Information

Steps to confirm if predicted miRNA is good or bad

Steps to confirm if predicted miRNA is good or bad


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I aligned all miRNAs available to the supercontigs of a particular genome with certain parameters (e value of 0.01 and a word match of atleast 7 as suggested in this paper). I have also isolated the pre-miRNAs (+100 nucleotides from either end of the match area). What would you suggest as ideal ab-initio methods to confirm that these miRNAs do exist in that particular genome? Some papers I found use mFOLD like this.

What would you suggest as best steps to confirm if a predicted miRNA/pre-miRNA is really present in the genome.


There won't be any duplicates in miRBase (for a given organism). Choose the taxon nearest to your organism if you are doing homology based discovery. If you want to take all/many organisms then you can usefastx_collapserto collapse redundant sequences. However you will lose the name of the miRNA. You can useawkalso for this and it will keep the sequence header from the first organism in the list.

awk '/>/{$0=h} !(/>/){if($0 in a){next} else{a[$0];print h"
"$0}}' organism1.fa organism2.fa… organismN.fa

Make sure that there are no extra newlines, otherwise you might need a small modification

To know if something is an expressed miRNA you would need to do a small RNA sequencing. There are some tools like mirdeep and mirSVR which you can use to discover miRNA sequences.


MiR-493 induction during carcinogenesis blocks metastatic settlement of colon cancer cells in liver

Liver metastasis is a major lethal complication associated with colon cancer, and post-intravasation steps of the metastasis are important for its clinical intervention. In order to identify inhibitory microRNAs (miRNAs) for these steps, we performed 𠆍ropout' screens of a miRNA library in a mouse model of liver metastasis. Functional analyses showed that miR-493 and to a lesser extent miR-493 * were capable of inhibiting liver metastasis. miR-493 inhibited retention of metastasized cells in liver parenchyma and induced their cell death. IGF1R was identified as a direct target of miR-493, and its inhibition partially phenocopied the anti-metastatic effects. High levels of miR-493 and miR-493 * , but not pri-miR-493, in primary colon cancer were inversely related to the presence of liver metastasis, and attributed to an increase of miR-493 expression during carcinogenesis. We propose that, in a subset of colon cancer, upregulation of miR-493 during carcinogenesis prevents liver metastasis via the induction of cell death of metastasized cells.


1. Introduction

Mutations are one of the fundamental forces of evolution because they fuel the variability in populations and thus enable evolutionary change. Based on their effects on fitness, mutations can be divided into three broad categories: the ‘good’ or advantageous that increase fitness, the �’ or deleterious that decrease it and the ‘indifferent’ or neutral that are not affected by selection because their effects are too small. While this simplistic view serves well as a first rule of thumb for understanding the fate of mutations, research in recent decades has uncovered a complex web of interactions. For example, (i) the effects of mutations often depend on the presence or absence of other mutations, (ii) their effects can also depend on the environment, (iii) the fate of mutations may depend on the size and structure of the population, which can severely limit the ability of selection to discriminate among the three types (making all seem nearly ‘indifferent’), and (iv) mutations' fate can also depend on the fate of others that have more pronounced effects and are in close proximity on the same chromosome.

A major theoretical goal in the study of the population genetics of mutations is to understand how mutations change populations in the long term. To this end, we have to consider many features of evolution and extant populations at both the phenotypic and molecular level, and ask how these can be explained in terms of rates and kinds of mutations and how they are affected by the forces that influence their fates.

We have increasing amounts of information at our disposal to help us answer these questions. The continuous improvement of DNA sequencing technology is providing more detailed genotypes on more species and observations of more phenomena at the genomic level. We are also gaining more understanding of the processes that lead from changes at the level of genotypes through various intermediate molecular changes in individuals to new visible phenotypes. Use of this new knowledge presents both opportunities and challenges to our understanding, and new methods have been developed to address them.

Brian Charlesworth has been at the forefront of many of the developments in the population genetics of mutations, both in the collection and analysis of new data and in providing new models to explain the observations he and others have made. This themed issue of Phil. Trans. R. Soc. B is dedicated to him to mark his 65th birthday. The authors of the accompanying papers have individually made important contributions to the field and have been directly associated with or indirectly influenced by his work.

In this collection of papers, various aspects are considered in detail, and in this introduction, we aim to provide an overview as a basis for the in-depth treatments that follow. We outline some of the theories that serve as the quantitative basis for more applied questions and have been developed with the main aims of: (i) measuring the rates at which different types of mutations occur in nature, (ii) predicting quantitatively their subsequent fate in populations, and (iii) assessing how they affect some properties of populations and therefore could be used for inference. The subsequent papers are broadly arranged in a continuum from specific questions of basic parameter estimation (strength of mutation, selection, recombination), via those that contribute a combination of biological theories and data on these parameters, to those which mostly address broader biological theories.

There is an enormous range of mutational effects on fitness, and wide differences exist in the strength of other evolutionary forces that operate on populations. This generates an array of complex phenomena that continues to challenge our capacity to mechanistically understand evolution. To make problems tractable, theoreticians have divided the parameter space into smaller regions such that specific simplifying assumptions can be made. These typically comprise assuming the absence of particular events (e.g. no recombination) or the presence of particular equilibria (e.g. mutation-selection balance). Subsequently, new theories are often developed in which these assumptions are relaxed so as to narrow the gap to reality, typically including more interactions between various evolutionary forces, albeit at the cost of becoming less tractable to analysis.

The dynamics of mutations are dominated by chance, yet we search for general principles that are independent of particular random events. This tension is reflected in the models used. All mutations start out as single copies and most are lost again by chance, so we can at best predict probabilities of particular fates but the stochastic models that can deal rigorously with randomness are often too complex to analyse for realistic scenarios. If we are interested only in the mean outcome of many individual random events, we may approximate the process by deterministic models that predict a precise outcome but these approximations can break down if only few individuals or rare events are involved.

To facilitate concise descriptions, there is a long history in population genetics of using mathematical symbols as abbreviations for various parameters and observations, but unfortunately there is no unique nomenclature. To try to meet our two conflicting goals of conciseness and readability, we list some important evolutionary parameters and their common abbreviations in tableਁ . Even so, for good reasons of history or local convention, some of these symbols are defined differently in some papers in this collection.

Tableਁ.

Some parameters in the population genetics of mutations*.

Umutation rate per generation per genome check context for effects of mutations
Ge, Geffective haploid genome size (all functional base pairs), total haploid genome size (with neutral sites)
μ, μ10, μ01mutation rate per locus or per site per generation, away from the preferred base, and back
κ, tn/tvmutational bias: μ10/μ01, transition/transversion ratio
re, rco, rgceffective recombination rate, cross over rate, gene conversion rate
sselection coefficient measures changes in fitness check context for exact definitions (homozygous or heterozygous positive or negative)
hdominance coefficient so that sh is the effect of heterozygous mutations
DME (or DFE)distribution of mutational effects on fitness
WAWrightian fitness of a genotype A (one of the many ways fitness can be defined)
εepistasis: interactions of mutational effects. If fitness is multiplicative, ε = WABWAWB
Ne, Neffective population size, census population size
mmigration rate
Pfixprobability of fixation of a (mutant) allele
Tfix, Tlosstime to fixation, loss in generations
KA, KS (or DN, DS)rate of DNA divergence per site between two species corrected for multiple hits (see context for method) substitutions can be non-synonymous (change amino acids), or synonymous (or silent) (check context)
πA, πS, θWA, θWSDNA diversity within a population per site: π is the average pairwise nucleotide diversity, θW is Watterson's estimate for explanation of indices see KA, KS
D, D′, r 2 measures of linkage disequilibrium (LD)
VGgenetic variance in a quantitative trait
VMincrement in VG from new mutations per generation
VEenvironmental variance in a quantitative trait

*For historical reasons and due to the limitations of the alphabet, several symbols have different meanings in different contexts. Examples: h 2 = heritability of a quantitative trait r = rate of selfing r = rate of population growth D = ‘Tajima's D’, where D < 0 may indicate population expansion or directional selection and D > 0 a bottleneck or balancing selection.

Naturally, a review of this length cannot cover all aspects of the population genetics of mutations. For example, mutation plays a pivotal part in coalescent theory (Hein et al. 2005) and in the construction of genotype–phenotype maps that are at the core of some efforts to understand adaptive landscapes, which provide a paradigm for understanding many broader aspects of population genetics from the perspective of individual mutations (�uses cancer or not’), as reviewed elsewhere (Loewe 2009). Here we focus almost entirely on how populations of individuals are changed by large numbers of mutations that have specified effects on fitness.

In ੲ of this paper, we discuss what is known about the diversity of mutations, and here and subsequently we refer to other papers in this themed issue that provide more in-depth information. In ੳ, we review some of the relevant theory in population genetics, starting with (i) simple theories that treat the fate of individual mutations in isolation before turning to more complicated models that consider (ii) linkage, (iii) epistasis, (iv) quantitative genetics approaches, and (v) challenges faced when attempting to integrate all these. Subsequently, we provide an overview of several general questions that have been resolved and others that remain (ੴ) and finally some conclusions (ੵ).


2. Experimental Procedures

2.1. Human Rhabdomyosarcoma Tumours

Seventeen patients treated for rhabdomyosarcoma in the Centre Léon Bérard were included in this study. Four frozen tumours and thirteen formalin-fixed paraffin-embedded tumours were obtained from biopsies realized at the diagnosis. Tumour diagnoses were realized by a referent anatomopathologist specialist for this pathology by immunohistochemistry, FISH, and qPCR.

2.2. Cancer Cell Lines

Five cancer cell lines were obtained from ATCC (Manassas, VA, USA): the two human osteosarcoma MNNG/HOS Cl #5 [R-1059-D] (reference CRL-15-47) and Saos-2 (HTB-85) cells, the chondrosarcoma cell line SW1353 (HTB-94) and the two Burkitt lymphoma Daudi (CCL-213) and Namalwa (CRL-1432) cells. Osteosarcoma and chondrosarcoma cells were grown in DMEM (Gibco, Carlsbad, CA, USA), supplemented with 10% decomplemented fetal calf serum (Lonza, Basel, Switzerland), 10 mL penicillin streptomycin (10 U/mL/10 μg/mL, Gibco, Carlsbad, CA, USA), and 5 mL L-glutamin (200 mM Gibco) at 37ଌ humidified atmosphere containing 5% CO2. Lymphoma cells were grown in RPMI (Gibco, Carlsbad, CA, USA). Cells were exposed to 100 nM RAD-001 (Novartis), 50 μM ifosfamide (ifos, Baxter) or 1 μM cisplatin (CDDP, TEVA) or 100 μM methotrexate (MTX, TEVA) for 24, 48, and 72 h.

2.3. Rat Osteosarcoma Model

Procedures for animal care were performed according to institutional and national guidelines. Animals were anesthetized throughout all surgical and imaging procedures with isoflurane/oxygen (2.5%/2.5%, v/v) (Minerve, Esternay, France). The transplantable orthotopic and metastatic rat osteosarcoma model has been previously described [21�]. This model mimics its human counterpart in terms of aggressiveness, metastatic spreading and chemoresistance phenotype [21�]. All the tumours obtained were classified as osteoblastic following histological analyses. Briefly, small tumour fragments (100 mm 3 ) taken from a hyperproliferative osteogenic tumour area were grafted on 3-weeks old immunocompetents Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA, USA). Using a lateral approach, a tumour fragment was placed contiguous to tibial diaphysis after periosteal abrasion then, the cutaneous and muscular wounds were sutured. Fourteen days after tumour transplantation, animals underwent a first 18 F − FDG PET Scan and were randomly assigned to a control group treated with saline solution or a treated group exposed to a subcutaneous dose of 10 mg/kg ifosfamide (ifos, Baxter, Deerfield, IL, USA), 7 days apart (at days 15 and 22 after tumour transplantation). A second 18 F − FDG PET Scan was performed 7 days after the second ifos administration. Animals were sacrificed one week after the end of the treatment. Tumour and normal tissue fragments (muscle, bone, and lung) were collected for RNA extractions.

