Urinary miR-221-3p and miR-324-5p in combination with albuminuria as a promising model for non-invasive diagnosis of pediatric celiac disease

Abstract

Celiac disease (CD) is a chronic autoimmune enteropathy triggered by gluten ingestion that can cause intestinal damage and extra-intestinal complications, including renal involvement in children. MicroRNAs (miRNAs) are small, stable, non-coding RNAs detectable in body fluids, and urinary miRNAs are particularly promising biomarkers due to their non-invasive collection and ability to reflect kidney function. While circulating miRNAs have been proposed as diagnostic markers of CD and indicators of adherence to a gluten-free diet (GFD), their role in monitoring renal alterations in pediatric patients remains unclear. Conventional measures, such as albuminuria, detect kidney damage but lack specificity for disease-related changes. Therefore, we aimed at exploring the possibility that urinary miRNAs coupled to albuminuria could be used to diagnose both CD and the presence of renal alterations. We analyzed urinary miRNA profiles in pediatric CD patients at diagnosis, on GFD, and in healthy controls. We found that urinary miR-221-3p and miR-324-5p and albuminuria are significantly dysregulated in children with CD compared to GFD and controls, identifying them as candidate biomarkers for diagnostic applications. Under GFD, their expression normalized supporting their potential role not only in monitoring renal recovery but also in objectively assessing dietary adherence. These findings highlight urinary miRNAs in combination with albuminuria as clinically relevant, non-invasive predictors for a non-invasive diagnosis of pediatric CD. More broadly, they suggest that urinary miRNA profiling could serve as a promising translational method to improve early detection, patient stratification, and longitudinal management across pediatric autoimmune disorders once fully validated in future larger cohorts.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-53245-5.

Introduction

Celiac disease (CD) is a chronic enteropathy that, in genetically predisposed individuals, elicit an inflammatory reaction in the small intestine after the ingestion of gluten proteins contained in wheat, rye and barley¹. The result is a severe damage of the bowel mucosa with consequence flattening of the villi, accompanied by the presence of atypical squamous epithelial cells, an increase in the number of lymphocytes and plasma cells in the lamina propria and hypertrophy of the crypts. In addition to the classic clinical symptoms, such as malabsorption and steatorrhea, patients with celiac disease can display extra-intestinal symptoms such as anemia, osteoporosis, arthritis and infertility with consequent impairment of the quality of life¹. Many efforts have been made to understand the pathogenesis of celiac disease and different studies were aimed at the identification of various targets for candidate therapies including the abolishment/reduction of the epithelial permeability, induction of gluten tolerance, gluten sequestering, peptidases and anti-IL15 therapies². However, to date the only therapy able to restore completely the clinical conditions and the intestinal damage is a gluten-free diet.

The gold standard for the diagnosis of CD is a combination of serological tests and duodenal biopsy. Recently, many researchers focused their interest on the identification of innovative diagnostic methods for CD by employing novel biological biomarkers, such as circulating miRNAs. These biomarkers have been reported to have the potential to replace invasive intestinal biopsies^3,4. MiRNAs, small non-coding RNAs ranging from 18 to 22nt in length, display high stability and tissue specificity, and play significant roles in various biological processes. Owing to these unique characteristics, miRNAs have recently emerged as a promising class of biomarkers⁵.

In our recently published observational prospective study, we studied the expression of circulating miRNAs in the sera of a cohort of CD patients at the time of diagnosis and while on gluten-free diet (GFD), compared to healthy controls. We found that three miRNAs (i.e., miR-192-5p, miR-215-5p and miR-125b-5p) either alone or as a combination of the three, were able to discriminate CD and GFD groups with very good accuracy and specificity. These finding suggested that circulating miRNAs can not only be used for the diagnosis of CD patients having a low TGA-IgA titer but also to monitor their adherence to a gluten-free diet³. To date, the detection of other non-invasive biomarkers such as fecal calprotectin) or urinary gluten immunogenic peptides (GIPs) are neither a reliable method to detect histological changes or atrophy in CD patients after GFD⁶ nor to identify hidden gluten exposures owing to their complex excretion kinetics⁷, respectively.

