation between miR-223 and the different variables. (Table 2). However, we observed significant p-values, and r-values of −0.25 for total IS and −0.30 for free IS, −0.27 for creatinine levels and 0.31 for hemoglobin. The previously reported independent positive association of miR-223 with hemoglobin, athough not reaching significance, was confirmed in all groups in the present study [17]. A significant correlation with EPO was also noted. There was no clear trend when looking at markers of endothelial dysfunction, whether MMP-7, VCAM-1, Sdc-1 or ANGPT2. We also looked at the effect of treatments with ACEi/ARBs as they have a known inhibiting effect on erythropoiesis. In our conditions, there was no correlation detected concerning these drugs (Table 2). We also compared the subgroup of patients with autosomal dominant polycystic kidney disease (ADPKD) versus other kidney pathologies. Indeed, ADPKD patients are much less likely to have kidney related anemia for the same kidney function [31]. Again, there was no difference detected between the two groups (Table 2). There also was no difference when comparing hemoglobin values between the ADPKD and other patients when using an ANOVA test (P 0.828) (see Table 3).

Table 3.

Expression of miR-223 as a function of different variables in patients with EPO treatment, without dialysis.

Parameter	Spearman r (related to miR-223)	p-value
Serum creatinine	−0.16	0.181
IS tot	−0.20	0.096
IS Free	−0.31	0.009∗∗
pCS Tot	0.04	0.736
pCS Free	−0.02	0.864
pCG Tot	−0.04	0.743
pCG Free	−0.08	0.481
IAA Tot	0.07	0.558
IAA Free	−0.02	0.853
CMPF	0.23	0.058
HA Tot	0.19	0.108
HA Free	0.18	0.124
Uric Acid	0.08	0.507
CRP	−0.17	0.148
Hemoglobin	0.21	0.071
MMP-7	−0.06	0.618
VCAM-1	−0.422	0.001∗∗
Sdc-1	−0.19	0.150
ANGPT2	−0.29	0.029∗
ACE-ARB	0.08	0.507
ADPKD	0.08	0.535
EPO	−0.22	<0.001∗∗∗

Abbreviations: Angiopoietin-2, ANGPT2; angiotensin-converting–enzyme (ACE) inhibitors; angiotensin-receptor blockers (ARBs); Autosomal dominant polycystic kidney disease (ADPKD); Chronic Kidney Disease, CKD; C-reactive protein, CRP; Erythropoietin, EPO; Indoxyl sulfate, IS; p-cresyl sulfate, pCS; P-cresol, pCG; indole-3-acetic acid, IAA; 3-carboxy-4-methyl-5-propyl-2-furan propionic acid, CMPF; Hippuric acid, HA; matrix metalloproteinase-7, MMP-7; Syndecan-1, Sdc-1; Vascular cell adhesion protein 1, VCAM-1.

Linear regression was performed to assess whether miR-223 expression is influenced by different variables (Supplementary Tables). The results of our multivariate analysis in the total population show that no model indicates a link between miR-223 expression and the variables studied. Indeed, these variables explain only 7 % of the variation in miR-223 expression (R2 = 0.07, Supplementary Table 2). Still we noticed negative spearman values with IS total, IS free, and serum creatinine (−0.25, −0.30 and −0.27 respectively). This result suggests that other factors, not studied here, influence the variation in expression of miR-223.

To further study the potential impact of EPO and/or dialysis treatment, taking into consideration that there is a difference between the groups, and the noted influence of EPO, the same approach was applied to the individual subgroups with EPO not on dialysis, without EPO on dialysis, with EPO and dialysis, and without EPO not on dialysis.

Concerning interactions between miR-223 levels and various parameters in the EPO group, not on dialysis (Table 3), the Spearman r value of two endothelial biomarkers was of interest. VCAM-1 was closer to −0.5 than in the total group, and the r value was close to significance (−0.12 vs −0.42). ANGPT2 displayed a significant p-value and a Spearman correlation of −0.31. Also free IS had a significant r value and a Spearman correlation of −0.31 (see Table 4).

Table 4.

Expression of miR-223 as a function of different variables in patients without EPO treatment, without dialysis.