2.4. RNA Extraction and Quantitative Real-Time PCR

FFPE tumours were lysed for 24 h in ATL buffer (Qiagen, France) supplemented with proteinase K (Qiagen) at 60ଌ in rotative agitation after different washes with toluene, ethanol, and tris/EDTA in this order. Total RNA was extracted from tumour or cell pellets using a single phenol/chloroform extraction protocol with Trizol, according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). Five hundred nanograms of total RNA were subjected to the microfluidic PCR technology performed by Applied Biosystems (Foster City, CA, USA). In brief, RNA was reversed transcribed, using multiplexed specific looped miRNA primers from the Taqman MicroRNA Reverse Transcription kit. The second step consists in a real-time quantitative PCR on TLDA: RT products are introduced through microchannels into miniature wells that are preloaded with dehydrated specific primers and probes. Recently, Applied biosystems released the second version of TLDA, consisting of two cards A and B. Analyses were performed for 377 miRNAs on card A and 290 on card B.

2.5. PCR Data Normalization

For each miRNA, the threshold cycle (Ct) was calculated by the ABI 7900 Sequence Detection System software (plate by plate manual Ct analysis with a threshold at 0.25 and automatic baseline). All further data manipulations were done using R scripts. A cutoff of 32 was applied to discard the late Ct values, except for RMS analysis. Around 60% of miRNAs passed the filtering criteria and were used for further analysis. For each TLDA, quality controls were performed on the raw data by checking internal controls and using box plot and scatter plot diagrams. Samples with any kind of problems were discarded so they would not introduce bias during the following normalization procedures. We tested different methods of normalization since the recommended “pseudo” normalization factor mammU6 plotted in each card was not stably expressed in our different samples. Normalization with the two most stable miRNAs identified by GeNorm was not optimal too. Finally, a global normalization by the median was chosen for its reliability over experiments. Tissues included in a given analysis were treated altogether, the normalization procedure being applied separately for the two types of card, A and B. Distribution of normalized data was checked with box plots and correlation plots. The following formula was used to correct Ct values of every card:

Through this approach, the new median value shared by all samples can be considered as a sort of perfect “virtual housekeeping gene”. Therefore, the standard Δ㥌t method can be used to determine the relative quantities (RQ) as follows:

For the Δ㥌t calculation, it was more relevant for the statistical analyses to use the mean of all 㥌t obtained across samples for each miRNA, instead of using the 㥌t of a reference sample

2.6. miRNA Target Predictions

We compiled 4 databases to determine miRNA targets: TargetScan 5.1, MiRanda, PICTAR, and the miRbase databases. These databases search the presence of conserved 8mer and 7mer sites on the 3′UTR parts of messenger RNA that match the seed region of each miRNA. It also predicts the efficacy of targeting for each matching site. We created our own database which regrouped each miRNA with the geneID of all their protein targets, for rat and human. We only conserved couples miRNA/geneID present in two databases at least.

2.7. miRNA-Regulated Cell Signalling Pathways Predictions

We used the “G-language microarray” web application, which allows the mapping of molecular dataset onto “Kyoto Encyclopedia of Genes and Genomes” (KEGG) pathway maps [24]. We first input miRNA-targeted proteins of interest and the sum of RQ values for all miRNAs that regulate these proteins, contained between 1 and 50 the software then generates KEGG data to create FLASH graphics of cell signalling pathways in which proteins are involved. The colour intensity of a highlighted protein varies with the strength of its regulation by miRNAs.

2.8. Proliferation Assay

Cells were plated in 96 well plates at 5000�lls/well and exposed to 100 nM RAD-001, 50 μM ifosfamide, 100 μM methotrexate, or 1 μM cisplatin or not (NT). Cell growth was measured 24, 48, and 72 h later with 20 μL Cell Titer Glo luminescent reagent (Promega, Madison, WI, USA) for 10 min. Luminescence was recorded using a Microbeta reader (PerkinElmer, Fremont, CA, USA).

2.9. Western Blot

Pelleted cells were resuspended in lysis buffer (Tris 50 mM pH 7.4, NaCl 250 mM, EDTA 5 mM, NaF 50 mM, Triton X-100 0.1%, orthovanadate 1 μM) plus protease inhibitors for 30 min on ice. After a centrifugation at 14000 rpm for 10 min, supernatants were boiled for 5 min in Laemmli sample buffer (Biorad, Hercules, CA, USA). Analysis of protein content was performed on 4%�% gradient gel. After electrophoretic separation, 30 μg proteins were electrotransferred on a polyvinylidene difluoride membrane (Immobilon P, Millipore corp., Bedford, MA, USA). The membrane was then blocked for 1 h at room temperature with blocking agent 0.2% in PBS/Tween 0.1%, probed overnight with a primary rabbit antibody against the protein of interest, and finally revealed with a secondary antirabbit antibody HRP conjugated (Upstate Biotechnology, Lake Placid, NY, USA) and ECL Advance system (GEhealthcare, Chicago, IL, USA). Primary antibody used was obtained from Cell Signaling (New England Biolabs, Beverly, MA, USA) used at 1/1000. The β actin was used as a reference (Sigma).

2.10. CGH Array

Oligonucleotide-based microarray analysis was performed using a custom-designed, 244K-feature whole-rat genome microarray manufactured by Agilent Technologies (Santa Clara, CA). Genomic DNA labeling, array hybridization, and washing were performed as specified by the manufacturer (Agilent Technologies). Results of aberration calls consisting of three or more consecutive oligos were then displayed using custom oligonucleotide CGH analysis software (Genespring).

2.11. Statistical Analysis

Normalized RQ data were directly input into the TIBCO Spotfire DecisionSite for Functional Genomics analysis software. We performed unsupervised hierarchical clustering to classify samples by groups. The selection of miRNAs useful to predict tumour response to treatment was statistically realized using ANOVA tests with P values of  .05 at least. Results were verified through supervised hierarchical clustering.

Data from miRNA lists of interest were then used as variables in a three-dimensional principal component analysis (PCA) performed with R 2.9.0 package to demonstrate their capabilities to distinguish types of tumours. PCA supplies a simplified three-dimensional picture to our multivariate dataset of miRNA RQ values. By mathematical combination of values according to their strength, three principal components are created that represent as much as possible the variability of the data. Thus, tumours possess three new coordinates in a three-dimensional space. According to their localization in this space, tumours form groups, and their subtypes can be predicted.


3. Discussion

An increased regenerative potential of the human heart following ischemic damage may reduce the occurrence of heart failure in patients. Here, we delineated the short- to long-term transcriptome response in zebrafish hearts after cryoinjury to gain insight into the mRNA dynamics and the driving miRNA regulators during zebrafish heart regeneration. To the best of our knowledge, this is the largest study of multi-omics data monitoring heart regeneration up to 160 dpi in zebrafish. While cryoinjury has been shown to be a good model to mimic myocardial infarction [6,8] the impact of the preceding surgery clearly affects the transcriptome and has to be taken into consideration, e.g., with the use of sham-operated controls. In addition, transcriptome changes with increasing age need to be adjusted using age-matched controls. Hence, age-matched and sham-operated controls are recommended to best study the regeneration related response in zebrafish.

However, preparing sham-operated and age-matched controls for every time point increases the operating time and expenses notably. While other studies used either sham-operated controls which recovered at one time point [12,16] or healthy controls [34], we used a control group with time-dependent weights of sham-operated fish (recovered at 1 dpi) and unoperated healthy fish covering all ages to extract particular heart regeneration changes induced by cryoinjury. Yet, this approach underlies the assumption of an exponential decay of the surgery influence and only sham-operated controls at every time point can exactly mimic the surgical impact in the transcriptomic changes. Sham-operated controls may be particularly important for studying the changes of the miRNAome of cryoinjured fish at late time points because even at 160 dpi, samples were still separated from healthy controls ( Figure 2 B). Alternatively, this might suggest a possible permanent change of the miRNAome after heart injury.

Using a mixed-control group, differential expression analysis showed the largest transcriptome changes at early time points (1𠄷 dpi) suggesting a major and fast response to cryoinjury. The most prominent pathways correspond to cell cycle and DNA replication processes, which is in accordance to previous studies [4,12,16] and the fact that regeneration occurs mainly by proliferating CM rather than differentiation processes [10,11]. During the intermediate response (14� dpi) cell cycle processes are still up-regulated, but ECM organization processes become more prominent, indicating a still ongoing remodeling of the scar tissue. At late time points (60� dpi) we observed no significant cell cycle alterations, suggesting a completion of the heart regeneration and tissue remodeling up to 60 dpi after cryoinjury. Yet, full functional recovery has not been accomplished, with heart function restoring processes still being up-regulated at 60� dpi, such as cardiovascular development, heart contraction and growth.

These results were further corroborated by the delineation of the transcriptome response using soft-clustering techniques. Here, proliferation regulators showed a fast increase up to 4𠄷 dpi and a slow decrease back to baseline until 45 dpi, which confirms previously observed phenotypical data of scar volume [7,8]. Once, the heart tissue is reestablished, processes restoring the heart function are activated. This can be seen for clusters 3 and 4, which did not return to baseline levels at 160 dpi and were being enriched for heart development and adhesion processes. Therefore, the transcriptome response of heart regeneration after cryoinjury is rather a long process up to 160 dpi [6], which is further in agreement with phenotypic observations of Hein et al. who still found limited radial wall displacement and the presence of scar tissue at 180 dpi [8].

Using spatially resolved transcriptome data [15], we assigned the dynamic clusters with their primary location sites in the injured heart. Genes involved in proliferation processes were highest expressed within the injury site of the heart, evincing the injured tissue to be the major source of proliferation, possibly through invading CM from the uninjured myocardium [13].

Since miRNAs are suggested to play a major role in regulation of heart regeneration [19], we investigated both the mRNAome and the miRNAome. By correlating mRNAs with miRNA dynamics, we identified miRNA target genes with visible repression (anti-correlation) in expression and therefore diminished false-positive binding sites that are often present in miRNA target-interaction data bases [21]. Among the top 10 miRNAs regulating the most genes, we found previously described and potential novel candidates.

We identified miR-144 to be up-regulated in the initial response (1� dpi Supplementary Figure S5) and targeting mostly genes in clusters 3 and 5 that were initially down-regulated. Interestingly, miR-144 was shown to play an important role in hematopoiesis and vascular development in zebrafish through inhibition of meis1 [35], but in our data both miR-144 and meis1 were up-regulated up to 21 dpi, suggesting a different mechanism. However, we could confirm enrichment of GO:0048534 hematopoietic or lymphoid organ development for down-regulated genes in the early, late and intermediate response, what matches to the dynamic of miR-144 and its role in hematopoiesis. While it could be hypothesized that these miRNAs hinder the regeneration process, we believe that the down-regulation of genes in clusters 3 and 5 belongs to a well-orchestrated regeneration process and is initially needed when proliferation is most important.

On the other hand, miR-218b was initially down-regulated and its target genes (up-regulated) were associated with DNA replication (Supplementary Figure S5). In cancer miR-218 acts as a tumor suppressor by inhibiting proliferation and migration in glioma cells [36], bladder cancer [37] and non-small cell lung cancer [38], indicating similar mechanisms regarding proliferation.

Moreover, the down-regulation of miR-148/152 family could be linked to the up-regulation of genes mostly involved in proliferation processes (cluster 2), suggesting an inhibitory role of proliferation during cardiac regeneration in zebrafish. Also, in carcinogenesis miR-148 was found to inhibit proliferation by targeting pro-proliferative genes and thereby acting as a tumor suppressor [39], pointing to a similar role and its multifunctional potential. It was also linked to suppress migration and invasion in tumor cells [39], which again plays an important part in heart regeneration [13]. At 160 dpi, miR-148/152 were up-regulated, suggesting a possible stopping mechanism of proliferation processes.

Furthermore, we identified down-regulation of the miR-19 family as key regulation of increased proliferation (cluster 2), with miR-19b being previously observed as down-regulated during zebrafish heart regeneration [24].

Among the known miRNA regulators of zebrafish heart regeneration, we confirmed the important role of miR-133 family, whose depletion has been shown to enhance CM proliferation [23]. MiR-26a has been shown to target cell cycle activators and inhibition stimulates cardiomyocyte proliferation in post-natal mouse hearts [25] underlining its importance in both zebrafish and mice heart regeneration.