Since gastrointestinal disorders are usually associated with an increased risk of kidney disease, we sought to investigate whether circulating miRNA expression profiles in CD patients correlate also with urinary miRNAs. CD patients may sometimes exhibit IgA nephropathy (IgAN), a glomerulonephritis that is characterized by dominating glomerular IgA deposits. The overexpression of the IgA1 mesangial receptor, also referred as the transferrin receptor (TfR), can be found in the mesangial cells of the glomeruli. Interestingly, this receptor is also overexpressed in CD patients. Moreover, IgAN patients that adhere to a gluten free diet, generally improve histological conditions and clinical parameters⁸.

As a result, we hypothesized the existence of a potential relationship between circulating and urinary miRNAs and therefore we were aimed at studying the potential role of urinary miRNAs and albuminuria as potential biomarkers of CD and potential renal diseases in CD patients^9,10.

In fact, miRNAs that originate from kidney and urinary tract cells, once filtered by glomeruli flow into urine bound to proteins or carried inside extracellular vesicles. Recently, the analysis of urinary miRNAs has been applied to detect and monitor acute and chronic renal diseases¹¹. Currently, tests such as creatinine and proteinuria, which assess glomerular filtration rate and structural damage to the glomerular filtration barrier, are used to monitor renal function. However, these traditional biomarkers are unable to discriminate among different types of damage, whereas the use of urinary miRNA could give this information at an early stage of progression in a non-invasive manner¹².

Thus, the aim of this study was to assess the clinical potential of urinary miRNAs coupled to albuminuria to detect not only early renal alterations in pediatric CD patients before and after GFD, but also to evaluate their effectiveness as non-invasive tools for the diagnosis of pediatric celiac disease.

Materials and methods

Study participants

Patients were enrolled in Rome (Italy) at the Hepatology, Gastroenterology and Nutrition Department of the Bambino Gesù Children’s Hospital.

Participants were eligible for inclusion based on the following criteria: age between 3 and 15 years, being positive for both TGA-IgA and EMA-IgA antibodies, having a duodenal biopsy showing Marsh score ≥ 3 or meeting international diagnostic criteria (applicable to the CD group), and being CD patients who had adhered to a GFD for a minimum of one year (the GFD group). Control subjects were chosen from children who had no personal history or diagnosis of inflammatory or autoimmune disorders, no family history of CD, and tested negative for CD-specific serological antibodies. Participants were excluded if they met any of the following criteria: current use of corticosteroids or hormonal treatments, presence of additional gastrointestinal disorders, immunodeficiency or autoimmune diseases, diabetes mellitus, severe psychiatric conditions, or physical disabilities that restricted physical activity. Individuals with severe or concurrent medical or psychiatric conditions that could interfere with protocol evaluation, those with diabetes, any type of complications, additional autoimmune-related disorders, or those receiving pharmacological treatment were not included in the study.

Urine sample collection

Fresh urines (first morning void) were collected in sterile containers and tubes and were processed within 4 h from the collection. Samples were kept at + 4 °C before processing and at -80 °C for long-term storage.

RNA isolation

Total RNA was isolated from a volume of 5mL using the Urine Total RNA Purification Maxi Kit - Slurry format (Norgen biotek), according to the manufacturer’s instructions. Extracted RNA was stored at -80 °C until use.

qPCR profiling of urinary miRNAs

A volume of 5 µl of the RNA samples was reverse transcribed to cDNA in a final volume of 25 µl using miRCURY LNA RT kit (QIAGEN) following the manufacturer’s instructions. The reactions were performed on a Proflex PCR system. qPCR assays were performed by using the FastStart SYBR Green Master mix (Roche) in a final volume reaction of 10 µl. The qPCR assay was performed by using miRCURY LNA miRNA focus Panel (QIAGEN), which contains primers for 179 miRNA, 6 interplate calibrators and 5 RNA spike-in controls, all in duplicate. QuantStudio 12 K was employed for qPCR reactions by using the following PCR program: a denaturation step at 95 °C (10 min) followed by 40 cycles (95 °C for 60 s and 60 °C for 60 s). The raw quantification cycle (Cq) values were analyzed using GenEx 6 Software (MultiD Analyses AB, Sweden) following manufacturer’s instructions. Quantitative PCR data were normalized (global mean expression of all genes)¹³ and a statistical one-way analysis of variance (ANOVA) method with multiple comparisons correction was then performed.