Parameter	Spearman r (related to miR-223)	p-value
Serum creatinine	−0.28	<0.001∗∗∗
IS tot	−0.21	<0.001∗∗∗
IS Free	−0.25	<0.001∗∗∗
pCS Tot	−0.17	<0.001∗∗∗
pCS Free	−0.17	<0.001∗∗∗
pCG Tot	−0.13	0.001∗∗
pCG Free	−0.15	<0.001∗∗∗
IAA Tot	−0.16	<0.001∗∗∗
IAA Free	−0.19	<0.001∗∗∗
CMPF	−0.15	<0.001∗∗∗
HA Tot	−0.10	0.02∗
HA Free	−0.10	0.015∗
Uric Acid	−0.17	<0.001∗∗∗
CRP	0.02	0.509
Hemoglobin	0.21	<0.001 ∗∗∗
MMP-7	−0.15	<0.001∗∗∗
VCAM-1	−0.13	0.002∗∗
Sdc-1	−0.12	0.007∗∗
ANGPT2	−0.11	0.015∗
ACE-ARB	−0.059	0.204
ADPKD	0.0007	0.987

Abbreviations: Angiopoietin-2, ANGPT2; angiotensin-converting–enzyme (ACE) inhibitors; angiotensin-receptor blockers (ARBs); Autosomal dominant polycystic kidney disease (ADPKD); Chronic Kidney Disease, CKD; C-reactive protein, CRP; Erythropoietin, EPO; Indoxyl sulfate, IS; p-cresyl sulfate, pCS; P-cresol, pCG; indole-3-acetic acid, IAA; 3-carboxy-4-methyl-5-propyl-2-furan propionic acid, CMPF; Hippuric acid, HA; matrix metalloproteinase-7, MMP-7; Syndecan-1, Sdc-1; Vascular cell adhesion protein 1, VCAM-1.

In the group of patients without EPO treatment, without dialysis, as in the total patient group, although numerous parameters appeared significant, no correlation was established between miR-223 expression and the different variables studied (Table 4). These results are confirmed by the multivariate linear regression test. Negative values close to 0.3 were however noticed with IS, both free and total, and serum creatinine. Thus, in this sub-group, only 6 % of the variation in miR-223 expression is explained by the variables included in our model (Supplementary Table 2) (see Table 5).

Table 5.

Expression of miR-223 as a function of different variables in patients with EPO treatment, with dialysis, Spearman test.

Parameter	Spearman r (related to miR-223)	p-value
IS tot	−0.39	0.146
IS Free	−0.30	0.276
pCS Tot	−0.34	0.207
pCS Free	−0.37	0.166
pCG Tot	−0.20	0.458
pCG Free	−0.19	0.480
IAA Tot	−0.38	0.160
IAA Free	−0.57	0.027∗
CMPF	−0.356	0.193
HA Tot	−0.357	0.191
HA Free	−0.34	0.207
Uric Acid	0.28	0.300
CRP	0.08	0.756
Hemoglobin	0.40	0.756

Abbreviations: Indoxyl sulfate, IS; p-cresyl sulfate, pCS; P-cresol, pCG; indole-3-acetic acid, IAA; 3-carboxy-4-methyl-5-propyl-2-furan propionic acid, CMPF; Hippuric acid, HA.

In the group of patients with EPO and with dialysis, a correlation was established between miR-223 expression and the level of the uremic toxin IAA free (Table 5). IS total and free also showed a clearer tendency than in the other groups. This analysis confirms a correlation between miR-223 expression and IAA free in the dialysis and EPO group. This variation in expression is at 32 % due to IAA free (Supplementary Table 2). Concerning IAA and other uremic toxins, R2 was up to 40 % in the group with EPO and with dialysis (Supplementary Table 2). This result could be explained by the fact that patients on dialysis have higher levels of these toxins (see Table 6).

Table 6.

Expression of miR-223 as a function of different variables in patients EPO-free on dialysis. Spearman test.

Parameter	Spearman r (related to miR-223)	p-value
IS tot	−0.42	0.07
IS Free	−0.35	0.136
pCS Tot	−0.38	0.109
pCS Free	−0.54	0.015∗
pCG Tot	−0.48	0.03∗
pCG Free	−0.46	0.043∗
IAA Tot	−0.13	0.584
IAA Free	−0.36	0.127
CMPF	−0.06	0.797
HA Tot	−0.40	0.084
HA Free	−0.43	0.061
Uric Acid	−0.25	0.293
CRP	−0.06	0.794
Hemoglobin	0.20	0.436

Abbreviations: Indoxyl sulfate, IS; p-cresyl sulfate, pCS; P-cresol, pCG; indole-3-acetic acid, IAA; 3-carboxy-4-methyl-5-propyl-2-furan propionic acid, CMPF; Hippuric acid, HA.