For miR-101a, we could corroborate similar expression dynamics in injured zebrafish hearts as previously reported [24]. While miR-101a is initially (1𠄳 dpi) down-regulated post-amputation and shown to enhance CM proliferation [24], it remains down-regulated up to 120 dpi post cryoinjury. At later time points post-amputation (7� dpi), miR-101 becomes up-regulated compared to uninjured hearts, which could be associated with scar tissue removal [24]. In cryoinjured fish, however, up-regulation does not occur until 160 dpi, suggesting a similar yet much slower regenerative response compared to resection experiments (Supplementary Figure S5).

Another known candidate is miR-29b that inhibits genes involved in ECM, confirming results in mice, where down-regulation of the miR-29 family was shown to regulate fibrosis after myocardial infarction [31].

Also, miR-142a has been found to play an important role in vascular integrity and developmental angiogenesis, with overexpression leading to a loss of these functions [40]. Interestingly, miR-142a is up-regulated during heart regeneration, suggesting miR-142 is an antagonist to heart regeneration. However, loss of these processes might also be necessary for the beginning of the response.

Last but not least, we identified miR-16c as novel player of heart regeneration in zebrafish with an association to cell cycle processes.

To research the cross-species potential of the identified candidates in proliferation processes of zebrafish heart regeneration, we compared our results to a differentiating H9c2 cell line that loses its proliferative capability with ongoing differentiation [33]. Interestingly, we found a significant association between the genes of cluster 2 in zebrafish and the down-regulated genes in differentiated rat CM-like cells, with both of them involved in proliferation processes. Furthermore, the target genes of the miRNAs involved in zebrafish heart regeneration were related to the down-regulated genes in the non-proliferative phenotype of the differentiated CM-like cells, indicating similar mechanisms regarding proliferation. However, the miRNAs associated with proliferation in the H9c2 cell line included only the miR-133 family as common player between zebrafish and rat, strengthening its important role in proliferation for both species and both experimental designs. Moreover, the analysis included miR-125a, which was found to inhibit proliferation in C2C12 myoblasts by targeting E2F3 [41], an interaction we could confirm in the H9c2 cell line ( Figure 6 D) and miR-128, which loss was identified to promote CM proliferation and heart regeneration in postnatal mice [42]. The discrepancies between the two studied organisms may be explained by (i) the experimental setups: in vivo vs. in vitro. Most miRNAs that played an important role in the in vivo setting of zebrafish were not expressed in the in vitro setting of the H9c2 cell line. In contrast, all miRNAs were expressed in an in vivo mouse model [25]. (ii) The cell composition: whole hearts vs. myoblasts that differentiate into CM-like cells, (iii) the differences in miRNA binding sites between fish and rat [22] and (iv) the proliferation trigger/stop: cryoinjury vs. differentiation. All of these can result in different miRNA expression, impeding comparative analysis between the two systems. Moreover, in order to overcome these discrepancies, we will carry out further validation experiments of predicted candidate miRNAs in zebrafish, e.g., using morpholino oligos to block miRNA, to understand the actual influence of the predicted miRNAs in cardiac regeneration.


3 Materials and methods

3.1 Datasets

3.1.1 MiRNAs and their target genes

A lot of public databases provide target information for miRNAs, such as TarBase ( Vlachos et al., 2015), TargetScan ( Lewis et al., 2003), PicTar ( Krek et al., 2005), miRanda ( John et al., 2004), DIANA-microT-v4 ( Reczko et al., 2012) and mirDB ( Wong and Wang, 2015). TarBase houses the experimentally validated miRNA-gene interactions, while the others provide computational tools for miRNA target prediction. In order to enable the analysis for large-scale miRNA dataset, we download miRNA target information from two databases, i.e. microRNA.org and mirDB, which adopt miRanda and MirTarget V3 as the prediction tool, respectively. The microRNA.org (released August, 2010) provides computationally predicted targets with good mirSVR scores for both conserved and non-conserved human miRNAs from www.microrna.org, including 1100 miRNAs (here the ‘good mirSVR score’ means the mirSVR value is less than –0.1) while mirDB (Version 5.0 released August, 2014) covers even more miRNAs (2588 human miRNAs). Note that these two databases have different settings of stringency for the predicted targets. Specifically, the average numbers of targets per miRNA in microRNA.org and mirDB are 717 and 4016, respectively, indicating that microRNA.org has a much looser confidence threshold for the identification of targets.

3.1.2 The construction of benchmark set for miRNA subcellular localization

Download all 7449 human miRNA entries with curated subcellular localization from the RNAlocate database, and merge them into 1048 unique miRNAs, as multi-locational miRNAs have multiples records in the database (We check aliases in miRBase.org )

Remove miRNAs that are not covered in microRNA.org and get 813 miRNAs, including 266 mono-locational ones and 547 multi-locational ones

Further remove three locations, i.e. endoplasmic reticulum, extracellular vesicle and nucleolus, because they have too few samples.

Data distribution of the benchmark set of miRNA subcellular localization

Location . # miRNA .
Cytoplasm 206
Microvesicle 329
Mitochondrion 294
Nucleus 319
Circulating 419
Exosome 712
Total label # 2279
Total miRNA# 813
Location . # miRNA .
Cytoplasm 206
Microvesicle 329
Mitochondrion 294
Nucleus 319
Circulating 419
Exosome 712
Total label # 2279
Total miRNA# 813

Data distribution of the benchmark set of miRNA subcellular localization

Location . # miRNA .
Cytoplasm 206
Microvesicle 329
Mitochondrion 294
Nucleus 319
Circulating 419
Exosome 712
Total label # 2279
Total miRNA# 813
Location . # miRNA .
Cytoplasm 206
Microvesicle 329
Mitochondrion 294
Nucleus 319
Circulating 419
Exosome 712
Total label # 2279
Total miRNA# 813

3.1.3 Benchmark datasets of miRNA–disease association

During the last decade, a lot of computational methods for the prediction of miRNA–disease associations have been developed and various public databases and benchmark datasets have merged, such as HMDD ( Li et al., 2013) and miR2Disease ( Jiang et al., 2009), which house the miRNA–disease associations reported in the existing literatures. In order to compare the proposed functional similarity metric with other metrics, we select two widely used benchmark sets, and name them as Data1 and Data2 as follows.

Data1 contains all the records in HMDD database (released on September 2009) created by Wang et al. (2010), including 1616 miRNA–disease associations. In order to generate the MISIM ( mi RNA sim ilarity) scores, Wang et al. (2010) merged the records with the same mature miRNAs (such as hsa-mir-376a-1 and hsa-mir-376a-2), then the set includes 1395 associations, covering 271 miRNAs and 137 diseases. In this set, 267 miRNAs have target information in microRNA.org . Thus, the final dataset used in our experiment includes 1388 associations, 267 miRNAs and 137 diseases.

Data2 was initially collected from HMDD and miR2Disease by Jiang et al. (2010), and contains 270 experimentally verified microRNA–disease associations. Then, Chen and Zhang (2013) used this dataset to validate their proposed methods, PBSI, MBSI and NetCBI, which also rely on MISIM, and they removed 19 miRNAs that are not covered in MISIM, thus remaining 242 miRNA–disease associations, including 99 miRNAs and 51 diseases. In order to compare our result with these three methods, we adopt the reduced set (242 associations) in our experiments, where all the miRNAs are covered in miRGOFS.

3.2 Methods

3.2.1 GO term similarity

As stated in Section 2.1, for IC-based methods, the closeness of a pair of GO terms is normally measured by the IC of their lowest common ancestors (LCAs). In this study, we take not only LCAs but also HCDs (highest common descendants) into consideration, which were often neglected in previous studies. The importance of descendant information is illustrated in Figure 1. For the two pairs, (A1, A2) and (B1, B2), the LCAs are their adjacent ancestor, E, but the local structure below the two pairs are very different. Apparently, A1 and A2 are more similar than B1 and B2.

An example of two pairs of nodes with the same ancestor but different local structure of descendants

An example of two pairs of nodes with the same ancestor but different local structure of descendants

According to the definition of LCA ( Bender et al., 2005), if a node is a common ancestor of x and y, and it is not an ancestor of any other common ancestor of x and y, then the node belongs to the LCA set. The HCD set can be defined analogously. For example, in Supplementary Figure S1 , GO: 0006119 is an HCD of GO: 0006091 and GO: 0016310, and GO: 000977 is also an HCD of them although it is lower than GO: 0006119 in the DAG.

It is straightforward to justify these two rules. Generally, for a pair of GO terms, the more LCAs/HCDs they share, the more similar they are. According to the intersection rule, more LCAs will lead to a smaller set with higher information content while by using the union rule, more HCDs lead to a larger set with lower information content. And, high IC of LCAs and low IC of HCDs result in large similarity score, as shown in Eq. (3).

To illustrate these two set operations, we extract a partial GO DAG from GO database as shown in Supplementary Figure S1 . Take the pair of GO: 0006091 and GO: 0016310 as an example, they are the LCAs for GO: 0042773 and GO: 0009777. The intersection set of their descendants includes GO: 0006119, GO: 0009777, descendants of GO: 0006119 and descendants of GO: 0009777, while the union set of their descendants consists of all descendants of GO: 0006091 and GO: 0016310. Obviously, the intersection set of LCAs would never be empty, which at least contains the query pair of nodes. However, two GO terms may have no HCD at all. In such case, Eq. (3) will degenerate to the sum of the first two terms.

3.2.2 Similarity between miRNAs

The miRGOFS method is implemented in C# with the task parallel library (TPL), which enables multiple threads to run in parallel. And the source code is available at https://github.com/yangy09/MiRGOFS.

3.2.3 Inference rule

Given the correlation values, we use them to draw ROC curve for performance evaluation in the prediction of miRNA–disease association, which has a large number of labels (diseases) while for miRNA subcellular localization, we treat it as a multi-label classification problem, and convert the correlation values into feature vectors as suggested in Zhou et al. (2017). Suppose there are a total of k locations, we generate a k-Dim vector for each miRNA, where the elements of the vector are the correlation values to the k locations respectively. Since the three GO categories, BP, MF and CC, yield different similarities for miRNAs, which may be complementary with each other, we generate a k-Dim vector for each GO category and combine them as a ( 3 × k ) -Dim vector.


How a non-coding RNA encourages cancer growth and metastasis

A mechanism that pushes a certain gene to produce a non-coding form of RNA instead of its protein-coding alternative can promote the growth of cancer, report researchers at the Medical University of South Carolina (MUSC) in an article published online ahead of print on August 21, 2017 by Nature Cell Biology. The non-coding RNA soaks up a microRNA that prevents epithelial-to-mesenchymal transition, one of the key features of tumor development.

From one gene, cells can often produce different forms of RNA. The exact pre-RNA copy of one strand of DNA in a gene must be cut and assembled into its final RNA form, or several forms, in a process known as alternative splicing. Yet while these alternative forms of RNA can encode different proteins, scientists are discovering that many types of RNA do not, instead performing vastly different functions that regulate cell fate and behavior. MicroRNAs, for example, home in on certain protein-coding RNAs and help degrade them.

It is another class, called long non-coding RNAs (lncRNA), that are of particular interest to Philip H. Howe, Ph.D., chair of the Department of Biochemistry & Molecular Biology, and the Hans and Helen Koebig Endowed Chair in Oncology at the MUSC Hollings Cancer Center. Howe and his team found that a pre-RNA for a protein called PNUTS can be alternatively spliced to form a lncRNA that contributes to cancer progression. The PNUTS lncRNA does not encode a protein, but rather soaks up like a sponge a certain microRNA that is usually tasked with preventing epithelial-to-mesenchymal transition, which is a key feature of tumor growth and metastasis.

Howe's group connected a number of dots to explain how this happens. First, they found that breast cancer cells contained more PNUTS lncRNA than normal breast epithelial cells -- a good initial sign that the non-coding RNA was associated with cancer development. Those cells were also more mesenchymal, meaning that they were more likely to form tumors.

