Comprehensive Survey of miRNA-mRNA Interactions Reveals That Ccr7 and Cd247 (CD3 zeta) are Posttranscriptionally Controlled in Pancreas Infiltrating T Lymphocytes of Non-Obese Diabetic (NOD) Mice

In autoimmune type 1 diabetes mellitus (T1D), auto-reactive clones of CD4+ and CD8+ T lymphocytes in the periphery evolve into pancreas-infiltrating T lymphocytes (PILs), which destroy insulin-producing beta-cells through inflammatory insulitis. Previously, we demonstrated that, during the development of T1D in non-obese diabetic (NOD) mice, a set of immune/inflammatory reactivity genes were differentially expressed in T lymphocytes. However, the posttranscriptional control involving miRNA interactions that occur during the evolution of thymocytes into PILs remains unknown. In this study, we postulated that miRNAs are differentially expressed during this period and that these miRNAs can interact with mRNAs involved in auto-reactivity during the progression of insulitis. To test this hypothesis, we used NOD mice to perform, for the first time, a comprehensive survey of miRNA and mRNA expression as thymocytes mature into peripheral CD3+ T lymphocytes and, subsequently, into PILs. Reconstruction of miRNA-mRNA interaction networks for target prediction revealed the participation of a large set of miRNAs that regulate mRNA targets related to apoptosis, cell adhesion, cellular regulation, cellular component organization, cellular processes, development and the immune system, among others. The interactions between miR-202-3p and the Ccr7 chemokine receptor mRNA or Cd247 (Cd3 zeta chain) mRNA found in PILs are highlighted because these interactions can contribute to a better understanding of how the lack of immune homeostasis and the emergence of autoimmunity (e.g., T1D) can be associated with the decreased activity of Ccr7 or Cd247, as previously observed in NOD mice. We demonstrate that these mRNAs are controlled at the posttranscriptional level in PILs.

An example of a disease resulting from defective negative selection, although environmental factors also compete, is type 1 diabetes mellitus (T1D), which is a prototype of autoimmune disease [6]. In this disease, self-reactive CD4 + and CD8 + T lymphocytes, which mainly recognize insulin and/or other pancreatic autoantigens (e.g., GAD 65), infiltrate the pancreas (here, these cells are termed pancreas-infiltrating T lymphocytes or PILs) and destroy the insulin producing beta-cells through inflammatory insulitis. This is a gradual process that takes several years in prone individuals or over five to eight months in nonobese diabetic (NOD) mice [7][8][9][10][11][12]. Specific autoantibodies also participate in the destruction of beta cells in T1D [13].
As we introduced in a previous work [14], the NOD mouse is an autoimmune mouse strain that represents the most appropriate animal model to immune tolerance and develop autoimmune T1D, which reflects at least a part of human T1D [15][16][17][18][19].
Susceptibility to T1D is often associated with the major histocompatibility humans and the mouse homolog I-Ag in the NOD mouse [20][21][22][23][24], although non-MHC genes, such as Ptpn22, also participate in susceptibility to this disease [25]. Moreover, the NOD mouse features insulin-dependent (Idd) susceptibility loci. However, the contribution of various other factors to the immunopathogenesis of T1D remains largely unknown [19,21,26].
Type 1 diabetes mellitus does not occur without the participation of CD3 + T lymphocytes with either a CD4 + or CD8 + phenotype [27,28]. The immunoregulatory genes or cells that control the aggressiveness of autoreactive in the periphery and peripheral tolerance is a matter of importance that was first recognized several years ago and has been the focus of attempts to better understand the molecular genetic basis of autoimmunity in T1D [29][30][31][32][33][34][35][36][37][38][39].
We previously demonstrated that the development of T1D in NOD mice follows the transcriptional levels of immune reactivity genes during the maturation of thymocytes to peripheral CD3 + T lymphocytes [14]. However, the fine control of this gene expression, e.g., posttranscriptional control involving microRNAs (miRNAs), that occurs in these cells, including PILs, is largely unknown. MiRNAs regulate the gene expression of a variety of biological processes at the posttranscriptional level and are considered to be the main fine-tuning controllers of normal and pathological development [40][41][42].
Taking into account the function of miRNAs as posttranscriptional regulators, in this work, we sought to identify the miRNAs that are differentially expressed as well as their mRNA targets by comparing samples of thymocytes, peripheral CD3 + T lymphocytes and PILs during the progression of insulitis in NOD mice.
To answer these questions, we comprehensively assayed, for the first time and in a comparative manner, the mirnome (miRNAs) and transcriptome (mRNAs) of these cells using microarray hybridizations.
The data were first analyzed by hierarchical clustering to identify the differentially expressed miRNAs and mRNAs, and a selected set of these RNA species were reanalyzed using a dedicated bioinformatics tool (GenMir++ algorithm) to evaluate their interaction networking.
From the set of miRNA-mRNA interactions that were found during the development of thymocytes into PILs, we identified interactions between miR-202-3p and the Ccr7 chemokine receptor and Cd247 (Cd3 zeta chain) mRNAs, which were previously found to be involved in the control of aggressive autoimmunity in T1D in NOD mice.

