Identification of differentially expressed PD effluent miRNAs between EPS and non-EPS in PD patients.
We aimed to develop a noninvasive diagnostic tool for EPS detection in PD patients. In total, 127 PD patients were enrolled in the study (Fig. 1). In the screening set, miRNA expression levels of PD effluent were profiled quantitatively by high-throughput real-time PCR arrays, which included 377 miRNA assays from 28 PD effluent samples (8 EPS and 20 non-PES). There is no suitable internal control of miRNA or noncoding RNA to normalize miRNA expression levels in PD effluent. The ratio of two miRNA expression levels from the same sample can be calculated to eliminate the normalization issue of extracellular miRNA expression. The values of the ratio of two miRNA expression levels between the EPS and non-EPS samples were analyzed by the fold change and the Student’s t-test (Fig. 2). Five ratios of miRNAs (miR-483-5p/miR-597, miR-422a/miR-518e, miR-202-3p/miR-597, miR-155-5p/miR-17-5p, and miR-597/miR-100-5p) were present with a high statistical significance and fold change between the EPS and non-EPS groups (Fig. 2). Furthermore, eight candidate miRNAs among the five ratios of miRNAs (miR-17-5p, miR-100-5p, miR-155-5p, miR-202-3p, miR-422a, miR-483-5p, miR-518e, and miR-597) were selected for further verification and the establishment of the prediction model.
We examined the expression of eight candidate miRNAs from the screening set by single qRT-PCR in the validation set, which included 127 PD effluent samples (56 EPS and 71 non-EPS). The expression levels of seven miRNAs among eight candidate miRNAs were significantly decreased in the EPS group (p value: miR-17-5p: 0.0049, miR-100-5p: 6.1E-09, miR-155-5p: 2.4E-10, miR-202-3p: 2.5E-06, miR-422a: 7.6E-09, miR-483-5p: 0.00066 and, miR-597: 0.0009) (Fig. 3). The area under the receiver operating characteristic curve (AUC) is the most commonly used performance measure to indicate the discriminative ability. To develop a miRNA signature-based model for assessing the risk of EPS, we carried out a ROC analysis. All miRNA combination ratios were calculated by randomly selecting two miRNAs expressed among seven candidate miRNAs in the PD effluent. We selected the top five AUC values: miR-422a/miR-17-5p, miR-202-3p/miR-483-5p, miR-422a/miR-483-5p, miR-202-3p/miR-155-5p, and miR-202-3p/miR-17-5p, with AUC values of 0.7115, 0.7438, 0.7310, 0.6962, and 0.707, respectively (Fig. 4). Furthermore, these five ratios of miRNAs contained only five miRNAs, and the expression levels of these top five miRNAs were also significantly decreased in the EPS group (Fig. 5).
The development of prediction model to identify EPS in PD patients
Next, we utilized multiple logistic regression calculation formulas by combining five ratios of miRNA expression levels to establish a proper model to estimate the EPS in PD patients. The predicted probability of EPS was calculated by: Logit P = -3.215 + (1.499 * miR-422a/miR-17-5p) + (1.415 *miR-202-3p/miR-483-5p) + (1.428 * miR-422a/miR-483-5p) + (1.521 * miR-202-3p/miR-155-5p) + (0.349 * miR-202-3p/miR-17-5p). As a result, the AUC value of the combined five ratios of miRNAs expressions from 127 effluents increased to 0.8929 with a sensitivity of 91.1% and a specificity of 69% with a cutoff value of >-0.8585 (95% CI = 0.8364 to 0.9476), compared with the use of each ratio of miRNAs alone (Fig. 6A). The signature score of the combined five ratios of miRNA expressions in EPS is significantly higher than that in non-EPS of PD patients (Fig. 6B).
Moreover, seven PD patients developed EPS during the interval of 0.5-4 years in our study, and their signature scores from non-EPS and EPS effluents were compared (Fig. 6C). Despite uncertain symptoms, our prediction model accurately predict thirteen results from fourteen samples. Nausea, vomiting and abdominal pain developed in patient 1. Esophagogastroduodenoscopy revealed duodenal ulcers with duodenitis and a partial gastric outlet obstruction when the measured score was − 3.215. Her symptoms improved after omeprazole therapy. Severe abdominal pain with rebounding pain developed one year later, and free air, bowel tethering, fluid loculation, and local bowel dilatation were confirmed by computed tomography. The intraoperative findings were that there was gangrene of the intestines, which was encapsulated by a leathery peritoneum and a dirty ascites. Patient 2 developed bacterial peritonitis in Pseudomonas aeruginosa infection, and the patient developed abdominal distension and intermittent pain. The diagnosis of EPS was suspected from our measured score and then was confirmed after noting a thickened peritoneum on abdominal computed tomography. Although the signature score in patient 3 without EPS was incorrect, the trend of the signature scores from low to high was noticed. The change in signature scores may indicate the development of EPS.