They next examined a ribonucleoprotein called hnRNP E1, which binds to pre-RNA and suppresses alternative splicing. Importantly, they knew that TGF-beta, which is released in large amounts by tumor cells, could prevent its binding, potentially allowing alternate forms to be made. Computer models predicted that this ribonucleoprotein could bind to PNUTS pre-RNA on its alternative splicing site. In lung and breast cancer cell lines, specially designed RNA probes confirmed that this exact splicing site was more exposed when the ribonucleoprotein was knocked down and that those cells had more PNUTS lncRNA. When cells were exposed to TGF-beta over time, PNUTS lncRNA was made in increasing amounts. It turns out that the ribonucleoprotein was bound more tightly with the alternative splice site. In normal conditions, this allowed PNUTS protein to be made, but in tumors, the alternative splice site became exposed and more lncRNA was made instead.

Yet the group wanted to confirm exactly how PNUTS lncRNA could encourage tumor formation. Additional computer simulations predicted that, based on its sequence, there were seven potential locations on the PNUTS lncRNA for microRNA-205 to bind. This microRNA binds and destroys a transcriptional regulator called ZEB1 that encourages cells to unstick from one another and spread -- a major step that allows epithelial-to-mesenchymal transition to occur. As predicted, without those potential binding locations, the lncRNA and the microRNA were unable to bind together. This helped cells stick together and spread less, even with TGF-beta added to push them to spread.

It appeared that PNUTS lncRNA was soaking up microRNA-205, which freed up ZEB1 to encourage cells to act more like tumors. To be sure that this was true, the group stuck fluorescent molecules to ZEB1 to track it and found that more of it was present when there were more PNUTS lncRNA.

As expected, preclinical models revealed that breast and lung tumors grew faster and larger when their cells contained more PNUTS lncRNA. By connecting all of the dots, Howe's group had shown that one gene can make either a protein-coding RNA or a long non-coding RNA. With TGF-beta, the lncRNA soaked up microRNA-205 like a sponge, freeing up ZEB to drive epithelial-to-mesenchymal transition, a critical event in the development and spread of cancer.

This is the first study to show exactly how TGF-beta drives cancer through formation of a long non-coding RNA. Howe and his team are conducting experiments to find other such long non-coding RNAs that follow this same mechanism in cancer, with the goal of developing therapies to target them.

"My prediction is that this mechanism didn't evolve to make just one long non-coding RNA," says Howe. "There are probably others that are generated in this same fashion."


Introduction

B-chronic lymphocytic leukemia (CLL), the commonest leukemia of Western countries 1 is characterized by elevated numbers of circulating clonal leukemic B cells. It was classically considered as resulting from accumulation of minimally self-renewing B cells as a consequence of defective apoptosis. 2, 3 Most peripheral CLL cells are typically in G0/G1 tending to be kinetically resting, phenotypically activated and functionally anergic with reduced levels of the B-cell antigen receptor components IgM, IgD and CD79b with most malignant B cells over expressing the antiapoptotic B-cell lymphoma 2 (Bcl-2) protein. 4, 5, 6 However, recent evidence indicates that CLL is not a static disease that results simply from accumulation of long-lived lymphocytes, but as a dynamic process in which cells proliferate and die, often at appreciable levels. 7 CLL shows a highly variable clinical course spanning from rapidly aggressive to completely indolent behaviors. 5 Several biological markers have been described that can predict the clinical course and are useful to segregate patients into two main groups according to the mutational status of the immunoglobulin heavy-chain variable-region gene (IgVH), the expression levels of the ZAP-70 tyrosine kinase and CD38+ expression, 8, 9, 10, 11 as well as the presence of cytogenetic abnormalities, such as 11q or 17p deletions. 12 Patients with clones having unmutated (NM) IgVH genes or with high expression of either CD38 or ZAP-70 and/or 11q or 17p deletions have frequently a rapidly fatal course despite aggressive therapy, whereas patients with mutated (M) clones or low CD38 or ZAP-70 expression and no deleterious cytogenetic abnormalities have an indolent course and frequently die by unrelated diseases. Despite recent advances in CLL biology the molecular mechanism involved in its initiation and progression remains elusive. Recently, a novel class of small noncoding RNAs dubbed microRNAs (miRNAs) have been identified in plants and animals. 13, 14 Mature miRNAs of 19–24 nt in length regulate target gene expression post-transcriptionally through base pairing within 3′-UTR regions of the target messenger RNAs inducing the degradation and/or translational inhibition. 13, 15 Although the biological functions of miRNAs are not yet fully understood, it has been demonstrated that they participate in the regulation of developmental timing, cell death, cell proliferation, fat metabolism, hematopoiesis and patterning of the nervous system. 16 The finding that more than half of the known human miRNAs are located at cancer-associated regions of the genome suggested that miRNAs might play a broad role in cancer pathogenesis. 17 Consistent with this notion is the observed aberrant expression of miRNAs in diverse cancer subtypes, including Burkitt's lymphoma, 18 colorectal cancer, 19 lung cancer, 20 breast cancer 21 and glioblastoma. 22 Increasing experimental evidence implicates miRNAs either as oncogenes or tumor suppressors. 23, 24, 25, 26 Recently, two different strategies were used to profile miRNA expression in CLL. 27, 28 By a microarray approach a specific expression signature consisting of 13 miRNAs, including miR-15a and miR-16, was associated with relevant prognostic factors. 29 Following the report of bcl-2 as a physiological target of these two miRNAs, 30 a role of miR-15a and miR-16 in CLL pathogenesis was also proposed through the upregulation of bcl-2. However, recent reports revealed that significant reduction in miR-15 and/or miR-16 expression levels in CLL was not paralleled by any significant change in bcl-2. 28 In another recent study using a cloning-based strategy, the aberrant expression profile of miRNAs in CLL showed that miR-21, miR-150 and miR-155 were significantly overexpressed. 28 Microarray methods are powerful for global analysis but they have technical limitations when used with miRNAs due to their short size and the sequence similarity between family members. Additionally and most importantly, microarray studies are directed toward reported miRNAs, excluding the analysis of unannotated candidates. Therefore, to gain insight into the role of miRNAs in CLL, we used a cloning-based approach combined with stringent northern blot assays to study small RNAs from CLL patients. Our results show the global fluctuation of miRNA expression levels in CLL cells compared to normal B cells and the identification of five novel miRNA candidates from CLL cells. The observed changes are associated with significant overexpression of miR-155 in combination with a reduction of miR-181a, let-7a and miR-30d. Notably, the differential profile of miRNA expression is associated with molecular prognostic factors and clinical course in CLL.


Abstract

MicroRNA (miRNA) alterations are involved in the initiation and progression of human cancer. The causes of the widespread differential expression of miRNA genes in malignant compared with normal cells can be explained by the location of these genes in cancer-associated genomic regions, by epigenetic mechanisms and by alterations in the miRNA processing machinery. MiRNA-expression profiling of human tumours has identified signatures associated with diagnosis, staging, progression, prognosis and response to treatment. In addition, profiling has been exploited to identify miRNA genes that might represent downstream targets of activated oncogenic pathways, or that target protein-coding genes involved in cancer.


MiRNA Profiling: How to Bypass the Current Difficulties in the Diagnosis and Treatment of Sarcomas

Sarcomas are divided into a group with specific alterations and a second presenting a complex karyotype, sometimes difficult to diagnose or with few therapeutic options available. We assessed if miRNA profiling by TaqMan low density arrays could predict the response of undifferentiated rhabdomyosarcoma (RMS) and osteosarcoma to treatment. We showed that miRNA signatures in response to a therapeutic agent (chemotherapy or the mTOR inhibitor RAD-001) were cell and drug specific on cell lines and a rat osteosarcoma model. This miRNA signature was related to cell or tumour sensitivity to this treatment and might be not due to chromosomal aberrations, as revealed by a CGH array analysis of rat tumours. Strikingly, miRNA profiling gave promising results for patient rhabdomyosarcoma, discriminating all types of RMS: (Pax+) or undifferentiated alveolar RMS as well as embryonal RMS. As highlighted by these results, miRNA profiling emerges as a potent molecular diagnostic tool for complex karyotype sarcomas.

1. Introduction

Sarcomas are rare malignant tumours arising in connective tissues like fat, muscle, bones, and cartilage. According to molecular cytogenetic alterations, sarcomas could be divided into two classes: (1) sarcomas with specific alterations (translocation, oncogenic mutation) including Ewing sarcoma, gastrointestinal stromal tumours, and alveolar rhabdomyosarcoma (2) sarcomas with complex karyotype like leiomyosarcoma, pleomorphic liposarcoma, or osteosarcoma. Osteosarcoma is the most frequent primary malignant bone tumours, characterized by its metastatic potent particularly in lung sites and its resistance to conventional treatments like chemotherapy and radiotherapy [1]. Even if the median survival of osteosarcoma patients has been improved through preoperative administration of chemotherapeutic agents, there are nowadays around 40% poor-responder patients [2]. In fact, osteosarcoma tumours often resist or relapse to presurgical chemotherapeutic treatment, and only few therapeutic options are possible and generally noncurative [3]. A second intensive cure of chemotherapy is currently administered in this case. Thus, it seems essential to develop a diagnosis tool to predict tumour response to chemotherapy to avoid the administration of inefficient drugs. There is also a need for efficient therapeutic alternatives based on the discovery of new targets involved in osteosarcoma tumourigenesis.

Rhabdomyosarcoma (RMS) is one of the most common soft-tissue sarcoma. Three types of RMS are observed: alveolar RMS (20%), embryonal RMS (eRMS, 60%), and pleomorphic RMS (20%). 70% aRMS present a specific translocation of the transcription factor Pax3 at the 3′end of FOXO1, creating a potent transcription factor able to induce myogenesis and survival [4]. 10% aRMS present a translocation of Pax7 with FOXO1 [5]. aRMS are of bad prognosis as compared to eRMS, particularly those with Pax3 fusion gene [6]. Thus, it appears primordial to obtain a diagnosis tool identifying precisely the RMS subtypes, and particularly discriminating Pax-aRMS from eRMS, difficult to separate according to patient survival characteristics, gene expression profiles, and CGH arrays [7].

Micro-RNAs (miRNAs) are promising diagnosis biomarkers with their tissue specificities and their involvement in oncogenic process [8]. miRNAs are noncoding small RNA molecules synthesized from intronic regions with a size range from 16 to 35 nucleotides. They are processed by specific complexes of proteins containing Drosha and Dicer to be matured and finally integrated in RISC complexes [9, 10]. Mature miRNAs match with complementary sequences in messenger RNAs resulting in translation inhibition and accelerated mRNA degradation [11]. miRNA expression levels are characteristic for one tissue to regulate gene expression during growth and development, as it was shown for skeletal tissue and muscle development [12–14]. Their expression is also deregulated in many cancers [15, 16], resulting in a tumour miRNA signature, which could be useful for their classification in line with their tissue origin and molecular alterations [17–19]. Thus, they currently constitute potent biomarkers for cancer diagnosis [18, 20] with their abilities to be detected in patient serum. A noninvasive diagnostic tool based on miRNAs for osteosarcoma could be very useful to adapt chemotherapy protocols to tumour biological specificities.

In this study, we performed the miRNA profiling of sarcoma cell lines, human or rat tumours, to assess if miRNAs could constitute potent biomarkers to surpass the current limitations for rhabdomyosarcoma diagnosis and osteosarcoma treatment. miRNA expression levels were determined using microfluidic cards performing high-throughput TaqMan Low Density Arrays (TLDA), a real-time quantitative PCR (RT-qPCR) assays based on TaqMan technology. We firstly studied the effects of different chemotherapeutic agents on osteosarcoma cell miRNA profiles we observed that these miRNA signatures were cell specific and drug specific. A CGH array of osteosarcoma tumours obtained from a rat model revealed that this miRNA signature, conserved in rat and human cells, was independent of chromosomal rearrangements, suggesting that miRNA profiles were linked to tumour phenotypes rather than to their genetic background. Of great interest, a miRNA signature was identified in rhabdomyosarcoma tumours from patients in accordance with the molecular translocation Pax3 or Pax7. This signature was in fact a potent tool to discriminate alveolar RMS (Pax-) from embryonal RMS, indistinguishable by the molecular techniques currently used. In conclusion, miRNA profiling constitutes a promising technology as an alternative or a partner of usual molecular techniques to overcome the present difficulties in diagnosis and treatment of sarcomas.