Finally, in EPO-free patients on dialysis, a non-parametric Spearman's test revealed a significant correlation between miR-223 expression and the pCS and pCG uremic toxins (Table 6). After testing for linear regression, more than 20 % of this variation was explained by pCS toxin and 22 % by pCG toxin (Supplementary Table 2). This result may be explained by the fact that dialysis patients generally have higher levels of these toxins.

4. Discussion

Our study suggests a relationship between uremic toxins levels, various parameters related to CKD and miR-223 seric levels. CKD is a progressive condition characterized by an impaired kidney function, leading to the accumulation of various uremic toxins in the bloodstream [4]. In recent years, emerging evidence has revealed intriguing interactions between miRNAs and specific uremic toxins, suggesting their involvement in CKD pathogenesis [13,14]. miRNAs have been found to regulate key pathways involved in kidney function, such as inflammation, oxidative stress, and fibrosis [16,32]. Dysregulation of miRNAs in CKD can lead to these processes going unchecked, exacerbating kidney damage and leading to the accumulation of uremic toxins. The exact mechanisms underlying the relationship between uremic toxins and miRNA levels, whether seric or cellular, are not yet fully understood. miRNAs play crucial roles in various biological processes, including erythropoiesis. We previously exposed UT7/EPO cells to uremic levels of IS, a uremic toxin, and showed that IS disturbs erythropoiesis [7]. We conducted in this present work a follow-up study to further investigate the molecular mechanism of action of IS. To specifically identify miRNAs dysregulated in the context of CKD-related anemia, UT7/EPO cells were incubated in the presence of IS, and a transcriptomics Nanostring screening method was used to select the most dysregulated miRNAs. We found miR-223 to be significantly increased and these findings were further validated using primary culture cells. Next, a link between serum levels of miR-223 and serum levels of IS and other uremic toxins was suggested in a large cohort of more than 500 patients with CKD at various stages. In this same cohort, we previously determined that miR-223 was as accurate as GFR to detect cardiac and overall in mortality in CKD patients [17]. Still creatinine levels and the CKD-EPI are of course used in clinic routine to check DFG.

Our present study links anemia, CKD and miR-223 cellular and blood levels. We speculate that increased seric levels of miRNA-223 could be due to either decreased renal clearance [16] or general metabolic breakdown [1], or a combination of both. A first link between miR-223 and uremic toxins has been described by others, who showed that IS induces organic anion transporter-1 (OAT1) via aryl hydrocarbon receptor (AhR) and epidermal growth factor receptor signaling, under the control of miR-223 [33]. We previously assessed the levels of miR-223 in the blood as a potential prognostic marker in CKD. We conducted our study on a large group of CKD patients at different stages of the disease, as well as a control group without CKD [17]. Our findings revealed that patients with miR-223 levels below the median had a lower survival rate, along with an increased incidence of cardiovascular and kidney-related events. However, the association between miRNA levels and mortality or disease events in CKD was influenced by the estimated glomerular filtration rate (eGFR). We did not however rule out the possibility of miR-223 playing a role in the development of CKD. Also the study by Ulbing et al.demonstrated an association between miR-223, IL-6, and eGFR [22]. Previously we found a clear relationship between miR-223 and hemoglobin, suggesting a role for miR-223 in CKD-related anemia [17]. miR-223 has also been linked to the anemia field by other studies. It has been shown to inhibit γ-globin expression in β-thalassemia [34,35], to have a role in the development of hypercoagulability in iron deficiency anemia [36] and in aplastic anemia [37]. Modulating miR-223 in CKD patients may thus hold promise in treating or preventing anemia. [21]. This may be done using gene therapy vectors, modified RNAs, CRIPSR Cas9 technology or a combination of the methods as discussed in another work [16]. In our study, we have also checked the link between miR-223 and its target LMO2 [30], confirming that downregulation of miR-223 is necessary to initiate erythroid differentiation and unblock LMO2 protein expression [30]. LMO2 is a transcription factor that regulates erythroid differentiation via activation of a specific gene program. Decreasing LMO2 expression in erythroid progenitors delays G1-S progression and stops EPO-dependent cell growth [38]. Fonelli et al. showed that increased expression of miR-223 reduces the mRNA and protein levels of LMO2, in turn affecting differentiation of CD34⁺ cells directed towards erythroid differentiation. This is in concordance with our results as we show that IS increases miR-223, in turn downregulating LMO2, and thus erythropoiesis (our present results and [6]).

