Pilot Study of Whole Blood MicroRNAs as Potential Tools for Diffuse Low-Grade Gliomas Detection

Abstract

Earlier diagnosis and longitudinal monitoring of diffuse low-grade gliomas (DLGG) increase overall survival by maximizing surgery efficacy and optimizing time for an adjuvant treatment when resection is incomplete. Presently, only imaging permits the non-invasive detection and monitoring of DLGG, but it lacks sensitivity. Measure of circulating microRNAs levels could represent a non-invasive alternative. We hypothesized that slow-growing DLGG induce overtime a systemic reaction impacting blood cells microRNA profiles, while the intact blood–brain barrier restricts the passage of tumor microRNAs into bloodstream. In 15 DLGG patients and 15 healthy controls, expression levels of 758 microRNAs were measured by the TaqMan OpenArray RT-qPCR platform, on preoperative whole blood, containing both cell-free and blood cells microRNAs. Normalized data were computed by a Student t test with a p value threshold allowing a 10% rate of false positive. Statistical analysis retained fifteen microRNAs, all overexpressed in patients. MiR-20a, miR-106a, miR-20b, and miR-93 belong to clusters genetically related. As miR-223 and miR-let7e, they target the transcription factor STAT3. MicroRNA expression levels were not correlated to preoperative tumor volume. A signature composed of miR-93, miR-590-3p, and miR-454 enabled to nearly perfectly separate patients from controls. Our study performed on a homogeneous cohort was designed accordingly to DLGG particularities and provided the first microRNAs signature proposal. Functional convergence on STAT3 and overexpression of miR-223, factors respectively involved in myeloid-derived suppressor cells and granulocytes, argued for a systemic peripheral response. Overexpressed microRNAs and tumor volume were uncorrelated, making a tumor origin elusive.

Electronic supplementary material

The online version of this article (doi:10.1007/s10571-017-0536-7) contains supplementary material, which is available to authorized users.

Introduction

Diffuse low-grade gliomas (DLGG) are slow-growing premalignant brain tumors, occurring mostly in young patients usually with an active life. These tumors evolve inexorably towards a higher grade of malignancy.

Precocious therapeutic management appears to be effective for increasing overall survival of these patients especially because it improves the rate of achievable total/supratotal resections that is a favorable factor in prognosis (Capelle et al. 2013). To this end, a preventive screening proposal in the relevant age population might be justified.

But, because of their extensive and infiltrative growth in the brain parenchyma, complete resection of the tumor is often impossible, so that a post-operative tumor residue will be left from which the tumor will steadily and slowly re-grow. To obtain the best chance of limiting and delaying the malignant transformation of residual tumor, we need to optimize time for an adjuvant treatment. For this purpose, close monitoring of the evolution of the residual tumor is essential (Duffau and Taillandier 2015). As DLGG are indolent tumors in the first phase of their course, systematic monitoring of patients will last several years.

Up-to-date imaging (as MRI or PET) is the only non-invasive method to presuppose the diagnosis of glioma, then confirmed by pathological examination, and to allow a longitudinal monitoring of residual tumor. So that imaging is an important part both in the diagnosis and therapeutic management of DLGG. Nevertheless MRI lacks sensitivity for accurately determining the actual extent of DLGG, particularly on the periphery of the lesion, where tumor cells density becomes lower than the sensitivity threshold of the technique because of the highly migratory nature of gliomas (Pallud et al. 2010).

In recent years the study of circulating biomarkers emerges as a powerful non-invasive way of detecting cancers and their evolution over time. The advantages of these biological methods compared to MRI are their cost-effectiveness, ease of implementation, and possibility of a frequent use. Among circulating biomarkers already exploited in oncology, microRNAs appear to be objectively attractive. They are secreted by many kinds of cells to serve as messengers between origin cells and recipient cells. Thereby microRNAs are retrieved in biologic fluids in a variety of forms. In blood, microRNAs are contained in blood cells or are cell-free, associated to cargo proteins, to HDL or included in extracellular vesicles. They are transcription regulators whose expression level is tightly regulated by the cancer disease progression suggesting that microRNAs expression profiles could specifically characterize a tumor. Furthermore microRNAs patterns contain more diagnostic information than single biomarkers.