Materials and Methods
Animals and assessment of type 1 diabetes mellitus Female non-obese diabetic (NOD) mice were born in specific pathogen free (SPF) conditions at the special mouse facility of the Ribeirão Preto Medical School and maintained in our laboratory in SPF mini-isolators with sterilized water and food ad libitum. The pre-diabetic mice were 1 month ± 2 week-old (1 mo) or 7 month ± 4 week-old (7 mo) mice. The 7 mo animals that became diabetic, have been identified and included in the study (group of diabetic 7 mo mice). Diabetes was assessed based on the blood glucose levels (animals exhibiting ! 250 mg glucose/dl blood were considered diabetic) using Accu Check-Active (Roche Diagnostics Brazil, São Paulo, Brazil). All experiments in this study were made in triplicate (samples collected from three animals of each phase of the disease) for microarray hybridizations as well as for qRT-PCRs.
The protocols used in this study were approved by the local ethics committee on animal experimentation (CEUA, University of São Paulo at Ribeirão Preto Campus, Brazil, Permit No. 120/2008).

Assessment of insulitis
The progression of insulitis was assessed by conventional hematoxylin and eosin (H&E) histology of the pancreas from pre-diabetic or diabetic NOD mice. The tissues were included in paraffin blocks, sliced at a 5-μm thickness, mounted on slides and stained with H&E. Histological examination and capturing of images were performed with a Leica model DMLB2 microscope coupled to a Leica DC300F camera. This approach allowed the observation of pancreatic islets with or without lymphocyte infiltration (data not shown).

Isolation of thymocytes, peripheral CD3 + T cells and PILs
The thymocytes or peripheral CD3 + T cells were isolated from the thymus or spleen, respectively, of pre-diabetic or diabetic NOD mice according to a previously described protocol [43]. In summary, thymi were dissected in DMEM/F10 medium, thymocytes were obtained by 2-3 passages of the thymic fragments through a 10-μm mesh nylon membrane (Sefar Inc., Depew, NY, USA), and pelleted thymocytes were suspended in PBS. Fluorescence activated cell sorting (FACS) analysis using a BD-FACSCalibur flow cytometer with a phycoerythrin (PE)-labeled anti-CD3 antibody indicated that this procedure yielded a thymocyte population with purity of ! 93%. These cells were used for total RNA extraction.
The peripheral CD3 + T cells were isolated from spleens of mice using magnetic beads for negative selection (Pan T-cell isolation kit, mouse, Miltenyi Biotec) according to the manufac-turer´s instructions. FACS analysis with the PE-labeled anti-CD3 antibody indicated that this procedure yielded a peripheral CD3 + T lymphocyte population of approximately 87% purity. These cells were used for total RNA extraction.
The T CD3 + pancreas-infiltrating lymphocytes (PILs) were isolated from 4 week-old prediabetic animals. Pancreata from pre-diabetic NOD female mice were dissected and macerated in RPMI 1640 medium. The pancreatic fragments were then incubated in 5 ml of 0.125% (w/v) type II collagenase and 0.1 DNase I in RPMI 1640 medium at 37°C for 15 minutes and gently agitated every 5 minutes with a 1 ml pipette. After digestion, the T lymphocytes were separated by 2-3 passages of the fragments and culture medium throughout a 10 um mesh nylon membrane (Sefar Inc, Depew, NY, USA). For T CD3 + cell recovery, the filtrate was centrifuged for 10 minutes at 1300 rpm at 4°C. Pelleted cells were suspended in RPMI 1640 medium and stained with a FITC-labeled anti-mouse CD3e antibody (Biolegend) for 30 minutes at 4°C. After that, T CD3 + cell population was separated by fluorescent-activated cell sorting in a FACS Aria III (BD Biosciences) flow cytometer. This procedure yielded approximately 70.8% purity of PIL T CD3 + cells (S1 Fig). These cells were used for total RNA extraction.

