A predictive serum miRNA signature impacts diffuse large B‐cell lymphoma cell viability via inhibition of EGLN1 and TXNRD1 regulators of ferroptosis

Summary

Diffuse large B‐cell lymphoma (DLBCL) is a heterogeneous disorder. Prognostic factors include genomic alterations and cell‐of‐origin (COO) subtypes, even though they cannot fully predict treatment response. MicroRNAs (miRNAs) deregulated in patient tumours and blood are promising non‐invasive biomarkers. Several circulating miRNAs were found to be correlated with progression‐free survival (PFS), independently of other prognosticators. However, miRNA signatures, rather than individual miRNAs, represent more reliable biomarkers and a better mirror of the disease. In this study, we identified circulating miRNAs differentially expressed between R‐CHOP refractory and responding subjects by small‐RNA sequencing on serum from 33 DLBCL patients. Among the identified miRNAs, the combined expression of three of them improved the predictive performance and was correlated with PFS. Two out of three miRNAs, miR‐421 and miR‐324‐5p, were also differentially expressed in tumour tissues based on treatment response. Overexpressing these miRNAs reduced cell proliferation, viability and resistance to R‐CHOP in the germinal centre B‐like COO subtype. EGLN1 and TXNRD1, regulators of oxygen metabolism and redox homeostasis, were identified as miRNA targets and the silencing or inhibition of these genes impaired cell viability and induced ferroptosis. These results support the application of a two‐miRNA signature and its targets for novel combined therapeutic interventions in DLBCL.

Diffuse large B‐cell lymphoma is a heterogeneous and aggressive pathology. MicroRNAs, deregulated in patient tumours and blood, are promising non‐invasive predictive biomarkers for this pathology. We identified the circulating miR‐200c/421/324 signature both predictive of response to R‐CHOP and correlated with PFS. MiR‐421 and miR‐324‐5p were also altered in tumour tissues based on treatment response. Their overexpression reduced cell proliferation, viability and resistance to R‐CHOP in the germinal centre B‐like cell‐of‐origin subtype. The miR‐421/324 signature elicited its biological effects by induction of ferroptosis through the modulation of EGLN1 and TXNRD1 target genes, involved in the regulation of the redox homeostasis.

INTRODUCTION

Diffuse large B‐cell lymphoma (DLBCL) is the most common non‐Hodgkin lymphoma, characterized by clinical and genetic heterogeneity, which affects both prognosis and treatment response. ¹ While most patients achieve remission with the current standard‐of‐care Rituximab, Cyclophosphamide, Doxorubicin, Vincristine and Prednisone (R‐CHOP) immunochemotherapy or with the recently introduced Pola‐R‐CHP protocol, ¹ , ² ≈35% of patients experience refractory/relapsed disease. ¹ The international prognostic index (IPI) stratifies patients into different risk groups. ³ Prognostic patient stratification further relies on the definition of the cell‐of‐origin (COO) subtype, being the activated B‐cell‐like (ABC) subtype related to a worse prognosis compared to the germinal centre B‐like (GCB) subgroup, and the identification of specific genetic alterations. ⁴ However, none of these stratification criteria may accurately predict response to first‐line therapy. Furthermore, with either classification system, a proportion of patients always fall in an unclassifiable subgroup. ⁵ Therefore, identifying better predictive biomarkers is crucial for improving survival.

MicroRNAs (miRNAs), key players in DLBCL biology, ⁶ are deregulated both in tumour tissues and patient blood, where they circulate in a stable form and can be easily detected, thus making them interesting non‐invasive biomarkers. ⁷ , ⁸ We previously identified serum miR‐22‐3p as a circulating prognostic biomarker that correlated with progression‐free survival (PFS). ⁷ , ⁹ However, it is widely recognized that single miRNA profiles provide low accuracy as cancer biomarkers, mostly due to the multifactorial nature of tumours and to the large number of targets for a single miRNA. ¹⁰ Thus, the evaluation of a multiple miRNA signature could be a more valuable and reliable approach to identify molecular markers that could mirror the complexity of the disease. ¹⁰ , ¹¹ In this study, we identified three serum miRNAs (miR‐200c‐3p, miR‐421 and miR‐324‐5p) whose combined expression showed to improve predictive performance over individual miRNAs. Interestingly, among these three miRNAs, miR‐421 and miR‐324‐5p were also differentially expressed in DLBCL tumour tissues based on treatment response. The functional analysis showed that their ectopic expression reduced cell proliferation specifically in GCB‐COO DLBCL, and not in the ABC subtype, suggesting a tumour‐suppressive role in a COO‐dependent context. Furthermore, we identified EGLN1 and TXNRD1, proteins affecting oxygen metabolism and redox homeostasis, ¹² , ¹³ as direct targets of miR‐421 and miR‐324, respectively. The inhibition of these targets impacts viability by triggering ferroptosis. Of note, inhibitors for the enzymes encoded by the target genes constitute already approved drugs for haematological or inflammatory diseases, providing hope for a possible drug repurposing approach for DLBCL treatment.