To our knowledge, no studies on circulating biomarkers were specifically devoted to low-grade gliomas. DLGG have specific characteristics that will determine the nature of the most relevant circulating biomarkers. These are slow-growing tumors that could induce, remotely, over time, a systemic reaction. Moreover, before malignant transformation, the blood–brain barrier remains intact and exchanges with the blood still highly regulated.

The assessment of microRNAs in whole blood takes into account both circulating cell-free microRNAs and those contained in blood cells. Cell-free circulating microRNAs released in bloodstream could originate from tumor cells while development of the tumor could peripherally induce changes in microRNA blood cells. Opening the scope of the analysis beyond plasma biomarkers may be particularly suited to specific features of DLGG.

Confirming the interest of achieving analysis of all circulating microRNAs in the broadest way (cell-free and cellular microRNAs), case-control studies have been successfully conducted on whole blood to search differential microRNA profiles between patients and healthy peoples in other cancer diseases (Hausler et al. 2010; Schultz et al. 2014).

Accordingly, we conducted a case-control study on 15 DLGG patients, without any histological sign of malignant transformation, and 15 healthy controls sex and age matched.

We performed the analysis of 758 microRNAs expression levels on whole blood.

Materials and Methods

Blood Samples

The blood was collected in PAXgene Blood RNA tubes (PreAnalytiX, GmbH Qiagen, DE), the day before surgery for the patients.

Each participant in the study gave informed consent. This study was approved by the local ethics committee and registered at the Agence Nationale de Sécurité des Médicaments (France) under the number 121951B-11.

MicroRNA Analysis

Total RNA extractions with a specific enrichment in small RNA species was performed with the PAXgene blood microRNAs kit (PreAnalytiX, GmbH Qiagen, DE) according to the manufacturer’s instructions. Concentration and integrity of the isolated RNA were assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). All samples had RNA integrity number between 6.6 and 8.1.

Reverse Transcription (RT) and Pre-amplification

MicroRNA expression was measured with the Life Technology TaqMan OpenArray^R system, a qPCR-based platform (Thermo Fischer Scientific, Waltham, MA USA).

Total RNA of each sample was reverse transcribed into single-stranded cDNA using separately Megaplex looped-RT primers of Pool A and Pool B corresponding to 377 human microRNAs each. The RT reaction was performed as follows: 40 cycles of a three-step program (2 min at 16 °C, 1 min at 42 °C, and 1 s at 50 °C) then enzyme was inactivated for 5 min at 85 °C.

Secondly, 2.5 µl of cDNA targets was pre-amplified with each Megaplex PreAmp Pool A and Pool B primers. The reaction was done using the following conditions: three single steps (10 min at 95 °C, 2 min at 55 °C, and 2 min at 72 °C) and 12 cycles of a two-steps program (15 s at 95 °C, 4 min at 60 °C) then enzyme was inactivated for 10 min at 99.9 °C.

TaqMan OpenArray MicroRNA Assay

We used Standard Human TaqMan^R OpenArray^R MicroRNA Panel and QuantStudio™ 12 K Flex Real-Time PCR chips (3072-well per chip hosting each a single RT-qPCR reaction with TaqMan Probes and specific PCR primers). The post–pre-amp dilution mixtures were loaded onto the chips using the AccuFill system. Three different samples were loaded on one chip using the standard MicroRNA.edt file. Then PCR was run on QuantStudio™ 12 K Flex real-time instrument. The thirty samples of the study (15 patients and 15 controls) were randomly distributed on ten chips.

Data were processed, checked, and exported with Expression suite V1.0.2 software.

