Abstract
B-cell lymphomas represent a diverse group of neoplasms classified primarily by histopatholgy and are often challenging to accurately diagnose. Despite having been recognized less than 20 years ago, microRNAs (miRNAs) have emerged as one of the most promising class of cancer molecular biomarkers and are particularly attractive as they can be readily detected in formalin-fixed paraffin-embedded biopsy material and biological fluids such as blood. Many of the identified B-cell lymphoma miRNA biomarkers also play crucial regulatory roles in normal B-cell development. Below we consider the identity, function, and biomarker potential of miRNAs in B-cell lymphoma and most importantly the barriers that remain to be overcome if they are really to become part of routine clinical practice.
Introduction
The first discovery of what we now know as microRNAs (miRNAs) came in 1993 from the laboratories of Victor Ambros in Dartmouth College and Gary Ruvkun in Harvard. They simultaneously published a description of lin-4, a previously identified locus in Caenorhabditis elegans involved in developmental timing, that appeared to have a direct function without encoding for a protein.1,2 Things went quiet for the next 7 years, until the Ruvkun lab identified, let-7a, a second sequence from C elegans, with similar properties to lin-4. 3 Unlike lin-4, however, the sequence of let-7 was found to be highly conserved in eukaryotic genomes and it was realized that many similar sequences were present in the genomes of higher species. The first use of the term miRNA was made in 2001 by Lee and Ambros in a publication where they identified a further 15 C elegans miRNAs. 4 Since that time, there have been more than 25 000 miRNAs identified in over 200 different species (http://www.mirbase.org), including more than 2500 human miRNAs.5,6
MicroRNAs are short non-coding (nc)RNAs of 18 to 24 nucleotides in length that bind to regions of complementarity generally located in the 3ʹ-UTR (untranslated region) of target genes. They primarily act as inhibitor molecules causing post-transcriptional inhibition or degradation, although in some instances, they may also act as gene activators. 7 It is estimated that two-thirds of human genes are directly regulated by miRNAs, 8 and as a consequence, miRNAs are involved in most, if not all, cellular processes under physiological conditions. Moreover, dysfunctional expression of miRNAs appears to be a hallmark of all cancer types,9,10 including B-cell lymphomas that are the focus of this review.
Lymphoma is a cancer of the lymphatic system arising from B cells or T cells that represents the fifth most common cancer type worldwide, affecting more than a million people. Lymphomas are a heterogeneous group of cancers that vary in presentation, prognosis, and pathogenesis. In the latest version of World Health Organization (WHO) classification, there were more than 100 different lymphoma types listed, most of which were B-cell lymphomas, but which can have very different clinical characteristics and treatment regimens. 11 As a consequence, correct classification of a given lymphoma is often challenging, and therefore there is a clear clinical need for better biomarkers for these diseases. MicroRNAs are particularly attractive candidates as biomarkers, as their expression can classify different tumours according to their diagnosis, subtype, and stage more accurately than messenger RNA expression profiles. 12 Moreover, due to their intrinsic stability, they can be reliably detected in routinely prepared formalin-fixed paraffin-embedded (FFPE) tissue. This stability also means they are readily detected in biological fluids such as blood, which has led to a great deal of interest in the use of miRNAs as biomarkers in liquid biopsies discussed below.
MiRNAs as lymphoma liquid biopsy biomarkers
Currently, the gold standard of B-cell lymphoma diagnosis depends on the histopathologic examination of surgically excised biopsy material. This procedure, however, is expensive, invasive, uncomfortable, and can be risky for patients. Therefore, there has been a great interest in the development of non-invasive cancer biomarkers, also known as liquid biopsies. MicroRNAs hold a great promise in this area, as not only can they be extracted from frozen and paraffin-embedded tissue but also from many different body fluids including blood,13,14 urine, 15 saliva,16,17 sputum,18,19 amniotic fluid, and even from tears. 20
Most of the attention has been focused circulating miRNAs in blood, either in whole plasma or within circulating extracellular vesicles such as exosomes.21,22 The first report of miRNAs in the blood of B-cell lymphomas, or indeed any cancer, came in 2007. 23 We found that levels of miR-21, miR-155, and miR-210 in the serum samples of patients with diffuse large B-cell lymphoma (DLBCL) compared with healthy controls were higher suggesting their usefulness as biomarkers. 24 Since this time, there have been many follow-up studies in blood of patients with lymphoma as described below and in Table 1.
