Alternative pre-mRNA splicing leads to potential biomarkers in diffuse large B-cell lymphoma – a systematic review

Diffuse large B-cell lymphoma (DLBCL) is a highly malignant haematological disease. Despite the introduction of the combination chemotherapy rituximab, cyclophosphamide, doxorubicine, vincristine and prednisone (R-CHOP), a large number of DLBCL patients have non-curable cancer associated with poor survival. Thus, research aiming to improve the outcome for these patients is necessary [1, 2]. The goal is to prescribe the correct medicine for each patient with the correct dose at the correct time, also known as personalised medicine. An approach towards this is the use of biomarkers. This approach could be based on alternative pre-mRNA splicing.

The classic flow of genetic information is from DNA through RNA via transcription, before the effector protein is generated by translation. The initial product of transcription is pre-mRNA that is modified to form many different transcripts in a process of selective inclusion or removal of exons. This mechanism is defined as alternative pre-mRNA splicing. Alternative pre-mRNA splicing ensures a high diversity of the resulting effector proteins because different protein isoforms with different functions can be generated from the same pre-mRNA [3, 4]. Misregulated alternative pre-mRNA splicing can, however, also contribute to malignant transformation, cancer progression and metastasis by activating oncogenes and inactivating tumour suppressors [5, 6]. In general, the characteristics of cancers, or “hallmarks of cancer”, have key elements that are alternatively spliced [5]. The role of alternative pre-mRNA splicing in DLBCL remains, however, largely unexplained [7, 8]. Based on the known role of alternative pre-mRNA splicing in the “hallmarks of cancer”, its use as a biomarker and target for a potential new class of anticancer therapeutics has been proposed [3, 5, 6].

Recognition of aberrations in splicing events and their associations with disease are widely acknowledged in a great number of human diseases [9], including neurological diseases [10], muscular dystrophy [11] and myelodysplastic syndrome [12]. Because deregulated alternative pre-mRNA splicing is known to occur in DLBCL, this has been proposed as a potential biomarker [13]. To our knowledge, the literature concerning alternative pre-mRNA splicing as a potential biomarker in DLBCL has, however, never been studied systematically.

A biomarker is defined by the National Institutes of Health (NIH) as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention” [14]. Biomarkers may be classified as diagnostic (identifying patients with an abnormal condition), prognostic (indicator for overall outcome) or predictive (therapeutic response prior to an intervention) [14, 15]. To assess the potential clinical usefulness of diagnostic biomarkers, a systematic approach guiding the process of biomarker development was developed by the Early Detection Research Network (ERDN) [16]. Currently, nothing similar has been designed for prognostic or predictive biomarkers.

Some biomarkers are already used in haematological, clinical indices such as the International Prognostic Index (IPI) [17]. The more recent classifications
of DLBCL into “activated B-cell-like” (ABC), “germinal-centre B-cell-like” (GCB) and an unclassified third type based on tumour gene expression profiling [18] are widely recognised, but not implemented in a routine clinical setting. Other promising novel biomarkers are miRNAs [19] and the B-cell associated gene signatures [20].

The ultimate goal when constructing a clinical test based on a biomarker, whether it is diagnostic, prognostic, or predictive, is that a high sensitivity and specificity for detecting and distinguishing between positive (diseased, poor prognosis, non-responder to treatment) and negative (non-diseased, good prognosis, responder to therapy) cases. Therefore, we find it relevant to evaluate biomarker studies in terms of their sensitivity and specificity to assess their potential contribution towards a shift into personalised medicine.

This systematic review aimed to evaluate alternative pre-mRNA splicing as clinically useful diagnostic, prognostic or predictive biomarkers in DLBCL by evaluating the strength and limitations of the study design, the evidence level, the potential sensitivity and specificity, and the potential contribution to personalised medicine.

METHODS

This review was organised according to the Preferred Reporting for Systematic Reviews and Meta-analyses (PRISMA) guidelines [21].

Search strategy

A systematic search for studies was performed in Embase, PubMed and Scopus. According to the Population-Intervention-Comparison-Outcome (PICO) approach [22], the search was structured by combining MeSH terms/EMTREE terms and/or free-text words related to the population (such as diffuse large B-cell lymphoma), intervention (such as RNA splicing) and outcome (such as biomarker) (Appendix A). No terms were searched for in the comparison category as it was not relevant for this review. The reference lists of all included studies were searched for additional studies that the electronic search strategy may have missed. The last literature search was performed on 30 June 2015.

Study selection

To identify relevant articles meriting full review, titles and abstracts retrieved by the electronic search were screened. Articles of interest that met the inclusion criteria (see below) were subsequently reviewed in full length before inclusion. Reports on alternative splicing as a potential diagnostic, prognostic or predictive biomarker in DLBCL were considered for inclusion if they had alternatively spliced pre-mRNAs or resulting protein isoforms as an endpoint. To narrow the review, it was predefined to report only alternative pre-mRNA splicing and not mutations resulting in deregulated splicing machinery leading to alternative pre-mRNA splicing. Moreover, papers reporting alternatively spliced pre-mRNAs or resulting proteins as a target for therapeutics were not included as pharmacodynamics were not the main focus in this review. Reports concerning protein isoforms resulting from post-translational modifications were also excluded. Only original, full journal publications were included. In addition, the selection was carried out without limitations regarding study design, publication year or language.