MATERIALS AND METHODS

Patient recruitment, serum collection and processing

Serum was collected at diagnosis from de novo DLBCL adult patients consecutively enrolled between 2015 and 2019 at the IRCCS‐Regina Elena National Cancer Institute (IRE). Formalin‐fixed and paraffin‐embedded (FFPE) tumour tissue was collected from 2017 to 2024 at IRE and IRCCS Policlinico Gemelli. Patients were uniformly treated with six courses of R‐CHOP every 21 days followed by two doses of rituximab. Baseline disease staging and treatment response assessment of non‐Hodgkin lymphoma were primarily performed according to the Lugano recommendations. The study was approved by the Institutional Ethics Committee of IRE (protocol number: RS 831/16) and Policlinico Gemelli (protocol number: 0003275/23). Informed written consent was obtained from all patients. Patient data and samples were treated according to the Declaration of Helsinki.

Small RNA sequencing

The small RNA‐Seq library was prepared as previously described ¹⁰ using RNA extracted from serum with the QIAseq miRNA Library Kit (Qiagen, Cat.no. 331502), following the manufacturer's instructions for serum samples. The library quality assessment was performed by the bioanalyser and the high sensitivity DNA kit (Agilent Technologies, Cat.no. 5067‐4626). The quantification of the libraries was performed with the Qubit fluorimeter and the dsDNA HS Assay Kit (ThermoFisher Scientific, Cat.no. Q32851), and the pool of libraries by quantitative polimerase chain reaction (qPCR). The final pool was diluted to 4 nM, then denatured and further diluted for sequencing following the manufacturer's instructions, sequencing 75 bp single read and 8 bp for the indices. The small RNA sequencing analysis is further detailed in the Data S1.

Functional experiments

OCI‐Ly‐19 and KARPAS‐422 GCB‐DLBCL cell lines, SU‐DHL‐2 and U2932 ABC‐DLBCL cell lines were used in this study.

Culture condition, transfection, cell proliferation, cell viability, cytotoxicity assays, western blotting, RNA isolation from cells, conditional medium and tumour tissues, analysis by quantitative real‐time PCR and luciferase assay techniques are all described in the Data S1; Tables S1 and S2.

Bioinformatic and statistical analysis

The analysis for miR‐421 and miR‐324‐5p putative target genes was performed using the TargetScan tool (https://www.targetscan.org), considering the predicted targets with a cumulative weighted content++ score < −0.05. The selected genes were further analysed using the miRDB tool (https://mirdb.org) and only those confirmed by this second analysis were considered for further evaluation (Table S1). The statistical analyses were conducted with MATLAB R2023b. The statistical significance of intergroup differences was estimated with the parametric Student's t‐test or the non‐parametric Mann–Whitney test, where appropriate. All results are shown as the mean ± standard deviation of the mean (SD) of at least three independent experiments. The statistical significance was considered to be p ≤ 0.05.

RESULTS

Global serum miRNA profiling in DLBCL patients responding versus refractory to R‐CHOP first‐line treatment