Preoperative Tumor Volume Assessment

Topography of the tumor was analyzed on a preoperative MR image (T1-weighted and spoiled gradient images obtained before and after gadolinium enhancement, and T2/FLAIR-weighted images). Preoperative tumor volumes were calculated using semiautomatic segmentation on dedicated software (Myrian, Intrasense France).

Data Preprocessing and Statistical Analysis

Following Livak and Schmittgen, when technical replicates are available, arithmetic mean of raw intensity value was computed and used for further analyses (Livak and Schmittgen 2001). In order to choose the most stable housekeeping gene, the NormFinder algorithm implemented in the Bioconductor package NormqPCR was applied. The RNU48 gene was used to normalize data. A two-tailed Student t test was applied on the obtained 2^−∆Ct values to compute raw significances. A p value threshold has been set so that it allows a 10% rate of false positives (FDR). This threshold was determined using mock groups: random labels (DLGG vs. control) were associated to samples and p values were computed using the same t test as the one for real data. All microRNAs with a p value lower than a chosen threshold will then be considered as false positives. Hence, it is possible to compute FDR value for any given threshold and to choose the threshold allowing an appropriate FDR. This process is performed 1000 times (1000 sets of mock labels are generated) so that results are not dependent on the labeling. A similar method has been used by Hausler et al. (2010).This non-parametrical assessment of FDR is more appropriate than a classical adjustment of p values, because t tests corresponding to different microRNAs do not meet the independence assumption underlying methods such as Benjamini–Hochberg adjustment.

In order to establish possible correlations between the expression level of the five microRNAs with the weakest FDR and either the preoperative volumes or the blood cell counts, we applied the standard correlation test using Student’s t distribution.

Finally, in the aim to combine information provided by microRNAs whose p value is lower than the determined threshold, we designed a multivariate signature. It uses a classification tree (Breiman et al. 1984) such as implemented in the rpart function of the rpart R package. Input data are the selected microRNAs, the chosen splitting index is Gini (Breiman et al. 1984), and the minimum number of observations that must exist in a node in order for a split to be attempted is set to 5.

Results

Patient Characteristics

The criteria of inclusion were a pathological diagnosis of WHO grade II glioma lacking any histological signs suggestive of a trend towards malignant transformation, a first surgery, and absence of any preoperatively treatment. Otherwise the patients of the study did not suffer from other central nervous system diseases or systemic immunological diseases that were clinically detectable at the time of inclusion.

Fifteen patients meeting the aforementioned criteria and operated consecutively at our institution between January 2014 and August 2015 were included in the study, whatever the molecular status of the tumors was. Tumors reassessed in accordance with the WHO 2016 classification include five oligodendrogliomas and ten astrocytomas, reflecting the classical distribution of low-grade gliomas according to the histological type. All the tumors were IDH1 R132H mutated while 33% of them were 1p 19q co-deleted.

Fifteen healthy subjects controls, matched for sex and age (with a maximum tolerance of 10 years) with patients, were also included in the study.

The sex ratio M/F was 7/8 in both populations. The median age was 32.5 years (26–53 of ranges) for patients and 37.5 years (26–62 of ranges) for controls.

Tumor and clinical patient characteristics are summarized in Table 1.

Table 1.

Tumor and clinical characteristics of patients

Case	Age at surgery (years); gender	Time between diagnosis and surgery (months)	Preoperative volume (cc)	Tumor location	Histological subtype	Co-deletion 1p19q	IDH mutation
1	40; F	2	88	L FrI	A	No	Yes
2	35; M	6	63	R FrTI	A	No	Yes
3	36; M	4	60	R FrTI	OD	Yes	Yes
4	30; M	12	90	R P	OD	Yes	Yes
5	26; F	5	12	R Fr	A	No	Yes
6	32; F	18	12	L Fr	OD	Yes	Yes
7	33; F	6	82	R I	A	No	Yes
8	33; M	6	56	L FrTI	A	No	Yes
9	45; F	10	32	L FrTI	A	No	Yes
10	54; M	4	113	L FrTI	A	No	Yes
11	41; F	7	117	L Fr	OD	Yes	Yes
12	42; F	2	84.5	L FrTI	A	No	Yes
13	32; F	3	26.5	R I	OD	Yes	Yes
14	62; M	13	24	R P	A	No	Yes
15	35; M	4	30	L Fr	A	No	Yes
Median	37.5	5.5	60
Range	26–62	2–18	12–117
Proportion of “Yes”						33%	100%