2. Experimental Procedures

2.1. Human Rhabdomyosarcoma Tumours

Seventeen patients treated for rhabdomyosarcoma in the Centre Léon Bérard were included in this study. Four frozen tumours and thirteen formalin-fixed paraffin-embedded tumours were obtained from biopsies realized at the diagnosis. Tumour diagnoses were realized by a referent anatomopathologist specialist for this pathology by immunohistochemistry, FISH, and qPCR.

2.2. Cancer Cell Lines

Five cancer cell lines were obtained from ATCC (Manassas, VA, USA): the two human osteosarcoma MNNG/HOS Cl #5 [R-1059-D] (reference CRL-15-47) and Saos-2 (HTB-85) cells, the chondrosarcoma cell line SW1353 (HTB-94) and the two Burkitt lymphoma Daudi (CCL-213) and Namalwa (CRL-1432) cells. Osteosarcoma and chondrosarcoma cells were grown in DMEM (Gibco, Carlsbad, CA, USA), supplemented with 10% decomplemented fetal calf serum (Lonza, Basel, Switzerland), 10 mL penicillin streptomycin (10 U/mL/10 μg/mL, Gibco, Carlsbad, CA, USA), and 5 mL L-glutamin (200 mM Gibco) at 37°C humidified atmosphere containing 5% CO2. Lymphoma cells were grown in RPMI (Gibco, Carlsbad, CA, USA). Cells were exposed to 100 nM RAD-001 (Novartis), 50 μM ifosfamide (ifos, Baxter) or 1 μM cisplatin (CDDP, TEVA) or 100 μM methotrexate (MTX, TEVA) for 24, 48, and 72 h.

2.3. Rat Osteosarcoma Model

Procedures for animal care were performed according to institutional and national guidelines. Animals were anesthetized throughout all surgical and imaging procedures with isoflurane/oxygen (2.5%/2.5%, v/v) (Minerve, Esternay, France). The transplantable orthotopic and metastatic rat osteosarcoma model has been previously described [21–23]. This model mimics its human counterpart in terms of aggressiveness, metastatic spreading and chemoresistance phenotype [21–23]. All the tumours obtained were classified as osteoblastic following histological analyses. Briefly, small tumour fragments (100 mm 3 ) taken from a hyperproliferative osteogenic tumour area were grafted on 3-weeks old immunocompetents Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA, USA). Using a lateral approach, a tumour fragment was placed contiguous to tibial diaphysis after periosteal abrasion then, the cutaneous and muscular wounds were sutured. Fourteen days after tumour transplantation, animals underwent a first 18 F - FDG PET Scan and were randomly assigned to a control group treated with saline solution or a treated group exposed to a subcutaneous dose of 10 mg/kg ifosfamide (ifos, Baxter, Deerfield, IL, USA), 7 days apart (at days 15 and 22 after tumour transplantation). A second 18 F - FDG PET Scan was performed 7 days after the second ifos administration. Animals were sacrificed one week after the end of the treatment. Tumour and normal tissue fragments (muscle, bone, and lung) were collected for RNA extractions.

2.4. RNA Extraction and Quantitative Real-Time PCR

FFPE tumours were lysed for 24 h in ATL buffer (Qiagen, France) supplemented with proteinase K (Qiagen) at 60°C in rotative agitation after different washes with toluene, ethanol, and tris/EDTA in this order. Total RNA was extracted from tumour or cell pellets using a single phenol/chloroform extraction protocol with Trizol, according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). Five hundred nanograms of total RNA were subjected to the microfluidic PCR technology performed by Applied Biosystems (Foster City, CA, USA). In brief, RNA was reversed transcribed, using multiplexed specific looped miRNA primers from the Taqman MicroRNA Reverse Transcription kit. The second step consists in a real-time quantitative PCR on TLDA: RT products are introduced through microchannels into miniature wells that are preloaded with dehydrated specific primers and probes. Recently, Applied biosystems released the second version of TLDA, consisting of two cards A and B. Analyses were performed for 377 miRNAs on card A and 290 on card B.

2.5. PCR Data Normalization

For each miRNA, the threshold cycle (Ct) was calculated by the ABI 7900 Sequence Detection System software (plate by plate manual Ct analysis with a threshold at 0.25 and automatic baseline). All further data manipulations were done using R scripts. A cutoff of 32 was applied to discard the late Ct values, except for RMS analysis. Around 60% of miRNAs passed the filtering criteria and were used for further analysis. For each TLDA, quality controls were performed on the raw data by checking internal controls and using box plot and scatter plot diagrams. Samples with any kind of problems were discarded so they would not introduce bias during the following normalization procedures. We tested different methods of normalization since the recommended “pseudo” normalization factor mammU6 plotted in each card was not stably expressed in our different samples. Normalization with the two most stable miRNAs identified by GeNorm was not optimal too. Finally, a global normalization by the median was chosen for its reliability over experiments. Tissues included in a given analysis were treated altogether, the normalization procedure being applied separately for the two types of card, A and B. Distribution of normalized data was checked with box plots and correlation plots. The following formula was used to correct Ct values of every card:

Through this approach, the new median value shared by all samples can be considered as a sort of perfect “virtual housekeeping gene”. Therefore, the standard ΔΔCt method can be used to determine the relative quantities (RQ) as follows:

For the ΔΔCt calculation, it was more relevant for the statistical analyses to use the mean of all ΔCt obtained across samples for each miRNA, instead of using the ΔCt of a reference sample

2.6. miRNA Target Predictions

We compiled 4 databases to determine miRNA targets: TargetScan 5.1, MiRanda, PICTAR, and the miRbase databases. These databases search the presence of conserved 8mer and 7mer sites on the 3′UTR parts of messenger RNA that match the seed region of each miRNA. It also predicts the efficacy of targeting for each matching site. We created our own database which regrouped each miRNA with the geneID of all their protein targets, for rat and human. We only conserved couples miRNA/geneID present in two databases at least.

2.7. miRNA-Regulated Cell Signalling Pathways Predictions

We used the “G-language microarray” web application, which allows the mapping of molecular dataset onto “Kyoto Encyclopedia of Genes and Genomes” (KEGG) pathway maps [24]. We first input miRNA-targeted proteins of interest and the sum of RQ values for all miRNAs that regulate these proteins, contained between 1 and 50 the software then generates KEGG data to create FLASH graphics of cell signalling pathways in which proteins are involved. The colour intensity of a highlighted protein varies with the strength of its regulation by miRNAs.

2.8. Proliferation Assay

Cells were plated in 96 well plates at 5000 cells/well and exposed to 100 nM RAD-001, 50 μM ifosfamide, 100 μM methotrexate, or 1 μM cisplatin or not (NT). Cell growth was measured 24, 48, and 72 h later with 20 μL Cell Titer Glo luminescent reagent (Promega, Madison, WI, USA) for 10 min. Luminescence was recorded using a Microbeta reader (PerkinElmer, Fremont, CA, USA).

2.9. Western Blot

Pelleted cells were resuspended in lysis buffer (Tris 50 mM pH 7.4, NaCl 250 mM, EDTA 5 mM, NaF 50 mM, Triton X-100 0.1%, orthovanadate 1 μM) plus protease inhibitors for 30 min on ice. After a centrifugation at 14000 rpm for 10 min, supernatants were boiled for 5 min in Laemmli sample buffer (Biorad, Hercules, CA, USA). Analysis of protein content was performed on 4%–12% gradient gel. After electrophoretic separation, 30 μg proteins were electrotransferred on a polyvinylidene difluoride membrane (Immobilon P, Millipore corp., Bedford, MA, USA). The membrane was then blocked for 1 h at room temperature with blocking agent 0.2% in PBS/Tween 0.1%, probed overnight with a primary rabbit antibody against the protein of interest, and finally revealed with a secondary antirabbit antibody HRP conjugated (Upstate Biotechnology, Lake Placid, NY, USA) and ECL Advance system (GEhealthcare, Chicago, IL, USA). Primary antibody used was obtained from Cell Signaling (New England Biolabs, Beverly, MA, USA) used at 1/1000. The β actin was used as a reference (Sigma).

2.10. CGH Array

Oligonucleotide-based microarray analysis was performed using a custom-designed, 244K-feature whole-rat genome microarray manufactured by Agilent Technologies (Santa Clara, CA). Genomic DNA labeling, array hybridization, and washing were performed as specified by the manufacturer (Agilent Technologies). Results of aberration calls consisting of three or more consecutive oligos were then displayed using custom oligonucleotide CGH analysis software (Genespring).

2.11. Statistical Analysis

Normalized RQ data were directly input into the TIBCO Spotfire DecisionSite for Functional Genomics analysis software. We performed unsupervised hierarchical clustering to classify samples by groups. The selection of miRNAs useful to predict tumour response to treatment was statistically realized using ANOVA tests with P values of .05 at least. Results were verified through supervised hierarchical clustering.

Data from miRNA lists of interest were then used as variables in a three-dimensional principal component analysis (PCA) performed with R 2.9.0 package to demonstrate their capabilities to distinguish types of tumours. PCA supplies a simplified three-dimensional picture to our multivariate dataset of miRNA RQ values. By mathematical combination of values according to their strength, three principal components are created that represent as much as possible the variability of the data. Thus, tumours possess three new coordinates in a three-dimensional space. According to their localization in this space, tumours form groups, and their subtypes can be predicted.

3. Results

3.1. miRNA Signatures of Osteosarcoma Cell Lines

In our recent study published in International Journal of Cancer, we showed that the two osteosarcoma Saos-2 and CRL-15-47 (15-47) cells mimic the biological response of human osteosarcoma and tumours obtained from a rat model. In fact, we identified in an osteosarcoma rat model a panel of 61 miRNAs discriminating tumours with a good response to ifosfamide from those with a bad response [25]. On the basis of this signature, we realized a principal component analysis allowing predicting tumour response. In this PCA diagram, we could notice that the Saos-2 cells were predicted as sensitive to ifosfamide contrary to 15-47 cells (Figure 3(b) [25]), according the results obtained by a proliferation assay (Figure 6(a) [25] and Figure S1). This was confirmed by a PCA analysis realized with the miRNA signature identified in human tumours (Figure S2). We so considered that these two cell lines were an interesting model to study the importance of miRNAs in cell response to treatment and to identify new therapeutic strategies.

3.2. miRNA Signatures of Human Cancer Cell Lines

We firstly performed a preliminary miRNA profiling on different cell models to compare the miRNA profiles of osteosarcoma cells used in our laboratory to perform in vitro experiments, Saos-2 and 15-47 cells, with the chondrosarcoma cells SW-1353 (chondro) and the Burkitt lymphoma Daudi and Namalwa cells. In a previous study, we identified 61 miRNAs involved in osteosarcoma cell response to treatment [25]. We only conserved these miRNAs to realize an unsupervised hierarchical clustering with the five cancer cell lines. As shown in Figure 1(a), this miRNA signature was representative of the two human osteosarcoma cell lines, since these two cells clustered together independently but closely to the chondrosarcoma cells. These three cell lines were classed in a distinct group from the two lymphoma cells Daudi and Namalwa. This confirmed that each cancer cell line presents a miRNA signature in accordance with their origin, as shown by others [15, 16].