Total RNA extraction and quality control
The total RNA was extracted from approximately 1 x 10 7 thymocytes, peripheral CD3 + T cells or PILs using the mirVana total RNA isolation kit (Ambion, NY, USA) according to the manu-facturer´s instructions. The parameters for the quality control of the RNA preparations were previously described [43]. RNA preparations were confirmed to be free of proteins and phenol by UV spectrophotometry.
The integrity of RNA species was assessed by microfluidic electrophoresis using Agilent 6000 RNA Nanochips and an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Only RNA samples that were free of proteins and phenol and featured an RNA Integrity Number (RIN) ! 9.0 were used (data not shown).

Microarray hybridizations
For mirnome (miRNA) profiling. We used total RNA for direct labeling with Cy3 using the Agilent miRNA Complete Labeling and Hybridization Kit (Agilent Technologies, Mississauga, ON, Canada). The Cy3-labeled RNA samples were hybridized to Agilent mouse 8 x 15 K-format oligonucleotide miRNA microarrays, and the slides were washed according to the manufacturer´s instructions (Agilent Technologies) and scanned using an Agilent DNA microarray scanner [43].
For transcriptome (mRNA) profiling. We used total RNA to synthesize dscDNA and cyanine 3 (Cy3)-CTP-labeled complementary amplified RNA (cRNA) using the Agilent Linear Amplification Kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufac-turer´s instructions. The (Cy3)-cRNA samples were hybridized to Agilent mouse 4 x 44 K oligonucleotide microarrays (Agilent Technologies) for 18 h at 60°C, washed according the manufacturer´s instructions and scanned using an Agilent DNA microarray scanner [43].

Microarray data analysis
The hybridization signals from the scanned miRNA or mRNA oligonucleotide microarrays were extracted using the Agilent Feature Extraction software, version 10.5. The expression profiles from independent preparations of thymocytes or PILs from pre-diabetic or peripheral CD3 + lymphocytes from pre-diabetic or diabetic animals were analyzed by comparing the microarray hybridizations of the respective samples (Fig 1). The numerical, quantitative microarray data were normalized to the 75th percentile and analyzed using the GeneSpring GX bioinformatics platform (http://www.agilent.com/chem/genespring) according to the default instructions and allowing for hierarchical clustering of samples of mice or types of RNAs through ANOVA statistical analysis (P < 0.05), with a fold change ! 2.0 and uncentered Pearson correlation metrics [44].
Dendrograms were used to represent the similarities and dissimilarities in expression profiles between samples, miRNAs or mRNAs, and a colored heat map was used to represent the variability of RNA expression. The GeneSpring platform or Panther gene classification system The samples analyzed were as follow: Thymocytes (n = 2) or pancreas-infiltrating lymphocytes (PILs) (n = 2) derived from pre-diabetic (1 mo) or peripheral CD3 + T lymphocytes derived from pre-diabetic (1 mo n = 2 or 7 mo n = 3) NOD mice. These cell types featured unique mirnome expression signature. The dendrograms and heat map were obtained using cluster and tree-view algorithms through the Agilent GeneSpring platform. Heat map legend: red = up-regulation, green = downregulation, black = unmodulated (Pearson correlation metrics).
(http://www.pantherdb.org/) was used to assess the biological functions of mRNAs. Information about the miRNAs was retrieved from the miRBase data bank (www.mirbase.org).
Quantitative reverse-transcription real-time PCR (qRT-PCR) Cd247 mRNA. The differential expression of Cd247 mRNA as observed by microarray hybridizations was confirmed using qRT-qPCR. Complementary DNA (cDNA) was synthesized from 500 ng of total RNA using High-Capacity RNA-to-cDNA Kit (Applied Biosystems -Life Technologies, USA) as recommended. The PCR products were amplified from cDNA samples using Applied Biosystems TaqMan Universal PCR Master Mix (Applied Biosystems-Life Technologies, USA) with an assay probe spanning the exon junction of the murine Cd247 (FAM/MGB probe; assay ID: Mn00446171_m1 from Applied Biosystems-Life Technology, USA) or for the Applied Biosystems mouse GAPDH endogenous control (FAM/MGB Probe, Non-Primer Limited; assay ID: Mm00446171_m1) that was used as constitutive reference.
miR-202-3p. For qRT-PCR confirmation of miRNA levels we used TaqMan MicroRNA Assay (Applied Biosystems-Life Technologies, USA) to evaluate the expression of miR-202-3p (accession number: MIMAT0000235; mature sequence: AGAGGUAUAGCGCAUGGGAAGA). The snoRNA202 (accession number: AF357327; sequence: GCTGTACTGACTTGATGAAAG TACTTTTGAACCCTTTTCCATCTGATG) was used as an endogenous control. This method uses two-step qRT-PCR. In the reverse transcription (RT) step, cDNA was reverse transcribed from 10 ng of total RNA samples using the miRNA-specific stem-loop primers (sequences above) from the TaqMan MicroRNA Assays and reagents from the TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems-Life Technologies, USA). PCR products were amplified from each cDNA sample using the TaqMan MicroRNA Assays (Applied Biosystems -Life Technologies, USA), which contains the specific probes labeled with FAM, together with the TaqMan Universal PCR Master Mix (Applied Biosystems-Life Technologies, USA), according to manufacturer's instructions.
Transcriptional expression levels were determined using a StepOne Real-Time PCR System (Applied Biosystems, USA). The ΔΔCT was used as a relative normalization method. We used the GraphPad Prism 5.00 tool (http://www.graphpad.com/prism/Prism.html) to run one-way or two-way ANOVA with Bonferroni's correction statistics.