L left, R right, Fr frontal, I insular, P parietal, T temporal, A astrocytoma, OD oligodendroglioma

Differentially Expressed MicroRNAs in DLGG Patients and Healthy Subjects

We assessed the expression levels of 758 different human microRNAs. We obtained results for 667 of them. We retained only the microRNAs with a maximum of two missing data. Thereby we subjected 291 microRNAs to statistical analysis. When a false-positive ratio (FDR) of 9% is considered, 5 microRNAs are found differentially expressed between cases and healthy controls (Fig. 1). By increasing the FDR threshold to 10%, we retrieved 15 microRNAs differentially expressed, all overexpressed in patients. The characteristics of these 15 microRNAs are summarized in Table 2. The level of overexpression of these microRNAs is not influenced by the 1p 19q molecular status of the tumors otherwise all IDH1 mutated (Supplemental Table 1).

Fig. 1.

Boxplots of the distribution of the normalized relative expression (∆Ct values) obtained for the 5 microRNAs (FDR 9%) according to their group (DLGG vs. controls). Each group consisted of 15 subjects

Table 2.

DLGG overexpressed microRNAs characteristics

miRNA	Chromosomal location	Expression fold change	p value	AUC (area under curve)
hsa.mir 93	7	2.37	0.001790	0.83556
hsa.mir 16	13	2.05	0.002560	0.80444
hsa.mir 190b	1	2.50	0.003243	0.83590
hsa.mir 223	X	2.08	0.003354	0.82667
hsa.mir 590.3p	7	3.86	0.003868	0.813333
hsa.mir 422a	15	2.05	0.005314	0.80444
hsa.mir 20b	X	2.52	0.006279	0.80889
hsa.mir 222	X	2.09	0.007817	0.77333
hsa.let.7e	19	2	0.008341	0.76444
hsa.mir 20a	13	2.5	0.008735	0.76889
hsa.mir 454	17	2.35	0.008934	0.75111
hsa.mir 26b	2	1.99	0.009243	0.78222
hsa.mir 106a	X	2.09	0.009376	0.79111
hsa.mir 886.5p	5	1.92	0.009662	0.76
hsa.let.7 g	3	2.13	0.009908	0.78222

Ten of them have already been linked to gliomas setting, while six of them were mentioned in the blood cells related literature. Five microRNAs detected by our study had never previously been mentioned in gliomagenesis (miR-190b, miR-422, miR-886-5p, miR-let7e, and miR-20b), contrary to what might be expected from microRNAs distinguishing patients from healthy subjects. One possible explanation is that our series was composed entirely of WHO grade II gliomas, while the results published to date relate to series overwhelmingly composed of higher grade gliomas. It is now well admitted that de novo glioblastomas and low-grade gliomas arise from distinct molecular pathways that should influence microRNAs profiles (Ohgaki and Kleihues 2011; Cohen and Colman 2015).

Nevertheless, we isolated miR-222, miR-223, miR-106a, miR-20a, and miR-454, already identified as circulating markers in high-grade gliomas.

Correlations Between Preoperative Tumor Volume, Blood Cells Counts, and MicroRNAs Expression Levels

Preoperative tumor volume, measured over the last MRI performed before surgery, corresponds to the largest measurable mass of tumor cells. Preoperative median volume was 60 cc, with a range from 12 to 117 cc.

To investigate if there was an influence of the tumor cells amount on the circulating microRNAs, we looked for a correlation between the preoperative tumor volume and the expression levels of the five most differentially overexpressed microRNAs in patients (miR-16, miR-223, miR-190b, miR-590-3p, miR-93). We found no significant correlation.

