University of Crete (UOC)
Μελέτη και κλινική αξιολόγηση των miRNAs ως βιοδεικτών στην οξεία λεμφοβλαστική λευχαιμία της παιδικής ηλικίας
Abstract
dc:descriptionIntroductionAcute lymphoblastic leukemia (ALL) is the most common neoplasm in the pediatric population. Despite notably high cure rates, relapse occurs in approximately 10–20% of children and adolescents with ALL and is associated with a significantly worse prognosis. Contemporary risk stratification protocols primarily rely on clinical and laboratory features and measurable residual disease (MRD). Nonetheless, there remains a critical need to identify additional biomarkers that better capture disease heterogeneity and enhance the prediction of adverse outcomes – particularly relapse. MicroRNAs (miRNAs or miRs), small non-coding RNAs that primarily regulate gene expression post-transcriptionally, are implicated in leukemogenesis, lymphoid differentiation, and chemoresistance. The clinical application of these molecules as biomarkers requires rigorous standardization of pre-analytical and analytical procedures, appropriate normalization techniques, and validation through independent studies. AimThe MICRO-CALLS study (MicroRNA levels in Childhood Acute Lymphoblastic Leukemia and Survival) aimed to investigate levels of selected miRNAs in relation to prognosis in children and adolescents with ALL and constitutes the subject of this dissertation. It comprises three phases: a preliminary phase, a pre-analytical phase, and the main experimental phase (quantification of selected miRNAs, evaluation of polymorphisms in miRNA biogenesis genes, and confirmation of findings using small RNA-seq). The preliminary phase aimed to identify microRNA signatures by molecular subtype in pediatric and adolescent ALL, based on immunophenotype and in cases with high relapse risk, using extensive RNA sequencing databases. Subsequently, the predictive performance of these signatures was evaluated to select candidate miRNAs for RT-qPCR validation within the study cohort. The pre-analytical phase aimed to evaluate factors that may influence RT-qPCR–measured miRNA levels, including specimen source (bone marrow versus peripheral blood; serum versus plasma), the interval from blood collection to miRNA extraction, the blast percentage in the initial sample, and potential differences between blood collected in EDTA tubes and coagulation tubes. The main experimental phase focused on identifying and clinically assessing miRNAs as prognostic biomarkers in pediatric and adolescent ALL, with a particular focus on relapse prediction and correlations with survival endpoints. Additionally, known polymorphisms in genes involved in miRNA biogenesis were genotyped to assess their relationships with measured miRNA levels and patient outcomes. Finally, small RNA sequencing was conducted to validate and enhance the RT-qPCR results and to evaluate the reproducibility of miRNA signatures across different methodologies. Materials and MethodsThis dissertation reports the MICRO-CALLS study, conducted in three phases: (i) preliminary in silico analysis and modeling in 123 children from the St. Jude Total XVI cohort (EGAD00001007530; St. Jude Children’s Research Hospital) using the JADBio autoML platform, (ii) a pre-analytical phase to identify variables that significantly affect miRNA expression levels and may act as confounders, and (iii) the main clinical–experimental phase focused on miRNA quantification by RT-qPCR, with confirmation by small RNA-seq, and on the investigation of associations between miRNA expression and polymorphisms in miRNA biogenesis genes (DROSHA, DICER1, and AGO1). A total of 142 children and adolescents with ALL diagnosed between 2004 and 2025 were included in the study, with follow-up conducted at the Department of Pediatric Hematology-Oncology at the University Hospital of Heraklion through December 7, 2025. The immunophenotype was B-ALL in 125 patients (88%) and T-ALL in 17 patients (12%). Relapse occurred in 13 patients (9.2%). Overall survival in the cohort was 92.3%. Clinico-laboratory parameters were systematically recorded: age and white blood cell count (WBC) at diagnosis, extramedullary central nervous system (CNS) involvement, peripheral blood blasts/μL on day 8 of induction (D8), MRD on days 15 and 33 of induction (MRD D15 and D33), cytogenetic features (karyotype, FISH, whole-exome sequencing [WES]), treatment protocol, and survival endpoints (overall survival [OS], event-free survival [EFS], relapse-free survival [RFS]), as