Reconstruction of miRNA-mRNA interaction networks
We used a protocol that was previously designed by our group and explained in detail [43]. Briefly, the miRNA-mRNA interaction networks were reconstructed on the basis of the respective miRNA and mRNA expression profiles to identify candidate miRNA-mRNA target pairs that were best supported by the expression data. The GenMir++ algorithm [45,46], which is available at http://www.psi.toronto.edu/genmir/, was used to establish such interactions. This algorithm calculates the scores to reproduce a mRNA profile using a weighted combination of the genome-wide, average normalized expression profile and the negatively weighted profiles of a subset of the miRNA regulators. The networks were graphically represented using the Cytoscape version 2.1 program (www.cytoscape.org). We used the TargetScanS database (http://www.targetscan.org) to verify the prediction of the selected target mRNAs.

Validations of miRNA-mRNA interactions
Determination of the hybridization minimum free energy (mfe). Using the miRNA-mRNA interaction networks generated by the GenMir++ algorithm, we selected pairs based on the biological function of the mRNA target. The annealing was validated through the use of the RNA-hybrid algorithm [47,48], which estimates the most favorable pairing between a given miRNA and its mRNA target by calculating the minimum free energy based on a thermodynamic state that postulates that an RNA duplex is more stable and thermodynamically stronger when the free energy is low [43,49]. This program is also available at http://bibiserv.techfak. uni-bielefeld.de/rnahybrid/.
Luciferase reporter gene assay (LRGA). We used a protocol previously designed in our group and explained in detail [43]. Briefly, the complementary oligonucleotide pairs containing a portion of the predicted miRNA-binding sites of the Ccr7 or Cd247 (CD3 zeta) 3´UTRs were synthesized by Integrated DNA Technologies (IDT, Coralville, IA, USA). The oligonucleotides were annealed and cloned into the pmirGLO vector (Promega Corporation, USA) between the XhoI/XbaI restriction sites of the polycloning site (PCS), resulting in the miRNA target region in the correct 5´to 3´orientation and immediately downstream of the luciferase gene. For selected targets, we introduced point mutations into the 7-nt seed-binding sequence. These constructs, named "pMIR-Ccr7" or "pMIRCd247" for the wild-type sequences and "pMIR-Ccr7(m)" or "pMIR-Cd247(m)" for the mutant sequences, were selected by colonypolymerase chain reaction (PCR) using a pair of primers flanking the vector PCS. We used Escherichia coli DH5α for cloning.
For the LRGA, 0.2 μg of each pmirGLO construct was transfected with 1.6 pmol of miR-202-3p or scrambled control miRNA (Thermo Scientific Dharmacon, Waltham, MA, USA) into human HEK-293T cells (6 x 10 4 cells/well) in a 96-well plate. Transfections were performed using Attractene Transfection Reagent (Qiagen, Hilden, Germany) according to the manufac-turer´s instructions. Transfected cells were incubated at 37°C in a 5% CO2 incubator, and 24 h after transfection, the cells were lysed in Passive Lysis Buffer. Firefly and renilla luciferase activities were measured in a Molecular Devices FlexStation 3 Luminometer using the Dual-Luciferase reporter system (Promega Corporation, USA) according to the manufacturer´s instructions. Our laboratory has national biosafety permission (National Technical Committee for Biosafety, Brasilia, Brazil, CTNBio, Permit No. 0040/98).