On the other hand, the participation of microRNAs from the blood cells in the total circulating microRNAs profile is presumably important. Blood cells belong to functionally distinct cell types characterized by specific microRNAs profiles. Therefore, in order to reveal a possible bias on microRNAs profiles, we sought, in patients only, a correlation between the expression level of the most significant microRNAs and platelet, leukocyte, and red blood cells counts. We found no significant correlation.

Diagnostic Tree with a Three MicroRNAs Signature

By applying the classification tree algorithm described in the Methods section to the normalized relative expression, we obtained a tree implying three microRNAs: miR-93, miR-590-3p, and miR-454. It nearly perfectly separates the 30 observations according to their disease status (DLGG vs. control) with only two errors (one in each group).

Hence, using such a multivariate signature allows to strongly enhancing the discriminative power of each microRNA taken separately. These three microRNAs have already been identified separately in high-grade gliomas. Their ability to distinguish DLGG patients from controls, when all three are used together as a signature, may reflect a feature of low-grade gliomas (Fig. 2). Of course, the tree model will have to be validated using more samples in both groups. For the moment, this signature can only claim to be considered as a proof of concept of the relevance of blood microRNA profiles in distinguishing people with a DLGG.

Fig. 2.

Classification tree allowing to separate DLGG patients from controls according to miR-93, miR-590-3p, and miR-454. At each level, the current dataset is split according to the value of each observation for the considered microRNAs with regards to the threshold indicated below the microRNAs name. Hence, samples having a normalized relative expression higher than 1.25 for miR-93 fall in the “YES” box, else they fall in the “NO” box. The tree leaves (subsets not to be split anymore) are shown in a green box when most of samples falling into it belong to the control group and in a red box when most of samples falling into it belong to the DLGG group. The two misclassified samples are indicated in orange (Color figure online)

MicroRNAs Overexpressed in Patients are Genetically and Functionally Clustered

Four microRNAs of our study belong to genetically related paralog clusters, i.e., the miR 17-92 cluster (miR-20a), the miR-106a-363 cluster (miR-106a and miR-20b), and miR-106b-25 cluster (miR- 93). Paralog clusters are characterized by a high conservation across species of the organization and content of microRNAs sequences. These three paralog clusters are characterized by the same seed sequence specific of the miR-17 microRNAs family. MicroRNAs bind specifically to their mRNA target by this seed sequence on a recognition site located in the 3′UTR part and exert their transcriptional regulation. Although the seed sequence is not the only determinant of the specificity of transcriptional regulation, it is a decisive prerequisite key. MicroRNAs that possess the same seed sequence will recognize and regulate the same mRNA targets. MiR-17 microRNAs paralog clusters are known to be operatively connected and redundant (Ventura et al. 2008).

Furthermore, the four previously mentioned microRNAs plus miR-222, miR-223, and miR-let7e target STAT3 mRNA thus forming a functional cluster (Cao et al. 2013; Zhang et al. 2011; Rao et al. 2015). STAT3 is a transcription factor involved in cancer and immunity (Yu et al. 2009). The signaling pathway of STAT3 has been widely implicated as a molecular mechanism underlying the maintenance of high-grade gliomas. In the DLGG, the expression of the phosphorylated STAT3 protein is very low as demonstrated by the study of Abou-Ghazal et al. performed on WHO grade II astrocytomas (Abou-Ghazal et al. 2008). Furthermore, Doucette et al. have studied the role of the STAT3 protein in the malignant transformation of gliomas. By inducing the expression of the STAT3 gene in the glioneuronal progenitor cells in mice, the authors show that this alone was insufficient to induce the emergence of malignant gliomas. But on the other hand, the STAT3 protein played a potentiating role on the transforming effect of PDGFB, a well-described initiator of DLGG (Doucette et al. 2012).

Thus seven, i.e., almost half, of the microRNAs differentially expressed between controls and patients are connected by already demonstrated genetic and/or functional links making sense in this selection.