well as an additional 47 variables examined in relation to measured miRNA levels. To better characterize unfavorable risk genetics (URG), WES was performed in ten diagnostic samples not previously assigned to a molecular subtype (“B-other”). Quantification was performed in bone marrow mononuclear cells (BMMNCs) using TaqMan Advanced miRNA Assays for 10 miRNAs: miR-16-5p, miR-21-5p, miR-24-3p, miR-125b-5p, miR-128-3p, miR-155-5p, miR-181a-5p, miR-191-5p, miR-223-3p, and miR-708-5p. Extraction efficiency was monitored using the exogenous spike-in cel-miR-39-3p, and hemolysis was assessed by measuring hsa-miR-451a. Amplification acceptance criteria (Cq thresholds for targets and reference genes) and inter-plate correction were applied according to MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) guidelines. Candidate reference miRNAs for normalization were miR-16-5p, miR-24-3p, and miR-191-5p. The most stable miRNA was miR-191-5p, and the optimal normalization pair was miR-191-5p with miR-16-5p; the geometric mean of the two was used as the normalization factor. Relative expression was calculated using the Livak 2−ΔΔCq method.Statistical analyses included parametric and nonparametric tests as appropriate, correlations (Spearman and Pearson), ROC analyses to define thresholds (Youden index), logistic regression, multivariable Cox regression adjusted for established risk factors, and survival analyses. Automated machine-learning approaches (JADBio) were also used to evaluate predictive models. Results The in silico analysis identified a 14-miRNA signature that distinguished T-ALL from B-ALL with excellent accuracy (AUC 99.5%), and the combined targeting analysis highlighted apoptosis as a dominant pathway. The same analysis indicated that an 18-miRNA signature could identify patients at high risk of relapse, a 5-miRNA signature could separate patients with URG, and additional subtype-specific miRNA signatures were identified. In this preliminary phase, miR-223 expression was linked to high-risk stratification and chemoresistance, and the role of miR-181a also emerged. The pre-analytical phase showed that delayed sample processing resulted in significant differences in measured miRNA levels (extraction within 1 hour vs storage for 6 or 24 hours, p < 0.01). Storage in sodium citrate tubes was associated with higher miRNA levels than in EDTA tubes, although this did not reach statistical significance. Paired-sample analysis (bone marrow, peripheral blood, serum, plasma) showed higher expression in bone marrow, particularly in BMMNCs (p < 0.01). No difference in miRNA expression was observed between serum and plasma.RT-qPCR revealed distinct miRNA expression profiles and associations with response and outcomes. Poor prednisone response (AAB D8 ≥ 1,000/μL) was associated with upregulation of miR-21-5p and downregulation of miR-125b-5p and miR-128-3p at diagnosis. Notably, miR-125b-5p and miR-128-3p levels showed a strong inverse relationship with WBC at diagnosis, and weighting by WBC and blast percentage yielded borderline significance for all observed associations (p = 0.03–0.05). MRD was significantly associated with five miRNAs. MRD D15 ≥ 10% was associated with high miR-21-5p and miR-181a-5p and low miR-125b-5p, miR-128-3p, and miR-708-5p. Among these, downregulation of miR-128-3p showed the strongest association (p < 0.001), followed by high miR-181a-5p. Similarly, MRD D33 ≥ 10−4 was significantly associated with high miR-21-5p and low miR-125b-5p, miR-128-3p, and miR-708-5p. The strongest and independent association was observed for miR-128-3p downregulation at diagnosis with high end-of-induction MRD (AUC 86.3%, 95% CI: 68–100%, p < 0.001; sensitivity 85.7%, specificity 91.2%), followed by a non-independent association for miR-125b-5p expression. Survival endpoints were also associated with five miRNAs. OS was significantly lower in children with high miR-181a-5p and low miR-125b-5p or miR-708-5p at diagnosis. Among these, markedly lower survival was observed in patients with miR-125b-5p downregulation at diagnosis (86.8% vs 94.2%, p = 0.002). In addition, a fivefold cumulative risk of death was estimated in the group with high miR-181a-5p (p = 0.02). Regarding RFS, several differences were noted compared with OS. Low miR-223-3p at diagnosis showed the strongest association with low RFS (72.4% vs 97.4%, p = 0.002), followed by high miR-155-5p (11.5-fold relapse risk; RFS 80.5% vs 98.3%, p = 0.006) and low miR-125b-5p (84.2% vs 94.2%, p = 0.036). Moreover, miR-223-3p