Statistical analysis
The LRA results are presented as the standard error of the mean (SEM). The differences were evaluated by one-way ANOVA followed by Student´s ttest (two groups). P < 0.05 was considered statistically significant.

Transcriptional profiles in thymocytes, peripheral T CD3 + lymphocytes and PILs
Microarray miRNA expression profiles (mirnome). From the set of murine miRNAs that were described and present in the microarray, 171 were identified as differentially expressed [p < 0.05, false discovery rate (FDR) = 0.05 and fold change ! 2.0] when we compared thymocytes from 1 mo-old and CD3 + peripheral T lymphocytes from 1 mo-old or 7 moold pre-diabetic mice with the corresponding cells from ! 7 mo-old diabetic mice or PILs from pre-diabetic 1mo-old NOD mice. The dendrogram and heat map depicted in Fig 1 show that these cell types feature unique miRNA expression signatures. From this set of miRNAs, we then selected 26 whose respective expression was exclusive in a given cell type of pre-diabetic or diabetic animals, i.e., induced/repressed in thymocytes, CD3 + peripheral T lymphocytes or PILs (Table 1). The up-or down-regulated miRNAs from each cell type were then reanalyzed using the GenMir++ algorithm.
Microarray mRNA expression profiles (transcriptome). From the whole set of sequences that were tested in the microarray, we initially identified a set of 33,459 mRNAs that were differentially expressed [p < 0.05, false discovery rate (FDR) = 0.05 and fold change ! 2.0] when comparing thymocytes from 1 mo-old and CD3 + peripheral T lymphocytes from 1mo-old or 7 mo-old pre-diabetic mice with the corresponding cells from ! 7 mo-old diabetic mice or PILs from pre-diabetic 1 mo-old NOD mice. From this set, we subsequently selected 2,070 mRNAs encoding proteins participating in biological processes associated with apoptosis, cell adhesion, cellular regulation, cellular component organization, cellular processes, development and the immune system.
The dendrogram and heat map depicted in Fig 2 show that these cell types feature unique expression signatures encompassing 73 mRNAs. In this work, we used a FACS sorted PIL preparation from which we focused those genes involved in the molecular/biological processes considered relevant for autoimmune reactivity for further analysis. The mRNAs of each cell type were then reanalyzed using the GenMir++ algorithm.

Reconstruction of miRNA-mRNA interaction networks
The 13 selected differentially expressed (up-or down-regulated) miRNAs and 2,070 selected differentially expressed (up-or down-regulated) mRNAs were reanalyzed using the GenMir++ algorithm to reconstruct the miRNA-mRNA interaction predictions. The predicted miRNA-mRNA interaction networks are depicted according to cell type: thymocytes (Fig 3A), CD3 + peripheral T lymphocytes ( Fig 3B) and PILs (Fig 3C).
We decided to further study the participation of Ccr7 and Cd247 (Cd3 zeta chain) in PILs that were simultaneously regulated by several miRNAs, including up-regulated miR-202-3p, Table 1. Modulated miRNAs according to cell type during the transition from pre-diabetic to diabetic NOD mice, their potential mRNA targets as predicted by the GenMir++ algorithm, the number of miRNA-mRNA interactions and GO biological processes of predicted target mRNAs. (#) Down-regutated, (") Up-regulated.
Comprehensive Survey of miRNA-mRNA Interactions in T Lymphocytes due to the involvement of these mRNAs in the control of cellular autoimmune reactivity. Because of their relevance to this study and their possible involvement in the control of autoreactivity, these miRNA-mRNA interactions were selected for further validation.

miR-202-3p or Cd247 mRNA expression levels as evaluated by qRT-PCR
We observe that while thymocytes develop into peripheral T CD3 + lymphocytes and PILs, the expression levels of miR-202-3p and Cd247 mRNA are opposite. While miR-202-3p levels are relatively lower in thymocytes and in peripheral T CD3 + lymphocytes, the CD247 mRNA levels are higher. The opposite occurs when we compare their expression in PILs. In these cells, miR-202-3p levels are high while Cd247 mRNA is low (Figs 4 and 5).