List of major miRNAs identified as biomarkers in B-cell malignancies.
Abbreviations: BL, Burkitt lymphoma; CLL, chronic lymphocytic leukaemia; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; HL, Hodgkin lymphoma; miRNA, microRNA; MALT, mucosa-associated lymphoid tissue; MCL, mantle cell lymphoma; PBMCs, peripheral blood mononuclear cells; SMZL, splenic marginal zone lymphoma.
*the minor strand of the mature form of the miRNA
Aberrant Expression of miRNAs in B-cell Lymphoma
Many of the miRNAs that have been identified as lymphoma biomarkers (Figure 1 and Table 1) also play key roles in normal B-cell lymphopoiesis. Frequently, these aberrantly expressed biomarker miRNAs also appear to be key drivers of lymphomagenesis.100,101 For example, miR-155 controls germinal centre (GC) development by controlling immunoglobulin production, after activation of the B-cell receptor (BCR), and is a requirement for high-affinity antibody formation.102,103 However, when overexpressed in a transgenic mouse model, the mice developed a high-grade lymphoma similar to DLBCL. 104 In a similar manner, the miR-17~92 controls pro–B-cell to pre–B-cell development via targeting of the proapoptotic protein BIM, 105 but when overexpressed in a murine MYC model, increased the aggressiveness of B-cell lymphomas.106,107 MiR-21 that targets tumour suppressor molecules including PTEN and PDCD4,108,109 when overexpressed in mice resulted in formation of B-cell lymphomas. 110 MiR-34a controls the transition of pro- to pre-B cell in haematopoietic stem cells via FOXP1 and SIRT1 targeting,111,112 and overexpression of this miRNA in mice abrogated lymphoma formation in a xenotransplant model.
Schematic diagram of the major B-cell lymphoma miRNA biomarkers that have been identified and their relationship to B-cell development. BL indicates Burkitt lymphoma; CLL, chronic lymphocytic leukaemia; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; HL, Hodgkin lymphoma; miRNA, microRNA; MALT, mucosa-associated lymphoid tissue; MCL, mantle cell lymphoma; PBMCs, peripheral blood mononuclear cells; SMZL, splenic marginal zone lymphoma.
In addition to the miRNAs mentioned above, miR-181 has long been recognized as a key regulator of GC B-cell differentiation,113,114 along with miR-150 that inhibits MYB downregulation. 115 The GC B cells are characterized by expression of markers BCL6, CD10, HGAL, and LMO2, as well as the absence of activated B-cell markers such as IRF4, PRDM1/BLIMP1, and XBP1. These transcription factors are also regulated at the level of miRNAs. For example, BCL6 is regulated by miR-30 family, miR-9 and let-7a, 116 whereas miR-155 regulates expression of HGAL and CD10 protein expression,117,118 and miR-223 regulates expression of LMO2. 119 In contrast, miR-125b and miR-155 regulate expression of the activated B-cell markers, IRF4 and PRDM1.116,120
The cause of aberrant miRNA expression in lymphoma (and other cancers) can result from many genomic events, such as chromosomal aberrations, epigenetic modifications, mutations in the sequence of miRNAs or their promoter regions, or factors that regulate synthesis or function of miRNAs (for further details see the work by Croce 121 ). Below, we discuss the aberrantly expressed miRNAs in different B-cell lymphoproliferative diseases that could facilitate the diagnosis, prognosis, and prediction of treatment response.