Urine analysis, albumin and creatinine determination

First, we estimated urine metabolites and other physico-chemical parameters, such as urobilinogen, bilirubin, ketones, blood, proteins, nitrites, leukocytes, specific gravity, pH, ascorbate and microalbumin, by using the qualitative and semi-quantitative urine strip method Screen Generic Urina 12, gently provided by ScreenItalia (www.screenitalia.it) evaluated by visual inspection. To evaluate more precisely the analytes content, we optimized the quantitative spectrophotometric dye-binding method for albumin¹⁴ and the Langley-Evans method for creatinine¹⁵. Briefly, albumin was determined by using bromophenol blue (BPB) 1mM and NaHCO3 1 M (buffer solution). The solution slightly changed its color from light blue to more intense blue-violet at higher concentrations. After a 10 min incubation at 25 °C, the color turned blue and fluorescence emission were registered at 630 nm (ex. at 580 nm).

Creatinine was determined by a 1:1 solution made by 3,5-dinitrobenzoic acid in alkaline 2-propanol. After 5 min incubation the solution turned its color to reddish-violet and fluorescence emission were registered at 480 nm (ex. at 400 nm).

Experiments were run in duplicate and albumin amount has been expressed in µg/mg creatinine (as normalizer). To assess the specificity of the albumin assay, we comparatively tested other proteins such as heparin, trypsin and BSA with the same protocol.

ROC curve analysis

ROC curves were calculated by using the library pROC in R Bioconductor (https://www.bioconductor.org) by using the function glm for model generation.

Bioinformatics analysis of miRNA targets

To find the targets of the dysregulated miRNAs (statistically significant miRNAs with − 2 < FC < 2) in CD vs. CTRL and CD vs. GFD, we adopted the bioinformatics workflow already reported in our previous papers. Briefly, the algorithm considered the three most common prediction algorithms for gene target prediction represented by TargetScan, miRanda and PITA. The integration of these algorithms allowed to investigate miRNA conservation (Targetscan), to correlate their profiles with gene expression data (miRanda) and even to calculate the steric occupation occurring during mRNA/miRNA duplex formation (PITA). Predictions obtained by this algorithm originate from a workflow with different underlying principles, and we demonstrated in many papers that its use gave very reasonable and affordable predictions. The list of target genes were then analyzed using the DAVID Bioinformatics tools (https://davidbioinformatics.nih.gov).

Statistics

All variables were expressed as mean ± standard deviation (SD). Qualitative data were expressed as percentage. Comparison of the mean among the groups was performed using ANOVA or non-parametric tests or χ-square (contingency tables). For statistical analyses, we employed SPSS 20.0 (SPSS, Inc., Chicago, IL, USA) and we considered significant a value of p < 0.05.

AI tools

We employed NotebookLM (Gemini 2.5 Flash) to revise the consistency of the discussion section and to revise some sentences to check the presence of language errors.

Results

Patients, urine collection and quality check of samples

A total number of 53 individuals were recruited. These subjects belonged to three groups: newly diagnosed celiac disease patients (CD, n = 18), CD on gluten-free diet (GFD, n = 20) and one group of healthy individuals (CTRL, n = 15) (Table 1).

Table 1.

Patients’ data, genetics and histological characterization of the cohort composed of CD patients, patients on gluten-free diet (GFD) and control (CTRL) individuals.

ENTIRE COHORT
Demographic characteristics	CTRL (n = 15)	CD (n = 18)	GFD (n = 20)	p-value
Gender	M = 8, F = 7	M = 6, F = 12	M = 10, F = 10	0.54
Age (years)	10.42 (3.8)	9.15 (4.3)	11.22 (4.2)	0.28
BMI	22.6 (7.6)	18.1 (4.5)	18.0 (3.2)	0.30
Total serum IgA (g/L)	---	1.2 (0.6)	1.2 (0.4)	0.95
Total serum TGA-IgA at diagnosis (U/mL)	---	5203 (9950)	---	0.04
Total serum TGA-IgA after gluten-free diet (U/mL)	---	---	34.37 (53.26)
EMA-IgA positivity (n, %) (at diagnosis)	---	n = 15 (83%)	n = 12 (60%)	---

Histology	CTRL (n = 15)	CD (n = 18)	GFD (n = 20)
Biopsy confirmed (at diagnosis)	---	n = 4	n = 8
Confirmed according to ESPGHAN guidelines (at diagnosis)	---	n = 14	n = 12
Not available/not performed	n = 15	n = 14	n = 12
Marsch type
0-2	---	---	n=1
3a	---	n=2	n=5
3b	---	---	---
3c	---	n=2	n=2
Genetics
DQ2/DQ8	---	n = 7	n = 9
Negative	---	---	---
n.a.	n = 15	n = 12	n = 7

The Marsch score of GFD patients refers to the score at diagnosis and not after the gluten-free diet. The positivity threshold for serum TGA-IgA is >7 U/mL (negative values are < 3.5 U/mL). The p-value reported for Total serum TGA-IgA at diagnosis is relative to the comparison of the levels of TGA-IgA between CD and GFD patients (reported in the following row).