Discussion

Relevance of the Newly Selected MicroRNAs in Gliomagenesis and Blood Cells Physiology

Among the ten microRNAs related to gliomas, miR-93, miR-222, miR-16, and miR-106a were repeatedly incriminated in tumor tissue and in glioma specific cell lines.

Several studies attribute an oncomir role to miR-93 and miR-222. In tumor tissue, their expression level is higher than in normal brain parenchyma and correlated to tumor grade (Jiang et al. 2015; Yang et al. 2015). Functional studies, performed in vitro on cell lines, report for both miR-93 and miR-222 a promoter role in angiogenesis, and a proliferative role (Jiang et al. 2015) while miR-222 induced in addition glial tumor cells invasion (Yang et al. 2015).

For miR-16 and miR-106a, the results reported in literature are more ambiguous. The results of functional studies on various cell lines pointed to a tumor-suppressing role for miR-16, exerted via an inhibition of invasion (Yang et al. 2014; Wang et al. 2014). In the case of miR-106a, functional studies in vitro on cell lines give contradictory results: an oncomir role in the study of Wang et al. by increasing the invasiveness of CD 133+ stem cells (Wang et al. 2015) and a suppressor tumor role, in other studies, by decreasing the proliferation of glial cells (Yang et al. 2011).

MiR-20a has already been described as an oncomir in gliomas, associated to miR-16 (Malzkorn et al. 2010) or to miR-106a (Wang et al. 2015). Our study highlights again this functional connection.

MiR-590-3p and miR-26b roles in gliomas are much less documented. Nevertheless, they are both considered as tumor suppressor microRNAs (Pang et al. 2015; Jiang et al. 2016).

Nevertheless, we isolated miR-222, miR-223, miR-106a, miR-20a, and miR-454, already identified as circulating markers in high-grade gliomas. An overexpression in plasma for miR-454-3p or in serum for miR-106a, miR-20a, and miR-222 was reported in patients with gliomas of various grades compared to healthy control subjects (Shao et al. 2015; Zhi et al. 2015; Zhang et al. 2016). On whole blood, miR-223 tended towards a statistical significant overexpression in 20 patients with glioblastomas (Roth et al. 2011).

When working on whole blood, measure of microRNAs includes also those contained in blood cells. Intriguingly miR-222, miR-223, miR-106a, and miR-20a are strongly associated with blood cells. Indeed, miR-223 is the most abundant microRNAs in platelets (Landry et al. 2009) and it plays an important role in other myeloid lineages. MiR-222, miR-106a, miR-93, and miR-20a are found in the monocyte macrophage lineage, while miR-20a, miR-106a, and miR-20b appear to be involved in the differentiation of Th17 cells (Honardoost et al. 2015).

Tumor Origin of Up-Regulated Circulating MicroRNAs Remains Elusive

To our knowledge, the correlation of the expression level of circulating microRNAs before surgery with preoperative tumor volume has been rarely investigated. Two studies compared the expression levels of circulating microRNAs in gliomas before and after surgery. The observed decrease in expression levels of these microRNAs after tumor removal was considered as an evidence for a tumor origin (Shao et al. 2015; Zhi et al. 2015). But surgery is likely to induce per se consequences on circulating microRNA patterns unrelated to tumor cell mass decrease. Conversely, MRI non-invasive measure of preoperative tumor volume does not interfere downstream with the microRNAs profiles.

MicroRNAs derived qualitatively from the tumor or the healthy cells are not different. Accordingly, only a modulation of their expression levels in patients by the amount of tumor cells could be expected. Presence of microRNAs in the bloodstream corresponds to two processes: tumor cells secret microRNAs to act as messengers or microRNAs are passively released during the breakdown of tumor cells. But the preoperative tumor volume and the bloodstream expression level of the five more differentially expressed microRNAs in patients were not correlated. This lack of correlation suggests that circulating tumor microRNAs were either undetectable under the method sensitivity threshold or absent. This finding blurs the possible tumor origin of the overexpressed microRNAs in DLGG.