downregulation at diagnosis appeared able to distinguish cases that subsequently relapsed with high accuracy and independently of other high-risk features (AUC 75.1%, p < 0.001), ranking just behind WBC ≥ 50,000/μL in multivariable Cox regression—with even larger differences when focusing on B-ALL. Consistently, low EFS in the cohort was significantly associated primarily with low miR-223-3p at diagnosis and secondarily with low miR-125b-5p, corresponding to a 7-fold (p = 0.004) and 3.5-fold (p = 0.03) increased risk of relapse or death, respectively. Additional significant associations between diagnostic miRNA levels and clinico-laboratory variables are discussed, highlighting frequent interactions. The JADBio platform was used to assess the combined predictive performance of all recorded variables for relapse and survival. The most reliable variable for relapse prediction was low miR-223-3p expression (mean accuracy 70.1%, AUC 63.9%). Additionally, a signature comprising miR-125b-5p levels and URG was identified as most specific for EFS prediction, but with only moderate mean accuracy (66.5%), despite high specificity (95.2%). Longitudinal assessment of miR-223-3p in ten patients (paired samples at diagnosis, D33, end of treatment ± relapse) confirmed its “protective” effect against relapse. Similarly, small RNA-seq documented a significant association of high miR-181a-5p and miR-155-5p, and low miR-223-3p, with relapse. Genotyping of DROSHA polymorphisms in 115 patients and 137 healthy controls showed that homozygotes for the minor allele of DROSHA rs642321 were associated with disease susceptibility (TT vs CC+CT, OR 4.5, 95% CI: 1.2–21.2, padj = 0.034). Likewise, homozygotes for the minor allele of DROSHA rs3805500 were associated with increased ALL risk (GG vs AA+AG, OR 2.7, 95% CI: 1.3–6.1, padj = 0.012). Among ALL patients, the rs3805500 AG genotype was significantly associated with relapse (OR 5.8, 95% CI: 1.6–24.3, padj = 0.011) and death (OR 5.0, 95% CI: 1.1–26.3, padj = 0.038). Patients carrying the rs3805500 AG and GG genotypes had lower survival indices. Polymorphisms in miRNA biogenesis genes showed significant effects on only two miRNAs: (i) homozygotes for the minor T allele of DROSHA rs642321 had increased miR-24-3p levels and higher Drosha expression, and were associated with age >10 years at diagnosis, T-ALL, TCF3::PBX1, and absence of hyperdiploidy; (ii) homozygotes for the major T allele of DROSHA rs10035440 showed upregulation of miR-181a-5p, higher Drosha expression, and association with poorer survival indices (OS, RFS, EFS).ConclusionsThe MICRO-CALLS study demonstrated that selected miRNAs may provide clinically useful prognostic information in pediatric and adolescent ALL. A key finding is the downregulation of miR-223-3p at diagnosis, which serves as an accurate and independent marker of relapse and diminished EFS, particularly in B-cell ALL. Furthermore, elevated miR-181a-5p, miR-155-5p, and miR-21-5p, along with decreased miR-125b-5p, miR-128-3p, and miR-708-5p, were associated with prognosis across multiple parameters, including glucocorticoid response, MRD, and survival outcomes. Concurrently, polymorphisms within genes involved in the microRNA biogenesis pathway appear to correlate with clinico-laboratory features of the disease, influencing expression levels and potentially the functionality of the associated ribonucleases and Argonaute proteins. The translational application of these findings requires independent validation in larger, multicenter cohorts and further standardization of pre-analytical and analytical variables. Looking ahead, integrating miRNA biomarkers with MRD and cytogenetic data may enhance relapse risk stratification and facilitate individualized treatment approaches in pediatric and adolescent ALL.
Degree
thesis:*- Grantor dc:publisher
- University of Crete (UOC)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kyriakidis, Ioannis
- Κυριακίδης, Ιωάννης
Subjects
dc:subject × 26- MicroRNAs
- Οξεία λεμφοβλαστική λευχαιμία
- Παιδιά
- Έφηβοι
- Βιοδείκτες
- Υποτροπή
- MRD
- Ανάλυση επιβίωσης
- Πρόγνωση
- Acute lymphoblastic leukemia
- Children
- Adolescents
- Biomarkers
- Relapse
- Survival
- Prognosis
- Βασική Ιατρική
- Ιατρική και Επιστήμες Υγείας
- Μοριακή ιατρική
- Κλινική Ιατρική
- Παιδιατρική, Νεογνολογία και Υγεία του παιδιού
- Basic Medicine
- Medical and Health Sciences
- Molecular Medicine
- Clinical Medicine
- Pediatrics, Perinatology and Child Health
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/61084