When considering the variance explained as a percentage (from a multiple regression model including all relevant variables), R² was up to 40 % in the group without EPO and without dialysis, suggesting a significant part of this miRNA in anemia-related CKD, and the action of uremic toxins. We know from the literature that miR-223 plays an important role in the proper functioning of erythropoiesis. It has been shown that expression of this microRNA increases during megakaryocytic and granulocytic differentiation [30]. Thus, the overexpression of this microRNA in CD34⁺ cells in the presence of IS, is consistent with the hypothesis of a phenotypic change in cells towards a megakaryocytic rather than erythrocytic profile. This may explain the differences in expression of this microRNA between the different groups of patients (with EPO and dialysis, with EPO without dialysis, without EPO and without dialysis and without EPO and without dialysis), but we cannot conclude that EPO has an effect on miR-223 unless independent from kidney function. Furthermore, the modest correlation between the variables studied (several toxins, VCAM-1, ANGPT2, hemoglobin) points out that other, unmeasured and more complex factors also influence miR-223 expression. In the absence of any significant independent relation, we can hypothesize that the univariate correlations may have been confounded.

Concerning the various uremic toxins, their correlations between miR-223 varied between the different subgroups. In all subgroups, values close to 0.3 were noted with IS. IAA was correlated with miR-223 mainly in the patients on dialysis and EPO treatment. This is interesting as IAA has been linked to cardiovascular morbidity and mortality in preclinical models of CKD, and CKD patients [39]. Finally, in the patients not treated with EPO, but on dialysis, a link was suggested between pCS and pCG, whether free or total. pCS and pCG both derive from the same molecule, p-cresol. The free fraction of PCG appears to have a comparable predictive power for mortality as PCS and IS do [26,40]. This would indicate that the absence of EPO treatment with dialysis favors a link between pCS and pCG toxins and miR-223 on one hand, and in the other hand, the presence of EPO treatment with dialysis favors a link between IAA and miR-223. Both could be speculated to increase cardiovascular risk. Interestingly, another compound related to UT, the modified amino acid lanthionine, induced a significant upregulation of miR-223 at concentrations comparable to those found in uremic patients [41].

A correlation between miR-223 and endothelial dysfunction biomarkers (ANGPT2 and VCAM-1) was found in the EPO group, without dialysis. Interestingly, the correlations were not the same whether patients were treated with EPO, or not. VCAM-1 and ANGPT2 levels were correlated with miR-223 in this subgroup. On the other hand, MMP-7 and SDC-1 were not associated with the miRNA in any subgroup. This heretofore undetected correlation between various biomarkers of endothelial dysfunction and miR-223 corroborates previous findings from Tabet et al. [42] These authors have demonstrated that HDL's anti-inflammatory properties are partly due to HDL delivery of miR-223 to endothelial cells, which in turn induces translational repression of ICAM-1. The same team has also found a profound link between miR-223 and cholesterol homeostasis in the general population [43], but also more importantly in obese patients [44]. We confirm this as we have already described a correlation between miR-223 serum levels and total cholesterol in a previous work [17]. As cholesterol transport, atherosclerosis and endothelial dysfunction are highly associated [45], we can suggest a role for miR-223 in endothelium homeostasis that would need to be further explored in future work. ANGPT2 expression is upregulated in animal models of kidney disease, and ANGPT2 was reported to have a role in arterial stiffness in CKD patients [41]. Glorieux showed a positive correlation between ANGPT2 and pCS (total and free) and free IAA in the present cohort [26]. However, sparse information exists in the literature on a link between various endothelial biomarkers and anemia. ANGPT2 expression was found to be increased in mouse brain after chronic moderate hypoxia [46].

Given our and other's work, miR-223 could be useful as a marker for CKD related-anemia, although much more work would need to be performed to determine its use compared to the clinical standard, hemoglobinemia [17]. Several potential non-invasive miRNA biomarkers have been identified in kidney disease and could enhance diagnostic accuracy for CKD status, predict prognosis, and monitor disease progression. However, large-scale clinical studies involving diverse patient populations are necessary to evaluate the clinical value of miRNAs.