Standard urinalysis was performed on all samples (n = 53) (Supplementary Table 2). All the analytes showed no statistically significant differences among groups, therefore for a fine determination of albumin level we decided to integrate the data with a more sensitive quantitative method. Urine volume for each sample was then measured to collect five ml for further analysis. Eight samples (i.e., one for CTRL, four for CD and three for GFD) were excluded owing to insufficient volume (less than 5 ml), whereas other six samples (i.e., two for CTRL, one for CD and three for GFD) were excluded after qPCR analysis owing to the limited number of circulating miRNAs identified (less than 100) (Fig. 1). These selection criteria are quite stringent and led to a high sample exclusion rate. We preferred to be more stringent during the discovery phase in order to find only relevant biomarkers that could be validated in future validation cohorts. We are also assessing new RNA extraction protocols in order to employ little volumes (approx. 2 ml) and allow the miRNA analysis also with a limited amount of urine.

Fig. 1.

Workflow of the analytical steps followed for qPCR analyses. The initial cohort was first filtered according to available urine volume then, after expression analyses by miRNA number (greater than 100 miRNAs). qPCR detected miRNAs were further filtered by their statistical significance (FDR p-value < 0.05), then according to their expression level (-2 > FC > 2). Only two miRNAs resulted highly significant and differentially modulated in CD compared to controls or to GFD patients.

Urinary miRNAs in CD patients are dysregulated compared to GFD and controls

We evaluated the dysregulation of urinary miRNAs by high-throughput quantitative PCR (qPCR) that employed very sensitive LNA primers¹⁶ designed and specifically optimized to detect 179 miRNAs. We obtained two lists from qPCR data: the first containing five statistically significant (p < 0.05) urinary miRNAs in CD patients (compared to CTRL) and the second one containing nine miRNAs (compared to GFD). Of these, 3 miRNAs in CD compared to CTRL and 8 miRNAs in CD compared to GFD were also dysregulated at least two-fold (Figs. 1 and 2).

Fig. 2.

Up- and down-regulated urinary miRNAs: (A) CD compared to controls (CTRL) with the correspondent FC values, (B) CD compared to GFD with the correspondent FC values and (C) GFD compared to controls with the correspondent FC values. ANOVA analysis performed on the three groups resulted in a list of significant miRNAs that are represented in the dendrogram depicted in (D) and the correspondent FC values are reported in the tables. Expression levels are colored according to the fold of change (FC) indicated by the correspondent scale.

A total number of seven miRNAs were upregulated (i.e., miR-100-5p, miR-30b-5p, miR-221-3p, miR-16-5p, let-7b-3p, miR-331-3p, let-7c-5p), whereas five miRNAs were downregulated (i.e., miR-32-5p, miR-185-5p, miR-28-5p, miR-324-5p, miR-1260a). Two dysregulated miRNAs are in common between these two comparisons (i.e., miR-221-3p and miR-324-5p) that were also dysregulated more than two-fold. By comparing GFD vs. Controls five miRNAs were upregulated (i.e., miR-2110, miR-28-5p, miR-874-3p, miR-185-5p and miR-378a-3p). Only two miRNAs (i.e., miR-28-5p and miR-185-5p) were also found in the group of dysregulated miRNAs in CD vs. GFD, although with an opposite fold-change.

By considering the three groups together (CTRL, CD and GFD) and performing ANOVA analysis, we obtained a final list of six statistically dysregulated urinary miRNAs (i.e., miR-185-5p, miR-32-5p, miR-30c-5p, miR-324-5p, miR-221-3p and miR-28-5p) (Fig. 2D). Interestingly, miR-221-3p and miR-324-5p were statistically significant (p-value (FDR) < 0.05) suggesting that the presence of these two miRNAs in the urine of CD patients could be correlated to some clinical characteristics.

CD patients display a mild albuminuria compared to GFD and controls

We assessed the amount of albumin (expressed in mg) present in the urine of CD, GFD patients and control individuals and we normalized it with the level of urinary creatinine (expressed in g) to consider the urine dilution. The first morning voids are more reliable than spot urine samples to assess microalbuminuria, and correlate better with the gold-standard method (i.e., urinary albumin excretion in a 24 h collection) according to the literature¹⁷.