Validation of miRNA-mRNA interactions
Minimal free energy (RNA-Hybrid algorithm). The consequential pairing of the M. musculus Ccr7 mRNA (acc NM_001838) target region and the miR-202-3p miRNA was predicted to occur at nucleotide positions 166-172 of its 760-bp 3´UTR (www.targetscan.org). This miR-202-30p Cd247 (M. musculus Cd247 mRNA, acc. NM_ 000734 or NM_198053) interaction has not been previously reported. To validate our findings, we used the RNA-Hybrid bioinformatics tool, which calculates the most favorable hybridization between a given miRNA and the predicted 3´UTR portion of its mRNA target by calculating the minimum free energy (MFE).
The respective miRNA-mRNA annealing and MFEs are presented in Fig 6. Artificially introduced mutations changed the MFE and likelihood of miRNA-mRNA hybridization.
Luciferase reporter gene assay (LRGA). The LRGA was used to confirm the occurrence of the miR-202-3p-Ccr7 or miR-202-3p-Cd247 3´UTR interactions within the cellular milieu. We confirmed that miR-202-3p interacts with Ccr7 or Cd247 by its hybridization to their respective 3´UTR sequences, which contain the predicted binding sites for this miRNA. Irrelevant miRNA sequences or artificially introduced point mutations into (base substitutions) or base deletions from the original 3´UTRs significantly abolished the miRNA interaction, thus demonstrating the necessity of specificity and the exactness of base pairing for hybridization within the nucleotides that were selected for mutation (Fig 7).

