Chronic B-cell lymphocytic leukaemia (CLL) is generally an indolent disease, with most patients surviving years without treatment although some progress more rapidly. Several markers, such as IGHV mutational status (Hamblin et al, 1999), cytogenetic abnormalities (Döhner et al, 2000) and recurrent gene mutations (Wang et al, 2011; Jeromin et al, 2014), have helped to better stratify patient risk, but few are routinely used by clinicians and they are still relatively unreliable, reflecting the clinical and physiopathological heterogeneity of the disease. Thus there remains room for improvement and new molecular markers. Our understanding of the various roles of small nucleolar RNAs (snoRNAs) in cancer is continually expanding. Besides their well-known function in ribosomal RNA modifications, snoRNAs have recently been described as new biomarkers in haematological cancers (Ronchetti et al, 2012, 2013; Valleron et al, 2012a,b). Notably, Ronchetti et al (2013) studied a cohort of Binet stage A CLL patients, reporting that a specific snoRNA signature was predictive of outcome for early stage CLL. In this study, we investigated the snoRNA expression profiles in CLL patients fully annotated for both the classical prognostic markers (IGHV mutational status, cytogenetic abnormalities, Binet stage) and recurrent somatic mutations (TP53, NOTCH1, SF3B1). The methods and patient characteristics are described in Data S1, Tables SI and SII. Using high-throughput quantitative polymerase chain reaction (Fluidigm, Les Ullis, France), we first determined whether snoRNA profiling (n = 221, covering more than two-thirds of previously described snoRNAs) could discriminate between CLL prognostic subgroups in an exploration set of 58 treatment-naive CLL and five normal B-cells. Unsupervised hierarchical clustering showed that patients did not cluster together when considering criteria such as IGHV mutational status, Binet stage, age, gender, karyotype, fluorescence in situ hybridization and NOTCH1/TP53/SF3B1mutation, but instead were scattered along the dendrogram (Fig S1). Supervised analysis also did not find a snoRNA signature specific to the conventional clinico-biological parameters, suggesting that snoRNA expression profiles were not associated with the aforementioned factors impacting on CLL outcome. The apparent contradiction between our results and those reported by Ronchetti et al (2013) could be explained by the differences between the two cohorts as we included more cases presenting adverse prognosis factors. However, the unsupervised analysis showed that normal B-cells clustered together, suggesting that snoRNA expression profiles are influenced by leukaemic phenotype. Sparse Partial Least Squares-Discriminant Analysis (sPLS-DA) is a mathematical method that can be used to determine the best gene set while also allowing discrimination between two groups. Thus we used sPLS-DA to identify snoRNAs that could discriminate between normal B-cells and CLL cases and obtained a minimal set of four snoRNAs (Fig 1A, B). SNORD35B, SNORD71, SNORD116-11 and SNORD116-25 were sufficient to separate healthy cells from leukaemic cells in hierarchical clustering analysis (Fig 1C). These results were confirmed in a validation cohort that included 56 new, treatment-naïve, CLL and five new normal B-cells, stressing the robustness of this signature (Fig 1D). We then investigated whether snoRNA profiles could be associated with differences in CLL progression. To this end we focused on treatment-free survival (TFS), which is the first prognostic parameter in the disease course, as some CLL patients remain therapy-free for many years while others rapidly progress and require treatment. The patients included in our exploration and validation cohorts had a median TFS of 34 and 30 months respectively (in keeping with the frequency of adverse biological risk factors in our population). We could not demonstrate any dysregulation of a specific set of snoRNAs based either on the median TFS of the exploration cohort or among the different subtypes, with the exception of the IGHV-mutated (IGHV-M) patients. Most IGHV-M patients display a prolonged TFS but snoRNA expression profiles enabled division into two prognostic subgroups. Using sPLS-DA we identified a set of 20 snoRNAs (Table SIII) that are globally overexpressed among IGHV-M patients with an unexpectedly short TFS (Fig S2). To increase the number of IGHV-M patients, the exploration and validation cohorts were merged to compare TFS between IGHV-M patients with high and low expression of the 20 snoRNAs (according to the signature expression index, Fig S3) and IGHV-unmutated (IGHV-UM) patients. Kaplan–Meier analysis confirmed that overexpression of the 20 snoRNAs was associated with a shorter TFS (median: 32 months) compared to underexpression of the 20 snoRNAs (median: 144 months) (Fig 2A). Addition of IGHV-UM to the Kaplan–Meier analysis indicated that the TFS of IGHV-M patients with high expression of these snoRNAs was similar to that of IGHV-UM patients (Fig 2B), leading us to postulate that IGHV-M patients with greater progression potential could be further classified using snoRNAs. However, due to the low number of IGHV-M cases investigated, this signature needs further validation in a larger prospective set of IGHV-M patients. We then investigated the putative function of these snoRNAs in CLL proliferation. One of the main criteria indicating the