The optimization of EPS prediction model with the clinical characteristics of patients
As shown in Table 1, there were 85 PD patients (34 EPS and 51 non-EPS) with clinical data. EPS patients tended to sustain more episodes of peritonitis (p < 0.001) as well as a longer PD duration (p = 0.004). Moreover, C-reactive protein (CRP) levels were significantly higher (p = 0.027) in the EPS group, which may implicate the potential inflammation of EPS by high glucose dialysate and presence of infection. The etiology of ESRD and comorbidities were similar between PD patients with EPS and non-EPS patients. Similar results showed an increased incidence of EPS with a longer duration of PD and active inflammation with a higher CRP value 16,22,23. The predictive value of the clinical characteristics with PD duration (cutoff value: > 10.5 years) and CRP levels (cutoff value: > 2.24 mg/dl) was estimated to detect EPS, and the AUC was 0.848 (Fig. 7A and 7B). ROC analysis with the five miRNA ratios was shown to distinguish non-EPS and EPS of PD patients with an AUC of 0.9426. For optimization, we combined the values of the five ratios of miRNAs with information on the duration of PD and the CRP value, which was calculated by: Logit P = -8.014 + (3.960 * PD duration) + (1.388 * CRP) + (1.989 * miR-422a/miR-17-5p) + (2.710 * miR-202-3p/miR-483-5p) + (2.081 * miR-422a/miR-483-5p) + (2.354 * miR-202-3p/miR-155-5p) + (0.357 * miR-202-3p/miR-17-5p). The accuracy of the detection of EPS was further improved, with an AUC of 0.9931, a sensitivity of 100% and a specificity of 94.1% (95% CI = 0.9819 to 1) (Fig. 7).
Table 1
Clinical characteristics of EPS and control patients
Factor | Non-EPS | EPS | P value |
n | 51 | 34 | |
Male: Female | 16:35 | 13:21 | 0.51 |
Age (year) | 58 ± 11.3 | 52.8 ± 14.2 | 0.06 |
Weight (kg) | 54.4 ± 8.3 | 51.9 ± 10.5 | 0.24 |
Height (cm) | 156.6 ± 6.7 | 156.3 ± 10.37 | 0.87 |
PD duration (year) | 9.4 ± 3.5 | 12.8 ± 5 | 0.004 |
Peritonitis (episodes) | 0.10 ± 0.30 | 0.45 ± 0.56 | < 0.001 |
PET | | | |
D/P creatinine | | | < 0.001 |
| H | 17.6% | 66.7% | |
| HA | 52.9% | 16.7% | |
| LA | 29.4% | 13.3% | |
| L | 0% | 3.3% | |
D/P glucose | | | < 0.001 |
| H | 9.8% | 33.3% | |
| HA | 51% | 33.3% | |
| LA | 39.2% | 16.7% | |
| L | 0% | 16.7% | |
Hemoglobin (g/dl) | 10.2 ± 1.4 | 10.1 ± 2.0 | 0.79 |
Albumin (g/dl) | 3.6 ± 0.3 | 3.5 ± 0.5 | 0.96 |
Calcium (mg/dl) | 9.7 ± 0.9 | 9.4 ± 0.8 | 0.12 |
Phosphorus (mg/dl) | 5.2 ± 1.2 | 4.6 ± 1.2 | 0.066 |
CRP (mg/dl) | 0.7 ± 0.8 | 25.2 ± 10.8 | 0.027 |
iPTH (pg/ml) | 517.3 ± 478.9 | 684.9 ± 710.6 | 0.2 |
EPS: encapsulated peritoneal sclerosis, PD: peritoneal dialysis, PET: peritoneal equilibration test, D/P: The ratio of solute concentrations in dialysate and plasma, CRP: C-reactive protein, iPTH: intact PTH, H, HA, LA, and L: high, high average, low average and low level of peritoneal transport status. |
High transport of peritoneal equilibration test with fast waste removal and limited ultrafiltration was found in EPS patients because of the pathologic changes in the peritoneum, such as angiogenesis and fibrosis (Table 1). Our score distribution was significantly positively correlated with the transport rate of creatinine, indicating these scores were linked to the functional impairment of peritoneal membrane. (Fig. 8).