Data extraction

To extract relevant information, a predesigned data abstraction form was used based on the Strengthening the Reporting of Observational studies in Epidemiology: Molecular Epidemiology (STROBE-ME) guideline [23]. The methodological quality was assessed by evaluating the limitations and strengths of each study [23-25] because no validated tool currently exists.

RESULTS

Search results and selected publications

Through database and reference list searching, 165 articles were identified (Figure 1). The search results from each database were imported into the reference manager programme Mendeley, and 27 duplicates were automatically removed. Therefore, 138 articles were the starting point for the analysis. By reviewing titles and abstracts, 22 papers fulfilled the inclusions criteria and were eligible and relevant for this analysis. A total of six studies were excluded from the systematic analysis because only their abstracts were available. Thus, 16 full text articles remained for analysis.

Study characteristics of included studies

The eight studies reporting potential diagnostic biomarkers [26-33] were observational and cross-sectional due to the expression level being reported at one time point [34] (Table 1). The data sources varied from two cell lines to 250 tissue samples, comparing normal cell lines or other neoplastic cell line samples.

Two studies reported a potential biomarker to be both prognostic and predictive [35, 36] (Table 2). In general, all the studies reporting potential prognostic or predictive biomarkers were observational identifiers for the level of biomarker expression associated with an outcome or response to treatment. A total of seven studies were characterised as longitudinal studies with retrospective data from 28 to 290 tissue samples, and/or cell lines and clinical data from databases or medical records [13, 35-40]. One study had a cross-sectional design that examined the prognostic value of the potential biomarker by comparing the expression level of the potential biomarkers with the number of patients who survived by the end of the study period [41].

To assess any contribution to the shift towards personalised medicine and the potential sensitivity and specificity, the statistical methods in these studies were evaluated. Only studies reporting potential prognostic or predictive biomarkers that described the statistical methods used were included (Table 1). In general, these studies used simple statistical methods, such as correlation coefficients. Confidence intervals were reported only in one of the studies [36]; however, the significance value was reported in all studies [13, 35-41]. None of the studies presented pre-study power calculations. The sensitivity and specificity of the tests were stated only in Nagel et al by receiver operating characteristics (ROC) curves [37].

Results from individual studies

Potential diagnostic biomarkers

Several potential diagnostic biomarkers were investigated in different ways and a few studies examined the same potential biomarker [32, 33]. Several studies identify the presence of a potential promising diagnostic biomarker by comparing the expression level to normal or other neoplastic cells [26-28]. Two studies explore the distribution of alternative pre-mRNA splicing events between ABC- and GCB-DLBCL [27, 32]. These potential biomarkers were therefore hypothesised to be related to a worsened or improved outcome for ABC- and GCB-DLBCL, respectively [18]. Additionally, two studies investigate the presence of alternative pre-mRNA splicing in two subtypes of DLBCL, namely primary mediastinal large B-cell lymphoma (PMLBCL) and primary central nervous system lymphoma (PCNSL) that generally have a very poor prognosis compared with systematic DLBCL [31, 33]. They report that alternative pre-mRNA splicing in particular is present in these lymphomas compared with DLBCL.

Potential prognostic biomarkers

When considering the expression of potential prognostic or predictive biomarkers in the identified studies, a predominance of studies reporting alternatively spliced variants of CD44 were observed. Several studies describe that the alternative splicing of CD44 was significantly correlated with clinically accepted prognostic staging methods such as Ann Arbor Stage and IPI in DLBCL [13, 35, 37-40].

The identified studies use different endpoints for evaluation of the prognostic potential of CD44 splice variants (Table 2). All of the studies [13, 35-41] describe a high expression of CD44 splice variants correlated with a poor prognosis, except for Wei et al, who report that CD44v6, a specific alternatively spliced variant of CD44, was associated with superior survival in a multiple Cox regression analysis [35]. Furthermore, two studies report conflicting evidence as to whether or not alternative pre-mRNA splicing of CD44 is associated with increased or decreased relapse rates [13, 38].

Potentially predictive biomarkers

Two studies [35, 36] report alternative pre-mRNA splicing as a drug-specific, potentially predictive biomarker (Table 2). Espinosa et al [36] concluded that the PKC-β II membrane protein predicts a decreased complete remission (CR) rate when DLBCL patients were treated with adriamycin-containing chemotherapy. Wei et al [35] report that positive expression of the CD44H protein predict reduced overall survival (OS) and event-free survival (EFS) when treated with CHOP, while the predictive
value is no longer present when treated with R-CHOP.

APPENDIX