Chronic lymphocytic leukaemia
Chronic lymphocytic leukaemia (CLL) is the most common haematologic malignancy worldwide 122 and was the first haematologic malignancy, or indeed any cancer to be associated with aberrant miRNA expression when in 2002, George Calin and colleagues reported that the frequently (55%) deleted locus, 13q14, encodes for the miR-15a/16-1 cluster, and that these miRNAs were downregulated in most of the patients with 13q(del) CLL. 33 These miRNAs act as tumour suppressors in CLL through targeting of the anti-apoptotic BCL2 protein 123 and the tumour suppressor TP53. 124 In contrast, miR-7-5p, miR-182-5p, and miR-320c/d are regulated by p53 in CLL. 34 Epigenetic silencing of the miR-15a/16-1 cluster is observed in 30% to 35% of patients with CLL, a feature mediated through HDAC1-3 overexpression, 125 suggesting that these patients might benefit from HDAC-inhibitor–based therapies. However, murine models of the 13q14 deletion suggest that other factors also contribute to the aggressiveness of the disease. 126 Furthermore, the closely related miR-15b/16-2 cluster also appears to modulate genes involved in proliferation and anti-apoptotic pathways. 127
Similar to miR-15a/16-1, miR-181b is also typically downregulated in CLL, and low expression of this miRNA has been related to poor prognostic outcome. 39 Consistent with this phenotype, levels of miR-181b correlate with treatment-free survival in CLL. 40
In contrast, miR-155 is overexpressed in CLL but was found to be lower in patients who responded to therapy compared with refractory patients, 47 suggesting its usefulness as a predictive biomarker for CLL. MiR-29 is also overexpressed in both indolent and aggressive CLL, when compared with normal counterpart, but its expression was found to be lower in aggressive CLL. 35 When miR-29 was overexpressed in murine B cells, the animals developed an indolent-type form of CLL. 128
MicroRNA expression profiling has been used to distinguish between aggressive and indolent CLLs, with high levels of miR-21 and miR-155 being associated with a higher mortality rate.40,41 In contrast, upregulation of miR-708 has been associated with a favourable prognostic outcome for patients with CLL that was shown to be linked to a reduction in the nuclear factor κB signalling pathway. 42 The proliferation status of a subset of peripheral blood cells–unmutated patients with CLL was linked with miR-22 overexpression via inhibition of PTEN and PI3K/AKT activation. 129
Recently, it has been described that low levels of miR-150 in tumour cells or alternatively high levels of this miRNA in (circulating) serum are related to poor prognosis in CLL. 43 In another study, levels of both miR-150 and miR-155 in the blood were associated with the prognostic outcome of CLL. 44 Moreover, high levels of miR-155 in extracellular vesicles derived from the serum samples of patients with CLL were found compared with healthy controls. 36 Filip et al 37 found that the serum of patients with CLL had higher levels of miR-34a, miR-31, miR-155, miR-150, miR-15a, and miR-29a than controls. Another study showed that levels of miR-192 in peripheral blood mononuclear cells (PBMCs) are downregulated in patients with CLL compared with controls, suggesting that this miRNA could be a diagnostic biomarker for early stage of CLL. 38 In CLL, proliferation centres, considered to drive the disease and play a role in progression of disease, had high levels of miR-155 and miR-92 and low levels of miR-150. 130
Hodgkin lymphoma
Hodgkin lymphoma (HL), first described in 1832 by Thomas Hodgkin, 131 is one of the most frequent lymphomas, accounting for 1% of total cancers worldwide. The defining characteristic of HL is that neoplastic cells typically account for less than 1% of the tumour mass. 132 Tumour cells in classical HL (cHL), known as Hodgkin and Reed-Sternberg (HRS) cells, lack functional BCR expression or typical B-cell markers and instead express CD15 and CD30 cell surface markers.133,134 Anke van den Berg’s lab was the first to identify miRNAs in HL, when they observed in 2003 that the non-coding BIC locus, subsequently found to encode for miR-155, was overexpressed in HL cell lines.25,26 Since this time, miR-155 has been shown to target several genes in HL cells including DET1 and NIAM, among others. 135
Apart from this miRNA, several others have been implicated in HL including miR-135a which was the first miRNA to be associated with survival in HL. 31 The patients with HL with low levels of miR-135a had shorter disease-free survival than those with high levels of this miRNA. JAK2 is directly targeted by miR-135a, and the overexpression of this miRNA increases apoptotic levels and decreases cell growth via Bcl-xL inhibition. 31 In addition, let-7 and miR-9 inhibition has been shown to block plasma cell differentiation, by decreasing levels of PRDM1/BLIMP1, as well as targeting Dicer and HuR. 136 In a complementary study, inhibition of miR-9 was observed to hamper cytokine production and consequent inflammatory cell attraction in HL cell lines. 137