To assess whether circulating miRNAs may discriminate R‐CHOP refractory patients from those responding to therapy, the serum‐miRNome was analysed by small RNA‐Sequencing in a cohort of 33 DLBCL patients (Table 1). Five serum miRNAs (miR‐27a‐3p, miR‐143‐3p, miR‐200c‐3p, miR‐324‐5p and miR‐421) were then selected as differentially expressed among the two groups of patients based on the area under curve (AUC) from a logistic regression higher than 0.70 in predicting treatment response (log rank test ≤0.05; Figure 1A). These selection criteria resulted in two out of five miRNAs downregulated (miR‐27a‐3p and miR‐143‐3p; Log2 fold change<−0.3) and three upregulated (miR‐200c‐3p, miR‐324‐5p and miR‐421; Log2 fold change≥0.6) in serum of refractory compared to responding patients (Figure 1A). The selected miRNAs were independent of other clinical and biological features. To evaluate whether combinations of the selected miRNAs could improve the predictive potential respect to individual miRNAs, we performed a combined ROC curve analysis considering the group of the downregulated miRNAs and that of the upregulated ones. As shown in Figure 1B, this latter combination has a higher AUC indicating a better discriminatory performance. Moreover, the miR‐200c‐3p, miR‐324‐5p and miR‐421 signature is significantly correlated with patient PFS, supporting the value of this circulating miRNA signature as a non‐invasive predictive and prognostic marker (Figure 1B). Next, we evaluated the expression of miR‐200c‐3p, miR‐324‐5p and miR‐421 in DLBCL tumour tissues of refractory (n = 13) and responding (n = 13) patients (Table 1). We found that two out of three miRNAs, miR‐421 and miR‐324, were significantly downregulated in tumours of refractory patients compared to the responding ones (Figure 1C). Since these results suggested a role of miR‐421 and miR‐324 in the biology of DLBCL, we focused our attention on the two‐miRNA signature hypothesizing an oncosuppressive effect.

TABLE 1.

Biological and clinical features of DLBCL patients.

Characteristics	Cohort for serum profiling		Cohort for FFPE profiling
	Refractory patients N = 15 (%)	Responsive patients N = 18 (%)	Refractory patients N = 13 (%)	Responsive patients N = 13 (%)
Mean age (range)	70 (31–84)	65 (28–77)	59 (31–87)	60 (28–79)
Gender M/F	7/8	10/8	5/8	6/7
Ann‐Arbore stage
I–II	0	5 (27.8)	4 (30.8)	3 (23.1)
III–IV	15 (100)	13 (72.2)	9 (69.2)	10 (76.9)
IPI
Low (0–1)	0	2 (11.2)	3 (23.1)	4 (30.8)
Intermediate (2–3)	7 (46.6)	13 (72.2)	6 (46.1)	8 (61.5)
High (4–5)	8 (53.4)	3 (16.6)	4 (30.8)	1 (7.7)
LDH
Normal	5 (33.3)	10 (55.5)	2 (15.4)	2 (15.4)
High	10 (66.7)	8 (44.5)	11 (84.6)	11 (84.6)
Cell of origin
GCB	6 (40)	8 (44.4)	4 (30.8)	6 (46.1)
Non‐GCB	5 (33.3)	5 (27.8)	7 (53.8)	5 (38.5)
N.A.	4 (26.7)	5 (27.8)	2 (15.4)	2 (15.4)

Abbreviations: GCB, germinal centre B‐like subtype; IPI, international prognostic index; LDH, lactate dehydrogenase; N.A., non‐assessable; Non‐GCB, non‐germinal centre B‐like subtype.

FIGURE 1.

Genome‐wide analysis and selection of a serum miRNA‐signature deregulated in diffuse large B‐cell lymphoma (DLBCL) patients based on treatment response. (A) Experimental workflow for the miRNA profiling; boxplots of the indicated serum miRNA levels in non‐responsive (NR) versus responsive (R) patients. (B) Fold change expression (FC) and area under the ROC curve (AUC) of the selected five miRNAs. ROC curves for predictive accuracy of the combinations of the downregulated and upregulated miRNAs. Kaplan–Meier plots for PFS based on the miR‐200c/miR‐324/miR‐421 combination (log‐rank test). (C) Plots of the indicated miRNA levels in tumour tissues of NR versus R patients. *p ≤ 0.05; ***p ≤ 0.001 (Mann–Whitney test).

Impact of the two‐miRNA signature on DLBCL cell proliferation, cell viability and sensitivity to R‐CHOP

To gain more insight into the role of the two miRNAs identified in DLBCL, we assessed the consequences of their overexpression in DLBCL cell models belonging to both ABC and GCB‐COO subtypes. Results show that the ectopic expression of the signature members reduces the proliferation and viability of GCB cell lines but not of the ABC subtype (Figure 2A,B). Notably, considering cell viability, we further observed a more pronounced effect compared to the mimic negative control–transfected cells when the entire two‐miRNA signature was expressed in combination. Altogether, these data strongly suggest that the two‐miRNA signature elicits an oncosuppressive effect in DLBCL, specifically for the GCB‐COO context.