5. Conclusions

One of the significant challenges in managing CKD is the accumulation of uremic toxins in the bloodstream as kidney function declines [26,47]. Anemia is a frequent complication of CKD, where it is linked to premature death due to CVD [48,49]. We found a correlation between IS levels, hemoglobin and miR-223 seric levels. More work is needed to clearly identify miR-223 are as reliable biomarker for early detection and monitoring of anemia-related CVD during CKD progression [22]. This could lead to more timely interventions to slow down kidney function decline. Importantly, miRNAs hold a good potential as non-invasive biomarkers, as they can be easily measured using minimally invasive techniques such as blood samples [17,24]. Moreover, miRNAs can provide dynamic information about disease progression and response to therapy, enabling personalized medicine approaches for CKD patients suffering from anemia. Indeed, by monitoring the expression levels of these miRNAs over time, clinicians can assess the efficacy of interventions and tailor treatment strategies accordingly. In conclusion, miRNAs have emerged as valuable blood biomarkers in CKD, providing insights into pathogenesis, prognosis, and treatment response of this disease. Finally, as uremic toxins are biomarkers roughly equivalent to hemoglobin levels, one could speculate that modulating miR-223 could have a therapeutic purpose in patients resistant to EPO treatments or victim of its numerous side effects. The link between miRNA blood levels and anemia in CKD is an intriguing area of research with promising implications for kidney disease management. While much work remains to be done to fully elucidate the mechanisms and identify specific miRNA biomarkers of CKD-related anemia, the potential for improved detection, monitoring, and personalized treatments makes this a promising avenue for future research in nephrology and clinical practice. Their non-invasive nature and potential for dynamic monitoring make them attractive candidates for clinical application.

CRediT authorship contribution statement

Emma Brisot: Writing – review & editing, Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Data curation. Pierre-Marie Leprêtre: Writing – review & editing, Software, Resources, Formal analysis, Data curation, Conceptualization. Eya Hamza: Writing – review & editing, Methodology, Formal analysis, Data curation. Ophélie Fourdinier: Writing – review & editing, Resources, Methodology. Benjamin Brigant: Writing – review & editing, Methodology, Investigation, Formal analysis. Hakim Ouled-Haddou: Methodology, Investigation, Formal analysis. Gabriel Choukroun: Writing – review & editing, Funding acquisition, Conceptualization. Ziad A. Massy: Writing – review & editing, Funding acquisition, Formal analysis, Conceptualization. Francis Verbeke: Writing – review & editing, Validation, Investigation, Formal analysis, Data curation, Conceptualization. Valérie Metzinger-Le Meuth: Writing – review & editing, Writing – original draft, Supervision, Project administration, Investigation, Formal analysis, Conceptualization, Conceptualization, Data curation, Formal analysis, Investigation, Project administration, Writing – original draft, Writing – review & editing. Griet Glorieux: Writing – review & editing, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Laurent Metzinger: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Data availability statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval statement

This research was granted authorization by the ethical committee of Ghent University Hospital (approval code 2010/033; B67020107926). All participants provided written consent after receiving detailed information about the study. The investigation followed the guidelines outlined in the Declaration of Helsinki and its subsequent revisions. Human hematopoietic cells were acquired from mobilized peripheral blood with the patients' informed consent, following the guidelines of the Helsinki protocol [27]. Concerning primary cultures, Human hematopoietic cells were obtained from mobilized peripheral blood after patients’ informed consent according to Helsinki protocol. This study followed French legislation in terms of non-interventional research.

Funding statement

This work was supported by the Société Francophone de Néphrologie, Dialyse et Thérapeutique (SFNDT), Vifor Pharma, and by the French Ministry of Research (MESRI) (E.H. PhD grant).

Declaration of competing interest

ZAM reports grants for CKD-REIN and other research projects from Amgen, Baxter, Fresenius Medical Care, GlaxoSmithKline, Merck Sharp and Dohme-Chibret, Sanofi-Genzyme, Lilly, Otsuka and the French government, as well as fees and grants to charities from GlaxoSmithKline, Astra Zeneca, Boehringer Ingelheim, and Bayer; these sources of funding are not necessarily related to the content of the present manuscript. GC reports grants from Amgen, Vifor Pharma, GSK and Astellas (honorarium for consultancy and conferences). Other authors had nothing to disclose.

Appendix A. Supplementary data

The following is the Supplementary data to this article.