Discussion
Type 1 diabetes mellitus (T1D) is an autoimmune metabolic disease with a strong genetic influence and association with environmental components. T1D begins with inflammatory insulitis, i.e., autoreactivity mediated mainly by CD4+ and/or CD8 + T lymphocytes that are directed toward insulin-producing pancreatic beta cells, in prone individuals or non-obese diabetic Comprehensive Survey of miRNA-mRNA Interactions in T Lymphocytes  (NOD) mice [15,23,[50][51][52]. Efforts are being made to better characterize the pancreatic autoantigens that are involved in insulitis as well as autoreactive immune cells [53][54][55].
In a previous study [15], we demonstrated the transcriptional modulation of immune reactivity genes during the development of T1D as thymocytes mature into peripheral CD3 + T lymphocytes in NOD mice, whereby genes involved in the negative selection of thymocytes, T-cell maturation, differentiation and autoreactivity are sequentially expressed.
However, two important gaps remain: the assessment of gene expression of the T lymphocytes that invade the pancreas (pancreas-infiltrating lymphocytes or PILs) and the evaluation of post-transcriptional control involving miRNAs.
In the present study, we sought to fill in these gaps by following a more comprehensive approach. Our hypothesis was twofold: first, we hypothesized that the modulation (up-or down-regulation) of miRNAs or mRNAs would be sufficient to hierarchize the different cell types (thymocytes, CD3 + peripheral T lymphocytes and PILs) as well as the different stages of NOD mice (pre-or diabetic); second, we examined the interaction between modulated miR-NAs and mRNAs from these cells during the evolution of T1D in these mice.
Our intention was to evaluate the expression signatures of these cells by focusing on known miRNA or mRNA sequences. The microarray method and bioinformatics pipeline that were used for this data analysis were adequate for this purpose. As proof of the suitability of these methods, the dendrograms and heat maps depicted in Figs 1 and 2 show that the different cell Moreover, to reconstruct the miRNA-mRNA interaction networks, the data were analyzed using the GenMir++ algorithm [46,47], which is based on Bayesian statistics for the analysis of large data sets originating from microarray hybridizations.
The results revealed that thymocytes, peripheral CD3 + T lymphocytes and PILs featured a large set of miRNAs as well as mRNAs that were differentially modulated (up-or down-regulated) ( Table 1). As a result, these cells, as well as the different phases of diabetic evolution in the mice (pre-or diabetic), were hierarchically clustered according to their respective miRNA or mRNA expression signatures. These data confirmed the first part of the original hypothesis.
Moreover, the demonstration of miRNA modulation corresponded to a unique aspect of this work and indicated that these cells undergo posttranscriptional control. By using Bayesian statistics through the GenMir++ algorithm, we were able to reconstruct miRNA-mRNA interaction networks and predict mRNA targets in an unbiased way, thus demonstrating that differentially expressed miRNAs interact with a large set of mRNA targets in each T lymphocyte phase that was assayed (Fig 3A-3C). This result confirmed the second part of the hypothesis.
As expected, the miRNAs that were selected for the reconstruction of the networks simultaneously interacted with several mRNA targets. For example, miR-101a, which was down- Comprehensive Survey of miRNA-mRNA Interactions in T Lymphocytes regulated in thymocytes, interacted with the following mRNA targets: Bub1, Bub1b, Cenpf, Endou, Rorc, Pax1, Epha2, Rag1 and Ptcra.
In addition, a given mRNA was regulated by several miRNAs. For example, in thymocytes, the Ccl3 mRNA simultaneously interacted with the following miRNAs: miR-29c (down-regulated), miR-345-5p and miR-34a (up-regulated).  Cotransfections with miR-202-3p or control (scrambled) miRNA were also performed to show the specificity of the miRNA, and co-transfections with the mutant (m) plasmid constructs were performed to show specificity of the 3´UTR for the interactions. The data are presented as the means and standard error of mean (SEM). The differences were evaluated by one-way ANOVA followed by Bonferroni´s test, and p < 0.05 was considered statistically significant when the wild-type 3´UTR pMIR was compared to the mutant (m) 3´UTR sequence in the presence of the respective miRNA mimic. Comprehensive Survey of miRNA-mRNA Interactions in T Lymphocytes Moreover, we observed that, independent of the expression profile (up-or down-regulation), miRNAs interacted with mRNA targets. The modulation of miRNAs was comparative among the cell types i.e. thymocytes, CD3 + peripheral T lymphocytes and PILs.
Regardless of their expression profile (up-or down-regulation), modulation of the miRNAs predicted the miRNA-mRNA interactions. This presumption was reinforced when we observed the opposite levels of miR-202-3p and Cd247 mRNA target during the development of thymocytes into peripheral T CD3 + lymphocytes and PILs (Figs 4 and 5).
The interactions of greater interest for this study were validated at two levels. In the first, we evaluated the minimum free energy (MFE) of the annealing between the miRNA and 3´UTR seed region of its mRNA target (Fig 6). The second level of validation involved the luciferase reporter assay, which confirmed that the miRNA-mRNA interaction could occur in the cell milieu (Fig 7).
As a whole, as they develop into CD3 + peripheral T cells and then into PILs, thymocytes exhibit miRNA interactions involving mRNA targets related to the immune system, apoptosis, cell adhesion, cellular regulation, cellular component organization, cellular processes, development, positive or negative thymic selection (although this miRNA regulation is absent in thymocytes) and immunological synapse, among others.
The T lymphocytes that were analyzed at different stages during the evolution of T1D in NOD mice underwent complex posttranscriptional control.
In our view, PILs are a cell type that should be better studied with respect to their posttranscriptional control of gene expression. Although efforts are being made to better understand the role of these cells [56], to date, little is known about their gene expression control.
In a more focused examination of PILs, we observed that miR-202-3p regulates the Ccr7 and Cd247 (CD3 zeta chain) mRNAs. This result is intriguing because Ccr7 is involved in the control of tolerance, and a previous study [57] demonstrated that mice lacking this chemokine receptor generated autoreactive T cells.
Moreover, Cd247 (CD3 zeta chain) was identified as a novel diabetes susceptibility gene whose function positively influenced the onset of autoimmune diabetes in mice when the gene was disturbed during TCR signaling [58].
Using the strategy of large-scale miRNA-mRNA interaction screening, in this study, we present evidence that these two mRNAs are under posttranscriptional control in PILs of NOD mice.
In our view, these results open new perspectives for further studies involving the use of miRNA antagonists (anti-Mirs) that are directed toward miR-202-3p and other interacting miRNAs, which could be further assayed in the control of insulitis in vivo in NOD mice and/or in vitro making use of the transwell assay to evaluate the migration of NOD mice-derived T lymphocytes toward pancreatic autoantigens.