(a)
(b)
(c)
(a)
(b)
(c) Cancer cell miRNA signatures were consistent with their tissue origin and with their sensitivity to ifosfamide. (a) This unsupervised hierarchical clustering only conserved the 61 miRNAs which discrimated osteosarcoma cells according to their response to treatment. Osteosarcoma cell lines clustered together near chondrosarcoma cells and independently to the lymphoma Daudi and Namalwa cells. Each row represents the relative levels of expression for each miRNA, and each column shows the expression levels for each sample. The red or green colour indicates relatively high or low expression, respectively, while grey squares indicate no expressed miRNA. (b) and (c) miRNA profiles after exposure to 50 μM ifosfamide for 24 h. (b) This unsupervised hierarchical clustering only conserved the 61 miRNA differently expressed in osteosarcoma cells according to their sensitivity to ifos following an ANOVA (

). (c) This supervised hierarchical clustering conserved miRNAs differently expressed in cells according to their response to treatment following an ANOVA (

3.3. miRNA Profiles in Response to Chemotherapeutic Agents Were Cell Specific

Then, we assessed if miRNA profiles were specifically modified in response to chemotherapy. We chose to expose osteosarcoma and lymphoma cells to ifosfamide, an alkylating chemotherapeutic agent currently used for paediatric osteosarcoma. A proliferation assay based on ATP measurement showed that the only Saos-2 cell line was moderately sensitive to 50 μM ifosfamide after 48 h exposure (proliferation inhibition around 30%) (Figure S1). Based on this observation, we decided to expose these cells to 50 μM ifosfamide for 24 h to realize miRNA profiling. On the basis of the panel of 61 miRNAs identified in our previous study [25], osteosarcoma cells were markedly different from lymphoma cells, confirming that miRNA profiles were cell specific as shown by the unsupervised hierarchical clustering in Figure 1(b). We could notice that Saos-2 cells present a unique miRNA signature in which the majority of miRNAs were overexpressed (in red in Figure 1(b)). A supervised hierarchical clustering realized following an ANOVA between the Saos-2 sensitive cells versus the resistant cells revealed that they effectively clustered according to their sensitivity to ifos: Saos-2 in one hand, independently to 15-47 cells and both lymphoma cells (Figure 1(c)). We confirmed this observation with the other chemotherapeutic agent ciplatin. As previously, cells were classified according to their susceptibility to CDDP on the supervised hierarchical clustering in Figure S3A (ANOVA ): the 15-47 and Namalwa cells, sensitive to CDDP based on the proliferation assay in Figure S3B, clustered together, independently to Daudi and Saos-2 cells refractory to this treatment.

3.4. Osteosarcoma Cell miRNA Profiles Were Specific of Each Chemotherapeutic Agent

Thus, since miRNA signatures of untreated as well as treated cells were cancer specific, we assessed if each chemotherapeutic drug induced a different miRNA profile in a same cell. As suggested previously for osteosarcoma cells, cisplatin and ifosfamide exposure resulted in quite different miRNA profiles. After a statistical analysis with an ANOVA , we only found two dicriminating miRNAs common to both miRNA signatures induced by ifos and CDDP in the two cell lines (Figure S3). In this context, we test a third cytotoxic agent currently administered in osteosarcoma pathology, the methotrexate. As shown in the unsupervised hierarchical clustering in Figure 2, only conserving the 61 miRNAs of interest for osteosarcoma response, as explained above, the miRNA signature in the two osteosarcoma cells Saos-2 and 15-47 strongly differed from those observed for ifsofamide and cisplatin. It is important to note that a majority of these miRNAs were overexpressed in both cell lines in response to MTX. This was relevant with their sensitivity to MTX as shown in the proliferation assay in Figure 2(b). In brief, it seems that discriminating miRNAs were generally overexpressed in the cells after exposure to a cytotoxic agent, to which they were sensitive, as it was also shown for ifosfamide in the Saos-2 cells (Figure 1(b)). This also confirmed that miRNAs predicting cell response to a treatment differed according to the drug.


(a)
(b)
(a)
(b) miRNA expression profiles osteosarcoma cells were specific for each chemotherapeutic agent. (a) This unsupervised hierarchical clustering conserved the miRNAs identified as discriminating for ifosfamide response after removing the miRNAs whose expression depends on the cell types. Each row represents the relative levels of expression for each miRNA, and each column shows the expression levels for each sample. The red or green colour indicates relatively high or low expression, respectively, while grey squares indicate no expressed miRNA. (b) Cell growth was measured by the Cell Titer GloLuminescent assay as described in Section 2.6 24, 48, and 72 h after exposure to 100 μM methotrexate. Results were represented as the mean % of proliferation normalized to untreated cells of two independent experiments realized in duplicate.

(a)
(b)
(a)
(b) mTOR inhibition by RAD-001 in chondrosarcoma resulted in cell proliferation inhibition. (a) Cell growth was measured by the Cell Titer GloLuminescent assay as described in Section 2.6 24, 48, and 72 h after exposure to 100 nM RAD-001. Results were represented as the mean % of proliferation normalized to untreated cells of two independent experiments realized in duplicate. A Fisher test was realized, NS corresponds to “nonsignificant”. (b) Western blot analysis of RAD-001 effects on the mTOR pathway in the chondrosarcoma and 15-47 osteosarcoma cells exposed or not (NT) to doxorubicin (dox). 30 μg protein extracts were analysed by Western blot with antibodies 1/1000 against actors of the mTOR pathway (phosphorylated form or not) (Cell signalling, Beverly MA).

On the basis of these preliminary in vitro results, we could suggest that miRNA profiles, due to their drug specificity, could be a potent tool to predict a cancer cell response to a treatment. Since osteosarcoma is currently resistant to conventional treatments, the prediction of its response to one agent could be a progress for this pathology.

3.5. Osteo- and Chondrosarcoma Cell Response to the mTOR Inhibitor RAD-001

As highlighted by these previous data, we were able to classify and predict osteosarcoma cell response to chemotherapy. Our algorithms were not only interesting for chemotherapeutic agents but also promising to identify new targeted therapies to encounter osteosarcoma resistance. Thus, we tested a potent drug for skeletal sarcoma treatment, which inhibits the pro-oncogenic protein mTOR, called RAD-001 (Everolimus, Novartis). mTOR is often aberrantly activated in cancers and, in particular in chondrosarcoma [26] and osteosarcoma [27]. mTOR signalling has been described as implicated in tumour development, metastasis, and drug resistance [28, 29] thus, mTOR targeting successfully inhibits tumour growth and renders them sensitive to conventional treatments [30, 31]. RAD-001, acting in a similar manner than rapamycin through the inhibition of mTORC1 complexes, is currently tested in various clinical trials for renal cell carcinoma (RECORD program), advanced papillary tumours (RAPTOR), metastatic neuroendocrine tumours (RAMSETE), or breast cancers (BOLERO).

Thus, we performed in vitro experiments on chondrosarcoma and osteosarcoma cells with 100 nM RAD-001. The Saos-2 and chondrosarcoma cell proliferation was reduced of 40% following exposure to RAD-001 during 72 h contrary to 15-47 cell growth (Figure 3(a)). In parallel, we realized Western blot with RAD-001 on chondrosarcoma and osteosarcoma cells concerning the major actors of the mTOR cell signalling pathways. This revealed that the mTOR pathway was inhibited by RAD-001 in chondrosarcoma cells contrary to 15-47 cells, in particular eIF4G and p70 S6 kinase whose phosphorylation level was decreased (Figure 3(b)).

Thus, we analysed if the miRNA signatures of these cells were different and could explain their differential response to RAD-001. We performed a supervised hierarchical clustering between untreated Saos-2, chondrosarcoma and 15-47 cells following an ANOVA with . This clustering revealed that 16 miRNAs discriminated the chondrosarcoma and Saos-2 cells in one hand and the 15-47 cells in the other hand (Figure 4). Except miR-146, Saos-2 and chondrosarcoma overexpressed these contributory miRNAs.


miRNA expression profiles of osteosarcoma and chondrosarcoma cells were consistent with their sensitivity to the mTOR inhibitor RAD-001. This hierarchical clustering only conserved miRNA differently expressed in tumours according to their response to treatment following an ANOVA (

Thereafter, as we have explained in our previous study on osteosarcoma [25], miRNA profiling constitutes a potent tool to identify miRNA-targeted cell signaling pathways through an in silico approach. In our case, we searched if these miRNAs shown as differently expressed in cells according to their response to RAD-001 potentially target the mTOR signalling pathway. We created a database, as described in Section 2.6 which determine the predicted targets for these miRNAs described in the miRbase. Then, we summed up the RQ values for each miRNA in the Saos-2 and chondrosarcoma cells sensitive to RAD-001 and concatenated with the geneID of their protein targets. We finally inserted these data in the G-language microarray web application, which connects miRNA targets according to their involvement in similar KEGG pathways, the mTOR pathway in this case. As shown in Figure 5, the mTOR pathway is targeted by these miRNAs and particularly its downstream proteins implicated in VEGF signaling and autophagy processes, in particular RICTOR, ATG1, and HIF-1a. Thus, the Saos-2 and chondrosarcoma cells overexpressed miRNAs that potentially inhibit mTOR signalling. Inhibition of these miRNAs through the use of Locked Nucleic Acid (LNA) and qPCR measurement of RICTOR, ATG1, and HIF1a could confirm this concept.


The discriminating miRNAs interfered with the mTOR pathway. Proteins in yellow, orange, and red colours represent targets of miRNAs a yellow square represents a weak repression, while red represents the maximal repression green squares were not targeted.

(a)
(b)
(c)
(d)
(e)
(f)
(a)
(b)
(c)
(d)
(e)
(f) CGH analysis of six tumours from an osteosarcoma rat model. Here, we represent six chromosomes among the twenty + X chromosomes present in rat genome. All the analyses were performed with the same untreated bone sample as a reference.

To resume, miRNAs constitute potent biomarkers to determine the susceptibility to a treatment and could be very useful to identify new therapeutic targets as an alternative of chemotherapy for chondrosarcoma and osteosarcoma often refractory to this treatment. In the next steps, we assessed if these observations were relevant in vivo, with a model of rat osteosarcoma and with patient samples.

3.6. Predictive miRNA Signature of a Rat Osteosarcoma Model Was Probably Not Related to DNA Aberrations

As described in other studies realized by members of our team [21, 22, 25], we possess a rat osteosarcoma model mimicking the human pathology concerning aggressiveness, chemoresistance and the apparition of lung metastases (see Section 2.6). The treatment of animals with ifosfamide results in two groups, the good versus the bad or moderate responders, in a proportion closer to that observed for patients. By miRNA profiling, we were able to distinguish tumours sensitive to ifosfamide from those refractory to this drug and above all to predict the response of untreated tumours with ten miRNAs through the use of statistical algorithms created in our lab [25]. Following these interesting data, we would like to confirm that this miRNA signature was specific of tumour response to treatment and not related to different tumour genetic backgrounds. We thus realized an analysis in CGH array with the same tumours used for miRNA profiling. We analysed two tumours of each type, untreated, treated with ifosfamide and good responder, or treated with ifosfamide and bad responder, as compared to the same untreated bone sample, the reference tissue in CGH analysis. The majority of chromosomal aberrations observed in CGH array was common to untreated tumours and treated tumours, regardless of their response to treatment (Figure 6). The few different abnormalities were essentially linked to individual tumour biological specificities.

We compiled all abnormalities and verified in our “home-made” database if any miRNA, identified as discriminating of tumour response, was located in these DNA regions. Interestingly, this in silico analysis also revealed that neither miRNA nor gene were present in the few differential aberrations observed in these tumours, in particular in the chromosome 4 (Figure S4), suggesting that the different miRNA profiles were rather linked to tumour response to treatment and not due to upstream chromosomal rearrangements. Although we could not rule out that some trans-acting proteins could be deregulated consequently to these aberrations, this suggested that all tumours were homogeneous and that an increase in some miRNAs in sensitive tumours were due to their upregulation and not to a genetic amplification.

It seems that molecular diagnosis based on miRNA profiling highlights the tumour behaviour, that is, in response to a treatment, and thus a phenotype rather than a genotype contrary to CGH array. These two molecular techniques could be a couple of choice to improve the care of patients with pathologies currently hardly to diagnose.