FIGURE 2.

Impact of miR‐421 and miR‐324 ectopic expression on diffuse large B‐cell lymphoma (DLBCL) proliferation and viability. (A) Cell proliferation of the indicated DLBCL cells after transfection with individual or combined miRNA‐mimics. Values are reported as fold change of the viable cells over the control. (B) Cell viability after miRNA‐mimic or control‐mimic transfection. The percentage of dead cells is expressed as fold change over the control. (C) miRNA levels in the extracellular fraction (EC) and dose–response curves of the indicated DLBCL cells after treatment with CHOP plus rituximab. EC values, normalized by the intracellular miRNA expression (IC) and by cell density, are reported as the mean of at least three experiments; bars represent the standard deviation. (D) Cytotoxicity assay of the indicated DLBCL cells after transfection with individual or combined miRNA‐mimics, and upon treatment with CHOP plus rituximab, or with the drug vehicle (DMSO). Cell death is expressed as fluorescent green area normalized by the total area of confluence; representative images of cell samples after 72h of treatment are reported; *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 (Student's t‐test).

To understand whether the inverse deregulation of the miRNA expression between refractory and responsive patients observed in tumour tissue and serum (Figure 1B,C) could be due to a higher release of these miRNAs by R‐CHOP‐resistant tumours, we evaluated their basal expression in GCB‐DLBCL cells (OCI‐Ly‐19 and KARPAS‐422) conditioned medium (extracellular fraction, EC). As shown in Figure 2C, taking into account the expression levels of the miRNA signature in the EC and normalizing by the basal intracellular miRNA expression (intracellular fraction, IC), we found a higher release rate of both miRNAs by the KARPAS‐422 cells, which concomitantly show a higher R‐CHOP resistance compared to OCI‐Ly‐19 (Figure 2C), thus suggesting a higher release of miRNAs in the EC from resistant tumours consistent with the data obtained in DLBCL patient samples. ⁷ , ¹⁴ , ¹⁵

To investigate the functional impact of the two‐miRNA signature on the response to R‐CHOP treatment, we performed a cytotoxicity assay by Incucyte live‐cell imaging system on GCB‐DLBCL cells overexpressing the miRNA signature members, alone or in combination, and treated with R‐CHOP drug mix. As shown in Figure 2D, in the KARPAS‐422 cells, which have a higher baseline resistance to treatment, the miRNA signature members, both alone and in combination, induced an increase in cell death upon treatment compared to the negative control. On the other hand, in the more drug‐sensitive cell line OCI‐Ly‐19, only the combined signature produced a sensitization to R‐CHOP, while the single miRNAs had a milder and non‐statistically significant effect in this cell line.

Identification of EGLN1 and TXNRD1 genes as novel direct targets of miR‐421 and miR‐324‐5p signature

To identify putative target genes of the miR‐421/miR‐324 signature, we performed an in silico target analysis. We sorted out 649 and 216 targets of miR‐421 and miR‐324, respectively (Figure 3A and Table S1). Among these, we first identified TXNRD1 as a common predicted target of miR‐421 and miR‐324. This gene encodes a pyridine nucleotide‐disulphide oxidoreductase, which reduces thioredoxins and plays a key role in redox homeostasis. ¹² , ¹⁶ , ¹⁷ However, a further evaluation of the mRNA modulation of this target gene upon miRNA overexpression led to the downregulation of the TXNRD1 transcript only by miR‐324 ectopic expression and not by miR‐421 overexpression. Thus, we looked for other putative targets of miR‐421 involved in molecular pathways connected to those regulated by TXNRD1, identifying EGLN1, which is also reported to regulate pathways linked to the redox state of cells. ¹³

FIGURE 3.