We found that the excreted albumin in CD patients is (85.40 ± 18.31 mg/g), whereas in GFD patients is (43.74 ± 9.43 mg/g), slightly higher than that of controls (25.76 ± 7.36 mg/g) (Fig. 3A). ANOVA analysis (p-value = 0.010) resulted in a post-hoc p-value = 0.004 for CD vs. CTRL and a p-value = 0.027 for CD vs. GFD. Values < 30 mg/g are considered normal, microalbuminuria is present with values between 30 and 300 mg/g, whereas albuminuria occurs with values > 300 mg/g. Out of the 15 CD patients that we analyzed, only two patients displayed normal albuminuria (12.4 and 17.3 mg/g), whereas the remaining patients were all above the limit of 30 mg/g. None of them displayed values greater than 300 mg/g.

Fig. 3.

(a) ANOVA analysis of excreted albumin (in mg/g creatinine) in the three groups of patients (CD, GFD and controls) and (b) The specificity of the quantification method toward albumin. Other proteins cannot be adequately detected as their concentrations are not linear.

Our results indicate that CD patients display a marked microalbuminuria compared to GFD and controls, and that after a gluten-free diet, patients recover and display a lower excretion of albumin. Of note, in all of the groups of our cohort some of the analyzed children declare to perform regular physical activity (i.e., sports, physical exercise at school, etc.) that usually lead to an increased level of albumin in urines. However, this had a negligible effect when considering the mean albumin levels. Interestingly, the colorimetric method that we employed for this kind of assessment is linear and highly specific to detect only albumin (Fig. 3B) and not other urinary proteins.

In fact, to validate the analytical method, we repeated the measurements by using different proteins (i.e., heparin, trypsin, the urinary protein IgG) and a mixture of heparin and albumin to assess the specificity of the assay toward albumin alone. Results showed that the method was effective in selectively quantifying only albumin. However, although an overestimation of albumin can be possible in our analysis, this eventually occurred in all the samples analyzed, and did not interfered with the statistical significance of results we obtained.

Urinary miR-221-3p and miR-324-5p in CD and GFD patients

Next, we plotted the relative expression of the two statistically significant urinary miRNAs (i.e., miR-221-3p and miR-324-5p) that are either expressed in CD vs. GFD and CD vs. controls for all the cohort that we considered (Figs. 2 and 4).

Fig. 4.

(a) Expression level of miR-221-3p in the three groups of patients (CTRL, CD and GFD) with the relative statistical significances. (b) Expression level of miR-324-5p in the three groups of patients (CTRL, CD and GFD) with the relative statistical significances.

Although the expression values are quite dispersed around the mean, the differences are statistically significant (ANOVA p-value (FDR) = 0.043 for miR-221-3p and p-value (FDR) = 0.041 for miR-324-5p) between the three groups (Fig. 4).

ROC curve analysis and predictive models for CD diagnosis

We next sought to investigate the predictive power of the two dysregulated urinary miRNAs coupled to the albumin level to predict patients with celiac disease, either CD and GFD or CD alone, compared to control individuals. We were aimed to find a candidate predictive model made by a suitable combination of traditional (i.e., albumin levels) and molecular biomarkers (i.e., urinary miRNAs) and able to suggest the presence of the disease. We therefore calculated the ROC curves for miR-324-5p, miR-221-3p and albumin amount to demonstrate that the individual variables were not able to predict satisfactorily the CD + GFD group compared to controls (Fig. 5A) or the CD group alone (Fig. 5B) compared to controls.

Fig. 5.

(a) ROC curves of miR-324-5p, miR-221-3p and albumin for the prediction of CD and GFD patients (CD+GFD vs. Controls). (b) ROC curves of miR-324-5p, miR-221-3p and albumin for the prediction of the CD group only (CD vs. Controls).

Next, we calculated the ROC curves of several models obtained by combining: (i) miR-324-5p with miR-221-3p, (ii) either miR-324-5p or miR-221-3p with albumin amount or (iii) both miRNAs and albumin levels, to investigate the diagnostic efficacy of the combination of these predictors.