A 25-miRNA signature that could differentiate between cHL and reactive lymph nodes was identified by Navarro et al 28 using chromogenic in situ hybridization. Gibcus et al 27 compared the expression of miRNAs between different HL cell lines and other B-cell lymphoma cell lines and described a 23-miRNA signature for HL, which included the overexpression of miR-17~92 cluster, miR-16, miR-21, miR-24, and miR-155 along with the downregulation of miR-150. Using microarrays, another group identified 134 differentially expressed miRNAs in HL cell lines and an overlapping signature of 100 miRNAs differentially expressed in tumour samples. 29 Moreover, they observed that the levels of miR-21, miR-30e, miR-30d, and miR-92b could differentiate patients with HL according to prognostic risk groups. Epigenetic modifications of miRNA sequences have also been associated with HL including hypermethylation of miR-124a which was associated with more aggressive HL, 32 and miR-9-2 methylation which is a common feature of this disease. 30 Navarro et al 138 recently observed that miR-34a and miR-203 are frequently methylated in HL cells. It has been recently found that the alteration of miRNAs related to the regulation of antioxidant enzymes is associated with an aggressive outcome of the disease. 139 In plasma, the levels of miR-494, miR-1973, and miR-21 were higher in patients with HL than controls, 21 and in another study, levels of miR-24, miR-127, miR-21, miR-155, and let-7a were higher in purified plasma exosomes from patients with HL than disease controls. 22
Diffuse large B-cell lymphoma
Diffuse large B-cell lymphoma is the most common B-cell lymphoma in Western countries, accounting for around 20% to 30% of cases. 11 Thanks to the routine implementation of R-CHOP therapy, the survival of patients with DLBCL has been greatly improved; however, a third of patients still relapse or have a refractory disease. 140 Diffuse large B-cell lymphoma is a heterogeneous disease both at the clinical and molecular level, with the existence of at least 2 different molecular subtypes: GC B-cell like (GC-DLBCL) and activated B-cell like (ABC-DLBCL). 141 These subtypes are also distinguishable at the miRNA profile level with ABC-type lymphoma being associated with high expression of miR-21, miR-146a, miR-155, miR-221, and miR-363, and GCB-type DLBCL with high expression of miR-421 and the miR-17~92 cluster.49-53,142 It has been described that miRNAs can predict differences between DLBCL and follicular lymphoma (FL)49,50 or DLBCL and Burkitt lymphoma (BL).54,75 Central nervous system (CNS) relapse is a complication of DLBCL that occurs in approximately 5% of patients, associated with low survival, miR-20a and miR-30d are correlated with CNS relapse in patients with DLBCL and therefore could be used for patient stratification. 58
As noted above, overexpression of miR-155 in mice is enough to cause development of a high-grade lymphoma, similar to DLBCL. 143 Indeed, when the same authors used an inducible expression system, removal of the miR-155 stimulus was sufficient to allow complete recovery of affected mice. 104 MiR-155 has also been linked with metastasis and prognosis in patients with DLBCL. 59 Apart from miR-155 overexpression, low expression of both miR-34a and miR-27b expression has also been linked with a worse prognostic outcome for patients with DLBCL.61,62 In addition, low levels of miR-21 have been linked with shorter relapse-free survival in both tumour tissue 50 and in serum from patients.24,66 As a consequence, levels of this miRNA have been proposed to act as an independent prognostic factor in DLBCL. 64 It has been suggested that miR-21 may contribute to increase viability and reduce apoptotic levels of tumour cells through targeting BCL2 and PTEN.144,145 Furthermore, miR-21 inhibition leads to an increase in the sensitivity of DLBCL cell lines to CHOP treatment and reduces tumour cell proliferation and invasion.63,65
Several studies have looked at the association between miRNA expression and prognostic outcome in R-CHOP-treated patients with DLBCL. Our study found that levels of miR-27a, miR-142, miR-199b, miR-222, miR-302, miR-330, miR-425, and miR-519 were linked with overall survival. 50 More recently, miR-125b and miR-130a were associated with resistance to R-CHOP in DLBCL, 67 and high expression of miR-155 has also been linked to treatment failure. 54 In vitro, overexpression of miR-199a and miR-497 resulted in increased sensitivity to rituximab, vincristine, and doxorubicin, drugs present in R-CHOP regimen. 68 Overexpression of miR-370-3p, miR-381-3p, and miR-409-3p also increased sensitivity to rituximab and doxorubicin. 69
Outside of the tumour itself, we observed that levels of miR-21, miR-155, and miR-210 in the serum samples of patients with DLBCL were differentially expressed when compared with serum samples from healthy controls. 24 Subsequent studies using plasma also observed increased levels of miR-124 and miR-532-5p along with decreased levels of miR-122, miR-128, miR-141, miR-145, miR-197, miR-345, miR-424, and miR-425. 55 Fang et al 146 found that miR-15a, miR-16, miR-29c, and miR-155 were upregulated and miR-34a was downregulated in the serum samples of patients with DLBCL, and more recently Yuan et al 67 found a good correlation between circulating levels of 8 miRNAs and their matched FFPE samples. High expression of serum miR-22 was associated with poor prognostic outcome. 57 Recently, next-generation sequencing (NGS) technology was used to identify 51 miRNAs that were differentially expressed in the serum samples of patients with DLBCL compared with control serum samples. 56 Three of these were validated by quantitative reverse transcription-polymerase chain reaction in a validation cohort. MiR-34a-5p was upregulated, whereas miR-323-3p and miR-431-5p were downregulated.