Identification and validation of EGLN1 and TXNRD1 as miR‐421 and miR‐324 direct targets. (A) Venn diagrams of putative targets selected by TargetScan (blue) and miRDB (green). (B) Scheme of the biological processes impacted by EGLN1 and TXNRD1. (C) EGLN1 and TXNRD1 mRNA and protein levels in diffuse large B‐cell lymphoma (DLBCL) cells after miRNA overexpression. Values are reported as fold change over the control. (D) Scheme of putative binding sites for miRNAs in the 3′UTR of their target genes and Firefly luciferase activity. Results are normalized by basal activity of luciferase empty vector and expressed as fold activation over the control. Bars indicate the standard deviation; *p ≤ 0.05, **p ≤ 0.01 (Student's t‐test).

To verify whether those predicted targets are regulated by the miRNA signature, we assessed the mRNA level of EGLN1 and TXNRD1 after overexpression of miR‐421 and miR‐324, respectively, in GCB‐DLBCL cells. We found a decrease of both transcripts upon miRNA ectopic expression in KARPAS‐422 but not in the OCI‐Ly‐19 cell line (Figure 3C). We then verified the expression of the EGLN1 and TXNRD1 proteins and observed a decreased level of these enzymes upon miRNA overexpression in both GCB‐DLBCL cells (Figure 3C). This result confirmed the regulation of EGLN1 and TXNRD1 by miR‐421 and miR‐324, respectively. Furthermore, these data suggest that, in the OCI‐Ly‐19 cells, the two‐miRNA signature regulates its targets at a post‐transcriptional level, since the protein levels were decreased without any changes in the mRNA.

To validate the direct interaction between miR‐421 and miR‐324, and their targets, a human EGLN1 and TXNRD1 3′UTR fragment with or without the target seed sequences (for EGLN1, we identified two putative binding sites termed as binding site 1 and 2) was cloned downstream of the firefly luciferase reporter gene and co‐transfected with the related miRNA mimic in HEK‐293 FT recipient cells. We found that the relative luciferase activity of the reporter with the wild‐type EGLN1 and TXNRD1 3′UTRs had a 17% and a 12% decrease upon co‐transfection with miR‐421 or miR‐324, respectively. Regarding EGLN1, we observed a modulation of luciferase activity only in the presence of binding site 1 and not with binding site 2 (Figure 3E), suggesting that only binding site 1 is targeted by miR‐421. As expected, there was no decrease in luciferase activity of the reporter containing a non‐targeted sequence, as a negative control (Figure 3E). Altogether, these findings demonstrate that the selected genes are novel direct targets of the signature.

Impact of the miRNA signature on ferroptosis programmed cell death

Since both EGLN1 and TXNRD1 are reported to regulate the redox state of the cells, ¹² , ¹³ we hypothesized that their inhibition, elicited by the miRNA signature, could induce ferroptosis, a programmed cell death process closely related to oxidative stress. ¹⁸ To evaluate this, the impact of the miRNA signature on the expression levels of GPX4, a marker of ferroptosis, was assessed. As shown in Figure 4A, we found in DLBCL cells a decrease of GPX4 mRNA and protein levels after overexpression of both miRNAs of the signature (Figure 4A). Next, to verify whether the viability decrease and the ferroptosis induction observed after miR‐421 and miR‐324 overexpression are due to the inhibition of their target genes, we silenced them by siRNA transfection. We found that EGLN1 and TXNRD1 silencing significantly increased cell mortality (Figure 4B). Concomitantly, we observed a decrease in the GPX4 marker upon EGLN1 and TXNRD1 gene silencing (Figure 4C), suggesting that both targets could recapitulate the effects of the miRNA signature overexpression on viability and ferroptosis induction. These results were further confirmed by the cell treatment with molidustat and auranofin, inhibitors of EGLN1 and TXNRD1, respectively. ¹⁹ , ²⁰ , ²¹ As shown in Figure 4D, both inhibitors determined a reduction of GPX4 at both mRNA and protein levels.

FIGURE 4.

Impact of miR‐421/miR‐324 signature and its targets on ferroptosis. (A) mRNA and protein levels with densitometric analysis of GPX4 upon overexpression of miR‐421 and miR‐324 or the negative control in diffuse large B‐cell lymphoma (DLBCL) cells. (B) EGLN1 and TXNRD1 protein levels after gene silencing in GCB‐DLBCL cells; cell viability after EGLN1 and TXNRD1 siRNA or negative control transfection. The percentage of dead cells is expressed as fold change over the control. (C) Protein levels and densitometric analysis of the GPX4 ferroptosis marker after EGLN1 and TXNRD1 silencing in the GCB‐DLBCL cells. (D) GPX4 mRNA and protein levels with densitometric analysis in GCB‐DLBCL cells treated with 10 μM molidustat or 0.4 μM auranofin versus the DMSO treated negative control. Values are reported as fold change over the control; bars represent the standard deviation. (E) Cell viability of DLBCL cells after transfection with miRNA‐mimics and upon treatment with 2 μM Ferrostatin‐1 (Fer‐1), or with the drug vehicle (DMSO); *p ≤ 0.05 (Student's t‐test).