Figure 6A reports the ROC curves of the four models employed to predict the group of CD and GFD, where the model obtained by the combination of miR-221-3p and albumin reached the highest performance (66.7% specificity, 90.5% sensitivity and an AUC of 0.799), whereas the addition of miR-324-5p in the model did not add any improvement. In fact, the specificity, sensitivity and AUC of these two models resulted the same. Figure 6B reports the ROC curves of the four models employed to predict the CD group alone. The best model with the highest performance resulted the one obtained by the combination of both miR-324-5p and miR-221-3p with albumin levels (89.5% specificity, 90.9% sensitivity and an AUC of 0.938).

Fig. 6.

(a) ROC curves of the models obtained by combining miR-324-5p and miR-221-3p (red line), miR-324-5p and albumin (blue line), miR-221-3p and albumin (green line) and both miRNAs and albumin for the prediction of CD and GFD patients. (b) ROC curves of the models obtained by combining miR-324-5p and miR-221-3p (red line), miR-324-5p and albumin (blue line), miR-221-3p and albumin (green line) and both miRNAs and albumin for the prediction of the CD group only.

Although these results are very promising for diagnostic applications, we are aware that a larger cohort of patients is needed to confirm these results and obtain a refined model. Notwithstanding, we are confident that these two urinary miRNAs that we have analyzed coupled to albumin levels could be effective biomarkers of CD in pediatric patients.

Finally, due to the good ability (65.0% specificity, 80.0% sensitivity and an AUC of 0.695, data not shown) of this model to predict also GFD patients that generally have low values of TGA-IgA due to prolonged gluten-free diet, we can hypothesize that the model could be able to diagnose also CD patients with low levels of TGA-IgA that generally should undergo intestinal biopsy. This would add a further advantage for these patients (i.e., to avoid invasive EGD) although to confirm this data, a dedicated recruitment of low-titer CD patients should be carefully performed.

Target genes of dysregulated urinary miRNAs

The prediction of the target genes of the seven upregulated (i.e., miR-100-5p, miR-30b-5p, miR-221-3p, miR-16-5p, let-7b-3p, miR-331-3p, let-7c-5p) and the five downregulated miRNAs (i.e., miR-32-5p, miR-185-5p, miR-28-5p, miR-324-5p, miR-1260a) in CD patients resulted in a number of 5790 and 5434 unique genes, respectively. However, we restricted the analysis only to those genes targeted by at least two miRNAs, leading to a number of 2153 and 1171 genes respectively for the upregulated and downregulated miRNAs. These genes were further classified according to their biological functions and pathways to obtain information about their role and the consequent role of the targeting miRNAs. The genes targeted by upregulated miRNAs belonged to axon guidance (hsa04360) (FDR = 0.005), MAPK signaling pathway (hsa04010) (FDR = 0.007), and mTOR signaling pathway (hsa04150) (FDR = 0.049) (Fig. 7A), and particularly involved in protein dephosphorylation (GO:0006470) (FDR = 0.04). No significant pathways were obtained with the list of genes targeted by downregulated miRNAs although they resulted involved in the positive regulation of transcription from RNA polymerase II promoter (GO:0045944) (FDR = 0.046) and chromatin remodeling (GO:0006338) (FDR = 0.046) (Fig. 7B).

Fig. 7.

Venn diagrams showing the genes targeted by upregulated (a) or downregulated (b) urinary microRNAs and their ontology classification into pathways and biological processes.

These findings are in line with previous papers suggesting that the involvement of MAPK and/or mTOR signaling are crucial pathways in inflammasome activation¹⁸, a key process in celiac disease (CD) pathogenesis, but also relevant in other autoimmune diseases^19,20.

Discussion

We and others reported that microRNAs (miRNAs) are crucial post-transcriptional regulators, playing significant roles in the development and progression of various kidney diseases, and their stability in biological fluids such as urine and plasma makes them promising non-invasive diagnostic and prognostic biomarkers, as well as potential therapeutic targets^21–26. In our previous publication, we have demonstrated that miR-192-5p, miR-215-5p and miR-125b-5p either alone or in combination, were able to discriminate CD and GFD groups with very good accuracy and specificity, but none of them were found significantly dysregulated in urines³. Therefore, we did not performed any correlation studies. Moreover, none of the circulating miRNAs previously found in sera were found dysregulated also in urine, suggesting that each biological fluids contain a peculiar profile of miRNAs that represent the specific surrounding environment in which these miRNAs are expressed.