Follicular lymphoma
Follicular lymphoma is the most common indolent B-cell lymphoma worldwide, and despite being essentially incurable, it has a median overall survival of ~20 years. However, nearly a third of patients with FL will suffer histologic transformation into a high-grade lymphoma often termed transformed FL (tFL), that is morphologically indistinguishable from DLBCL, with a much worse prognosis than the antecedent FL.147,148 We identified a signature of 6 miRNAs (miR-223, miR-217, miR-222, miR-221, and let-7i and let-7b) that could distinguish between de novo DLBCL and tFL. 50 Subsequently, miR-31 and miR-17-5p have also been identified as being differentially expressed between FL and tFL. 70
The t(14;18) translocation resulting in the constitutive expression of the anti-apoptotic BCL2 protein is the genetic hallmark of more than 90% of FL cases. 149 Using microarrays, a signature of 17 miRNAs was identified when comparing t(14; 18)-positive and t(14; 18)-negative FL cases. Downregulation of miR-16, miR-26a, miR-101, miR-29c, and miR-138 was associated with changes in the expression of target genes related to cell cycle control, apoptosis, and B-cell differentiation. 71 It has been demonstrated that miRNA expression differs between pathogenic and non-neoplastic tissue, such as miR-9 and miR-155. 49 Another study found a subset of 44 miRNAs which discriminates between FL and follicular hyperplasia, and the same study also described a 23-miRNA signature that was associated with an improved response to chemotherapy. 72 Moreover, miR-494 was found overexpressed in FL compared with a potentially confounding diagnosis of nodal marginal zone lymphoma. 73
Finally, one study analysed bone marrow smears from patients with FL and showed that 39 miRNA were decreased and 27 miRNA were increased significantly; among these, miR-451 showed the greatest decrease and miR-338-5p the greatest increase in patients with FL. 74
Burkitt lymphoma
Burkitt lymphoma most commonly affects children and adolescents and is a highly aggressive lymphoma with a very poor prognosis that often involves extra-nodal sites. Burkitt lymphoma is characterized by overexpression of the MYC oncogene and is associated with the t(8:14) translocation in most of the cases (>90%). 11 However, there are few cases that lack the t(8:14) translocation but have MYC overexpressed. 76 The authors suggest that miR-34b could be responsible for MYC overexpression in these cases. 76 In further studies, additional miRNAs have been identified as being differentially expressed between t(8:14)-positive and t(8:14)-negative cases by downregulation of miR-29 family members,79,80 miR-9 81 and miR-34b, 76 and upregulation of miR-513a-5p and miR-628-3p.77,80 Furthermore, levels of MYC-regulated miRNAs, such as the let-7 family, miR-155, miR-146a, miR-29, and the miR-17~92 cluster, can distinguish BL from other B-cell lymphoma types.75,81-83,150 Recently, NGS was used to identify 49 differentially expressed miRNAs between BL cases and normal GC B cells, many of which can target MYC. 84 Furthermore, miR-181b was found downregulated in BL cases, and the authors propose that it may function as a tumour suppressor. 85 In an earlier study, significantly lower expression of miR-155, miR-21, and miR-26a was observed between classical BL and cases with intermediate features between BL and DLBCL (DLBCL/BL). 78
Most of the endemic BL cases (>90%) are associated with Epstein-Barr virus (EBV) infection11,151 that has been shown to regulate several miRNAs, including miR-21, miR-146a, miR-155, miR-10a, and miR-127 in BL cases.152–155 In addition, EBV itself encodes for miRNAs that can interfere and compete with endogenous expression of miRNAs.156,157 In paediatric BL levels of cplasma, miR-21 and miR-23a were associated with both diagnosis and prognosis. 158
Mantle cell lymphoma