To further confirm that the miRNA signature members promote ferroptosis, we overexpressed them in GCB‐DLBCL cells in the presence or absence of Ferrostatin‐1 (Fer‐1), an inhibitor of ferroptosis. ²² As shown in Figure 4E, treating cells with Fer‐1 partially rescued the decrease in cell viability induced by the miRNA signature members.

Altogether, these data showed that the prognostic miR‐421/miR‐324 signature has an oncosuppressive role in GCB‐DLBCL cell lines, impairing cell viability by inducing ferroptosis and silencing of EGLN1 and TXNRD1 targets.

DISCUSSION

DLBCL clinical and molecular heterogeneity has prompted further investigation into biomarkers that can accurately predict survival. ⁸ Identifying early biomarkers for refractoriness to first‐line therapy is crucial for guiding alternative treatment approaches in refractory patients. In this context, circulating miRNAs are promising non‐invasive biomarkers due to their stability and ease of detection in body fluids like blood. ⁸ , ⁹ , ¹⁰ , ¹¹ In this study, we applied a genome‐wide approach to find a specific miRNA signature correlated with response to treatment. Among the identified miRNAs, we focused on the miR‐200c‐3p/miR‐324‐5p/miR‐421 combination since it led to an improvement in the predictive performance, compared to individual miRNAs, supporting the value of a combined approach for the use of circulating miRNAs as prognostic/predictive biomarkers.

Two out of three miRNAs (miR‐421 and miR‐324) were also modulated in DLBCL tumours based on treatment response, with lower levels in refractory patients, thus suggesting an oncosuppressive role. Even if miR‐421 has primarily been studied as an oncomiR in gastric cancer, ²³ it has also been reported to play a role as a tumour suppressor in other malignancies ²⁴ , ²⁵ , while miR‐324‐5p has shown oncosuppressive effects in several tumours, ²⁶ , ²⁷ including haematological malignancies. ²⁸ , ²⁹ However, both miR‐421 and miR‐324‐5p have scarcely been studied in a DLBCL context. Our results confirmed an oncosuppressive role for the two‐miRNA signature, as it inhibited cell proliferation and viability in DLBCL cells, particularly in the GCB‐DLBCL subtype, indicating a different function of the identified miRNAs in distinct COO contexts. Notably, a contrasting trend in the levels of these miRNAs was observed between patient blood and tumour tissue. This has also been reported for other tumours, ¹⁴ , ²⁷ , ³⁰ , ³¹ , ³² including DLBCL, ⁷ and may reflect a selective secretion by tumour cells into the extracellular space. ³¹ Consistently, we found that higher miR‐421/‐324 release from DLBCL cells correlated with greater cells resistance to R‐CHOP, thus supporting the fact that the miRNA released into the biofluids (serum) of DLBCL patients may reflect the clinical outcome. In regard to the effects of miR‐421 and miR‐324 on the GCB‐DLBCL cell response to R‐CHOP treatment, we found a sensitizing effect of both miRNAs in the KARPAS‐422 cell line, which, notably, has a higher resistance to R‐CHOP treatment. Instead, in a more sensitive model such as OCI‐Ly‐19 cell line, only the combination of the two‐miRNA signature had an advantage over the R‐CHOP treatment.