Cell-free tumor DNA (ctDNA) found in plasma or serum is another circulating marker increasingly used in patients with cancer. It takes advantage of its specificity in indicating a tumor origin because it carries genetic alterations of tumor cells that occur at negligible frequencies in normal cell populations. In a comparative study of fifteen types of neoplasms, Bettegowda et al. reported the lack of sensitivity of ctDNA assessment in gliomas. CtDNA was detected in less than 10% of patients that appeared very disappointing for the detection of gliomas (Bettegowda et al. 2014). In patients harboring gliomas of different grades, Boisselier et al. reported a specificity of 100% but a mean sensitivity of 60% for the plasma detection of cell-free IDH1 mutated circulating DNA. When focusing on DLGG group, the sensitivity was lower (37.5%) and detection of mutated circulating DNA restricted to tumors with the larger preoperative volume (Boisselier et al. 2012).

The issue of circulating markers in gliomas remains challenging to date unlike in other tumors such as lung cancers. These difficulties could be related to the particular location of gliomas isolated from the bloodstream by the blood–brain barrier. The decrease in sensitivity in the detection of cell-free DNA demonstrated by the Boisselier et al. study when passing from high-grade gliomas to low-grade gliomas suggests that brain isolation induced by the blood–brain barrier may be an explanatory factor. It is noteworthy that in DLGG the blood–brain barrier remains intact unlike in glioblastomas.

Experience with ctDNA suggests that the sensitivity of analytical methods will be a major challenge in identifying a circulating marker suitable for DLGGs insofar as circulating markers directly derived from tumor cells are searched. Consequently, one way of resolving this difficulty would be to take an interest in a circulating signature that would not be restricted to markers directly derived from the tumor but rather would be composed of markers induced at a distance and specifically by the growing of the tumor.

Some Arguments for a Tumor-Induced Signature in Circulating Blood Cells

DLGG are slow-growing tumors inducing brain plasticity (Duffau 2005). Tumor invasion triggers major neuronal reorganizations both in the tumor surroundings and in remote brain areas. Furthermore, the DLGG slow growth can maintain thanks to the establishment of an immune-tolerance state. Multiple sclerosis presents some similarities to DLGG since important overhauls of the white matter and a deregulation of the immune response are also experienced. A case-control study, performed on whole blood cells, report that the microRNA profiles are different between multiple sclerosis patients and healthy subjects (Junker et al. 2011). In DLGG, if neuronal reshaping and immune tolerance have repercussions on the peripheral blood microRNAs signatures is largely yet unknown, but could be hypothesized.

The immune escape of tumor involves a population of immature myeloid cells with the capacity to suppress the innate and adaptive immune responses, the myeloid-derived suppressor cells (MDSCs)(Marvel and Gabrilovich 2015). The increase of MDSCs in the blood of patients with glioblastomas compared to healthy subjects has been reported by several studies (Raychaudhuri et al. 2011; Gielen et al. 2015; Dubinski et al. 2016).The STAT3 transcription factor is considered as one of the main regulators of the function and expansion of MDSCs (Dufait et al. 2016). Seven microRNAs of our study target the STAT3 gene.

Two of them (miR-20a and miR-223) have been directly related to the MDSCs functioning. MiR-20a appears as a negative regulator of the immunosuppressive properties of MDSCs by inhibiting the expression of STAT3 gene (Zhang et al. 2011). MiR-223, in a mouse model, suppresses the differentiation of stem cells from the bone marrow in MDSCs, reducing their number (Liu et al. 2011). Moreover, in MDSCs the expression level of miR-20a and miR-223 would be repressed by tumor-derived factors (Zhang et al. 2011; Liu et al. 2011). Both studies suggest a tumor suppressor role for miR-20a and miR-223. However, the overexpression of these microRNAs in DLGG patients compared with healthy subjects may be explained by the overall increase of the number of MDSCs in patients like in glioblastomas.