Mantle cell lymphoma (MCL) accounts for 5% to 10% of non-Hodgkin lymphomas 159 and has the worst prognosis of any B-cell lymphoma.160,161 Nearly all MCL (>90%) cases contain the t(11:14) translocation leading to overexpression of cyclin D1 (CCND1).162,163 It has been demonstrated that miR-15/16 and miR-17~92 are involved in CCND1 deregulation.86,87 The former miRNA (miR-15b) additionally involved in the transformation of classical to aggressive MCL. 89 A miRNA signature of 95 miRNAs was identified that could differentiate between differing clinical subtypes of MCL.88,90 Low miR-29 together with high miR-20b and miR-18b levels; high expression of miR-129, miR-135, miR-146a, miR-424, and miR-450; and low expression of miR-222 or low miR-223 levels have been associated with poor prognosis in MCL.88,91–94
Other B-cell lymphomas
Splenic marginal zone lymphoma (SMZL) is a rare indolent B-cell lymphoproliferative disorder characterized by the 7q32 deletion. It has been demonstrated that this chromosomal aberration triggers the downregulation of 8 miRNAs (miR-29a, miR-29b1, miR-96, miR-129, miR-182, miR-183, miR-335, and miR-593) in SMZL cases. 95 MiR-127, miR-139, miR-335, and miR-411 were also found downregulated in SMZL cases, whereas miR-451 and miR-486 were upregulated. 96
Mucosa-associated lymphoid tissue (MALT) lymphoma is a multifocal disease that involves the MALT, frequently of the stomach, and is frequently associated with chronic inflammation as a result of Helicobacter pylori infection. 11 On one hand, a signature of 27 miRNAs has been identified that can distinguish between gastritis and MALT lymphoma cases.97,98 On the other hand, miR-142 and miR-155 were found overexpressed in MALT lymphoma lesions compared with surrounding non-tumour mucosae. The expression levels of miR-142-5p and miR-155 were significantly increased in MALT lymphomas resistant to H pylori eradication than in cases showing complete remission after H pylori eradication. The expression levels of miR-142-5p and miR-155 were also associated with the clinical courses of gastric MALT lymphoma cases. 99
Discussion and Future Directions
Despite the rapid growth of literature proposing miRNAs as B-cell lymphoma biomarkers, we are still far from the clinical implementation. Most of the miRNA biomarker studies to date are single centre with a retrospective design, with not enough power in most cases (Table 1). As a consequence, many reports are non-overlapping or even contradictory. These differences are probably due to variation in the handling of the material and the technical methodology used in each study.
The choice of the starting material (whole blood, PBMCs, serum, plasma, fresh of FFPE biopsy material) is of vital importance for the experimental design as it will generate different expression profiles.164-166 Sample collection and handling procedures are also crucial, and in the case of liquid biopsies, they should be optimized to reduce the time between phlebotomy and processing and to avoid excessive haemolysis which could lead major differences in the levels of miRNAs.167–169
It should also be taken into account that differences in the miRNA purification procedure are a source of variability. 170 In addition, miRNA detection technique (qRT-PCR, microarrays, or NGS), along with the lack of a standard approach to normalization or a suitable endogenous reference gene for miRNA studies, can influence results significantly.13,24,171–175 It is therefore necessary to establish a standardized approach to miRNA biomarker studies alongside a systematic and comprehensive comparison of these confounding factors to ensure that the potential of these molecules is effectively realized in the clinic and live up to the hyperbole.
Footnotes
Acknowledgements
We apologize to the authors of the many studies who were not included in this review due to space limitations.
Funding:
The author(s) disclosed receipt of the following financial support for the research, authorship, and/ or publication of this article: This work is supported by grants from the Ikerbasque Foundation for Science, Ministerio de Economía y Competitividad of Spanish central government and FEDER funds (PI12/00663, PIE13/00048, DTS14/00109, PI15/00275), Departamento Desarrollo Económico y Competitividad y Departamento de Sanidad of Basque government, Asociación Española Contra el Cancer (AECC), and the Diputación Foral de Gipuzkoa (DFG).
Declaration of conflicting interests:
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Author Contributions
CS and EA contributed equally to this work.