Interestingly, we found that CD patients have a slightly higher excretion of albumin compared to healthy individuals, although the level is below the 300 mg/g limit. That is likely due to intensive physical exercise declared by these patients or is a consequence, not completely explored, of CD. To assess the level we proposed for the first time a novel spectrofotometric/fluorimetric method, highly selective for albumine.

Quantitative PCR was used to assess the level of urinary miRNAs of CD patients (compared to controls) and this is another novelty as this exploration was nevev carried out before for this disease.

In urine, the dysregulated miRNAs belongs to many important biological patways related to regulatory or signaling processes in which surrounding tissues may be involved in. However, in order to completely unravel their role, a deep functional investigation (that is beyond the scope of this work) should be required in the future.

By focusing only on the two urinary miRNAs dysregulated in CD either vs. CTRL or GDF (i.e., miR-221-3p and miR-324-5p), is it possible to summarize and discuss some data. The first miRNA, miR-221-3p, has been found in the urine of rats treated with thick ascending limb (TAL) toxicants, aligning with its expression localization within the kidney²². Laser capture microdissection experiments further confirmed that miR-221-3p is preferentially enriched in TAL and/or collecting duct (CD) fractions compared to proximal tubule segments. This site-specific expression suggests its potential value in discriminating the underlying pathology of toxicant-induced lesions in the kidney or a more general kidney damage²².

In chronic kidney disease (CKD), particularly diabetic kidney disease (DKD), some studies indicated miR-221 upregulation in urine and urine exosomes²⁴, whereas another study reported that in diabetic nephropathy (DN), miR-221 is markedly up-regulated in extracellular vesicles (EVs) derived from podocytes under high-glucose conditions²⁷. These podocyte-derived EVs, containing miR-221, play a key role in mediating proximal tubular epithelial cell (PTEC) dedifferentiation²⁷. This process is facilitated by miR-221 directly targeting DKK2, an inhibitor of the Wnt signaling pathway, which subsequently leads to the nuclear accumulation of β-catenin and activation of Wnt/β-catenin signaling. Of note, inhibition of miR-221 in diabetic mice alleviates proximal tubule cell injury, reducing collagen fiber deposition in the tubule interstitium and reversing the abnormal expression of PTEC dedifferentiation-related proteins. The authors demonstrated that the direct transfer of fluorescently labeled miR-221 from podocytes to PTECs, especially under high-glucose conditions, further supports its role in mediating this cellular damage²⁷. These mechanistic insights emphasize the pathogenic potential of miR-221 in renal injury and underscore the need to explore its role beyond diabetic contexts, including autoimmune-related renal stress such as in CD.

In other renal contexts, where miRNAs were examined in association with microalbuminuria in a group of 120 heavy metals exposed adolescents, urinary miR-221 levels were marginally higher in the microalbuminuric group compared to the normoalbuminuric control group, but this difference was not statistically significant (p = 0.077) and the level of urinary albumin excretion did not correlate with urinary miR-221 levels (Spearman’s r = 0.155, p = 0.09)²⁸.

Beyond its mechanistic involvement, although not studied in urine, elevated serum hsa-miR-221 levels were reported significantly increased in DN patients, particularly those with macroalbuminuria, suggesting its strong potential as a diagnostic and prognostic biomarker for DN²⁹. Interestingly, these levels are also higher in diabetic patients without nephropathy compared to non-diabetic CKD patients and healthy controls, indicating its broader significance in diabetes pathogenesis and in particular, in contributing to macrovascular and diabetes complications by affecting specific genes (i.e., adiponectin)²⁹.

In parallel, miR-324-5p is intimately involved in renal dysfunction and sodium retention. Its expression is increased under hypertonicity in renal KC3AC1 cells and in the kidneys of furosemide-treated mice, where it contributes to a reduction in Mineralocorticoid Receptor (MR) expression³⁰. This miRNA directly targets Nr3c2 (the gene encoding MR) and Elavl1 (HuR) transcripts, which may alter MR signaling by destabilizing these critical mRNAs. Overexpression of miR-324-5p in renal KC3AC1 cells results in a decrease in MR expression at both mRNA and protein levels, consequently compromising MR-mediated target gene expression³⁰. Its co-expression with MR in aldosterone-sensitive distal nephron segments, including connecting tubules (CNT), cortical collecting duct (CCD), and outer medullary collecting duct (OMCD), as well as in other segments like glomeruli (Glo) and proximal convoluted tubules (PCT), points to a broad regulatory role in renal function. The observed increase in miR-324-5p in furosemide-treated mice, where luminal tonicity is elevated, reinforces its physiological importance in controlling MR expression in response to changes in tonicity³⁰.