3.7. Rhabdomyosarcoma miRNA Profiles Were Correlated to their Histological Subtypes

Finally, to corroborate the previous idea considering that miRNA profiling could be very helpful for uncertain diagnoses, we performed the miRNA profiling of rhabdomyosarcoma samples. In fact, we recently showed that miRNA profiling was reliable for osteosarcoma diagnosis on 29 formalin-fixed paraffin embedded (FFPE) biopsies of patients [25]. Based on the expression level of a panel of five miRNAs, we successfully separated good responders from bad responders to treatment. So, we assessed if our TLDA platform was also competitive for RMS diagnosis. We obtained seventeen tumours including alveolar RMS patients, (Pax3+) (3 patients) or Pax7+ (2), embryonal RMS patients (6) and negative fusion aRMS (6). All these tumours were diagnosed through the use of immunohistochemistry, FISH and qPCR, which were validated by a referent anatomopathologist (Table S5). A supervised hierarchical clustering on rhabdomyosarcoma tumours following an ANOVA with a P value between the four types of RMS, revealed that tumours clustered according to their molecular alterations Pax3/FOXO1, Pax7/ FOXO1 or no translocation, on the basis of the expression level of 10 miRNAs (Figure 7(a)). (Pax+) tumours, particularly those (Pax3+) overexpressed all these miRNAs. Then, we performed a statistical analysis with these ten miRNAs based on Principal Component Analysis, a method which allows studying the variability between a set of variables. This consists of assigning a new system of three coordinates to each contributory miRNA by a mathematical procedure. Then, RQ values of each miRNA are adjusted for each tumour by the new coefficients obtained previously and summed up. Thus, a 3-dimension PCA diagram was realized with the three new coordinates for each tumour (Figure 7(b)). Through this mathematical representation, we could distinguish (Pax+) from fusion negative aRMS and eRMS. eRMS also constitutes an independent group with a high value of component 2 (represented in the y-axis on Figure 7(b)). The fusion negative aRMS constitute a separate group even if some samples were difficult to classify in accordance with their uncertain diagnosis. Even if the number of samples was low for each subset, a statistical analysis showed a significant P value between (Pax3+) and (Pax7+) and between (Pax+) and (Pax−) tumours, 0.05 and 0.0005 respectively (Figure S6).


(a)
(b)
(a)
(b) Rhabdomyosarcoma miRNA signatures were consistent with their molecular alterations. (a) This supervised hierarchical clustering only conserved miRNA differently expressed in the different subtypes of RMS following an ANOVA (

We showed that miRNA profiling was a potent tool to discriminate fusion negative aRMS from embryonal RMS. miRNAs could be useful biomarkers to improve the diagnosis of this type of RMS, since fusion negative aRMS are currently molecularly indistinguishable from eRMS [7].

4. Discussion

miRNA signatures are observed for many types of cancers, that is, sarcoma [19], breast and prostate cancers [18, 32]. These signatures constitute potent diagnosis and prognosis tools for chronic lymphocytic leukemia [33], colon adenocarcinoma [34], or lung cancers [35]. Here, we showed that osteosarcoma cell lines also expressed miRNA patterns different from those of chondrosarcoma and lymphoma cells (Figure 1(a)) and which allow us to discriminate cell response to chemotherapeutic treatment (Figures 1(b), 2, and S3). In addition, osteosarcoma miRNA signatures were cell and drug specific (Figure 2(a)). This drug specificity of osteosarcoma has also been observed by Song et al. with U2-OS osteosarcoma tumour xenografts, in which different miRNAs were deregulated in response to the chemotherapeutic agents doxorubicin, cisplatin, and ifosfamide only 3 miRNAs were commonly found deregulated in response to all drugs [36]. With their specificity, miRNAs constitute promising biomarkers to anticipate the tumour response to a treatment of interest. As we have recently shown, through miRNA profiling, we were able to predict osteosarcoma tumour response to chemotherapy for rat tumours as well as for patient FFPE biopsies [25]. Here, we showed that miRNA profiles of osteosarcoma cells were in accordance with their response to the mTOR inhibitor, RAD-001 (Figures 3 and 4). The miRNAs deregulated in response to this drug in sensitive cells, effectively targeted the mTOR pathway, in particular the downstream proteins eIF4G and p70 S6 kinase (Figure 3(b)), and potentially RICTOR, ATG1 and HIF1a, which might be validated by qPCR analysis (Figure 5).

In brief, miRNAs appeared very useful for the identification of new exciting therapeutic approaches through the targeting of some miRNA protein targets or some miRNAs involved in tumour development themselves. In future, we would like to confirm the implication of these miRNAs in treatment response in vitro through the use of miRNA mimics or inversely of Locked Nucleic Acid (LNA) against these miRNAs. As mentioned in this study, we possess an interesting in vitro osteosarcoma model, on which we could test the miRNA functionality in the presence of the different drugs used in this work. Following the validation of miRNA involvement in vitro, we would also test these mimics or LNAs in vivo in the model of rat osteosarcoma. This approach has been successfully employed in rhabdomyosarcoma through the conditional expression of miR-206 in mice [37] and could become a potent therapeutic strategies [38].

In addition to the identification of new targets, miRNA also constitute an interesting alternative to the conventional molecular technologies routinely used for cancer diagnosis. In fact, osteosarcoma present complex karyotypic alterations rendered them difficult to diagnose with current diagnostic methods, like CGH array [39]. With the rat osteosarcoma model, we confirmed that tumours presented numerous long chromosomal aberrations (Figure 6). These abnormalities were generally common to all tumours, regardless to their susceptibility to treatment and neither miRNAs of interest nor genes were located in these regions (Figure S4). Even if some proteins involved in the regulation of miRNA expression (trans-acting factors or epigenetic regulating factors) could be deregulated following these mutations, the miRNA profiles observed in rat tumours might be correlated to the effects of the cytotoxic drugs on the miRNA machinery and no to upstream DNA rearrangements. Even if miRNAs could be submitted to epigenetic regulation like methylation or acetylation, this only concerns 5% to 10% miRNAs, and we could consider that this process is minor for the miRNA signature of osteosarcoma tumours and cells based on 61miRNAs [40, 41]. Enthusiastically, our work is the first suggesting that miRNA signatures were not correlated to DNA amplifications, as it was observed for neuroblastoma [42] or in mixed lineage leukemia [43]. Although our cohort was not fully satisfying, it appeared that miRNA profiling could predict tumour response to treatment by reflecting tumour biological specificities and not genotypic characteristics. This work also highlights miRNA measurement as an interesting partner to CGH array in the case of pathologies with unstable karyotypes. In the same way, Selvarajah et al. was the first to suggest a combination of CGH array and interphase FISH to better understanding osteosarcoma pathogenesis [44].

miRNA patterns were not only related with osteosarcoma phenotypic properties but also with rhabdomyosarcoma histological subtypes. By miRNA profiling, we were able to discriminate the different subtypes of rhabdomyosarcoma: Pax3+ or Pax7+ or fusion negative, classically difficult to diagnose by histological analysis (Figure 7(a)). This miRNA pattern was unique since all miRNAS identified as discriminating are no or weakly described in the literature. Very interestingly, on the basis of their miRNA profiles, our algorithms allow us to discriminate embryonal RMS from fusion negative aRMS (Figure 7(b)). It was in agreement with the work of Wachtel et al. identifying different expression profiles linked to aRMS (Pax+), fusion negative aRMS, and eRMS [45].

Altogether, it seems that miRNA measurement is advantageous for sarcoma with complex karyotype, since fusion negative RMS, similarly to osteosarcoma, are characterized by a complex karyotype linked to allelic imbalance, loss of heterozygoty and heterogeneous gene expression profiles. Although the molecular classification of fusion negative RMS is always controversial, our work corroborates the study of Davicioni et al. suggesting that Pax/FOXO1 dictates a specific expression signature in RMS by oligonucleotide microarray expression profiling [46]. Inversely, this differs from the recent work of Williamson suggesting that fusion negative alveolar rhabdomyosarcoma is difficult to distinguish from embryonal rhabdomyosarcoma concerning patient survival characteristics, gene expression profiles, and CGH arrays [7]. In fact, our work and theirs were not totally contradictory, since they only focused on genomic analysis. As suggested for osteosarcoma, miRNA patterns reflect the phenotypic tumour properties rather than its genetic and could be a promising alternative for RMS diagnosis to surpass the current limitations of molecular analysis combined to traditional histopathology.

Thus, it seems that miRNA profiling could be very useful for osteosarcoma and rhabdomyosarcoma diagnosis. Here, we showed that on the basis of ten miRNAs, we were able to separate the different subtypes of RMS. We have previously suggested that a panel of five miRNAs was statistically sufficient to distinguish the potent response of osteosarcoma patients to treatment [25]. The TLDA technology presents numerous advantages including its need for few amount of total RNA and the possible analysis of FFPE samples, as it was previously shown by others [47–49]. This method is especially useful to detect circulating miRNAs in patient serum, an emerging field these two past years [50, 51]. A blood-based molecular diagnosis tool through miRNA profiling from patient serum could be a major advance for osteosarcoma, requiring a biopsy for its diagnosis, which could result in a secondary amputation.

Altogether, these promising results open up the way to a new diagnosis tool based on miRNA for osteosarcoma as well as rhabdomyosarcoma, which could improve patient survival in both cases through the prediction of patient response to chemotherapy and the precise identification of RMS subtypes, respectively.

Abbreviations

CDDP:Cisplatin
Chondro:Chondrosarcoma cells
Ct:Threshold cycle
Dox:Doxorubicin
FFPE:Formalin-fixed paraffin embedded
ifos:Ifosfamide
KEGG:Kyoto encyclopedia of genes and genomes
LNA:Locked nucleic acid
NT:Non treated
PCA:Principal component analysis
RAD:RAD-001
RMS:Rhabdomyosarcoma
RQ:Relative quantity
RT-qPCR:Real-time quantitative PCR
TLDA:Taqman low density array
15-47:Human osteosarcoma cells CRL-15-47.

Acknowledgments

This work has been supported by the Ligue Nationale contre le Cancer (grants from the Ain department), the Institut National du Cancer, and the Conticanet Consortium. The authors acknowledge Imaxio, which realized the CGH analysis.