To delve deeper into the molecular mechanisms regulated by the two‐miRNA signature, we looked for potential miRNA‐regulated genes and found TXNRD1 and EGLN1 as novel direct targets of miR‐324 and miR‐421 respectively. TXNRD1 codes for an NADPH‐dependent reductase involved in reductive cellular pathways and redox regulation, ¹² , ³³ and it is overexpressed and associated with poor prognosis in several cancers. ¹² , ³³ , ³⁴ EGLN1 is a member of the EGLN prolyl hydroxylases family of oxygen sensors and regulators of the hypoxia‐inducible factor (HIF1α). ³⁵ , ³⁶ EGLN isoforms have been described with both pro‐oncogenic and tumour‐suppressive functions, depending on the tumour type. ³⁵ , ³⁶ Notably, EGLN1 has been reported to regulate the redox state of cells by inhibiting lipid peroxidation in cooperation with the transcription factor c‐Myc. ¹⁶

Since redox regulation has an impact on cell survival inducing ferroptosis, ¹⁸ we investigated the impact of the miRNA signature on this programmed cell death pathway. We found that both miR‐421 and miR‐324 decrease GPX4 levels, a marker that is negatively correlated with ferroptosis. Similar results were observed inhibiting EGLN1 and TXNRD1 targets by siRNA and small molecule inhibitors, molidustat and auranofin, ¹⁹ , ²⁰ confirming their involvement in miRNA‐induced ferroptosis. Lastly, the treatment with Fer‐1 partially reduced the viability decrease produced by miR‐421 and miR‐324 in DLBCL. On the one hand, this result further corroborated the induction of ferroptosis by the two‐miRNA signature. On the other hand, this partial rescue may suggest that other molecular mechanisms might contribute to the signature‐associated effects in DLBCL cells. Dysregulated ferroptosis is linked to the development and recurrence of several tumours, including DLBCL, ³⁷ and it is also being explored for novel therapeutic options. ³⁸ Interestingly, DLBCL cells from the GCB subtype appear more sensitive to ferroptosis‐inducing stimuli, potentially due to their high ALOX5 and low GPX4 levels, positive and negative regulators of ferroptosis, respectively. ³⁷ , ³⁹ This could explain the differential effects of the miRNA signature across COO subtypes.

Altogether, the results of this study support the value of the miR‐200c‐3p, miR‐324‐5p and miR‐421 as non‐invasive biomarkers of response to first‐line R‐CHOP, with possible applications in treatment‐guiding predictive models. Furthermore, the oncosuppressive role of the miR‐421/miR‐324 signature in the GCB‐DLBCL subtype suggests its potential as a therapeutic target through its impact on ferroptosis. The translation of these data into a clinical setting may pave the way for miRNA‐based complementary therapies in DLBCL patients.

AUTHOR CONTRIBUTIONS

Conceptualization, M.G.R. and G.R.; software, A.S.; formal analysis, A.S.; investigation, G.V., A.B.D.M., M.D.G. and F.G.; resources, V.L., F.M., F.P., M.N., S.H. and F.B.; data curation, M.T. and E.P.; writing—original draft preparation, G.R.; writing—review and editing, M.G.R.; supervision, M.G.R. and A.M.; funding acquisition, M.G.R. All authors have read and agreed to the published version of the manuscript.

FUNDING INFORMATION

This study was supported by the Italian Ministry of Health PNRR M6/C2/Investment 2.1 “Rafforzamento e potenziamento della ricerca biomedica del SSN” 2022 (grant No. PNRR‐MAD‐2022‐12376707, CUP: H83C22000800001) funded by the European Union‐ NextGenerationEU to M.G.R.; Associazione Mia Neri Foundation supported by ‘Cassa Sovvenzioni e Risparmio Dipendenti Banca d'Italia’ to M.G.R.

CONFLICT OF INTEREST STATEMENT

No conflicts of interest declared.

ETHICS STATEMENT

The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of the IRCCS‐Regina Elena National Cancer Institute (protocol code RS 831/16 approved in date 12/07/2016) and of the Policlinico Gemelli (protocol number: 0003275/23 approved in date 26/01/2023).

PATIENT CONSENT STATEMENT

Informed consent was obtained from all subjects involved in the study. Patient data are anonymized.

Supporting information

Data S1.

Table S1.

Table S2.

Regazzo G, Vari G, Marchesi F, Díaz Méndez AB, Di Giuliani M, Sacconi A, et al. A predictive serum miRNA signature impacts diffuse large B‐cell lymphoma cell viability via inhibition of EGLN1 and TXNRD1 regulators of ferroptosis. Br J Haematol. 2025;207(3):802–812. 10.1111/bjh.70017

DATA AVAILABILITY STATEMENT

The data underlying all findings of this study are publicly available at https://gbox.garr.it.