need to initiate treatment is increased lymphocyte doubling time (Hallek et al, 2008). Thus, a shorter TFS can be related to increased cell proliferation. Therefore, we treated primary cells from five CLL patients with immunostimulators (interleukin 2 and CpG oligodeoxynucleotides) to induce proliferation (Fig S4A). We also induced the proliferation of normal B-cells using lipopolysaccharide and anti-IgM (Fig S4B) to assess the potential abnormal regulation of snoRNAs in CLL. Following proliferation, quantification showed that 11–17 of the 20 snoRNAs had increased expression in CLL, depending on the patient (Fig S5). Similarly, 15–18 snoRNAs had increased expression in normal B-cells, depending on the donor (Fig S6). This suggests that most of the 20 snoRNAs could be functionally relevant in a proliferation context. Interestingly, comparison of the fold change in expression of the 20 snoRNAs between CLL and normal B-cells shows that only seven were consistently upregulated following proliferation in both normal B-cells and CLL. Two snoRNAs (SNORA80; SNORD1A) were consistently downregulated in CLL (Fig S5). These results indicate that some snoRNAs are deregulated in response to proliferation in CLL. Nevertheless, as these 20 signature snoRNAs are globally upregulated in short TFS IGHV-M patients, our results imply that the signature reflects more than just the proliferation potential of the leukaemic cells. Other factors affecting TFS, such as increased resistance to apoptosis, should also be explored to assess their impact on snoRNA expression. Taken together, these results suggest that snoRNA expression profiles could be used as new biomarkers to refine the classification of IGHV-M CLL and should be further explored to ascertain the functional link between progression and non-coding RNA modulation. The authors thank the patients who participated in this study, Frédéric Martins and Jean-José Maoret (GeT-Purpan genomic platform, Toulouse, France). PB is supported by the Institut Universitaire de France and the LABEX TOUCAN (Laboratoire d'Excellence Toulouse Cancer). LB is supported by the Fondation ARC (Association pour la Recherche sur le Cancer). LB, WV, SG, CQ and OZ performed the experiments; LB, WV, MB, LY and PB designed the study; AQ-M, LY and FD collected and assembled the clinical records; LB, WV and MB analysed and interpreted the data; LB, MB, LY and PB prepared the first draft and finalized the manuscript; and all authors contributed to the writing of the manuscript and gave final approval. The authors declare no competing financial interests. Data S1. Supplementary methods. Fig S1. SnoRNA expression profiles in CLL and CD19+ control samples. Fig S2. A 20 snoRNA signature identifies low- and high-risk CLL among IGHV-mutated patients. Fig S3. 20 snoRNA signature expression index. Fig S4. CFSE dilution after proliferation induction in CLL and normal B-cells. Fig S5. Expression levels of the 20 snoRNAs in the signature after proliferation induction in CLL. Fig S6. Expression levels of the 20 snoRNAs in the signature after proliferation induction in normal B-cells. Table SI. Patient characteristics of the exploration and validation cohorts. Table SII. List of the primers used in the Fluidigm experiment. Table SIII. Name, targets, host gene and localization of the 20 snoRNAs from the signature. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
Les petits ARN nucleolaires (snoARN) forment une famille d'ARN non codants majoritairement impliques dans la maturation des ARN ribosomaux. De nouvelles fonctions ont ete decrites rendant leur etude pertinente dans le cancer. Deux pistes ont ete investiguees dans le modele des cancers hematopoietiques : i) les proteines de fusion oncogeniques recurrentes ont-elles un impact sur le profil d'expression des snoARN ? ii) l'expression des snoARN peut-elle etre correlee aux caracteristiques clinico-biologiques ? Pour suivre la premiere piste, nous avons analyse les profils d'expression de snoARN dans le modele des leucemies aigue myeloblastiques. Nous avons demontre que les snoARN sont deregules dans les cellules neoplasiques. Une signature specifique des snoARN SNORD112-114 a ete mise en evidence chez les patients atteints de leucemie aigue promyelocytaire. L'expression ectopique de ces snoARN apparait liee a la proteine oncogenique PML-RARalpha. De plus, le snoARN prototype SNORD114-1 influence positivement la croissance cellulaire. Nos donnees suggerent donc l'implication de snoARN dans la tumorigenese. La seconde piste a ete menee sur un ensemble heterogene de lymphomes a pronostic pejoratif, les lymphomes T peripheriques. Nous avons recherche un lien entre le profil d'expression des snoARN et les caracteristiques cliniques d'une cohorte de patients atteints des trois sous-types majeurs de lymphome T peripherique. Nous avons constate que les snoARN sont significativement sous exprimes dans les cellules neoplasiques versus cellules saines. Nous avons egalement determine que les snoARN pouvaient servir d'outil diagnostique et d'outil pronostique pour certain sous-types.
Tumor-intrinsic immuno-resistance is a prerequisite for emergence of follicular lymphomas. Here we show that in vitro, such cells are more resistant to immune cytolysis when grown as follicle-mimicking tridimensional aggregates than when grown as cell suspensions. So in patients, this innate adaptation to tumor immunity might precede its selective pressure.