References

  1. L. B. Rozeman, A. M. Cleton-Jansen, and P. C. W. Hogendoorn, “Pathology of primary malignant bone and cartilage tumours,” International Orthopaedics, vol. 30, no. 6, pp. 437–444, 2006. View at: Publisher Site | Google Scholar
  2. S. Ferrari and E. Palmerini, “Adjuvant and neoadjuvant combination chemotherapy for osteogenic sarcoma,” Current Opinion in Oncology, vol. 19, no. 4, pp. 341–346, 2007. View at: Publisher Site | Google Scholar
  3. A. J. Chou and R. Gorlick, “Chemotherapy resistance in osteosarcoma: current challenges and future directions,” Expert Review of Anticancer Therapy, vol. 6, no. 7, pp. 1075–1085, 2006. View at: Publisher Site | Google Scholar
  4. F. G. Barr, N. Galili, J. Holick, J. A. Biegel, G. Rovera, and B. S. Emanuell, “Rearrangement of the PAX3 paired box gene in the paediatric solid tumour alveolar rhabdomyosarcoma,” Nature Genetics, vol. 3, no. 2, pp. 113–117, 1993. View at: Publisher Site | Google Scholar
  5. R. J. Davis, C. M. D'Cruz, M. A. Lovell, J. A. Biegel, and F. G. Barr, “Fusion of PAX7 to FKHR by the variant t(113)(p36q14) translocation in alveolar rhabdomyosarcoma,” Cancer Research, vol. 54, no. 11, pp. 2869–2872, 1994. View at: Google Scholar
  6. P. H. B. Sorensen, J. C. Lynch, S. J. Qualman et al., “PAX3-FKHR and PAX7-FKHR gene fusions are prognostic indicators in alveolar rhabdomyosarcoma: a report from the Children's Oncology Group,” Journal of Clinical Oncology, vol. 20, no. 11, pp. 2672–2679, 2002. View at: Publisher Site | Google Scholar
  7. D. Williamson, E. Missiaglia, A. de Reyniès et al., “Fusion gene-negative alveolar rhabdomyosarcoma is clinically and molecularly indistinguishable from embryonal rhabdomyosarcoma,” Journal of Clinical Oncology, vol. 28, no. 13, pp. 2151–2158, 2010. View at: Publisher Site | Google Scholar
  8. A. Esquela-Kerscher and F. J. Slack, “Oncomirs—microRNAs with a role in cancer,” Nature Reviews Cancer, vol. 6, no. 4, pp. 259–269, 2006. View at: Publisher Site | Google Scholar
  9. G. Hutvágner and P. D. Zamore, “A microRNA in a multiple-turnover RNAi enzyme complex,” Science, vol. 297, no. 5589, pp. 2056–2060, 2002. View at: Publisher Site | Google Scholar
  10. Y. Lee, C. Ahn, J. Han et al., “The nuclear RNase III Drosha initiates microRNA processing,” Nature, vol. 425, no. 6956, pp. 415–419, 2003. View at: Publisher Site | Google Scholar
  11. J. G. Doench and P. A. Sharp, “Specificity of microRNA target selection in translational repression,” Genes and Development, vol. 18, no. 5, pp. 504–511, 2004. View at: Publisher Site | Google Scholar
  12. Z. Li, M. Q. Hassan, M. Jafferji et al., “Biological functions of miR-29b contribute to positive regulation of osteoblast differentiation,” Journal of Biological Chemistry, vol. 284, no. 23, pp. 15676–15684, 2009. View at: Publisher Site | Google Scholar
  13. T. Sugatani and K. A. Hruska, “MicroRNA-223 is a key factor in osteoclast differentiation,” Journal of Cellular Biochemistry, vol. 101, no. 4, pp. 996–999, 2007. View at: Publisher Site | Google Scholar
  14. K. Chen and N. Rajewsky, “Natural selection on human microRNA binding sites inferred from SNP data,” Nature Genetics, vol. 38, no. 12, pp. 1452–1456, 2006. View at: Publisher Site | Google Scholar
  15. J. Lu, G. Getz, E. A. Miska et al., “MicroRNA expression profiles classify human cancers,” Nature, vol. 435, no. 7043, pp. 834–838, 2005. View at: Publisher Site | Google Scholar
  16. S. Volinia, G. A. Calin, C. G. Liu et al., “A microRNA expression signature of human solid tumors defines cancer gene targets,” Proceedings of the National Academy of Sciences of the United States of America, vol. 103, no. 7, pp. 2257–2261, 2006. View at: Publisher Site | Google Scholar
  17. G. A. Calin, M. Ferracin, A. Cimmino et al., “A microRNA signature associated with prognosis and progression in chronic lymphocytic leukemia,” New England Journal of Medicine, vol. 353, no. 17, pp. 1793–1801, 2005. View at: Publisher Site | Google Scholar
  18. A. Israel, R. Sharan, E. Ruppin, and E. Galun, “Increased microRNA activity in human cancers,” PLoS One, vol. 4, no. 6, Article ID e6045, 2009. View at: Publisher Site | Google Scholar
  19. S. Subramanian, W. O. Lui, C. H. Lee et al., “MicroRNA expression signature of human sarcomas,” Oncogene, vol. 27, no. 14, pp. 2015–2026, 2008. View at: Publisher Site | Google Scholar
  20. J. C. Wittig, J. Bickels, D. Priebat et al., “Osteosarcoma: a multidisciplinary approach to diagnosis and treatment,” American Family Physician, vol. 65, no. 6, pp. 1123–1136, 2002. View at: Google Scholar
  21. A. Dutour, D. Leclers, J. Monteil et al., “Non-invasive imaging correlates with histological and molecular characteristics of an osteosarcoma model: application for early detection and follow-up of MDR phenotype,” Anticancer Research, vol. 27, no. 6B, pp. 4171–4178, 2007. View at: Google Scholar
  22. A. Dutour, J. Monteil, F. Paraf et al., “Endostatin cDNA/cationic liposome complexes as a promising therapy to prevent lung metastases in osteosarcoma: study in a human-like rat orthotopic tumor,” Molecular Therapy, vol. 11, no. 2, pp. 311–319, 2005. View at: Publisher Site | Google Scholar
  23. M. Allouche, H. G. Delbruck, and B. Klein, “Malignant bone tumours induced by a local injection of colloidal radioactive 144Cerium in rats as a model for human osteosarcomas,” International Journal of Cancer, vol. 26, no. 6, pp. 777–782, 1980. View at: Google Scholar
  24. K. Arakawa, N. Kono, Y. Yamada, H. Mori, and M. Tomita, “KEGG-based pathway visualization tool for complex omics data,” In Silico Biology, vol. 5, no. 4, pp. 419–423, 2005. View at: Google Scholar
  25. A. Gougelet, D. Pissaloux, A. Besse et al., “miRNA profiles in osteosarcoma as a predictive tool for ifosfamide chemotherapeutic treatment response,” International Journal of Cancer. In press. View at: Google Scholar
  26. R. E. Brown, “Morphoproteomic portrait of the mTOR pathway in mesenchymal chondrosarcoma,” Annals of Clinical and Laboratory Science, vol. 34, no. 4, pp. 397–399, 2004. View at: Google Scholar
  27. Q. Zhou, Z. Deng, Y. Zhu, H. Long, S. Zhang, and J. Zhao, “mTOR/p70S6K signal transduction pathway contributes to osteosarcoma progression and patients' prognosis,” Medical Oncology, vol. 27, no. 4, pp. 1239–1245, 2010. View at: Publisher Site | Google Scholar
  28. B. H. Jiang and L. Z. Liu, “Role of mTOR in anticancer drug resistance: perspectives for improved drug treatment,” Drug Resistance Updates, vol. 11, no. 3, pp. 63–76, 2008. View at: Publisher Site | Google Scholar
  29. X. Wan, A. Mendoza, C. Khanna, and L. J. Helman, “Rapamycin inhibits ezrin-mediated metastatic behavior in a murine model of osteosarcoma,” Cancer Research, vol. 65, no. 6, pp. 2406–2411, 2005. View at: Publisher Site | Google Scholar
  30. I. Beuvink, A. Boulay, S. Fumagalli et al., “The mTOR inhibitor RAD001 sensitizes tumor cells to DNA-damaged induced apoptosis through inhibition of p21 translation,” Cell, vol. 120, no. 6, pp. 747–759, 2005. View at: Publisher Site | Google Scholar
  31. S. Mabuchi, D. A. Altomare, D. C. Connolly et al., “RAD001 (Everolimus) delays tumor onset and progression in a transgenic mouse model of ovarian cancer,” Cancer Research, vol. 67, no. 6, pp. 2408–2413, 2007. View at: Publisher Site | Google Scholar
  32. S. Ambs, R. L. Prueitt, M. Yi et al., “Genomic profiling of microRNA and messenger RNA reveals deregulated microRNA expression in prostate cancer,” Cancer Research, vol. 68, no. 15, pp. 6162–6170, 2008. View at: Publisher Site | Google Scholar
  33. G. A. Calin, M. Ferracin, A. Cimmino et al., “A microRNA signature associated with prognosis and progression in chronic lymphocytic leukemia,” New England Journal of Medicine, vol. 353, no. 17, pp. 1793–1801, 2005. View at: Publisher Site | Google Scholar
  34. A. J. Schetter, S. Y. Leung, J. J. Sohn et al., “MicroRNA expression profiles associated with prognosis and therapeutic outcome in colon adenocarcinoma,” Journal of the American Medical Association, vol. 299, no. 4, pp. 425–436, 2008. View at: Publisher Site | Google Scholar
  35. S. L. Yu, H. Y. Chen, G. C. Chang et al., “MicroRNA signature predicts survival and relapse in lung cancer,” Cancer Cell, vol. 13, no. 1, pp. 48–57, 2008. View at: Publisher Site | Google Scholar
  36. B. Song, Y. Wang, Y. Xi et al., “Mechanism of chemoresistance mediated by miR-140 in human osteosarcoma and colon cancer cells,” Oncogene, vol. 28, no. 46, pp. 4065–4074, 2009. View at: Publisher Site | Google Scholar
  37. R. Taulli, F. Bersani, V. Foglizzo et al., “The muscle-specific microRNA miR-206 blocks human rhabdomyosarcoma growth in xenotransplanted mice by promoting myogenic differentiation,” Journal of Clinical Investigation, vol. 119, no. 8, pp. 2366–2378, 2009. View at: Publisher Site | Google Scholar
  38. R. N. Veedu and J. Wengel, “Locked nucleic acids: promising nucleic acid analogs for therapeutic applications,” Chemistry and Biodiversity, vol. 7, no. 3, pp. 536–542, 2010. View at: Publisher Site | Google Scholar
  39. J. A. Fletcher, M. C. Gebhardt, and H. P. Kozakewich, “Cytogenetic aberrations in osteosarcomas: nonrandom deletions, rings, and double-minute chromosomes,” Cancer Genetics and Cytogenetics, vol. 77, no. 1, pp. 81–88, 1994. View at: Publisher Site | Google Scholar
  40. L. Han, P. D. Witmer, E. Casey, D. Valle, and S. Sukumar, “DNA methylation regulates microRNA expression,” Cancer Biology and Therapy, vol. 6, no. 8, pp. 1284–1288, 2007. View at: Google Scholar
  41. Y. Saito and P. A. Jones, “Epigenetic activation of tumor suppressor microRNAs in human cancer cells,” Cell Cycle, vol. 5, no. 19, pp. 2220–2222, 2006. View at: Google Scholar
  42. I. Bray, K. Bryan, S. Prenter et al., “Widespread dysregulation of MiRNAs by MYCN amplification and chromosomal imbalances in neuroblastoma: association of miRNA expression with survival,” PLoS One, vol. 4, no. 11, Article ID e7850, 2009. View at: Publisher Site | Google Scholar
  43. S. Mi, Z. Li, P. Chen et al., “Aberrant overexpression and function of the miR-17-92 cluster in MLL-rearranged acute leukemia,” Proceedings of the National Academy of Sciences of the United States of America, vol. 107, no. 8, pp. 3710–3715, 2010. View at: Publisher Site | Google Scholar
  44. S. Selvarajah, M. Yoshimoto, O. Ludkovski et al., “Genomic signatures of chromosomal instability and osteosarcoma progression detected by high resolution array CGH and interphase FISH,” Cytogenetic and Genome Research, vol. 122, no. 1, pp. 5–15, 2008. View at: Publisher Site | Google Scholar
  45. M. Wachtel, M. Dettling, E. Koscielniak et al., “Gene expression signatures identify rhabdomyosarcoma subtypes and detect a novel t(22)(q35p23) translocation fusing PAX3 to NCOA1,” Cancer Research, vol. 64, no. 16, pp. 5539–5545, 2004. View at: Publisher Site | Google Scholar
  46. E. Davicioni, F. G. Finckenstein, V. Shahbazian, J. D. Buckley, T. J. Triche, and M. J. Anderson, “Identification of a PAX-FKHR gene expression signature that defines molecular classes and determines the prognosis of alveolar rhabdomyosarcomas,” Cancer Research, vol. 66, no. 14, pp. 6936–6946, 2006. View at: Publisher Site | Google Scholar
  47. A. B. Hui, W. Shi, P. C. Boutros et al., “Robust global micro-RNA profiling with formalin-fixed paraffin-embedded breast cancer tissues,” Laboratory Investigation, vol. 89, no. 5, pp. 597–606, 2009. View at: Publisher Site | Google Scholar
  48. K. R. M. Leite, J. M. S. Canavez, S. T. Reis et al., “miRNA analysis of prostate cancer by quantitative real time PCR: comparison between formalin-fixed paraffin embedded and fresh-frozen tissue,” Urologic Oncology: Seminars and Original Investigations. In press. View at: Publisher Site | Google Scholar
  49. A. Liu, M. T. Tetzlaff, P. VanBelle et al., “MicroRNA expression profiling outperforms mRNA expression profiling in formalin-fixed paraffin-embedded tissues,” International Journal of Clinical and Experimental Pathology, vol. 2, no. 6, pp. 519–527, 2009. View at: Google Scholar
  50. X. Chen, Y. Ba, L. Ma et al., “Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases,” Cell Research, vol. 18, no. 10, pp. 997–1006, 2008. View at: Publisher Site | Google Scholar
  51. P. S. Mitchell, R. K. Parkin, E. M. Kroh et al., “Circulating microRNAs as stable blood-based markers for cancer detection,” Proceedings of the National Academy of Sciences of the United States of America, vol. 105, no. 30, pp. 10513–10518, 2008. View at: Publisher Site | Google Scholar

Copyright

Copyright © 2011 Angélique Gougelet et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


Watch the video: Biogenesis of miRNAs and mode of action (January 2023).