MiR-223 is most strongly expressed in granulocytes. MiR-223 physiologically finely regulates both the generation and function of granulocytic cells with a control on the inflammatory response (Johnnidis et al. 2008). The important role of miR-223 both in monocytes immune cells and neutrophils is consistent with the frequent association observed between cancer and inflammation. The link between cancer and inflammation involves also the STAT3 transcription factor targeted by several microRNAs of our study (Yu et al. 2009).

Our pilot study with a limited number of cases and controls makes difficult to draw already affirmative conclusions in the direction of a peripheral immune signature of DLGG. However, presumptive elements seem already to be pointed by the joint selection, among several hundred microRNAs tested, of a few microRNAs functionally connected to MDSCs and STAT3 pathway.

Advantages of a Whole Blood MicroRNAs Study in DLGG

Working on whole blood allowed us to analyze indiscriminately microRNAs of three possible origins: first microRNAs contained inside blood cells and also circulating extracellular microRNAs that can originate from the tumor or healthy cells.

Firstly, multiple artifacts could spoil pre-analytical phases of plasma or serum microRNAs studies. Pritchard et al. warned about observed variations in plasma or serum microRNAs profiles attributable to normal blood cells. These variations were greater in magnitude than many of the differences assessed between cancer patients and controls and considered as significant (Pritchard et al. 2012). Working on whole blood allowed benefiting from the information contained in blood cells microRNAs while avoiding a contaminant confounding effect from them.

Second, remote changes in peripheral blood cells induced by glioblastomas beyond the confines of the CNS have been described. Nilsson et al. isolated, in the circulating blood platelets of glioblastoma patients, an EGFRvIII RNA whose origin can only be the tumor tissue, evidencing an unsuspected transfer process from cancer cells to platelets. Furthermore, whole RNA profiles of platelets isolated from control subjects and from glioblastomas patients were different (Nilsson et al. 2011). It can be hypothesized that cancer cells could have transferred microRNAs too to platelets.

Nevertheless, one might object that the much greater amount of cellular microRNAs contained in blood cells may mask or dilute the tumor-specific cell-free microRNAs associated to exosomes or to protein complexes. But in the end, information contained in the cellular fraction could compensate this limitation especially in the case of DLGG. Indeed, in the DLGG context of highly regulated blood–CSF exchanges, tumor microRNAs release remains very uncertain. Hence, extending the search for whole blood could lift the limitation in peripheral detection of the tumor that would be linked to an undetectable level of tumor microRNAs.

Conclusion

To our knowledge, this is the first study of microRNAs profiles performed on a homogeneous cohort including only DLGG without histological signs of malignant transformation. It is important to restrict discovery cohorts to DLGG patients in order to identify a specific marker because the relevant markers are likely not the same as for high-grade gliomas.

We adapted our approach to peculiarity of DLGG. Indeed, one of our main hypotheses is that the slow and insidious development of DLGG requires prolonged modulation of the immune system that should affect the peripheral blood cells, including their microRNAs profiles. First interesting tracks in connection with MDSCs, inflammation, and more broadly with the STAT3 signaling axis have been pointed. Furthermore, we identified a combination of three microRNAs able to nearly perfectly differentiate subjects with DLGG from healthy ones in our learning cohort.

Our study is at now only a pilot feasibility study, with a limited number of cases and patients, and these initial results have to be confirmed on a larger number of DLGG patients. Whole blood assessment of microRNAs, whose levels could reflect a tumor release as well as functional modifications induced by tumor development in peripheral immune cells, is an innovative concept.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Compliance with Ethical Standards

Conflict of interest

All authors declare that they have no conflict of interest.

Ethical Approval

All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This study was approved by the local ethic committee (Centre de Protection des Personnes de Nîmes,France) and registered with the national regulatory agency (Agence Nationale de Sécurité des Médicaments, France) under the number 121951B-11.

Informed Consent

Informed consent was obtained from all individual participants included in the study.