Furthermore, although not studied in urine, miR-324-3p was reported as the most prominently upregulated miRNA in microdissected glomeruli in progressive nephropathy and is also highly expressed in cortical tubules³¹. This miRNA putatively targets the prolyl endopeptidase (Prep), an enzyme involved in the synthesis of antifibrotic peptides like Ac-SDKP. Its upregulation is associated with reduced Prep expression and activity, lower urine Ac-SDKP levels, and increased collagen deposition, thereby contributing significantly to renal fibrosis³¹.

Notably, both urinary miR-221-3p and miR-324-5p, which were significantly dysregulated in untreated pediatric CD patients, normalized in GFD patients. This trend suggests that these urinary miRNAs may potentially serve a dual role: reflecting subclinical renal involvement as well as dietary adherence as they respond to gluten withdrawal.

However, since miR-221-3p and miR-324-5p are known to participate in various inflammatory and metabolic pathways, we cannot exclude that their dysregulation may not be entirely kidney-specific.

Therefore, further functional studies in cellular or animal models (to investigate the mechanistic role of miR-221-3p and miR-324-5p in kidney injury or repair) could clarify whether these miRNAs actively contribute to renal pathology or recovery, or whether they merely reflect downstream effects of intestinal inflammation.

Conclusions

The coordinated dysregulation and functional impacts of both miR-221 and miR-324 highlight their multifaceted involvement in the pathogenesis of various kidney diseases, making them compelling candidates for novel diagnostic and prognostic biomarkers as well as attractive therapeutic targets for different forms of kidney injury and chronic kidney conditions^{22,23,25,27,29–31}. However, translating these findings into clinical practice faces several challenges, primarily due to the observed inconsistencies in miRNA profiling studies, heterogeneity in patient populations, the diversity of biological sample types analyzed (e.g., serum, plasma, whole blood, urine, urinary exosomes, kidney tissue biopsies), varied sample handling and processing procedures, different miRNA extraction protocols, a wide array of detection and quantification techniques (e.g., RT-qPCR, microarray, next-generation sequencing), and the lack of standardized normalization controls^23–25. Future studies are needed also to reduce the required urine volume for RNA extraction, thereby minimizing sample exclusion and improving feasibility for clinical use. Therefore, our proof-of-concept study request further large-scale clinical trials enrolling diverse patient cohorts and rigorous experimental studies are essential to validate the clinical utility of urinary miR-221 and miR-324, confirm their accuracy and generalizability, assess their organ specificity, and ultimately evaluate their translatability for human clinical detection and therapeutic application in various urinary diseases^22–25. At the end, if validated in larger longitudinal multicenter pediatric cohorts, urinary miR-221-3p and miR-324-5p could be employed not only as biomarkers to monitor renal health and adherence to a gluten-free diet but also to confirm their translational potential as innovative diagnostic tools in pediatric CD.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CD

Celiac disease

GFD

Gluten-Free Diet

miRNA

microRNA

qPCR

quantitative Polymerase Chain Reaction

CTRL

Control

TGA-IgA

Tissue transglutaminase IgA

EMA-IgA

Endomysial antibodies IgA

Author contributions

Conceptualization, A.P., C.F., S.P.B. and A.B.; methodology, A.B and A.M; software, A.P.; validation, S.S., M.R-N., C.F., S.P.B.; formal analysis, C.F., S.P.B., C.M, E.C; investigation, A.P., A.B, A.M.; resources, F.F., A.M.; data curation, S.S., M.R-N.; writing—original draft preparation, A.P., A.B, C.F., S.P.B., F.F., A.M.; writing—review and editing, S.S, M.R-N., C.M., E.C., F.F.; visualization, C.M., F.F.; supervision, A.M., A.B.; project administration, A.P., A.M., A.B.; funding acquisition, A.B, A.M. All authors have read and agreed to the published version of the manuscript.

Data availability

The data presented in this study are available on request from the corresponding author due to privacy and ethical reasons, as the dataset contains clinical data from patients.

Competing interests

The authors declare no competing interests.

Ethical approval

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of Bambino Gesù Children’s Hospital-IRCCS (protocol code Progetto PNRR-POC-2022-12375779; 15/03/2023).

Informed consent

Written informed consent was obtained from a parent and/or legal guardian of all subjects involved in the study for study participation and to publish this paper.